id stringlengths 3 9 | source stringclasses 1 value | version stringclasses 1 value | text stringlengths 1.54k 298k | added stringdate 1993-11-25 05:05:38 2024-09-20 15:30:25 | created stringdate 1-01-01 00:00:00 2024-07-31 00:00:00 | metadata dict |
|---|---|---|---|---|---|---|
259375619 | pes2o/s2orc | v3-fos-license | Compilation of a simple chemistry application to quantum error correction primitives
,
I. INTRODUCTION
Quantum error correction (QEC), the study of how many noisy physical qubits are used to represent a smaller number of less noisy logical qubits, has seen significant recent developments in a number of directions.One such success is experimental demonstrations of error correction successfully suppressing errors on a real-world quantum device [1].Another recent development is in careful resource estimates, which have allowed for more accurate estimates of the resources a quantum computer requires to solve problems of significant interest, from estimating chemical properties [2][3][4][5][6] to factoring RSA integers [7,8].These developments together have helped define both the current state of our abilities to suppress noise on quantum devices, and where we need to get to in order to solve key industrial problems.
There are some natural next steps following the experimental demonstration of a logical quantum memory.Natural follow-ups include implementing basic logical gates: implementing Pauli gates through transversal operations, non-Pauli Clifford gates through lattice surgery techniques [9][10][11][12][13][14][15], and non-Clifford gates initially through error mitigation techniques [16] and later through magic state distillation [17][18][19].Eventually, a natural goal will be to demonstrate small-scale quantum algorithms, showing that these logical operations can be used to solve a toy application.Understanding the resources required for such an algorithm is important for knowing the point at which small applications can start being solved on fault-tolerant quantum computers, as well as helping us understand the constant factors in the scaling of large quantum algorithms.A number of algorithms have recently been proposed that are aimed specifically at this regime, referred to as "early fault-tolerant" algorithms [20][21][22][23][24]; it is therefore particularly relevant to assess how challenging even minimal applications will be to perform using fault-toleration operations.* alex.moylett@riverlane.com In this work, we estimate the resources required for implementing a small quantum algorithm on a fault tolerant quantum computer, including detailed consideration of how to perform each required operation using lattice surgery.The application we choose is quantum phase estimation (QPE) applied to finding the groundstate energy of the hydrogen molecule.This application is sufficiently small that related circuits without QEC have already been successfully run on current quantum hardware [25,26].We investigate optimisations of this algorithm at a variety of levels, including algorithmic [2,[27][28][29], gate decompositions [30], compilation to lattice surgery primitives [9][10][11][12][13][14] and generation of magic states [19].Our final resource estimates are presented in Fig. 1, looking at different physical error rates and techniques which trade off time and space resource requirements.It is worth noting that when implemented on the surface code, even this small application requires hundreds of physical qubits and thousands of QEC rounds.This shows the significant prefactor associated with quantum error correction, and suggests that in early fault-tolerance further techniques will be required to yield small-scale algorithmic demonstrations [16].
The rest of this paper is laid out as follows.In Section II, we review the algorithms and chemical system to be considered, and present the logical quantum circuit.In Section III, we describe how to decompose the logical quantum circuits into operations from the Clifford+T gate set and how to implement these gates on the surface code using lattice surgery primitives.In Section IV we estimate the overhead introduced by quantum error correction.Finally, we conclude with some open questions and further directions for research in Section V. FIG.1: Estimated cost in physical qubits and time for calculating the ground-state energy of a hydrogen molecule on an error-corrected quantum computer with varying physical error rates, using iterative quantum phase estimation.Methods for implementing logical gates through either directly implementing Clifford and T gates or moving Clifford gates through the circuit are described in Sections III B and III C, respectively.Further details about estimating these resource requirements are presented in Section IV C.
of quantum phase estimation (QPE) [31].QPE is one of the key proposed quantum algorithms for calculating ground and excited-state energies in electronic structure problems.Provided an initial trial state can be prepared that has a sufficiently good overlap with the true ground state (which is usually the case for molecular systems), QPE is capable of obtaining energy estimates to a desired precision in polynomial time with system size.However, the algorithm requires high circuit depths for non-trivial examples, and so has seen less attention compared to variational quantum algorithms in current NISQ applications.For fault-tolerant applications, however, it is often regarded as the algorithm of choice.
We focus on the "textbook" [32] and iterative (semiclassical) QPE algorithms [33][34][35][36].The textbook QPE algorithm is perhaps the best known QPE approach, the circuit for which is presented in Fig. 2a.The algorithm allows one to measure the eigenphases of some unitary U up to m bits of precision; doing so requires m ancilla qubits, in addition to the n data qubits needed to represent U .At the end of the circuit, an inverse quantum Fourier transform (QFT) is performed and the ancilla qubits are measured.If the input state |ψ⟩ is an exact eigenstate of U , then the measured bits will yield the bits of the corresponding eigenphase.For a non-exact |ψ⟩, the probability of obtaining the desired phase will depend on the overlap between |ψ⟩ and the corresponding exact eigenstate.
The inverse QFT can also be performed in a semiclassical manner [37].Using such a semi-classical QFT, the resulting phase estimation algorithm is performed iteratively, obtaining one bit of information about the phase from each iteration.We refer to this approach as iterative quantum phase estimation [36].Iterative QPE has many of the benefits of the textbook approach, in- FIG. 2: QPE circuits used in this paper.In both cases, the state |ψ⟩ is over n qubits.In (b), the circuit is iterated backwards from k = m (in the initial iteration) to k = 1 (in the final iteration).The rotation angle in iteration k is ω k = −π(0.xk+1 x k+2 . . .x m ), with ω m = 0 in the initial iteration.While the ancilla is measured at the end of each iteration, the data qubits remain coherent throughout.
cluding a Heisenberg-limited running time O(ϵ −1 ) for a precision of ϵ, but has the significant benefit that it uses only a single ancilla qubit.
We briefly give some analysis of the iterative QPE approach here.We are interested in estimating the eigen-values of a Hamiltonian where P j are n-qubit Pauli operators and c j are coefficients.We denote the eigenvalues and eigenvectors of H by {λ j ; |Ψ j ⟩}.We multiply H by a constant t such that −0.5 ≤ λ j t ≤ 0.5 for all j, which can always be achieved by choosing 1/t = 2 j |c j |.We then work with the unitary The eigenvalues of U are e 2πiϕj , where the range 0 ≤ ϕ j ≤ 1 can be chosen.It is then simple to obtain λ j t from ϕ j , which only differ due to the wrapping of phases; the normalization of Ht above is chosen to avoid potential ambiguity in this wrapping.Therefore each ϕ j can be written in binary as In iterative QPE the bits ϕ jk are measured directly using the circuit in Fig. 2b.The circuit is performed for m iterations in order to obtain m bits of precision for ϕ j , starting with k = m and iterating backwards to k = 1.
After each controlled-unitary operation, an R z (ω k ) gate is applied to the ancilla with angle which depends on the measurement results from previous iterations (and ω m = 0 in the initial iteration).The data qubits are prepared in an initial state |ψ⟩, which should be an approximation to the exact state whose energy is to be estimated.We write |ψ⟩ in the eigenbasis of H by |ψ⟩ = j ν j |Ψ j ⟩. ( The state of the qubits before the first measurement Consider the simple case where ϕ j can be represented by exactly m bits, so that ϕ j = 0.ϕ j1 ϕ j2 . . .ϕ jm 00 . ... In this case exp(i2 m πϕ j ) = exp(iπϕ jm ) exactly, and the state of the system before measurement is 1 2 Thus the probabilities of measuring the ancilla as 0 or 1 are Provided that |ν j | is sufficiently large for the desired state |Ψ j ⟩, the desired bit will be measured with high probability.The measurement will also project away the contribution from those states |Ψ j ⟩ for which ϕ jm does not match the measured result.It is simple to continue this process for subsequent iterations to k = 1.After the final iteration, the probability that all of the bits for the desired ϕ j were measured is |ν j | 2 .Therefore, for a sufficiently good initial state, and a sufficient number of repetitions, the ground-state energy can be measured with high probability.Further clear analysis is given in [36].
In addition to the textbook and iterative QPE methods, there has been recent progress on statistical phase estimation methods [21,23,[38][39][40].Compared to the above approaches, such statistical methods allow shorter circuit depth [22,23] and ready combination with error mitigation techniques [26], in exchange for performing many circuits.It has been suggested that these methods are particularly appropriate for early-fault tolerant quantum computers.We do not consider such methods here, but note that they would be interesting to investigate further in the context considered here.
B. Hamiltonian simulation via Trotterization
In this section we briefly discuss first and second-order Trotterization, and present an optimization to the latter.
We consider n-qubit Hamiltonians of the form of Eq. 1.We specifically denote H j = c j P j , so that In Trotter schemes more generally, each H j might correspond to a sum of commuting Pauli terms, rather than a single Pauli contribution.We are concerned with implementing an operator U = e iHt , controlled on an ancilla qubit.The well-known firstand second-order Trotter approximations, U 1 and U 2 , are and which have errors O(t 2 ) and O(t 3 ) compared to the exact U , respectively.Let us consider the number of single-qubit rotations needed to implement the controlled U 1 and U 2 unitaries, as required to perform QPE.Each controlled Pauli rotation, which we shall denote by W j , can be rewritten 3: Circuit diagrams demonstrating reduction of the controlled time evolution operator in QPE with second-order Trotterization.(3a) The time evolution operator controlled on an ancilla, and acting on an initial trial state |ψ⟩, which can be equivalently replaced by (3b) in phase estimation circuits.For the second-order Trotter formula, this can be further reduced to the circuit (3c).Lastly, each pair of boxed terms can be expressed as a single multi-qubit Pauli rotation (3d).
as (16) which is a product of two multi-qubit Pauli rotations.These can be reduced to a single-qubit rotation each after conjugation through an appropriate Clifford [40].Therefore the cost of each controlled Pauli rotation is 2 single-qubit rotations plus Cliffords, and the number of single-qubit rotations for U 1 is 2L per Trotter step.
At first glance it appears that for a given t, the secondorder formula requires 4L single-qubit rotations to implement.In fact in QPE circuits this is not the case, and the second-order formula can also be implemented with 2L rotations, as for the first-order formula, but with better error suppression.This trick was introduced in Ref. [27], and is known as directionally-controlled phase estimation.It was expanded on in Refs.[2,28] and also used in Ref. [29].
We briefly give a derivation of the directionallycontrolled approach.The general procedure is presented in Fig. 3.We consider the controlled time evolution operator in Fig. 3(a).The state of the qubits at the end of this circuit is Now, note that we can apply e −iHt/2 to the data qubits in Fig. 3(a) without affecting any measurement outcomes; since this operator commutes with all controlled-e iHt gates, it can be moved to the end of the circuit where it has no effect on the measurement of the ancilla.With this additional operator applied, the final state of the qubits is and we see that we can work with circuit in Fig. 3(b) instead.
We next expand e iHt/2 via its Trotter formula, where K is the number of terms in the Trotter product formula, equal to L for the first-order formula and 2L for second-order formula.Then, For even-order Trotter formulas the string of operators , and the expansion is unchanged when the order of the terms is reversed.Therefore, for the second-order Trotter formula (but not the first-order formula) we can write which is equivalent to the circuit in Fig. 3(c).Lastly, note that the paired operators in Fig. 3(c) can each be expressed as which can be reduced to a rotation on a single qubit plus Clifford gates.Therefore, application of the secondorder Trotter formula in QPE can be performed with 2L rotations, which is equal to the number required for the first-order Trotter formula.In addition, the Trotter expansion is applied to the operator e iHt/2 instead of e iHt , resulting in lower Trotter error.
C. The hydrogen molecule
We next define the Hamiltonian that we will consider throughout this paper.As an application of QPE, we will consider the common task of finding the ground-state energy of an electronic structure Hamiltonian.Such a Hamiltonian can be defined in second-quantized form as where p, q, r and s label spin orbitals.The coefficient h 0 defines the nuclear-nuclear contribution (which is just a number due to the Born-Oppenheimer approximation), and h pq and h pqrs are one-and two-body integrals, respectively.The form of these integrals are well known from quantum chemistry [41].
In this paper we are concerned with compiling a minimal chemistry problem to lattice surgery operations, including visualization of the patch layout.We therefore consider the hydrogen molecule H 2 in a STO-3G basis, which is a prototypical minimal molecular example, consisting of 2 electrons in 2 spatial orbitals, or 4 spin orbitals.We use an equilibrium geometry with an internuclear distance of 0.7414 Å.
The fermionic Hamiltonian in Eq. 23 must be mapped to a qubit Hamiltonian for use in QPE.Because the minimal basis for H 2 consists of 4 spin orbitals, direct mappings will result in a Hamiltonian with 4 qubits.However, as shown by Bravyi et al. [42], the qubit Hamiltonian for this problem can be reduced to just a single-qubit operator.This can be seen from symmetry arguments; the H 2 Hamiltonian (in this non-relativistic approximation) commutes with spin and particle-number operators, and also has spatial symmetry.Each of these symmetries allows one qubit to be tapered.More precisely, labelling the bonding and antibonding orbitals as ψ g and ψ u , and ordering the spin orbitals as ψ g↑ , ψ g↓ , ψ u↑ , ψ u↓ (that is, using a spin-interleaved arrangement), the only determinants that contribute to the ground-state wave function are |1100⟩ and |0011⟩, and these two states can be represented by a single qubit.A more general approach for tapering qubits due to Z 2 symmetries is given in Ref. [42].
The Hamiltonian used takes the form with c 1 = 0.78796736 and c 2 = 0.18128881, and we have neglected the identity contribution.Note that an identical qubit Hamiltonian was considered in Ref. [25], which performed textbook QPE on a neutral-atom quantum computer.
D. Overall logical circuit
Using the second-order Trotter formula techniques described in Section II B, we derive logical circuits for both textbook and iterative quantum phase estimation.We choose a time step t = π/(c 1 + c 2 ), where c 1 and c 2 are defined in Eq. 24, in order to ensure that eigenvalues of Ht are in the range [−π, π].In this simple application, we take just a single time step in the Trotter expansion of e iHt .We also perform QPE for just three bits of accuracy in the energy.These simplifications will of course lead to large errors in the final energy estimate; indeed, after removing rescaling factors, using three bits of precision means that the energy can only be estimated to precision (c 1 + c 2 )/4 = 0.242 Ha.Here, we are primarily interested in understanding the required circuits in terms of lattice surgery primitives.Increasing the number of Trotter steps, or bits of precision, does not provide further insight beyond increasing the circuit depth (and number of ancilla qubits, in the case of textbook QPE).
To implement e −ic1 Z⊗Zt/4 and e −ic2 Z⊗Xt/2 we use rotation operations R Z⊗Z and R Z⊗X , defining R P (θ) = e −iP θ/2 .Thus we have rotation angles θ 1 = tc 1 /2 and θ 2 = tc 2 for the Z ⊗ Z and Z ⊗ X rotations, respectively.We also make a minor optimisation by combining pairs of R Z⊗Z (θ 1 ) rotations into a single R Z⊗Z (2θ 1 ) rotation where possible.Figures for the logical circuits are provided in Appendix A.
For both iterative and textbook QPE circuits, the gates can be grouped into three types: Pauli gates such as the X gate, non-Pauli Clifford gates such as the Hadamard and S † gates, and non-Clifford gates such as the two-qubit rotations and T † gates.These different types of gates require different techniques to be implemented on the surface code, which we shall detail further in Section III.
III. IMPLEMENTING LOGICAL GATES
In this section we discuss how to implement the logical circuits presented in Section II D using operations available on the surface code.The surface code represents logical qubits as patches of d × d data qubits, where d is the distance of the code [43,44].Stabilisers consist of weight-4 X and Z measurements on the patch, and logical X and Z observables are defined along the horizontal and vertical boundaries of the patch [9].The surface code has proven to be a popular candidate for fault tolerant quantum algorithms, due to both its high threshold and low connectivity requirements.In particular, a number of resource estimation papers use the surface code as the basis for estimating the overhead from quantum error correction [4][5][6][7][8].
This section proceeds as follows.First, we approximately decompose the logical gates into a sequence of Clifford and T gates.Then we consider two potential methods for implementing these gates: in Section III B, we implement the Clifford and T gates directly using native lattice surgery operations; whereas in Section III C, we use commutation relations to remove Clifford operations from the circuit, at the cost of needing to implement more general T -like operations.
A. Decomposition to Clifford and T gates
Ideally we would want to implement logical quantum operations transversely on our error-correcting code, applying the operation to each physical qubit(s) in turn.Unfortunately, the Eastin-Knill theorem shows that this is not possible for any quantum error-correcting code [45].In the case of the surface code, the logical gates which can be implemented transversely are single-qubit Pauli gates if the code distance is odd.Other gates within the Clifford group can be implemented on the surface code via lattice surgery operations such as patch deformation [9], but non-Clifford gates such as the T gate cannot be implemented in an error-corrected fashion.
However, it is possible to approximately decompose an arbitrary unitary operation into a sequence consisting of Clifford gates and the single-qubit T gate.This was shown for arbitrary gates originally using the Solvay-Kitaev theorem [46], and a number of improvements have been subsequently shown for both single-and multi-qubit gates [30,47].
In the case of the QPE circuits in Section II D, the circuits contain a mixture of Clifford and non-Clifford gates.As Clifford gates can be implemented on the surface code via lattice surgery and patch-deformation techniques, such as the ones that we shall describe in Section III B, we only need to decompose the non-Clifford gates.Both the textbook and iterative QPE circuits consist of a series of two-qubit Pauli rotations as part of the Trotter expansion.In textbook QPE, there are controlled phase gates after the two-qubit rotations, to implement the inverse quantum Fourier transform.In iterative QPE, the two-qubit rotations are followed by (classically conditioned) single-qubit phase gates to implement a semi-classical version of the inverse Fourier transform.
We decompose the non-Clifford gates in two steps.First, we exactly compile the two-qubit operations into Clifford gates and single-qubit Z rotations and phase gates using circuit identities presented in Figure 4. Second, we use the gridsynth software package to approximately decompose the single-qubit Z rotations into sequences of one-qubit Clifford and T gates [30].Note that the two-qubit Z ⊗ Z rotations require a different decomposition to the controlled phase gates, due to differences in local phases.In comparison, single-qubit Z rotations are equivalent to single-qubit phase gates up to a global phase R Z (θ) = e −iθ/2 P (θ), and can therefore be decomposed using the same techniques.An example of using gridsynth to approximately decompose a single-qubit Z rotation into the Clifford and T gate set is provided in Figure 5.
There is a trade-off to be made between the accuracy of decompositions generated by gridsynth and the number of gates required.gridsynth approximates a single rotation R Z (θ) up to error ϵ in the operator norm with typically 3 log 2 (1/ϵ)+O(log(log(1/ϵ))) non-Clifford gates [30].To get an understanding of how this extends to a whole circuit, we ran simulations of the textbook and iterative QPE circuits with gridsynth decompositions of varying accuracy, from 1 bit to 32 bits.For each number of bits of accuracy, we generate 1000 circuits with the single-qubit rotations decomposed to that degree of accuracy, and simulate each circuit 10,000 times.
The results are presented in Figure 6.In Figure 6a, we take the total variation distance between the output distributions of the decomposed circuits with that of the perfect QPE circuit.From this we see that for both textbook and iterative QPE, the total variation distance reduces quickly to approximately 8.3 × 10 −3 at 10 bits of precision per gate decomposition, but tails off beyond this value.This is due to finite precision used when estimating the total variation distance from samples.In the following results, we choose 10 bits of precision for the decomposition of phase gates, as it provides sufficient overall total variation distance for purpose of this circuit.
We also present the number of gates required for each gate decomposition accuracy in Figure 6b.For 10 bits of precision, there are approximately 1,300 and 1,000 logical gates for textbook and iterative QPE, respectively.Fewer logical gates can also be used at the cost of increased error; for example, fewer than 1,000 logical gates can be achieved with 5 bits of precision per rotation: 870 gates for textbook QPE, and 740 gates for iterative QPE.The total variation distance at 5 bits of precision is 2.4%.
The results in Figure 6b also show that for this particular circuit iterative QPE requires fewer gates than textbook QPE regardless of decomposition accuracy.This is due to the fact that the inverse QFT step of textbook QPE requires two-qubit controlled phase rotations around fixed angles θ.These are subsequently decomposed into smaller rotations θ/2 and −θ/2, which are then approximately decomposed using gridsynth.In comparison, iterative QPE works with single qubit phase rotations θ which are classically controlled.As these single angles are larger than those used for the single-qubit rotations in textbook QPE, fewer gates are required to decompose them up to a desired accuracy.For this particular QPE circuit, which is only performed to three bits of accuracy, the smallest rotation angle required beyond the Hamiltonian simulation step in the iterative QPE circuit is −π/4, which can be implemented as a single T † gate.Hence for the rest of this paper we shall primarily focus on iterative QPE.
Finally, for anyone curious to see an example of the complete logical circuit, we have included example QASM circuits in the supplementary material to this paper [48], including an iterative QPE circuit with phase gates decomposed up to 10 bits of precision.This circuit features a total of 1029 operations, of which there are 13 X gates, 169 Z gates, 34 CNOT gates, 411 Hadamard gates, 13 S/S † gates, 386 T /T † gates, and three measurements in the Z basis.This is the circuit we will estimate the resources for in Section IV.Note that gridsynth is a randomised process, and so different runs might produce different gate decompositions than presented here.Next, we consider methods to implement Clifford and T gates in the logical circuit.These can either be applied directly, or can be moved to the end of the logical circuit [12].In this section we first discuss the time and space cost of directly implementing both Clifford and T gates.Both of these estimates will be calculated in terms of the code distance d.The approach of moving Clifford operations will then be considered in Section III C.
R(θ)
The simplest gates to implement on the surface code are single-qubit Pauli gates.These operations can be implemented by either applying the corresponding Pauli gate to all data qubits if the distance d is odd, or, if the distance d is even, by tracking their values in software.Due to their simplicity, we shall not focus on how to implement them in this section.Likewise, preparation and measurement of a logical qubit in either the Z or X basis can be done in a single QEC round by preparing or measuring all data qubits in that basis.In many cases these operations can even be implemented at a cost of no additional QEC rounds, by preparing the data qubits at the start of the following round or measuring the data qubits at the end of the preceding round.Note however that preparing or measuring a logical qubit in the Y basis is more complicated [14].
It is important to note that not every Clifford gate presented in Section III will be directly implemented.Any sequence consisting of only Z, S/S † , and T /T † gates can be implemented at the cost of implementing a single T gate (as shown in Appendix B).Thus we can think of the sequences of gates generated by gridsynth such as those shown in Figure 5 as equivalent to sequences of alternating Hadamards and T -like gates.
Before discussing how to implement non-Pauli gates, we present how our logical qubits are arranged on a quantum processor with nearest-neighbour connectivity.For iterative QPE, we have two logical qubits, each of which is represented by a d × d patch.The primary lattice surgery operations we utilise are for implementing joint Z ⊗ Z measurements.We arrange our logical qubits as d × d patches such that performing joint measurements with the horizontal observable is easy.We also introduce two additional spaces of d × d data qubits, which can be used as both routing space for performing joint measurements with the vertical operator, and for additional qubits required for implementing logical gates.We have the layout in Figure 7, which for distance d uses a total of (2d + 2) 2 data qubits, or 2(2d + 2) 2 physical qubits including those used for measurement.
CNOT gate
A CNOT gate between a control qubit c and target qubit t can be implemented based on two-qubit joint Pauli measurements [9,11], see Figure 8. Namely, an auxiliary qubit a is initialised in the |+⟩ state, followed FIG.7: Layout of logical qubits as distance d = 3 surface code patches which can be used when directly implementing Clifford and T operations.Orange dots represent qubits used for measuring stabilisers, which are represented by squares and triangles.X and Z stabilisers are coloured in grey and blue, respectively.Data qubits are not shown, but lie on the corners of the stabilisers.Additional qubits lie outside this space for generating states required for T gates, as detailed in Section IV A. by two joint measurements: Z c ⊗ Z a and X t ⊗ X a .Finally, the auxiliary qubit is measured out in the Z basis, and Pauli corrections are applied based on the outcomes.This operation can be implemented on our patches via the protocol shown in Figure 9.In Figure 9b, we use the routing space to initialise an additional patch in the |+⟩ state.We then use a merge-and-split operation between the horizontal boundaries of the control patch and the auxiliary patch to perform the Z ⊗ Z measurement, and at the same time grow and shrink the target patch to move it into the routing space.Next, we use another merge-and-split operation between the vertical boundaries of the target patch and the auxiliary patch to implement the X ⊗ X measurement.Finally, we measure out the auxiliary patch and at the same time use patch growing and shrinking to move the target qubit back to its original space.The remaining Pauli operations can either be applied transversely at the start of the next operation if the distance d is odd, or simply tracked in software if the patch distance d is even.The operations for growing and joining patches require d QEC rounds in order to protect the code from both qubit and measurement errors, the operations for splitting and shrinking patches as well as the single-qubit logical X measurement each require a single QEC round, and the Pauli operations at the end of the circuit are effectively free, meaning a total of 3d + 4 QEC rounds are required to implement the CNOT gate.
Hadamard gate
The Hadamard gate is a Clifford gate whose role is to swap the X and Z observables of a qubit.Naïvely, this can be achieved on a surface code patch by applying a Hadamard operation transversely to all data qubits on the patch, as shown in Figure 10a.However, this has the side-effect of swapping the X and Z stabilisers as well as the logical observables, resulting in a different patch to the one we started with and making joint patch operations such as those used for the CNOT in Section III B 1 more complicated.This effect of swapping the stabilisers can be seen by comparing the patches in Figures 10a and 10b.
If we rotated the patch by 90 degrees around the central data qubit after applying the transversal Hadamard gates, then we would have implemented the logical Hadamard gate.However, this is not possible on a physical device.Instead, we use a patch deformation technique, which we present in Figures 10b-10f, to achieve the same effect [49].First in Figure 10b, we grow the patch into a longer one with length 2d + 1.At the same time we move the corner at the top right in the original patch to the top left in the longer patch.Next in Figure 10c, we use patch deformation to move the corner on the bottom-right up to the top-right.At this stage the logical observables have changed directions from vertical to horizontal and vice versa.Next, we shrink the patch down in Figure 10d.Now we have the X and Z logical observables swapped, with the stabilisers in their original positions, but the whole patch has been shifted upwards.
To move this patch back to its original position, we start by growing and shrinking the patch in Figures 10e and 10f, but this leaves the patch one row of stabilisers higher than it originally was.To correct this, we use two rounds of SWAP gates to swap the data qubits with neighbouring measurement qubits, as shown in Figure 10f.
The most expensive parts of this process are the stages that involve patch growing and corner movement, which require d QEC rounds each.Since in general two-qubit gates are much noisier than one-qubit gates, the transversal Hadamard at the start of this sequence does not require any QEC rounds.Finally, patch shrinking and transversal SWAP gates each require a single QEC round, thus requiring a total of 3d + 4 QEC rounds.
S/S † gate
The S gate, also known as the √ Z gate, is a Clifford gate that applies a phase of i to the |1⟩ state.Like the Hadamard gate, this gate also cannot be implemented transversely on the surface code.
There are various ways of implementing the S gate using patch deformation, similarly to implementing the Hadamard in Section III B 2 [14,49].However, these require extending the X observable of a patch, and therefore require moving the patch into the routing space and back.Instead, we consider a different technique, which uses an additional patch in the |Y + ⟩ = (|0⟩ + i |1⟩)/ √ 2 state [12].We then perform a joint Z ⊗ Z measurement between this qubit and our qubit, and measure this auxiliary qubit in the X basis.Finally, we apply a Z correction depending on the outcomes of the two measurements.A circuit describing this operation is presented in Figure 11.
Note that unlike Z and X basis states, the |Y + ⟩ state cannot be generated in a single QEC round.Instead, we utilise a different technique to generate Y basis states in d/2 + 2 rounds with no additional qubits [14].With this additional patch, we can implement the logical S gate using the process described in Figure 12.Generating the |Y + ⟩ state in Figure 12a takes d/2 + 2 QEC rounds, the joint measurement in Figure 12b takes d QEC rounds, and measuring the |Y + ⟩ state in the X basis takes a single QEC round.Any Z correction can be applied in software at no additional cost, so the total number of QEC rounds required is 3d/2 + 3. Finally, note that the S † = SZ gate can also be implemented at no additional cost, by simply inverting the conditions under which the Z correction is applied.
T /T † gate
The T gate is a non-trivial gate to implement on the surface code as it cannot be performed either transversely or via lattice surgery operations such as patch deformation.Instead, we introduce an auxiliary qubit initialised in the |T ⟩ = (|0⟩+e iπ/2 |1⟩)/ √ 2 state.With this |T ⟩ state prepared, we can implement the T gate using techniques similar to those for implementing the S gate in Section III B 3. The circuit is presented in Figure 13 [12].First, we perform a joint Z ⊗ Z measurement between the data qubit and the auxiliary qubit.Next, we perform an S gate conditioned on the result of this measurement outcome.Finally, we measure out the auxiliary qubit in the X basis, and depending on this measurement outcome apply a final Z gate to the data qubit.However, while the |Y + ⟩ state can be prepared on the surface code in a fault-tolerant way in d/2 + 2 QEC rounds, the |T ⟩ state cannot be prepared on the surface code in an errorcorrected fashion, and thus additional work is required in order to prepare a high-quality |T ⟩ state.We shall detail this further in Section IV A. FIG.9: Implementation of a CNOT via lattice surgery operations, with the left patch of (9a) being the control qubit and the right patch being the target qubit.Green dots represent stabiliser measurements whose outcomes produce the result of the joint logical Pauli measurement.
We can implement this circuit on our patch layout using the process shown in Figure 14.Note that the patch for the |T ⟩ state is not stored in the routing space like the |Y + ⟩ is in Figure 12.This is because unlike the |Y + ⟩ state, the |T ⟩ state cannot be generated in a fault tolerant process, and instead needs to be generated elsewhere and stored outside of the routing space until it is required.Also note that the patch for the |T ⟩ state in Figure 14a is rotated compared to the patches for our data qubits, such that the vertical observable on the auxiliary patch matches the horizontal observable on our data patches.We use this to perform a joint Z ⊗ Z measurement between our auxiliary patch and our data patch via mergeand-split operations in Figure 14b.Finally, in Figure 14c we measure our auxiliary patch in the X basis, and at the same time we potentially apply an S correction using the methods described in Section III B 3. As with the CNOT gate presented in Section III B 1, the Z operation is effectively free as it can be either tracked in software or implemented transversely.The joint measurement requires d QEC rounds, the X measurement requires a sin-gle QEC round, and the S correction requires 3d/2 + 3 QEC rounds, leading to a total of 5d/2 + 4 QEC rounds to implement a logical T gate.
Finally, it is worth noting that other sequences of gates can also be implemented using these techniques with no extra cost.In general, any sequence consisting of only T /T † , S/S † , and Z gates can be implemented using the protocol above at the cost of implementing a single T gate.Further details are provided in Appendix B.
C. Moving Clifford gates
In this section we will consider another way of implementing the logical circuit from Section II on the surface code based on [12].This technique offers the benefit of only needing to think about how to implement the non-Clifford gates, but at the cost of increasing the complexity of implementing such gates.
state is required.Note that the S † gate can be implemented by inverting the condition under which the Z gate is applied.
Pauli product rotations
The key to this implementation method is that the logical gates we want to implement can be realised as rotations in a particular single-or multi-qubit Pauli basis.More formally, an n-qubit quantum gate can be implemented as a sequence of rotations R Pj (θ j ) = e −iPj θj /2 for suitably chosen P j ∈ {I, X, Y, Z} ⊗n and θ j [50].The simplest example of this phenomenon is the Pauli-gates themselves, which can be implemented as P = R P (π).Similarly, the T and S gates are both single-qubit rotations in the Z basis, and can thus be realised as T = R Z (π/4) and S = R Z (π/2), respectively.Singlequbit Pauli measurements, although not rotations around a Pauli basis, can also be seen as operations which project a state into a Pauli basis.In the case of QPE for example, measurements project a state into the Z basis.
The remaining gates to translate into this picture are the CNOT and Hadamard gates.Although not as easy to see as the gates listed above, both of these gates can be implemented as sequences of Pauli π/2 rotations given in Figure 15 [12].The Hadamard gate can be decomposed as ), up to a global phase.The CNOT can be written as a joint π/2 Z ⊗ X rotation, followed by a −π/2 Z rotation on the control qubit, and a −π/2 X rotation on the target qubit.This is similar to the circuit used in Figure 8, but with Pauli π/2 rotations rather than Pauli measurements.FIG.12: Implementing an S gate on a logical patch.In (12a), a patch in an S state has been initialised in the routing space using the methods provided in Ref. [14].Green dots represent stabiliser measurements whose outcomes produce the result of the joint logical Pauli measurement.
Moving Pauli rotations
The benefit of describing operations as rotations in a Pauli basis is that it becomes easier to understand how to transform them without modifying the outcome of the circuit.For example, in Figure 16, a π/2 rotation in the X basis is moved past a π/4 rotation in the Z basis.The result is that the Z rotation is transformed into a π/4 rotation in the iXZ = i(−iY ) = Y basis.These transformations can be applied more generally as well, the rules for which we discuss in Appendix C. The benefit of these transformations to the circuit is that we can move all π and π/2 Pauli rotations, which correspond to Pauli and Clifford operations, past the final measurement operation of the circuit.Operations beyond this point do not affect the outcome of our circuit, and therefore do not need to be implemented.Thus we have reduced our circuit to only involving π/4 Pauli rotations, which correspond to a generalisation of T gates, and joint Pauli measurements.We shall now look at how to implement these more general operations.
Implementing π/4 joint Pauli rotations
First we shall show how to reduce the π/4 joint Pauli rotations to joint Pauli measurements.These will then be implemented using a particular patch layout and lattice surgery operations in Section III C 4.
A circuit for implementing π/4 rotations is presented in Figure 17.This can be seen as a generalisation of the T gate circuit in Figure 13, where now the single-qubit Z basis has been replaced with a general multi-qubit basis P .The auxiliary qubit required for this operation is the same |T ⟩ = (|0⟩+e iπ/4 |1⟩)/ √ 2 state from Section III B 4. Because the rotation basis has generalised, so too have the corrective gates performed after the measurement.Now, instead of single-qubit S and Z gates we have more general π/2 and π rotations in an arbitrary Pauli basis P .The implementation of the π rotation is still a Pauli operation, and can be either tracked in software or implemented transversely as before.As for the π/2 rotation, one can account for this by employing the same techniques as described in Section III C 2 in an online fashion, moving the rotation past the final round of measurements to effectively remove it from the circuit and adjusting the subsequent operations accordingly [12].
Implementing joint Pauli measurements
Finally we discuss how to implement general Pauli measurements between patches on a surface code.The specific arrangement we use is given in Figure 18a.Note that this patch has more routing space than the one in 7, this is because the more general operations require access to both the horizontal and vertical observables of the patches.This results in six logical patches arranged on a grid of (3d+4)×(2d+2) data qubits, or 2(3d+4)×(2d+2) physical qubits total.FIG.14: Implementing a T gate on a logical patch.In (14a), a patch in a T state has been provided in some additional space generated by a magic state factory using methods described in Section IV A. Green dots represent stabiliser measurements whose outcomes produce the result of the joint logical Pauli measurement.S correction is not shown, but occurs in (14c) after the Pauli X measurement.
The most challenging operations to implement are those which include the Y basis of a qubit.This is because the Y basis does not correspond to the horizontal or vertical observable on a surface code patch, but is instead a product of both the horizontal and vertical observables.One option is to decompose π/4 rotations which involve the Y basis of a qubit into a sequence of π/4 and π/2 rotations which only act on the X and Z bases [12].However, doing so introduces π/2 rotations which cannot be moved past the π/4 rotation without reintroducing the Y basis, so such rotations would need to be implemented.Instead, we utilise another technique from [13] to implement Y basis measurements directly via lattice surgery operations.Some example measurements for implementing Pauli π/4 rotations in the Y ⊗ X and Z ⊗ Y bases are given in Figure 18.These joint measurements require d QEC rounds, followed by a single QEC round to measure the auxiliary patch in the X basis.These two sets of measurement results give us the corrections to move past future operations.
Here we utilise some lattice surgery techniques not used in Section III B. First, we add weight-five stabilisers, known as twist defects, which involve a Y Pauli term on one of the qubits.To ensure the surrounding stabilisers commute with the twist defects, we utilise two other lattice surgery techniques: first, we add domain walls, which are denoted by half-blue-half-grey squares and act as a combination of X and Z stabilisers; and second, we add elongated weight-four stabilisers, which are denoted by blue and grey rectangles.It is important to note that although these techniques allow for direct implementation of joint measurements involving the Y basis, there is an additional cost in that measuring these longer stabilisers requires additional connectivity compared to the layout used in Section III B. These extra connections between measurement qubits are not uniform, and shown by arrows in Figure 18.In general, for distance d a total of 4d extra connections are required for implementing this algorithm, which connect four columns of adjacent measurement qubits.
IV. ERROR CORRECTION OVERHEADS
We are now ready to discuss the cost of implementing these logical gates on the surface code.There are two primary sources of error which contribute to the probability of a failure at the error-correction level: first, errors from generating |T ⟩ states, which we shall explore in Section IV A; and second, errors from a logical failure on a qubit, which we shall explore in Section IV B.
A. Generating |T ⟩ states
Both of the methods used in Section III require additional qubits initialised in the |T ⟩ state.It is possible to initialise a surface code patch into an arbitrary qubit state |ψ⟩, by initialising one data qubit of the patch in the |ψ⟩ state, followed by d rounds of measurements [9].However, initialising a data qubit into an arbitrary state means that this qubit is initially unprotected from errors, so this method cannot be implemented in a way that reduces the logical error probability below the physical error probability.In fact, it can be shown that there is no fault-tolerant way of initialising non-stabiliser states such as the |T ⟩ state on the surface code [51].
Even though patches cannot be initialised in the |T ⟩ state in a way that suppresses errors, it is possible to use distillation protocols to reduce the error probability of |T ⟩ states.These protocols take multiple noisy |T ⟩ states and output a smaller number of |T ⟩ states with a reduced error probability [12,[17][18][19][52][53][54][55].For example, if it is possible to generate 15 |T ⟩ states each with error probability p, it is possible to distill these into a single |T ⟩ state with error probability 35p 3 [17].It is also possible to concatenate these factories to reduce the error probability even further.For example, if the 15to-1 protocol is used to generate 15 |T ⟩ states each with error probability 35p 3 , these can then be used in another 15-to-1 protocol to generate a single |T ⟩ state with error probability 35(35p 3 ) 3 = 1, 500, 625p 9 [12].The cost with these protocols is that reducing the error probability requires additional resources in terms of both time and number of qubits.A summary of several protocols and their associated costs is provided in Ref. [19].We also provide some example resource estimates for 15-to-1 factories in Table I, generated using code from Ref. [19].
When choosing a suitable protocol, there are multiple factors that we need to consider.First, we need to consider the overall logical failure probability from faulty |T ⟩ state generation.This means that if our logical circuit uses m T gates -and therefore requires m |T ⟩ stateswe need to choose a probability of distilled state failure p dist such that m × p dist is within our error bounds.
The second aspect we need to consider is the time required to generate each |T ⟩ state.In order to avoid logical qubits remaining idle as we wait for |T ⟩ states to be generated, we need to ensure that |T ⟩ states are gen-erated fast enough that they are available as and when they are needed.This depends on both the number of QEC rounds required to generate the |T ⟩ states, but also the number of QEC rounds required to implement these logical operations.If we implement Clifford and T gates directly as described in Section III B, the circuit primarily consists of alternating sequences of Hadamard gates, which take 3d + 4 QEC rounds, and T -like gates, which take between d + 1 and 5d/2 + 4 QEC rounds, depending on whether or not an S gate correction is required.This means that when implementing Clifford and T gates directly, a |T ⟩ state needs to be produced at least once every 4d + 5 QEC rounds.In comparison, when Clifford operations have been moved through the circuit as described in Section III C, the only operations required are a single joint Pauli measurement and a single X basis measurement, meaning that a |T ⟩ state must be produced every d + 1 QEC rounds.If a single distillation protocol cannot generate states fast enough, multiple instances of the protocol can be run in parallel to generate states more frequently, at the cost of increasing the number of physical qubits [12].As we show in Appendix D, up to four factories can be placed around the two corners at the top of the routing space.It is possible to add even more factories beyond these four, but doing so could require additional space for routing and storage of |T ⟩ states.On the other hand, if a logical |T ⟩ state can be generated faster than required, additional storage space is required to protect the state from errors while it waits to be consumed, which can be included as part of the routing space estimates.
B. Estimating code distance
To reduce the probability of a logical error occurring on one of our logical qubits, we can tweak the code distance d.A higher distance will reduce the probability of getting a sequence of physical errors which lead to a logical error, but comes at the cost of increasing both the number of physical qubits per logical qubit, and the number of QEC rounds per logical operation.In the case of the surface code, the probability of a logical error on a single logical qubit per code cycle assuming a depolarising noise model can be estimated as where p is the physical error probability [10,12,54].For the purpose of this application, we want to choose a sufficiently high d that the probability of a logical error occurring on any qubit during any QEC round is within our error bound.We use Eq. 25 to approximate our probability of a logical error at any point in the computation as where n data is the number of surface code patches for our data qubits, n route is the number of additional patches used for routing [56], and n meas is the number of QEC rounds.Given these parameters and physical error probability p, we can pick a distance by choosing an appropriate d such that Eq. 26 is within our target failure probability.
C. Results
We are now ready to estimate error correction overheads for our iterative quantum phase estimation circuit.As a recap, our circuit consists of 13 X gates, 169 Z gates, 34 CNOT gates, 411 Hadamard gates, 13 S/S † gates, 386 T /T † gates, and three Z basis measurements.As previously described, X and Z gates are free as they can be implemented transversely at the start of a QEC round.Of the S and S † gates, one is used in a sequence of T gates, and can therefore be implemented as a T -like gate.This leaves our costing as 411 Hadamard gates, 34 CNOT gates, 386 T -like gates, 12 S/S † gates, and three measurements.
We also need to make assumptions on the error correction requirements of our algorithm.We assume physical errors correspond to depolarising noise with a physical error probability ranging between 10 −4 and 2 × 10 −3 .We also assume a target failure probability of 1%, though a higher target probability can be used to reduce overheads [7,18].This target failure probability is split evenly, so the probability of errors occurring from faulty |T ⟩ state preparation is at most 0.5%, and the probability of logical errors happening on the qubits used in the logical circuit is also at most 0.5%.For 386 T -like gates, the required error rate per T gate in order to meet this error budget is 1.3 × 10 −5 .This is a higher error probability than what is seen from many distillation techniques [18,19], so instead we use code from [19] to look for smaller factories which still fit within our target failure probability.Note that both factories presented in Table I suffice at error rates 10 −3 and 10 −4 .
Our results are presented in Fig. 1.To help explain these resource estimates, the rest of this section will provide detailed costings for physical error rates of 10 −3 and 10 −4 .These physical error rates are commonly used when estimating the resource requirements of faulttolerant quantum algorithms [4,12,19].For ease of reading, a summary of these results is presented in Table II.
Cost of directly implementing Clifford and T gates
Using the estimates described in Section III B, we note that there are four logical patches to consider when estimating code distance.In terms of time requirements, CNOT and Hadamard gates require 3d + 4 rounds, S/S † gates require 3d/2 + 3 rounds, T -like gates require up to 5d/2 + 4 rounds, and Z basis measurements require a single round.This brings our total number of rounds to 2318d + 3363.Using Eq. 26, we find that for a physical error probability of 10 −3 , distance d = 12 achieves a logical error probability of 3.9 × 10 −3 , requiring 1,352 physical qubits for the patches and 31,179 QEC rounds.The factory in Table I produces a |T ⟩ state with error probability 8.1×10 −6 on average once every 31.3QEC rounds, meaning a single factory is sufficient.This factory uses 2,066 physical qubits, along with 288 physical qubits for storing |T ⟩ states.Combined with our 1,352 qubits for the logical circuit and routing, this leads to a total of 3,706 physical qubits.The additional logical qubit for storing |T ⟩ states increases the probability of a logical error to 4.9 × 10 −3 , leading to a total error probability of 8.1 × 10 −3 .
For a physical error probability of 10 −4 , an error probability of 6.9 × 10 −4 can be achieved with distance d = 6, which requires 14,953 QEC rounds and 288 physical qubits.A |T ⟩ state needs to be generated every 29 rounds with an error probability of 1.3 × 10 −5 .For this physical error probability, the factory in Table I produces a |T ⟩ state on average every 18.05 rounds with error probability 4.7×10 −6 .We use a single factory, which requires 522 physical qubits, along with 72 physical qubits for storing |T ⟩ states.Adding in our 392 qubits for the logical circuit and routing, this gives us a total of 986 physical qubits.The additional storage space for data qubits increases the probability of a logical error on qubits used in the quantum circuit to 8.6355 × 10 −4 , leading to a total error probability of 2.6 × 10 −3 .
Cost of moving Clifford gates
If we choose to move Clifford gates through the circuit, we are left with a total of 386 Pauli π/4 rotations, each of which requires d+1 QEC rounds, and three joint Pauli measurements, which require d QEC rounds each.Therefore our total number of QEC rounds is 389d + 386.We also have six logical patches allocated for both the logical circuit and routing.
For a physical error probability of 10 −3 , distance d = 11 achieves a logical error probability of 2.8 × 10 −3 , requiring 1,776 physical qubits and 4,665 QEC rounds.A |T ⟩ state needs to be produced once every 12 QEC rounds.We use the same 15-to-1 factory as in Table I, however a single factory is not sufficient for producing one |T ⟩ state every 12 rounds.Instead, we use three factories, which produce a single |T ⟩ state on average once every 10.4 QEC rounds and require 6,198 physical qubits for implementing the factories.The additional three logical qubits for storing |T ⟩ states increase the probability of a logical error on a qubit used in the quantum circuit to 4.2 × 10 −3 , leading to a total error probability of 7.3 × 10 −3 .The total number of physical qubits is 8,700.
For a physical error probability of 10 −4 , distance d = 5 achieves a logical error probability of 1.4 × 10 −3 at a cost of 2,331 QEC rounds and 456 physical qubits.Using the same factory as in Table I produces a |T ⟩ state with sufficiently low failure probability every 18.05 rounds, but a |T ⟩ is required every 6 rounds.Arranging four factories around the data qubits is sufficient to remove this bottleneck.These four factories require 2,088 physical qubits, and 200 physical qubits for storing |T ⟩ states.Adding this to our 456 physical qubits for the logical patches and routing leads to a total of 2,744 physical qubits.The extra four logical qubits for storing |T ⟩ states increase the probability of a logical error on a qubit used in the quantum circuit to 2.3 × 10 −3 , which means the total error probability is 4.1 × 10 −3 .
D. Analysis
As we can see from Fig. 1, there are still some significant overheads introduced from quantum error correc-tion.The most optimistic error rates still require hundreds of physical qubits and thousands of QEC rounds, while at an error rate of 0.2% this circuit requires tens of thousands of qubits and QEC rounds.
From the detailed costings of Section IV C, we can identify several bottlenecks with these approaches.For physical qubits, the overhead mostly comes from |T ⟩ state factories: at a physical error rate p = 10 −4 a single factory requires 522 physical qubits, nearly twice as many as required by the data qubits when implementing Clifford and T gates directly.This is even more prominent when moving Clifford gates through the circuit, where of the 2,744 physical qubits required at error rate 10 −4 , 2,288 are for preparing and storing |T ⟩ states.Although using fewer factories can reduce the number of physical qubits, this creates time bottleneck as the data qubits need to remain idle while |T ⟩ states are prepared.
Although it is expected that the overhead from such factories will become a less significant factor as we move towards larger quantum computations [19], for early fault-tolerant quantum circuits these overheads are likely to be more costly.This could be improved via more efficient small footprint factories like the ones presented in [19], as well as the use of error-mitigated T gates [16].
When moving Clifford operations through the circuit, another space overhead comes from joint Pauli operations in the Y basis.These require additional routing space and extra connectivity.Optimising the circuit to remove such measurements would also therefore reduce the routing overhead.
In time complexity, a significant bottleneck is the long sequences of Hadamard and T gates which come from gridsynth decompositions.Of the 846 logical operations implemented in this circuit, 797 are either Hadamard or T gates.The Hadamard gate is especially expensive, requiring 3d+4 QEC rounds.In practice, this means that more than half the QEC rounds are spent implementing Hadamard gates: at a physical error rate of 10 −4 , 9,042 of the 17,271 QEC rounds are spent implementing logical Hadamard gates.Time requirements can also be further reduced in general by using gate-based teleportation to execute gates in parallel, though this comes at a cost of more physical qubits [12].
Finally it is worth emphasising that there are other ways in which these resource estimates can be improved above the quantum error correction layer, such as the use of different quantum algorithms [57] and decomposition techniques [47].
V. CONCLUSION
As we enter the era of early fault-tolerant quantum computers, where quantum error correction is able to suppress errors on a logical qubit and basic logical gates are demonstrable, it is essential for us to understand the progress required for large-scale fault-tolerant quantum algorithms.Understanding the requirements of small ap-plications is an important step in the process.In this work, we have analysed a minimal application: estimating the ground-state energy of the hydrogen molecule.We have used several techniques to reduce the estimated resources to approximately 900 physical qubits and 15,000 QEC rounds through implementing Clifford and T operations directly, and approximately 2,700 physical qubits and 2,331 QEC rounds when implementing general Pauli π/4 rotations.
It is worth emphasising that even for this small application, the numbers of physical qubits and gates required is several orders of magnitude larger than what has been performed experimentally so far.There are a number of further optimisations which can be made across the quantum computing stack in the hope of reducing these estimates.At the algorithmic level, techniques such as qubitisation have been shown to produce asymptotically shorter quantum circuits [58][59][60][61], and could potentially offer improvements even for this minimal example [57].Statistical phase estimation methods can allow reduced circuit depth in exchange for performing more samples [22,23,40], and are often stated as being particularly appropriate for the early fault-tolerant era for this reason.At the gate synthesis level, alternative techniques have produced circuits with a smaller T count, at the cost of additional logical qubits [47].When implementing π/4 joint Pauli rotations, the number of QEC rounds can be further reduced by implementing non-commuting rotations in parallel on separate patches before using teleportation to combine them, though this comes at a cost of more physical qubits [12].Finally, improvements can be made to the implementation of non-Clifford gates which are more targeted towards early fault-tolerant quantum devices, such as the use of error mitigation when implementing faulty T gates [16], avoiding the need for magic state distillation factories.Algorithms such as statistical phase estimation may remain well suited even in the presence of error mitigation [26].
A final note is that these estimates assume that quantum computers are affected specifically by depolarising noise [7,10,12].While depolarising noise is easy to mathematically model, the physical noise that affects real-world devices is more complex and cannot necessarily be captured by such a model.An important direction of future work is investigating other more realistic noise models such as leakage and deriving similar scaling formulae to that presented in Eq. 25.
CODE AVAILABILITY
The source code for generating & running the logical circuits, and estimating resources, is available on GitHub [48].
FIG. 6 :
FIG.6: Performance of gridsynth on textbook and iterative QPE, for increasing bits of precision in the gridsynth decomposition (with 3 bits of precision used in each QPE circuit).(6a) Comparison of the decomposed circuits to the exact circuits in terms of total variation distance of the output distributions.(6b) The number of gates in the overall circuit.
B
. Directly implementing Clifford and T gates
FIG. 8 :
FIG.8: Circuit for implementing a CNOT gate through joint Z ⊗ Z and X ⊗ X measurements.An additional qubit initialised in the |+⟩ state is required.
FIG. 10 :FIG. 11 :
FIG.10: Implementation of a Hadamard gate on the right logical qubit via a transversal Hadamard gate, a series of patch deformations, and two transversal SWAP gates.Arrows denote SWAP gates between pairs of neighbouring qubits.In (10f), the first QEC round is to shrink the patch, and the subsequent two QEC rounds occur after each round of SWAP gates.
|TFIG. 18 :
FIG. 18: Layout of logical patches for implementing joint Pauli measurements.In (18a), the two qubits used in the logical circuit are at the bottom, and an auxiliary qubit is initialised in the |T ⟩ state at the top.Example joint measurements required for implementing π/4 Pauli Y ⊗ X and Z ⊗ Y operations in d + 1 QEC rounds are presented in (18b) and (18c), respectively.The auxiliary |T ⟩ state is always measured in the Z basis as part of the joint measurement.Green dots represent stabiliser measurements whose outcomes produce the result of the joint logical Pauli measurement.Twist defects are presented in yellow.Arrows between neighbouring measurement qubits show extended connectivity than what is required for the methods presented in Section III B.
FIG. 22 :
FIG. 22: Moving n-qubit π/2 Pauli rotations past other Pauli operations.(22a) A π/2 rotation in Pauli basis P can be swapped with a π/4 rotation in a commuting Pauli basis P ′ .(22b) A π/2 rotation in a Pauli basis P can also be swapped with a π/4 rotation in a non-commuting basis P ′ , by modifying the basis of the π/4 rotation to iP P ′ .These rules also apply to moving π/2 rotations past Pauli measurements, as shown in (22c) and (22d).
( a )FIG. 24 :
FIG.24: Arrangements of |T ⟩ state factories around the routing spaces for (24a) implementing Clifford and T operations directly and (24b) commuting Clifford operations.Green space denotes storage space for |T ⟩ states produced by the factories, which are denoted in yellow.Note that the full factories are not shown due to size.
gates into Clifford operations and single-qubit Z rotations.Note that while single-qubit Z rotations are equivalent to single-qubit phase gates up to a global phase, the two-qubit Z ⊗ Z rotation is different from a controlled phase gate due to local phases.
c) Controlled phase gate FIG.4: Decompositions of parameterised (4a) Z ⊗ Z rotations, (4b) Z ⊗ X rotations, and (4c) controlled-phase FIG.5: Example approximate decomposition of a π/8 Z rotation into a sequence of single-qubit Clifford gates and T gates using the gridsynth software package with three bits of precision.Global phases have been omitted, and some optimisations have been applied to combine multiple S and T gates.
TABLE I :
Resource estimates for some example 15-to-1 |T ⟩ state factories at physical error rates 10 −3 and 10 −4 .
TABLE II :
Detailed resource estimates required to perform the iterative QPE circuit described in the main text for the hydrogen molecule, considering physical error rates of 10 −3 and 10 −4 .Resource estimates for the |T ⟩ state factories are in TableI. | 2023-07-10T06:41:17.098Z | 2023-07-06T00:00:00.000 | {
"year": 2023,
"sha1": "4e4c9a6c4fc7c36474080253931dc1c4517ad1a5",
"oa_license": "CCBY",
"oa_url": "http://link.aps.org/pdf/10.1103/PhysRevResearch.6.013325",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "702f5dd1942432cfc7b718f2ffde7cfa7fe0830d",
"s2fieldsofstudy": [
"Chemistry",
"Computer Science"
],
"extfieldsofstudy": [
"Physics"
]
} |
261678657 | pes2o/s2orc | v3-fos-license | Is the Genetic Landscape of the Deep Subsurface Biosphere Affected by Viruses?
Viruses are powerful manipulators of microbial diversity, biogeochemistry, and evolution in the marine environment. Viruses can directly influence the genetic capabilities and the fitness of their hosts through the use of fitness factors and through horizontal gene transfer. However, the impact of viruses on microbial ecology and evolution is often overlooked in studies of the deep subsurface biosphere. Subsurface habitats connected to hydrothermal vent systems are characterized by constant fluid flux, dynamic environmental variability, and high microbial diversity. In such conditions, high adaptability would be an evolutionary asset, and the potential for frequent host–virus interactions would be high, increasing the likelihood that cellular hosts could acquire novel functions. Here, we review evidence supporting this hypothesis, including data indicating that microbial communities in subsurface hydrothermal fluids are exposed to a high rate of viral infection, as well as viral metagenomic data suggesting that the vent viral assemblage is particularly enriched in genes that facilitate horizontal gene transfer and host adaptability. Therefore, viruses are likely to play a crucial role in facilitating adaptability to the extreme conditions of these regions of the deep subsurface biosphere. We also discuss how these results might apply to other regions of the deep subsurface, where the nature of virus–host interactions would be altered, but possibly no less important, compared to more energetic hydrothermal systems.
INTRODUCTION
Viruses play a crucial role in marine biogeochemical cycles, microbial ecology, and evolution. Several recent reviews (Suttle, 2005;Rohwer and Thurber, 2009;Kristensen et al., 2010) have highlighted our current understanding of the viral impact on the marine environment. Generally, viruses influence the marine environment in three ways: first, through altering biogeochemical cycles by "shunting" the microbial loop through lysis of hosts (Suttle, 2005); second, by modifying the diversity and abundance of their hosts, particularly those that are most abundant, through what is dubbed "kill the winner" (Thingstad and Lignell, 1997); and third, by altering the genetic content of their hosts. Via the latter mechanism, viruses can fundamentally alter the course of evolution in their microbial hosts.
One means by which viruses can manipulate the genetic content of their hosts is by facilitating the process of horizontal gene transfer through transduction. This occurs when, in the process of virion synthesis, host genetic material is also incorporated into the viral genome. The newly synthesized viruses can then transfer the previous host's genetic material into a new host upon infection. It has been suggested that this could be an important mechanism for horizontal gene transfer in the marine environment: one study estimated that up to 10 14 transduction events per year occur in Tampa Bay Estuary alone (Jiang and Paul, 1998). Additionally, viral-like transducing particles known as gene transfer agents (GTAs) are increasingly recognized as an important mechanism for horizontal gene transfer: it has been estimated that GTA transduction rates are over one million times higher than previously reported viral transduction rates in the marine environment (McDaniel et al., 2010). GTAs and their potential impact on the deep subsurface biosphere will be discussed below.
Viruses may be particularly important in facilitating horizontal gene transfer between phylogenetically distinct lineages. Fully sequenced genomes of archaea and bacteria indicate that horizontal gene transfer does occur between lineages that are distantly related, including between domains. Overall, it is thought that interdomain horizontal gene transfer is a relatively common phenomenon, with approximately 3% of genes in bacterial genomes and 4-8% of genes in archaeal genomes involved in transfer events (Yutin and Koonin, 2009). The genome of Methanosarcina mazei, for example, contains many genes of possible bacterial origin including a bacterial chaperonin system (Deppenmeier et al., 2002). Thermophiles appear to have a particularly high percentage of transferred genes (Beiko et al., 2005;Yutin and Koonin, 2009), though it is unclear if this is due to a propensity for acquiring distantly related genes or to the close proximity of archaea and bacteria in high temperature habitats like vents. For example, up to 16.2% of the genes in Aquifex aeolicus had a best hit to archaeal species (Deckert et al., 1998;Yutin and Koonin, 2009), which is much higher than the average for bacteria. The genome of Thermotoga maritima, isolated from geothermally heated sediment, contains genes related to oxygen reduction that are likely to have www.frontiersin.org been transferred through a single transfer event from a member of the Thermococcales (Nelson et al., 1999). Several Thermococcales strains have been found to contain mobile genetic elements or virus-like particles (Prieur et al., 2004), and thus there is a strong possibility that viruses are responsible for these transfer events. Viral transduction and GTAs are attractive as possible mechanisms for transfer events such as these because other mechanisms of genetic transfer do not seem capable of explaining the observed genomic similarities among distantly related species. Conjugation is a specialized process limited to particular lineages, and transformation often transfers only small amounts of genetic material, rather than multi-gene cassettes. While viral host range varies depending on viral type, some, especially GTAs, are thought to be capable of infecting distantly related hosts.
Viruses can manipulate the genetic content of their hosts not only by horizontal gene transfer, but also by expression of viralencoded genes during the course of infection. The life cycle of lysogenic viruses involves a stage in which the virus integrates its own genome into the host genome, lying latent within the host genome for several generations until it is induced by an environmental stressor or other signal. Many integrated prophage have been found to encode what are termed fitness factors or lysogenic conversion genes: genes expressed by the prophage that can promote host fitness (Paul, 2008). These fitness factors can enhance host survivability in various environmental conditions. The cholera toxin genes in Vibrio cholerae, for example, are expressed by a filamentous bacteriophage that has integrated into the V. cholerae genome (Waldor and Mekalanos, 1996). Studies have shown that prophage genes are upregulated in response to changing environmental conditions (Smoot et al., 2001) or during biofilm formation (Whiteley et al., 2001). Even cryptic prophage, which have been integrated in the host genome for so long that they have decayed and are no longer active as phage, can carry genes that improve survivability during osmotic, oxidative, and acid stresses, and influence biofilm formation . Viruses infecting marine cyanobacteria, or cyanophage, have been found to carry genes for both photosystems I and II (Mann et al., 1993;Lindell et al., 2004Lindell et al., , 2005Millard et al., 2004;Sullivan et al., 2005;Sharon et al., 2009). These genes are expressed during viral infection, and are thought to enhance phage fitness by supplementing the host's photosynthetic machinery (Lindell et al., 2005).
While a large amount of work has been dedicated to understanding the viral impact on the marine environment, the implications for the deep subsurface biosphere have been barely explored. Table 1 summarizes current research on viral abundance, production, and diversity in the deep subsurface biosphere, as well as shallow sediments, methane hydrates, and the deep-water column to provide a basis for comparison.
As can be seen in Table 1, only a few studies have focused on deep subsurface viruses. Most research to date in the deep ocean has focused on viral production and abundance in the water column and surface sediments. None of these studies, to our knowledge, has focused on the genetic and evolutionary implications of viral infection on the deep subsurface biosphere. Yet the viral impact on the evolution of bacterial and archaeal hosts could have even more profound implications in the deep subsurface than in the upper water column. A primary reason for this is that it is generally accepted that lysogeny is a more important viral lifestyle under suboptimal conditions, when host or nutrient abundance is low (Paul, 2008). This has been demonstrated in laboratory conditions, in which lysogenic viruses have a competitive advantage over lytic viruses in nutrient-limited media (Levin and Lenski, 1983). In natural populations, lysogeny becomes a more common lifestyle during seasons or in regions where host abundance is low (Jiang Table 1 | Summary of previous work on viral abundance, activity, and diversity in various environments of the deep subsurface biosphere, deep ocean, and sediments.
Environment Work on viruses to date Reference
Surface marine sediments High viral production in benthic ecosystems: may be responsible for up to 80% of cell mortality, thus releasing large amounts of carbon through the "viral shunt." Viral diversity in sediments is fairly high, and showed a higher incidence of lysogenic than lytic phages Danovaro et al. (2008), Middelboe et al. (2006, Siem-Jørgensen et al. Deep-water column Viral abundance generally tracks bacterial abundance, but the virus:cell ratio at depth varies. In some areas, the ratio increases with depth Hara et al. (1996), Parada et al. (2007), Steward and Preston (2011) Metagenomic work characterizing viral diversity found most viral sequences had matches to bacteriophages in the Podo-, Sipho-, and Myoviridae, with a few hits to eukaryotic sequences Frontiers in Microbiology | Extreme Microbiology and Paul, 1998;McDaniel et al., 2002;Williamson et al., 2002;Weinbauer, 2004). Therefore, lysogeny may be a favored lifestyle in the deep subsurface due to difficulties in finding a new host after newly synthesized virions are released. In deep sediments, the main difficulties are likely to be low cell abundance and immobility. In extreme environments, lysogeny may be favored because of reduced viability of viral particles outside the cell. As shown in Table 1, lysogeny appears to be a common lifestyle in sediments and hydrothermal vents. If lysogeny is indeed more common in the deep subsurface biosphere, this would increase the proportion of cells harboring prophage, potentially increasing the number of cells expressing fitness factors encoded by the prophage.
The impact of viruses on the genetic landscape of their hosts may be particularly pronounced in diffuse flow fluids of hydrothermal systems. In these environments, high temperature hydrothermal fluid mixes with seawater both in the subsurface and above the seafloor, resulting in gradients of temperature, pH, and chemical and mineralogical composition (Baross and Hoffman, 1985). A schematic representation of the vent environment, and the accompanying gradients, is shown in Figure 1A. The constant fluid flux through these gradients enables potentially frequent contact between diverse microbial communities.
The microorganisms inhabiting diffuse flow fluids are tremendously diverse in terms of taxonomy, metabolism, and thermal Edwards et al. (2011), Huber et al. (2003, and Baross and Hoffman (1985). www.frontiersin.org regime. Studies of the population structures of archaea and bacteria in diffuse fluids have found thousands of phylotypes within both domains (Huber et al., 2007). The abundance of reduced compounds in vent fluids allows microbial metabolisms to take advantage of a wide range of energy sources, including hydrogen, reduced iron, sulfur, as well as organic compounds. There is, therefore, a wide range of potential hosts for viruses to infect in diffuse flow fluids.
In the extreme and dynamic conditions of hydrothermal vents, genetic exchange could provide a fitness advantage to any organism able to acquire a novel function that is useful in a changing environment. Evidence already exists suggesting that vent habitats are conducive for genetic exchange. As mentioned above, genomic studies have shown that rates of horizontal gene transfer between thermophiles are higher than between other groups, including between domains. This may be due to the predominant physiology of vent organisms: microbial communities often form biofilms on the surfaces of or within vent structures (Jannasch and Wirsen, 1981;Schrenk et al., 2003). Subsurface sediments and crustal habitats also provide abundant surfaces for biofilm formation. The high cell density in a biofilm increases rates of host contact, thus facilitating higher rates of viral infection, and can also foster genetic exchange through transformation. A recent metagenomics study found that transposase sequences in biofilms in the Lost City hydrothermal field are 10 times more abundant than in metagenomes from other environments (Brazelton and Baross, 2009). If this result is characteristic of subsurface biofilms, then genetic exchange in the deep subsurface biosphere is likely to be of great importance.
It has been suggested that viruses may act as a reservoir of genes that can be used as a mechanism for adaptation by their cellular hosts (Goldenfeld and Woese, 2007), effectively expanding the "pan-genome" to include the viral assemblage. Here, we argue that viral-mediated genetic exchange is particularly important as a means to adapt to frequently changing conditions in the diffuse flow vent environment. We will review what is known about viruses in the vent environment, and examine what evidence exists that viruses play a role in modifying the genetic content (and therefore the fitness) of their cellular hosts. We conclude by discussing the larger implications for viruses in the deep subsurface biosphere in general.
VIRUSES AS A "GENETIC REPOSITORY" IN HYDROTHERMAL VENT SYSTEMS EVIDENCE FOR VIRAL ACTIVITY IN DIFFUSE FLOW SYSTEMS
The few studies that have focused on viruses in diffuse flow vent environments have indicated that viruses play an important role in influencing microbial ecology in these ecosystems. One study quantified viral abundance in diffuse flow fluids, finding that on average there were approximately 10 7 viral-like particles (VLPs) per milliliter of fluid, or about 10 times as many viruses as cells, which is comparable to other marine ecosystems (Ortmann and Suttle, 2005). Another study, focusing on VLP counts at shallow hydrothermal vents, found that viruses were about five times more abundant than cells, though found that VLP counts increased with distance from the vents (Manini et al., 2008). The authors suggested that one possible explanation for this trend was a change in viral lifestyle from lysogenic to lytic, resulting in an increase in the number of apparent viruses. Additionally, diffuse fluids from 9˚North, a hydrothermal vent field on the East Pacific Rise, have been found to harbor a higher incidence of inducible prophage than nearby ambient seawater, indicating that lysogeny is a more common lifestyle for vent phage than those in other marine environments (Williamson et al., 2008). A viral metagenome at Hulk vent on the Juan de Fuca Ridge showed that vent viruses are relatively diverse and have the potential to infect hosts from a wide range of taxonomic groups and thermal regimes (Anderson et al., 2011).
Genome sequences of bacteria and archaea may also show evidence of extensive viral activity in diffuse flow hydrothermal vents. Within bacterial and archaeal genomes are regions called clustered regularly interspaced palindromic repeats (CRISPRs), which are thought to facilitate the immune response to viral infection (Barrangou et al., 2007;Brouns et al., 2008;Sorek et al., 2008;van der Oost et al., 2009;Hovarth and Barrangou, 2010;Labrie et al., 2010;Marraffini and Sontheimer, 2010). These loci consist of a series of short repeats, about 20-50 bp long, and are interspersed by a series of spacer regions, each about 25-75 bp in length. These spacer regions are created to match a short sequence on an invading element, such as a virus or a plasmid. The newly synthesized spacers are then inserted between direct repeats on the CRISPR locus (Makarova et al., 2003;Bolotin et al., 2005;Haft et al., 2005;Mojica et al., 2005;Pourcel et al., 2005;Marraffini and Sontheimer, 2008;Hale et al., 2009). If a virus or plasmid invades a cell that possesses a CRISPR spacer matching a sequence on that invader, the CRISPR system will mobilize the immune response. This occurs through the formation of a Cascade complex from nearby Cas (CRISPR-associated) genes in conjunction with small RNAs derived from the CRISPR spacers, which recognize and bind target DNA (Jore et al., 2011). The invading nucleic acid is cleaved as a result (Garneau et al., 2010).
Each of these CRISPR regions therefore acts as a record of previous viral infection, with each spacer thought to represent at least one independent infection event in the history of that strain. Interestingly, it has been observed that thermophilic strains, on average, have a higher number of CRISPR loci in their genomes than mesophiles or psychrophiles (Makarova et al., 2003;Anderson et al., 2011; Figure 2). While there is no definitive explanation for the abundance of CRISPRs in thermophiles, this does indicate that viral infection plays an important role in the evolution and ecology of thermophilic microbial communities. Moreover, the CRISPR immune mechanism itself is unique among viral immunity systems in that it responds in a sequence-specific manner to the invasion of foreign genetic material, rather than through prevention of phage adsorption, blocking phage DNA entry, or random restriction modification. Thus, the abundance of CRISPRs in thermophiles indicates that at least the entry, if not the successful takeover, of foreign genetic material between different hosts is a relatively common phenomenon in high temperature environments. CRISPR loci can also be found in the genomes of bacteria and archaea isolated from other environments in the deep ocean and sediments ( Table 2), though there is a clear correlation with temperature. This may serve as evidence that viruses play a particularly important role in the evolution of microbes in high temperature environments. This includes diffuse flow hydrothermal vents but could also include other regions in the subsurface with broad temperature gradients.
Frontiers in Microbiology | Extreme Microbiology
A further piece of evidence for the important role of viruses, one that is not restricted to high temperature organisms, is the presence of prophage in the genomes of sequenced archaea and bacteria. As mentioned above, lysogenic phage can provide supplementary metabolic functions while integrated into the genome as prophage. Table 2 lists the bacteria and archaea from diffuse flow fluids and other deep ocean habitats that have been found with integrated prophage in their genomes. Two of the listed bacterial isolates, one from the deep-water column and one from marine sediments, each possess seven prophage in their genomes. This represents a significant amount of genetic material. Considering the tendency of bacteria to select for faster reproduction rates and thus for smaller genomes (Carlile, 1982), the presence of seven prophage regions on the genome represents a potential fitness cost (through slower replication rates) that must be offset by a fitness benefit, such as the presence of fitness factors expressed by the prophage. Considering the prevalence of prophage in the genomes of isolates from diffuse flow fluids and deep sediments, it would not be surprising if prophage are present in most subsurface organisms.
EVIDENCE FOR VIRAL-MEDIATED HORIZONTAL GENE TRANSFER
Having established that viruses are abundant and likely to be active in diffuse flow hydrothermal fluids, we now turn to evidence that viruses mediate the exchange of genes between hosts in these regions. Here, we will do this primarily through analysis of a marine hydrothermal vent viral metagenome, or virome, collected from Hulk vent in the Main Endeavor Field on the Juan de Fuca Ridge, previously described in Anderson et al. (2011). We examine evidence for viral-mediated horizontal gene transfer in the vent environment first by searching the marine vent virome for genes related to lysogeny and gene insertion; and then by using comparative metagenomics to determine which genes are selected to be maintained in the viral gene pool.
GENES RELATED TO LYSOGENY AND GENE INSERTION
As mentioned above, a higher proportion of bacteria and archaea in diffuse flow hydrothermal vents appear to contain inducible prophage within their genomes than bacteria and archaea from other marine environments (Williamson et al., 2008), indicating that lysogeny is an important lifestyle for viruses in diffuse flow fluids. To test whether this finding is supported by publicly available metagenomic data, we searched for genes associated with lysogeny in the marine vent virome from Hulk and compared the number of matches with that of metagenomes from other environments. It is important to note that while most lysogenic phage are integrated into cellular genomes as prophage, the Hulk vent sample would have experienced substantially colder temperatures as well as decreased pressures prior to the filtration steps that removed cells and captured viral-sized particles. This environmental shock may have induced many of the prophage within the microbial community, resulting in their being captured in the viral size fraction. While this has not yet been demonstrated experimentally, it www.frontiersin.org CRISPR loci were identified with CRISPRFinder (Grissa et al., 2007a,b) and prophage were identified with ProphageFinder (Bose and Barber, 2006). Modified from a table in Orcutt et al. (2011).
is an important consideration in the sampling of metagenomes from relatively extreme environments.
To determine the overall abundance of genes related to lysogeny, we first created a database of lysogeny-associated proteins with Pfam seed sequences (Finn et al., 2010). These proteins include phage integrases, repressors, and antirepressors expressed during the prophage stage, regulatory proteins that trigger the switch between the lysogenic and lytic stages, and proteins involved in phage integration and excision. We queried this database with the marine vent virome using BLASTX, as well as a set of other viral and cellular metagenomes from the MG-RAST database (Meyer et al., 2008).
Metagenomes with the highest percentages of sequencing reads matching lysogeny domains were cellular rather than viral ( Table 3). These hits were most likely matches to prophage incorporated into cellular genomes. Two of the top three metagenomes were sampled from environments that are considered "extreme," namely, a highly acidic mine drainage and the high-pH, high temperature Lost City hydrothermal field. This supports the hypothesis stated earlier that lysogeny becomes a more common viral lifestyle in extreme environments because viruses have a higher chance of survival as prophage than as virions under harsh conditions. Also in the top three was the whale fall cellular metagenome. While it is unclear exactly why this metagenome had a relatively high percentage of lysogeny-related domains, one possibility is that since whale falls are relatively rare events, the organisms colonizing whale falls must endure extreme periods of relative starvation, followed by periods of plenty. Any viruses infecting these taxa would benefit from a lysogenic lifestyle in order to survive these extreme periods of starvation, during which cells would most likely either be dormant or replicating extremely slowly.
Fourth on the list was a cellular metagenome sampled from farm soil, in which lysogeny may be the favored lifestyle for viruses as a result of the difficulties in encountering a new host within the sediment matrix, an environment in which mobility is likely to be significantly impaired. However, sequences matching lysogeny domains were not particularly abundant in cellular metagenomes derived from deep Peru Margin sediments, though some depth horizons had a higher abundance than others (Table 3). At this point it is unclear whether this is due to differences in viral lifestyle or simply to a lower abundance of viruses in Peru Margin sediments.
In contrast, the sequences matching lysogeny domains in the Yellowstone hot springs and marine vent viromes (fifth and sixth Frontiers in Microbiology | Extreme Microbiology on the list, respectively) are most likely derived from lysogenic viruses that entered the lytic stage due to an environmental stressor, possibly resulting from the sampling process. Interestingly, these two viromes, both sampled from high temperature environments, had the highest proportion of sequencing reads matching lysogeny domains of all the viromes analyzed here. In contrast, viromes from more temperate environments, such as the Bay of British Columbia or the Sargasso Sea, had a much lower abundance of lysogeny sequences. This parallels the relative abundance of lysogeny-related domains in extreme cellular metagenomes, and again may indicate that natural selection favors lysogenic viruses in extreme environments. The lysogeny-associated domains chosen for this analysis were selected from genes that are uniquely associated with prophage. However, this would exclude genes that serve other roles in both viruses and cells but may also be crucial in facilitating horizontal gene transfer. This includes genes required for integration of DNA into host or viral genomes, such as DNA ligases. Interestingly, 1.25% of the sequencing reads in the marine vent virome had a match to a DNA ligase, which is almost 10 times higher than any of the 17 other cellular or viral metagenomes analyzed here (Table 3). Moreover, these ligases were especially enriched in the subset of the marine vent virome considered more likely to be "viral" -that is, those reads that were assembled into contigs with an average coverage of at least eight, or within contigs in which a majority of reads were categorized as either "unknown" or "viral" by MG-RAST (Anderson et al., 2011). Therefore, the abundance of ligases appears to be distinctly viral in character.
Most of the potential ligase sequences in the marine vent virome had matches to NAD-dependent ligases rather than ATPdependent ligases. ATP-dependent ligases are fairly widespread among eukaryotes, bacteria, and archaea, while NAD-dependent ligases are generally characteristic of bacteria, though they have been seen in some viruses, eukaryotes, and archaea (Doherty and Suh, 2000;Tomkinson et al., 2006). Furthermore, they appear to be characteristic of marine rather than terrestrial metagenomes, for unknown reasons. To determine whether the ligases in the marine vent virome were most closely related to ligases from a particular group of organisms, we constructed a phylogenetic tree of NAD-dependent DNA ligases including ligase-matching www.frontiersin.org sequences from the marine vent virome, using a reference tree of previously characterized DNA ligases as a constraint (Figure 3). The reference viral ligases do not necessarily group together on this tree, suggesting there is not a viral ligase type that is necessarily distinct from bacterial, archaeal, or eukaryotic ligases. Similarly, the sequences from the marine vent virome are scattered across the tree, with some closely related to ligases found in viruses, and others more closely related to bacterial or archaeal ligases. Yutin and Koonin (2009) have previously highlighted the non-monophyletic nature of viral DNA ligases. This may attest to the fluid nature of viral genomes in that they frequently pick up genes from their hosts and may also be a further indication of the high diversity of the marine vent viral assemblage. While several vent virome sequences were grouped with DNA ligases found in viruses, the majority grouped most closely with a DNA ligase from Rickettsia felis, a Gram-negative bacterium in the Alphaproteobacteria group. R. felis is closely related to SAR11, one of the most common bacterial lineages in marine environments. While SAR11 sequences were not common in the marine vent virome, it is possible that ancestral viruses in the marine vent environment had acquired ligases from SAR11 or similar groups, which then became more abundant in the viral gene pool due to positive selection.
While it is unclear exactly what ecological role these ligases play, their high abundance suggests that they play a uniquely important role in the vent viral assemblage. DNA ligases repair double-stranded breaks in DNA by catalyzing the synthesis of phosphodiester bonds between 5 -phosphoryl and 3 -hydroxyl groups (Lehnman, 1974). DNA ligases are thus important for DNA replication, recombination, and repair across all domains as well as viruses. However, the high abundance of DNA ligases in the vent virome is unusual and may reflect a specific role in the vent ecosystem, possibly in the integration of viral genomes into the host genome through recombination. It is therefore plausible that the abundance of these genes is indicative of prevalent horizontal gene transfer in these regions of the deep subsurface biosphere.
Finally, GTAs may act as an important mechanism for viralmediated horizontal gene transfer. These phage-like particles were originally discovered in marine Rhodobacteria, but have since been discovered in various other bacterial and archaeal species, particularly in the marine environment (Lang and Beatty, 2000;Matson et al., 2005;Stanton, 2007;Biers et al., 2008, Leung et al., 2010. GTAs randomly incorporate segments of the host genome into a viral capsid, then transfer this to new hosts, including phylogenetically unrelated organisms, without resulting in lysis of the host cell. It has been suggested that GTAs are defective phage Metagenomic reads are colored red. All ligase protein sequences were obtained from NCBI (accession numbers listed). Trees were constructed in RAxML by incorporating metagenomic sequences into a constraint tree of references sequences based on the phylogeny of Yutin and Koonin (2009). Trees imaged with TreeViewX. (Lang and Beatty, 2000;Matson et al., 2005;Stanton, 2007); if so, it would seem that GTAs have effectively lost their parasitic nature and have instead been usurped by the host for the purposes of gene exchange. While the only sequenced GTA genes are from the Rhodobacter capsulatus (Lang and Beatty, 2000) and Brachyspira hyodysentariae (Matson et al., 2005) GTAs, the marine vent virome included six sequencing reads with matches to known GTA genes. While these are few, the scarcity of sequences from GTAs in public databases to date prohibits a larger-scale search for GTAs in viral metagenomes. Nevertheless, GTAs may play a significant role in transferring genes between hosts in the marine environment and appear to be present in diffuse flow fluids. In the dynamic, extreme conditions of the deep subsurface, selection may be particularly strong for cells that harbor GTAs as a mechanism for obtaining advantageous genes. It has been hypothesized that a large portion of marine viromes may consist of GTAs carrying poorly conserved bacterial genes (Kristensen et al., 2010) and may thus contribute a large portion of the poorly conserved "cloud" of genes in the viral gene pool.
ENRICHMENT FOR FUNCTIONAL GENES IN THE VENT VIROME
This "cloud" of genes, or genetic reservoir, consists of an overlapping pool of genes derived from both viruses and cells. Genes in the viral genetic reservoir are expressed in cellular hosts through either horizontal gene transfer or prophage insertion, and then maintained in the gene pool through positive selection. Thus, the genes maintained in the viral metagenome, in addition to genes for viral synthesis, packaging, and maintenance, are likely to consist of non-housekeeping genes that provide a selective advantage in the event of a shift in environmental conditions. This is expected to be particularly the case in the gradient-dominated, dynamic vent environment.
An initial example of this can be found through analysis of assembled contigs of the marine vent virome from Hulk (assembly and contig analysis described in Anderson et al., 2011). Among the lysogeny domains matching sequences in the marine vent virome was the XerD tyrosine recombinase protein, commonly associated with lysogenic phage and transposases. This gene was assembled onto a contig that also contains the universal stress protein UspA. UspA is known to be phosphorylated in response to stasis stress in Escherichia coli (Nyström and Neidhardt, 1994;Freestone et al., 1997), and thus is induced when growth conditions are not optimal. It would be advantageous to both host and prophage for the prophage-encoded protein to induce a response in the host when growth conditions are poor. As mentioned, prophage have been found to encode genes to enhance host survivability in stressful conditions, and this sequence may be one example of many such instances in the vent ecosystem.
While this is one example of a potential fitness factor encoded by a lysogenic phage, comparative metagenomics can give a more large-scale picture of which functional genes are enriched in the viral fraction. To do this we compared the relative enrichment of several gene categories in the marine vent virome and a cellular metagenome sampled from a sulfide chimney at Mothra hydrothermal field on the Juan de Fuca Ridge (Xie et al., 2011;Figure 4). Each metagenome was analyzed through the MG-RAST pipeline and annotated using the KEGG Orthology database (Meyer et al., 2008). For this analysis, as before, we used only the vent virome subset that we consider to be convincingly viral: that is, sequencing reads assembled into contigs with an average coverage of at least eight or composed of reads identified as "unknown" or "viral" by MG-RAST. The virome was relatively enriched in gene categories such as replication and repair, nucleotide metabolism, and translation, which are important for the synthesis of viral genetic material during the lytic stage. Interestingly, however, the virome was particularly enriched in genes related to energy metabolism, with nearly 12% of all identifiable sequencing reads matching this category in the virome, compared to 7% in the cellular metagenome. Genes in this category include genes used for oxidative phosphorylation, photosynthesis, methane metabolism, carbon fixation, and sulfur and nitrogen metabolism. It seems plausible that these genes are maintained in the viral genetic reservoir through positive selection. For example, if a cell were to be flushed into a region in which the abundance of its conventional electron donor was limited, the acquisition of a gene allowing it to utilize an alternative electron donor, via either transduction or prophage expression, would increase the fitness of that cell. The relative enrichment of genes related to energy metabolism in the virome may be evidence of this type of selection.
Other sequences in the marine vent virome had matches to metabolic genes that have also been found in viruses from other environments. For example, 11 sequencing reads in the marine vent virome had close matches to PhoH, a phosphate starvationinducible protein that has previously been found in Prochlorococcus and Synechococcus phage genomes (Sullivan et al., 2009). It has been hypothesized that PhoH, when expressed by phage, could aid the host in phosphorus scavenging during the infection stage. In hydrothermal systems, phosphate is removed from seawater through water-rock reactions in the deep subsurface, thus limiting phosphate availability to native microbial communities (Wheat et al., 1996). The presence of PhoH in the marine vent virome suggests that vent viruses may, in a manner similar to the cyanophage, assist in phosphate acquisition during the course of infection. Similarly, eight reads closely matching transaldolases, also found on cyanophage genomes, were found in the marine vent virome. It is thought that these transaldolases may play a role in metabolizing carbon substrates to assist in energy production and synthesis of important compounds, and thus may be yet another example of fitness factors expressed by phage (Sullivan et al., 2005).
This positive selection for genes that can increase host fitness would suggest that the genes found in the marine vent virome can give an indication of the unique environmental conditions from which it was sampled, such as limitations or extremes in factors such as nutrients, energy availability, temperature, or pH. For example, 301 sequencing reads in the marine vent virome were annotated by MG-RAST as uptake [NiFe]-hydrogenases, enzymes involved in the oxidation of hydrogen to produce energy (Vignais and Billoud, 2007). Hydrogen oxidation is known to be an important metabolic strategy in hydrothermal vent systems, where it provides a greater energy return than the oxidation of methane or sulfur (Amend and Shock, 2001), which are also common reduced compounds in vent environments. The [NiFe]hydrogenases encoded in the marine vent virome could enable www.frontiersin.org (Meyer et al., 2008), and reads were annotated with the KEGG Orthology database, release 56.
utilization of an important alternative energy source for viral hosts if transferred into their genome via transduction, or if expressed by prophage as a fitness factor, which could bolster host metabolism during viral infection.
To present a larger-scale illustration of how the genetic profile of a virome might act as a signature of the environment from which it was sampled, we compared the relative enrichment of functional gene categories in the marine vent virome with a set of 42 other viromes, initially analyzed by Dinsdale et al., 2008;Figure 5). As in the previous analysis, only the convincingly viral subset of the marine vent virome from Hulk was used. In this analysis, genes were annotated with the SEED subsystems in MG-RAST to make a more direct comparison with the study by Dinsdale et al. (2008). The results show that the marine vent virome is particularly enriched in genes related to regulation and cell signaling (enriched by 300%) and RNA metabolism (enriched by 230%). The genes assigned to the "cell signaling" category include those related to biofilm formation and quorum sensing, regulation of virulence, and sensing environmental stimuli. For example, 118 sequencing reads in the marine vent virome have close matches to genes in the "regulation of virulence" category. Most of these matched the BarA-UvrY two-component system, a system regulating virulence in pathogenic E. coli (Herren et al., 2006). The BarA-UvrY system functions by sensing changes in environmental conditions to induce a metabolic switch, a function that would be useful in the rapidly changing conditions of the vent environment.
We expect that the pool of genes undergoing positive selection for retention in the virus genetic reservoir would represent those genes that may be occasionally necessary, though not strictly required, in the environment from which it was sampled. Thus, we expect accessory genes such as those related to secondary metabolisms to be selected for maintenance in the virome, rather than housekeeping genes such as ribosomal proteins. In addition to representing the "cloud" of poorly conserved genes from both cellular and viral pangenomes (Kristensen et al., 2010), we expect positive selection to customize this "genetic reservoir" to match the needs of the hosts drawing from it.
EXTRAPOLATING THE VIRAL IMPACT FROM HYDROTHERMAL VENTS TO OTHER REGIMES OF THE DEEP SUBSURFACE BIOSPHERE
We have suggested here that viruses are likely to play an important role in modifying the genetic content of their hosts in diffuse flow fluids of hydrothermal systems. We now turn to a discussion of whether the characteristics of viruses in diffuse fluids is likely to be the case for other regimes of the deep subsurface biosphere. The vent ecosystem has several unique attributes that distinguish it from other deep subsurface habitats, so extrapolation from the vent subsurface to the rest of the deep subsurface biosphere must be done cautiously. The two most distinctive attributes of the vent system are the extreme gradients in temperature, pH, redox state, chemical composition, and mineralogy, and the constant fluid flux Frontiers in Microbiology | Extreme Microbiology between microenvironments. On the ridge axis, hot hydrothermal fluid mixing with cool, oxygenated seawater facilitates interaction among diverse cellular communities and their accompanying viral assemblages. For this reason, we have argued that this environment may be conducive to gene flow from one microenvironment to the next via viral transfer. However, the impact of viruses is likely to extend to other provinces of the deep subsurface biosphere, which we detail below.
CRUSTAL ENVIRONMENTS: UNSEDIMENTED RIDGE FLANKS, CRUSTAL OUTCROPS, SEAMOUNTS, ARC SYSTEMS
This fluid flux from one environment to the next is characteristic of many regimes in the deep subsurface biosphere. As sediments covering the seafloor tend to be relatively impermeable, fluid flux depends on outcropping of igneous crust . Fluids flux most vigorously at mid-ocean ridges at the hydrothermal systems discussed above, and degree of fluid flux decreases with distance from the ridge axis, as shown in Figure 1B. However, these environments are not completely stagnant. On ridge flanks and recharge zones, which can extend hundreds of kilometers from the ridge axis, oxygenated seawater flows into the ocean floor, with residence times of days to years (Johnson et al., 2010). Some of this fluid emerges in discharge zones in the form of hydrothermal vents on ridge axes, while some fluids may circulate more locally. Offaxis, fluid flows through seamounts; and farther afield, fluids may circulate in a more restricted manner (Edwards et al., 2005). Overall, however, the volume of fluid flux is large, as at least 60% of the oceanic crust is hydrologically active, and the fluid-crust reservoir in which these processes might be expected to occur is approximately 10 times the size of the sedimentary reservoir . Fluid-flux-dominated environments, therefore, constitute the majority of the deep subsurface. This fluid flux facilitates transfer of chemicals throughout the crust, thus exposing microbial communities to varying conditions and also transporting microorganisms from one region to the next. The regions of the oceanic crust dominated by fluid flux are therefore analogous to the hydrothermal systems described above, though characterized by fewer extremes in temperature, pH, and redox conditions. By means of this flux, viruses or their accompanying hosts are transported between biomes, and can facilitate horizontal gene transfer between organisms native to drastically different environments. Viral metagenomics has suggested that while local diversity of viral assemblages is high, global diversity may be low, suggesting that viruses frequently migrate between biomes (Breitbart and Rohwer, 2005). One study found that viruses from sediments, soils, and lakes were able to propagate in marine waters, suggesting that viruses have a broad enough host range that they can successfully move between biomes (Sano et al., 2004). Fluid flux between the seafloor-ocean interface, within sediments, and within cracks in the Earth's crust may act as a conduit for viruses or the microbes that bear them, thus potentially sharing genes among habitats. Moreover, if biofilms are hotspots for horizontal gene transfer and viral infection (as discussed above), we may expect that regions of the subsurface characterized by biofilm formation, which could www.frontiersin.org include any habitat with surfaces available for attachment, will be exposed to higher rates of gene transfer than in environments with lower cell density.
SEDIMENTED ENVIRONMENTS: ABYSSAL PLAINS, CONTINENTAL MARGINS
As shown in Figure 1B, some provinces of the deep subsurface biosphere, particularly those characterized by deep sedimentation, experience more restricted fluid flux and therefore have potentially limited contact between hosts of different environment niches. Yet even in these regions, viruses may alter host fitness through lysogeny. As stated above, lysogeny appears to become a more predominant lifestyle in regions with suboptimal conditions, such as low nutrient abundance, low host replication rates, or low host contact rates. Some sedimented regions of the deep subsurface are characterized by particularly low cell abundance, such as in sediments within oligotrophic gyres (D'Hondt et al., 2009), or are exposed to extremely limited organic matter, nutrient, or free energy availability (Schrenk et al., 2010). Moreover, the lack of mobility in the sediment matrix may encourage a lysogenic lifestyle as well, as viruses may have difficulty in contacting a new host within a sediment matrix, especially in regions with low cell abundance. Therefore, the lysogenic lifestyle is likely to be much more common among viruses in these regions, and the archaeal and bacterial inhabitants of these regions have an even higher likelihood of expressing fitness factors encoded by prophage. Initial studies of prophage in the deep subsurface biosphere seem to support this case (Engelhardt et al., 2011). These prophage may aid in host survivability. For example, in deeply buried marine sediments with limited organic carbon or other nutrients, viruses may carry genes to aid in scavenging these compounds or in providing secondary metabolisms to take advantage of alternative energy or nutrient sources. However, more work needs to be done on sequencing viral or cellular isolates from these regions to gain further support for this hypothesis.
One final viral influence that is likely to impact all provinces of the deep subsurface biosphere is the input of viruses from surface waters. Marine sediments receive large inputs of allochthonous material daily, much of it bearing particle-associated microbes that sink through the water column. If these microbes carry prophage or lytic viruses in the process of replicating, these viruses could potentially encounter a different host in its new sediment-bound habitat. These viruses, delivered from the upper water column, could then deliver genes from more pelagic habitats to the deep subsurface. One might even expect induction of prophage to occur more frequently in sinking microbes as they are exposed to increased pressures, which has been found to induce prophage in E. coli (Aertsen et al., 2004). In this sense, viruses may serve as a highway for gene exchange between the surface marine realm and the deep subsurface biosphere.
CONCLUSION
The presence of lysogeny-related genes as well as potentially advantageous cellular functional genes in the virome of a marine, diffuse flow hydrothermal vent bolster the hypothesis that viruses modify the genetic landscape of the hydrothermal habitat, both through expression of prophage genes during lysogenic infection and through the process of transduction. Indeed, this habitatas well as several others in the deep subsurface biosphere -is uniquely poised for fostering an almost symbiotic relationship between viruses and their hosts. As mentioned above, deep sea microbes have higher numbers of integrated prophage and transposable elements in their genomes Vezzi et al., 2005;Ivars-Martinez et al., 2008), and lysogeny appears to be favored over lysis in the extreme regions of the marine realm (Paul, 2008). In regions of the deep subsurface biosphere with low host contact rates, low nutrient flux, or rapidly changing conditions, expression by latent prophage or viral transduction may play a crucial role in bolstering host fitness.
More work certainly remains to be done to determine the scope of the viral impact on the deep subsurface biosphere. Sequencing the genomes of archaeal and bacterial isolates from various regions of the deep subsurface can indicate what prophage elements are present. Viral metagenomics can indicate what types of genes are harbored by viruses, and therefore can indicate which genes may either be transferred into host genomes or expressed during the prophage stage. Thus far, the evidence suggests that viruses play a crucial role in enhancing the fitness of their hosts by modifying their genetic content. They may help their hosts adapt to the unique challenges of the various habitats of the deep subsurface biosphere, acting as a genetic repository for new, adaptive functions. Viruses are emerging as a profound evolutionary force whose impact we have yet to fully assess, particularly in the realms of the deep. | 2016-05-12T22:15:10.714Z | 2011-11-09T00:00:00.000 | {
"year": 2011,
"sha1": "f294c76d826042a04878395510c80eeb49baec34",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fmicb.2011.00219/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "f294c76d826042a04878395510c80eeb49baec34",
"s2fieldsofstudy": [
"Environmental Science",
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
52267419 | pes2o/s2orc | v3-fos-license | Zoledronic Acid and Leuprorelin Acetate , Alone or in Combination , Similarly Reduce Proliferation and Migration of Prostate Cancer Cells In Vitro
We studied the effects of zoledronic acid (ZA) and leuprorelin acetate (LA) separately and combined on a human cell line derived from a bone metastasis of prostatic adenocarcinoma (PC3). In particular, we focused on the effects that drugs given singularly or in association may play on tumor evolution and metastatization. Cell proliferation, 2Dand 3D-migration were studied in basal conditions and under attractive stimuli exerted by bone marrow mesenchymal stem cells (BM-MSC). Either drug decreased PC3 proliferation, though ZA was much more cytotoxic than LA. However, LA cytotoxic concentrations are higher than those usually reached in vivo. Subtoxic concentration of either drug inhibited migration especially under BM-MSC medium stimuli via an Akt-dependent mechanism. The capability of either drug to inhibit cellular migration is in line with their well-known effect in limiting metastatization. Intriguingly, no additive effect on the antiproliferative activity or in hindering migration is observed when the drugs are administered concomitantly, compared to the effects of each drug alone.
Introduction
GnRH receptors are present both in prostatic malignant tissue and carcinoma derived cell lines, suggesting that GnRH analogues such as leuprorelin acetate (LA) [28], normally used to treat advanced prostate cancer (PCa), can exert a local action [10].An inhibition of proliferative activity in presence of GnRH agonists has been found in androgen independent and androgen dependent prostatic cancer derived cell lines [23] and in epithelial cells isolated from prostatic tumors [2].
Moreover, LA antagonizes the metastatic behavior of both androgen-independent cell lines, DU145 and PC3, by decreasing their capacity to invade a reconstituted matrix and to migrate in response to haptotactic stimuli [5].Bone is the preferential site for PCa metastases.Bone metastases likely depend on crosstalk between metastatic cells, bone matrix, osteoblasts and osteoclasts, and cellular components of the bone marrow microenvironment.Among these, bone marrow mesenchymal stem cells (BM-MSCs) play a paramount role in the so-called metastatic niche [14,22].In particular, BM-MSC paracrine factors may allow tumor cells to access cellular niches favoring tumor cell survival and growth [14,22].
Administration of IV bisphosphonates, such as zoledronic acid (ZA), can prevent bone loss and modulate tumor cell migration and invasion [11] in breast and prostate cancer bone metastases in patients treated with GnRH agonists [20].
The beneficial effects of bisphosphonates in PCa patients are attributed to their inhibitory activity on osteoclast-mediated bone resorption, which is markedly increased in lytic and in blastic metastatic bone disease [9].They likely act on two essential stages of metastatic process: (1) tumor cell adhesion to bone extracellular matrices and (2) invasion through extracellular matrix, via modulation of cell adhesion molecules and extracellular matrix interacting molecules [4].
In addition to their antiosteoclast effects, bisphosphonates directly inhibit in vitro the growth of breast [6], pancreas [30], and prostate cancer cells, in a concentration-and time-dependent manner [17].
Despite the absence, to our knowledge, of in vitro and achieved preclinical studies investigating any potential interaction in the coadministration of GnRH analogues and bisphosphonates, this association is considered the gold standard of metastatic PCa therapy.Particularly, pharmacodynamic drug interactions could be the result of overlapping mechanisms of action.Therefore, we wondered whether drug interactions can positively or negatively modify both drug effect and/or toxicity.A better understanding of these potential interplays could be used to increase the therapeutic effect of a combination regimen or to minimize its toxicity.We, therefore, examined the antiproliferative and antimigratory effect of a GnRH analogue, LA, and of a bisphosphonate, ZA, alone or in combination, on PC3 cells in vitro, in basal and stimulated conditions, obtained by exposing PC3 cells to the medium conditioned by BM-MSCs.Moreover, we analyzed the role of these two drugs on Akt phosphorylation, which is one of the most notable mechanisms involved in cell proliferation and migration [16].PC3 cells, isolated from bone metastases of PCa, were chosen because they mimic the castration-resistant phase of PCa patients treated with both GnRH analogues and bisphosphonates [15] and because they are thoroughly recognized as a representative of a highly metastatic PCa cell line, on account of, among other aspects, their lack of AC-tubulin [29] and increased levels of β1 integrins and integrin-induced autophosphorylation of FAK [18].PC3 cells are frequently used to study PCa bone metastasis both in vivo and in vitro [12,13].
With this in vitro study, besides confirming the direct cytostatic effect of either drugs (ZA or LA), we also show, for the first time, that both drugs limit cell migration induced by BM-MSCs and that there is no additive effect when these drugs are used concomitantly.Therefore, we suggest that it is worthwhile to consider a therapeutic regimen in which the two classes of drugs are included in early stages when there are not yet signs of bone metastases in patients with PCa.
Bone marrow mesenchymal stem cells (BM-MSC) isolation and production of BM-MSC conditioned medium
Bone marrow cells were harvested from femurs of adult rats as detailed in the work of Raimondo et al. [25].Animals received care in accordance with the Italian law (DL-116, 27 January 1992), which complies with the Guide for the Care and Use of Laboratory Animals by the US National Research Council.
We previously demonstrated that BM-MSCs isolated with this procedure are CD90 + and CD34/CD45 -and that under opportune stimuli they can differentiate [7,25].We also previously reported that BM-MSCs stimulate PCa cell migration [22].BM-MSCs were therefore included in the present study according to the described protocols.
BM-MSC conditioned medium was collected after three days of culture, centrifuged to eliminate cells and cellular debris, and freshly used or frozen [22].
Proliferation assay
PC3 cells were seeded into flat-bottomed 96-well microplates (1,000/100 µL culture medium/well) and allowed to attach overnight in complete medium before drugs addition.Drugs were added to culture medium, alone or in combination, testing various concentrations from 2.5 µM to 50 µM (ZA) [19] and 0.5 µM to 100 µM (LA) [23] for 24-96 h, according to protocols.
2D migration assay-wound healing
PC3 cells were seeded in a 6-well plate at 30,000 cells/well.When confluent, we performed a cross "wound" in each well with a p1000 tip [8], then wells were washed thrice with PBS and medium replaced without (control) or with single (ZA 5 µM and LA 100 µM) or combined drugs (ZA 5 µM plus LA 100 µM).
We photographed "wounds" on time-laps every 45 min to highlight migration until a maximum of 6 h and, as a final time point, at 24 h.
Experiments were repeated three times and every time five different points for each experimental condition were considered.Images were analyzed using ImageJ software: migration was evaluated considering the reduction of wound area, and then compared to controls.
3D migration assay
The transwell migration assay was used to measure the three-dimensional movements of cells.Migration assays were performed in transwells (BD Falcon cell culture inserts incorporating polyethylene terephthalate (PET) membranes with 8.0 µM pores, 1 × 10 5 pores/cm 2 ).
When assessing spontaneous migration, culture medium was RPMI.When assessing migration in stimulated conditions, control medium was αMEM.When tests were performed in the presence of Wortmannine, cells were preincubated for 30 min at 37 °C with 100 nM.
Cells (10 5 ) resuspended in 200 µL of culture medium containing 2% FBS were seeded in the upper chamber of a transwell.The lower chamber (a 12-well plate) was filled with fresh or conditioned culture medium containing 2% FBS and placed in the incubator.After 6 h, cells were treated detailed by Gambarotta et al. [8].Wells were photographed using a BRESSER MikroCam 3 Mpx camera, with an optical microscope (Leica DC 100) at 100×.Five pictures were randomly chosen per well, and used to count the migrated cells with ImageJ software using cell-counter plug-in.Results from different experiments (performed at least three times in duplicate) were expressed as mean ± SE.
In order to avoid any cytotoxic effect of potentially confounding migration results, we performed a cytotoxicity test as previously described at the same time and same conditions of every migration test.
Western blotting
Since Akt phosphorylation plays a role of paramount importance in cell migration [16], we tested the effects of ZA or LA on the phosphorylation levels of Akt (expressed as pAkt/Akt ratio) in the presence of studied drugs and in the presence of Wortmannine (100 nM), used as a positive control.
Cells were seeded in 10 cm diameter Petri dishes, cultured until subconfluence, when drugs were added (ZA 5 µM and LA 100 µM).After 6 h incubation, cells were collected and treated as detailed by Mognetti et al. [22] and immunoblotted according to Penna et al. [24].In parallel proliferation tests-as previously described-performed in the same culture conditions in order to exclude any confounding element, no significant cytotoxicity was appreciable at the chosen concentration after 6 h incubation.
Phosphorylation levels of Akt were expressed as ratio pAkt/Akt.All data were expressed as percentage modification relative to control conditions.
Statistics
Statistical analyses were performed by one-way or two-way ANOVA, and P < .05 was considered significant.If not differently specified, data are expressed as percentage mean ± standard deviation compared to control.
Inhibition of proliferative activity
ZA strongly inhibited cell proliferation (Figure 1(a)); cell growth was inhibited after 48 h already by 20 µM ZA, and even more strikingly after 72 h culture.Yet, no significant effect on cell proliferation was induced by LA at any concentration tested in the first 72 h (Figure 1(b)).significant inhibition was induced by concentrations greater than 10 µM only after 96 h treatment (Figure 1(b)).
The simultaneous presence of the two drugs neither increased nor decreased the effect of the single drugs at any considered time point.An example is reported in Figure 2, where a constant amount of LA (100 µM, i.e., the highest concentration tested) was added to increasing concentrations of ZA.As in Figure 1, LA (100 µM) alone did not induce appreciable variation of cell proliferation (triangle in Figure 2).
To further analyze drug interactions, the concentrations of LA (from 1 µM to 100 µM) were increased in the presence of 5 µM ZA, and cells were grown for 72 h in the presence of the two drugs.Also in this case no significant variation of cell growth was observed after the addition of increasing concentrations of LA (data not shown).
Inhibition of migratory activity
Two-dimensional migration was inhibited both by ZA and LA, with significant effects already visible after 1.5 h (Figure 3).The simultaneous presence of the two drugs did not relevantly change the cell behavior, except between 2 and 3 h incubation, when it significantly quenched LA effects.On the other hand, no significant difference was evidenced between the drug combination and ZA alone, or between the two single drugs.The same trend is respected after 24 h incubation, when the three conditions are significantly different from control but not different between each other.
Since no difference in the rate of migration was evidenced when experiments were performed in RPMI or αMEM, we pooled data and globally referred to them as control.
Three-dimensional migration was significantly inhibited after 3 h (data not shown) and 6 h (Figure 4) by ZA and by LA, both in basal and in stimulated conditions.Again, the concurrent presence of the two drugs did not increase their inhibitory effect (even, it quenched LA effect).
In the same conditions, we measured Akt phosphorylation and revealed that both ZA and LA decrease it in basal and stimulated conditions.
Inhibition of Akt phosphorylation and that of 3D migration are not proportional (no direct correlation could be established), nevertheless the trend is clearly the same.Importantly, 3D migration rate was decreased by the inhibition of Akt phosphorylation (100 nM Wortmannine), both in basal and stimulated conditions.
Discussion
A major source of morbidity in patient with prostate cancer is the widespread bone metastases, which often grow in a castration-resistant manner and still are a major topic in the research for novel treatment strategies.Monotypic in vitro cultures are commonly employed to study drug effects on cell proliferation and migration.Cell lines are still the main experimental model used in many publications, despite the fact that animal models have become widely available in recent years.PC3 cells are androgen independent, proliferate normally in androgen deprived media in vitro and possess high migratory capability.In addition, PC3 xenograft tumors proliferate rapidly and are quite invasive.For those reasons, PC3 cells are thoroughly recognized as an appropriate model for the of the more aggressive forms of PCa.
The use of complementary assays, such as wound healing and transwell motility assays, is commonly used to assess in vitro metastatic potentials [31].Moreover, our coculture system permits to study the drug effects on the migration of PCa cells in conditions that more resemble the in vivo tumor microenvironment, by providing the tumor cells with contextual signals.This model has already been validated [22] and proved to be reliable in reproducing the chemotactic effect exerted by BM-derived cells on PCa.
This in vitro study strongly supports that ZA has cytotoxic effects on a PCa cell line derived from a bone metastases, an effect that is much less evident for LA.Yet, either drug limits cell migration in a similar extent.Intriguingly, no relevant additive effects can be appreciated between the two drugs in affecting cell migration and proliferation.The lack of additive effects as well as the similar effects on Akt phosphorylation suggests that diminution of the effect of every single drug is highly unlikely in vivo.
A direct cytotoxicity of ZA has been already described [17], though never in combination with LA.In our hands, ZA toxicity was already significant after 48 h culture at 20 µM, regardless of LA cotreatment.In a previous study [19], the toxicity of ZA was observed at lower concentrations probably because of different experimental conditions.It is of note that, in vivo, the effective local concentrations of bisphosphonates at sites of active bone resorption are much higher than serum levels and may reach up to 10 −3 M in the resorption lacunae [27].It is possible therefore to infer that in treated patients the close association of cancer cells with actively resorbing osteoclasts and the bone matrix will result in exposure of tumor cells to bisphosphonates concentrations sufficient to induce their apoptosis [17].
It is much more difficult, on the other hand, to expect, in vivo, a direct cytotoxicity attributable to LA, since concentrations needed, as defined by our study and by other groups [5,23], cannot be reached with therapeutical dosages.
When used at subtoxic concentration, both ZA and LA significantly inhibit 2D migration, already after 1.5 h.This property is even more striking when cells are exposed to a chemotactic stimulus.Intriguingly, this phenomenon is probably very close to what happens in vivo, where bone microenvironment strongly influences cancer cells motility [3,32].
Caraglia et al. [1] described a decrease in the activation of PI3K/Akt pathway following treatment with ZA, suggesting that it may act upstream of the metastasis, by inhibiting PCa cell migration towards metastatic sites through interference with this pathway mostly implicated in cell migration [16].Our study concurs with that of Caraglia et al. by suggesting that inhibition of migration provoked by ZA and LA might be mediated by inhibition of Akt phosphorylation, which, on the other hand, is increased when cells are exposed to a chemotactic stimulus fostering their motility.
We conclude that ZA is more cytotoxic than LA in vitro.However, due to the effective dosage of the two drugs, we suggest that such an effect is likely, in vivo, for ZA only.Nevertheless, both compounds strongly reduce basal and BM-MSC stimulated cell motility in vitro.Therefore, they are likely implicated in reduction of PCa cells motility, with great fall out on bone metastases formation.This antitumoral activity is in line with the recently demonstrated beneficial effects of these two compounds in PCa patients with bone metastases [26].
The clinical implications might be the possible increase in GnRH analogues dosage in castration-resistant patients, to increase its antiproliferative effect, and an early use of ZA before the appearance of the metastasis, with the aim of inhibiting PCa cell proliferation, promoting apoptosis and preventing cell migration.
Importantly, when these two drugs are used in association we do not observe unfavorable interactions between them on the studied parameters.However, we cannot rule out that association treatment regimen that enhances or inhibits the effect of the two drugs exists in vivo.The experience with this association has scarcely been studied in vivo under a pharmacological point of view; therefore studies such as ours are needed.
Figure 2 :
Figure 2: Proliferation assay.Effect of the simultaneous administration of LA (100 µM, constant concentration) plus ZA (2.5 to 40 µM, dotted line) on PC3 proliferation compared to cytotoxicity of 100 µM LA alone (triangle) and ZA alone (2.5 to 40 µM, black line).Results are expressed as percentage of control ±SE.
Figure 3 :
Figure 3: Two-dimensional migration.The wound healing was quantified every 45 min in time laps up to 6 h (a) and at 24 h (b), and is displayed as basal conditions (CTRL) or as the effect of 5 µM ZA, 100 µM LA or a combination of both drugs (5 µM ZA + 100 µM LA).Data are expressed as relative to control conditions ±SE.Experiments were performed three times in duplicate.P values are reported in the table; n.s.means difference being not statistically significant. | 2018-09-12T19:58:51.481Z | 2014-09-01T00:00:00.000 | {
"year": 2014,
"sha1": "2a245d0996c78200d1577134acc6ebddb4e1f495",
"oa_license": "CCBY",
"oa_url": "https://iris.unito.it/bitstream/2318/1552644/1/Mognetti%202014.pdf",
"oa_status": "GREEN",
"pdf_src": "Anansi",
"pdf_hash": "2a245d0996c78200d1577134acc6ebddb4e1f495",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Materials Science"
]
} |
54457019 | pes2o/s2orc | v3-fos-license | Regmex: a statistical tool for exploring motifs in ranked sequence lists from genomics experiments
Background Motif analysis methods have long been central for studying biological function of nucleotide sequences. Functional genomics experiments extend their potential. They typically generate sequence lists ranked by an experimentally acquired functional property such as gene expression or protein binding affinity. Current motif discovery tools suffer from limitations in searching large motif spaces, and thus more complex motifs may not be included. There is thus a need for motif analysis methods that are tailored for analyzing specific complex motifs motivated by biological questions and hypotheses rather than acting as a screen based motif finding tool. Methods We present Regmex (REGular expression Motif EXplorer), which offers several methods to identify overrepresented motifs in ranked lists of sequences. Regmex uses regular expressions to define motifs or families of motifs and embedded Markov models to calculate exact p-values for motif observations in sequences. Biases in motif distributions across ranked sequence lists are evaluated using random walks, Brownian bridges, or modified rank based statistics. A modular setup and fast analytic p value evaluations make Regmex applicable to diverse and potentially large-scale motif analysis problems. Results We demonstrate use cases of combined motifs on simulated data and on expression data from micro RNA transfection experiments. We confirm previously obtained results and demonstrate the usability of Regmex to test a specific hypothesis about the relative location of microRNA seed sites and U-rich motifs. We further compare the tool with an existing motif discovery tool and show increased sensitivity. Conclusions Regmex is a useful and flexible tool to analyze motif hypotheses that relates to large data sets in functional genomics. The method is available as an R package (https://github.com/muhligs/regmex). Electronic supplementary material The online version of this article (10.1186/s13015-018-0135-2) contains supplementary material, which is available to authorized users.
Modified Rank Sum Statistic
Let s 1 s 2 , . . . , s N be a list of sequences ranked according to an experimental setting, and let n i denote the number of observed motifs in s i . Under the null model, we assume n i ∼ po(λ i ), with λ i = − ln(1 − p), where p is the probability of observing at least one motif in the sequence. This follows from the probability mass function of the the Poisson distribution If we think of motif occurrences as a Poisson process, where our "time axis" is composed of consecutive intervals of length λ i ordered according to the experimental rank, motif occurrences are now, under the null hypothesis, uniformly distributed on the interval [0, λ.], where λ. = N i=1 λ i . We now calculate a score rm, corresponding to the mid point of the interval (sequence) in which a motif was observed.
We associate the score with motif occurrences in the sequence list. Under the null hypothesis, the probability of observing a motif in a sequence is proportional to the interval length, and thus the expectation is that motif scores are uniformly distributed across the whole interval [0, λ.]. Under the null model, the score for motif occurrences is thus normally distributed with mean λ./2 and variance λ. 2 /12. We calculate the test statistic
Brownian Bridge Method
This method is a re-implementation of the method developed by Jacobsen et al. [1] and recently implemented in cWords [2]. Our implementation differs in the calculation of the sequence dependent motif p-values. The method calculates the max value D of a running sum of mean adjusted log scores of the ranked sequence dependent p-values p i (SSP i ) where ls i = − ln(p i + α) and α is a score dampening factor of 10 −5 .ls is the mean of the log scores.
The running sum has the form of a bridge (starting and ending in 0), and the maximum value is compared to the theoretical distribution of the absolute maximum M of a Brownian bridge under the null model [3] where n is the number of sequences in the sequence list.
Random Walk Statistics
The random walk (RW) method is inspired by the way RW theory is used in the BLAST algorithm to estimate significance of observed homologies between sequences [4]. In that case matches and mismatches become steps in a walk. Here, it is the list of sequence specific p-values (SSPs) for a motif that are transformed into steps in a walk. The maximum value of the walk is compared to the probability distribution of maxima under the null model. Under the null model, sequence specific p-values (SSPs) are assumed to be uniformly distributed between 0 and 1. This situation is expected for (indefinitely) long random sequences. This condition, however, does not hold in general. For shorter sequences and longer motifs, the SSPs tend to discretize to small intervals specific for the number of observed motifs (see Figure S2). In particular, the p-value for observing "0 or more" motif occurrences is always 1. However, for random sequences, the cumulative distribution of discrete SSPs approximates that of the uniform distribution in the SSP values [5] (see Figure S2). Because the discrete SSPs depend highly on both the sequence and the motif, we normalize for the discrete effect. This is done by drawing p-values from the interval between the probability of observing n obs or more and n obs + 1 or more motifs. E.g. for a 6-mer not observed in a 1000 bases random sequence, we draw between p(n obs ≥ 0) = 1 and p(n obs ≥ 1) = 0.2. In this way the modified SSPs for random motifs in random sequences will follow the uniform distribution as required under the null model.
The modified p-values are now transformed into scores in a walk according to a scoring scheme based on the expected motif density when the motif is enriched (foreground) relative to the background. The scoring scheme ensures a negative drift under the null model by taking steps defined by the likelihood ratio: Here, , f g and bg are expected foreground (default = 0.2) and background (default = 0.05) motif densities in the sequence list.
These default values can be changed by the user.
We are interested in the distribution of maximum values that a random walk visits before absorption, under the null model. The random walk is a particular case of a discrete-time Markov chain. We calculate the distribution of interest by building a Markov model with states −1, 0, 1, . . . , M , where M is the observed maximum height of the walk, and assign transition probabilities according to the scoring scheme. From the resulting probability distribution of maxima under the null model, we derive the p-value for the observed maximum height. Alternatively, we can calculate the p-value by using a geometric-like (Gumbel) distribution approximation for the random walk maxima as given in [6]. Figure S3: Quantile-quantile plots for observed and expected p-values of a 6-mer motif run using Regmex in the Brownian Bridge setting. The comparison employed a data set of mir-430 overexpression in zebrafish that was also used in the original presentation of Sylamer. The gene rank was randomly re-sampled and the 10-mer AATGCCCGGT was spiked into the re-sampled sequence rank: A single motif was inserted 100 times randomly among the top 500 ranked sequences and two motifs were inserted 50 times among the first 100 sequences. | 2018-12-09T14:12:32.518Z | 2018-12-01T00:00:00.000 | {
"year": 2018,
"sha1": "692f749553a9a535a78d35c0d0cec40b999b6647",
"oa_license": "CCBY",
"oa_url": "https://almob.biomedcentral.com/track/pdf/10.1186/s13015-018-0135-2",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "692f749553a9a535a78d35c0d0cec40b999b6647",
"s2fieldsofstudy": [
"Biology",
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
252225009 | pes2o/s2orc | v3-fos-license | Emotional-based pedagogy and facilitating EFL learners' perceived flow in online education
Given the fundamental role of emotional intelligence (EI) in learning, especially in virtual learning contexts where individuals experience more stress and anxiety, the need to understand and recognize one's own feelings and the mutual feelings of peers has gained more importance. Flow as the ultimate state in harnessing emotions in the service of performance and learning has been introduced as the main reason for one's willingness to perform activities which are connected to no external motivation. In this regard, the present study was conducted to first introduce a new educational program to enhance the EI level in the English as a foreign language (EFL) online education environment and next to investigate the possibility of raising EFL learners' perceived flow. To achieve these goals, the study recruited a sample of 67 EFL learners who were next divided into experimental (n = 32) and control (n = 35) groups. The experimental group received the EI intervention over 10 weeks and the control group received the ordinary online EFL instruction. Data were collected through EI and flow questionnaires and semi-structured interviews which focused on learners' perception of the EI intervention and signs of enhanced flow. Statistical analysis of the data showed a positive effect of the program on the learners' EI and their perceived flow. The study emphasizes the role of applying positive emotions in making language learners more engaged in online classroom tasks.
Introduction
Along with the global outbreak of COVID-19 and the following social distancing and the physical gap between people as a strategy to decrease the dissemination of the disease, many businesses shifted to online platforms and schools closed their doors. Accordingly, many institutions and educational settings opted for alternative methods of teaching and learning process to cope with this dramatic change and online instruction delivery gained more attention. Such a change in the routine behavior may have profound effects on the learners' perceptions, attitudes, and feelings as they have to not only accept the situation but also adopt themselves and try to continue their routine lives. The imbalance brought .
to life and the difference caused in the type of emotions felt made some learners detached from their paths of meaning and purpose and resulted in a lack of focus on tasks and activities presented in online classes. In addition, the current lockdown was associated with slowing down of time, which is thought to be brought by boredom and sadness (Droit-Volet et al., 2020;Li and Dewaele, 2020;Resnik and Dewaele, 2021). As these changes in individuals' affect are hypothesized to have lasting negative impacts (Chadi et al., 2022;Samji et al., 2022), a number of researchers focused on ways to reduce boredom and improve emotional well-being (e.g., Rankin et al., 2019;Patsali et al., 2020). The literature investigating learners' psychological factors during this lockdown has focused on such factors as learners' mental health (Patsali et al., 2020;Saha et al., 2020;Shigemura et al., 2020), learners' attitude (Unger and Meiran, 2020), emotion regulation (Restubog et al., 2020), boredom (Derakhshan et al., 2021b;Kruk et al., 2022), and anxiety and coping strategies (Hyland et al., 2020;Savitsky et al., 2020). One of the variables related to anxiety and motivation is flow which has been the focus of few number of studies (Dewaele et al., 2022;Wang and Huang, 2022) in educational contexts especially during this pandemic. Flow is a state characterized by the experience in which one is subconsciously immersed in an activity, have active participation in learning tasks with a feeling of pleasure, and nothing but the action would be of importance. So, following previous studies which showed the effectiveness of flow for boosting emotional states during the stressful periods of uncertainty (Rankin et al., 2019), two promising candidates for the successful adaptation to online education could be emotional intelligence and flow. Therefore, applying a mixed-methods design, the present study aimed to investigate the effectiveness of introducing emotional activities to boost the learners' perceived flow.
Review of literature Emotional intelligence
Emotions, and more specifically positive emotions, and their influential role in learning a second language was quite a neglected area of enquiry till the last three decades . However, since the 2010s, following the establishment of the International Association for the Psychology of Language Learning, there was a surge of interest in positive and negative emotions in second/foreign language contexts (Abdolrezapour, 2017a,b;Derakhshan et al., 2021a;Ghanbari and Abdolrezapour, 2021) and various definitions have been proposed for emotions consistent with the researchers' viewpoints. In this article, we will use the definition proposed by Reeve (2009) as "short-lived subjective-physiological-functional-expressive phenomena that orchestrate how we react adaptively to the important events in our lives" (p. 9), which attribute a multidimensional, complex nature to emotions as subjective feelings and social phenomena. And the focus is on academic emotions, defined as emotions that are "directly linked to academic learning, classroom instruction, and achievement" (Pekrun et al., 2002, p. 91).
Currently, the two most widely investigated emotions in the language learning context are foreign language enjoyment (FLE) and foreign language classroom anxiety (FLCA) (Resnik and Dewaele, 2020). In this respect, Dewaele and Alfawzan (2018) attention was drawn to the link between FLE and FLCA and language learners' performance and found that FLE was a better predictor of performance than FLCA, which implies that it is crucial for language teachers to put premium on boosting learners' positive emotions to ensure enhanced acquisition. This being the case and considering the relationship found between FLE, FLCA, and emotional intelligence (Li and Xu, 2019;Resnik and Dewaele, 2020), and the malleability of EI (Abdolrezapour, 2017a,b), the best way to help learners regulate their emotions is through the field of emotional intelligence (Goetz and Bieg, 2016).
The most broadly acknowledged and widely accepted definition of emotional intelligence (EI), considered as a facet of general intelligence, has been proposed by Salovey and Mayer in 1990, who defined emotional intelligence as "the ability to monitor one's own and others' feelings and emotions, to discriminate among them, and to use this information to guide one's thinking, and actions" (Salovey and Mayer, 1990, p. 189). This investigation follows the Goleman's EI framework (2005), according to which EI is composed of five characteristics: (1) knowing one's emotion, (2) managing one's emotion (i.e. handling fear, anxiety), (3) motivating oneself (emotional control, the ability to delay gratification), (4) recognizing emotions in others, and (5) handling relationships.
Considering the positive link between emotional intelligence and learners' academic performance and the social nature of emotions (Reeve, 2009;Ghanbari and Abdolrezapour, 2021), there is a general agreement among scholars that nurturing learners' emotional intelligence allows them to be connected with others and learn in a more effective way and increase their chances of success both in school and future life (Goetz and Bieg, 2016). Accordingly, if class activities are emotionally and personally relevant to students' lives, corresponding to the characteristics of emotional intelligence, they would engage learners' attention, encourage cooperative learning and would lead to deeper learning and more commitment to learning processes (Abdolrezapour and Tavakoli, 2012).
As in each subject and context, learners' emotional intelligence was found to be prominent (e.g., Domitrovich et al., 2017;Corcoran et al., 2018), in an online setting, learners' emotion-related personality traits are influentially engaged. So, researchers' attention was attracted to the contributions of learners' and teachers' emotional intelligence to their . /fpsyg. . performance in a language education context (Buzdar et al., 2016;Alenezi, 2020;Li and Dewaele, 2020;Fraschini and Tao, 2021;Zhao and Song, 2022). In this regard, Derakhshan et al. (2021b) confirmed that EFL teachers and students encounter more difficulties in online classes and their teacher participants pointed to the importance of EI in online educational contexts.
In the same vein, Buzdar et al. (2016) confirmed the large predictive contribution of students' emotional intelligence in explaining variance in their readiness for online education. Fraschini and Tao (2021), acknowledging the need to investigate EFL learners' emotion in a virtual context, pointed to the link between positive emotions and learning achievements with some teacher characteristics. As for the possibility of enhancing learners' EI in a foreign language learning context, Abdolrezapour (2017b) studied the role of emotions in computer-mediated learning and provided evidence that emotions could be successfully exploited in EFL classrooms through a number of activities presented by the language instructor. Following this line of research, and Zhao and Song (2022), who found differences between learners' emotions of face-to-face classes and those of online learning, the current attempt aimed to shed light on the effectiveness of emotional intelligence in an online education context and its relation to the learners' flow.
Flow
Flow is described as "a subjective state that people report when they are completely involved in something to the point of forgetting time, fatigue, and everything else but the activity itself " (Csikszentmihalyi and Rathunde, 1993, p. 59). It refers to a state of optimal experience or peak performance in which the individual has an intense engagement in an activity. This state has been linked to the feeling of enjoyment and pleasure and has been introduced as the main reason for one's willingness to execute actions which are linked to no external motivation; the leading cause of which was known to be an internal locus of control, a personality construct that refers to people's belief about the action-outcome relationship (Rotter, 1966). According to Goleman (1995), flow state is emotional intelligence at its best and it is the ultimate possible state in harnessing the emotions in the service of performance and learning. In such status, one would experience positive emotions aligned with the task at hand. This concept has been found influential in various fields including sports (as mentioned in a review by Swann et al., 2012), gaming (e.g., Bressler and Bodzin, 2016), research activities (Hudock, 2015), and work-related activities (e.g., Nakamura and Csikszentmihalyi, 2009). Khan and Pearce (2015) correlated this state with concentration, intrinsic motivation, and enjoyment to perform actions at hand, which would affect learners' perceived learning (Hung et al., 2015) and satisfaction (Buil et al., 2018). Furthermore, in educational contexts, the state of flow has been regarded as a prerequisite of tasks demanding higher-order thinking, in that it provides a higher level of concentration and focus on the task and various studies pointed to its correlation with learning outcomes (e.g., Everett and Raven, 2015).
Following the pioneering study of Egbert (2003), who pointed that the state of flow involves a particularly intense focus in an activity to the extent that one may even lose selfconsciousness and a track of time, a number of researchers focused on the efficiency of flow in the foreign language context (e.g., Aubrey, 2017; Dewaele and MacIntyre, 2019;Liu and Song, 2021;Dewaele et al., 2022;Zhao and Khan, 2022). In her study, Egbert (2003) maintained that some preconditions must exist for the flow experience to occur including (a) a balance of skills and challenge, (b) intense concentration, (c) clear goals, (d) immediate feedback, (e) a sense of control, and (f) interest. Later, studying the experience of positive flow and anti-flow by 232 Spanish foreign language learners from around the world, Dewaele and MacIntyre (2019) pointed to more experience of positive flow, which was linked to a higher degree of multilingualism, high relative standing in the group, age, and years of study.
A series of empirical studies also examined flow in the e-learning contexts (Rodriguez-Ardura and Meseguer-Artola, 2017; Li et al., 2021;Liu and Song, 2021;Wang and Huang, 2022;Zhao and Khan, 2022) which shows the importance of providing learning programs that raise the students' flow states as a key feature for academic success in the virtual education environment. In this regard, Pearce (2005) investigated the possibility of making e-learning activities motivating and engaging, yet still productive. He applied two different methods to measure flow experiences and pointed to the dynamic nature of the students' flow experiences. More recently, Wang and Huang (2022) developed a 14-item foreign language flow scale in a Chinese context and found that the flow state is composed of three dimensions including enjoyment, boredom, and anxiety. However, the online education, especially during the COVID pandemic, was found to be more accompanied by students' boredom and different variables have been found influential in this regard including the topics assigned, activity types (i.e., repetitiveness and monotony of activities used), excessive teacher control, lack of learner participation, as well as overchallenging tasks (Derakhshan et al., 2021b;Zawodniak et al., 2021;Kruk et al., 2022).
In general, previous studies have all pointed to the importance of flow in various learning contexts. It was found that it would lead to learners' interest and satisfaction in the tasks and their desire to experience more challenging tasks. However, applying tasks that enhance learners' flow had been a neglected area of study in foreign language learning situations. Flow or the sense of optimal state is linked to the peak performance and subsequent happiness which .
points to the higher correlation with affective rather than cognitive aspects of flow-related processes. Thus, following Derakhshan et al. (2021b), who proposed adopting livelier class, more teacher-student interactions, and improved interpersonal relationships to lower students' boredom in EFL online classes, incorporating emotional training programs in this investigation is hypothesized to support enhanced flow.
Emotional intelligence and flow
Flow state is clearly a key construct of positive psychology and a highly emotional experience. As already noted, the flow experience is characterized by a condition in which the individual is completely immersed in actions with a high degree of intrinsic motivation and it has been found related to some positive experiences including well-being, enjoyment, affect, and satisfaction. It is now well-established from a variety of studies that EI and flow are linked (Culver and Yokomoto, 1999;Li and Dewaele, 2020;Wang and Shaheryar, 2020;Rakei et al., 2022). In this part of the study, we wish to discuss a number of flow-related studies dealing with a wide range of concepts associated with different components of emotional intelligence as proposed by Petrides et al. (2006), including well-being, self-control, emotionality, and sociability.
There is a consensus among psychologists that emotional psychological well-being (Csikszentmihalyi, 2014;Rankin et al., 2019;Lynch and Troy, 2021) is positively linked to one's experience of flow. Moreover, Kuhnle et al. (2012) pointed that self-control is a good predictor of flow experiences in eighth graders. This being the case, Rivkin et al. (2018) found that higher levels of flow experiences would enhance self-control demands and would ultimately result in higher well-being. The authors also claimed that the experience of flow revealed higher levels of intrinsic motivation (Rivkin et al., 2018). Finally, as for social skills, Walker (2010) examined whether the social flow was preferred to solitary flow and the results of his survey study revealed that social flow was more enjoyable than solitary flow. The author concluded that students like to do the tasks together rather than alone. In addition, social flow experiences were found to be stimulated by positive collective gatherings (Zumeta et al., 2016), which in turn promoted personal well-being and social cohesion.
Considering the context-dependent nature of flow (Ghasemi et al., 2021), there is still a need to investigate EFL learners' flow in the transition to online education during the COVID-19 pandemic. Moreover, the existing body of research suggested the association between emotional intelligence components and flow in various contexts, including the foreign language setting (Dewaele et al., 2022), and studies on boredom in online classes during the COVID-19 pandemic pointed to the paramount importance of individuals' emotional experiences in second language context Kruk et al., 2022). Nevertheless, the possibility of enhancing EFL learners' experience of flow through EI intervention is unexplored and such relation in an online setting would provide useful insights for the language teachers. Thus, the main objective of this study was to investigate whether EFL learners' engagement with emotional intelligence tasks would be associated with a higher degree of flow. Accordingly, two research questions were posed in the current investigation: 1. Is it possible to nurture learners' emotional intelligence in an online EFL class? 2. What is the effect of using emotional tasks in improving EFL learners' perceived flow?
Methodology Design of the study
This article is an attempt to enhance EFL learners' perceived flow through applying emotional intelligence activities. To do so, the teacher adopted various strategies to encourage social interaction and cooperative learning, motivate learners, and achieve deeper and more permanent learning. Therefore, the study adopted a mixed-method design to assess the effects of two teaching approaches on the learners' perceived flow in an EFL context. To this aim, we had two groups of learners going through different interventions. Figure 1 below shows the different stages of the study.
Participants
A convenience sample of 67 Iranian EFL learners, studying English in a private language institute, participated in the study. They had been enrolled in two different upperintermediate conversation classes and were randomly assigned to experimental (n = 32) and control (n = 35) groups. There were both male and female participants in both groups and they varied in age from 18 to 23 years old (M = 19.43, SD = 2.57). Participants mostly had studied English for about 6 to 8 years. Though the placement tests taken by the institute could provide enough assurance for their homogeneity regarding their general language ability prior to the study, the students were asked to take a Preliminary English Test (PET) and the results of the independent samples t-test confirmed their homogeneity. The participants were informed that participation was entirely voluntary, and were assured of the confidentiality of their responses.
FIGURE
Design of the study.
Instruments
To achieve the purposes of the current study, the following instruments and materials were utilized: on a 7-point Likert scale ranging from "Completely Disagree (number 1)" to "Completely Agree (number 7)." Cooper and Petrides (2010), applying item response theory, confirmed the validity of TEIQue-ASF. Following previous studies in the Iranian EFL context (e.g., Abdolrezapour and Tavakoli, 2012; Ghanbari and Abdolrezapour, 2021) which confirmed the reliability and validity of the scale, the instrument was validated in the particular context of the study. The reliability of the test was found to be high (Cronbach's α = 0.84). c. Egbert's (2003) inventory of learners' flow: To measure learners' state of flow, Egbert's (2003) inventory of learners' flow was used (see Appendix A). The questionnaire required students to reflect on experiences during the preceding task and consisted of 14 items in four dimensions: four items for interest (e.g., "When doing this task, I was totally absorbed in what I was doing."), three items for control (e.g., "This task allowed me to control what I was doing."), three items for focus (e.g., "I would do this task even if it were not required."), and four items for a balance of skills and challenges (e.g., "This task made me curious"). Learners were required to rate themselves on a sevenpoint Likert-type scale with response options ranging from 1 (strongly disagree) to 7 (strongly agree). Four of the questions were reversely scored. The higher score represents a higher level of flow experience and more concentration on learning tasks. In this study, the Cronbach's alpha values of the questionnaire and the four dimensions, i.e., interest, control, focus, and the balance of skills and challenge, were 0.85, 0.78, 0.84, 0.89, and 0.79, respectively. This instrument was supplemented with questions collecting demographic information including the participants' age, gender and selfreport level of language proficiency.
Tasks
As the experience of flow differs in different skills, tasks, and activities (Ghasemi et al., 2021), to ensure higher comparability across the responses provided by learners, they were required to answer the items based on their experience of the language tasks and activities during which they had interactions with their friends. Accordingly, an oral narrative task was used, in which subjects were required to narrate a story based on a sequenced set of picture prompts. The tasks were previously piloted with a group of EFL learners of the same level and the expert colleagues confirmed their suitability for these participants based on Egbert (2003) criteria of the conditions facilitating the occurrence of flow, i.e., the perception of appropriate challenges, clear goals, an interesting task, sufficient time, immediate feedback, a sense of control and a chance to focus.
Semi-structured interviews
Following Egbert (2003, p. 508), who posited that "there is no objective way to measure flow precisely, " in addition to the flow questionnaire, to gain a more in-depth understanding of learners' experiences and feelings, an interview guide was developed addressing key issues related to the experience of flow. A significant advantage of this data collection approach is its suitability for revealing situation-specific factors or cultural characteristics of flow (Swann et al., 2012). Thus, seven EFL learners were then required to answer a number of questions in a semi-structured interview (see Appendix B). The incentive behind having semi-structured form of interview was to have a number of open-ended questions with a general guide with. In this way, we let the participants direct the interview and help address issues that were not previously considered by the authors. The interviews were audio-recorded with the consent of the participant and the recordings were transcribed and analyzed.
The EI intervention
As mentioned before, connecting learning tasks and activities to learners' personal lives would lead to deeper learning. Keeping this is mind and following Egbert (2003) who claimed that "teachers can theoretically facilitate the flow experience for students by developing tasks that might lead to flow" (p. 513), a 10-week intervention was developed for the experimental group which was integrated into the regular syllabus. So that the subjects received the regular English instruction in an hour and then 1-hour EII, whereas the control group received their regular 2-hour English instruction. The EII involved storytelling, reflective activities, group discussions, and experience-sharing. Emotional tasks and activities adopted in the study comprised the following components and were based on the five characteristics of EI proposed by Goleman (1995): 1. Helping learners know and manage their feelings: this characteristic of emotional intelligence gains more importance in online education where many learners find it difficult to cope with the many feelings they experience. Thus, learners were taught to identify their emotions and they were instructed to feel them rather than avoid them. So, the teacher encouraged learners to talk about the emotions they had during the hard times in the online education and their friends were encouraged to listen and give directions to them. 2. Guiding learners on strategies to know others and handle relationships: as most relationships are supposed to be limited in online education, there is an urgent need to work on strategies to maintain healthy relationships with peers and the teacher. Thus, to cultivate friendship among the individuals, they were guided to maintain a sense of self. Then, they were asked to discuss strategies they adopted to both stay connected to old relationships and form new ones in online classes. This strategy was hypothesized to help learners stay in tune with the teacher and the peers' conversations.
3.
Teach learners monitor their progress: to enhance the learners' self-motivation, they were required to do some goalsetting tasks. They were also guided to revise and adjust the goals if they were not able to meet them.
Data collection procedure
First, the participants were pretested on language proficiency level, EI and flow states to ensure the sample homogeneity in terms of the variables under investigation. Next, those in the experimental group went through the EI intervention and those in the control group received the regular class activities with no special focus on emotional concepts and notions. For each session, the teacher had alternate exercises for both groups so that if the primary tasks were not interesting, a different approach or different tasks could be adopted. In addition, to gain a more in-depth understanding of the participants' classroom behavior and of their classroom learning experiences, seven students from the experimental group were chosen to participate in semi-structured interviews. The interviews were in Persian and were conducted by one of the researchers (a Persian native speaker) and lasted around 30-40 min each. Finally, post-tests (i.e., EI and flow measures) were administered to both groups.
Data analysis
Learners' EI and perceived flow were assessed using the quantitative methods and the SPSS software. So that first descriptive statistics were used and the obtained scores were checked in terms of the normality of distribution using measures such as Kurtosis and Skewness. Then, mixed between-within group ANOVAs were performed on the scores obtained from the pretests and posttests to examine the learners' improvement over time. Then, the learners' interviews were qualitatively analyzed to find the common themes within the transcripts. We adopted a three-step grounded-theory analysis (GTA) (Corbin and Strauss, 2008) of the interview responses, including open coding, axial coding, and selective coding. In the open coding stage, the themes expressed by the participants were coded by each researcher and data were organized based on the phrases and ideas that were stated and re-stated in the interviews. In this stage, if a component was discovered which was not in previous studies, the authors had to discuss it and reach an agreement to whether include it as a main category or a subcategory. Then, in the axial coding stage, relationships among the codes were identified and the links between categories and subcategories were checked. Finally, in the selective coding stage, the transcripts and the selected codes were checked several times to select the codes underlying the students' perceived flow. To ensure the reliability of the obtained results, both intra-coder (by each researcher) and inter-coder (by both researchers) reliability checks were done. Moreover, to enhance the accuracy of the data, member checking or participant validation was performed, where participants were asked to check the transcripts of their interviews and correct any misinterpretations of their responses.
Results and discussion
First, to check initial homogeneity of the groups, independent samples t-tests were run. The results of descriptive statistics and t-tests on EI and flow pretests are provided in Table 1. As shown, the results confirmed the comparability of the two groups prior to the intervention.
The impact of the intervention on learners' EI
To find the effect of the intervention on learners' EI, a mixed between-within group ANOVA was conducted on the pre-and posttest scores of the two groups, with the type of condition or treatment (i.e., experimental or control group) as the between-subjects factor and time (pretest and posttest scores) as the within-subjects factor. Table 2 displays the EI posttest descriptive statistics for each group, which point to the better performance of the experimental group on the posttest (Mean difference 24.72). The results of a mixed between-within group ANOVA showed main effects for time, F (1, 65) = 262.177, p < 0.05, the interaction between the time and treatment condition, F (1, 65) = 246.708, p < 0.05 (Table 3), and also the treatment condition, F (1, 65) = 64.796, p < 0.05 (Table 4). These findings provide evidence that the learners improved over time as a result of the treatment condition and also that the two treatment conditions engendered differential effects on the learners' improvement in EI.
Thus, the mixed between-within group ANOVA confirmed the effectiveness of exposing language learners to emotional intelligence intervention on their emotional states. In other words, while a significant difference was observed between the pre-test and posttest EI scores of the experimental group, the EI scores of the control group did not increase significantly from pre-test to post-test conditions.
The impact of the intervention on learners' perceived flow
As mentioned before, this article was an attempt to enhance EFL learners' perceived flow. So, to find a plausible answer to the second research question which aimed to investigate the impact of the EG activities on learners' perceived flow and to study the potential differences between the two groups of participants as well as the changes in the individual members of each group over time, a mixed between-within group ANOVA was run. Once again, the type of treatment was taken as the between-subjects Level of significance is considered as p < 0.05. Level of significance is considered as p < 0.05. factor and time was considered as the within-subjects factor. Learners' levels of perceived flow in the posttest are brought in Table 5, which, again, shows the better performance of the experimental group. The results of a mixed between-within group ANOVA showed main effects for time, F (1, 65) = 22.682, p < 0.001 and the interaction between the time and treatment condition, F (1, 65) = 22.355, p < 0.001 (Table 6); as well as the treatment condition, F (1, 65) = 4.559, p < 0.001 (Table 7). According to the results of this part of the study, the students differed in their flow condition based on the particular group they belonged to. While a significant improvement was found in the flow scores of the students in experimental group, the flow condition of the students in the control group did not differ significantly from pre-test to post-test occasions.
Semi-structured interviews
This part of the article provides the information obtained from interviewing seven EFL learners participating in the experimental intervention. As mentioned in the previous section, the interviews were transcribed and coded applying the GTA. From the in-depth interview data, which aimed to provide insight into the signs of flow experience, several coding themes were identified, including enjoyment, engagement with the tasks, reduced self-consciousness, and intense concentration. Table 8 shows the frequency of the final categories observed in the interview transcript of each single student. It should be pointed that if a student pointed to a particular category several times, it was only counted once.
Here, the analysis of the interviews along with some excerpts from the participants that showed the existence of flow state after completing language activities would be provided. As noted above, the adopted tasks and activities were mainly based on the Goleman (1995) model of emotional intelligence; however, based on the results of semi-structured interviews, these tasks led to the main dimensions of flow including enjoyment, engagement with the tasks, concentration, reduced self-consciousness and absence of time alertness. However, as stated above, some components were taken as the subcategory, rather than a new category to conform with the literature. Thus, relationships among the codes were identified and absence of time alertness was considered as the sub-category of reduced self-consciousness. Moreover, participants (either implicitly or .
Representative extracts Frequency Percentage
Enjoyment This enjoyable experience was one of a kind.
The task was interesting and manageable.
100%
Engagement with the tasks I was willing to do the activity several times.
I knew what I was expected to learn.
86%
Loss of self-consciousness I wasn't worried about making mistakes.
I was lost in the activity. I felt like time flew.
71%
Intense concentration I was totally absorbed in the tasks. 6 86% explicitly) referred to the balance of skill and challenge, clear and focused goals, and the feedback provided by the teacher as the main antecedents of their perceived flow. In general, the EI intervention was found to engage the learners' interest as some commented that they liked the goals of the tasks and the feedback received made them sure that they were performing the tasks properly as one commented: I really like to replicate the tasks as the teacher was telling us exactly what we had to do and she said that we should have a goal in listening and speaking tasks. The ongoing feedback that she was giving us made the course much more interesting. We knew whether we are moving toward the goal or we have to revise our performance.
This part was in line with what Egbert (2003) posited, i.e., tasks presented in language classroom should have clear goals to induce learners' flow. Participants also proposed that they were more interested in doing the tasks and experienced loss of selfconsciousness which is reflected in the excerpt below stated by another learner: The activities were connected to our lives so we were more engaged. As for me, I had lower stress and my mind was relaxed and I could perform with much more ease.
Learners also enjoyed the collaborative tasks and the experience of talking about emotions. They liked the degree of challenge they had, which was once again in line with what Egbert (2003) mentioned, i.e., to attain learners' flow, there should be a balance of skills and challenge. An example of such experience can be observed in the following comment provided by one learner: The tasks were neither hard nor easy; sometimes we had some collaborative tasks for which we had to discuss the answer in the WhatsApp group and share the results with other classmates. In such cases and the times when we were .
listening to the stories of our friends, we couldn't perceive the passage of time.
Furthermore, the tasks and activities offered to language learners resulted in their intense concentration; an instance is shown in the following comment: For some tasks, we were given a time to talk together and share our feelings about the things happened to a friend. The thing that I liked about such tasks was that the teacher didn't interrupt us and we could focus on the activity, sometimes we even forgot that we were taking part in a formal class. The tasks were wonderful.
Discussion
As noted above, this study was an attempt to, first, explore the impact of the application of an EI intervention program on the EFL learners' emotional intelligence level, and second, to examine the extent to which this intervention would impact the learners' perceived flow. The results pertaining to the first research question, which concerned the possibility of nurturing learners' emotional intelligence over the 10week program, showed that exposing EFL learners to the intervention positively affected their emotional intelligence level. The results of this part of study confirmed the literature regarding the possibility of nurturing emotional intelligence in various educational settings (Abdolrezapour, 2017a; Ghanbari and Abdolrezapour, 2021) and were in line with the studies which pointed to the predictive contribution of learners' emotional intelligence in explaining variance in their readiness for online education (Buzdar et al., 2016;Abdolrezapour, 2017b;Fraschini and Tao, 2021). Thus, teaching intrapersonal emotional factors such as managing one's emotions, knowing one's emotions and motivating oneself in addition to the interpersonal emotional characteristics such as understanding others' feelings and strategies to handle ones' emotions would be beneficial to language learners in online educational settings.
The second research question addressed the effect of the EI intervention on nurturing EFL learners' flow. The findings provided evidence that the learners improved over time as a result of the treatment condition and also that the two treatment conditions engendered differential effects on the learners' improvement in flow, which confirmed Pearce (2005) claim of the possibility of making e-learning activities motivating and engaging. In line with the L2 literature that pointed to the task type and task engagement as the common boredom-inducing factors in online education (Derakhshan et al., 2021b;Zawodniak et al., 2021;Kruk et al., 2022) who suggested adopting educational programs with improved interpersonal relationships and more teacherstudent interactions, incorporating the EI intervention with the different topics and activities made language learners more willing to participate. Moreover, according to Csikszentmihalyi (1998), when individuals participate in an activity for its own sake, i.e., it is so satisfying that they are inclined to repeat it at higher levels of challenge, their perceived flow would increase. So, it seems that learners' exposure to various emotional tasks and activities made them more interested and eager to participate in activities and this accordingly increased their flow.
Reasons for the difference in learners' perceived flow over a 10-week period were provided in the semi-structured interview sessions with the open-ended questions, which were intended to look into the experimental group's perceptions of the EI intervention and signs of enhanced flow. Our qualitative data showed that participants were more willing to participate in educational tasks and share their feelings and thoughts with their classmates. The intervention applied in this study led to enjoyment, engagement with the tasks, reduced self-consciousness, and intense concentration. So, considering the results obtained in this part of the study and the literature (e.g., Rodriguez-Ardura and Meseguer-Artola, 2017; Dewaele and MacIntyre, 2019;Li et al., 2021;Liu and Song, 2021;Wang and Huang, 2022;Zhao and Khan, 2022) which confirmed the efficiency of flow in the foreign language context, it can be a good strategy to expose language learners to teachable techniques for managing emotions, knowing one's emotions, handling relationships, and motivating oneself.
Conclusion
During traumatic circumstances, such as the COVID-19 pandemic, many individuals might not be in a desired emotional state to focus on educational tasks and activities. In such situations, online educational programs, capable of driving students' interest and effort, lead to higher learning outcomes. Learners who experience flow are more eager to participate actively in learning tasks and have one predetermined learning objective; they set a goal for themselves and feel more pleasure and are much more satisfied. In such a state, they do not consider other thoughts and distractions. Thus, to ensure learning success, it is necessary to develop learning tasks and activities that yield favorable opportunities to all language learners through providing an optimal level of challenge, control, and interest.
Following Goleman (1995), who posited that flow state is resulted from the highest level of emotional intelligence, students' eagerness and performance in online learning can be ameliorated through regulating their emotions and raising their emotional intelligence. And as learners with higher levels of EI have higher self-awareness, self-management, and are more capable of handling social relationships, they can ultimately .
/fpsyg. . perform better in online educational courses that inherently abound with internal distractions (including learners' feelings and interference from homes such as younger siblings and parents), and external ones (such as other students and technical complexities such as difficulties experienced in connecting to the net).
In line with previous studies (Abdolrezapour, 2017a,b;Ghanbari and Abdolrezapour, 2021), this study introduced some emotional activities to provide opportunities for language learners to work on emotional intelligence, and such activities can be adopted in various educational contexts. Generally, focusing on learners' emotional states and designing tasks to discuss such issues in the class will allow learners to see the lighter and more humorous side of things, and to regain some qualities as persistence, motivation, willingness, and cheerfulness that characterize successful learners. In addition, it would increase their inner peace and strength, and in doing so, it helps them be able to control and reduce the stresses that accompany some learning contexts, especially virtual ones. When working on learners' emotional intelligence, they can develop their vision and lower their stress, fear, and disappointment. They can have a much more positive attitude and experience lower levels of negative feelings. Thus, considering the teacher's crucial role in setting the conditions for ensuring positive flow (Dewaele and MacIntyre, 2019), it is suggested that teachers apply various educational tools and different tasks to make learners more engaged and interested and increase their motivation.
As with all other studies, this study has a number of limitations that need to be acknowledged. First, we had a small sample for semi-structured interviews and only from the experimental group, which might delimit the generalizability of the study. Future attempts can strengthen the generalizability by conducting similar research with more in-depth investigation of learners' views from both the control and experimental groups. Also, the generalizability of the obtained results is subject to certain limitations as the study used a convenience sample. Finally, to expand the use of emotional intelligence intervention in various educational settings and especially in online settings, the provision of a more comprehensive educational program can also be a productive venue for future research.
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Ethics statement
The study was reviewed and approved by Research Ethics Committee of the Salman Farsi University of Kazerun. All procedures performed in the study were in accordance with the ethical guidelines of the 1964 Helsinki declaration. In addition, at the onset of the study, our participants were informed that the data of the study will be used only for research purposes and that all data will be used anonymously. The participants provided their written informed consent to participate in this study.
Author contributions
PA prepared the approach and instruments, she was also the major contributor in writing the manuscript. NG was the instructor of the two classes, collected the required data, and interpreted all data. Both authors read, revised, and approved the final manuscript. | 2022-09-15T14:19:57.274Z | 2022-09-14T00:00:00.000 | {
"year": 2022,
"sha1": "df042d082a5a0e4ed86ba7822e815d93705c044a",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Frontier",
"pdf_hash": "df042d082a5a0e4ed86ba7822e815d93705c044a",
"s2fieldsofstudy": [
"Linguistics"
],
"extfieldsofstudy": [
"Medicine"
]
} |
158429281 | pes2o/s2orc | v3-fos-license | Philosophical interpretation of knowledge and information: knowledge value and information diversity in modern communication
The article examines the development of the intellectual potential of modern society. The purpose of the article is to determine the main dimensions and philosophical foundations of knowledge and information in the context of the problems of communication development of society. The methodology of the article are based on the analysis and generalization of scientific-research works of scientists in the field of information and knowledge development of society, cognitive philosophy, communication development of society. The authors define the basic philosophical foundations of modern understanding of the content of knowledge and information in the context of communication social development. It is noted that knowledge and information are the organizing principle of the post-industrial information society. Conclusions about the correlation between cognitive and social content of the concepts of knowledge and information are made.
Introduction
The most important sources of rapid development of social processes include the quantitative growth of information flows and the improvement of communication technologies. The most important driving force behind these changes has been Information (Ursul, 1975, Webster, 2002. Informatization promotes penetration of information and communication technologies in all spheres of human life and activity (Mayer, 2006, Pushkarev, Pushkareva, 2015. Become more and more widespread new, especially information technology (Mayer, 2005). Formed covering the planet of communication and transportation networks, the flows of capital, strengthens the processes of migration (Choi, 2015).
In order to understand the peculiarities of the development of modern society, it is necessary first of all to understand the place and peculiarities of the functioning of knowledge and information.
Knowledge and information is the organizing principle of the post-industrial information society.
The beginning of the information society put the statistical report, which appeared in the late 50-ies of XX century in the United States Commerce Department. It was said that for the first time in the history of the number of employees exceeded the number of production workers. The new state of social development was defined as: -"Postindustrial" (D. Bell, 1973), -"Technological" (J. P. Grant, 1983), -"Programmable" (A. Touraine), -"Superindustrial, The Third Wave" (A. Toffler, 1980), -"Post-Capitalist" (R. Dahrendorf), -"Informational" (Y. Masuda, 1981), -"Knowledge Society" (M. Castells, 2001, P. Drucker, 1993a, 1993b. In the works of D. Bell, A. Toffler and other researchers the prospects of human development were associated with the recognition of the increasing role of theoretical knowledge and information in the modern world (Pushkarev, Pushkareva, 2015).
The technical basis of the post-industrial society is the development of computer technologies and means of communication, radically transforming the various spheres of modern society.
Modern researchers noted that: -Communicational technologies have increased the amount and types of communications; -More communications increase the complexity of society, accelerating social change; -Increased complexity leads to the possibility of proposing the arrival of a new era (Rodríguez, Busco, Flores, 2015).
The modern stage of development is characterized by rapid growth of scientific and technological progress, as well as problems related to the development of material production (Vlasyuk, 2016), the need to move to low-waste technologies and to civilization with a minimum of energy costs .
New social practices, often contributing to a sharp renewal of established social relations and professions, tend to acquire advanced technological forms, become increasingly capacious in intellectual and information content (Knyazev, Buyankina, Zukov, 2017).
Modern researchers are actively exploring the concepts of knowledge society and information society: characterize the current state and direction of the knowledge society, its connection to related ideas of digital economy, e-government, and others, and detail implications for business and other organizations, and for society at large (Phillips, Ching-Ying, Hameed, El Akhdary, 2017). Modern society is actively looking for new approaches to educational strategies in accordance with the accumulated experience and General development trends in the context of globalization (Auer, 2015, Diaz-Fernandez, 2017, Smajs, 2016. Modern researchers are searching for the concepts of sustainable development of society (proposed as a positive vision for a more sustainable society (Heylighen, 2017, Vlasyuk, 2016, so and negative vision (Smajs, 2016).
The researchers consider some philosophical background of the education process, which have a significant impact on its creation until today (deals with some philosophical backgrounds of the process of education, which have a major impact on its creating till today) (Masuda, 1981). The penetration of information technologies in almost all spheres of social life is closely connected with the processes of globalization. Globalization as a manifestation of the modern scientific and technological revolution is reflected in the organic connection of science with production, the transformation of production into a large-scale innovation process, the emergence of a new method of production based on information technologies (Pisonova, 2016). Accordingly, modern society is becoming more dependent on technologies (more dependent on technologies), new opportunities and ways of integrating technologies into the educational process are created every day (Laal, 2013, Ushakova, 2006, Wohlin, Smite, Brede Moe, 2015, Zlate, Enache, 2015. This era is characterized by a special form of social organization in which the latest technologies for obtaining, processing and transmitting information have become a fundamental source of efficiency, productivity and power (Knyazev, Buyankina, Zukov, 2017).
Socio-political problems as a result of uneven and suboptimal development of society today have reached a new aggravation.
If we want to understand what is happening to man and society today, we must answer the questions of what modern knowledge and information is in terms of their essence, meaning and structure.
The purpose of the article is to determine the main dimensions and philosophical foundations of knowledge and information in the context of the problems of communication development of society.
Materials and Methods
The methodology of the article are based on the analysis and generalization of scientific-research works of scientists in the field of information and knowledge development of society, cognitive philosophy, communication development of society. Results
The Information Concepts
Let's start with a broader concept -information. Information today in most cases in relation to any individual is not only sufficient, but also excessive.
The total amount of information, knowledge today is growing avalanche and can be described as an information explosion.
The spiritual crisis of man-made civilization against the background of the accelerating development of scientific thought and information explosion, the consequences of which are still difficult to predict, is becoming more acute.
Information is becoming increasingly uncertain. This refers to its ability to act not only positively, but also negatively (for example, the risk of terrorist threats increases).
Information (from lat. Information -explanation, presentation), originally referred to the information transmitted by people orally, in writing or in another way (using conventional signals, technical means, etc.).
From the middle of the XX century -is a general scientific concept, including the exchange of information between people, man and machine, automatic and automatic; the exchange of signals in the animal and plant world; one of the basic concepts of cybernetics.
Within the boundaries of the system-cybernetic approach information is viewed in the context of three fundamental aspects of any cybernetic system (Tabl.): Information related to the implementation in the system of a certain set of processes of reflection of the outside world and the internal environment of the system by collecting, storing and processing the respective signals; management, taking into account the processes of functioning of the system, the direction of its movement under the influence of the information received and the degree of achievement of their goals; organizational, which characterizes the structure and the degree of perfection of the control system itself in terms of its reliability, survivability, completeness of functions, perfection of structure and cost-effectiveness of management processes in the system (Gritsanov, 1998): 274).
Table. Functions and content of information in the context of a systematic approach
In modern times, information begins to act not just as a social memory, but as a working tool, as a means of decision.
The phenomenon of social information at the beginning of the third Millennium is extremely complex, with many specific characteristics.
Social information in its quantity has become almost immense, infinite in relation to an individual subject. On quality-so versatile that there are difficulties with its systematization.
The Knowledge Concepts
Understanding the specifics of the development of any social processes is impossible without understanding the place and role of knowledge in modern information civilization.
Accordingly, it is further necessary to determine what we mean by knowledge and how knowledge and information are correlated.
Currently, many scientists have dramatically changed their attitude to the role of knowledge. Publications have shown a steady tendency to differentiate it in the General information flow and to place it at the centre of the production and consumption of information.
Questions about what is knowledge, what is the meaning of knowledge, what we need to know, why (for what) to know, worried mankind since when it became able to philosophize. At different stages of history, it responded to him differently. Humanity has gone from knowledge (singular) to knowledge (plural).
In the early days of the development of society knowledge was of a General nature. Today knowledge has become deeply specialized by necessity. Previously, such a concept as "a person with knowledge" was not used, but it was said: "an educated, scientific person". In other words, a person of wide erudition, who has sufficient knowledge to conduct a conversation or write on a variety of common topics and at the same time does not know any specific practical activities.
Socrates believed that the purpose of knowledge is self-knowledge and selfdevelopment. The results are to the person. Protagor also argued that the purpose of knowledge-to be able to say what you need and how to. In the middle Ages, the education system included grammar, logic and rhetoric, that is, it was also aimed at developing the ability to speak argumentatively.
In the modern philosophical understanding knowledge is "an objective reality given in the consciousness of man, which in its activity reflects, perfectly reproduces the objective natural connections of the real world" (ibid.).
It arises as a result of interaction of the subject and object on the basis of experience of social and historical practice of acquisition, development, deepening and expansion, improvement and reproduction of information on the relations both in the sphere of life and in the sphere of consciousness. Knowledge is, firstly, abilities, abilities, skills gained through acquaintance or practical use of algorithms of activity; secondly, information that is important for cognitive and practical activities; third, the "special epistemological unit of man's relationship to the real world, corresponding (in the dialectical unity of theory and practice) the operationalization of abstract constructs of consciousness" (Chupahin, 2004: 133).
Scientific revolutions invariably led to constructive changes in the system of knowledge, influencing the ideological development and understanding of the laws of the world. As a result, the press of time, the process of division of labor, especially intensively occurred in recent centuries, split the once holistic knowledge of nature, man and society into a lot of knowledge.
Differentiation and specialization of knowledge is increasing, and specialists of even close areas understand each other less and less.
The very nature of knowledge has been modified: the key provision has been given to knowledge, which in the form of an invention or organizational improvement are involved in the practical processing of resources.
At the same time, the structural and functional characteristics of scientific knowledge and its integration abilities are complicated (Knyazev, Buyankina, Zukov, 2017).
Modern knowledge of mankind is a huge array of information, which is even difficult to imagine at once, and even more so, to comprehend and assess in detail.
Knowledge in modern society is information of practical value, serving to obtain concrete results. What we call knowledge today is constantly proving its importance and being tested in practice. Moreover, the results are manifested outside the person-in society, the economy, in the development of knowledge itself. To obtain significant results in any field requires knowledge of highly specialized. That is why the tradition that started in antiquity, is preserved today in the education system (Pushkarev, Pushkareva, 2016).
Processing of complex information
The history of cognitive science began earlier than what we now call science, because the development of cognitive abilities was a requirement of the practice of human survival in a changing environment (Abrarova, Salikhov, 2016). Any scientific idea trying to explain any side of the reality existing outside the person needs objectification. Under the objectification of knowledge to understand it in the sense of the world. One form of objectification can be visualization. In modern science and technological processes visualization is often used to represent the dynamics of any abstract quantities (physical parameters, economic characteristics, etc.) in the processing of complex information about the structure of objects and systems (Trofimov, 2017).
The relationship between Knowledge Concepts and Information Concepts
The first is the expansion, the total; in this case, knowledge is the entire array of information available to society at a certain time period of its development. And then the term "knowledge" turns out to be, in essence, synonymous with the aggregate information of the society (or any particular social community) (Ushakova, 2006).
The second position is more narrow: knowledge is seen as a certain part of the total information, often having a specific quality, which should be understood as a positive, vital meaning concluded in the relevant part of the information. In the second case, the information of meaningless or destructive pastime, the corresponding types of activity is no longer knowledge, understood as the fundamental basis of social existence (Ushakova, 1998).
At the same time, the development of technological and infrastructural features of information radically change the face of social practices and social relations in General. All this makes a very diverse not only the subject of the interaction of knowledge and information, but also the problem space of the application of this interaction in the socio-economic, humanitarian, scientific and educational spheres of society (Knyazev, Buyankina, Zukov, 2017).
In relation to personality Knowledge is a form of existence and systematization of the results of human cognitive activity. Knowledge is objectified by sign language means. At the same time, knowledge is the result of the process of cognition of reality, which has been confirmed in practice; adequate reflection of objective reality in the human consciousness (ideas, concepts, judgments, theory); it is fixed in the signs of natural and artificial languages.
The core of knowledge forms the General Outlook as an understanding of the world of people, nature and their own life world (ideological aspect), as well as the General practical orientation of the person (praxiological aspect as existential readiness of the person to act in a certain direction). A "knowledge shell" forms blocks of social and professional knowledge of people (professional and pragmatic aspect-for society) .
Conclusions
Information in modern conditions in relation to any individual is not only sufficient, but also redundant.
Modern knowledge of humanity is a vast array of information that is difficult to assess competently; differentiation and specialization of knowledge is increasing.
Information is becoming more and more uncertain, that is, it affects social processes not only positively, but also negatively.
Knowledge has a vital meaning in the relevant part of the information. In this case, the information of meaningless or destructive pastime, the corresponding types of activity is no longer knowledge.
Information is not just a quality of social memory, but also an effective tool, a means of decision-making.
Knowledge in modern society is information of practical value, serving to obtain concrete results. | 2019-05-20T13:05:23.740Z | 2018-01-01T00:00:00.000 | {
"year": 2018,
"sha1": "7bf77cabbee835ca18d68592ffe1dbfdbbaa7650",
"oa_license": null,
"oa_url": "http://www.xlinguae.eu/files/XLinguae3_2018_17.pdf",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "a4d3a9c849b6afc05804e3a977b4e7cd48c00107",
"s2fieldsofstudy": [
"Philosophy"
],
"extfieldsofstudy": [
"Sociology"
]
} |
221763539 | pes2o/s2orc | v3-fos-license | CHESTNUT COVER AUTOMATIC CLASSIFICATION THROUGH LIDAR AND SENTINEL-2 MULTI-TEMPORAL DATA
Chestnut ( Castanea sativa Mill.) managed forests in Galicia (Northwestern Spain) have important cultural, economic and ecosystem values. However, due to rural exodus chestnut stands are being degraded. In order to take restoration and conservation measures knowledge of these forests' location, expanse and stage is needed. The available Spanish official cartography is based on photointerpretation which is inaccurate in terms of chestnut forest location and classification. However, remote sensing has recently been proven to be an effective tool for this purpose. Sentinel 2 multi-temporal classification is recently acquiring importance as a method to classify tree species. This project intends to detect chestnut forests using LiDAR and Sentinel 2 multi-temporal data and to compare these results with those obtained using the official cartography. It also intends to assess how the use of different phenological stages could improve classification results. The results obtained provide an overall accuracy of 76% when a threemonth combination is used: (March, July and September) leaf-off stage, flowering and leaf-on stage. Overlapping of the current map and the official cartography lead to an accuracy and precision increase; highlighting the utility of the presented methodology to acquire knowledge about chestnut forests location. * Corresponding author
INTRODUCTION
Information about forest resource distribution remains essential for policy makers and land management. European chestnut (Castanea sativa Mill.) covers more than 2.5 million ha in Europe of which 98.3 % are managed forests used for wood or fruit production purposes (Conedera et al., 2004). Since the first half of the last century, a decreased importance of this species as a staple food and a progressive depopulation of rural areas has caused the abandonment of chestnut agriculture (Zlatanov et al., 2013), an abandonment which has been accelerated due to pathogens (Baser and Bozoglu, 2020) and invasive species (Battisti et al., 2014). However, in recent decades chestnut tree management has once again become viable and shows great potential as a source of income thanks to research in biological control and genetics (Acquadro et al., 2019;Trapiello et al., 2017). Galician chestnut forests (Northwestern Spain) have been managed for chestnut production for centuries (Fernández et al., 1998). The result is forests with important ecological, ecosystem and cultural values. In recent years, rural abandonment has led to stand degradation and to a loss of knowledge of their distribution. However these stands still have the potential to become an important source of income for rural areas and to keep providing ecosystem, recreational and cultural services (Roces-Díaz et al., 2018). In order to recover the ecosystem, knowledge of its current expanse and stage is needed. The current official source of forest location, distribution, expanse and classification in Spain is the Spanish Forest Map (MFE). The Spanish cartographic official source (MFE) is elaborated through photointerpretation and field work on a 1:25000 scale (Miteco, 2014). It consists of a set of polygons where the type of forest, along with its principal and secondary species composition are indicated. Chestnut forest pure stands are classified as a "castañar" type of forest. However, when mapping chestnut forest stands, the method and resolution used can become a problem leading to surface overestimation or underestimation. This compromises decision making when it comes to chestnut production or conservation measures.
Nowadays, tree species distribution can be acquired using remote sensing techniques. Medium spatial resolution satellites (especially Landsat 8 and Sentinel 2) are the most cost-efficient tool and the current trend in tree-species-classification-related studies (Fassnacht et al., 2016). Of the open-data satellites, Sentinel 2 is the one which obtains images with the highest spatial resolution (10 m to 20 m) and the highest spectral resolution (up to 13 bands). Three of the 13 bands are red edge bands, which are not available from satellites like Landsat, and have been shown to be important in tree species classification (Immitzer et al., 2016). The free availability and the high revisit time (5 days) of these satellites, which allow for the acquisition of multiple images corresponding to different vegetation stages at the same area, has recently led to the incorporation of multitemporal analysis to vegetation classification related studies (Fassnacht et al., 2016). These analyses are especially suitable for boreal and temperate climates where changes in plant phenology throughout the year can help to discriminate between different species.
In addition to satellite images, LiDAR sensors play an important role in land cover classification as well. Thanks to their ability to estimate vegetation height they are a tool that efficiently discriminates classes with spectral similitudes (Sothe et al., 2018). Recently available LiDAR data on National Geographic databases have led to the incorporation of this technology in vegetation studies (Barrett et al., 2016).
This study aims to classify areas covered by chestnut forests and to analyze the potential of using multi-temporal information for this purpose. It also intends to provide a tool which can be used to improve the accuracy of the official cartography based on open-access data. Therefore, comparisons between the official maps and the results obtained from this study will be performed.
MAIN BODY
The study area is in Galicia (Northwestern Spain). Although knowledge of chestnut location and expanse is relevant in the whole Galician region, a small study area was selected to develop the method and study its potential. The selected area was the municipality of Folgoso do Courel, due to its abundance of chestnut managed forests. The municipality covers an area of 193 km 2 and is a mountainous area with small, steep valleys with different aspects. (See Figure 1 The area is covered fundamentally by temperate mixed forest including 237 different species of dicotyleoneous plants (Quercus pyrenaica, Betula alba, Quercus suber, Quercus robur, Fagus sylvatica, Alnus glutinosa, Corylus avellana, Ilex acquifolium, Juglans regia, Prunus avium, Salix sp., Fraxinus sp.,...).The region's climate has a mediterranean influence, therefore species such as Quercus ilex can be found as well. However, Castanea sativa pure stands, due to management, cover an important part of the area. There are plantations of coniferous trees as well, mainly of: Pinus pinaster, Pinus radiata and Pinus sylvestris. (Xunta de Galicia/USC, 2005.) Chestnut phenology in the study area is not homogeneous due to differences in height and aspect. These variations are increased due to the presence of different chestnut cultivars (Pereira-Lorenzo et al., 2001). However, as a general guideline senescence occurs in November (assessed through field work), sprouting from mid to late May according to local forest technicians and blooming from mid-June to late July (Fernández-López et al., 2013).
Data acquisition and analysis
Sentinel 2 is one of the ESA's (European Space Agency) satellites constellations equipped with a multispectral camera. It samples 13 spectral bands, which range from a 10 m to a 60 m spatial resolution. The temporal resolution is up to 5 days. Images covering different phenological stages of the vegetation were used. Initially images from every season and every month were evaluated in order to optimize the multi-temporal analysis. However several months were finally discarded due to the presence of too much clouds, snow and shadows. The LiDAR data were obtained from the Spanish Geographical Information System (IGN). Data is freely available. LiDAR point clouds were acquired in 2016 using an airborne laser scanner (ALS). The sensor was a LEICA ALS80 which allowed for a nominal point density of 0.5 points/m 2 . Georeferecing was executed on the ETRS89 georeference system with a Root Mean Square Error (RMSE) of 0.3 m in the horizontal directions and 0.2 m in the vertical directions (Miteco, 2019).
Methodology
The present study methodology is based on an image classification process. However, as the study area has other land covers besides forest, a study area filtering step was performed using LiDAR data. Only areas with a height tall enough to be a forest were selected. The selected height threshold was 10 m, based on analyses of the heights of chestnut areas. This step was performed using a calculated CHM (Canopy height model). To obtain the CHM, ground points were identified on the LiDAR point cloud and for the rest of the points the height above the ground was computed. This allowed for the generation of a normalized point cloud from which the CHM was created. CHM resolution was selected accordingly to match one of the Sentinel images (20 m). Pixels with a height value below 10 m were erased from the study area.
Different Sentinel 2 band combinations were created and visually analyzed to assess differences between seasons. Figure 2 shows the combination of bands 11/8A/4 for the selected months where it is possible to see the variations in radiometry over the course of the seasons, especially between March-May and July-September. A multi-temporal classification approach was chosen as it has been shown that there is a potential for differences in radiometry due to phenology. The algorithm employed was Breiman's Random Forest (Breiman, 2001). The Random Forest is an algorithm that is extensively and successfully employed in other multi-temporal vegetation-classificationrelated studies such as (Hościło and Lewandowska, 2019) and (Persson et al., 2018a). To carry out the multi-temporal study several models based on the different month combinations were created. In order to acquire information about the relevance of including several months in the classification, four models were also created using just one month for contrast.
The Random forest algorithm was provided for training areas. There were two classes of training areas: chestnut and others. Others included the rest of forested areas present in the study area area such as other broadleaf forests and coniferous forests. Training areas were created through field work. Random transects were created on the road-accessible parts of the study area in order to look for pure stands that would be suitable training areas. Care was taken to ensure the transects covered different aspects, heights and slopes. Training area locations were marked with a GPS with centimeter accuracy (GEOMAX Zenith15). Finally, a set of 73 polygons were marked: 48 corresponding to other tree covers (Pine sp., Quercus ilex, Quercus pyrenaica, Betula alba, Alnus glutinosa,…) and 25 corresponding to chestnut forests (Castanea sativa). Bands combinations on different months were created to asses differences on radiometric answer between species and months. Figure 3 is an example of this step results. There it is possible to see band combination 11/8A/4 result on different forests for March image. Colorations result allow to see that there are differences between species radiometric behaviour.
Castanea sativa Quercus pyrenaica
Mixed forest Quercus ilex Pinus sp. Random forest models were applied to the whole study area. A cross-validation was performed in order to explore multitemporal results and to select the most efficient model. For this purpose, 100 random points were selected within the study area. Random points ground truth was obtained from photointerpretation of the PNOA image (PNOA, 2019).
Once the most accurate model was selected, results were crossreferenced with the official cartography of the study area in order to assess the differences.
Results and discussion
Cross-validation of the results acquired by random forest modelling reveals differences between the results obtained when using different months as predictive variables. These results are presented in Table 3. A higher overall accuracy (76%) was obtained when using a combination of three months to create the model. The effectiveness of Sentinel 2 time series has been noted previously in multiple projects (Grabska et al., 2019;Hościło and Lewandowska, 2019;Persson et al., 2018). Previously mentioned studies classified forest stands according to all of the species present in an area using multi-temporal images and they obtained accuracies close to 85% when using the best month combinations. Accuracies obtained in this case study are lower (around 10%). However, it was difficult to find pure stands of other broadleaves in order to have enough training areas. Future studies should focus on increasing field work efforts for this purpose.
Comparing cases where higher and lower Overall Accuracies were obtained, differences accounted for less than a 10%. Larger differences exist in Users and Producers Accuracy. Several of the models do not allow an excess of 50% for these values, which could result in a random prediction of a class. Single month models for July and September periods are the least accurate. However, these months combined with March (Model 5) provide the best result. According to results, March ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume V-3-2020, 2020 XXIV ISPRS Congress (2020 edition) This contribution has been peer-reviewed. The double-blind peer-review was conducted on the basis of the full paper.
https://doi.org/10.5194/isprs-annals-V-3-2020-425-2020 | © Authors 2020. CC BY 4.0 License. and May are important months for the prediction of chestnut presence, revealing the importance of including images from senescence periods. Persson et al. (Persson et al., 2018) also note the importance of the leaf-off period, however they attribute it to images from early senesce which couldn't be included in this study due to cloud and shadow problems. They confirm that adding leaf-on periods helps to improve these results as well. However, in order to perform a better interpretation of the results it would be necessary to obtain phenology information about the rest of the broadleaves present in the study area. Model 5 revealed that chestnut cover in the Municipality of Folgoso do Courel amounts to 2,677 ha. Figure 4 shows the results obtained. The MFE chestnut area estimation greatly differs from the results presented in this study. The MFE polygons of chestnut pure stands cover an area of 3,208 ha. When overlapped, the MFE and Model 5 only match on 42% of the area (1,359 ha). The remaining 58% (916 ha) are classified as other tree species.
ID
On the other hand, there are 402 ha of chestnut according to the MFE that are in fact not covered by forest. According to Model 5, 1,318 ha of chestnut forest are not currently classified by the official cartography. The results of these comparisons were evaluated through field work and photointerpretation. Figure 5 shows an example of an area that is not covered by trees that is included in the MFE chestnut polygons. Figure 6 shows an example of other tree covers that are included in the MFE chestnut polygons. Figure 7 shows an example of an area classified as a chestnut area by Model 5 but that is not included in the MFE chestnut polygons. The comparisons performed suggest that the present methodology based on remote sensing techniques can improve the Spanish official cartography. The potential of remote sensing to improve the currently-available information about Spanish forests has been claimed before by (Gómez et al., 2019), therefore the present is a case study which supports this statement. The methodology appears to be replicable in other study areas, however the presence of different tree species could affect the results.
CONCLUSIONS
Sentinel 2 multi-temporal analysis through random forest algorithms linked with LiDAR data has been shown to be an efficient tool for performing chestnut forest cartography. The main advantages of the proposed methodology are the use of open-access data and the use of an automatic process. However, in some situations dependence on training areas could be an issue. The inclusion of months with different phenological stages helps to increase mapping accuracy, as previous studies have indicated. However, the acquisition of data for the wide range of months needed to carry out a complete analysis of the potential of multi-temporal approach for the detection of chestnut plantations was hindered by the presence of shadows due to the topography and climatic conditions of the study area. Further studies should try to record a wide range of months. Additionally, they should focus on links between the results obtained and the different species phenology in order to draw conclusions about which are the most critical phenological stages to include. On the other hand, chestnut forests in the municipality of Folgoso do Courel were detected and more accurately estimated using new methodologies than by using the official cartography sources, showing that new methods are available that could improve the official cartography. At the same time, a ready-touse product was obtained which can be used as a decisionmaking tool when it comes to chestnut forest management. | 2020-08-06T09:07:46.150Z | 2020-08-03T00:00:00.000 | {
"year": 2020,
"sha1": "6e754e05fc6192f6aeda9e0d318a629c7fe0ee76",
"oa_license": "CCBY",
"oa_url": "https://www.isprs-ann-photogramm-remote-sens-spatial-inf-sci.net/V-3-2020/425/2020/isprs-annals-V-3-2020-425-2020.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "1a397cd5fe5e51a5962c3afc9568529e06278b0a",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Geography"
]
} |
262995634 | pes2o/s2orc | v3-fos-license | Impact of Phantom Size on Low-Energy Virtual Monoenergetic Images of Three Dual-Energy CT Platforms
The purpose of this study was to compare the quality of low-energy virtual monoenergetic images (VMIs) obtained with three Dual-Energy CT (DECT) platforms according to the phantom diameter. Three sections of the Mercury Phantom 4.0 were scanned on two generations of split-filter CTs (SFCT-1st and SFCT-2nd) and on one Dual-source CT (DSCT). The noise power spectrum (NPS), task-based transfer function (TTF), and detectability index (d’) were assessed on VMIs from 40 to 70 keV. The highest noise magnitude values were found with SFCT-1st and noise magnitude was higher with DSCT than with SFCT-2nd for 26 cm (10.2% ± 1.3%) and 31 cm (7.0% ± 2.5%), and the opposite for 36 cm (−4.2% ± 2.5%). The highest average NPS spatial frequencies and TTF values at 50% (f50) values were found with DSCT. For all energy levels, the f50 values were higher with SFCT-2nd than SFCT-1st for 26 cm (3.2% ± 0.4%) and the opposite for 31 cm (−6.9% ± 0.5%) and 36 cm (−5.6% ± 0.7%). The lowest d’ values were found with SFCT-1st. For all energy levels, the d’ values were lower with DSCT than with SFCT-2nd for 26 cm (−6.2% ± 0.7%), similar for 31 cm (−0.3% ± 1.9%) and higher for 36 cm (5.4% ± 2.7%). In conclusion, compared to SFCT-1st, SFCT-2nd exhibited a lower noise magnitude and higher detectability values. Compared with DSCT, SFCT-2nd had a lower noise magnitude and higher detectability for the 26 cm, but the opposite was true for the 36 cm.
Among the various DECT platforms, one manufacturer has developed two platforms for obtaining two-photon spectra during acquisitions: the dual-source CT (DSCT) platform and the split-filter CT (SFCT) platform [1,16].For the DSCT platform, two X-ray tube/detector pairs (95 • offset for 3rd generation) are used to acquire the image datasets.The X-ray tube "A" uses low kVp (70, 80, 90, and 100 kVp) and the other X-ray tube "B" uses high kVp with (Sn150 kVp) or without a tin filter (140 kVp).With this platform, spectral acquisitions can be made with five pairs of kVps, and the choice of these pairs depends on the clinical indication for the CT examination and the patient's morphology.For the SFCT platform, the photon beam is split into two-energy spectra using two filters placed at the outlet of the X-ray tube used.The gold filter (Au) is used for the low-energy spectrum and the tin filter (Sn) for the high-energy spectrum.In the first-generation platform (SFCT-1st), the gold filter is 0.05 mm thick and the tin filter, 0.6 mm.In the second-generation platform (SFCT-2nd), the thickness of the gold filter has been increased to 0.07 mm, and the thickness of the tin filter has been increased to 0.70 mm.For the SFCT-1st, only 120 kVp can be used to acquire the spectral image datasets, whereas 120 and 140 kVp can be used for the SFCT-2nd.
Two studies have highlighted the differences in the spectral performance between SFCT-1st and DSCT [16,17].Using two X-ray tubes provides good spectral separation for DSCT.Otherwise, with SFCT, spectral separation is more limited, due to the overlap between the low-and high-energy spectra generated by the two filters.These differences in the spectral performance translate into greater image noise and poorer lesion detectability on the virtual monoenergetic images (VMIs) at low energy levels [16,17].Only one study has evaluated the impact of modifying the thickness of the two filters on the spectral performance of SFCT-2nd, compared with SFCT-1st [19].The authors demonstrated that the modifications in the thickness to the gold and the tin filters of SFCT-2nd improved spectral separation and also the spectral performance (especially, the lowest values of noise magnitude and the highest values of the detectability indexes) compared to SFCT-1st for the low-keV VMIs.All three studies were carried out using a phantom with a fixed morphology.In addition, the spectral image quality is also influenced by patient morphology and the associated dose level.Euler et al. evaluated the impact of the phantom diameter and dose level on the spectral performance of the DSCT platform [20].However, no similar studies have been conducted on the two generations of SFCT platforms.
The purpose of our study was to compare the spectral performances of three DECT platforms according to the phantom diameter.To achieve this goal, the spectral performances of the three DECT platforms were assessed by carrying out a task-based image quality assessment on the VMIs at low energy levels using the three largest sections of an image quality phantom.
Phantom
The three largest sections (26, 31, and 36 cm diameter) of the Mercury v4.0 phantom (Sun Nuclear, Melbourne, FL, USA) were used to perform the task-based image quality assessment (Figure 1A).These three sections have diameters corresponding to the abdomen of patients with a body mass index of 18, 27, and 36 kg/m 2 , respectively [21].Each section of the phantom is made up of a homogeneous zone with which the Noise Power Spectrum can be calculated (NPS; Figure 1B) and a section with five inserts for calculating the Task-based Transfer Function (TTF), particularly on the iodine insert at 10 mg/mL (Figure 1C). on the clinical indication for the CT examination and the patient's morphology.For the SFCT platform, the photon beam is split into two-energy spectra using two filters placed at the outlet of the X-ray tube used.The gold filter (Au) is used for the low-energy spectrum and the tin filter (Sn) for the high-energy spectrum.In the first-generation platform (SFCT-1st), the gold filter is 0.05 mm thick and the tin filter, 0.6 mm.In the second-generation platform (SFCT-2nd), the thickness of the gold filter has been increased to 0.07 mm, and the thickness of the tin filter has been increased to 0.70 mm.For the SFCT-1st, only 120 kVp can be used to acquire the spectral image datasets, whereas 120 and 140 kVp can be used for the SFCT-2nd.
Two studies have highlighted the differences in the spectral performance between SFCT-1st and DSCT [16,17].Using two X-ray tubes provides good spectral separation for DSCT.Otherwise, with SFCT, spectral separation is more limited, due to the overlap between the low-and high-energy spectra generated by the two filters.These differences in the spectral performance translate into greater image noise and poorer lesion detectability on the virtual monoenergetic images (VMIs) at low energy levels [16,17].Only one study has evaluated the impact of modifying the thickness of the two filters on the spectral performance of SFCT-2nd, compared with SFCT-1st [19].The authors demonstrated that the modifications in the thickness to the gold and the tin filters of SFCT-2nd improved spectral separation and also the spectral performance (especially, the lowest values of noise magnitude and the highest values of the detectability indexes) compared to SFCT-1st for the low-keV VMIs.All three studies were carried out using a phantom with a fixed morphology.In addition, the spectral image quality is also influenced by patient morphology and the associated dose level.Euler et al. evaluated the impact of the phantom diameter and dose level on the spectral performance of the DSCT platform [20].However, no similar studies have been conducted on the two generations of SFCT platforms.
The purpose of our study was to compare the spectral performances of three DECT platforms according to the phantom diameter.To achieve this goal, the spectral performances of the three DECT platforms were assessed by carrying out a task-based image quality assessment on the VMIs at low energy levels using the three largest sections of an image quality phantom.
Phantom
The three largest sections (26, 31, and 36 cm diameter) of the Mercury v4.0 phantom (Sun Nuclear, Melbourne, FL, USA) were used to perform the task-based image quality assessment (Figure 1A).These three sections have diameters corresponding to the abdomen of patients with a body mass index of 18, 27, and 36 kg/m 2 , respectively [21].Each section of the phantom is made up of a homogeneous zone with which the Noise Power Spectrum can be calculated (NPS; Figure 1B) and a section with five inserts for calculating the Task-based Transfer Function (TTF), particularly on the iodine insert at 10 mg/mL (Figure 1C).
For each CT system, the acquisition and reconstruction parameters usually used in clinical practice for abdomen DECT examination were selected (Table 1).For each CT system, the automatic tube current system was disabled and the tube currents were set to obtain a volume CT dose index (CTDI vol ) of 12.0, 9.1, and 6.9 mGy for the 36 cm, 31 cm, and 26 cm diameter phantoms, respectively.These dose levels correspond to those usually used on these three CT systems for DECT abdomen-pelvis examinations, according to the patient's morphology.Each acquisition was repeated 10 times.The raw data were reconstructed using Level 3 of the iterative reconstruction algorithm ADMIRE (ADvanced Modeled Iterative REconstruction) on each CT system.For all systems, the quantitative reconstruction kernel (Qr40), a slice thickness of 1 mm, and a 380 mm field of view were used.For each acquisition, the VMIs for the four lowest energy levels (40/50/60/70 keV) were reconstructed on Syngo.viasoftware (VB60A_HF03) with the specific Monoenergetic Plus application.
Task-Based Image Quality Assessment on VMIs
A task-based image quality assessment was performed on the VMIs using the iQMetrix-CT software (v1.1) developed by the French Society of Medical Physicists [22].For each CT system and each energy level, the Noise Power Spectrum (NPS), Task-based Transfer Function (TTF), and detectability index were computed based on all the data from the 10 acquisitions.A single calculation for each of these metrics was carried out per energy level.
Noise Power Spectrum
For each energy level, the NPS was computed in 180 consecutive axial slices (18 slices for each of the 10 acquisitions) by placing four square ROIs (Figure 1B).The sizes of the four square ROIs were adjusted according to the phantom's sections: 80 × 80 pixels for the 26 cm, and 104 × 104 pixels for 31 cm and 120 × 120 pixels for the 36 cm diameters.
To quantify the changes in noise magnitude, the square root of the area under the NPS1D curve (HU) and the magnitude of the NPS1D peak (HU 2 •mm 2 ) were measured.To quantify the changes in noise texture, the average spatial frequency (f av , mm −1 ) of the NPS curve and the spatial frequency of the NPS peak(s) (f peak , mm −1 ) were measured.
Task-Based Transfer Function
For each VMI, the TTF was computed on iodine inserts at 10 mg/mL (Figure 1C) using the circular edge technique [23].To minimize the image noise effect, the TTF was computed from 160 consecutive axial slices (16 slices for each of the 10 acquisitions).
The TTF values at 50% (f 50 , mm −1 ) were used to quantify the changes in spatial resolution.
Detectability Index
A non-prewhitening observer model with an eye filter (d' NPWE ) was used to calculate the detectability index (d') for one 10 mm diameter clinical task approaching the contrast of enhanced vascular or strongly enhancing parenchymal structures like, for example, a hepatocellular carcinoma [20].Based on the variations in contrast between the phantom's background material and the iodine insert according to the energy level, the contrast was set at 1260, 820, 550, and 390 HU for the energy levels ranging from 40 to 70 keV.
The shape of the signal was circular and the contrast profile was Gaussian [24].The interpretation conditions for calculating the d' were a zoom factor of 1.5, a 500 mm viewing distance, and the Eckstein visual function [25].
Results
The TTF curves obtained for each CT system, each phantom's diameter, and each keV are presented in the Appendix A. The axial images of the iodine insert at 10 mg/mL obtained with the 31 cm diameter for the three CT systems according to the energy level are depicted in the Appendix B. In the Results section, the values are expressed as means ± standard deviations (SD).
Noise Magnitude
For all CT systems and all diameters of the phantoms, the noise magnitude decreased as the energy level increased (Figure 2A).From 40 to 70 keV, the noise magnitude decreased on average for all diameters of the phantoms by −59.0% ± 3.3% for SFCT-1st, −57.1% ± 1.6% for SFCT-2nd, and −55.4% ± 0.8% for DSCT.For all CT systems and all energy levels, the noise magnitude increased as the phantoms' diameters increased.From the diameter of 26 cm to 31 cm, the noise magnitude increased on average for all energy levels by 40.5% ± 3.2% for SFCT-1st, 36.3% ± 2.7% for For all CT systems and all energy levels, the noise magnitude increased as the phantoms' diameters increased.From the diameter of 26 cm to 31 cm, the noise magnitude increased on average for all energy levels by 40.5% ± 3.2% for SFCT-1st, 36.3% ± 2.7% for SFCT-2nd, and 32.3% ± 1.0% for DSCT and from 31 cm to 36 cm, 41.4% ± 6.7%, 31.6% ± 1.7%, and 17.8% ± 1.0%, respectively.
Whatever the phantom diameter and energy level, the highest noise magnitude values were found with SFCT-1st.For all energy levels, the noise magnitude was higher with DSCT than with SFCT-2nd for the diameters of 26 cm (10.2% ± 1.3%) and 31 cm (7.0% ± 2.5%), and the opposite for the diameter of 36 cm (−4.2% ± 2.5%).
Noise Texture
For all CT systems and all phantom diameters, the average NPS spatial frequencies (f av ) increased as the energy level increased (Figure 2B).From 40 to 70 keV, the f av values increased on average for both SFCTs by 20.1% ± 0.3% for 26 cm and 23.5% ± 1.9% for 31 cm.For the 36 cm diameter phantom, the variations in the f av values were more marked for SFCT-1st (from 0.13 to 0.22 mm −1 ) than for SFCT-2nd (from 0.17 to 0.22 mm −1 ).For DSCT, the f av values increased for all phantom diameters by 11.0% ± 2.7% on average from 40 to 70 keV.
Whatever the phantom diameter and energy level, the highest f av values were found with DSCT.For all energy levels, the f av values were similar with SFCT-1st and SFCT-2nd for the diameters of 26 cm and 31cm but for 36 cm, higher f av values were found with SFCT-2nd, and all the more so at low keV levels.
With both SFCT systems and all phantom diameters, two NPS peaks were present on the curves for 40, 50, and 60 keV (and 70 keV only for SFCT-1st and the 36 cm diameter): one at a spatial frequency of 0.02 to 0.04 mm −1 and another at 0.14 to 0.17 mm −1 (Figure 3 and Table 2).With DSCT, two NPS peaks were present at 40 keV for all diameters and at 50 and 60 keV only for the 36 cm diameter phantom.The magnitude of the low frequency NPS peak was higher than the high frequency NPS peak for all systems, but the magnitude difference between these two peaks decreased as the energy levels increased.
Table 2. Values of noise power spectrum peak(s) and their respective spatial frequencies (f peak ) obtained for all energy levels (keV) on the three CT systems.
Task-Based Transfer Function
For all CT systems and all diameters of the phantoms, the TTF values at 50% (f50) increased as the energy level increased (Figure 4 and A1).From 40 to 70 keV, the f50 values increased on average for all the diameters of phantoms by 7.6% ± 0.6% for SFCT-1st, 6.3% ± 0.3% for SFCT-2nd, and 5.2% ± 0.5% for DSCT.
Whatever the phantom diameter and energy level, the highest f50 values were found with DSCT, except at 70 keV for the 36 cm diameter phantom.For all energy levels, the f50 values were higher with the SFCT-2nd platform than with the SFCT-1st platform for the
Task-Based Transfer Function
For all CT systems and all diameters of the phantoms, the TTF values at 50% (f 50 ) increased as the energy level increased (Figures 4 and A1).From 40 to 70 keV, the f 50 values increased on average for all the diameters of phantoms by 7.6% ± 0.6% for SFCT-1st, 6.3% ± 0.3% for SFCT-2nd, and 5.2% ± 0.5% for DSCT.
Whatever the phantom diameter and energy level, the highest f 50 values were found with DSCT, except at 70 keV for the 36 cm diameter phantom.For all energy levels, the f 50 values were higher with the SFCT-2nd platform than with the SFCT-1st platform for the 26 cm diameter (3.2% ± 0.4%) and the opposite for the 31 cm diameter (−6.9% ± 0.5%) and 36 cm (−5.6% ± 0.7%).
Whatever the phantom diameter and energy level, the lowest d' values were found with SFCT-1st.For all energy levels, the d' values were lower with DSCT than with SFCT-2nd for the 26 cm diameter phantom (−6.2% ± 0.7%), similar for the 31 cm diameter phantom (−0.3% ± 1.9%), and higher for the 36 cm diameter phantom (5.4% ± 2.7%).
Discussion
In the present study, for the first time, the performances of three dual-energy CT (DECT) platforms were compared according to the diameter of the phantoms.To achieve this goal, a task-based image quality assessment of the low-energy-level virtual monoenergetic images was performed.The results showed that, despite increasing the dose level, the greater the phantom's diameter, the greater the image noise and degradation of the spatial resolution, noise texture and detectability.We also found that the lower keV led to an increase in contrast and detectability, despite an increase in noise magnitude and a degradation in noise texture and spatial resolution.Finally, the highest detectability was found with the SFCT-2nd platform for the smallest phantom diameter and, for the dual-source CT platform, with the largest diameter phantom.
The outcomes of the NPS confirmed that the noise magnitude increased as the phantoms' diameter increased [20] and the energy levels decreased [16,17,19].For all diameters of the phantoms, the noise magnitude values were higher with SFCT-1st than with SFCT-2nd and DSCT.Similar outcomes were found between SFCT-1st and SFCT-2nd using AuSn120 kVp using another phantom [19].The noise magnitude values were higher with DSCT than with SFCT-2nd for all diameters, except for the 36 cm diameter phantom.Regarding noise texture, the f av values increased as the energy levels increased and as the phantom's diameter decreased.For all diameters of the phantoms and for all keVs, the best noise texture was found with DSCT, resulting in "fine" noise.Similar noise texture was found with both SFCTs for the 26 cm and 31 cm diameters, but the highest f av values were found with the SFCT-2nd for the 36 cm diameter phantom.These results are consistent with previously published studies on the SFCT-1st which used a different image quality phantom and a different reconstruction kernel (Br40) [16,17] and on DSCT [20].We also found the presence of two peaks on some NPS curves for all DECT platforms.For all phantom diameters, these two peaks were present on all the NPS curves of both SFCT platforms at 40, 50, and 60 keV.For these two platforms, the magnitude of the NPS peak at low frequencies was higher than the NPS peak's magnitude at high frequencies, where the lower the keV, the greater the phantom diameter.This phenomenon was more pronounced with the SFCT-1st than with the SFCT-2nd.With DSCT, these two peaks were present at 40 keV for the three phantom diameters and at 50 and 60 keV only for the 36 cm diameter phantom.In addition, the magnitude of the two NPS peaks was similar for both spatial frequencies.The presence of these two NPS peaks had been found in previous studies [16,17,19] and may be related to the distortions in the images or artifacts on the VMIs at the lowest energy levels for these platforms.
The spatial resolution outcomes confirmed that, for the iodine insert at 10 mg/mL, the TTF values at 50% decreased as the phantom's diameter increased [20].We found that the f 50 values decreased as the energy levels decreased.Similar outcomes were found for SFCT-1st and DSCT [16,17,20].However, Greffier et al. found that the f 50 values were not changed from 40 to 70 keV for SFCT-1st and SFCT-2nd [19].This study was performed with a different phantom and inserts containing lower iodine concentrations (e.g., 2 and 4 mg/mL).These results were directly related to the noise magnitude and the contrast variations of the insert used according to the phantom's diameter and the energy levels.TTF was calculated using the circular edge technique [23], which generates an Edge Spread Function (ESF) from the difference in the HU values between the phantom's background material and the insert.The ESF therefore depends on the noise conditions in the image, as well as on the differences in the HU values between the background and the insert.We found that, for the same contrast, the image noise increased as the phantom diameter increased, altering the ESF, and therefore, the resulting TTF.Furthermore, for the same phantom diameter, the reduction in keV resulted not only in an increase in image noise, but also in an increase in contrast.Our results seem to suggest that the impact of noise on the ESF was more marked than that of the contrast, since the f 50 values tended to decrease as the low keVs decreased.Finally, based on these variations in contrast and image, the highest f 50 values were found for DSCT, particularly for the diameters of 26 and 31 cm.With the SFCT platforms, the f 50 values were higher with the SFCT-2nd than with the SFCT-1st for the 26 cm diameter phantom, and the opposite was true for the other diameters.
The results of the detectability index (d') values calculated for the simulated lesions show that the d' values peaked at 40 keV and decreased as the energy levels increased.Similar outcomes were obtained by Euler et al. for DSCT [20] for the same phantom and by Greffier et al. for SFCT-1st and DSCT [17] and for SFCT-2nd for other image quality phantoms [19].The d' values were also found to decrease as the phantom's diameter increased.For all diameter phantoms and energy levels, the lowest detectability was found with the SFCT-1st platform.The highest detectability was found with the SFCT-2nd for the 26 cm diameter phantom and for DSCT for the 36 cm diameter phantom.Similar d' values were found for DSCT and SFCT-2nd for the 31cm diameter phantom.As the contrast values were set for each energy level, these outcomes were directly related to the noise magnitude outcomes.Indeed, unlike the TTF, the NPS curves are not normalized and the variations in noise magnitude have a strong impact on the detectability results.
The results of our study show that the spectral performance of the three DECT platforms understudy varied according to the phantoms' diameters and the technology used.Although the dose level increased between the 26 and 36 cm diameter phantoms, increasing the phantom's diameter was found to alter the detectability, spatial resolution, amplitude, and noise texture.Furthermore, different results could be obtained with the higher dose levels, particularly for the 36 cm diameter phantom.However, to represent clinical practice as closely as possible, the dose levels were defined from the dose levels of abdomen CT examinations performed on the three CT systems used for patients with BMIs similar to those extrapolated for each phantom's diameter.Concerning the technology used, we were able to highlight the different results depending on the DECT platform used.Compared to the two other DECT platforms, the SFCT-1st platform, with a limited spectral separation, presented the worst detectability and noise magnitude results and was the most influenced by the increase in phantom diameter.The technological evolutions brought to the SFCT-2nd platform by thickening the Au and Sn filters for better spectral separation have led to better results than with the SFCT-1st platform [19].These improvements have brought the detectability results closer to those obtained with DSCT and even better for the 26 cm diameter phantom.However, these results must also be interpreted with caution as, in this study, only one pair of kVps (100/Sn150 kVp) out of the five available was studied, and different results might be obtained with another pair of kVps.In all cases, the results of this study show that the best detectability was obtained at 40 keV for the three CT systems used and therefore suggest that the low keV VMIs can be used in clinical routine.In order to guarantee optimal image quality on the keV-based VMIs, special attention must be paid to the dose level used for overweight patients.The results obtained on the phantoms in this study must now be confirmed via specific clinical applications on patients.
This study has certain limitations.The acquisitions were performed with only one quantitative reconstruction kernel and only one level of the ADMIRE algorithm was used.Other parameter combinations may provide different outcomes.However, the acquisition and reconstruction parameters used in this study were those used in clinical routine on all DECT systems assessed.Also, we did not use a tube current modulation system on the three CT systems.This choice made to guarantee similar dose levels in each phantom diameter for each CT system assessed.Finally, only one task function was chosen to model the detection of the contrast-enhanced lesion, with the contrast defined according to the contrast value variations relative to the VMIs for the one iodine insert available in the Mercury v4.0 phantom.This method was chosen to take into account the variations in the contrast of the simulated lesion according to the energy level.However, a clinical contrast-enhanced study on the lesions of a known specific size should really be carried out now to confirm our results.
Conclusions
The outcomes of the present study confirmed that the spectral performance depends on the DECT platform and the phantom's diameter.Increasing the phantom's diameter increased the image noise and degraded the spatial resolution, noise texture, and detectability, despite the increase in dose level.Changing the thickness of the tin and gold filters on the SFCT-2nd platform has led to a lower noise magnitude and better detectability than with the SFCT-1st.Compared with the DSCT platform, the SFCT-2nd had the lowest noise magnitude and highest detectability for the smallest phantom diameter, and the opposite was true for the largest diameter.Further patient studies are now required for specific clinical applications to confirm the results obtained on the phantoms in this study.
Figure 1 .
Figure 1.(A) Front X-ray image of the 26, 31, and 36 cm diameter sections of the Mercury v4.0 phantom.(B) Example of the four regions of interest (ROIs, in red) placed in the homogenous part of the Figure 1.(A) Front X-ray image of the 26, 31, and 36 cm diameter sections of the Mercury v4.0 phantom.(B) Example of the four regions of interest (ROIs, in red) placed in the homogenous part of the 31 cm diameter section and used on VMIs to assess the noise power spectrum (NPS).(C) Example of the ROI placed around the iodine insert of 10 mg/mL (in blue) and the phantom's background material (in red) and used to compute the task-based transfer function (TTF).
14 Figure 2 .
Figure 2. Values of noise magnitude and average Noise Power Spectrum spatial frequencies (fav) obtained for the two generations of split-filter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Figure 2 .
Figure 2. Values of noise magnitude and average Noise Power Spectrum spatial frequencies (f av ) obtained for the two generations of split-filter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Figure 3 .
Figure 3. Normalized noise power spectrum (nNPS) curves obtained for the two generations of splitfilter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Figure 3 .
Figure 3. Normalized noise power spectrum (nNPS) curves obtained for the two generations of split-filter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Figure 4 .
Figure 4. Values of task-based transfer function at 50% (f50) obtained for the two generations of splitfilter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Figure 5 .
Figure 5. Detectability index (d′) values for the simulated lesion obtained for the two generations of split-filter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Figure 4 .
Figure 4. Values of task-based transfer function at 50% (f 50 ) obtained for the two generations of split-filter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Figure 4 .
Figure 4. Values of task-based transfer function at 50% (f50) obtained for the two generations of splitfilter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Figure 5 .
Figure 5. Detectability index (d′) values for the simulated lesion obtained for the two generations of split-filter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Figure 5 .
Figure 5. Detectability index (d ) values for the simulated lesion obtained for the two generations of split-filter CT platforms (SFCT) and the dual-source CT platform (DSCT) according to phantom diameter and energy level (keV).
Table 1 .
Acquisition parameters used for the three CT systems. | 2023-09-27T15:07:15.849Z | 2023-09-25T00:00:00.000 | {
"year": 2023,
"sha1": "0f96cb4af4bb555aff304486d92910fe4c5f4587",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2075-4418/13/19/3039/pdf?version=1695620693",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "56c0911c6c79b98935502aff26f8d05f596270ce",
"s2fieldsofstudy": [
"Physics",
"Medicine"
],
"extfieldsofstudy": []
} |
13043865 | pes2o/s2orc | v3-fos-license | Condylar Aplasia and Hypoplasia: A Rare Case
Aplasia of condyle is very rare, when this condition not seen as a part of a syndrome. We report a case of condylar aplasia on the right side and hypoplasia on the left side in a 21-year-old female. The patient reported to the department with a chief complaint of underdeveloped lower jaw. Clinical examination, conventional radiographs, and 3D CBCT images revealed complete absence of condyle on the right side and hypoplasia on the left side.
Introduction
The temporomandibular joint (TMJ) is one of the most complex joints of the human body. It is considered a ginglymus diarthrodial joint capable of both rotational and translatory movements. It consists of the mandibular condyle and the articular eminence of the temporal bone. The condyle is very special because the expression of mandibular growth is provided by mandibular condyle. In compared to other diarthrodial joints, during prenatal life the TMJ lags morphologically behind other synovial joints in both the timing of its appearance and its progress, so that at birth the joint is still largely underdeveloped. The TMJ first appears in the 8th week of gestation, when two separate areas of mesenchymal blastemas appear near the eventual location of the mandibular condyle and glenoid fossa [1,2]. Bone and cartilage are first seen in the mandibular condyle at approximately the 10th gestational week. First condylar blastema developed from which the mandibular condyle cartilage, the aponeurosis of the lateral pterygoid muscle, and the disc and capsule component composing the lower portion of the joint are derived. Next is the temporal blastema, which eventually forms the articular surface of the temporal component and the structures of the upper portion of the joint. The mandibular condyle and temporal blastemas begin their growth at relatively distant sites; they then move towards each other as the joint develops by the 12th week. At birth, the articular surfaces of both the mandibular condyle and temporal bones are covered with fibrous connective tissue. Later, this tissue is slowly converted to fibrocartilage as the fossa deepens and the mandibular condyle develops under functional influences [3,4].
Growth disturbances in the development of mandibular condyle may occur in utero late in the first trimester and may result in disorders such as aplasia or hypoplasia of the mandibular condyle. As compared to hypoplasia, hyperplasia of the mandibular condyle is not visible at birth and seems to be gradually acquired during growth [5].
Case Report
A 21-yr-old female was presented to the oral medicine and radiology department with a chief complaint of underdeveloped lower jaw, which was first noticed during childhood and gradually progressed. Due to unfavorable socioeconomic conditions, it was not possible to get the treatment done for the patient. At the anamnesis there was no history of any trauma or any systemic diseases. Patient's parents gave a history of consanguineous marriage. There was no family history of the present problem.
General Physical examination did not reveal any abnormalities. Her vital signs were within normal limits. Extraoral examination revealed facial asymmetry with severe retruded mandible giving a bird face appearance (Figures 1 and 2). Mouth opening was restricted (10 mm) with no deviation or deflection ( Figure 3). On palpation condyles were not detected on both sides. Intraorally there was crowding in upper and lower anterior teeth with increased overjet and overbite ( Figure 4). Molar relation was Angle's Class II bilaterally. She had deep palate and generalized enamel hypoplasia (fluorosis). No other important clinical extraoral or intraoral findings were observed. Based on clinical findings, a provisional diagnosis of bilateral ankylosis and differential diagnosis of bilateral condylar hypoplasia or aplasia were given. inconclusive ( Figure 6). Lateral skull views showed severe retruded and micrognathic mandible (Figure 7). Findings of panoramic radiograph were confirmed by lateral skull radiographs. CBCT was advised for additional information. CBCT also confirmed the findings of OPG and lateral skull radiographs (Figures 8, 9, 10, and 11). After radiographic confirmation patient was advised complete systemic evaluation and referred to general medicine, cardiology, ophthalmology, ENT, and orthopedics to rule out any syndromes. Medical evaluation revealed no abnormalities. Based on the clinical and radiographic findings, a final diagnosis of nonsyndromic agenesis of the right condyle and hypoplasia of the left condyle was given. Patient was referred to oral surgeon and orthodontist for the best possible treatment.
Discussion
The congenital deformities and developmental abnormalities of the mandibular condyle can be classified as hypoplasia or aplasia, hyperplasia, and bifidity. Hypoplasia or aplasia of the mandibular condyle indicates underdevelopment or nondevelopment associated mainly with various craniofacial abnormalities. These may be either congenital or acquired [5].
Congenital (primary) condylar hypoplasia is characterized by unilateral or bilateral underdevelopment of the mandibular condyle and usually occurs as a part of some systemic condition originating in the first and second branchial arches, such as Mandibulofacial dysostosis (Treacher Collins syndrome), Hemifacial microsomia (first and second branchial arch syndrome), Oculoauriculovertebral syndrome (Goldenhar syndrome), Oculomandibulodyscephaly (Hallermann-Streiff syndrome), Hurler's syndrome, Proteus syndrome, Morquio syndrome and Auriculocondylar syndrome [5][6][7][8]. As a rule, in each of these conditions some soft tissue manifestations accompany the condylar agenesis and/or condylar malformations [9]. Acquired (secondary) condylar hypoplasia takes place if the condyle is injured during active growth, because of which development may be arrested. The most common causes are mechanical injury, such as trauma (before the age of 2), infection of the joint itself or the middle ear, childhood rheumatoid arthritis, radiotherapy, and parathyroid hormone-related protein deficiency which affect bone formation and chondrocyte differentiation [5,9,10].
Several authors confirmed that mandibular deficiency can occur without any defined etiology [11]. Aplasia of the mandibular condyle without any other facial malformations is an extremely rare condition [8].
The TMJ develops from initially widely separated temporal and condylar blastemata which appear at about the 8th week of conception. Eventually they grow towards each other and ossify to form a functional joint by about the 20th week of intrauterine life [5]. In our case, total absence of the condyle and glenoid fossa on the right side and hypoplastic condyle and glenoid fossa on the left side constitute an evidence that the defect originated in the prenatal period.
Various treatment approaches have been proposed for treating condylar aplasia and possibilities for influencing mandibular growth. Most of the time it is treated by multimode with the help of oral surgeon, general surgeon, plastic surgeon, and orthodontist [9,15,16].
The treatment could then be a costochondral graft transplant, preferably before the growth spurt, orthognathic surgery at the end of the growth period, or both [16]. Krogstad reported that effective results were obtained through the application of a form of orthodontic activator which aimed to swing the mandible to the unaffected side and promote formation of a mandibular condyle, albeit irregular in shape [9]. Surgery is often required, but the timing and regimen of this choice is still an issue to be resolved [15].
Conclusion
In conclusion we report a rare case of total condylar aplasia on the right side and condylar hypoplasia on the left side, not related to any clear pathological disorder. This case of unknown etiology was thoroughly examined; based on clinical and radiographic findings, we suggest that this case is of congenital origin. Nonsyndromic condylar hypoplasia and aplasia are exceedingly rare conditions and very few case reports are published till date. In this context, our case is an important addition to the literature. Early detection and prompt treatment are imperative to restore esthetics and thus provide psychologic benefit to these patients. | 2018-04-03T06:10:22.593Z | 2013-03-24T00:00:00.000 | {
"year": 2013,
"sha1": "face1aa136441d89be9ff7157e4d06899437db58",
"oa_license": "CCBY",
"oa_url": "http://downloads.hindawi.com/journals/crid/2013/745602.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "2f422fd1791fca57c73a8e9c013ca4e705ca642e",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
246781949 | pes2o/s2orc | v3-fos-license | Divorce and Its Psychological and Social Effects on the Children of Divorced People from Their Viewpoint in the Jordanian Society
This study aimed to identify the psychological and social effects of divorce on children from their viewpoint, and to achieve the aim of the study A questionnaire was developed consisting of (17) items that were distributed to a sample of (100) children of the divorced, and they were randomly selected. The researchers used the descriptive and analytical method for the study. The study concluded that feeling fear and anxiety about the future and the accumulation of worries and diseases is one of the most psychological effects that children of divorced men are exposed to, and it leads the individual to a feeling of inferiority and lack of self-esteem. While the most social factors affecting the members of the study sample were that divorce leads children to bad company due to poor family control over them. Then society viewed them as an inferior view that they are not socially qualified. The results of the study also showed that there were no statistically significant differences attributed to the age or gender variable, while statistically significant differences appeared due to the scientific qualification variable.
Introduction
Divorce is a social phenomenon since the emergence of humanity and the rise of ancient human civilizations, and its names have multiplied, such as separation, divorce, abandonment, neglect and others. Although divorce is a big problem that works to disintegrate the family and marital life, but sometimes it is a solution to marital problems for fear of exacerbating them and leading to crimes such as violence, suicide, rape and other sex. Project, mental illness, depression and other crimes and diseases are present in all societies, whether Arab or Western, and also in different religions (Al-Janabi, 1983). The social, economic, political and technological changes have played a major role in the disintegration of marital life and the increase in divorce cases, which generate negative effects on the individual, the family and society. The children of divorced people are the victims of divorce and family disintegration, as they face many problems such as frustration, anxiety and psychological diseases. And the phenomenon of divorce is not limited to a particular social class, but has become occurring in all classes, whether poor, rich, educated or uneducated, especially after the emergence of modern technology such as (the Internet, Facebook, Twitter) and other social networking sites that have played a prominent role in the instability of life. Marital status and the increase in divorce rates every year, both in the Arab community in general and the Jordanian society in particular as the cases of divorce for the year 2020 reached (22780) in the Jordanian society, according to what was stated by the Chief Justice Dr. Abdel-Hafez Al-Rabtah on Wednesday 3/2/2021 in a press conference. And that this phenomenon has become threatening the stability of the family and society, both in terms of social, economic and psychological (Khudair, 2004). Especially on the life of children, their upbringing, their care and their psychological, cognitive and behavioral development, and for this reason the phenomenon of divorce has become in need of treatment and attention through educational, social, psychological and legal studies carried out by a team of specialists and those interested in this field to work on rehabilitating and educating young people coming to marriage about the concept of marriage and its importance and the rights of each one They are for the sake of family cohesion and the success of marital life, which reduces divorce cases in society.
Study Problem
The problem of divorce has become in the increase in divorce cases in the Arab community in general and the Jordanian society in particular, according to the statistics issued by official and private institutions in Jordan, and it has become a social, psychological and economic problem due to its negative effects on society, the family and children, and the problem of studying is summarized by knowing the psychological and social effects of divorce Which is reflected on the children from the point of view of the study sample.
Importance of the Study
This study derives its importance in theory in preserving the cohesion and cohesion of society by studying the phenomenon of divorce, which has become a major problem as a result of the rapid changes, cultural risks and revolution in the world of communication and technology, which leave negative effects on society and the children of divorced people as a result of family disintegration stemming from divorce. In practice, the results of this study will contribute to presenting a clear picture by knowing the most important psychological and social implications of divorced children, which helps decision-makers and officials who have a relationship with the issue of family and divorce in developing strategies and counseling and awareness programs for young people and couples to reduce divorce and make them aware of its effects and damages on the family and children And society.
Study Questions
1) What are the most important psychological effects of divorce that affect the children of divorced people?
2) What are the most important social effects of divorce that affect the children of divorced people?
3) Is there a statistically significant effect of the demographic variables (gender, age, educational qualification) from the viewpoint of the study sample individuals regarding the fields of study?
1.4 Objectives of the Study 1) Identify the most important psychological effects of divorce that affect the children of divorced people.
2) Identify the most important social effects of divorce that affect the children of divorced people.
3) Knowing is there a statistically significant effect of the demographic variables (gender, age, educational qualification) from the viewpoint of the study sample individuals regarding the fields of study.
Time Limit
The study is limited to the psychological and social effects of divorce from the viewpoint of the children of divorced women for the 2020/2021 academic year.
Place and Objective Limit
This research is limited to determining the psychological and social effects of divorce, Capital Governorate (Amman).
The Human Limit
This research is limited to the children of divorced people in the Capital Governorate (Amman).
The results of the study are determined by the nature of the tool and its validity and reliability indications.
Procedural Concepts
Divorce: is the separation of the husband from the wife through verbal or writing, and its official approval by the Sharia judge and its registration.
Psychological effects: every negative behavior that affects the psyche of the child as a result of divorce.
Social effects: all social behavior affected by the child's marital status as a result of divorce.
Divorced: They are the husband and wife separated from each other officially by the judge of the Sharia court.
Children of divorced:
They are the children of the husband and wife separated from each other officially by the judge of the Sharia court.
Idiomatic Concepts
Sharia divorce: raising the marriage immediately or in the event with a specific wording, in other words meaning: dissolving the marriage bond and ending the marital relationship, and divorce is called single will (Al-Sartawi, 2012: 169).
The definition of the United Nations: A court ruling separating the spouses gives each of them the right to remarry according to the laws followed in their countries (Al-Shalabi, 1992).
Factors Leading to Divorce
A. Cultural and social factors: Cultural and social factors are represented by values, customs, traditions, religion, language, social systems, ethics, economic, social, cultural, political and technological changes that play an important role in destabilizing the social structure and family disintegration that generates divorce through the different concepts of the spouses, such as the difference in character, socialization, interference of the couple's family and bad relationship, Whereas, the bad relationship between husband and wife may be one of the reasons leading to divorce, especially when each husband or wife tries to impose their views on the other and the difference in the role of women through increasing educational and work opportunities and the emergence of feminist movements to demand equality between men and women, as there are also other factors that lead to Divorce, such as poor choice, moral corruption, marital infidelity, frigidity in the husband or wife, work of women, neglect of marital duties, non-procreation, social class differences, etc. (Shukri et al., 2011).
B. Psychological factors: These are the psychological variables that are related to the human personality of the individual mainly, especially the motives, emotions, instincts, the absence of feelings of love, affection, and affection between them, and the coercion in choosing a husband or wife, and early marriage may generate psychological effects on the spouses such as quarrels and tensions between them and the lack of satisfaction of sexual desire as a result of the psychological effects that occur when Husband to hate her body after her birth or pregnancy without Considerate the physiological changes that happens to the wife and perhaps bad words from both parties, or the disease of the husband or wife.
C. Economic factors: The poor economic situation and family unemployment may prevent the continuation of married life because of the living requirements that the spouses cannot provide, especially since most women have the desire to buy and shop in imitation of girlfriends, relatives and celebrities, which makes them look for work to help the husband, and this may lead to neglecting the affairs of her home And her children at the expense of her new social position at work, the failure of marriage to be based on sound foundations, and the difference between husband and wife in their outlook on life and the level of culture and social status (Al-Khashab, 1985).
D. Technological factors: The modern means of communication and social media sites such as Facebook, Twitter and the Internet have played a major role in the disintegration of marital relations and divorce despite their importance, especially among groups of young people who have become rebellious through them against the behaviors, values and social trends recognized, generating depression and isolation (Al-Zubaidi et al., 2009). It also led in many cases to the disintegration of social and family relations and distrust between the spouses, causing separation from each other, and some studies indicated that the high rate of divorce was due to the use of social networking sites through messages and pictures of sexual and emotional temptation, introverted isolation and suspicion between the spouses (Al- Khaqani, 2015).
Psychological and Social Effects on Divorced Children
Psychological effects: Children may be born as a result of divorce worries and their feeling of fear and anxiety about the future and the society's view of them as divorced children, and there have been cases as a result of this feeling of fear and others subjected to psychological problems such as introversion and isolation from people as well as losing the sympathy of tenderness as a result of being away from the mother's tenderness if they are with the father and the tenderness of the father If they are with the mother, the divorce also affects the psychological and physical health of the children, especially if they are five or six years old this is due to the lack of interest of their parents in them as a result of family problems, and viewing of both parties or one of them is through family and social centers and institutions, which leads to a decline in children's morale, such as crying, despair, anxiety, trouble and the need for financial matters this is because the father and mother lose their livelihood (Khudair, 2004).
The child as a result of divorce is affected by problems and ruptures that have a negative impact on the child, family and society. Among these effects: The child has internal conflicts as a result of the breakdown of family life, so this child carries aggressive motives towards parents and the rest of society. In many cases, the child moves from the headquarters of the broken family to live strangely with his father or mother, and thus faces great difficulties in adapting to the stepmother or stepfather. The child may make several comparisons between his parents and the new parents, which makes him in a state of continuous psychological disorder. According to this new situation, the child must adapt to different home environments in terms of economic, social and cultural levels which affects the child's personality greatly he creates a personality unstable, and oscillating. like parents, the child bears the burden of constantly thinking about the problem of separation. The child draws continuous comparisons between his disintegrating family and the family life that the rest of the children live, which generates a feeling of frustration, or it may win him an aggressive tendency towards everyone, especially children of healthy families. The child is exposed to disorder and anxiety as a result of his lack of awareness of the underlying goals of the conflict between the parents or the reasons for trying to use it -by his parents -to launch an attack on each other and use it as a tool to achieve victory over the other side. This disorder in childhood leads to a disturbance in the emotional and mental development of the child, so that a person with a shaky or ailing personality emerges to society, which harms the entire society (Al-Jaafari, 1999).
Social effects: Divorced children may be affected by the negative view of society towards them in terms of behavior, behavior, disappointment and frustration expected for them in the future, which increases them only complication and fear of the future, and the separation of the father from the mother may affect the early socialization of children in the first years of his life that leads to disintegration and homelessness As a result of the lack of direct supervision of children by parents and living in the streets, homelessness, delinquency, and their loss of a sense of security, protection and stability, and this makes them think as a result of these conflicts, dropping out of schools, deviation in behavior, weak self-confidence, inability to intellectual development and social interaction with their friends, and their deprivation of proper social upbringing They resort to homelessness and the use of drugs and intoxicants to keep away from the concerns of the family, and all of this is one of the negative effects that divorce leaves on the children (Ahmed, 2001).
Among the most important social effects on children is also family disintegration, which works to loosen the bonds of family building, weak social interactions between family members, disturbance of expectations of their roles, feelings of alienation and insecurity, and this is the result of the social changes that society is going through (Al-Shabiki, 2006). Divorce also generates negative effects on children, such as losing the breadwinner and financial source, increasing economic pressures, and also the parental life of guidance, knowledge, counseling and assistance through parents, as children who live with their parents have more financial expenses than the spouses are separated as a result of limited income and financial resources after divorce. For both spouses, this will affect the livelihood of the children in terms of education, quality of schools, housing, health care, and relationship with friends (Teachman and Paasch, 1994).
Previous Studies
Al-Zahrani (2021), this study aimed to identify the relationship of emotional divorce with psychological stress and self-efficacy among a sample of couples in Jeddah, Saudi Arabia. The researcher used the descriptive correlative approach, the emotional divorce scale, the spouses' psychological stress scale, and the self-efficacy scale. The study sample consisted of (198 husbands and wives), and the results of the study found a positive relationship between psychological stress and emotional divorce. While there was a negative relationship between self-efficacy and emotional divorce, and there were also differences between the averages of both husbands and wives in being affected by psychological stress in favor of the wives' sample. There are no differences between husbands and wives in the level of self-efficacy. And the presence of a significant effect of the high level of psychological stress, And the low level of self-efficacy in predicting the emergence of emotional divorce in spouses. Najdawi's study (2018) aimed to try to understand the phenomenon of emotional divorce in Jordanian society, where the study was conducted on an intentional sample consisting of (17) women who reported emotional divorce. The researchers used the qualitative approach through an in-depth interview to collect data one of the results of this study was that emotional divorce exists in Jordan and it is difficult to accurately identify its numbers. One of the reasons for this divorce was the difference in expectations in marriage for both spouses, the concept of marriage for men and women, cultural factors, and the prevalence of individualism among men. And one of the reasons that pushed the woman to accept the matter reality and continuation of this marriage not to request an official divorce is to preserve the social image and fear for the future of children and parents. Rababaa study (2015) Vol. 16, No. 2;2022 divorce are poor choice at a rate of (83%), interference by the wife's family by (77%), interference by the husband's family by (70%) and failure to perform the wife's rights to the husband at a rate of (63%). A Kheira Study (2015) It aimed to identify the self-image of children of divorce by highlighting the defect that children of divorce suffer in their perception of their self-image by studying this phenomenon on three cases of adolescents who are victims of the divorce of their parents through the application of a set of tools from observation, interview and GPS testing one of the most important results of this study was that children of divorce suffer from a feeling of anxiety, tension, excitement, feelings of deprivation, a sense of inferiority and insecurity, and the feeling of difference creates a feeling that makes them think that people look at them differently this makes the teenager frustrated and lack of confidence in social relationships, especially if he was at a young age. A study (Vitanen, 2014) in the United States of America, entitled The Divorce Revolution and Generalized Trust: Evidence from the United States 1973 -2010, the study clarified the relationship between marriage laws and trust in society. As easy as what is known as the culture of easy divorce, whenever individuals have less confidence in society, which in turn will be reflected in social cohesion on the one hand, and affiliation and loyalty to society on the other hand. The issue of trust, whether on the individual level (self-confidence) or on the social level (confidence in society), its value and criteria are extremely important, and therefore the results of this study are in agreement with many studies conducted in the United States on confidence in society and the factors affecting it. A study (Valenzuela et al., 2014) Social network sites, marriage well-being and divorce Survey and state-level evidence from the United States: This study aimed to reveal the relationship between the use of social networks and marital happiness on the one hand, and divorce rates on the other hand, and its correlation with social networks. The results showed that the increased use of social networks has a negative relationship on marital happiness. And it works to increase and raise the level of problems in marital relations, and thinking about the divorce decision. The results of this study are in agreement with the results of the national study that was titled Facebook from 2008-2010. The study found that the increase in social media websites increases divorce rates directly in various parts of the United States with different factors causing divorce, and the two studies also focused on how to convert negative factors for social communication into positive factors. Masoud's study (2013) aimed to identify the most important (social, economic and psychological damages resulting from divorce for girls in the early stage of marriage in slums, And testing the practice of the meaning treatment model with Islamic content in the service of the individual to mitigate the harms resulting from divorce for girls in the early stage of marriage in slum areas, and The researchers relied on more than one tool consistent with the nature and quality of the study, such as the questionnaire and the interview, and the study community was one of the divorced women who hesitated to the Model Association for Service. The individual, family and society, and the study sample consisted of 127 cases who meet the conditions, The researchers reached several results, the most important of which are: The problems dealt with in that study within the framework of the meaningful treatment model for social problems due to poor relations with her family, frequent blame for her and family monitoring of her actions, and economic problems such as the financial burdens that the family bears on the newly married divorced woman as a result of separation from the husband. Divorce, psychological problems, and psychological stress that falls on divorced women and isolation as a result of society's view of them. The study of the Sheikh and others (2013) aimed at identifying the causes of the phenomenon of divorce and its implications, and the proposed solutions to treat this phenomenon. The study tool was questionnaires distributed on a sample consisting of (62) divorced (husband and wife) from the community of Medina, The most important results of this study: The most important social reasons for divorce are the lack of dialogue within the family in a democratic manner, as well as marriage imposed by families, and culturally the main reason for divorce is to watch programs contrary to customs and values through the media, and the most important psychological reasons leading to divorce It is infidelity, and the most important religious reason was the failure to respect the rights of the spouses to marital rights and duties.
What Distinguishes the Study from Previous Studies
Previous studies talked about divorce and its factors, and the implications for both spouses and children, such as the study (Sheikh et al., 2013;Masoud, 2013), and the studies also showed the most important effects on children of divorce, such as the study (Khaira, 2015). The current study agreed with most of the studies on the factors leading to divorce, such as (Al-Najdawi study 2018, and the fourth in 2015) and the implications for children, such as the study (Khair, 2015). This study was distinguished from previous studies by knowing the implications for children due to divorce.
Study Methodology
The researchers used the descriptive and analytical approach to conduct this study, due to its blameness with the nature of the study.
Study Population
The study population consisted of children of divorced people of different genders, groups and age qualifications in the Capital Governorate (Amman).
Study Sample
A random sample of 100 children of divorced people was selected in the Capital Governorate (Amman). directly and also through social media via electronic questionnaire It is as follows:
Study Tool
The study tool consisted of a questionnaire containing (17) paragraphs that were applied to a number of divorced children in order to find out the psychological and social effects of divorce on children by reviewing the theoretical concepts and previous studies that are related to the topic. The level of the answer for each paragraph will be graded according to (the five Likert scale) And defining them in five levels as follows: very agree and give (5) grades and a level in agreement and gives (4) degrees and a neutral level and gives (3) degrees and a level that does not agree and gives (2) two degrees and a level that does not agree strongly and gives (1) one degree, and a scale Judging the results divided into (high, medium, low), depending on the tool categories.
Tool Validation
The researchers carried out to verify the validity of the tool presenting the tool to a number of arbitrators specialized in the field of psychology, education, Islamic law and sociology, in order to know what the paragraphs measure, and the extent of the scale's relevance to the variables to be measured.
Tool stability: The values of stability transactions (Cornbach Alpha) were extracted for the study areas and were as follows: The tool as a whole. 0. 73 The results of the study showed a rise in stability transaction ratios, showing a high consistency between single-field paragraphs and the paragraphs of the tool as a whole. This makes them acceptable for scientific research purposes.
Study Procedures
The questionnaire was distributed to the study sample after explaining its objectives, and that their answers will be treated with complete confidentiality, and will be used only for scientific research purposes only.
Statistical Treatment
For the purpose of answering the research questions and achieving the desired goals, the following statistical treatments were used: 1) To answer the first and second questions, the arithmetic means and standard deviations were extracted.
2) To answer the third question to extract the effect of gender, the T-Test was used for two independent samples. To extract the effect of age and scientific qualification, the One Way ANOVA test was used.
3) To extract the stability of the tool, the stability factor (Cornbach Alpha) was used.
4) Frequencies and percentages of sample count were extracted.
The Answer to the First Question: What are the Most Important Psychological Effects of Divorce that Affect Children of Divorced People
The mean and standard deviations of the answers of the study sample individuals were related to the field of: Psychological effects, and the results were as follows: Fear and anxiety about the future and the accumulation of worries and diseases. It appears from the previous table that the arithmetic averages were high for the items according to the answers of the individuals of the study sample, and the total arithmetic mean came (4. 15), with a standard deviation of (0.56).
The table shows that Paragraph No. (7) which states that "divorce leads to bad company of children due to poor supervision over them," came first with an arithmetic average of (4.40), While each of Paragraph No. (5), which states that "the individual (males and females) finds it difficult to relate in the future," came in the second rank with an arithmetic mean of (4.36), and it came in the rank before The last paragraph No. (8), which states that "divorce and lack of affection in children lead to delinquency towards violence," with an average arithmetic of (3.94), and came in the last rank, Paragraph No.
First: Sex
The T-test was used for two independent samples, and the results were as follows: Field The results of the study indicate that there are no statistically significant differences between the answers of the male and female respondents with regard to the different answers of the study sample individuals according to gender. Where the value of the level of significance in all areas of the study was not statistically significant at the level of significance (α≤0.05).
Second: Age
Arithmetic averages were extracted for the answers of the study sample individuals according to the age variable, and the answers were as follows: To find out if these differences are statistically significant or not, a One Way ANOVA analysis was performed, and the results were as follows: It appears from the table that the differences between the answers of the study sample members according to age were not statistically significant at the level of significance (α≤0.05) This indicates that there are no statistically significant differences between the answers of the study sample individuals due to the age variable.
Third: Academic Qualification
Arithmetic averages were extracted for the answers of the study sample according to the scientific qualification variable, and the answers were as follows: It appears from the table that there are apparent differences between the averages of the responses of the study sample individuals depending on the scientific qualification variable, To find out if these differences are statistically significant or not, a One Way ANOVA analysis was performed, and the results were as follows: It appears from the table that the differences between the answers of the study sample members according to the difference in scientific qualification were statistically significant at the level of significance (α≤0.05) This indicates the existence of statistically significant differences between the answers of the study sample individuals due to the scientific qualification variable. To find out the source of these differences, a procedure was carried out shafih test, and the results were as follows: It appears from the table that the statistically significant differences were between My classes are secondary and below and bachelors degrees in the first field, As for the second field, the differences were between the secondary and lower categories, the bachelors degree, the secondary and lower categories, and the higher studies.
Discussing the Results
The results of the study concluded that the psychological factors that most affect the children of divorced people are the feeling of fear and anxiety about the future and the accumulation of worries and diseases. Then comes the individual's feeling of inferiority and lack of self-appreciation. The results of the study also found that the most social factors affecting the study sample members as a result of divorce are bad company due to poor control over them, difficulty in future association, and society's inferiority view towards divorced children as not being socially qualified. The results of the study also showed that there were no statistically significant differences due to the age or gender variable, while statistically significant differences appeared due to the scientific qualification variable.
Study Recommendations
1) Activating the role of family, social and educational institutions in directing and guiding those who are about to get married, such as family reform offices in Sharia courts in Jordan.
2) Educating young people about the rights of spouses in marital life and the negative effects of divorce on spouses, children and society.
3) Educating spouses about the importance of confidentiality and privacy in marital life, especially with regard to marital cohabitation. 4) Accustom the spouses to solve family disputes and problems on their own without the intervention of the spouses' family or reformers except in the case of necessity.
5) Rational thinking in choosing a husband or wife without the influence of family or friends in the selection process away from emotions, sexual instincts and shame when making the decision to approve the marriage.
6) Giving religion, morals, and social and educational values of great importance in the process of choosing a husband or wife. | 2022-02-13T16:23:00.110Z | 2022-02-11T00:00:00.000 | {
"year": 2022,
"sha1": "3f2183364636207c6a7262beeb7b76794512e9c2",
"oa_license": "CCBY",
"oa_url": "https://ccsenet.org/journal/index.php/mas/article/download/0/0/46738/49944",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "207c934cc1f67ce4955f4bebcd5e0b97c3349d80",
"s2fieldsofstudy": [
"Sociology",
"Psychology"
],
"extfieldsofstudy": []
} |
119311335 | pes2o/s2orc | v3-fos-license | Khovanov homology of a unicolored B-adequate link has a tail
C. Armond, S. Garoufalidis and T.Le have shown that a unicolored Jones polynomial of a B-adequate link has a stable tail at large colors. We categorify this tail by showing that Khovanov homology of a unicolored link also has a stable tail, whose graded Euler characteristic coincides with the tail of the Jones polynomial.
The following is an easy corollary of the results of section 7.7 of [Kho99]: Theorem 1.1. The numbers χ L and κ L are topological invariants of a B-adequate framed link L, because they do not depend on the choice of representative B-adequate diagram D for L. Moreover, if B-adequate framed links L and L ′ differ only by framing, then (1.1) For an unframed B-adequate link L we define the minimal crossing number χ ! L as the minimum among the numbers χ D for B-adequate framed diagrams D representing L. Thus defined, the bracket is framing-dependent: The Jones polynomial of a framed link L is the Kauffman bracket of its diagram: J L (q) = L .
1.3. Cables and coloring. We introduce coloring of tangle and link components through cabling and Jones-Wenzl projectors. A cable of a strand is depicted by using a thicker line with the label indicating the number of strands, and the Jones-Wenzl projector is depicted by a box: For a positive integer N let J N,L (q) denote the unicolored Jones polynomial of L, that is, all components of L are colored by the same color N. A coloring of a link component by N means that we assign the (N + 1)-dimensional irreducible representation of SU(2) to it. Equivalently, the color N means that the link component is N-cabled and we place the Jones-Wenzl projector on this cable.
In this paper we consider unicolored links, that is, links, all of whose components are colored by the same number N. Their colored Jones polynomial J N,L (q) is a Laurent polynomial of q 2 up to an overall factor: if L is presented by a diagram D, then q 1 2 χ D N 2 +κ D N J N,L (q) ∈ Z[q ±2 ].
1.4. Homological notations. Let A be a finitely generated additive category: objects of A are finite sums of elements of a finite set A. Let Kom + (A) denote the homotopy category of its complexes bounded from below: an object of Kom + (A) is a chain where A i = α∈A m i,α α and m i,α ∈ Z ≥0 are the multiplicities of generators. The notation m for multiplicity is treated in this paper as an arbitrary constant, so the appearance of m in different expressions does not imply that there is a relation between the multiplicities, unless it is stated specifically. The special multiplicities appearing in a presentation of the categorified Jones-Wenzl projector are denoted by µ.
We use a non-standard notation for the translation functor: hA = A[1], which allows us to define a functor p(h) for any polynomial p(x) with integer non-negative coefficients. In Draft of Htail22new: 5/5/2014 12:13 particular, we use a functor ¶ i j © h based on a combinatorial polynomial (1.4) We also use a non-standard notation for the cone of two complexes: (1.5) in order to emphasize the fact that the cone Cone(A → B) can be presented as a sum hA⊕B deformed by an extra differential A f − → B. Moreover, when we work with bi-graded Khovanov complexes, there may be some confusion about which of two gradings is homological, but our non-standard notation (1.5) specifies all degree shifts explicitly.
The homological order |O| h of an object O ∈ Kom + (A) is the minimum number m, for which O can be presented by a complex (1.3). Consider a direct system of complexes of Kom + (A): A 0 → A 1 → · · · . If this system is 'Cauchy', that is, if for the cones B i = Cone(A i−1 → A i ) there is a limit lim i→∞ |B i | h = ∞, then there exists a direct limit lim → A i .
Since A i ∼ Cone(h −1 B i → A i−1 ), the direct limit lim → A i can be viewed as a result of attaching the complexes B i one after another to the initial complex B 0 = A 0 , hence we use the following notation for the complex lim → A i : (1. 6) In fact, if all B i are 'homologially minimal' representatives of their equivalence classes, then the sum ∞ i=0 B i is well-defined (every chain object is finitely generated) and lim → A i is homotopy equivalent to ∞ i=0 B i defomed by adding extra differentials B i f ij − → B j for all pairs i > j.
We refer to the r.h.s. of eq.(1.6) as a multi-cone, and we also use a similar notation for the complex (1.3): A = · · · → h i A i → · · · ∞ i=m . Note the use of the functor h to set explicitly the correct homological degree of the chain object A i in the multi-cone.
If a multi-cone A is generated by complexes B a : (1.7) (where m ij,a are multiplicities) but we do not care how those complexes are arranged within the multi-cone, then we use a 'lump sum' notation because, as a complex, A is a sum of B a with total multiplicities i m ij,a deformed by an extra differential depicted as .
If the category A is abelian, then we can compute the homology of the multi-cone (1.6) with the help of the filtered complex spectral sequence. The E 1 term of this spectral sequence is the sum of homologies of B i : E 1 = ∞ i=0 H(B i ) and it is determined by the lump sum form of the multi-cone.
Remark 1.2. Since subsequent terms in the spectral sequence get only smaller, there is a bound on the homological order of the homology of (1.6) in terms of homological orders of its constituent complexes: In particular, for the lump sum multi-cone (1.7) 1.5. Khovanov homology. In defining Khovanov complexes [Kho99] for tangles we follow the cobordism based approach of D. Bar-Natan [BN05], albeit with a different grading convention. We still have two degrees: h-degree deg h and q-degree deg q , and we use the notations h and q for their translation functors (these functors increase the corresponding degrees by 1). The q-degree is the genuine homological degree: it takes values in Z and its parity determines the sign factors. The h-degree is 'pseudo-homological', it takes values in 1 2 Z and it has no impact on signs, however it is the h-degree shift functor h which is present explicitly in the Khovanov bracket.
In our notations, Khovanov bracket of a crossing and of the unknot are where h and q are degree shift functors, while s is the morphism corresponding to the saddle cobordism. Note that deg h s = −1, while deg q s = 1, so s is odd.
Thus defined, Khovanov bracket is invariant under the first Reidemeister move only up to a degree shift: (1.9) Relations (1.8) transform into the relations (1.2) after the substitution h → q, q → −q, hence in our notations the graded Euler characteristic of Khovanov homology of a framed link equals its Jones polynomial: Draft of Htail22new: 5/5/2014 12:13 We will use the Khovanov bracket notation − very sparingly, because it clutters the pictures, especially when the diagrams are big. Nevertheless, we hope that the distinction between diagrams and their Khovanov complexes will be clear. Actually, we blur this distinction further by allowing the presence of categorified Jones-Wenzl projectors within diagrams, since, strictly speaking, projectors are not diagrams but rather complexes within Bar-Natan's universal category.
1.6. A categorified Jones-Wenzl projector. An (a, b)-tangle is an embedding of circles and segments, the segment endpoints coinciding with initial a points or final b points. Imagine that the tangle goes from the bottom up. Depending on the position of its endpoints, the segment is either straight, or a cap, or a cup. If one of its endpoints is initial and the other is final (so the segment goes straight through the tangle), then the segment is straight, if both endpoints are initial, then the segment is a cap, and if both of its the segments are final, then the segment is a cup.
The width |τ | wd of a tangle τ is the number of its straight segments. An (a, a) tangle has an equal number of cups and caps, we call this number a width deficit and denote it as |τ | df . Obviously, |τ | df = 1 2 (a − |τ | wd ).
A Temperley-Lieb (TL) tangle is a flat tangle which contains no circles. Let T a be the set of all (a, a) TL tangles. The categorified Jones-Wenzl projector a was constructed independently by Frenkel, Stroppel and Sussan [FSS], Cooper and Krushkal [CK12] and by the author [Roz]. It satisfies three essential properties: it is a projector: (1.10) it annihilates cups and caps: and it has a presentation as a cone of an identity braid and a complex a generated by TL tangles with positive width deficit and with non-negative h-degree and q-degree shifts (1.13) and µ ij,γ are multiplicities.
Draft of Htail22new: 5/5/2014 12:13 1.7. Khovanov bracket of colored tangles. We define Khovanov bracket of colored tangles by cabling tangle components and adding at least one categorified Jones-Wenzl projector to each tangle component. This means that we allow semi-infinite complexes which may extend infinitely far into positive homological degree.
The colored Khovanov bracket is independent of the framing up to a degree shift: (1.14) 2. Results 2.1. Overview. Let H ∞ (L) be the direct limit of the direct system determined by the sequence of maps associated with the direct limit (2.8) are isomorphisms onH Kh i,• (L N ) for i ≤ N − 1, hence the direct limit H ∞ (L) is finite-dimensional in every bi-degree, it satisfies the bounds H ∞ i,j (L) = 0 at the conditions (2.2)-(2.5) and dim H ∞ 0,0 (L N ) = 1. (2.10) Remark 2.4. To define the maps (2.7) we have to choose a diagram D representing L, however we expect that the maps can be defined canonically, that is, independently of that choice.
2.1.3.
Relation to the tail of the Jones polynomial. The bounds (2.2) and (2.3) on H ∞ (L) mean that the graded Euler characteristic of the tail homology is well-defined, because in its presentation as an alternating sum of homology dimensions there is only a finite number of non-trivial terms for any given value of i + j.
The bound (2.3) indicates thatH Kh i,j (L N ) and H ∞ i,j (L) are trivial when i + j < 1 2 i − 1 2 χ ! L , hence their high h-degrees contribute only to coefficients at high powers of q in the graded Euler characteristic. Since the map (2.9) is an isomorphism at low h-degrees, we come to the following: Theorem 2.5. The graded Euler characteristic of the tail homology determines the lower powers of q in the unicolored Jones polynomial of a B-adequate link: This means that the tail homology categorifies the tail of the unicolored Jones polynomial of B-adequate links studied by C. Armond [Arm] and by S. Garoufalidis and T. Le [GL]. Corollary 2.7. If L β is a circular closure of a connected negative braid β, then the tail homology of L β is that of an unknot: Proof. It is easy to see that L β is B-adequate and a reduced B-diagram of L β consists of a single circle without struts. ✷ Corollary 2.8 (Invariance under strut doubling). If L is a B-adequate link and L ′ is constructed by performing a replacement Proof. Obviously, L and L ′ have the same B-reduction. ✷ The latter corollary is a categorification of a similar property of the tail of the unicolored Jones polynomial observed by C. Armond and O. Dasbach [AD, Arm]. This property suggests that a single crossing plays the role of a categorified Jones-Wenzl projector in the tail homology. In order to make this statement precise, tail homology has to be defined for knotted graphs, which may include both finite and infinite colors, so that the essential property of contracting cups/caps can be formulated. We hope to address this issue in a subsequent paper. Meanwhile, we prove in Appendix that for large N a crossing of two N-cables is, indeed, homologically close to the Jones-Wenzl projector placed on two parallel N-cables.
Technicalities.
We prove most statements of the previous subsection not just for Badequate links, but for any link diagram. However, we conjecture that the results are trivial for B-inadequate diagrams, because their tail homology is trivial, if defined as a direct limit of the system that we construct. We expect that B-inadequate links also have tail homology, but the proof that the tail of their Khovanov homology stabilizes in the limit of large color requires new ideas.
Draft of Htail22new: 5/5/2014 12:13 2.2.1. Shifted Khovanov homology. Let D be a diagram of a tangle which may include single lines, cables and Jones-Wenzl projectors. We define n × (D) to be the total number of single line crossings in D (that is, a crossing between an a-cable and a b-cable contributes ab to n × (D)). The following is an obvious corollary of eq. (1.8) and the fact that, according to (1.12) and (1.13), | | h = 0: Theorem 2.9. The complex D has a lower homological bound: Let D be a diagram of a link which may include single lines, cables and Jones-Wenzl projectors. Define n • (D) to be the total number of circles in the diagram constructed from D by replacing the Jones-Wenzl projectors with identity braids and performing B-splicings on all crossings. Now we define the shifted Khovanov homology of D: The following is a particular case of Theorem 2.9: For a link diagram D let D N denote the corresponding unicolored diagram (that is, every link components is N-cabled and contains at least one Jones-Wenzl projector). Then, obviously, n × (D N ) = N 2 χ D and n • (D N ) = Nκ D , sõ (2.12) Hence, if D is B-adequate and represents a link L, thenH Kh (D N ) coincides with the shifted homology defined by eq.(2.1). Now Theorem 2.1 is a corollary of Theorem 2.10 and the following: Theorem 2.11. A shifted homology of a unicolored diagram D N has a bound:H Kh i,j (D N ) = 0 if one of the following conditions is satisfied: Theorem 2.2 is a special case of the following: Theorem 2.12. For any link diagram D there is a sequence of degree preserving maps This theorem implies that the direct system formed by maps (2.16) has a limit which is finite-dimensional in every bi-degree.
Conjecture 2.13. If the diagram D is not B-adequate, then the direct limit (2.17) is trivial: 2.3. Discussion. We conjecture that B-adequate links have a tri-graded homology H ∼ (L), which has an additional 'b-grading', whose zero-degree part coincides with the tail homology H ∞ (L): This homology should have a family of mutually anti-commuting differentials d N , where a L is a constant determined by L.
There are three reasons to formulate this conjecture. The first reason is that the proof of Theorem 2.2 is based on numerous long exact sequences (3.2), in which the 'correction homology' starts at homological degree proportional to N.
The second reason is that, according to Garoufalidis and Le [GL], the tail of the Jones polynomial of B-adequate links has a 'telescopic' structure. They show that if L is alternating, then there exists a family of Laurent series Φ n (q) = m a m q m such that for any k > 0 the combined series F k (q) = k n=0 Φ n (q) is a better approximation for the tail of the colored Jones polynomial that just the first term Φ 0 (q) = J L,∞ (q). With the help of the colored Kauffman bracket (cf. (3.10)) we can prove a similar result for all B-adequate links and we expect that the 2-variable series is the bi-graded Euler characteristic of the tri-graded homology H ∼ (L).
The third reason for our conjecture comes from the paper by Gukov and Stošić [GS]. Based on QFT models of Khovanov homology, they suggest that its dependence on color should be Draft of Htail22new: 5/5/2014 12:13 similar to the dependence of the SU(n) homology on n: this homology may be presented as a homology of a special differential acting on SU(N) homology if N > n. We suggest to go half step further. Ultimately, the SU(N) homology may be presented, at least, conjecturally, from the tri-graded HOMFLY-PT homology with the help of special differentials d n and we expect that a similar process may work for the tail homology.
We expect that the formation of a stable tail of a unicolored B-adequate link is a general feature which originates in the tri-graded homology when the Young diagram describing the color has a very large value of one of the differences between the lengths of rows or columns. In particular, it could be easy to follow the tail formation in case when the diagram consists of a single very large column.
Witten suggested [Wit] that a series of the form (2.19) should represent the graded Euler characteristic of Khovanov homology in the background of a flat U(1)-reducible SU(2) connection in the link complement. We conjecture that if a link can be presented as a circular closure of a totally negative braid, then the tail homology coincides with the one related to the flat U(1)-reducible SU(2) connection.
2.4. Acknowledgements. The author thanks Eugene Gorsky for sharing the results of his unfinished research and, in particular, the conjecture about the structure of Khovanov homology of a colored unknot and its stabilization in the limit of high color.
This work was supported in part by the NSF grants DMS-0808974 and DMS-1108727.
Five tools
The proof of Theorems 2.11 and 2.12 requires five tools: local transformations, purging, braid straightening, colored Khovanov bracket and recurrence relations between categorified Jones-Wenzl projectors. Local transformations relate homologies of similar diagrams. Purging gets rid of redundant TL tangles in complexes, which contain Jones-Wenzl projectors, thus improving estimates of homological order. Straightening a braid is a simple observation that braiding within a cable attached to a projector results only in degree shifts. The colored Khovanov bracket is a special presentation of a crossing of two cables attached to Jones-Wenzl projectors. Finally, recurrence relations relate Jones-Wenzl projectors on N and N + 1 strands.
3.1. Local replacments and local transformations. A local replacement is a pair of tangles τ i and τ f , which may contain single and cabled lines, as well as Jones-Wenzl projectors. Both tangles should have the same sets of incoming legs and the same sets of outgoing legs. Hence if an initial diagram D i contains the tangle τ i attached by its legs to the rest of the Draft of Htail22new: 5/5/2014 12:13 diagram, then we can construct a final diagram D f by replacing τ i with τ f . If τ i or τ f is not an actual diagram, but rather a complex of diagrams within the universal category, the local replacement still makes sense as a construction of D f from D i .
A local transformation is a local replacement together with a specified degree preserving Up to a degree shift, the 'correction' complex τ c may be the categorification complex of an actual tangle τ c , or it may be just a convenient shortcut.
Let D i be a diagram of a link which contains τ i and let D f and D c be the diagrams constructed by replacing τ i with τ f and τ c . The relations (3.1) imply a long exact sequence For all local transformations considered in this paper, there are relations hence eq. (3.2) turns into the following sequence of degree preserving maps between shifted homologies: . This exact sequence implies the following: Purging. Purging is a process of using eq. (1.11) to remove constituent TL tangles of a complex, whose cups or caps are connected directly to a Jones-Wenzl projector. Let T a,b be the set of all (a,b) TL tangles and let T ⊃ is a subset of (a, b) TL tangles, which contain no cups, but only caps and straight segments.
1 The unnatural direction of morphism is chosen for future convenience.
Draft of Htail22new: 5/5/2014 12:13 Proposition 3.2. There is a homotopy equivalence 3.3. Straightening a braid attached to a Jones-Wenzl projector. Proof. We prove this theorem by induction over a. At a = 1 it amounts to Khovanov bracket (1.8). Suppose that it holds for some a and consider the crossing of two (a + 1)cables. We split each cable into an a-cable and a single line and apply eqs. to the crossing of a-cables: (3.11) The categorification complex of a constituent tangle of the resulting multi-cone can be simplified: Here the first homotopy equivalence follows from eq. (3.6) and the second one is the application of Khovanov bracket (1.8) to the crossing of two single lines. We substitute eq.(3.12) for every constituent tangle in both multi-cones of eq. (3.11). The lump sum multi-cone transforms into the r.h.s. of eq.(3.10) for the intersection of two (a + 1)-cables with the help of a simple identity Associativity of the cone operation implies that the second multi-cone of eq. (3.11) can be brought to the linear form of the r.h.s. of eq.(3.8), so it remains to verify inequalities (3.9). The first inequality follows from the lump sum multi-cone formula (3.10) which we have just proved. Let us verify the second inequality for two tangles of the cone (3.12) after they appear through the substitution in the second multi-cone of eq. (3.11). Since every constituent tangle of (3.11) is replaced by a cone of two tangles, the second tangle of the cone (3.12) will appear at the multi-cone position i ′ = 2i and the second inequality of (3.9) Draft of Htail22new: 5/5/2014 12:13 for it obviously holds. The first tangle of (3.11) appears at the position i ′ = 2i + 1 and it carries k ′ = k + 1. The inequality 2i ′ + 1 ≥ 2 k ′ − 1 follows easily from the assumed inequality i ≤ 2 k − 1. ✷ 3.5. Recurrence relations for categorified Jones-Wenzl projectors.
Proposition 3.6. A larger Jones-Wenzl projector absorbs a smaller one: . (3.13) Proof. In view of eqs.(1.12) and (1.13) for a = N, this equivalence is a result of purging the smaller projector with the larger one. ✷ Let us introduce a shortcut notation: where the complex is defined by eq.(1.12).
Proposition 3.7. Thus defined, the complex has a multi-cone presentation where µ ij are the multiplicities of the Temperley-Lieb tangle inside the dotted box, with which it appears in the r.h.s. of eq. (1.13).
Proof. We purge the complex with the help of two N-strand Jones-Wenzl projectors. The tangle in the dotted box is the only (N + 1, N + 1) TL tangle which is not contracted when sandwiched between them. Proof. Consider a sequence of homotopy equivalences The first homotopy equivalence comes from eq.(3.13), the second follows from eq.(1.12) and the last one follows from eqs.(3.14) and (1.10). ✷ Theorem 3.9. The (N + 1)-strand categorified Jones-Wenzl projector can be presented as the following cone: Proof. Consider the composition of the line winding around the N-cable with the left portion of the complex which generates the multi-cone (3.15): Here the first equivalence is purely topological: the projector is moved left along the cable, the second equivalence uses eq. (1.9) to remove two framing kinks on the single line and the third equivalence follows from eq. (3.6). The equivalence (3.19) comes from applying equivalence (3.20) to every constituent complex in the multi-cone (3.15). ✷ Proof of Theorem 3.9. Eq. (3.18) follows from a sequence of homotopy equivalences: Here the first equivalence follows from eq.(3.6), the second equivalence follows from eq.(3.16), the third equivalence follows from eqs. Theorem 3.11. The (N + 1)-strand categorified Jones-Wenzl projector is homotopy equivalent to a cone in which the complex has the following multi-cone presentation: If we connect the endpoints of the upper single line in eq. (3.21) and apply the framing relation (1.9) to the last diagram in that relation, then we come to the following corollary Corollary 3.13. There is a homotopy equivalence ( The map f N of eq.(2.16) is a composition of many maps between Khovanov homologies of a sequence of diagrams related by local transformations, the first diagram in that sequence being D N +1 and the last being D N (recall that maps go backwards).
We use three types of local transformations, which are based on the following local replacements: The transition from the diagram D N +1 to D N is performed in two stages. At the first stage we apply the first replacement of (4.2) to every crossing of D N +1 . The result is the diagramD N , which consists of two parts connected at Jones-Wenzl projectors. The first part is the N-cabled diagram D N and the second part consists of non-intersecting circles formed by single lines appearing in the final diagrams of replacements (I) of (4.2). These single line circles go along the B-circles. We orient them clockwise and assume that in our pictures the clockwise orientation corresponds to the direction from the left to the right. The circles are attached to D N at the Jones-Wenzl projectors and those junctions have four possible forms: At the second stage of the transition from D N +1 to D N we remove the single circle lines ofD N one-by-one. In order to remove a particular circle we select an 'initial' Jones-Wenzl projector on it and then detach the single lines from other projectors going clockwise. During this process, the single line between the initial and current Jones-Wenzl projectors are kept on the same side of the B-circles. If the current projector has the incoming and outgoing single lines on the opposite sides of the B-circle (third and fourth type of (4.3)) then, prior to detachment, we perform the following transformation for the junction of the third type (and a similar transformation for the fourth type): In these pictures the left projector is initial, the right projector is current, the first homotopy equivalence comes from the Reidemeister moves, while the second equivalence comes from eq.(3.6). Note that the single line between the initial and current projectors is kept always above the rest of the diagram.
After the single lines attached to the current projector are brought to the same side of the B-circle, we detach the single line from that projector by the local replacement (II) of (4.2) and pass to the next projector on the single line.
The single line is kept above the rest of the diagram, so once it is detached from all projectors except the initial one, it can be contracted to a small loop attached to that initial projector with the help of Reidemeister moves. The final step is the removal of that loop by the replacement (III) of (4.2). After all single line circles are removed, the diagramD N becomes D N .
Our transition from D N +1 to D N is generally similar to that used by C. Armond [Arm], but the details are different. In particular, we do not replace (N + 1)-cable crossings by projectors, but rather apply replacements (I) of (4.2) directly to the crossings.
4.2.
Local transformations generate isomorphisms at low h-degrees. We describe the local transformations related to replacements (4.2) and show that the corresponding maps (3.3) between shifted homologies are isomorphisms at low h-degrees, thus proving Theorem 2.12. Proof. Let D c be the diagram constructed by performing the local replacement τ i τ c on the current vertex. We estimate the homological order of H Kh (D c ) with the help of Theorem 2.9:
Local transformation I. Set
2 of τ c in eq. (4.5)) and the claim of the theorem follows from Proposition 3.1. ✷ has a homological bound |τ | h ≥ − 1 2 N 2 .
Local transformation II. Set
Remark 4.4. This bound is better than the crude bound of Theorem 2.9. In fact, it coincides with that bound, if we neglect the intersections between the single line and the N-cables.
Proof of Lemma 4.3. Applying eq.(3.10) to the N-cable crossing in τ we get the presentation Proof of Proposition 4.2. Let D c be the diagram constructed from D i by replacing the current projector (τ i of eq.(4.6)) with the tangle complex τ c of eq.(4.6). By Proposition 3.1, we have to prove the bound: Since the complex τ c of eq.(4.6) is a multi-cone (3.15) generated by an 'elementary' tangle then, according to Remark 1.2, it is sufficient to prove where D e is the diagram constructed by replacing τ i in D i with τ e .
Consider a tangle within D e which consists of the right half of τ e and the cable crossing which follows the current projector and transform its complex with the help of two homotopy equivalences: (4.10) The first equivalence comes from sliding the upper left projector down right along its Ncable, and the second equivalence comes from eq.(3.6). The dashed line indicates the possible presence of another single line which has not been removed yet, however, it plays no role in these calculations.
Let D ′ e denote the diagram D e in which the left tangle of eq.(4.10) has been replaced by the right tangle, then (4.11) We would like to estimate |D ′ e | h with the help of Theorem 2.9. In doing so we would have to take into account possible crossings coming from the stretch of the single line between the initial projector and the left projector of the tangle τ e of eq.(4.8) and N-cables participating in the crossings attached to the current single line circle. These new crossings are generated by the Reidemeister moves involved in the first homotopy equivalence of eq.(4.4): when a single line is flipped to the other side of the circle, it may come across the N-cable crossings, from which parts of this line originate through replacements I of (4.2) (see the picture (4.7) of the tangle τ ). However, Remark 4.4 indicates that these crossings between the single line and the Draft of Htail22new: 5/5/2014 12:13 N-cables may be ignored when applying the estimate of Theorem 2.9, so |D ′ e | h ≥ − 1 2 n ′ × (D ′ e ), where n ′ × (D ′ e ) is the number of single line intersections within D ′ e , except those which we can ignore.
The cable intersection of the left tangle of eq.(4.10) involves two N-cables, while the same intersection in the right tangle involves a N-cable and a (N − 1)-cable, hence n ′ × (D ′ e ) = n × (D i ) − N and the inequality (4.9) follows from eq.(4.11). ✷
Proposition 4.5. Let D i be a diagram constructed by removing some single line circles from D N and by detaching the 'current' single line circle from the all of its projectors, except the initial one, to which it is attached as in the picture (4.12) of tangle τ i . Let D f be the diagram D i from which this circle is completely removed. Then the degree preserving map (3.3) is an isomorphism onH Kh i,• for i ≤ 2N − 2.
Proof. Since the complex τ c of eq.(4.12) is a multi-cone (3.25) generated by the elementary tangle τ f of eq. (4.12), then, according to Remark 1.2, the claim of this proposition would follow from the bound H Kh i,• (D f ) = 0 for i ≤ − 1 2 n × (D i ) − 1. The latter follows from Theorem 2.9 coupled with an obvious relation n × (D i ) = n × (D f ). ✷ 4.3. Proof of Theorem 2.12. Of all three types of local transformations considered in Propositions 4.1, 4.2 and 4.5, it is the local transformation (4.6) which yields the weakest estimate of the homological degrees at which the map (2.16) is an isomorphism, and this is the estimate of Theorem 2.12 ✷
Proof of Theorem 2.11
In order to compute the shifted homologyH Kh (D N ), we apply the colored Khovanov bracket (3.10) to all crossings of D N . As a result, this diagram turns into a multi-cone of flat diagrams of a special kind. Let V be the set of crossings of D. A state of D N is a The diagrams D s for all states s generate a multi-cone presentation of D N , henceH Kh (D N ) can be computed by spectral sequence, and its E 1 term is a sum of appropriately shifted homologies H Kh (D s ): where ||s|| = v∈V s 2 v . Hence, a component E 1 i,j of bi-degree i, j (both are homological and have nothing to do with filtration!) has the form As we already noted in Remark 1.2, further steps of spectral sequence may only reduce homology, hence E 1 i,j = 0 impliesH Kh i,j (D N ) = 0. Moreover, all differentials have bi-grading (-1,1), hence E 1 i+1,j−1 = E 1 i−1,j+1 = 0 impliesH Kh i,j (D N ) = E 1 i,j . These arguments imply that Theorem 2.11 follows from the proposition Proposition 5.1. E 1 i,j = 0 if one of the following conditions is satisfied: Thus we proved Theorem 2.12 ✷ In view of eq.(5.1), Proposition 5.1 follows from the next one: Draft of Htail22new: 5/5/2014 12:13 Proposition 5.2. H Kh i,j (D s ) = 0 if one of the following two conditions is satisfied: Proof of Proposition 5.1. Conditions (5.2)-(5.4) follow easily from the conditions (5.6)-(5.8).
In order to prove eq. (5.5), observe that according to eq. (5.5), Since i ′ ≥ 0 by eq. (5.6) and k ≥ 0 by eq. (5.1), then in view of the bound (5.8) with χ in D = 0 we conclude that non-trivial contributing homology exists only for i ′ = k = 0, so j ′ = −||s|| − Nκ D and i = ||s||. Thus we proved that E 1 i,−i is a sum of homologies H Kh 0,−||s||−N κ D (D s ) with ||s|| = i, hence eq. (5.5) follows from eq. (5.9) and from the fact that the state s with ||s|| = 0 is unique (it corresponds to B-splicing all crossings in D N ) and its multiplicity in the presentation of E 1 i,−i is one, because complete B-splicing has multiplicity one in eq.(3.10). ✷ Proof of Proposition 5.2. First of all, we observe that the bound (5.6) follows from the fact that D s has no crossings, while the formulas (1.12), (1.13) for the categorified Jones-Wenzl projector contain only non-negative shifts of h-degree.
The proof of other bounds requires a simplification of the complex, whose homology yields Khovanov homology H Kh (D s ). We cut the diagram D s into pieces (tangles), simplify their Khovanov complexes and then glue those complexes back together. In other words, a tangle contains no cups, but only caps and straight segments.
Lemma 5.3. The Khovanov categorification complex of the colored tangle diagram τ s,c can be presented in the form where the diagrams τ are of one of two types depicted in Fig. 1: with N 1 , N 2 ≤ N.
By gluing the complexes (5.10) back together we get a complex such that H Kh (D s ) = H( D s ). The 'circle diagrams' D • which result from gluing the diagrams of Fig. 1 at the strut line cutting points and replacing projectors with complexes (1.12) and (1.13), consist of multiple single line circles. In view of the second formula of eq.(1.8), the lowest q-degree in the homology H Kh (D s ) may be bounded by the highest number of circles in those circle diagrams.
The circles in circle diagrams are of three types. The first type is jumping circles: they contain at least one strut line. The circles of the second and third type stay within the same B-circle. A straight circle goes along its B-circle, passing straight through each Jones-Wenzl projector on its way. A winding circle changes its direction at least twice, because it contains at least one cup and one cap of a constituent TL tangle coming from one of projectors.
Let us prove the inequalities (5.7) and (5.8) by finding upper bounds for the numbers n j , n s and n w of jumping, straight and winding circles respectively in a circle diagram.
We begin with n j . A jumping circle must contain at least two strut lines of an adequate crossing or at least one strut line of an inadequate crossing, so the number of jumping circles n j has a bound: (5.11) where V ad , V in ⊂ V are the subsets of B-adequate and B-inadequate crossings. The obvious inequalities 2s v ≤ s 2 v + 1, (the third inequality uses the fact that s v is integer) indicate that the bound (5.11) implies to other bounds: (5.12) which means that the first three terms in the r.h.s. of the inequality (5.7) and the first two terms in the inequality (5.8) bound the negative contribution of jumping circles to the q-degree of H Kh (D s ).
Next we prove the bound n s,c + n w,c ≤ N, where n s,c and n w,c are the numbers of straight and winding circles within any given B-circle c. It implies the bound n s + n w ≤ Nκ D and combined with the bounds (5.12) they imply the bounds (5.7) and (5.8). In order to prove the bound (5.13), we observe that in the first diagram a straight circle contains one strand from the N 1 -cable and one strand from the N 2 -cable, while a winding circle contains at least two strands from one of these cables, hence there is a bound n s,c + n w,c ≤ 1 2 (N 1 + N 2 ) ≤ N. (5.14) The second diagram is treated similarly, if we set N 2 = 0 in the previous argument. Thus we proved the bounds (5.7) and (5.8).
It remains to prove eq. (5.9). Since this time D is B-adequate, the second inequality of (5.12) becomes n j ≤ ||s||. Since we consider only homology of zeroth h-degree, then according to eqs. (1.12) and (1.13), we may replace Jones-Wenzl projectors with identity braids, so there is only one circle diagram D • contributing to H Kh 0,−||s||−N κ D (D s ), and this circle diagram has no winding circles: n w = 0. Furthermore, n s ≤ Nκ D , but if ||s|| = 0, then there is at least one pair of strut lines in D s , so n s < N and n j + n s < ||s|| + Nκ D , hence H Kh 0,−||s||−N κ D (D s ) = 0. If ||s|| = 0, then D s has no strut lines and consists of disjoint N-cabled circles, so the relevant circle diagram D • consists of Nκ D single-line circles, and H Kh 0,−N κ D (D s ) = Q follows from the second equation of (1.8). ✷ Proof of Lemma 5.10. We prove the lemma by 'purging' categorified Jones-Wenzl projectors appearing in the tangle τ s,c . In order to bring the complex τ s,c to the form (5.10) with diagrams τ depicted in Fig. 1, we insert two extra Jones-Wenzl projectors side-by-side at any place on the cable which runs along the B-circle. Then we go from the front one (relative to the clockwise orientation) to the back one in the clockwise direction, purging each preexisting projector that appears on our way. It is easy to prove by induction that after every projector purge we get a multi-cone presentation whose constituent diagrams τ have one of two possible forms between the front projector and the first unpurged projector (which lies in the pictures to the left of the dashed line) depicted in Fig. 2 The proof of this lemma is similar to proofs of Section 4: we show that D N can be constructed from D ′ N with the help of a local transformation and prove the homological smallness of the correction diagram.
We need a simple corollary of Theorem 3.5 Corollary 6.2. The Khovanov bracket formula (3.8) for the colored crossing can be recast in the form Proof. According to the second inequality of (3.9), k(0) = 0, hence the the right tangle of the cone (6.1) is the one that appears at i = 0 in the multi-cone (3.8). According to eq. (3.10), this tangle has multiplicity one, so this is the only place where it may appear in that multi-cone, and it has a zero shift of h-degree. ✷ Draft of Htail22new: 5/5/2014 12:13 Proof of Lemma 6.1. Since D ′ is constructed from D by a removal of a single crossing (strut of B(D)), we set so that D i = D and D f = D ′ , while the relation (3.1) comes from (6.1) if we set m f = 0 and n f = − 1 2 N 2 . Since n × (τ i ) = N 2 , in view of Proposition 3.1 it remains to establish the bound H Kh i,• (D c ) = 0, if i ≤ N. Let D c,k be the diagram D c in which τ c is replaced by a constituent tangle τ c,k from the r.h.s. of eq.(6.2), in which one strand of a k-cable is separated from the others: The lump sum presentation (6.2) of τ c allows us to use Remark 1.2: it is sufficient to establish a boundH because the tangle τ c,k has an extra h-degree shift h k 2 in the lump sum formula (6.2) and k 2 − k ≥ 0.
Consider the portion of D c,k between the left k-cable of τ c,k and another crossing which connects the same B-circles: · · · · · · (6.5) Let D ′ c,k be the diagram constructed from D c,k by replacing the subdiagram (6.4) with (6.5). According to Propositions 4.1 and 4.2 there is a map of shifted Khovanov homologies H Kh (D ′ c,k ) →H Kh (D c,k ), which is an isomorphism onH Kh i,• (D c,k ) for i ≤ N − k. We are going to show thatH Here the first homotopy equivalence comes from sliding k-cable projectors to the left along N-cables and then contracting double projectors into single ones, while the second homotopy comes from eqs.(1.14) and (3.6). Let D ′′ c,k be the diagram constructed from D ′ c,k by replacing the left tangle of eq.(6.7) with the right tangle. Since n × (D ′′ c,k ) = n × (D ′ c,k ) + (N − k) 2 − N 2 , homotopy equivalence (6.7) implies the isomorphism of shifted Khovanov homologies H Kh (D ′ c,k ) = h k(2N −k) q kHKh (D ′′ c,k ) and the bound (6.6) follows from Theorem 2.10. Here the first homotopy equivalence comes from eq.(3.6), while the second equivalence comes from eq.(1.12).
In order to put a bound on the homological order of D c , we purge the gray box in τ c , that is, we contract all constituent TL tangles, whose cup or cap is connected directly to Jones-Wenzl projectors sitting on N-cables. After the purge, the complex of τ c takes the form We used homotopy equivalence (6.7). Note that there are no tangles with k = 0, because the complex (1.13) does not contain identity braids. | 2012-04-02T21:03:00.000Z | 2012-03-26T00:00:00.000 | {
"year": 2012,
"sha1": "25544b288d08b7490fc6da47c72ca6a6d06a076d",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1203.5741",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "25544b288d08b7490fc6da47c72ca6a6d06a076d",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
158367463 | pes2o/s2orc | v3-fos-license | Producer firms, technology diffusion and spillovers to local suppliers: Examining the effects of Foreign Direct Investment and the technology gap
In this paper, we conduct a detailed examination of the effects of Foreign Direct Investment (FDI) and the technology gap on local technology dissemination and spillovers. Using unique firm level data from surveys among FDI firms and domestic producer firms and a random sample of their suppliers in the North East of Mexico, we present new evidence showing that the effects of FDI and the technology gap are multi-faceted. FDI firms are significantly more involved in technology transfers. Their suppliers are more likely to experience positive spillovers, even when we control for the support that they receive. A large technology gap has a negative effect on technology diffusion. However, suppliers of FDI firms are more likely to receive several types of technology transfer when the technology gap with their client firms is large. As for local spillovers, we find that a large technology gap fosters positive spillovers, especially among suppliers of FDI firms and among those suppliers that are best suited to absorb new technologies.
Introduction
Multinational enterprises (MNEs) are a key driving force behind the international dissemination of new technologies (Iammarino and McCann, 2013;Keller, 2010). By engaging in Foreign Direct Investment (FDI), MNEs introduce new technologies into many developed and developing economies (Dunning and Lundan, 2008). Not only do FDI firms enhance the technology stock of a host economy, additional positive effects arise when FDI firms generate positive spillovers. Through channels including demonstration effects, inter-firm linkages and labour turnover domestic firms can absorb new technologies from FDI firms, culminating into positive productivity effects (Blomstro¨m and Kokko, 1998;Go¨rg and Greenaway, 2004).
Despite the growing belief that FDI spillovers constitute an important source of new technologies, the evidence from traditional econometric studies is very mixed. There is little consensus on the prevalence of intra-industry spillovers (Hanousek et al., 2011;Irsova and Havranek, 2013). Evidence on inter-industry spillovers is more prominent, as several studies report a positive relation between FDI and productivity of domestic firms in upstream industries. This suggests that positive spillovers are likely to materialise among suppliers of FDI firms (Blalock and Gertler, 2008;Jordaan, 2008a). However, the evidence is far too heterogeneous to conclude that such inter-industry spillovers occur automatically or in all settings.
Recent research is characterised by several approaches that try to obtain more evidence and a better understanding of FDI spillovers. One line of investigation that focuses on the geographical dimensions of these externalities finds that FDI spillovers are more pronounced at the regional than at the national level (Driffield, 2006;Menghinello et al., 2010;Monastiriotis and Jordaan, 2011). Importantly, the positive effect of regional proximity applies in particular to inter-industry spillovers between FDI and local suppliers (Blalock and Gertler, 2008;Jordaan, 2008b).
A second research line is characterised by using case studies or small scale surveys to identify the size and nature of linkages between FDI and local suppliers. These more qualitative studies often provide detailed information on the types of linkages that FDI firms create and the nature and degrees of technology transfers that their suppliers may benefit from (e.g. Domanski and Gwosdz, 2009;Liao and Wei, 2013;Potter et al., 2002Potter et al., , 2003. A recent example is Ivarsson and Alvstam (2011) who provide detailed interviewbased evidence showing that a large Swedish home furnishing retailer provides a variety of types of support to its suppliers in China and South-East Asia.
The technology gap between FDI and domestic firms receives only limited attention in these research lines. This is striking, as it is traditionally seen as a potentially key factor influencing spillovers (Crespo and Fontoura, 2007;Sjo¨holm, 1999). Furthermore, studies that do consider the technology gap as a driver of FDI spillovers usually analyse its effect on intra-industry spillovers. Its effect on inter-industry spillovers to local suppliers, the type of externalities that is believed to be more prominent, has received much less attention in the literature.
In this context, the purpose of this study is to examine whether and how FDI and the technology gap affect technology transfers and spillovers among local suppliers in a developing country setting. Our contribution to the literature is three-fold. First, we use data obtained from detailed surveys among producer firms and local suppliers in the North East of Mexico. This allows us to study the effects of FDI and the technology gap both on the creation of technology transfers and on spillovers among local suppliers. Second, the heterogeneous nature of the evidence on FDI spillovers indicates that it is not clear whether FDI firms necessarily differ from domestic firms in their degree of technology dissemination. The rich nature of the dataset allows for a consistent comparison of FDI and domestic firms, enabling us to shed new light onto this issue. Third, we re-examine the concept of the technology gap and conduct a detailed analysis of its effects on both technology transfers by producer firms and inter-industry spillovers among local suppliers.
The paper is constructed as follows. Section ''FDI spillovers and the technology gap'' presents a selective review of the literature on the technology gap and FDI spillovers which we use to inform our research questions. Section ''Technology dissemination and spillovers: Is foreign ownership important?'' describes the dataset and investigates whether foreign ownership matters for local technology diffusion and spillovers. Section ''FDI, technology gap and technology dissemination'' presents a multivariate analysis of drivers of local technology dissemination, focusing on the effects of FDI, the technology gap and their interaction. In section ''FDI, technology gap and spillovers'', we use data from the supplier survey to identify and examine the main drivers of local spillovers. Section ''Summary and implications for theory and policy making'' summarises our findings and discusses implications for theory and policy making.
FDI spillovers and the technology gap Technology gap, technology absorption and catch up
The origins of the interest in the technology gap can be traced back to research from the 1950s-1960s that established that capital and labour accumulation explained only a limited part of economic growth in the major developed economies (Abramovitz, 1989). Instead, technological change became progressively seen as the key factor in processes of growth and development (Fagerberg, 1994). Although neo-classical economics acknowledged the importance of technology, it was perceived as growing at a non-variable rate and available to all countries as a public good (Solow, 1957). Another interpretation deviated strongly from these assumptions, arguing that countries differ markedly in their levels of technology creation, adoption and usage, making international technology gaps a common feature (Fagerberg, 1987).
Accepting the existence of international technology gaps generates the question whether such gaps affect processes of growth and development (Kemeny, 2011). Gershenkron (1962) introduced the concept of ''relative backwardness'', suggesting that countries that are technologically backward can achieve higher growth rates due to their initial disadvantage. 1 By adopting new technologies from leading countries, lagging countries can benefit from large productivity improvements (Abramovitz, 1989). Baumol (1986) and Abramovitz (1986) provide early corroborating evidence of the existence of such catch up effects. However, their findings also show that these effects materialised mainly among the more developed economies, suggesting that these processes do not occur automatically. Also, Kemeny (2011) presents findings suggesting that the technology gap between advanced and lagging countries may not have decreased since the 1970s, questioning the prevalent nature of international processes of technological catch-up. Nelson and Phelps (1966) represent an early recognition of the notion that the technology gap fosters catch up processes under certain conditions. In their model of technology diffusion, the effect of the technology gap is specified as interacting with the level of human capital of the technology adopter. When a country possesses a limited level of human capital, it will be less able to benefit from technology diffusion, possibly preventing catch up processes to materialise. In broader terms, this suggests that lagging countries will be better able to benefit from relative backwardness when they possess a high level of absorptive capacity (Abreu, 2011;Levinthal, 1989, 1990). Absorptive capacity reflects ''the capacity to identify, assimilate and exploit knowledge from the external environment'' (Cohen and Levinthal, 1989: 569). 2
Technology gap and FDI spillovers
In contrast to the original interpretation of the technology gap in the literature on international convergence, the standard interpretation in the FDI spillovers literature argues that a small rather than a large technology gap favours spillovers. This interpretation follows Cantwell (1989), who studied the effects of US firms in the UK for the 1950s-1970s. His findings show that only those domestic firms that were technologically advanced were successful in competing with and benefitting from the presence of US firms. From this, the technology gap has come to be interpreted as a direct inverse indicator of the level of absorptive capacity of domestic firms. Small technological differences are taken to imply that domestic firms can absorb and implement new technologies, allowing positive spillovers to materialise.
Some studies present evidence in line with the notion that a small gap favours spillovers. Kokko (1994) estimates FDI spillovers in Mexican manufacturing industries and finds that positive spillovers only materialise in industries with a small productivity gap between domestic and FDI firms. Kokko et al. (1996) and Taki (2005) present similar evidence for Uruguay and Indonesia. Girma and Go¨rg (2007) estimate FDI spillovers in the electronics and engineering industries in the UK and find that a decreasing technology gap may favour FDI spillovers. Liao and Wei (2013) present findings from a firm level survey of the ICT industry in Suzhou, China that indicate that the technology gap may lower the occurrence of technology linkages.
These findings notwithstanding, it seems clear that the notion that a small technology gap favours FDI spillovers is strongly at odds with its original meaning in the literature on technological catch up. Applying the original interpretation of the technology gap as reflecting the degree of relative backwardness, technological differences between FDI and domestic firms can be seen as an indicator of the potential scope of positive spillovers (Findlay, 1978;Wang and Blomstro¨m, 1992). When the technology gap is large, there is a large potential for domestic firms to learn and benefit from new technologies, fostering positive spillovers.
Incorporating absorptive capacity directly into the concept of the technology gap is also not in line with the literature on international convergence. As Nelson and Phelps (1966) and Levinthal (1989, 1990) clearly argue, positive catch up effects from large technological differences may be conditional or at least facilitated by the capacity of countries or firms to absorb new technologies. However, this is very different from treating the size of technological differences as a direct inverse indicator of the level of absorptive capacity of technology receiving countries and firms.
Furthermore, one could argue that large technological differences generate incentives for domestic firms to try and absorb new technologies. As Blalock and Gertler (2009) for instance argue, a large technology gap implies that domestic firms can start absorbing new technologies by picking ''low hanging fruits'', technologies with relative low marginal costs and high marginal returns. It may also be the case that domestic firms are incentivised to absorb new technologies in response to the increase in competition that the entry of FDI firms creates. With respect to inter-industry spillovers, domestic firms may try to become suppliers to FDI firms that are technologically advanced, given the substantial benefits that they may obtain from learning new technologies from their client firms.
Several studies present findings that support our alternative interpretation of the expected effect of the technology gap on FDI spillovers. Jordaan (2005Jordaan ( , 2008b, Sjo¨holm (1999) and Imbriani et al. (2014) estimate FDI spillovers in Mexico, Indonesia and Italy and find that positive spillovers only occur in industries with a large technology gap. For Mexico, Blomstro¨m and Wolff (1994) present evidence of a direct positive effect of the productivity gap between FDI and domestic firms on productivity of domestic firms. Be´ke´s et al. (2009) and Peri and Urban (2006) similarly report a direct positive productivity effect of the technology gap for Hungary, Italy and Germany. In extension of this, some studies examine the interaction between the presence of FDI and the technology gap. Jordaan (2005Jordaan ( , 2008b finds for Mexico that FDI spillovers are enhanced in industries that are characterised by a large technology gap and a high FDI presence. Similar findings of positive interaction effects between technological differences and FDI are presented by Blalock and Gertler (2009) and Blalock and Simon (2009) for Indonesia, Girma (2005) for the UK and Castellani and Zanfei (2003) for France, Italy and Spain.
Research questions
We distil the following three research questions from the selective literature review. First, do FDI firms create positive spillovers? Given the considerable degree of heterogeneity of the evidence, this is a matter of empirical verification. In line with recent research, we focus on inter-industry spillovers that may occur between FDI firms and local suppliers. Second, does the technology gap between FDI firms and their suppliers foster spillovers? In contrast to the traditional interpretation of the technology gap in the FDI spillovers literature, we interpret the technology gap as reflecting the scope for positive spillovers to materialise. When the technology gap is large, suppliers of FDI firms have ample opportunity and scope to learn and benefit from new technologies that are incorporated into the production processes of their client firms. Third, do the effects of type of ownership and the technology gap on local spillovers interact? If FDI and the technology gap influence spillovers, it may be that there are additional positive effects on spillovers through their interaction. It may even be the case that such interaction effects are required for any positive spillovers to materialise.
Technology dissemination and spillovers: Is foreign ownership important? Data
For our empirical analysis, we use data that was obtained through two surveys among producer firms and local suppliers. The main aim of these surveys was to obtain detailed information on the scale and nature of inter-firm linkages between producers and local suppliers and the impacts that these linkages create among suppliers. The surveys were carried out in 2000-2001 in the Metropolitan Area of Monterrey in the state of Nuevo Leon, constituting the main agglomeration of economic activity in the North East of Mexico. This regional economy has a long history of manufacturing activity and is characterised by a substantial level of foreign participation in modern and trade oriented industries, offering a good setting to study the operations and effects of FDI firms (Jordaan, 2011a(Jordaan, , 2009).
The producer survey was carried out among both FDI and domestic producer firms. 3 Using information from local industry organisations and the American Chamber of Commerce in Mexico, 180 large producer firms were identified. Of these firms, 32 Mexican and 50 FDI firms completed the survey. Second, the dataset contains information from a survey among a random sample of local suppliers. To prevent the construction of a biased sample, the producer firms were not used to obtain contact details of their suppliers. Instead, a list of 1100 manufacturing firms employing fewer than 150 employees was created, representing the potential pool of local suppliers. A telephone survey was carried out among these firms to identify firms that operate as supplier to the producer firms. After a three-month period, all the firms had been contacted and 356 firms had participated. Of these firms, 300 indicated to operate as supplier to the producer firms. All these firms were invited to participate in the supplier survey and 100 firms completed the survey. 4
Comparing FDI and Mexican producer firms
Earlier studies that have used the dataset compare degrees of local sourcing and frequencies of support provided by FDI and Mexican producer firms (see Jordaan, 2011aJordaan, , 2011b. These studies find that FDI firms are significantly more engaged in the provision of several types of support to their local suppliers. In the present study, we extend upon these earlier findings in two ways. First, we rely on data from the supplier survey to determine the effect of FDI on technology transfers. Data from the producer survey may contain biases as it consists of self-reported indicators of technology dissemination. By using information from the firms that are on the receiving end of the technology transfers, such biases are avoided. Second, we use propensity score matching techniques to ascertain the effect of foreign ownership. Unconditional comparisons of frequencies of support may be biased when there are supplier characteristics that are related to a firm being a supplier to FDI firms and the receipt of support. The use of propensity score matching techniques allows for a more unbiased identification of the effect of foreign ownership on local technology transfers. To determine whether FDI firms differ from domestic firms in local technology dissemination, we estimate the average treatment effect of a local supplier producing for FDI client firms. The problem that we face is that for a given outcome variable Y, we only observe outcome Y (1) if a firm is a supplier to FDI firms or Y (0) if the supplier produces for Mexican firms. If we could observe both outcomes for the same supplier, the average treatment effect would resort to the supplier sample average Y (1)-Y (0). In the absence of this, we need to obtain an indicator of Y (0) that allows us to identify the treatment effect.
To do so, we use propensity score matching techniques (Heckman et al., 1997;Rosenbaum and Rubin, 1983). This involves estimating the probability (propensity score) that a firm is a supplier to FDI firms according to a set of co-variates. Firms in the group of suppliers to FDI firms are then matched to firms in the group of suppliers of Mexican firms, using the propensity scores as matching criterion. The average treatment effect of being a supplier to FDI firms is then obtained as the average difference in the outcome variable between the two matched sets of firms (Abadie and Imbens, 2016; Caliendo and Kopeneig, 2006). Table 1 contains the findings from the supplier survey on local technology dissemination and spillovers. The first part of the table lists a variety of types of support. 5 In line with findings from other studies (e.g. Gentile-Lu¨decke and Potter et al., 2003), the provision of product designs and specifications occurs most frequently, followed by assistance with production and quality control. Other types of support that are relatively prominent include the provision of special tools and machinery, assistance with diversification of products, recommending suppliers to other firms and assistance with exporting. Given that we only classify suppliers that indicate to receive technology transfers frequently as being the recipient of support from their client firms, the overall impression is that local technology dissemination is an important feature of the inter-firm relationships between producer firms and their suppliers.
Findings on the effects of foreign ownership
Next, the findings from the unconditional comparisons between suppliers of the two types of producer firms indicate that FDI firms are more supportive. Looking at technological support, the percentage of suppliers of foreign-owned firms that receive special tools, machinery and training is significantly larger than among suppliers of Mexican firms. The findings for organisational support also show that FDI firms are significantly more supportive in several cases. The exception is assistance with start-ups and expansion of existing plants, where suppliers of Mexican firms benefit more often.
The last column of Table 1 presents the average treatment effect of being a supplier to FDI firms. 6 The impression that FDI firms are more engaged in technology transfers is clearly confirmed. Looking at the results for technological support, the size of the estimated treatment effect indicates that the effect of foreign ownership is the strongest for training programs, followed by the provision of machinery and special tools. For organisational support, we identify a positive treatment effect regarding recommending suppliers to other firms and assistance with exporting activities. Again, the only type of support where being a supplier of Mexican firms generates a positive effect is support with start-ups and business expansion. The second part of Table 1 presents findings on the importance of technological and organisational support. The survey enquired about the importance of the support that a firm received for being a successful supplier to the producer firms. The motivation behind this is to assess whether there is a bias caused by suppliers self-selecting into becoming a supplier of FDI firms. It may be that firms that need support try to become a supplier of foreign-owned firms, when they believe that these firms offer more support. If this is the case, the identified higher level of supportiveness of FDI firms can be partially explained by this selection effect. The findings suggest that such a bias is absent, as there are no significant differences between the two groups of suppliers regarding the importance of technological or organisational support.
The last rows of Table 1 report findings on the overall spillover impacts that suppliers experience. A significantly larger percentage of suppliers of FDI firms indicate to have experienced a large positive impact on their performance from their business dealings with their client firms. When distinguishing between a large technological or organisational impact, the results show that foreign ownership matters only for creating large technological improvements among the suppliers. 7 These findings are corroborated by the estimated treatment effects. There is a clear positive treatment effect regarding overall performance improvement for suppliers of FDI firms. Comparing the treatment effects concerning organisational and technological improvement shows that only the latter effect is positive and significant.
FDI, technology gap and technology dissemination
The findings in the previous section provide clear indications that foreign ownership matters for technology dissemination and local spillovers. To obtain evidence on the effects of FDI, the technology gap and their interaction, we turn to multivariate analysis. In this section, we use data from the producer survey to identify drivers of technology transfers. Section five presents our analysis of drivers of spillovers among local suppliers.
Regression model
As dependent variable, we use a dummy variable that takes the value of 1 if a producer firm offers a type of support frequently and 0 otherwise. The binary nature of this variable requires the use of logistic regression techniques. The odds that producer firm i provides support can be defined as the ratio of the probability that the firm provides support (P) i over the probability that the firm does not offer support (1-P) i . The log of this ratio, the logit, can be used as dependent variable in a regression model that can be estimated with maximum likelihood techniques: The variables FDI, TechGap and their interaction are the main variables of interest. FDI is a dummy variable taking the value of 1 when a firm is foreign-owned. TechGap captures the technology gap between a producer firm and its suppliers, measured on a Likert scale as the degree that technological differences with suppliers prevent a producer firm from increasing its level of local sourcing. 8 The interaction term between FDI and TechGap tests whether foreign firms differ from domestic firms in their response to a technology gap with local suppliers.
The vector Xi contains a set of additional control variables. We control for the age and size of producer firms as modern and larger firms may be more supportive (Jordaan, 2011a). We also control for the level of local sourcing and the importance of material inputs in a producer firm's production processes. A high level of local integration into the local economy may foster the provision of support. Similarly, producer firms that rely heavily on using material inputs may be more inclined to offer support to suppliers that produce this type of input.
Next, we control for the nature of the production process of the producer firms. We include the share of blue collar workers in a firm's labour force to capture the overall reliance on assembly style production (Robertson, 2007). In relation to this, we also control for effects of the Maquiladora program (Ramirez, 2003). First generation Maquiladora firms are primarily geared to benefitting from tax breaks on imported inputs and focusing on labour intensive activities. More recently however, second and third generation Maquiladora firms are characterised by a stronger reliance on local suppliers (Sargent and Matthews, 2008). To capture these two opposing features, we add a dummy variable labelled Maquiladora to capture the effect of a firm participating in the Maquiladora program. To capture first generation Maquiladora firms, we include a dummy variable that identifies those Maquiladora firms that have been participating in the program for at least 15 years.
Empirical findings on drivers of technology dissemination
We confine our analysis on drivers of technology dissemination in three ways. First, we are primarily interested in technological support. This type of support is directly linked to actual production processes of suppliers and the findings in Table 1 show that suppliers of FDI firms are significantly more likely to experience large technological improvements. Second, although we have data for 82 producer firms, several firms did not complete all the sections in the survey on the various types of technology transfer that we identified. As we are unable to assess whether the non-response of these firms reflects the non-importance of technology dissemination or an unwillingness to reveal this type of information, we use data for the 66 firms that answered all the relevant survey questions. Third, local suppliers consist of two different types. Suppliers of material inputs produce inputs that producer firms purchase to integrate directly into their production processes. Suppliers of production services perform a variety of treatments on intermediate inputs that are sent to them by producer firms. After treatment, the inputs are returned and re-integrated into the production process of the producer firms. We confine ourselves to discussing the results for drivers of technology dissemination to suppliers of material inputs, as this type of local supplier is used more heavily by the producer firms. Also, suppliers of material inputs receive support more frequently (Jordaan, 2011a), suggesting that producer firms assess technological improvement of these suppliers to be more important. 9 The findings are presented in Table 2. The first column shows the findings for the overall level of technological support. Firm size carries a positive coefficient, indicating that larger firms have more resources to devote to support. Unexpectedly, local sourcing has a negative effect on the provision of technological support. 10 A possible explanation for this effect is that producer firms with a high degree of local sourcing provided more support in an earlier phase when they were increasing their use of local suppliers. The estimated effect of the two Maquiladora variables indicates a structural difference between mature and younger Maquiladora firms, as younger Maquiladora firms are more likely to create technology transfers.
Turning to the main variables of interest, the positive effect of foreign ownership confirms that FDI firms are more likely to offer technological support. 11 The estimated effect of TechGap is negative, indicating that a large technology gap lowers the likelihood that technological support is offered. This may explain findings from econometric studies that identify a negative effect of the technology gap on spillovers. However, rather than that a large technology gap reflects a low level of absorptive capacity of domestic firms, our findings indicate that a large gap lowers the provision of technological support, resulting in lower spillovers.
In contrast, the estimated effect of the interaction between FDI and TechGap is positive. This suggests a difference in the effect of the technology gap on the supportiveness of FDI and Mexican firms. Adding the negative coefficient of the technology gap to the positive coefficient of the interaction variable indicates that the supportiveness of FDI firms is negatively affected by the technology gap. However, the positive coefficient of the interaction variable shows that the negative effect of the technology gap on supportiveness is smaller among FDI firms than among domestic firms. 12 The next set of findings concerns drivers of various types of technological support. The estimated effects of FDI, TechGap and their interaction are similar for these different types of support. FDI firms are more likely to offer support, whereas the technology gap lowers the probability of support. The effect of the interaction variable indicates that the technology gap has two distinct effects on technology dissemination by FDI firms. Regarding the provision of machinery, the negative effect of the technology gap on technology dissemination is less strong among FDI firms. For the other three types of support, the positive interaction effect between the technology gap and FDI is larger than the negative effect of the technology gap. This indicates that when it comes to the provision of product designs, special tools and training, FDI firms interpret the technology gap as an indicator of the scope of suppliers to improve, leading them to offer more support. This provides an explanation for findings in econometric studies that identify a positive effect of the interaction between FDI and the technology gap on productivity spillovers.' 13,14 To further assess the effects of the technology gap we examine whether the likelihood that producer firms provide support changes under increasing sizes of the technology gap. The findings are presented graphically in Figure 1. The top left of Figure 1 shows the findings for overall technological support. The difference between FDI and Mexican producer firms is evident, with FDI firms always being more likely to offer support. However, the figure also shows that the likelihood of providing support decreases when the technology gap becomes larger. The likelihood of FDI firms providing support decreases faster than that of Mexican firms. At the largest size of the technology gap, the difference between the two types of firm has disappeared.
Looking at the findings for the various types of technological support, the results for the provision of product designs and specifications and machinery are similar to what we find for overall technological support. The findings for special tools and training are very different, however. Again, FDI firms show a higher likelihood of providing support.
However, the likelihood of providing these two types of support increases (for both types of firm) when the technology gap becomes larger. This suggests that producer firms use these two types of support to bridge large technological differences with their suppliers. 15
FDI, technology gap and spillovers
We use data from the supplier survey to identify drivers of local spillovers. The dependent variable is a dummy variable capturing whether a supplier has experienced a large positive impact from its business dealings with its client firms. Defining (p) i as the probability that a large positive impact has occurred and (1Àp) i as the probability that supplier i has experienced either a small or no positive impact gives the following regression model: The variables Techsupport and Orgsupport control for the levels of support that a supplier receives. FDIsupplier is a dummy variable capturing whether a supplier works for FDI or Mexican client firms. As Blomstro¨m and Kokko (1998) note, FDI firms benefit from the support that they offer when the suppliers improve their performance. However, it is unlikely that FDI firms will be able to reap all the benefits from the support that they provide, resulting in positive spillovers among their suppliers. By controlling both for the level of support that suppliers receive and their supplier status, we ensure that FDIsupplier picks up pure spillover effects. The inclusion of the variable TechGap controls for the possibility that suppliers experience larger positive effects when the scope for improvement is large. We also include the interaction variable between FDIsupplier and TechGap to test whether FDI spillovers are magnified under large technological differences.
Next, we include controls for the type of contract and the degree of own investment that suppliers make to benefit from relations with their client firms. The dummy variable Figure 1. Type of ownership, technology gap and technology transfers.
Contract takes the value of 1 when client firms only purchase inputs via purchasing orders, not requiring the exchange of information, mutual coordination, etc. The dummy variable Owninvestment takes the value of 1 when a supplier indicates that own investments have been important to benefit from its dealings with its client firms. The vector Supplier contains several firm characteristics that are related to a supplier's capacity to absorb new technologies. We control for firm size under the assumption that a large supplier has more resources to absorb new technologies. We control for the age of the suppliers and expect a positive effect, as experienced suppliers may have developed skills and experience in identifying and absorbing new technologies. We also include a variable capturing the degree of specialisation of a supplier's production processes. We expect that specialised suppliers are better able to identify and absorb new technologies. Finally, we include interactions terms between the technology gap and these supplier characteristics. By including these interaction terms, we test the ideas proposed by Nelson and Phelps (1966) and Levinthal (1989, 1990) that the capacity to absorb technologies facilitates the occurrence of spillovers when the technology gap is large. Table 3 presents the findings on the drivers of spillovers. 16 The first column with findings shows the results from estimating model (2) without the interaction terms. Most of the variables carry significant coefficients with expected signs. Suppliers that receive more support are more likely to experience a large positive impact. Own investment and firm size also increase the likelihood of a positive impact. The effect of the variable Contract is negative, indicating that the relation between suppliers and their client firms needs to involve more than simple purchasing contracts for spillovers to arise.
Empirical findings on drivers of local spillovers
The estimated effect of FDIsupplier is positive. This constitutes strong evidence that FDI firms generate positive spillovers, given the inclusion of the other control variables. The estimated effect of the technology gap is also positive. Again, this goes against the standard interpretation of the technology gap and suggest that positive spillovers are more likely to arise when there is a large scope for suppliers to improve.
The next column shows the findings from estimating the model with the interaction terms. The positive effect of being a supplier to FDI firms persists. Looking at the estimated effects of the technology gap and the interaction terms suggests that large technological differences only foster spillovers among those suppliers that possess a sufficient capacity to absorb new technologies. The effect of the technology gap has turned negative and insignificant. In contrast, the effect of the interaction variables between the technology gap and firm size, experience and specialisation are significant and positive. This indicates that firms that possess these characteristics can benefit from the large scope for improvement that is captured by the technology gap. There is no additional positive effect of the technology gap among suppliers of FDI firms, however.
The second set of findings in Table 3 contains the results from estimating the model with an alternative dependent variable capturing whether suppliers have experienced large improvements that are directly linked to their production processes. The results from the model that omits the interaction terms confirm the positive effects of being a supplier to FDI firms and the technology gap. The main change in findings that results from using the alternative dependent variable is revealed in the findings from the model that includes the interaction variables. The inclusion of these variables does not affect the significant positive effect of the technology gap. This suggest that there is an independent positive effect of the technology gap on spillovers, but only when we consider spillovers that are directly related to a supplier's production processes. The estimated effect of the interaction between FDIsupplier and the technology gap is also positive, indicating that suppliers of FDI firms experience additional spillovers when there is a large scope to improve. These findings are in line with results from econometric studies that identify positive productivity effects from FDI, the technology gap and their interaction. In addition, large and specialised suppliers are also more likely to experience a positive technological impact when the technology gap is large.
To complete the analysis, we examine whether the likelihood that spillovers occur is affected by different sizes of the technology gap. The findings are presented graphically in Figure 2. The left-hand side of the figure shows the results for the model that uses a large general positive impact as dependent variable; the right-hand side shows the findings from the model with a large technological impact as dependent variable. Suppliers of FDI firms are more likely to experience spillovers at all the sizes of the technology gap, irrespective of the type of spillovers. However, the relationship between the technology gap and the likelihood that spillovers materialise differs between the two models. The likelihood that a large general positive impact occurs decreases when the technology gap increases in size.
In contrast, the likelihood of a large technological impact increases when the technology gap becomes larger. This underlines the findings in Table 3 that the technology gap exercises an independent impact on spillovers, conditional on the type of impact. Our findings indicate that the positive effect of the technology gap is particularly pronounced when we consider spillovers that have a direct impact on production processes of the suppliers. 17
Summary and implications for theory and policy making
In this paper, we provide novel evidence on the effects of FDI, the technology gap and their interaction on local technology dissemination and spillovers. We find that foreign ownership is important, as FDI firms are significantly more involved in the creation of technology transfers and their suppliers experience positive spillovers. Regarding the technology gap, we find that large technological differences lower technology transfers. In contrast, a large technology gap fosters positive spillovers, especially among suppliers that are best suited to absorb new technologies. As for interaction effects, FDI firms are more likely to offer several types of support to their suppliers when the technology gap is large. Also, positive spillovers of a technological nature are fostered when there is a large technology gap between FDI firms and their suppliers. Importantly, most of the findings support our alternative interpretation of the technology gap. In contrast to the standard interpretation of the technology gap as a direct inverse indicator of absorptive capacity of domestic firms, the technology gap is better interpreted as reflecting the scope for learning and improvement among suppliers. Our findings contain several implications for theories on FDI, technology dissemination and spillovers. First, FDI spillovers are generally explained by foreign-owned firms possessing better or more modern technologies, which generate positive productivity effects among suppliers that absorb these technologies. Our estimations corroborate the presence of such effects. However, we also find that FDI firms are significantly more engaged in actively transmitting new technologies to their suppliers. More work is needed to explain why this is the case. FDI and domestic producer firms both seem to share the same motive in improving their suppliers, as it results in better inputs. Therefore, next to the possession of higher levels of technology, foreign ownership must be connected to other motives that can explain why FDI and domestic firms differ in technology dissemination. Second, our findings indicate the importance of a better understanding and analysis of inter-firm linkages. Irrespective of type of ownership, we find that inter-firm linkages between producer firms and suppliers are characterised by the transmission of a variety of types of technology. More theory making is required to identify conditions that drive supportive inter-firm linkages. Most research on FDI that looks at local linkages focuses almost exclusively on how being part of a multinational organisation affects the use of local suppliers and technology dissemination. Although clearly important, our findings indicate that type of ownership is only one of several variables that are related to local technology transfers. Therefore, the effects of a range of firm level characteristics on the level and nature of inter-firm linkages need to be examined more closely. In doing so, the creation and effects of technology transfers need to be examined for both technology sending and receiving parties. A second avenue of research is to improve our understanding of how the regional economic environment creates the right setting to allow supportive linkages to materialise. This involves an analysis of a range of regional economic, industrial and institutional characteristics that is likely to influence supportive inter-firm linkages.
Third, with the necessary caveats, our analysis provides clear evidence that a large technology gap fosters technology dissemination and spillovers. This clearly challenges the theoretical interpretation of the technology gap as being directly and inversely linked to absorptive capacity of domestic firms. Not only do our findings indicate that the technology gap should be reinterpreted, they also imply that the absorptive capacity of spilloverreceiving firms should be treated as a separate concept. Our findings suggest that a large technology gap fosters positive spillovers. Although this effect may be conditional or at least fostered by a high level of absorptive capacity of domestic firms, this does not mean that absorptive capacity should be directly incorporated into the concept of the technology gap. Instead, more work is needed to establish conditions under which the technology gap exercises positive effects and to identify and understand firm level characteristics that improve the capacity of firms to absorb and implement new technologies.
Finally, our findings have several implications for regional policy making. Many regional governments actively attract FDI firms in the expectation that these firms foster local technological development. Although our results provide general support for such expectations, it is important to recognise that our findings are obtained in a regional economy with mature industries and a sufficiently large local supplier base. Attracting FDI into regions where industries have not reached a sufficient scale or maturity will not generate similar effects, at least not in the short run. In such cases, long term regional development policies will need to be deployed to facilitate the development of these industries, which at some stage can foster the development of inter-firm linkages and technology dissemination.
Our results on the effect of the technology gap indicate that regional governments need to have a clear understanding of regional industries and their technological development. Overall, our findings suggest that the best strategy to promote technology dissemination is to attract FDI firms that are technologically more advanced than domestic firms. The problem with this policy is of course that technologically more advanced firms are less inclined to locate in regions with a lower level of technological development. On the other hand, technology dissemination by foreign firms that are at the same technological level of domestic firms will be much more limited. This indicates the need for a solid understanding of both the processes of technological development that regional industries are experiencing and the level of technology incorporated in FDI firms. Regional governments will therefore have to actively screen potential new inward FDI for their capacity to disseminate local technologies and contribute to overall regional technological development. Strategic decisions whether to incentivise individual FDI firms to locate in a particular region should be based on such screening.
In relation to this, regional governments will also need to provide a set of policies that aid domestic firms in the absorption and implementation of new technologies. Governments can implement policies to increase the effectiveness of the support that FDI firms offer their suppliers. For instance, governments could work together with regional industry associations to collect and share information on the capacity of FDI firms to provide support and to identify the technological needs of local suppliers. The justification of such policies would not only be that they would improve the functioning of inter-firm relations between suppliers and FDI firms, but also that other firms and industries in the regional economy would benefit from local technology absorption and implementation. Another type of policy would be to design programmes that actively help local suppliers improve their capacity to absorb and implement new technologies. The importance of local technology dissemination may warrant regional governments to take an active role in this. Given our findings, such a policy would need to extend beyond general development programs at the industry and/or regional level and focus specifically on firm level characteristics of local suppliers. Contemporary thinking on drivers of regional development is strongly focused on the dynamics of regional industries and their interconnections. Our findings indicate that additional policies are necessary, targeting individual FDI firms and suppliers to improve the effectiveness of technology dissemination and absorption through inter-firm linkages. 5. The online appendix contains the definitions of all the variables used in this study. 6. We use the following variables to estimate the propensity score: age, industry size, firm size, type of contract, export status, number of client firms, level of imports, perceived technology gap and car and chemical industry dummies. To assess the quality of the matching, we examine whether the covariates are balanced. In the online appendix, we report the standardised differences and variance ratios for the original data and the matched sample. In the matched sample the standardised differences are close to 0 and the variance ratios are close to 1, suggesting that the matching exercise has balanced the covariates. The boxplot of the propensity scores of the raw and matched data also indicate this. 7. A large technological impact refers to improvements of actual production processes. A large organisational impact captures improvements in supplier performance not directly linked to production processes. 8. We use this definition of the technology gap as we learned during our fieldwork that the issue of technological differences between producer firms and suppliers is most important when producer firms consider increasing the use of local suppliers. 9. Findings on drivers of technological support to suppliers of production services are reported in the online appendix. 10. Giroud and Marek (2012) report a similar finding in their study of technology transfers in several transition economies. 11. We also examined whether the percentage of foreign ownership affects technology transfers. This is not the case, indicating that the effect of foreign ownership is captured by the dichotomous distinction between FDI and domestic producer firms. 12. Following the suggestion of one referee, we investigated whether the car industry differs from the other industries. We estimated the model separately for producer firms in the car industry and firms in the other industries. The estimated coefficients of the variables FDI, Technology Gap and their interaction do not differ between the two estimations. We also estimated the model for the full sample of firms, adding interaction terms between a car industry dummy and the variables FDI and Technology Gap. Both these interaction variables carry insignificant coefficients, suggesting that our findings do not obscure structural differences between the car industry and the other industries. 13. The estimated effect of the FDI variable for the different types of support could be biased if FDI firms as a group provided more positive answers to the questions on support. To assess this issue, we performed a confirmatory factor analysis with the four types of support. The findings indicate that FDI and Mexican producer firms load similarly onto a single latent factor of supportiveness with coefficients that are not significantly different, suggesting that such a bias is not affecting the results. 14. Although the findings are somewhat weaker, the results in the online appendix indicate that the effects reported in Table 2 do not apply exclusively to support offered to material input suppliers. 15. The figures for suppliers of production services are shown in the online appendix. 16. We report the findings for the full sample of suppliers. The online appendix contains the findings for the subsamples of suppliers of material inputs and production services. Overall, findings for the subsamples are in line with the findings reported in Table 3. 17. The figures for the subsamples of suppliers of material inputs and production services are shown in the online appendix. | 2019-05-20T13:02:48.977Z | 2017-09-24T00:00:00.000 | {
"year": 2017,
"sha1": "d98f6c9d6e4666fe71d1dc397cbe9d755ab9b03b",
"oa_license": "CCBYNC",
"oa_url": "https://journals.sagepub.com/doi/pdf/10.1177/0308518X17731942",
"oa_status": "HYBRID",
"pdf_src": "Sage",
"pdf_hash": "b7f16453eeceddb5ee4e44a73fd151730d4363f0",
"s2fieldsofstudy": [
"Economics"
],
"extfieldsofstudy": [
"Business"
]
} |
221624418 | pes2o/s2orc | v3-fos-license | Rapid and dynamic processing of face pareidolia in the human brain
The human brain is specialized for face processing, yet we sometimes perceive illusory faces in objects. It is unknown whether these natural errors of face detection originate from a rapid process based on visual features or from a slower, cognitive re-interpretation. Here we use a multifaceted approach to understand both the spatial distribution and temporal dynamics of illusory face representation in the brain by combining functional magnetic resonance imaging and magnetoencephalography neuroimaging data with model-based analysis. We find that the representation of illusory faces is confined to occipital-temporal face-selective visual cortex. The temporal dynamics reveal a striking evolution in how illusory faces are represented relative to human faces and matched objects. Illusory faces are initially represented more similarly to real faces than matched objects are, but within ~250 ms, the representation transforms, and they become equivalent to ordinary objects. This is consistent with the initial recruitment of a broadly-tuned face detection mechanism which privileges sensitivity over selectivity.
H umans are incredibly skilled at both detecting and recognizing faces, with a significant region of cortex dedicated to face processing 1 . Despite this expertise, sometimes we spontaneously perceive faces where there are none -for example in inanimate objects, such as in a tree or a piece of fruit. This phenomenon, known as face pareidolia, can be conceptualized as a natural error of the face-detection system and has recently been demonstrated behaviorally in macaque monkeys 2,3 , suggesting that the perception of illusory faces arises from a fundamental feature of the primate face-detection system, rather than being a uniquely human trait. Despite substantial progress in uncovering the primate face processing system 1,4-7 it is still not understood what constitutes a face for visual cortex, and what neural mechanism elicits errors of face detection in ordinary objects.
Here, we combine noninvasive neuroimaging tools with high temporal (MEG) and spatial (fMRI) resolution as well as behavioral ratings and model-based analyses in order to understand how illusory faces are processed in the human brain. Critical to our approach here is the use of a yoked stimulus design. For each illusory face we found a matched object image, which was semantically and visually similar, but which did not contain an illusory face (Fig. 1). The matched set of objects facilitates examination of how the presence of an illusory face modulates the brain's representation of an object. In terms of the spatial distribution of responses, previous findings suggest a considerable degree of abstraction in the visual selectivity of face-responsive brain regions 5,6,[8][9][10][11] . The sensitivity of face-selective regions to abstract faces 5,8,9 suggests these regions are likely sensitive to illusory faces in inanimate objects, but it is an open question whether this sensitivity is specific to face-selective regions, or whether it is widespread throughout category-selective cortex, including regions selective to objects 12,13 . This makes illusory faces in objects particularly interesting in terms of their category membership as they are perceived as both an object and as a face. Importantly, natural examples of illusory faces are visually diverse and do not require any assumptions to be made about the key features that drive the brain's response to face stimuli. This means that illusory faces are potentially revealing about the behaviorally relevant tuning of the face-detection system.
While understanding the spatial organization of responses to illusory faces will clarify the role of higher-level visual cortex in face perception, identifying the temporal dynamics of how illusory faces are processed is critical to understanding the origin of these face-detection errors. Human faces are rapidly detected by the human brain 12,13 , but it is not known to what extent illusory face perception relies upon the same neural mechanisms. One possibility is that certain arrangements of visual features (such as a round shape) rapidly activate a basic face-detection mechanism, leading to the erroneous perception of a face. Alternatively, illusory face perception may arise from a slower cognitive reinterpretation of visual attributes as facial features, for example as eyes or a mouth. If illusory faces are rapidly processed, it would suggest the face-detection mechanism is broadly tuned and weighted toward high sensitivity at the cost of increased false alarms. Here we exploit the high temporal resolution of MEG in order to distinguish between these alternative accounts in the human brain.
We find that face-selective regions are sensitive to the presence of an illusory face in an inanimate object, but other occipital-temporal category-selective visual regions are not. In addition to this spatially restricted response, we discover a transient and rapidly evolving response to illusory faces. In the first couple of 100 ms, illusory faces are represented more similarly to human faces than their yoked nonface object counterparts are. However, within only 250 ms after stimulus onset, this representation shifts such that illusory faces are indistinguishable from ordinary objects. In order to enhance our understanding of what is driving this early face-like response to illusory faces, we implement a model-based analysis to compare the brain's response to behavioral ratings of "faceness" and the output of computational models of visual features. We find that the brain's representation correlates earlier with the visual feature models than the behavioral model, although the behavioral model explained more variance in the brain's response overall than the computational models. Together, our results demonstrate that that an initial erroneous face-like response to illusory faces is rapidly resolved, with the representational structure quickly stabilizing into one organized by object content rather than by face perception.
Results
Recording responses to illusory faces in the human brain. We recorded neuroimaging data (fMRI and MEG) in two separate experiments while human participants viewed 96 photographs including 32 examples of illusory faces in inanimate objects, 32 matched nonface objects, and 32 human faces ( Fig. 1a; Supplementary Fig. 1). The nonface objects were yoked to the illusory face examples, so that each illusory face was paired with a matched nonface object that was of the same object identity and as visually similar as possible (see Online methods). The stimulus set was validated by asking workers on Amazon Mechanical Turk (n = 20) to "Rate how easily can you can see a face in this image" on a scale of 0-10. Human faces (M = 9.96, SD = 0.10), were rated as more face-like than both illusory faces (t(31) = 26.15, p < 0.001, two-tailed, Bonferroni correction for k = 3 comparisons) and matched objects (t(31) = 143.57, p < 0.001). Importantly, illusory faces (M = 6.27, SD = 0.79) were rated as significantly more face-like than matched nonface objects (M = 0.70, SD = 0.36), demonstrating that faces are spontaneously perceived in the images we selected (t(31) = 39.68, p < 0.001; Fig. 1b).
We collected fMRI and MEG data for the same stimulus set to benefit maximally from the higher spatial resolution of fMRI and the finer temporal resolution of MEG (Fig. 1c). Due to the relative temporal sluggishness of the fMRI BOLD response, in the fMRI experiment (N = 21), images were shown for 300 ms with a 3.7 s interstimulus interval and additional randomly inserted blank trials of 4 s duration. In the MEG experiment (N = 22), images were shown for 200 ms with a 1-1.5 s variable interstimulus interval. In order to maintain alertness, in both experiments participants judged whether the image on each trial was rotated slightly to the left or right by 3°.
We used a multivariate pattern analysis approach [14][15][16][17] to analyze the neuroimaging data and extracted the brain activation patterns associated with viewing each of the 96 visual images (Fig. 1d). For the fMRI data, we extracted the spatial patterns across voxels in response to each of the visual stimuli, providing finer spatial resolution to complement the high temporal resolution of the MEG data. For the MEG experiment, we isolated the neuromagnetic signal from −100 to 1000 ms relative to stimulus onset on each trial and extracted the spatiotemporal patterns across the whole brain across all 160 sensors as a function of time. These patterns across voxels (fMRI) or sensors (MEG) were then used for the specific analyses detailed below.
Illusory faces modulate responses in face-selective cortex. Our first aim was to characterize which of the category-selective regions in higher-level visual cortex are sensitive to the presence of an illusory face in an inanimate object. In each participant, we identified the face-selective fusiform face area (FFA) and occipital face area (OFA), an object-selective lateral occipital (LO) area, and the scene-selective parahippocampal place area (PPA) based on independent functional localizer runs (Fig. 2a). We used a cross-decoding analysis to test which brain regions had activity that could distinguish between the three categories of experimental stimuli (Fig. 1d). A linear support vector machine was trained to classify which category of stimulus a subject was viewing based on the patterns of BOLD activation across voxels in a given region of interest. The classifier's performance was tested on data from stimulus exemplars of each category not used in the training set, thus correct classification required generalization to new stimuli, ruling out an explanation based on visual features associated with specific images in the training set. Statistical significance was assessed using one-sample t-tests (one-tailed) against chance decoding performance (50%). The FDR adjustment was made to all reported p values to control for multiple comparisons. In all four regions, human faces could be distinguished from objects with an illusory face (FFA: t (15) Supplementary Fig. 2). These results were further supported by a whole-brain decoding searchlight (Fig. 3), which showed that the main areas of successful decoding of human faces from objects (Fig. 3a) and illusory faces from objects ( Fig. 3b) align with functionally defined FFA and OFA, with some additional area located on the ventral surface between these ROIs. To further characterize the brain's response to illusory faces, we applied representational similarity analysis (RSA) 18,19 . We constructed dissimilarity matrices by correlating the patterns of fMRI BOLD activation between each pair of stimuli and taking 1correlation to convert to a measure of pattern dissimilarity (Fig. 2c). For visualization, we applied multidimensional scaling (MDS) to the resulting matrices for each region of interest and plotted the first two dimensions (Fig. 2d). Note that in FFA and OFA, human faces are represented similarly to each other, and illusory faces are more similar to human faces than matched objects are (see blue regions in Fig. 2c). In comparison, in LO and PPA, all objects are grouped together regardless of whether they contain an illusory face or not. This grouping is consistent with the lack of illusory face cross-decoding in these regions. Overall, this analysis demonstrates that illusory faces are represented uniquely compared to both human faces and nonface objects in face-selective cortex. In contrast, illusory faces are not distinct from ordinary objects in either object or scene-selective cortex.
Illusory faces are rapidly decoded from whole-brain activity. One of our primary goals was to determine whether illusory faces are processed rapidly based on activation of a broadly tuned facedetection mechanism by their low-level visual features, or alternatively, whether they result from a slower cognitive process requiring the reinterpretation of the image. Individual stimuli were readily decodable from the MEG whole-brain activation patterns after stimulus onset with similar decoding within each category ( Supplementary Fig. 3). To address the processing of illusory faces, we focused on the category level and evaluated how quickly illusory faces could be distinguished from both human faces and matched objects from time-varying whole-brain activity measured with MEG. We used a cross-decoding approach (training and testing on different exemplars), so that successful classification required the classifier to generalize to new stimuli. We generated predictions for the cross-decoding results based on the two possible accounts (Fig. 4). If illusory faces activate a rapid neural pathway based on low-level visual features, we would expect peak decoding at a similar time to that for human faces (green line, Fig. 4). Alternatively, if illusory face perception requires a slower cognitive process based on reinterpreting the image, peak decoding should occur later relative to stimulus onset (orange line, Fig. 4). In both cases, we would expect reduced performance for cross-decoding illusory faces from matched objects compared to decoding real faces vs. objects because the illusory faces share many more visual and semantic features with the objects; and consequently their brain activation patterns will be more similar, making the classification problem more challenging.
To reveal at what time the presence of an illusory face in an object could be decoded from the brain activation patterns, we trained a separate binary LDA classifier to discriminate between each of the three pairs of our three stimulus categories (i.e., human faces vs. objects, illusory faces vs. objects, and illusory faces vs. human faces). Importantly, here we used a crossdecoding approach, training the classifier on brain activation patterns in response to a subset of the 96 stimuli, and testing the classifier's performance on generalizing to brain activation patterns elicited by a new subset of stimuli not used in training (Fig. 1d). This means that successful classifier performance in discriminating each of the three stimulus categories (human faces, illusory faces, nonface objects) from the time-varying activity across MEG sensors could not be explained by sensitivity of the classifier to brain signatures associated with the visual properties of particular images, since successful performance requires generalizing to new examples of each category.
All three categories could be decoded from each other within the 200 ms stimulus presentation duration (Fig. 5a). As outlined in Fig. 4, we were interested in the relative time course of decoding for illusory faces vs. objects in comparison to that for human faces vs. objects. Human faces could be discriminated from nonface objects soon after stimulus onset (red line), with two peaks in classifier performance at~160 and~260 ms. Importantly, decoding of illusory faces from objects also peaked at~160 ms (blue line). This is consistent with rapid processing of Fig. 1 for all 96 visual stimuli. Full resolution versions of the stimuli used in the experiment are available at the Open Science Framework website for this project: https://osf.io/9g4rz. b Behavioral ratings for the 96 stimuli were collected by asking N = 20 observers on Amazon Mechanical Turk to "Rate how easily can you can see a face in this image" on a scale of 0-10. Illusory faces are rated as more facelike than matched nonface objects. Error bars are ±1 SEM. Source data are provided as a Source data file. c Event-related paradigm used for the fMRI (n = 16) and MEG (n = 22) neuroimaging experiments. In both experiments the 96 stimuli were presented in random order while brain activity was recorded. Due to the long temporal lag of the fMRI BOLD signal, the fMRI version of the experiment used a longer presentation time and longer interstimulusintervals than the MEG version. To maintain alertness the participants' task was to judge whether each image was tilted slightly to the left or right (3°) using a keypress (fMRI, mean = 92.5%, SD = 8.6%; MEG, mean = 93.2%, SD = 4.8%). d Method for leave-one-exemplar-out cross-decoding. A classifier was trained to discriminate between a given category pair (e.g., illusory faces and matched objects) by training on the brain activation patterns associated with all of the exemplars of each category except one, which was left out as the test data from a separate run for the classifier to predict the category label. This process was repeated across each cross-validation fold such that each exemplar had a turn as the left-out data. Accuracy was averaged across all cross-validation folds. Source data are provided as a Source data file. c Representational dissimilarly matrices (96 × 96) for all stimuli for the four regions of interest. The dissimilarity is calculated by taking 1-correlation (Spearman) between the BOLD activation patterns for each pair of stimuli. The colorbar range is scaled to the max and min of the dissimilarity values for each ROI for visualization. White lines indicate stimulus category boundaries. Insets show 3 × 3 matrices for each ROI averaged by category, excluding the diagonal. Source data are provided as a Source data file. d Visualization of the dissimilarity matrices in (c) using multidimensional scaling. The first two dimensions following MDS are plotted, each of the points representing the 96 stimuli is colored according to its category membership. Proximity of the points represents more similar brain activation patterns for the stimuli. Note that in the FFA and OFA, the illusory faces are more separated from the matched objects and closer to the human faces compared to LO and PPA.
illusory faces based on visual features, instead of a slower cognitive reinterpretation of the image (compare to predictions in Fig. 4). A result we did not anticipate was the more transient nature of the result for decoding illusory faces from objects compared to real faces. Notably the second peak at~260 ms observed for human faces is absent for illusory faces, and although decoding remains significant after onset for most of the 1100 ms analysis window, its magnitude is vastly reduced relative to that for real faces. We investigate this further using RSA to probe the brain's representation of illusory faces in more detail.
An early face response followed by rapid reorganization. The decoding results are indicative of a dynamic change in how illusory faces are represented relative to real faces and similar nonface objects over a relatively brief time period of~200 ms. To investigate this further, we applied RSA 18,19 . For each pair of 96 stimuli, we correlated their time-varying MEG activation patterns across all sensors at each time point. This produced a representational dissimilarity matrix (RDM, Fig. 5b, Supplementary Movie 1), which depicts the relative similarity between the brain's representation of each pair of stimuli (1-Spearman's R). Here we focus on three timepoints of interest based on the cross-decoding results in Fig. 4a: t 1 (130 ms) corresponds to the first time point after all three category comparisons were significant, t 2 (160 ms) is when the first decoding peak occurs for all three category pairs, and t 3 (260 ms) is the time of the second decoding peak for decoding human faces from all objects with or without a face, which is notably absent for decoding illusory faces from nonface objects. We additionally used MDS to produce a visualization of the dissimilarity matrices, plotting the first two dimensions (Fig. 5c, Supplementary Movie 2). Each of the points representing the 96 stimuli is colored according to its category membership. Closer proximity of the points represents more similar brain activation patterns for the associated stimuli.
The most striking feature of the representational structure is how it changes dynamically across these three timepoints, as visualized in the dissimilarity matrices (Fig. 5b) and MDS plots (Fig. 5c). At 130 ms the human faces are already more similar to each other than to illusory faces or objects (Fig. 5b), as shown by the clustering of the human face exemplars in the MDS plot (Fig. 5c). Notably, a few illusory face exemplars are clustered with the human faces (Fig. 5c). By 160 ms, the illusory faces have started to segregate from the matched objects (Fig. 5c), and have become more similar to the human faces than the matched objects are (Fig. 5b). There is a categorical shift in the representational organization between 160 and 260 ms: initially the illusory faces are distinct from the matched objects, but only 100 ms later, the illusory faces are grouped with the matched objects (Fig. 5c). As with the rapid onset of category-level crossdecoding (Fig. 5a), this shows that illusory faces are rapidly processed. These errors of face detection are initially treated more like real faces than matched objects are, however, the human brain rapidly resolves this detection error and they are represented more similarly to objects in less than a quarter of a second.
The contribution of visual features to rapid face detection. The decoding and RSA results show an early response to illusory faces in the patterns of brain activity measured with MEG. This is consistent with engagement of a rapid process for face detection (see Fig. 4) rather than a slower cognitive reinterpretation of the image. Presumably a fast face-detection process would be based on simple or coarse visual features associated with real faces. If this is the case, we might expect that the early stages of the MEG response to illusory faces would be driven by simple or coarse visual features in the images. Here we aim to examine the extent to which two computational models of visual features can explain the brain's response to illusory faces by comparing their representation of the stimuli to that measured by behavioral ratings of "face-ness" and the brain activation patterns measured by MEG. We selected two well-established computational models that emphasize different visual features in an image: Graph-Based Visual Saliency 20 (GBVS) and the GIST 21 visual feature model. Prior research has suggested that features important for face recognition typically consist of high contrast regions [22][23][24] . Computational models of visual saliency such as GBVS aim to predict where a human observer would look in an image. Given that regions of high saliency frequently correspond to areas of high contrast, we reasoned that this class of model may approximate the type of visual characteristics relevant to the face-detection mechanism. In contrast, the GIST 21 visual feature model is based upon characterizing the spatial distribution of content across an image, and may be able to capture different low-level visual features associated with illusory faces.
Before testing the correlation with the MEG data, we examined the representational dissimilarity matrices (RDMs). First, we constructed a dissimilarity matrix based on the behavioral ratings of "face-ness" for the stimuli (Fig. 6a). Notably, the behavioral matrix reveals that illusory faces are more similar to human faces than the matched objects are. Further, different illusory face exemplars are also more similar to each other than they are to the matched objects. These relationships are expected from the behavioral results showing that illusory faces were rated as more face-like than matched objects, but less so than human faces (Fig. 1b). Next, we built separate dissimilarity matrices for the two models by correlating the saliency maps generated from the GBVS model (Fig. 6b) and the gist descriptors produced by the GIST model (Fig. 6c) for each pair of images. To test whether the representation of illusory faces generated from the GBVS and GIST models is distinct from that of human faces and matched objects we averaged the dissimilarity across all 32 exemplars of each category, producing 3 × 3 category dissimilarity matrices (Fig. 6d). For the GIST model, the category-averaged RDM revealed that illusory faces were on average more similar to human faces than the matched objects (mean difference = −0.0696, p = 0.012, permutation test with FDR corrected p values). This was not the case for the GBVS saliency model (mean difference = 0.0063, p = 0.701). However, illusory faces were not more similar to each other on average than they were to matched objects for either the GBVS (mean difference = −0.0000004, p = 0.649) or GIST (mean difference = −0.0450, p = 0.092) models. Overall, these results suggest the GIST visual feature model is sensitive to some of the distinguishing features of illusory faces, however, neither the GIST or GBVS models capture the clear perceptual differences revealed by the behavioral ratings of how easily a face can be perceived in each image (Fig. 6c).
To examine to what degree these model representations explain the time-varying representation of the stimuli measured with MEG, we correlated each of the three model RDMs (Fig. 6a-c: behavior, saliency, and visual features) with the time-varying MEG RDM (Supplementary Movie 1). Importantly, we removed the human faces from both model and MEG RDMs for this analysis. A preliminary analysis showed that human faces produce a strong neural response, which dominated the results, and consequently only confirmed that human faces were represented differently than the other stimuli. As our focus here is on how illusory faces modulate the representation of an object, including human faces would inflate any observed correlations between the brain representations and the models. The saliency RDM significantly correlated with the MEG RDM from 85-125 ms after stimulus onset (Fig. 6f), with one discreet peak at 110 ms. In comparison, a more sustained correlation was observed with the visual feature RDM, starting at 95 ms after stimulus onset and continuing throughout the stimulus presentation. Consistent with the observed differences in the category-averaged RDMs (Fig. 6d), this difference in the time course of the correlation with the MEG RDMs suggests that the saliency and visual feature models pick up on different aspects of the stimuli. The behavioral RDM based on the face ratings had a sustained correlation with the MEG RDM from 120 ms, with a peak at 170 ms (Fig. 6b). The correlation with behavior also reached much closer to the noise ceiling, an estimate of the maximum correlation expected with the MEG data 25 . The peak correlation with behavior occurred around the time of the first decoding peak at~160 ms (Fig. 5a), which is notably later than the correlation with the saliency or visual feature models. The finding that the correlation with behavior was stronger overall than for either computational model suggests that these models capture an initial stage of processing, but not the full time course.
An early face-like response in FFA. In order to link the MEG and fMRI data, we used a fusion approach 27 . We correlated the fMRI RDMs for each of the four regions of interest with the timevarying MEG RDM at every time point (Fig. 7a). As for the RSA model comparison with saliency and behavior, we removed human faces from this analysis. This makes the analysis more conservative, and any significant correlations are reflective of the representation of objects with and without an illusory face. The representation in the FFA significantly correlated with the MEG data for a brief period post stimulus onset (160, 175 ms). The correlation was not significant at any time for OFA, LO, or PPA. This suggests that the face response to illusory faces demonstrated by the peak correlation with behavior at 170 ms (Fig. 6b) and in cross-decoding at 160 ms (Fig. 5a) in the MEG data has a likely origin in FFA.
Discussion
Our results reveal the representation of natural errors of face detection in the human brain. With fMRI and multivariate pattern classification methods, we found that a face-like response to illusory faces was restricted to face-selective regions in occipital-temporal cortex. In scene and object areas, illusory faces were instead represented more similarly to objects. Exploiting the much higher temporal resolution of MEG, we showed that illusory faces are initially represented as more face-like in their brain activation patterns than similar matched objects. However, onlỹ 100 ms later, illusory faces are represented as more object-like, consistent with a rapid resolution of the detection error. Thus, although illusory faces in objects are perceived as having a persistent dual identity (face, object), their neural representation quickly shifts in its relative weighting of these two identities over time. This is consistent with the rapid activation of a broadly tuned face-detection system, which is tolerant of substantial visual variance in the definition of facial features. Following the initial face-detection process, which possibly occurs via a subcortical route 28 , the subsequent resolution of the error is likely driven by processing in cortical areas involved in face recognition 1,4,5 .
It is currently debated to what degree representations in ventral temporal cortex are driven by the appearance or visual properties of an object vs. its semantic meaning [29][30][31][32][33][34] . Recently it was demonstrated that in ventral occipitotemporal cortex, objects that look like another object (for example, a cow-shaped mug) have BOLD activation patterns that are more similar to what object they look like (e.g., a cow) than to their actual object identity (e.g., a mug) 30 . Similarly, it has been debated whether object representations in ventral temporal cortex are organized by an overall semantic principle such as animacy 14 or real-world size 35,36 , or by visual properties that co-vary with these categories 29,31,32,34 . Our data offer interesting insights into this issue. Despite sharing more visual properties and their semantic identity with the yoked objects, the illusory faces were represented as more similar to human faces than the yoked objects were in face-selective FFA and OFA, and for a brief transient period in the first quarter of a second as revealed by whole-brain MEG. This suggests that the response to stimuli in which meaning and visual appearance are dissociated is not homogenous throughout ventral visual cortex, and further, that these representations exhibit temporal variability rather than stability over time in the brain's response.
Our results are able to adjudicate between two alternative mechanisms for illusory face perception, providing evidence for a rapid detection mechanism rather than a slower cognitive reinterpretation of the object as a face. The early timing of the correlation between the MEG data and the saliency and visual feature models relative to that with behavior suggests that illusory face perception is at least partially driven by particular low or midlevel features. However, neither computational model fully explains the behavioral or brain data. The differences in timing reflected by the transient correlation with the GBVS saliency model and the more sustained correlation with the GIST visual feature model suggest that the two models isolate different visual characteristics that might drive the erroneous face-detection response to illusory faces at different stages of processing. Previous fMRI investigation of face detection in visual noise has suggested the eye and mouth regions are important features in modulating the response of face-selective visual areas, consistent with the idea of a simple template for face detection 37 . A rapid subcortical route for face detection has been proposed based on multiple lines of evidence 28 , involving the superior colliculus, pulvinar, and amygdala. The amygdala has been implicated in orienting toward faces in primates, and amygdala lesions impair this tendency for both real and illusory faces 3 . Along with the timescale of the early confusion we observe between real and illusory face processing, this is suggestive of a possible role for the amygdala in falsely detecting faces in everyday objects.
Beyond revealing a rapid neural mechanism underlying illusory face perception, our data offer insight into the distinction between face detection and face recognition in the visual system. Visual detection and recognition have competing requirements, which must be adequately weighted in a biological system in order to optimally direct the behavior of the organism. To improve detection a system needs to maximize sensitivity, yet recognizing an object as belonging to one category or another requires precise tuning to visual features. Neurons in faceselective cortical areas are tightly tuned to particular features associated with real faces 6,11,38 , and this tight tuning might be expected to result in poor detection. In contrast, the existence of face pareidolia in a wide range of objects is suggestive of a broadly tuned face-detection system. Our results suggest that this broadly tuned face detector results in compelling errors, yet these errors are quickly resolved, likely via a rapid subcortical route for face detection. The role of individual face-selective regions in the human brain is still being revealed 1 . If rapid face detection is served by a subcortical route, processing in cortical areas may be focused on different aspects of the complex task of face recognition. This functional anatomical distinction may explain how the human visual system balances the competing requirements of face detection vs. recognition.
Although we observe a rapid face-like response to illusory faces in objects, there are substantial differences in the representation of human faces vs. illusory faces. The fMRI data showed that in face-selective FFA and OFA, human faces have more similar within-category activation patterns than illusory faces, which are less similar to each other. Further, in nonface areas such as LO and PPA, the illusory faces group with objects rather than human faces. This is similar in the time domain revealed by MEG, in which illusory faces have a more transient face-like response than human faces, and are much more similar to nonface objects within a couple 100 ms. The fact that human faces and illusory faces are readily decoded from each other in both fMRI and MEG is also indicative of differences in their representation. Overall, this difference is consistent with the interpretation of illusory faces as a quickly resolved error of a broadly tuned face-detection system 2,3 . Thus although illusory faces elicit a face-like response, they are not represented in the same way as human faces within the visual system.
Our unexpected finding that the face-like response to illusory faces is not only rapid, but is also brief, underscores the importance of investigating temporal dynamics in understanding the neural mechanisms of visual perception. Our results are consistent with prior MEG results suggesting differential stages of information processing of faces that unfold over a few 100 ms 39 . Recently, an MEG study using famous faces reported that certain features such as gender and age were evident in the whole-brain representation of the face before identity 40 . Similarly, neurons in macaque temporal cortex are known to respond to global information about faces in the earliest part of the response, with finerscale information about identity or expression emerging later 41 . Broadly consistent with our finding of a rapid face-like response to illusory faces, an EEG study reported an increased neuronal response across frontal and LO cortex within 500 ms to white noise patterns in which observers mistakenly thought they saw a face 42 . Our result reveals a transformation of the representational space that occurs in a fraction of a second. Together these results are indicative of the type of insights into neural mechanisms that are inaccessible at the slower temporal resolution of the fMRI BOLD response 43 . A curious feature of illusory faces is that the percept of a face persists well beyond the initial few 100 ms in which we now know they are represented in the brain as more face-like. This suggests that the MEG data are reflective of the initial processing that leads to the misperception of a face in these objects.
In sum, illusory faces are represented uniquely in the brain compared to both real human faces and similar objects that do not have illusory facial features. The presence of an illusory face initially results in a rapid face-like response to an inanimate object, possibly via a subcortical route. However, this detection error is rapidly resolved and the brain's representation transforms such that illusory faces are represented more similarly to matched objects than human faces within a couple of 100 ms. This result underscores the importance of considering temporal dynamics in understanding human cognition.
Methods
Participants. All imaging experiments were approved by the Human Research Ethics Committee of Macquarie University and participants received financial compensation for their time. The online experiments were conducted on Amazon Mechanical Turk following guidelines set by the NIH Office of Human Subjects Research Protections, and participants were also compensated for their time. All experiments were conducted in accordance with the Declaration of Helsinki and informed consent was obtained from each participant. In total, 22 participants (8 male, 14 female, mean age 26.2 years, range 18-41 years) completed the MEG experiment. In total, 21 participants (11 male, 10 female, mean age 25.4 years, range 20-36 years) participated in the fMRI experiment. N = 4 participants were removed from the fMRI analysis due to inability to define category-selective regions from their localizer data and N = 1 was excluded due to failing to complete the experiment, leaving N = 16 fMRI datasets for analysis. In total, 20 participants (12 female, 7 male, 1 other) completed the behavioral experiment online via Amazon Mechanical Turk. Note that the human face images used in the experiments are not shown in the paper because we do not have the rights to publish them. The original face stimuli used in the experiments are available at the Open Science Framework website for this project: https://osf.io/9g4rz. The human faces shown in this figure are similar photographs taken of lab members who gave permission to publish their identifiable images.
Visual stimuli. The stimulus set consisted of 96 photographs sourced from the internet (32 illusory faces, 32 matched nonface objects, 32 human faces). For each of the 32 examples of illusory faces in inanimate objects (e.g., bell peppers, backpack, coffee cup), we selected a matched object of the same type and as visually similar as possible, but which did not contain an illusory face (Fig. 1a). The illusory face images have a high degree of variance in the visual appearance of their illusory facial features, compared to human facial features. In addition, the illusory face images have greater variance in visual properties such as color, orientation, and face size, compared to the controlled face images typically used in experiments on face processing. For these reasons, the set of 32 human faces was selected to have a high degree of variance across age, gender, race, facial expression, and head orientation in order to match the high variance of the illusory face image set. Human faces were not individually matched to the illusory faces because of the ambiguity in defining parameters such as age, race and gender for illusory faces. However, we did match the number of images containing more than one face across human and illusory face sets. Two of the illusory face images contained two faces, so we included two images in the human face set that also had two faces. All images were cropped to a square image and resized to 400 × 400 pixels, but no other manipulations were made. A whole-brain volume containing 42 slices was collected. In total, 128 volumes were collected per localizer run (5 min each), and 218 per experimental run (9 min each). One localizer run was collected immediately after the structural scan, followed by the seven experimental runs, and then the second localizer run. In total the scanning session took <90 min. Stimulus presentation scripts were written in MATLAB using functions from the Psychtoolbox [44][45][46] fMRI category-selective localizer: two independent localizer runs were used to define the category-selective regions FFA, OFA, LO, and PPA individually in each participant. The functional localizer stimuli were color pictures of faces, places, objects, and scrambled objects (480 × 480 pixels). In total, 54 images for each category were selected from The Center for Vital Longevity Face Database 47 , the SUN397 database 48 , and the BOSS database 49,50 respectively. Scrambled objects for localizing object-selective region LO 51 were pregenerated in MATLAB by randomly scrambling each object image in an 8 × 8 grid and saving the resulting image. Localizer runs began with a 4 s fixation period before the first stimulus block. For each stimulus class, there were 3 unique blocks of 18 images. Participants performed a 1-back task, pressing a key each time an image was repeated twice in a row. Every time a block was run, the images were presented in a random order and two random images were repeated twice for the 1-back task. Each of the 20 images within a block (18 unique + 2 repeats) was shown for 600 ms followed by a 200 ms interstimulus interval. The 16 s stimulus blocks alternated with 10 s fixation blocks. Each of the four stimulus categories was repeated three times per 5-min localizer run, once per unique image set. The order of block presentation was in a pseudorandom order, with two different orders counterbalanced across runs for each participant.
fMRI experimental runs: the fMRI experiment used a rapid event-related design 14 , in order to measure the BOLD activation patterns for each individual stimulus. Each of the 96 stimuli was shown once per run, in random order. In each 4 s trial, stimuli were presented for 300 ms at the start of the trial on a gray background. For the remaining time in each trial after stimulus offset, a gray background with a fixation cross was shown. In addition to the 96 stimulus trials, 32 blank null trials (4 s duration each) were inserted randomly in the sequence for each run and each participant. Finally, four null trials were inserted at the start and end of each run. This produced 136 trials total per run for a run duration of 9 min. Each participant completed seven experimental runs.
A task unrelated to face detection was used in order to maintain participants' alertness during the fMRI experiment. Each image was presented tilted slightly by 3°to the left or right of center (Fig. 1c) and participants' task was to report the direction of tilt with a keypress using a Lumina MRI-compatible button response pad. At the end of each run, participants received on-screen feedback on their performance on the rotation task (accuracy, reaction time, number of missed trials). For each run and participant, the tilt of each stimulus was pseudo-randomly allocated such that tilt direction was counterbalanced equally within each of the three stimulus categories (faces, illusory faces, and matched objects). Averaged across participants and runs, the mean task accuracy was 92.5% (SD = 8.6%). Eye movements were not recorded, however, the brief stimulus duration prevented any substantial eye movements during stimulus presentation.
fMRI preprocessing: fMRI data were preprocessed using the AFNI 52 software package. EPIs were slice-time corrected, motion-corrected, and co-registered to the participant's individual anatomical volume. Spatial smoothing of 4 mm full-width at half-maximum was applied to the localizer runs only, no smoothing was conducted on the experimental runs. All analyses were conducted in the native brain space of each participant.
Functional ROI definition. Four functional ROIs were defined in each participant's brain from the independent localizer data: FFA, OFA, LO, and PPA. Data from the two independent localizer runs were entered as factors into a GLM in AFNI 52 to estimate the beta weights for faces, scenes, objects, and scrambled objects. Contrasts of faces-scenes, scenes-faces, and objects-scrambled objects produced tmaps used to define the boundaries of each ROI. Cortical reconstruction was performed using Freesurfer 6.0 from the structural scan for each participant. Inflated surfaces were visualized in SUMA 53 for functional ROI definition, with the results of the GLM overlaid as t-maps. FFA and OFA were defined as the contiguous cluster of voxels in the fusiform gyrus and LO area, respectively, produced from the contrast between faces vs. scenes. LO was defined as activation on the LO surface from the contrast between objects and scrambled objects. Finally, PPA was defined as the peak cluster of activation in the parahippocampal gyrus produced by the contrast between scenes vs. faces. To ensure unique ROIs, any overlapping voxels in the four regions were preferentially allocated to the faceselective regions, and any voxels allocated to both FFA and OFA were removed from both ROIs. The mean size of each ROI averaged across participants was 271 voxels for FFA (range: 86-577), 190 voxels for OFA (range: 101-316), 337 voxels for LO (range: 70-543), and 393 voxels for PPA (range: 277-712).
fMRI multivariate pattern analysis. Decoding analysis was performed using The Decoding Toolbox (TDT) 54 and MATLAB. Decoding was performed with a linear SVM in each ROI on the beta weights estimated in a GLM using AFNI 52 for each of the 96 stimuli, producing a separate beta weight for each run (i.e., 7 beta weights per stimulus). For cross-validation we used a combined leave-one-run-out and leave-one-exemplar-out procedure as implemented in TDT 54 . Cross-decoding for each category pair (faces vs. objects, faces vs. illusory faces, illusory faces vs. objects) was performed by training the classifier on the data for all exemplars except one pair (N−1 = 31), which served as the test set. Leave-exemplar-out cross-validation was performed by repeating this process iteratively such that each exemplar was in the test set once. Classifier accuracy was averaged across all cross-validation folds for each ROI and category pair (Fig. 2b). Statistical significance was assessed using one-tailed t-tests with control for multiple comparisons (at α < 0.05) implemented using the FDR procedure for adjusting p values described by Benjamini and Hochberg 55 and implemented with MATLAB's mafdr function.
RSA was conducted using MATLAB. RDMs for each ROI and participant were constructed by correlating the beta coefficients for each pair of stimuli across voxels in the ROI, and taking 1-correlation (Spearman's rho) to convert to a dissimilarity measure 18,19 . The mean of individual participants' RDMs produced one RDM per ROI (Fig. 2c). MDS was performed using the MATLAB function cmdscale on the average RDM across participants for each ROI, with the first two dimensions plotted for visualization (Fig. 2d).
A whole-brain decoding searchlight was conducted using the Newton linear SVM classifier implemented in TDT 54 with a searchlight radius of 3 voxels. We used a leave-exemplar-out cross-validation approach as implemented for the ROI decoding analysis, however, to improve performance for the searchlight we used the faster Newton implementation of SVM and did not implement leave-one-runout cross-validation in addition to leave-exemplar-out. The searchlight was conducted in each participant's native brain space, and then the results were mapped on to surface nodes in SUMA 53 for visualization of the average group decoding maps for all participants (Fig. 3).
MEG data acquisition. MEG data were acquired using a 160-channel (axial gradiometers) whole-head KIT MEG system (Model PQ1160R-N2, KIT, Kanazawa, Japan) at the KIT-Macquarie Brain Research Laboratory (Sydney, Australia). Recordings were collected with a 1000 Hz sampling rate and filtered online between 0.03 and 200 Hz. Participants lay in a supine position inside the MEG scanner within a magnetically shielded room (Fujihara Co. Ltd, Tokyo, Japan). Head position was tracked using five marker coils placed on a fitted elastic cap worn on the participant's head. A photodiode tracked the exact onset of each visual stimulus and was activated by the presentation of a small white square on the screen during each stimulus presentation.
MEG experimental design. The experimental script was written in MATLAB using functions from the Psychtoolbox [44][45][46] . Visual stimuli were presented in the MEG via a projector at 114 cm viewing distance and subtended 4°of visual angle. The 96 visual stimuli were presented in random order in 6 runs for each subject. Each run contained 4 repeats of each of the 96 stimuli, for a total of 384 trials per run and 2304 trials in total (24 repeats of each stimulus). A break occurred halfway through each run and participants pressed a key when they were ready to proceed. Images were presented in the center of the screen for 200 ms on a mid-gray background. As for the fMRI experiment, eye movements were not recorded, however, the brief stimulus duration prevented any substantial eye movements during stimulus presentation. The interstimulus interval was jittered and randomly varied between 1-1.5 s on each trial. Each image was presented tilted slightly by 3°t o the left or right of center (Fig. 1c) and participants' task was to report the direction of tilt with a keypress on an MEG-compatible button response pad. The tilt of each stimulus was carefully counterbalanced within and across runs so that tilt direction and/or motor response would not be an informative cue for multivariate pattern analysis. Averaged across participants and runs, the mean task accuracy was 93.2% (SD = 4.8%).
MEG preprocessing. MEG analysis was conducted using MATLAB (The Math-Works), including functions from the CoSMoMVPA 26 and FieldTrip 56 toolboxes. We performed minimal preprocessing on the data 17,57 . Trials were downsampled to 200 Hz (5 ms) temporal resolution and for each trial an epoch from −100 to 1000 ms relative to the onset of the visual stimulus was used for analysis. Principal components analysis was performed to reduce the dimensionality of the data, and the components explaining 99% of the variance were retained for each participant's dataset. No further preprocessing was conducted on the data prior to analysis 17 .
MEG multivariate pattern analysis. All analyses were conducted on the principal components across all 160 sensors individually for each subject in sensor-space, with the results then pooled. For cross-decoding we used a leave-one-exemplar-out cross-validation approach. An LDA classifier was trained to discriminate between two of the three categories (e.g., human faces vs. objects) on all but one exemplar of each category (i.e., n = 31 images per category). We chose LDA for MEG decoding based on the finding that SVM and LDA produce comparable results for MEG data using the PCA preprocessing pipeline we apply here, yet LDA is generally faster and thus suited to handling MEG timeseries data 17 . This process was repeated in an iterative manner across each cross-validation fold so that each combination of exemplars was used as test data. Classifier accuracy was averaged over each crossvalidation fold, and then averaged across participants (Fig. 5a). Statistical significance was assessed using Threshold-Free Cluster Enhancement as implemented in the CoSMoMVPA 26 toolbox.
RDMs for each time point (from −100 ms before stimulus onset to 1000 ms after stimulus onset, in 5 ms increments) and participant were constructed by correlating the whole-brain patterns across sensors (after PCA) for each pair of stimuli, and taking 1-correlation (Spearman's rho) to convert to a dissimilarity measure 18,19 . The mean of individual participants' RDMs produced one RDM per time point (Fig. 5b; Supplementary Movie 1). MDS was performed using the MATLAB function cmdscale on the average RDM across participants for each time point, with the first two dimensions plotted for visualization ( Fig. 5c; Supplementary Movie 2). Comparison between the time-varying MEG RDM and behavioral, visual feature, and saliency RDMs (Fig. 6f) was made by correlating the RDMs using Kendall's tau-a 25 . This approach was also used for the fMRI-MEG fusion 27 in Fig. 7.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Data availability
The original experimental stimuli and the empirical RDMs (fMRI, MEG, behavioral, output of computational models) are publicly available at the Open Science Framework website for this project: https://osf.io/9g4rz. Source data are provided with this paper. | 2020-09-12T13:06:44.292Z | 2020-09-09T00:00:00.000 | {
"year": 2020,
"sha1": "5dce72987dc352339a24d229514d19299f8c2a80",
"oa_license": "CCBY",
"oa_url": "https://www.nature.com/articles/s41467-020-18325-8.pdf",
"oa_status": "GOLD",
"pdf_src": "SpringerNature",
"pdf_hash": "524fbd6ccff401df1d36d5dbf6b34d306dbf9ed3",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Medicine",
"Computer Science"
]
} |
243938668 | pes2o/s2orc | v3-fos-license | Conservative Hybrid Automata from Development Artifacts
The verification of cyber-physical systems operating in a safety-critical environment requires formal system models. The validity of the verification hinges on the precision of the model: possible behavior not captured in the model can result in formally verified, but unsafe systems. Yet, manual construction is delicate and error-prone while automatic construction does not scale for large and complex systems. As a remedy, this paper devises an automatic construction algorithm that utilizes information contained in artifacts of the development process: a runtime monitoring specification and recorded test traces. These artifacts incur no additional cost and provide sufficient information so that the construction process scales well for large systems. The algorithm uses a hybrid approach between a top-down and a bottom-up construction which allows for proving the result conservative, while limiting the level of over-approximation.
INTRODUCTION
Hybrid systems connect the discrete, logical world of computers with the continuous, unpredictable real world. In the last decades, they became an inevitable part of society by controlling essential infrastructures such as power plants, commercial airplanes, and cars. A mathematical model for such systems are hybrid automata, which combine information regarding their discrete control structure and continuous physical behavior. They can be used in all phases of Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from permissions@acm.org. HSCC'22, May 4-6, 2022, Milan, Italy © 2022 Association for Computing Machinery. ACM ISBN 978-x-xxxx-xxxx-x/YY/MM. . . $15.00 https://doi.org/10.1145/nnnnnnn.nnnnnnn development. Before deployment, they allow for analysis and verification of critical properties. During deployment, time-bounded prediction and identification of anomalous or unexpected behavior based on the model is possible. Lastly, recorded mission data enables a post-mortem analysis as well. Albeit indisputably beneficial for the development process, designing a hybrid automaton to properly reflect the semantics of the system is a delicate process. Thus, unsurprisingly, several approaches aim at automatically constructing either hybrid automata or their simpler cousins, timed automata, based on execution traces of the system. These traces are usually a development artifact as they get recorded during test runs. Estimation methods, most prominently machine-learning, yield promising results in terms of reconstructing the correct discrete structure of the automaton and approximating the continuous dynamics. Their great accuracy notwithstanding, the direction of the approximation is unclear, resulting in over-and under-approximation. While this suffices for conveying the gist of the system, it does limit its capabilities in terms of safety-critical analyses. For this reason, we propose a construction algorithm for conservative hybrid automata, i.e., a guaranteed over-approximation of the original system.
Rather than relying solely on execution traces, the construction employs another development artifact: a formal runtime monitoring specification. This specification is crucial for safety-critical systems as it enables dynamic verification methods -runtime verification -in which a dedicated component monitors the behavior of the system during the execution. When the monitor deems this information indicative of a malfunction, it terminates the execution. Note that the constraints imposed by the specification encompass the entire execution and change depending on the state of the system. Hence, to judge the situation accurately, the monitorand by proxy the specification -needs to keep track of different operational phases.
This comes to show that both the execution traces and the specification manifest expertise acquired during the development process: a) keeping track of operational phases requires the specification to contain information regarding the discrete control structure and b) since the execution traces are recorded test runs, they are expected to satisfy relevant coverage criteria. Hence, not only do they cover the "average", expected behavior including initialization and termination, they also cover the extreme and corner case behavior. Consequently, both the specification and the traces constitute indispensable resources.
The construction algorithm proposed in this paper first extracts the implicit discrete control structure from the specification, resulting in a strong over-approximation of the system. It then proceeds by a) refining this information based on discrete control signals contained in the execution traces and b) enriching it by analyzing the evolution of samples over time. The resulting automaton arXiv:2111.05613v2 [cs.FL] 21 Jan 2022 is an under-approximation. Hence, the last step of the algorithm merges control modes within the automaton based on discrete evidence found in the traces to finally obtain an over-approximation of the original system. This mixture of a top-down and bottomup construction results in an automaton that is both a provable over-approximation and retains a high level of precision.
Apart from being conservative, the construction distinguishes itself from existing approaches in two major ways. First, the specification roughly indicates the general discrete structure of the constructed automaton. This alleviates the need to second-guess the structure in its entirety, reducing revisions to local sub-structures. This pushes scalability far beyond * -based approaches [5] like Medhat et al. [14] in which significant time is spent to determine the discrete structure. Secondly, the construction reduces the level of over-approximation by merging modes of an under-approximation only if needed. This can result in more fine-grained refinements than when successively widening dynamics until the language of the automaton encompasses every input trace [18].
An empirical evaluation validates three major claims. First, the construction requires few traces to produce decent results. For a fourteen-mode automaton, for example, seven hand-picked or on average 35 random traces suffice for a perfect reconstruction. Secondly, the precision -while not flawless -comes close to the optimal result for adequate input data. Thirdly, the construction algorithm scales extraordinarily well. Even large automata with over 1000 modes can be constructed within mere seconds. All three benefits are the result of relying on development artifacts in form of test traces and a runtime monitoring specification: a readily available resource often left under-utilized.
All in all, the contributions of this paper are: • A three step construction for conservative hybrid automaton: First, it extracts a discrete over-approximation of the system from a runtime monitoring specification. This approximation is successively enriched with continuous information and refined into an under-approximation by incorporating data from execution traces. Lastly, it merges modes based on discrete evidence found in test traces until obtaining an overapproximation. • A correctness proof of the construction: under realistic assumptions on the input traces, the constructed automaton subsumes the language of the original system projected onto the behavior exposed through the set of input traces. The projection is required to eliminate parts of the system that are not exposed to the outside such as unreachable modes. • Experimental results showcasing the quality and scalability of the conservative construction. Even small sets of input traces stemming from random walks allow for precise constructions. Moreover, automata with thousands of modes can be constructed in a matter of seconds.
MOTIVATION
The foundation for the conservative construction is a set of traces generated from an unknown system and a runtime monitoring specification thereof. As a running example, consider the automaton depicted in Figure 1a, a simple model for an aircraft. The system starts in a takeoff mode and resides there until reaching cruising altitude. Here, it can go straight or adjust its course via left and right curves until it attempts a landing. Under windy conditions, the descend-phase is elongated as a precaution. A specification for the aircraft imposes several constraints depending on the current state of the system. For example, during takeoff, the specification requires the aircraft to accelerate; while traveling it requires a stable altitude; during landing it requires the landing gear to be lowered. An analysis of the specification hence yields a state machine with coarse information on different execution phases as well as conditions on phase-changes. The state machine is depicted in color in Figure 1a, superimposed by the aircraft automaton. As can be seen, the specification does not distinguish between maintaining course or adjusting it; the requirements on the system remain the same. Yet, it contains no indication regarding the continuous behavior of the system.
To fill these gaps, the conservative construction then successively enriches the specification automaton with information extracted from the set of traces. The whole process is illustrated in Figure 1c. It first translates the specification automaton A Φ into a hybrid automaton H + 1 . For each step of each trace, it adds more modes into the automaton while maintaining the structure provided by the specification. The result, i.e., H + |Π | , is by design overly restrictive: it consists of a single, isolated path for each trace. A merge process based on the discrete behavior of modes remedies this problem and finally produces H + .
The key point behind H + is that it is conservative, i.e., under certain assumptions on the specification and traces, H + over-approximates H . The assumptions are three-fold: the specification needs to a) be a coarse abstraction of the actual system, b) agree with the system on phase changes, and c) the set of traces needs to encompass sufficient information on the discrete behavior. While these assumptions seem strong, they are tailored for the use case at hand such that they are expected to be satisfied naturally. The first and second assumption concern the specification, which is handcrafted specifically for the system. Hence, the specifier must have had knowledge regarding the abstract control structure, e.g. through which phases the system traverses during a mission. Evidently, the concrete control structure refines the abstract structure for more fine-grained control. This abstract control structure manifests itself in the specification. Here, each mode of the abstract structure imposes a different set of requirements on the system. A violation of such a requirement constitutes a safety-hazard. Hence, engineers need to define precisely when the requirements on the system changes. As a result, the specification needs to contain the same precise criterion for a phase change as the system itself.
Regarding the third criterion, recall that the set of traces is a development artifact that arose from the testing process. Thus, not only do they cover large parts of the system's regular execution, they represent executions in which the system was purposefully coerced into extreme and corner case behavior. Moreover, the empirical evaluation revealed that even small sets of random walks through the underlying system already produces an adequate trace set. Here, "small" means on average 32 traces for a fourteen state automaton and two traces for a seven state automaton. This reasoning shows why these assumptions, while strong, are realistic when considering carefully engineered, safety-critical system. A formal description of the assumptions and the proof that H + is conservative can be found in Section 5. The empirical evaluation of the approach in Section 6 allows for validating three claims. a) The construction requires a low volume of input traces -especially when compared to machine-learning approaches. For the running example, the construction requires as little as three traces of length eight. b) It scales linearly for increasing dimension and quadratically in the number of traces and size of the original/constructed system. Constructing a three-dimensional automaton with 2 10 modes based on a specification with nine states and 512 traces requires less than a second. c) Though over-approximating, the constructed automaton has a high level of precision. For the aircraft, the construction is identical to the original apart from a merge of the two terminal modes.
PRELIMINARIES
Definition 3.1 (Interval). I = R 2 denotes an -dimensional rectangle where I 1 = I is an interval over R. The multiplication of a -and an ℓ-dimensional rectangle yields a ( + ℓ)-dimensional rectangle. Addition and multiplication of intervals with scalars are geometric translation and scaling, respectively. Definition 3.2 (Convex Hull). Let be a convex set. The convex hull of two convex sets ⊆ and ⊆ is the minimal convex set covering both and . Further, let S be a non-empty set of convex sets. Conv(X) computes the convex set covering all elements of X by applying the convex hull iteratively in arbitrary order. Conv(X) yields the identity for singleton inputs.
Hybrid Automata
An -dimensional Multi-Rectangular Hybrid Automaton H is a 6tuple ( , Λ, flow, , , ) over R . denotes the finite set of discrete control modes. A state ∈ × R of the automaton consists of a discrete mode and a continuous state. Here, = ( , ) ∈ × R is the initial state and Λ is the finite set of action labels. flow : → I defines the dynamics of modes. When entering one, a random value is drawn from the respective interval for each dimension. ⊆ × Λ × is the finite set of edges containing discrete, labeled transitions between two modes. Lastly, : → I assigns guard conditions to edges. A transition can only be taken in a state if its continuous component lies within the rectangle. Figure 1a shows the easy-to-grasp visual representation of a hybrid automaton where each white rectangle constitutes a mode. The text inside it defines the dynamics as differential equations, edge labels state guard conditions. Semantics. Hybrid automata allow for two kinds of transitions: control mode changes according to and delays according to the flow of the current mode during which the system state evolves continuously. More formally, the semantics of a multi-rectangular hybrid automaton H are defined based on valid omniscient traces . . , , , , +1 is valid for an automaton ( ⊲ H ) iff: a) the trace starts in the initial state of H , i.e., = 0 . b) the first discrete transition starts in the initial mode, so 1 = ( , , ) for some and . c) all guards are satisfied: ∀1 ≤ ≤ : ∈ ( ). d) all delay transitions are valid, i.e., for 0 ≤ ≤ and +1 = ( , , ), the state changes according to the flow: +1 ∈ ( + · flow( )). The language of an automaton is the set of valid traces:
Definition 3.3 (Observable Traces
). An observable trace is an omniscient trace stripped of its information regarding source and target modes: For the remainder of the paper, unless stated otherwise, a trace refers to an observable trace and we only consider finite languages.
Notation. Any automaton with decoration such as H + will be implicitly destructed into its components with the same decoration, e.g. + denotes the modes of H + . ( ) denotes the th component of the -dimensional vector for 0 < ≤ . For a finite set , | | denotes the cardinality of . The length | | of a trace ∈ R × R ≥0 × R × (Λ × R ≥0 × R ) is the number of timed transitions occurring in it, i.e., + 1. A trace of length + 1 is implicitly destructed into the following components: = . . , , , , +1 . Moreover, mode is a member of an omniscient trace if it reaches the mode at least once: A step of a trace is the combination of a delay and a discrete transition. Further, let Π be a sequence of traces in arbitrary order. Then, denotes the th entry of the sequence with ≤ |Π|.
Bisimulation. Discrete bisimulation on two (hybrid) automata is defined conventionally by disregarding any continuous behavior or behavior not shared among the automata.
CONSTRUCTING CONSERVATIVE AUTOMATA
The construction proceeds in three steps: First, it extracts information from the specification to obtain a finite state machine A Φ and a table mapping discrete control mode changes to conditions for undergoing such a change. The automaton is a coarse abstraction of the underlying system. The second step transforms it into a hybrid automaton H + and iteratively refines it by extracting information regarding the continuous behavior from the input traces. By design, the refinement overshoots its goals, resulting in an abstraction that is too fine. As a remedy, the third step merges parts of the automaton to construct a conservative automaton. The alternation in coarseness has the effect that the resulting automaton is an over-approximation without being overly permissive.
Extracting Discrete Information from the Specification
The requirements on the system change depending on its current state. For example, during the landing of an airplane, the landing gear must be lowered whereas it is required to be retracted when on traveling altitude. Hence, the specification needs to keep track of relevant parts of the system state to impose the proper restrictions. This process of keeping track induces an abstract state machine that lacks any information on the continuous dynamics since the monitor solely relies on external input data such as sensor readings. Each abstract state may summarize several concrete modes of the actual system. In the plane example, the requirements on the abstract mode "in full flight" apply to both the control mode "no wind" and "tail wind" even though they have different continuous dynamics. By assumption, the contrary is false: a change of requirements on the system is always accompanied by a change in concrete modes. Intuitively, a change of requirements is strongly linked to an action or reaction of the system: approaching a geofence imposes new constraints and hence prompts a reactionary change of course to satisfy them. A formalization of these assumptions follows in Section 5.1. The extraction of the abstract automaton varies depending on the specification language. This paper uses the RTLola [9,17] monitoring framework since it was designed for and integrated into safety-critical cyber-physical systems [1,6,8]. An RTLola specification consists of input streams representing data the monitor receives from the system, output streams and triggers. With output streams the specifier declares how to process input data with the goal of analyzing the state of the system. Lastly, triggers define conditions on output streams. The satisfaction of a trigger condition prompts the monitor to emit a message to the system, informing it about a violation of a safety requirement or the detection of a phase change.
A specification can keep track of the current set of requirements imposed on the system by using an output stream Φ . The value of Φ indicates in which abstract state the system is. Assume there are two abstract states Φ 1 and Φ 2 , and a state transition occurs under some condition . Then, the Φ stream has the following shape: Here, last( Φ ) provides the last value of the Φ . For more possible abstract states and transitions, the conditional statement can be extended accordingly. In addition, a state change is accompanied by a respective trigger: The trigger checks for a change in Φ from Φ 1 to Φ 2 and emits this information coupled with the name of the respective transition.
An analysis of the output stream and trigger declarations yields two artifacts: In the construction of A Φ , Φ and Φ are the domain and initial value of Φ , respectively. Then, for each trigger as the one stated before, Φ contains the edge ( Φ 1 , , Φ 2 ) and Γ Φ ( ) = . Note that the following assumes A Φ to be free of unreachable states and related edges. This is the case in sensible specifications and can easily be enforced by pruning the respective parts of the graph.
Extracting Continuous Information from Traces
While the specification provides information about the system's discrete structure, the traces reveal how the continuous state of the system evolves over time. They also reveal mode changes within a single abstract state. This information allows for transforming A Φ into a more fine-grained automaton with annotated dynamics in each mode. For this, the transformation iteratively constructs an automaton H + , processing each position of all traces in separation. This requires to keep track of two maps: a concrete mode-map : Π → that maps each trace to the mode of the constructed automaton in which it currently resides, and an abstract mode-map : → Φ mapping each concrete mode to an abstract one in the specification automaton A Φ . Definition 4.2 (Construction Initialization). The construction starts with a quasi-empty hybrid automaton H + 1 that is structurally similar to A Φ , a concrete mode-map 1 and an abstract mode-map 1 defined as: Here, 1 denotes the neutral element with respect to the multiplication of intervals. Moreover, the solve-function computes the singular interval representing the linear dynamics exhibited by a delay transition: Thus, H + 1 already incorporates the information of each trace regarding their first delay transition.
After the initialization, the procedure successively incorporates information contained in further positions of the traces.
Intuitively, for each position of each trace the construction (i) adds a new mode with the dynamics exhibited by the delay transition, (ii) adds a new edge from to ′ for the discrete transition, and (iii) updates the mode maps accordingly. The latter means that if the transition was accompanied by a step in A Φ , maps the ′ to the respective abstract mode and looks up the guard from the specification. Otherwise, it maps ′ to the same abstract state as with a vacuous guard indicating a lack of information.
Merging Modes
Evidently, following the procedure yields an automaton with at most | | · |Π| modes arranged as a tree as can be seen in Figure 1c. It transformed the overly coarse specification automaton into an overly fine hybrid automaton. To find the sweet spot between both extremes, the next construction step merges modes within an abstract state provided they are sufficiently similar. Suppose some relation ∼ ∃ captures this notion of similarity. Then, intuitively, the construction deems any two modes ̸ ∼ ∃ ′ sufficiently dissimilar such that they must represent different modes in the original system. For this, let ∼ denote the relation induced by for a constructed hybrid automaton, 1 ∼ 2 indicates that both modes refine the same abstract state, i.e., ( 1 ) = ( 2 ).
Definition 4.4 (Action Similarity).
For a constructed hybrid automaton H + , two modes 1 , 2 ∈ + are action-similar if they share some discrete characteristics and reside in the same abstract state of the specification. Assume there are some modes ′ 1 , ′ 2 ∈ + and action ∈ Λ + .a Note that by construction ∼ is coarser than ∼ ∃ . Terminal modes need further attention: consider the automaton in Figure 1a. There are two identical traces in the language of the automaton starting in Takeoff and traversing Straight, but ending in different Landing modes. Based on these traces the construction cannot distinguish the two terminal modes, since the difference in modes is unobservable. In fact, there is no finite set of traces for which they can be distinguished with certainty. This forces the construction to merge them as can be seen in Figure 1b.
Two modes are merge-similar iff they are either action-similar or terminal-similar: ∼ = ∼ ∃ ∪ ∼ ⊥ Merge Operation. The merge operation now minimizes the automaton with respect to ∼ by building the quotient automaton. Formally, a merge operates on an equivalence relation ≈ over the set of modes where each equivalence class ⊆ will be replaced by a single representative ⟦ ⟧ ≈ . By slight abuse of notation let ⟦ ⟧ ≈ = ⟦ ⟧ ≈ for ∈ . Moreover, if context permits, the subscript may be omitted. The representative conserves the language of each mode contained in by retaining discrete transitions and computing the convex hull for its continuous components.
Putting it Together: Construction Algorithm
The overall construction algorithm now proceeds as outlined in Algorithm 1: First, the procedure extracts information from the specification, constructs the initial automaton and refines it successively by iterating over the traces. After processing all traces completely, the procedure computes and applies the merges with respect to action-and terminal-similarity.
Time Complexity. The construction process consists of three phases: extraction, construction and merging. Recall that the dimensionality, i.e., the number of continuous state variables is . The first phase scales linearly in the size of the specification O (|Φ|). The second phase construct an automaton with a single mode per step of any trace. Its size and the running time of the construction scales linearly with the number and length of traces as it creates a new mode per step of any trace. It is also linear in the dimension since the dynamics of each dimension have to be computed separately per mode. Hence, the complexity is in O ( · | | · |Π|). Lastly, the complexity of the last phase depends on the complexity of a single merge, which is linear in the dimension, and the number of merges. The latter is quadratic in the size of H + |Π | , which in turn is linear in the number and length of traces: O ( · |H + |Π | | 2 ) = O ( · | | 2 · |Π| 2 ). This, however, only describes the worst case. The procedure compares each mode against each other with respect to ∼ . In the best case, all elements of an equivalence class are identified successively by chance. In this case, the process is quadratic in the number of equivalence classes: Ω( · |∼ | 2 ). Here, |∼ | denotes the number of equivalence classes induced by ∼ with |A Φ | ≤ | ∼ | = | + | ≤ | + |Π | |. In conclusion, the overall asymptotic running time is dominated by the merge procedure: O ( · | | 2 · |Π| 2 )
CORRECTNESS OF CONSTRUCTION
The validation of the construction requires a proof that the constructed automaton is -under certain assumptions on the input -indeed conservative. For this, a key criterion is that the automaton over-approximates the discrete and continuous behavior of the original system when projected down to the parts that contributed to the inputs. Evidently, if the original system encompasses parts that were neither reflected in the specification, nor traversed in the input traces, the constructed automaton cannot reconstruct it.
Hence, this section first formalizes requirements on the input data. Then, a definition of projection automata enables proving that the constructed automaton subsumes the language of the projected original system.
Requirements on Input Data
The construction of the conservative automaton relies on the quality of the input traces and specification. Hence, they need to satisfy three criteria: (i) the specification must be an abstraction of the real system, (ii) its trigger conditions must be at least as restrictive as the respective conditions on mode changes, and (iii) the trace set needs to traverse every control mode of the system sufficiently often to capture the discrete behavior.
Definition 5.1 (Adequacy of Input Data).
A specification Φ and trace set Π are adequate for a hybrid automaton H iff they satisfy three criteria.
(1) The specification induces a coarser automaton A Φ than the original, i.e., ∃ ∼ Φ : H ↓ ∼ Φ ≈ A Φ . (2) Assume Φ = ⟦ ⟧ | ∈ P . For any discrete transition that is both in H and A Φ , the specification contains a mode change condition that is at least as permissive as the guard of the respective transition in H .
For every mode in H , let = indeg( ) + outdeg( ) be the number of input and output actions of in Π. The trace set needs to contain more than ( − 1)/2 traversals through . Here, a trace traverses through a mode if its omniscient counterpart contains two subsequent edges first ending and then starting from .
2 Evidently, these criteria depend on the original hybrid automaton, which seemingly contradicts the premise of this paper regarding the unavailability of such a model. Nevertheless, the criteria are designed in a way that they are either satisfied naturally or can be satisfied without access to all formal details of the system. First, consider 5.1.1 and 5.1.2. These criteria restrict the specification, which was hand-crafted for the underlying system. Here, a reasonable specification summarizes control modes that are subject to the same requirements; at the same time, the specification needs to capture changes in the abstract state precisely to impose the correct sub-specification on the system. Thus, even without perfect knowledge of the inner workings and dynamics of the system, the first two criteria can be ensured. Consider the third criterion, which is concerned with the trace set. A thorough testing process demands that all discrete paths 1 through the system are tested at least once. Moreover, the system has a fixed control interface, represented by Λ. As a result, it is reasonable to assume that the number of times each control mode is traversed during the development exceeds the threshold required by Criterion 5.1.3. This again does not rely on knowledge about the exact mode structure nor dynamics of the underlying system.
The exact threshold for the third criterion seems arbitrary but is anchored in graph theory, the impact of which can be seen in the next lemma. Proof. By reduction on the graph connectivity problem. Let G( , Π) = ( , ) be a graph where is the set of labels of incoming or outgoing edges of in H . connects two action labels , ′ if there is a trace in Π that reaches via and leaves via ′ . The solution of the graph connectivity problem states that and ′ are necessarily connected if | | exceeds | | · (| | − 1)/2. This threshold corresponds to Criterion 5.1.3. Recall that ∼ ∃ relates all modes with at least one common incoming or outgoing edge label. Thus, since ∼ ∃ ⊆∼ , all respective modes are merged in H + and by Definition 4.6, the resulting mode ⟦ ⟧ ∼ retains these transitions. Lastly, ∼ refines ∼ Φ , hence ( ) = (⟦ ⟧ ∼ ), which concludes the proof. □
Projection Automata
The assessment of the quality of the reconstruction depends on the projection of the original system onto the set of traces. This first requires a definition of projections on automata.
Definition 5.3 (Projection Automata).
The projection of an automaton H down to a set of omniscient traces Π is an automaton H | Π with the following constituents.
Here, for ∈ {min, max}, the min and max values for guards and flows are: Intuitively, the projection strips the automaton of any information not reflected in the set of traces. This reduces the sets of modes, edges, and transition labels. By definition, the initial state occurs in all traces and thus remains the same. Guards and flows are reduced to the maximum and minimum value exhibited by some trace.
Note that the projection automaton H | Π is not meant to be constructed at any point; it merely serves as theoretical point of reference for the quality of the construction. It is easy to see that in general the projection reduces the expressiveness of an automaton, i.e., L (H ) ⊇ L ( H | Π ). This, however, is not necessarily the case as the following theorem shows. The proof of Theorem 5.4 can be found in Appendix B. Note that the language equality cannot be extended to identical or isomorphic automata since H can contain unreachable modes that are not reflected in its language and thus not in any trace. The theorem emphasizes the generality of the conservative construction: For an appropriate trace set, the projection of an automaton perfectly resembles the original system. Since the constructed automaton is conservative with respect to this very projection, it is also conservative with respect to the original system. This is independent of the exact structure of the underlying system.
Construction Guarantees
The first observations are that application of a merge and iterations of the construction do not reduce the language of an automaton. The proof of the lossless merge can be found in Appendix B.
After showing the property of a merged automaton, we state that our construction is lossless. Lemma 5.6 (Lossless Construction). Given a set of traces Π and specification Φ. For any iterations and , if ≥ , then the set of edges, the flow, and the transition guards only grow over the iterations:
Proof. This lemma follows directly from the construction step (Definition 4.3). □
This suffices to prove that the language of the constructed automaton at least includes all input traces. Theorem 5.7 (Input Trace Inclusion). Given an adequate set of traces Π and specification Φ, the language of a constructed automaton H + subsumes Π, i.e., Π ⊆ L (H + ).
The proof of Theorem 5.7 can be found in Appendix B. A stronger classification of the language of H + requires some insight into its discrete structure in relation to the projection automaton of the original system. Specifically, H | Π has a finer discrete structure than H + . Lemma 5.8 (Discrete Refinement). For an adequate specification Φ and set of traces Π for a hybrid automaton H , the reconstruction H + is coarser than the projection of H onto Π, i.e., ∃ ∼ + : H + ↓ ∼ + ≈ H | Π .
Proof. The proof proceeds in two steps. First, for an arbitrary trace through H | Π it generates a trace ′ through H + . Second, it constructs the equivalence relation ∼ + based on these trace pairs.
Step 1: For a given ∈ L ( H | Π ), the proof inductively constructs ′ with ′ ∈ L (H + ) such that the observable traces for and ′ are equal. Moreover, for any step : . The induction base is trivial since both traces originate in the fixed initial state, which corresponds to the initial state of the specification automaton. Induction Step: Suppose the observable traces for Note that the refinement can be a true refinement since the merge criterion might falsely relate modes that are distinct in H | Π but share some discrete behavior, as discussed before. Corollary 5.9. ∼ + is finer than ∼ .
EXPERIMENTS
The empirical evaluation shows the scalability and precision of the approach presented in this paper. It is based on a prototype implementation in Rust 2 and the code is open source. All experiments were conducted on an Intel i5-7200u with 8GB RAM.
Aircraft System
As a first proof of concept, consider the running example from Figure 1a. For adequate input traces, the output will always be structurally equal with varying dynamics. This can be seen in Figure 1b, which shows the results of two construction processes. The dynamics in black are constructed from three hand-picked traces of length eight. Two of these traces travers all three Travel modes, whereas the last one skips the course adjustment modes and loops in one of the Landing modes instead. As can be seen, by picking the state values for the traces in such a way that they represent the extreme behavior, the reconstruction of the dynamics is perfect. Conversely, the constructed dynamics based on an adequate trace set of ten traces obtained by conducting random walks on the original system is shown in gray. The traces can be found in Appendix A. Evidently, the reconstruction closely resembles the original system both structurally and in terms of dynamics despite being based on a small set of random traces.
Scalability
Recall the complexity of each step of the construction algorithm, i.e., extraction, construction, and merging, from Section 4.4. The extraction only requires a single pass over the specification and is thus negligible. The construction and merges depend on the dimensionality of the system and the number and length of traces. The merge also depends on the number of equivalence classes with respect to ∼ in the best case, which is the size of the output automaton.
For this reason, the scalability evaluation considers exactly these three factors: dimensionality, number and length of traces, and output-size. To this end, it automatically generates an automaton with matching specification and adequate trace set. The automaton is shaped like a binary tree of variable depth (scales the length of traces) where each of the 2 +1 − 1 nodes is a control mode with dynamics of variable dimension (scales the dimensionality). The specification summarizes a variable number of modes with equal depth (scales the output size) and generates a variable number of adequate traces (scales the number of traces) enabling the respective merges. Figure 3a and Figure 3b show the running time and memory consumption for varying sizes of the original automaton. The green line represents runs where ∼ only equates identities, prohibiting any merges. For the blue line, two modes are equal if they have the same depth in the underlying automaton. The number and size of the traces required for an adequate trace set scales linearly with the size and the depth, respectively. Independent of the existence of merges, the running time lies below a second for automata with less than 2 10 modes and increases steadily afterwards -as expected considering the asymptotic complexity. Even for automata with 2 15 modes, the construction terminated after less than an hour (green line) or half an hour (blue line). The memory consumption behaves similarly, starting to rise significantly around 2 7 owing both to the increased number of traces stored in memory, and the resulting size of H + |Π | . Note that the memory consumption almost exclusively stems from the construction process; merging only deallocates memory. In terms of running time, the lion's share comes from merging due to the difference in time-complexity. At a size of 2 9 modes the running time of the construction process amounts to 3.061% and decreases to 0.015% for automata with 2 15 modes.
The dimensionality impacts the running time to a lesser extent. Raising the dimension from 1,000 to 7,000 for an automaton size of 10 10 increases the running time from around 4 s to 14 s. The limiting factor here is the memory consumption: each additional dimension increases the memory consumption of every guard condition, mode, and trace step. As a result, 10,000 dimensions requires around 5 GB of memory, which is also reflected in a relatively steep increase in running time to 96 s. Lastly the number of traces has an almost identical impact on both the running time and memory consumption. The scale of the impact lies in-between the one of the dimensionality and the size of the automaton. Raising the number of traces from 3,000 to 16,000 increases the running time roughly 40fold. Detailed results for the dimensionality and number of traces can be found in Appendix C.
Comparison Against Other Approaches
The construction of a hybrid automaton requires the determination of both the discrete structure and continuous behavior. Regarding the former, the conservative construction relies on the information provided by the specification and refines abstract states according to trace information. This restricts revisions to a local level. In absence of such a specification, other approaches resort to learning algorithms. Medhat et al. [14] use a modification of Angluin's * algorithm [5] to learn the discrete structure separately from the dynamics. Tappler et al. [20] on the other hand learn a timed automaton with genetic programming. Note that these automata alleviate the need to infer dynamics. We will evaluate the conservative construction against both of these approaches. Anguin's * and Clustering. Medhat et al. [14] use an adaptation of * and clustering to identify the discrete and continuous behavior of a system, respectively. In their case study, they use a Simulink model of a closed-loop engine timing control system. Figure 2a shows an approximate representation of the system as hybrid model. Their construction uses eight traces to generate an automaton that resembles the underlying system for new traces up to an error of 2.6%. Note that a detailed comparison is exacerbated by a lack of information regarding the running time, memory consumption, and length of the input traces. The conservative construction can perfectly reconstruct the automaton with one hand-picked trace of length ten within less than 1 ms. When using random walks, an average of 35 traces of length fifteen suffices for the perfect reconstruction. In this example, the specification always summarizes modes belonging to a certain operation of the system, i.e., a drop, jump, ramp-up and the stable configuration.
Genetic Programming. Tappler et al. [20] use genetic programming to successively adapt a candidate automaton to encompass all input traces. As an example, they consider a timed automaton modeling a car alarm system (see Figure 2b). A sufficiently precise reconstruction requires 2,000 randomly generated traces and took a mean of around 100 min. When using seven hand-selected traces, the conservative construction can perfectly reconstruct the system within less than 1 ms, disregarding resets. With random walks, an average of 35 traces of length fifteen is necessary for the perfect reconstruction. Note that the nature of the car alarm system does not allow for a meaningful summary of several modes in the specification. However, in terms of performance, the specification is irrelevant owing to the small size of the system.
RELATED WORK
The theory of hybrid automata was first studied by Henzinger [11] as the real-time extension of timed automata [3]. Learning the complex structure of timed and hybrid automata is a line of research that resulted in deterministic and stochastic reconstruction algorithms.
Niggemann et al. [15] present the tool HyBUTLA that builds prefix trees of the traces and applies merges when appropriate. Since Angluin's * algorithm [5] is a prominent solution for learning discrete automata, several extensions for timed automata were proposed [4,10]. Based on that, Medhat et al. [14] split the learning process of a hybrid automaton into two steps. They first learn the discrete model of the automaton with * and then capture the dynamics using clustering. Both of these techniques can potentially be replaced or integrated into different frameworks. Hence, their approach is complementary to the conservative construction: substituting the clustering for the simpler solve function can yield better precision at the price of conservativeness. Focusing on the medical application domain, HyMN [12] learns patient specific parameters for hybrid automata deterministically. Soto et al. [18] synthesize a hybrid automaton with an online algorithm without relying on a specification as discrete template. While precision is very high and completeness is shown, learning a trace prompts a global analysis of the previously learned hybrid automaton, which incurs performance penalties. The conservative construction avoid this complexity by using the specification automaton and the adequacy criterion. This way, revisions are local and still retain correctness guarantees. Other approaches for learning hybrid automata using mathematical models for node identification were proposed by Summerville et al. [19] and Breschi et al. [7].
If large datasets of traces are available, stochastic learning of hybrid automata is feasible. Tappler et al. [20] use genetic programming to reconstruct timed automata both in an offline and online setting [2]. Santana et al. [16] build hybrid automata with the Expectation-Maximization algorithm to iteratively define the model parameters. An unsupervised learning approach was presented by Lee et al. [13], whereas Birgelen and Niggemann [21] use self-organizing feature maps. Despite the success of machine learning, the results do not provide provable guarantees.
CONCLUSION
This paper presented a construction algorithm for conservative hybrid automata from development artifacts in the shape of a runtime monitoring specification and pre-recorded execution traces. The construction is validated mathematically by proving that the result is an over-approximation under certain assumptions on the inputs. An additional empirical evaluation revealed both the extraordinary scalability of the construction and that even randomly generated inputs regularly satisfy the input requirements. Considering that -in a realistic setting -these inputs are high-quality artifacts acquired during development of the system, they should no longer be left under-utilized. Treating them as the valuable assets they are allows for constructing precise, conservative hybrid automata in a scalable fashion.
A ADEQUATE TRACES OF AIRCRAFT SYSTEM
This section of the appendix presents the adequate traces used to learn the version of the aircraft system depicted in Figure 1b with the dynamics written in black. Proof. The proof selects traces from L (H ) enabling the perfect projection. For each in , Π * contains a trace with ∈ . This immediately entails that the projected edge set, set of actions, set of modes, and initial mode are accurate when disregarding unreachable entities. For each mode, Π * encompasses four traces per dimension: one minimizing and one maximizing the flow and continuous state variable of the mode and dimension. The minimization and maximization is over the set of traces rather than over the mode itself. As a result, the projection of the flow is perfect. Lastly, for each mode, outgoing edge, and dimension, there are two traces in Π * which maximize and minimize the continuous state value before taking the transition. Hence, the projection of the guard condition is lossless in terms the language of the automaton.
Proof. By contradiction: Assume there is a trace ∈ L (H + ) \ L (H + ↓ ≈ ). As the merge operation is defined by unifying modes, either (1) takes a discrete transition or (2) traverses a continuous state not permitted in the merged automaton. Definition 4.6 "bends" edges such that they originate and end in the respective representatives. Hence, the construction retains all edges up to elimination of duplicates due to set semantics, ruling out (1). Regarding (2), Definition 4.6 builds the convex hull for all flow and guard definitions of the merged states. The convex hull is at least as permissive as its constituents, rendering a less permissive behavior impossible, which concludes the proof. □ Theorem 5.7 (Input Trace Inclusion). Given an adequate set of traces Π and specification Φ, the language of a constructed automaton H + subsumes Π, i.e., Π ⊆ L (H + ).
C ADDITIONAL EVALUATIONS RESULTS
This section presents supplementary material for the evaluation in Section 6. Figure 4 shows that the construction scales extremely well for an increasing dimension, both in terms of running time and memory consumption assuming the number of modes is a constant 2 11 − 1. The increase in running time mainly stems from the merge operation, which computes the convex hull of the dynamics for each dimension separately. However, the impact is rather low since the running time is less than 100 s even for 10 4 dimensions. Similarly, the memory consumption increases solely owing to the necessity to store the multi-dimensional dynamics of each mode. Figure 5 shows the result when constructing five-dimensional automata with the same number of modes, but varying number of traces. Each trace has a length of eleven. Increasing the number of traces results in an increase in running time since the construction has to traverse the current automaton for each traces. It also increases the memory consumption because traces lead to the creation of new modes. However, the construction scales well in both performance metrics with a running time of around 17 min and memory consumption of 200 MB for 16,000 traces. | 2021-11-11T02:31:54.989Z | 2021-11-10T00:00:00.000 | {
"year": 2021,
"sha1": "7349467e4e51a2c78fc253c5d605bf71db1f3c88",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "7349467e4e51a2c78fc253c5d605bf71db1f3c88",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
118551678 | pes2o/s2orc | v3-fos-license | SrFeAsF as a parent compound for iron pnictide superconductors
We have successfully synthesized the fluo-arsenide SrFeAsF, a new parent phase with the ZrCuAsSi structure. The temperature dependence of resistivity and dc magnetization both reveal an anomaly at about T_{an} = 173 K, which may correspond to the structural and/or Spin-Density-Wave (SDW) transition. Strong Hall effect and magnetoresistance were observed below T_{an}. Interestingly, the Hall coefficient R_H is positive below T_{an}, which is opposite to the cases in the two parent phases of FeAs-based systems known so far, i.e., LnFeAsO (Ln = rare earth elements) and (Ba, Sr)Fe_2As_2 where the Hall coefficient R_H is negative. This strongly suggests that the gapping to the Fermi surfaces induced by the SDW order is more complex than we believed before that it removes the density of states on some Fermi pockets and leaves one of the electron pockets less-gapped or un-gapped. Our data clearly show that it is possible for the parent phase to have electron-like or hole-like charge carriers.
We have successfully synthesized the fluo-arsenide SrFeAsF, a new parent phase with the ZrCu-SiAs structure. The temperature dependence of resistivity and dc magnetization both reveal an anomaly at about Tan = 173 K, which may correspond to the structural and/or Spin-Density-Wave (SDW) transition. Strong Hall effect and moderate magnetoresistance were observed below Tan. Interestingly, the Hall coefficient RH is positive below Tan, which is opposite to the cases in the two parent phases of FeAs-based systems known so far, i.e., LnFeAsO (Ln = rare earth elements) and (Ba, Sr)Fe2As2 where the Hall coefficient RH is negative. This strongly suggests that the gapping of the Fermi surface induced by the SDW order leaves one of the hole pockets fully or partially ungapped in SrFeAsF. Our data show that it is possible for the parent phases of the arsenide superconductors to display dominant carriers that are either electronlike or holelike. The discovery of superconductivity in the quaternary compound LaFeAsO 1−x F x which is abbreviated as the FeAs-1111 phase, has attracted great attentions in the fields of condensed matter physics and material sciences. 1 The family of the FeAs-based superconductors has been extended rapidly. As for the FeAs-1111 phase, most of the discovered superconductors are characterized as electron-doped ones and the superconducting transition temperature has been quickly raised to T c = 55∼ 56 K via replacing lanthanum with other rare earth elements. 2,3,4,5,6,7 Meanwhile, the first holedoped superconductor La 1−x Sr x FeAsO with T c ≈ 25 K was discovered, 8,9 followed with the observation of superconductivity in hole-doped Nd 1−x Sr x FeAsO 10 and Pr 1−x Sr x FeAsO. 11 Later on, (Ba, Sr) 1−x K x Fe 2 As 2 which is denoted as FeAs-122 for simplicity 12,13,14 , and Li x FeAs as an infinite layered structure (denoted as FeAs-111) were discovered. 15,16,17 It is assumed that the superconductivity both in the FeAs-1111 phase and FeAs-122 phase is intimately connected with a Spin-Density-Wave (SDW) anomaly in the FeAs layers. 12,18 For undoped LaFeAsO, an SDW-driven structural phase transition around 150 K was found. 19 It seems that any new parent phase will initiate a series of new superconductors by doping it away from the state with features of a bad metal and the SDW order.
In this paper, we report the discovery of a new FeAsbased layered compound SrFeAsF which has the ZrCu-SiAs structure. As we know SrZnPF is a compound with the ZrCuSiAs structure 20 . We replace the ZnP sheets with FeAs sheets and get a new compound of SrFeAsF. The compound SrFeAsF has the tetragonal space group P4/nmm at 300 K. Both the resistivity and the dc magnetic susceptibility exhibit a clear anomaly at about 173 K, which is attributed to the structural and/or SDW transition. Surprisingly, a positive Hall coefficient R H has been found implying a dominant conduction by holelike charge carriers in this parent phase.
The SrFeAsF samples were prepared using a two-step solid state reaction method, as used for preparing the LaFeAsO samples. 21 In the first step, SrAs was prepared by reacting Sr flakes (purity 99.9%) and As grains (purity 99.99%) at 500 o C for 8 hours and then 700 o C for 16 hours. They were sealed in an evacuated quartz tube when reacting. Then the resultant precursors were thoroughly grounded together with Fe powder (purity 99.95%) and FeF 3 powder (purity 99%) in stoichiometry as given by the formula SrFeAsF. All the weighing and mixing procedures were performed in a glove box with a protective argon atmosphere. Then the mixture was pressed into pellets and sealed in a quartz tube with an Ar atmosphere of 0.2 bar. The materials were heated up to 950 o C with a rate of 120 o C/hr and maintained for 60 hours. Then a cooling procedure to room temperature was followed.
The dc magnetization measurements were done with a superconducting quantum interference device (Quantum Design, SQUID, MPMS7). For the magnetotransport measurements, the sample was shaped into a bar with the length of 3 mm, width of 2 mm and thickness of about 0.9 mm. The resistance and Hall effect data were collected using a six-probe technique on the Quantum Design instrument physical property measurement system (PPMS) with magnetic fields up to 9 T. The electric contacts were made using silver paste with the contacting resistance below 0.05 Ω at room temperature. The data acquisition was done using a DC mode of the PPMS, which measures the voltage under an alternative DC current and the sample resistivity is obtained by averaging these signals at each temperature. In this way the contacting thermal power is naturally removed. The temperature stabilization was better than 0.1% and the resolution of the voltmeter was better than 10 nV.
The X-ray diffraction (XRD) pattern for the sample Sr-FeAsF is shown in Fig. 1. One can see that all the main peaks can be indexed to the FeAs-1111 phase with the tetragonal ZrCuSiAs-type structure. Only small amount of SrF 2 impurity phase was detected. By using the software of Powder-X, 22 we took a general fit to the XRD data of this sample and the lattice constants were deter- mined to be a = 4.004Å and c = 8.971Å. It is clear that the a-axis lattice constant of this parent phase is slightly smaller than that of the LaFeAsO system, while the c-axis one is much larger, 1,21 indicating a completely new phase in the present system since the radii of Sr 2+ is larger than that of La 3+ .
In Fig.2 (a) we present the temperature dependence of resistivity for the SrFeAsF sample under magnetic fields up to 9 T. A rather large value of the resistivity is observed. An upturn in the low-temperature regime can be seen under all fields, representing a weak semiconductor like behavior for the present sample. It is unclear at this moment whether this behavior is intrinsic in nature, or it is due to the weak localization effect, or some other effect. This curve also reveals an anomaly at about T an = 173 K, which may correspond to the structural and/or SDW transition, as has been found in the parent phase of LnFeAsO (Ln = rare earth elements) and (Ba, Sr)Fe 2 As 2 . 1,12 Fig.2 (b) shows the zero field cooled dc magnetization of the same sample at 5000 Oe. A clear anomaly at about 173 K in the magnetization curve confirms the structural and/or SDW transition observed in the resistivity data. Above 173 K, the magnetization exhibits a rough linear temperature dependence, which may be a common effect in the FeAs-based systems and was explained as due to short range correlation of the local moments. 23 To get a comprehensive understanding to the conducting carriers in the SrFeAsF phase, we measured the Hall effect of the present sample. The inset of Fig. 3 shows the magnetic field dependence of Hall resistivity (ρ xy ) at different temperatures. In the experiment, ρ xy was taken as ρ xy = [ρ(+H) -ρ ( field dependence of ρ xy was observed in the temperature regime below 75 K, while the linear behavior appeared above 100 K. This may suggest that a multi-band effect or a complicated scattering mechanism (perhaps magnetic related) emerged in the low temperature regime. The temperature dependence of the Hall coefficient R H is presented in the main frame of Fig.3. One can see that R H remains positive in wide temperature regime and decreases monotonically in the temperature regime below about 160∼ 170 K and it becomes slightly negative above that temperature. The sign changing of R H and the temperature dependent behavior may be related to the structural and/or SDW transition as revealed by the resistivity data, considering that this change occurred at temperatures close to T an . It is worth noting that the positive Hall coefficient R H in this sample SrFeAsF is quite unique because in the two parent phases of FeAsbased systems known so far, i.e., LnFeAsO (Ln = rare earth elements) and (Ba, Sr)Fe 2 As 2 , the Hall coefficient R H is negative. This strongly suggests that the gap- ping to the Fermi surfaces induced by the SDW order is more complex than we believed before, and the case in SrFeAsF is that it removes the density of states on some Fermi pockets and may leave one of the hole pockets partially or fully ungapped. Our data clearly show that it is possible for the parent phase to have electronlike or hole-like charge carriers. It is well known that in the conventional metals the Hall coefficient R H is almost independent of temperature. The strong temperature dependence of R H below T an in our data suggests either a strong multi-band effect or the variation of the charge carrier densities, or both effects collectively contribute to the Hall signal in the present parent phase of SrFeAsF. The magnetoresistance (MR) is a very powerful tool to investigate the properties of electronic scattering. 24,25 Field dependence of MR for the present sample at different temperatures is shown in the main frame in the top part of Fig 4. One can see a moderate MR effect up to 9% under the field of 9 T at 2 K. This is a rather large magnitude compered with the F-doped LnFeAsO samples. 21,26 The semiclassical transport theory has predicted that the Kohler's rule will be held if only one isotropic relaxation time is present in a solid state system. 27 The Kohler's rule can be written as where ρ(H) and ρ 0 represents the longitudinal resistivity at a magnetic field H and that at zero field, respectively. Equation (1) ple is revealed in the inset of the top part of Fig.4. An obvious violation of the Kohler's rule can be seen from this plot. This behavior may indicate a multi-band effect or a gradual gapping effect to the density of states by the SDW ordering in the present sample. Temperature dependence of MR under the field of 9 T is shown in the bottom part of Fig 4. Rather similar to that observed in the R H vs T plot, ∆ρ/ρ 0 decreases monotonically in the low temperature regime below about 200 K and a minimum appears around T an . This may provide another evidence of the influence of the structural and/or SDW transition on the behavior of the conducting charge carriers.
In summary, a parent phase, namely SrFeAsF, with the ZrCuSiAs structure was synthesized successfully using a two-step solid state reaction method. An anomaly at about 173 K can be observed from the data of the resistivity and dc magnetization, which is ascribed to the structural and/or SDW transition. Also strong Hall effect and moderate MR were observed below T an . We found that the Hall coefficient R H is positive below T an , displaying an opposite behavior comparing to the cases in the two parent phases of FeAs-based systems known so far, i.e., LnFeAsO (Ln = rare earth elements) and (Ba, Sr)Fe 2 As 2 where the Hall coefficient R H is negative. This suggests that the gapping to the Fermi surfaces induced by the SDW order may remove the density of states on some Fermi pockets and leave one of the hole pockets partially or fully ungapped in the present parent phase. Our results clearly show that it is possible for the parent phase to have electron-like or hole-like charge carriers. We also observed a moderate magnetoresistance up to 9% under the field of 9 T. The violation of the Kohler's rule along with the strong temperature dependence of R H may suggest a multi-band and/or a spin scattering effect in this system. By doping strontium with lanthanum, we found superconductivity in Sr 1−x La x FeAsF, which will be presented separately. 28 Note added: When we were finalizing this paper, we became aware that a paper was posted on the website on the same day of our submission. That paper reports also the synthesizing of the compound SrFeAsF and a different set of data. 29 This work is supported by the Natural Science Foundation of China, the Ministry of Science and Technology of China (973 project: 2006CB01000, 2006CB921802), the Knowledge Innovation Project of Chinese Academy of Sciences (ITSNEM). | 2008-10-30T03:33:22.000Z | 2008-10-14T00:00:00.000 | {
"year": 2008,
"sha1": "e1fa794ebfdc129c15464aff3ffcb49ca6bf9743",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/0810.2475",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "e1fa794ebfdc129c15464aff3ffcb49ca6bf9743",
"s2fieldsofstudy": [
"Materials Science",
"Physics"
],
"extfieldsofstudy": [
"Materials Science",
"Physics"
]
} |
126671654 | pes2o/s2orc | v3-fos-license | Rare earth elements in surface sediments of a marine coast under heavy anthropogenic in fl uence: The Bohai Bay, China Estuarine, Coastal and Shelf Science
Surface sediments along four river-to-sea transects from the coastal Bohai Bay were assessed using a four-step sequential extraction procedure to determine the concentrations of rare earth elements (REEs) and the geochemical fractions in which those elements were present. The concentrations of light REEs were slightly depleted than those of the middle REEs and the heavy REEs. No signi fi cant Ce anomaly was observed whereas slight Eu-positive anomaly was found. The sequential extraction results revealed that the geochemical fractions of REEs followed the order of residual > reducible > oxidizable > acid soluble. The residual fraction accounted for the majority of the total concentrations of REEs. Middle REEs were more easily leached than other REEs. No obvious river-to-sea transportation was observed and the concentrations of REEs in surface sediments of the coastal Bohai Sea were consistent with the data from the Post-Archean Australian Shale. This study further demonstrated on the basis of our previous infer-ence that the geochemical features of REEs in coastal sediments were generally unaffected by heavy anthropogenic effects.
Introduction
The rare earth elements (REEs) are a coherent group of trace elements in the periodic table whose chemical properties are similar and change gradually and systematically with the increase of atomic number (Henderson, 1984;Dubinin, 2004). The coherent and predictable behavior of REEs, along with their sensitivity to pH, redox conditions, and their tendency to participate in adsorption/ desorption reactions, allows them to be used for the interpretation of diagenetic redox conditions in palaeoenvironmental research (Shields and Stille, 2001), discriminating sediment provenance and evolution (Grousset et al.,1998;Xu et al., 2011Xu et al., , 2012, and elucidating seawater circulation patterns (Tachikawa et al., 1999), hydrothermal fluxes (German et al., 1999), and the historic oxygenation of the oceans (Henderson, 1984;Murray et al., 1992).
REEs exist naturally in coastal sediments but they can accumulate therein though river runoff, groundwater discharge and atmospheric deposition due to their use in industry and agriculture (Strauch et al., 2008;Kulaksiz and Bau, 2011). Their environmental contamination has already been found in some mineralized areas and soils affected by the long-term application of sludge (Essington and Mattigod, 1990). Their effects on human health have also been documented, e.g. Gd accumulations can lead to kidney failure (Ergun et al., 2006), anaphylactic shock (Idee and Corot, 2008) and finally death (Kay, 2008). In recent years, the environmental behaviors of REEs have attracted increasing attention. Investigations of the background concentrations and geochemical fractionation of REEs in the environment including sediments, soils, and biological samples have been performed worldwide (e.g. Zhang et al., 1998;Leleyter et al., 1999;Cao et al., 2000;Zhang and Ishii, 2000;Yang et al., 2002;Borrego et al., 2005).
Approximately 97% of REEs are produced in China (U.S. Geological-Survey, 2011) and the industrial and agricultural usage of REEs has increased rapidly (Xie, 1991). Although REEs geochemistry in core sediments from the East China Sea (Xu et al., 2011) and the Bohai Bay (Xu et al., 2010) were studied, information about REEs especially the extent of anthropogenic effects on them in the Chinese coastal areas remains very scarce. Xu et al. (2012) investigated the fraction characteristics of REEs in surface sediments of the central Bohai Bay and deduced that the Haihe River may be a predominant source of REEs based on the similarity of their distribution patterns in different geochemical fractions in surface sediments between the Bohai Bay and the Haihe River.
In order to clarify the potential eco-environmental effects of extensive and intensive human activities from the surrounding areas, an investigation was carried out in the northwestern Bohai Bay, covering its intertidal and sublittoral zones and the lowermost reaches of the major rivers to which it connects. The geochemical characteristics of some environmental quality assessment related trace metals in surface sediments of this region indicated a general state that they were subject to apparent anthropogenic influence (Gao and Chen, 2012;Gao and Li, 2012), which was also the condition in many other areas along the coastal waters of the Bohai Sea . The results of this investigation also indicated that the REEs in the intertidal surface sediments of this region were generally unaffected by the heavy anthropogenic effects (Zhang et al., 2014). Although rivers transports both industrial and domestic waste effluents from the land to the sea, the role of rivers in conducting REEs from the land to the Bohai Bay has not yet been estimated. In the present study, both the concentrations and the fractionation of REEs in surface sediments of the coastal Bohai Bay from the major rivers it connects with to its coastal waters were determined to further evaluate the response of REEs to the anthropogenic effects from the surrounding areas in this region.
Description of the study area
The Bohai Bay is located to the west of the Bohai Sea and covers an area of 1.58 Â 10 4 km 2 , about one-fifth of the total area of the Bohai Sea (Fig. 1). The Bohai Bay is a shallow water basin with an average depth of 12.5 m (range 5.6e34.0 m) and a semi-diurnal tidal range of 0.5e2.5 m. The width of the tideland of the Bohai Bay is 3e5 km and the mean velocity of the residual current is less than 0.1 m/s. The exchange of water between the Bohai Bay and the Bohai Sea is weak so the physical self-cleaning capacity of the Bohai Bay is very poor (Tao, 2006). The bottom has a very mild-slope bottom and most of its sediments are fine mud.
The coastal region surrounding Bohai Bay is one of the most densely populated and industrialized zones in China, where some megacities, i.e. Beijing, Tianjin and Tangshan, are located. Several rivers scatter along the coast of the Bohai Bay ( Fig. 1), of which the Haihe River dominates land runoffs flowing into the Bohai Bay although its annual freshwater discharge had decreased from 8.20 Â 10 8 m 3 averaged over 1960e2010 to 3.76 Â 10 8 m 3 averaged for 2010 and 2011 (The Ministry of Water Resources of P.R. China, 2011). Other smaller rivers with only minor runoffs are the Dou River, Jiyun River, Chaobai River, Yongdingxin River, Duliujian River, Qingjinghuang River, Ziyaxin River, and Beipaishui River. These rivers bring in both municipal sewages and agricultural drainage from surrounding area, leading to the deterioration of the Bay's ecosystem.
Sampling
A total of 42 surface sediment samples (top 0e5 to 0e10 cm) were collected from the northwestern coast of the Bohai Bay in May 2008 (Fig. 1). The sampling stations were arranged along the major rivers of this area extending from the land to the sea and formed four transects. The water depths of all sampling sites in marine region are less than 10 m. The samples from the marine region were collected with a stainless steel grab sampler, and the samples from rivers were collected using a plastic spatula. The samples were placed in acid-rinsed polyethylene zipper bags and stored at~4 C in the dark. Lab analyses were performed immediately after landing to minimize storage time.
Analytical methods
The fractionations of REEs were obtained using the sequential extraction procedure described by Rauret et al. (1999). The scheme partitions elements into four operationally defined geochemical fractions, including acid-soluble (exchangeable and bound to carbonates), reducible (bound to Fe/Mn oxyhydroxides), oxidizable (bound to organic matter and sulfides) and residual (contained within lithogenic minerals). The details of the sequential extraction protocol used in this study have been described elsewhere in Gao et al. (2008Gao et al. ( , 2010 and Zhang et al. (2014).
Sample drying can reportedly change the solid phase distribution of trace elements (Rapin et al., 1986;Hjorth, 2004). Additionally, the element concentrations in sediments depend strongly on the grain size (Horowitz and Elrick, 1988;Howari and Banat, 2001), and a triturating treatment can potentially alter the extractability of elements (Gilliam and Richter, 1988). Therefore, wet and unground sediments were used herein in the sequential extraction procedure to reduce errors.
After the third step of extraction, a mixture of concentrated HF, HNO 3 and HClO 4 (5:2:1; Li et al., 2000) was used to digest all of the remaining REEs in the residues instead of the so-called pseudototal digestion with aqua regia, which was used by Rauret et al. (1999). The total digestion of three randomly selected samples was performed by the same method as that used to determine the concentrations of REEs in the residual fraction. The total concentrations of REEs in sediments were estimated by summing up the results for the four fractions, and they were 90e110% of the values obtained in the total digestion experiment.
Inductively coupled plasma mass spectrometry (Thermo X Series II) was used to determine the concentrations of REEs in this work. Plasma CAL calibration standards (SCP Science, Canada) were used to make up a linear calibration curve. Five multi-element standards were used to make up a linear calibration curve. Standard solutions were run after every 30 samples to monitor and correct fluctuations of instrumental drift. The Chinese national geostandard samples of GSS-1 and GSS-8 were processed in the same way as the sample was processed to control the analytical quality. The results matched the reference values, and the differences were all within 10% (most of them within 5%). All plasticware and glassware were pre-cleaned by soaking them in 10% HNO 3 (v/v) for at least two days. They were then soaked and rinsed with deionized water. All chemicals used in the experiment were guaranteed to be reagent-grade or better. Blank determinations were conducted throughout the processes of total concentration analyses and sequential extraction. All data were corrected for the dry weight of the sample.
Sample's water content, expressed as percentage, was calculated by the difference of wet and dry weight divided by wet weight. Dry weight was determined gravimetrically by heating sample at 105 C until a constant value was obtained. For each sample, the percentages of the following three groups of grain sizes were determined <4 mm (clay), 4e63 mm (silt), and >63 mm (sand) using an Malvern Mastersizer 2000 laser diffractometer that could analyze particles with sizes between 0.02 and 2000 mm. The total organic carbon (TOC) content in the sediments was obtained by subtracting the inorganic carbon content from the total carbon, which was determined using a Shimadzu TOC-V CPH /SSM-5000A and an Elementar Vario MACRO Cube CHNS analyzer, respectively.
General characteristics of sediments
The sediments were predominantly composed of silt (average 66.1%, range 57.0%e78.7%), followed by clay (average 28.6%, range 11.6e40.6%) and sand (<5%). For the three sediments from the Duliujian River, their fine fraction (clay þ silt) contents decreased downriver, and the one from the station nearest to the sea (DLJ-1) had the lowest value among all the studied samples. The TOC contents varied from 0.9 to 7.2% of the dry sediment weight with an average of 2.3%. The high TOC contents (>2.5%) were all recorded at the riverine sampling stations. However, the difference among the samples from the marine region was insignificant (1.6% ± 0.3%). The TOC contents showed a clear decreasing trend from 5.9% to 1.1% seaward at the Dou River-M transect. Detailed information about these can be found in Gao and Chen (2012) and . Table 1 presents the concentrations of REEs in surface sediments from the coastal Bohai Bay. At all stations, the concentrations of REEs followed the order of Ce >La >Nd > Pr > Dy > Sm>
REE concentration in bulk sediments
The concentrations of light REEs (LREEs; La to Eu), heavy REEs (HREEs; Gd to Lu), and the total concentration of REEs (SREEs) at each sampling site are presented in Fig. 2. At DH-M transect, SREEs concentration was found to be the lowest at DH-3 (120.9 mg g À1 ), increasing seaward up to 182.8 mg g À1 at M-2, then decreasing at M-3 and M-4, and finally rising to a maximum of 193.1 mg g À1 at M-5.
At YDX-K transect, SREEs concentration increased from 182.0 mg g À1 at the upstream station YDX-5 to 222.7 mg g À1 at the downstream station YDX-1, which was a maximum value among all the stations. At the marine stations from K-2 to K-8, the concentrations of SREEs displayed minor variations (153.5e168.2 mg g À1 ).
The average SREEs (172.6 mg g À1 ) in the surface sediments of the study area is consistent with the REEs abundance of the Chinese loess (171 mg g À1 ; Zhao et al., 1990) and higher than those of the East China Sea sediment (109.03 mg g À1 ), the Yellow Sea sediment (123.01 mg g À1 ), and pelagic sediment (125.93 mg g À1 ) (Zhu et al., 2006). This result shows that REEs have obvious terrigenous succession.
However, no decreasing trend was found from the river to the sea along each transect and also no significant difference was found among the four transects, indicating that land runoff was not a significant way to input REEs to the Bohai Bay. This finding is quite different from other heavy metals such as Cd, Cr, Cu, Ni, Pb and Zn in this area, which show an obvious decreasing trend in the riverto-sea direction caused by fluvial transportation from surrounding land (Gao and Chen, 2012).
Atmospheric transport and deposition plays an important role in introducing trace elements into coastal ocean. Dissolved atmospheric input is suggested to account for about 30e130% of REEs input to the oceans by rivers (Greaves et al., 2003). For the East China Sea and Yellow Sea, atmospheric fluxes of trace elements may exceed their riverine inputs . Wang (2003) studied the atmospheric REEs concentration above Beijing city and found that the enrichment factor of REEs against earth crust ranged from 3 to 10 and industrial plants (such as Yanshan Petrochemical Plant) were marked as significant point sources. So, we speculate that the REEs in this studied area may be introduced mainly via atmospheric transportation from surrounding anthropogenic activities.
Normalization analysis
To eliminate the so-called "Oddo-Harkins" effect, i.e. the evenodd variation in the natural abundance of elements and their isotopes, measured REE concentrations are normalized to their corresponding concentrations in the upper continental crust (Nozaki et al., 2000;Wood et al., 2006). The REE concentrations of Post-Archean Australian Shale (PAAS) (Taylor and McLennan, 1985), which was commonly used in environmental and marine-related studies, were chosen for normalization.
These results showed that the sampled area was not heavily Table 1 Concentrations of REEs along four transects in surface sediments from the coastal Bohai Bay. For convenience of comparison, the concentrations of REEs in sediments of the Changjiang River and Huanghe River (Yang et al., 2002) and Po River (Kramer et al., 2002) are also presented. Unit is mg g À1 for all elements. Zhang et al. (2014), probably due to their similar material source and sedimentary environment evolution. In surface sediments from the coastal Bohai Bay, the normalized concentration ratios of La/Gd and La/Yb were lower than 1.00 except at DLJ-2 and DLJ-3 (both were 1.03) (Fig. 2). The total average value of La/Gd was 0.80 ± 0.06 and La/Yb was 0.86 ± 0.08. The negative ratios indicated that the concentrations of LREEs were slightly depleted than the concentrations of middle REEs and HREEs. The total average value of La/Gd (0.80) was quite in line with those in the sediments from the Changjiang (0.80), Huanghe (0.80) and Po Rivers (0.80), while the total mean ratio of La/Yb (0.86) was 4e28% lower (1.20, 1.10, and 0.90 respectively) (Yang et al., 2002;Kramer et al., 2002).
Ce and Eu anomalies are widely observed because these two elements are present in nature not only as trivalent ions, like all other REEs, but also with other valences (Ce 4þ and Eu 2þ ). Ce 3þ is subject to be oxidized to Ce 4þ in oxygenated aqueous environment and Ce 4þ tends to be more rapidly removed by particle scavenging than other REEs. Consequently, a relative depletion of Ce with respect to its REE neighbors La and Pr must occur (Sholkovitz et al., 1994;Nozaki et al., 2000). The anomalous behavior of Ce mentioned above is usually indicated by the so-called Ce anomaly (dCe), which is calculated as dCe ¼ 2(Ce/Ce PAAS )/[(La/La PAAS ) þ (Pr/ Pr PAAS )]. Diagenetic remobilization of Eu is possible under conditions of reduction to Eu 2þ at low oxidation potential and Eu is reprecipitated in the 3 þ state in environments of high oxidation potential, yielding positive Eu anomalies in highly oxidized sediments (Macrae et al., 1992). Eu anomaly (dEu) is calculated as the ratio of dEu ¼ 2(Eu/Eu PAAS )/[(Sm/Sm PAAS ) þ (Gd/Gd PAAS )]. The distributions of dCe and dEu are quite uniform with global average values of 0.99 ± 0.05 and 1.05 ± 0.05, respectively, showing no significant Ce anomaly and slight Eu anomaly (Fig. 2).
The dCe in our present study are quite in line with those of the Changjiang River (1.00), Huanghe River (1.00), Po River (1.00) (Yang et al., 2002;Kramer et al., 2002) and the intertidal zone of the Bohai Bay reported in our early study (Zhang et al., 2014). Hoyle et al. (1984) suggested that Ce behaved coherently with its neighboring elements as a result of estuarine mixing so that Ce anomaly is not generated in estuary. The same result was also found in Elimbah Creek (Australia) by Lawrence and Kamber (2006).
Compared to the significant Eu-negative anomalies in the central Bohai Bay (Xu et al., 2012), the slight Eu-positive anomalies may be attributed to the well-oxygenated sediment in the coastal Bohai Bay due to shallow overlying water column. In addition, different nature of source rocks can lead to variation of Eu anomaly, sometime from negative to positive (Armstrong-Altrin et al., 2012, 2015. However, due to lack of data, we cannot discuss the possibility of source rocks-caused variation of Eu anomaly from coastal to central Bohai Bay herein.
Fractionation
Besides the total concentrations of metals, their physicochemical forms in sediments are also critical for evaluating their biogeochemical behaviors (Maiz et al., 2000;Gleyzes et al., 2002). Fig. 2 presents the percentages of the four geochemical fractions of REEs in surface sediments from the coastal Bohai Bay. At each site, the largest portion of REEs was associated with the crystalline lattices of minerals (residual fraction), followed by the fraction bound with FeeMn oxides (reducible fraction), the fraction bound with organic matter (oxidizable fraction), and the fraction of exchangeable and bound with carbonates (acid soluble fraction) sequentially. Samples from the Duliujian River (DLJ 1e3) showed high residual fraction (ranging from 69.0% (Ho) to 83.1% (La)) and low reducible and oxidizable fractions. Samples from the upstream Dou River (DH-2 and DH-3) displayed a minimum acid soluble fraction with a range of 0.24% (Ce)e2.8% (Eu). Samples from other sites presented similar percentages for each fraction, without apparent difference between the riverine and marine regions.
Averaged percentages of each REE geochemical fractions from the sampling sites in both the riverine and marine regions were presented in Fig. 4. In the riverine region, the average value of residual fraction accounted for 54.4% of their total concentration. The second largest was reducible fraction (32.2%), followed by oxidizable fraction (11.1%) and acid soluble fraction (2.3%) (Fig. 4a). In the marine region, the fractions of REEs followed the same order as mentioned above with 52.9% in the residual fraction (2.7% lower compared to riverine region), 34.0% in reducible fraction (5.7% higher), 9.6% in the oxidizable fraction (13.7% lower) and 3.5% in acid soluble fraction (50.4% higher) (Fig. 4b).
The sequence of average concentrations of REEs in different geochemical fractions in surface sediments of the coastal Bohai Bay coincides with that of the central Bohai Bay reported by Xu et al. (2012), although different sequential extraction method was used in their work and the acid soluble fraction was divided into exchangeable fraction and bound with carbonates fraction. Dominance of residual fraction indicated that most REEs were bound to mineral lattices in surface sediments of the coastal Bohai Bay, which was consistent with the finding that clay minerals are important reservoirs of REEs in Chinese coastal marine sediments (Zhao and Yan, 1994).
The percentages of residual fractions from La to Lu took on concave shapes, while other fractions took on convex shapes (Fig. 4), suggesting that middle REEs are more easily leached than other REEs. Many mechanisms have been presented to interpret the phenomenon of middle REEs enrichment in the non-residual fractions. Gosselin et al. (1992) proposed one potential fractionation mechanism as solideliquid exchange, which includes primary ion exchange or adsorption/desorption on mineral surfaces. Wang et al. (1994) interpreted middle REE enrichment in the nonresidual fractions as being caused by the stability of the partially filled 4f electron shells in the REEs. The middle REE enrichment in the acid soluble and reducible fractions might be explained as a result of the middle REE enrichment in carbonate and FeeMn coatings (Palmer, 1985). The middle REE enrichment in the oxidizable fraction may be explained by the complexing ability of REEs with several kinds of carboxylic acids which roughly follow convex shapes (Zhang et al., 1998).
Correlation analysis
Correlation matrix for REE parameters, grain sizes and TOC are presented in Table 2. In riverine samples, LREEs, HREEs, and SREEs are positively related to clay and negatively related to silt, sand and TOC. While in marine samples, they are positively related to silt and negatively related to sand. LREEs/HREEs, (La/Gd)/ P and (La/Yb)/ P are negatively related to clay and positively related to sand in riverine samples, whereas they are positively related to sand and negatively related to TOC in marine samples. dCe in riverine samples is positively related to clay and negatively related to sand, while in marine samples it is positively related to silt and TOC and negatively related to sand. dEu is positively related to TOC in riverine samples.
Both reducible and oxidizable REEs in riverine samples are positively related to clay and negatively related to sand, whereas residual REEs are negatively related to clay and positively related to sand. They present similar characteristics in marine samples. TOC in riverine samples is slightly related to oxidizable REEs in a positive way but in marine samples weakly related to residual REEs.
These results suggested that both grain size and TOC are key determinants to riverine and marine samples. Clay fraction is a more important carrier of REEs in riverine samples compared to that in marine samples. Since REE are strongly associated with organic matter, the negative relationship between LREE, HREE and REE and TOC in riverine samples may be attributed to the removal of REE-organic colloids, whereas the loss of statistic relationship between them in marine region may be due to coagulation of REEcontaining colloids caused by pH and salinity change during estuarine mixing (Hoyle et al., 1984;Shlokovitz, 1995) and/or frequent anthropogenic disturbation such as dredging (Feng et al., 2010).
Conclusions
This study investigated the concentrations of REEs and their chemical forms in surface sediments along four river-to-sea transects from the coastal Bohai Bay. Residual fractions of all REEs occupied the majority of their concentrations, whereas the exchangeable fraction was bound to Fe/Mn oxides, organic matter and sulphides, and carbonate sequentially. Middle REESs are more easily leached than other REEs. LREEs were slightly depleted compared to middle REES and HREES. The correlation analysis revealed that grain size and TOC were two important factors for the geochemical distributions of REEs in surface sediments of the coastal Bohai Bay. REEs were more abundant in sediments with finer grains. REE compositions in the coastal Bohai Sea were consistent with the data from the Post-Archean Australian Shale and no obvious river-to-sea transportation was found, indicating that the Bohai Bay was not highly polluted by REEs. À0.56 b À0.43 c a P < 0.001. b 0.001 < P < 0.01. c 0.01 < P < 0.05. | 2019-04-23T13:33:26.760Z | 2015-10-05T00:00:00.000 | {
"year": 2015,
"sha1": "e03609722064109d3f6b12f461b2a87ef395216b",
"oa_license": "CCBYNCSA",
"oa_url": "http://ir.yic.ac.cn/bitstream/133337/9302/4/Rare%20earth%20elements%20in%20surface%20sediments%20of%20a%20marine%20coast%20under%20heavy%20anthropogenic%20influence%20The%20Bohai%20Bay,%20China.pdf",
"oa_status": "GREEN",
"pdf_src": "Adhoc",
"pdf_hash": "55def29e65e1c046b8c2e6f486468ffcc88e0bb7",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Geology"
]
} |
15203984 | pes2o/s2orc | v3-fos-license | CRISPR-Mediated Slamf1Δ/Δ Slamf5Δ/Δ Slamf6Δ/Δ Triple Gene Disruption Reveals NKT Cell Defects but Not T Follicular Helper Cell Defects
SAP (SH2D1A) is required intrinsically in CD4 T cells to generate germinal center responses and long-term humoral immunity. SAP binds to SLAM family receptors, including SLAM, CD84, and Ly108 to enhance cytokine secretion and sustained T cell:B cell adhesion, which both improve T follicular helper (Tfh) cell aid to germinal center (GC) B cells. To understand the overlapping roles of multiple SLAM family receptors in germinal center responses, Slamf1Δ/Δ Slamf5Δ/Δ Slamf6Δ/Δ triple gene disruption (Slamf1,5,6Δ/Δ) mice were generated using CRISPR-Cas9 gene editing to eliminate expression of SLAM (CD150), CD84, and Ly108, respectively. Gene targeting was highly efficient, with 6 of 6 alleles disrupted in 14 of 23 pups and the majority of alleles disrupted in the remaining pups. NKT cell differentiation in Slamf1,5,6Δ/Δ mice was defective, but not completely absent. The remaining NKT cells exhibited substantially increased 2B4 (SLAMF4) expression. Surprisingly, there were no overt defects in germinal center responses to acute viral infections or protein immunizations in Slamf1,5,6Δ/Δ mice, unlike Sh2d1a-/- mice. Similarly, in the context of a competitive environment, SLAM family receptor expressing GC Tfh cell, GC B cell, and plasma cell responses exhibited no advantages over Slamf1,5,6Δ/Δ cells.
Introduction
Germinal center (GC) responses are critical for the generation of high affinity antibodies (Abs), memory B cells, and long-lived plasma cells. T follicular helper (Tfh) cells are subset of CD4 T cells specialized in supporting and regulating GC responses [1,2]. Tfh cells provide help signals that support GC B cell survival, proliferation, and somatic hypermutation and promote plasma cell differentiation via receptors like CD40L and through secreted factors like IL-21 [3,4]. These cognate interactions between Tfh cells and GC B cells are regulated by SLAM family receptors and SLAM-associated protein (SAP, Sh2d1a). Therefore, understanding how these receptors modulate Tfh cells and regulate GCs can be useful for developing rational vaccine immunology as well as potential therapeutics for autoantibody-associated autoimmune diseases.
Defects in Sh2d1a in humans causes X-linked lymphoproliferative disease (XLP), a disorder with high mortality characterized by the inability to clear infections and dysregulated T cell responses [5]. SAP deficient humans and mice have impaired generation of germinal centers (GCs), isotype-switched memory B cells, and circulating antibodies [6][7][8][9][10]. These defects can be rescued in mice via transfer of SAP sufficient CD4 T cells, demonstrating the critical cell intrinsic role of SAP in CD4 T cells [7,11]. Therefore, understanding the role of SAP in CD4 T cells is critical to the mechanistic understanding of defective humoral responses in XLP patients.
SAP binds to SLAM family (SLAMf) receptors, a family of nine receptors selectively expressed on cell types of the hematopoietic lineage. CD4 T cells express the SAP-binding SLAMf receptors SLAM (SLAMF1), Ly9, CD84 (SLAMF5), and Ly108 (SLAMF6) [5] and these receptors regulate different T cell functions. All four receptors are homophilic ligands. Single SLAMf KO mice have modest, if any, defects in the magnitude of Tfh or GC responses [12][13][14][15], in stark contrast to the severe defects observed in SAP-deficient animals. Slamf1 -/-Tfh cells have significantly reduced IL-4 expression [15,16], but the overall magnitude of the Tfh and GC responses are unchanged. The generation of pseudo double knockout conditions was useful to examine phenotypes involving interacting cells and SLAMf redundancies. For example, examining interactions between CD84 KO CD4 T cells with Ly108 KO B cells led to the finding that CD84 and Ly108 both contributed to sustained T cell:B cell adhesion [13]. Similarly, it was shown that SLAM and Ly108 were the primary receptors involved in NKT cell development by using Ly108 KO NKT cell precursors with SLAM KO thymocytes expressing CD1d [17]. Likewise, there may be additional roles that SLAMf receptors play that are unknown due to redundancy. Furthermore, Ly108 transmits both positive and negative signals [12,18,19], additionally confounding the interpretation of single SLAMf KO mice. Slamf6 -/-Sh2d1a -/mice exhibit substantially rescued GC Tfh cells and germinal center responses, demonstrating that Ly108 transmits potent negative signals in the absence of SAP. Ly108 transmits positive signals in NKT cells [12], NK cells [20], and CD8 T cells [18,19], but this was not directly observable in CD4 T cells. Thus, generating multi-SLAMf receptor gene deficient mice is a useful way to gain a more comprehensive understanding of SLAMf receptor function. However, because the SLAMf genes are located adjacent to each other on chromosome 1 in a large cluster, it has been very challenging to make multi-SLAMf receptor knockouts and this has hindered research in this area. A Slamf1 -/-Slamf5 -/-Slamf6 -/-(Slamf1,5,6 -/-) mouse has recently been reported [21,22]; however that genetic modification involved deletion of 140,000 bp of the genome, which could have unpredictable effects on gene regulation of the many neighboring SLAMf genes. Therefore, we employed CRISPR-Cas9 gene editing strategies to generate multi-gene deficient mice to examine the roles of multiple SLAM family receptors.
Ethics statement
The mouse experiments were conducted in compliance with the La Jolla Institute for Allergy and Immunology Animal Care Committee (Office of Laboratory Animal Welfare assurance number A3779-01), who approved all animal care and protocols used (Protocol license number AP006-SC1-0612). The mouse care and use protocol adheres to the Public Health Service (PHS) Policy on the Humane Care and Use of Laboratory Animals (Department of Health and Human Services) and the Guide for the Care and Use of Laboratory Animals (eighth edition). Mice were anesthetized with isoflurane and euthanized with a carbon dioxide gas chamber followed by cervical dislocation.
Generation of SLAM Δ/Δ , CD84 Δ/Δ , and Ly108 Δ/Δ mice using CRISPR-Cas9 gene editing technology The in vitro molecular and cellular biology was performed by Applied Stem Cell, Inc. Guide RNAs were selected using optimized CRISPR design by the Feng Zhang lab (crispr.mit.edu). Guide RNAs were further selected based on the criteria that they target the second exon of each receptor, target all isoforms of each receptor, and be unique for the targeted sites with up to two base pair mismatches. Also, 5'G motifs [23] and 3' purines were preferred [24]. Oligos for each of the gRNAs were cloned into the gRNA expression vector pBT-U6-Cas9-2A-GFP (or pX330 hSpCas9 vector with 2a-EGFP from the Feng Zhang lab). To test the activity of each gRNA, the gRNA expressing vectors were transfected into mouse N2A cells and the Surveyor assay was performed according to the manufacturer's instructions. Linearized pBT-T7-Cas9 plasmid was used as the template for in vitro transcription (IVT) using mMESSAGE mMA-CHINE T7 ULTRA kit (Life Technologies). T7 promoter was added to each gRNA template by PCR, gel purified, and used as a template for IVT using MEGAshortscript T7 kit (Life Technologies). Cas9 mRNA and gRNAs were purified using MEGAclear kit (Life Technologies) and eluted in RNA elution buffer. To test the activity of Cas9 mRNA, Cas9 mRNA was translated into protein using 1-Step Human IVT kit (Thermo Scientific) per instructions. An in vitro cleavage assay showed >95% IVT Cas-9 activity. An injection mix of 50 ng/μl Cas9 mRNA, 50 ng/μl SLAM-gRNA, 50 ng/μl CD84-gRNA, and 50 ng/μl Ly108-gRNA was injected into 150-250 one-cell embryos from C57BL/6J mice by the UCSD Stem Cell Core. These embryos were implanted into C57BL/6J surrogate mothers, and pups were genotyped by DNA sequencing and phenotyping by flow cytometry. DNA sequences were analyzed using Sequencher and diagrammed using SnapGene.
For Bcl6 staining, cells were stained for surface markers, then fixed and permeabilized with fixation and permeabilization buffer (eBioscience) per manufacturer's protocol, and stained with anti-Bcl6 monoclonal antibody (K112-91, BD Biosciences). For CXCR5 staining, cells were stained with purified rat anti-mouse CXCR5 (2G8, BD Biosciences) in PBS + 0.5% BSA + 2% normal mouse serum + 2% FCS (FACS buffer) for 1 hour, followed by goat anti-rat (H +L, Jackson Immunoresearch) in FACS buffer for 30 minutes, followed by other surface stains with washes in between each step.
Enzyme-linked immunosorbent assay (ELISA)
ELISAs were performed as previously described [26]. Briefly, Maxisorp plates were coated with Vaccinia virus antigen or HIV envelope trimers (YU2-gp140-F) overnight at 4°C. The plates were blocked with PBS + 0.1% Tween 20 + 0.5% BSA, mouse serum was added to plates, and HRP-labelled anti-mouse IgG (fragment specific Fcγ) was added with washing steps using PBS + 0.1% Tween 20 in between each step. Colorimetric development was performed with the TMB substrate kit and stopped with H 2 SO 4 , followed by measurement of absorption at 450 nm. Analysis was performed to measure endpoint titers (0.1 OD above background) and Area under curve (AUC). AUC analysis better accounts for both the quantity and quality of the IgG, as it accounts for the shape of the curve. AUC total peak area above baseline calculations (Graphpad Prism 6.0) were done for each individual sample, log transformed.
Statistics
Prism 6 Software was used to plot geometric mean and geometric standard error of the mean for log-based graphs, or mean and standard error of the mean for linear-based graphs. Statistical analysis was performed using Mann-Whitney t-tests.
Results
Generation of Slamf1 Δ/Δ Slamf5 Δ/Δ Slamf6 Δ/Δ triple gene disruption (Slamf1,5,6 Δ/Δ ) mice using CRISPR-Cas9 technology To determine the role of multiple SLAM family (SLAMf) receptors in GC responses, Slamf1 Δ/Δ Slamf5 Δ/Δ Slamf6 Δ/Δ triple gene disruption (Slamf1,5,6 Δ/Δ ) mice were generated. Given the sequential location of these genes at the same gene locus, the rare occurrence of chromosomal crossover between them rendered it unfeasible to create triple knockout (TKO) mice by crossing single SLAMf receptor single knockout (KO) mice. Thus, CRISPR-Cas9 gene-editing technology was employed to knockout all three receptors. CRISPR-Cas9 uses guide RNAs (gRNAs) and the Cas9 DNA endonuclease to cut specific sequences in the genome [24,[27][28][29][30][31]. Unique 20 base pair sequences were designed to target each SLAMf receptor and gRNA sequences were chosen that had no off-target gene matches with up to two base pair mismatches ( Table 1). The efficiency of gene targeting using these gRNAs was 30-34 percent as quantified in vitro via Surveyor assay (Fig 1A). One-cell embryos were then injected with Slamf1, Slamf5, and Slamf6 gRNAs and Cas9 mRNA and embryos were implanted into surrogate mothers. This resulted in 23 viable pups, which were sequenced for mutations in Slamf1, Slamf5, and Slamf6 and/or phenotyped for the expression of the receptors SLAM, CD84, and Ly108 (Table 2). Error-prone DNA repair resulted in insertions, deletions, or a combination of the two, at the targeted cut site and this led to frameshifts and stop codons ( Fig 1B). Strikingly, null mutations were generated in all 6 gene alleles for 14 of 23 pups. The remaining 9 pups were then partially characterized, with 8 of 9 confirmed to be at least double knockout (DKO). We then established three separate breeder lines (Table 2; breeders: #4 and #21, #11 and #23, #14 and #22), which each gave identical immunological results (S1 Fig). In sum, CRISPR-Cas9 targeting of SLAM family genes in embryos was highly efficient. Table 2. doi:10.1371/journal.pone.0156074.g001 Slamf1 Δ/Δ Slamf5 Δ/Δ Slamf6 Δ/Δ triple gene disruption mice generated similar frequencies of peripheral T
cells and B cells but decreased frequencies of NKT cells
To determine if the SLAMf receptors impacted the generation of peripheral lymphocytes in Slamf1,5,6 Δ/Δ mice, peripheral CD4 T cells, CD8 T cells, B cells, and NKT cells were quantified. Surface expression of SLAM, CD84, and Ly108 was absent on CD4 T cells, CD8 T cells, and B cells in Slamf1,5,6 Δ/Δ mice compared to wild-type (WT) mice (Fig 2A and 2B). Residual signal by the anti-CD84 mAb binding B cells suggests a crossreactivity of the mAb for another antigen on B cells, as no signal was apparent on CD4 T cells or CD8 T cells, and homozygous Slamf5 stop mutations were confirmed in the genome (S2 Fig). Frequencies of peripheral CD4 T cells, CD8 T cells, and B cells were similar (Fig 2C and 2D), suggesting that SLAM, CD84, and Ly108 are not required for the development of peripheral T and B cells. In contrast, peripheral NKT cells were reduced approximately 3-fold in liver and 2.5-fold in spleen of Slamf1,5,6 Δ/Δ mice compared to WT (p = 0.0002) (Fig 2E and 2F). Surprisingly, this NKT phenotype was less severe than that seen in SAP-deficient mice [32,33] and moderately more severe than Slamf6 -/-(Ly108-deficient) mice [12,17]. Previously, it was thought that SLAM and Ly108 were the primary SLAMf receptors involved in NKT cell development [17]. The data here, and from the Slamf1,5,6 -/mice [21], suggest that additional SLAMf receptors, or SAP binding receptors other than SLAM, CD84, and Ly108, contribute to NKT cell development.
Absence of Tfh or germinal center defects in Slamf1,5,6 Δ/Δ mice SAP expression in CD4 T cells is necessary for GC responses [7,35]. Because SAP binds to SLAM, CD84, and Ly108 on CD4 T cells, the impact of these receptors on GC responses was investigated. GC responses were measured in WT and Slamf1,5,6 Δ/Δ mice at 8 days post lymphocytic choriomeningitis virus (LCMV) infection. Frequencies of T follicular helper (Tfh) cells, GC Tfh cells, GC B cells and plasma cells (PCs) were comparable between WT and Slamf1,5,6 Δ/Δ mice (Fig 4A and 4B). To determine if this was a representative result, vaccinia virus (VACV) infection was used as a second acute viral infection model. Overall Tfh, GC Tfh, GC B cells, and plasma cell responses to VACV were comparable between WT and Slamf1,5,6 Δ/Δ mice (Fig 4C and 4D). We also found similar VACV-specific IgG titers in the serum of WT and Slamf1,5,6 Δ/Δ mice (Fig 4E), consistent with the comparable plasma cell responses seen in the spleen. This is in contrast to a report using Slamf1,4,5 -/genomic locus deletion mice, where a slight increase in antibody responses was observed in the context of NP hapten immunization [22].
We considered the possibility that GC defects in Slamf1,5,6 Δ/Δ mice may be compensated by other pathways in the context of inflammatory acute viral infections, even though no compensation is seen in SAP-deficient mice. We therefore immunized mice with HIV envelope (Env) protein (YU2 gp140-F) in mild adjuvant as an independent model. Again, no differences were observed in Tfh, GC Tfh, or GC B cell responses between WT and Slamf1,5,6 Δ/Δ mice (Fig 5A and 5B). Additionally, there were no differences in HIV Env-specific IgG titers in serum ( Fig 5C). Thus, in Slamf1,5,6 Δ/Δ mice, where all cells are deficient for SLAM, CD84, and Ly108 expression, no quantitative defects in Tfh or GC responses were observed in three independent models.
The roles of many factors can only be observed in vivo in the context of a competitive environment between WT and deficient cells. Therefore, bone marrow (BM) chimeras of B6.SJL (CD45.1 +/+ ) reconstituted with WT (CD45.2 +/+ ) and Slamf1,5,6 Δ/Δ CD45.2 +/+ BM cells were tested for T and B cell responses after LCMV infection (Fig 6A). At 7/8 days post LCMV (Fig 6B-6E). There were no significant differences between the frequencies of WT and Slamf1,5,6 Δ/Δ cells (Fig 6F-6I). A second set of BM chimera experiments were performed with VACV infections. When Tfh cell, GC Tfh cell, and GC B cell responses were measured at 7 days post VACV infection (Fig 7A-7C), there were also no significant differences between the frequencies of WT and Slamf1,5,6 Δ/Δ cells (Fig 7D-7F). Thus, expression of SLAM, CD84, and Ly108 was not required for the initiation and maintenance of germinal center responses.
Discussion
SLAMf receptors and SAP are required for NKT cell development, with the primary receptors thought to be SLAM and Ly108 [17,21]. In Slamf1,5,6 Δ/Δ mice, NKT cell development is defective, but not completely abolished as seen in SAP-deficient mice [32,33]. This partial defect in Slamf1,5,6 Δ/Δ mice, which confirms a recent report using Slamf1,5,6 -/genomic locus deletion mice [21], suggests that other receptors are involved in NKT cell development. We then proceeded to further characterize the NKT cells that develop in Slamf1,5,6 Δ/Δ mice. We show a functional defect in NKT cells that develop and moreover, we find that 39% of splenic NKT cells in Slamf1,5,6 Δ/Δ mice aberrantly express 2B4/SLAMF4. We therefore conclude that 2B4 expression during development may largely compensate for loss of SLAM and Ly108 for NKT cell development, and that the 2B4 expression is then maintained in mature Slamf1,5,6 Δ/Δ NKT cells.
SLAMf receptors have been shown to play roles in CD4 T cell function, including cytokine secretion and sustained interactions with B cells. Thus, SLAMf receptors impact Tfh cell help to GC B cells, and in turn, GC B cells help maintain Tfh cells in GCs [2]. SAP is necessary for GC responses and long-term humoral immunity, and SAP expression in CD4 T cells aids in the recruitment and retention of GC Tfh cells within GCs [36]. For these reasons, we expected that multiple gene disruption in SLAM, CD84, and Ly108 would lead to defects in GCs. Therefore, the lack of overall defects in GC responses after viral infection and protein immunization even in the context of competition was unexpected. A recent report using Slamf1,5,6 -/mice suggested that SLAM, CD84, and Ly108 negatively regulated antibody responses by recruiting SHP-1 to the B cell receptor (BCR) complex [22]. SHP-1 recruitment to the BCR is required for the maintenance of GCs [37], but it is not known how SLAM family receptors are involved in this regulation. Wang et al. show an approximately 2-fold increase in antigen specific IgG titers and plasma cell frequencies in Slamf1,5,6 -/mice. In contrast, we observed no enhancement of plasma cell or Ab responses in the Slamf1,5,6 Δ/Δ mice in three independent systems. Furthermore, Wang et al. show enhanced antibody titers from Slamf1,5,6 -/-B cells transferred into Rag-1 -/mice whereas we show similar plasma cell responses in WT and Slamf1,5,6 Δ/Δ bone marrow chimeras. Differences in these results may relate to differences in experimental systems, with some systems more capable of revealing subtle differences than others. Yet, the use of alternatively generated SLAMf receptor deficient mice and different experimental systems is beneficial to our understanding of the roles of these complex receptors in the immune response. Both Terhorst and colleagues and our group found no overt differences in Tfh cells and GC B cell responses. This lack of phenotype in Slamf1,5,6 Δ/Δ mice did not correspond with CRISPR-Mediated Slamf1 Slamf5 Slamf6 Triple Gene Disruption the severe GC defects seen SAP -/mice. This suggests that the SAP -/defect is almost exclusively caused by potent negative signaling via Ly108 in the absence of SAP [12]. Our data also suggests that SLAM, CD84, and Ly108 have functionally redundant roles with other receptors involved in GC responses. Additional receptor mediated interactions impact the recruitment and maintenance of GC Tfh cells, such as chemokine receptors, integrins, and ICOS:ICOSL interactions between Tfh cells and B cells [38]. Thus, SLAMf receptor expression impacts GC responses likely by enhancing T:B cell interactions. Meanwhile, other receptors may perform partially redundant functions in T:B interactions that help orchestrate the complex interactions occurring in GCs. Lastly, because SLAMf receptors have the potential to signal both positively and negatively through immunotyrosine switch motifs (ITSMs) [5], the interpretation of these SLAMf receptor TKO phenotypes is complicated by impacts on functions downstream of both positive and negative signaling. Further exploration using multiple SLAMf receptor deficient mice to understand these phenotypes and the mechanisms underlying these phenotypes would be worthwhile. (Table 2), which were confirmed genotypically as Slamf1,5,6 Δ/Δ mice. CD84 expression on CD4 T cells and B cells from WT mice (blue) are shown as a comparison to CD84 expression in Slamf1,5,6 Δ/Δ mice (dashed gray). (EPS) | 2018-04-03T01:40:35.560Z | 2016-05-25T00:00:00.000 | {
"year": 2016,
"sha1": "a2b4bef8dd623b81daa7c37921df9b873df5ca7a",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0156074&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a2b4bef8dd623b81daa7c37921df9b873df5ca7a",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
4508278 | pes2o/s2orc | v3-fos-license | Sex matters in Massive Parallel Sequencing: Evidence for biases in genetic parameter estimation and investigation of sex determination systems
Using massively parallel sequencing data from two species with different life history traits -- American lobster (Homarus americanus) and Arctic Char (Salvelinus alpinus) -- we highlighted how an unbalanced sex ratio in the samples combined with a few sex-linked markers may lead to false interpretations of population structure and thus to potentially erroneous management recommendations. Multivariate analyses revealed two genetic clusters that separated males and females instead of showing the expected pattern of genetic differentiation among ecologically divergent (inshore vs. offshore in lobster) or geographically distant (east vs. west in Arctic Char) sampling locations. We created several subsamples artificially varying the sex ratio in the inshore/offshore and east/west groups, and then demonstrated that significant genetic differentiation could be observed despite panmixia for lobster, and that Fst values were overestimated for Arctic Char. This pattern was due to 12 and 94 sex-linked markers driving differentiation for lobster and Arctic Char, respectively. Removing sex-linked markers led to nonsignificant genetic structure (lobster) and a more accurate estimation of Fst (Arctic Char). We further characterized the putative functions of sex-linked markers. Given that only 9.6% of all marine/diadromous population genomic studies to date reported sex information, we urge researchers to collect and consider individual sex information. In summary, we argue that sex information is useful to (i) control sex ratio in sampling, (ii) overcome “sex-ratio bias” that can lead to spurious genetic differentiation signals and (iii) fill knowledge gaps regarding sex determining systems.
Introduction
Recently, the revolution in massively parallel sequencing (MPS) technology has led to the production of many genome-wide datasets, whereby thousands of markers can be easily and inexpensively genotyped in hundreds of individuals for both model and nonmodel species (Davey et al. 2011;Andrews et al. 2016).Several MPS studies based on either RAD-sequencing or Genotype-By-Sequencing (GBS) techniques have demonstrated that these markers bring unprecedented insights on the causes and consequences of population structuring (reviewed in Narum et al. 2013).The strength of such methods comes from its supposedly random sampling of the entire genome (Davey et al. 2013).While the random distribution of markers achieved by these methods is advantageous in many regards, it has one over-looked result that could have consequences for inferences of population structure: some of the markers identified will be located on sex chromosomes, or in regions linked to sex, in species with genetic sex determination.Indeed, Wright (1931) pointed out this bias in genetic parameter estimations, particularly when sampling populations with varying sex ratios or in the presence of sex-biased dispersal.Despite the potential importance of these biases, few MPS studies have focused on the analysis of sex-linked markers (but see Gamble & Zarkower 2014;Kafkas et al. 2015;Brelsford et al. 2016;Larson et al. 2016) and to our knowledge, none have investigated the influence of sex-linked markers on inferences of population structure observed.
In addition to the importance of avoiding potential biases, detecting sex-linked markers in MPS datasets can also provide valuable information on sex determination (Pan et al. 2016).Sex is common to almost all living animals and often leads to the evolution of male and female dimorphism, both at the genetic and phenotypic level (Bell 1982).Diverse mechanisms acting at the scale of the genome, chromosomes or cells underlie the morphological, physiological and behavioral differences between males and females.Moreover, sex determination systems vary tremendously among and within taxa (Bachtrog et al. 2014), highlighting the challenges in determining the selective forces driving sex determination.In general, the diversity of sex determination systems reported in fish (particularly teleosts) and crustaceans is much more pronounced than that observed in mammals and birds (Bachtrog et al. 2014).Yet, the characterization of the genetic architecture of sex determination in these taxonomic groups has been limited to a few studies (Legrand et al. 1987).The access to new genomic approaches, which are increasingly being used in non-model marine and aquatic organisms (Kelley et al. 2016), offers new prospects to investigate the molecular basis of sex determination in this diverse group.
The identification of sex-linked markers can also provide a wealth of other useful information for management, conservation, and aquaculture (Pan et al. 2016).First, sexlinked markers can assist in the identification of the sex of an individual, particularly in cases with an absence of clear sexual dimorphism (e.g., at young life history stages).In aquaculture practices, this can help farmers to maintain equal sex ratios of breeding adults and to implement efficient breeding programs (Martínez et al. 2014).Second, sex information is often important to include as a covariate in genetic models for finding loci linked to specific traits in order to reduce residual variation (Broman & Sen 2009).Third, knowing the sex of individuals may facilitate the demonstration of sex-biased dispersal, i.e., when individuals of one sex are more prone to disperse (Prugnolle & De Meeûs 2002).Sex-biased dispersal is widely spread among vertebrates and can have important ecological and evolutionary consequences, but there is still little research on this topic in marine organisms, such as fishes and crustaceans, compared to mammals and birds (Mossman & Waser 1999).
Here, we present two empirical examples that illustrate how an unbalanced sex ratio combined with a few sex-linked markers can lead to false interpretations of population structure and to erroneous management recommendations, especially in species with high connectivity as frequently observed in marine and diadromous organisms.Our initial goal was to separately investigate population structure between two groups of American lobsters (Homarus americanus) occupying inshore and offshore habitats, and between Arctic Char (Salvenius alpinus) collected from two geographically separated regions (east and west) in the Canadian Arctic.In both cases, preliminary multivariate analyses mainly revealed two genetic clusters corresponding to male and female individuals instead of being related to inshore/offshore groups of lobsters or to east/west groups of Arctic Char.To further understand the clustering, we identified sexlinked markers driving the genetic differentiation between male and female in American lobster and Arctic Char.To demonstrate the potential impacts of sex-linked markers on the population genetic analysis, we tested for both species how different numbers of sexlinked markers and ratios of samples from each sex can cause biased inferences of population structure.Finally, using the set of sex-linked markers identified, we found potential candidate genes or chromosomal regions linked to sex for American lobster and Arctic Char.We conclude with an exhaustive literature search demonstrating that very few studies performed on marine and diadromous species report sex information, and we argue, in light of our findings, that collecting this information can be critical to avoid biases, especially in high gene flow species.
Sampling and molecular techniques
American lobster: Commercial fishers collected 203 American lobsters (100 males and 103 females) from 13 sites including eight inshore sites and five offshore sites along the Atlantic coast of North America (Figure 1A; Table S1).The sex of all specimens was determined visually from obvious external morphological differences.Genomic DNA was then extracted using Qiagen Blood and Tissue kits.DNA quality was confirmed using visual inspection on 1% agarose gel followed by quantification with Quantit Picogreen dsDNA assay kits.RAD-sequencing libraries were prepared following the protocol from Benestan et al. (2015).Each individual was barcoded with a unique sixnucleotide sequence and 48 individuals were pooled per library.Real-time PCR was used to quantify the libraries.Single-end, 100 bp sequencing was performed on an Illumina HiSeq2000 platform at the Genome Québec Innovation Centre (McGill University, Montréal, Canada).
Arctic Char: Samples of 290 adult anadromous Arctic Char (142 males and 148 females) were collected from six rivers located on southern Victoria Island, Nunavut, Canada (Figure 1B; Table S2).Sex was determined visually by observation of the gonads for a subset (n = 174) and based on a genetic assay for another subset (n = 116), as described in Moore et al. (2016).In brief, the genetic sex was inferred based on the PCR assay described in Yano et al. (2013).Six individuals of known sex (three males and three females) were used as controls.Genomic DNA was extracted using a salt-extraction protocol modified from Aljanabi and Martinez (1997).DNA quality and quantity were checked on 1% agarose gels and using PicoGreen assays (Fluoroskan Ascent FL, Thermo Labsystems), respectively.Libraries were prepared based on a GBS protocol modified from Mascher et al. (2013)
Bioinformatics and genotyping
Both the American lobster and Arctic Char libraries were de-multiplexed using process_radtags in STACKS (v.1.29 for American lobster and v.1.40for Arctic Char) (Catchen et al. 2013).Raw sequencing data was checked in FASTQC (Andrews 2015).
Reads were truncated to 80 bp for lobster and 70 bp for Arctic Char and adapter sequences were removed with CUTADAPT (Martin 2011).
American lobster: Loci were identified allowing a maximum of three nucleotide mismatches (M = 3), according to Ilut et al. (2014) and a minimum stack depth of three (m = 3), among reads with potentially variable sequences (ustacks module in stacks, with default parameters).Then, reads were clustered de novo to create a catalogue of putative RAD tags (cstacks module in STACKS, with default parameters).In the populations module of STACKS v.1.29 and following consecutive filtering steps, SNPs were retained when they were genotyped in at least 80% of the individuals and found in at least 9 of the 12 sampling sites.Potential paralogs were excluded by removing markers showing heterozygosity > 0.50 and 0.30 < F IS < -0.30 within sites.Only SNPs with a global minor allele frequency > 0.02 were retained for the analysis.The resulting filtered VCF files were converted into the file formats necessary for the following analyses using PGDspider v.2.0.5.0 (Lischer & Excoffier 2012).
Arctic Char: SNPs were identified by first mapping the reads to the genome of the closely related Rainbow Trout (Oncorhynchus mykiss; Berthelot et al. 2014) using GSNAP v2016-06-09 with a minimum of 90% read coverage (-min-cov 90), tolerating 2 mismatches (-m 2) and setting an indel penalty to 2 (-i 2).A subsequent trimming step was conducted with SAMtools v1.2 (Li et al., 2009) to remove unmapped and multimapped reads using flags -F 1797 and -F 4, and a minimum mapping quality (MAPQ) of 1, respectively.The binary alignment files (bam) were then used as input for downstream analysis.Genotypes were obtained using STACKS v.1.40integrated in a workflow developed in our laboratory (Benestan et al. 2016a).The catalog of loci was created allowing no mismatches among loci in cstacks (n=0) and a minimum stack depth of four (-m 4).SNPs were retained if at least 50% of the individuals were genotyped for the marker (-r 0.5) and the locus was present in at least four populations (-p 4).Potential paralogs were excluded by removing markers showing heterozygosity > 0.60 and F IS < -0.40 F IS > 0.40 within samples.Only SNPs with a global minor allele frequency > 0.01 were retained for the analysis.
Discriminant Analysis of Principal Component (DAPC)
For American lobster, Discriminant Analysis of Principal Components (DAPC) was performed in the R package adegenet (Jombart et al. 2010).The optimal number of discriminant functions (n=60) to retain was evaluated according to the optimal α-score obtained from the data (Jombart et al. 2010).For Arctic Char, a Principal Component Analysis was performed in adegenet.As population differentiation was pronounced enough to be observed with the PCA, a DAPC was not conducted.
Sex outlier loci detection
American lobster and Arctic Char: Outlier loci corresponding to the most divergent markers between sexes were identified with a level of differentiation between sexes exceeding random expectations using F st -based outlier analyses.Outlier SNPs were detected with BAYESCAN v. 2.1 (Foll & Gaggiotti 2008).BAYESCAN runs were implemented using permissive prior model (pr_odds) of 10, including a total of 10,000 iterations and a burn-in of 200,000 steps.For both species, these outlier analyses were conducted on the entire data set separated by sex.
Sex ratio and sex-linked marker influence on index of genetic differentiation (F st )
To determine the extent to which differing sex ratio influences the detected genetic structure, different proportions of male and female American lobsters or Arctic Char were subsampled from inshore or east and offshore or west, respectively, keeping a total of 50 individuals per group.This generated a gradient of six different sex ratio datasets, representing different sampling bias scenarios, from the most balanced (sex ratio = 25:25/25:25) to the most unbalanced sex ratio (sex ratio = 0:50/50:0).
Considering the three most unbalanced sex-ratio datasets (i.e., 0:50/50:0, 5:45/45:5, 10:40/40:10), we removed sex-linked markers (i.e., here outlier SNPs) according to their F st values (in descending order) and we estimated F st between offshore/inshore for the American lobster and east/west for the Arctic Char.We calculated F st values using the function fst_WC84 in assigner R package (Gosselin et al. 2016).
Marker annotation and genomic position
American lobster: There is no reference genome or high-density linkage map available for American lobster and so the approximate locations or associated linkage groups of the sex-linked SNPs could not be determined.Probable proximity between markers was determined by linkage disequilibrium (LD) analysis by calculating LD between pairs of SNPs using the geno-r2 command available in VCFTOOLS (Danecek et al. 2011).The LD data frame obtained with VCFTOOLS was then transformed into an LD matrix to be analyzed using the heatmap command in the R environment (Team 2013).In order to determine what genes are associated with these sex-linked markers, the 12 candidate SNPs (outliers identified by BAYESCAN) were queried using BLAST against the transcriptome of the American lobster (F.Clark and S. Greenwood, University of Prince Edward Island, personal communication; see details in Benestan et al. 2016b).Six of the 12 candidate SNPs were distributed among six different contigs in the transcriptome data.
The associated contigs were used as queries in a BLAST search against the SWISS-PROT database (Bairoch & Apweiler 2000).A minimal E-value threshold of 1 x 10 -6 and percent similarity of at least 70% were used.This yielded a set of two candidate SNPs associated with known genes.Gene ontology (GO) annotation terms were then associated to the candidate SNPs using SWISS-PROT accessions.
Arctic Char: There is no reference genome available yet for Arctic Char, but there is a high-density linkage map available for the closely related Brook Char (Sutherland et al. 2016).To obtain approximate positions of the sex-linked SNPs from Arctic Char, the MapComp method (Sutherland et al. 2016) was used to pair all of the Arctic Char markers with mapped Brook Char markers using the Atlantic Salmon genome (Lien et al. 2016;GenBank: GCA_000233375.4) as the intermediate reference genome.This method connects markers from two different linkage maps by mapping the markers to a reference genome, then pairing markers that map uniquely to the same place or close to each other in the reference genome.This was done as previously described (Sutherland et al. 2016), but with ten iterations to permit more than one anonymous marker pairing with a single mapped marker, as previously described (Narum et al. in review) but with a 1 Mbp maximum distance between the paired markers on a reference genome.This yielded approximate positions for determining the number and identity of linkage groups associated with sex in Arctic Char.To determine which genes are associated with these linked markers, the sex-linked markers were used in a BLAST query against the annotated Atlantic Salmon genome (Lien et al. 2016); NCBI Genome ICSASG_v2 reference Annotation Release 100).
Literature search for marine and diadromous species population genomic studies
We performed an exhaustive literature search to document the proportion of population genomics studies that have reported sexing the species analyzed.More specifically, we conducted a literature search of population genomics studies on marine/diadromous species published in peer-reviewed journals from January 2010 to 15 November 2016 using the ISI Web of Knowledge bibliographic database (Thomson Reuters, http://thomsonreuters.com)using search keywords (i) "genomics" AND "marine" AND "SNP" yielded 22 hits, (ii) "population structure" AND "marine" AND "SNP" yielded 47 hits, (iii) "RAD-sequencing" AND "marine" yielded 39 hits and (iv) "population genomics" AND "marine" yielded 243 hits, (v) "population genomics" AND "anadromous" OR "catadromous" yielded 11 hits.From these hits, several criteria were used to determine which studies to include in our analyses.First, the paper needed to focus on a marine animal and use a set of more than 1,000 SNP markers.Second, the paper needed to refer to population genomics or related areas such as outlier identification because these are the target areas of research likely to be influenced by the sex ratio bias in sampling.After removing studies on non-marine or non-animal organisms, or those with too low density of markers, a total of 38 and 14 publications were retained for marine and diadromous species, respectively (listed in Table 1 and 2).
Artefactual population structure caused by sex-linked markers
For American lobster, using 1,717 filtered SNPs, Discriminant Analysis of Principal Components (DAPC) was performed on the 203 individuals successfully genotyped to investigate the extent of population structuring between offshore and inshore locations.
Instead of finding significant genetic differences between inshore and offshore samples, the first axis of the DAPC highlighted a significant genetic differentiation between sexes (F st = 0.0057, P-value = 0.0009), explaining 16.04% of the total genetic variation (Figure 2A).
For Arctic Char, using 6,147 filtered SNPs a principal components analysis (PCA) of genotypes from 290 individuals identified strong clustering that explained 5.74% of the total genetic variation between two groups not corresponding to any particular geographic region (Figure 2C).By using the data on phenotypic and genetic sex, it was clear that samples mainly clustered by sex in this PCA (Figure 2C) and that this genetic differentiation was modest but significant (F st = 0.0132, P-value = 0.0002).
Delineating the influence of sex ratio on F st in panmictic or anadromous species
A DAPC and a PCA were run on datasets containing only males for offshore or east region and only females for inshore or west locations for American lobster and Arctic Char, respectively (Figure 2B,D).As expected, the DAPC for American lobster showed a highly significant signal of genetic differentiation between inshore and offshore samples with a F st value in the range typically seen in many marine species (F st = 0.0056, 95% CI inf = 0.0027 and CI sup = 0.0088, P-value < 0.05), which in reality resulted from the extremely skewed sex ratio of this artificial dataset (Figure 2B,D).This outcome contrasts with the panmictic structure observed between inshore and offshore (F st = 0.0001, CI inf = -0.0004and CI sup = 0.0006, P-value > 0.05) when sex ratio is balanced (sex ratio in the original dataset is equal to 25:25/25:25).As expected, F st between inshore and offshore was highest and most significant when sex ratio was completely unbalanced, i.e., sex ratio equal to 0 (F st = 0.0055, CI inf = 0.0030 and CI sup = 0.0092, Pvalue < 0.05).F st remained significantly elevated until the sex ratio was 15:35/35:15 (F st < 0.001, CI inf < 0; Figure 3A).Following the same method described above to simulate differing sex ratio datasets, F st between east and west Arctic Char locations was highest and most significant when sex ratio was completely unbalanced, i.e., sex ratio equal to 0:50/50:0 (F st = 0.0215, CI inf = 0.0194 and CI sup = 0.0242, P-value < 0.05).F st then gradually decreased with increasingly even sex ratios until it reached F st = 0.0064 (CI inf = 0.0055 and CI sup = 0.0072; P-value < 0.05) with a sex ratio of 25:25/25:25 (Figure 3B).(A) American lobster.F st between offshore and inshore according to sex ratio proportion when subsampling 100 individuals with a sex ratio ranging from a complete unbalanced sex ratio (i.e., sex ratio equal to 0:50/50:0) to a perfectly balanced sex ratio (i.e., sex ratio equal to 25:25/25:25).The horizontal black dashed line indicates the threshold below which F st values are no longer significant at P < 0.05.(B) Arctic Char.F st between east and west according to the sex ratio proportion when subsampling 100 individuals with a sex ratio ranging from a complete unbalanced sex ratio (i.e., sex ratio equal to 0:50/50:0) to a perfectly balanced sex ratio (i.e., sex ratio equal to 25:25/25:25).F st was still significant for the anadromous, but was overestimated in the skewed sex ratio cases.In both panels, the vertical limits of the box represent one standard deviation around the mean (n = 10 individual subsample iterations), the horizontal line within the box is the median, and the whiskers extend from the box to the 25 th and 75 th percentiles.
Sex-linked markers in genome-wide datasets
Identifying sex-linked markers in American lobster and Arctic Char Out of the 1,717 SNPs initially considered for the American lobster, BAYESCAN identified 12 highly differentiated markers between the sexes (Figure S1).These 12 markers have a BAYESCAN F st of 0.0800 on average between the sexes (range = 0.1567-0.1167)whereas the remaining 1,705 SNPs have on average a F st of 0.0030 on average (range = 0.0032-0.0101).
Out of the 6,147 markers initially considered for Arctic Char, BAYESCAN identified 94 markers contributing to the male/female separation (Figure S1).These 94 markers show a BAYESCAN F st of 0.0421 between the sexes (range = 0.0039-0.1140)whereas the remaining 6,053 markers were on average 0.0019 between the sexes (range = 0.0019-0.0036).
Delineating the influence of sex ratio on F st in panmictic or anadromous species
We investigated the influence of the number of these 12 and 94 sex-linked markers on the index of genetic differentiation (F st ) calculated between inshore/offshore or east/west for both species, where sex ratio in sampling was unbalanced at different degrees (0:50/50:0, 5:45/45:5, 10:40/40:10).For American lobster, we observed high and significant F st values when no sex-linked marker was removed for the three scenarios.Then, F st progressively decreased with the removal of sex-linked markers (in descending order regarding their F st values) until reaching a small and non-significant value when we removed at least 11 out of 12 sex-linked markers for the most extreme scenario (0:50/50:0; Figure 4A).For Arctic Char, F st progressively decreased from 0.0215 to 0.0064 on average, considering all scenarios, which suggest that F st is more than two-fold smaller when sex-linked markers are removed from the data set (Figure 4B).This decrease reached a plateau when 80 sex-linked markers were removed, which corresponds to almost the totality (n = 94) of the sex-linked markers found.1).The dashed line in black indicates the threshold below which F st values are no longer significant at P < 0.05.Sex ratio of 0.4 and 0.5 were not included in this analysis because F st values were not significant in these cases (see Figure 4A).(B) Artic Char.The line graph displays the influence of sex-linked markers on (as a function of the number of sexlinked markers removed from the analysis considering three sampling scenario with different degrees of sex ratio bias (0:50/50:0, 5:45:45:5, 10:40/40:10).Sex-linked markers are removed in descending order according to their F st values.
Characterizing sex-linked markers in American lobster
Linkage disequilibrium (LD) calculations for the 12 sex-linked markers in American lobster revealed two clusters of markers in high LD (Figure S2).One of the clusters includes seven markers with the strongest genetic differentiation between the sexes (F st > 0.40; Table 3).Six of these markers displayed heterozygosity excess in males (H O = 0.49, H O ranging from 0.16 to 0.63) and heterozygosity deficit in females (H O < 0.02; H O ranging from 0.00 to 0.29), thus providing evidence for a male heterogametic system.
The identities of genes nearby the sex-linked SNPs in lobster were further explored in the six contigs containing the six sex-linked SNP markers, which were located in sequences that had a significant match (more than 90% of nucleotide identity) in the American lobster transcriptome.The polymorphisms associated with two of these sequences both occurred in the 3'UTR region of the genes annotated by SWISSPROT database.These genes were sulfotransferase family cytosolic 1B member 1 (hereafter SULT1B1) and pre-mRNA-splicing factor cwf19 (hereafter cwf19), and are involved in steroid metabolism and mRNA splicing, respectively.Both genes that were previously reported to influence sex determination in fishes (Devlin & Nagahama 2002), namely in European Eel (Anguilla anguilla; Churcher et al. 2015) and Greenland Halibut (Scophthalmus maximus; Ribas et al. 2015a).
Using BLAST to align the 94 sex-linked markers against the Atlantic Salmon (Salmo salar) reference genome (Lien et al. 2016;GenBank GCA_000233375.4) consistently identified the Atlantic Salmon chromosomes homologous to the Brook Char chromosomes that were assigned using iterative MapComp.An additional nine of the 49 non-positioned markers aligned against the Atlantic Salmon chromosomes corresponding to the four highly sex-linked chromosomes, Ssa01, Ssa10 and Ssa09 (Ssa09 corresponds to a fused metacentric chromosome that corresponds to BC35 and BC38; Sutherland et al. 2016).Four non-positioned markers were assigned to chromosomes not identified as the four highly sex-linked chromosomes.Often the markers that had not received positions with iterative MapComp either did not have significant alignments or had many equal alignments in the Atlantic Salmon genome.Using BLAST against the annotated Atlantic Salmon genome, 28 of the 94 markers were found within a gene.For the remaining markers not found in a gene but with significant alignments, the closest upstream and downstream genes were identified along with the distance from the marker to the gene.Two sex-linked markers positioned on the Brook Char sex chromosome (BC35), SNP 86986 and SNP 87087, were on either side of transcription factor SOX-11-like, a member of the SRY-related HMG-box gene family associated to sex determination (Graves 1998;Woram et al. 2003).This gene was the closest annotated gene to these markers in the downstream or upstream direction, respectively, although the distance was large (~280 kb in each direction).Other identified genes containing sex-linked markers are involved in chromosome segregation and recombination (e.g., nipped-B-like protein, nuclear pore complex protein Nup93, bloom syndrome protein and centrosomal protein of 164 kDa), putative sex-specific activities (e.g., talin-2), or transcription factor activity (e.g., retinoic acid receptor RXR-alpha).
Discussion
Sex-ratio bias in genotyping-by-sequencing studies Sex-linked markers are expected to be present in all massively parallel sequencing genomic datasets developed on species with a genetic basis for sex determination (Gamble & Zarkower 2014).Despite the ubiquity of these sex-linked markers across taxa, very few population genomic studies on marine or diadromous species have reported information on the sex of samples being analyzed (see details below).However, our results clearly demonstrate that the occurrence of such markers jointly with an unbalanced sex ratio in sampling can lead to the observation of a spurious or biased population structure.This, in turn, may result in misinterpreting the biology of the species being investigated and possibly leading to improper management recommendations.For instance, in the case of the lobster study here, with an unbalanced sex ratio this could have led to the conclusion that inshore and offshore lobsters comprise two genetically distinct stocks (and therefore distinct management units) while in reality they comprise a single panmictic unit.This bias is particularly critical for high gene flow species characterized by very weak population structuring, which is typical of many marine and diadromous species alike.In such cases, only a few highly differentiated markers (here 0.7% and 1.5% of the total filtered markers for American lobster and Arctic Char, respectively) can generate a signal of significant genetic differentiation or inflate the signal in the cases of panmictic or low population differentiation, respectively.These outcomes highlight the importance of collecting sex information of individual samples to draw accurate conclusions about population structure of non-model species using genome-wide data sets.
Moreover, sex ratio is obviously an important characteristic of a population and is tightly linked to its dynamics.Therefore, gaining this information is valuable for an efficient and well-designed management plan, especially considering that sex ratio can vary widely in nature.For instance, sex-biased dispersal will strongly affect sex ratio, and this is widespread in birds and mammals (Pusey 1987) but still poorly investigated in marine organisms (Burgess et al. 2015).Identifying sex-linked markers for identifying the genetic sex of sampled individuals may enable further studies documenting sexbiased dispersal (Yano et al. 2013) as well as overcoming the influence of an unbalanced sex ratio on the analyses of genetic structure.
Addressing bias in sex ratio for population genomic studies on marine and diadromous species Marine and diadromous population genomics studies in animals have become increasingly frequent in recent years, going from a single published article in 2010 to 52 (38 for marine and 14 for diadromous species) articles in 2016 (based on our selective criteria; Table 1).The literature search we performed indicates that in only 9.6% of all studies (5/52; Galindo et al. 2010;Bruneaux et al. 2013;Johnston et al. 2014;Benestan et al. 2015;2016b) information was reported about the sex of the sampled individuals.
Most of these studies have a sample size comparable to those of the present study ( 118and 359 samples on median for marine and diadromous MPS studies respectively) as well as a comparable number of individuals sampled per location (median N per location range = 20-38).Therefore, all of these studies could potentially have been susceptible to the biases identified here.In the majority of these studies, the number of markers genotyped was higher than ours (7,688 and 9,107 SNPs on median for marine and diadromous MPS studies respectively) but since we demonstrate that only 12 and 94 sexlinked markers (0.7% and 1.5% of our total initial MPS datasets) were sufficient to create a signal suggestive of genetic structure, a greater number of markers will not overcome the influence of a small proportion of sex-linked markers in a high gene flow system such as that observed in the majority of marine species.
Since many MPS studies currently under way may not have access to sex information, one alternative way of overcoming the potential bias resulting from sex ratio differences would be to statistically assess the presence of two genetic clusters not associated with geography or other a priori factors hypothesized to influence genetic structure.Then, one could run a BAYESCAN defining groups based on the two observed clusters and assess the level of heterozygosity shown by the outlier markers found, as we did for the American lobster.However, the heterozygosity method will only work if the sex-linked markers are within a sex determining region that is not present on the alternate sex chromosome (i.e.only on Y), or if the sex chromosomes are largely heteromorphic.
Nevertheless, this could help MPS studies to ensure that this bias is not present when interpreting patterns of genetic structure.
Sex determination in the American lobster
In crustaceans, as in many other species, sex is determined either by male (XX/XY) or female heterogamety (ZZ/ZW).However, sex chromosomes are difficult to identify in crustaceans because of the large number of chromosomes (e.g., on average 110 chromosomes for American lobster; Hughes 2014) and the small chromosome size (Legrand et al. 1987).Although markers associated with sex determination can be identified by approaches such as that used here, or as conducted in salmon lice (Lepeophtheirus salmonis, Carmichael et al. 2013) but see also (Gamble & Zarkower 2014), most of the crustacean sex determining systems are poorly understood and understudied (Legrand et al. 1987).Taking advantage of RAD-sequencing, we provide the first evidence of a male heterogametic system in the American lobster (XX/XY), which is in agreement with one review reporting that male heterogamy is more common in Subphylum Crustacea than in the majority of other invertebrate species (Legrand et al. 1987).In addition, we demonstrate the potential to efficiently uncover the sex chromosome system of a non-model species using a genome-wide dataset and analysis of heterozygosity excess or deficit.
Candidate genes involved in sexual differentiation in American lobster
We identified two candidate genes linked to sex in American lobster: SULT1B1, which is involved in steroid metabolism, and cwf19, which acts on pre-RNA splicing.Steroids play important roles in regulating physiological functions related to reproduction and sex differentiation in fishes (James 2011).More broadly, several publications have identified that sulfotransferase genes, such as SULT1, are linked to sex determination in house mouse (Mus musculus; Dunn et al. 1999), mussels (Mytilus galloprovincialis; Atasaral Şahin et al. 2015), European Eel (Anguilla anguilla; Churcher et al. 2015) and Turbot (Scophthalmus maximus; Ribas et al. 2015b).For instance, sulfotransferase 6B1-like gene (SULT6B1) was expressed at higher levels in the livers of sexually mature European Eel males relative to females, which was hypothesised as indicating that this gene may be associated with pheromonal communication during the reproduction of this species (Churcher et al. 2015).In addition, one sulfotransferase gene (hs3st1l2) was a candidate gene for sex determination in Turbot, being associated with differential expression between sexes at sexual maturity (Ribas et al. 2015a).Interestingly, this study also identified cwf19 gene as a putative sex determining gene in the Turbot (Ribas et al. 2015b).
Both candidate polymorphisms occurred in the 3'UTR region of SULT1B1 (SNP 2879519) and cwf19 gene (SNP 1525332).In particular, the polymorphism located in the 3'UTR region of SULT1B1 displayed heterozygosity excess in males and heterozygosity deficit in females (see Table 3).The 3'UTR regions have an important role in posttranscriptional control of gene expression, and thus may affect the level of protein being expressed (Hesketh 2004).Several studies have shown that polymorphisms in 3'UTR region modulate the level of transcription of genes (Barrett et al. 2012).Here, polymorphisms found in SULT1B1 and cwf19 gene may thus affect transcription, as was documented for European Eel (SULT6B1 was overexpressed in liver of sexually mature males with a fold change of 7.8; Churcher et al. 2015) and Turbot (cwf19 was underexpressed in turbot females with a fold change of -1.7 ; Ribas et al. 2015b).
Although the functional annotation for these two genes in American lobster is unknown, these markers may provide information on the sex determination system of this species, but would require further work.
Chromosomes and genes associated with sex-linked markers in Arctic Char
There is no high-density linkage map available yet for Arctic Char, but low-density linkage maps have indicated that the sex chromosome is homologous between Arctic Char and Brook Char and is homologous to Rainbow Trout RT-25 (Timusk et al. 2011).
However, Arctic Char may have a metacentric sex chromosome (Timusk et al. 2011) whereas in this mapping family, Brook Char has an acrocentric sex chromosome (BC35; Sutherland et al. in prep).BC35 corresponds to the ancestral chromosome 15.1 determined from Northen Pike (Esox lucius, Rondeau et al. 2014), which corresponds to Sex-linked markers in genome-wide datasets RT-25b (Sutherland et al. 2016), confirming the previous homology observations (Timusk et al. 2011).Six sex-linked markers are grouped on BC35.Considering that the Arctic Char sex chromosome is expected to be metacentric, the Arctic Char chromosome corresponding to BC35 is probably fused with another acrocentric chromosome that is linked to sex here.Other linkage groups, BC13 (14.1),BC15 (19.1) and BC38 (1.2) also show substantial linkage to sex with between 8-12 sex-linked markers being present on each.Interestingly, BC13 (14.1) is the homeologous chromosome to the Rainbow Trout sex chromosome (omySex; 14.2; Palti et al. 2015;Sutherland et al. 2016), and BC15 (19.1) is homologous to the neo-Y chromosome of Sockeye Salmon (Faber-Hammond et al. 2012).The presence of sex-linked markers on these chromosomes related to sex determination in other salmonids indicates the importance of comparative genomics for characterizing the sex chromosomes of the salmonids, which will be possible as more genomes become available.
Sex-linked markers were located on BC35 on both sides (~280 kb up or downstream) of the SOX-11-like gene (markers 86986 and 87087), which is interesting given the role of the Sox (SRY-related) family in sex determination (Graves 1998).This is the closest annotated gene down-stream to marker 86986 or up-stream to 87087.
Several other sex-linked markers were within genes related to recombination and chromosome segregation, which is interesting given the differences in recombination rate between the sexes (i.e., heterochiasmy) in the salmonids (Sakamoto et al. 2000).Genes containing sex-linked markers that were related to recombination included nipped-B-like protein (on BC13), involved in holding sister chromatids together during cell division (Losada 2014) bloom syndrome protein (on BC15), involved in homologous recombinational repair of double strand breaks during meiosis to suppress crossovers, centrosomal protein of 164 kDa (CEP164; on BC38), a centrosomal protein involved in cell cycle and chromosomal segregation (Sivasubramaniam et al. 2008), and nuclear pore complex protein Nup93 (BC15), with a range of activities including transcription regulation and chromosome segregation (Ibarra & Hetzer 2015).A sex-linked marker was also identified near centrosomal protein kizuna (BC12) involved in establishing mitotic centrosome architecture.Sex-linked markers were also within genes related to transcription factor activity, including retinoic acid receptor RXR-alpha (BC13), a Sex-linked markers in genome-wide datasets member of the steroid and thyroid hormone receptor superfamily involved in sex differentiation in many organisms (Lv et al. 2013) and in SOX-mediated gene expression regulation (Nikčević et al. 2008).Several sex-linked markers were near genes involved in Wnt signaling, which is important for sex determination (Jordan et al. 2001), including Wnt-7b-like (BC15) and frizzled-9-like (BC38).Finally, a sex-linked marker was present in talin-2 (BC15), which has a truncated version known to be specifically expressed in testes and kidneys and expressed in elongating spermatids (Debrand et al. 2009).These genes linked to sex in Arctic Char may provide a better understanding of sex determination and heterochiasmy within the salmonids.
Conclusion
In summary, these results indicate the importance for population genomics studies to collect sex information about individual samples when possible in order to (i) control sex ratio in sampling, (ii) overcome "sex-ratio bias" that can lead to spurious genetic differentiation signals and (iii) fill knowledge gaps regarding sex determining systems.If morphological sex is difficult to determine at some life stages, the identification of sexlinked markers for screening samples may provide a useful alternative solution.Here the exploration of sex-linked markers provided information regarding the sex determination system as well as genes that may be involved in sex dimorphism in American lobster.
Furthermore, using comparative genomics within the salmonids allowed us to identify chromosomes that may harbor genes involved in sex determination this ecologically and economically valuable salmonid species, including several chromosomes, which have already been associated with sex in other salmonid species.
We are grateful to the lobster fishers, the Ekaluktutiak Hunters and Trappers Organization, and the employees of Kitikmeot Foods Ltd without whom this project would have been impossible.We thank G Gerlach for providing the American lobster
Data accessibility
All the raw reads will be submitted to NCBI's Short Read Archive and the qualityfiltered genotypes will be submitted as vcf files to Dryad upon article acceptance.Code to position Arctic Char anonymous markers on the Brook Char genetic map, combine them with BAYESCAN Fst values, and draw the Manhattan plots can be found on GitHub at the following link: https://github.com/bensutherland/salp_anon_to_sfon.Table 1.Marine population genomics studies describing the name of the study, the organism studied, the method used to produce the genetic markers, the number of individuals sampled (N), the number of genetic markers (SNPs), the number of individuals sampled per location (N POP ), the index of genetic differentiation observed among the location studied (F st ).
Tables
using PstI and MspI (details can be found in Perreault-Payette et al. in press).Specimens were individually barcoded with unique six-nucleotide sequences and pooled with 48 individuals per library.Libraries were each sequenced on two Ion Torrent Proton P1v2 chips.
Figure 2 .
Figure 2. Discriminant Analysis of Principal Components (DAPC) and Principal Components Analysis (PCA) of genetic differentiation depending on the sampling scenario (A and C).Results of the DAPC (A) and the PCA (C) performed on lobster and Arctic Char respectively with sex information included.Individuals from the inshore/east and offshore/west regions are represented by different shape symbols, and male and female are represented by black and white symbols, respectively.(B and D) Results of the DAPC (B) and the PCA (D) performed on lobster and Arctic Char respectively, but using hypothetical datasets in which only males were sampled in one of the location (offshore and west respectively) and only female in the other location (inshore and east respectively) showing a false signal of population differentiation driven by differences in sex ratios.
Figure 3 .
Figure3.Boxplots showing the influence of sampling sex ratio on F st. (A) American lobster.F st between offshore and inshore according to sex ratio proportion when subsampling 100 individuals with a sex ratio ranging from a complete unbalanced sex ratio (i.e., sex ratio equal to 0:50/50:0) to a perfectly balanced sex ratio (i.e., sex ratio equal to 25:25/25:25).The horizontal black dashed line indicates the threshold below which F st values are no longer significant at P < 0.05.(B) Arctic Char.F st between east and west according to the sex ratio proportion when subsampling 100 individuals with a sex ratio ranging from a complete unbalanced sex ratio (i.e., sex ratio equal to 0:50/50:0) to a perfectly balanced sex ratio (i.e., sex ratio equal to 25:25/25:25).F st was still significant for the anadromous, but was overestimated in the skewed sex ratio cases.In both panels, the vertical limits of the box represent one standard deviation around the mean (n = 10 individual subsample iterations), the horizontal line within the box is the median, and the whiskers extend from the box to the 25 th and 75 th percentiles.
Figure 4 .
Figure 4.The effect of sex-linked markers on the index of genetic differentiation (F st ).(A) American lobster.The line graph displays the influence of sex-linked markers on F st as a function of the number of sex-linked markers removed from the analysis considering three sampling scenario (10:40/40:10, 5:45/45:5, 0:50/50:0).Sex-linked markers are removed in descending order according to their F st values (see Table1).The dashed line in black indicates the threshold below which F st values are no longer significant at P < 0.05.Sex ratio of 0.4 and 0.5 were not included in this analysis because F st values were not significant in these cases (see Figure4A).(B) Artic Char.The line graph displays the influence of sex-linked markers on (as a function of the number of sexlinked markers removed from the analysis considering three sampling scenario with different degrees of sex ratio bias (0:50/50:0, 5:45:45:5, 10:40/40:10).Sex-linked markers are removed in descending order according to their F st values.
Figure 5 .
Figure 5. Manhattan plot of BAYESCAN F st between the sexes for Arctic Char markers positioned on the Brook Char genetic map.Arctic Char markers without positions were assigned positions on the Brook Char linkage map using multiple iterations of MapComp to identify linkage groups that were associated with sex in Arctic Char.Plotting the BAYESCAN F st along with marker positions indicates four linkage groups show strong linkage to sex: BC13, 15, 35 and 38.All positioned markers are displayed, and crosses indicate significant BAYESCAN F st markers.Markers that are not associated with sex have very low F st and can be seen along all of the linkage groups at the bottom of the graph.
samples.The NSERC CFRN funded the lobster portion of this research, while Fisheries and Oceans Canada and the Nunavut Wildlife Management Board funded the Arctic Char part of the work.Access to the Biomina Galaxy server hosted by the Biomina research center at the University of Antwerp, Belgium, made possible the analysis of the lobster transcriptome data.Funding for the lobster transcriptomics program was provided by the P.E.I.Atlantic Shrimp Corporation Inc. (12-LSC-035) and the Atlantic Lobster Sustainability Measures Program.Sequencing of the lobster transcriptome was performed by Genome Québec.L. Benestan was supported by a doctoral fellowship from NSERC CFRN and Réseau Aquaculture Québec (RAQ), and funds from LB's Canadian Research Chair in Genomics and Conservation of Aquatic Resources.
Table 2 .
Anadromous or catadromous (in the case of eels) population genomics studies describing the name of the study, the organism studied, the study goal, the method used to produce the genetic markers (Method), the number of individuals sampled (N), the number of genetic markers (SNPs), the number of individuals sampled per location (N POP ), the index of genetic differentiation observed among the location studied (F st ).
Table 3 .
American lobster.Observed heterozygosity (H o ), expected heterozygosity (H e ), inbreeding coefficient (F is ), P-value associated to inbreeding coefficient (P-value) and genetic differentiation index (F st ) between sexes (females, n=100; males, n=103) for 12 sex-linked markers identified with BAYESCAN.Markers showing the strongest genetic differentiation between both sexes and belonging to the same LD the cluster are in bold (see FigureS2). | 2018-04-03T01:06:56.285Z | 2016-12-22T00:00:00.000 | {
"year": 2016,
"sha1": "3b88fd1b0797657d3f2b3e816f71ac062f577054",
"oa_license": "CCBY",
"oa_url": "https://www.biorxiv.org/content/biorxiv/early/2016/12/22/096065.full.pdf",
"oa_status": "GREEN",
"pdf_src": "Anansi",
"pdf_hash": "51fc07234f5cce77eeeb71748c09e1563b3f1bc8",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
3344111 | pes2o/s2orc | v3-fos-license | Ablation of caspase-1 protects against TBI-induced pyroptosis in vitro and in vivo
Background Traumatic brain injury (TBI) is a critical public health and socioeconomic problem throughout the world. Inflammation-induced secondary injury is one of the vital pathogenic parameters of TBI. Molecular signaling cascades of pyroptosis, a specific type of cellular necrosis, are key drivers of TBI-induced inflammation. Methods In this study, mice with genetically ablated caspase-1 (caspase-1−/−) were subjected to controlled cortical impact injury in vivo, and primary neuron deficient in caspase-1 through siRNA knockdown and pharmacologic inhibition was stimulated by mechanical scratch, equiaxial stretch, and LPS/ATP in vitro. We evaluated the effects of caspase-1 deficiency on neurological deficits, inflammatory factors, histopathology, cell apoptosis, and pyroptosis. Results During the acute post-injury period (0–48 h) in vivo, motor deficits, anti-inflammatory cytokines (TGF-β and IL-10), pro-inflammatory cytokines (IFN-γ, IL-1β, and IL-18), and blood lactate dehydrogenase (LDH), as well as pyroptosis-related proteins (caspase-1, caspase-1 fragments, caspase-11 and GSDMD), were increased. Caspase-1 was activated in the cortex of TBI mice. Inflammatory activation was more profound in injured wild-type mice than in caspase-1−/− mice. In vitro, mechanical scratch, equiaxial stretch, and LPS/ATP-induced neuron pyroptosis, apoptosis, LDH release, and increased expression of inflammatory factors. The effects of mechanical and inflammatory stress were reduced through inhibition of caspase-1 activity through siRNA knockdown and pharmacologic inhibition. Conclusion Collectively, these data demonstrate that pyroptosis is involved in neuroinflammation and neuronal injury after TBI, and ablation of caspase-1 inhibits TBI-induced pyroptosis. Our findings suggest that caspase-1 may be a potential target for TBI therapy. Electronic supplementary material The online version of this article (10.1186/s12974-018-1083-y) contains supplementary material, which is available to authorized users.
Background
Traumatic brain injury (TBI) is a type of physical brain insult caused by blunt mechanical force [1]. According to the World Health Organization, TBI is a critical public health and socioeconomic problem throughout the world. Considering the incidence of TBI increases annually, TBI will be a major health problem and a substantial cause of disabilities by 2020 [2,3]. Treatment of TBI can be very costly due to the need for on-going treatment of long-term effects, which can include psychiatric vulnerability and neurological disorders. The prevalence of neurological dysfunction following TBI has been increasingly appreciated [4,5]. For example, the risk of clinical depression is approximately1.5 times higher in TBI survivors compared to the general population. Furthermore, the risk for developing Alzheimer disease is 2.3 to 4.5 times greater, depending on whether the TBI is moderate or severe [6]. Worse, still no effective treatment options are available, and little is known about the complex cellular response to TBI.
TBI occurs when the brain is exposed to external forces that induce focal and/or diffuse injuries that include axonal shearing, edema, vascular damage, and neuronal death [7][8][9]. Injuries from TBI can range from mild to severe [10]. The post-TBI primary insult typically leads to secondary damage and neuronal death caused by inflammation [8,9,11], oxygen radical formation [12,13], calcium release [14], and mitochondrial dysfunction [15,16]. To date, TBI research outcomes have not translated into clinical therapeutic approaches [17]. Thus, there is an increasing need for further research into the signaling cascades activated by TBI.
Neuroinflammatory signaling leads to a specific type of inflammatory necrosis called pyroptosis, which is characterized by cell swelling, lysis, and release of proinflammatory cytokines and intracellular contents [18]. Pyroptosis is involved in pathological processes during viral or bacterial infection (or in the presence of products from pathogens such as lipopolysaccharides (LPS) or viral DNA) [19,20], myocardial ischemia, lung and kidney injury, and cerebral stroke [21][22][23][24]. This type of inflammatory necrosis occurs predominantly in the phagocytes, macrophages, monocytes, and dendritic cells, as well as various other cell types such as T cells. The pro-inflammatory caspase subfamily, including caspase-1 in humans and mice, caspase-4 and caspase-5 in humans, and caspase-11 in mice, mediate pyroptosis [19]. Li et al. found that chemical neurotoxicity-induced pyroptosis of the cortical neurons is dependent on caspase signaling through the Erk1/2-Nrf2/Bach1 signal pathway [25]. However, the specific mechanism of neuronal pyroptosis in TBI has not yet been elucidated.
To better understand the mechanism of brain injury responses, we used an in vivo mouse model of TBI and in vitro cellular models of mechanical stress and inflammation. Caspase-1 knockout (KO) mice (caspase-1 −/− ), siRNA knockdown, and pharmacologic inhibition were used as tools to assess the involvement of caspase-1 in downstream inflammation following TBI and mechanical stress. We found that genetic ablation, expression reduction, and pharmacologic inhibition of caspase-1 decreased TBI-and mechanical injury-induced behavioral neurological deficits, inflammatory cytokine protein and mRNA expression levels, and neuronal apoptosis. Ultimately, our results lead us to conclude that mechanical stress caused by TBI activates caspase-1 which contributes to TBI-induced pyroptosis.
Animals and TBI models
In this study, we used caspase1-deficient (caspase1 −/− ; Jackson Laboratories, Bar Harbor, ME, USA) adult male mice (8-10 weeks old) and age-matched C57Bl/6 (WT) male mice. Mice were housed in the animal care facility under a 12-h light-dark cycle, with ad libitum access to food and water. All surgical procedures were carried out in accordance with protocols approved by the Institutional Animal Care and Use Committee. The animals were anesthetized with intramuscular injection of chloral hydrate (5%, 0.1 ml/10 g). Mice were fixed in a stereotaxic frame, and the scalp was shaved and cleaned with iodophor. The left lateral aspect of the skull was exposed by retracting the skin and the surrounding soft tissue. Traumatic brain injury (TBI) was induced by dropping a 200-g steel weight with a flat end from a height of 5 cm onto the left lateral skull. The free-falling weight produced a moderate contusion injury of the left parietal cortex. Sham-injured control animals underwent a similar anesthesia protocol as the TBI animals with the exception that the scalp was not exposed and no weight was dropped. About 25-30% of the mice died in the first seconds after the trauma, but there was no delayed mortality or prostration in the surviving mice. The body temperatures of the animals were monitored by a rectal probe and fixed in a range of 37 ± 0.5°C using a heating pad. The animals were returned to their quarters after recovery from anesthesia and were allocated randomly into groups, with 18 mice in each group.
Neurobehavioral training and evaluation
Neurological deficits were assessed using well-established, modified neurological severity scoring (mNSS), open-field and Rotarod testing at 12, 24, and 48 h after TBI. Experimenters were blinded to the TBI or sham treatments as well as the genotype of the mice. Each behavioral test was repeated twice with four different trials to validate the data. The mNSS test consists of ten different tasks that can evaluate the motor (muscle status, abnormal movement), sensory (visual, tactile, and proprioceptive), balance, and reflex functions of mice. Neurological function was graded from 0 to 18 (0 = normal function; 18 =maximal deficit). One point was scored for each abnormal behavior or for the lack of a tested reflex. Therefore, higher scores imply greater neurological injury.
The open-field test was used to determine general activity levels and to measure anxiety-like behavior. Animals were monitored under moderate lighting for 15 min in a 40-cm 2 open field using video tracking software (ANY-Maze, Stoelting, USA). General activity was evaluated by determining the total of distance traveled. Anxiety-like behavior was assessed based on the pattern of exploration in the open field (center vs. periphery).
Fine motor coordination and learning were assessed using an accelerating Rotarod apparatus (RWD Life Science, China). On the day before the injury, mice were trained on the Rotarod for three consecutive trials at a slow rotational speed (4 rpm/min) for 1 min to adapt to the rod, followed by four additional trials with an accelerating rotational speed (from 4 to 40 rpm in 5 min) to obtain baseline latency. On each testing day, the mice were given four 300-s accelerating Rotarod trials with an inter-trial interval of 30 min. The average latency to the first fall off the rod was recorded. Passive rotation, accompanying the rotating rod without walking, were also considered as a fall.
Primary neuronal culture and injury models
The primary neuronal culture was prepared as follow: Cortex from embryonic day 14 BALB/c mice was used for cultures. After decapitation of the mice and removal of the meninges, the cortex was digested by papayotin, and the cells were treated with DNase I (Sigma) prior to centrifugation at 1000×g for 3 min. Neurons were cultured on poly-L-lysine-coated dishes at a density of 1 × 10 6 cells per 10-cm dish in minimum essential medium with 10% horse serum. The medium was changed to neuronal base medium (supplemented with 2% B27, 0.2 mM L-glutamine, and 1% penicillin-streptomycin) on the following day. The medium was changed every 3 days, and cultured cells were used for experiments on post-culture days 7-14. Subsequently, the primary neuron was collected and detected the Iba1 (microglia) and GFAP (astrocytes) expression through Western blotting (WB) and NeuN stain to assess neuronal purity.
We then performed sets of three experiments to examine the effects of mechanical and inflammatory stress on neurons in vitro: (1) neurons were incubation for 5 h with LPS and 5 mM ATP for 30 mins; (2) neurons were scratched by a 200-μl yellow gunshot; and (3) equiaxial stretch (12% strain, 1.0 Hz frequency) was applied to cultured neurons for 4 h, by a Flexcell® FX-5000™ Tension System (Flexcell, USA). In the stretch model, neurons were inoculated in 6-well plates (BioFLEX®). At the same time, we studied the role of caspase-1 in neuronal pyroptosis by caspase-1 siRNA knockdown or adding the caspase-1specific inhibitor, Belnacasan (VX-765, 10 μM, for 24 h). After treatments, cells were used for flow cytometry, immunofluorescence, and protein or RNA extraction.
Transient transfection
The primary neuron was respectively infected with caspase-1 shRNA (m) lentiviral particles (Santa Cruz, sc-29922-V) according to the manufacturer's protocol (Santa Cruz Biotechnology, Inc.). Plate targets cells in a 24-well plate 48 h prior to viral infection. The cells should be approximately 50% confluent before infection. Add 0.5 ml of complete optimal medium with caspase-1 shRNA lentiviral particles (MOI = 5) (without polybrene), incubate cells for 48 h, and then change complete optimal medium without lentiviral. At 72 h after infection, the RNAi efficiency was determined by qRT-PCR (Additional file 1: Table S1 and Additional file 2: Figure S1).
Apoptosis analysis by Annexin V-PI double staining
Neurons were stained with Annexin-V and propidium iodide (PI) using the Annexin-V-FITC kit (BD Biosciences) according to the protocol of the company. Briefly, cell culture medium was collected separately, and the treated cells were digested and washed twice with ice-cold PBS and re-suspended in the binding buffer at a concentration of 1 × 10 5 per 100 μl. Then, the cells were supplied with 5 μl of Annexin V-FITC and 5 μl of PI and incubated in the dark at RT for 10 min. Next, 400 μl of binding buffer was added into each sample for flow cytometry analysis (BD Accuri. C6 Plus, BD Biosciences).
LDH release detection
Mice were sacrificed under deep anesthesia for blood collection (centrifugation at 3000g, 10 min), and the blood serum was measured for lactate dehydrogenase (LDH) release. For the primary neuron, supernatant from serum-free media was filtered using 0.2-μm syringe filters to use for LDH release detection. The detection was using a commercially available kit (Solarbio, Beijing). One hundred microliters of the blood serum or supernatant was transferred to 96-well plates, and then the reaction mixture was added and incubated in the dark for 30 min at RT. LDH concentration was quantified by measuring the absorbance at 490 nm.
Caspase-1 activity detection
The caspase-1 activity was detected with Caspase-1 Activity Assay Kit (Solarbio, Beijing). The specific steps were conducted according to the manufacturer's instructions: First, 2-10 × 10 6 cells were lysed in 50-100 μl lysis buffer on ice for 10 min. For tissue samples, 3-10 mg tissues were added to 100 μl lysis and were homogenated with a tissue homogenizer and centrifuged. The supernatant was retained. Protein concentrations were detected using the Bradford method, ensuring that the protein concentration was 1-3 μg/μl. A standard curve was prepared using the pNA standard. Then, the optical density of specimen was read on a microplate reader (Molecular Device) at 405 nm. The percentage of caspase-1 activity changes was calculated by the radio of OD405 of the experimental wells to that of the normal wells.
TUNEL staining
The cerebral cortex was harvested, and apoptosis was determined by terminal deoxynucleotidyl transferasemediated dUTP-biotin nick end labeling (TUNEL) staining, using a TUNEL Apoptosis Assay Kit (R&D, Switzerland). TUNEL staining was performed with fluorescein-dUTP for apoptotic cell nuclei and 4′,6diamidino-2-phenylindole (DAPI) to stain all cell nuclei. apoptosis index (AI), the number of TUNEL-positive cells divided by the total cells per field, was examined. Each AI was assessed in 20 randomly selected fields.
Immunohistochemical analysis
Mice were euthanized by injection of 5% chloralic hydras in accordance with our approved animal protocol, and the brain tissue was removed. The brain tissues were immediately perfused with 4% paraformaldehyde in PBS after removal and embedded in paraffin for sectioning. Following rehydration of the paraffin section and two washes in PBS, endogenous peroxidase activity was blocked using 3% H 2 O 2 for 10 min. Antigen crosslinking was conducted in an autoclave for 3 min, and then the samples were washed twice in PBS. All preparations were then treated with goat serum blocking reagent (Abcam, ab7481) for 45 min. Excess reagent was removed with a quick rinse with PBS. Sections were incubated with primary antibodies overnight. Following two washes in PBS, the secondary antibody (Abgent, ASS3403) were added to all slides for 30 min. Excess reagent was removed, and the slides were washed in PBS and incubated with DAB (Solarbio, DA1010) for 5-10 min until the desired color appeared. All preparations were counterstained with hematoxylin (Solarbio, G1120) for 30 s. Protein expression was determined in five adjacent sections per sample.
To extract protein from the cortical brain tissues, the frozen brain samples were homogenized in ice-cold buffer containing 20 mM HEPES (pH 7.0), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, Proteinase Inhibitor Cocktail (Roche Applied Sciences), and Halt Phosphatase Inhibitor Cocktail (Thermo Scientific) for 30 s. The samples were sonicated through the Cell Ultrasound Cruizer (Thermo Scientific) for 10 s and centrifuged at 15,000 rpm for 10 min, and total proteins were recovered in the supernatants.
Immunofluorescence assays
Primary neurons were subjected to immunofluorescence by the protocol described below. Cells were fixed by 4% PFA for 20 min at RT followed by PBS wash for three times (5 min each). The cells were then incubated by the blocking buffer (0.2% TritonX-100 and 5% BSA dissolved in PBS) for 1 h. After PBS wash for three times, the cells were then incubated with the primary antibodies of NeuN (Abcam, 1:300,), dissolved in the blocking buffer at 4°C for 24 h. PBS wash was carried out followed by the treatments with the Alexa Fluor 488/568-conjugated secondary antibody (Abgent, 1:1000) dissolved in the blocking buffer for 2 h at RT. After incubation with secondary antibodies, cells were washed and counterstained with DAPI solution (1 μg/ml in the blocking buffer) for 10 min at RT. Cells were analyzed by fluorescence microscope (Leica, IX71).
Statistical analysis
All the data were presented as the mean ± SEM and were analyzed using SPSS statistical software (version 22.0, IBM). Non-parametric data from the mNSS test were analyzed using the Kruskal-Wallis H analysis followed by a Mann-Whitney U test. One-way analysis of variance (ANOVA) with repeated measures was used to analyze Rotarod data. The remaining biochemical data were analyzed using a two-way ANOVA. Each experiment was repeated three times, and significance was determined with two-tailed Student's t test or one-way ANOVA. Error bars represent ± SEM. p value lower than 0.05 was considered as significant.
TBI-induced neurological deficits
To assess neurological deficits post-TBI, we conducted a time course of behavioral testing before TBI and 12, 24, and 48 h following TBI. Prior to surgery, there was no difference in mNSS and Rotarod scores between the groups of mice. In the mNSS test (Fig. 1a), mice in the sham group showed no significant differences at 12, 24, and 48 h; however, TBI caused significantly higher mNSS relative to the sham group (p < 0.001). The mNSS scores were maximal at 24 h and gradually decreased, albeit insignificantly, with time in TBI group. The scores remained significantly different between the TBI and sham groups at 48 h post-injury (p < 0.05). Caspase-1 −/− -TBI mice exhibited lower mNSS compared to the wide-type (WT) mice subjected to TBI (WT-TBI) (p < 0.01). To examine spontaneous locomotor activity in response to a novel environment, all mice were tested in the open-field behavioral task (Fig. 1c). WT-TBI mice spent a shorter time in the perimeter zone and traveled less total distance than sham group (p < 0.01). A modest yet statistically significant difference in perimeter zone and total travel distance was revealed between WT-TBI and caspase-1 −/− -TBI mice (p < 0.05). In the Rotarod test (Fig. 1b), the sham groups exhibited the best performance relative to the TBI groups (p < 0.001). The scores were minimal at 24 h and following gradually increased with time in WT-TBI group. Caspase-1 −/− -TBI mice had higher scores relative to the WT-TBI group (p < 0.05). The brains of the mice were assessed postmortem. TUNEL staining Fig. 1 Neurological deficits after TBI. Neurological effects of TBI were assessed by a mNSS, b Rotarod, and c open-field behavioral task tests prior to and 12, 24, and 48 h post-TBI. a mNSS scores were significantly higher in the TBI groups relative to those in the sham groups. The mNSS scores were maximal at 24 h and gradually decreased with time in the TBI group. No significant difference was observed between the caspase-1 −/− and WT-TBI group at 12 h. However, the caspase-1 −/− group showed significantly lower mNSS scores relative to the WT-TBI group at 24 and 48 h post-TBI. b Rotarod scores were minimal at 24 h and gradually increased with time in WT-TBI group. Relative to WT-TBI mice, TBI-induced impaired Rotarod performance was attenuated in the caspase-1 −/− mice at 24 and 48 h after TBI. c In the open-field task, all mice traveled the same distance in the center zone compared to the perimeter zone. WT-TBI mice spent less time in the perimeter zone of the open-field chamber compared to caspase-1 −/− -TBI mice. Total travel distance was significantly different between WT-TBI and caspase-1 −/− -TBI mice. d TUNEL staining was used to detect cortical damage in our TBI model. Apoptosis index, the number of TUNEL-positive cells divided by the total cells per field, was examined. Each AI was assessed in 20 randomly selected fields. Data are presented as the mean ± SEM. **p < 0.01 and ***p < 0.001 versus sham group, #p < 0.05 and ##p < 0.01, compare to TBI group (n = 11) (mNSS, Rotarod, and open-field) was used to detect cortical damage. The green fluorescence was dramatically enhanced in WT-TBI group compared to the sham group (p < 0.001). Apoptosis was significantly reduced in the brains of caspase-1 −/− -TBI mice relative to the WT-TBI group (p < 0.05; Fig. 1d).
Inflammatory cytokine levels in the cortex following TBI
To compare the protein levels of inflammatory-associated mediators in the cortex, an array of inflammatory cytokines were measured using ELISA. All detected cytokine levels were the lowest in the sham groups compared to the other groups (Fig. 2). Protein levels of the anti-inflammatory cytokines TGF-β at 12 h (p < 0.01), 24 h (p < 0.001), and 48 h (p < 0.001) and levels of IL-10 at 12 h (p < 0.001), 24 h (p < 0.001), and 48 h (p < 0.001) post-TBI increased in the WT-TBI group compared to the sham group (Fig. 2a, b). Furthermore, levels of the pro-inflammatory cytokines IFN-γ, IL-1β, and IL-18 were all augmented significantly at 24 h (p < 0.001) and 48 h (p < 0.01) in the TBI group relative to the sham group Fig. 2 Inflammatory cytokine expression in the cortex following TBI. a-e Relative to the sham group, concentrations of the anti-inflammatory cytokines a TGF-β1 and b IL-10 and pro-inflammatory cytokines c IFN-γ, d IL-1β, and e IL-18 were significantly increased in the region of the contusion at 12, 24, and 48 h after TBI, whereas IFN-γ level barely had changed at 12 h after TBI. Caspase-1 −/− -TBI mice had a less dramatic increase in anti-inflammatory cytokines and pro-inflammatory cytokines than the WT-TBI mice. f mRNA expression of inflammatory cytokines, TLR4, MyD88, and NLRP3, at 24 h post-TBI. TBI induced a marked increase in TLR4, MyD88, and NLRP3 mRNA expression in the injured WT-TBI mice brains relative to the sham group, but caspase-1 −/− only exhibited mild increases. Data are presented as the mean ± SEM. **p < 0.01 and ***p < 0.001, compared to sham group, #p < 0.05 and ###p < 0.001 versus TBI, n = 6/group (Fig. 2c-e). Meanwhile, the levels of these proteins, IFN-γ, IL-1β, and IL-18, peaked at 24 h and declined significantly by 48 h (p < 0.05, relative to 24 h). Post-TBI at 24 h, protein levels of TGFβ and IL-10 significantly increased in injured caspase-1 −/− mice as compared to WT-TBI (p < 0.05). IFN-γ in caspase-1 −/− -TBI mice reduced as compared to WT-TBI group (p < 0.05) at 48 h. There was no significant difference in protein levels of IL-1β and IL-18 in caspase-1 −/− -TBI mice compared to sham. Gene expression related to the inflammation in the cortex region of the brain tissues was analyzed using qRT-PCR. Consistent with the ELISA results, TBI induced dramatic increases in mRNA expression of inflammatory cytokines compared to the sham groups at 24 h post-injury. The mRNA expression levels of TLR4, MyD88, and NLRP3 all ascended at 24 h post-TBI (p < 0.001) (Fig. 2f). However, NLRP3 mRNA was dramatically reduced in caspase-1 −/− -TBI mice as compared to WT-TBI group (p < 0.001).
TBI-induced pyroptosis in mice model
To determine whether pyroptosis was occurring in the murine TBI model and presented time-dependence, inflammation was assessed in the cortex by WB, caspase-1 activity assay kit, and immunohistochemistry. As shown in Fig. 3a, WB assay was performed to quantify pyroptosis-related protein expression at 0, 1, 2, 4, 8, 12, and 24 h post-TBI. Caspase-1 p45 was elevated at 8 h (1.4-fold), 12 h (1.9-fold), and 24 h (2.0-fold) compared to the brain tissue from uninjured animals. The p20 and p10 fragments of caspase-1 increased gradually-24 h to 5.6-folds and 1.5-folds, respectively. Meanwhile, caspase-11 and GSDMD protein expression increased after TBI. At 24 h post-TBI, caspase-11 exhibited a 3.5-fold increase, and GSDMD expression increased 3.6-fold relative to normal brain tissue. As shown in Fig. 3b, caspase-1 (including p10, p20, and p45), caspase-11, and GSDMD expression were a significant reduction in caspase-1 −/− -TBI mice when relative to WT-TBI mice at 24 h. Immunohistochemistry was used to detect caspase-1, caspase-11, and GSDMD protein distribution in the trauma region and the surrounding brain tissue at 24 h post-TBI. As shown in Fig. 3c, caspase-1 and caspase-11 were mainly expressed in the injured area in WT-TBI mice, while they were less expressed in caspase-1 −/− -TBI mice. They had a high degree of co-localization and gradually diffused outward radially to the surrounding tissue, and the expression gradually decreased in a radial distribution from the injury to the surrounding area. Consistent with the WB results, caspase-1 activity significantly changed after TBI. Following TBI, caspase-1 activity was significantly greater in cortical samples from TBI animals than non-injured animals (Fig. 3d). For instance, at 24 h post-injury, caspase-1 had 15.4-fold greater activity in WT-TBI mice as compared to sham controls. The caspase-1 activity was less affected in the caspase-1 −/− -TBI mice. At 24 h post-injury, the caspasedeficient mice exhibited only a 2.3-fold increase in the activity relative to the normal brain tissue (Fig. 3d). LDH is found in almost all body tissues. It plays an important role in cellular respiration, the process by which glucose from food is converted into usable energy for cells. Although LDH is abundant in tissue cells, blood levels are normally low. While the tissues are damaged by injury, they release more LDH into the bloodstream. Interestingly, serum LDH was tested at 24 h following TBI, since cerebral injury increases cellular LDH leakage. Serum LDH concentration increased approximately 2.0-fold in WT-TBI mice relative to the sham group (p < 0.001) and was near normal levels in caspase-1 −/− -TBI mice relative to the WT-TBI mice (p < 0.01) (Fig. 3e).
Neuron injury-induced apoptosis
Primary neuron culture was potential contamination with another type of cells, such as glial cells. The primary neuron was collected and detected the Iba1 (microglia) and GFAP (astrocytes) expression through WB to assess neuronal purity. At 7 days, Iba1 and GFAP were expressed, and they were disappeared following culture at 10 and 14 days (Fig. 4a). At 10 days, there were no obvious protein bands of Iba1 and GFAP, when 30, 50, and 70 μg protein were loaded respectively (Fig. 4b). Furthermore, we have also stained the primary culture cells by markers for neurons (NeuN) and confirmed that more than 90% of cells were neurons (Fig. 4c).
In order to continue to explore the correlation between neuron damage and pyroptosis, we stimulated cultured primary neuron mechanically with scratch and stretch or chemically with LPS/ATP treatment. Apoptosis was detected by an Annexin-V-FITC kit. To validate these methods of mechanical and chemical injury, we assessed the apoptosis rate over time. Scratch, stretch, and LPS/ ATP all significantly induced apoptosis of primary neuron (Fig. 5a). Apoptosis scores were 47.14 ± 5.51 for scratch stimulation, 46.56 ± 3.54 for equiaxial stretch, and 45.37 ± 3.87 for LPS/ATP stimulation. The culture supernatant was collected to detect LDH release by scratch, stretch, and LPS/ATP stimulation. Supernatant LDH concentration increased approximately 1.7-fold, 1.6-fold, and 2.2-fold by scratch, stretch, and LPS/ ATP stimulation, respectively, relative to normal group (p < 0.001) (Fig. 5b).
Caspase-1 effected neuron injury
In order to determine whether caspase-1 resides in upstream cellular signaling pathways that lead to pyroptosis after mechanical and chemical stress and injury, we utilized caspase-1 siRNA to knockdown expression of caspase-1 as well as a pharmacological inhibitor, VX-765. Caspase-1 siRNA and inhibition with VX-765 reduced neuron apoptosis following mechanical scratch, equiaxial stretch, and (See figure on previous page.) Fig. 3 Pyroptosis in the murine model of TBI. a Expression of pyroptosis-related proteins in response to TBI was assessed by Western immunoblotting (WB). After 0, 1, 2, 4, 8, 12, and 24 h post-TBI, caspase-1, caspase-11, GSDMD, and caspase-1 protein fragments were severally elevated in WT-TBI mice compared to sham controls. All data are expressed as means ± SD, n = 5. Post-TBI at 24 h, caspase-1 −/− mice exhibited reduced TBI-induced pyroptosis. b In caspase-1 −/− mice, pyroptosis-related proteins were less prominent than the WT-TBI group, which demonstrated in WB results. All data are expressed as means ± SEM, n = 5. c Micrograph at × 20 magnification of caspase-1, caspase-11, and GSDMD immunostaining in the cortex in the vicinity of the lesion. As shown in the immunohistochemistry results, caspase-1, caspase-11, and GSDMD distributed in trauma organization and surrounding. The cortex showed a strong increase in caspase-1 and GSDMD staining in the WT-TBI compared to caspase-1 −/− -TBI at 24 h post-TBI. However, caspase-11 was no obvious difference between caspase-1 −/− -TBI and WT-TBI. One representative experiment of four was shown. d Caspase-1 activity was stimulated after TBI. In contrast, TBI-induced increase in caspase-1 activity was significantly reduced in caspase-1 −/− mice (###p < 0.001, WT-TBI vs. caspase-1 −/− -TBI). e Serum LDH was tested at 24 h following TBI since cerebral injury increases cellular LDH leakage (p < 0.001). Data are presented as the mean ± SEM. *p < 0.05 and ***p < 0.001 versus sham group, ###p < 0.001 versus TBI group, n = 6 Fig. 4 The purity of in vitro primary neuron culture. a Primary neuron was collected at 7, 10, and 14 days, respectively; 50 μg protein per lane was used to detect Iba1 and GFAP expression. At 7 days, Iba1 and GFAP were expressed, they were disappeared following culture at 10 and 14 days. b 30, 50, and 70 μg protein per lane were used to detect Iba1 and GFAP expression at 10 days. As before, there was no obvious protein bands of Iba1 and GFAP. c Primary neurons were stained by the neuronal marker NeuN at 10 days. To confirm the purity of in vitro primary neuron culture, cells were stained by NeuN, and DAPI was used to counterstain the cellular nucleus. Almost all cells were stained by NeuN. One representative experiment of four was shown LPS/ATP stimulation (Fig. 7a). After VX-765 treatment, apoptosis rates were 26.36 ± 3.16 cells/h for scratch, 23.13 ± 2.13 cells/h for equiaxial stretch, and 21.79 ± 1.51 cells/h for LPS/ATP. Similarly, with siRNA knockdown of caspase-1, apoptosis scores were 23.54 ± 3.11 cells/h for scratch, 16.82 ± 1.28 cells/h with equiaxial stretch, and 17.72 ± 1.11 cells/h with LPS/ATP. In comparison, the apoptosis scores for neurons with fully functional caspase-1 were significantly greater-41.36 ± 6.44 cells/h for scratch, 47.38 ± 4.61 cells/h for equiaxial stretch, and 44.62 ± 4.81 cells/h for LPS/ATP. The culture supernatant was collected to detect LDH release. Supernatant LDH concentration observably increased after the scratch, stretch, or LPS/ATP stimulation, relative to normal culture (p < 0.01), whereas the leaked LDH significantly reduced in caspase-1 siRNA and VX-765 groups, relative to control group (p < 0.05) (Fig. 7b). Enhanced caspase-1 activity by neuronal injury was inhibited by caspase-1 siRNA and VX-765 inhibition (Fig. 7c). Expression of caspase-1, caspase-11, and GSDMD decreased in caspase-1 siRNA and VX-765 groups, and caspase-1 p45 and its fragments (p10 and p20) significantly reduced in the culture supernatant, simultaneously (Fig. 7d). In order to demonstrate the reduction in the pro-form of caspase-1 levels by VX-765 was not an artifact due to direct VX-765 neurotoxicity, VX-765 groups, and a vehicle control (DMSO as the menstruum) for VX-765 treated with neuron. CCK8 results demonstrated that there was no neurotoxicity by VX-765 at 10 μM (Additional file 4: Figure S3). Further, ELISA results showed decreased secretion of neuroinflammatory mediators, such as IL-1β and IL-18, when the caspase-1 was inhibited (Table 1).
Discussion
Results of this study demonstrate that pyroptosis modulates responses to the acute phase of TBI through caspase-1. Caspase-1 is expressed and activated in neurons and therefore is likely to be activated via the mechanical damage. During the acute post-traumatic period (0-48 h), neurological deficits, inflammatory cytokine, and pyroptosis are greater in the brain-injured mice. To date, there are scant data regarding the regulating of pyroptosis during the acute phase of TBI. Therefore, we performed a time-effect study for damage on post-TBI sensory-motor deficits and motor dysfunctions, which has significantly higher mNSS score and minimum Rotarod test score at 24 h. We found that acute injury significantly reduced neuron functional deficits as early as 24 h post-TBI. For instance, a previous study using this murine TBI model observed higher neurological deficits at 24 h than at 72 h and found that brain injury is most pronounced after 24 h [26]. Inflammatory correlation factors are also detected in the acute phase. A range of anti-inflammatory cytokines and pro-inflammatory cytokine secretion were all increased and continued high secretion from 24 to 48 h, such as TGF-β, IL-10, IFN-γ, IL-1β, and IL-18. Simultaneously, mRNA of the upstream inflammatory regulator TLR4, MyD88, and NLRP3 are also highly expressed at 24 h post-TBI. Apoptosis and immunostaining of the damaged cortex further validated our murine TBI model. There was significant apoptosis, expression of caspase-1, caspase-11, and GSDMD proteins around the contusion site, and blood LDH release at 24 h post-TBI. These data suggest that pyroptosis is likely involved in neuroinflammation with TBI. Apoptoses mediated by caspase-induced cleavage events are key features of pyroptosis [27]. Adamczak et al. verify neuronal pyroptosis Fig. 5 Neuron injury-induced apoptosis in cultured primary neuron. a Flow cytometer results for apoptosis following scratch-, stretch-, and LPS/ ATP-induced neuron damage. Scratch, stretch, and LPS/ATP stimulation induced significant apoptosis in cultured primary neuron (***p < 0.001, relative to the normal group). b Culture supernatant was collected to detect LDH release by scratch, stretch, and LPS/ATP stimulation. Supernatant LDH concentration increased by scratch, stretch, and LPS/ATP stimulation (p < 0.001). All data are expressed as means ± SEM; one representative experiment of five was shown which is mediated by the AIM2 inflammasome, is an important cell death mechanism during the CNS infection and injury [28]. NLRP3 inflammasome plays a key role in the secondary phase of TBI, and lack of the NLRP3 inflammasome will improve recovery following TBI [29][30][31]. AIM2 and NLRP3 as sensors, they are involved in activation of caspase-1 which catalyze neuroinflammation and pyroptosis. However, we directly demonstrate whether pyroptosis involves the acute phase of TBI and whether caspase-1 deficiency will reduce neuroinflammation and neuron injury in vivo and in vitro.
On the cell level, we stimulated primary neuron through scratch, stretch, or LPS/ATP, to discuss the mechanism of pyroptosis on TBI. Mechanical scratch can directly damage the neurons; inflammatory cytokines in damaged neurons may release rapidly and thus affect the activity of other normal neurons. As for the mechanical stretch, it is commonly used to study the biomechanical stimulation of cardiovascular disease and bone-related (See figure on previous page.) Fig. 6 Neuron injury-induced inflammation and pyroptosis. Relative to the control group (cultured primary neuron without mechanical or chemical treatment), scratch, stretch, and LPS/ATP stimulation induced significant increases in pro-inflammatory cytokines: a IFN-γ, b IL-6, c IL-1β, and d IL-18. e mRNA expression of TLR4, MyD88, and NLRP3 after scratch, stretch, and LPS/ATP treatments. Expression of these transcripts increased by 4.5-fold, 4.0-fold, and 9.0-fold, respectively, compared to control neurons. f Caspase-1 activity was enhanced at 10.3-fold, 12.4-fold, and 21.6-fold following scratch, stretch, and LPS/ATP stimulation, compared to the control group. g Cell lysate and culture supernatant were collected; pyroptosis-related protein expression was detected by WB after scratch, stretch, and LPS/ATP stimulation in cultured primary neuron. Caspase-1 and its fragments (p10 and p20) leaked to the culture supernatant. The histogram was used to analyze protein expression in the cell lysate. Data are presented as the mean ± SEM. ***p < 0.001 versus normal group; one representative experiment of five was shown Fig. 7 Caspase-1 deficiency attenuated neuron injury-induced apoptosis and pyroptosis. a Mechanical scratch, equiaxial stretch, and LPS/ATP stimulation induced apoptosis and caspase-1 knockdown with siRNA and pharmacologic inhibition with VX-765 reduced cellular apoptosis following scratch, stretch, and LPS/ATP. b Supernatant LDH concentration was detected when caspase-1 was a deficiency (siRNA and VX-765). c Caspase-1 activity was enhanced with scratch, stretch, and LPS/ATP stimulation of cultured primary neuron. Caspase-1 deficiency reduced the caspase-1 activity enhancement. d Caspase-1 deficiency decreased pyroptosis-related protein expression and reduced caspase-1 and its fragments (p10 and p20) leakage. The histogram was used to analyze protein expression in the cell lysate. Data are presented as the mean ± SEM. **p < 0.01 and ***p < 0.001 versus normal group, #p < 0.05, ##p < 0.01, and ###p < 0.001 versus control group, one representative experiment of five was shown [43,44]. The neuron/axles may be damaged under equiaxial stretch (12% strain, 1.0 Hz frequency) for 4 h. In some reports, LPS usually has been used to induce inflammation and pyroptosis of macrophages [23,36]. In this work, we tried to incur pyroptosis by LPS/ATP in the primary neuron. Scratch or stretch and LPS/ATP, all caused neuron damage, neuron apoptosis, LDH leakage, and a lot of inflammatory cytokine release, and pyroptosisrelated proteins were also triggered. Caspase is a protease involved in cell apoptosis, with more than ten subtypes. Caspase-1 regulates cell apoptosis by shearing Bcl-XL and the inflammatory response of related cytokines which are mediated by the shear of the precursors of cytokines [45]. Caspase-1 is the unique caspase that produces active cytokines in the precursors of IL-1β and IL-18 [19]. Caspase-11 shears 45kd-caspase-1 precursor protein to produce p10 and p20 fragments, which form a heterogenous dimer and then form a tetrameric active enzyme [46]. In the case of caspase-1 suppression, caspase-1 activity was significantly inhibited, it weakens the injure and LDH leakage of neurons which induced by scratch, stretch, or LPS/ATP, and it decreased the expression or the release of inflammatory factors. Although previous studies had suggested that caspase-1 was the main enzyme implicated in the cleavage of pro-IL-1β and pro-IL-18 into the biologically active cytokine, caspase-1 and its fragments (p10 and p20) became less and mature of IL-1β and IL-18 reduced when caspase-1 was suppressed in the primary neuron. In the meantime, caspase-1 suppression aroused decrease of GSDMD, which formed large pores in the membrane that drive swelling and membrane rupture. Ultimately, caspase-1 deficiency reduced secretion of mature IL-1β and IL-18, which abated neuroinflammation [47].
Our results align with what is known about the role of caspase-1 in other brain injury responses. Previous researches show that caspase-1 KO mice and transgenic mice expressing a dominant-negative caspase-1 construct are partially resistant to cerebral disease and stroke. In the permanent stroke model, the proteolytic activity of caspase-1 is increased 30 min after occlusion with a second wave of activation 12 h later [48]. The increase of caspase-1 expression has been described in neurons and astrocytes after thromboembolic stroke and observed later in microglia (24 h post-stroke) [49]. Other studies have demonstrated the role of caspase-1 after stroke using transgenic mice. Caspase-1 KO mice and mice expressing dominant-negative caspase-1 exhibit a reduction of brain damage relative to wild type after experimental stroke [50]. Moreover, intracerebroventricular administration of caspase-1 inhibitors provides protective effects in experimental stroke models [51,52]. Beneficial effects of caspase-1 inhibition are also observed in a model of oxygen and glucose deprivation in rat organotypic hippocampal slices [53]. Some studies indicate that the caspase-1 mRNA expression in the cortex was significantly increased in AD patients [54]. Altogether, these studies as well as our study show that caspase-1 is a key regulator of underlying mechanisms of brain injury and disease.
Our study experimentally demonstrates that TBI induces pyroptosis, and caspase-1 deficiency reduces neuroinflammation and neuronal damage in the acute phase. These findings may provide strategies for clinical treatment of TBI and other neuroinflammatory conditions. There is an increasing realization that neuroinflammation occurs within the whole gamut of the central nervous system pathologies [55]. Regulating inflammation to control TBI damage is possible through drug treatments [56]. NLRP3 inflammasome and interleukin-1 receptor antagonist treatments have been used to mitigate neuroinflammation after TBI [30,57]. These treatments were able to alleviate the nerve damage and promote neurological recovery after TBI. Electro-acupunctureinduced TLR4 signaling inhibition has been used to promote hippocampal neurogenesis and neurological recovery post-trauma [58].
Conclusion
Overall, our findings indicate that TBI induces pyroptosis, and caspase-1 is a significant player in neuroinflammation. The caspase-1 deficiency drives an anti-inflammatory response in the injured cortex, and depressed pyroptosis alleviates some of the neuroinflammation and neurological deficits involved in the acute phase of TBI. Therefore, inhibition of caspase-1 may alleviate the severity of TBI and improve the efficacy of these other blocking agents.
Additional files
Additional file 1: Table S1. Primers of caspase-1 for RT-qPCR. (DOCX 15 kb) Additional file 2: Figure S1. Caspase-1 mRNA expression level in the primary neuron. The primary neuron was infected with caspase-1 shRNA (m) lentiviral particles (Santa Cruz, sc-29922-V) according to the manufacturer's protocol (Santa Cruz Biotechnology, Inc.). At 72 h after infection, The RNAi efficiency was determined by qRT-PCR. Caspase-1 mRNA was significantly reduced by RNAi. Data were represented as means ± SEM, one representative experiment of four was shown. ***p < 0.001, versus control. (DOCX 24 kb) Additional file 3: Figure S2. Uncropped blots for the pro-form and cleaved fragments of caspase-1 in figures. (DOCX 877 kb) Additional file 4: Figure S3. VX-765 neurotoxicity detected by CKK8 assay. In order to demonstrate the reduction in the pro-form of caspase-1 levels by VX-765 was not an artifact due to direct VX-765 neurotoxicity, a VX-765 alone and a vehicle control for VX-765 treated with neuron. The primary neuron was seeded in 96-well according to the primary neuronal cultures. At 10 days, cells were supplied with a fresh medium and treated with VX-765 (0, 1, 5, and 10 μM). After 24 h of cultivation respectively, then 10 μl of the CCK8 reaction solution (Dojindo Laboratories) was added to each well. After 4 h incubation at 37°C in the 5% CO2 incubator, the absorbance at 450 nm was measured by Thermo MK3 (Thermo Scientific). The primary neuron was supplied with a fresh medium and treated with VX-765 at 0, 1, 5, and 10 μM (DMSO as the menstruum). After 24 h of cultivation respectively, neurotoxicity was detected by CCK8 assay. Data were represented as means ± SEM, one representative experiment of five was shown. (DOCX 43 kb) | 2018-02-20T04:22:32.419Z | 2018-02-19T00:00:00.000 | {
"year": 2018,
"sha1": "20b0207ce0defadfe43cf9477cff67852220594f",
"oa_license": "CCBY",
"oa_url": "https://jneuroinflammation.biomedcentral.com/track/pdf/10.1186/s12974-018-1083-y",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "20b0207ce0defadfe43cf9477cff67852220594f",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
258052077 | pes2o/s2orc | v3-fos-license | The therapeutic window of intravenous immunoglobulin (IVIG) and its correlation with clinical outcomes in Kawasaki disease: a systematic review and meta-analysis
Background The optimal therapeutic window to start intravenous immunoglobulin (IVIG) for Kawasaki disease (KD) is highly debatable. We aimed to summarize the existing literature to evaluate the therapeutic window of IVIG treatment and its correlation with clinical outcomes in KD patients. Methods We searched the databases from inception to August 26, 2022, without language restrictions. The primary outcomes were initial IVIG resistance and coronary artery lesions (CALs) in acute phase. Secondary outcome was CALs during 1–2 months of follow-up. Results 27 studies involving 41,139 patients were included in this study. Very low-quality evidence showed that the earlier IVIG treatment within 4 days had a higher IVIG-resistance rate (RR, 1.80; 95% CI, 1.50–2.15; P < .00001; I2 = 75%) than the late treatment. Very low-quality evidence showed that IVIG treatment for more than 7 days was associated with a higher risk of CALs in acute phase(RR, 0.57; 95% CI, 0.40–0.80; P = .001; I2 = 76%). There was a lower risk of CALs during 1–2 months follow-up for those who started IVIG administration within 10 days from the onset. Conclusions Overall, IVIG treatment within 7 days of illness seems to be the optimal therapeutic window of IVIG. IVIG treatment within 7 days is found to be effective for reducing the risk of coronary artery lesions and cardiac sequelae in KD patients. The early IVIG treatment within 4 days should be vigilant for the IVIG resistance although large multi-center randomized trials with well design are needed. Supplementary Information The online version contains supplementary material available at 10.1186/s13052-023-01451-6.
Background
Kawasaki disease (KD) is an acute, febrile vasculitis of unknown aetiology that primarily affecting children < 5 years of age [1]. Major classical features of KD include fever, conjunctivitis, erythema of the lips, oral mucosa, changes in extremities, rash, and cervical lymphadenopathy. KD has been reported in more than 60 countries worldwide since first described in Japan in 1974 [2]. The incidence of KD is increasing, and the highest relative risk is in Asian children, especially in eastern Asia [3]. KD has surpassed acute rheumatic fever as the most common cause of acquired heart disease in developed countries [4].
The most common sequel caused by KD is coronary artery lesions (CALs), including coronary artery dilatation and coronary artery aneurysm (CAA) [5]. CAA can develop in around 25% of untreated KD children [1]. Intravenous immunoglobulin (IVIG) in conjunction with aspirin remains the first-line drug therapy for KD patients [1]. The mortality rates of KD have been markedly decreased since the IVIG therapy was introduced in 1983 [6], and the incidence of CALs in those treated appropriately is dropped to 4% [1]. However, several latest clinical guidelines on the optimal timing of IVIG administration and if IVIG can be given earlier remain inconclusive [1,7,8]. Some previous retrospective studies reported that IVIG treatment within 5 days of illness reduces the risk of cardiac sequelae, suggesting the efficacy of the earlier initiation of IVIG treatment [9][10][11]. On the contrary, other studies reported that IVIG use at ≤ 4 days of illness had no benefit in preventing coronary complications but was instead associated with increased IVIG resistance [12,13]. Therefore, the association between the earlier timing of IVIG administration of disease onset and the risk for IVIG unresponsiveness remains controversial and debatable [14].
Given this background, there is a need to qualitatively and critically appraise available evidence and evaluate where gaps exist. Thus, we conducted a systematic review and meta-analysis to investigate the optimal timing of IVIG administration in KD patients by comparing the clinical outcomes of early and conventional IVIG treatment.
Methods
This systematic review and meta-analysis aimed to critically appraise the therapeutic window of IVIG treatment and its correlation with clinical outcomes in KD patients. We performed this systematic review based on Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [15]. The protocol was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO; registration number: CRD42022356138).
Data sources and search strategy
We systematically searched databases, including PubMed (MEDLINE), Cochrane Library (CENTRAL), Embase, Web of Science, and Chinese Biomedical Literature database (CBM) from database inception to August 26, 2022, without language restrictions. An e-mail alert was received weekly based on a previously developed search strategy saved in PubMed for any new potential studies. For completed and ongoing trials, we also searched trial registries, including ClinicalTrials.gov and the international Clinical Trial registry platform (ICTRP). Finally, we hand-searched reference lists of identified articles for further eligible trials. A search strategy was built based on MeSH terms and free-text words based on the following terms: Mucocutaneous Lymph Node Syndrome, Kawasaki disease, intravenous immunoglobulin, and gamma-Globulins. A detailed search strategy can be seen in Supplementary Material, Appendix S1.
Eligibility criteria
Studies that met all of the following criteria were included: (a) Patients with KD (diagnosed using any recognized diagnostic criteria) undergoing early (≤ day 4), conventional (days 5-10) or late (>day 10) IVIG treatment; (b) study design: randomized controlled trials (RCTs), cross-sectional study, cohort, case-control; (c) with adequate data on necessary basic characteristics and outcomes.
Exclusion criteria were as follows: (a) KD patients complicated with severe infection, allergy, autoimmune diseases, or collagen disease; (b) case reports, reviews, guidelines, opinions, editorials, letters, animal studies, and comments.
Study selection
Study selection was managed using Covidence software (https://www.covidence.org/). Two independent reviewers (ZZ and XMZ) evaluated articles for potential inclusion by screening titles and abstracts. The full texts of those identified as being relevant would then be assessed to determine eligibility for final inclusion. The results would be discussed between each assessment to reach a consensus on the interpretation of the inclusion criteria. Any disagreements regarding study eligibility would be resolved by consensus, and a third reviewer (WT) was consulted if necessary. If the information required to assess eligibility were unavailable or unclear, the relevant study authors would be contacted for clarification. Duplicate publications were identified and removed using EndNote software version X7 (Clarivate Analytics). The identified publications were analyzed using criteria based on largest sample size, the maximum correspondence with the inclusion criteria, and a minimal risk of bias.
Data extraction
Two authors (JL and CMY) independently extracted data in duplicate using a predefined data extraction form (Excel, Microsoft Corporation, USA). We did a thorough pilot test before the formal data collection to ensure consistency in the data extraction process and that all necessary information was collected. The following data were extracted from the studies selected for inclusion, as follows: (a) general characteristics of the study (author names, publication date, study design, study period, setting, funding and country); (b) patient demographic features (including sample size, groups, age, gender, the proportion of complete KD cases, recurrent cases and patients who received IVIG treatment); (c) outcome measures and analyses; (d) IVIG treatment protocol and definitions of IVIG resistance.
The primary outcomes were the incidence of initial IVIG resistance and CALs in the acute phase. The secondary outcome was CALs during 1-2 months follow-up.
Risk of bias and grade certainty assessment
Two reviewers (GFP and HLL) independently assessed the risk of bias based on articles and protocols. Any discrepancies were solved by discussion and intervention of a third reviewer (QT) whenever necessary. We used the Risk of Bias Tool 2.0 (RoB 2.0) developed by the Cochrane Collaboration for randomized trials [16]. This tool evaluates five domains of bias: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported results. The RoB 2.0 for each of the 5 domains is described as low, some concerns, or high. The risk of bias was assessed using the Newcastle-Ottawa scale (NOS) for observational cohort and case-control studies [17]. There were three grouping items as follows: selection, comparability, and exposure/outcomes. A study can be awarded a maximum of one star for each numbered item within the selection and outcome categories. A maximum of two stars can be given for comparability. More stars are equalling lower risk. The NOS scale awards a maximum of nine stars to each study. The summary score given to studies can be categorized as poor (0-3 points), fair (4-6 points), or high (≥ 7 points) quality. The quality of evidence of the included studies was assessed according to The Grading of Recommendations Assessment, Development and Evaluation (GRADEpro Guideline Development Tool, available online at gradepro.org) [18].
Statistical analysis
Results were assessed using forest plots and presented as RRs for the binary outcomes. Heterogeneity was assessed by I² tests, and substantial heterogeneity was defined as I² greater than 50%. An inverse variance, fixed-effects method was used if I² was less than 50%, and a randomeffects method was used if I² was 50% or greater. We performed subgroup analyses based on research types, study location and type of CALs (coronary artery dilation or CAA). Sensitivity analysis was performed by evaluating the pooled estimate after omitting a study each time. Publication bias was assessed by visually inspecting a funnel plot and quantified by the Egger test. Statistical analyses were conducted using ReviewManager version 5.2 (RevMan; Copenhagen: The Nordic Cochrane Center, the Cochrane Collaboration, 2014, London, UK) and Stata statistical software version 14.0 (StataCorp, College Station, Texas) software. A two-sided p-value less than 0.05 was set as the threshold for statistical significance.
Literature search and study selection.
We identified 8914 articles by the initial database search, including 201 from Pubmed, 4412 from Embase, 246 from the Cochrane database, 3710 from Web of Science, and 345 from CBM. After removing duplicates (n = 2692) and articles based on screening of the title and abstract screening (n = 6179), 43 full-text articles were assessed for eligibility. Furthermore, two articles were identified through citation searching. After screening of full texts, 18 articles that did not meet the inclusion criteria were excluded (reasons for exclusion are summarized in Supplementary Material, Appendix S2). A total of 27 studies [9][10][11][12][13] met the inclusion criteria and were analyzed for the systemic review and meta-analysis ( Fig. 1).
Assessments of risk of bias
All observational studies were of fair-to-high quality, and all randomized studies were at some concerns. The results for each quality assessment by the study are presented in Supplementary Material, Appendix S4 and Supplementary Material, Appendix S5.
Certainty of the evidence
We rated the certainty of the evidence of primary outcomes using by GRADE approach. The analysis results showed that the overall quality of evidence was very low, mainly due to more observational designs we included and inconsistency (I 2 > 50%). The outcomes and assessments were presented as a summary of findings in Supplementary Material, Appendix S6.
Sensitivity analysis
We conducted several sensitivity analyses to confirm the robustness of our findings. In the sensitivity analysis, the total findings did not change significantly. All sensitivity analyses indicated that the results of the meta-analysis were robust and the heterogeneity reduced to some extent (Supplementary Material, Appendix S8 and S9).
Subgroup analysis
Subgroup analyses were performed based on research types and the study location (Supplementary Material, Appendix S10 and S11). For the subgroup of research types, the pooled analysis of randomized trials, which had a broader confidence interval, smaller sample size and higher heterogeneity (RR, 1.97; 95% CI, 0.76-5.07; P = .16; I 2 = 81%), showed that early initiation of IVIG treatment within 4 days was not associated with a higher risk of initial IVIG resistance. The pooled analysis from 13 observational studies showed that early IVIG treatment (≤ 4 days ) was associated with a higher risk of initial IVIG resistance (RR, 1.74; 95% CI, 1.45-2.09; P < .00001; I 2 = 74%), but there was evidence of heterogeneity across the studies. The subgroup analyses of CALs in the acute phase showed similar findings based on research types. There were no significant differences in the occurrence of CALs (≤ 4 days VS > 4 days) in both RCTs (RR, 0.71; 95% CI, 0.38-1.33; P = .28; I 2 = 0%) and observational studies (RR, 0.94; 95% CI, 0.78-1.14; P = .51; I 2 = 74%).
Publication Bias
Funnel plots were generated for outcomes with 10 or more studies to evaluate publication bias (Supplementary Material, Appendix S16 and Supplementary Material, Appendix S17). The Egger test did not demonstrate significant publication bias for initial IVIG resistance (4-day cut-off point: P = .36, 10-day cut-off point: P = .90) and CALs in the acute phase (4-day cut-off point: P = .25, 10-day cut-off point: P = .57).
Discussion
This systematic review and meta-analysis of 27 studies, including 41,139 participants, comprehensively summarized the available published literature and focused on the optimal therapeutic window of IVIG treatment across a wide range of populations. There are several main findings of our study. First, very low-quality evidence showed that the early IVIG treatment (≤ 4 days) had a higher IVIG resistance rate than the late treatment (≥ 5 days). Further subgroup analysis with heterogeneity from observational studies showed similar results. However, the pooled results of the RCTs subgroup showed that early IVIG treatment within 4 days was not associated with a higher risk of initial IVIG resistance. In contrast to the studies performed in China and Canada that found no significant difference in IVIG resistance, studies conducted in Japan and South Korea showed an increased risk of IVIG resistance with early IVIG treatment within 4 days. Second, early IVIG treatment within 7 days was associated with a lower occurrence of CALs in the acute phase in the overall estimate. Third, for the secondary outcomes, there was a lower risk of CALs during 1-2 months follow-up for those who started IVIG administration within 10 days from the onset.
IVIG is the first-line treatment of KD with well-established therapeutic effects in preventing coronary artery abnormalities [41]. However, the criteria for when to provide IVIG are unclear and differ from the latest guidelines. The latest 2017 American Heart Association (AHA) and 2018 Italian Society of Pediatrics guidelines recommend that IVIG be administered to KD patients within the first 10 days of illness and, if possible, within the first seven days of illness because the rate at which KD patients develop aneurysms increases significantly after the ninth day of illness [1,7]. Similarly, the 2020 Japanese Circulation Society stated that IVIG was most frequently administered on the 5th day of illness [8]. There is no suggestion on the optimal timing of IVIG and if it can be given earlier. Our study showed that IVIG treatment within 4 days of illness was associated with increased initial IVIG resistance, which is in line with the results of a previous meta-analysis [42]. It is worth noting that subgroup analyses for the RCTs revealed that early IVIG treatment (≤ 4 days) was not associated with a higher risk of initial IVIG resistance. Meanwhile, the pooled results from observational studies with heterogeneity showed that early IVIG therapy (≤ 4 days ) was associated with a higher risk of IVIG resistance. Thus, these results must be interpreted cautiously since most of the observational studies were retrospective, with potential selection and information bias. More large, well-conducted RCT evaluating the relationship between the early IVIG treatment and IVIG resistance would answer this question.
In the nationwide surveys of KD in Japan, patients treated early (≤ day 4 of illness) were more likely to require retreatment with IVIG [12]. Several studies conducted in Japan showed similar results [13,21]. On the other hand, recent studies [9,10] showed that Chinese KD patients who had earlier IVIG treatment administration within 4 days might not increase the higher incidence of IVIG resistance. Further subgroup analyses confirmed that unlike the studies performed in China that found no significant difference in IVIG resistance, studies conducted in Japan showed an increased risk in the rate of IVIG resistance with early IVIG treatment within 4 days. One possible explanation is that the early group showed a higher rate of IVIG resistance because there might be more patients with severe inflammation or atypical clinical course, respectively. Another reason for the difference may be that KD's characteristics vary between ethnicities.
The most common sequela of KD is CAL, which is speculated to be caused by acute systemic inflammation [43]. The early prevention of CALs is important to improve outcomes in KD patients, as CALs severely impair the life quality of KD patients. It is not surprising that our results showed that patients treated earlier (≤ 7 days) had a lower rate of occurrence of CALs in the acute phase. Furthermore, KD patients who received IVIG therapy more than 10 days after the onset had a higher incidence of CALs during 1-2 months of follow-up. Our study highlights the importance of early intervention to prevent coronary artery complications in treating KD. To minimize cardiac sequelae, avoiding any delay in IVIG treatment is crucial because earlier inflammatory suppression may contribute to avoiding developing CAA [44].
Strengths and Limitations
Our study had strengths: First, we used standard Cochrane protocols and had the largest cumulative sample size to date compared to the previous reports. Second, we analyzed long-term follow-up data on primary and secondary outcomes and included all IVIG infusion time points. Third, no language limit was applied; we included studies published in English and Chinese. Fourth, the GRADEpro approach was used to rate the certainty of evidence.
This study had several limitations to be noted. First, Most of the articles we included were retrospective observational studies with potential selection and information bias. In addition, the sample size of included RCTs was small, and all were carried out in China. Second, the clinical heterogeneity of studies, particularly the participant characteristics, IVIG treatment protocol, and the diagnostic criteria of IVIG resistance and CALs, was quite varied, potentially leading to substantial heterogeneity. Finally, the certainty of the evidence for all outcomes was very low.
Conclusion
The findings of this systematic review and meta-analysis pooling data from multiple countries demonstrate that IVIG therapy within 7 days of illness may be the optimal therapeutic window of IVIG. IVIG treatment within 7 days is found to be effective for reducing the risk of coronary artery lesions and cardiac sequelae in KD patients. The early IVIG treatment within 4 days should be vigilant for the IVIG resistance although large multi-center randomized trials with well design are needed. | 2023-04-11T13:56:25.161Z | 2023-04-11T00:00:00.000 | {
"year": 2023,
"sha1": "d43226832f692b54c566fe9446196b908e5f2567",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Springer",
"pdf_hash": "d43226832f692b54c566fe9446196b908e5f2567",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
215773042 | pes2o/s2orc | v3-fos-license | Molecular Dynamics Simulations Reveal Canonical Conformations in Different pMHC/TCR Interactions.
The major defense system against microbial pathogens in vertebrates is the adaptive immune response and represents an effective mechanism in cancer surveillance. T cells represent an essential component of this complex system. They can recognize myriads of antigens as short peptides (p) originated from the intracellular degradation of foreign proteins presented by major histocompatibility complex (MHC) proteins. The clonotypic T-cell antigen receptor (TCR) is specialized in recognizing pMHC and triggering T cells immune response. It is still unclear how TCR engagement to pMHC is translated into the intracellular signal that initiates T-cell immune response. Some work has suggested the possibility that pMHC binding induces in the TCR conformational changes transmitted to its companion CD3 subunits that govern signaling. The conformational changes would promote phosphorylation of the CD3 complex ζ chain that initiates signal propagation intracellularly. Here, we used all-atom molecular dynamics simulations (MDs) of 500 ns to analyze the conformational behavior of three TCRs (1G4, ILA1 and ILA1α1β1) interacting with the same MHC class I (HLA-A*02:01) bound to different peptides, and modelled in the presence of a lipid bilayer. Our data suggest a correlation between the conformations explored by the β-chain constant regions and the T-cell response experimentally determined. In particular, independently by the TCR type involved in the interaction, the TCR activation seems to be linked to a specific zone of the conformational space explored by the β-chain constant region. Moreover, TCR ligation restricts the conformational space the MHC class I groove.
Introduction
Cytotoxic T cells recognize and kill virus-infected cells and cancer cells upon T-cell antigen receptor (TCR) interaction with major histocompatibility complex (MHC) class I proteins presenting viral and tumor antigens, respectively [1][2][3]. MHCs class I are composed of one α-chains, which form the binding site for a nine/ten residue-long peptide, and of a non-covalent bound β2 microglobulin ( Figure 1). To recognize pMHC, T cells use a clonally distributed αβ dimer with Ig-like variable domains, Vα and Vβ. Together, Vα and Vβ form the pMHC binding site composed of six loops homologous to antibody complementarity determining regions (CDRs) 1, 2 and 3 [4,5]. CDR1 and CDR2 have limited variability, while CDR3s are hypervariable. VαVβ orientates diagonally relative to Combining experimental and theoretical studies, we recently found that the bound peptide can affect the conformation of the MHC I binding groove, suggesting a different presentation of the antigens, which seems to be related to different CTLs responses [10]. However, in that study we did not provide a modelling of the pMHC interacting with the TCR. To better characterize such a complex molecular network, here we present the first exhaustive computational study of five pMHC-TCR complexes, modelled in a heterogeneous lipid bilayer (see Video S1 in SI, and Figure 1).
Complexes Modelling
Using the crystallographic structures (PDB ID: 4MNQ; PDB ID: 2BNR) the transmembrane alpha helices were built by means of the Modeller Software version 9.19 (Accelerys, San Diego, CA, USA) [11], following the amino acid sequences provided by the Uniprot database [12] (Uniprot entries P01848 and P01850 for the TCR α and β chains respectively; Q9MY51 for the α chain of the HLA-A* 02:01). The heterogeneous lipid bilayer was built by means of the CHARMM-GUI Membrane Builder web software (Harvard, Cambridge, MA, USA) [13], with a composition of: 1-palmitoyl-2-oleoyl-snglycero-3-phosphocholine (POPC) at 90%, phosphatidylinositol (4,5)-bisphosphate (PIP2) at 7% and 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (POPS) at 3%. Such a lipid composition is based on the work of Chavent et al. [14] and Zech et al. [15] to approximate the lipid composition of a mammalian cell membrane [16]. These studies reported an essential role of the PIP2 in receptor activation and TCR triggering. Note that recent findings reported that cholesterol can inhibit T cell activation and thus, this lipid was not included in our membrane model [17,18]. The modelled complexes were then manually inserted in the membrane using the VMD Software version 1.9.3 [19] (University of Ilinois, Champaign, IL. USA). Combining experimental and theoretical studies, we recently found that the bound peptide can affect the conformation of the MHC I binding groove, suggesting a different presentation of the antigens, which seems to be related to different CTLs responses [10]. However, in that study we did not provide a modelling of the pMHC interacting with the TCR. To better characterize such a complex molecular network, here we present the first exhaustive computational study of five pMHC-TCR complexes, modelled in a heterogeneous lipid bilayer (see Video S1 in SI, and Figure 1).
Complexes Modelling
Using the crystallographic structures (PDB ID: 4MNQ; PDB ID: 2BNR) the transmembrane alpha helices were built by means of the Modeller Software version 9.19 (Accelerys, San Diego, CA, USA) [11], following the amino acid sequences provided by the Uniprot database [12] (Uniprot entries P01848 and P01850 for the TCR α and β chains respectively; Q9MY51 for the α chain of the HLA-A* 02:01). The heterogeneous lipid bilayer was built by means of the CHARMM-GUI Membrane Builder web software (Harvard, Cambridge, MA, USA) [13], with a composition of: 1-palmitoyl-2-oleoyl-snglycero-3-phosphocholine (POPC) at 90%, phosphatidylinositol (4,5)-bisphosphate (PIP2) at 7% and 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (POPS) at 3%. Such a lipid composition is based on the work of Chavent et al. [14] and Zech et al. [15] to approximate the lipid composition of a mammalian cell membrane [16]. These studies reported an essential role of the PIP2 in receptor activation and TCR triggering. Note that recent findings reported that cholesterol can inhibit T cell activation and Cells 2020, 9, 942 3 of 15 thus, this lipid was not included in our membrane model [17,18]. The modelled complexes were then manually inserted in the membrane using the VMD Software version 1.9.3 [19] (University of Ilinois, Champaign, IL, USA).
MD Simulations
The systems were solvated with the TIP3P water model [20] and neutralized with Na+ and Clions at physiological concentration (0.15 M). The exceeding solvent was manually removed, excluding the water molecules within a range of 2 Å from lipids. An energy minimization step was performed using the steepest descent algorithm without position restraints. After the minimization, a series of equilibration steps were performed: (1) a NPT equilibration of 40 ps was run to allow the packing of the lipids around the protein, using an integration time step of 0.2 fs; then the NPT equilibration was extended until 2 ns. (2) an NVT equilibration of 40 ps was performed and then extended until 4 ns, increasing the time step at 1 fs. (3) A latest NPT simulation 500 ns long was run with a time step of 2 fs. The V-rescale thermostat [21] and the Parrinello-Rahman barostat [22] were used with a τT = 0.1 ps and a semi-isotropic coupling with a τp = 5 ps, respectively. The temperature, considering the melting point of the lipid composition, was kept constant at 305 K. The electrostatic interactions were calculated using the particle mesh Ewald method [23] with a cut-off of 1.2 nm. A cut off of 1.2 nm was used for the van der Waals interactions. The simulations were performed using the additive for lipids CHARMM-36 force field [24,25], and the Gromacs Software version 2018.1 (University of Groningen, Groningen, Netherlands) [26]. For each system, we performed a Molecular Dynamics simulation lasting 500 ns, resulting in an aggregated time of four microseconds.
Structural Analysis
The root mean square deviation (RMSD) is a statistical measure of the average distance between a selected group of atoms, with respect to a reference structure. According to the following equation (Equation (1)): where r i (t) is the position of the atom i at the time t, N is the total number of atoms in the group considered and r i 0 is the position of atom i in the reference structure. The RMSD calculation was performed considering the alpha carbons, choosing the first frame of the simulation as reference. The root mean square fluctuation (RMSF) is a statistical measure of the deviation between the position of the atom i (or a group of atoms, e.g., a residue) r i (t), and the initial structure r i 0 considering the time interval T (Equation (2)): Such a measure allows to detect and quantify the displacement of the different protein regions along the MD simulation.
Essential Dynamics
The essential dynamics technique is a statistical method based on the principal component analysis [27]. Briefly, the covariance matrix of the atomic positions is built from the MD simulations on a selected group of atoms (usually C-alpha). From the diagonalization of such a matrix, a set of eigenvectors and associated eigenvalues is obtained. The eigenvectors represent the principal motion directions of the system and, therefore, they are used to describe the "essential" protein modes, which often represent the functional ones. In this way, the fastest motions present in the simulations, which describe biologically not relevant motions (i.e., vibrations), are excluded making possible to represent the protein dynamics in a reduced space-as defined by the eigenvectors-which approximate well the overall molecular motions.
The essential subspace, describing the overall motion, is mostly confined within the first 2 eigenvectors, in the case of study. Combining two (or more) trajectories of different systems (having equal alpha carbons numbers) it is possible to obtain common eigenvectors defining the subspace explored by the different proteins. The projections of the MD trajectory on the first 2 eigenvectors (i.e., principal components), allow the comparison of the conformations assumed by the proteins during the simulation. We compared the conformational behavior of the pMHC-TCR systems in study, analyzing the entire complex and then the single regions.
The gmx covar and gmx anaeig tools of the Gromacs Software 2018.1 [26] were used to build the covariance matrix and to calculate the 2d projections with respect to the first 2 eigenvectors.
Cross-Correlation Matrix
The cross-correlation is the correlation between the entries of two random vectors X and Y, while the correlations of a random vector X are the correlations between the entries of X itself, those forming the correlation matrix of X. In such a matrix, the correlations of the various temporal instances of X with itself are known as autocorrelations, and they are arranged on the matrix diagonal. Outside the diagonal, there are the cross-correlations between X and Y across the time, which assume the value between +1 and −1. We considered that the regions are correlated when such a value is greater than 0.75, and they are anti-correlated from −1 to −0.25. The cross-correlation matrix was computed by means of the Bio3d package of the R Software version 3.5.3 [28,29] (University of Michigan, Ann Arbor, MI, USA).
Construction of the Zernike Descriptor
For each MD frame we calculated the molecular surface and the electrostatic potential by means of PDB2PQR [30] and Bluues [31] software package. Then, we extracted by means of a voxelization procedure the three Zernike 3D functions (3DZD) [32,33], representing the shape, the positive electrostatics and the negative electrostatics of the selected region, i.e., the binding groove. Such a procedure was recently implemented and applied in our recent work on similar systems [34,35].
Network Analysis
To investigate the topological and structural properties of the different systems, we have adopted a graph theory approach. To this end, we have selected about 100 frames for each simulation and each structure has been represented as a network, where each residue is a node (or vertex) of the graph. Two nodes are joined by a link (or edge) if the distance between the alpha carbon of the corresponding residues is lower than a threshold (6 Å). To analyze the contribution of each residue within the protein network, we have considered two local descriptors: "closeness centrality" and "degree". The "closeness centrality" of a node is a local network parameter, measuring the mean number of steps required to reach any other protein residue (or node), starting from that node. The node "degree" is one of the most used topological descriptors in graph-theory based approach, counting the number of links of that node (in other words the node degree is the number of residues closer than the cutoff to the examined residue). To highlight the network characteristics from a structural point of view, each residue of the complex was colored in accordance with the corresponding value of the network parameter. For both the network parameters, a boxplot was used 164 to represent the distribution of the values for each residue [36].
Membrane Analysis
The membrane thickness was computed by means of the GridMAT-MD program [37]. Such a tool splits the lipid bilayer into a grid. To obtain a better resolution, a grid of 100 × 100 points was selected to perform the analysis. The distance between the two bilayers was calculated using the gmx mindist tool of the Gromacs Software [26].
Results
Due to the availability of kinetic data present in the literature [38][39][40][41][42][43], we modelled the HLA-A*02:01 containing the peptide ESO9C (the tumor antigen NY-ESO157-165 fragment-SLLMWITQC [40,42]), the mutated ESO4D (a.a. sequence SLLDWITQV [42]), the hERT540-548 antigen (a human telomerase reverse transcriptase epitope-sequence ILAKFLHWL [38,41,43]) and the mutated analogous (a.a. sequence ILAKFAHWL, named below as hTERT-6A [41]). The complexes with the NY-ESO-derived peptides were modelled in presence of the 1G4 TCR, while the hTERT derivatives with the ILA1 TCR and the ILA1α1β1. This latter TCR is mutated in the CDR2β and CDR3α loops, as shown in Table 1. In addition, the A*02:01-ESO9C, the 1G4 and ILA1α1β1 were separately simulated in a membrane environment, as reference. The simulation of the complexes in a membrane environment makes it possible to study the effect of the lipidic phase on the pMHC-TCR dynamical behavior, which was recently indicated to play an important role in the receptor activation [44]. Table 1. The simulated complexes and the corresponding experimental binding affinity data [41,42]. The MHC class I is the HLA-A*02:01. The last two lines report the sequence differences between the ILA1 and the ILA1α1β1 (the mutated residues are reported in red).
Complex with HLA-A*0201
Peptide Sequence
Conformational Analysis
Due to the complexity of the simulated systems, the RMSDs of the pMHC-TCR complexes reach a plateau after several ns ( Figure S1), and, thus the first 50 ns were removed from the analysis. To understand the complex dynamics behavior of the simulated systems, we first looked at the RMSF analysis, which provides a simple representation of the fluctuation behavior ( Figure 2). Such an analysis shows that the more fluctuating regions are almost conserved, despite relevant differences in the sequences between the simulated systems ( Figure S2). A more detailed analysis of the RMSF on specific regions, points out that the 1G4 construct is more affected by the peptide changes, especially in the Vβ region ( Figure S2, panel C).
analysis, which provides a simple representation of the fluctuation behavior ( Figure 2). Such an analysis shows that the more fluctuating regions are almost conserved, despite relevant differences in the sequences between the simulated systems ( Figure S2). A more detailed analysis of the RMSF on specific regions, points out that the 1G4 construct is more affected by the peptide changes, especially in the Vβ region ( Figure S2, panel C).
Collective Motions
The overall motions of the simulated systems have been compared by means of the Essential Dynamics method. Such an analysis shows that a large part of the motions can be represented by the first few principal eigenvectors. In fact, the first two eigenvectors describe about 50% of the system total variance, making possible to describe the main conformational behavior in the subspace described by the first two eigenvectors. Therefore, we firstly projected the conformations explored by the bound and unbound states of the alpha carbon of the 1G4 and the ILA1α1β1 complexes and then all the bound simulations were compared to each other.
Comparison between the Bound and Unbound States
Combining the MD trajectories of the 1G4 systems, we performed the analysis on the TCR alpha carbons, and, separately, on the variable and constant regions. As shown in Figure S3, no large conformational rearrangements occur in the time range of the MD simulations, and the overall motions of the bound form of the 1G4 TCR are comparable with respect to the unbound one. On the contrary, the Vβ regions of the bound systems project on different regions with respect to the unbound one (Figure 3a,b). Here the CDR2β and CDR3β regions are the movement-determining sites. This is in line with the work of Housset et al. [45], where a significant conformational change of the CDR3β loop, necessary for the binding to the pMHC was found. Interestingly, the Cα region projections show a larger movement of such a region in the bound state, enclosing conformational rearrangements, which are different for the 1G4-ESO9C and ESO4D ( Figure S3, panel e,f). This observation seems in line with the conformational change previously observed in the Cα region of the TCR by means of fluorescence spectroscopy [46]. Even more interesting, the Cβ region projections are quite overlapped in the essential subspace, suggesting a similar conformational behavior of the TCR (Figure 3). The projections of the TCR conformations sampled by the ILA1α1β1 simulations ( Figure S3b), show that such an unbound state explores different conformation with respect to the corresponding bound systems, e.g., ILA1α1β1-hTERT and hTERT6A. The analysis of the variable regions underlines the presence of different conformations of the Vα between the bound and the unbound state ( Figure S3d), and even more in detail, a conformational change occurs in the ILA1α1β1-hTERT. On the contrary, the Vβ conformations explore the same region of the conformational space ( Figure 3c). Finally, the constant region projections discriminate between the bound and unbound ILA1α1β1 TCR (Figure 3d). Cells 2020, 9, 942 7 of 15
Comparison Between the Bound States
The essential dynamics of the five pMHC-TCR complexes show that the first eigenvector is able to discriminate between the two different TCR types. Noteworthy, such an eigenvector principally describes the motions of the variable and the constant regions, suggesting that differences in TCR overall motions are essentially due to the dynamical behavior of such regions (Figure 4a and Figure S4). Interesting, all the Cβ regions conformations explore similar areas on the essential subspace, except for the ILA1 which has no binding affinity for the HLA-A*0201:hTERT-6A (Figure 5d). Finally, the binding groove alpha carbons projections (residues 1-175 of the HLA-A*0201 alpha chain) reveal that similar regions are explored by all the simulated systems (Figure 4b). Similar results were obtained by i) the RMSD matrix of the binding groove of a reference simulation with respect to another ( Figure S5) and ii) by means of Zernike polynomials-based approach (see method). The latter, which essentially describes the pockets shapes, confirms the geometrical shape similarity for all the binding grooves ( Figure S6). Contrary to our previous results [10], this might suggest a quite-rigid conformation of the pMHC pocket, which seems to be unaffected by the sequence of the peptide.
Comparison between the Bound States
The essential dynamics of the five pMHC-TCR complexes show that the first eigenvector is able to discriminate between the two different TCR types. Noteworthy, such an eigenvector principally describes the motions of the variable and the constant regions, suggesting that differences in TCR overall motions are essentially due to the dynamical behavior of such regions (Figure 4a and Figure S4). Interesting, all the Cβ regions conformations explore similar areas on the essential subspace, except for the ILA1 which has no binding affinity for the HLA-A*0201:hTERT-6A (Figure 5d). Finally, the binding groove alpha carbons projections (residues 1-175 of the HLA-A*0201 alpha chain) reveal that similar regions are explored by all the simulated systems (Figure 4b). Similar results were obtained by (i) the RMSD matrix of the binding groove of a reference simulation with respect to another ( Figure S5) and (ii) by means of Zernike polynomials-based approach (see method). The latter, which essentially describes the pockets shapes, confirms the geometrical shape similarity for all the binding grooves ( Figure S6). Contrary to our previous results [10], this might suggest a quite-rigid conformation of the pMHC pocket, which seems to be unaffected by the sequence of the peptide.
Peptide Interactions
The MD simulations also allow us to quantitatively investigate the interaction between the peptide and the two macromolecular units. The analysis of the hydrogen bonds (Table 2), involving the four peptides, confirms the presence of two anchor sites, located at the two ends of the peptides, which fix the ligand to the HLAs ( Figure 6). In fact, the hydrogen bonds between the same residues of the binding groove and the residue P1-P2 and P8-P9 of the peptide are always maintained. Curiously, these h-bonds were found in our previous work on two HLA-B*27 subtypes [10], suggesting that, independently by the peptide sequence and the HLA type, the two ends of the ligand establish a similar interaction with conserved pocket residues. On the other hand, the hydrogen bonds between the peptide and the TCRs are specific of the complexes ( Table 2 and Table S1): in the case of the 1G4-ESO9C, the peptide forms one and three h-bonds with the alpha and beta chains. The mutations 4D-9V, in the ESO peptide, increase the number of h-bonds between the alpha and beta
Peptide Interactions
The MD simulations also allow us to quantitatively investigate the interaction between the peptide and the two macromolecular units. The analysis of the hydrogen bonds (Table 2), involving the four peptides, confirms the presence of two anchor sites, located at the two ends of the peptides, which fix the ligand to the HLAs ( Figure 6). In fact, the hydrogen bonds between the same residues of the binding groove and the residue P1-P2 and P8-P9 of the peptide are always maintained. Curiously, these h-bonds were found in our previous work on two HLA-B*27 subtypes [10], suggesting that, independently by the peptide sequence and the HLA type, the two ends of the ligand establish a similar interaction with conserved pocket residues. On the other hand, the hydrogen bonds between the peptide and the TCRs are specific of the complexes ( Table 2 and Table S1): in the case of the 1G4-ESO9C, the peptide forms one and three h-bonds with the alpha and beta chains. The mutations 4D-9V, in the ESO peptide, increase the number of h-bonds between the alpha and beta
Peptide Interactions
The MD simulations also allow us to quantitatively investigate the interaction between the peptide and the two macromolecular units. The analysis of the hydrogen bonds (Table 2), involving the four peptides, confirms the presence of two anchor sites, located at the two ends of the peptides, which fix the ligand to the HLAs ( Figure 6). In fact, the hydrogen bonds between the same residues of the binding groove and the residue P1-P2 and P8-P9 of the peptide are always maintained. Curiously, these h-bonds were found in our previous work on two HLA-B*27 subtypes [10], suggesting that, independently by the peptide sequence and the HLA type, the two ends of the ligand establish a similar interaction with conserved pocket residues. On the other hand, the hydrogen bonds between the peptide and the TCRs are specific of the complexes ( Table 2 and Table S1): in the case of the 1G4-ESO9C, the peptide forms one and three h-bonds with the alpha and beta chains. The mutations 4D-9V, in the ESO peptide, increase the number of h-bonds between the alpha and beta chains respectively. Moreover, the introduction of mutations in the CDR3α region in ILA1 determines the formation of a larger number of h-bonds, whereas the mutations in the CDR2β remove such bonds. However, in all the pMHC-TCR complexes the h-bonds formed between the peptide and the TCR are located in the central region (P4-P5) of the peptide. On the contrary, in the 1G4 simulations the solvent exposure per residue shows similar behavior for all the peptides (Table S2). Table 2. Hydrogen bonds between the peptide and the binding groove. The a.a. residues involved in the interaction are listed on the left. The residue of the peptide which makes the h-bonds is labelled with "P", and the associate number denotes its position in the peptide sequence. Cells 2020, 9, 942 9 of 15 chains respectively. Moreover, the introduction of mutations in the CDR3α region in ILA1 determines the formation of a larger number of h-bonds, whereas the mutations in the CDR2β remove such bonds. However, in all the pMHC-TCR complexes the h-bonds formed between the peptide and the TCR are located in the central region (P4-P5) of the peptide. On the contrary, in the 1G4 simulations the solvent exposure per residue shows similar behavior for all the peptides (Table S2). Table 2. Hydrogen bonds between the peptide and the binding groove. The a.a. residues involved in the interaction are listed on the left. The residue of the peptide which makes the h-bonds is labelled with "P", and the associate number denotes its position in the peptide sequence.
Network and Cross-Correlation Matrix Analysis
The cross-correlation (CC) maps show a main common coupling behavior between the main regions of the complexes (Figure 7). In fact, to a different extent, the Cα regions show an anticorrelation with the Cβ ones and a significant correlation between the α chain of the HLA and the Cα, Vβ, and Cβ is observed in all the systems. In the case of ILA1, the sequence changes do not affect the main correlation pattern, being that the extent of such correlation only slightly increased upon the mutations. Concerning the 1G4 systems, the modification of two a.a. residues in the peptide essentially decreases the extent of the motion correlation and anticorrelation.
As expected, the network analysis confirms that the regions of the TCR and pMHC directly interacting mainly contribute to the protein interaction network ( Figure S7).
Network and Cross-Correlation Matrix Analysis
The cross-correlation (CC) maps show a main common coupling behavior between the main regions of the complexes (Figure 7). In fact, to a different extent, the Cα regions show an anticorrelation with the Cβ ones and a significant correlation between the α chain of the HLA and the Cα, Vβ, and Cβ is observed in all the systems. In the case of ILA1, the sequence changes do not affect the main correlation pattern, being that the extent of such correlation only slightly increased upon the mutations. Concerning the 1G4 systems, the modification of two a.a. residues in the peptide essentially decreases the extent of the motion correlation and anticorrelation.
Interaction Energies
The interaction energies between selected regions of the complexes-involving the peptide and/or the binding groove-have been analyzed ( Figure 8). As shown in Figure 8a, the distributions of such energy profiles all show favorable interactions between these regions and comparable values. The sole relevant difference was observed in the TCR-binding groove interaction, where the A*0201:hTERT-6A:ILA1 shows a sharp peak of the distribution in an energy range significantly higher than the other complexes. Interestingly, experimental data reports that no binding is observed in such a system. It is also worth noting that the single a.a. mutation provides a remarkable shift of As expected, the network analysis confirms that the regions of the TCR and pMHC directly interacting mainly contribute to the protein interaction network ( Figure S7).
Interaction Energies
The interaction energies between selected regions of the complexes-involving the peptide and/or the binding groove-have been analyzed ( Figure 8). As shown in Figure 8a, the distributions of such energy profiles all show favorable interactions between these regions and comparable values. The sole relevant difference was observed in the TCR-binding groove interaction, where the A*0201:hTERT-6A:ILA1 shows a sharp peak of the distribution in an energy range significantly higher than the other complexes. Interestingly, experimental data reports that no binding is observed in such a system. It is also worth noting that the single a.a. mutation provides a remarkable shift of the interaction energies as observed in both the 1G4 and ILA1α1β1 systems. The pocket electrostatic similarity computed by means of the Zernike Descriptor ( Figure S8) shows a different electrostatic behavior of the pHLA-A*0201 interacting with ILA1. This might suggest that the TCR induces a modification of the electrostatic properties of the pMHC. Cells 2020, 9, 942 11 of 15 the interaction energies as observed in both the 1G4 and ILA1α1β1 systems. The pocket electrostatic similarity computed by means of the Zernike Descriptor ( Figure S8) shows a different electrostatic behavior of the pHLA-A*0201 interacting with ILA1. This might suggest that the TCR induces a modification of the electrostatic properties of the pMHC.
Membrane Behavior
To characterize the membrane properties, we measured the thickness of the lipid bilayers and the evolution of the distance between the two bilayers-hosting the pMHC and the TCR-along the trajectories. For the calculation of the membrane thickness (see methods for details), we compare the thickness variation between the two halves of each trajectory. As shown in Figure S9-S10, no substantial differences were observed for such a property, i.e., the values are all within an interval of 3/4 nm. The distances between the two bilayers-computed by selecting the headgroup of each lipid ( Figure S11) show a limited fluctuation around an average value.
Discussion
To characterize the conformational behavior of the interaction between the pMHC class I with the TCR, we performed an extended set of Molecular Dynamics simulations of five pMHC-TCR complexes, modelled in a heterogenous lipid environment. The chosen systems consist of the same MHC class I (HLA-A*02:01) binding to different peptides, which are recognized by three different TCRs: 1G4, ILA1 and ILA1α1β1. Such complexes differ from each other in the binding affinity reported in previous experimental studies, and all the systems present at least a mutation in the peptide or in the TCR sequence. In addition, we performed MDs of the unbound forms of the HLA-A*02:01, 1G4 and ILA1α1β1, to highlight the effects of the TCR interaction on the complex structuraldynamical behavior. To the best of our knowledge, this is the first computational study considering different pMHC-TCR complexes simulated in a lipid bilayer.
Membrane Behavior
To characterize the membrane properties, we measured the thickness of the lipid bilayers and the evolution of the distance between the two bilayers-hosting the pMHC and the TCR-along the trajectories. For the calculation of the membrane thickness (see methods for details), we compare the thickness variation between the two halves of each trajectory. As shown in Figure S9 and S10, no substantial differences were observed for such a property, i.e., the values are all within an interval of 3/4 nm. The distances between the two bilayers-computed by selecting the headgroup of each lipid ( Figure S11) show a limited fluctuation around an average value.
Discussion
To characterize the conformational behavior of the interaction between the pMHC class I with the TCR, we performed an extended set of Molecular Dynamics simulations of five pMHC-TCR complexes, modelled in a heterogenous lipid environment. The chosen systems consist of the same MHC class I (HLA-A*02:01) binding to different peptides, which are recognized by three different TCRs: 1G4, ILA1 and ILA1α1β1. Such complexes differ from each other in the binding affinity reported in previous experimental studies, and all the systems present at least a mutation in the peptide or in the TCR sequence. In addition, we performed MDs of the unbound forms of the HLA-A*02:01, 1G4 and ILA1α1β1, to highlight the effects of the TCR interaction on the complex structural-dynamical behavior. To the best of our knowledge, this is the first computational study considering different pMHC-TCR complexes simulated in a lipid bilayer.
pMHC Influence on the TCRs Conformations
As expected, the presence of the pMHC affects the conformational behavior of the TCR in all the simulated systems. A more detailed analysis points out that this is verified in all the TCR regions, except for the Cβ ones which are quite unaffected by the presence of the HLA. The comparison between the different systems shows that the different TCR types can be discriminated by the explored conformations basin. In fact, the principal component analysis clearly demonstrates that the 1G4 and the ILA1α1β1 MD structures sample distinct regions of the conformational space, whereas the differences within the same TCR and different epitopes are quite limited. Very interestingly, the unique complexes showing no binding affinity for the A*0201:hTERT6A explores different regions of the conformational spaces at both global and local level.
TCR Effects on the Binding Groove Behaviour
Concerning the binding groove dynamics, our MD simulations point out that the MHC class I simulated in complex with different TCRs show that their conformational behavior is quite unaffected by both the TCR types and the sequence of the bound peptides. Such results were also confirmed by the RMSD and by the Zernike analysis, the latter providing a high pocket shape similarity in all the cases. On the other hand, some differences come out by the local analysis of the h-bonds between the TCR and the peptides. That is, the numerous h-bonds established between the ILA1α1β1and hTERT-6A are not found in the case of the ILA1 bound to the same peptides. Considering that the simulations have shown a limited conformational variability of the binding groove, we can argue that the TCR is the main determinant of the peptide interaction.
Dynamical Coupling of the pMHC-TCR
Although the kinetic assays reported differences in the binding activity, the pMHCs-TCRs exhibit similar interaction patterns. In detail, a remarkable correlation between the MHC alpha-chain and the whole TCR-which are not directly connected-was found in all the pMHC-TCR complexes. Moreover, the mutations of two residues in the peptide (ESO4D to ESO9C) induces a severe decrease of the correlation patterns, in both the TCR variable and TCR constant regions as well as in the MHC alpha-chain one. Such analysis suggests that a larger correlation between the variable and constant regions of the TCR does not necessarily imply an increased activation of the receptor. In fact, the 1G4-ESO4D, which experimentally shows a higher value of the K D than the 1G4:ESO9C, presents a minor release of IFNγ [40] and a higher interchain correlation. In other words, a limited coupling between the pMHC and the TCR motions could imply a larger receptor response.
As expected, the interaction energies between the binding groove and the ILA1 T-cell receptor show higher values with respect to the other complexes, despite the very similar interaction energy distributions between the hTERT6A:ILA1 and the other peptide-TCRs. This might suggest that the complementarity between the MHC class I and the TCR is one of the key factors regulating this complex recognition process.
Conclusions
We performed several MD simulations to understand the contribution of each character involved in the TCR triggering. During the recognition process the MHCs adapt themselves to the TCRs surface, assuming similar conformations. On the other hand, different peptides induce specific changes in the TCRs structures: a single-residue mutation on the peptide determines conformational changes in all the TCR regions, excluding the Cβ one. Moreover, double-residues mutations are also responsible for a modification of the pMHC-TCR coupling motions. In addition, the TCR behavior seems to be linked to the its specific sequence: the TCR type determines changes in the variable regions and sequence alterations in the CDR2β and CDR3α variable regions (e.i ILA1α1b1) induce clear conformational changes especially in the Cβ region.
Accordingly, we argue that although the MHC is essential in the antigen presentation to the TCR, the activation process is mainly influenced by the peptide sequence, leaving the MHC conformational behavior quite unaffected. The ability of the TCR to modulate its response based on the peptide sequence/structure is then probably linked to the structural-dynamical behavior of the Vα, Vβ and Cα regions. Due to the time scale of the Cβ conformational rearrangements, (which should have a relevant role in the process [8]) we were able to detect modification only between radically different situations. i.e., binding vs. no binding. Despite the present limitations of the MD approach which are essentially due to a limited time scale with respect to the pMHC-TCR activation process, we think that our outcomes suggest a new key-reading of the role of each character involved in the pMHC recognition.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4409/9/4/942/s1, Video S1: pMHC/TCR in membrane, Figure S1: The RMSD computed on the alpha carbon of the systems, Figure S2: The RMSF of the Tableprojections of the total TCR alpha carbons of the bound states, Figure S5: RMSD matrices, Figure S6: The Zernike Shape similarity of the binding groove, Figure S7: Network analysis, Figure S8: The Zernike pocket electrostatic similarity Figure S9: Membrane Thickness for the 1G4 complexes, Figure S10: Membrane Thickness for the ILA1/ILA1α1β1 complexes, Figure S11: Distance between the bilayers, Table S1: The Hydrogen bonds between the peptide and the TCR are shown, Table S2: Peptide Solvent exposure. Funding: This research was funded by "Sapienza" University of Rome, and PRACE grant number 2019204862. Financial support was from WellCome Trust grants: WT200844/Z/16/Z to Oreste Acuto. | 2020-04-16T09:13:58.654Z | 2020-04-01T00:00:00.000 | {
"year": 2020,
"sha1": "b0d55da8f6b8caefd884c8ec87cd40d7f6e46852",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2073-4409/9/4/942/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "1bb87a29560896d0c6650f9136699ca6fafb1566",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Chemistry",
"Medicine"
]
} |
8243292 | pes2o/s2orc | v3-fos-license | Retrospective Cohort Study of the Prevalence of Lumbosacral Transitional Vertebra in a Wide and Well-Represented Population
Purpose. The aim of this study is to determine the prevalence of lumbosacral transitional vertebra (LSTV) in a well-represented general population. Methods. For a retrospective cohort study, abdominal radiographs of adult subjects were queried with clear visibility of the vertebral body articulation of the last rib, all lumbar transverse processes, and complete sacral wings. Exclusion criteria included any radiologic evidence of previous lumbosacral surgery that would block our view. A total of 6200 abdominal films were reviewed, and 3607 were identified as being suitable for the measurement of the desired parameters. Results. A total of 3607 subjects were identified as eligible for the study, and 683 (18.9%) were classified as positive for a lumbosacral transitional vertebra. The prevalence of sacralization and lumbarization was found as 17.2% and 1.7%, respectively. The average age at the time of the study was 39.5 ± 15.2 years (18–86 years). Conclusions. As a result of different opinions, LSTV retains its controversial status. Our prevalence study of the general population will provide assistance for resolution of the controversy. Prevalence studies of the general population with a wide participation will shed light on comparative studies.
Introduction
Lumbosacral transitional vertebra (LSTV) is a congenital vertebral anomaly of the L5-S1 junction in the spine [1]. LSTV occurs because of sacralization in which one or both of the transverse processes of the fifth lumbar vertebra exhibits fusion with the first sacral segment or as a result of lumbarization in which the first sacral segment exhibits an abnormal transverse process similar to that of the lumbar vertebra. This modification may contribute to incorrect determination of a vertebral segment, which can lead to surgery at the wrong level and unresolved symptoms [1].
LSTV is evaluated with lumbosacral anteroposterior radiographs, and abdominal radiographs may also show this anomaly. Abdominal radiographs, as radiographic examinations for abdominal indications, have many clinical uses and are most often indicated for patients who exhibit signs of intestinal obstruction or visceral perforation. In this study, a different purpose was pursued. We used abdominal radiographs which were with clear visibility of the last rib's vertebral body articulation, all lumbar transverse processes, and the complete sacral wing to provide radiographic images of the lumbar spine. For a study evaluating LSTV cases, the use of these radiographs was desirable as they are likely to be Table 1: Castellvi radiographic classification system of sacralization [7]. A unilateral TP height greater than or equal to 19 mm Type Ib Both processes heights greater than or equal to 19 mm Type IIa Presence of unilateral articulation between the TP and the sacrum Type IIb Presence of bilateral articulation between the TP and the sacrum Type IIIa Unilateral fusion of the TP and the sacrum Type IIIb Bilateral fusion of the TP and the sacrum Type IV Unilateral type II transition (articulation) with a type III (fusion) on the contralateral side TP: lowest lumbar transverse process.
Type Ia
representative of the general population. The coexistence of low back pain and LSTV can create bias for this prevalence study. In order to eliminate low back pain population, we did not prefer lumbosacral radiographs. Several studies have been reported about the LSTV in a back pain population [2][3][4][5][6]; however, the prevalence studies with a wide participation of members of the general populations are lacking. Therefore, our study aims to establish the prevalence rates for LSTV in the wide and well-represented population.
Methods
A retrospective cohort study was conducted by reviewing the abdominal radiographs of 6200 patients over a oneyear period after the institution review board has been approved. These were identified from four cities (İstanbul, Diyarbakır, Konya, and Samsun) in different geographic regions of our country (Turkey). Inclusion criteria were subject's age older than 18 years at time of radiographs and abdominal radiographs available with clear visibility of the last rib's vertebral body articulation, all lumbar transverse processes, and the complete sacral wing. Exclusion criteria consisted of any radiologic evidence of previous lumbosacral surgery that would block our view. A total of 6200 abdominal radiographs from the past year (2011) were reviewed, and 3607 were identified as being suitable for the measurement of the desired parameters as we described the clear visibility in this retrospective cohort study.
Data collection consisted of the subject's age at the time of imaging, gender, bilateral height measurement of the lowest lumbar transverse processes, and the classification of the sacralization. Three spine specialists and one radiologist performed all of the measurements and classified the cases, using a systemized approach according to the Castellvi [7] radiographic classification system (Table 1) to decrease variability; in addition, the reviewers consulted among themselves. They read all the radiographs and evaluated the reproducibility of the descriptions. They determined the interobserver reliability. Digital films were downloaded to an imaging processing program for standardization of the measurements.
The prevalence of LSTV, sacralization, and lumbarization was reported separately. The anomalies were compared according to gender. Statistically significant differences were evaluated using the chi-square test for categorical variables (gender), and significance was set at < 0.05.
Results
A total of 1843 female and 1764 male subjects were identified as eligible for the study; the average age was 39.5 ± 15.2 years (18-86 years). Average ages of the subjects according to the classification types were shown in Table 2. Of these subjects, 683 were classified as positive for LSTV (Figure 1), with a gender distribution of 314 (46%) women and 369 (54%) men. The prevalence was found as 18.9% (Table 3).
A total of 623 subjects were classified as positive for sacralization, with a gender distribution of 276 (44.5%) women and 344 (55.5%) men, for a prevalence of 17.2% ( Table 4). The most common anatomical variant was Castellvi type Ia (5.5%). A total of 63 lumbarizations were classified, for a prevalence of 1.7%. Statistically significant differences were found between the two sex groups in subjects with LSTV ( = 0.002) and sacralization ( < 0.001). Prevalence of lumbarization was higher in women but the difference was not statistically significant ( = 0.088). Higher incidences of type Ia and type Ib were found in men and these differences were statistically significant ( = 0.016, < 0.001) when compared with the women (Table 4).
Discussion
Lumbosacral transitional vertebral frequency of the general population has been given a wide range of percentages ranging from 4% to 36% [1,[8][9][10][11]. Most studies have evaluated the relationship between low back pain and LSTV [4,5,7,[12][13][14][15]. Few further studies have been published regarding racial differences [11,16,17]. The wide range observed may be explained by differences in diagnostic criteria, imaging techniques, and confounding factors between the investigated population samples. Hsieh et al. [16] found a prevalence of 4% in a population consisting primarily of Chinese patients when using anteroposterior (AP) plain radiographs for diagnosis. However, they excluded Castellvi type I. Erken et al. [17] also used AP plain radiographs for diagnosis but did not exclude subtypes of LSTV. They found a prevalence of 35.9% in the Turkish population sample. Apazidis et al. [11] examined 211 subjects (107 men and 104 women) and found that 75 were classified as positive for transitional lumbosacral vertebra, with a gender distribution of 40 (19%) men and 35 (16.6%) women. They reported the prevalence at 35.6% in the American general population. Hsieh and Erken both used lumbosacral radiographs for their studies. The study of Apazidis et al. used kidney-urinary bladder radiographs. In the present study, we used abdominal radiographs instead of lumbosacral graphs. The coexistence of low back pain and LSTV can create bias for this prevalence study. In order to eliminate low back pain population, we did not prefer lumbosacral graphs. The prevalence of LSTV in the general population was found as 18.7% in our study.
Igbinedion and Akhigbe [18] reported a higher incidence of sacralization in men and a higher incidence of lumbarization in women, but no statistical correlation was established between transitional vertebra and gender. Magora and Shwartz [19] found an overall incidence of 0.65% lumbarization and 20.8% sacralization, and the incidence was markedly lower in women. They reported no direct relation between sacralization and age, sex, or ethnic community. Olanrewaju [20] showed that sacralization is predominant in males. Despite the higher number of females present in the sample population, more males were seen to have LSTV and also presented with sacralization. He recorded a relationship between LSTV and gender. In our study, the occurrence of type Ia and Ib was statistically higher in men. LSTV has been reported to occur with a higher statistical prevalence in male subjects. The significant sex effect reflects body size gender dimorphism in humans. This reveals the uniqueness of each case and may reflect the numerous and complicated genetic and developmental factors involved. The width of the last lumbar transverse process is likely related to the sacral characteristics. Having a statistically significantly higher prevalence of type 1a and 1b in male subjects can be explained by gender dimorphism.
The LSTV frequency in the low back pain (LBP) population ranges from 6% to 37% [4,5,7,[12][13][14][15]. The relationship between LBP and LSTV is not clear. Numerous studies have found no significant correlation between transitional vertebrae and low back pain [3,5,6,9], while others have found such a correlation [7,8,12,15,19]. Wigh and Anthony Jr. [12] and Castellvi et al. [7] advocated a relationship between LSTV and LBP. Frymoyer et al. [3] determined similar rates of radiological abnormalities in three groups of patients-no LBP, moderate, and severe. Similarly, Otani et al. [9] reported the incidence of transitional vertebra to be 13% in patients with LBP and 11% in their control group. Our study presented a prevalence of LSTV in a well-represented general population as a contribution to help resolve this controversy.
Conclusion
As a result of different research opinions, LSTV retains its controversial status. In this study, we wanted to find the prevalence of this pathology and the relationship between the genders. Our retrospective cohort study of the prevalence of a wide and well-represented population was found compatible with the literatures in which a wide range of frequency in the previous studies have been reported. The limitation of our study was literally not reflecting the general population. Prevalence studies of the population with wide participation such as the present case will help to shed light on future comparative studies. | 2016-05-04T20:20:58.661Z | 2013-06-24T00:00:00.000 | {
"year": 2013,
"sha1": "fab8c5c5a833ca94af3ad5a860ebd884c9f4fa9c",
"oa_license": "CCBY",
"oa_url": "https://downloads.hindawi.com/archive/2013/461425.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "fab8c5c5a833ca94af3ad5a860ebd884c9f4fa9c",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
225288564 | pes2o/s2orc | v3-fos-license | The Application of Discovery Learning Models in Learning to Write Descriptive Texts
The purpose of this study was to describe the application of discovery learning models in learning to write descriptive texts. The research method uses descriptive methods. Data collection techniques using observation techniques, test writing descriptive text, and documentation. Data analysis techniques by reducing data, data presentation, data interpretation, conclusions. The results showed that the steps of the discovery learning model in learning to write descriptive texts are to describe topics, ask questions, formulate hypotheses, carry out investigations, conduct observations, organize or analyze, and present the results of investigations. Then, the students' ability to write descriptive text obtained an average value of 76.8 with good criteria after applying the discovery learning model. The learning to write descriptive text is carried out to instill the value of responsibility to students and improve the ability to write text descriptive students based on text, images, and places, so students are not passive.
. Then, learning the Indonesian language and literature is essentially learning to communicate, both verbally and in writing (Kusmiarti et al., 2020). For this reason, teachers need a good understanding of the learning model used to improve student learning outcomes.
Indonesian learning in the 2013 revised curriculum is oriented towards text-based learning models (Manshur et al., 2018;Noermanzah & Friantary, 2019). Indonesian language learning is not just language knowledge, but as a text that serves as a source of user self-actualization in the socio-cultural context (Dewantara et al., 2019). In achieving text-based language, learning models are needed that emphasize mastery of the text. Darmansyah (2010) & Sutikno (2014 explained that the learning model is a way of organizing the content of lessons, delivery of lessons, and management of learning activities.
Teachers can use various learning resources that support the creation of the learning process's effectiveness and efficiency to achieve better learning outcomes.
One effort to improve learning outcomes is to invite students to learn based on needs rather than compulsion. The teacher can influence students' mindset to find their abilities through learning experiences by what is seen and felt. This can lead students to determine ideas in writing a text so that the students can be creative and can describe or paint a thing based on the imagination of the things found (Karto et al., 2019;Rahmawati, Syafryadin, Widiastuti, 2019). However, idea could be obtained by reading (Martina, Syafryadin, Utama, 2020). Therefore, to support writing a text, students need to read a lot.
Indonesian language learning in class VII semester 1 has core competencies and basic competencies. The fourth core component is trying, processing, and presenting in the realm of concrete (using, decomposing, assembling, modifying, and making). Abstract domains (writing, reading, calculating, drawing, and composing) by what is learned in schools and sources others are the same in point of view/theory. The basic competence to 3.1 is to identify information in the descriptive text about the object (school, tourist attractions, historic sites, and atmosphere of local art performances) that are heard and read. The indicators to be achieved are able to (1) determine
Ariyana, Enawar Ramdhani, Sulaeman The Application of Discovery Learning
Models in Learning to Write Descriptive Texts the general characteristics of the descriptive text that is read and heard; (2) determine the descriptive text that is read and heard; (3) get information from the descriptive text that is read and heard, and (4) express the descriptive text verbally.
Descrptive text is the text that contains a form of discourse that seeks to present a thing or object of conversation that makes the object as if the reader saw the object for himself or as if it was in front of the reader's head (Amrolani et al., 2014). A text with the main idea is conveyed by clearly describing the object, place, or event that is being the topic to the reader. Therefore, the reader seems to feel what is being expressed in the text directly. A descriptive is a form of writing related to the writers' efforts to provide details of the object being discussed or written (Kosasih, 2019).
Descriptive is a form of discourse that presents an object as if the readers saw the object themselves (Suryati, 2019). Descriptive provides an image of something that is experienced, for example, a scene, person or sensation.
Descriptive provides a descriptive of an event or events and problems. Through descriptive writing, a writer tries to transfer the messages of his observations and feelings to the reader by exposing the nature and all the details that exist in an object. Rani et al., (2006: 37) explain that discourse descriptive is a type of discourse aimed at the recipient of the message in order to form an image or imagination about something. He further stated that the psychological aspect that can digest descriptive discourse is emotion because with emotion a person can form an image or imagination about something. Haryati (2018) also explains the descriptive paragraph as a form of describing a situation with sentences, so that it creates a living impression. The depiction or painting must be presented in a lively manner, so that what is depicted lives on in the reader's imagination. The descriptive emphasizes its expression through a series of words. Descriptive allows us to see a visualization of the object. Roughly speaking, it can be said that the descriptive focuses on the appearance of the goods. In the descriptive we see the cultivated object in a living and concrete way. impressions/perceptions to create sensations to produce an impression based on reason and imagination (Budiyono, 2016). Components of the descriptive text are (1) The contents of the descriptive text are broken down into details of the parts of the object. (2) The contents of the descriptive text concretely use special words. (3) The contents of the descriptive text are personal with emotional content so that they use emotional words strong. (4) Students are able to master the linguistic aspects (punctuation and use of letters according to PUEBI). Students must master these components so students can create or compile descriptive texts in oral and written (Sahri, 2015).
Improving the ability to write descriptives also needs a learning model that can enable students' ability to understand the descriptive text. The selection of the right learning model will help increase the ability to write descriptive texts (Sulaeman & Ariyana, 2018). One learning model that can improve writing skills is the discovery learning model. Discovery learning is a learning model that tends to ask students to make observations, experiments, or scientific actions to get conclusions from the results of these scientific actions (Fajri, 2019). With discovery learning, students can learn independently and in groups with a scientific process through the stages of observing, questioning, trying, reasoning, and communicating (Yusnawarni, 2017).
Through Discovery Learning, students can determine their information and concepts about learning material (Seung, 2009). Students are invited to think critically to describe objects in an interesting way. This is under Juliyanti & Suryani's research (2018) that the discovery learning model can increase students' understanding of the main ideas and ideas in the form of descriptive text. With the help of discovery learning models, students can write descriptives well, especially detail, and describe an event so that the reader can see, hear, and experience what is described. Suryati (2019) explains that the use of the discovery learning method in compiling a written descriptive text is more suitable to be applied compared to other learning methods. This is because the discovery learning method is formed from ideas in the form of mind maps that can be adapted to the structure of the descriptive text, namely the general descriptive and descriptive The discovery learning model has also been shown to improve the writing skills of fantasy text stories for students of class VII5 of SMP Negeri 3 Ambon (Souhuwat, 2019). Some of the relevant research results above show that the discovery learning model can provide understanding to students in writing descriptives and writing fantasy text. For that reason, the discovery learning model will be tried to apply to learn to write descriptives in grade VII students of SMPN 14 Tangerang City. The formulation of the problem in this study is, "How is the application of discovery learning models in learning to write descriptives in grade VII students of SMPN 14 Tangerang City?" To describe the application of discovery learning models in learning to write descriptives in grade VII students of SMPN 14 Tangerang City. The results of this study are expected to explain Indonesian language teachers' steps to write a descriptive with a discovery learning model and to explain the ability to write a descriptive of grade VII students of SMPN 14 Tangerang City.
Research Methodology
This research uses a descriptive method. The descriptive method used is expected to be able to explain data quantitatively and qualitatively (Creswell, 2014). The discovery learning models application was used in learning to write the descriptive text for grade VII students of SMPN 14 Tangerang City. Using descriptive methods is expected to be able to clearly describe the steps of learning to write descriptive text using the discovery learning model and the ability to write descriptive text after the discovery learning model is applied. The assessment indicators write descriptive text, namely the aspects of the suitability of the title with the contents, the use and writing of the spelling, choice of words or diction, sentence structure, inter-sentence integration, and integration between paragraphs. Then, the rating classification is 80-100 with very good criteria, 70-79.9 with good criteria, 60-69.9 with enough criteria, and values> 59.9 with fewer criteria.
Data analysis techniques in this study were steps, namely, reducing observation and test data, presentation of test and observation data, interpretation of observation and test data, conclusions. The data validity test uses the triangulation of data collection techniques in the form of observation, tests, and documentation. Then, in the analysis of test data, the inter-rater validity test of the two assessors' ability to write descriptive texts is used. Also, in each activity to analyze the data so that the data is analyzed scientifically.
Findings
From the results of the analysis of observational data, tests, and documentation of the learning implementation plan, the application of the discovery learning model to learning to write descriptives in grade VII students of SMPN 14 Tangerang City showed good results. The Indonesian language teachers implemented the Indonesian language learning to manage the class well, starting from exploration, elaboration, and confirmation activities. They applied the discovery learning model following the steps of the discovery learning model compiled in the learning implementation plan document.
Indonesian language teachers implemented the steps of learning to write descriptive text using discovery learning models in class VII SMPN 14 Tangerang City. Namely, describe topics, ask questions, formulate hypotheses, carry out investigations, conduct observations, organize or analyze, and display the results of investigations. What is interesting is that when making observations, the teacher assigns activities carried out in several shopping centers in the city of Tangerang so that students are motivated to make observations and enthusiasm to express their ideas and can organize writing in the form of between sentences and between paragraphs to form a good descriptive text.
The weaknesses in the ability to write descriptive text, there are still some students who still experience errors in writing spelling, standard words, and the use of punctuation.
Discussion
The application of discovery learning models in learning to write descriptives in grade VII students of SMPN 14 Tangerang City showed good results. The results of observations, the Indonesian language teachers who implemented Indonesian language learning, can manage the class well, starting from exploration, elaboration, and confirmation activities. The applied the discovery learning model following the steps of the discovery learning model that is compiled in the learning implementation plan document. The Then, from the results of the ability to write descriptive text after the implementation of discovery, learning models obtained an average value of 76.8 with good criteria. The students can write titles according to content, use spelling, pour words or diction according to the characteristics of paragraph descriptives, arrange sentences according to the structure of effective sentences, and integrate between sentences and between paragraphs to form a good descriptive text. This is also under Juliyanti & Suryani (2018) 's research that the discovery learning model can motivate students to write descriptive texts by investigating, observing, and organizing sentence by sentence into integrated paragraphs.
The application of discovery learning models in learning to write descriptives in grade VII students of SMPN 14 Tangerang City shows that the teacher's role as a facilitator is also important, especially in guiding students individually, in groups, and classically (Abdullah, 2005). Also, the ability to The Discovery learning model will gain several advantages if done well and by the needs of students, especially in developing cognitive abilities, affective, and psychomotor students. This is because the discovery learning model refers to the application of a scientific approach that requires students to be active in developing their ideas and creativity. This is following the opinion of Roestiyah (2008) that the advantages of using discovery learning models are 1. this model helps students develop, increase readiness, and mastery of skills in the cognitive process/introduction of students; 2. students gain knowledge that is very personal/individual so that they can be sturdy/deep that is left in the souls of the students; 3. can arouse the enthusiasm of student learning; 4. this model can provide opportunities for students to develop and progress according to their respective abilities; 5. able to direct the way students learn, so they have more strong motivation to study harder; 6. helps students to strengthen and increase their confidence with their discovery process; and 7. student-centered is not on the teacher, and the teacher is only a study partner or facilitator. professionals in mastering the material and mastering the art of managing an active, innovative, and fun class. This is in accordance with the opinion of Darmadi (2010) and Permadi & Arifin (2013) that competent teachers, both professional competence, pedagogical competence, social competence, and personality competence have a very good chance of creating highachieving students because generally teachers are used as role models for their students.
Conclusion and Suggestion
The application of discovery learning models in learning to write descriptives in grade VII students of SMPN 14 Tangerang City showed good results. The observations results on the Indonesian language teachers who implement Indonesian language learning, they are able to manage the class well, starting from exploration, elaboration, and confirmation activities. The steps of the discovery learning model in learning to write descriptives are to explain topics, ask questions, formulate hypotheses, carry out investigations, conduct observations, organize or analyze, and present the results of investigations.
Then, from the results of the ability to write descriptive text after the implementation of discovery, learning models obtained an average value of 76.8 with good criteria. The students can write titles according to content, use spelling, pour words or diction according to the characteristics of paragraph descriptives, arrange sentences according to the structure of effective sentences, and integrate between sentences and between paragraphs to form a good descriptive text.
From the results of this study, it is expected that the discovery learning model can be used as one of the learning models that Indonesian language teachers can use to improve the ability to write student descriptive texts. The discovery learning model is also expected to be followed up by other researchers to be applied to other materials about writing skills or other language skills.
Ariyana, Enawar Ramdhani, Sulaeman The Application of Discovery Learning
Models in Learning to Write Descriptive Texts | 2020-09-10T10:06:40.492Z | 2020-09-02T00:00:00.000 | {
"year": 2020,
"sha1": "67da359a79d828fc23b998b5fdbc92f0bfa9ca48",
"oa_license": "CCBYSA",
"oa_url": "https://ejournal.unib.ac.id/index.php/JEET/article/download/12154/6119",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "fb191b9fa877ad0dbff4b63de4ce24e83ee2bfda",
"s2fieldsofstudy": [
"Education"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
216867195 | pes2o/s2orc | v3-fos-license | Quantifying alpha clustering in light nuclei from binding energies
What is the origin of nuclear clustering and how does it emerge from the nuclear interaction? While there is ample experimental evidence for this phenomenon, its theoretical characterization directly from nucleons as degrees of freedom remains a challenge. In this work, it is shown that the degree of one- and two-alpha (${ {}^{4}\text{He} }$) clustering in a given nucleus can be quantified empirically using only the binding energies of its partition subsystems. The proposed clustering measures are parameter-free and correctly identify alpha clustering features in light nuclei and long-lived excited states such as the Hoyle state in $^{12}$C and the ${ {0}_{2}^{+} }$ state of $^{14}$C at 6.59 MeV. It is revealed that in light nuclei ranging from ${ {}^{6}\text{Li} }$ to ${ {}^{14}\text{C} }$, state-of-the-art density functional theory and \textit{ab initio} approaches fail to adequately capture alpha clustering. Stringent constraints on binding energies are then provided by back-propagating 10\% relative uncertainties on the experimental one-alpha clustering measure using a parallel Markov chain Monte Carlo algorithm. It is demonstrated that the nuclei $^{6,7}$Li, $^7$Be, $^{10,11}$B, and $^{11}$C are particularly sensitive to alpha clustering despite not being the most clustered systems identified. Using results on $^{10}$B, a strong case is made for a link between three-body forces and alpha clustering. This study provides the first quantification of alpha clustering based on binding energies only, as well as new and practical constraints for future optimizations of nuclear forces, potentially helping with the current issues in medium-mass nuclei.
I. INTRODUCTION
Nuclear clustering [1,2] refers to an emergent phenomenon by which the atomic nucleus effectively acquires a molecular-like structure and dynamics due to the formation of tightly bound clusters of protons and neutrons at low energy [3][4][5][6][7][8][9]. This phenomenon is fairly common and particularly salient in light nuclei close to cluster emission thresholds [10,11], as summarized in the famous Ikeda diagrams [12].
Light clusters are an important ingredient for the equation of state of nuclear matter [13] and the symmetry energy at low temperatures and densities [14][15][16], effectively impacting supernovae explosions and pasta phases in the crust of neutrons stars [17].
A proper understanding of nuclear clustering is still incomplete and calls for a closer link between experiment and theory [9]. Fundamental approaches based on nucleons as degrees of freedom play a special role in this enterprise since they offer the possibility to compute the degree of clustering of a A-body system defined by the probability ⟨A P (A)⟩ 2 where P (A) is a partition of the system.
Those approaches, given a model of nuclear forces, can theoretically solve the many-body problem exactly. In practice, only numerically controlled solutions can be provided due to finite computational resources, generating systematic errors. They include ab initio approaches and, it can be argued, density functional theory.
At low energy, calculating the degree of clustering precisely can prove challenging due to the presence of competing emergent phenomena like pairing, deformation, or collective motion, but also single-particle effects responsible for shell closures and the "magic numbers".
Fortunately, in light nuclei, precise calculations are fea-sible, and modern many-body forces adjusted on twoand three-body data are expected to give satisfactory results, while in medium-mass and heavier systems it is unclear whether or not it should be the case. However, are modern nuclear forces really describing clustering reasonably well? There are deep reasons to believe that alpha ( 4 He) clustering specifically is an important feature of the nuclear interaction. Nuclear matter is near a phase transition between a nuclear liquid and a Bose-condensed gas of alpha particles [18][19][20][21], largely controlled by the strength of the interaction and its locality [20]. A similar conclusion was reached in an earlier work [22] connecting clustering with the depth of the nuclear mean-field, which itself depends on the strength of the interaction.
From a different perspective, it was also shown that the Wigner SU(4) symmetric part of the interaction, which controls the ground state of the alpha particle, dominates in large nuclear systems [23]. Moreover, it was demonstrated that adjusting a simplistic interaction on alphaalpha scattering [24] improved binding energies beyond light nuclei.
Thus, it could be argued that nuclear forces must capture alpha clustering properly to be valid beyond light nuclei, where the mean-field dominates and its symmetries and geometry affect properties of clustered states [25,26].
In this work, a new and simple model-independent way to quantify indirectly the degree of nuclear clustering using only binding energies is introduced. Then, the quality of various state-of-the-art density functional theory (DFT) calculations and ab initio results based on modern nuclear forces is assessed against alpha clustering in light nuclei. Finally, a Markov chain Monte Carlo algorithm is used to provide constraints on binding energies by back-propagating uncertainties on the surrogate clustering measure.
II. FROM BINDING ENERGIES TO ALPHA CLUSTERING
Intuitively, the degree of clustering, measured by the probability ⟨A P (A)⟩ 2 , should be high (low) when the energy required to break the A-body system into its partition P (A) is low (high), but also when the energy required to break the individual clusters in P (A), in the presence of the other clusters, is high (low).
In the particular case of one-alpha clustering, i.e. P (A) = α + a where symbolically a = A − α is treated as a cluster, the energy to break The energy necessary to break the a-body cluster in the presence of the alpha is approximately the energy Σ α (A) necessary to remove the a nucleons one-by-one from the A-body system. In practice, proton and neutron separation energies cancel each other along any path connecting a given nucleus to the alpha particle, giving The goal here is not to determine the exact behavior of the probability p α (A) = ⟨A α + a⟩ 2 with Q α (A) and Σ α (A), but only to find a quantity varying roughly in the same way and which is rather sensitive when approaching complete factorization or p α (A) = 1. With those requirements, the simplest surrogate measure of one-alpha clustering proposed is: where the absolute value ensures that the measure stays positive for unbound systems such as 8 Be which are clustered by definition. One notes that, unlike p α (A) which cannot exceed one by definition, r 1 (A, α) diverges when Q α (A) tends to zero, making this measure sensitive near complete factorization as desired.
The dependence on Q α (A) of this simple form can be approximately justified using perturbation theory for two weakly interacting clusters as shown in Appendix A.
The clustering ratio r 1 is model-independent in the sense that there are no parameters and different manybody approaches using the same interaction should provide the same energy spectra and hence the same clustering ratios.
Long-lived excited states can also be approximately considered by removing their excitation energy E * (A) to Σ α (A). Moreover, the clustering ratio for two identical clusters in A can be defined similarly to the one-cluster ratio: where E(A) − E( 8 Be) is the energy to remove all the nucleons above 2α one-by-one. The main interest in these two extensions is to study (excited) systems with two alpha particles such as the Hoyle state in 12 C which is critical for nucleosynthesis in stars. Results for the oneand two-alpha clustering ratios are given in Tab. I for all nuclei where an estimate can be provided. As expected, the helium isotopes do not exhibit any particular pattern since the excess neutrons around the alpha core are not clustered. In the other isotopic chains, the ratio r 1 is maximal for 7 Li, 8 Be, 9 B, and 12 C, while the ratio r 2 is maximal for 12 Be, 12 B, and 12 C. This means that the latter have a pronounced 2α + (A − 2α) structure, except perhaps for the ground state of 12 C for which r 1 and r 2 are maximal simultaneously, indicating a 3α structure as expected. These results are consistent with what is already known about the structure of light nuclei and can serve as a reliable basis to test nuclear models.
The Hoyle state at 7.65 MeV above the ground state of 12 C, yields one-and two-alpha clustering ratios of r 1 = 193.8 and r 2 = 48.29, respectively. These large values are consistent with the characteristic clustered nature of the Hoyle state. However, the fact that r 2 is significantly different than r 1 suggests a two-plus-one-alpha structure rather than a three-alpha structure. This seems consistent with previous ab initio calculations [29][30][31][32], claiming that the Hoyle state has a "bent-arm" shape made of two-plus-one alphas.
In Ref. [33], it was suggested that the 0 + 2 state of 14 C at 6.59 MeV could have a similar structure than the Hoyle state. This hypothesis is supported by the clustering ratios r 1 = 12.97 and r 2 = 3.887 of this state, but the magnitude of clustering seems quite reduced in comparison.
Binding energies given by many-body approaches can be used as well. In this work, both DFT and ab initio results are used. The former are based on energy density functionals adjusted over thousands of ground state energies across the nuclear chart and in some cases other ob-servables as well. They are not designed to give accurate results in light nuclei, but they are efficient at capturing bulked properties of nuclei such as deformation which are affected by clustering. The latter are based on modern nuclear forces usually adjusted on two-and three-body data only (but not always), and they should be precise in the light sector, but they can miss emergent properties of nuclei if those are not fully encoded into the interaction.
There are, of course, other well-known methods based on nucleons as degrees of freedom that could not be used in this work but should be mentioned, such as the antisymmetrized molecular dynamics [34,35] and the fermionic molecular dynamics [36,37], both being particularly well suited for studying clustering and with a long history summarized in Ref. [9].
In Fig. 1, the DFT results from Refs. [38,39], denoted SLy4, UNEDF0, UNEDF1, SkP, and SkM * , were used to compute the one-alpha clustering ratios of several Li, Be, B, and C isotopes and compared to experiment [27, 28] (see Tab. I). Only the results from the SV-min energy density functional are not shown as they were not of sufficient quality. Also shown are the ab initio Green's function Monte Carlo (GFMC) results based on the so-called AV18+IL7 high-precision two-and three-body potentials compiled in Ref. [32], as well as the GFMC results in Ref. [40] based on the local chiral two-and three-body potentials including intermediate ∆-excitations and denoted NV2+3Ia; and the no-core shell model (NCSM) results based on the JISP16 [41] and Daejon16 [42][43][44] two-body interactions, as well as the LENPIC interactions of 2018 [45] (two-body) and 2019 [46,47] (two-and three-body).
In some ab initio approaches, the occasional absence of results in very light nuclei (A ≤ 5) was compensated using experimental data assuming they would be described correctly anyway.
The first observation is that DFT models do capture alpha clustering to a large extent but, like most other approaches presented here, they tend to underestimate it except in 10,14 C where they overestimate it. Those models are efficient at describing bulk properties of nuclei and so they represent the minimum that more precise approaches should be able to achieve.
The second observation is that the ab initio approaches based on two-body forces only, i.e. the NCSM results using the JISP16, Daejon16, and LENPIC(18) interactions, give very different results depending on how they have been optimized.
The LENPIC (18) interaction is fitted only on two-body data and gives mixed results, with alpha clustering overestimated in 7 Li and 9 Be but vastly underestimated in 8 Be.
The JISP16 and Daejon16 interactions were fitted on few-body data as well as several ground-state energies in larger systems with A ≤ 16, including 6 Li and 10 B (see next section). However, the former did not benefit from modern fitting techniques at the time it was designed and lacks precision as compared to the Daejon16 interaction. For that reason, the JISP16 results stand out and do not match experiment. This outcome can seem puzzling because this interaction can describe (n, α) and (p, α) scattering data [48], as well as alpha-alpha scattering as shown in Ref. [49] using a new ab initio method tailored for clustering. The Daejon16 interaction, on the other hand, gives a near-perfect match to experiment. This interaction, unlike JISP16, is based on chiral effective field theory potentials but should be able to describe the same scattering data because of the way it was built. The only real difference is the quality of its optimization on binding energies. It is thus possible to build a pure two-body interaction compatible with the one-alpha clustering measure as well as (n, α), (p, α), and alpha-alpha scattering data, providing that some information beyond few-body systems is included in the optimization.
Finally, the ab initio results including two-and threebody forces adjusted on few-body data only, i.e. the NCSM-LENPIC (19) and both GFMC results, give the correct trends and are consistent with each other even if they underestimate clustering significantly in 8 B and 12 C (outside the 10% relative uncertainty band). The comparison between the LENPIC interactions of 2018 and 2019 in 9 Be provides clear evidence of the crucial role played by three-body forces when fitting only few-body data, despite an underperformance of the latter in 13,14 C.
Overall, the results in Fig. 1 present the first evidence of a systematic deficiency in most modern nuclear forces with respect to alpha clustering. On the one hand, ab initio approaches based on two-body forces give satisfactory results solely when adjusted on binding energies of A ≤ 16 nuclei, at odd with the ab initio rationale. On the other hand, those including three-body forces and adjusted on A ≤ 3 data show consistency but underestimate alpha clustering in critical nuclei.
From a strict ab initio perspective, either current forces are not constrained enough to encode clustering effects, or, as was suggested in Ref. [9], those effects come at higher orders in chiral effective field theory, or the nuclear interaction should be redesigned as to explicitly account for the existence of clusters in nuclear matter.
Can alpha clustering be used to improve nuclear forces? When experimental Q α values are used in place of the theoretical ones when calculating r 1 , all models fall into the 10% relative uncertainty band, except for the DFT results which are not expected to perform well in light nuclei by design. This suggests that only a marginal improvement on binding energies is required. Which energies in particular need to be improved is the topic of the next section.
III. FROM ALPHA CLUSTERING TO BINDING ENERGIES
The one-alpha clustering measure introduced in the previous section offers the opportunity to identify which nuclei's ground-state energies are the most important to constraint many-body approaches with respect to alpha clustering.
If success is defined as being compatible with the experimental clustering ratios r exp 1 (A) provided in Tab. I within 10% relative uncertainties, the question is: what are the constraints on binding energies? This is a standard Bayesian inference problem where uncertainties on a fixed output must be back-propagated on the input so that the latter is compatible with the former. In this work, this is done using a Markov chain Monte Carlo (MCMC) algorithm based on the Metropolis-Hastings method as commonly used in in different fields of physics.
The problem is analogous to a network calculation in nuclear astrophysics [50], albeit computationally much less expensive. To compute one set of r 1 (A) values, all the binding energies of all the nuclei in Tab. I, as well as those of their partition subsystems, are required. Those energies are initialized randomly around the corresponding experimental values ±30% to start relatively far from the ideal values. Then they are updated by the MCMC algorithm at each step, which is itself guided by whether or not the r 1 (A) values are approaching the experimental ones within the accepted uncertainties, as defined by the χ 2 measure at a step i: where ∆E controls the "speed" at which the parameter space is explored. The MCMC run is stopped when the energies are stable and the χ 2 (i) does not decrease significantly. The two hyper-parameters controlling these criteria are the number of MCMC steps N steps and ∆E. In the present work, sufficient convergence was found for N steps = 5000 and ∆E = 10 keV.
At the end of one MCMC run, one set of binding energies (input) compatible with the experimental onealpha clustering ratios (output) is obtained. The principle of parallel MCMC is to repeat this process many times to sample the distribution of the input compatible with the output, and then to extract mean values and standard distribution of individual binding energies, assuming Gaussian distributions.
In practice, the binding energies of 2,3 H and 3,4 He were kept fixed to their experimental values since they are very well reproduced theoretically. One additional exception was considered for 8 Be to ensure the stability of the MCMC algorithm. Its binding energy was allowed to change, but its clustering ratio was not included in the χ 2 because of its very large value.
The results from the averaging of 20 runs of 50 independent MCMC runs with 5000 steps each are shown in Tab. II. In principle, a single run of 50 independent MCMC runs would be enough, but the additional averaging procedure ensures a better precision and allows to check that the error on the mean energies and their standard deviations is very small.
The 30 binding energies are well constrained by the optimization of the 24 clustering ratios r 1 as indicated by the mean values µ(E). When looking at standard deviations, nuclei fall into two groups defined by σ(E) ≈ 100 − 250 keV and σ(E) < 30 keV. The latter contains only the six nuclei 6,7 Li, 7 Be, 10,11 B, and 11 C whose ground-state energies are the most constrained by the one-alpha clustering measure. They are all made of one or two alphas and either 2 H, 3 H, or 3 He; and except for 7 Li, they are surprisingly not the most clustered nuclei identified earlier ( 7 Li, 8 Be, 9 B, and 12 C) but their direct neighbors.
Knowing that the binding energies of 5 He, 5 Li, and 8 Be do not exhibit a strong sensitivity to alpha clustering makes the situation of the JISP16 interaction less puzzling. Also, the reason why the Daejon16 interaction matches experimental r 1 values is because it reproduces the binding energies of the nuclei 6 Li and 10 B.
The case of 10 B is quite revealing. It was found in ab initio calculations [9,51,52] that this nucleus is particularly sensitive to three-body forces. It is also known that 9 Be and 9 B, which are one proton and one neutron away of 10 B, respectively, have ground states very close to thresholds and hence are prone to clustering [53,54], supporting the idea that 10 B is clustered and has a α − α − d structure. The results in Tab. II show that 10 B is one of the most sensitive nuclei with respect to alpha clustering, and the results in Fig. 1 show that three-body forces are critical to describe alpha clustering in light nuclei. Together, these elements are strongly indicative of a link between three-body forces and alpha clustering. Another anecdotal argument in this direction comes from the AV18+IL7 interaction. It was shown in Ref. [55] that the "Illinois" (IL) family of three-body forces used in the GFMC-AV18+IL7 calculations were able to describe correctly (n, α) scattering data, and that the IL7 version, in particular, yielded an improved ordering of the lowlying states in 10 B [32].
IV. CONCLUSIONS
In conclusion, the empirical one-and two-alpha clustering measures introduced in this work, based solely on binding energies, allow to correctly identify the nuclei 7 Li, 8 Be, 9 B, and 12 C as being the most clustered in their respective isotopic chains, and to characterize excited states of interest such as the Hoyle state.
It was shown that, according to the one-alpha clustering measure, state-of-the-art ab initio approaches based on modern nuclear forces significantly underestimate alpha clustering in key nuclei such as 8 Be and 12 C.
Finally, relative uncertainties on the one-alpha clustering measure were back-propagated on binding energies using a MCMC method and revealed that the ground states energies of 6,7 Li, 7 Be, 10,11 B, and 11 C are particu-larly sensitive to alpha clustering.
By examining the literature on 10 B together with the results of the present work, a strong case was made for a link between three-body forces and alpha clustering.
This work opens many new avenues ranging from the extension of the clustering measures beyond 14 C and to other types of clusters, to the systematic study of clustered excited states, and to the testing of current and future DFT models and ab initio interactions. Perhaps, the most interesting opportunity is the development of new two-and three-body forces in the spirit of the NNLO sat interaction [56] including the binding energies of some of the six nuclei identified in their optimization. Such interactions could give satisfying results in medium-mass nuclei studied at the Facility for Rare Isotope Beams (FRIB) [57]. The results presented can also help experiments by telling which isotope to use in a reaction depending on the clusters of interest, and even which energy to choose by plotting the clustering measures as a function of the excitation energy to determine when clustering is maximal (not shown).
In all fairness, the measures introduced are empirical and mostly provide a posteriori tests of the presence of clusters. For these reasons, a genuine attempt was made at quantifying alpha clustering to reveal real issues in nuclear models, but conclusions were mostly limited to qualitative statements.
Moreover, the one-alpha clustering measure does not discriminate between a meaningful model and a fictitious one if both are perfectly fitting binding energies of light nuclei. However, to quote J. von Neumann "With four parameters I can fit an elephant, and with five I can make him wiggle his trunk", and so, as for all models, good judgment matters.
A picture is emerging where alpha clustering, threebody forces, and observables in medium-mass nuclei are interrelated. Alpha clustering is an important feature of the nuclear interaction and should be used when developing and evaluating modern nuclear interactions as was recently done in Ref. [58] for three-body forces on (n, α) scattering.
Appendix A: One-alpha clustering measure approximation in perturbation theory The isospin symmetric approximation of the one-alpha clustering measure r 1 (A, α) ≈ 1 + E(A − α) Q α (A) can be justified approximately using perturbation theory. Assuming two non-interacting clusters α and a with their non-orthogonal ground and first excited states α, a⟩ and α, a * ⟩, respectively, the state of the interacting system A = α + a writes in the first order of perturbation theory: where ε is a small parameter and V is the interaction between the two clusters. The interaction matrix elements will be written using the short notation ⟨α, a * V α, a⟩ = V a * ,a for convenience. The energy of the total system in its first excited state is: Using the fact that E(α, a) = E(α) + E(a) (noninteracting clusters) and denotinig the energy difference between the ground state and the first excited where one recognizes the Q α -value. The overlap between the interacting and factorized systems then writes: ⟨α, a A⟩ = 1+ εV a * ,a δE(A) − Q α (A) + εV a * ,a * ⟨α, a α, a * ⟩ . (A4) For small perturbations, i.e. ε << 1, one obtains: where C is a small constant. The degree of clustering is then given by the probability p α (A) = ⟨A α, a⟩ 2 , indicating possible corrections in 1 Q 2 α (A) to the one-alpha clustering measure r 1 (A, α).
ACKNOWLEDGMENTS
I am grateful to M. Piarulli for sharing the binding energies obtained using the GFMC-NV2+3Ia interaction, as well as to P. Maris and J. Vary for sharing the last update of the binding energies obtained using the NCSM-JISP16, Daejon16, and LENPIC (19) interactions, and for the ensuing discussions. I also want to thank N. Vassh and S. König for clarifying the use of the parallel MCMC algorithm and for pushing me to develop the two-alpha clustering measure during their productive visits at Argonne National Laboratory. My appreciation also goes to R. Wiringa, H. Hergert, S. Bogner, and W. Nazarewicz for their patience and useful comments on clustering and nuclear forces. Finally, I would like to thank T. Li and S. Wang for pointing out errors in the original manuscript. This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Nuclear Physics, under the FRIB Theory Alliance award de-sc0013617. An award of computer time was provided by the Institute for Cyber-Enabled Research at Michigan State University. | 2020-05-01T01:00:56.751Z | 2020-04-30T00:00:00.000 | {
"year": 2020,
"sha1": "335ef0b8180759e318b694c60bb20f48185ba452",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "335ef0b8180759e318b694c60bb20f48185ba452",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
248806978 | pes2o/s2orc | v3-fos-license | Transcriptomic and Metabolomic Analysis Reveal Possible Molecular Mechanisms Regulating Tea Plant Growth Elicited by Chitosan Oligosaccharide
Chitosan oligosaccharide (COS) plays an important role in the growth and development of tea plants. However, responses in tea plants trigged by COS have not been thoroughly investigated. In this study, we integrated transcriptomics and metabolomics analysis to understand the mechanisms of chitosan-induced tea quality improvement and growth promotion. The combined analysis revealed an obvious link between the flourishing development of the tea plant and the presence of COS. It obviously regulated the growth and development of the tea and the metabolomic process. The chlorophyll, soluble sugar, and amino acid content in the tea leaves was increased. The phytohormones, carbohydrates, and amino acid levels were zoomed-in in both transcript and metabolomics analyses compared to the control. The expression of the genes related to phytohormones transduction, carbon fixation, and amino acid metabolism during the growth and development of tea plants were significantly upregulated. Our findings indicated that alerted transcriptomic and metabolic responses occurring with the application of COS could cause efficiency in substrates in pivotal pathways and hence, elicited plant growth.
Introduction
The quantity and quality of tea leaves are mainly affected by the metabolites in tea leaves, such as tea polyphenols, theanine, caffeine, vitamins, volatile oils, polysaccharides, and minerals, which are closely related to the growth process of tea trees [1]. Tea plant pests and diseases and environmental pressure cause major economic losses for tea. Traditional pesticides can increase the yield of tea, but their residues would be harmful to the local ecosystem and the quality of the tea plant [2][3][4][5][6]. The application of natural-origin chitosan has been reported safe for the environment and beneficial to the quality and yield of crops [7][8][9][10].
Chitosan oligosaccharides (COS) are derived from the shells of shrimp and other sea crustaceans and possess good water solubility and a wide range of biological activity [11]. Chitosan has generated great attention in a range of fields due to its exceptional biological activities, especially in agricultural uses like antimicrobial action [12,13], plant growth and development promotion [14][15][16], plant defenses elicitation [13], and resistance against various abiotic stresses [12,14,17]. For example, it augments plant height, leaf number, fruit weight, number, and yield [18][19][20]. Chitosan can improve the growth shoot and root length in beans [21]. The promotion function of chitosan application is also verified in several other plants, such as potatoes [22], wheat [23], and rice [24]. Chitosan can increase chlorophyll content and thus photosynthesis in rape seedlings [25] or up-regulate a series of primary C and N metabolic pathways in the leaves of wheat seedlings [26]. We have previously reported that chitosan has obvious promoting effects on tea plant growth [27] and eliminates the cold stress on the tea plant [28]. This biodegradable, economical, and renewable oligosaccharide is more appealing in organic agriculture and benefits plant yields and ecological biodiversity.
Most previous studies have been limited to the apparent effects of chitosan on plant physiology and growth characteristics. However, no research has confirmed the metabolic responses of chitosan on tea plant growth and the yield promotion of chitosan in the tea tree. At the same time, plant histological techniques are widely used for the discovery of functional genes, identification of metabolic pathways, and modification of plant cell characteristics [29][30][31]. Metabolite profiling technology is critical to the comprehensive analysis of the plant growth mechanism [32][33][34]. Multiparametric metabolic response of plants can be achieved by metabolomics, where metabolic accumulation and change could be detected by chromatography interfaced with mass spectrometry [32].
However, despite the elicitors function of chitosan discovered to date, alterations or responses in tea plants trigged by it have not been thoroughly investigated. Therefore, biochemical analysis of the tea plant by integrating transcriptomics and metabolomics analysis will aid our understanding of the mechanisms of chitosan-induced tea quality improvement and growth-promoting.
In this study, we compared metabolic profiles of control and chitosan-treated leaves and relationships within the metabolic network of the tea plant. Using this approach, we determined that a chitosan-induced metabolic strategy can increase the normal regulation of amino and nucleotide sugar metabolism, which enables intrinsic growth promotion.
Yield Measurement of Tea Leaves
The data of tea yield, including bud density, were measured as shown in Table 1. The concentration of COS at 2.0 g/667 m 2 (1:800) performed better but with no significant difference. The average bud densities were 122.3, 128.8, and 126.9 buds for the three concentrations 1.6, 2.0, and 3.2 g/mu (1 mu = 667 m 2 ), which increased by 18.97, 25.29, and 23.44% when compared to CK. Accordingly, the mean 100-bud weight was 7.68, 7.77, and 7.70 g and increased by 13.66, 15.06, and 14.03% compared to CK. Actual tea leave yield was 33.96, 36.33, and 34.66 g/m 2 and increased by 15.7, 23.8, and 18.1%, respectively. In sum, the presence of COS can significantly increase the tea plant yield. In contrast, the concentration of COS at 2 g/667 m 2 performed better, though with no significant difference. The value represents the mean ± SD of three biological repeats. Different letters indicate significant differences at p < 0.05.
Among the three tested concentrations, 800-time dilution performed best. Both the higher and lower concentrations spoiled the effect rather than increasing the production yield. Therefore, we applied and checked the most effective 800-time for their function.
As shown in Figure 1, more fresh shoots from the tea bushes were exposed in the COS treatment than in that of the control group. In addition, the promoted emergence of new and tender leaves could be compared by bud density and 100-bud weight.
As shown in Figure 1, more fresh shoots from the tea bushes were exposed in COS treatment than in that of the control group. In addition, the promoted emergence new and tender leaves could be compared by bud density and 100-bud weight.
Changes of Chlorophyll, Soluble Sugar, and Amino Acid Content in Tea Leaves
The effects of COS treatment on the contents of the total amino acids, chloroph and soluble sugar in the tea leaves were investigated. The values for those tested para eters in the COS group were obviously higher than in the control group ( Figure 2 Among them, the amino acids in the COS-treatment group were 60.77 mg/g FW a 39.86% higher than in the CK group (43.45 mg/g FW). Similarly, the chlorophyll conte in the COS treatment was 0.41 mg/g FW and 64% higher than in the control (0.18 mg FW), with great significance ( Figure 2B). In addition, the soluble sugar content of C treatment was 83.37 mg/g FW and increased by 17.76% compared to the control gro ( Figure 2C).
Changes of Chlorophyll, Soluble Sugar, and Amino Acid Content in Tea Leaves
The effects of COS treatment on the contents of the total amino acids, chlorophyll, and soluble sugar in the tea leaves were investigated. The values for those tested parameters in the COS group were obviously higher than in the control group ( Figure 2A). Among them, the amino acids in the COS-treatment group were 60.77 mg/g FW and 39.86% higher than in the CK group (43.45 mg/g FW). Similarly, the chlorophyll content in the COS treatment was 0.41 mg/g FW and 64% higher than in the control (0.18 mg/g FW), with great significance ( Figure 2B). In addition, the soluble sugar content of COS treatment was 83.37 mg/g FW and increased by 17.76% compared to the control group ( Figure 2C). Figure 1, more fresh shoots from the tea bushes were exposed in the COS treatment than in that of the control group. In addition, the promoted emergence of new and tender leaves could be compared by bud density and 100-bud weight.
Changes of Chlorophyll, Soluble Sugar, and Amino Acid Content in Tea Leaves
The effects of COS treatment on the contents of the total amino acids, chlorophyll, and soluble sugar in the tea leaves were investigated. The values for those tested parameters in the COS group were obviously higher than in the control group ( Figure 2A). Among them, the amino acids in the COS-treatment group were 60.77 mg/g FW and 39.86% higher than in the CK group (43.45 mg/g FW). Similarly, the chlorophyll content in the COS treatment was 0.41 mg/g FW and 64% higher than in the control (0.18 mg/g FW), with great significance ( Figure 2B). In addition, the soluble sugar content of COS treatment was 83.37 mg/g FW and increased by 17.76% compared to the control group ( Figure 2C).
Transcriptome Sequencing and Assembly
In this study, both COS and CK group samples were sequenced by the Illumina Nova seq platform. A total of 70-106 million raw reads were obtained for the COS group three repeats and 76 to 96 million raw reads for the CK group, respectively. After filtering, 66.2-100.0 million clean reads were obtained for the COS group and 70-90 million reads for the control. In all groups, Q20 was 100%, Q30 was above 99%, and the average GC contents of the CK group and COS group were also above 46%. The results indicated that the sequencing is of high quality and could be used for further analysis ( Table 2).
Differentially Expressed Gene (DEG) Analysis
The biological repeatability between samples was analyzed using Spearman analysis to verify the high correlation of gene expression levels between samples and proved that a perfect Spearman correlation occurred based on the RPKM (Reads per Kilobase per Million) of different samples. Genes with p-value < 0.05 and |log2 (Fold Change)| > 1 were defined as DEGs (genes that were differentially expressed between control and COS). A total of 5806 DEGs were identified between two groups of tea leave samples, of which 3262 up-regulated genes and 2544 down-regulated genes were found (Table S1).
Gene Ontology (GO) Annotation
Gene ontology (GO) provides the representation of gene product attributes and covers three domains: cellular component (CC), molecular function (MF), and biological process (BP). The GO annotation for the two groups revealed that a major part of DEGs was involved in BP, and MF covered DEGs with high rich factors ( Figure 3 and Table S2). The main CC categories (the parts of a cell or its extracellular environment) were "intracellular organelles" (GO:0043229), "cytoplasm" (GO:0005737)", and "organelles part" (GO:0044422), and so on. The main MF category (the elemental activities of a gene product at the molecular level) was the expression of genes related to transferase activity at the molecular level, such as "methyltransferase activity" (GO:0008168), "transferase activity" (GO:0008168), "serine hydrolase activity" (GO:0017171), and "ubiquitin-protein transferase activity"(GO:0004842). In terms of BP, the main categories focus on metabolic processes or molecular events, such as the "protein metabolic process" (GO:0019538), the phosphorus metabolic process (GO:0006793), and the cellular amino acid metabolic process (GO:0006520).
Kyoto Encyclopedia of Genes and Genomes (KEGG) Annotation
The degree of enrichment of KEGG was measured by the abundance factor, p-value, and the number of genes in the pathway. The significance of enrichment is shown on the horizontal coordinates; the greater the value (−log10 (p-value)), the more significant the enrichment. The KEGG pathway is shown on the longitudinal coordinates in Figure 4. The size of the dots indicates the number of different genes contained in the KEGG pathway, and the color of the dots indicates the degree of rich factor enrichment. These enriched pathways include "starch and sucrose metabolism" (ko00500), "glycolysis/gluconeogenesis" (ko00010), "biosynthesis of amino acids" (ko01230), "plant hormone signaling" (ko04075) ( Figure 4, Table S3). The DEGs were significantly enriched in pathways for "starch and sucrose metabolism", "phytohormone signal transduction", " Phenylpropanoid biosynthesis", "glycolysis/gluconeogenesis", "glycerophospholipid metabolism", "fatty acid metabolism", "cysteine and methionine metabolism", "the carbon fixation of photosynthetic organisms", "the biosynthesis of amino acids", "the biosynthesis of unsaturated fatty acids", and other genes closely related to plant growth and development, which were all significantly up-regulated.
Kyoto Encyclopedia of Genes and Genomes (KEGG) Annotation
The degree of enrichment of KEGG was measured by the abundance factor, p-value, and the number of genes in the pathway. The significance of enrichment is shown on the horizontal coordinates; the greater the value (−log10 (p-value)), the more significant the enrichment. The KEGG pathway is shown on the longitudinal coordinates in Figure 4. The size of the dots indicates the number of different genes contained in the KEGG pathway, and the color of the dots indicates the degree of rich factor enrichment. These enriched pathways include "starch and sucrose metabolism" (ko00500), "glycolysis/gluconeogenesis" (ko00010), "biosynthesis of amino acids" (ko01230), "plant hormone signaling" (ko04075) ( Figure 4, Table S3). The DEGs were significantly enriched in pathways for "starch and sucrose metabolism," "phytohormone signal transduction," " Phenylpropanoid biosynthesis," "glycolysis/gluconeogenesis," "glycerophospholipid metabolism," "fatty acid metabolism," "cysteine and methionine metabolism," "the carbon fixation of photosynthetic organisms," "the biosynthesis of amino acids," "the biosynthesis of unsaturated fatty acids," and other genes closely related to plant growth and development, which were all significantly up-regulated. The differentially expressed genes (DEGs) involved in response to COS were summed up and analyzed. Many DEGs are found related to pathways related to the growth and development of tea plants, including plant hormone signal transduction, starch, and sucrose metabolism, phenylpropanoid biosynthesis, glycolysis/gluconeogenesis, and carbon fixation in photosynthetic organizations, and biosynthesis of amino acids. The DEGs involved in the main metabolic pathways, including plant hormone signal transduction, carbohydrate metabolism, photosynthesis, amino acid metabolism, and other physiological and biochemical processes related to tea growth and development, are displayed in Tables S4-S9. These DEGs were characterized as genes encoding auxin-responsive GH3 gene family (GH3), SAUR family protein (SAUR), auxin influx carrier (AUX1), histidinecontaining phosphotransfer protein (AHP), alpha-amylase (AMY), sucrose synthase (SUS), fructose-1,6-bisphosphatase I (FBP), trehalose-phosphatase (otsB), sucrose phosphate synthase (SPS), sucrose synthase (SS), malate dehydrogenase (MDH), carbonic anhydrase (CA), glutamine synthetase (GLUL), glutamate dehydrogenase(GDH), tryptophan (trpB) synthase and so on. Among them, glutamine synthetase (GLUL) and tryptophan (trpB) synthase are DEGs specifically belonging to tea in biosynthesis amino acids. The differentially expressed genes (DEGs) involved in response to COS were summed up and analyzed. Many DEGs are found related to pathways related to the growth and development of tea plants, including plant hormone signal transduction, starch, and sucrose metabolism, phenylpropanoid biosynthesis, glycolysis/gluconeogenesis, and carbon fixation in photosynthetic organizations, and biosynthesis of amino acids. The DEGs involved in the main metabolic pathways, including plant hormone signal transduction, carbohydrate metabolism, photosynthesis, amino acid metabolism, and other physiological and biochemical processes related to tea growth and development, are displayed in Table S4-S9. These DEGs were characterized as genes encoding auxin-responsive GH3 gene family (GH3), SAUR family protein (SAUR), auxin influx carrier (AUX1), histidine-containing phosphotransfer protein (AHP), alpha-amylase (AMY), sucrose synthase (SUS), fructose-1,6-bisphosphatase I (FBP), trehalose-phosphatase (otsB), sucrose phosphate synthase (SPS), sucrose synthase (SS), malate dehydrogenase (MDH), carbonic anhydrase (CA), glutamine synthetase (GLUL), glutamate dehydrogenase(GDH), tryptophan (trpB) synthase and so on. Among them, glutamine synthetase (GLUL) and tryptophan (trpB) synthase are DEGs specifically belonging to tea in biosynthesis amino acids.
Validation of qRT-PCR (Quantitative Real-Time Polymerase Chain Reaction)
To validate the accuracy and repeatability of the transcriptome analysis, qRT-PCR ( Figure 5) was performed on a set of DEGs selected randomly from the above acquired DEGs. Results showed that the qRT-PCR expression profiles were consistent with their abundance changes identified by RNA-seq (Table S1), which verified the reproducibility and credibility of RNA-seq data. Results confirmed that AMY, SUS, SAUR, and GH3 were significantly upregulated by exogenous COS. To validate the accuracy and repeatability of the transcriptome analysis, qRT-PCR ( Figure 5) was performed on a set of DEGs selected randomly from the above acquired DEGs. Results showed that the qRT-PCR expression profiles were consistent with their abundance changes identified by RNA-seq (Table S1), which verified the reproducibility and credibility of RNA-seq data. Results confirmed that AMY, SUS, SAUR, and GH3 were significantly upregulated by exogenous COS. To examine metabolite patterns elicited by COS, metabolites were extracted from the leaves of the control and COS groups, followed by detection, respectively. Metabolomic analysis of the tea samples by UPLC-QTOF MS in positive mode (ESI + ) and negative mode
Metabolome Data Analysis 2.4.1. Effect of COS on Differential Changes in Tea Metabolites
To examine metabolite patterns elicited by COS, metabolites were extracted from the leaves of the control and COS groups, followed by detection, respectively. Metabolomic analysis of the tea samples by UPLC-QTOF MS in positive mode (ESI + ) and negative mode (ESI − ) mode detected. Principal components analysis (PCA) score plots (n = 6 for each control and COS-treated sample from tea plant)) revealed discriminative metabolic levels between the tea plant sample of COS and the control group in both pos and neg-modes ( Figure 6). The score scatter plots classified samples as COS treatment "HAI7_2" and control "KB7_1", with 47.57 and 55.91% of the variability contained in PC1 in positive and negative mode, respectively.
Effect of COS on Differential Changes in Tea Metabolites
To examine metabolite patterns elicited by COS, metabolites were extracted from the leaves of the control and COS groups, followed by detection, respectively. Metabolomic analysis of the tea samples by UPLC-QTOF MS in positive mode (ESI + ) and negative mode (ESI -) mode detected. Principal components analysis (PCA) score plots (n = 6 for each control and COS-treated sample from tea plant)) revealed discriminative metabolic levels between the tea plant sample of COS and the control group in both pos and neg-modes ( Figure 6). The score scatter plots classified samples as COS treatment "HAI7_2" and control "KB7_1", with 47.57 and 55.91% of the variability contained in PC1 in positive and negative mode, respectively. By setting the threshold for significantly differential metabolites (DMs) screening at variable importance in the projection (VIP) > 1.0, p < 0.05, a total of 416 DMs were obtained. The content of each metabolite was leveled to complete hierarchical linkage clustering as presented in the heatmap. Each sample was visualized in a single column (COS treatment sample: H7_2_1-H7_2_6; control: K7_1_1-K7_1_6), and each metabolite is represented by a single row [ Figure 7, Table S4]. Of which 108 DMs were significantly upregulated and 116 DMs were significantly downregulated in positive mode, 90 were upregulated, and 112 were downregulated in negative mode (Table 3). By setting the threshold for significantly differential metabolites (DMs) screening at variable importance in the projection (VIP) > 1.0, p < 0.05, a total of 416 DMs were obtained. The content of each metabolite was leveled to complete hierarchical linkage clustering as presented in the heatmap. Each sample was visualized in a single column (COS treatment sample: H7_2_1-H7_2_6; control: K7_1_1-K7_1_6), and each metabolite is represented by a single row [ Figure 7, Table S4]. Of which 108 DMs were significantly upregulated and 116 DMs were significantly downregulated in positive mode, 90 were upregulated, and 112 were downregulated in negative mode (Table 3).
KEGG Analysis of DMs Response to COS Treatment
KEGG pathways were organized and analyzed to explore the functions of COS on the DMs related to tea plant development (Figure 8). The color of the point represents the p-value of the hypergeometric test. The smaller the value, the more reliable and statistically significant the result is. The size of the dots represents the number of differential metabolites in the corresponding pathway; the larger size indicates a higher significant differential of metabolites in the pathway. It could be seen that the DMs were mainly enriched in the following six pathways, "biosynthesis of secondary metabolites," "Phenylpropanoid biosynthesis," "flavone and flavonol biosynthesis," "alpha−Linolenic acid metabolism," "Biosynthesis of unsaturated fatty acids," "Arachidonic acid metabolism," and "Glutathione metabolism" (Figure 8, Table S10). DMs enriched in these metabolic pathways may play an immediate physiological role and promote and influence growth and development. Among them, the most enrichment of differential metabolites pathway is the biosynthesis of secondary metabolites, and 33 DMs were found in it (Table S11), including amino acids (L-Histidine, L-Ornithine, L-Asparagine), organic acids (Loganic acid, Ferulic acid, Ascorbic acid), flavonoids (Rutin, Kaempferol, Quercetin), and other compounds. The above results showed that COS treatment could significantly increase the production of secondary metabolites in tea plants, thereby promoting the growth and development of tea plants.
KEGG Analysis of DMs Response to COS Treatment
KEGG pathways were organized and analyzed to explore the functions of COS on the DMs related to tea plant development (Figure 8). The color of the point represents the p-value of the hypergeometric test. The smaller the value, the more reliable and statistically significant the result is. The size of the dots represents the number of differential metabolites in the corresponding pathway; the larger size indicates a higher significant differential of metabolites in the pathway. It could be seen that the DMs were mainly enriched in the following six pathways, "biosynthesis of secondary metabolites", "Phenylpropanoid biosynthesis", "flavone and flavonol biosynthesis", "alpha−Linolenic acid metabolism", "Biosynthesis of unsaturated fatty acids", "Arachidonic acid metabolism", and "Glutathione metabolism" (Figure 8, Table S10). DMs enriched in these metabolic pathways may play an immediate physiological role and promote and influence growth and development. Among them, the most enrichment of differential metabolites pathway is the biosynthesis of secondary metabolites, and 33 DMs were found in it (Table S11), including amino acids (L-Histidine, L-Ornithine, L-Asparagine), organic acids (Loganic acid, Ferulic acid, Ascorbic acid), flavonoids (Rutin, Kaempferol, Quercetin), and other compounds. The above results showed that COS treatment could significantly increase the production of secondary metabolites in tea plants, thereby promoting the growth and development of tea plants.
Combination of the Metabolic Profiling and Transcriptomic Analysis
The transcriptomic analysis showed that DEGs in pathways of the plant hormone signal transduction, starch, and sucrose metabolism, phenylpropanoid biosynthesis, glycolysis/gluconeogenesis, and carbon fixation in photosynthetic organizations, and biosynthesis of amino acids of tea plants were significantly enriched in COS treatment. Metabolomics data indicated a total of 1339 differential metabolites were enriched in different metabolic pathways, including the most enriched biosynthesis of secondary metabolites and phenylpropanoid biosynthesis. A combined metabolic and transcriptomic analysis was performed to investigate the potential for the regulation of related metabolites. The result showed that the DEGs and DMs were significantly up-regulated in phenylpropane metabolism pathways. It may refer that the differential genes and metabolites in phenylpropane metabolism play an important role in the growth and development of tea plants. The DEGs and DMs in the phenylpropanoid biosynthesis pathway were summarized and sorted through the KEGG database (Table S12). The metabolic intermediates and end products of the phenylpropanoid biosynthesis pathway were mapped to the known KEGG pathway. The overall trends of the phenylpropanoid biosynthesis pathway suggested an increase both in metabolic intermediates (such as ferulic acid and coniferyl-aldehyde) and end products (such as coniferin and sinapyl alcohol) by spraying COS on tea plants (Figure 9). . The abscissa is x/y (x: number of differential metabolites in the corresponding metabolic pathway, y: total metabolites identified in this pathway). The higher the value, the higher the enrichment of differential metabolites in this pathway.
Combination of the Metabolic Profiling and Transcriptomic Analysis
The transcriptomic analysis showed that DEGs in pathways of the plant hormone signal transduction, starch, and sucrose metabolism, phenylpropanoid biosynthesis, glycolysis/gluconeogenesis, and carbon fixation in photosynthetic organizations, and biosynthesis of amino acids of tea plants were significantly enriched in COS treatment. Metabolomics data indicated a total of 1339 differential metabolites were enriched in different metabolic pathways, including the most enriched biosynthesis of secondary metabolites and phenylpropanoid biosynthesis. A combined metabolic and transcriptomic analysis was performed to investigate the potential for the regulation of related metabolites. The result showed that the DEGs and DMs were significantly up-regulated in phenylpropane metabolism pathways. It may refer that the differential genes and metabolites in phenylpropane metabolism play an important role in the growth and development of tea plants. The DEGs and DMs in the phenylpropanoid biosynthesis pathway were summarized and sorted through the KEGG database (Table S12). The metabolic intermediates and end products of the phenylpropanoid biosynthesis pathway were mapped to the known KEGG pathway. The overall trends of the phenylpropanoid biosynthesis pathway suggested an increase both in metabolic intermediates (such as ferulic acid and coniferyl-aldehyde) and end products (such as coniferin and sinapyl alcohol) by spraying COS on tea plants (Figure 9). . The abscissa is x/y (x: number of differential metabolites in the corresponding metabolic pathway, y: total metabolites identified in this pathway). The higher the value, the higher the enrichment of differential metabolites in this pathway.
Discussion
In the past few decades, numerous studies have confirmed that chitosan oligosaccharides play an important role in promoting plant growth and development [16,[21][22][23] and plant defense elicitation [13] in various crops. This present study was designed to provide more information and expand the understanding of how this elicitor COS works on tea plants, as in research [27,28] we have focused on in the past few years. We explored the function of COS on tea plants by integrating analysis of transcriptomics and metabolomics besides the yield indexes and some physiological parameters. The bud density, 100-budweight, and actual yield in the COS treatment group were significantly higher than those of the control group. The contents of total amino acids, chlorophyll, and soluble sugar in Figure 9. Schematic of a portion of phenylpropanoid biosynthesis pathway in tee plant. Among them, the differential metabolites are marked in red and green, red means up-regulated, and green means down-regulated. Differentially expressed genes are marked next to related enzymes in the form of a heat map. The gene scale bar corresponds to the range of relative transcriptional levels of enzymes.
Discussion
In the past few decades, numerous studies have confirmed that chitosan oligosaccharides play an important role in promoting plant growth and development [16,[21][22][23] and plant defense elicitation [13] in various crops. This present study was designed to provide more information and expand the understanding of how this elicitor COS works on tea plants, as in research [27,28] we have focused on in the past few years. We explored the function of COS on tea plants by integrating analysis of transcriptomics and metabolomics besides the yield indexes and some physiological parameters. The bud density, 100-budweight, and actual yield in the COS treatment group were significantly higher than those of the control group. The contents of total amino acids, chlorophyll, and soluble sugar in the tea plant were ascertainably improved t by 17.76%, 39.86%, and 64%, respectively. The actual yield (biomass) of tea was increased by 23.8%. These results concluded that COS could indubitably and significantly promote tea plant growth and yield. The chlorophyll content was remarkably increased in the COS-treated tea plant. It extensively corroborated the previous reports on tea plants [27] and other crops [16,17,35].
The chlorophyll content is an important index to measure the rate of plant photosynthesis [36]. COS could improve the content of photosynthetic pigments in various plants [37,38]. COS could induce the synthesis of phytohormones, such as gibberellins and auxin, to enhance growth and development [39] and enhance endogenous concentration [40]. Our present RNA sequencing indicated that COS activated some important metabolic activities and cell processes in tea plants. In addition. they significantly changed the expression level of genes involved in plant hormone signal transduction, photosynthesis, carbon metabolism, and amino acid metabolism. AUX1, GH3, and SAUR are the important genes encoding auxin, and the AHP gene which encodes cytokinin. Additionally, the DELLA protein (DELLA) and phytochrome interacting factor 3 (PIF3) encoding gibberellin were significantly upregulated. The above results show that spraying COS on tea plants can regulate plant growth by regulating the expression of genes related to plant hormones.
The synthesis of carbon compounds (activated by carbonic anhydrase) is directly related to photosynthesis and the growth and development of plants [41]. In the present research, COS upregulated photosynthetic organisms, starch and sucrose metabolism, and Glycolysis/Gluconeogenesis, as important components of plant carbon metabolism, played an important role in plant growth and development. The overexpressing of cytosolic FBP and SPS both resulted in the accumulation of sucrose, as reported [42,43]. Sucrose synthase (SUS) is a class of enzymes that are widely considered to be the main pathway for sucrose carbon to enter plant cell metabolism and play a key role in plant development. In our research, the chlorophyll content in the COS treatment group was increased and the genes encoding CA, FBP, SPS, and SUS were also significantly upregulated (Table S9). The above results indicated that COS application on tea plants could enhance the photosynthetic CO 2 fixation and the accumulation of photosynthetic assimilates, which as an endogenous correlation, further contributes to the improvement of the photosynthesis and growth of tea plants. Similarly, the content of soluble sugar in the COS treatment group was significantly higher than that in the control group.
Apart from the enhanced photosynthesis, the key genes involved in carbon and nitrogen metabolism were also upregulated by COS. It has been well-documented that carbon and nitrogen are both primary nutrients for plant growth and crop yields [44]. In the present study, genes encoding trehalose-phosphatase, PEPC, MDH, and glutamate dehydrogenase were significantly upregulated (Table S9), which was consistent with the results obtained in rice plants [45]. Studies have proved that the net accumulation of organic acids could act as a C-skeleton for the synthesis of amino acids in nitrogen assimilation [46]. In our study, amino acid metabolism was significantly activated, and key enzymes in related metabolic pathways were significantly upregulated, such as GLUL, trpB, GS, and GDH. Metabolomics data also showed up-regulated expression of amino acid compounds, such as L-Tyrosine, L-Histidine, L-Asparagine, and L-Ornithine (Table S4). The amount of total amino acids extracted from the tea leaves increased correspondingly in the COStreated group. DEGs upregulated in the amino acid pathway together with the association differential metabolites of amino acid are tea-specific responses for COS treatment. In brief, COS could appreciably enhance the gene expression concerned with photosynthesis, carbon metabolism, nitrogen, amino acid metabolism, and metabolic response modification in tea plants, especially secondary metabolism, hence promoting the growth of tea plants. These related genes are the potential targets of chitosan-mediated growth stimulation in various plants [44,47,48].
Altogether, in this study, the transcriptome and metabolome of tea leaves from the control and COS treatment were performed by association analyses. Information about the differential expressed genes and metabolites was obtained. Furthermore, the transcriptome datasets and analysis reported here will facilitate functional genomics, gene discovery, and transcriptional regulation of COS in tea plants. It provided a systemic insight into the mechanism of chitosan oligosaccharides on tea plant growth promotion and formed a theoretical basis for the application of chitosan oligosaccharides in tea plants.
Field Experiment, Sample Collection and Yield Estimate
Field tests were conducted at a tea garden in Meitan (Longitude: 107.48 latitude: 27.7) in Guizhou Province, China. The tested tea plants (Camellia sinensis var. Sinensis, Qianmei 601.) were around 10 years old and free of pesticide application during the test.
5% COS agents were purchased from Hainan Zhengye Zhongnong High-tech Co., Ltd., Haikou, China. The commercial agent was diluted with water 500, 800, and 1000 times in volume, and water consumption was 32 kg for one mu (667 m 2 ). The actual dose of COS was 3.2, 2.0, and 1.6 g/667 m 2 , respectively. The adjacent plot with no COS spraying served as a control (CK). There were three replicates in the randomized complete block for every spraying application in the tea garden. Seven days after spraying COS, yields in the control and COS-treated group were allowed to estimate in randomly selected tea plant rows in the center of the field. The number of buds in each of 0.1 m 2 (0.33 m × 0.33 m) was counted, and the mean bud density (buds per square meter) for each plot was calculated. The obtained tea buds were randomly selected, and the weight of the 100 tea buds was measured. All measurements were repeated three times. Finally, the tea buds picked in each treatment group were quickly frozen with liquid nitrogen and stored in a −80 • C freezer.
Determination of Chlorophyll, Soluble Sugar and Amino Acid Content
In the controls and COS-treated leaves, the content of chlorophyll soluble sugar and amino acids was measured and expressed as mg/g FW (fresh weight). Among them, chlorophyll and soluble sugar content were determined by following the provisions in bioassay kits, as previously reported [28]. The amino acid content was determined according to GB/T8314-2013 determination of total free amino acids in tea. The involved bioassay kits were purchased from Solarbio, Beijing, China.
Transcriptomic Profiling
RNA sequencing was used to analyze the transcriptome of samples from the control and COS treatment groups, treated at a concentration of 2.0 g/667 m 2 (n = 3 for both groups, six samples in total) as previous description [28]. The extraction of total RNAs was done by using TRIzol (Invitrogen, Carlsbad, CA, USA). The concentration of isolated RNA was determined by NanoDrop TM OneC spectrophotometer (Thermo Fisher Scientific Inc., Carlsbad, CA, USA). The quality and integrity of RNA were checked by running a 1.5% agarose gel. The cDNA synthesis was finished by using 2 ug DNAse-treated RNA using a cDNA synthesis kit (Thermo Fisher Scientific Inc., Carlsbad, CA, USA). The mRNA libraries were generated using KC-Digital TM stranded mRNA libraryprep kit for Illumina ® (Catalog NO. DR08502, Seqhealth Technology Co., Ltd., Wuhan, China), and the library quality was assessed on the Agilent Bioanalyzer 2100 system. The kit eliminates duplication bias in PCR and sequencing steps by using a unique molecular identifier (UMI) of eight random bases to label the pre-amplified cDNA molecules. The library products corresponding to 200-500 bps were enriched, quantified, and finally sequenced on Illumina Novaseq 6000 (Illumina, San Diego, CA, USA). RNA extraction, library preparation, and data analysis of high-throughput sequencing were completed by Kangce Technology Co., Ltd.
in Wuhan, China. For RNA-Seq data analysis, raw sequencing data were first analyzed by FastQC, and low-quality reads were filtered by Trimmomatic (version 0.36) and discarded. The reads contaminated with adaptor sequences were trimmed, and then clean reads were mapped to the reference genome of Camellia sinensis using STATR software (version 2.5.3a). Estimation of gene expression level using Fragments Per Kilobase Million (FPKM) method. All assembled unigenes were annotated in the following databases: GO and KEGG. Differentially expressed genes (DEGs) between the two groups were identified using edgeR software (version 3.12.1). FDR-corrected p < 0.05 with a fold change ≥ 2 was used to determine whether the differential genes were statistically significant. Differential gene GO, and KEGG data analysis was performed using KOBAS software (version 2.1.1) with a corrected p < 0.05 to determine if the enrichment was statistically significant. To infer the putative functions of DEGs, we conducted GO enrichment analysis using top GO and KEGG pathways and enrichment analysis using cluster Profiler(version R 3.6.0, October 2019).
Extraction and Detection of Metabolites
Tea leaves (100 mg) from the COS treatment (HAI7_2) and control (KB7_1) were ground with liquid nitrogen, and the homogenate was resuspended with prechilled 80% methanol and 0.1% formic acid using a vortex. The samples were incubated on ice for 5 min and then centrifuged at 15,000× g, 4 • C for 20 min. Some supernatants were diluted to a final concentration containing 53% methanol with LC-MS-grade water. The samples were subsequently transferred to a fresh Eppendorf tube and were then centrifuged at 15,000× g, 4 • C for 20 min. Finally, the supernatant was injected into the LC-MS/MS system analysis [49]. UHPLC-MS/MS analyses were performed using a Vanquish UHPLC system (Thermo Fisher Scientific GmbH, Dreieich, Germany) coupled with an Orbitrap Q Exactive TM HF-X mass spectrometer (Thermo Fisher, Dreieich, Germany) by Novogene Co., Ltd. in Beijing, China. The raw data files generated by UHPLC-MS/MS were processed using Compound Discoverer 3.1 (Thermo Fisher, Dreieich, Germany) to perform peak alignment, peak picking, and quantitation for each metabolite. Peaks were then matched with the mzCloud (https://www.mzcloud.org/, accessed on 26 October 2020), mzVault, and MassList databases to obtain accurate qualitative and relative quantitative results. Statistical analyses were performed using the statistical software R (R version R-3.4.3), Python (Python 2.7.6 version), and CentOS (CentOS release 6.6); when data were not normally distributed, normal transformations were attempted using of area normalization method. The metabolomic sampling method is the same as the transcriptome analysis experiment, and three biological replicates are set for each treatment.
For clustering heat maps, the data were normalized using z-scores of the intensity areas of differential metabolites and were plotted by pheatmap package in R language. The correlation between differential metabolites was analyzed by "cor ()" in R language (method = "Pearson"). Statistically significant correlation between differential metabolites was calculated by "cor. test ()" in R language. A p-value < 0.05 was considered statistically significant, and correlation plots were plotted by corrplot package in R language. The functions of these metabolites and metabolic pathways were studied using the KEGG database.
Validation of qRT-PCR
To verify the reliability of transcriptome sequencing results, we selected seven DEGs associated with tea plant defense for qRT-PCR analysis. Gene-specific primers (Table S13) were designed. Additionally, qRT-PCR was performed on a Light Cycler 96 System in a 20 mL reaction volume under the following parameters: 95 • C for 15 s, 60 • C for 30 s, and 72 • C for 30 s for 40 cycles. GAPDH was used as a reference gene, and relative gene expression was calculated using the 2 −∆∆Ct method [50].
Statistical Analysis
Data were expressed as the mean ± standard error, and the data were subjected to a one-way analysis of variance (ANOVA A) (p < 0.05) followed by a significant difference test (LSD) using SPSS statistics v16.0 (SPSS Inc., Chicago, IL, USA). Data Availability Statement: All data in the manuscript are available from the corresponding author upon request.
Conflicts of Interest:
The authors declare no conflict of interest. | 2022-05-16T15:04:52.907Z | 2022-05-01T00:00:00.000 | {
"year": 2022,
"sha1": "e15a80013bf956c8defb98419edf2e09d15f4ed5",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1422-0067/23/10/5469/pdf?version=1652839095",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "99f5a1a314bc3d3b62e2840c4e3681587b0a4b27",
"s2fieldsofstudy": [
"Environmental Science",
"Agricultural and Food Sciences"
],
"extfieldsofstudy": [
"Medicine"
]
} |
4784642 | pes2o/s2orc | v3-fos-license | Social Interventions to Prevent Heat-Related Mortality in the Older Adult in Rome, Italy: A Quasi-Experimental Study
This study focuses on the impact of a program aimed at reducing heat-related mortality among older adults residing in central Rome by counteracting social isolation. The mortality of citizens over the age of 75 living in three Urban Areas (UAs) located in central Rome is compared with that of the residents of four adjacent UAs during the summer of 2015. The data, broken down by UA, were provided by the Statistical Office of the Municipality of Rome, which gathers them on a routine basis. During the summer of 2015, 167 deaths were recorded in those UAs in which the Long Live the Elderly (LLE) program was active and 169 in those in which it was not, implying cumulative mortality rates of 25‰ (SD ± 1.4; Cl 95%: 23–29) and 29‰ (SD ± 6.7; Cl 95%: 17–43), respectively. Relative to the summer of 2014, the increase of deaths during the summer of 2015 was greater in UAs in which the LLE program had not been implemented (+97.3% vs. +48.8%). In conclusion, the paper shows the impact of a community-based active monitoring program, focused on strengthening individual relationship networks and the social capital of the community, on mortality in those over 75 during heat waves.
Introduction
Heat-related mortality is a growing public health concern worldwide because of the increased number of older adults highly susceptible to the effects of heat. In fact, heat-related mortality amongst those over 75 is becoming a major issue because of the high prevalence of frail individuals, who show a higher risk of death as a consequence of psycho-physical impairment and/or lack of socio-economic resources [1][2][3]. In this context, the heat waves that hit Southern Europe during the summers of 2003 and 2015 were the most relevant episodes in causing an increase of deaths in older adults [4][5][6][7][8][9][10][11][12]. Prevention programs implemented after the summer of 2003 seem to have had a slight impact on mortality during the milder summers, while no impact has been noted during the hottest summers [13]. Apart from age, gender is one of the main factors related to mortality during heat waves, with a higher impact being recorded for older women than for older men, though this differential seems to be reduced in the over 75 population [14,15].
Several factors prevent an accurate assessment of the reduction in the impact of heat waves following the implementation of heat-related mortality prevention programs; in practice, most prevention programs are aimed at the general population of a city (or even a country), so that most studies are based on historical research rather than on a comparison of those sections of a population that are included or not included in prevention activities [16,17]. Moreover, a comparison between different areas, even within the same city, can only be approximate because of locally different climatic conditions (temperature, pollution, humidity and other variables), which can affect the impact of heat waves on health [18,19]. Over the years, however, a reduction in heat-related mortality has been noted, probably arising from interventions, such as campaigns on healthy life-styles and on adaptation strategies during heat waves, which were aimed at the population at large [12,13].
The Long Live the Elderly (LLE) program has been running for 13 years in some Urban Areas (UAs) of central Rome [20]. It was started in 2004 with the purpose of mitigating the effects of extreme atmospheric phenomenon such as cold and/or hot spells, after the severe heat wave of 2003, which especially hit older adults in Italy and throughout Europe. The program is aimed at those over 75 and focused on strengthening social relations around isolated and/or sick individuals, in order to increase social capital at both individual and community levels. Social capital could be defined as resources embedded in a social structure that are accessed and/or mobilized in purposeful actions [21,22]. In fact, the role played by social isolation is well known as a risk factor for the occurrence of heat-related deaths, especially in people who are suffering from physical impairment due to various diseases [23][24][25]. At the same time, the presence of a supportive social environment has been associated with a reduced mortality [26][27][28]. To our knowledge, however, the impact on heat-related mortality of strengthening a weak social relationship network or even constructing such a network around socially-isolated individuals is not known. In other words, the question is whether a program aimed at identifying isolated and/or sick individuals, and then supporting them with social interventions, shows the same positive impact on the mortality due to extreme climatic episodes as a "natural" network of relationships. The purpose of this study is to assess the impact of the LLE program by comparing UAs in which the program is active with those within the same administrative district of the city where it is not.
Population and Settings
The study design is a quasi-experimental retrospective cohort study. The impact of the intervention program has been assessed by comparing the death rate between two populations living in the first administrative urban district of Rome ( Figure 1, Table 1). The two populations were divided between cases (6483 people residing in the UAs served by the LLE program) and controls (5724 people residing in the remaining UAs of the first district), comparable for age and gender.
Definition of Heat Waves
The definition of heat wave made by the Heat Health Watch Warning Systems (HHWWS), and implemented in Italy at the national level by the national prevention program of heat-related risk effects on population health, results from a combination of two factors: the air-mass model (a combination of air temperature, dew point temperature, atmospheric pressure, wind speed and direction) with the maximum apparent Temperature (T appmax ), which is a discomfort index based on air dew point temperature, calculated using the following formula [29,30]: T apparent = −2.653 + 0.994(T air ) + 0.0153(T dew point ) Both models have been matched with the retrospective analysis of mortality data [29][30][31] to define the thresholds associated with negative impacts on the population; Level 0 is made up by the values of both factors not associated with an increase of mortality. Level 1 is associated with an increase of mortality lower than 20% of the standard mortality for that period of the year. Level 2 is defined as the T appmax values associated with an increase of mortality higher than 20% or to an air-mass model associated with any increase of mortality. Level 3 is the persistence of Level 2 for three or more days [31]. The Civil Protection Department, which, together with the Italian Ministry of Health, implements the national program for the prevention of heat effects on health, releases alerts aimed at the general population living in 34 cities with more than 200,000 inhabitants; special attention is given to messages aimed at older adults. The program is modelled on national campaigns on life-style and on the identification of susceptible subgroups (mainly older adults requiring higher care needs, identified through clinical and administrative data managed by the local administrative bodies). Such subgroups are the target of local prevention activities. When a heat wave is declared, based on the information gathered through the HHWWS, GPs are supposed to telephone their patients, who would already have been included in the prevention plan, and visit them at home in if necessary. During the summer of 2015, 23 alerts (Level 2 or more) were recorded in the city of Rome, while during the years 2014 and 2016, there were 2 and 5 alerts, respectively [13].
The Intervention
In 2004, the LLE program, aimed at the population over 75, was implemented in the "Testaccio" and "Trastevere" UAs of Rome and in 2011 in that of "Esquilino"; the three main goals were: (a) to contact all those over 75 for the purpose of offering them a periodic assessment of their social and health needs, health promotion campaigns (i.e., "Tips for hot weather"), assistance in handling bureaucratic matters or seeking formal or informal care and to provide details of the office, active from 8 a.m.-5 p.m., Monday-Friday, to contact if necessary; (b) to strengthen the community network around sick and/or socially-isolated individuals by involving people living or working near them in volunteer care actions; and (c) to increase the awareness of the community about the needs of older adults.
The LLE program promotes a proactive approach to reach the whole targeted population, so as to avoid some individuals being neglected because of a lack of awareness of their care needs. Based on a list provided by the municipality, all those over 75 receive a letter and then a phone call to obtain their consent to be part of the program; the percentage of refusal is lower than 2%. If the client accepts, then a multidimensional evaluation of their care needs is performed, and the service begins. According to the risk of a negative event, as assessed by the multidimensional evaluation, an individual care plan is drafted (including the services needed, even if the program will not provide these services, but will facilitate their provision to the client), and the client is included on the list for periodical phone calls: the higher the risk of negative events, the more frequently the person will be called, with a maximum frequency of once every two weeks, unless specific actions need to be taken. The activities of the program are intensified when a heat wave occurs: all those over 75 assisted by the program are traced by phone, and if necessary, the staff intervene with a home visit, bringing food and/or medicines as necessary, or involving the client's network of relationships. It must be taken into account that a part of the population took holidays out of the city and that some of the declared residents were not in fact living at their official home addresses. Throughout the year, the operators act as a liaison between the older adults assisted by the program and the community, in order to increase the social capital of both. Healthy 75 year olds, however, are also contacted at least three times a year and during climatic emergencies, in order to monitor their situation.
Data Analysis
This analysis is focused on the death rate during the summer of 2015 and is based on data collected through standard procedures and periodically published by the Statistical Office of the Municipality of Rome (total population and total number of deaths by month, from June 2014-December 2016, in Rome's first administrative district). However, a special dataset for this study was provided to the authors by the Statistical Office of the Municipality (see the Supplementary Materials-file_data.xls): published data are usually aggregated by administrative district, but in this case, data were broken down according to the UAs that made up the first administrative district. Because of the limited and homogeneous geographical areas taken into consideration [18], it was assumed that climatic conditions were uniform in central Rome. No account was therefore taken of meteorological variations between the UAs. Moreover, the analysis does not focus on the impact of a single heat wave or on the accumulation of several heat waves, but on the whole summer period. The assumption is that the impact of heat waves on those sections of the population showing a high prevalence of chronic diseases, like older adults, could persist longer than at any other age, as is often the case with other diseases affecting older adults. This is one of the reasons resulting in hospitalizations, the duration of which is 40% greater for those over 75 than for the 45-64 age group, independent of the disease causing admission [32]. Moreover, illness caused by or worsened by weather conditions could have serious consequences, including death, several days after the end of the heat wave; in the case of a series of heat waves, it would be very difficult to differentiate between the direct effect of climatic conditions in summer and indirect effects, i.e., general poor health, mediating the impact of climate on an individual's health status in old age. All the deaths recorded from June-September 2015 are compared to those registered in the same period of 2014, in order to allow a homogeneous framework. Ideally, the period before and after an event should have been longer, in order to set up an average death rate as a baseline against which to measure the 2015 death rate variations. Unfortunately, the statistical office of the municipality could not supply death rates broken down by UA before May 2014.
Conceptual issues are raised by the variation of death rates where services have been implemented to take care of segments of the population, like those over 75, with a higher risk of death [13]. In fact, the analysis should take into account the potential protective effect of the program, as well as the potential mortality displacement effect due to the pre-summer weather condition. The winter of 2014-2015 saw an increase of mortality, mainly amongst those over 80 [10], with a potential impact on the size of the pool of individuals susceptible to extreme climate events in the following months. These factors have been taken into consideration by including in the analysis the pre-summer 2015 mortality (October 2014-May 2015) as a control factor (Tables 2 and 3). Moreover, to give a comprehensive picture of the mortality trend, the death rate observed in October-November 2015 and the variation with respect to the summer 2015 mortality are shown ( Table 2).
Age-specific death rates of those over 75 and the percentage variations of age-specific death rates for summer 2015 have been compared with those of summer 2014 in order to assess the impact on mortality of the greater number of heat alerts in 2015. A list of variables (see Supplementary Materials, Table S1) has been matched with the 2014-2016 death rates in order to select the ones showing a statistically-significant association to be included in the final multivariable model. Because of the well-known impact of socio-economic resources on mortality, several variables have been tested to include in the analysis as a proxy of the citizens' socio-economic status; among the data available per urban areas, the property tax valuation has been chosen because of its relationship with mortality. Differences between the average mortality rates in the UAs where the LLE program was active and those in which it was not have been assessed by non-parametric tests and univariate regression. A multivariable linear regression model has been applied to check differences in mortality variations between UAs adjusted to the pre-summer 2015 mortality and to those variables that showed a statistically-significant association with mortality, including the intervention as a dummy variable (presence vs. absence). All the analyses were carried out before and after weighting the UAs populations for their mean values during the summer 2015. With this procedure, ethical approval of the study was not required
Results
The two populations were made up of 6483 cases and 5724 controls, comparable for age and gender (Table 1). From June-September 2015, 4720 older adults were contacted by the LLE program, by phone; 2963 received more than one phone call (2.22 phone calls per person on average) and 502 (10.6% of the total number) more than four phone calls. The discrepancy between the official residents and the citizens contacted by the program is mainly due to a real residence different from the official one and to the absence of the citizen due to spending holidays outside the city. The authors assume that this discrepancy should not affect cases and controls differently. The LLE program operators carried out 660 interventions other than phone calls for 143 older adults (3.0%): most of these interventions were "visiting the patients when admitted to the hospital" (111, 16.2% of the interventions other than phone calls) and "bringing food/medicine" at home (101, 15.3%). Other frequent interventions were "contact relatives/friends" (88 cases) and "search for a paid assistant" (73 cases): these interventions all together represent about 54% of the total number during the summer of 2015.
During the summer of 2015, 167 deaths were recorded in the UAs in which the LLE program was active, and 169 in those in which it was not, implying cumulative mortality rates of 25 (SD ± 1.4; Cl 95%: 23-29) and 29 (SD ± 6.7; Cl 95%: 17-43), respectively. The increase of deaths during the summer of 2015, compared to the summer of 2014, was more prominent in UAs in which the LLE program had not been implemented (+97.3% vs. +48.8%, Table 2). This was also the case for the average increase of deaths in comparison with the mortality recorded during the months before the summer of 2015 (35.4% vs. 6.6% from October 2014-May 2015), as well as for the average reduction during the months of October and November 2015 (−35.6% vs. −12.5%). The standard deviation between UAs of the summer of 2015's average death rate was smaller in the UAs assisted by the LLE program than in the control UAs (Moses rank test <0.001- Figure 2). The difference of the summer and autumn percentage of cumulative mortalities between cases and controls ( Table 2: ∆1 + ∆2 = 19.1% vs. 71.0%) can be taken as a measure of the resilience of the two populations during heat waves: the higher the resilience, the lower the cumulative variation.
If the cases had had the same death rate as the controls, the expected number of deaths during summer 2015 would have been 192, implying that 25 deaths were averted, a 13% reduction in mortality. The multivariable regression model (Figure 3, Table 3), weighted by the number of residents in each urban area, adjusted for the proportion of those over 90 and the pre-summer of 2015 mortality per urban areas, confirmed the statistical significance of the lower increase of mortality recorded in the urban areas served by the LLE program (β = 0.217; p < 0.001). Interestingly, the mean property tax valuation is associated with the overall mortality from May 2014-December 2016 (Pearson correlation = 0.829, p = 0.021), but is not associated with either the 2015 or the summer of 2015 mortality (Table S1). The gender balance, as the percentage of females in the population, did not show a statistically-significant difference according to the presence or not of the LLE program (62.0% in the LLE program population vs. 58.5% among the controls). Likewise, no association was observed between the share of females per UA and the summer of 2015 mortality. A marginal effect of gender on the overall mortality (cumulative mortality from June 2014-December 2016) can be observed, most probably as an effect of the higher percentage of females in the population.
Discussion
An increase of mortality during the summer of 2015 was recorded in all the UAs, even if the percentage increase varied. The main differences observed among residents followed by the LLE program were the reduction of the expected mortality by about 13% during the summer, compared with residents in other UAs of central Rome, and the narrowing of the death rate variation among the UAs. The main factors affecting mortality were the age of residents in each UA, the pre-summer mortality and the implementation of the LLE program. There are many factors demonstrating the link connecting loneliness and a poor community network with higher mortality rates and vice versa [23][24][25][26][27][28]. In this case, we note a reduction in mortality, due to a program aimed at strengthening the individuals' social relations, as well as the social capital at the community level. The program acts as a shock-absorber, able to reduce the mortality variation by implementing monitoring and targeted care activities at the community level; these are intensified during heat waves. In such periods, all the participants are called at least once, especially during the 2015 summer compared with other summers, due to the higher number of alerts. The impact of the LLE program on mortality is shown also by the reduction of mortality recorded in the months following the heat waves (Table 2), due to the harvesting effect (excess of mortality in previous months) that was reduced where the LLE program was operating. The smoothing of the mortality curve has already been acknowledged as the result of adaptation, through a period of time, to climatic changes (i.e., the increased frequency of heath waves) and of the prevention campaigns rolled out in Rome [13]. However, this was a historical comparison, which also included years with no severe heat waves. Any effect of the prevention campaign during extreme climatic events (like the heat waves that occurred in 2015) was not reported. In this case, the authors noted a smoothing of the mortality curve among residents benefitting from the LLE program, with a reduction of the mortality peak, as well as of the cumulative variation of mortality during the whole year, which could be seen as a measure of the population's resilience to heat waves. Resilience can be defined as the capacity to cope with the consequences of disasters [33][34][35]: the resilience of a community is increased when the susceptibility to an extreme event is reduced, mainly by improving the residential environment, either by dealing with the social environment or with the physical environment. In this case, the smoothing of the mortality rate curve observed in the LLE program UAs can be considered a measure of the increase of resilience that results in a mitigation of the harvesting effect of summer heat waves.
The mean property tax valuation as a proxy of the residents' socio-economic status, was associated with mortality rates only if the whole period of observation was considered, while no association was observed with the 2015 mortality rates alone. This could be a further indirect demonstration of the positive impact of the LLE program: in fact, the provision of social services seems to break the link between higher mortality and lower socio-economic status reported by several authors [36][37][38]. The gender balance did not affect either the summer 2015 mortality or the 2015 increase of mortality, probably because the difference in the gender balance between cases and controls is small, close to 5% [13].
A longer period should be considered than the one more often allowed for in the analysis of heat wave events and their impact (mortality and morbidity) on populations of older adults. It was noted that the impact of cold snaps lasts up to three weeks after the event ends [39,40], while for heat waves, the results are generally not conclusive [41,42]. Since the causes of mortality amongst older adults are the same in either hot or cold weather (mainly cardiovascular and respiratory diseases), it is likely that the lag time could be similar in the two cases, at least for older adults and for similar durations of intense hot or cold spells. In fact, compared with younger people, the reaction to physical stress can be delayed in older adults, as it is to many other stimuli [43]. In Italy, the hospital length of stay is about 40% longer among those over 75 than among the 45-64-year-old age group, independently of the disease leading to admission [32]. It is likely that the onset, as well as the exacerbation of a disease caused by climatic events can have medium-/long-term consequences for older adults more frequently than is generally the case with younger people, especially when the interval between heat waves is short, in which case the negative effects caused by the first heat wave could still be active when the second begins. It is worth noticing that, during the summer of 2015, 23 alerts were recorded in Rome, which is on average one alert every six days. The increase of hospital admissions during hot days and heat waves has been reported [29,44]. However, there is also evidence of an undesirable correlation between temperature increases and hospitalization rates, or visit rates to emergency departments, by those over 65 (though there may well be a different temporal relationship from that between heat waves and morbidity in those over 75). The present study was not designed to take into account the impact of heat waves on hospitalization because it is based on data provided by the Municipality Statistical Office, which does not gather information about hospital care. Further ad hoc studies on the impact of social care on older adults' use of hospital care are needed in order to shed more light onto this issue.
The results of this study should be treated with caution, because of its limitations: first of all, the data reported here are based on the assisted people's administrative residence, which does not correspond to the place where they live. In fact, it is estimated that, in Rome, at least 10% of the people living in UAs are not resident, while the same percentage are resident, but living in a different place. Since mortality data are based on the official home address, some inaccuracies might occur, although the authors assume that they would affect all the UAs equally. A second limitation could be the lack of individual information on factors that can affect mortality, such as the presence of air conditioning in the house or the vacation time spent outside Rome during the summer of 2015. These conditions, which are likely to be enjoyed by the more affluent people among those under observation, can be taken into consideration by including in the model a proxy of the socio-economic status like the mean property tax valuation in a given UA. However, this variable did not affect the 2015 mortality, even if it showed a strong correlation with the overall 2014-2016 mortality. This could be an effect of the interventions, based on an assessment of multidimensional frailty, including socio-economic conditions, and focused on frail people, including those lacking socio-economic resources. The program probably reduces inequalities, providing support to those most in need, based on psycho-physical impairment and/or a lack of socio-economic resources. We do not know the immediate causes of deaths, even for those that occurred during heat waves, which are usually not so different from those under normal circumstances (mainly cardiovascular and/or respiratory diseases), so we cannot establish a direct relationship between heat waves and deaths that occurred during the summer. Again, the authors do not think this limitation should have different effects between different UAs.
Conclusions
In conclusion, the paper shows the impact of a community-based active monitoring program focused on strengthening individual relationship networks and the social capital of the community on the mortality of those over 75 during heat waves. The mitigation of mortality during the 2015 heat wave summer compared to non-heat wave summers is likely to be the outcome of the increase of resilience resulting from the implementation of the LLE program throughout the year and especially during climatic crises. Investments in social services associated with assessments of frailty carried out routinely in order to draft individual care plans could be a good strategy for improving people's health and quality of life, at least during heat wave episodes.
Supplementary Materials: The following are available online at http://www.mdpi.com/1660-4601/15/4/715/s1, File_data: a special dataset for this study, provided by the Statistical Office of the Municipality. Table S1: Variables tested against the mortality indicators (univariate linear regression).
Acknowledgments:
The authors wish to acknowledge the Statistic Office of the Municipality of Rome for the provision of population data, which made this analysis possible, and Michael Bennet for his valuable contribution to revising the language of the paper | 2018-04-26T23:46:28.678Z | 2018-04-01T00:00:00.000 | {
"year": 2018,
"sha1": "3f3bae66bba69174ff5e909e753b287ff9a0740d",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1660-4601/15/4/715/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "03c82ee3da79d630a9d372d6da00927314096899",
"s2fieldsofstudy": [
"Environmental Science",
"Medicine",
"Sociology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
92987607 | pes2o/s2orc | v3-fos-license | The connection between cosmology and neutrino physics
Cosmology provides a unique and very powerful laboratory for testing neutrino physics. Here, I review the current status of cosmological neutrino measurements. Future prospects are also discussed, with particular emphasis on the interplay with experimental neutrino physics. Finally I discuss the possibility of a direct detection of the cosmic neutrino background and its associated anisotropy.
Introduction
Neutrino physics provides one of the prime examples of the interplay between particle physics and cosmology. Because neutrinos are so abundant, the cosmic neutrino background contributes significantly to the cosmic energy density at all times, and therefore have a profound influence on the evolution of our Universe. At early times, around the epoch of neutrino decoupling at T ∼ 1 MeV, they influence the formation of light nuclei, and later they have an influence on cosmic structure formation. This also means that precision cosmology can be used to probe details of neutrino physics, such as the absolute value of neutrino masses, and the presence of light or heavy sterile neutrinos.
Here I will focus on the use of cosmology to probe the mass of neutrinos. As it turns out the main influence of light neutrinos on structure formation comes via their contribution to the energy density. At late times when at least some of the mass eigenstates are non-relativistic, the energy density in neutrinos can be quantified via the sum of neutrino masses, m ν . To a very good approximation the effect of neutrinos on structure formation can be described using just this one parameter, i.e. it does not matter how the mass is distributed internally between the different states (see [1,2] for a thorough discussion of this).
Interestingly, cosmology is sensitive to a different combination of mass eigenstates than other experimental probes. In neutrinoless double beta decay experiments, the important parameter is the coherent sum [3,4] m ββ = c 2 13 c 2 12 m 1 + c 2 13 s 2 12 m 2 e iφ 2 + s 2 13 m 3 e iφ 3 , which allows for phase cancellation. The current best upper bound on m ββ comes from the Heidelberg-Moscow experiment and is m ββ < 0.27 eV (90% C.L.) [5,6].
However, the most direct current upper bound on the neutrino mass comes from the final data analysis of the Mainz experiment, and yields m ν β ≤ 2.3 eV at 95% C.L. [7], where is effective parameter measured in beta decay spectra. In this case the mass states are weighed with their mixing with ν e .
Assuming just the three active neutrino species this corresponds to an upper bound on the sum of neutrino masses of ∼ 7 eV. The current neutrino temperature is T ν,0 ∼ 1.7 × 10 −4 eV so that any mass eigenstate heavier than this is non-relativistic at present. The contribution of neutrinos of any mass eigenstate, i, to the current energy density is given by Ω ν,i h 2 = m ν,i /93 eV, where Ω is the density parameter and h the Hubble parameter in units of 100 km s −1 . The total neutrino contribution to the energy density is therefore Ω ν,i h 2 = i m ν,i /93 eV, where the sum is over all non-relativistic states.
The current upper bound on the dark matter density is roughly Ω dm < ∼ 0.1 so that for m ν , neutrinos would make up a very large fraction of the dark matter density. However, this is strongly ruled out by observations because of the free streaming property of neutrinos. Being light particles, neutrinos are relativistic approximately until the epoch of matter radiation equality. This means that all neutrino structures inside the horizon at this epoch have been erased, and if neutrinos constituted all the dark matter structure formation would have been impossible. This possibility is clearly excluded, and the argument can be refined to set constraints on the neutrino mass using precision measurements of cosmic structure formation.
Current constraints on neutrino properties
The most important data set currently used to constrain cosmology is the measurement of the CMB anisotropy by the WMAP experiment [8]. However, some parameters do not have a significant impact on the CMB and must be constrained by adding other data. One example is the neutrino mass which leads to a suppression of fluctuation power on scales smaller than the free-streaming scale. In linear theory the relative change in the matter power spectrum is roughly with f ν = Ω ν /Ω m . The transition occurs smoothly around the free-streaming scale, given very approximately by [45] A detailed discussion of these issues can be found in [2]. This damping of power is best constrained by either large scale structure (LSS) surveys which directly measure the spectral shape of the matter power spectrum, or by measurements of the fluctuation power amplitude on small scales. The LSS surveys like SDSS are generally more robust because they do not explicitly rely on measurements in the very non-linear regime, but formally stronger constraints can be obtained by using data from the Lyman-α forest or from the cluster mass function.
A very large number of papers have been dedicated to the study of how neutrino properties can be constrained using such data (see [8,41,42] for an incomplete list). Most studies have focussed on two effective parameters, m ν , and the effective number of neutrino species, N ν . The latter parameter can be defined in a number of way, but the most widely used definition in this context is to have the total neutrino mass, m ν distributed among N ν species of equal mass. The standard model predicts N ν = 3.04 with the 0.04 coming from incomplete neutrino freeze-out and finite temperature effects around e + e − annihilation [9]. However, this number can be changed by the presence of sterile neutrinos, a finite neutrino chemical potential or other light particles such as axions.
Here I show an example calculation using the latest cosmological data: CMB measurements from WMAP-5 [8], as well as the matter power spectrum measured from the SDSS-DR7 LRG data [42]. In addition I add a prior from the latest estimate of the Hubble parameter [43].
Another important issue in cosmological parameter estimation is the complexity of the model parameter space. Most studies have focussed on neutrino constraints on top of the simplest vanilla ΛCDM model such that either m ν or N ν is fitted in addition to the usual cosmological parameters: Ω m , the matter density, Ω b , the baryon density, H 0 , the Hubble parameter, n s , the scalar spectral index, A s , the scalar amplitude, and τ , the optical depth to reionization. However, this approach may be too naive, it could well be that other physics like dark energy with a non-trivial equation of state, w, has to be added together with neutrino physics in the fit. Some papers have discussed the possible degeneracy between neutrino parameters and other cosmological parameters.
In Fig. 1 I show the contraint on m ν = m ν /N ν and N ν from WMAP-5, SDSS-DR7 LRG, and H 0 data. The mass constraint on each mass state is m ν < ∼ 0.2 eV at 2σ, compatible with other recent analyses. It should again be stressed that the bound can be strengthened by adding other data sets which are, however, less robust. The preferred value of N ν is larger than the value predicted by the standard model, again in accordance with other recent findings. However, N ν = 3.04 is within the 95% contour which is 3.03 < N ν < 7.59 and even though the result is interesting and could be pointing to the presence of for example a new light sterile state of small mass, the statistical significance is very weak.
In Fig. 2 I show an analysis with the same data, but now adding the equation of state of dark energy, w, as a free parameter. The figure shows a significant degeneracy between m ν , N ν , and w. Even though there is a degeneracy the formal bound on m ν does not change much. The same is true for the lower bound on N ν which is 2.97 at 95% C.L., but the upper bound is relaxed significantly from 7.59 to 8.70.
Future constraints on neutrino properties
New and much more precise CMB measurements will be provided by the Planck satellite, launched in May 2009 CHANGE. In the coming years there will also be a large number of new cosmological experiments designed for precision measurements of structure formation. There are two ways of improving current data, either measuring larger effective volumes, or going to higher redshifts where structures are 0.043 eV more linear. In practise many of the proposed experiments will do both. One example is measurements of weak gravitational lensing by the Large Synoptic Survey Telescope (LSST) which will have an effective volume many times larger than the SDSS and measure out to redshifts of order 1.5 [44].
One very powerful way to discriminate massive neutrinos from other physical effects acting to suppress fluctuations on small scales is to look at how the power spectrum evolves with time on scales comparable to the free-streaming scale. Unlike other effects the suppression produced by neutrinos is redshift dependent, and by measuring at different redshifts it is possible to disentangle the effect of massive neutrinos from other effects coming from e.g. dark energy.
This type of measurement can be achieved with the LSST because the redshift of source galaxies can be determined reasonably well from photometry. It is therefore possible to do tomographical measurements of strucuture formation. It has been estimated that using data from Planck and LSST it will be possible to probe the sum of neutrino masses down to below 0.1 eV at 95% C.L., i.e. it will be possible to tell the difference between the normal and the inverted hierarchy. Table 1 shows the huge increase in sensitivity which can be achieved by doing weak lensing tomography. By binning the LSST data an improvement on σ( m ν ) of almost a factor 4 can be achieved. Other large scale structure surveys in the coming decade may reach a comparable sensitivity.
On the other hand, improving CMB measurements beyond Planck will not have a big impact on the determination of m ν because in the context of m ν CMB data are mainly used to break degeneracies with other parameters. For N ν this is not the case because a change in N ν has a very direct impact on the CMB spectrum. This is caused mainly by the change in the time of matter-radiation equality. From a detailed analysis Ref. [46] estimated the sensitivity of Planck to be σ(N ν ) ∼ 0.3 in a very general 11-parameter model, almost an order of magnitude better than WMAP.
Non-linear structure formation and neutrinos
Many of the upcoming precision experiments will probe structure formation in the quasi-linear regime 0.1 h Mpc −1 < k < 1 h Mpc −1 , and to fully exploit their potential it is necessary to calculate observables in this range to the required precision. In terms of the power spectrum this is ∼ 1 − 2% for k > 1 h Mpc −1 .
N-body simulations of non-linear structure formation with neutrinos included have shown that neutrinos provide an additional, non-linear suppression of power beyond the usual linear theory ∆P/P ∼ 8Ω ν /Ω m . Fig. 3 shows the power spectrum suppression for various values of the neutrino mass. Very interestingly, the damping relative to a pure ΛCDM model is larger than in linear theory, an effect caused by the interplay between the neutrino thermal velocity and the gravitational virial velocity [39]. The maximum relative suppression (which does not occur at asymptotically high k) is given approximately by ∆P P ∼ −9.6f ν .
This additional suppression, which is unique to neutrinos, might provide a smoking gun signature for the presence of massive neutrinos. It will be almost impossible to mimic with other physics because it requires the presence of a second dark matter component with very specific thermal properties.
Measuring the relic neutrino background
Standard model physics likewise predicts the presence of a Cosmic Neutrino Background (CνB) with a well defined temperature of T ν ∼ (4/11) 1/3 T γ . While it remains undetected in direct experiments, the presence of the CνB is strongly hinted at in CMB data. The homogeneous CνB component has been detected at the 4-5σ level in the WMAP data (see e.g. [8,[10][11][12][13][14]). Furthermore, this component is known to be free-streaming, i.e. to have an anisotropic stress component consistent with what is expected from standard model neutrinos (see [15][16][17][18][19][20][21]). Finally the standard model neutrino decoupling history is also confirmed by Big Bang Nucleosynthesis (BBN), the outcome of which depends on both the energy density and flavour composition of the CνB.
While this indirect evidence for the presence of a CνB is important, a direct detection remains an intriguing, but almost impossible goal. The most credible proposed method is to look for a peak in beta decay spectra related to neutrino absorption from the CνB [22][23][24], although many other possibilities have been discussed [25][26][27][28][29][30][31][32]. The neutrino absorption method was first investigated by Weinberg [22], based on the possibility that the primordial neutrino density could be orders of magnitude higher than normally assumed due to the presence of a large chemical potential. Although a large chemical potential has been ruled out because it is in conflict with BBN and CMB [33][34][35][36][37], the method may still work and recently there has been renewed interest in detecting the CνB using beta unstable nuclei.
Although the direct detection of the CνB is already very challenging, one might speculate on the possibility that in the more distant future anisotropies in the CνB will be detectable.
Anisotropy of the background
In Ref. [38] the anisotropy of the CνB was calculated, and the result is shown in Fig. 4. For massive neutrinos the anisotropy increases dramatically on large scales and reaches O(1) for a mass of order 0.05-0.1 eV. Assuming that the CνB anisotropy is to be detected using a beta decay experiment the actual measured anisotropy will be a linear combination of the mass eigenstates, weighed with their mixing with ν e . Since the anisotropy grows very strongly with increasing mass, the main signal will almost solely come from the most massive eigenstate, even if it has very small mixing.
Conclusions
Cosmology remains one of the main tools for the study of neutrino physics. Currently cosmology provides the most stringent upper bound on the neutrino mass, and even though the exact number is model dependent a very conservative upper bound on the sum of neutrino masses can be put at m ν < ∼ 0.6−0.7 eV at 95% C.L. More aggressive use of data leads to more stringent bounds but in that case it is also necessary to rely on less well controlled effects from non-linear structure formation. The other important parameter which can be probed using structure formation data is the energy density in neutrinos at early times, quantified by the parameter N ν , the effective number of neutrino species. The current bound from the same set of data is 3.03 < N ν < 7.59 at 95% C.L. Intriguingly, the preferred value of N ν is consistently higher than the standard model value of 3.04, but not at more than approximately 2σ.
In the future a range of different experiments will improve the sensitivity to neutrino parameters. Most important for a precision determination of m ν will be measurements of the matter power spectrum using larger volumes and going to higher redshifts than current surveys. During the next decade the most precise data will probably come from weak lensing survey of the LSST telescope. Together with measurements of the CMB anisotropy by the Planck satellite it has been estimated that a 1σ uncertainty on m ν of ∼ 0.04 eV can be achieved. In the more distant future it may be possible to decrease this error bar significantly by large scale measurements of 21-cm fluctuations at very high redshift, using for example the proposed FFTT project [47]. This could in principle increase the sensitivity by another factor of a few, making a precise neutrino mass determination possible even for the normal hierarchy. However, it should be stressed that there are many currently unadressed systematics involves in this, and it remains unclear if 21-cm surveys can ever reach this sensitivity.
In conclusion, cosmology is an important and complementary laboratory for probing neutrino physics. Some neutrino parameters, like the neutrino mass, are in principle much easier to measure using precision cosmological data than in direct laboratory experiments. Furthermore, since cosmology is measuring a different effective mass quantity than beta or double beta decay experiments it remains an intriguing possibility that they will yield different and seemingly incompatible results. For example it may be the case that cosmology provides a stringent upper limit while for example a beta decay experiment shows positive evidence for a non-zero m ν . Such a possibility could point to non-standard physics such as right handed currents masking as a neutrino mass in the beta decay experiment. | 2010-03-22T10:33:43.000Z | 2010-03-22T00:00:00.000 | {
"year": 2010,
"sha1": "a01be06755e6d2f88fc1a6176a35834c5940a859",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "a01be06755e6d2f88fc1a6176a35834c5940a859",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
267136731 | pes2o/s2orc | v3-fos-license | Bisphosphonates do not affect healing of a critical-size defect in estrogen-deficient mice
Bisphosphonates (BP) are anti-resorptive drugs that are widely used to prevent bone loss in osteoporosis. Since inhibition of bone resorption will cause a decrease in bone formation through a process called coupling, it is hypothesized that extended treatment protocols may impair bone healing. In this study, β-tri‑calcium-phosphate (βTCP) ceramics were inserted into critical-size long bone defects in estrogen-deficient mice under BP therapy. The study assessed the benefits of coating the ceramics with Bone Morphogenetic Protein-2 (BMP2) and an engineered BMP2 analogue (L51P) that inactivates BMP antagonists on the healing process, implant resorption, and bone formation. Female NMRI mice (11–12 weeks of age) were ovariectomized (OVX) or sham operated. Eight weeks later, after the manifestation of ovariectomy-induced osteoporotic bone changes, BP therapy with Alendronate (ALN) was commenced. After another five weeks, a femoral critical-size defect was generated, rigidly fixed, and βTCP-cylinders loaded with 0.25 μg or 2.5 μg BMP2, 2.5 μg L51P, and 0.25 μg BMP2/2.5 μg L51P, respectively, were inserted. Unloaded βTCP-cylinders were used as controls. Femora were collected six and twelve weeks post-implantation. Histological and micro-computer tomography (MicroCT) evaluation revealed that insertion of cylinders coated with 2.5 μg BMP2 accelerated fracture repair and induced significant bone formation compared to controls (unloaded cylinders or coated with 2.5 μg L51P, 0.25 μg BMP2) already six weeks post-implantation, independent of estrogen-deficiency and BP therapy. The simultaneous administration of BMP2 and L51P (0.25 μg BMP2/2.5 μg L51P) did not promote fracture healing six and twelve weeks post-implantation. Moreover, new bone formation within the critical-size defect was directly linked to the removal of the βTCP-implant in all experimental groups. No evidence was found that long-term therapy with ALN impaired the resorption of the implanted graft. However, osteoclast transcriptome signature was elevated in sham and OVX animals upon treatment with BP, with transcript levels being higher at six weeks than at twelve weeks post-surgery. Furthermore, the transcriptome profile of the developing repair tissue confirmed an accelerated repair process in animals treated with 2.5 μg BMP2 implants. L51P did not increase the bioefficacy of BMP2 in the applied defect model. The present study provides evidence that continuous administration of BP does not inhibit implant resorption and does not alter the kinetics of the healing process of critical-size long bone defects. Furthermore, the BMP2 variant L51P did not enhance the bioefficacy of BMP2 when applied simultaneously to the femoral critical-size defect in sham and OVX mice.
Introduction
Osteoporosis is a major health issue in aging societies and affects hundreds of millions of people worldwide, approx.80 % among them post-menopausal women (Willers et al., 2022).Dysregulated bone turnover with a negative remodeling balance results in loss of bone mass and microarchitectural deterioration, leading to increased fragility and fracture risk (Zebaze et al., 2010;Parfitt et al., 1983;Compston et al., 2019;Liu et al., 2019).
Bisphosphonates (BP) are the most frequent therapy to prevent bone loss in post-menopausal osteoporosis (Liu et al., 2019;Bone et al., 1997;Cranney et al., 2002;Hosking et al., 1998;Liberman et al., 1995).BP bind with high affinity to bone mineral and exhibit long-term action on bone metabolism (Rodan and Fleisch, 1996;Russell, 2006).Bone turnover is efficiently attenuated by BP since the suppression of osteoclastic bone resorption upon cellular uptake of the drugs will lead, through a coupling mechanism, to a decrease in bone formation (Luckman et al., 1998;Weinstein et al., 2009;Hughes et al., 1995).This "frozen" bone state in extended treatment protocols may impair bone healing in elderly patients that are prone to fractures (Diab and Watts, 2012;McClung et al., 2013;Watts and Diab, 2010;Odvina et al., 2005;Molvik and Khan, 2015).Furthermore, large bone defects frequently require the implantation of bone grafts, such as calcium phosphate-based ceramics like β-tricalcium phosphate (β-Ca 3 (PO 4 ) 2 , βTCP) (Galois et al., 2002;Chazono et al., 2004;Bohner, 2000).The manufacturing process of βTCP grafts enables the modification of pore size and connectivity, mechanical stability, and porosity.This allows optimization of the biomaterial to support the invasion by bone cell lineages and stimulate angiogenesis (Bohner et al., 2020;Malhotra and Habibovic, 2016).Essentially, calcium phosphate-based materials are already approved for clinical application and widely used in orthopedic surgery and dentistry (Dorozhkin, 2010;Rolvien et al., 2017;Roca-Millan et al., 2022;Tanaka et al., 2017;Guillaume, 2017).
The replacement of grafted materials by newly formed bone is essential to ensure the restoration of the mechanical stability of the damaged site.BP inhibit osteoclast-mediated bone resorption, and possibly block implant resorption.Various in vivo approaches studied the effects of BP on fracture healing in rodents and reported varying outcomes, ranging from delayed bone healing to superior mechanical stability of the defect tissue (Gerstenfeld et al., 2009;Hauser et al., 2018a;Hadjiargyrou, 2022).However, in several rat studies, delayed bone unions and impaired healing were observed in the presence of BP (Li et al., 2001;Li et al., 1999;Kidd et al., 2011;Manabe et al., 2012).In particular, ALN treatment caused delayed bone remodeling and implant removal, as well as increased callus sizes in rats (Cao et al., 2002;Fu et al., 2013;Hauser et al., 2018b).
Since the capacity of bone to heal decreases with age, bone grafts were combined with osteoinductive growth factors to support bone regeneration (Roddy et al., 2018;Schmidmaier et al., 2009;Newman and Benoit, 2016).Recombinant human Bone morphogenetic proteins (BMP) like BMP2 are approved for the treatment of non-healing defects and the stimulation of bone fusions (Lowery and Rosen, 2018;Gillman and Jayasuriya, 2021;Gautschi et al., 2007).The spatial and temporal availability of the administered BMP2 might not be optimal to support the healing process and the upregulation of BMP antagonists can further decrease the bioefficacy of the applied growth factor.To enhance the bioavailability of exogenously administered BMP2, supraphysiological doses are used in clinical procedures to compensate for the short action of the growth factor owing to the fast release kinetics and diffusion (Mesfin et al., 2013;Canalis et al., 2003;Dean et al., 2010).Increasing concentrations of the applied BMP2, thus, did not always lead to sufficient bone formation and were even associated with inflammatory responses, ectopic bone formation, and osteolysis (James et al., 2016;Carragee et al., 2011;Nguyen et al., 2017).
A possible strategy to overcome the limitations of the low bioefficacy of BMP2 is presented by the synthetic BMP2 variant L51P, with a leucine to proline substitution at amino acid position 51 interrupting the major binding site with the BMP receptor type I. L51P exhibits high binding affinity to BMP antagonists like noggin, chordin and gremlin without the activation of BMP signaling (Keller et al., 2004).L51P efficiently blocked BMP antagonists in vitro (Keller et al., 2004;Albers et al., 2012;Khattab et al., 2014) and enhanced the bioactivity of BMP2 in vivo (Hauser et al., 2018b;Khattab et al., 2014;Sebald et al., 2012;Khattab et al., 2019).Moreover, L51P reduced the necessary amount of BMP2 to induce bone formation in a rat calvaria defect model (Khattab et al., 2014) and in critical-size defects fitted with βTCP ceramic implants (Hauser et al., 2018b;Sebald et al., 2012).
The present study aimed to evaluate implant resorption and bone formation during the healing process of a femoral critical-size defect in estrogen-deficient mice fitted with βTCP-ceramics.The bioefficacy of BMP2, combined with the BMP variant L51P, with BP therapy and estrogen-depletion, was assessed.
Design of the animal study
This study was performed in accordance with Swiss Federal regulations and approved by the Cantonal Veterinary Office (permit numbers BE86/18 and BE84/21 to WH).Animals were kept under specific pathogen-free conditions in the Central Animal Facility of the University of Bern.Female outbred NMRI (Han) mice (11-12 weeks, Charles River, Sulzfeld, GER) were assigned to one of the 20 experimental groups representing the combinations of surgical procedures, treatments, and two time points: sham/ovariectomy (OVX), Vehicle (Veh)/Alendronate (ALN) therapy, implant coating (0 μg BMP2/0 μg L51P; 2.5 μg L51P; 0.25 μg BMP2; 0.25 μg BMP2/2.5 μg L51P; 2.5 μg BMP2).Samples were collected six and twelve weeks after application of the femoral criticalsize defects.
The experimental setup of the study is depicted in Fig. 1.Briefly, animals underwent either OVX or sham surgery (Hauser et al., 2018a).Veh/ALN treatment was initiated eight weeks later, and continued until sacrifice of the animals (Sato et al., 1991).Five weeks after the onset of Veh/ALN treatment, a critical-size defect (3.5 mm) was applied in the left femur of all animals.The defect was filled with βTCP-cylinders loaded with 2.5 μg L51P, 0.25 μg BMP2, 0.25 μg BMP2/2.5 μg L51P, 2.5 μg BMP2 or unloaded controls (0 μg BMP2/0 μg L51P).The defect site was rigidly fixed using a titanium osteosynthesis system (MouseFix™ plate 6 hole, RIS.401.130;RISystem AG, Davos, CH).Animals were sacrificed six and twelve weeks after setting the defect.Groups for histology and MicroCT included 2-5 animals (101 and 105 animals in total, respectively), and for RNA sequencing 3 animals (96 animals in total).
Ovariectomy
The removal of ovaries (OVX) to generate the estrogen-deficient mouse model for postmenopausal osteoporosis was performed as previously described (Hauser et al., 2018a).Briefly, two 0.5 cm flank incisions at the mid-dorsum were applied to locate the ovaries.During the OVX procedure, the oviducts were ligated with an absorbable thread (Coated VICRYL® 5-0, Ethicon, Zug, CH), and the ovaries were removed.Subsequently, the peritoneum was closed with absorbable sutures, and the skin with non-absorbable polypropylene threads (PROLENETM 5-0, Ethicon).Sham animals underwent the identical surgical procedure without ligation of the oviducts and removal of the ovaries.At the end of the study, uteri were collected, and uterus dry weight was assessed after drying in air for 24 h.
The human recombinant BMP2 and L51P proteins were expressed in Escherichia coli and were kindly provided by Prof. W Sebald (University of Würzburg, GE).
Femoral critical-size defect
The critical-size defect was created in the left femur of all animals 13 weeks after sham/OVX surgery and five weeks after the onset of Veh/ ALN therapy, as described previously (Manassero et al., 2012).After anesthesia, the skin was shaved and disinfected, and a longitudinal incision was made into the left thigh parallel to the femur.In the next step, a six-hole titanium plate developed for osteosynthesis in murine models was mounted onto the femur with four interlocking screws into the two outer plate holes on both sides.Subsequently, to create a reproducible mid-femoral critical-size defect (3.5 mm), a selfconstructed guiding tool was fixed to the two inner screws.Two Gigli saws were used to cut two osteotomies, and the defect site was rinsed with sterile physiological saline solution.To bridge the bone defect, βTCP-cylinders with varying coatings of BMP2 and L51P were press-fit into the defects and fixed with a non-absorbable thread.Subsequently, the thigh muscle tissue was realigned and stitched with an absorbable thread, and the wound was closed with non-absorbable sutures.To support the recovery from the surgery and the anesthesia, an antidote mixture was injected, and analgesia was applied for 72 h.Animals were monitored daily according to a pre-defined score sheet.Load bearing or movement was not restricted after surgery.
The integrity of the surgical site and the placement of the βTCPimplant were examined post-operatively by high-resolution radiography (MX-20, Faxitron X-Ray Corporation, Edimex, Le Plessis, FR) (Suppl.Fig. 1).In total, 260 animals underwent surgery; 17 mice were excluded from the analysis because of incorrect plate or implant positioning, and five animals were euthanized due to health issues independent of the surgical protocol.
MicroCT
Tissues for MicroCT analysis and histology were fixed in 4 % paraformaldehyde in phosphate-buffered saline for 24 h, rinsed with tap water, and transferred to 70 % ethanol.Femora with a critical-size defect were examined by MicroCT analysis (MicroCT40, SCANCO Medical AG, Brüttisellen, CH), by placing the long axis of the bone orthogonally to the axis of the X-ray beam (Bouxsein et al., 2010).The Xray tube was operated at 70 kVp and 57 μA with an integration time of 300 ms.The analysis was performed at a resolution of 8 μm.The composition of the defect tissue, including the total volume of the repair tissue, the volume of mineralized tissue consisting of newly formed bone and remaining implant material, and the bone volume/total volume (BV/TV) was assessed using the built-in software from Scanco (Scanco Module 64-bit; V5.15).
The examination of vertebrae was carried out with identical settings of the X-ray tube at a resolution of 6 μm.The measurements were recorded perpendicularly to the longitudinal axis of the vertebrae.The region of interest in the center of the vertebral body of the lumbar vertebra four (L4) was defined manually by using the built-in software from Scanco.
Histology and histomorphometry
For histomorphometric analysis, tissue samples were embedded in methyl methacrylate as described previously (Wernike et al., 2010).
Ground sections of approx.600-800 μm were cut with a diamond saw (Leco VC-50, Leco Corporation, St. Joseph, Michigan, USA).After grinding to a thickness of approx.300 μm and polishing, sections were stained with toluidine blue mixed with MacNeal's tetrachrome solution and fuchsin red (0.05 %).Images were captured with a digital microscope (VHX-6000, Keyence, Mechelen, BE).The quantification of the remaining implant material and the newly formed bone in the defect site was done by manual selection (Photoshop, version 24.0) based on color gradients (Dempster et al., 2013).
RNA sequencing 2.9.1. Isolation of total RNA
For RNA extraction and library preparation, the tissue between the two inner central screws containing the remaining implant material and newly formed bone tissue was dissected, placed in RNALater® (Sigma-Aldrich), and stored at − 20 • C. For total RNA extraction, samples were transferred into microtubes containing 1 ml TRIzol™ Reagent (Invi-trogen™) and metal beads for tissue disruption in a benchtop tissue homogenizer (TissueLyser 2, Qiagen, Hilden, DE).
RNA was isolated according to the manufacturer's instructions using the NucleoSpin® RNA Plus Kit (Macherey-Nagel, Oensingen, CH), and genomic DNA was removed by digestion with the provided DNase.The quantity and quality of the purified total RNA was assessed using a Thermo Fisher Scientific Qubit 4.0 fluorometer with the Qubit RNA BR Assay Kit (Q10211, Thermo Fisher Scientific, Waltham, Massachusetts, USA) and an Advanced Analytical Fragment Analyzer System using a Fragment Analyzer RNA Kit (DNF-471, Agilent Technologies, Santa Clara, California, USA), respectively.
F. Strunz et al.
Preparation of libraries, sequencing, and mapping to reference genome
In total, 32 experimental groups with 3 biological replicates (96 samples in total) each underwent RNA sequencing.Sequencing libraries were prepared with 500 ng input RNA using an Illumina TruSeq Stranded mRNA Library Prep kit (20020595, Illumina, San Diego, California, USA) in combination with TruSeq RNA UD Indexes (20022371, Illumina) according to the manufacturer's guidelines.Pooled cDNA libraries were sequenced paired-end using a shared Illumina NovaSeq 6000 S4 Reagent Kit (300 cycles; 20028312, Ilumina) on an Illumina NovaSeq 6000 instrument.This run was performed in the NovaSeq Xp workflow using a NovaSeq XP 4-Lane Kit v1.5 (20043131, Illumina).On average, the run produced 32.7 million reads/library.The quality of the sequencing run was assessed using Illumina Sequencing Analysis Viewer (lllumina version 2.4.7), and all base call files were demultiplexed and converted into FASTQ files using Illumina bcl2fastq conversion software v2.20.The quality control assessments, generation of libraries, and sequencing were conducted by the Next Generation Sequencing Platform of the University of Bern.The quality of the sequencing data was assessed by FastQC version (v.) 0.11.9 (Babraham Bioinformatics -FastQC A Quality Control tool for High Throughput Sequence Data, n.d.).The reference genome (assembly GRCm39) and associated annotation were obtained from the Ensembl genome database (Ensembl genome browser 109, n.d.).The reference genome was indexed using STAR (v.2.7.10) (Dobin et al., 2013).STAR was also used to calculate the table of counts containing the number of reads per gene.
Differentially expressed genes and GO enrichment analysis
Principal component analysis was performed based on the 500 genes with the most variable expression over the whole transcriptome to reveal differences with respect to estrogen-deficiency, BP therapy, time point after implantation, and the availability of BMP2 and L51P.The differential gene expression analyses to determine the effects of BP therapy, estrogen-deficiency, and implant coating were carried out using the Bioconductor package DESeq2 (v.1.32.0)(Love et al., 2014) in R (v. 4.1.0)(R: The R Project for Statistical Computing, n.d.).To increase stringency, only differently expressed genes (DEG) with a two-fold differential expression were considered for further analysis.Differences in expression with a false discovery rate smaller than 0.05 were considered significant.Gene ontology (GO) enrichment analysis was performed when >100 genes were significantly differentially expressed.To identify GO terms, the enrichGO function from the clusterProfiler package (v.4.0.5)(Wu et al., 2021) was used, along with the Bioconductor package org.Mm.eg.db (Carlson, n.d.) containing the genome-wide annotation for mouse.Changes in transcript levels encoding osteoclast and osteoblast traits with respect to BP therapy and time points six and twelve weeks after application of the defect were investigated.After removing the dependence of the variance on the mean using DESeq2::vst(), heatmaps of selected marker genes were produced using pheatmap (v.1.0.12)(Kolde, 2019).
Statistical analysis
The statistical analyses were performed using GraphPad Prism 9 for Windows (GraphPad Software, San Diego, California, USA).The unpaired t-test was used for the analysis of body weight and uterus dry weights, and one-way or two-way analysis of variance (ANOVA) with Tukey post-hoc was applied to analyze MicroCT and histomorphometric data.P values smaller than 0.05 were considered significant.
Characterization of the post-menopausal osteoporosis model and the effects of BP therapy
The suitability of the murine post-menopausal osteoporosis model in NMRI mice was assessed by analyzing body weight, uterus dry weight, and bone architecture upon estrogen-deficiency.The average body weight was significantly increased in OVX animals at 13, 19, and weeks after OVX surgery when compared to sham animals (p < 0.01) (Suppl.Fig. 2a).The average uterus dry weight of OVX animals was significantly lower compared to sham controls (p < 0.001) (Suppl.Fig. 2b).
The effects of OVX and ALN therapy on bone mass and structure were assessed by MicroCT analysis of the vertebral body of L4 in samples harvested 19 and 25 weeks after OVX/sham surgery, equivalent to and 17 weeks after the onset of BP/Veh therapy, respectively (Suppl.Fig. 3).The BV/TV of L4 was decreased by 32.4 ± 26.0 % (p < 0.01) in OVX/Veh animals when compared to sham/Veh controls.ALN therapy increased the BV/TV on average by 55.3 ± 35.0 % (p < 0.01) in the OVX/ALN animals as compared to the OVX/Veh group.In sham/ALN animals, the BV/TV was 56.1 ± 22.8 % (p < 0.001) higher than in the sham/Veh group.BP therapy preserved bone mass and structure in the vertebral body of L4 upon estrogen-depletion.Treatment with ALN led to significantly higher BV/TV in both sham and OVX animals, a significant difference, however, remained between the sham/ALN and OVX/ ALN groups, with lower values in estrogen-deficient animals.
Bone healing process in the femoral critical-size defects
For the time point of six weeks post-surgery, implant coatings 0.25 μg BMP2, 0.25 μg BMP2/2.5 μg L51P and 2.5 μg BMP2 were included in the experimental design and loading with 0.25 μg BMP2 was used as the negative control.For the time point twelve weeks post-surgery, all implant coatings were included in the evaluation of bone formation and implant removal by histology and MicroCT analysis.Unloaded cylinders together with implants coated with 2.5 μg L51P and 0.25 μg BMP2 were used as negative controls to assess the induction of bone growth by 2.5 μg BMP2 and 0.25 μg BMP2/2.5 μg L51P implants.
Histological analysis of the femoral repair site
The healing process of the critical-size defect was visualized histologically six (Fig. 2) and twelve (Fig. 3) weeks post-implantation.Bone formation developed from the proximal and distal sides of the bone defect with endosteal and periosteal reactions.The press-fit integration Fig. 2. Histological analysis of the femoral repair site six weeks post-implantation.Ground sections were prepared from the methyl methacrylate-embedded repair tissues isolated six weeks after the application of the critical-size defect in sham and OVX animals with Veh and ALN therapy.Bone defects were fitted with βTCP-ceramics loaded with 0.25 μg BMP2 (a-d), 0.25 μg BMP2/2.5 μg L51P (e-h), or 2.5 μg BMP2 (i-l).Sections were stained with a mixture of toluidine blue and MacNeal's tetrachrome solution.The coating with 0.25 μg BMP2 did not induce bone healing in sham and OVX animals (a-d).The combination of 0.25 μg BMP2/2.5 μg L51P caused minor bone formation coupled with implant removal in the sham and OVX groups (e-h).Implants loaded with 2.5 μg BMP2 led to the strongest induction of bone healing in all experimental groups (i-l).The femur is oriented with the distal end on the left side and the proximal part on the right.The scale bar represents 1 mm.and stable fixation of the βTCP-ceramics promoted the direct ingrowth of bone tissue into the interconnected pores of the graft material.
Six weeks post-surgery, coating with 0.25 μg BMP2 did not induce bone formation within the defect in sham and OVX animals, independent of BP therapy (Fig. 2a-d).Twelve weeks post-surgery, 0.25 μg BMP2 induced bone healing from the defect ends in sham animals (Fig. 3i, j), whereas OVX groups showed minor bone formation (Fig. 3k, l).Additional control groups at the time point twelve weeks, unloaded (Fig. 3a-d) and 2.5 μg L51P (Fig. 3e-h), displayed minimal bone formation within the defect in all animal groups.In contrast, ceramics loaded with 2.5 μg BMP2 induced bone healing and implant removal independent of estrogen-depletion and BP treatment.Moreover, newly formed bone emanating from the adjacent bone cortex was growing around the βTCP-implant, nearly reaching from the proximal to the distal defect end both six (Fig. 2i-l) and twelve (Fig. 3q-t) weeks postimplantation.
Six weeks post-implantation, the combination of 0.25 μg BMP2/2.5 μg L51P caused minor bone formation, mainly arising from the proximal defect end in sham and OVX animals (Fig. 2e-h).Twelve weeks postimplantation, 0.25 μg BMP2/2.5 μg L51P induced minimal bone growth in OVX animals (Fig. 3o, p), whereas considerable bone formation coupled with implant removal was detected in sham animals (Fig. 3m, n).In general, the histological analysis showed that only 2.5 μg BMP2 had osteoinductive effects leading to enhanced bone formation within the critical-size defect, whereas all other coatings did not stimulate the healing process.
Histomorphometry of the developing repair tissue
Histomorphometric analysis was performed to assess newly formed bone within the defect site (Fig. 4a, b), remaining implant material (Fig. 4c, d), and mixed tissue representing implant material mixed with ingrown bone (Fig. 4e, f).Six weeks post-implantation, significant differences between the 0.25 μg BMP2 group and other coatings were observed only in the OVX/ALN group.The exposure to 2.5 μg BMP2 resulted in 3-fold higher bone formation in estrogen-deficient animals under BP therapy (p = 0.0069) (Fig. 4a).Twelve weeks postimplantation, no significant differences in the formation of bone were detected within the defect in sham and OVX animals, independently of BP therapy or implant coating (Fig. 4b).However, in sham animals under BP treatment, coating with 2.5 μg BMP2 and 0.25 μg BMP2/2.5 μg L51P resulted in 2-fold and 3-fold higher bone formation, respectively, compared to negative controls (unloaded, 2.5 μg L51P, 0.25 μg BMP2).Six weeks post-implantation, coating with 2.5 μg BMP2 significantly reduced the remaining βTCP-material by almost 50 % in OVX animals with Veh (p < 0.05) and ALN (p < 0.01) therapy compared to 0.25 μg BMP2 (Fig. 4c).
Twelve weeks post-implantation, coating with 2.5 μg BMP2 and 0.25 μg BMP2/2.5 μg L51P resulted in an approx.60 % decrease of the remaining βTCP-implant in OVX/Veh animals compared to defects fitted with 2.5 μg L51P or 0.25 μg BMP2 cylinders (p ≤ 0.05) (Fig. 4d).Six weeks post-implantation, the analysis of the tissue fraction comprised of implant material with ingrown bone, revealed a nearly 5-fold increase of the mixed tissue in OVX/Veh animals treated with 2.5 μg BMP2 when compared to animals treated with 0.25 μg BMP2 implants (p = 0.01).A similar effect was observed in OVX/ALN animals, which received 0.25 μg BMP2/2.5 μg L51P implants (p < 0.05) (Fig. 4e).Twelve weeks postimplantation, loading with 0.25 μg BMP2/2.5 μg L51P led to 5-fold increase in the volume of mixed tissues in the sham/ALN group compared to animals treated with unloaded controls (p < 0.05) (Fig. 4f).The detected amount of mixed tissue correlated with new bone formation and higher implant removal.The histomorphometric analysis showed that the removal of the βTCP-material was directly coupled to the amount of bone formation within the critical-size defect.It further confirmed the osteoinductive effect of implants coated with 2.5 μg BMP2.
MicroCT analysis of the developing repair tissue
MicroCT analysis was performed to assess the volume of the repair tissue (Suppl.Fig. 4a, b), the volume of mineralized tissue consisting of newly formed bone and remaining implant (Suppl.Fig. 4c, d), and the BV/TV of the repair tissue (Suppl.Fig. 4e, f).The average volume of repair tissue remained constant at both time points when animals received unloaded implants or cylinders loaded with 2.5 μg L51P, 0.25 μg BMP2, and 0.25 μg BMP2/2.5 μg L51P.In contrast, exposure to 2.5 μg BMP2 increased the average volume of the repair tissue by approx.60 % in OVX/ALN animals compared to all other coatings at both time points (p < 0.05) (Suppl.Fig. 4a, b).Six weeks post-implantation, 2.5 μg BMP2 led on average to 57.9 ± 34.0 % more mineralized tissue in OVX/ALN animals when compared to 0.25 μg BMP2 (p < 0.001) (Suppl.Fig. 4c).Twelve weeks post-implantation, 2.5 μg BMP2 resulted in higher amounts of mineralized tissue in all groups, particularly in animals treated with BP (Suppl.Fig. 4d).Six weeks post-implantation, coating with 2.5 μg BMP2 led to approx.30 % lower BV/TV of the defect tissue in OVX/Veh animals in comparison to 0.25 μg BMP2 and 0.25 μg BMP2/ 2.5 μg L51P (p = 0.02) (Suppl.Fig. 4e).Twelve weeks post-implantation, BP therapy increased the BV/TV of the defect tissue independent of the implant coating (Suppl.Fig. 4f).The data demonstrates that 2.5 μg BMP2 stimulated the formation of mineralized tissue, and the amount of repair tissue partially correlated with a higher BV/TV, particularly in combination with BP.
RNA sequencing
To evaluate the molecular mechanisms regulating bone repair, RNA Ground sections were prepared from the methyl methacrylate-embedded repair tissues isolated twelve weeks after the application of the critical-size defect in sham and OVX animals with Veh and ALN therapy.Bone defects were fitted with unloaded (0 μg BMP2/0 μg L51P) βTCP-ceramics (a-d), or with implants coated with 2.5 μg L51P (e-h), 0.25 μg BMP2 (i-l), 0.25 μg BMP2/2.5 μg L51P (m-p), or 2.5 μg BMP2 (q-t).Sections were stained with a mixture of toluidine blue and MacNeal's tetrachrome solution.Unloaded implants (a-d) and βTCPceramics coated with 2.5 μg L51P (e-h) resulted in minimal bone formation in all experimental groups.The coating with 0.25 μg BMP2 induced bone healing in sham animals (i, j), whereas OVX groups (k, l) showed minor bone formation.
The combination of 0.25 μg BMP2/2.5 μg L51P caused considerable bone formation coupled with implant removal in sham groups (m, n) and induced minimal bone growth in OVX animals (o, p).Implants loaded with 2.5 μg BMP2 led to the strongest induction of bone healing in all groups (q-t).The femur is oriented with the distal end on the left side and the proximal part on the right.The scale bar represents 1 mm.
Fig. 4.
Histomorphometry of newly formed bone and remaining implant material.Histomorphometric analysis of the developing repair tissue was performed by manual selection on one histological section per animal.Assessment of the total bone area (a, b), the remaining implant material (c, d), and mixed tissue comprised of implant material mixed with ingrown bone (e, f) in all groups six and twelve weeks after application of the defect, respectively.Statistical differences were determined within the treatment groups regarding the effect of different implant coatings using a two-way ANOVA with Tukey post-hoc testing.Differences were termed statistically significant with values of p ≤ 0.05 (*), p ≤ 0.01 (**).Data presented as mean ± standard deviation with 2-4 biological replicates.
F. Strunz et al. was isolated from tissues within the defect site at six and twelve weeks after the introduction of the critical-size defect, and a global RNA sequencing analysis was performed.On average, 32.7 million reads/library with an overlap of approx.86 % (minimum 82 %, maximum 89 %) with the annotated genome (GRCm39, ENSEMBL) were obtained.
Clustering, differentially expressed genes and GO enrichment analysis
The first two axes of the principal component analysis, based on the 500 genes with the most variable expression over the whole transcriptome, did not reveal a distinct clustering of the samples in dependence of time point after the application of the critical-size defect, BP therapy, or estrogen-deficiency.In addition, no clustering was observed, when comparing impact of the bioavailability of BMP2 and L51P on the transcriptome at six and twelve weeks post-surgery (Fig. 5).
To further explore the gene expression profile of the developing repair tissues, the numbers of differentially expressed genes (DEG) and the overrepresentation of DEG in the gene ontology (GO) enrichment analysis were assessed.This allows the evaluation of potential changes in the transcriptome linked to bone regeneration with respect to estrogen-deficiency, BP therapy, the effects of BMP2 and L51P, and the time point post-implantation.
Exposure to 2.5 μg BMP2 led to a similar transcriptome profile six
and twelve weeks post-implantation.Tissue samples harvested six or twelve weeks after MouseFix™ surgery were grouped together regarding the coatings of the respective implants.Focusing on the differences in the temporal evolution of the bone healing process in the presence of BMP2 and L51P between the time points six and twelve weeks post-surgery, the parameters estrogen-depletion and BP therapy were not considered in the analysis (Suppl.Table 1).
The differential gene expression analysis revealed minor changes on the level of the transcriptome in samples with implants coated with 2.5 μg BMP2 with respect to the time after surgery (98 DEG).In contrast, in animals treated with 0.25 μg BMP2/2.5 μg L51P implants, 934 DEG were detected with respect to the time after surgery.Twelve weeks postsurgery, treatment with 0.25 μg BMP2/2.5 μg L51P resulted in the upregulation of biological processes linked to muscle regeneration (GO:0003012 "muscle system process", GO:0060537 "muscle tissue development", GO:0014706 "striated muscle tissue development").In addition, in animals treated with 0.25 μg BMP2 implants, 157 DEG were detected with respect to the time after surgery.Twelve weeks postsurgery, treatment with 0.25 μg BMP2 resulted in the downregulation of biological processes linked to extracellular matrix organization (GO:0030198 "extracellular matrix organization") and bone repair (GO:0001503 "ossification", GO:0031214 "biomineral tissue development", GO:0110148 "biomineralization", GO:0061448 "connective tissue development", GO:0051216 "cartilage development").Six weeks post-surgery unloaded and 2.5 μg L51P implants were not included in the experimental design.Taken together, the comparison of the transcriptome profiles of repair tissues, demonstrates that 2.5 μg BMP2 accelerated the repair process.
Exposure to 2.5 μg BMP2 accelerated the bone repair process.
Tissue samples harvested six or twelve weeks after MouseFix™ surgery were grouped together regarding the coatings of the respective implants.The parameters estrogen-depletion and BP therapy were not considered in the analysis to detect the impact of the implant coating on the transcriptomic level within the time points post-surgery (Table 1).Six weeks post-surgery, exposure to 2.5 μg BMP2 led to changes in the differential gene expression profile as compared to 0.25 μg BMP2 (145 DEG).The GO enrichment analysis revealed the downregulation of GO terms associated with bone repair (GO:0061448 "connective tissue development", GO:0048705 "skeletal system morphogenesis", GO:0060348 "bone development"), cartilage development (GO:0051216 "cartilage development", GO:0002062 "chondrocyte differentiation"), and extracellular matrix organization (GO:0030198) in animals treated with 2.5 μg BMP2.The comparison between 2.5 μg BMP2 and 0.25 μg BMP2/ 2.5 μg L51P revealed 3 DEG.Twelve weeks post-surgery, exposure to 2.5 μg BMP2 led to changes in the differential gene expression profile as compared to unloaded controls (1242 DEG) and implants coated with 2.5 μg L51P (1341 DEG), 0.25 μg BMP2 (175 DEG) or 0.25 μg BMP2/2.5 μg L51P (381 DEG).The GO enrichment analysis revealed the downregulation of GO terms associated with extracellular matrix organization (GO:0030198 "extracellular matrix organization") and bone healing (GO:0061448 "connective tissue development", GO:0051216 "cartilage development") in animals treated with 2.5 μg BMP2.No changes on the transcriptome level were detected in the defect tissue between animals that received 0.25 μg BMP2/2.5 μg L51P, 0.25 μg BMP2, 2.5 μg L51P, or unloaded implants.The transcriptome analysis between defect tissues exposed to 2.5 μg BMP2 and all other coatings, further confirmed the conclusion that 2.5 μg BMP2 accelerated the healing process.
Estrogen-deficiency induced inflammatory responses within the repair tissue twelve weeks post-surgery.
To detect the general effects of OVX on the transcriptome profile of the repair tissue six and twelve weeks post-surgery, samples from OVX animals were compared to samples from sham animals.Implant coating and BP therapy were not considered in the analysis.Six weeks after the defect, estrogendeficiency did not cause changes in the differential gene expression analysis.Twelve weeks post-surgery, 378 DEG were detected between the samples from OVX and sham animals.OVX samples showed upregulation of GO terms linked to B cell mediated immune responses (GO:0019724 "B cell mediated immunity", GO:0002455 "humoral immune response mediated by circulating immunoglobulin", GO:0050853 "B cell receptor signaling pathway").Twelve weeks post-surgery, estrogen-depletion induced inflammatory responses in the developing repair tissue.This resulted in significant changes in the function and regulation of B-and T-cells twelve weeks post-surgery in OVX animals.
BP therapy altered osteoclast differentiation within the repair tissues six weeks post-surgery.
To detect the general effects of BP treatment on the transcriptome profile of the developing repair tissues at time points six and twelve weeks, samples of ALN treated animals were compared to samples from animals receiving Veh only.The implant coating and estrogen-depletion were not considered in the analysis.Six weeks after the application of the defect, 151 DEG were detected within the defect tissues between the samples from Veh and ALN treated animals.The tissues harvested from ALN treated animals showed upregulation of the GO term "multinuclear osteoclast differentiation" (GO:0072674).Twelve weeks after application of the defect, the comparison of the repair tissue revealed only 95 DEG between Veh and ALN treated animals.Six weeks post-surgery, BP therapy altered the function and regulation of osteoclasts in the developing repair tissue.
BP therapy led to significant upregulation of selected osteoclast genes.
To detect the general effects of BP treatment on the expression of genes essential for osteoclast development and function at time points six and twelve weeks, samples of ALN treated animals were compared to samples from animals receiving Veh only (Table 2).The implant coating and estrogen-depletion were not considered in the analysis.ALN therapy resulted in significantly higher levels of transcripts encoding selected markers of osteoclast lineage cells compared to control animals receiving Veh only.Transcripts encoding osteoclast traits like Oscar, Ocstamp, Dcstamp, Slc9b2, Mmp9, Ctsk, Fos, and Acp5 showed upregulation upon ALN therapy at both time points post-surgery.To visualize the effect of BP therapy on the differential gene expression of the selected osteoclast marker genes, heatmaps were rendered for both time points (Suppl.Fig. 5).
3.3.2.6.Increased expression levels of osteoclast related genes in the developing repair tissue six weeks post-surgery.To further investigate the temporal evolution of genes essential for osteoclast development and function, samples collected six and twelve weeks after implantation of βTCP-implants were compared either under ALN therapy or Veh treatment (Table 3).
The implant coating and estrogen-depletion were not considered in the analysis.Levels of transcript of selected osteoclast marker genes (Oscar, Ocstamp, Acp5, Calcr, Slc9b2, Dcstamp) were higher at six weeks as compared to twelve weeks in both Veh and ALN treated animals.The relevance of the respective time point was further emphasized in heatmaps displaying the relative gene expression of the selected osteoclast markers (Suppl.Fig. 6).
Increased expression levels of osteoblast related genes in the developing repair tissue six weeks post-surgery.
To investigate the temporal evolution of the expression of genes essential for osteoblast development and function, samples from animals treated either with ALN or Veh were collected six and twelve weeks after implantation of βTCPimplants and compared (Table 4).Implant coating and estrogendepletion were not considered in the analysis.The transcript levels of selected osteoblast marker genes (Col1a1, Cola1a2, Sp7, Bglap, Sparc, Ibsp, Alpl, Tnfsf11, Runx2, Bgn) were higher at six weeks as compared to twelve weeks in both Veh and ALN treated animals.Heatmaps of the relative gene expression of the selected osteoblast marker genes emphasize the relevance of the respective time point (Suppl.Fig. 7).
Table 1
Effect of 2.5 μg BMP2 on the bone healing process at the transcriptome level.
Differential gene expression analysis, of the developing repair tissue in animals treated with 2.5 μg BMP2 compared to all other coatings six and twelve weeks post-surgery (n = 12).
Discussion
Osteoporosis is a major health issue for aging societies and affects hundreds of millions of people worldwide, predominantly postmenopausal women (Willers et al., 2022).Anti-resorptive drugs, in particular BP, are the therapy of choice to prevent bone loss in osteoporosis (Liu et al., 2019;Bone et al., 1997;Cranney et al., 2002;Hosking et al., 1998;Liberman et al., 1995).The consequences of long-term therapies with BP are controversially discussed as the "frozen" bone state and blocked osteoclast activity may impair bone healing and removal of bone grafts (Diab and Watts, 2012;McClung et al., 2013;Watts and Diab, 2010;Odvina et al., 2005;Molvik and Khan, 2015).The present study aimed to investigate the repair process of a critical-size femoral defect, fitted with βTCP-implants, in a murine model for postmenopausal osteoporosis treated with BP with respect to (i) bone repair, (ii) modulation of cellular and chemical resorption of βTCP-implants, and (iii) the effects of BMP2 and L51P on the healing process.
In the present study, estrogen-depletion induced by OVX was utilized to generate a murine model for post-menopausal osteoporosis in outbred NMRI mice.The increase in body weight and significant shrinkage of the uterus in OVX animals provided evidence of estrogen deprivation.Even though cancellous bone is more sensitive to bone loss induced by estrogen deficiency than is cortical bone, the changes in bone metabolism will be the same over the whole skeleton.MicroCT analysis confirmed the ovariectomy-induced osteopenic bone changes, and BP treatment, starting eight weeks after OVX, increased bone mass.Elevated bone density was also detected upon BP therapy in sham controls.BP bind to the bone surface, preferentially at bone sites with increased turnover, affecting bone modeling and remodeling systemically (Sato et al., 1991).
The critical-size defect was applied mid-femur to allow the stable fixation of the defect by the osteosynthesis system.Histological analysis of the critical-size defect six weeks post-implantation demonstrated that 2.5 μg BMP2 accelerated fracture repair and induced bone formation compared to controls (unloaded, 2.5 μg L51P, 0.25 μg BMP2).Binding of 2.5 μg BMP2 to the implant led to new bone formation around the βTCPimplant from the proximal to the distal defect end, irrespective of estrogen-deprivation or BP therapy.Exposure to 0.25 μg BMP2/2.5 μg L51P, however, did not elevate bone formation or promote fracture healing, indicating that L51P did not increase the bioefficacy of BMP2 in the applied defect model.
Efficient bone repair requires specific action of growth factors with precise temporal and spatial distribution (Dumic-Cule et al., 2018;Czech and Oyewumi, 2021).Thus, the bioefficacy of the applied BMP2 is affected by multiple factors beyond the presence of BMP antagonists.Approx.60 % of the applied BMP2 and L51P are released from the βTCPceramics within the first 24 h (Sebald et al., 2012).In cases of large defects, this rapid release may not be sufficient to stimulate bone regeneration (El Bialy et al., 2017).As a result, supraphysiological BMP2 doses only marginally support healing in later repair phases.Additionally, simultaneous administration of L51P and BMP2 may not ideally contribute to the healing process, as the upregulation of BMP antagonists occurs with a chronological delay (Montjovent et al., 2013).The effect of L51P may also depend on the species and age of the applied rodent model, as combined administration of BMP2 and L51P significantly improved fracture healing in retired breeder rats (Hauser et al., 2018b;Sebald et al., 2012).
Both histological and MicroCT evaluation confirmed that new bone formation within the critical-size defect was directly correlated to the removal of the βTCP-implant.This coupling resulted in a comparable volume of total repair tissue among all experimental groups.BP therapy partially enhanced bone healing, leading to elevated levels of repair tissue and higher amounts of mineralized tissue, particularly in animals treated with 2.5 μg BMP2 implants.Overall, in the applied fracture model, the administration of BP did not affect the kinetics of the bone repair process.This observation contrasts with the anticipated outcome, since it was hypothesized that bone formation, through coupling, would be decreased in animals under BP therapy, consequently leading to a delayed healing process compared to animals not subjected to BP treatment.Twelve weeks post-surgery, the implant had not been fully removed, and the critical-size defect had not completely healed, lacking a continuous cortical bone structure or marrow cavity in all experimental conditions.Therefore, distinct differences due to ALN treatment might become detectable later in the healing process only.
The effect of BP on fracture healing remains controversial, with clinical studies showing minor or no influence on bone repair in humans (Begkas et al., 2019;Kates and Ackert-Bicknell, 2016;Gao et al., 2021;Xue et al., 2014).The outcome in rodent models varied from delayed healing to improved mechanical stability of bone defects, depending on the agent, dosage, and treatment protocol (Hadjiargyrou, 2022).In rat fracture models, BP caused delayed unions and reduced healing: in particular, ALN impaired implant removal and increased the callus size (Li et al., 2001;Li et al., 1999;Kidd et al., 2011;Manabe et al., 2012;Cao et al., 2002;Fu et al., 2013;Hauser et al., 2018b).Species-dependent differences and characteristics of the applied model, including the age of the animals and the type of fracture, might explain the discrepancy in
Table 3
Temporal evolution of the expression of selected osteoclast genes in the defect site under Veh or ALN treatment.Temporal evolution (6 weeks vs. 12 weeks after implantation of βTCP-ceramics) of the expression of transcripts encoding markers of the osteoclast trait in critical-size femoral defects under ALN therapy or Veh treatment.Analysis performed by DESeq2 (n = 18-30).the impact of BP on bone regeneration (Hadjiargyrou, 2022).Independently of the cellular resorption, the degradation of the βTCP implant might be partially attributable to a chemical dissolution process due to a decrease of the pH levels within the healing defect.This process can be initiated by a mechanism called intrinsic osteoinduction caused by the implantation of the βTCP-material itself (Bohner et al., 2020;Bohner and Miron, 2019).The implanted graft acts as a nucleation site for the precipitation of carbonated apatite resulting locally in lower concentrations of phosphate and calcium ions.This biomimetic apatite layer stimulates the differentiation of osteoblast precursor and stem cells initiating bone formation linked to the gradual breakdown of the implant material (Bohner et al., 2020;Bohner and Miron, 2019).
In the present study, transcriptome analysis of the repair tissue showed no data clustering due to the extensive variables in the experimental setup.The severe intervention of creating the femoral defect activated intrinsic healing processes, outweighing the effects of estrogen-deficiency and BP therapy, as seen in previous studies (Hauser et al., 2018a).In that study, the mode of fixation was found to have the strongest impact on the healing process.Rigid compared to non-rigid fixation of femoral osteotomies initiated different developmental programs, endochondral healing compared to direct bone formation, respectively.However, other parameters did not significantly differentiate the experimental groups (Hauser et al., 2018a).
Changes on the level of transcription related to implant coating were detectable when implants were coated with 2.5 μg BMP2.Differential gene expression analysis indicated accelerated repair, and GO terms associated with bone healing and tissue regeneration were downregulated twelve weeks post-surgery.This finding aligns with the histological analysis, showing significant bone formation already six weeks after defect treatment with 2.5 μg BMP2.The temporal evolution of selected osteoclast and osteoblast genes showed higher expression levels at six weeks compared to twelve weeks.This indicates a more dynamic repair process at six weeks, with prominent cellular responses from both cell lineages.In addition, transcriptome analysis suggested that bone regeneration was almost complete, or the healing capacity was already exhausted six weeks after surgery.The strongest causative factor affecting bone regeneration and the composition of the transcriptome was implant coating, specifically the presence of 2.5 μg BMP2, followed by BP therapy and estrogen-depletion.
The efficacy of the BP therapy was also confirmed on the transcriptome level.Transcripts encoding osteoclast lineage markers at both six and twelve weeks post-surgery showed increased expression levels in the presence of BP, regardless of estrogen-depletion and implant coating.As BP therapy causes osteoclast inactivity (Rodan and Fleisch, 1996;Russell, 2006), the lack of a feedback mechanism that is required to control osteoclastogenesis leads to the continuous recruitment of osteoclasts to the repair site (Weinstein et al., 2009).An increase in the recruitment of osteoclast precursors and enhanced fusion of progenitors forming giant osteoclasts may also contribute to the high expression levels of transcripts encoding the osteoclast marker genes (Hauser et al., 2018a).
Nevertheless, bone formation within the defect stayed linked to implant resorption, further indicating the chemical dissolution of the βTCP-graft due to pH changes in areas with active bone healing.The metabolic activity of osteoblasts might cause the acidification of the microenvironment, facilitating the dissolution of the implanted graft.Moreover, recruited osteoclasts might release factors like Sphingosine-1phosphate stimulating osteoblast activity (Kim et al., 2020), indirectly fostering the osteogenic environment to promote the healing process, and coupled implant removal.
The present study investigated two key aspects in reconstructive orthopedic surgery.Firstly, it demonstrated that removal of βTCP-ceramics from a repair site can also occur under BP treatment.The osteogenic environment within a defect undergoing repair might initiate the dissolution of the calcium phosphate-based material outweighing the action of the BP therapy.This observation corroborates the results in humans, where the influence of long-term BP treatment on bone repair was minor or absent.Secondly, application of supraphysiological doses of BMP2 remained superior in promoting bone formation.In the used fracture model, L51P did not enhance the bioefficacy of BMP2 when administered simultaneously on βTCP-grafts.As various factors influence the bone repair process, it is not excluded that an adjustment of the applied L51P/BMP2 doses might result in the desired osteoinduction.Thus, further studies would be necessary to investigate the efficacy of L51P in vivo and to discover the optimal dosage and timing of L51P application to enhance bone regeneration.There are, however, also limitations of the utilized fracture model.The developing repair tissue was only examined six and twelve weeks post-surgery and no biomechanical testing of the partially healed defects could be performed.Despite these limitations, however, the data does not provide evidence that BP therapy does impair the healing process of large bone defects treated with βTCP-implants.
Declaration of competing interest
None.
Fig. 3 .
Fig. 3. Histological analysis of the femoral repair site twelve weeks post-implantation.Ground sections were prepared from the methyl methacrylate-embedded repair tissues isolated twelve weeks after the application of the critical-size defect in sham and OVX animals with Veh and ALN therapy.Bone defects were fitted
Fig. 5 .
Fig. 5. Implant coatings did not result in distinct data clustering.Principal component analysis based on the 500 genes with the most variable expression did not reveal a distinct clustering of the samples taken six (a) weeks or twelve (b) weeks after the application of the critical-size defect in dependence of the bioavailability of BMP2 and L51P.
Table 2
Differential expression of selected osteoclast genes in the defect site six and twelve weeks post-implantation in ALN-treated compared to Veh-treated animals.Effect of ALN therapy on the expression of transcripts encoding markers of the osteoclast trait during bone healing in critical-size femoral defects at six or twelve weeks after implantation of βTCP-implants.Differential gene analysis was performed by DESeq2 (n = 18 for 6 weeks, n = 30 for 12 weeks).
Table 4
Temporal evolution of the expression of selected osteoblast genes in the defect site under Veh or ALN treatment.Temporal evolution (6 weeks vs. 12 weeks after implantation of βTCP-ceramics) of the expression of transcripts encoding markers of the osteoblast trait in critical-size femoral defects under ALN or Veh treatment.Analysis performed by DESeq2 (n = 18-30). | 2024-01-24T17:36:12.224Z | 2024-01-01T00:00:00.000 | {
"year": 2024,
"sha1": "444a19aa2b7f2f290e69fcc7ae86a40914e5eb3e",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1016/j.bonr.2024.101739",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "2eacea21b88e34daed84a49d19402f806a91c425",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
236934344 | pes2o/s2orc | v3-fos-license | Prospective association of serum adipocyte fatty acid‐binding protein with heart failure hospitalization in diabetes
Abstract Aims Adipocyte fatty acid‐binding protein (AFABP) is associated with cardiovascular diseases in type 2 diabetes. Whether circulating AFABP levels are associated with the risk of heart failure (HF) in type 2 diabetes remains undefined. We investigated the prospective association of circulating AFABP levels with incident HF hospitalization in type 2 diabetes, and its relationship to the use of sodium glucose co‐transporter 2 inhibitors (SGLT2i) which reduce HF risk. Methods and results Baseline serum AFABP level was measured in 3322 Chinese participants without known history of cardiovascular diseases or hospitalization for HF, recruited from the Hong Kong West Diabetes Registry. Its association with incident HF hospitalization was evaluated using multivariable Cox regression analysis. Use of SGLT2i was included as a time‐dependent covariate. Among these 3322 participants (52.9% men; mean age 60.0 ± 12.6), 176 (5.3%) developed HF hospitalization over a median follow‐up of 8 years. Seven hundred and thirty‐one (22%) were started on SGLT2i during the study period (empagliflozin 55.1%, dapagliflozin 44.2%, canagliflozin 0.4%, and ertugliflozin 0.3%). Serum AFABP levels were significantly higher in participants who developed HF hospitalization than those who did not (men: 14.8 vs. 8.3 ng/mL; women: 21.5 vs. 14.6 ng/mL; all: 18.6 vs. 10.9 ng/mL, P < 0.001). In multivariable Cox regression analysis, baseline serum AFABP level was significantly associated with incident HF hospitalization [hazard ratio (HR) 1.38, 95% confidence interval (CI) 1.06–1.80, P = 0.019] independent of the use of SGLT2i, in a model also consisting of age; sex; body mass index; smoking status; duration of diabetes; hypertension, dyslipidaemia; atrial fibrillation; presence of chronic kidney disease and albuminuria; glycated haemoglobin and high‐sensitivity C‐reactive protein levels; and use of metformin, insulin, aspirin, furosemide, and beta‐blockers at baseline. High cumulative defined daily dose (cDDD) of SGLT2i was protective of incident HF hospitalization (HR 0.10, 95% CI 0.01–0.68, P = 0.019). The addition of circulating AFABP level to a clinical model of conventional HF risk factors provided significant improvement in the category‐free net reclassification index (11.5%, 95% CI 1.6–22.1, P = 0.02) and integrated discrimination improvement (0.3%, 95% CI 0.1–1.7, P = 0.04). A dose‐dependent reduction in cumulative incidence of HF hospitalization in response to SGLT2i, based on cDDD, was more clearly observed in participants with a higher baseline AFABP level above the sex‐specific median (P for trend <0.01). Conclusions Circulating AFABP level is independently associated with incident HF hospitalization in type 2 diabetes and is potentially helpful in risk stratification for the prevention of HF hospitalization.
Introduction
Type 2 diabetes is associated with an increased risk of cardiovascular diseases (CVD), including heart failure (HF). Although reductions in cardiovascular deaths due to ischaemic heart disease and stroke have been observed in recent decades, especially among older individuals with type 2 diabetes, 1 mortality related to HF did not change significantly and remained high. In epidemiological studies, type 2 diabetes almost doubled the risk of incident HF hospitalization, regardless of the presence of coronary artery atherosclerosis. 2 Moreover, type 2 diabetes was a significant predictor of adverse outcomes in clinical trials of HF. 3 Both hyperglycaemia and insulin resistance could lead to left ventricular hypertrophy, myocardial fibrosis, and diastolic dysfunction, through various mechanisms including lipotoxicity, inflammation, formation of advanced glycation end products and endoplasmic reticulum stress. 2 The introduction of sodium glucose co-transporter 2 inhibitors (SGLT2i) has led to a paradigm shift in the management of type 2 diabetes and HF. 4 Canagliflozin, dapagliflozin, and empagliflozin had been consistently shown to reduce the rates of HF hospitalization in several landmark randomized controlled trials. [5][6][7][8] The latter two, in particular, were demonstrated to lower the risks of worsening HF and cardiovascular deaths in patients with reduced left ventricular ejection fraction (LVEF), even in patients without diabetes. 9,10 Sotagliflozin, which is a dual inhibitor of SGLT2 and SGLT1, was recently shown to reduce cardiovascular deaths in patients with type 2 diabetes and decompensated HF, 11 as well as HF hospitalization in patients with type 2 diabetes and chronic kidney disease, with and without albuminuria. 12 Adipocyte fatty acid-binding protein (AFABP) is a lipid chaperone protein that is abundantly secreted by the adipocytes and is also highly expressed in the macrophages and endothelial cells. 13,14 Pre-clinical studies have revealed that AFABP is pro-inflammatory and promotes vascular dysfunction. Our group previously demonstrated that AFABP could perpetuate lipopolysaccharide-induced inflammatory responses in macrophages through its interaction with c-Jun NH2-terminal kinase (JNK) and activator protein-1 forming a positive feedback loop. 15 In mice, the anti-atherogenic effects of adipose-specific JNK inactivation were abolished by the continuous infusion of recombinant AFABP. 16 Moreover, genetic ablation of AFABP protected mice from cardiac dysfunction secondary to diabetes and myocardial ischemia/ reperfusion injury. AFABP, on the other hand, reduced endothelial nitric oxide synthase phosphorylation and increased superoxide anions formation. Consequently, endothelial dysfunction ensued which could further induce oxidative stress, cardiac inflammation, hypertrophy, and fibrosis which impaired myocardial contractility. 17 Circulating AFABP level has been shown to have prognostic importance in cardio-renal events and mortality outcomes in patients with and without type 2 diabetes. 18,19 However, with regard to HF, most studies that demonstrated a positive relationship between circulating AFABP levels and ventricular dysfunction were of cross-sectional design. 20,21 Moreover, although circulating AFABP level had been shown in the Cardiovascular Health Study as a modest but independent predictor of incident HF, the study was conducted among older community-dwelling individuals aged 65 years or above, of whom less than 20% had diabetes and was performed during the period when SGLT2i was still not available. 22 Therefore, we conducted this prospective study to investigate the association of circulating AFABP level with HF hospitalization and its relationship to the use of SGLT2i, using an exclusively diabetic population with increased HF risk.
Study participants
All participants were recruited from the Hong Kong West Diabetes Registry (HKWDR), which comprised patients who had type 2 diabetes and were being followed-up regularly at the medical specialist clinics of the Hong Kong West Cluster since 2008. All Chinese patients were invited during enrolment to the registry to participate in a prospective cohort study that aimed to identify the risk factors, including genetic and serum biomarkers, of diabetic complications as described previously. 19 The study protocol was approved by the institutional review board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster. Written informed consent was obtained from all recruited participants prior to any study-related procedures.
In the current study that evaluated the role of circulating AFABP levels and incident HF hospitalization, participants who had history of CVD, HF hospitalization, or structural heart diseases at baseline were excluded. Moreover, participants who were non-Chinese, had end-stage renal disease, on renal replacement therapy or had received a kidney transplant, or had follow-up for less than 1 year were also excluded.
Clinical and biochemical assessments
All participants had regular assessment for the development of chronic diabetic complications every 12 to 18 months, in addition to their usual follow-up for diabetes every 4 to 6 months according to a management protocol. This involved both clinical and laboratory investigations to determine their control of diabetes, its related cardiovascular risk factors, and the presence of chronic diabetic complications.
The participants attended each assessment after an overnight fast of at least 8 h. During the baseline assessment, demographic data, including age, sex, smoking, and drinking status were obtained. Detailed medical, family, and drug histories were assessed using a standardized questionnaire. Anthropometric parameters, including body weight, height, body mass index (BMI), waist circumference, and blood pressure were measured. Fasting blood was drawn for plasma glucose, lipids, and glycated haemoglobin (HbA1c) levels and stored in aliquots at À70°C for assays of biomarkers of diabetic complications. Serum creatinine level was measured, and estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation. Albuminuria status was assessed using at least two random urine samples on two separate occasions within 6 months, and was categorized accordingly (urine albumin to creatinine ratio < 3 mg/mmol [A1], 3-30 mg/mmol [A2], and >30 mg/mmol [A3]). Serum high-sensitivity C-reactive protein (hsCRP) level was measured with a high-sensitivity, particle-enhanced immune-turbidimetric assay (Roche Diagnostics, GmbH, Mannheim, Germany).
Serum AFABP was measured with a monoclonal antibodybased enzyme-linked immunosorbent assay (Antibody and Immunoassay Services, University of Hong Kong) that had been used in other regional and international studies. 23 The intra-assay and inter-assay precision coefficients of variability of the AFABP enzyme-linked immunosorbent assay were <4.1% and <4.5%, respectively, and the lowest detection limit was 0.39 ng/mL.
Definitions of outcomes and clinical variables
All outcomes were recorded and verified from the Hong Kong Hospital Authority database or their private practitioners as of 30 June 2020. HF hospitalization, the primary outcome of interest in this study, was defined as the first recorded hospitalization with HF as the principle diagnosis coded by a physician based on the Ninth edition of the International Classification of Diseases (ICD). The diagnosis were adjudicated and reviewed by two physicians independently, taking into account a constellation of symptoms (shortness of breath, orthopnoea, and paroxysmal nocturnal dyspnoea), signs (ankle oedema and basal chest crepitations), chest X-ray or transthoracic echocardiography (ECHO) findings (pulmonary congestion and suboptimal LVEF), and treatment of HF (use of intravenous or increased dose of oral diuretics). Disagreements between the two adjudicating physicians were resolved by a third.
Hypertension was defined as blood pressure ≥ 140/ 90 mmHg or the use of anti-hypertensive medications. Dyslipidaemia was defined as fasting triglyceride (TG) ≥ 1.69 mmol/L, high-density lipoprotein cholesterol < 1.04 mmol/L in men and <1.29 mmol/L in women, low-density lipoprotein cholesterol ≥ 2.6 mmol/L or the use of lipid-lowering agents. For the exclusion criteria of the study, known CVD at baseline was defined as any history of myocardial infarction, stroke, transient ischemic attack, HF, and coronary arterial revascularization based on the ICD-9 codes 410-412, 414, 428-429, and 430-438. Moreover, known structural heart disease at baseline was defined as any history of chronic rheumatic heart disease, cardiomyopathy and congenital heart disease based on the ICD-9 codes 393-398, 425, and 745-726.
Sodium glucose co-transporter 2 inhibitors exposure
Since our prospective cohort commenced in 2008 and SGLT2i only became available locally after 2015, use of SGLT2i was treated as a time-dependent covariate in the analysis. In all study participants, prescriptions of SGLT2i including canagliflozin, dapagliflozin, empagliflozin, and ertugliflozin were identified. SGLT2i usage was defined by the continuous prescription for at least 90 days during the study period from the initiation of medication to the development of outcome or end of observation, whichever earlier, and classified based on their cumulative daily defined dose (cDDD) as described by the World Health Organization. 24 cDDD reflects the average daily maintenance dose of a drug prescribed for the main indication in an adult. In this study, cDDD was calculated as the sum of dispensed defined daily doses of all prescribed SGLT2i during the study period. One defined daily doses of SGLT2i is equivalent to daily use of canagliflozin 200 mg, dapagliflozin 10 mg, empagliflozin 17.5 mg, or ertugliflozin 10 mg.
Statistical analysis
All data were analysed with IBM SPSS Statistics 26.0 and R Version 3.2.3 (http://www.r-project.org). Data that were not normally distributed as determined using the Kolmogorov-Smirnov test, such as serum AFABP, TG, and hsCRP levels, were log-transformed before analysis. The values were reported as means ± standard deviation, medians with interquartile range if skewed data, or percentages. Comparisons between groups were analysed by χ 2 test for categorical variables, and independent t-test or analysis of variance for continuous variables as appropriate. Multivariable Cox regression analysis was performed to evaluate the associations between baseline serum AFABP levels and incident HF hospitalization. The Cox regression analysis was repeated with use of SGLT2i included as a time-dependent covariate. The variables included in the Cox regression models were those that were either biologically relevant or statistically significant in the univariate analysis. The hazard ratio (HR) for circulating AFABP level referred to the risk of 3966 C.H. Lee et al. developing HF hospitalization per unit difference in the log-transformed serum AFABP level measured in ng/mL. Proportional hazards assumption was checked and verified using a global goodness-of fit test proposed by Schoenfeld. The predictive performance of the various models was evaluated using C-statistics, category-free net reclassification improvement, and integrated discrimination improvement. Furthermore, an exploratory subgroup analysis was performed to evaluate if a dose-dependent association was present between the cumulative incidence of HF hospitalization and use of SGLT2i in participants with baseline AFABP levels above and below the median. In all statistical tests, two-sided P values < 0.05 were considered significant.
Baseline serum AFABP levels were significantly higher in participants with incident HF hospitalization than those without
A total of 3322 participants were included in this study, after excluding 1920 with history of CVD, HF hospitalization, or structural heart diseases at baseline. Higher quartiles of serum AFABP levels were significantly associated with older age; longer duration of diabetes; higher BMI; hsCRP levels; prevalence of hypertension; dyslipidaemia; atrial fibrillation; eGFR < 60 mL/min/1.73 m 2 ; albuminuria; as well as use of multiple medications including metformin, sulphonylurea, glitazones, insulin, aspirin, statin, fibrate, angiotensinconverting enzyme inhibitor or angiotensin receptor blocker, furosemide, and beta blockers at baseline (all P < 0.001) ( Table 1).
Over a median follow-up of 8 years, 176 (5.3%) of the 3322 participants developed incident HF hospitalization. Among these 176 participants, 127 (72.2%) had ECHO performed following their admission: 20, 11, and 96 participants had LVEF of <40%, 40-49%, and ≥50%, respectively. Table 2 summarizes their baseline characteristics. Participants who had incident HF hospitalization were significantly older, with higher BMI, prevalence of hypertension, atrial fibrillation, and albuminuria at baseline than those without. Moreover, their duration of diabetes was also significantly longer with higher baseline HbA1c, TG and hsCRP, and lower eGFR levels than those without incident HF hospitalization. Furthermore, a significantly higher proportion of participants with HF hospitalization were on insulin, aspirin, angiotensin-converting enzyme inhibitor/angiotensin receptor blocker, beta-blockers, and furosemide at baseline than those who without.
Use of SGLT2i during the study period was associated with reduced risk of incident HF hospitalization
All study participants were not on SGLT2i at baseline. However, SGLT2i was started in 731 participants (22%) during the study period (empagliflozin 55.1%, dapagliflozin 44.2%, canagliflozin 0.4%, and ertugliflozin 0.3%). SGLT2i user had a significantly lower risk of incident HF hospitalization than those who did not (HR 0.24, 95% CI 0.04-0.74, P = 0.014). Among the SGLT2i users, the cDDD of SGLT2i was significantly lower in those who developed incident HF hospitalization compared with those who did not (139 vs. 686, P = 0.002). Moreover, compared with non-users, cDDD of ≥180, which was equivalent to the use of dapagliflozin 10 mg daily for 180 days, was significantly associated with a lower risk of incident HF hospitalization (HR 0.08, 95% CI 0.01-0.55, P = 0.010). (Table 3)
When use of SGLT2i was included in the multivariable Cox regression model as a time-varying covariate, serum AFABP level remained independently associated with a higher risk of incident HF hospitalization (HR 1.38, 95% CI 1.06-1.80, P = 0.019), such that each unit increase in the log-transformed serum AFABP level was associated with a 38% increase in the risk of HF hospitalization. On the other hand, use of SGLT2i with cDDD ≥ 180 was associated with a significantly lower risk of incident HF hospitalization (HR 0.10, 95% CI 0.01-0.68, P = 0.019) ( Table 4).
Moreover, the addition of circulating AFABP level to the clinical model, which consisted of age; sex; BMI; smoking status; duration of diabetes; hypertension; dyslipidaemia; atrial fibrillation; eGFR < 60 mL/min/1.73m 2 ; albuminuria; HbA1c; hsCRP levels; and use of metformin, insulin, aspirin, beta-blockers, and furosemide at baseline, provided significant improvement in the net reclassification improvement
Sensitivity analyses
Several sensitivity analyses were performed to further evaluate the association between serum AFABP level and incident HF hospitalization in type 2 diabetes. Among the 3322 study participants, only 1065 (32.1%) had ECHO performed at baseline. However, we identified 57 participants without history of HF hospitalization but with reduced LVEF < 50% at baseline. In a sensitivity analysis excluding these 57 participants, serum AFABP remained independently associated with incident HF hospitalization (HR 1.38, 1.04-1.82; P = 0.026) after multivariable adjustments. (Supporting information, Table S1) Next, we conducted a sensitivity analysis including all participants with and without history of CVD and HF hospitalization at baseline (N = 5242). Again, baseline serum AFABP level was independently associated with incident HF hospitalization (HR 1.33, 1.11-1.60, P = 0.002), in a model consisting of age, sex, BMI, ever smoking, duration of diabetes, hypertension, dyslipidaemia, atrial fibrillation, eGFR < 60 mL/min/1.73m 2 , presence of albuminuria, HbA1c, use of metformin, insulin, aspirin, furosemide, beta-blockers, spironolactone, hsCRP levels, history of CVD and HF, as well as an interaction term between history of HF and use of furosemide at baseline. Consistently, the association was significant (HR 1.33, 1.10-1.60, P = 0.003) even after further adjustments for the use of SGLT2i ( Table 5). Since SGLT2i was not available locally until 2015, (Table S3). Lastly, the results were also similar in a sensitivity analysis in which patients with atrial fibrillation at baseline were excluded.
Differences in the cumulative incidence of HF hospitalization among the study participants stratified by their baseline serum AFABP levels and cDDD of SGLT2i
In an exploratory analysis to examine the interactions between the use of SGLT2i and incident HF hospitalization in relation to their baseline serum AFABP level, the study 95% CI, 95% confidence interval; ACEI, angiotensin converting enzyme inhibitors; AFABP, adipocyte fatty acid-binding protein; ARB, angiotensin II receptor blockers; BMI, body mass index; BP, blood pressure; DPP4i, dipeptidyl peptidase-4 inhibitor; eGFR, estimated glomerular filtration rate; FG, fasting glucose; HbA1c, glycated haemoglobin; HDL-C, high density lipoprotein-cholesterol; HF, heart failure; HR, hazard ratio; hsCRP, high-sensitivity C-reactive protein; LDL-C low density lipoprotein-cholesterol; TG, triglyceride. Data were presented as mean ± standard deviation or median (25th to 75th percentile). a Log-transformed before analysis. b Sex-adjusted hazard ratio (95% confidence interval) and P value.
AFABP predicts HF independent of use of SGLT2i participants were divided into two groups with their baseline serum AFABP level above and below the sex-specific median.
In each group, the participants were further stratified by their cDDD of SGLT2i. Not surprisingly, participants who had the highest cDDD of SGLT2i had significantly lower cumulative incidence of HF hospitalization compared with those who did not use SGLT2i throughout the study period. This finding was evident in both groups with AFABP above (0.36% vs. 8.44% on no SGLT2i, P < 0.001, respectively) or below the sex-specific median (0% vs. 2.98% on no SGLT2i, P < 0.001, respectively). On the other hand, a dose-dependent reduction in the cumulative incidence of HF hospitalization in response to SGLT2i, based on cDDD, was more clearly seen in participants with a high baseline AFABP level above the sex-specific median (P for trend < 0.01 and <0.05 for AFABP above and below sex-specific median, respectively) ( Figure 1).
Discussions
In this study, we observed that serum AFABP level was independently associated with incident HF hospitalization in an exclusively diabetic population with increased HF risk, regardless of the use of SGLT2i during the study period. Moreover, we also explored the potential of employing baseline serum AFABP level as a tool to prioritize patients with type 2 diabetes who might benefit more from the use of SGLT2i in preventing HF hospitalization.
Compared with most previous studies on AFABP and HF, the prospective, contemporary, and homogenous nature of our study cohort is one of its strengths. Although serum AFABP level had been shown to correlate positively with circulating concentrations of N-terminal fragment of pro-B-type natriuretic peptide, an established marker of HF, 21,25,26 the associations between serum AFABP level and left ventricular remodelling is less straight forward. Left ventricular hypertrophy and diastolic dysfunction contribute significantly to the pathogenesis of HF in diabetes, especially in those without known ischaemic heart disease like our study participants. 2 In several cross-sectional studies, high circulating AFABP level was associated with the presence of left ventricular systolic and/or diastolic dysfunction, 20,27,28 as well as increasing severity of clinical HF. 21 However, there were also studies reporting an inverse relationship. 26,29 These conflicting observations could possibly be a result of the inherent limitations of a cross-sectional study design, or differences in the study populations, or both. It was recently shown that the association between circulating AFABP level and left ventricular mass followed a slightly U-shaped curve, with positive association observed only at high circulating AFABP levels, 26 which were commonly found among individuals with type 2 diabetes. 30 Indeed, a recent prospective ECHO study in patients with type 2 diabetes demonstrated that circulating AFABP level was associated not only with left ventricular remodelling and diastolic dysfunction at baseline but also with the longitudinal increase in left ventricular mass, E/e' ratio, and the development of major adverse cardiovascular events. 31 Therefore, in contrast to our cohort with exclusively individuals with type 2 diabetes, variations in the percentage of patients having type 2 diabetes in the study populations in previous studies could have led to their observed differences in the association between circulating AFABP level and LV remodelling. Moreover, our findings have extended the understanding of the prospective associations between circulating AFABP level and HF hospitalization, reported previously in the Cardiovascular Health Study. 22 We demonstrated that, despite the beneficial effects of SGLT2i in reducing the risks of HF hospitalization, the longitudinal association between baseline serum AFABP level and incident HF hospitalization remained significant and was independent of the use of SGLT2i in patients with type 2 diabetes. In fact, AFABP, in addition to being a lipid chaperone protein, has also been implicated in endoplasmic reticulum stress, JNK activation, lipotoxicity, and systemic inflammation, 16 which are all key pathways in the pathogenesis of diabetic cardiomyopathy. 2 Moreover, in vitro studies had also shown that adipocytederived AFABP could exert a negative inotropic effect on rat cardiomyocytes and inhibit their contractility. 32 Together with the clinical data, these preclinical findings provide a mechanistic link that may account for the association between AFABP and the development of HF hospitalization. On the other hand, in our exploratory analysis, we found that the use of SGLT2i appeared to be more effective in reducing the cumulative incidence of HF hospitalization in those with higher HF risk as reflected by their baseline AFABP level. The conclusions from this explorative analysis remained speculative and likely hypothesis generating. However, SGLT2i, beyond glucose-lowering and diuresis, has also been proposed to possess other beneficial effects to the heart, such as reduction of inflammation and oxidative stress, inhibition of sodium-hydrogen exchange, as well as improvement in myocardial remodelling and bioenergetics, 33 which could have offset the detrimental effects of AFABP, with the benefits being more readily observed in those with higher AFABP levels.
There are several limitations in our study. First, the observational study design did not allow for any causal relationship to be inferred between high circulating AFABP level and the development of incident HF hospitalization. Secondly, due to the registry design and the fact that ECHO is not routinely performed in our clinical practice, ECHO data were not available in a significant proportion of participants both at baseline and at the time of HF hospitalization. This rendered it difficult to perform subgroup analysis on the associations of serum AFABP level with different HF phenotypes. Thirdly, neither brain BNP nor N-terminal fragment of pro-BNP level was measured in our study. Furthermore, although some participants might have been managed as outpatients without hospitalization, it is difficult to be certain of the diagnosis of HF based on outpatient records in which the documentation by clinicians can be highly variable. Therefore, in the present study, we have adopted a harder clinical endpoint of HF hospitalization, which is a commonly used outcome in large-scale randomized control trials. 5-8 Lastly, because serum AFABP level was measured only once in all participants, it was possible that changes in serum AFABP level could have occurred during the study period.
Nonetheless, while we previously demonstrated that circulating AFABP had the potential to become a novel prognostic marker of adverse renal and mortality outcomes in diabetes, 19,34 the present study further suggested that it might also be usefully employed for HF risk stratification in patients with type 2 diabetes. Further research is required not only to validate our findings but also to confirm if serum AFABP level can be employed to prioritize the use of SGLT2i in HF prevention. These are important especially in areas where health care resources are limited, and in this era when SGLT2i is increasingly advocated and prescribed in patients with type 2 diabetes for its cardiorenal benefits. Table S1. Sensitivity analysis showing the association between baseline circulating AFABP levels and incident HF hospitalization (N=3265 after exclusion of study participants with LVEF <50% at baseline). Table S2. Use of SGLT2i and incident HF hospitalization in participants who survived and remained free of outcome events in 2015. Table S3. Sensitivity analysis showing the association between baseline circulating AFABP levels and incident HF hospitalization in participants who survived and remained free of outcome events in 2015 (N=5035). | 2021-08-07T06:18:13.356Z | 2021-08-06T00:00:00.000 | {
"year": 2021,
"sha1": "e15721b31713655873d167df2944a3c6e12ebd77",
"oa_license": "CCBYNCND",
"oa_url": "https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/ehf2.13472",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "5bb8e2cd5c9a5075bde7e1ceb4a8298053420bbf",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
259167721 | pes2o/s2orc | v3-fos-license | Simultaneous LC–MS determination of glucose regulatory peptides secreted by stem cell–derived islet organoids
For studying stem cell–derived islet organoids (SC‐islets) in an organ‐on‐chip (OoC) platform, we have developed a reversed‐phase liquid chromatography–tandem mass spectrometry (RPLC–MS/MS) method allowing for simultaneous determination of insulin, somatostatin‐14, and glucagon, with improved matrix robustness compared to earlier methodology. Combining phenyl/hexyl‐C18 separations using 2.1 mm inner diameter LC columns and triple quadrupole mass spectrometry, identification and quantification were secured with negligible variance in retention time and quantifier/qualifier ratios, negligible levels of carryover (<2%), and sufficient precision (±10% RSD) and accuracy (±15% relative error) with and without use of an internal standard. The obtained lower limits of quantification were 0.2 µg/L for human insulin, 0.1 µg/L for somatostatin‐14, and 0.05 µg/L for glucagon. The here‐developed RPLC–MS/MS method showed that the SC‐islets have an insulin response dependent on glucose concentration, and the SC‐islets produce and release somatostatin‐14 and glucagon. The RPLC–MS/MS method for these peptide hormones was compatible with an unfiltered offline sample collection from SC‐islets cultivated on a pumpless, recirculating OoC (rOoC) platform. The SC‐islets background secretion of insulin was not significantly different on the rOoC device compared to a standard cell culture well‐plate. Taken together, RPLC–MS/MS method is well suited for multi‐hormone measurements of SC‐islets on an OoC platform.
INTRODUCTION
The development of organoids, that is, laboratory-grown 3D organ models, is a rapidly growing field with broad implications on biomedical research [1,2].By combining organoids with microfluidics in an organ-on-chip (OoC) device, it has been possible to improve aspects of organ functionality in vitro compared to static 3D culture systems [3,4].Our research group focuses on applying separation science and mass spectrometry technology for studying metabolism of organoids under static conditions and on OoC devices [5,6].
Pancreatic islets are composed of several endocrine cells, the majority of the cells being insulin-producing beta cells, somatostatin-14-producing delta cells, and glucagon-producing alpha cells [7].The precise regulation of glucose homeostasis is controlled by the hormone secretion from these cells [8].In type 1 and 2 diabetes mellitus, both insulin secretion and glucagon secretion are impaired [9,10].Somatostatin-14 is a paracrine inhibitor of the secretion of insulin and glucagon; however, the role of somatostatin-14 in diabetes is not yet fully understood [11].
Stem cell-derived pancreatic islet organoids (SC-islets) are an emerging alternative for disease modeling of diabetes and cell replacement therapy [12,13].Similar to natural islets, SC-islets differentiated from embryonic stem cells consist of multiple types of endocrine cells [14,15].SC-islets should therefore be suitable for disease modeling and may serve as a source for islet transplantation in betacell replacement therapy for type 1 diabetes patients [13,15].Another important hormone produced in pancreatic islets is urocortin-3, which is a general maturation marker for both alpha and beta cells in humans [16,17].Urocortin-3 would also be beneficial to study in disease modeling, as it is a marker for dedifferentiated beta cells under diabetic conditions (i.e., altered phenotype with loss of insulin capabilities) [17].
The characterization of SC-islets, both under static conditions and on-chip, would benefit from a sensitive and robust methodology for the simultaneous determination of multiple hormones.We have previously developed a sensitive reversed phase liquid chromatography-tandem mass spectrometry (RPLC-MS/MS) method for measuring insulin secreted from SC-islets [18].However, we experienced limitations regarding sample matrix compatibility, that is, the method was compatible with Krebs buffer (a balanced salt solution used to mimic physiological conditions), but not compatible with cell medium commonly applied in cultivation of organoids.We hypothesized that, by improving the LC separation, the method could be improved to also include other hormones and be suitable for several biologically relevant sample matrices.
We here describe an expanded and more versatile RPLC-MS/MS method for simultaneous determination of insulin, somatostatin-14, and glucagon secreted by SCislets and show that the method is compatible with supernatant collected from SC-islets cultivated in a pumpless, recirculating OoC (rOoC) device [4].We show that the SC-islets display an insulin response that is dependent on glucose concentrations.In addition, we were able to detect and quantify the release of somatostatin-14 and glucagon from the SC-islets, confirming that the SC-islets contain functional beta-, delta-, and alpha cells.
Islet maturation cell medium was prepared in-house by adding 1% penicillin/streptavidin, 2% BSA, 10 mg/L of heparin, 10 µM of ZnSO4, 0.1% of trace elements A and B stock solution, 1% of GlutaMAX stock solution, and 1% of MEM NEAA stock solution to basal MCDB131 cell medium.
Preparation of individual hormone solutions and calibration solutions
Aqueous water solutions of human insulin, somatostatin-14, glucagon, urocortin-3, and bovine insulin (applied as internal standard for human insulin) were prepared individually by dissolving 1 mg of peptide powder in 1 mL of 0.1% FA in water.The 1 g/L stock solutions were further diluted to working solutions consisting of 10 mg/L of each individual peptide and divided into 100 µL aliquots.Aliquots of human and bovine insulin were kept at −20 • C until use or for a maximum of 3 months.Aliquots of somatostatin-14, glucagon, and urocortin-3 were kept at −80 • C until use.All solutions containing proteins were prepared in protein low binding tubes from Sarstedt.
Separate standard solutions of 10 mg/L of somatostatin-14, glucagon, and urocortin-3 in a 1 + 1 mixture of ACN and water were prepared for direct injections on the MS.
Solutions with human and bovine insulin in Krebs buffer and islet maturation cell medium were prepared in the same manner as described for water-based solutions, with the exception being the amount of FA: 0.5% FA in Krebs buffer and 1% FA in cell medium.Krebs buffer and cell medium solutions were spiked with separate water-based solutions of 10 mg/L somatostatin-14, glucagon, and urocortin-3 and further diluted with the appropriate matrix to obtain the desired concentrations.
For assessment of the LC-MS method and the preparation of calibration solutions, freshly thawed working solutions of the peptides were further diluted to the desired µg/L concentration with the experiment appropriate matrix and spiked with bovine insulin to a concentration of 5 µg/L unless stated otherwise.Quality controls (QC) were prepared in the same manner.
Cell culture, differentiation, and glucose stimulated insulin secretion for stem cell-derived islets
The SC-islets examined in this study are prepared according to a previously described differentiation protocol [18].In brief, SC-islets were generated from the human pluripotent cell line H1 (WA01, WiCell) with a stepwise differentiation protocol.Following the differentiation, the cells were aggregated as spheroids on an orbit-shaker for 7 days prior to analysis.For static conditions in Krebs buffer, batches of 30 SC-islets were placed in 24-well cell culture plates, and hormone secretion was assessed by exposure to: 1 mL Krebs buffer with 2 mM glucose for 60 min at 37 • C, 1 mL Krebs buffer with 20 mM glucose for 60 min at 37 • C, and 1 mL Krebs buffer containing 20 mM glucose plus 30 mM KCl for 30 min at 37 • C. Up to 900 µL of supernatant was collected.
Insulin in the supernatant was quantified with human insulin enzyme-linked immunosorbent assay (ELISA) kit (Mercodia) and with the LC-MS/MS method described in this study.Prior to injection on the LC-MS/MS system, the collected supernatants were spiked with a total of 5 µg/L of bovine insulin, and 100% FA was added to a total of 0.5%.
Adsorption experiments on a poly(methyl methacrylate)-based organ-on-chip platform
Experiments were performed on the poly(methyl methacrylate) (PMMA)-based rOoC platform, which was fabricated in-house using laser-cutting and thermobonding as previously described in Ref. [4].The rOoC platform consists of two nested circuits of perfusion channels separated by two organoid chambers.Channels are separated from the organoid chambers by a step reservoir with a height of 300 µm.
To evaluate surface adsorption in the rOoC platform, a solution consisting of 2 mg/L of human insulin in Krebs buffer was incubated on-chip and in standard cell culture 24-wells plate from Corning for 20 h and compared to an aliquot stored in the freezer.
2.5
Background secretion experiments from stem cell-derived islets cultivated on-chip and in static system SC-islets were cultured in islet maturation cell medium in an rOoC platform with a modified design compared to Ref. [4], where the chamber for SC-islets was placed under the perfusion channel to facilitate media and oxygen exchange.In the organoids chamber, batches of 3−6 SC-islets were embedded in extracellular matrix (Geltrex, Gibco, cat.n A1569601) and cultured for 7 days under perfusion.The cell medium in the channels was collected after 24 h of incubation on days 5 and 7.In the off-chip control culture, batches of 14−19 SC-islets were cultured in 24-well plates for 7 days, and cell medium was collected after 24 h of incubation on days 5 and 7.
Gel electrophoresis
A water-based solution consisting of 10 mg/L human insulin and 10 mg/L BSA was used as a protein marker during gel electrophoresis.was applied to the gel by a power supply from Delta Elektronika.Subsequently, the gel was washed four times with water for 5 min on a shaker.Following the wash, the gel was covered with Imperial protein stain (from Thermo Fischer Scientific) and left on shaking for 15 min.Before photographing the gel using a smartphone camera, the gel was washed with water four times for 5 min, followed by washing with gentle shaking in water overnight (18 h).
Liquid chromatography-mass spectrometry instrumentation
The LC-MS system applied in this study has been used for insulin determination previously [18].In brief, the conventional LC system was a modified Agilent 1100 series pump employing only shielded fused silica connectors (shielded fused silica nanoViper sheathed in polyetheretherketone tubing from Thermo Fisher).Injection was achieved by coupling a 6-port-2-position valve, with a 50 µm inner diameter (id) × 550 mm (1.08 µL) shielded fused silica loop or a 20 µL shielded fused silica loop.A 250 µL glass syringe was used for injection, following injection the syringe and the loop were washed by flushing the syringe three times with 50/50 MeOH/water, before washing the loop twice with the MeOH/water solution.Prior to the next injection, the syringe was washed once with 0.1% FA in water and the loop was washed twice with 0.1% FA in water.The column setup consisted of an Accucore phenyl/hexyl guard column (2.1 mm id × 10 mm, 2.6 µm, 80 Å) attached within a Uniquard drop-in holder to the InfinityLab Poroshell EC-C18 separation column (2.1 mm id × 50 mm, 2.7 µm, 120 Å).
The applied MS was a TSQ Quantiva triple quadrupole MS equipped with a heated electrospray ionization source (H-ESI-II probe) both from Thermo Fisher.The vaporizer temperature was set to 210 • C, and a spray voltage of 3.5 kV was applied to the H-ESI-II probe.Sheath gas was set at 20 Arb (approx.2.7 L/min), whereas auxiliary gas was set at 9 Arb (approx.7.5 L/min), and sweep gas was not applied.The ion transfer tube temperature was kept at 275 • C.
Optimized gradient settings for LC separation
The mobile phase reservoirs A and B contained 0.1% FA in water added 1% DMSO and 0.1% FA in ACN added 1% DMSO, respectively.A 150 µL/min solvent gradient was started at 1% B, quickly increased to 25% B in 1 min, then linearly increased to 32.5% B in 6 min, and kept at 32.5% B for 4 min.In the following step, the %B was quickly increased to 80% B and kept at 80% for 2 min, before quickly decreased to 40% B, and kept at 40% B for 3 min, before being further decreased to 1% B and kept at 1% B for 7 min.The gradient had a total runtime of 23 min, including column re-equilibration for 7 min at 1% B.
Settings for selected reaction monitoring
The collision energies and radio frequency (RF) lens voltage were optimized using the compound optimization provided in Xcalibur.The transitions, used in selected reaction monitoring, including collision energies and RF lens voltage, are listed in Table 1.The collision gas (argon) pressure in q2 was 0.53 Pa, and the cycle time was set to 1 s (equal to 77 ms or 91 ms dwell time per transition with and without urocortin-3, respectively).[18].The four analytes and the internal standard were successfully separated in a water-based solution with the originally applied gradient (8 min separation window from 1% B to 60% B [18]), shown in Figure 1A.However, when the analytes were dissolved in Krebs buffer, the separation of urocortin-3 from BSA was not possible, see Figure 1B.BSA is the main component TA B L E 1 Selected reaction monitoring (SRM) transitions used for human insulin, bovine insulin (internal standard for human insulin), human somatostatin-14, human glucagon, and human urocortin-3, including quantifier/qualifier status, precursor ion, product ion, and collision energy.
Compound Identity
Precursor ion (m/z) shows the programed gradient and the gradient due to system delay.The system delay of 6.5 min in the gradient delivery was estimated by running the analysis under non-retained conditions starting at 60% B, see Section S2 for more details.
in both Krebs buffer and in the cell medium that is used for culturing SC-islets, containing 0.1% BSA and 2% BSA, respectively.BSA produces a multitude of interfering peaks in the m/z range from 1000 to 1500 [18].Effort was put toward the optimization of the gradient, in order to achieve a better separation of the peptide hormones from BSA.By applying a shallow gradient from 25% B to 32.5% B, with an 1.25% B increase per minute, followed by an isocratic step at 32.5% B for 4 min, the BSA interference was removed from the insulins, somatostatin-14, and glucagon in both Krebs buffer (Figure 1C) and cell medium (Figure 1D).Urocortin-3 remained impossible to separate from the BSA in the Krebs buffer in the current setup, as shown in Figure 1C, and was not examined further in this study (see Conclusions sections for further discussion).In addition, the collision gas pressure was increased from 0.33 Pa used in the original method to 0.53 Pa due to significantly increased peak area for human insulin and glucagon, see Section S1 for more details.
In conclusion, by optimizing the applied gradient in the LC separation, the existing RPLC-MS/MS method could be improved to include three of the main hormones produced in islets: insulin, somatostatin-14, and glucagon, while not being limited by the sample matrix.However, the current column setup was not suitable for the inclusion of a more retained hormone, urocortin-3, which was not separated from the matrix components.
3.1.1
Guard cartridge: not one-size-fits-all for determination of intact peptides In the previous section and Olsen et al. [18], an Accucore phenyl/hexyl guard cartridge (80 Å, 2.6 µm particles, 2.1 mm inner diameter × 10 mm) combined with a Poroshell EC-C18 separation column (120 Å, 2.6 µm particles, 2.1 mm inner diameter × 50 mm) was applied in the LC-MS system.Traditionally, the guard column is packed with the same particles and stationary phase as the separation column.However, it is commonly accepted that combining different stationary phases on a two-column setup, with a trapping column and a separation column, can improve the separation [19].The combination of an Accucore phenyl/hexyl trapping column and Poroshell C18 separation column has previously been successfully applied for the determination of insulin in urine matrix [20].
To examine in detail if the phenyl/hexyl guard cartridge significantly affected the separation, the separation was compared with and without the phenyl/hexyl guard.In addition, an equivalent Accucore C18 guard cartridge (80 Å, 2.6 µm particles, 2.1 mm inner diameter × 10 mm) was assessed as an alternative to the phenyl/hexyl guard.Asymmetry factors and peak areas obtained on different column setups for the peptides are summarized in Table 2.
The asymmetry factor obtained for human insulin increased from 2.3 (N = 3, technical replicates on LC-MS; Figure 2A) with the phenyl/hexyl guard to 5.6 (Figure 2B) on the C18 guard cartridge, and the peak area was reduced from 7.79 × 10 3 on the phenyl/hexyl to 6.71 × 10 3 on the C18.Without applying a guard cartridge, the asymmetry factor for human insulin was 1.8 (Figure 2C), and the peak area was 8.05 × 10 3 .By one-factor analysis of variance (ANOVA), there was no significant difference in the obtained peak area of human insulin with or without a guard cartridge or type of guard.However, the tailing effect for human insulin obtained on the C18 guard was significantly larger compared to that on the phenyl/hexyl guard or without applying a guard.Additionally, there was no signal detected of human insulin in a subsequent blank injection on the setup without guard or with the phenyl/hexyl guard, but the C18 guard contributed to a 5% carryover of human insulin.
For bovine insulin, the asymmetry factor was not significantly different for the various column setups; however, the peak area was significantly higher with the phenyl/hexyl guard compared to the C18 guard.The peak area obtained with the phenyl/hexyl guard was not significantly different from the peak areas obtained without a guard.
For somatostatin-14, the asymmetry was not significantly different for the different column setups; however, the peak area was significantly increased without a guard cartridge being applied compared to the phenyl/hexyl guard, information summarized in Table 2.There was no significant difference between the C18 and the phenyl/hexyl guard or C18 and without guard applied concerning peak areas obtained for somatostatin-14.
For glucagon, the peak areas were significantly higher when applying the phenyl/hexyl guard cartridge compared to no guard cartridge applied or the C18 cartridge, whereas the asymmetry factor was not significantly different for the different column setups, summarized in Table 2.
The differences in the obtained chromatographic performances show that there is no "one-size-fits-all" approach when selecting guard cartridge for the determination of various intact peptides, and large effects for asymmetry factor and peak areas can be seen based on changing the stationary phase.Based on our findings, it is hard to conclude why the C18 guard cartridge caused dramatic tailing for human and bovine insulin, and if there was an effect of having particles with smaller pore sizes in the guard cartridge compared to the separation column (80 vs. 120 Å).When considering the use of a guard cartridge, TA B L E 2 Chromatographic performances by injection of a peptide mix consisting of 5 µg/L of bovine and human insulin, somatostatin-14, and glucagon.the LC-MS method was intended for experiments with supernatant collected directly from SC-islets possibly containing cell debris and particles, which would not be beneficial to inject directly onto the more expensive separation column without the protection from a guard cartridge.
Columns
In conclusion, we continued to employ a phenyl/hexyl guard cartridge based on the significantly higher peak areas obtained for glucagon and the guard cartridge's protection of the separation column when analyzing unfiltered samples collected directly from SC-islets.
3.1.2
The success of intact hormone analysis is dependent on the type of separation column All three hormones and the internal standard have a minimum of one sulfur-containing amino acid residue (i.e., cysteine and methionine), which may cause adsorption on stainless steel column housing and contribute to band broadening [21,22].To see if this was the case in our setup and if the chromatographic performance could be improved, a Premier column (C18, 2.6 µm, 90 Å) with modified metal surfaces in the column housing and filters to reduce non-defined adsorption was compared to the Poroshell column employed in previous sections [23].For our particular application, the peaks of insulins and glucagon obtained on the Premier column were broader and not sufficiently separated from BSA or other matrix components in the Krebs buffer, Figure 2D.The lack of separation indicated that the gradient optimized for the Poroshell column (C18, 2.7 µm, 120 Å) was not suitable for the Premier column.In the attempt to optimize the separation, the peak area of glucagon was greatly reduced for each subsequent injection (from a peak area of approx. 2 × 10 5 to a peak area <1 × 10 4 , results not shown), and none of the insulins were successfully separated from BSA (Figure 2E).On the Accucore phenyl/hexyl-Poroshell C18 column setup, an equivalent injection would give a stable peak area of glucagon around 250 000.For somatostatin-14, the Premier column setup was able to provide equivalent chromatographic performance as the phenyl/hexyl-Poroshell C18 column setup.
Based on the lack of separation from BSA and the loss of signal for glucagon on the Premier column, the Premier column was not found appropriate for our application and was not examined further.One of the major differences between the Premier column and the Poroshell column is the pore size of the particles of 90 and 120 Å, respectively.In Section 3.1.1,the guard cartridges with 80 Å particle pore size were sufficient for sample loading.However, from the comparison of the Premier and Poroshell column, it is clear that larger pores in the separation column were necessary to be able to separate the intact hormones from BSA and other matrix components [18,24].
In conclusion: Of the columns tested, the originally applied Poroshell C18 separation column was found to be the best option for the simultaneous determination of the three peptide hormones in complex matrices.
Satisfactory accuracy and precision in determining the concentration of insulin, somatostatin-14, and glucagon in quality controls in Krebs buffer
Glucose-stimulated insulin secretion is a standard characterization experiment of insulin response in SC-islets using Krebs buffer [25,26].The experiments require highly accurate determinations of hormone concentrations in Krebs buffer.Therefore, the method's repeatability concerning linearity, precision, and accuracy in determination of the concentration of the three hormones was examined by establishing calibration lines in the range from 5 to 15 µg/L in Krebs buffer over 3 days.The precision and accuracy for the determination of hormone concentrations were evaluated based on concentration found in QCs containing 10 µg/L of each analyte.
The established calibration lines for human insulin (with bovine insulin as internal standard, see Figure 3A), somatostatin-14 (Figure 3C), and glucagon (Figure 3E) did not show signs of heteroscedasticity (i.e., difference in variation of the response depending on the concentration level) as the residuals appear to fall randomly around the x-axis in the residual versus concentration plots in Figure 3B,D,F, respectively [27].The hypothesis of homoscedasticity was confirmed with an F-test comparing the variance in the response of the standard with the lowest and highest concentration levels for each of the analytes.
Bovine insulin was a suitable internal standard for human insulin due to similar structures, similar retention time, and the variation of the peak area of bovine insulin was 4% RSD (N = 5) on day 1, 3% RSD (N = 5) on day 2, and 3% RSD (N = 5) on day 3.The variation in the peak area of bovine insulin was much smaller during these examinations compared to the examination done in Section 3.1.1,as changes to the system were not done between injections allowing for better stability.Determination of somatostatin-14 and glucagon was done without the use of bovine insulin as an internal standard, due to large differences in the structures of the analytes compared to bovine insulin, and different retention times.These differences indicate that bovine insulin is not a suitable internal standard as it cannot compensate for differences in matrix effect (different retention times) nor the transfer from ions in solution into gas phase occurring during the ESI process (different structures).Bovine insulin could compensate for differences in injection volume; however, there is a relatively small variation between repeated injections that such a compensation was not deemed necessary.The retention time variance over the 3 days was ≤0.5% RSD (N = 24) for all of the hormones, see Figure 3G.
In the determination of QC analyte concentration using the same standard solutions to establish a calibration curve, the accuracy was within ±10% relative error (N = 3, per day), ±10% RSD (N = 3) intraday precision, and ±10% RSD (N = 9) interday precision.Determination based on a single injection of the QC, only somatostatin-14 was not determined with accuracy within ±15% relative error, as shown in Figure 3H for injection number nine.The average concentration of each analyte found in the QCs on the 3 separate days was not significantly different determined with one-way ANOVA.
To summarize, the RPLC-MS/MS method, featuring a phenyl/hexyl guard and Poroshell C18 separation col-umn combined with triple quadrupole mass spectrometry, was successful in the simultaneous determination of the three analytes; human insulin (including bovine insulin as internal standard), somatostatin-14, and glucagon in Krebs buffer with sufficient accuracy, precision, and repeatability.
Stem cell-derived islet organoids show glucose concentration-dependent insulin response and potassium-dependent release of somatostatin-14 and glucagon in Krebs buffer
As mentioned in the previous section, the secretion of peptide hormones in SC-islets can be examined by exposing the SC-islets to Krebs buffer containing various amounts of glucose.It has also been shown that the three major cell types found in islets, beta-, delta-, and alpha cells release a large pool of stored hormones when exposed to high levels of potassium due to direct membrane depolarization [11,28,29].Hence, to show compatibility between supernatant collected from SC-islets in Krebs buffer and RPLC-MS/MS, we attempted to measure the three hormones in Krebs buffer collected from SC-islets exposed to (1) a low amount of glucose (2 mM), (2) a high amount of glucose (20 mM), and (3) a combination of 20 mM glucose and 30 mM KCl.
The concentration of insulin determined with RPLC-MS/MS in the supernatant collected from SC-islets challenged with low glucose (Figure 4A) was on average 2.2 µg/L (RSD = 13%, n = 4 batches of SC-islets, N = 1), whereas there was 4.1 µg/L insulin (RSD = 10%, n = 4, N = 1) secreted by SC-islets challenged with high glucose (Figure 4B).The stimulation index of insulin secretion during the glucose challenge was 1.86 (RSD = 5%, n = 4, N = 1).The amount of insulin determined in the samples collected with 20 mM glucose and 30 mM KCl (Figure 4C) was 18 µg/L (RSD = 14%, n = 4, N = 1), which was eight times higher than the response in low glucose.Simultaneously, the reliability of the determination of insulin using the RPLC-MS/MS method was supervised by QCs, to see if the reported concentrations were within 15% relative error.QCs were included at 2 µg/L (N = 2), 8 µg/L (N = 3), and 18 µg/L (N = 2), and the insulin concentration was determined within 15% relative error for each injection.
The insulin concentration in the supernatant collected from the SC-islets was also determined with an established ELISA method, which found the following insulin concentrations: 2.6 µg/L (low glucose, RSD = 5%, n = 4, N = 1), 4.5 µg/L (high glucose, RSD = 13%, n = 4, N = 1), and 20 µg/L (KCl, RSD = 13%, n = 4, N = 1).An independent two sample t-test, at 95% confidence, showed that the concentrations determined by LC-MS/MS were not significantly different from the concentrations determined with ELISA.A large-scale comparison study consisting of several experiments should be examined to strengthen or disprove the indication that LC-MS/MS and ELISA provide similar determinations of insulin.
The same samples were also simultaneously examined for somatostatin-14 and glucagon.Neither the quantifier nor the quantifier transition of somatostatin-14 and glucagon was detected in the supernatant from SC-islets exposed to low (Figure 4A) or high glucose (Figure 4B); however, in supernatant collected after exposure to KCl, detectable signals were obtained for both hormones and transitions (Figure 4C).For somatostatin-14, the average peak area was 1.3 × 10 3 (RSD = 7%, n = 4, N = 1) after KCl exposure, which was lower than the peak area of 1.9 × 10 3 obtained for the calibration standard with the least amount of somatostatin-14 of 0.25 µg/L (Figure 4D).For glucagon, the concentration was determined to be 0.28 µg/L (RSD = 18%, n = 4, N = 1) after KCl exposure.The QCs examined for somatostatin-14 and glucagon at 2 µg/L and 8 µg/L were determined within 11% relative error (RSD < 10%), showing that the RPLC-MS/MS method for somatostatin-14 and glucagon has sufficient precision and accuracy without the use of an internal standard.
For insulin, it was beneficial that the samples had been examined by a clinically approved ELISA kit prior to being analyzed by the LC-MS method [30], as a suitable calibration concentration range could easily be selected.For the other hormones, somatostatin-14 and glucagon, a gold standard for the determination of the concentrations has yet not been established [31][32][33].A calibration concentration range was selected without prior information about the expected concentration of analytes in the sample, and the concentration in the samples was found to be below or around the lowest concentration calibration standard (0.25 µg/L).Therefore, the samples collected with 20 mM glucose and 30 mM KCl were reexamined with a calibration from 0.1 to 3 µg/L for somatostatin-14, and 0.05 to 3 µg/L for glucagon, including QCs at 0.4 µg/L.The concentration for somatostatin-14 could now be determined and was found to be 0.27 µg/L (RSD = 20%, n = 4, N = 1) in the samples collected with KCl.For glucagon, the concentration was determined with the new calibration curve to be 0.31 µg/L (RSD = 18%, n = 4, N = 1), which was not significantly different from the concentration determined with the first calibration curve based on an independent two sample t-test (95% confidence).The QCs examined at 0.4 µg/L were all within 10% relative error for somatostatin-14 (N = 3), and 11% relative error for glucagon (N = 3).
The carryover was less than 1% for all of the analytes and the internal standard, shown in a representative chromatogram from a blank injection of 0.5% FA in Krebs buffer following injections of the standards used to establish the curve in Figure 4E.In addition, the carryover was equal to less than 20% of the peak area obtained in the calibration standard with the smallest concentration of the analytes, giving a lower limit of quantification of 0.2 µg/L for human insulin, 0.1 µg/L for somatostatin-14, and 0.05 µg/L for glucagon.The retention time variation was less than 0.2% RSD (N = 26) for all of the analytes and the internal standard.
A challenge when examining supernatant collected from SC-islets after exposure to different levels of glucose and KCl is the change in the sample matrix.In this study, Krebs buffer without glucose or KCl was used as the solution for the preparation of the calibration standards.Possible effects of glucose and KCl in the samples have not been examined.The ratio of the quantifier and qualifier transitions obtained for each hormone is shown in Figure 4F.For bovine and human insulin, the quantifier/qualifier ratio was not significantly different in samples with different amounts of glucose or KCl (determined with one-way ANOVA).In addition, for all analytes and bovine insulin, an independent two sample t-test, at 95% confidence, confirmed that there was no significant difference in the quantifier/qualifier ratio obtained in the samples compared with the quantifier/qualifier ratio obtained in the calibration standards and QCs.The identification and quantification in the RPLC-MS/MS method were secured by negligible variance in retention time and quantifier/qualifier ratios, negligible levels of carryover (<1%), and sufficient precision and accuracy.
In conclusion, the RPLC-MS/MS method demonstrates sufficient quantification limits for the determination of insulin secretion in the supernatant of SC-islets (n = 30).In addition, we show that the SC-islets obtained through our differentiation protocol [18] have obtained a glucose dependent insulin secretion in response to 2 and 20 mM glucose.The method can also determine the production of somatostatin-14 and glucagon in 20 mM glucose and 30 mM KCl.However, better sensitivity is needed to determine the secretion of somatostatin-14 and glucagon in SC-islets (n = 30) stimulated by glucose alone.
3.4
Insulin could be determined in background secretion from a small number of stem cell-derived islet organoids cultivated in a pumpless, recirculating organ-on-a-chip device Hormone secretion from isolated mice or human islets-onchip has been determined with the use of, for example, luminescent immunoassay (AlphaLISA) [34], or ELISA [35].However, to the authors' knowledge, the combination of human SC-islets, organ-on-a-chip device, and hormone secretion determination with LC-MS has not been previously reported.As a proof-of-concept for combining RPLC-MS/MS determination of intact hormones from SCislets cultured on an OoC device, we wanted to examine the background secretion of the hormones.In addition, we wanted to show that the RPLC-MS/MS method was versatile concerning the applied sample matrix and therefore did not change the islet maturation cell medium (will be referred to as cell medium) with Krebs buffer for this experiment.
The background secretion over 24 h (in cell medium containing 5.5 mM glucose) from 3 to 6 SC-islets cultivated on-chip in a rOoC device was compared to 14−19 SC-islets cultivated in a standard cell culture well-plate.
To avoid introducing variation in the quantification, bovine insulin was not applied as an internal standard during the determination of human insulin in cell medium, due to a higher variation in the peak area (RSD > 7%, N = 6) in cell medium compared to Krebs buffer (RSD < 5%, see Section 3.2).At day 5 on the rOoC, an average of 5 µg/L of insulin was secreted per SC-islet (RSD = 52%, n = 4, N = 1), whereas on day 7, an average of 3 µg/L of insulin was secreted per SC-islet (RSD = 30%, n = 4, N = 1).Similarly, on the well-plate at day 5, an average of 7 µg/L of insulin was secreted per SC-islet (RSD = 24%, n = 4, N = 1), whereas on day 7, an average of 3.9 µg/L of insulin was secreted per SC-islet (RSD = 7%, n = 4, N = 1).An independent two sample t-test, at 95% confidence, confirmed that there was no significant difference in insulin secretion per SC-islet on the rOoC compared to the cell culture wellplate.The high variation in the determination of insulin secretion per SC-islet on the rOoC device (54% on day 5 and 30% on day 7) compared to the 24 well-plate (24% on day 5 and 7% on day 7) can be explained by the number of SC-islets included on the different devices.There were 3−6 SC-islets on the rOoC, whereas there were 14−19 SCislets included on the well-plate.Individual differences in the SC-islets may affect the reliability of the measurements when examining a small batch of SC-islets, and the lower RSD values on the 24 well-plate indicate that a representative batch of SC-islets should be closer to 20 individual SC-islets.
Concerning background secretion of somatostatin-14 and glucagon in 3−6 SC-islets cultivated on rOoC device, the detection limits were not sufficient for the quantification of secretion from a limited number of SC-islets stimulated by only glucose, for details see Section S3.We were able to determine the secretion of glucagon in the samples collected from 14 to 19 SC-islets on the cell culture well-plate.
The composition of the SC-islets was confirmed with flow cytometry quantification and immunostaining (Figure 5A), for details see Sections S4.The SC-islets consisted of >66% insulin-positive cells, >17% somatostatin-positive cells, and >22% glucagon-positive cells, whereas >95% of the cells were endocrine cells (i.e., cells which can secrete hormones).The multicellular SC-islets have a composition, which is similar to the distribution of the cell types in human islets [13,15].
To summarize, the determination of secreted insulin from a limited number SC-islets (n < 6) cultivated in cell medium on an rOoC device was possible with the applied RPLC-MS/MS method.It was found that the insulin secretion in the SC-islets was not significantly different on the rOoC device compared to the secretion occurring on a standard cell culture well-plate.
F I G U R E 5 (A) Representative immunostaining images of SC-islets stained for C-peptide (green), glucagon (red), somatostatin-14 (magenta), and cell nuclei with Hoechst 333242 (white).Scale bar = 25 µm.Representative chromatograms obtained in selected reaction monitoring (SRM) mode of supernatant from SC-islets cultivated (B) in a 24 well-plate and (C) on the recirculating organ-on-chip (rOoC) device, and (D) calibration standard with 100 µg/L human insulin, 0.1 µg/L somatostatin-14, and 0.1 µg/L glucagon with 5 µg/L bovine insulin in 1.0% formic acid (FA) in cell medium.The first transition for each hormone is the quantifier, whereas the second transition is the qualifier.(E) Coomassie blue stained protein bands were found by gel electrophoresis in the following samples: (1) 10 mg/L of human insulin and 10 mg/L of bovine serum albumin (BSA) in water, (7) 1.0% FA in cell medium, (8) pooled supernatant collected on day 5 from SC-islets cultured in a 24 well-plate, (9,10) two replicates of supernatant from SC-islets incubated on rOoC collected on day 5.The picture of the gel has been cropped, where Lanes 2−6 are not included; however, the picture of the gel is provided in raw format in Section S5. (F) The ratio of quantifier/qualifier transitions for bovine insulin (squares), human insulin (triangles), somatostatin-14 (diamonds), and glucagon (circle) obtained for calibration standards and QCs (dark gray), and samples (light gray).Both transitions of somatostatin-14 were not detected in any of the examined samples.
3.4.1
Discussion: concerning challenges with cell medium, the organ-on-chip device, and the applied liquid chromatography-mass spectrometry method In the present method, an interfering peak (eluting after 13 min in standard solutions in cell medium) was eluting closer to the analytes and co-elute with glucagon in cell medium incubated with SC-islets on well-plate (Figure 5B) and on an rOoC device (Figure 5C).In comparison, there is a baseline separation of the analytes and the interfering peak in the standard solution prepared with fresh cell medium spiked with the analytes and internal standard (Figure 5D).The separation in the samples may be affected by other sample matrix components introduced following incubation of the SC-islets on the 24 well-plate or in the rOoC device.By gel electrophoresis, it was possible to confirm the presence of various proteins in the size range between human insulin and BSA (see Lane 1, Figure 5E) in supernatant collected from SC-islets that were either cultivated in a 24 well-plate (Lane 8) or in the rOoC device (Lanes 9 and 10).The observations suggest that the extra sample matrix components are released into the cell medium by the SC-islets or the extracellular matrix used for embedding the SC-islets on the two devices.Indeed, in cell medium, neither incubated with SC-islets nor been in contact on either of the devices, significantly fewer protein bands were visible in the same size range.Gel electrophoresis was also used to compare cell medium with Krebs buffer concerning protein content, showing that there were more sample matrix components present in cell medium compared to Krebs buffer, see Section S5 for more details.The gel is overloaded but is shown here to give an indication of the complexity of the matrix, as the amount applied to the gel is similar to that injected on the LC column.
It is worth noting that during the preliminary examination of the rOoC device with Krebs buffer, there was a significant loss of human insulin following incubation on the rOoC device compared to standard cell culture well-plates, see details in Section S6.
In the current experiment, the reliability of the determination of hormones with the RPLC-MS/MS method was not affected by the changes to the separation due to the presence of additional sample matrix components.The identification was secured by a stable ratio of the quantifier and qualifier transitions obtained for each hormone in calibration standards and the sample shown in Figure 5F, as there was no significant difference in the ratio obtained in calibration standards and samples.In addition, the retention time of each hormone varied ≤0.5% RSD (N = 30) and the carryover was <2% for glucagon, <0.5% for human insulin, and <0.1% carryover for bovine insulin and somatostatin-14.Additionally, all QCs (examined before, within, and after the sample set) at 25 µg/L of human insulin and 0.8 µg/L of somatostatin-14 and glucagon were determined within ±15% relative error (N = 5).The only exception was for the determination of glucagon, where the relative error was 17% in the second replicate of the QC.
To summarize, the RPLC-MS/MS offers a reliable determination of insulin, somatostatin-14, and glucagon in a complex matrix (cell culture supernatant in this study) without the use of internal standard, with negligible variation in retention time, repeatable quantifier/qualifier transition ratios, and negligible levels of carryover.With the growing complexity of the cell medium samples following incubation with the SC-islets embedded with extracellular matrix on the 24 well-plate and in the rOoC device, we are approaching a limit where the RPLC-MS/MS method alone is not sufficient for reliable determination.In the case of even more complex samples, the inclusion of sample preparation steps might become necessary.
CONCLUDING REMARKS
The study was dedicated to the determination of multiple peptide hormones, insulin, somatostatin-14, and glucagon at a low µg/L range, in samples collected from SC-isletson-chip using liquid chromatography-mass spectrometry.
It has been shown that liquid chromatography is needed to separate the target peptides from the interfering BSA present in the sample matrices used to study SC-islets.However, when the complexity of the samples grows and large amounts of proteins are present, chromatography and mass spectrometry may not be enough for successful peptide determination (e.g., urocortin-3 coelutes with BSA).The interference of BSA points to the need for the inclusion of sample preparation steps prior to RPLC-MS/MS analysis, for example, electrophoresisbased extractions of target peptides.Sample preparation may be of utmost importance when the field moves toward studying multiple organoid types on the same OoC system, for example, metabolism-on-chip using liver and islet organoids.
F I G U R E 1
Separation of a peptide mix consisting of bovine insulin, human insulin, somatostatin-14, glucagon, and urocortin-3 in different sample matrices.With the original steep gradient: (A) 1.08 µL injection of 250 µg/L peptide mix in 0.1% formic acid (FA) in water and (B) 20 µL injection of 2 µg/L peptide mix in 0.5% FA in Krebs buffer.With an optimized shallow gradient: (C) 20 µL injection of 5 µg/L peptide mix in 0.5% FA in Krebs buffer and (D) 20 µL injection of 5 µg/L peptide mix in 1.0% FA in cell medium.The first panel above each separation
Note:
The obtained asymmetry factor (A s ) and peak area of the peptides are shown for the following guard cartridge options: Accucore phenyl/hexyl, Accucore C18, and no guard cartridge, combined with the Poroshell EC-C18 separation column (N = 3).F I G U R E 2Comparison of peak shape obtained for human insulin on the following column setups: (A) Accucore phenyl/hexyl guard-Poroshell C18, (B) Accucore C18 guard-Poroshell C18, and (C) only the Poroshell C18 column.Attempted separation of 5 µg/L peptide mix in 0.5% Krebs buffer on the Premier column with different gradients: (D) insufficient separation of insulins and glucagon from bovine serum albumin (BSA) with the same gradient as applied for the Poroshell column.(E) Following attempted gradient optimization, still insufficient separation of insulins and BSA, and low signal intensity for glucagon.
F I G U R E 3
Established calibration lines in the range from 5 to 15 µg/L for (A) human insulin with bovine insulin as internal standard, (C) somatostatin-14, and (E) glucagon, where day 1 is represented with diamonds, triangles for day 2, and squares for day 3. Residual versus concentration plots for each curve belonging to (B) human insulin, (D) somatostatin-14, and (F) glucagon, where day 1 is represented with diamonds, triangles for day 2, and squares for day 3. (G) The retention time for bovine insulin (squares), human insulin (triangles), somatostatin-14 (diamonds), and glucagon (circles) over the 24 injections of calibration standards and QCs.(H) Determined concentration in QCs for human insulin (triangles), somatostatin-14 (diamonds), and glucagon (circles) in the nine injections.
F I G U R E 4
Representative chromatograms obtained in selected reaction monitoring (SRM) mode of supernatant from SC-islets exposed to (A) 2 mM glucose, (B) 20 mM glucose, and (C) 20 mM glucose with 30 mM KCl. (D) Calibration standard with 0.25 µg/L human insulin, somatostatin-14, and glucagon with 5 µg/L bovine insulin in 0.5% formic acid (FA) in Krebs buffer and (E) blank injection of 0.5% FA in Krebs buffer.The first transition for each hormone is the quantifier, whereas the second transition is the qualifier.(F) The ratio of quantifier/qualifier transitions for bovine insulin (squares), human insulin (triangles), somatostatin-14 (diamonds), and glucagon (circle) obtained for calibration standards and QCs (dark gray), and samples (light gray).
Accucore phenyl/hexyl- Poroshell C18 Accucore C18-Poroshell C18 No guard-Poroshell C18 Bovine insulin
4 (4% RSD) 4.44 × 10 4 (3% RSD) 4.57 × 10 4 (1% RSD) Financial support was obtained from the Research Council of Norway through its Centres of Excellence funding scheme, project number 262613, and partly from the UiO:Life Science convergence environment funding scheme.S.R.W. is a member of the National Network of Advanced Proteomics Infrastructure (NAPI), which is funded by the Research Council of Norway INFRASTRUKTUR-program (project number: 295910). | 2023-06-16T13:11:27.334Z | 2023-06-12T00:00:00.000 | {
"year": 2023,
"sha1": "a76212a46898766df994553c1bcad667bfe91f69",
"oa_license": "CCBY",
"oa_url": "https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/elps.202300095",
"oa_status": "HYBRID",
"pdf_src": "Wiley",
"pdf_hash": "46b999b14810c7521c8f2bc56e46736ed7ac0e0c",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology"
]
} |
242537737 | pes2o/s2orc | v3-fos-license | An Exploration of ECG Signal Feature Selection and Classification using Machine Learning Techniques
Abstract-This effort examines and likens a collection of active methods to dimensionally reduction and select salient features since the electrocardiogram database. ECG signal classification and feature selection plays a vital part in identifies of cardiac illness. An accurate ECG classification could be a difficult drawback. This effort also examines of ECG classification into arrhythmia kinds. This effort discusses the problems concerned in Classification ECG signal and exploration of ECG databases (MIT-BIH), pre-processing, dimensionally reduction
I. INTRODUCTION
Cardio vascular diseases affect 17.9 million (estimated 31%) people worldwide every year. ECG Signal could be a medical examination that detects heart irregularity by through calculating the heart's electrical and muscle activity. A cardiac creates small electrical impulses that feast over cardiac muscle. Those impulses are often identified with ECG instrument. ECG instrument records the electrical action of the cardiac and shows this information on ECG graph sheet. That information is understood by a health care provider. ECG assistance to seek out the reasons for sign or heart pain and additionally helps to observe abnormal cardiac rhythm (heart defects). Usually healthy cardiac have a typical form. Some abnormalities with in the cardiac rhythm otherwise injury to the cardiac muscle will modification the electrical action of heart, therefore form of ECG signal gets modified. Clinician might suggest associate electro cardiogram for patients who might in danger of cardiac condition as a result of household history of heart condition, smoking etc. The cardiac disorders which will be detected using ECG take in irregular cardiac rhythm and distended heart. Electrocardiogram is one of the most effective cardiac illness diagnostic tools. Owing to high death rate of cardiac illness, initial recognition and accurate discernment of ECG signal stands important of the action of patients. Cardiogram signals classification exploitation optimized techniques will offer significant contribution to doctors to verify the diagnosing and recognition of heart illness sorts will facilitate classifying the abnormality of a patient. When classifying the abnormality, the cardiac illness is noticed and therefore recovering treatment of the patient is done. Precise cardiogram classification into heart disease kinds offers enough info to recognition the cardiac illness. Classification of cardiogram signals is a difficult issue due to classification process problem. Foremost classification of ECG problems, normalization of ECG features is lacking, changeability, originality, absence of best classification and inconsistency in graphs of patients. Emerging the foremost suitable classifier that's accomplished of classifying illness, ECG signal classifier main application is detect the cardiac illness diagnosis.
II. BACKGROUD KNOWLEDGE
The main function of ECG is heart's electrical and muscle activity. The electrodes mounted on skin are measured. It measures the pulse rate, rhythm as well as indirect evidence of blood flow to the heart muscle. To generate twelve leads of electrical heart views, ten electrodes required. Cardiac signals classification is one of the important roles in cardiac illness medical diagnosis and drawback is diagnosing cardiac illness using electro cardiogram is normal signal vary for every person and occasionally single illness has different signs on dissimilar patients cardiogram signals. The above drawbacks complicate the cardiac illness diagnosis. So use of classifier techniques can develop the new patient's cardiac arrhythmia diagnosis. Cardiac arrhythmia is a category of disorders where the heart beat is too fast otherwise too slow irregular (>120 beats/min -Tachycardia, <60 beats/min -Bradycardia). Table.3. displays the detailed exploration of classification arrhythmia.
III. MATERIALS AND METHODS
In this effort ECG classification consist of six blocks
Output Block
Classification of Arrhythmia and performance measures B. Pre-processing Block: Filtering was used to pre-process the cardiac signals. High-frequency noise and low-frequency noise should be eliminated. Different kinds of noise affect cardiac signals for example high frequency noise include electromyogram noise, white Gaussian additive noise, power line interference and low frequency noise include baseline wandering. [6]. Many investigators had applied different noise removal (pre-processing) techniques such as wavelet transform based technique, curvelet transform based technique and Adaptive digital filters. Thakor and Zhu [32] had done the noise removal by a digital filter using continuous otherwise unity reference input. That was used to remove low frequency noise. Yet, this type of filter isn't consistent for applications that require diagnostic ECG analysis.
IV. RESULT ANALYSIS AND DISCUSSION
In this section we described results of two blocks (i) pre-processing block and (ii) dimensionally reduction block and exploration of ECG signal classification. Signals of normal and cardiac arrhythmia are extracted from the MIT-BIH database. In this pre signal processing block, simple approach to power line interference reduction is to consider a filter identified by a complex-conjugated pair of zeros at the interfering frequency in the unit circle. Implements (Matlab) the two IIR (notch) filters, one to eliminate the low frequency noise with 0.68 Hertz and another one eliminate the high frequency noise with 50 Hertz. The results of the pre-processing block displays in figure.4. Figure.4. Simulation result of ECG signal with low and high frequency noise, ECG signal with low frequency noise removed and ECG signal with high frequency noise removed.
In this dimensionally reduction block, linear and non-linear approach to high dimensional data to low dimensional reduction data reduction is to consider linear technique and non-linear technique. Advantage of linear technique is simple geometric representation and desirable computing properties and advantage of non-linear technique is sparse matrix used so computational efficiency high and polynomial speedup. Samples are divided into equal sets, 80% of training and 10% of testing. High dimensional data displays in figure.5 using matlab, results of dimensionally reduction displays in figure.6 using linear technique and results of dimensionally reduction displays in figure.7 using Non-linear technique. Feature extraction carried out Pan-Tompkins algorithm. Accuracy 80-85% Cascade FFNN with BPNN algorithm [27] Pre-processing and feature selection carried out.
V. CONCLUSION
This effort showing a relative table estimating the performance of various algorithms that was carried out before the ECG signal classification and machine learning technique area of those classifiers is also shown in Table.4. This feature selection is done in completely involuntary technique. These results show that, the different classification system significantly improves the detailed fitness attained by the various classifiers and it is heftiness against the issue of limited availability of training beat that may characterize unusual occurrence pathologies. Refining the accurateness of analyzing the heart illness in advance is important with in the case of human monitoring system. So, forthcoming efforts can extend the diagnosis accuracy of the heart illness. All of those approaches can offer information to the person who reads about the finest signal pre-processor block, dimensionally reduction block, feature selector block and classifier block machine learning techniques aimed at arrhythmia classification. | 2020-01-30T09:03:24.102Z | 2020-01-10T00:00:00.000 | {
"year": 2020,
"sha1": "5d23d4a5abc780e9afdf12371fbe9ab5957d4db9",
"oa_license": null,
"oa_url": "https://doi.org/10.35940/ijitee.c8728.019320",
"oa_status": "GOLD",
"pdf_src": "Unpaywall",
"pdf_hash": "36f8714c0edea79a7465983cd3dc4c7ebbd72a30",
"s2fieldsofstudy": [
"Computer Science",
"Medicine"
],
"extfieldsofstudy": []
} |
122949992 | pes2o/s2orc | v3-fos-license | A Review on Extension of Lagrangian-Hamiltonian Mechanics
rastogivikas@yahoo.com Sant Longowal Institute of Engineering and Technology This paper presents a brief review on Lagrangian-Hamiltonian Mechanics and deals with the several developments and extensions in this area, which have been based upon the Mechanical Engineering Department principle of D’Alambert or the other. It is not the intention of the authors to attempt to 148106 Longowal, Punjab, India provide a detailed coverage of all the extensions of Lagrangian-Hamiltonian Mechanics, Amalendu Mukherjee whereas detailed consideration is given to the extension of Noether’s theorem for nonconservative systems only. The paper incorporates a candid commentary on various amalendu@iitkgp.ac.in extensions including extension of Noether’s theorem through differential variational principle. The paper further deals with an extended Lagrangian formulation for general Anirban Dasgupta class of dynamical systems with dissipative, non-potential fields with an aim to obtain invariants of motion for such systems. This new extension is based on a new concept of anir@iitkgp.ernet.in umbra-time, which leads to a peculiar form of equations termed as ‘umbra-Lagrange’s Indian Institute of Technology Mechanical Engineering Department equation’. This equation leads to a simple and new fundamental view on Lagrangian 721302 Kharagpur, W. Bengal, India Mechanics and is applied to investigate the dynamics of asymmetric and continuous systems. This will provide help to understand physical interpretations of various extensions of Lagrangian-Hamiltonian Mechanics.
complex physical systems, and are preferable compared to the Newtonian approach.However, the Lagrangian approach has certain limitations.The elimination of the constraint forces from the Lagrange's formulation does not allow one to directly calculate these forces.They can, however, be determined using an indirect approach.Besides this, Lagrange's equation suffers heavily in the presence of time fluctuating parameters, non-potential fields, general dissipation and gyroscopic forces.Derivation of the Lagrange's equations of motion for nonconservative and dissipative system (Rosenberg, 1977;Meirovitch, 1970;Whitaker, 1959) is essentially patchwork.This hinders the analysis of such systems, which the Lagrangian can afford.Nevertheless, the greatest advantage of Lagrangian formulation is that it brings out the connection between conservation laws and important symmetry properties of dynamical systems.Knowledge of conservation laws is of great importance in the analysis of dynamical systems as they lead to a complete integrability of dynamical system.The fundamental symmetries motivated the study of conservation laws from geometrical and group-theoretical point of view.The theorem of Emmy Noether (Noether, 1918) is one of the most fundamental justifications for conservation laws.Her theorem tells us that conservation laws follow from the symmetry property of nature.
Introduction
From the late seventeenth century to the nineteenth century classical mechanics (Goldstein, 1980;Sudarshan and Mukunda, 1974) was one of the main driving forces in the development of physics, interacting strongly with developments in mathematics, both by borrowing and lending.In fact, mechanics and indeed all theoretical science is a game of mathematical make-believe.The topics developed by its main protagonists, Newton, Lagrange, Euler, Hamilton and Jacobi among several others form the basis of classical mechanics.
Since the last few decades, the subject of classical mechanics itself was undergoing a rebirth and expansion with strong developments in mathematics.There has been an explosion of research in the classical dynamical systems, focused on the discovery of advanced mathematics (e.g.Lie Algebra, differential geometry, etc.) (Sattinger and Weaver, 1986;Bluman and Kumei, 1989;Gilmore, 1974).The aforementioned occurrences in the second part of the 20 th century have radically changed the nature of the field of classical mechanics.The first development has led to modeling and analysis of complex, multi-bodied (often elastic bodied) structures, such as satellites, robot manipulators, turbo machinery and vehicles.The second has led to the development of numerical techniques to derive the describing equations of motion of a dynamical system, integration, simulation and obtaining the response.This new computational capability has encouraged scientists and engineers to model and numerically analyze complex dynamical systems, which in past either they could not be analyzed, or were analyzed using gross simplifications.
The prospect of using computational techniques to model a dynamical system has also led dynamicists to reconsider existing methods of obtaining equations of motion.The methods of Lagrange (Lagrange, 1788) and Hamilton (Baruh, 1999) are used to carry out the primary task of deriving the equations of motion.Generalized coordinates which do not necessarily have to be physical coordinates are used as motion variables in these methods.This makes the Lagrangian-Hamiltonian approach more flexible than the Newtonian, as Newtonian approach is implemented using physical coordinates.The use of Lagrange's formulation of dynamics offers the quickest way of deriving system equations for The objective of the paper is to present the developments in the field of Lagrangian-Hamiltonian Mechanics with particular regard to extension of Noether's theorem.In recent years, the authors have attempted to develop alternative method to the construction of the first integrals of dynamical systems by means of extended Noether's theorem.Typical contributions in this area are given in reference (Rastogi, 2005;Mukherjee et al., 2006;Mukherjee et al., 2007;Mukherjee et al., 2009).Investigating such alternatives has been applied to analyze the dynamics through invariance of the action integral for some engineering applications, which are rarely applied.
Evolution of Classical Lagrangian-Hamiltonian Mechanics
The major contribution in classical mechanics came from Lagrange (1788).The contributions of Lagrange put the field of analytical mechanics into a structured form now known as Lagrangian mechanics.In the original derivation, Lagrange's equations were written for conservative systems only, and applicable when the system is closed, constraints are integrable, and there are no gyroscopic forces.Hamilton (Baruh, 1999;Gantmachar, 1970;Calkin, 2000) has developed the most general principle of least action and showed that the Lagrangian with time integration provided the definition of action and minimization of this action integral established the Lagrange's generalized equation.The main advantage of this new formulation is that it holds for any system subject to constraints and independently of the co-ordinates, which are chosen to represent the motion.However, the problem of dissipation was handled by Rayleigh (Gantmachar, 1970;Jose and Saletan, 1998), who attempted to enlarge the scope of Lagrange's equation to incorporate dissipative forces in this generalized equation.He added velocity dependent potential through virtual work done by the dissipative elements and then re-encapsulated in an extended formula.In this formulation, the velocity's dependent potential should not be brought inside the scope of total derivative with respect to time, otherwise an unrealistic momentum and inertia would enter in the equation.This is the reason why the velocity's dependent Rayleigh potential fails in the case of gyroscopic forces.
The next problem is to deal with Nonholonomic systems in classical mechanics as to determine the equation of motion for constrained systems.When physical constraints are imposed on an unconstrained set of particles, forces of constraints are engendered, which ensure the satisfaction of the constraints.The equation of motion developed for such constrained systems is based on the principle of D'Alambert, and later elaborated by Lagrange (1788) through Lagrange multipliers.Since its initial formulation by Lagrange, the problem of constrained motion has been vigorously and continuously worked on by various scientist including Volterra, Boltzmann, Hamel, Whittaker and Synge to name a few.Gauss (1829) explained a new general principle for the motion of constrained mechanical systems referred to as Gauss's principle, by making use of acceleration.Gibbs (1879) and Appell (1899) independently discovered a new equation, which is known as the Gibbs-Appell equations of motion (Appell, 1899).Pars (1979) also referred to the Gibbs-Appell equations as the most comprehensive equations of motion so far discovered.Routh Gantmachar (1970) proposed the equations of motion in a potential field taking a part of the Lagrangian variables and a part of the Hamiltonian variables, called as Routh variables.Lie (Hassani, 1999;Olver, 1986) introduced the group theory for canonical transformations by considering infinitesimal transformations.
Extensions of Lagrangian-Hamiltonian Mechanics through Variational Principle and Group-theoretical Approach
Apparently, the first to notice the connection of conservation laws to invariance properties of dynamical systems was Jacobi (1884), who has derived the conservation law for linear and angular momentum from the Euclidean invariance of the Lagrangian.Emmy Noether (1918) formulated a theorem to find the invariants of the dynamical system and showed a relationship between symmetry aspects of conservation laws and invariance properties of space and time, i.e., their homogeneity and isotropy.These fundamental symmetries motivated the study of conservation laws from geometrical and group-theoretical point of view.Most of the results of conservation laws of classical mechanics based on Noetherian approach could be found in the research papers of Hill (1951), and Desloge and Karch (1977), where it has been applied as a reliable tool to find new conservation laws of dynamical systems.
The physics associated with the classical conservation laws widely attracted the investigations in this field, intriguing problems of classical mechanics by engineers and theoretical physicists, who formulated newer types of constant of motion.In their several papers, Vujanovic (1970), Djukic andVujanovic (1975) andVujanovic (1978) have investigated this field of analytical mechanics and developed a new approach to obtain constants of motion.Vujanovic (1970) has established a group-variational procedure for finding first integral of dynamical systems.Djukic and Vujanovic (1975) have proposed a Noether's theorem for mechanical system with non-conservative forces.Primarily, this theory was based on the idea that the transformations of time and generalized coordinates together with dissipative forces determine the transformations of generalized velocities.Vujanovic (1978) has reported a method for finding the conserved quantities of nonconservative holonomic systems based on the differential variational principle of D'Alembert, which was equally valid for both conservative and non-conservative systems.His research work has shown that the existence of first integrals mainly depends on the existence of solutions of partial differential equations, known as Killing equations (Hassani, 1999;Olver, 1986).
The above procedures, however, do not have generality of the Noether's theorem, as it mainly depends on the particular structure of the special class of problems being attempted.However, our choice to relate the alternative method of umbra Lagrangian mechanics is motivated by the fact that Noether's theorem, extended by Bahar et al. (1987) tackles both the aspects which are of considerable importance in the study of conservation laws.On phenomenological level, it shows the connection of conservation laws of some non-conservative system to the symmetries of space and time.On the other hand, it also possesses a pragmatic value as it could be used in engineering applications.
The significant work in this direction was reported by Bahar and Kwatny (1987), who provided a useful method based on a differential variational principle (Vujanovic, 1978) in order to extend Noether's theorem to constrained-nonconservative dynamical systems, which includes the influence of dissipation and constraints, and thus making it suitable for use in engineering applications.The main focus of their research work was primarily concerned with the extension of the notion of variation, which also included variation in time, thus leading to non-contemporaneous variation (NCV).Here the use of NCV is limited to first order terms, and was denoted by Δ as the convention adopted in Vujanovic (1978), whereas δ is the contemporaneous variation (CV).The symbol δ also defines a simultaneous or Lagrange's variation.A representative point A that is on the actual path at time t and an infinitesimal point B on the varied path at the same time t are correlated by
) (t q
where and are coordinates of points B and A respectively.The geometrical interpretation non-contemporaneous variation (NCV) may be easily seen in Fig. 1.The following definition was used The Non-contemporaneous variation of any function , , is given by the expression Putting Eq. ( 2) in ( 3) and following the procedure as given in reference (Bahar and Kwatny, 1987;Vujanovic, 1978), one may obtain ( ) ( ) Equation ( 4) demonstrates that the usual commutatively rule does not extend to NCV.For the derivation of the Noether's theorem, one may consider the variational expression as , which defines the integrand of the action integral governing the motion of such systems.This Lagrangian can be defined up to an additive term Then, one may get and still satisfies the Lagrangian equation of motion identically.L can be replaced by to have: Equation ( 13) is a conserved quantity.Now considering the linear infinitesimal one parameter transformation as followed in Bahar [29], one may obtain the conservation law Thus expression (5) has been defined by authors (Bahar and Kwatny, 1987) Expression ( 6) in non-contemporaneous may be written as Expression (7) will finally reduce to The last two terms of expression (8) may also be written in NCV as well.
(b) Contemporaneous Transformation (CT): Following the usual process to rewrite expression (6) as and following the procedure as in reference (Bahar and Kwatny, 1987;Vujanovic, 1978), one may obtain the following equation: The first integrals or conserved quantities may be obtained if right hand side of Eq. ( 10) can be made to vanish, then bracketed quantity under time derivative sign become a constant i i and Ρ i are the linear infinitesimal one-parameter transformations.This may also be obtained by following usual approach and generalized killing equations, which has been obtained in reference (Djukic and Vujanovic (1975).Many such examples of the Noether's theorem are contained in Vujanovic and Jones (1989).A variety of methods have been developed for the search of conservation laws such as methods of integrating factors, also termed as direct or ad hoc procedure as reported by Sarlet and Bahar (1980), and Djukic and Sutela (1984).Other methods were based on similarity variables (Jones and Ames, 1967) and transformation approach as presented by Crespo da Silva (1974).In this way, some procedures of group-theoretical approach with considerable generality have been established, which related the existence of first integrals to the symmetries of certain mathematical objects and served for describing the dynamical systems.
Several other studies concerned with the symmetry aspects of Lagrangian and Hamiltonian formalism have been considered in the review papers of Katzin andLevine (1976), andFokas (1979).A generalization of Noether Theorem in classical mechanics has been attempted by Sarlet and Cantrijn (1981).Another class of methods, in the spirit of finding invariants of motion for time-dependent parameters, are primarily established by few researchers such as Lewis and Leach (1982), who have reported an approach of finding exact invariants for one-dimensional time-dependent classical Hamiltonians, and as Sarlet (1983), who has developed a methodology of finding first integrals for one-dimensional particle motion in a non-linear, time-dependent potential field.Motivated by the research works of Vujanovic (1970), Vujanovic (1978) and Djukic and Vujanovic (1975), Simic (2002) has analyzed polynomial conservation laws of one-dimensional non-autonomous Lagrangian dynamical system and demonstrated that final form of dynamical system and corresponding conservation law depends on the solution of the so-called potential equation, which will be presented as variables.However, the structure of symmetry transformation, which generated particular class of conservation laws, could be prescribed independent of potential equation.In this Lagrangian function, generality of Noether's theorem is not being considered, which may be suitable to obtain invariants of any general class of systems.
Variational principles (Gelfend and Fomin, 1963) and principle of virtual work continued to attract interest of the researchers and have great importance in physics and mathematics.These principles helped in establishing connections and applications of these disciplines, and in devising diverse approximation techniques.Arizmendi et al. (2003) developed a variant of the usual Lagrangian, which describes both the equations of motion and the variational equations of the system.The required Lagrangian is defined in an extended configuration space comprising both the original configurations of the system and all virtual displacements joining any two integral curves.An extremal principal for obtaining the variational equations of a Lagrangian system is reviewed and formalized by Delgado et al. ( 2004) by relating the new Lagrangian function (Arizmendi et al., 2003) needed in such scheme to a prolongation (Hassani, 1999;Olver, 1986) of the original Lagrangian.In their work, they considered an N-degree of freedom dynamical system described by an autonomous non-singular Lagrangian function , a = 1.2…N defined in the tangent bundle TQ of its configuration Manifold Q.Now, an extended configuration space D (D'Alambert's configuration manifold) was considered, comprising of both the original configuration of the system plus all possible "virtual displacements" joining, in a first approximation, any two of the extremal paths of the original system.With the help of L, they defined new Lagrangian ) where and q are given configuration displacements and velocities, ε is virtual displacement and are virtual velocity.It is worth mentioning that even nonconservative systems can also be handled by using a prolonged Lagrangian function and Noether's theorem in this extended space, obtained by them.It is not appropriate to provide all details of this extension, as it basically finds its applications in relativistic theories.
Alternative Method for Extending Lagrangian-Hamiltonian Mechanics
As detailed in the previous section, the procedure and methods developed by various researchers did not consider the generality of Noether's theorem, as it was mainly focused on the particular structure of the special class of dynamical problems being studied.So, it is necessary to extend the scope of Lagrangian and Noether's theorem, which includes the influence of dissipation and sometimes constraints, thus making it suitable for the larger and complex engineering applications.To overcome the limitations and enlarging the scope of Lagrangian-Hamiltonian mechanics, a new proposal of additional time like variable 'umbra-time' was made by Mukherjee (1994) and this new concept of umbra-time leads to a peculiar form of equation, which is termed as umbra-Lagrange's equation.A brief and candid commentary on idea of umbra Lagrangian is given by Brown (2007).This idea was further consolidated by presenting an important issue of invariants of motion for the general class of system by extending Noether's theorem (Mukherjee, 2001).This notion of umbra-time is again used to propose a new concept of umbra-Hamiltonian, which is used along with the extended Noether's theorem to provide an insight into the dynamics of systems with symmetries.The advantages of using such Lagrangian are many ways as one may get the both aspects of the problem.It provides a great insight of the dynamical system through extended Noether's theorem and on the other hand, it gives a pragmatic value since it could be used as a reliable tool for derivation of new conservation laws for many engineering problems, where the physicist can play a leading role.One of the most important insights gained from the umbra-Lagrangian formalism is that its underlying variational principle (Rastogi, 2005) is possible, which is based on the recursive minimization of functionals.In this direction, Rastogi (2005) also defined all these notions in an extended manifold comprising of real time, and umbra and real time displacements and velocities.The umbra Lagrangian theory has been used successfully to study invariants of motion for non-conservative mechanical and thermo-mechanical systems [48].In another paper, the authors applied umbra Lagrangian to study dynamics of an electromechanical system comprising of an induction motor driving an elastic rotor (Mukherjee et al., 2009).This system was symmetric in two sets of coordinates, one set was mechanical or geometrical, and the other symmetry was in electrical domain.Recently, Mukherjee et al. (2009) presented the extension for Lagrangian-Hamiltonian Mechanics for continuous systems and investigated the dynamics of an internally damped rotor through dissipative coupling.Some basic concepts of umbra-Hamiltonian theory may be given in Appendix A for ready reference.The concept of umbra-Lagrangian may be represented as shown in Fig. 2 and briefly expressed as follows: (a) D'Alembert's basic idea of allowing displacements, when the real time is frozen, is conveniently expressed in terms of umbra-time.(b) Umbra-time may be viewed as the interior time of a system.(c) Potential, kinetic and co-kinetic energies stored in storage elements like symmetric compliant and inertial fields can be expressed as functions in umbra-time (umbra-displacements and umbra-velocities).(d) The effort of any external force, resistive element or field, gyroscopic element (treated as anti-symmetric resistive field), transformer or lever element, anti-symmetric compliant field and sensing element depends on displacements and velocities in real time.The potentials associated with them are obtained by evaluation of workdone through umbra-displacements.In formulating the umbra-Lagrangian for a system, two classes of elements are generally required: (a) storage elements, whose energies are defined in terms of umbra-displacements and umbravelocities, and (b) rest of the elements for which the efforts returned are evaluated entirely in terms of real time and their umbra-potential are obtained by umbra-displacement of the corresponding element.These two categories of elements can be identified through breaking the system into its basic entities or dynamical units.Bond graphs (Mukherjee and Karmakar, 2000;Karnopp et al., 1990) may be one of the tools for representing the dynamics of the system and obtaining the expressions for either the classical or the umbra-Lagrangian as provided in details in Appendix B.
The broad principle, on which the creation of umbra-Lagrangian and other relevant energies (Mukherjee, 2001;Rastogi, 2005;Mukherjee et al., 2009) are based, can be summarized as follows: (a) All temporal fluctuations of parameters are in real time.(b) The co-kinetic and potential energies would normally be evaluated taking generalized velocities and displacements as function of umbra time and it is assumed that there are n generalized co-ordinates.All definitions are separately given in Nomenclature.
The umbra-potential (for potential forces only) is defined as For example, umbra-potential for a damper with time varying damping coefficient may be written as where a bold face letter represents a vector quantity.As an example, the umbra-potential energy for a spring with time varying stiffness can be written as It is significant to note that in classical approach, one may incorporate dissipative forces through Rayleigh potentials, which in linear case can be defined as Likewise, the umbra-kinetic energy is defined as In such case, the anti-symmetric part of [R] in Eq. ( 24), if present, has no contribution to .In classical approach, such anti-symmetric part is identified as gyroscopic force and subjected to a set of alternative treatment.However, in present approach, as considered in Eq. ( 22), it can include both the dissipative (symmetric part) as well as the gyroscopic effects (anti-symmetric part) through the resistive field, for which the corresponding umbra-potential becomes and the umbra co-kinetic energy as (25) For instance, the umbra co-kinetic energy for a time varying mass can be represented as (d) The umbra-potential associated with external generalized forces may be incorporated as (26) (c) The umbra-potential associated with generalized resistive fields is evaluated, which is based on the philosophy that resistive fields open the system, and thus they observe the states of motion in real time as an external observer.The force generated by them does work on the system through umbra generalized displacements To illustrate, one may find the umbra-potential for any external force ( ) The total umbra-potential may be obtained by summing-up all the potentials represented by Eqs. ( 17), ( 22) and ( 26) and expressed as Using Eq. ( 30) along with Eqs. ( 33) and (34) in the previous equation ( 35), the extended Noether's theorem may be obtained and written as and the umbra-Lagrangian would, therefore, be . ... 1 m j = with New Lagrange's equations for a general class of systems may be given as In terms of the differential one-forms and the above relation may be expressed as Noether's theorem (Noether, 1918) states that, if the Lagrangian of a system is invariant under a family of single parameter groups, then each such group renders a constant of motion.The extended Noether's theorem, as discussed in paper (Mukherjee et al., 2009) may lead to a constant of motion, or trajectories, on which some dynamical quantity remains conserved.
The term on the left-hand side is the classical Noether term while the term on the right-hand side is additional and termed here as modulatory convection term.This modulatory convection term is made zero to obtain the conserved quantity.So, whenever the extended Lagrangian is found invariant, there is either the general conserved quantity or a trajectory on which such quantity remains conserved.The aforementioned methodology may be explained with two simple examples provided in the next subsections.
The umbra-Lagrangian may be defined on extended manifold, which consists of real displacements and velocities as well as umbra-displacements and velocities and real time (Mukherjee et al., 2009) Example 1. Simple mass-spring-damper system with time fluctuating parameters Here, the super dot ( ) denotes a derivative with respect to real time or umbra time, depending on the context.Unlike the classical formulation, this analysis requires single but extended manifold comprising of both umbra and real displacements and velocities and real time.The umbra-Lagrangian of a system admits several oneparameter transformation groups, and then the infinitesimal generator (Hassani, 1999;Olver, 1986) corresponding to j • Let us consider a simple mass spring damper system as shown in Fig. 3. Using the aforementioned procedure, the umbra-Lagrangian for this system may be expressed as th parameter (or group) may be decomposed as follows: The equation of motion obtained from the above umbra-Lagrangian through Eq. ( 30) may be written as The general forms for and would be 1 coupling of strength γ as shown in Fig. 4(a).The umbra-Lagrangian of the system may be written as Applying the second theorem of umbra-Hamiltonian and finding Again, the corollary of the second theorem of umbra-Hamiltonian gives and the symmetry (invariance) condition for umbra-Lagrangian may be expressed as , and on substituting the expression of yields Hence, is a constant of motion.Through Eq. ( 36), one obtains Both Examples illustrated in this paper provide an overview of the whole concept.It is apparent throughout the paper that the proposed extension of Lagrangian-Hamiltonian mechanics in terms of umbra philosophy gives a new dimension for analyzing the dynamical systems with non-conservative and non-potential forces.
The paper presented a brief review on the literature in Lagrangian-Hamiltonian Mechanics.Most of the research papers and books available in this field are incorporated, which undoubtedly enhanced and enriched the field of Mechanics.In this paper, authors have presented a brief review on extension of Lagrangian-Hamiltonian Mechanics.Various previous extensions on the subject matter were discussed with a particular regard to extension of Noether's theorem with nonconservative and nonholonomic systems for general class of systems.After review on literature on this subject matter, the following points are concluded: i) The procedure and methodology developed by other researchers don't have any generalization of Noether's theorem, as it has been mainly applied on the particular structure of the problems, which were rather mathematical without much physical interpretations of the real system.In this way, there is a substantial loss of generalization of the theorem, which may be applied to any engineering problems.However, in recent years, few researchers have applied the generalized Noether's theorem in few engineering applications.
where C is a constant of integration.The first term is the moment of momentum, and the second term is contributed by the gyroscopic coupling.
The umbra-Hamiltonian (discussed in Appendix A) of the system may be expressed as (ii) In contrast to all the previous extensions, the philosophy developed by the authors has addressed the issue of nonconservative and dissipative forces by assuming a new Lagrangian, which find wider applications for engineering problems.The authors have devised a new methodology to find invariants of motions of the dynamical systems.Gauge transformations [48], bi-symmetric rotor- motor system [49], dynamics of the rotor with internal damping [50] and few others are the applications already published in archival literature.
(iii) It is noteworthy to say that the alternative methods developed by the authors give more transparent physical interpretations, which enable the analyst to make further use of these first integrals in stability analysis.
(iv) In this article, the authors intended to provide critical evaluations of other extensions, which are rarely applied in the realworld problems.
Appendix A
Concept of umbra-Lagrangian and umbra-Hamiltonian Mukherjee (1994) introduced a concise and modified form of Lagrange's equation and manifested the use of this new scheme to arrive at system models in the presence of time fluctuating parameters, general dissipation and gyroscopic couplings, etc.In this scheme, real and virtual energies (or work) are separated by introduction of an additional time like parameter, which is termed as 'umbra-time'.The prefix 'umbra' was appended to all type of energies, and corresponding Lagrangian was termed as the "umbra Lagrangian".The basic idea presented in reference (Mukherjee, 1994;2001) leading to umbra-Lagrangian and umbra-Lagrange's equation may be briefly expressed as follows: (a) Umbra-time is the beholder of D'Alembert's basic idea of allowing displacements, when the real time is frozen.(b) Umbra-time may be viewed as the interior time of a system.(c) Potential, kinetic and co-kinetic energies stored in storage elements like symmetric compliant and inertial fields can be expressed as functions in umbra-time (umbra-displacements and umbra-velocities).(d) The effort of any external force, resistive element or field, gyroscopic element (treated as anti-symmetric resistive field), transformer or lever element, anti-symmetric compliant field and sensing element depends on displacements and velocities in real time.The potentials associated with them are obtained by evaluation of workdone through umbra-displacements.The broad principle on which the creation of umbra-Lagrangian and other relevant energies (Mukherjee, 1994;Mukherjee, 2001;Rastogi, 2005;Mukherjee et al., 2006;Mukherjee et al., 2007;Mukherjee et al., 2009) are summarized in the section IV.However, the umbra-Hamiltonian (Mukherjee, 2001) Now, it is easy to verify that umbra-Lagrangian of Eq. (B1) renders right equation of motion through Eq. (30) for the system.
Moreover, such research will open new horizons for the physics Nomenclature F(t = external force with time fluctuation * H = umbra-Hamiltonian of the system K = stiffness of the spring in N/m L = Lagrangian of the system * L = umbra-Lagrangian of the system R = damping coefficient of the damper in N-s/m V = infinitesimal generator of rotational SO (2) group j
as physical motivation and simply considered this equation to undergo two different transformations given as follows: (14) (a) Non-Contemporaneous Transformation (NCT): The variational expression given in Expression (5) may be written in the form where ξ , τ
and real displacement and real time.The fact that the given umbra-Lagrangian is invariant under the j
, one obtains If the umbra-Lagrangian admits the one-parameter rotational group, then the infinitesimal generators of the rotational SO (2
FigFigure 4 .
Fig. 4(a) Fig. 4(b) Figure 4. (a) Schematic diagram of two oscillators with gyroscopic coupling; (b) Bond graph model for system represented by (a) with artificial flow sources to obtain umbra-Lagrangian.
Figure 6
Figure 6(a).Causal bond graph model of system 5(a) with artificial flow sources.
may be represented as | 2019-03-04T06:51:01.587Z | 2011-03-01T00:00:00.000 | {
"year": 2011,
"sha1": "9f7adb133ac680148cfca259e85f580c5f4b43ef",
"oa_license": "CCBYNC",
"oa_url": "http://www.scielo.br/pdf/jbsmse/v33n1/04.pdf",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "9f7adb133ac680148cfca259e85f580c5f4b43ef",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
3076987 | pes2o/s2orc | v3-fos-license | Basaloid Squamous Cell Carcinoma
Squamous cell carcinoma is the second most common cancer of the skin. This tumor arises predominantly in sun exposed actinically damaged areas. Implicated as predisposing factors, in addition to sunlight, these are industrial carcinogens, chronic ulcers, and ionizing radiation [1]. They are common cancer in immunocompromised and renal transplant patients [2]. Squamous cell carcinomas are known to be the most prevalent malignant tumor of the head and neck region [3]. They are also reported in many organs including cervix, lung, bladder, uterus, ovary, esophagus and teratomas [4;5;6,7].
Introduction
Squamous cell carcinoma is the second most common cancer of the skin. This tumor arises predominantly in sun exposed actinically damaged areas. Implicated as predisposing factors, in addition to sunlight, these are industrial carcinogens, chronic ulcers, and ionizing radiation [1]. They are common cancer in immunocompromised and renal transplant patients [2]. Squamous cell carcinomas are known to be the most prevalent malignant tumor of the head and neck region [3]. They are also reported in many organs including cervix, lung, bladder, uterus, ovary, esophagus and teratomas [4;5;6,7].
Squamous cell carcinoma is characterized by squamous cells with large nuclei and abundant eosinophilic cytoplasm. The cells exhibit prominent intracellular bridges and variable keratin formation, depending on the degree of differentiation. Poorly differentiated tumors lack keratinization and usually form solid sheets of cells with marked pleomorphism to the extent that require special studies to establish the nature of the tumor.
Histologic variants of squamous cell carcinoma
Several histologic variant of squamous cell carcinoma are identified. These variants are based on certain morphological features accordingly, which may or may not have prognostic implications. The following are the most reported variant in the literature and include basaloid, warty verrucous, papillary, spindle cell, adenosquamous, clear cell, acantholytic and lymphoepithelioma-like type.
Spindle cell carcinoma is rather rare and is composed of atypical spindle cells with whorled arrangement (Fig 1), which mostly come from immunosuppressed renal transplant patients. The tumor needs to be differentiated from desmoplastic melanoma, atypical fibroxanthoma or metastatic carcinoma with spindle cell features. Immunohistochemistry is of value in differentiating these entities [8].
Clear squamous cell carcinoma is another variant first described as squamous cell carcinoma with extensive hydropic changes. The cells appear glassy looking, due to accumulation of fluid, and can be easily mistaken for sebaceous cell carcinoma. The differential also includes other clear cell tumors such as clear cell acanthoma, clear cell hidradenoma, metastatic renal cell carcinoma, balloon cell nevus and melanoma [9]. Verrucous squamous cell carcinoma presents with rather non dysplastic epithelium with hyperkeratosis and elongation of the rete pegs. This is in contrast to the papillary variant of squamous cell carcinoma, which is characterized by malignant looking epithelium with papillary or exophytic architecture [10]. Adenosquamous cell carcinoma is very rare subtype, which is composed of admixed adenocarcinoma with squamous cell carcinoma. Mucin stain usually highlights the adenocarcinoma component.
Basaloid basal cell carcinoma is a rare variant of squamous cell carcinoma with more cases, which have been published since its first description in 1986 by Wain et al (11). These tumors affect both sexes but with predominance of male patients. They are frequently seen in the aerodigestive tract with most of the cases to be found in the tongue, floor of the mouth, the pyriform sinus, tonsil, and larynx [12]. These tumors have also been described in a variety of sites including nasopharynx, trachea, skin, cervix, bladder, thymus, anus, conjunctiva, and lung [13,14,15,16]. Clinically, patients have similar presentation to conventional squamous cell carcinoma depending on the site of the lesion.
Etiology and pathogenesis of basaloid cell carcinoma is similar to conventional squamous carcinoma. Most patients have a long history of smoking and alcohol drinking. In some cases there was a history of previous radiation to the head and neck region [17]. Both represent independent risk factors for the development of squamous cell carcinoma.
Smokeless tobacco and other exogenous carcinogens such as occupational, environmental and nutritional factors may also play role in the pathogenesis of this cancer. EBV was detected in few cases using in situ hybridization technique from nasopharyngeal sites.
Recent studies detected a higher frequency of HPV and HSV in basaloid tumors than in conventional squamous cell carcinomas of the head and neck [18]. Basaloid squamous cell carcinoma in non smoker young patients revealed infection with HPV, high risk genotype 16. The expression is so significant, to the extent that it led some authors to consider the expression may be important for the diagnosis of this type of squamous cell carcinoma. The prognosis of HPV induced carcinoma appeared to have better outcome than the HPV negative cases [19]. It is not practical to perform in situ hybridization and sequencing techniques on every single case of basaloid squamous carcinoma as this is technically demanding and can be performed mostly in special centers [20]. The cell of origin of these tumors has been suggested to be a multipotential cell, which is able to differentiate into multiple cell type. However, the most acceptable origin for these cells is that they are from the surface epithelium since there is dysplastic or carcinoma in situ changes with direct continuity within the invasive component.
The tumor is considered by many authors as high grade with more aggressive behavior [11.16]. These lesions are capable of distant metastases, deep invasion, local recurrence and lymph node involvement. The most common sites for distant metastasis are the lung and liver. Multifocal disease includes other sites in the head and neck, which were also documented [17] However, some controversy is still present regarding the prognosis and conflicting results, which have appeared in recent literature. Some published papers claimed they have similar prognosis to traditional squamous cell carcinoma [21]. The majority of these cases are found at an advanced stage, which could explain the poor clinical outcome and prognosis. No general guidelines are present regarding the management of this disease; however, most published reports recommend a combination of surgery and postoperative radiotherapy, in order to prevent local recurrence and distant metastases [12,14].
Pathology of basal squamous cell carcinoma
Macroscopic appearances of these tumors show flat or slightly raised or polypoid exophytic lesions with or without a central ulceration in most cases reported in the literature [3] Microscopic examination of these lesions show characteristic invasive growth appearance, shared by most lesions. Generally they are composed of ribbons and or cords of basaloid cells with peripheral palisading and closely resemble traditional basal cell carcinoma figure 2. This lesion comes from rare urinary bladder flat lesion from a 66-year-old man seen on cystoscopy. In addition, the cellular arrangements of these lesions can closely mimic adenoid cystic carcinoma, due to the glandular or cribriform pattern, and have a tendency to have intracellular deposition of eosinophilic hyaline material figure 3. One of the major features of this tumor is that the cells exhibit high nucleocytoplasmic ratio and often have dense hyperchromatic nuclei and comedonecrosis may be seen in these tumor, figure 4, which was seen in a patient who presented with nasal sinus mass. This appearance represents common features of these lesions. Mitotic figures may be high and may include atypical forms. Careful search will reveal focal squamous differentiation with intercellular bridges or keratin formation, which is important for the accurate pathological assessment of these tumors. Another important feature of these lesions is dysplasia of the surface epithelium in cutanous neoplasm. Sometimes the tumor show true neural type rosette formation and other tumors may exhibit spindle shaped pleomorphic cells with elongated nuclei. Vascular or lymphatic invasion may als o b e p r e s e n t . I n r e c e n t p u b l i c a t i o n o f cutanous basaloid squamous cell carcinoma, this tumor may also have rather large pleomorphic cells with big nuclei widely scattered throughout the lesion. These pleomorphic cells present no significant biological behavior. The immunoprofile of these tumors show consistent positive staining to high molecular weight cytokeratin antibody 34βE12, KL1, and MNF116, and focal staining for vimentin, EMA, CAM5.2, CK7, CEA, S100 and GFAP, and negative immunostaining for CK20, chromogranin, synaptophysin, BCL2, and Ber-EP4. Actin staining was positive in the basaloid cells and some cases were positive for CD99. More recent studies confirm strong and diffuse staining for P63 immunomarker in this tumor figure 5. Lastly, electron microscopic examination of samples may show tonofilaments and desmosome and do not demonstrate any characteristic findings, as the malignant cells show mainly undifferentiated cellular features, and the organelles are rather poorly developed (22). The differential diagnosis of these tumors includes adenoid cystic carcinoma, small cell neuroendocrine carcinoma and other carcinomas depending on the anatomical sites.
Adenoid cystic carcinoma is characterized by basaloid-looking cells with predominant myoepithelial cells forming cribriform, solid or tubular structure. The tumor is slowly growing, less aggressive and with infrequent lymph node metastasis. Perineural invasion is a common feature of this lesion [23]. Immunohistochemistry show positive staining of the myoepthelial cells for S100, actin and calponin. The epithelial ductal cells of the tumor stain for cytokeratin, CEA and EMA. The stromal hyaline material can be highlighted with collagen IV and laminin. Small cell carcinoma is a more aggressive lesion with a different treatment approach. The tumor cells are positive for chromogranin, synaptophysin and dotlike staining for cytokeratin. These tumors are negative for 34βE12 marker, which is normally present in basaloid squamous cell carcinoma [17]. Skin basal cell carcinoma share histologic features with basaloid squamous cell carcinoma and need to be differentiate; however it lacks surface epithelial dysplasia, pleomorphism and the comedonecrosis seen in basaloid squamous cell carcinoma [17]. Adenosquamous carcinoma, which comprised of squalors and glandular differentiation, can have surface epithelial dysplasia. These lesions contain mucin and lack basaloid cells and peripheral palisading [24].
In conclusion, this variant of squamous cell carcinoma is reported in many sites and organs and present unique pathological and clinical features. This neoplasm is currently under more investigation to determine the nature and clinical behavior. The pathology of this entity is characterized by closely packed basaloid-looking cells with scanty cytoplasm. The cells are arranged in ribbons or with trabecular pattern. Occasional foci of squamouslooking cells are identified. The immunoprofile of these tumors are helpful to distinguish them from basal cell carcinoma, adenoid cystic carcinoma and small cell neuroendocrine carcinoma. The tumor cells are positive for epithelial marker 34B E12, EMA and P63.
Management of basaloid squamous cell carcinoma which is considered by many authors as more aggressive tumor requires radical excision followed by locoregional radiation and chemotherapy. For advance cases combination of radiotherapy and chemotherapy is a logical approach to control the disease. The overall disease free survival rate statically is slightly lower than in the conventional type of squamous cell carcinoma. Meanwhile some studies concluded that the prognosis is comparable to conventional type but the number of cases is to small to draw a definite conclusion. Metastatic disease is recorded in many patients with basaloid squamous cell carcinoma and it is advisable to perform metastatic work up. In conclusion the disease appears in most reported cases in the literature as more aggressive and capable of distant metastasis. [17,25]. This book points to some new areas for investigation on squamous cell carcinoma (SCC). Firstly, the features and management of some specific SCC is discussed to give the readers the general principles in dealing with these uncommon and sophisticated conditions. Some new concepts in adjuvant therapy including neoadjuvant therapy and gold nanoparticle-based photo dynamic therapy are introduced. Secondly, a detailed discussion of molecular aspects of tumor invasion and progression in SCC is provided with the emphasis on the roles of some important factors. The role of tumor microenvironment in head and neck SCC is specifically discussed. Thirdly, the roles of cancer stem cells (CSC) in cancer therapy of SCC are described. Molecular mechanisms involving therapeutic resistance and new therapeutic strategies targeting CSC are discussed in detail. Finally, other aspects concerning SCC are included, which involve the assessment, genetic manipulation and its possible clinical implications for the treatment of SCC. | 2017-04-01T16:56:20.999Z | 2012-02-03T00:00:00.000 | {
"year": 2020,
"sha1": "7bc72a2a6184db3ac7875b20835504b8a1c231fc",
"oa_license": "CCBY",
"oa_url": "https://www.qeios.com/read/ZWKKSZ/pdf",
"oa_status": "HYBRID",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "958ce11e499eedc1a9aa698e4086450e778e3085",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
99685013 | pes2o/s2orc | v3-fos-license | Photocatalytic Studies if Tio2/Sio2 Nanocomposite Xerogels
TiO2 is well recognized as a valuable material with application as a white pigment in paints, as filler in paper, textile and in rubber/plastics [1]. Due to low cost, non-toxicity, stability and other best characteristics TiO2 attracts a great attention. TiO2 has wide applications in various fields like antireflection opticals, coatings, waste water purifications, catalyst supporting, ceramics senser element, as a photocatalyst, in electric devices like (in lithium based battery), as a base in high quality paints, paper, plastics [2]. Titania has excellent biocompatibility with respect to bones implants and applications in electrochromic devices [3]. Titania shows good photocatalytic applications due to which it gained tremendous demands and green energy and environmental protection. Many other oxides like iron oxides, zinc oxides etc. also shows the similar behavior due to photocatalytic activity of titanium dioxide it play a wide role in different fields like air, waste water purification, good UV blocking properties weakening of the organic fibers [4]. Silica doped in to the titania matrix increase the photocatalytic activity because the silica doping decrease particle size and also increase the specific surface area and thermal stability of titania particle towards anatase to rutile phase conversion [5]. SiO2TiO2 materials are used in different fields like as catalyst supporting materials, acidic catalyst for many reactions, selective reduction, as an anti-reflective materials for coatings or sensing nanoimprints photonic crystals [6-9]. Dielectric mirrors and low loss waveguides solids of low thermal expansion coefficient, bioactive solids self-cleaning coatings solids of controlled acidity and photocatalysts [10-14]. The TiO2-SiO2 mixed oxides catalytic activity was studied and observed that the TiO2SiO2 have better photocatalytic activity as compared to TiO2 and SiO2 which was confirmed through LLS and UV results.
Introduction
TiO 2 is well recognized as a valuable material with application as a white pigment in paints, as filler in paper, textile and in rubber/plastics [1]. Due to low cost, non-toxicity, stability and other best characteristics TiO 2 attracts a great attention. TiO 2 has wide applications in various fields like antireflection opticals, coatings, waste water purifications, catalyst supporting, ceramics senser element, as a photocatalyst, in electric devices like (in lithium based battery), as a base in high quality paints, paper, plastics [2]. Titania has excellent biocompatibility with respect to bones implants and applications in electrochromic devices [3]. Titania shows good photocatalytic applications due to which it gained tremendous demands and green energy and environmental protection. Many other oxides like iron oxides, zinc oxides etc. also shows the similar behavior due to photocatalytic activity of titanium dioxide it play a wide role in different fields like air, waste water purification, good UV blocking properties weakening of the organic fibers [4]. Silica doped in to the titania matrix increase the photocatalytic activity because the silica doping decrease particle size and also increase the specific surface area and thermal stability of titania particle towards anatase to rutile phase conversion [5]. SiO 2 -TiO 2 materials are used in different fields like as catalyst supporting materials, acidic catalyst for many reactions, selective reduction, as an anti-reflective materials for coatings or sensing nanoimprints photonic crystals [6][7][8][9]. Dielectric mirrors and low loss waveguides solids of low thermal expansion coefficient, bioactive solids self-cleaning coatings solids of controlled acidity and photocatalysts [10][11][12][13][14]. The TiO 2 -SiO 2 mixed oxides catalytic activity was studied and observed that the TiO 2 -SiO 2 have better photocatalytic activity as compared to TiO 2 and SiO 2 which was confirmed through LLS and UV results.
Sample preparation
The xerogels can be synthesized by Sol-Gel process in which metal alkoxide is used as a precursor source that undergoes catalyzed hydrolysis and condensation to get nano scale materials of that metal [15]. TTIP was used as a precursor and TEOS was added as an organic solvent. In this synthesis HCl and NaOH were used as catalysts.
TiO 2-SiO 2 mixed oxide
TiO 2 -SiO 2 xerogels was synthesized by Sol-Gel process. TEOS (Tetraethylorthosilicate) and TTIP (Titanium tetra isopropoxide) were used as precursor. It was observed that the TTIP hydrolysis rate is much faster than TEOS [16]. The synthesis consists of two steps. The first step is the synthesis of SiO 2 sol, in which 1:6 TEOS and Methanol were mixed together i.e. 7 ml of TEOS and 43 ml of methanol. TEOS were added drop wise slowly to the methanol with continuous stirring. Then 0.05 mol L -1 HCl was added drop wise to the sol in order to adjust the acidic pH at 2. Then the sol was allowed to stirrer for about 2 an hour in order to get the homogeneous sol. In the second step TiO 2 Sol synthesis, 1:14 solution of TTIP and DIW were mixed with continuous stirring until the homogeneous sol of TiO 2 were obtained. Then in 1:1 of TiO 2 and SiO 2 sols were mixed with continuous stirring in order to get homogeneous sol and then 0.05 mol L -1 NH 4 OH solution was added to this mixture drop wise for the adjustment of pH. The pH of the homogeneous mixtures (sol) was observed by the pH-meter continuously. At last the mixtures (sol) were allowed to stirrer for some time to get the homogeneous mixture (sol).
Gel and xerogel preparation
In this method 3ml of TEOS, the silica precursor was taken in the reaction beaker and 6ml of ethanol was added dropwise with continue stirring. Further 3 ml of water, 4 ml of acetic acid and 3 ml of TTIP (Titania precursor) were added with continuous stirring for 15 min at room temperature. The prepared sol changed into a gel which was placed in the oven at 65 for 1 h which resulted in the conversion of gel in to xerogel.
FTIR result
The FTIR spectra of the synthesized nanocomposite (TiO 2 /SiO 2 ) were recorded by Perkin Elmer series 100 FTIR spectrometer with a 5 cm -1 resolution. FTIR spectrum was recorded at 4000-450 cm -1 . The absorption at 1074 cm -1 (Figure 1) is the characteristics peak for Si-O-Si. The peak observed at 801 cm -1 is due to Si-O-Si symmetric stretching. The broad absorption at 1633 cm -1 match with OH bending vibrations and is attributed to chemisorbed water. The well defined peak at 3441 cm -1 shows OH stretching vibrations. The peak observed at 923 cm -1 corresponded Si-O-Ti vibrations. The band observed at 450-610 cm -1 is due to Ti-O Stretching.
Many characteristic FTIR peaks were observed in Figure 2. The bands observed at 3391 cm -1 and 1557cm -1 were due to the OH bending and stretching vibrations respectively. The peak observed at 1165 cm -1 correspond to Si-O-Si antisymmetric stretching vibration. The peak obtained at 923 cm -1 correspond to Si-O-Ti vibration. The peak observed at 801 cm -1 is due to the Si-O-Si symmetric stretching vibration. The absorption at 1410 cm -1 is due to C-H interaction of Si-R structure unit (Figure 3).
XRD measurement
The XRD pattern of synthesized nanocomposites (TiO 2 /SiO 2 ) was collected in the range of 10-60 2θ (degree) shown in Figure 4. The XRD patterns of the synthesized materials indicate that the TiO 2 -SiO 2 nanocomposites are essentially non-crystalline and have amorphous structure which can be accomplished from the broad characteristic diffraction peak between 2θ ~ 20θ° and 30 θ°.
The Figures 5 and 6 also show the same XRD results of the synthesized TiO 2 /SiO 2 nanocomposites. These samples were also synthesized by sol-gel method with different concentrations of precursors and solvents used. The hump in XRD pattern indicate that the TiO 2 -SiO 2 nanocomposites are essentially non-crystalline and have amorphous structure which can be accomplished from the broad characteristic diffraction peak between 2θ ~ 20θ° and 30 θ° while the Figure 2 shows very low crystallinity at 2θ and 25 θ°.
Catalytic activity
To check the catalytic activity of TiO 2 -SiO 2 the p-nitrophenol was reduced by NaBH 4 to p-aminophenol in aqueous medium. The procedure adopted was used as a given amount of mixed oxides was added to 1 ml of p-nitrophenol (0.08 mmol L -1 ) for initiation of reduction. For mixture preparation 1 ml of aqueous solution of NaBH 4 (1.5 mmol L -1 ) was added to reaction chamber. The p-nitrophenol reduction was determined by studying the absorbance at 440 nm with respect to time. This reaction was selected due to simplicity and formation of single product. By using time dependence uv-vis spectra the reduction process was checked. The appearance of a new peak at 300 nm was observed which confirm the p-nitrophenol reduction to p-aminophenol which is shown in Figure 6.
LLS result (hydrodynamic radii)
The LLS results (hydrodynamic radii) of the three samples are given in Table 1 (hydrodynamic radius) of the sample TSX 3 were measured during the conversion of sol in to gel and the LLS results (hydrodynamic radii) of the TSX 2 and TSX 17 were calculated after the gel was formed.
SEM results
The Figure 7 shows the SEM image of synthesized containing TiO 2 -SiO 2. From SEM image, it is clear that in most part of the sample large particles is obtained, which gives the evidence of the agglorimation of small particles. Along large agglomarited particle small size individual particles can also be absorved.
EDX results
The EDX spectrum is showing the elemental composition of the three elements in different percentage which is silicon (Si) and titanium (Ti) and oxygen (Figure 8). The EDX spectra has confirmed all the constituents that were added in the synthesized sample.
Conclusion
(TiO 2 /SiO 2 ) nanocomposites were synthesized by sol-gel process where Tetraethylorthosilicate (TEOS) and Titanium tetraisopropoxide (TTIP) were used as precursors in the presence of organic solvent. The XRD results confirmed that the synthesized TiO 2 -SiO 2 mixed oxide xerogel is non-crystalline and having amorphous nature. In the TiO 2 / SiO 2 nanocomposites, materials, in the amorphous SiO 2 matrix TiO 2 nanocrystals are present in highly dispersed form. The amorphous SiO 2 and Ti-O-Si bond formation and TiO 2 -SiO 2 mixed oxides give rise effectively stability of TiO 2 anatase form. It also bound crystallites growth, significantly increase surface area. So, the increase in surface area caused the improvement in photocatalytic activity of TiO 2 enhance surface properties, catalytic and photoactivity. The TiO 2 -SiO 2 mixed oxides catalytic activity was also studied and observed that the TiO 2 -SiO 2 have better photocatalytic activity as compared to TiO 2 and SiO 2 which was confirmed through LLS and UV results. | 2019-04-08T13:12:03.562Z | 2017-02-25T00:00:00.000 | {
"year": 2017,
"sha1": "aba3e4ef43e5dd3397b442570d46c9e4b315944b",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.4172/2155-9872.1000348",
"oa_status": "HYBRID",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "531b542180f59ea34ad90c5accd0d099cec8e25b",
"s2fieldsofstudy": [
"Chemistry",
"Materials Science"
],
"extfieldsofstudy": [
"Chemistry"
]
} |
119241686 | pes2o/s2orc | v3-fos-license | Characterizing Spinning Black Hole Binaries in Eccentric Orbits with LISA
The Laser Interferometer Space Antenna (LISA) is designed to detect gravitational wave signals from astrophysical sources, including those from coalescing binary systems of compact objects such as black holes. Colliding galaxies have central black holes that sink to the center of the merged galaxy and begin to orbit one another and emit gravitational waves. Some galaxy evolution models predict that the binary black hole system will enter the LISA band with significant orbital eccentricity, while other models suggest that the orbits will already have circularized. Using a full seventeen parameter waveform model that includes the effects of orbital eccentricity, spin precession and higher harmonics, we investigate how well the source parameters can be inferred from simulated LISA data. Defining the reference eccentricity as the value one year before merger, we find that for typical LISA sources, it will be possible to measure the eccentricity to an accuracy of parts in a thousand. The accuracy with which the eccentricity can be measured depends only very weakly on the eccentricity, making it possible to distinguish circular orbits from those with very small eccentricities. LISA measurements of the orbital eccentricity can help constraints theories of galaxy mergers in the early universe. Failing to account for the eccentricity in the waveform modeling can lead to a loss of signal power and bias the estimation of parameters such as the black hole masses and spins.
The Laser Interferometer Space Antenna (LISA) is designed to detect gravitational wave signals from astrophysical sources, including those from coalescing binary systems of compact objects such as black holes. Colliding galaxies have central black holes that sink to the center of the merged galaxy and begin to orbit one another and emit gravitational waves. Some galaxy evolution models predict that the binary black hole system will enter the LISA band with significant orbital eccentricity, while other models suggest that the orbits will already have circularized. Using a full seventeen parameter waveform model that includes the effects of orbital eccentricity, spin precession and higher harmonics, we investigate how well the source parameters can be inferred from simulated LISA data. Defining the reference eccentricity as the value one year before merger, we find that for typical LISA sources, it will be possible to measure the eccentricity to an accuracy of parts in a thousand. The accuracy with which the eccentricity can be measured depends only very weakly on the eccentricity, making it possible to distinguish circular orbits from those with very small eccentricities. LISA measurements of the orbital eccentricity can help constraints theories of galaxy mergers in the early universe. Failing to account for the eccentricity in the waveform modeling can lead to a loss of signal power and bias the estimation of parameters such as the black hole masses and spins.
I. INTRODUCTION
Binary systems of compact objects will be ubiquitous sources for the Laser Interferometer Space Antenna (LISA) [1,2]. Observations have shown that today there are massive black holes in the center of nearly all galaxies [3][4][5]. When galaxies collide their central black holes sink to the center of the merged galaxy and begin to orbit one another, losing energy and angular momentum in the form of gravitational waves [6]. Gravitational wave (GW) emission rapidly erases any initial eccentricity [7,8], so it has long been thought that eccentricity could be ignored when modeling the signals from massive black hole binaries [9]. More recently, however, it has been shown that the mechanisms that may harden the binary to the point where the gravitational wave emission takes over all tend to drive up the eccentricity [10][11][12][13][14][15][16][17][18][19]. The question then is whether significant eccentricity survives until the final year or so before merger. Figure 1 shows the eccentricity evolution as a function of orbital frequency for systems that enter the gravitational wave dominated evolution stage at frequency f gw with eccentricities of e gw = 0.95 and e gw = 0.5. The tracks are computed using the leading-order Peters and Matthews [8] evolution equations. These equations predict that once the eccentricity drops below e ∼ 0.3, it decays as e ∼ f −19/18 -in other words, roughly a factor of ten in eccentricity is lost per decade of frequency. The rate of decay is slower for systems with very high eccentricities, allowing them to maintain significant eccentricity for longer. For typical LISA black hole binaries, the GW decay time drops below the Hubble time when the systems are 3 to 5 decades in frequency from the LISA band, so unless e gw is very close to unity, a purely GW driven orbital evolution results in nearly circular orbits in the LISA band. On the other hand, if the hardening mechanism (e.g. gas dy-namics or stellar scattering) continues to dominate the dynamics until the system is close to the LISA band, then interesting eccentricities can be maintained. In a recent study [20], Sesana has shown that stellar scattering produces LISA sources with eccentricities in the range e 0 = 10 −3 → 0.2. Moreover, as shown in Figure 8 of Ref. [20], the distribution of eccentricities depends on the component masses in a particular way, so measuring this distribution can help constrain black hole merger models. While the distribution of component masses and spins as a function of luminosity distance will likely play a more important role in studying galaxy -black hole co-evolution [21][22][23], the eccentricity distribution may provide useful additional constraints. These considerations suggest that it is desirable to include the effects of eccentricity in the gravitational waveform, bringing the total number of parameters needed to describe a black hole inspiral to seventeen: the redshifted black hole masses (m 1 , m 2 ), the distance to the source (D L ), the initial radial eccentricity and semi-major axis (e 0 , a 0 ), the dimensionless spin parameters for the two black holes (χ 1 , χ 2 ), the source sky location (cos θ, φ), the initial orientation of the angular momentum and spin vectors (cos θ L , φ L ) , (cos θ S1 , φ S1 ) , (cos θ S2 , φ S2 ), and two initial phase parameters (n 0 , φ 0 ). Failing to include the eccentricity in the waveform model will lead to a loss of signalto-noise [24,25] and to biases in the recovery of the other parameters [26,27]. The instantaneous gravitational waveforms describing the inspiral of a black hole binary with arbitrary spins, masses, and orbital eccentricity were calculated by Kidder [28] to second post-Newtonian order (2 PN, or order v 4 /c 4 in the relative velocity), and extended to 2.5 PN order by Faye, Blanchet and Buonanno [29,30]. Majár and Vasúth [31] later introduced a convenient framework for expressing the waveforms in a form amenable to producing detection templates. Together with a solution to the conservative equations of motion, adiabatically advanced with the angular momentum and energy dissipation equations [28], we can build the time dependent gravitational waveforms for a general binary black hole system with the full seventeen parameters necessary to describe the system. LISA observations of binary black hole inspirals can be compared to these waveforms to produce posterior distributions for the model parameters. These observations will allow us to constrain galaxy merger scenarios [21][22][23], and allow us to test General Relativity in the dynamical strong field regime [32]. Left unaccounted for, the effects of orbital eccentricity may be mistaken for a departure from General Relativity (in particular, even small eccentricities that produce negligible power in higher harmonics can lead to easily detectable changes in the phase evolution of the signal).
Black hole binary systems in eccentric orbits may also be detected by ground based gravitational wave detectors such as the Laser Interferometer Gravitational wave Observatory (LIGO) and Virgo [33]. Some models even predict that inspiral signals may enter the LIGO band with e > 0.9 and that eccentric templates will be necessary to detect such sources [34]. Current LIGO data analysis uses circular templates and may need to be generalized to include eccentricity.
We have previously described a method for combining the instantaneous gravitational waveforms for eccentric binary systems with a post-Keplerian solution to the equations of motion that is adiabatically advanced using the the orbit-averaged dissipation and spin precession equations to build ready-to-use gravitational waveforms for the general case of a spinning black hole binary system in an eccentric orbit [35]. Here we present the results of a parameter estimation study for spinning eccentric binary black hole sources for the LISA mission. This is an extension of the Lang and Hughes LISA parameter estimation study of spinning binary black holes in circular orbits [36] and is the first to include the full seventeen parameters that describe a general black hole binary inspiral.
II. WAVEFORM MODEL
The equations needed to numerically calculate time dependent gravitational waveforms for binary black hole systems have been computed to 2.5PN order in the amplitude and phase [28][29][30][31]. The resulting system of equations could be numerically evolved to produce waveforms for our study, but the computational cost of resolving the motion of the black holes on the orbital timescale is too large for the parameter estimation study we wish to perform. Taking advantage of the separation of time scales in the waveform model we have developed an efficient waveform generator at 1.5 PN order in the amplitude and phase [35]. These waveforms include the effects of periastron precession, the precession of the orbital plane due to spin-orbit coupling, and higher harmonics from the higher order mass and current multipole moments. Our waveform model does not include the effects of spinspin coupling which enter at 2PN order. In future work we plan to extend the waveform model and our parameter estimation study to higher Post-Newtonian order.
The fastest time scale for the system is the orbital time scale, which to leading order is given by Kepler's law: where M = m 1 + m 2 is the total mass of the system and a is the semi-major axis. Periastron advance enters at 1 PN order, with a fractional advance per orbit of k = 3M 2 µ 2 /L 2 , where µ = m 1 m 2 /M is the reduced mass of the system, L ≃ µ aM (1 − e 2 ) is the orbital angular momentum and e is the eccentricity of the orbit. The ratio of the periastron precession timescale to the orbital timescale is given by The orbital angular momentum L precesses with angular velocity ω prec = S eff /r 3 [35], where The effective spin vector has magnitude S eff ∼ M 2 since the magnitude of the individual spins is given by S i = χ i m 2 i where the dimensionless spin parameter 0 ≤ χ i ≤ 1. The ratio of the precession time scale and the orbital timescale is given by We see that the precession of the orbital plane enters at 1.5 PN order relative to the orbital time scale. The precession of the orbital angular momentum due to spinorbit coupling results in a modulation of the amplitude of the gravitational waveform at the solar system barycenter.
The time dependent gravitational waveform h+(t) at the solar system barycenter for a binary black hole system with e0 = 0.3 (Source 1 in Table II) during the final year before merger.
The rate at which the binary black hole system loses energy and angular momentum due to gravitational wave emission defines the decay time scale and the ratio The loss of energy and angular momentum results in the decay of the semi-major axis and the radial eccentricity of the system. Orbit averaged expression for these decay rates can be found in Ref. [35]. The effects of spin-orbit induced precession of the orbital plane and the overall sweep up in frequency and amplitude as the system inspirals over the course of a year are apparent in Figure 2 for a system with e 0 = 0.3 and other source parameters given in Table II for Source 1. At times well before the merger of the system M/a is small and we find that We take advantage of this separation of the relevant time scales to make our waveform calculations more efficient. Since the decay and precession timescales are much longer than the orbital timescale we can start with a solution to the orbital equations of motion that neglects dissipation. This allows us to use Nyquist sampling of just a couple of samples per orbit for the quantities that depend on the dissipation and precession equations and saves considerable computational cost. Considering only times early in the evolution of the system we have a clean separation of time scales and the different processes can be treated differently in the calculations. This simplifies the problem by allowing an adiabatic treatment where the dissipation is assumed to be small over the course of an orbit and the eccentricity and semi-major axis are treated as constant while calculating an individual orbit. The approximations involved in exploiting the separation of timescales introduce small errors in the waveforms. The largest of these comes from using the orbitaveraged spin precession equations, which neglects small periodic changes in the spin orientations that occur on the orbital timescale, but these changes are effectively 2.5 PN order contributions to the higher harmonics. The averaging also introduces small errors in the long term secular evolution that scale as the ratio of the averaging time scale (in our case the orbital timescale) and the timescale of the terms being averaged (such as the spin precession timescale). These errors are multiplicative, and so represent higher PN order terms that can be discarded at 1.5 PN order.
As the system approaches merger the various time scales become comparable and our waveform model breaks down. We adopt the termination condition 2πM f orb = 0.01, which corresponds to an expansion parameter M/a ≈ 0.05. Denoting the time when this condition is met as t s , and the orbital frequency at this time as f s , we taper the waveform smoothly to zero by multiplying the waveform with a half-Hann filter: Our termination condition is conservative in terms of the signal to noise ratio (SNR) LISA will be able to extract from this type of source. Most of the SNR comes from times near merger, so the extension of the validity of the waveform closer to merger results in a big increase in SNR. Our study here is thus a pessimistic estimate of how well LISA will be able to determine the various source parameters. In the case of the radial eccentricity parameter however, the circularization of the waveform toward merger means that most of the eccentricity information is encoded at times well before merger. While increased SNR would improve the determination of the other source parameters, we find that the eccentricity is not highly correlated with the other parameters (see Figure 13). Our choice for when to truncate the waveform thus should not have much of an effect on our study of how well LISA will be able to determine the eccentricity of black hole binary systems.
We have tested our waveform generator in various limits against other codes. In the circular limit, and with dissipation turned off, we found precise agreement with the 1.5 PN limit of the spinning black hole code developed by Cornish, Hughes, Lang and Nissanke that is described in Ref. [37]. We do not expect, and nor do we find, precise agreement when dissipation is included. This is because our eccentric waveform generator evolves the semi-major axis, while the circular orbit code evolves the orbital frequency, which leads to numerical differences at 2 PN order. In the 0-PN limit (hence no spin effects and no higher harmonics) we find precise agreement with the Peters and Matthews waveforms [7].
III. LISA RESPONSE
We simulate the LISA response to a gravitational wave signal plus instrument noise and confusion noise due to galactic binary sources of gravitational waves. We adopt the standard ecliptic coordinate system with origin at the barycenter. The individual data streams from the six LISA phase meters can be combined to cancel out the laser phase noise and form Time Delay Interferometry (TDI) variables [38]. The Michelson style TDI variables {X, Y, Z} can be used to construct three noise orthogonal data streams that are similar to the {A, E, T } variables described in Ref. [39].
Processing the gravitational waveform through the LISA response function imparts additional amplitude and frequency modulations on the one year timescale of the LISA orbits. These effects can be seen in Figure 3, which shows the A-channel response to the Barycenter signal shown previously in Figure 2. Figure 2. The amplitude modulation is due to a combination of the antenna pattern sweeping around as LISA orbits the Sun, and the spin induced precession of the orbital plane. The overall gravitational wave amplitude grows as the system spirals in and nears merger.
The one-sided noise spectral density of the detector in the A and E channels is given by [41]: where u = 2πf /L, S p n (f ) is the position noise and S a n (f ) is acceleration noise, and we have taken the limit of symmetric noise in the detector. The confusion noise due to gravitational wave sources in the galaxy has been estimated by direct simulation of the LISA response [43] to a synthetic galaxy [42], followed by the removal of resolvable systems [44]. An approximate fit to the resulting confusion noise estimate is given by The total noise is then taken to be the sum of the the two contributions: S n (f ) = S inst (f ) + S conf (f ).
IV. PARAMETER ESTIMATION
The goal of a parameter estimation study is to find how accurately we can determine the values of the source parameters for a given signal. We take a Bayesian approach and use the Markov Chain Monte Carlo (MCMC) technique to compute the posterior distribution function p( x|s) describing the model parameters x that we infer from data s. Our results establish how well the various source parameters will be able to be determined by the LISA mission, including the orbital eccentricity.
The use of MCMC techniques in gravitational wave data analysis is now a familiar technique for parameter estimation in gravitational wave astronomy [45][46][47][48][49][50]. The result of a well constructed MCMC is a set of samples from the posterior distribution. The number of samples from a particular region of parameter space is proportional to the posterior weight contained in that region. The uncertainty in each parameter is given by quantiles of the marginalized posterior distribution (e.g. the halfwidth of the 90 th percentile equates to a 2-σ error if the distributions are Gaussian).
By Bayes theorem, the posterior distribution is given by the product of the prior distribution p(x) and likelihood p(s|x), normalized by the evidence p(s) = p(x)p(x|s) dx. For Gaussian noise, the likelihood of the data s having being generated by a gravitational wave signal h(x) is given by where C is a normalization constant that does not depend on the signal or the template. Here we have used the noise weighted inner product where S n (f ) is the one-sided noise spectral density. The prior probability density for x reflects our knowledge of the source parameter, however ill-formed, before we analyze the data. For example, we assume a uniform prior for angular parameters such as the sky location and the initial orientation of the angular momentum vector, such that the cosine of the co-latitude is uniformly distributed in the range [−1 : 1] and the azimuth is uniformly distributed in the range [0 : 2π].
The primary mode of the posterior distribution yields the best fit values for the source parameters, according to the current data and our prior knowledge. In many instances the posterior distribution is multi-modal, and the quantiles used to estimate the parameter uncertainties may cover disjoint regions in parameter space. Even when the bulk of the posterior weight lies in a single, contiguous region, the posterior distribution may not be well approximated by a Gaussian distribution. Nonetheless, a Gaussian approximation to the posterior distribution often provides a reasonable estimate of the parameter estimation errors, which can be efficiently computed using the Fisher information matrix Γ ij , which measures the expectation value of the curvature of the posterior distribution about the mode: We employ a parallel tempered [52] Metropolis-Hastings [53,54] MCMC routine to explore the PDF. The Markov chain starts at parameter values x and transitions to y with probability Here q(x|y) is the proposal distribution, which is the function that generates proposals for moves from x to y. The performance of an MCMC algorithm is quite sensitive to the choice of proposal distribution, and care must be taken to ensure that the chains do not get stuck on local maxima of the PDF. We employed several techniques to ensure rapid exploration of the full parameter space: local coordinate transformations to uncouple the parameters; moves that exploit symmetries of the likelihood surface to encourage jumps between local maxima; and parallel tempering to encourage wide exploration of the posterior [50,51,56]. The number of iterations spent at each parameter value is proportional to the posterior density, and histograms of the parameters visited by the chain provide an estimate of the posterior distribution.
We have found that drawing from a variety of proposal distributions provides a set of jump proposals that tend to produce an MCMC that efficiently maps out the desired PDF and provides accurate parameter uncertainties even for very large search spaces. Our parameter estimation study thus uses several proposal distributions, including parallel tempering and Fisher matrix proposals. In this high dimensional parameter space we found it advantageous to use the Fisher matrix approximation to the posterior to propose jumps along single eigen-directions as part of the mixture of jump proposals.
V. PARAMETER ESTIMATION WITH LISA
We can use our time dependent gravitational waveforms [35] and established MCMC techniques to study how well LISA will be able to measure the full set of seventeen parameters necessary to describe a spinning binary black hole system in an eccentric orbit. We are especially interested in determining when eccentric orbits can be distinguished from circular orbits.
We chose to study signals in their final year prior to merger, with an observation time that extends just beyond the merger. In order to choose initial parameter values for a system that will merge in one year, we need to calculate an initial semi-major axis based on the lifetime estimate for a system with some given initial radial eccentricity. We use the leading order, 0 PN expression for a 0 [8] a 0 = 4 µM 2 5 T c as an initial guess, and apply a bisection routine to the full numerical orbital evolution to find the value of a 0 that yields a merger time T c of one year.
We use parameter ranges consistent with typical LISA sources, given in Table I. The masses are given in terms of the mass of the Sun, M ⊙ = 1.9891 × 10 30 kg, and the luminosity distance D L is given in units of Gigaparsecs. The dimensionless spin parameters χ 1 and χ 2 combine with the black hole masses to give the magnitudes of the spins, S i = χ i m 2 i . There are initial orientation parameters for the orbital angular momentum vector L → (cos θ L , φ L ), as well as the spin vectors S i → (cos θ Si , φ Si ). The final two parameters, n 0 and φ 0 , are initial phases related to the mean motion and orbital phase. In the circular limit these parameters are degenerate, but for eccentric orbits we have to specify the initial periastron position. In the present study we have ignored the possibility that gas dynamics may partially aligned the spins with the orbital angular momentum [57], which would restrict the prior ranges and reduce the degree of orbital precession. The impact of this partial alignment on parameter estimation has been considered for circular orbits [58], and it would be interesting to extend this study to include eccentricity.
Here we study several representative cases to establish the parameter recovery errors and to study correlations between the parameters. The high dimension of the parameter space makes it difficult to perform a comprehensive study -if we were to choose just two values of each parameter we would need to perform 2 17 ∼ 10 5 parameter estimation studies. Each MCMC run involves ∼ 100, 000 iterations with ∼ 8 parallel chains, and takes about a week to run on a single 2.66 GHz Intel processor, so we are limited in the number of examples we can consider. We perform MCMC parameter estimation studies of several representative examples, varying the mass ratio, sky location, distance, eccentricity, and dimensionless spin parameters. We only looked at a few initial spin and orbital orientations since we do not expect these to have a significant effect on the results -unless the initial orientations are very special the system will explore a wide range of orientations during the orbital evolution. To be able to explore the parameter space more widely would take a faster code. One possibility is to use the Fisher information matrix approximation to the posterior, which is many orders of magnitude faster than a full MCMC study. In preparation for such a study we compare the Fisher matrix approximation to the MCMC derived posterior distributions and find that the approximation is fairly reliable so long as the initial eccentricity exceeds e 0 ∼ 0.01.
Our first study focuses on determining when the eccentricity one year before merger is distinguishable from zero. We studied two systems that only differed in sky location (and hence in SNR), and considered initial eccentricities in the range e 0 ∈ [0.001, 0.2], see Table II for a list of the other source parameters. Marginalized posterior distributions for e 0 are shown in Figure 4 for Source 1, and Figure 5 for Source 2. We see that the error in the measured value of e 0 gets smaller as e 0 gets larger, but the dependence on e 0 is fairly weak, and never larger than ∆e 0 ∼ 0.001. Our criteria for deciding if the eccentricity is distinguishable from zero is to see if their is any weight in the posterior distribution at e 0 = 0 (this test is motivated by the Savage-Dicke density ratio estimate for the model evidence [43]). For Source 1 we see that the ex- II: Injected parameter values for two sets of sources studied with a range of values for e0. The results of the parameter estimation study for Source 1 are given in Figure 4 and the results for Source 2 are given in Figure 5. amples with e 0 ≥ 0.005 are clearly distinguishable from circular, while the e 0 = 0.002 case is on the margin of detectability. For Source 2, which has higher SNR due to a more favorable sky location, the e 0 = 0.002 case is clearly distinguishable from circular. The Fisher matrix approximation to the posterior distribution is computed at the maximum a posteriori probability (MAP) value of the parameters, and is found to work well for eccentricities e 0 > 0.05, but breaks down for small eccentricities. The MCMC derived posterior distributions are much flatter than a Gaussian distribution, and we attribute the failure of our Fisher matrix estimates to only including the leading order, quadratic curvature terms in the Fisher matrix calculation. This study suggests that LISA observations of eccentric black hole binary systems will be able to measure the eccentricity of the system and distinguish eccentric systems from circular systems to parts in a thousand. A similar study was performed by Porter and Sesana [27] for non-spinning eccentric binary black hole systems. Their results suggest that LISA will be able to measure the eccentricity to parts in 10 −4 for such sources.
We find that the other source parameters are also measured quite well, as illustrated in Figure 6 and Figure 7. Marginalized posterior distributions are shown for the chirp mass M c = (m 1 m 2 ) 3/5 /(m 1 + m 2 ) 1/5 and reduced mass µ = m 1 m 2 /(m 1 + m 2 ), the distance to the source, the initial radial eccentricity, and the two sky location parameters. We compare the Fisher matrix approximation to the marginalized posterior distributions computed from the MCMC runs and find excellent agreement for all the parameters (except for the eccentricity in Figure 6). Table II. Table II As a start to exploring the large parameter space of eccentric binary black hole systems we consider a few representative examples below. We find that in general physical parameters are well constrained by LISA observations and that the Fisher matrix makes fair estimates of the parameter errors. In addition to the examples below with varied spin and distance parameters we also studied systems with a range of mass ratios m 1 /m 2 ∈ [1, 5] and total masses M ∈ [10 5 M ⊙ , 10 7 M ⊙ ] with similar results. We also compared the parameter estimation errors with those obtained when the eccentricity is held fixed at zero, and saw only small (less than 50%) changes in the error estimates.
The dimensionless spin parameters χ 1 and χ 2 are varied in Table III parameters. The magnitude of the spin vectors is related to the mass of the black hole S i = χ i m 2 i . The posterior distribution for several of the source parameters for these cases are given in Figures 8 and 9. The Fisher approximation is a reasonable prediction of the width of the posterior distribution for the spin of the more massive body, but does a poor job for the less massive body. This discrepancy was seen in many other examples that we looked at. The cause of the discrepancy is presently not understood. Two lower SNR examples are shown in Figure 10 and Figure 11, and we see good agreement with the Fisher matrix estimates.
We do not expect the eccentricity to be highly corre- lated with the other source parameters since the higher harmonics introduced in the waveform due to eccentricity cannot be simulated by changes in other parameters or their combinations. We indeed find that the initial eccentricity is not correlated with the other parameters. Compare the distribution of values for the two masses in Figure 12 to the distribution of mass and eccentricity values in Figure 13. The two masses are highly correlated since it is the total mass of the system and the ratio of the masses that appear in the waveform. The distribution of eccentricity versus the other source parameters is similar to that seen in Figure 13.
VI. CONCLUSION
Our studies of the response of the LISA detector to the gravitational wave signal from spinning binary black 14 hole systems in eccentric orbits show that the eccentricity should not be neglected for LISA data analysis and parameter estimation. We find that LISA can determine the eccentricity of the system one year before merger to parts in a thousand. This result depends only weakly on the initial value of the eccentricity, indicating that LISA will be able to distinguish between eccentric and circular orbits at the same level e 0 ∼ 10 −3 . The construction of the gravitational waveforms for spinning binary black hole systems in eccentric orbits to 1.5 PN order in [35] establishes the framework for the extension of this work to higher post-Newtonian order. Binary black hole waveforms that include eccentricity will be necessary for several of the LISA science goals including constraining galaxy merger scenarios and testing general relativity in the strong field regime near supermassive black holes. This work builds the foundation for further studies including a comprehensive exploration of the parameter space including source sky location, distance, spins, masses, and mass ratios. For sources with initial radial eccentricity greater than e 0 ∼ 0.02 the Fisher matrix is a good approximation to the posterior distribution. The very large parameter space could be studied quickly using the Fisher approximation, although it is not as useful for the low initial eccentricity cases.
The equations of motion and instantaneous gravitational waveforms have been calculated to the next post-Newtonian order and these pieces can be included in a future parameter estimation study. The 2 PN effects include the spin-spin coupling of the two black holes and thus corrections to the precession and evolution of the system.
These waveforms can be also be used to study how well the Advanced LIGO-Virgo network and proposed Einstein Telescope will be able to measure eccentricity and what level of bias could be expected from using circular templates for parameter estimation.
VII. ACKNOWLEDGEMENTS
This work was supported by NASA grants NNX07AJ61G and NNX10AH15G. We are grateful to Alberto Sesana and Edward Porter for extensive and informative discussion of the mechanisms responsible for producing black hole binaries with eccentric orbits. | 2011-02-12T14:35:45.000Z | 2010-06-18T00:00:00.000 | {
"year": 2010,
"sha1": "57595629180ca1b5d18eeda863c1a32d124e549e",
"oa_license": "publisher-specific, author manuscript",
"oa_url": "https://link.aps.org/accepted/10.1103/PhysRevD.83.083001",
"oa_status": "HYBRID",
"pdf_src": "Arxiv",
"pdf_hash": "57595629180ca1b5d18eeda863c1a32d124e549e",
"s2fieldsofstudy": [
"Physics",
"Geology"
],
"extfieldsofstudy": [
"Physics"
]
} |
59392673 | pes2o/s2orc | v3-fos-license | Influence of silencing the MC 4 R gene by lentivirus-mediated RNA interference in bovine fibroblast cells
Melanocortin receptor 4 (MC4R) is a key element in the mechanisms used to regulate both aspects of keeping the balance between energy uptake and energy expenditure. MC4R was knocked down by lentivirus-mediated shRNA expressing plasmids, which were controlled by the U6 promoter in bovine fibroblast cells, and the expression of MC4R was examined by the real time-PCR and Western blot analysis. Real time-PCR analysis was used to characterize the expression of Leptin, POMC, AGRP, MC3R and NPY gene. The relative genes [leptin, proopiomelanocortin (POMC), agouti-related peptide (AGRP), MC3R and neuropeptide Y (NPY)] expression level seemed to be closely associated with the MC4R gene in bovine fibroblast cell lines (BFCs). The levels of both MC4R mRNA and protein were significantly reduced by RNA interference (RNAi) mediated knockdown of MC4R in BFCs cells transfected with plasmid-based MC4R-specific shRNAs. The finding of this study demonstrated that vector based siRNA expression systems were an efficient approach to the knockdown of the MC4R gene expression in bovine fibroblast cells and they provided a new molecular basis for understanding the relationship of MC4R and other genes, which were responsible for the regulation of energy homeostasis by the melanocortin system.
INTRODUCTION
The melanocortin receptor 4 (MC4R), a G-protein coupled receptor, plays a pivotal role in controlling meal size and energy homeostasis in mammals (Govaerts et al., 2005;Adan et al., 2006).The accumulating evidence implicated the components of the CNS melanocortin system, which encompassed several receptors and their peptide ligands in the central regulation of feeding and energy balance (Lu et al., 1994;Fan et al., 1997).MC4R was one of the five melanocortin receptors (MCRs) and was shown to be widely distributed throughout numerous *Corresponding author.E-mail: simmenta@yahoo.com.cn.
#These authors contributed equally to this work.brain regions (Mountjoy and Wong, 1997).In spite of the evidence, presented about the role of several important factors in energy metabolism, the regulation and integration of the satiety pathway and the energy homeostasis have not been completely elucidated.These important factors are neuropeptide Y (NPY), leptin, proopiomelanocortin (POMC), Agouti-related peptide (AGRP) and their respective receptors (Pelleymounter et al., 1995;Stephens et al., 1995;Hu et al., 1996;Boston et al., 1997;Clement et al., 1998;Friedman and Halaas, 1998), whereas only few researches have been conducted on the relationship of MC4R with the expression of upstream regulatory genes in bovine.
RNA interference (RNAi) is a wide and powerful tool which will lead to sequence-specific gene silencing at the post-transcriptional level.In mammalian cells, synthetic siRNAs can be transferred into the cells to induce transient gene silencing, whereas short hairpin RNAs (shRNAs) could be driven by RNA polymerase III promoters and are subsequently cleaved by the Dicer to produce siRNAs in order to introduce mRNA degradation; but there are lots of problems related to the method of delivery, especially low efficiency of transfection in primary cells.However, lentivirus can stably transfect to many cells, such as primary cells, suspension cells and stem cells, at high efficiency, and also sustain long-term gene expression by integrating itself into the host genome.Therefore, lentiviral vector is considered as the most suitable delivery method for this study.
In the recent studies (Chen et al., 2000;Ste et al., 2000), knockout of the MC4R gene was usually carried out in vitro or in vivo of mice.However, differences may exist between the bovine MC4R mRNA and the other species mRNA in terms of the length and spatial structure, so as to affect the effectiveness of shRNAs screening and relative regulation genes.Therefore, it is necessary to perform shRNAs screen in primary bovine cells which naturally knockdown the MC4R gene expression.
In this study, a shRNA-expressing lentiviral vector was constructed to target the bovine MC4R gene, and the resulting interference effect was analyzed in bovine fibroblast cells (BFCs).Subsequently, the resulting induction along with the expression of leptin, POMC, AGRP and NPY genes were studied.
Isolation and culture of cell
Auris tissues, blocked from 20 month-old male local yellow bovine in the Animal Center of Institute of animal science -Chinese academy of agriculture science, were resected under aseptic conditions.Bovine fibroblast cells and human embryonic kidney 293 T cells (HEK-293T) were cultured in Dulbecco's Modified Eagle Medium (DMEM; Gibco BRL, USA) and supplemented with 10% fetal bovine serum (FBS; Gibco BRL, USA) at 37°C under 5% CO2.Cell cultures, at 80% confluence, were detached with trypsin-EDTA and plated into 12-well plates or 6-well plates at a density of 5 x 10 4 cells/cm 2 for transfection only.
Real-time PCR analysis
Total RNA was extracted by Tiangen RNeasy mini kit (Tiangen, China) according to the manufacturer's instructions.About 2 to 5 µg of the total RNA were reversely transcribed by Quant one step RT-PCR kit (Tiangen, China) with oligo(dT)18 primers.Real-time RT-PCR was performed by an ABI Prism TM 7900HT Sequence Detection System (Applied Biosystems, USA).Then, each PCR reaction with a total volume of 25 µL contained 2 µL of cDNA, 0.5 µL of 10 µmol L -1 primers, 9.5 µl ddH2O and 12.5 µl SYBR Premix Ex Taq (2×) (TaKaRa, China).The relative expression levels of all genes were normalized to 18S rRNA expression levels, respectively, and all primers are shown in Table 1.
Western blot analysis
The entire cellular protein was obtained with BCA Protein Assay Kit based on the manufacturer's instructions (Bestbio, China).The aliquots were separated on SDS-PAGE (10%) and transferred to nitrocellulose membranes.The MC4R antibody (SC6879-C19) was purchased from Santa Biotechnology (Santa Cruz Biotechnology Inc., USA).
Then, the membranes were incubated with goat anti-MC4R protein antibody at room temperature for 2 h and stained by the secondary antibody mouse anti-goat (SC2354) immunoglobulin G, horseradish peroxidase (HRP) conjugate (Santa Cruz Biotechnology Inc., USA) for 1 h.Anti-GAPDH antibody was obtained from Sigma Biotechnology as a positive control (Sigma, USA).
Statistical analysis
Statistical analyses were performed by student's t-test using SPSS statistical software (SPSS Inc., USA).P < 0.05 was considered as statistically significant.
Detection of MC4R silencing of Lvsh-MC4R in FBCs
The Auris tissue block was minced into pieces and digested in the presence of type II collagenase.Then, the resulting pellet was cultured for 96 h, in which most of the cells were slender with fibroblast shape under microscope.Generated lentivirus was further used to infect FBCs cells to obtain the most effective target virus (Figure 2a).The total RNA was extracted after 72 h later, and the Real-time PCR results that infection with Lvsh-MC4R-1,-2,-3 or -4 could inhibit MC4R mRNA expression by approximately 71, 8, 11 or 15%, respectively, in contrast to the infection with Lvsh-MC4R-NC (Figure 2b).As the transfection efficiency of the bovine fibroblast cells was around 80%, the result suggested that Lvsh-MC4R-1 markedly suppressed MC4R mRNA expression, while Lvsh-MC4R-2, -3 and -4 did not influence the levels of MC4R mRNA.Subsequently, the result from the western blot, when compared to the infection of LvshMC4R-NC (Figure 2c), further confirmed, that the Lvsh-MC4R-1 could significantly suppress the expression of MC4R at both levels of mRNA and protein, and maintain satisfactory silencing effect in BFCs.
Effect of Lvsh-MC4R on expression of regulatory genes
In order to observe the effect of MC4R silencing on the other genes of energy metabolism, Lvsh-MC4R1 was selected as the knockdown group (KD), which was compared with the negative control group (NC) to detect the expression level of the four upstream regulatory genes in all the samples.As shown in Figure 3, the Leptin, POMC and NPY genes expression in the KD group was detected to be appreciably lower than that of the NC group; but the expression of MC3R gene in NC group was found to be significantly lower than the KD group, especially, where it was found that the expression of AGRP gene in the KD group was much significantly higher than that of the NC group.
DISCUSSION
In this study, lentiviral vectors are used to silence bovine MC4R gene expression not only in HEK-293T cells, butalso in bovine fibroblast cells.Human U6 promoter was used to drive siRNA expression from the DNA templates and four target sites were designed.Finally, it was only one effective target site (Lvsh-MC4R1) that could down-regulate the MC4R gene expression in bovine fibroblast cells.Lentiviral vectors were chosen as the shRNA delivery vehicle to suppress the MC4R gene in bovine fibroblast cells, for the reason that lentivirus can integrate itself into the genome of host cells and sustain long-term gene expressions (Gropp and Reubinoff, 2007).Generally, a lentiviral vector expression system consists of three plasmid vectors, which include lentiviral plasmid (transfer vector), packaging vector and VSVG Fig. 1 The identities of pGCSIL-GFP-shMC4R vectors and LvshMC4R.A. The recombinant pGCSIL-GFP-shMC4R-1,-2,-3,-4 plasmids were identified by PCR.Lane 1, 3, 4, PCR products of the recombinant pGCSIL-GFP-shMC4R-1,-2,-3,-4 plasmids, Lane 2, PCR products of the empty vector; Lane 5, DNA ladder (DL 2000), 500 bp (up) and 250 bp (down); the positive clones were 352bp and the negative clones were 318 bp.B. 293T cells were infected with Lv-shMC4R (KD) or -NC (NC), and phase contrast or GFP expression which was examined after 72 h.expression plasmid, and a packaging cell line (pseudovirion producing cell line).In this study, there are four lentiviral plasmids that are successfully constructed to target MC4R and a negative control using the three plasmid expression system.The result, shown In Figure 2, suggested that the lentiviral vector expression system could effectively drive siRNA and transcribe it from the DNA templates into the bovine fibroblast cells.
In contrast to all other hormone systems known so far, the melanocortin system was the only one that was not only regulated by agonistic peptides, but also by two endogenously occurring neuropeptides that were blocked from ligand-induced MCR signalling: agouti and "agoutirelated protein" (AGRP).Agouti binds to almost all MCR with high affinity, whereas AGRP exhibits rather selective binding towards the MC3R and MC4R (Ollmann et al., 1997).This study indicated that AGRP impaired the appetite controlled in a melanocortin-independent manner, and it was defined as a classic competitive antagonist, since binding of AGRP prevented or displaced the binding of melanocortins to the MC3R or MC4R (Tolle and Low, 2008;Wu et al., 2008).However, this study herein found out that the expression of AGRP gene, presented significantly, was high after MC4R gene was silenced.This result supports the suggestion that mechanisms, other than competitive antagonism of MC4R signaling, should also be considered when discussing the physiological effects of AGRP on the molecular level.As the receptors for NPY and MC3R are known, the members of the G-protein-coupled receptor family were down regulated in hypothalamic regions under conditions that were known by enhancing the local NPY release (Widdowson et al., 1997).In this study, mRNA level of NPY gene was down regulated, and oppositely, that of MC3R gene was up regulated, suggesting that MC3R and MC4R genes were competitively bound with NPY.Adipose tissue-derived hormones, such as leptin, increase POMC expression in α-MSH-releasing neurons located in the arcuate nucleus of the hypothalamus (Shimizu et al., 2007).Various peripheral and central factors could signal energy balance to the POMC neurons, as ARC POMC neurons express the functional OB-Rb leptin receptor, while leptin apparently stimulates these neurons (Cheung et al., 1997;Schwartz et al., 1997).However, the expression level of leptin and POMC went down as the MC4R was knocked down in this study, suggesting that they might be the medium in the signalling pathway of the melanocortin system.
In summary, one out of the four designed target sites reduced the MC4R gene expression by lentivirus mediated in bovine fibroblast cells, and the expressions of a series of genes of the CNS melanocortin system were detected.The results of this study provided a new molecular basis for understanding the mechanism of energy homeostasis, and also laid the foundation for further study of the biological functions of MC4R gene and other related genes.
Fig. 2
Fig.2The effect of infection with Lv-shMC4R and LvshMC4R-NC in BFCs. A. BFCs were infected with Lv-shMC4R (KD) and LvshMC4R-NC (NC), and representative phase contrast or GFP expression was examined after 72 h.B. BFCs were infected with Lv-shMC4R or Lvsh-NC, and the level of MC4R mRNA was detected by Real-time PCR 72h later.mRNA levels in NC were taken as 1.0.C. MC4R protein levels were detected by western blot 72h later.Each treatment was established in duplicate and each sample was examined in duplicate.All data were expressed as means ± SEs
Table 1 .
Parameters of oligo-nucleotide primer pairs for PCR primers. | 2018-12-28T03:04:14.516Z | 2011-08-03T00:00:00.000 | {
"year": 2011,
"sha1": "d4944f6c3ed753e9028f54bad0bcb23f2919c79d",
"oa_license": "CCBY",
"oa_url": "https://academicjournals.org/journal/AJB/article-full-text-pdf/EEB12C425877.pdf",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "3f7a5c28d9250d83a2c81fe07f2171e2fc3d7f59",
"s2fieldsofstudy": [
"Biology",
"Medicine",
"Agricultural and Food Sciences"
],
"extfieldsofstudy": [
"Biology"
]
} |
221586387 | pes2o/s2orc | v3-fos-license | RoIFusion: 3D Object Detection from LiDAR and Vision
When localizing and detecting 3D objects for autonomous driving scenes, obtaining information from multiple sensor (e.g. camera, LIDAR) typically increases the robustness of 3D detectors. However, the efficient and effective fusion of different features captured from LIDAR and camera is still challenging, especially due to the sparsity and irregularity of point cloud distributions. This notwithstanding, point clouds offer useful complementary information. In this paper, we would like to leverage the advantages of LIDAR and camera sensors by proposing a deep neural network architecture for the fusion and the efficient detection of 3D objects by identifying their corresponding 3D bounding boxes with orientation. In order to achieve this task, instead of densely combining the point-wise feature of the point cloud and the related pixel features, we propose a novel fusion algorithm by projecting a set of 3D Region of Interests (RoIs) from the point clouds to the 2D RoIs of the corresponding the images. Finally, we demonstrate that our deep fusion approach achieves state-of-the-art performance on the KITTI 3D object detection challenging benchmark.
I. INTRODUCTION
Object detection of 3D bounding boxes is one of the fundamental challenges of situational awareness and 3D environmental perception for autonomous systems (e.g. autonomous vehicles, robots, unmanned aerial vehicles, etc.). In fact, autonomous systems need to perceive objects in their surrounding environment using different sensors (e.g. cameras, LI-DAR) for navigation and obstacle avoidance. In the past few years, 2D object detection for computer vision [7], [9], [22]- [24], [31], [32] has made significant progresses, especially with the advent of Convolutional Neural Network (CNN) technology [15]. However, 3D object detection remains an open challenge, especially when multiple sensors are used to obtain a more reliable and robust information.
Recently, many researchers focused on the exploitation of LIDAR-only methods for 3D object detection due to the advantages that the point clouds provide precise depth information and dense geometric shape feature [2], [3], [25], [29], [30], [43]. PointRCNN [34] builds a two-stage architecture to directly process dense 3D point clouds and estimate 3D bounding boxes from all the foreground points. VoxelNet [53], SECOND [47] convert the point clouds to voxels before applying standard CNNs to achieve the same result. Pixor [48], Complex-YOLO [36], and Birdnet [1] operate deep CNNs on Bird-Eye-View (BEV) maps for 3D object classification and bounding box regression.
However, a standard point cloud is incapable of offering texture information and high resolution of an object, which are actually beneficial to capture discriminative features. In contrast, the images provide rich color and texture information, but with a lack of depth and scale information without the application of complex and computationally intensive algorithms (i.e. stereography). For example, small objects (e.g. pedestrians) detected at long-distance generates only few points in the point cloud, which makes the classification or localization of these objects very difficult with only a LIDAR. Meanwhile, in the image domain, texture and color features of small objects can be still visible, due to the higher spatial resolution of images, and likely to be captured by existing mature 2D CNNs technology. As a result, the fused features, leveraging the advantages from both point clouds and images, are beneficial in exploiting more reliable representations and improving the performance of the 3D object detection architecture.
However, it is still challenging to develop an efficient and effective sensor fusion method due to the viewpoint misalignment caused by the properties of the point clouds and the images. In order to address this issue, early method MV3D VOLUME 4, 2016 1 arXiv:2009.04554v1 [cs.CV] 9 Sep 2020 [5] and AVOD [16] perform 3D bounding box regression on fused 2D images and 2D Bird-Eye-View (BEV) feature maps, although quantization for BEV generation gives rise to a lot of geometric information losses. Frustum Pointnets [28] and Pointfusion [46] project 2D bounding boxes from the imagebased 2D detector onto the point clouds to coarsely cluster potential foreground points. At this point, the PointNets are applied for 3D boxes estimation, but this procedure heavily relies on the performance of the 2D detectors. PointPainting [40] feeds the pixel-wise semantic features captured from image-based semantic segmentation model onto corresponding point-wise semantic features in the point cloud to boost the performance of the 3D object detection.
It can be observed that the main disadvantage of dense point-pixel fusion method [40] is that they are leading to a considerable amount of redundant computations. Meanwhile, using a BEV-image fusion method allows the deep learningbased fusion of the feature maps captured from an individual viewpoint but with geometric information losses. However, it is the authors assumption that it is not strictly necessary to densely fuse the whole point clouds with images. Conversely, it is feasible to generate a small set of potential Region of Interests (RoIs), followed by the application of a deep fusion method only on those local regions used for 3D object detection. The advantages of this fusion method are that it considerably reduces the computation cost and allows an easy alignment of the viewpoints on the local regions.
Motivated by these observations, we hereby present an efficient and lightweight deep fusion method for 3D object detection for point clouds and images. Our main contributions can be summarized as follows: • We propose a lightweight deep fusion neural network, named RoIFusion, aiming at efficiently and effectively fusing the point clouds and the images for 3D object detection. • We propose a keypoints generation layer for the estimation of the keypoints on the objects guided by the fusion of the point clouds and the images, followed by a voting layer used to generate the center points of the objects. • We propose a RoIFusion layer to aggregate the 3D RoIs generated from the center points with the corresponding 2D RoIs which are obtained by projecting the 3D RoIs to the images. • We evaluated our model on the KITTI dataset [10] and achieved state-of-the-art results compared with respect to other outstanding methods.
II. RELATED WORK A. LIDAR-ONLY METHODS FOR 3D OBJECT DETECTION
Many existing methods explored the possibility of detecting the objects with 3D bounding boxes only using point clouds, as they provide accurate geometric information. It is possible to broadly classify these methods into four subcategories: projection-based methods, volumetric-based methods, pointnets-based methods, and point-voxel methods.
a: Projection-based methods Several works [1], [37], [48] apply 2D CNNs directly to Bird-Eye-View (BEV) projected from the raw point clouds in order to estimate the 3D bounding box and orientation of an object. FVNet [52] projects the raw point clouds to the front view, which is then fed to a proposal generation network and a refinement network to estimate the parameters of the 3D bounding box (i.e. object location, size, and orientation). This method allows for building a lightweight neural network for real-time applications. However, it ignores the size and the location of the objects and suffers from lots of geometric information losses during quantization. As a result, it is unlikely to exploit sufficient discriminative features for 3D object detection.
b: Volumetric-based methods
Volumetric-based methods convert the raw point clouds to standard 3D grids and represents the point clouds as voxels.
For instance, VoxelNet [53] learns discriminative voxel-wise features for 3D region proposal generation and then proceeds to solve the 3D bounding box regression problem. However, many empty voxels are generated during the voxelization process, which leads to large computational cost because of the processing of those empty cells. In order to address this problem, SECOND [47] improved VoxelNet [53] by proposing an efficient method, named sparse convolution [12], to ignore the empty voxels. Finally, PointPillars [17] converts the raw point clouds to a set of stacked pillars and then encodes the same to 2D pseudo-images, which can be used as input for 2D CNNs for 3D bounding box regression.
c: Pointnets-based methods
PointNets [29], [30] models are efficient in the exploitation of point cloud features. PointRCNN [34] sets an example in the classification and regression of 3D bounding box directly from dense point clouds. Specifically, it firstly applies PointNet++ [30] to extract dense semantic features for all the points, and then generates 3D region proposals for all the foreground points. Successively, the second stage is applied to refine predictions. However, the dense processing leads to quite heavy computational costs.
d: Point-voxel methods
In order to achieve high detection performance but to also reduce the computational costs, several works [6], [26], [33], [35] introduced two-stages neural networks for 3D object detection. In the first stage, they coarsely localise the objects and estimate the parameters of the bounding box from the voxel grids generated from the raw point clouds. In the second stage, they introduce a refinement module that leverages the PointNets to refine the 3D bounding box. The methods leverage the sparsity of the voxel grids and the ability of PointNets to carry out feature extraction and to gradually detect the 3D objects starting from coarse to more refined representations.
Finally, a recent one-stage method, 3DSSD [49] abandons the refinement stage and builds a one-stage anchor-free neural network to directly regress 3D bounding box from the estimated candidate 3D RoIs.
B. IMAGE-ONLY METHODS FOR 3D OBJECT DETECTION
In the past few years, 2D object detection has made great progress. However, estimating 3D bounding box directly from 2D images is still quite difficult due to the lack of depth information in single camera images. Mousavian et al. [27] estimates the pose and 3D bounding box by learning the geometric constraints from the 2D bounding box. Wang et al. applies LIDAR-only 3D detectors on the Pseudo-LiDAR representations converted from the estimated image-based depth maps. Stereo R-CNN [20] applies Faster R-CNN [32] on both the left and right images and predicts 3D bounding boxes by learning the projection relations between the associated 2D left-right bounding boxes and 3D bounding box corners.
C. MULTI-SENSOR FUSION METHODS FOR 3D OBJECT DETECTION
In order to get the best of both worlds, there are several works attempting to fuse point clouds and 2D images with various strategies. Early works such as MV3D [5] and AVOD [16] firstly used off-the-shelf 2D feature extractors to capture the feature maps from the images and the multi-view representations of the point clouds (e.g. Bird Eye View and Front View), which are then typically fused together by a sum or a concatenation operation. A Region Proposal Network (RPN) is then applied to the fused feature maps to generate 3D bounding box proposals, followed by a refinement network for final 3D bounding box prediction. The advantages of this method are that mature 2D object detector and 2D feature extractor technologies are available to be applied on the multiview representations of the point clouds. Furthermore, the features from different sensors can interact over the stacked layers, as these features are normally obtained from similar or even the same neural networks. Liang et al. [21] utilizes the continuous convolution method to fuse the feature maps of the images and BEVs. Specifically, this approach proposes a continuous fusion layer that aggregates each pixel feature in the image feature maps with the features of the neighbouring points in the BEV feature maps to learn a fused local region, which allows to extract sufficient discriminative features for 3D object detection.
In order to narrow the searching space, Frustum pointnets [28] and Frustum convnet [44] introduced the concept of 3D bounding frustums. The 2D bounding boxes are obtained from mature 2D detectors, and then the 3D frustums are used to trim the point cloud data. Finally, Pointnets methods are applied to the trimmed point clouds for carrying out the 3D bounding box regression task. Similarly, Pointfusion [46] aggregates the global features of the image obtained from an off-the-shelf 2D feature extractor with the dense semantic features of the point cloud, which are captured from Pointnet [29]. Finally, PointPainting [40] densely aggregates the output of the image segmentation neural network with the point clouds before applying LIDAR-only 3D detectors to boost the performance of the 3D object detection task.
III. ROIFUSION ARCHITECTURE
In this section, we introduce and describe our RoIFusion neural network for 3D object detection as shown in Fig. 2, which uses both raw point clouds and 2D images as input. Our goal is to leverage the fusion information captured from both sensor modalities to classify and localize the objects within the oriented 3D bounding boxes. In particular, we firstly propose a fused keypoints generation layer (FKG layer) to estimate a set of 3D keypoints from the point clouds, followed by a RoIs fusion layer to fuse the 3D RoI features in the point clouds with the 2D RoI features in the images by the 3D/2D RoI pooling operation respectively. At last, a prediction layer is proposed to predict the parameters of the oriented 3D bounding box.
A. FUSED KEYPOINTS GENERATION (FKG) LAYER
Instead of densely generating 3D region proposals relying on all the foreground points, we only estimate a small set of 3D keypoints on the objects to generate the RoIs for deep fusion. As illustrated in part (a) of Fig. 2, our FKG layer takes the raw point clouds and the RGB images as input, and combines point-guided keypoints and pixel-guided keypoints that are generated by individually performing the LIDAR-only point cloud segmentation model and image-only segmentation network respectively. As a result, we obtain a set of keypoints on the objects leveraging both the point cloud and the image information.
We can define the point cloud as shown in Eq. (1), where x i denotes the i-th point with the 3D space coordinates [x i , y i , z i ] and the measured reflectance r i . As a result, the dimension of the point cloud data set is N × 4. At each layer, a set of points are processed and a new set with higher-level but fewer points is generated. Finally, we obtain a small number of points that are treated as keypoints.
With respect to the downsampling strategy, we use an iterative farthest point sampling (FPS) method to select the points for the subset. Let us suppose an empty subset X1, a random point is firstly picked and added to X1, then the point having the farthest 3D geometric Euclidean distance is iteratively added to X1 until the expected M points are picked. The FPS strategy, named D-FPS, has a better coverage of the whole point set than random sampling. In order to preserve sufficient foreground points and filter out the background, inspired by 3DSSD [49], we decided to also employ a specific FPS strategy, named F-FPS, which calculates the Euclidean distances of the semantic features for the points selection. The F-FPS method is beneficial to preserving foreground points (e.g. points on the objects) and removing the useless background, such as points on the ground. Finally, we follow [49] and combined both FPS strategies together to efficiently capture sufficient foreground points as the keypoints. Considering the fact that the colour and texture representations are useful to localize objects within point clouds, especially for small objects that are difficult to be detected by LIDARonly detectors, we capture the segmentation features and corresponding scores using an image segmentation network, which then is used to guide the keypoints selection as shown in Fig. 3.
The detailed procedure to extract the pixel-guided keypoints is shown in Alg. 1.Firstly, we generate the point clouds segmentation features X s using a Feature Propagation (FP) layer as shown in part (e) of Fig. 2. In particular, we leverage the output of the SA layer as input, and upsample the points by interpolating the point features using the inverse squared Euclidean distance weighted average function as shown in Eq. (2). Furthermore, we concatenate the interpolated point features with the skip linked point features from the corresponding SA layer. As a result, our FP layer outputs the 3D geometric points and corresponding semantic features with the same number of points as the raw point cloud.
where ω i (x) = For what concerns the image processing, it is a common choice in literature to use mature 2D feature extractors to capture the feature maps from the RGB images. However, these feature maps are unlikely to localize the objects in the images. As a result, we use a lightweight image segmentation neural network DeepLabv3 [4] to efficiently capture pixelwise segmentation features I s and segmentation scores S, which allows to ignore the background and conduct the keypoints selection. It is worth pointing out that our RoI-Fusion model is agnostic to the development of the image segmentation models.
After that, we project the point cloud to the image viewpoint to paint the point cloud with the segmentation features and the segmentation scores from the relevant pixels, then we mask the painted point cloud X img with the foreground image segmentation scores and map to the point clouds segmentation features to generate all the foreground segmentation features X (obj) s for the point cloud. At last, we use the F-FPS as our downsampling strategy to further select a small set of point segmentation featuresX (img) with the dimension size of M 2 × F , where M 2 and F are the number of keypoints and corresponding features respectively.
c: Keypoints fusion
We finally obtain the fused keypointsX ∈ R M,F by aggregating the point-guided keypoints with the pixel-guided keypoints along with the channel of the number of the keypoints, where M is the number of fused keypoints. We note that the points on small objects are likely to be selected due to the fact that a part of the points are captured based on the image segmentation scores.
B. ROIS FUSION LAYER
After the implementation of the FKG layer intertwined with the point cloud segmentation network and the image segmentation network, we obtain a set of keypoints scattered over the objects, which are also used to predict the center of the objects before we generate the RoIs. Considering that these keypoints are on the objects, inspired by [8], we use Algorithm 1. Pixel-guided keypoints generation.
Input: The point clouds X ∈ R N,4 .
The images I ∈ R W,H,3 . Homogenous transformation matrix T ∈ R 4,4 . Camera projection matrix M ∈ R 3,4 . Output: Pixel-guided keypoints featuresX (img) ∈ R M2,F . 1: Apply for point cloud segmentation network to obtain segmentation features X s ∈ R Fp . 2: Apply for image segmentation network to obtain segmentation features I s ∈ R W,H,Fi and segmentation scores S ∈ R W,H,C . 3: X img = P rojection(M, T, X). 4: X obj , index = M ask(X img , S). the spatial location and features of the keypoints to estimate the corresponding center of the objects. As shown in part (b) of Fig. 2, these high-level keypoints are used to generate 3D RoIs in the point cloud view and corresponding 2D RoIs in the camera view, followed by a RoI fusion operation to obtain the fused RoI features for further bounding box regression and object classification. Specifically, we build a subnetwork VoteNet with a single layer to learn the spatial offset between predicted center points and corresponding ground truth. We treat each center point as the centroid of the 3D bounding box of the object.
a: 3D RoIs generation and pooling
The 3D RoIs are generated for the center points we previously obtained. Successively, we apply a 3D RoI pooling layer to VOLUME We encode our RoIs using the axis-aligned 3D bounding boxes. Specifically, the centroid of each RoI (x (c) , y (c) , z (c) is parametrized using the obtained center point. The length l and the width w of the RoI are set to the enlarged length size of the objects to cover all the orientation scenarios as shown in Fig. 4. We finally use an enlarged height of the objects as the height h for each RoI. As a result, the dimension of the RoI is defined as (x where η is the parameter for extended size of the RoI. After that, we shift the points inside each 3D RoI to the relative locations based on the center points for better local features learning and then apply a subnetwork equipped with stacked Multi-Layer-Perceptron (MLP) layers on the cluster the points inside the 3D RoIs to extract the local RoI pooling features. Our 3D RoIs are then projected to the image to generate the corresponding 2D RoIs, followed by a 2D RoI pooling layer, inspired by [32], to learn the local texture features for the 2D RoIs.
c: RoI pooling features fusion
We finally fuse the point cloud 3D RoI and the image 2D RoI by aggregating the pooling features along with feature dimension axis as shown in Fig. 6. Specifically, we define a fusion strategy by concatenation as in Eq. (3): where F fuse is the fused feature from the 3D RoI pooling features F
C. PREDICTION LAYER
The prediction layer, inspired by [49], use an anchor-free method to directly predict the offset between the center points and corresponding ground truth of the center of the 3D bounding box for regression. Besides, we also directly regress the 3D bounding box size from the fused RoI features. For the orientation regression, we follow the method introduced in [28] that utilizes a hybrid classification and regression formulations to estimate the orientation angle of the 3D bounding box. In particular, we pre-define H equally split angle bins and use the output of the RoI fusion layer to classify the angle bins, and then regress residual with respect to the classified bin.
IV. MODEL STRUCTURE
The model structure is presented in Fig. 2. In our experiments we randomly choose N = 16384 points from the raw point cloud. We then apply the SA layer Fig. 5(a), the FP layer Fig. 5(b), and the image segmentation network DeepLabv3 to capture M = 256 keypoints. The hyper-parameters of the SA layer and the FP layer are represented in Fig. 5(a) and Fig. 5(b) respectively. Successively, we employ a single layer Votenet with filters (128) to estimate the center points for the 3D bounding box of the objects. The dimensions of the 3D RoIs are set to [h = 1.8m, w = 5.0m, l = 5.0m], [h = 1.8m, w = 1.0m, l = 1.0m], [h = 1.8m, w = 1.8m, l = 1.8m] for the car, pedestrians, and cyclists objects respectively. We set the constant extended value η = 1.0m.
V. EXPERIMENTS
In this section, we evaluate our deep fusion method on the widely used KITTI 3D object detection benchmark [10], [11]. We firstly introduce the KITTI dataset and explain the detailed training settings. Then, we demonstrate our results by comparison with recent state-of-the-art 3D detectors. We only test our model on the car category due to the large amount of data after preprocessing. However, we evaluate all the categories when we compare our model to the backbone model 3DSSD [51] to present the effectiveness of our fusion method. Finally, we analyse the efficiency of our fusion method and visualize some representative results for our 3D object detection model.
A. DATASET
The KITTI dataset [10] contains both 2D images and 3D point clouds with the corresponding annotations for the cars, pedestrians, and cyclists categories in an urban driving scenario. The sensors used for data collection are: 2 grayscale cameras, 2 color cameras, and 1 Velodyne HDL-64E LIDAR. We only used the point clouds data and images from the left color camera to train our fusion model. The dataset provides 7481 samples for training and 7518 samples for testing. As standard good practice, we further split the KITTI training dataset into 3712 samples for training and 3769 samples for validation. We evaluated our model on the validation dataset following the easy, moderate, and hard difficulty classification levels officially introduced by KITTI. Specifically, in order to align the performance of the algorithms and cover most of the traffic scene scenarios, the object detection task is divided into three levels for validation and testing with respect to the different size, occlusion, and truncation level as shown in Table 1. Besides, the average precision (AP) metric is used when we compare our results with other different models.
B. TRAINING SETTINGS
We used the Adam [14] algorithm as our training optimizer. The batch size was set to 4 on a NVIDIA 1080Ti GPU. The learning rate was initially set to 0.002, and then was divided by 10 at 40 epochs. Our model has been trained for a total of 50 epochs.
C. RESULTS
We can firstly compare our model to the backbone network 3DSSD [51] on the validation dataset to show the effectiveness of our fusion strategy as shown in Table 3. The bottom line indicates the difference between our model and 3DSSD for 3D car detection. It shows that our model outperforms 3DSSD in all the categories and all the difficulty levels, which convincingly shows the efficiency of our RoI fusion method. As shown in Table 2, our model also achieves the best performance compared to recent state-of-the-art fusion methods on the test dataset. We choose moderate difficulty as the main average precision (AP) metric, and compare our model to BEV-image fusion MV3D [5] and AVOD [16], our deep fusion method outperforms all others by a large margin. For the frustum method, our method outperforms F-PointNet [28] by 9.12%. Besides, our model significantly outperforms point-pixel-wise fusion method PointRCNN [34] by 7.84%. We also visualize some examples for prediction results and corresponding ground truth as shown in Fig. 9 for better representation.
D. ABLATION STUDY
We carried out several ablation experiments to investigate the effectiveness of extended value for RoI size and different VOLUME 4, 2016 In order to cluster sufficient points around the objects, we enlarge the RoI size by an extended value eta for more contextual local features. Table 4 shows that the model achieves the best performance when η = 1.0. Besides, we notice that there is a significant drop of performance when no extended size (i.e. η = 0) is used, especially for hard difficulty level of detection. It is assumed that the larger size of the box also provides sufficient information, but is likely to involve more redundant and harmful information. In contrast, smaller size only could provide part information of the cars, which is insufficient to predict the parameters of the 3D bounding box. For the hard detection level objects, they normally are occluded by other objects or far away from the sensor, which leads to very few points on the objects. As a result, involving more surrounding points is beneficial to object classification and regression.
b: Effects of the fusion strategy
We further investigate the effectiveness for the different fusion strategies. In addition to the concatenation operation as described in Section III-B0c, we also employ operations, such as sum, max operation, and compare the results for different choices. As shown in Table 5, the concatenation operation for RoI features fusion achieves 91.36% 82.74% 80.22% performance for easy, moderate, and hard difficulties respectively. The results show that the concatenation operation could fuse more discriminative features from the 3D RoIs and corresponding 2D RoIs. This can be linked to the fact that both the sum operation and the max operation could obtain signature features, but the concatenation operation allows to keep all the features from different sensors, which then is likely to allow capturing more useful features for classification and regression. Qualitative results on the KITTI validation dataset. The predicted objects and the ground truth objects are shown in red and green bounding boxes respectively. We also project the bounding boxes to the RGB images for better visualization. As we know that LIDAR-only methods are not efficient to detect the objects that include few points, such as small objects and objects in the long distance. In order to better compare our fusion method to the LIDAR-only model 3DSSD [51] which only provide car detection model, we detect the cars in the various distance range for the different size of object, and then compare the recall and the accuracy to the 3DSSD. As illustrated in the Fig. 7 and the Fig. 8, when the objects are within 20 meters, our results are nearly no different to the 3DSSD. However, the performance becomes worse when the objects are beyond 20 meters for both models, but our fusion model performs much better than 3DSSD when detect the objects located in the long distance. The results convincingly prove that our fusion method successfully predicts more true positive objects in the long distance range by learning sufficient colour and texture information for the point clouds.
d: Inference time
We tested the inference time on KITTI validation dataset with a NVIDIA 1080Ti GPU, and then compare to existing fusion methods in Table 6. Our model achieves the best tradeoff compared to BEV-image fusion method AVOD [16] and frustum method F-PointNet [28]. We also note that our RoI fusion method is much better than point-pixel fusion method PointPainting [40] in terms of both accuracy and inference time.
E. QUANTITATIVE ANALYSIS
The comparison of our model with other multi-sensor methods as shown in Table 2, shows that our model is more efficient and effective. RoI fusion enables to globally fuse the local area from the point clouds and the images, which make it easier to align the viewpoint when we concatenate the 3D/2D RoI pooling features. In contrast, the point-pixel fusion is unlikely to obtain discriminative features due to the fact that the point clouds are sparse and irregular, but the images distribute as the standard grid. Compared to BEV-image fusion, our model outperforms others by a large margin due to the information losses during BEV generation.
VI. CONCLUSION
In this paper, we propose a novel deep fusion method, named RoIFusion, to efficiently fuse the point clouds and the images VOLUME 4, 2016 for 3D object detection. We build a lightweight neural network to generate 3D RoIs from the point clouds and 2D RoIs from the images, and then employ a 3D RoI pooling layer and a 2D RoI pooling layer to obtain the geometric features and the texture features respectively for a potential local area in both point clouds and the images. Finally, we fuse them together to predict the oriented 3D bounding box for the detected object. Our fusion method is flexible and could combine any other LIDAR-only segmentation networks and image segmentation networks. The state-of-the-art performance of our model convincingly show that the fusion method proposed can successfully boost the performance of 3D object detection. | 2020-09-11T01:00:30.250Z | 2020-09-09T00:00:00.000 | {
"year": 2020,
"sha1": "57cd7548aa931d2365de924b6b1de00d14fe9210",
"oa_license": "CCBY",
"oa_url": "https://ieeexplore.ieee.org/ielx7/6287639/9312710/09393330.pdf",
"oa_status": "GOLD",
"pdf_src": "Arxiv",
"pdf_hash": "57cd7548aa931d2365de924b6b1de00d14fe9210",
"s2fieldsofstudy": [
"Computer Science",
"Engineering",
"Environmental Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
255171639 | pes2o/s2orc | v3-fos-license | Caputo–Fabrizio fractional model of MHD second grade fluid with Newtonian heating and heat generation
In this research article the heat transfer of generalized second grade fluid is investigated with heat generation. The fluid flow is analyzed under the effects of Magneto hydrodynamics over an infinite vertical flat plate. The Newtonian heating phenomenon has been adopted at the boundary. For this purpose the problem is divided into two compartments i.e. momentum equation and energy equations. Some specific dimensionless parameters are defined to convert the model equations into dimensionless system of equations. The solutions for dimensionless energy and momentum equations are obtained by using the Laplace transform technique. From obtained results by neglecting magneto hydrodynamic effects and heat source some special case are achieved which are already published in literature. The case for which the fractional parameter approaches to the classical order is also discussed and it has been observed that it is convergent. Finally, the influences of different physical parameters are sketched graphically. It has been observed that for increasing values of Prandtl number the velocity and temperature decreases, for increasing values of Grashof number the velocity of the fluid increases. Also it has been investigated that for increasing values of fractional parameter the velocity and temperature of the fluid increases.
List of symbols u [LT −1 ] Fluid velocity T [θ]
Temperature t [T] Time C p [L 2 MT −1 θ −1 ] At constant pressure the specific heat g [LT −2 ] Acceleration due to gravity k [MLT −3 θ −1 ] Thermal conductivity or heat conduction of the fluid Kinematic viscosity ρ [ML −3 ] Fluid's density β T [θ −1 ] Coefficient of volumetric expansion of thermal T w [θ] At plate the fluid temperature T ∞ [θ] The fluid temperature away from the plate Gr Grashof number of thermal α 1 Second grade parameter (Dimensional) γ Second grade parameter (Dimensionaless) Pr Basically the non-Newtonian fluids are divided into three main groups according to their actions with shear stress i.e. integral type, differential type and rate-type. (1) The integral type fluids are those fluids, whose shear stress is www.nature.com/scientificreports/ hardly dependent upon the shear rate. (2) the fluids whose shear strain and shear rate are related to each other are called differential type. (3) those fluids which have the properties of viscosity and elasticity are known as rate type. Grade Second fluids belongs to differential type which is more famous among the various popular models of the non-Newtonian fluids. The pioneers who designed the second grade fluid model were Coleman and Noll 1 . Subsequently, this framework was used for the analysis of different problems whose construction is relatively simple. The second-order Rivlin-Erickson equations has been applied to explain the pattern of a non-Newtonian fluid flowing unsteadily upon a flat surface as like the Couette flow and Poiseuille flux 2 . Many authors have been studied some of the unsteady flows of the second grade fluid 3,4 . Derivatives are mostly used to formulate the real world problems into mathematical models. In particular, fractional derivatives are more suitable for some wellknown problems than regular derivative. In recent past the application of fractional order derivatives has been expanded in distinct fields. Especially, in dynamics of fluid, viscoelasticity, bioengineering, electrochemistry, finance, fluent currents tracers and in signal processing. Particularly, fractional derivatives are more suitable for some major problems than regular derivatives [5][6][7] . The fractional derivative models are used widely in different directions like, polymers for glass transition and glass state referable that the fractional derivative models can easily explains the complex behavior of a viscoelastic fluid [8][9][10][11] . Free convection flow of generalized viscous fluid upon a vertical plate with chemical and Newtonian heating is investigated in 12 . The fractional second grade fluid investigated by using the Caputo fractional derivatives 13 . In the near past Caputo and Fabrizio have presented a new fractional operator know as Caputo-Fabrizio operator, which has been used in many theoretical real word phenomena 14 . The generalized second grade fluid is investigated with Caputo-Fabrizio differential operation by adopting the Laplace transformation technique, and obtained the exact solutions to the problem 15 . Exact analytical for viscous fluid with non-singular kernal differential operator is gained 16 . Due to the rising concern of fractional derivative modeling many fractional models have been modeled using the existing models of fluid [17][18][19] . The convection heat-mass transfer of generalized grade second fluid is analyzed, and results achieved by Caputo-Fabrizio is compared Atangana Baleanu fractional operator 20 . The author analyzed heat transfer in convective flow of second grade fluid subjected to Newtonian heating using Atangana Baleanu fractional and Caputo Fabrizio fractional derivative. They also carried out the comparison of the two approaches 21 . The heat transfer in MHD flow of generalized second grade fluid with porosity in the medium by adopting Caputo Fabrizio derivative 22 . Heat transfer during the unsteady magneto hydrodynamic flow of a differential-type fluid in Forchhiemer medium was analyzed numerically 23 . The unsteady magneto hydrodynamic flow of viscoelastic fluid flowing in a porous medium 24 . Heat transfer during the incompressible time-dependent flow of Maxwell viscoelastic fluid by some stretching surface with chemical reaction and radiation source was investigated in 25 . The analysis of rate type anomalous Nano-fluid with Caputo non-integer order derivative flowing unsteadily was studied in 26 . The two-dimensional and two-directional MHD flow of fractional viscoelastic fluid was analyzed in 27 . The authors studied the unequal diffusivities of chemical species in a Forchhiemer medium by using Scott Blair model of viscoelastic fluid with unsteady convection in 28 . The author studied the effects of mixed convection with thermal radiation and chemical by using the space-time coupled Cattaneo-Friedrich Maxwell Model with Caputo fractional derivatives in a porous medium 29 . The authors used the fractional calculus to analyzed the thermo-diffusion phenomenon numerically in a Darcy medium 30 . Khan and Rasheed studied the numerical implementation and error analysis with variable heat flux of coupled non-linear fractional viscoelastic fluid in 31 . Mumtaz et al. 32 studied the computational simulation of viscoelastic model of Scott Blair to the hybrid fractional Nanofluid in a porous Darcy medium. The main aim of this article is to extend the application of Caputo Fabrizio fractional derivative to the second grade fluid with magneto hydrodynamic effects in addition to the heat generation. Also the considered Newtonian heating is adopted at the boundary. The exact analytical solution has been achieved by using Laplace transformation on the dimensionless equations of the problem with suitable initial and boundary conditions. From obtained results by neglecting magneto hydrodynamic and heat source some special case are achieved which are published in literature. The case for which the fractional parameter approaches to the classical order also discussed and it has been observed that it is convergent. The exact solution of the problem is represented graphically to visualize the effects of physical parameters like time fractional, Magneto hydrodynamic, Prandtl number, Eta and Grashof Number etc.
Mathematical analysis of the problem
Consider the second grade fluid of unsteady flow over in an infinite upright plate with Newtonian heating at the boundary, flow direction is x-axis and y-axis is perpendicular to the flat plate. When t= 0 both the fluid and the plate are at rest and the fluid temperature is T ∞ . But as time start i.e., for t ≻ 0 then temperature is and fluid velocity becomes u(0, t) = H(t)cosωt . The temperature and velocity are dependent on time t and y only. Now by usual Boussinesq's approximation 16 . the unsteady flow is governed by the following set of partial differential equations. The flow of the fluid is represented by the following governing equations. The schematic diagram used in fluid flow problem is represented geometrically by Fig. 1.
The initial (ICs) and boundary (BCs) conditions are
Dimensionless variables
The following dimensionless variables are utilized to gain a system of dimensionless governing equations from the set of dimensional governing equations.
Using these dimensionless variables given in Eq. (6) in Eqs. (1)-(2) and dropping out the star (*) notation, the governing Eqs. (1)-(2) take the simplest forms www.nature.com/scientificreports/ To find a time-fractional order derivative model just interchange the time derivative of classical order with the time derivative of order α ∈ [0, 1] , then as a result the following system of governing equations come into being: The appropriate non-dimensional initial and boundary conditions are The fractional operator used in this problem is Caputo-Fabrizio which is defined as under in (14) for α ∈ [0, 1],
Numerical results and discussion
By using Mathcad software different physical parameters were drawn to analyze the effects of fluid velocity and temperature. The parameter Alpha α in Fig. 2, Eta η 1 in Fig. 3, and Prandtl number Pr in Fig. 4 are sketched for temperature field, while for velocity field the Alpha α in Fig. 5, Eta η 1 in Fig. 6, Grashof number Gr in Fig. 7, ϕ(y, t; Pr γ , αγ ) − ϕ(y, t; a 1 , a 2 ) dτ www.nature.com/scientificreports/ Magneto hydrodynamic MHD in Fig. 8 and Prandtl number Pr in Fig. 9 are presented with different values of time t. Figure 2 is drawn to show the effects of alpha α for temperature profile in which it is observed that by increasing the value of alpha α, the temperature is also increases. In this way the consistency of thermal boundary layer is increases with the parameter alpha α and time t. Figure 3 is sketched to check the effect eta η 1 for temperature profile in which it is observed that by increasing the value of eta η 1 the temperature decreases, the consistency of thermal boundary layer also decreases with the parameter eta η 1 and time t. Figure 4 is sketched to examine the effects of the Prandtl number Pr in which it is noticed that by increasing the values of the parameter Pr, the temperature profile decreases, as Prandtl number is the ratio of momentum diffusivity to thermal conductivity by increasing the Prandtl number thermal conductivity decreases which cause to decrease the temperature of the fluid. The Fig. 5 is drawn to examine the effects of fractional parameter alpha α on the velocity profile of the fluid, and it is concluded that the velocity of the fluid increases with increasing values of fractional parameter alpha α . Figure 6 is drawn to show the effects of η 1 on fluid velocity. From Fig. 6 it is noticed that velocity of the fluid have inverse relation with the parameter eta η 1 , the velocity of the fluid decreases with the increasing values of η 1 . Figure 7 is sketched to examine the effect of Grashof Gr number for velocity profile, here it is noticed that by increasing the value of Grashof number the fluid velocity is increases, because Grashof number is the ratio of www.nature.com/scientificreports/ inertia to viscous force, Grashof number is inversely proportional to viscous force, so increase in Grashof number cause decrease in viscosity. It is obvious that for low viscosity the velocity is higher. That is why for increasing values of Grashof number the fluid velocity increases. Figure 8 is drawn to show the effects of magneto hydrodynamic MHD on fluid velocity, here it is noticed that by increasing the value of Magneto hydrodynamic the motion of fluid is decreases. Figure 9 are sketched to examine the effect of Prandtl number Pr on velocity fluid, where we noticed that by increasing the value of Prandtl number, the velocity of the fluid is decreases. Figure 10 is shown in comparison with published result obtained in 16 . Figure 11 is shown for α=1. From obtained results by neglecting magneto hydrodynamic and heat source some special case are achieved which are published in literature published by Shah and Khan in 16 . The case for which the fractional parameter approaches to the classical order also discussed and it has been observed that it is convergent. www.nature.com/scientificreports/
Conclusion
The considered study is about analyze the unsteady natural convection flow of generalized second grade fluid with magneto hydro dynamic effects and Newtonian heating in addition to heat generation. Some special cases of the obtained solution are discussed from which some well-known results are found in the published literature which are similar to published in 16 www.nature.com/scientificreports/
Data availability
The datasets used and analyzed during the current study available from the corresponding author on reasonable request. www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. | 2022-12-28T16:09:59.779Z | 2022-12-01T00:00:00.000 | {
"year": 2022,
"sha1": "3be035c98b894523bf4031a2a5ae3a94c9c48c92",
"oa_license": "CCBY",
"oa_url": "https://www.nature.com/articles/s41598-022-26080-7.pdf",
"oa_status": "GOLD",
"pdf_src": "Springer",
"pdf_hash": "404393da5ffef0fd2c340097208d6c84315703c9",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": []
} |
119715517 | pes2o/s2orc | v3-fos-license | Split injectivity of A-theoretic assembly maps
We construct an equivariant coarse homology theory arising from the algebraic $K$-theory of spherical group rings and use this theory to derive split injectivity results for associated assembly maps. On the way, we prove that the fundamental structural theorems for Waldhausen's algebraic $K$-theory functor carry over to its nonconnective counterpart defined by Blumberg--Gepner--Tabuada.
Introduction
For a group G, let P be the total space of a principal G-bundle and let A denote the functor of nonconnective A-theory (taking values in the ∞-category of spectra). Then P gives rise to an Or(G)-spectrum A P sending a transitive G-set S to the spectrum A(P × G S). We will show the following split injectivity results for A P .
Theorem. Let G be a group and assume
(1) G is finitely generated; (2) G admits a finite-dimensional model for EG; (3) one of: Similar results about the K-and L-theoretic assembly maps for discrete group rings were originally obtained by Carlsson-Pedersen [CP95] and subsequently generalized by Bartels and Rosenthal [BR07], Ramras, Tessera and Yu [RTY14] and Kasprowski [Kas15]. The analog of Theorem 1.2 in the case of discrete group rings is originally due to Bartels [Bar03].
More precisely, we will show in Theorem 5.17 that A P is a (hereditary) CPfunctor, a notion introduced in [BEKWb]. We then apply results from [BEKWb] in order to deduce the above theorems in Section 5.4.
Recall from [BEKWb,Definition 1.5] that the family of subgroups FDC consists of those subgroups H of G such that the family {F \H | F ≤ H finite} has finite decomposition complexity as defined by Guentner, Tessera and Yu in [GTY12,GTY13].
For more details on this and a slightly more general result see [BEKWb, Theorem 1.11]. We also obtain the following A-theoretic analog of [BEKWb, Theorem 1.15].
1.4. Theorem. We assume the following: (1) G admits a finite-dimensional model for EG; (2) G is relatively hyperbolic to groups H 1 , . . . , H n ; is an equivalence. Then the A-theoretic assembly map for the family of finite subgroups of G α Fin AP : colim Finally, our methods imply the following result which was previously obtained by Barwick 1.5. Theorem. If G is a finite group, then the Or(G)-spectrum A P extends to a spectral Mackey functor.
To prove Theorem 5.17, we recast the definitions of coarse versions of A-theory given by Weiss [Wei02] and Ullmann and Winges [UW] in the setting of bornological coarse spaces. Since this construction relies on a sufficiently well-behaved version of nonconnective algebraic K-theory, Section 2 discusses the properties of the nonconnective K-theory functor introduced by Blumberg-Gepner-Tabuada [BGT13] as a functor on Waldhausen categories. Specifically, we show that the validity of the Additvity, Fibration, Approximation and Cofinality theorems are preserved in passing to the nonconnective version.
The necessary translation of the categories of controlled retractive spaces from [UW] to the setting of bornological coarse spaces is done in Section 3. Using the results from Section 2, we are then able to give a streamlined proof of the fact that coarse A-homology is a coarse homology theory in Section 4.
The final Section 5 establishes the last properties needed to obtain Theorem 5.17. In particular, we discuss the construction of transfer maps.
Algebraic K-theory
The algebraic K-theory functor originally defined by Waldhausen [Wal85] takes values in connective spectra. For the applications in Section 4 and Section 5, we require a nonconnective version of algebraic K-theory. A nonconnective version of A-theory can be derived from Waldhausen's connective K-theory functor using the methods of Sections 3 and 4, cf. [UW,Section 5]. However, we prefer to base our discussion on the axiomatic approach of Blumberg-Gepner-Tabuada [BGT13].
Waldhausen's K-theory functor is particularly useful to prove structural results. The key tools in Waldhausen's approach are the Additivity theorem [Wal85, Theorem 1.4.2], the Fibration theorem [Wal85, Theorem 1.6.4], the Approximation theorem [Wal85, Theorem 1.6.7] and the Cofinality theorem [TT90,1.10.1], [Vog90,Theorem 1.6]. In the present section we provide analogs of these theorems for the universal nonconnective K-theory of Blumberg-Gepner-Tabuada: the Additivity theorem (Corollary 2.36) holds as a corollary of the Fibration theorem (Theorem 2.35); the appropriate analog of the Approximation theorem is recalled in Theorem 2.16, and a version of the Cofinality theorem is given in Theorem 2.30.
We also discuss the compatibility of K-theory with infinite products-the connective case was originally established by Carlsson [Car95]-in Section 2.3.
2.1. Waldhausen categories as right-exact ∞-categories. In order to be able to employ the theory developed in [BGT13], we consider a class of Waldhausen categories (called homotopical) whose homotopy theory can be adequately described in terms of ∞-categories.
2.1. Definition. Let C be a Waldhausen category.
(1) C admits factorizations if every morphism in C can be factored into a cofibration followed by a weak equivalence; we assume no functoriality.
(2) C is homotopical if it admits factorizations and the weak equivalences satisfy the two-out-of-six property.
Recall that the two-out-of-six property means that if are composable morphisms such that both x 2 •x 1 and x 3 •x 2 are weak equivalences, then also x 1 , x 2 and x 3 (and hence also x 3 • x 2 • x 1 ) are weak equivalences.
2.2.
Remark. Homotopical Waldhausen categories as defined above are precisely the Waldhausen categories considered in [BGT13]. The term "homotopical" has been borrowed from [DHKS04,Chapter 5]. Let us comment shortly why it is sensible to restrict to this class of Waldhausen categories.
The existence of factorizations in C is the most natural condition to guarantee that the ∞-categorical localization C[wC −1 ] admits all finite colimits, which makes it amenable to the methods of [BGT13]; see also [BGT13,Definition 9.30].
Moreover, note that equivalences in any ∞-category satisfy the two-out-of-six property. In order to ensure that notions defined in terms of Waldhausen categories match up with their ∞-categorical counterparts, it is natural to require that the localization C → C[wC −1 ] detects weak equivalences. By [Cis,Corollary 7.5.19], this is the case if and only if the weak equivalences in C satisfy the two-out-of-six property, see also Proposition 2.8(1). See also [BM11, Lemma A.2.3, Theorem B.5.1 and Theorem 6.4] and [Wei99] for similar observations with respect to the hammock localization.
Let Wald ho denote the category of homotopical Waldhausen categories and exact functors. As an auxiliary tool, we introduce the category RelCat of relative categories and functors between relative categories. There is a forgetful functor u : Wald ho → RelCat, (C, co C, wC) → (C, wC).
Composing with the functor RelCat → Cat ∞ which sends a relative category to its localization, u induces a functor 2.4. Remark. In this remark we provide a point-set model for the functor (2.3). We model ∞-categories by quasi-categories and write N hc (C) for the ∞-category represented by the homotopy-coherent nerve of a fibrant simplicial category C. In order to apply this to an ordinary category C we consider it as a simplicial category with discrete mapping spaces. By abuse of notation, we also use C to denote N hc (C) for an ordinary category C in the main body of the paper.
as the homotopy coherent nerve of the subcategory of cofibrant-fibrant objects in sSet + .
Let C be an ∞-category and let wC be a wide subcategory. Recall that a localization of C at wC is a functor l : C → C[wC −1 ] such that restriction along l defines for every ∞-category D an equivalence of ∞-categories where Fun wC denotes the full subcategory of functors sending all morphisms in wC to equivalences. Localizations always exist and are essentially unique [Cis, Proposition 7.1.3]. Note that l * restricts to an equivalence on maximal Kan complexes, and thus induces an equivalence where Map wC Cat∞ denotes the collection of all components containing functors which send all morphisms in wC to equivalences. It follows formally that the localization l is equivalently described by the latter universal property. Consequently, any fibrant replacement of (C, wC) as a marked simplicial set models the ∞-categorical localization l : Composing u with the functor and any fibrant replacement functor R : sSet + → sSet + produces a concrete model by the previous discussion. . A functor C → D is exact if it preserves zero objects, sends pushouts along cofibrations in C to pushouts in D, and maps all morphisms in wC to equivalences in D.
Proof. The localization functor l is exact by the dual of [Cis, Proposition 7.5.6].
If l(f ) is an equivalence for some morphism f : X → Y in C, l(f ) defines an isomorphism in the homotopy category of ℓ(C). Applying [Cis, Corollary 7.5.19] twice, it follows that there exist morphisms g : Y → X and f ′ : X → Y such that both f g and gf ′ are weak equivalences in C. By the two-out-of-six property, it follows that f is a weak equivalence. This proves part (1).
Part (2) The goal of the present section is to show that U W loc satisfies the obvious analogs of the fundamental theorems of connective Waldhausen K-theory. For the most part, these are direct consequences of localization results from [ In the following argument we use the explicit model R • L • u of the functor ℓ ′ given in Remark 2.4. Then we must check that the natural morphism (2.12) colim is a weak equivalence in the marked model category structure on sSet + , where Q is the cofibrant replacement in the projective model category structure on Fun(I, sSet + ), and the subscript (−) I stands for point-wise application of the corresponding functor. By inspection, u : Wald ho → RelCat preserves filtered colimits, and so does L : RelCat → sSet + because the nerve preserves filtered colimits. Consequently we have a factorization (2.13) colim of (2.12). We now consider the following commutative diagram We now again use that marked equivalences are preserved under filtered colimits in order to deduce that the first arrow is a weak equivalence, too.
2.14. Corollary. The functor U W loc commutes with filtered colimits. Proof. U loc commutes with filtered colimits by definition, and Stab commutes with filtered colimits since it is a left adjoint. Hence the corollary is a consequence of Proposition 2.11.
Let f : C → D be an exact functor between homotopical Waldhausen categories.
2.15.
Definition. The functor f : C → D satisfies the approximation property if the following conditions are satisfied: (1) f preserves and detects weak equivalences; (2) For every object C of C and morphism y : Recall the functor ℓ from (2.6). Proof. This is a special case of the dual of [Cis, Proposition 7.6.15].
As a first application of Theorem 2.16, we establish a criterion to decide whether full inclusions of homotopical Waldhausen categories induce a fully faithful functor in Cat Rex ∞ . This is the first step in proving the Cofinality theorem, which is our next goal.
2.17. Definition. A subcategory C ⊆ D of the homotopical Waldhausen category D is a homotopical Waldhausen subcategory if (C, C ∩ co D, C ∩ wD) is also a homotopical Waldhausen category.
2.18. Remark. Note that being a homotopical Waldhausen subcategory is a stronger assumption than being a 'subcategory with cofibrations and weak equivalences' in the sense of [Wal85, page 321 and 327] which is in addition homotopical: If C ⊆ D is a subcategory with cofibrations, then co C ⊆ C ∩ co D, whereas Definition 2.17 requires co C = C ∩ co D.
2.19. Definition. The inclusion C ⊆ D admits a mapping cylinder argument if for every morphism C → D in D such that C belongs to C and D is the target of a weak equivalence from an object of C there exists a factorization Note that the weak equivalence C ′ ∼ → D is not necessarily the one in the assumption on D. A similar condition has been considered by Sagave [Sag04, Theorem 3.1].
2.20. Remark. The name of the property defined in Definition 2.19 comes from the observation that this property is typically verified using a mapping cylinder construction, see for example the proof of Corollary 4.16.
2.21. Remark. If C ⊆ D admits a mapping cylinder argument, then the following seemingly stronger condition is also satisfied: Every diagram in D, in which C 1 and C 2 are objects in C, can be extended to a commutative diagram of the shape in which C ′ is an object in C. This follows from Definition 2.19 by applying it to the morphism C 1 ∨ C 2 y+w − −− → D and rewriting the resulting diagram Let C ⊆ D be a full homotopical Waldhausen subcategory of the homotopical Waldhausen category D. Recall the functor ℓ from (2.6).
2.23. Proposition. If the inclusion C ⊆ D admits a mapping cylinder argument, then ℓ(C) → ℓ(D) is fully faithful.
Proof. The proof consists of the following steps: (1) We consider the saturation C of C in D defined as the full subcategory consisting of all objects of D which are connected by a sequence of zig-zags of weak equivalences with an object of C. We first show in Lemma 2.25 that C is a homotopical Waldhausen subcategory of D.
(2) We then show in Lemma 2.26, using Theorem 2.16, that the inclusion C → C induces an equivalence ℓ(C) (3) Finally, we show in Lemma 2.27 that the functor ℓ(C) → ℓ(D) induced by the inclusion C → D is fully faithful. This implies the assertion since ℓ(C) → ℓ(D) is the composition of an equivalence and a fully faithful functor.
As a preparation, we show that a sequence of zig-zags of weak equivalences in D can be reduced to a single inward-pointing one.
2.24. Lemma. If two objects D and D ′ in D are connected by a sequence of zig-zags of weak equivalences, then there exists a zig-zag of weak equivalences Proof. Cf. [BM11, Lemma 5.10] for a very similar argument. We claim that we can replace outward pointing zig-zags of weak equivalences by inward-pointing zig-zags of weak equivalences. If D and D ′ in D are connected by a sequence of zig-zags of weak equivalences, then we can apply the claim repeatedly to parts of the sequences and finally compose the maps in order to obtain the desired single inward-pointing zig-zag of weak equivalences connecting D and D ′ .
We now show the claim. We consider an outward pointing zig-zag of weak equivalences in D. Then we choose a factorization of the morphism D ∨ D ′′ id +x − −− → D into a cofibration followed by a weak equivalence. We have a commuting diagram : : D which shows that the morphism y is a cofibration and a weak equivalence. Similarly, the commuting diagram shows that z is a cofibration and a weak equivalence. Since z is cofibration, we can form a push-out square The lower horizontal arrow is a weak equivalence because it is the pushout of a weak equivalence. We now see that is the desired inward-pointing zig-zag of weak equivalences by composing the first two morphisms.
Let C be the saturation of C in D, i.e. the full subcategory consisting of all objects of D which are connected by a sequence of zig-zags of weak equivalences with an object of C.
2.25.
Lemma. If C ⊆ D admits a mapping cylinder argument, then C is a full homotopical Waldhausen subcategory of D.
Proof. Since C obviously admits factorizations, we must only show that C is closed under pushouts along cofibrations.
Let the diagram C 2 ← C 0 C 1 in C be given. In the following, we describe step by step how to construct the diagram The first column is the original span.
The second column is obtained from the first by factoring the morphism C 0 → C 2 into a cofibration followed by a weak equivalence. By Lemma 2.24, there exists a zig-zag Taking the indicated pushouts produces the third column of the diagram. The fourth column arises from the third by factoring the morphisms C 0 → C 1 ⊔ C0 D 0 and C 0 → C ′ 2 ⊔ C0 D 0 into a cofibration followed by a weak equivalence. Again using Lemma 2.24 and the fact that C ′′ 2 ∈ C, we choose a zig-zag The fifth column arises canonically by composing existing morphisms. For the next column, we use that C ⊆ D admits a mapping cylinder argument (note that D 2 is the target of a weak equivalence from the object C of C) in order to factor the morphism C 0 → D 2 as The final column is then obtained by factoring the morphism C 0 → C ′ 2 in C into a cofibration followed by a weak equivalence; in particular, we have C 2 ∈ C.
Note that the final column has the same shape as the fourth column, with the difference that C 0 and C 2 both lie in C. Repeating the argument that gave the right half of the diagram with C ′ 1 in place of C ′′ 2 finally yields a zig-zag of weak equivalences between the original span and a span in C. By virtue of the gluing axiom in a Waldhausen category, it follows that C 2 ⊔ C0 C 1 lies in C.
2.26.
Lemma. If C ⊆ D admits a mapping cylinder argument, then the inclusion C → C induces an equivalence ℓ(C) → ℓ(C).
Proof. We verify that the assumptions of Theorem 2.16 are satisfied. The saturation (C, C ∩ co D, C ∩ wD) of C in D is a homotopical Waldhausen category by Lemma 2.25. It remains to show that the inclusion C → C has the approximation property, see Definition 2.15.
The inclusion of course preserves and detects equivalences. In order to verify the second condition 2.15.2, we consider a morphism y : C → C with C ∈ C and C ∈ C. By Lemma 2.24, we find a zig-zag of weak equivalences in In particular, C ′ is the target of a weak equivalence from an object of C and we can use that C ⊆ D admits a mapping cylinder argument in order to factorize the composed map This determines the desired commutative square Lemma. If C ⊆ D admits a mapping cylinder argument, then the inclusion C → D induces a fully faithful functor ℓ(C) → ℓ(D).
Proof. C is a homotopical Waldhausen category by Lemma 2.25.
Noting that the opposite of a homotopical Waldhausen category is a category with weak equivalences and fibrations, we apply dualized versions of the results in [Cis,Chapter 7].
[Cis, Example 7.6.4] shows that there exists a calculus of fractions at each object of D. This allows us to apply [Cis, Corollary 7.2.9]: Let l : D → ℓ(D) be the localization functor. For two objects D 1 and D 2 in D, we have a canonical equivalence where the colimit is indexed by the full subcategory W D (D 2 ) of the comma category D 2 /D spanned by the weak equivalences with domain D 2 . Applying the same formula for C, the inclusion functor C → D induces for every two objects C 1 , C 2 in C the canonical map Since C is a full subcategory of D which is closed under weak equivalences, both sides are colimits of the same diagram. Hence ℓ(C) → ℓ(D) is fully faithful.
This finishes the proof of Proposition 2.23. Recall the idempotent completion functor
Lemma. The idempotent completion functor induces functors
Proof. We begin by showing that Idem(C) is right-exact for any C ∈ Cat Rex ∞ . The argument is very similar to the proof of [Lur, Corollary 1.1.3.7], but easier.
By [Lur09, Lemma 5.4.2.4], Idem(C) is equivalent to the full subcategory of κcompact objects in Ind κ (C). Since zero objects are compact and the collection of κ-compact objects is closed under finite colimits, this shows that Idem(C) is right-exact and that the canonical functor C → Idem(C) is exact.
Hence the idempotent completion functor restricts to a functor The universal property of Idem(C) implies that the canonical functor C → Idem(C) induces an equivalence Let C ⊆ D be a full homotopical Waldhausen subcategory and let D be an object of D.
We further say that D is dominated by C if every object of D is dominated by C.
Theorem (Cofinality theorem). Let C ⊆ D be the inclusion of a full homotopical Waldhausen subcategory. Suppose that
(1) the inclusion C ⊆ D admits a mapping cylinder argument, see Definition 2.19; (2) D is dominated by C, see Definition 2.29.
2.31. Lemma. The stabilization functor Stab : Cat Rex ∞ → Cat ex ∞ restricts to a functor Proof. Let C be an idempotent complete, right-exact ∞-category. Since Idem Rex is a left adjoint functor, formula (2.9) implies so Stab preserves idempotent completeness as claimed.
Corollary. There is a canonical equivalence
Proof. Since the various inclusion functors form a commutative square To conclude, we discuss the Fibration theorem. The Additivity theorem will follow from this as a corollary.
Proof. We use the colimit description (2.9) of Stab(C) and Stab(D), respectively. Let i n : C → Stab(C) and j n : D → Stab(D) denote the n-th structure morphism of the colimit systems. For any two objects X and Y of Stab(C), there exist some natural number n and objects X, Y ∈ C satisfying X ≃ i n (X) and Y ≃ i n (Y ).
Since Stab(f )(X) ≃ j n (f (X)) and using that f is exact and fully faithful, we have So Stab(f ) is fully faithful.
2.35. Theorem (Fibration theorem). Let (C, vC) and (C, wC) be homotopical Waldhausen categories having the same underlying category with cofibrations such that vC ⊆ wC. Denote by C w the full subcategory of C spanned by those objects which are w-equivalent to zero.
Since vC ⊆ wC, we have a commutative diagram in which l v and l w are the respective localization functors, Σ ∞ v and Σ ∞ w are instances of the counit Σ ∞ , and p and p s are induced by the identity on C.
Denote by C 0 ⊆ C the full subcategory spanned by those objects whose image under Σ ∞ w • l w is zero. Since l w is exact by Proposition 2.8(1), C 0 inherits Waldhausen structures (C 0 , vC 0 := C 0 ∩ vC) and (C 0 , wC 0 := C 0 ∩ wC). As above, we abbreviate C 0,v := ℓ(C 0 , vC 0 ), and similarly for C 0,w , C s 0,v and C s 0,w . Since C 0 is closed under w-equivalences and vC ⊆ wC, the inclusions (C 0 , vC 0 ) ⊆ (C, vC) and (C 0 , wC 0 ) ⊆ (C, wC) both admit a mapping cylinder argument (see Definition 2.19). It follows from Proposition 2.23 that the induced functors are also fully faithful by Lemma 2.34. Recall from [NS,Theorem I.3.3] the Verdier localization of stable ∞-categories: Given a stable ∞-category C and a full stable subcategory D, the localization C/D of C at all morphisms whose cofiber lies in D is a stable ∞-category. For any stable ∞-category E, restriction along the localization functor C → C/D induces an , where the right hand side denotes the full subcategory of those exact functors which vanish on D.
By the universal property, p s induces an exact functor on Verdier localizations. We show that p is an equivalence by constructing an inverse equivalence q.
Consider the functor Since a morphism in a stable ∞-category is an equivalence if and only if its cofiber vanishes, this proves that L v sends all morphisms in wC to equivalences. By the universal property of localization and stabilization, there exists an induced functor q s : Since the functor C s 0,v → C s 0,w is essentially surjective, q s vanishes on C s 0,w and hence induces an exact functor To verify that pq ≃ id also holds, we show that L w : C → C s w /C s 0,w , which is defined analogously to L v , enjoys a universal property. Let D be a stable ∞-category.
By the universal property of the Verdier localization, we have an equivalence Let f : C w → D be a functor vanishing on C 0,w , and let f : Using the colimit description (2.9) of the stabilization, there exists some natural number n such that X ≃ i n (X) for some X ∈ C s 0,w , where i n : C 0,w → C s 0,w is the n-th structure map of the colimit. By assumption, Σ n X gets mapped to zero in D. Since Σ is an equivalence on C s 0,w , it follows that f (X) ≃ 0, so f vanishes on C s 0,w . This proves that the functor (Σ ∞ w ) * is an equivalence. Finally, restriction along l w induces a functor denotes the full subcategory of Fun ex (C, D) containing those functors which send all objects in C 0 to zero (see Definition 2.7 for the notion of exactness of a functor C → D). Since the localization C 0 → C 0,w is essentially surjective, l * w is also an equivalence. Consequently, L w induces an equivalence Choosing D = C s w /C s 0,w and tracing through the definitions, we find that pq corresponds to L w under this equivalence. Therefore, pq ≃ id, so p is an equivalence. Since we have a commutative diagram the lower line is a Verdier sequence. Consider now the Verdier sequence By definition, the first map in this sequence is zero, so C s 0,w ≃ 0. It follows that C s w → C s w /C s 0,w is an equivalence, and hence Observe that C w ⊆ C 0 . We next claim that the induced functor is an equivalence. Since C w is closed under v-equivalences, the inclusion C w ⊆ C 0 admits a mapping cylinder argument. It follows from Proposition 2.23 and Lemma 2.34 that Stab(ℓ(C w , vC w )) → C s 0,v is fully faithful. Let X be an object in C 0 , which is equivalent to saying that Σ ∞ w (l w ( * → X)) is an equivalence. Using the colimit description (2.9) of stabilization, this implies that 0 → Σ n l w (X) is an equivalence in C w for some natural number n. Since p is exact, we have p(Σ n l v (X)) ≃ Σ n l w (X) in C w . As l v is essentially surjective, there exists some Y ∈ C such that l v (Y ) ≃ Σ n l v (X). In particular, p(0 → l v (Y )) is an equivalence. Since p • l v ≃ l w and l w detects weak equivalences by Proposition 2.8(1), we conclude that Y is weakly contractible with respect to wC, i.e. Y ∈ C w . Therefore, some iterated suspension of each object in C 0,v lies in the essential image of the functor ℓ(C w , vC w ) → ℓ(C 0 , vC 0 ). Hence Stab(ℓ(C w , vC w )) → Stab(ℓ(C 0 , vC 0 )) is essentially surjective.
We conclude that is a Verdier sequence.
induces an equivalence Consider the subcategory of weak equivalences w q S 2 C consisting of those morphisms whose image under q is a weak equivalence. By Theorem 2.35, we obtain a fiber sequence . The projection functor q : S 2 C → C admits the section If we consider the Waldhausen structure (S 2 C, w q ), this section is an inverse up to weak equivalence, so q induces an equivalence U W loc (S 2 C, w q ) ∼ − → U W loc (C). An object X Y ։ Z lies in S 2 C wq precisely if Z is weakly contractible. Hence, the functor i : C → S 2 C wq sending X to X X ։ * is a right-inverse to s| S2C wq and a left-inverse up to weak equivalence. It follows that i induces an equivalence U W loc (C) ∼ − → U W loc (S 2 C wq ). Therefore, we obtain a fiber sequence . Since i is split by s and M loc is stable, the claim follows.
2.3. Algebraic K-theory and infinite products. Since K is a colimit-preserving functor [BGT13, Theorem 9.8], all structural results established about U W loc in Section 2.2 carry over for K W . We refrain from stating them explicitly.
To finish our discussion of the general properties of algebraic K-theory of homotopical Waldhausen categories, we address one of its more exotic properties, namely its compatibility with infinite products. This was originally shown for connective algebraic K-theory by Carlsson in the setting of Waldhausen categories with a cylinder functor [Car95].
2.38. Theorem. Let (C i ) i∈I be a family of homotopical Waldhausen categories. Then the canonical map is an equivalence.
At the time of writing, we are not aware of an analogous statement being true for U W loc . We derive Theorem 2.38 from the analogous statement for stable ∞-categories [KW, Theorem 1.3].
2.39. Lemma. Let (C i ) i∈I be a family of right-exact ∞-categories. Then the canonical map is an equivalence.
2.40. Remark. Note that the canonical functor Stab( i∈I C i ) → i∈I Stab(C i ) is not essentially surjective, and hence not an equivalence. The claim of Lemma 2.39 boils down to the assertion that any sequence (Σ 2ni ) i∈I of iterated suspension functors induces the identity after applying U loc . The standard argument to show that Σ 2n induces the identity map after applying U loc requires an increasing number of applications of additivity as n grows, and thus runs into problems for such an infinite sequence.
However, there is a slight variation of the argument which only requires a fixed number of applications of additivity, regardless of n. For example, taking the coproduct of the cofiber sequences gives the cofiber sequence Similarly, we have a cofiber sequence Noting that the two cofiber sequences differ only in the second term, and that the second term is given by id ⊔S and Σ 4 ⊔ S for some endofunctor S, respectively, it follows that K(id) ≃ K(Σ 4 ).
Proof of Lemma 2.39. Recall that we have an equivalence for every right-exact ∞-category C. In particular, i∈I Stab(C i ) admits the following description: Let N denote the set of natural numbers, and let N I be the set of functions I → N, equipped with the partial ordering such that α ≤ β if and only if α(i) ≤ β(i) for all i ∈ I. Then for the functor D : N I → Cat Rex ∞ which satisfies D(α) ≃ i∈I C i for all α, and sends α ≤ β to the functor i∈I Σ β(i)−α(i) : Since i∈I Stab(C i ) ≃ Stab( i∈I Stab(C i )), it suffices to consider the map As Stab commutes with filtered colimits in Cat Rex ∞ and U loc commutes with filtered colimits in Cat ex ∞ , the canonical map is an equivalence, and the map we are interested in corresponds to the structural inclusion U loc (Stab(D(0))) → colim where 0 denotes the constant map 0 : I → N.
To prove the claim, it is enough to show that all maps in the diagram U loc (Stab(D)) are equivalences. Let (n i ) i∈I be an arbitrary sequence of even natural numbers. Consider the endofunctor
Then we have a cofiber sequence of exact functors
in which the first transformation is given by (id and 0 alternate). Moreover, there also exists a cofiber sequence in which the first transformation is given by . . ⊕ Σ ni−1 i∈I . By virtue of the Additivity theorem, we have U loc (id ⊕S) ≃ U loc ((Σ ni ) i∈I ⊕ S), and conclude that U loc (id) ≃ U loc ((Σ ni ) i∈I ).
Since the subset of all functions I → 2N is cofinal in N I , this suffices to show that the diagram U loc (Stab(D)) is essentially constant, and thus proves our claim.
Proof of Theorem 2.38. Unravelling the definition of the functor K W , we can factor the comparison map as The first map is an equivalence by [
Controlled retractive spaces over a bornological coarse space
To produce a coarse variant of A-theory, we have to transfer the notion of controlled retractive spaces from [Wei02] and [UW] to the setting of bornological coarse spaces. While this is relatively straightforward, we try to make our treatment selfcontained (modulo the terminology introduced in [BE, Section 2] and [BEKWa, Sections 2 and 3], which we will use freely throughout). The main deviation from [UW] in our treatment lies in the proof of the gluing lemma for controlled equivalences. 3.3. Definition. Let X be a G-set. An X-labeling of K is a G-equivariant function Let (X, U) be a G-coarse space. Let (K, λ K ) and (L, λ L ) be X-labeled G-CWcomplexes relative W , and let ϕ : K → L be a G-equivariant and cellular map relative W .
3.4. Definition. The map ϕ is (X, U)-controlled if there exists an entourage U in U such that {(λ L (e ′ ), λ K (e)) | e ∈ ⋄ K, e ′ ∈ ⋄ ϕ(e) } ⊆ U. If ϕ is the identity map on K, we say that (K, λ K ) is an (X, U)-controlled G-CWcomplex (or simply controlled G-CW-complex if the G-coarse space (X, U) is clear from context).
We denote by C G (X, U; W ) the category of (X, U)-controlled G-CW-complexes relative W , and (X, U)-controlled, G-equivariant and cellular maps.
3.5. Remark. Definition 3.4 requires maps to be uniformly controlled by a single entourage, whereas [UW, Definition 2.3] enforces this condition only on each skeleton. See Remark 3.24 for further discussion.
Definition. A subcomplex inclusion is a morphism of the form
in C G (X, U; W ) such that j is a subcomplex inclusion.
Proof. Let L ⊔ K ′ K be the usual pushout in G-CW-complexes. Note that L ⊔ K ′ K arises from L by successively attaching cells from K which do not lie in K ′ , so there is a canonical identification We define a labeling on L ⊔ K ′ K by The universal property is easy to verify.
3.8. Definition. Let (K, λ) be an object in C G (X, U; W ), and let L be a G-CWcomplex. Define (K, λ) ⊗ L as the G-CW-complex given by the pushout The product ⊗ of Definition 3.8 defines a functor The following proposition summarizes some properties of the functor ⊗.
(3) The pushout-product-axiom holds: The induced map This follows from standard considerations about CW-complexes, and we omit the proof.
3.10. Lemma. Let (K ′ , λ K | K ′ ) ֒→ (K, λ K ) be a subcomplex inclusion, and suppose that L ′ ⊆ L is an inclusion of CW-complexes which is also a homotopy equivalence. Then for all k > 0, one argues cell by cell to show that there exists a deformation retraction of sk k (K, K ′ )⊗L onto (sk k−1 (K, K ′ )⊗L) ⊔ sk k−1 (K,K ′ )⊗L ′ (sk k (K, K ′ )⊗L ′ ), where sk k (K, K ′ ) denotes the relative k-skeleton of the pair (K, K ′ ). On each cell e, the trace of this deformation retraction is contained in the subcomplex generated by e. Since K is a controlled G-CW-complex, this proves that the deformation retraction is a controlled homotopy.
The desired deformation retraction on the whole complex K is then obtained by stacking these homotopies. It is controlled since the control of K is measured by a single entourage.
3.11. Corollary (Controlled homotopy extension property). If (K ′ , λ K | K ′ ) ֒→ (K, λ K ) is a subcomplex inclusion, then any controlled map extends to a controlled map is a Kan fibration for any controlled G-CW-complex (L, λ L ).
Proof. We have to show that the lifting problem always has a solution. By definition, the given lifting problem corresponds to the extension problem Since the inclusion |Λ n i | ⊆ |∆ n | is a homotopy equivalence, the claim follows from Lemma 3.10.
3.13. Corollary. C G (X, U; W ) is enriched in Kan complexes.
In the next step, we introduce a variant of the simplicial enrichment we have just discussed.
3.14. Definition. Let Y ⊆ X be a G-invariant subset of X and let (K, λ) be a controlled G-CW-complex relative W .
The restriction (K, λ)| Y of (K, λ) to Y is defined as the subcomplex where K ′ denotes the largest (G-invariant) subcomplex of K satisfying λ(⋄ K ′ ) ⊆ Y .
Let Y = {Y i } i∈I be a big family of G-invariant subsets of X. We introduce the simplicially enriched category C G (X, U; W ) Y of controlled G-CW-complexes mod Y: It has the same objects as C G (X, U; W ), and its morphism spaces are given by where the colimit is taken along the obvious restriction maps. The composition operation in this category is defined as follows. Let ϕ : (K, λ K )| X\Yi → (L, λ L ) and ψ : gives a well-defined composition operation and generalizes readily to higher simplices.
If Y is the trivial big family {∅}, the simplicially enriched category C G (X, U; W ) Y coincides with C G (X, U; W ). The structure map of the colimit provides a simplicially enriched functor is a weak equivalence for all (M, λ M ).
For Y = {∅}, we call ϕ simply a controlled equivalence.
3.16. Remark. For any category C enriched in Kan complexes and morphism ϕ : K → L in C, the following are equivalent: (1) The restriction map ϕ * : C(L, M ) • → C(K, M ) • is a weak equivalence for all M in C.
3.17. Example. For every object (K, λ) in C G (X; W ) and every Y i ∈ Y, the inclusion map (K, λ)| X\Yi → (K, λ) is a controlled equivalence mod Y.
Theorem (Gluing lemma). Suppose we have a commutative diagram
in which all vertical morphisms are controlled equivalences mod Y, and in which ι 1 and ι 2 are subcomplex inclusions. Then the induced morphism is a controlled equivalence mod Y.
Proof. To increase legibility, we suppress the labelings from notation. Let r ∈ {1, 2} and let Y i ∈ Y. Denote by ϕ * r (L r | X\Yi ) the largest subcomplex of K ′ r satisfying ϕ r (ϕ * r (L r | X\Yi )) ⊆ L r | X\Yi . Moreover, denote by K i r the largest subcomplex of K r such that the following two conditions are satisfied: ( Since directed colimits commute with finite limits in simplicial sets, we obtain a pullback square In particular, we obtain isomorphisms It follows that the commutative square • is a pullback. All corners of this square are Kan complexes by Corollary 3.13, and the restriction map ι * r is a Kan fibration by Corollary 3.12. Hence, the square is a homotopy pullback.
Since the transformation of homotopy pullback squares induced by ψ K and ψ L is an equivalence on all but the top left corner by assumption, it follows that the induced map on the top left corner is also an equivalence. This proves the claim of the theorem.
3.2. Controlled retractive spaces. Let W be a G-space and let (X, U) be a G-coarse space.
3.19.
Definition. An (X, U)-controlled retractive space (K, λ, r) over W is an (X, U)-controlled G-CW-complex (K, λ) over W together with a G-equivariant retraction r : K → W to the structural inclusion W → K.
A morphism of (X, U)-controlled retractive spaces is an (X, U)-controlled morphism of (X, U)-controlled G-CW-complexes which is additionally compatible with the chosen retractions.
The (X, U)-controlled retractive spaces and their morphisms form a category R G (X, U; W ).
3.20. Definition. Define the cofibrations co R G (X, U; W ) ⊆ R G (X, U; W ) to be the collection of all morphisms which are isomorphic to a morphism given by a subcomplex inclusion.
3.21. Definition. Let Y be a big family of G-invariant subsets of X. A weak equivalence mod Y is a morphism which is sent to a controlled equivalence mod Y by the canonical functor R G (X, U; W ) → C G (X, U; W ). Denote by h Y R G (X, U; W ) the collection of all weak equivalences mod Y.
If Y = {∅}, we typically omit Y from notation.
Proof. The category R G (X, U; W ) has a zero object given by the controlled G-CWcomplex (W, ∅) together with the retraction id W . The unique map (W, ∅, id W ) → (K, λ, r) is a cofibration for every controlled retractive space, and all isomorphisms are cofibrations. The existence of pushouts along cofibrations follows from Lemma 3.7 together with the observation that the pushout of retractive spaces inherits a retraction by the universal property of the pushout. This also implies that co R G (X, U; W ) is a subcategory.
By definition, all isomorphisms are h Y -equivalences. The gluing lemma for h Yequivalences is precisely Theorem 3.18 Since the weak equivalences are pulled back from a fibrant simplicially enriched category, they are closed under retracts and satisfy the two-out-of-six property. Moreover, we can use the product construction from Definition 3.8 to show that all morphisms in R G (X, U; W ) admit a factorization into a cofibration followed by a weak equivalence: If ϕ : (K, λ K , r K ) → (L, λ L , r L ) is a morphism, define the mapping cylinder of ϕ by the pushout Since the left vertical subcomplex inclusion is a weak equivalence, Theorem 3.18 implies that the right vertical morphism is also a weak equivalence. Hence, the map π : M (ϕ) → (L, λ L , r L ) induced by (K, λ K , r K ) ⊗ [0, 1] → (K, λ K , r K ) ϕ − → (L, λ L , r L ) and id L via the universal property of the pushout is a weak equivalence. The cofibration (K, λ K , r K ) ⊗ {0} ֒→ (K, λ K , r K ) ⊗ [0, 1] induces a cofibration (K, λ K , r K ) ֒→ M (ϕ), and it is easy to check that the composition (K, λ K , r K ) ֒→ M (ϕ) π − → (L, λ L , r L ) equals ϕ. We will usually abbreviate notation and write (R G (X, U; W ), h Y ) for the Waldhausen category (R G (X, U; W ), co R G (X, U; W ), h Y R G (X, U; W )).
Note that R G (X, U; W ) admits arbitrary coproducts. Since we wish to take the algebraic K-theory of the Waldhausen category (R G (X, U; W ), h Y ), we impose additional finiteness properties on objects in R G (X, U; W ). In order to do so, we have to additionally assume that X comes equipped with a bornology.
3.23. Definition. Let (X, B, U) be a G-bornological coarse space. An object (K, λ) of C G (X, U; W ) is called locally finite if λ −1 (B) is a finite set for every bounded subset B of X.
A controlled retractive space is locally finite if it is locally finite as an object in C G (X, U; W ). Denote the full subcategory of locally finite controlled retractive spaces by R G lf (X, B, U; W ). To save space, we will typically suppress the bornology and coarse structure on X from now on and write R G lf (X; W ) for R G lf (X, B, U; W ). Since R G lf (X; W ) is closed under pushouts along cofibrations, it forms a full homotopical Waldhausen subcategory (R G lf (X; W ), h Y ) of (R G (X; W ), h Y ). 3.24. Remark. The finiteness condition introduced in Definition 3.23 is weaker than the notion of finiteness employed in [UW, Definition 3.3]: (1) We make no requirements about the image of the retraction map. This condition is irrelevant for the question whether the algebraic K-theory of R G lf (X; W ) defines a coarse homology theory (as a functor of X), and only affects how this theory behaves as a functor with respect to W . One can impose conditions on the images of the retraction maps without affecting the discussion in Section 4 and Section 5 except Proposition 5.15. For example, one could require the images of the retraction maps to be contained in a G-compact subset of W in order to obtain a functor in W which is compatible with directed colimits.
(2) We do not require complexes to be finite-dimensional. This is important in our context as it produces a strongly additive coarse homology theory, see Proposition 5.15. In [UW], this additional requirement ensures that locally finite complexes are uniformly controlled by a single entourage. Moreover, finite-dimensionality plays an important role in the proof of the Farrell-Jones conjecture for finitely VCyc-amenable groups in [ELP + 18]. As illustrated by [UW,Sections 4 and 5], the proof that the algebraic K-theory of R G lf (X; W ) produces a continuous coarse homology theory also goes through if we impose finite-dimensionality as an additional requirement (but we see no reason why the resulting theory should still be strongly additive).
To conclude this section, we discuss the functoriality of R G (X; W ) and R G lf (X; W ) in X. Let f : (X, U X ) → (Y, U Y ) be a morphism of G-coarse spaces. Since f is controlled, we obtain an induced exact functor Evidently, this defines a functor If X and Y are G-bornological coarse spaces and f is in addition proper (i.e. a morphism a G-bornological coarse spaces), then the exact functor f * restricts to an exact functor Consequently, we have a functor
Coarse A-homology
The goal of this section is to prove that the functor (1) E is coarsely invariant : For every G-bornological coarse space X, the morphism E({0, 1} max,max ⊗ X → X) is an equivalence.
(4) E is u-continuous: For every G-bornological coarse space X, the natural map colim is an equivalence, where X U denotes the G-bornological coarse space obtained from X by replacing the coarse structure by the smallest coarse structure containing U .
Definition ([BEKWa, Definition 5.15]
). An equivariant coarse homology theory E is continuous if the following holds: For every filtered family Y = {Y i } i∈I of G-invariant subsets such that for every G-invariant, locally finite subset F of X there exists some i ∈ I with F ⊆ Y i , the canonical map E(Y) → E(X) is an equivalence.
Recall Definition 2.10 of the functor U W loc and Definition 2.37 of K W . 4.3. Definition. Universal coarse A-homology (relative to the base space W ) is the functor Coarse equivariant A-homology (relative to the base space W ) is the functor Proof. Since f and f ′ are close, the identity morphism on underlying spaces defines a natural isomorphism f * (K, λ, r) 4.6. Corollary. The functor UAX G W is coarsely invariant. Proof. This follows directly from Lemma 4.5 since isomorphic functors induce equivalent maps and morphisms {0, 1} max,max ⊗X → Y correspond to pairs of close maps X → Y . 4.7. Lemma. The functor UAX G W is u-continuous. Proof. Since each controlled retractive space and each controlled morphism is controlled by a single entourage, the category of controlled retractive spaces can be written as a filtered colimit in Wald ho . The claim now follows from Corollary 2.14.
4.8. Proposition. Let X be a G-bornological coarse space. Suppose there exists a sequence (s n : X → X) n∈N of morphisms such that the following holds: (1) s 0 = id X ; (2) n (s n × s n+1 )(diag(X)) is an entourage of X, where diag(X) denotes the diagonal of X; (3) n (s n × s n )(U ) is an entourage of X for every entourage U ; (4) for every bounded subset B of X exists some natural number n 0 with B ∩ s n (X) = ∅ for all n ≥ n 0 . Then R G lf (X; W ) admits an Eilenberg swindle. Proof. Each morphism s n induces an endofunctor (s n ) * . By assumption (3), we obtain the exact endofunctor We claim that S preserves locally finite objects: Suppose (K, λ) is locally finite, and write S(K, λ) = (K ∞ , λ ∞ ). Let B be a bounded subset of X. Assumption (4) implies that there exists some natural number n 0 such that Hence, S(K, λ) is locally finite, and S restricts to an endofunctor . Using assumptions (1) and (2) together with Lemma 4.5, it follows that there exists a natural isomorphism id R G lf (X;W ) ∨S ∼ = S, so S is an Eilenberg swindle on R G lf (X; W ). 4.9. Remark. If f : X → X is a morphism which implements flasqueness of X in the sense of [BEKWa, Definition 3.8], then the sequence (f n ) n≥0 satisfies the assumptions of Proposition 4.8. If G is the trivial group, the assumption of Proof. If f : X → X implements flasqueness of X, we can apply Proposition 4.8 by Remark 4.9. Then the Additivity theorem (Corollary 2.36) implies UAX G W (X) ≃ 0.
4.11. Definition. Let X be a G-bornological coarse space and let Y = (Y i ) i∈I be a big family of G-invariant subsets of X. Then we define 4.12. Remark. In the situation of Definition 4.11 we can and will identify R G lf (Y; W ) with the full subcategory of R G lf (X; W ) spanned by those objects (K, λ, r) for which there exists some i ∈ I with λ(⋄ K) ⊆ Y i .
is closed under pushouts, it forms a full homotopical Waldhausen subcategory of R G lf (X; W ). 4.13. Construction. Let X be a G-bornological coarse space and consider a zig-zag Forgetting the retractions, we find by Remark 3.16 a controlled map of relative G-CW-complexes β : (L, λ L ) → (L ′ , λ L ′ ) as well as controlled homotopies id L ≃ αβ and id L ′ ≃ βα relative W . Let be the factorization of βϕ obtained by applying the mapping cylinder construction from the proof of Proposition 3.22. Since there exists a controlled homotopy αβϕ ≃ ϕ, the controlled maps ϕ and α on the ends of the mapping cylinder extend to a controlled map ψ : (M, λ M ) → (L, λ L ). Then ψ becomes a morphism of retractive spaces by equipping M with the retraction r M := r L ψ. Since ψ restricts to ϕ on K, the inclusion ι is then also a morphism of retractive spaces. In particular, we have a factorization in R G (X; W ). If (K, λ K , r K ) and (L ′ , λ L ′ , r L ′ ) are objects of a full homotopical Waldhausen subcategory R of R G (X; W ) which is closed under tensoring with [0, 1], then (M, λ M , r M ) also belongs to R.
Recall the notion of domination from Definition 2.29. Let furthermore R G lf (X; W ) h Y denote the full subcategory of R G lf (X; W ) on the objects which are equivalent to zero mod Y.
4.14. Proposition. Let X be a G-bornological coarse space and let Y be a big family of G-invariant subsets of X. Then every object in such that the composition ψϕ is a controlled equivalence.
Proof. Let (K, λ, r) be controlled contractible mod Y. By Remark 3.16, this means that there exists a controlled homotopy η : (K, λ)| X\Yi ⊗ [0, 1] → (K, λ) between the canonical inclusion and the zero map (K ′ , λ)| X\Yi r − → W ֒→ K for some i ∈ I. By the controlled homotopy extension property 3.11, the map extends to a controlled homotopy η : (K, λ)⊗[0, 1] → (K, λ). Let U be an entourage such that η is U -controlled. Since Y is a big family, there exists some j ∈ I such that U [Y i ] ⊆ Y j . The endpoint η 1 of the homotopy η differs from the retraction map only on cells labelled by points in Y i . Hence, η 1 factors as This gives rise to a domination
the assumptions of Theorem 2.16
Proof. Using Remark 3.16, it is easy to see that the first part of the approximation property holds.
To verify the second part of the approximation property, let ϕ : (K, λ K ) → (L, λ L ) be an arbitrary morphism in R G lf (X; W ) such that (K, λ K ) lies in R G lf (Z; W ) (which is considered here, for simplicity, as a full subcategory of R G lf (X; W )). Choose an entourage U such that ϕ is U -controlled. Let i ∈ I be such that Z ∪ Y i = X, and let j ≥ i sich that Since X \ Y i ⊆ Z, the morphism ϕ ′ lies in R G lf (Z; W ). In particular, we may form the pushout . By Theorem 3.18 (Glueing Lemma) and Example 3.17, the right vertical morphism is a controlled equivalence mod Z ∩ Y. By the universal property of the pushout, we obtain a factorization of the canonical inclusion (L, λ L )| X\Yi (L, λ L ) as By the two-out-of-three property of weak equivalences, it follows that ϕ ′′ is a controlled equivalence mod Y. Moreover, we have a factorization of ϕ as which verifies the second part of the approximation property.
Proof. Let X be a G-bornological coarse space, and let (Z, Y) be an equivariant complementary pair. Consider the commutative diagram By Theorem 2.35, both rows induce cofiber sequences in M loc upon application of U W loc . Moreover, the induced map of cofibers is an equivalence by Proposition 4.15 and Theorem 2.16. Therefore, we have a pushout square in M loc . Construction 4.13 shows that both inclusion functors admit a mapping cylinder argument. Therefore, Proposition 4.14 and Theorem 2.30 imply that these functors induce equivalences upon application of U W loc , which yields the desired pushout square.
Proof of Theorem 4.4. The functor UAX G W is an equivariant coarse homology theory by Corollary 4.6, Lemma 4.7, Corollary 4.10 and Corollary 4.16.
Let Y = {Y i } i∈I be a filtered family of G-invariant subsets such that for every G-invariant, locally finite subset F of X there exists some i ∈ I with F ⊆ Y i . Since for every locally finite retractive space (K, λ) the image λ(⋄ K) of the labeling is a locally finite subset of X, it follows that R G lf (X; W ) is the filtered union Continuity follows from Corollary 2.14.
Since K commutes with filtered colimits, it is immediate that AX G W is also a continuous equivariant coarse homology theory.
Split injectivity of the A-theoretic assembly map
The goal of this section is to use the equivariant coarse homology theory AX G W constructed in Section 4 to derive the split injectivity results stated in the introduction.
To avoid a barrage of definitions at the beginning, the proof of the main result (Theorem 5.17) is split into a sequence of individual statements, each of which addresses another additional property of coarse A-homology.
5.1. The relation to classical A-theory. First of all, let us relate coarse Ahomology to the classical A-theory functor due to Waldhausen [Wal85, Section 2].
Let Q be a topological space. The objects of the category of retractive spaces R(Q) are CW-complexes K relative Q equipped with a retraction r : K → Q to the inclusion Q ֒→ K. Morphisms in this category are cellular maps over and under Q. Cofibrations are those morphisms which are isomorphic to subcomplex inclusions, and weak equivalences are those morphisms which, as maps relative Q, are homotopy equivalences. The full homotopical Waldhausen subcategory R f (Q) of finite retractive spaces over Q is spanned by those retractive spaces which arise from Q by attaching only a finite number of cells.
The ordinary (nonconnective) A-theory functor is given by the composition where Top denotes the ordinary 1-category of topological spaces and continuous maps.
5.1. Remark. In this remark we explain the functoriality of R f (Q) with respect to the space Q.
and the retraction f * r : f * K → Q ′ is determined by the universal property of the pushout. The latter is also used to define R f (f ) on morphisms (K, r) → (K ′ , r ′ ) in R f (Q). Note that the construction K → f * K can be made strictly functorial by suitable choices on the point-set level: Since we require that Q is contained in K (as the (−1)-skeleton of K), the pushout f * K can be chosen to have the underlying set Q ′ ∪ (K \ Q), equipped with the appropriate topology and induced filtration.
Let now G be a discrete group and let P be the total space of a principal Gbundle. Then P gives rise to the functor and thus to the Or(G)-spectrum Let G can,min be the G-bornological coarse space whose G-set is G, equipped with the minimal bornology and the coarse structure generated by the subset {G(F ×F ) | F ⊆ G finite} of the power set of G × G.
Proposition. There is a zig-zag of equivalences
The proof of Proposition 5.2 relies on the following construction which we will reuse later on. 5.3. Definition. Let I be a set. Define the topological space I + to have underlying set I ⊔{+}, and define a non-empty subset of I + to be closed if and only if it contains the distinguished point +.
Let X be a G-bornological coarse space and consider an object (K, λ, r) of R G lf (X; P ). Let π 0 (X) denote the G-set of coarse components of X. Due to the control conditions imposed on (K, λ), we observe that for every coarse component X 0 of X the set λ −1 (X 0 ) is the set of relative open cells of a subcomplex of K. Hence, λ induces a continuous map λ 0 : K → π 0 (X) + , (which sends P to {+}) and we have a canonical identification (the wedge sum indicates the coproduct of CW-complexes relative P ) Proof of Proposition 5.2. Let us abbreviate We furthermore define the functor F ′ c : Or(G) → Wald ho by S → R G lf (G can,min ; P × S), where we use that R G lf (G can,min ; −) is a functor on topological spaces with a Gaction in the same way as explained in Remark 5.1.
Let S ∈ Or(G), and let (K, r, λ) be an object in F ′ c (S). The composed map K r − → P × S p − → S induces a G-equivariant function λ S : ⋄ K → S. By taking the pushout P × S / / p K P / / p * K of K along the projection map, we obtain a retractive space over P . Define a labeling p * λ on p * K by setting p * λ : ⋄ p * K ∼ = ⋄ K → G × S, e → (λ(e), λ S (e)).
It is easy to check that (p * K, p * r, p * λ) is a controlled complex over G can,min ⊗ S min,max . Since (K, r, λ) is locally finite over G can,min , there exist only finitely many G-cells in K. So (p * K, p * r, p * λ) is also locally finite over G can,min ⊗S min,max . It is straightforward to check that this construction extends to a functor F ′ c (S) → F c (S).
Conversely, let (K, r, λ) be an object in F c (S). Since π 0 (G can,min ⊗ S min,max ) ∼ = S, we have seen that the complex K canonically decomposes as a coproduct (relative P ) Set K := s∈S K s . Since each summand K s is a retractive space over P , K is canonically a retractive space over P × S. Moreover, there is an induced labeling Since {1} × S is bounded in G can,min ⊗ S min,max , the complex K contains only finitely many G-cells. Hence, K together with the labeling λ is an object in F ′ c (S). It is again straightforward to check that this construction extends to a functor F c (S) → F ′ c (S).
Moreover, the functors F c (S) → F ′ c (S) and F ′ c (S) → F c (S) are easily seen to be inverse to each other. One checks that the collection of functors Recall from [Wal85, Section 2.1] the equivariant version of the category of retractive spaces: The category of finite G-retractive spaces R G f (P ) has as objects (free) G-CW-complexes K relative P equipped with an equivariant retraction r : K → P such that K arises from P by attaching finitely many free G-cells. Morphisms in this category are G-equivariant and cellular maps over and under P . Cofibrations are morphisms isomorphic to subcomplex inclusions, and weak equivalences are those morphisms which, as maps relative P , are (equivariant) homotopy equivalences. Similar to the explanation in Remark 5.1, we obtain a functor F ′ u := R G f (P × −) from Or(G) to Wald ho .
Since the underlying CW-complex of any object in F ′ c (S) contains only finitely many G-cells, forgetting labelings defines a natural transformation F ′ c → F ′ u . In order to construct an inverse we consider an object (K, r) of F ′ u (S). Since G acts freely on P all G-cells of K are free. We then choose a base point in every G orbit on ⋄ K and define the equivariant labeling λ : ⋄ K → G such that its sends the chosen base points to 1 ∈ G. Note that (K, λ, r) belongs to F ′ c (S) since ⋄ K/G is finite. We define the inverse equivalence such that it sends (K, r) to (K, λ, r).
Finally, taking quotients by the G-actions induces a natural equivalence F ′ u → F u , see [Wal85, Lemma 2.1.3]. In sum, we obtain the desired zig-zag of natural equivalences By applying K W , Proposition 5.2 has the following immediate consequence.
Corollary.
There is an equivalence of Or(G)-spectra AX G P (G can,min ⊗ (−) min,max ) ≃ A P (−). 5.2. Transfers. In the next step, we show that coarse A-homology admits a certain amount of contravariant functoriality. To describe this additional functoriality, we use the ∞-category GBornCoarse tr introduced in [BEKWc, Section 2.2], which comes equipped with a functor ι : GBornCoarse → GBornCoarse tr . 5.5. Definition ([BEKWc, Definition 2.53]). An equivariant coarse homology theory E : GBornCoarse → C admits transfers if there exists a functor For our purposes, the precise definition of GBornCoarse tr is not relevant. Instead, we will rely on an explicit criterion to describe functors from GBornCoarse tr to the nerve of a strict 2-category. 5.6. Definition ([BEKWc, Definition 2.14]). Let W and X be G-bornological coarse spaces. A bounded covering w : W → X is a map of G-sets satisfying the following properties: (1) w is controlled and bornological.
(2) The coarse structure on W coincides with the coarse structure In particular, w * K is a G-CW-complex relative P and inherits the retraction Note that the set of relative open cells ⋄(w * K) fits into a pullback diagram / / π 0 (X) of G-sets. Since the restriction of w to each coarse component of W is an isomorphism of coarse spaces, it makes sense to define the labeling w * λ : ⋄(w * K) → W, e → w| −1 q(e) (λ(p(e))). Note that the map w * K → π 0 (W ) + in (5.8) is given by (w * λ) 0 . Define w * (K, λ, r) := (w * K, w * λ, w * r).
By the universal property of the pullback, we obtain an exact functor w * : R G lf (X; P ) → R G lf (W ; P ). If w is an isomorphism on the underlying G-coarse spaces, we can choose w * K := K, we let the top horizontal map in (5.8) be the identity on K, and we let the left vertical map in (5.8) be given by π 0 (w −1 ) + • λ 0 . It follows that w * = w −1 * in this case.
Let now v : V → W and w : W → X be two bounded coverings. Then the universal property of the pullback produces for every (K, λ, r) in R G lf (X; P ) a natural isomorphism a v,w : (w • v) * (K, λ, r) → (v * • w * )(K, λ, r) which can be chosen to be the identity transformation if both v and w are isomorphisms on the underlying G-coarse spaces. Note that, again by the universal property of the pullback, the relation (u * • a v,w )a u,wv = (a u,v • w * )a vu,w holds for every three composable bounded coverings u : U → V , v : V → W and w : W → X.
5.12. Definition. Let (X i , U i , B i ) i∈I be a family of G-bornological coarse spaces. The free union free i∈I X i of this family is the following G-bornological coarse space: (1) The underlying G-set is the disjoint union i∈I X i .
(2) The bornology is generated by those subsets B for which B ∩ X i ∈ B i for all i and B ∩ X i is non-empty for only finitely many i. (3) The coarse structure is generated by the entourages i∈I U i for all families of entourages (U i ) i∈I with U i ∈ U i .
Let E be an equivariant coarse homology theory. Let (X i ) i∈I be a family of G-bornological coarse spaces. Since (X j , j =i∈I X i ) is a coarsely excisive pair, we obtain by excision for every j ∈ I a projection map p j : E( i∈I X i ) → E(X j ). 5.13. Definition. Let C be a cocomplete stable ∞-category which admits all products. An equivariant coarse homology theory E : GBornCoarse → C is strongly additive if for every family of G-bornological coarse spaces (X i ) i∈I the map is an equivalence. 5.14. Remark. Let (X i ) i∈I be a family of G-bornological coarse spaces. The inclusion inc j : X j → free i∈I X i of the j-th component is a bounded covering. If the equivariant coarse homology theory E admits transfers, it follows formally that the projection map p j is equivalent to the transfer inc * j along inc j . See also [BEKWc, Remark 2.62]. 5.15. Proposition. The equivariant coarse homology theory AX G P is strongly additive.
Proof. Let (X i ) i∈I be a family of G-bornological coarse spaces. Using Remark 5.14, the comparison map factors as In view of Theorem 2.38, it suffices to observe that the transfer functors inc * i induce an exact equivalence This follows directly from the definitions.
5.4. Split injectivity results. We conclude this section by summarizing the results of the previous subsections and explaining how the axiomatic framework of [BEKWb] applies to prove split injectivity results for the A-theoretic assembly map.
5.16. Definition. A functor M : Or(G) → Sp is a CP-functor if there exists an equivariant coarse homology theory E such that the following holds: (1) M is equivalent to E(G can,min ⊗ (−) min,max ); (2) E is continuous; (3) E is strongly additive; (4) E extends to an equivariant coarse homology theory with transfers.
We call M a hereditary CP-functor if M • Res ϕ is a CP-functor for every surjective homomorphism ϕ : G → Q, where Res ϕ : Or(Q) → Or(G) denotes the functor restricting group actions along ϕ.
5.17. Theorem. For every discrete group G and principal G-bundle P , the functor A P is a hereditary CP-functor.
Proof. A P is a CP-functor for every discrete group G and every principal G-bundle P by Corollary 5.4, Theorem 4.4, Proposition 5.15 and (5.11). Let ϕ : G → Q be a surjective homomorphism. Since there exists an isomorphism P × G Res ϕ (S) ∼ = Ind ϕ (P ) × Q S which is natural in S, we see that A P •Res ϕ ≃ A Ind ϕ (P ) . Consequently, the previous paragraph implies that A P is a hereditary CP-functor. | 2018-11-28T22:24:26.000Z | 2018-11-28T00:00:00.000 | {
"year": 2018,
"sha1": "feffe941507c733921961f498410fb551efce123",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1811.11864",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "5c1e63fca6343bc627221f896f2c891fb106839d",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
16272777 | pes2o/s2orc | v3-fos-license | Intraoperative herniation of an L5-S1 disc during microdiscectomy and transforaminal lumbar interbody fusion: a case report
Introduction We report the progression of an intraoperative L5-S1 lumbar disc herniation that occurred during a routine microdiscectomy and transforaminal lumbar interbody fusion, which, to the best of our knowledge, has never been previously reported in the literature. The objective of this report is to bring to light the possibility of a lumbar disc herniating intraoperatively, and to demonstrate that accompanying neurologic involvement can be detected and subsequently addressed with the aid of neurophysiologic monitoring. Case presentation A 36-year-old African American woman, who had previously undergone minimally invasive microdiscectomy for a right L5-S1 herniated nucleus pulposus with full recovery, presented with a large reherniation of the L5-S1 disc on the right side. During her operation, while a tap was followed into the L5 left pedicle, there was a sudden profound spasm of our patient’s legs and back that lasted for the duration of 15 seconds, culminating in the loss of all somatosensory evoked potentials in our patient’s lower extremities. Exploration of this previous microlaminotomy site revealed a massive disc extrusion protruding through the microlaminotomy. Immediate removal of this extruded disc material restored all somatosensory evoked potentials and our patient awoke with no neurologic deficits. Conclusions An intraoperative disc herniation in the lumbar spine, though very rare, can occur and can result in neurologic compromise as evidenced by the loss of somatosensory evoked potentials. By identifying the event, it can be remedied by evaluating the disc visually, removing extruded fragments and decompressing nerve roots with recovery of somatosensory evoked potentials and normal neurologic function postoperatively. If neurophysiological monitoring shows there is a sudden loss of response, then consideration should be given to the possibility of an acute intraoperative herniation.
Introduction
Surgical intervention may be indicated when nonsurgical management of a symptomatic herniated lumbar disc, including physical therapy and epidural steroid injections, fails. A small percentage of patients who undergo primary lumbar discectomy develop symptoms of failed back surgery syndrome postoperatively, with recurrent lumbar disc herniation as a leading cause [1]. Current literature indicates that the chances of recurrent lumbar disc herniation following a primary discectomy is 5-15 % [2,3,4,5,6]. Transforaminal lumbar interbody fusion (TLIF) is a well-accepted method for the treatment of recurrent lumbar herniation, demonstrating improved clinical outcome and a low surgical complication rate [7].
In the case described, a 36-year-old woman undergoing a microdiscectomy and TLIF for a recurrent L5-S1 disc herniation developed acute further herniation of this disc intraoperatively, as evidenced by the loss of somatosensory evoked potentials (SSEPs). To the best of our knowledge, such a neurologic complication during surgery has never been previously described in the literature.
Case presentation
A 36-year-old African American woman with a history of occasional backaches for years had developed right leg pain for 2 months that was profound and incapacitating. Our patient had previously undergone minimally invasive microdiscectomy for a right L5-S1 herniated nucleus pulposus and made a full recovery, with no significant back pain and slight right leg pain (1-2/10). She had no symptoms for 6 weeks until she presented again complaining of difficulty walking and severe lower back pain radiating to her right lower extremity. Her neurologic examination demonstrated positive straight leg rising at 30°on the right, decreased pinprick over the lateral aspect of the right foot, and absent ankle jerk on the right. Repeat magnetic resonance imaging (MRI) showed a large reherniation of the same L5-S1 disc on the right side (Fig. 1a).
Due to worsening symptomatology and the size of the herniation, our patient was inclined to proceed with surgery. An L5-S1 secondary minimal-access endoscopic microdiscectomy and TLIF with O-arm computed tomography (CT) imaging and frameless stereotaxic guidance was performed. The approach to the discectomy was to be done at L5 on the right, so a decision to first place pedicle screws on the left side percutaneously, followed by a right side discectomy and meticulous disc fragment extraction, then subsequent placement of screws on the right side, was made.
Intraoperative neurophysiological monitoring (IONM) was performed with SSEPs of bilateral tibial nerves, free-run electromyography (EMG), and triggered EMG via pedicle screw stimulation. The tibial nerves were stimulated with surface electrodes at the level of the medial malleolus of the ankles. EMGs were monitored with disposable needle electrodes in the following muscles of both lower extremities: vastus medialis, tibialis anterior, abductor hallucis.
Following the aforementioned outline to place pedicle screws percutaneously on the left side first, k-wires were positioned and the tap was followed into the L5 left pedicle. At this time, all free-run EMG activity remained quiet (Fig. 2a). While tapping, there was a sudden and profound spasm of our patient's legs and back that lasted for the duration of 15 seconds, as evidenced by our EMG (Fig. 2b). The frameless guidance system and intraoperative CT imaging confirmed that the tap was accurately positioned in the pedicle. At this time, IONM indicated a loss of all SSEPs in our patient's lower extremities ( Fig. 3). Consideration was given to the possibility of further extrusion of the L5-S1 disc, resulting in a cauda equina compression. A minimal-access tubular approach was immediately taken to expose the previous microlaminotomoy site, where extruded disc material was protruding. Several large disc fragments were extracted, and the dura and neural roots were then visualized. Upon completion of the decompression the SSEPs all began to recover. An interbody device was placed for fusion in standard fashion (with no loose disc fragments remaining), and then the pedicle screws on the right side Fig. 1 Clinical diagnosis and surgical treatment. Preoperative sagittal magnetic resonance imaging (MRI) depicting the L5-S1 disc herniation immediately before surgery. The herniated disc can clearly be seen protruding into the spinal cord (a). Postoperative sagittal computed tomography (CT) scan immediately following surgery and transverse X-ray image of screws and interbody device at L5-S1 1 week later, both demonstrating accurate screw placement into the vertebral pedicles. It is clear that there was no protrusion of the screws into the spinal cord, and the pedicle was therefore properly tapped (b and c) were placed followed by rod placement with the Sextant system (Medtronic Sofamor Danek USA, Inc., Memphis, TN, USA) A follow-up intraoperative O-arm CT scan was performed, revealing the screws and interbody device to be in excellent position concurrently with no response in the IONM (indicating that the nerve roots were not injured from the pedicle screws). Positioning was confirmed by X-ray 1 week later ( Fig. 1b and 1c). Our patient awoke with normal strength and sensation in her lower extremities.
Discussion
To the best of our knowledge, this is the first reported case of intraoperative disc herniation that occurred while performing a secondary discectomy and fusion. It is clear from our SSEP and EMG recordings that a separate neurologic event occurred, which resulted in cauda equina compression and loss of sensory pathways (Figs. 2 and 3). The finding of disc material herniating through the previous microlaminotomy with several large extruded fragments detected, along with recovery of the sensory pathways upon removal of these fragments, is evidence for the occurrence of a further herniation intraoperatively (Fig. 3). We presume that the spasms that occurred resulted in an increase of intradiscal pressure, which resulted in this further herniation, resulting in compression of the lumbar nerve roots and leading to a loss of SSEPs.
An occurrence of intraoperative spasms can usually be explained by anesthetic lightening, electrocautery stimulation, anatomical changes, or static electrical discharge from a surgical tool. The spasms were unlikely related to anesthetic lightening, because our electroencephalography (EEG) recordings showed our patient to be in deep sleep at the time of spasm. Bipolar electrocautery was not being used at the time of spasm either, eliminating the possible spread of an electrical system as a cause. Additionally, direct irritation of the nerve root from possible anatomical changes was ruled out by the intraoperative CT scan and IONM monitoring, because the frameless stereotaxic guidance system ensured that there was no breach of the pedicle screws into the nerve roots, and there was no neurophysiological response to pedicle screw placement in IONM. When tapping the pedicle, it is possible to misdirect the tap, and it could theoretically damage one or multiple nerve roots, but it was clear that this was not the case in this instance, again as indicated by our postoperative scans showing the pedicle screws in proper placement ( Fig. 1b and 1c). Another anatomical possibility is that the slight force used during left pedicle tapping was sufficient enough to exacerbate a reherniation through the previous annulotomy site, considering this was already a mobile segment.
It is also possible that there was a discharge of static electricity from the surgical instrumentation as it was being placed over the k-wire, which resulted in nerve stimulation. The buildup of static electricity on surgical equipment has been recognized for years now [8], however, we found no specific literature reference to lower extremity muscle spasms related to static electrical discharge. An antistatic silicone rubber mat for retaining surgical instruments was used during the operation, but other surgical devices (such as the surgical probe) may still have accumulated in charge during the operation. It Fig. 2 Intraoperative electromyography recordings displaying patient spasms. Baseline intraoperative free-run EMG activity before muscle spasms displaying no muscle activity (a), and EMG burst activity during intraoperative leg muscle spasms (b). The activity during the muscle spasms is most pronounced in the quadriceps, although all EMGs showed substantial activity at this time. EMG electromyography, QUAD vastus medialis, TA tibialis anterior, AH abductor halluces is possible that none of the foregoing hypotheses are correct and that another, yet to be appreciated, event occurred.
A recent investigation of 100 consecutive TLIFs revealed that the incidence of minor surgical complication in these procedures was 16 %, and permanent surgical complication was very low [9]. The use of IONM during decompression/reconstruction surgeries is an additional factor that contributes to patient safety and potentially alerts the surgeon if there is a complication. Nonetheless, there is controversy over the usefulness of IONM in degenerative disc surgeries performed in the lumbar spine [10,11,12]. It is well understood by surgeons that IONM may not always indicate intraoperative neural deficits because motor evoked potentials or SSEPs can remain unaffected. Such was the case in a recent report on an intraoperative disc herniation that occurred in a 17-year-old patient during posterior spinal fusion for correction of Scheuermann's kyphosis [13]. However, IONM is still preferred to an intraoperative wake-up test, and is effective in certain spine operations. A case series directed at Athens University Medical School demonstrated the usefulness of IONM in allowing for intraoperative pedicle screw correction, revealing a greater reduction in the number of misdirected pedicle screws and thus neurologic complications when IONM was used [14]. Specifically, it was found that IONM had a predictive value of 98.73 % for indicating misdirection of pedicle screws in posterior thoracolumbar spinal fusion.
In the ongoing dialogue over whether IONM is necessary in spinal decompression and reconstruction procedures, this case demonstrates the significant value Fig. 3 Intraoperative neurophysiological monitoring displaying loss of tibial somatosensory evoked potentials and recovery. Loss of tibial SSEPs during intraoperative leg muscle spasms and partial recovery after decompression; the traces are in descending order of time, with the first trace in red as the baseline. The SSEP amplitudes in both tibial responses are dramatically reduced following "EMG burst activity", as marked between the crosshairs on each trace. It is thought that the muscle spasms led to increased intradiscal pressure and then intraoperative herniation, followed by loss of tibial SSEPs. The "recovery" of SSEPs occurred following decompression of the herniated disc. The last three traces of tibial SSEPs actually look worse, possibly due to the addition of inhalational anesthesia toward the end of the operation. Full clinical recovery of function may be attributed to additional time for the patient to recover from the effects of anesthesia prior to moving her limbs. EMG electromyography, SSEP somatosensory evoked potential of IONM in indicating the onset of an unanticipated and otherwise unrecognized event leading to profound neural compression. If the herniation had occurred intraoperatively at another disc level, without IONM there would have been no indication of its occurrence until postoperative deficits were evident. The ability of IONM to allow us to address this intraoperative herniation allowed for the avoidance of permanent neurologic injury.
Conclusions
This case presented with an unusual set of circumstances: (1) muscle spasms in the lower extremities during percutaneous lumbar pedicle screw tapping; (2) loss of tibial nerve SSEPs that occurred soon thereafter as well as EMG activation; (3) exploration of the previous microlaminotomy with identification and removal of large extruded fragments; and (4) immediate recovery of tibial SSEPs that correlated with normal neurologic examination postoperatively and quieting of the EMGs.
Neurosurgeons and anesthesiologists should be wary of possible conditions and hazards that can cause a patient to go into spasm during an operation. The likelihood of an intraoperative disc herniation is very small, but the possibility exists. In this case, the use of IONM was deemed critical in addressing a surgical complication rapidly, thus avoiding permanent neurologic injury.
Consent
Written informed consent was obtained from the patient for publication of this case report and any accompanying images. A copy of the written consent is available for review by the Editor-in-Chief of this journal. | 2017-06-23T14:42:54.472Z | 2015-11-27T00:00:00.000 | {
"year": 2015,
"sha1": "cecc35829771d6a8bbdba99fb34c430d1ea27fc4",
"oa_license": "CCBY",
"oa_url": "https://jmedicalcasereports.biomedcentral.com/track/pdf/10.1186/s13256-015-0766-6",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "d90f890e6786c187860c59e85ca72f230318d3b8",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
47018305 | pes2o/s2orc | v3-fos-license | Reconciling multiple genes trees via segmental duplications and losses
Reconciling gene trees with a species tree is a fundamental problem to understand the evolution of gene families. Many existing approaches reconcile each gene tree independently. However, it is well-known that the evolution of gene families is interconnected. In this paper, we extend a previous approach to reconcile a set of gene trees with a species tree based on segmental macro-evolutionary events, where segmental duplication events and losses are associated with cost \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta $$\end{document}δ and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda $$\end{document}λ, respectively. We show that the problem is polynomial-time solvable when \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta \le \lambda $$\end{document}δ≤λ (via LCA-mapping), while if \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta > \lambda $$\end{document}δ>λ the problem is NP-hard, even when \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\lambda = 0$$\end{document}λ=0 and a single gene tree is given, solving a long standing open problem on the complexity of multi-gene reconciliation. On the positive side, we give a fixed-parameter algorithm for the problem, where the parameters are \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta /\lambda $$\end{document}δ/λ and the number d of segmental duplications, of time complexity \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$O\left(\lceil \frac{\delta }{\lambda } \rceil ^{d} \cdot n \cdot \frac{\delta }{\lambda }\right)$$\end{document}O⌈δλ⌉d·n·δλ. Finally, we demonstrate the usefulness of this algorithm on two previously studied real datasets: we first show that our method can be used to confirm or raise doubt on hypothetical segmental duplications on a set of 16 eukaryotes, then show how we can detect whole genome duplications in yeast genomes.
Basic notions
For our purposes, a rooted phylogenetic tree T = (V (T ), E(T )) is an oriented tree, where V (T ) is the set of nodes, E(T ) is the set of arcs, all oriented away from r(T ), the root. Unless stated otherwise, all trees in this paper are rooted phylogenetic trees. A forest F = (V (F ), E(F )) is a directed graph in which every connected component is a tree. Denote by t(F ) the set of trees of F that are formed by its connected components. Note that a tree is itself a forest. In what follows, we shall extend the usual terminology on trees to forests. For an arc (x, y) of F , we call x the parent of y, and y a child of x. If there exists a path that starts at x and ends at y, then x is an ancestor of y and y is a descendant of x. We say y is a proper descendant of x if y = x, and then x is a proper ancestor of y. This defines a partial order denoted by y ≤ F x, and y < F x if x = y (we may omit the F subscript if clear from the context). If none of x ≤ y and y ≤ x holds, then x and y are incomparable. The set of children of x is denoted ch(x) and its parent x is denoted par(x) (which is defined to be x if x itself is a root of a tree in t(F )). For some integer k ≥ 0, we define par k (x) as the k-th parent of x. Formally, par 0 (x) = par(x) and par k (x) = par(par k−1 (x)) for k > 0. The number of children |ch(x)| of x is called the out-degree of x. Nodes with no children are leaves, all others are internal nodes. The set of leaves of a tree F is denoted by L(F ). The leaves of F are bijectively labeled by a set L(F ) of labels. A forest is binary if |ch(x)| = 2 for all internal nodes x. Given a set of nodes X that belong to the same tree T ∈ t(F ), the lowest common ancestor of X, denoted LCA F (X), is the node z that satisfies x ≤ z for all x ∈ X and such that no child of z satisfies this property. We leave LCA F (X) undefined if no such node exists (when elements of X belong to different trees of t(F )). We may write LCA F (x, y) instead of LCA F ({x, y}). The height of a forest F , denoted h(F ), is the number of nodes of a longest directed path from a root to a leaf in a tree of F (note that the height is sometimes defined as the number of arcs on such a path -here we use the number of nodes instead). Observe that since a tree is a forest, all the above notions also apply on trees.
Reconciliations
A reconciliation usually involves two rooted phylogenetic trees, a gene tree G and a species tree S, which we always assume to be both binary. In what follows, we will instead define reconciliation between a gene forest G and a species tree. Here G can be thought of as a set of gene trees. Each leaf of G represents a distinct extant gene, and G and S are related by a function s : L(G) → L(S), which means that each extant gene belongs to an extant species. Note that s does not have to be injective (in particular, several genes from a same gene tree G of G can belong to the same species) or surjective (some species may not contain any gene of G). Given G and S, we will implicitly assume the existence of the function s.
In a DL reconciliation, each node of G is associated to a node of S and an event -a speciation (S), a duplication (D) or a contemporary event (C) -under some constraints. A contemporary event C associates a leaf u of G with a leaf x of S such that s(u) = x. WA B I 2 0 1 8 5:4
Reconciling Multiple Genes with Segmental Duplications
A speciation in a node u of G is constrained to the existence of two separated paths from the mapping of u to the mappings of its two children, while the only constraint given by a duplication event is that evolution of G cannot go back in time. More formally: Definition 1 (Reconciliation). Given a gene forest G and a species tree S, a reconciliation between G and S is a function α that maps each node u of G to a pair (α r (u), α e (u)) where α r (u) is a node of V (S) and α e (u) is an event of type S, D or C, such that: 1. if u is a leaf of G, then α e (u) = C and α r (u) = s(u); 2. if u is an internal node of G with children u 1 , u 2 , then exactly one of following cases holds: • α e (u) = S, α r (u) = LCA S (α r (u 1 ), α r (u 2 )) and α r (u 1 ), α r (u 2 ) are incomparable; • α e (u) = D, α r (u 1 ) ≤ α r (u) and α r (u 2 ) ≤ α r (u) Note that if G consists of one tree, this definition coincides with the usual one given in the literature (first formally defined in [18]). We say that α is an LCA-mapping if, for each internal node u ∈ V (G) with children u 1 , u 2 , α r (u) = LCA S (α r (u 1 ), α r (u 2 )). Note that there may be more than one LCA-mapping, since the S and D events on internal nodes can vary. The number of duplications of α, denoted by d(α) is the number of nodes u of G such that α e (u) = D. For counting the losses, first define for y ≤ x the distance dist(x, y) as the number of arcs on the path from x to y. Then, for every internal node u with children {u 1 , u 2 }, the number of losses associated with u in a reconciliation α, denoted by l α (u), is defined as follows: The number of losses of a reconciliation α, denoted by l(α), is the sum of l α (·) for all internal nodes of G. The usual cost of α, denoted by cost(α), is d(α) · δ + l(α) · λ [16], where δ and λ are respectively the cost of a duplication and a loss event (it is usually assumed that speciations do not incur cost). A most parsimonious reconciliation, or MPR, is a reconciliation α of minimum cost. It is not hard to see that finding such an α can be achieved by computing a MPR for each tree in t(G) separately. This MPR is the unique LCA-mapping α in which α e (u) = S whenever it is allowed according to Definition 1 [3].
Reconciliation with segmental duplications
Given a reconciliation α for G in S, and given s ∈ V (S), write D(G, α, s) = {u ∈ V (G) : α e (u) = D and α r (u) = s} for the set of duplications of G mapped to s. We define G[α, s] to be the subgraph of G induced by the nodes in D(G, α, s). Note that G[α, s] is a forest.
5:5
Here we want to tackle the problem of reconciling several gene trees at the same time and counting segmental duplications only once. Given a set of duplications nodes D ∈ V (G) occurring in a given node s of the species tree, it is easy to see that the minimum the number of segmental duplications associated with s is the minimal number of parts in a partition of D in which each part does not contain comparable nodes. See Figure 1.(4) for an example. This number coincides [1] with h α (s) := h(G[α, s]), i.e. the height of the forest of the duplications in s. Now, denoted(α) = s∈V (S) h α (s). For instance in Figure 1, under the mapping µ in (2), we haved(µ) = 6, because h µ (s) = 1 for s ∈ {A, B, C, E} and h µ (F ) = 2. But under the mapping α in (3),d(α) = 4, since h α (A) = 1 and h α (F ) = 3. Note that this does not consider losses though -the α mapping has more losses than µ.
The cost of α is cost SD (G, S, α) = δ ·d(α) + λ · l(α). If G and S are unambiguous, we may write cost SD (α). We have the following problem : Most Parsimonious Reconciliation of a Set of Trees with Segmental Duplications (MPRST-SD) Instance: a species tree S, a gene forest G, costs δ for duplications and λ for losses. Output: a reconciliation α of G in S such that cost SD (G, S, α) is minimum.
Note that, when λ = 0, cost SD coincides with the unconstrained ME score defined in [20] (where it is called the FHS model).
Properties of multi-gene reconciliations
We finish this section with some additional terminology and general properties of multi-gene reconciliations that will be useful throughout the paper. The next basic result states that in a reconciliation α, we should set the events of internal nodes to S whenever it is allowed.
Lemma 2.
Let α be a reconciliation for G in S, and let u ∈ V (G) such that α e (u) = D. Let α be identical to α, with the exception that α e (u) = S, and suppose that α satisfies the requirements of Definition 1. Then cost SD (α ) ≤ cost SD (α).
Proof.
Observe that changing α e (u) from D to S cannot increased(α). Moreover, as dist(α r (u), α r (u 1 )) and dist(α r (u), α r (u 2 )) are the same as in α for the two children u 1 and u 2 of u, by definition of duplications and losses this decreases the number of losses by 2. Thus cost SD (α ) ≤ cost SD (α), and this inequality is strict when λ > 0.
Since we are looking for a most parsimonious reconciliation, by Lemma 2 we may assume that for an internal node u ∈ V (G), α e (u) = S whenever allowed, and α e (u) = D otherwise. Therefore, α e (u) is implicitly defined by the α r mapping. To alleviate notation, we will treat α as simply as a mapping from V (G) to V (S) and thus write α(u) instead of α r (u). We will assume that the events α e (u) can be deduced from this mapping α by Lemma 2.
Therefore, treating α as a mapping, we will say that α is valid if for every v ∈ V (G), Since we are assuming that S and D events can be deduced from α, the LCA-mapping is now unique: we denote by µ : where v 1 and v 2 are the children of v. Note that for any valid reconciliation α, we have α(v) ≥ µ(v) for all v ∈ V (G). We also have the following, which will be useful to establish our results. Let v 1 and v 2 be the two children of v. If α(v) = LCA S (α(v 1 ), α(v 2 )), then v must be a duplication, by the definition of S events. The same holds if α(v 1 ) and α(v 2 ) are not incomparable. Thus assume α(v) = LCA S (α(v 1 ), α(v 2 )) > µ(v) and that α(v 1 ) and α(v 2 ) are incomparable. This implies that one of
Lemma 4.
Let α be a mapping from G to S, and let v ∈ V (G). Suppose that there is some , and so α(v ) = µ(v). We must then have α(v ) = µ(v) for every node v on the path between v and v. In particular, v has a child v 1 with α(v) = α(v 1 ) and thus v is a duplication. If instead α(v) > µ(v), then v is a duplication by Lemma 3.
The Shift-down lemma will prove very useful to argue that we should shift mappings of duplications down when possible, as it saves losses (see Figure 2). For future reference, do note however that this may increase the height of some duplication forest G[α, s].
Proof. Let v 1 and v 2 be the children of v, and denote t := α(v), t 1 := α(v 1 ) and t 2 := α(v 2 ). Moreover denote α := α[v → s]. Let P be the set of nodes that appear on the path between s and t, excluding s but including t (note that s is a proper descendant of t but an ancestor of both t 1 and t 2 , by the validity of α ). For instance in Figure 2, P = {x, t}. Observe that |P | = k. Under α, there is a loss for each node of P on both the (v, v 1 ) and (v, v 2 ) branches. (noting that v is a duplication by Lemma 3). These 2k losses are not present under α . On the other hand, there are at most k losses that are present under α but not under α, which consist of one loss for each node of P on the (par(v), v) branch (in the case that v is not the root of its tree -otherwise, no such loss occurs). This proves that l(α ) ≤ l(α) − k.
An illustration of the Shift-down lemma can be found in Figure 2.
Figure 2
The Shift-down lemma in action. Here t = par 2 (s), and remapping v from t to s saves 2 losses -4 losses are saved below v and 2 are added above.
3
The computational complexity of the MPRST-SD problem We separate the study of the complexity of the MPRST-SD problem into two subcases: when λ ≥ δ and when λ < δ. The following theorem states that, when λ ≥ δ, the MPR (ie the LCA-mapping) is a solution to the MPRST-SD problem.
Theorem 6. Let G and S be an instance of MPRST-SD, and suppose that λ ≥ δ, Then the LCA-mapping µ is a reconciliation of minimum cost for G and S. Moreover if λ > δ, µ is the unique reconciliation of minimum cost.
Proof. Let α be a mapping of G into S of minimum cost. Let v ∈ V (G) be a minimal node of G with the property that α(v) = µ(v) (i.e. all proper descendants v of v satisfy α(v) = µ(v)). Note that v must exists since, for every leaf l ∈ L(G), we have α(l) = µ(l).
This possibly increases the sum of duplications by 1, so thatd(α ) ≤d(α) + 1. But by the Shift-down lemma, l(α ) ≤ l(α) − 1. Thus we have at most one duplication but save at least one loss.
If λ > δ, this contradicts the optimality of α, implying that v cannot exist and thus that α = µ. This proves the uniqueness of µ in this case.
The case of δ > λ.
We show that, in contrast with the λ ≥ δ case, the MPRST-SD problem is NP-hard when δ > λ and the costs are given as part of the input. More specifically, we show that the problem is NP-hard when one only wants to minimize the sum of duplication heights, i.e. λ = 0. Note that if λ > 0 but is small enough, the effect will be the same and the hardness result also holds -for instance, putting δ = 1 and say λ < 1 2|V (G)||V (S)| ensures that even if a maximum number of losses appears on every branch of G, it does not even amount to the cost of one duplication. The hardness proof is quite technical, and we refer the interested reader to the Appendix for the details.
We briefly outline the main ideas of the reduction. The reduction is from the Vertex Cover problem, where we are given a graph G and must find a subset of vertices V ⊆ V (G) of minimum size such that each edge has at least one endpoint in V . The species tree S and the forest G are constructed so that, for each vertex v i ∈ V (G), there is a gene tree A i in G with a long path of duplications, all of which could either be mapped to a species called y i or another species z i . We make it so that mapping to y i introduces one more duplication than mapping to z i , hence we have to "pay" for each y i . We also have a gene tree C h in G for each edge e h ∈ E(G), with say v i and v j being the endpoints of e h . In C h , there is a large set of duplications D under the LCA-mapping µ. We make it so that if we either mapped the duplications in A i or A j to y i or y j , respectively, then we may map all the D nodes to y i or y j without adding more duplications. However if we did not choose y i nor y j , then it will not be possible to remap the D nodes without incurring a large duplication cost. Therefore, the goal becomes to choose a minimum number of y i 's from the A i trees so that for each edge e h = {v i , v j }, one of y i or y j is chosen for the tree C h . This establishes the correspondence with the vertex cover instance.
The above hardness supposes that δ and λ can be arbitrarily far apart. This leaves open the question of whether MPRST-SD is NP-hard when δ and λ are fixed constants -WA B I 2 0 1 8 5:8 Figure 3 The construction of S and T from S and the set of gene trees G1, . . . , G k (here k = 3). The black squares indicate the path of k − 1 duplications that must be mapped to r(S ).
Reconciling Multiple Genes with Segmental Duplications
in particular when δ = 1 + and λ = 1, where < 1 is some very small constant. We end this section by showing that the above hardness result persists even if only one gene tree is given. The idea is to reduce from the MPRST-SD show hard just above. Given a species tree S and a gene forest G, we make G a single tree by incorporating a large number of speciations (under µ) above the root of each tree of G (modifying S accordingly), then successively joining the roots of two trees of G under a common parent until G has only one tree.
Theorem 8. The MPRST-SD problem is NP-hard for λ = 0 and for given δ > λ, even if only one gene tree is given as input.
Proof. We reduce from the MPRST-SD problem in which multiple trees are given. We assume that δ = 1 and λ = 0 and only consider duplications -we use the same argument as before to justify that the problem is NP-hard for very small λ. Let S be the given species tree and G be the given gene forest. As we are working with the decision version of MPRST-SD, assume we are given an integer t and asked whether cost SD (G, S, α) ≤ t for some α. Denote n = |L(G)| and let G 1 , . . . , G k be the k > 1 trees of G. We construct a corresponding instance of a species tree S and a single gene tree T as follows (the construction is illustrated in Figure 3). Let S be a species tree obtained by adding 2(t + k) nodes "above" the root of S. More precisely, first let C be a caterpillar with 2(t + k) internal nodes. Let l be a deepest leaf of C. Obtain S by replacing l by the root of S. Then, obtain the gene tree T by taking k copies C 1 , . . . , C k of C, and for each leaf l of each C i other than l, put s(l ) as the corresponding leaf in S . Then for each i ∈ [k], replace the l leaf of C i by the tree G i (we keep the leaf mapping s of G i ), resulting in a tree we call T i . Finally, let T be a caterpillar with k leaves h 1 , . . . , h k , and replace each h i by the T i tree. The resulting tree is T . We show that cost(G, S, α) ≤ t for some α if and only if cost(T, S , α ) ≤ t + k − 1 for some α .
Notice the following: in any mapping α of T , the k −1 internal nodes of the T caterpillar must be duplications mapped to r(S ), so that h α (r(S )) ≥ k − 1. Also note that under the LCA mapping µ for T and S , the only duplications other than those k − 1 mentioned above occur in the G i subtrees. The (⇒) is then easy to see: given α such that cost(G, S, α) ≤ t, we set α (v) = α(v) for every node v of T that is also in G (namely the nodes of G 1 , . . . , G k ), and set α (v) = µ(v) for every other node. This achieves a cost of t + k − 1.
As for the (⇐) direction, suppose that cost SD (T , S , α ) ≤ t + k − 1 for some mapping α . Observe that under the LCA-mapping in T , each root of each G i subtree has a path of 2(t + k) speciations in its ancestors. If any node in a G i subtree of T is mapped to r(S ), then all these speciations become duplications (by Lemma 4), which would contradict cost SD (T , S , α ) ≤ t + k − 1. We may thus assume that no node belonging to a G i subtree is mapped to r(S ). Since h α (r(S )) ≥ k − 1, this implies that the restriction of α to the
5:9
G i subtrees has cost at most t.
More formally, consider the mapping α from G to S in which we put because α does not contain the top k − 1 duplications of α , and cannot introduce longer duplication paths than in α .
We are not done, however, since α is a mapping from G to S , and not from G to S.
We will remap every such node to r(S) and show that this cannot increase the cost. Observe that if v ∈ Q, then every ancestor of v in G is also in Q. Also, every node in Q is a duplication (by invoking Lemma 3).
Consider the mapping α * from G to S in which we put α Moreover, the height of the r(S) duplications under α * cannot be more than the height of the forest induced by Q and the duplications mapped to r(S) under α . In other words, Therefore, the sum of duplication heights cannot have increased. Finally, because α * is a mapping from G to S, we deduce that cost SD (G, S, α * ) ≤ cost SD (G, S , α ) ≤ t, as desired.
An FPT algorithm
In this section, we show that for costs δ > λ and a parameter d > 0, if there is an optimal In what follows, we allow mappings to be partially defined, and we use the ⊥ symbol to indicate undetermined mappings. The idea is to start from a mapping in which every internal node is undetermined, and gradually determine those in a bottom-up fashion. We need an additional set of definitions. We will assume that δ > λ > 0 (although the algorithm described in this section can solve the λ = 0 case by setting λ to a very small value).
We say that the mapping α : That is, if a node is determined, then all its descendants also are. This also implies that every ancestor of a ⊥-node is also a ⊥-node. Reconciling Multiple Genes with Segmental Duplications v ∈ V (G), then α is called complete. Note that if α is partial and α(v) = ⊥, one can already determine whether v is an S or a D node, and hence we may say that v is a speciation or a duplication under α. Also note that the definitions ofd(α), l(α) and h α (s) extend naturally to a partial mapping α by considering the forest induced by the nodes not mapped to ⊥.
If α is a partial mapping, we call α a completion of α if α is complete, and α(v) = α (v) whenever α(v) = ⊥. Note that such a completion always exists, as in particular one can map every ⊥-node to the root of S (such a mapping must be valid, since all ancestors of a ⊥-node are also ⊥-nodes, which ensures that r(S) = α (v) ≥ α (v ) for every descendant v of a newly mapped ⊥-node v). We say that α is an optimal completion of α if cost SD (α ) is minimum among every possible completion of α. For a minimal ⊥-node v with children v 1 and v 2 , we denote µ α (v) = LCA S (α(v 1 ), α(v 2 )), i.e. the lowest species of S to which v can possibly be mapped to in any completion of α.
The following Lemma forms the basis of our FPT algorithm, as it allows us to bound the possible mappings of a minimal ⊥-node.
Proof. Let α * be an optimal completion of α and let α : This contradicts the optimality of α * .
A node v ∈ V (G) is a required duplication (under α) if, in any completion α of α, v is a duplication under α . We first show that required duplications are easy to find.
, and equality follows.
Lemma 11 and Lemma 12 allow us to find minimal ⊥-nodes of G that are the easiest to deal with, as their mapping in an optimal completion can be determined with certainty.
Lemma 11. Let v be a minimal ⊥-node under α. If v is not a required duplication under α, then α * (v) = µ α (v) for any optimal completion α * of α.
Proof.
Let v 1 , v 2 be the children of v, and let α * be an optimal completion of α. Since v is not a required duplication, by Lemma 10 we have α(v 1 ) < µ(v) and α(v 2 ) < µ(v) and, as argued in the proof of Lemma 10, α(v 1 ) and α(v 2 ) are incomparable. We thus have R. Dondi, M. Lafond and C. Scornavacca
5:11
is a valid mapping, and v is a speciation under this mapping. Henced(α * [v → µ(v)]) ≤d(α * ). Then by the Shift-down lemma, this new mapping has fewer losses, and thus attains a lower cost than α * .
Lemma 12.
Let v be a minimal ⊥-node under α, and let Proof. Let α * be an optimal completion of α. Denote s := µ α (v), and assume that α * (v) > s (as otherwise, we are done). Let α = α * [v → s]. We have that l(α ) < l(α * ) by the Shiftdown lemma. To prove the Lemma, we then show thatd(α ) ≤d(α * ). Suppose otherwise thatd(α ) >d(α * ). As only v changed mapping to s to go from α * to α , this implies that h α (s) > h α * (s) because of v. Since under α * , no ancestor of v is mapped to s, it must be that under α , v is the root of a subtree T of height h α (s) of duplications in s. Since T contains only descendants of v, it must also be that h αv (s) = h α (s) (here α v is the mapping defined in the Lemma statement). As we are assuming that h α (s) > h α * (s), we get h αv (s) > h α * (s). This is a contradiction, since h α * (s) ≥ h α (s) = h αv (s) (the left inequality because α * is a completion of α, and the right equality by the choice of α v ). Then l(α ) < l(α * ) andd(α ) ≤d(α * ) contradicts the fact that α * is optimal.
We say that a minimal ⊥-node v ∈ V (G) is easy (under α) if v falls into one of the cases described by Lemma 11 or Lemma 12. Formally, v is easy if either v is a speciation mapped to µ α (v) under any optimal completion of α (Lemma 11), ord(α) =d(α[v → µ α (v)]) (Lemma 12). Our strategy will be to "clean-up" the easy nodes, meaning that we map them to µ α (v) as prescribed above, and then handle the remaining non-easy nodes by branching over the possibilities. We say that a partial mapping α is clean if every minimal ⊥-node v satisfies the two following conditions: [C1] v is not easy; [C2] for all duplication nodes w (under α with α(w) = ⊥), either α(w) ≤ µ α (v) or α(w) is incomparable with µ α (v).
Roughly speaking, C2 says that all further duplications that may "appear" in a completion of α will be mapped to nodes "above" the current duplications in α. The purpose of C2 is to allow us to create duplication nodes with mappings from the bottom of S to the top. Our goal will be to build our α mapping in a bottom-up fashion in G whilst maintaining this condition. The next lemma states that if α is clean and some lowest minimal ⊥-node v gets mapped to species s, then v brings with it every minimal ⊥-node that can be mapped to s. Lemma 13. Suppose that α is a clean partial mapping, and let α * be an optimal completion of α. Let v be a lowest minimal ⊥-node under α, and let s := α * (v). Then for every minimal ⊥-node w such that µ α (w) ≤ s, we have α * (w) = s.
Suppose first that s = µ α (v). Note that since α is clean, v is not easy, which implies that h α (s) = h α (s) + 1. Since v is a lowest minimal ⊥-node, if w is a minimal ⊥-node such that µ α (w) ≤ s, we must have µ α (w) = s, as otherwise v would not have the 'lowest' property. Moreover, because v and w are both minimal ⊥-nodes under the partial mapping α, one cannot be the ancestor of the other and so v and w are incomparable. This implies that mapping w to s under α cannot further increase h α (s) (because we already increased it by 1 when mapping v to s). Thusd(α ) =d(α [w → s]), and w is easy under α and must be mapped to s by Lemma 12. This proves the α * (v) = µ α (s) case. Figure 4 An illustration of one pass through the algorithm. The species tree S is left and G has two trees (middle) and has partial mapping α (labels are the lowercase of the species). Here α is in a clean state, and the algorithm will pick a lowest minimal ⊥-node (white circle) and try to map it to, say, H. The forest on the right is the state of α after applying this and cleaning up. Now assume that s > µ α (v), and let w be a minimal ⊥-node with µ α (w) ≤ s. Let us denote s := α * (w). If s = s, then we are done. Suppose that s < s, noting that h α * (s ) > 0 (because w must be a duplication node, due to α being clean). If s = µ α (v), then w is also a lowest minimal ⊥-node. In this case, using the arguments from the previous paragraph and swapping the roles of v and w, one can see that v is easy in α[w → s ] and must be mapped to s < s, a contradiction. Thus assume s > µ α (v). Under α * , for each child v of v, we have α * (v ) ≤ µ α (v) < s , and for each ancestor v of v, we have α * (v ) ≥ α * (v) = s > s . Therefore, by remapping v to s , v is the only duplication mapped to s among its ancestors and descendants. In other words, because h α * (s ) > 0, we haved(α * [v → s ]) ≤d(α * ). Moreover by the Shift-down lemma, l(α * [v → s ]) < l(α * ), which contradicts the optimality of α * .
The remaining case is s > s. Note that h α * (s) > 0 (because v must be a duplication node, due to α being clean). Since it holds that v is a minimal ⊥-node, that α is clean and that s > µ α (v), it must be the case that α has no duplication mapped to s (by the second property of cleanness). In particular, w has no descendant that is a duplication mapped to s under α (and hence under α * ). Moreover, as s = α * (w) > s, w has no ancestor that is a duplication mapped to s. Thusd(α * [w → s]) ≤d(α * ), and the Shift-down lemma contradicts the optimality of α * . This concludes the proof.
We are finally ready to describe our algorithm. We start from a partial mapping α with α(v) = ⊥ for every internal node v of G. We gradually "fill-up" the ⊥-nodes of α in a bottom-up fashion, maintaining a clean mapping at each step and ensuring that each decision leads to an optimal completion α * . To do this, we pick a lowest minimal ⊥-node v, and "guess" α * (v) among the δ/λ possibilities. This increases some h α (s) by 1. For each such guess s, we use Lemma 13 to map the appropriate minimal ⊥-nodes to s, then take care of the easy nodes to obtain another clean mapping. We repeat until we have either found a complete mapping or we have a duplication height higher than d. An illustration of a pass through the algorithm is shown in Figure 4.
Notice that the algorithm assumes that it receives a clean partial mapping α. In particular, the initial mapping α that we pass to the first call should satisfy the two properties of cleanness. To achieve this, we start with a partial mapping α in which every internal node is a ⊥-node. Then, while there is a minimal ⊥-node v that is not a required duplication, we set α(v) = µ α (v), which makes v a speciation. It is straightforward to see that the resulting α is clean: C1 is satisfied because we cannot make any more minimal ⊥-nodes become speciations, and we cannot create any duplication node without increasing cost SD because α has no duplication. C2 is met because there are no duplications at all.
See below for the proof of correctness. The complexity follows from the fact that the algorithm creates a search tree of degree δ/λ of depth at most d. The main technicality R. Dondi, M. Lafond and C. Scornavacca
5:13
Algorithm 1 FPT algorithm for parameter d. 1: procedure SuperReconcile (G, S, α, d) G is the set of input trees, S is the species tree, α is a clean partial mapping, d is the maximum value ofd(α).
4:
else if α is a complete mapping then
Theorem 14. Algorithm 1 is correct and finds a minimum cost mapping
Proof. We show by induction over the depth of the search tree that, in any recursive call made to Algorithm 1 with partial mapping α, the algorithm returns the cost of an optimal completion α * of α havingd(α * ) ≤ d, or ∞ if no such completion exists -assuming that the algorithm receives a clean mapping α as input. Thus in order to use induction, we must also show that at each recursive call done on line 15, α is a clean mapping. We additionally claim that the search tree created by the algorithm has depth at most d. To show this, we will also prove that every α sent to a recursive call satisfiesd(α ) =d(α) + 1. The base cases of lines 3 -5 are trivial. For the induction step, let v be the lowest minimal ⊥-node chosen on line 7. By Lemma 9, if α * is an optimal completion of α and s = α * (v), then µ α (v) ≤ s ≤ par δ/λ (µ α (v)). We try all the δ/λ possibilities in the for-loop on line 9. The for-loop on line 11 is justified by Lemma 13, and the for-loop on line 13 is justified by Lemma 11 and Lemma 12. Assuming that α is clean on line 15, by induction the recursive call will return the cost of an optimal completion α * of α havinĝ d(α * ) ≤ d, if any such completion exists. It remains to argue that for every α sent to a recursive call on line 15, α is clean andd(α ) =d(α) + 1 .
Let us first show that such a α is clean, for each choice of s on line 9. There clearly cannot be an easy node under α after line 13, so we must show C2, i.e. that for any minimal 2 There is a subtlety to consider here. What we have shown is that if there exists a mapping α of minimum cost cost SD (G, S) withd(α) ≤ d, then the algorithm finds it. It might be that a reconciliation α satisfyingd(α) ≤ d exists, but that the algorithm returns no solution. This can happen in the case that α is not of cost cost SD (G, S).
5:14
Reconciling Multiple Genes with Segmental Duplications ⊥-node w under α , there is no duplication z under α satisfying α (z) > µ α (w). Suppose instead that α (z) > µ α (w) for some duplication node z. Let w 0 be a descendant of w that is a minimal ⊥-node in α (note that w 0 = w is possible). We must have µ α (w) ≥ µ α (w 0 ). By our assumption, we then have α (z) > µ α (w) ≥ µ α (w 0 ). Then z cannot be a duplication under α, as otherwise α itself could not be clean (by C2 applied on z and w 0 ). Thus z is a newly introduced duplication in α , and so z was a ⊥-node under α. Note that Algorithm 1 maps ⊥-nodes of G one after another, in some order (z 1 , z 2 , . . . , z k ). Suppose without loss of generality that z is the first duplication node in this ordering that gets mapped to α (z). There are two cases: either α (z) = s, or α (z) = s.
Suppose first that α (z) = s. Lines 10 and 11 can only map ⊥-nodes to s, and line 13 either maps speciation nodes, or easy nodes that become duplications. Thus when α (z) = s, we may assume that z falls into the latter case, i.e. z is easy before being mapped, so that mapping z to α (z) does not increase h α (α (z)). Because z is the first ⊥-node that gets mapped to α (z), this is only possible if there was already a duplication z 0 mapped to α (z) in α. This implies that α(z 0 ) = α (z) > µ α (w 0 ), and that α was not clean (by C2 applied on z 0 and w 0 ). This is a contradiction.
We may thus assume that α (z) = s. This implies µ α (w) < α (z) = s. If w was a minimal ⊥-node in α, it would have been mapped to s on line 11, and so in this case w cannot also be a minimal ⊥-node in α , as we supposed. If instead w was not a minimal ⊥-node in α, then w has a descendant w 0 that was a minimal ⊥-node under α. We have µ α (w 0 ) ≤ µ α (w) < s, which implies that w 0 gets mapped to s on line 11. This makes µ α (w) < s impossible, and we have reached a contradiction. We deduce that z cannot exist, and that α is clean.
It remains to show thatd(α ) =d(α) + 1. Again, let s be the chosen species on line 9. Suppose first that s = µ α (v). Then h α[v→s] (s) = h α (s) + 1, as otherwise v would be easy under α, contradicting its cleanness. In this situation, as argued in the proof of Lemma 13, each node w that gets mapped to s on line 11 or on line 13 is easy, and thus cannot further increase the height of the duplications in s. If s > µ α (v), then h α[v→s] (s) = 1 = h α (s) + 1, since by cleanness no duplication under α maps to s. Here, each node w that gets mapped on line 11 has no descendant nor ancestor mapped to s, and thus the height does not increase. Noting that remapping easy nodes on line 13 cannot alter the duplication heights, we get in both cases thatd(α[v → s]) =d(α) + 1. This proves the correctness of the algorithm.
As for the complexity, the algorithm creates a search tree of degree δ/λ and of depth at most d. Each pass can easily seen to be feasible in time O(δ/λ · n) (with appropriate pre-parsing to compute µ α (v) in constant time, and to decide if a node is easy or not in constant time as well), and so the total complexity is O( δ/λ d n · δ λ ).
Experiments
We used our software to reanalyze a data set of 53 gene trees for 16 eukaryotes presented in [10] and already reanalyzed in [19,1]. In [1], the authors showed that, if segmental duplications are not accounted for, we get a solution havingd equal to 9, while their software (ExactMGD) returns a solution withd equal to 5. We were able to retrieve the solution with maximum height of 5 fixing δ ∈ [28, 61] and λ = 1, but, as soon as δ > 61, we got a solution with maximum height of 4 where no duplications are placed in the branch leading to the Tetrapoda clade (see [19, Fig. 1 The species tree phylogeny for the yeast data set described in [2]. Numbers at the internal nodes are meaningless and are only used to refer to the nodes in the main text. We also reanalyzed the data set of yeast species described in [2]. First, we selected from the data set the 2379 gene trees containing all 16 species and refined unsupported branches using the method described in [12] and implemented in ecceTERA [11] with a bootstrap threshold of 0.9 and δ = λ = 1. Using our method with δ = 1.5, λ = 1 we were able to detect the ancient genome duplication in Saccharomyces cerevisiae already established using synteny information [13], with 216 gene families supporting the event. Other nodes with a signature of segmental duplication are nodes 7, 6, 13 and 2 (refer to Fig. 5) with respectively 190, 157, 148 and 136 gene families supporting the event. It would be interesting to see if the synteny information supports these hypotheses.
Conclusion
This work poses a variety of questions that deserve further investigation. The complexity of the problem when δ/λ is a constant remains an open problem. Moreover, our FPT algorithm can handle data sets with a sum of duplication height of about d = 30, but in the future, one might consider whether there exist fast approximation algorithms for MPRST-SD in order to attain better scalability. Other future directions include a multivariate complexity analysis of the problem, in order to understand whether it is possible to identify other parameters that are small in practice. Finally, we plan to extend the experimental analysis to other data sets, for instance for the detection of whole genome duplications in plants.
In the following, we will assume that λ = 0 and δ = 1. We reduce the Vertex Cover problem to that of finding a mapping of minimum cost for given G and S. Recall that in Vertex Cover, we are given a graph G = (V, E) and an integer β < n and are asked if there exists a subset V ⊆ V with |V | ≤ β such that every edge of E has at least one endpoint in V . For such a given instance, denote V = {v 1 , . . . , v n } and E = {e 1 , . . . , e m } (so that n = |V | and m = |E|). The ordering of the v i 's and e j 's can be arbitrary, but must remain fixed for the remainder of the construction. Let K := (n + m) 10 , and observe in particular that β < n K. We construct a species tree S and a gene forest G from G. The construction is relatively technical, but we will provide the main intuitions after having fully described it. For convenience, we will describe G as a set of gene trees instead of a single graph. Figure 6 illustrates the constructed species tree and gene trees. The construction of S is as follows: start with S being a caterpillar on 3n + 2 leaves. Let (x 1 , y 1 , z 1 , x 2 , y 2 , z 2 , . . . , x n , y n , z n , x n+1 ) be the path of this caterpillar consisting of the internal nodes, ordered by decreasing depth (i.e. x 1 is the deepest internal node, and x n+1 is the root). For each i ∈ [n], call p i , q i and r i , respectively, the leaf child of x i , y i and z i . Note that x 1 has two leaf children: choose one to name p 1 arbitrarily among the two. Then for each edge uv of S, graft a large number, say K 100 , of leaves on the uv edge (grafting t leaves on an edge uv consists of subdividing uv t times, thereby creating t new internal nodes of degree 2, then adding a leaf child to each of these nodes, see Figure 6). We will refer to these grafted leaves as the special uv leaves, and the parents of these leaves as the special uv nodes. These special nodes are the fundamental tool that lets us control the range of duplication mappings.
Finally, for each i ∈ {1, . . . , n}, replace the leaf p i by a tree P i that contains K distinguished leaves c i,1 , . . . , c i,K such that dist(c i,j , r(P i )) = K for all j ∈ [K], and such that dist(c i,j , lca(c i,j , c i,k )) ≥ K/2 for all distinct j, k ∈ [K]. By Lemma 15, each P i can be constructed in polynomial time. Note that the edges inside a P i subtree do not have special leaves grafted onto them. This concludes the construction of S. WA B I 2 0 1 8
5:18
Reconciling Multiple Genes with Segmental Duplications . . . We proceed with the construction of the set of gene trees G. Most of the trees of G consist of a subset of the nodes of S, to which we graft additional leaves to introduce duplications -some terminology is needed before proceeding. For w ∈ V (S), deleting w consists in removing w and all its descendants from S, then contracting the possible resulting degree two vertex (which was the parent of w if p(w) = r(S)). If this leaves a root with only one child, we contract the root with its child. For X ⊆ L(S), keeping X consists of successively deleting every node that has no descendant in X until none remains (the tree obtained by keeping X is sometimes called the restriction of S to X).
The forest G is the union of three sets of trees A, B, C, so that G = A ∪ B ∪ C. Roughly speaking A is a set of trees corresponding to the choice of vertices in a vertex cover, B is a set of trees to ensure that we "pay" a cost of one for each vertex in the cover, and C is a set of trees corresponding to edges. For simplicity, we shall describe the trees of G as having leaves labeled by elements of L(S) -a leaf labeled s ∈ L(S) in a gene tree T ∈ G is understood to be a unique gene that belongs to species s.
The A trees. Let A = {A 1 , . . . , A n }, one tree for each vertex of G. For each i ∈ [n], obtain A i by first taking a copy of S, then deleting all the special y i z i leaves. Then on the resulting z i y i branch, graft 10K leaves labeled q i . Then delete the child of z i that is also an ancestor of r i (removing z i in the process). Figure 6 bottom-left might be helpful.
5:19
As a result, under the LCA-mapping µ, A i has a path of 10K duplications mapped to y i . One can choose whether to keep this mapping in A i , or to remap these duplications to z i . The B trees. Let B = {B 1 , . . . , B n }. For i ∈ [n], B i is obtained from S by deleting all except 10K − 2 of the special r i z i leaves, and grafting a leaf labeled r i on the edge between r i and its parent, thereby creating a single duplication mapped to r i under µ.
we list the set of leaves that we keep from S. Keep all the leaves of the P i subtree of S, and keep a subset of the special leaves defined as follows: -keep 9K of the special x i y i leaves; -keep (j − i − 1)10K + K of the special y i z i leaves; -keep 9K of the special x j y j leaves; -keep all the special y j z j leaves. No other leaves are kept. Next, in the tree obtained by keeping the aforementioned list of leaves, for each k ∈ [K] we graft, on the edge between c h,k and its parent, another leaf labeled c h,k . Thus C h has K duplications, all located at the bottom of the P i subtree. This concludes our construction.
Let us pause for a moment and provide a bit of intuition for this construction. We will show that G has a vertex cover of size β if and only if there exists a mapping α of G of cost at most 10Kn + β. As we will show later on, two A i trees cannot have a duplication mapped to the same species of S, so these trees alone account for 10Kn duplications. The r i duplications in the B i trees account for n more duplications, so that if we kept the LCAmapping, we would have 10Kn + n > 10Kn + β duplications. But these r i duplications can be remapped to z i , at the cost of creating a path of 10K duplications to z i . This is fine if A i also has a path of 10K duplications to z i , as this does not incur additional height. Otherwise, this path in A i is mapped to y i , in which case we leave r i untouched, summing up to 10K + 1 duplications for such a particular A i , B i pair. Mapping the duplications of A i to y i represents including vertex v i in the vertex cover, and mapping them to z i represents not including v i . Because each time we map the A i duplications to y i , we have the additional r i duplication in B i , we cannot do that more than β times. Now consider a C h tree. Under the LCA-mapping, the c h,k duplications at the bottom enforce an additional K duplications. This can be avoided by, say, mapping all these duplications to the same species. For instance, we could remap all these duplications to some y i node of S. But in this case, because of Lemma 4, every node v of C h above a c h,k duplication for which µ(v) ≤ y i will become a duplication. This will create a duplication subtree D in C h with a large height, and our goal will be to "spread" the duplications of D among the y i and z i duplications that are present in the A i trees. As it turns out, this will be feasible only if some y i or y j has duplication height 10K, where v i and v j are the endpoints of edge e h corresponding to C h . If this does not occur, any attempt at mapping the c h,k nodes to a common species will induce a chain reaction of too many duplications created above. We now proceed with the details.
Proof. Let G and β be a given instance of vertex cover, and let G and S be constructed as above. Call a node v ∈ V (G) an original duplication if v is a duplication under µ. If µ(v) = s, we might call v an original s-duplication for more precision. For T ∈ G and t ∈ V (S), suppose there is a unique node w ∈ V (T ) such that µ(w) = s. We then denote w WA B I 2 0 1 8
5:20
Reconciling Multiple Genes with Segmental Duplications by T [t]. In particular, any special node that is present in a tree T ∈ G satisfies the property, so when we mention the special uv nodes of T , we refer to the special nodes that are mapped to the corresponding special uv nodes in S under µ. For example in the C h tree of Figure 6, the indicated set of (j −i−1)10K +K nodes are called special y i z i nodes as they are mapped to the special y i z i nodes of S under µ.
We now show that G has a vertex cover of size β if and only if G and S admit a mapping α of cost at most 10Kn + β.
(⇒) Suppose that V = {v a1 , . . . , v a β } is a vertex cover of G. We describe a mapping α such that for each i ∈ [n]: all its internal node descendants duplications. We map them to the set {y j } ∪ j−1 l=i {y l , z l }. This case is illustrated in Figure 7.
More specifically, the longest path from w to a descending internal node contains 9K + (j − i − 1)10K + K + 9K + K = (j − i + 1)10K, where we have counted the special x j y j nodes, the special y i z i nodes, the special x i y i nodes and the P i subtree nodes. We have h α (y j ) + j−1 l=i (h α (y l ) + h α (z l )) = 10K + (j − i)10K = (j − i + 1)10K, just enough to map the whole subtree rooted at w to duplications. It is easy to see that such a mapping can be made valid by first mapping the 9K special x j y j nodes to y j , then the other nodes descending from w to the rest of {y j } ∪ j−1 l=i {y l , z l }. This is because all these nodes are ancestors of the special y i z i nodes, the special x i y i nodes and the P i nodes (except y i , but we have h α (y i ) = 0 anyway).
We have constructed a mapping α with the desired duplication heights, concluding this direction of the proof. Let us proceed with the converse direction.
(⇐): suppose that there exists a mapping α of the G trees of cost at most 10Kn + β. We show that there exists a vertex cover of size at most β in G. For some X ⊆ V (S), define h α (X) = x∈X h α (x). For each i ∈ [n], define the sets Our goal is to show that the Y i 's for which h α (Y i ) > 9K correspond to a vertex cover. The proof is divided into a series of claims.
Proof. Consider the A i tree of A, and let y * i be an original y i -duplication in A i . If α(y * i ) ≥ x i+1 , then every node of A i on the path from y * i to A i [x i+1 ] is a duplication. This includes all the K 100 special z i x i+1 nodes, contradicting the cost of α. Thus y i ≤ α(y * i ) < x i+1 . | 2018-06-11T13:57:30.000Z | 2018-06-01T00:00:00.000 | {
"year": 2019,
"sha1": "b912d39c1ddb189b549d87386c8e356658bf0108",
"oa_license": "CCBY",
"oa_url": "https://almob.biomedcentral.com/track/pdf/10.1186/s13015-019-0139-6",
"oa_status": "GOLD",
"pdf_src": "ArXiv",
"pdf_hash": "588bf9e3bcee36eba281551a45ce03310a91a274",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": [
"Mathematics",
"Medicine",
"Biology",
"Computer Science"
]
} |
245019090 | pes2o/s2orc | v3-fos-license | Utilization of Tephrosia vogelii in post-mining land reclamation
Reclamation is an effort to repair the damaged mining sector to provide optimal benefits. This review article aims to propose a natural product technology approach to the reclamation of post-coal mining land by increasing the content of soil organic matter so that the post-mining land becomes fertile and can be used for agricultural land. The method used is an observation and literature study. Literature searches were using them ScienceDirect, Scopus, PubMed, and Google Scholar databases. Land-use change of post-mining site to agricultural land provides prospective economic value to the community. However, the land quality of the post-mining area should be improved before food crop production. Tephrosia vogelii is a member of the Papilionoideae subfamily is important for land revegetation. They can fertilize the soil because of mutualistic symbiosis with the bacteria Rhizobium sp. which is capable to fix nitrogen in the soil and hence provide available nitrogen. Thus, the use of post-mining land as agriculture can improve the economy and sustainability as well as protect life and the environment with natural product technology.
Introduction
Land requirements are increasing with the increase of population. Likewise, land problems increasing with more intensive mining. A lot of lands have been neglected and degraded after mining activities. Before changing land use of mining area to another use, especially agriculture, rehabilitation of postmining land is required. The mining sector in Indonesia contributes to forest destruction up to 10%, now 2 million hectares per year are destructed [1]. The forest vegetations are often intended for revegetation and further reclamation of post-mine land used since mining land is usually located in the forest [2]. After land reclamation and rehabilitation, the post-mining land can be used for agriculture, ecotourism, settlements or wetlands.
In carrying out post-mining land reclamation, things that need to be considered and carried out in rehabilitating/reclaiming post-mining land are the impact of changes from mining activities, land reconstruction, revegetation, prevention of acid mine drainage, drainage arrangements, and postmining land use [3]. Returning the land to its initial conditions takes a long time, due to damaged soil components and low organic matter content causing low activity and microbial population [4]. The selection of plants that are suitable for land conditions is one of the determinants of reclamation success. Plants that are often chosen for soil revegetation on post-mining areas are usually planted with Leguminosae, Gamal (Gliricidia sepium) and sengon (Albizia chinensis) [5]. However, the selection of plants must be in accordance with the conditions of the post-mining land.
The genus Tephrosia belongs to the family Leguminosae or Fabaceae and the sub-family Papilionoideae. According to Polhill, et al. [6], which are shrubs, herbs, and trees. The genus Tephrosia is known to have the ability to fertilize the soil because it can symbiotic mutualism with the bacteria Rhizobium sp. which can fix nitrogen in the soil, as reported by Harborne [7]. Tephrosia is discovered all through tropics and subtropics.
Tephrosia vogelii is one of the Tephrosia species. In West Java, this species is known as pig nut or kacang babi, as reported by Susanto [8]. This plant is known as a woody herb with branching stems that are downy white or brown, or a small tree with green leaves, spirally arranged and dense, ranging from 0.5-4 m in height. This study aims to propose a natural product technology approach to the reclamation of post-coal mining land with an increase in the content of soil organic matter so that the post-mining land becomes fertile and can be used for agricultural land
Method
The type of data used by the author in this study is primary data by making observations and observations, while secondary data is data obtained from journals, documentation books, and other references related to the study of revegetation plant selection for the success of reclamation of postmining land. The collected data is then analyzed using descriptive analysis methods. The descriptive analysis method is a research method that is carried out by describing the facts that are later followed by analysis, and providing adequate understanding and explanation. Literature searches were using the ScienceDirect, Scopus, PubMed, and Google Scholar [9]. The research data is then presented systematically, starting from the criteria for selecting tree species for post-mining land, revegetation plant types based on historical land use and describing examples of post-mining land reclamation, especially coal mining, which has been widely practiced by large companies engaged in mining in Indonesia, and using other keywords that are natural product technology, Tephrosia vogelii, soil organic matter, fixing nitrogen by Rhizobium sp., fertile agricultural land.
Characteristics of post-mining area
Many activities in coal mining can damage structure, texture, porosity and bulk density as important soil physical characteristics for plant growth. Each phase in this process is believed to be associated with different sets of environmental damage covering physical, chemical and biological effects to mining land. The ex-mining area could be classified as degraded land, referring to the land that loses its natural productivity due to a decrease in land quality, from the point of view of agriculture land usage. The degraded land of the post-mining area comprises the deterioration of the physical and chemical properties of the soil, a drastic species reduction in some flora, fauna and soil microorganisms. The poor land quality causes the decline in soil fertility and leads to a more compacted structure, which physically affects its capability of performing penetration, a condition which is unfavorable for the plant to grow. For this purpose, it is critical in restoring mined lands to repair the soil quality through a reclamation process, or even to improve the land productivity after surface mining. As an important part to remediate the land degradation, there are many methods of reclamation, such as revegetation, and improving land quality by using fertilizer to increase essential nutrients and reach the optimum of pH levels by adding an appropriate amount of lime based on the soil type and intended pH, as well as the use of microbes for soil rejuvenation. The objective of the treatments is to enhance soil health and quality [10]. The chemical characteristics of the soils such as pH, organic matter, exchangeable bases, effective cation exchange capacity, and exchange acidity are found to be conducive to rice production and other agricultural crops; the soil quality can improved by farmers and therefore soil cannot be permanent limitations for rehabilitation of post-mining area [11].
Land revegetation
The definition of the post-mining is an activity planned, systematic, and continue after the end of the part or all mining activities to restore the function of the natural environment and social tasks according to the local conditions in the mining area [12]. This is in accordance to Mining Law No. [15].
Post-mining land actually has the opportunity to be used as agricultural areas to meet food needs and overcome post-mining environmental problems. Agricultural technology innovation interventions are believed to be able to improve the physical, chemical, and soil's biological properties, allowing it to be used as an ideal medium for agricultural cultivation. Apart from the biophysical aspects, reclamation efforts should also consider the socio-economic aspects of the community, such as land ownership status, farmer knowledge and skills, and farming cost feasibility. As a result, it is critical to develop and operationalize the reclamation of post-mining land for agricultural use [16].
The problem in post-mining land is low productivity due to poor physical and chemical properties of the soil, due to low water absorption, large porosity, infertile soil due to the pH of the soil that is too acidic. In addition, low nitrogen and phosphorus, cation exchange capacity (CEC) in the soil, and low alkaline (K, Ca, Mg, and Na) and Al, in addition to the soil's high concentration of heavy metals and toxic compounds. Many ways can be done to repair the damage caused by mining, one of them is the Legum Cover Crop (LCC) which can live on damaged soil and the incorporation of organic compounds capable of improving the physical and chemical properties of the post-mining soil. Plant roots will grow if the soil has appropriate air and circulation. Good circulation will occur if the soil has a loose consistency and the soil structure has developed [17].
Soils with high organic matter content generally have loose consistency. Dense and hard soil due to loads from heavy equipment will greatly reduce the air infiltration rate, the amount of air and air in the soil, and of course inhibit the growth of plant roots. It is recognized that applying high doses of organic matter or compost to the soil or planting holes to increase soil looseness is often difficult to meet due to materials. Therefore, the use of organic material substitute compounds, such as humic compounds can be done. Chemical fertility is related to the availability of nutrients and the optimum level of soil acidity for plant growth. To increase the success of revegetation, topsoil is often sprinkled with a thickness of 50 -100 cm over the post-mining land that has been laid out with the assumption that the topsoil is chemically and physically fertile. The topsoil for reclamation is very infertile soil, as reflected by the very low organic matter content with the damaged soil structure making it easy to compact [18].
The improvement the chemical fertility of topsoil can be done by combining the use of compost and basic fertilizers that are commonly used. The humic acid compounds as a substitute for compost have significantly improved the performance of ground cover crops. Chemical fertility in soil with acidic pH is strongly influenced by the presence of toxic elements for plant growth caused by the oxidation of sulfide derivative compounds, such as pyrite. Sulfur can be toxic to plants. Sulfur dioxide in the atmosphere in any sizable concentration kills plants [19]. The chemical properties of soil containing sulfide minerals can be identified by measuring the pH in H 2 O 2 [20]. In very acidic environmental conditions, heavy metals in overburden or tailings will be more easily dissolved and carried away by surface runoff, thereby contaminating surface water and subsurface water. In such
Tephrosia vogelii (T. vogelii) for land revegetation
Legum cover crop plants on post-mining land reclamation are useful for protecting soil from erosion damage and for producing large amounts of organic matter, this Leguminosae plant has roots that can improve soil physical and chemical properties and able to absorb heavy metals on post-mining land, one of the Leguminosae plants that are often used is T. vogelii. In addition to the use of cover crop legumes, to increase soil fertility, it is necessary to add organic material, organic matter is a very important material in creating soil fertility, both in terms of physics, chemistry, and soil biology, so that plants can produce secondary metabolites [22]. Organic matter is an excellent aggregate stabilizer. Part of the soil's cation exchange capacity (CEC) comes from organic matter [17].
Tephrosia vogelii is a leguminous plant that grows a lot in Africa, there are bacteria Rhizobium japonicum that live in a symbiotic mutualism in its roots that can improve the nitrogen content in the soil. Tephrosia can improve soil fertility, like firewood, as an insecticide. Besides that, it is also used as a medicine for skin diseases and intestinal worms. Leaves of this plant contain rotenone one of the toxic compounds to fish and ruminants [23]. [24].
Tephrosia vogelii is a shrub, grows in savanna, grasslands, forest edges and shrubs, deserts, and empty fields and can grow in acid soils, with the presence of nodules this plant can fix nitrogen from the environment and release it slowly in the soil where so that this plant can grow on poor soil and is not susceptible to disease because of the content of chemical compounds that cause plants to survive The root arrangement of T. vogelii is generally very good, the root growth is straight into the soil and has many branch roots, at the root nodules containing the bacterium R. japonicum, which can bind N from the air which is then used to fertilize land [7]. Good pH tolerance as a condition for growing T. vogelii plants, which is between 5 -6.5, grows well on various types of soil, especially Andosol soil with good drainage and soil aeration. T. vogelii can grow well in places with a temperature range of 12-27 o C, in places that are open and with rainfall of 850-2,650 mm3 per month. Therefore T. vogelii is mostly grown in areas with a sea level of 2,100 m [23].
Tephrosia vogelii can be used as a crop plant for sustainable agriculture because the plant can reduce the use of pesticides, herbicides, and fertilizers [26], even this plant can withstand raindrops that fall directly and reduce the speed of water flow on the soil surface, restraining the rate of erosion, soil, adding soil organic matter produced from the remains of leaves, stems, and roots, carrying out transpiration so that it can withstand evaporation from the soil directly [27], increasing soil fertility where this plant has root nodules resulting from the symbiosis of plant roots with Rhizobium bacteria that can bind nutrients N from the air so that it can be available for farming [7].
Tephrosia vogelii can suppress weed growth because it has allelopathic properties. This is very beneficial in terms of crop cultivation because it can reduce the cost of chemical weed control and indirectly creates sustainable agriculture. Another advantage of this T. vogelii plant is that it can suppress or reduce diseases caused by fungi or bacteria the same as Cassia grandis [28] [29]. Postmining land reclamation methods can be updated and developed with information system technology with Structured Query Language (SQL) for the manufacture of artificial intelligence [30] and linked to an Internet of Things (IOT) based website that is connected to sensors and soil fertility measurement tools, water control, weed control, pest and disease control, thereby increasing capability, accuracy with the use of more diverse types of parameters [31].
Conclusions
Tephrosia vogelii which is one of the leguminous plants has advantages in helping post-mining land reclamation in terms of increasing soil fertility, controlling water, controlling weeds, controlling pests, diseases and creating sustainable agriculture. | 2021-12-11T20:07:34.885Z | 2021-11-01T00:00:00.000 | {
"year": 2021,
"sha1": "26d05b9a6cef553651915bb032652c17a1f3deb6",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1088/1755-1315/882/1/012067",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "26d05b9a6cef553651915bb032652c17a1f3deb6",
"s2fieldsofstudy": [
"Agricultural And Food Sciences"
],
"extfieldsofstudy": [
"Physics"
]
} |
252120077 | pes2o/s2orc | v3-fos-license | Boosting Bismuth(III) Complexation for Targeted α‐Therapy (TAT) Applications with the Mesocyclic Chelating Agent AAZTA
Abstract Targeted α therapy (TAT) is a promising tool in the therapy of cancer. The radionuclide 213BiIII shows favourable physical properties for this application, but the fast and stable chelation of this metal ion remains challenging. Herein, we demonstrate that the mesocyclic chelator AAZTA quickly coordinates BiIII at room temperature, leading to a robust complex. A comprehensive study of the structural, thermodynamic and kinetic properties of [Bi(AAZTA)]− is reported, along with bifunctional [Bi(AAZTA‐C4‐COO−)]2− and the targeted agent [Bi(AAZTA‐C4‐TATE)]−, which incorporates the SSR agonist Tyr3‐octreotate. An unexpected increase in the stability and kinetic inertness of the metal chelate was observed for the bifunctional derivative and was maintained for the peptide conjugate. A cyclotron‐produced 205/206Bi mixture was used as a model of 213Bi in labelling, stability, and biodistribution experiments, allowing the efficiency of [213Bi(AAZTA‐C4‐TATE)]− to be estimated. High accumulation in AR42J tumours and reduced kidney uptake were observed with respect to the macrocyclic chelate [213Bi(DOTA‐TATE)]−.
I.2 Synthetic Procedures
The protected and oxidised octapeptide linked to the Wang resin 1, precursor of octreotate, was prepared by standard Fmoc protocol as described by A. L. Petersen et al. in an automatic peptide synthesiser. [3,4] As starting material Wang Resin preloaded with H-Thr(tBu)-OH (115 mg, 0.87 mmol/g) was used. The bifunctional chelating agent (BFCA) AAZTA-C4-COOH tetra t-butyl ester was prepared as described by Manzoni et al. [5] Compound 1 and AAZTA-C4-COOH tetra t-butyl ester (134 mg, 0.200 mmol, 2 eq) were suspended in DMF (2 mL) in a SPPS manual reactor. The coupling was performed with HATU (76.0 mg, 0.200 mmol, 2 eq) and collidine (48.5 mg, 0.400 mmol, 4 eq) at room temperature in 3 h. After completion of the reaction, the resin was extensively washed with DMF. The peptide was cleaved from the resin and fully deprotected by treatment with 2 mL of the mixture TFA/H 2 O/TIS (95:2.5:2.5) at room temperature for 18 h. The product (AAZTA-C4-TATE) (Figure 3) was purified by preparative HPLC using as eluent a variable mixture of (A) 0.1% TFA in H 2 O, (B) 0.1% TFA in CH 3 CN. Gradient profile: isocratic at 25% of B for 15 min, linear gradient from 25% to 45% of B in 5 min, linear gradient from 45% to 100% in 5 min. Flow rate; 20 mL/min. The same elution program was applied to the HPLC analysis to determine the purity of the final product, obtained as a single peak. (Figure S1) The solvent was removed in vacuo and the product finally lyophilized after dissolution in water to give a white solid (21.
II.1. Experimental
The chemicals used for the experiments were of analytical grade. Bi(ClO 4 ) 3 solution was prepared by dissolving Bi 2 O 3 (99.9%, Fluka) in excess HClO 4 . The concentration of the Bi(ClO 4 ) 3 solution was determined by complexometry with the standardized Na 2 H 2 EDTA solution and xylenol orange as indicator at about pH 1. The H + concentration of the Bi(ClO 4 ) 3 solution was determined by using pH potentiometric titration in the presence of twofold excess of Na 2 H 2 EDTA. The concentration of the ligand stock solutions was determined by pH-potentiometric titration in the presence and absence of a large (40-fold) excess of CaCl 2 . The pH-potentiometric titrations were made with standardized 0.2 M NaOH. The concentration of the AAZTA-C4-TATE stock solution was determined by spectrophotometry with a standardized Bi(ClO 4 ) 3 solution by following the formation of the [Bi(AAZTA-C4-TATE)]complex at 280 nm and pH=7 (formation of [Bi(AAZTA-C4-TATE)]complex is rapid process). The solutions of the Bi III complexes were prepared by the reaction of equivalent amounts of Bi(ClO 4 ) 3 and the ligand. The Bi(ClO 4 ) 3 solution was added dropwise to the vigorously mixed solution of AAZTA, AAZTA-C4-COOH or AAZTA-C4-TATE. The pH of the solution was controlled, and it was set to 6.5-7.0 by addition of concentrated NaOH solution. Caution! Dry powder perchlorate samples are prone to explosion and should be handled with special care, especially in larger quantities. Titrando titration workstation equipped with a Metrohm-6.0234.110 combined electrode was used for pH measurements and titrations. Equilibrium measurements were carried out at a constant ionic strength (0.15 M NaClO 4 ) in 6 mL samples at 25 °C. The solutions were stirred, and N 2 was bubbled through them. The titrations were made in the pH range of 1.7-12.0. KH-Phthalate (pH=4.005) and borax (pH=9.177) buffers were used to calibrate the pH meter, For the calculation of [H + ] from the measured pH values, the method proposed by Irving et al. was used as follows. [6] A 0.01M HClO 4 solution was titrated with standardized NaOH solution at 0.15 M NaClO 4 ionic strength. The differences (A) between the measured (pH read ) and calculated pH (-log[H + ]) values were used to obtain the equilibrium H + concentration from the pH values measured in the titration experiments (A=-0.040). For the equilibrium calculations, the stoichiometric water ionic product (pK w ) was also needed to calculate [OH -] values under basic conditions. The V NaOH -pH read data pairs of the HClO 4 -NaOH titration obtained in the pH range 10.5-12.0 were used to calculate the pK w value (pK w =13.68). The stability constant of [Bi(AAZTA)]and [Bi(AAZTA-C4-COO -)] 2was determined by studying the competition reaction between AAZTA 4or AAZTA-C4-COO 5and NTA 3for Bi III with spectrophotometry at the absorption band of Bi III -complexes at pH=7. Prior to CZE analysis all buffers were filtered through a 0.45 m syringe filter and stored in refrigerator at +4ºC. In CZE the sample solutions were introduced at the anodic end of the capillary in normal mode (50 mbar, 6 s). The effective length of the capillary was 56 cm. The capillary was preconditioned with the buffer electrolyte (150 mM disodium hydrogen phosphate, pH=7.4) for 3 minutes. The separation was performed at 10 °C with the application of 20 kV voltage. After analysis, postconditioning (0.1 M NaOH (3 min) and buffer washing (3 min) were run to remove all possibly adsorbed materials from the capillary. In all measurements, 5 mM DMSO as internal standard was applied in order to correct the migration time of components on the electropherogram. The detection was carried out by on-column DAD measurement at 200 nm. The electropherograms were recorded and processed by ChemStation B.04.02 version (Agilent). Table S1, in each case with a precision better than 4%. The LOD and Linear range values obtained at 200 nm are summarized in Table S1. i=1, 2 …6 (S1) 25.51 (logKi H -logK3 H ) a Ref. [8] ; b Ref. [9] ; c Ref.
The protonation sequence of AAZTA 4was determined in detail by both spectroscopic and potentiometric methods. [9] The first protonation occurs at both endo-and exocyclic nitrogen atoms of the ligand (all nitrogen atoms are partially protonated). The second proton binds to the exocyclic nitrogen, with a simultaneous shift of the first proton to the exocyclic nitrogen owing to the electrostatic repulsion between the protonated sites. Further protonations take place at the ring-carboxylate groups located on the unprotonated exocyclic nitrogen, the unprotonated exocyclic nitrogen and/or the carboxylate pendant arms, respectively. According to the similarities It can be assumed that the protonation sequence of AAZTA-C4-COOH is similar to that of the parent AAZTA. A comparison of the protonation constants obtained in the presence of the same background electrolyte (0.15 M NaClO 4 , Table S2) indicated that the logK 3 H value of AAZTA-C4-COOH is slightly larger than the corresponding protonation constants of AAZTA. By taking into account the protonation constant of n-valeric acid (logK 1 H =4.69, 1.0 M NaClO 4 , 25C). [13] it is reasonable to assume that the 3 rd protonation of AAZTA-C4-COOH occurs at the side-chain carboxylate. The comparison of the logK i H values in Table S2 reveals that the logK 1 H value of AAZTA determined in 0.15 M NaClO 4 or NaCl solution is significantly lower than the value obtained in 0.1 M KCl or Me 4 NCl solutions due to the competition between the first protonation process and the Na + -complexes formed in the presence of Na + ion. [9] II.3 Equilibrium Studies of the Bi III -AAZTA -NTA system.
The stability, protonation and conditional stability constants of the Bi III -complexes formed with AAZTA, AAZTA-C4-COOH, DTPA, DOTA and NTA ligands are defined by Eqs. (S2) -(S6) The stability constants of the Bi III -polyamino-polycarboxylate complexes are generally determined by spectrophotometry with the use of the competition reactions by Brion, [12,14] by H + or by OHions at very acidic and basic conditions, respectively. [15,16]
IV.2 1 H and 13 C NMR Studies of [Bi(AAZTA)] -Complex.
In order to acquire a deeper insight into the solution structure of [Bi(AAZTA)] -, variable temperature (VT) multinuclear 1D and 2D NMR studies have been performed. 1D and 2D NMR studies of [Bi(AAZTA)]obtained in the temperature range 273-333 K are shown in Figures S11-S15. The assignment of the 1 H and 13 C NMR signals of the [Bi(AAZTA)]complex is shown in Figure S10-S12.
IV.3 1 H and 13 C NMR Studies of [Bi(AAZTA-C4-COO -)] 2--Complex.
In Single crystal XRD data collections were performed at the X-ray diffraction beamline (XRD1) of the Elettra Synchrotron, Trieste (Italy). [19] The crystals were dipped in NHV oil (Jena Bioscience, Jena, Germany) and mounted on the goniometer head with kapton loops (MiTeGen, Ithaca, USA). Complete datasets were collected at 100 K (nitrogen stream supplied through an Oxford Cryostream 700 -Oxford Cryosystems Ltd., Oxford, United Kingdom) through the rotating crystal method. Data were acquired using a monochromatic wavelength of 0.700 Å, on a Pilatus 2M hybrid-pixel area detector (DECTRIS Ltd., Baden-Daettwil, Switzerland). The diffraction data were indexed and integrated using XDS. [20] Two different datasets have been merged for [Bi(HAAZTA)], collected from random orientations of the same crystal. Semi-empirical absorption corrections and scaling were performed on datasets, exploiting multiple measures of symmetry-related reflections, using SADABS program. [21] Both the samples were prone to radiation damage, especially for the thin needles of ([Bi(AAZTA)] -), that show also limited diffraction power (best dataset resolution of ~0.86 Å). The structures were solved by the dual space algorithm implemented in the SHELXT code. [22] Fourier analysis and refinement were performed by the fullmatrix least-squares methods based on F 2 implemented in SHELXL (Version 2018/3) [23] . The Coot program was used for modeling. [24] Anisotropic thermal motion refinement have been used for all atoms with occupancy greater than 50%. Hydrogen atoms were included at calculated positions with isotropic U factors = 1.2·U eq or U factors = 1.5·U eq for methyl and hydroxyl groups (U eq being the equivalent isotropic thermal factor of the bonded non hydrogen atom). Hydrogen atoms for solvent water molecules have not been included in the refined models since it was not possible to locate them unambiguously in electron-density peaks of Fourier difference maps (contributions of these missing H atoms are still included in the properties reported in Table S4). Geometric and thermal motion parameters restrains (DFIX, DANG, SIMU) have been applied on disordered and poorly defined fragments. Pictures were prepared using Ortep3, [25] CCDC Mercury [26] and Pymol [27] software. Essential crystal and refinement data are reported below (Table S4) In both [Bi(AAZTA)]complexes, Bi III is octacoordinated by 7 donor atoms of the AAZTA chelating agent (three amino N and four carboxylate O donor atoms), whereas the eight coordination site is occupied by a carboxylate bridging two neighbor metal centers, as previously reported for [Gd(AAZTA)] -. [28] Selected bond distances and angles for [Bi(AAZTA)]are shown in Table S5. Entries with * refer to oxygens coordinated to Bi III -ions, whereas ** refer to oxygens of carboxylates bridging two Bi III centers. Naming scheme in use is reported in Figure S22 (solvent water molecules and disordered fragments are omitted for clarity). Figure S23). Table S6 (Several hydrogen bonds involved by the solvent water molecules ("O" atom labels) are excluded, since the corresponding hydrogens could not be located into electron density maps). Figure S22B). The [Bi(HAAZTA)] complex form dimers with units related by crystallographic inversion centers and linked by hydrogen bonds between one protonated and one deprotonated carboxylate groups (d OH···O = 2.508(5) Å, Figure S24B). Selected bond distances and angles for [Bi(HAAZTA)] are shown in Table S7. The lengths of the carboxylate C-O bond are also reported to confirm the assignment of their protonation states. Entries with * refer to oxygens bound to a Bi III -ion, while ** refer to the protonated oxygen atoms that give rise to H-bonded dimers formed by [Bi(HAAZTA)]. Naming scheme in use is reported in Figure S22 (solvent water molecules and disordered fragments are omitted for clarity). Based on the Fourier electron density maps and the length of C-O bond (Table S7), the proton of [Bi(HAAZTA)] complex is localized on the non-coordinated ring carboxylate-O atom. Table S8. (Several hydrogen bonds involved by the solvent water molecules ("O_11/12/13" atom labels) are excluded, since the corresponding hydrogens could not be located into electron density maps).
VI.1 Material and Methods
Suprapur nitric acid (s.p. HNO 3 ), ≥ 96% ethanol and the human plasma type AB from male AB were purchased from Merck. Ultrapure ammonium acetate (CH 3 CCONH 4 ) was supplied by VWR. TK-200 resin (a trioctyl-phosphine oxide (TOPO) based extractant) was bought from TRISKEM. The Strata-X 33u Polymeric Reversed Phase cartridge was purchased from Phenomenex. Glass macrofiber chromatography paper impregnated with silica gel (iTLC-SG) was bought from Agilent Technologies. All other reagents and solvents were bought from ROTH. Labelling experiments were followed by Raytest miniGita Star TLC scanner with beta-positron detector.
VI.2 Production of 205/206 Bi III
The 205/206 Bi isotope mixture was produced in a GE PETtrace cyclotron with 16 MeV proton beam on natural Pb-foil target (99.995%, 0.9 by 0.9 cm, 0.25 mm thick). A 60 min irradiation with a 10 μA beam current yielded approx. 60 MBq activity. After a 24 hours decay period, the irradiated Pb target was dissolved in 7 M suprapur HNO 3 (2 mL). Solutions were prepared with ultrapure (u.p.) water. Pb(NO 3 ) 2 precipitation was observed after the solution was concentrated to circa 1 mL. The solution was separated from the solid and diluted to 10 mL with u.p. water and filtered with Millipore 0.
VI.4 Production of [ 205/206 Bi(AAZTA-C4-TATE)]and [ 205/206 Bi(DOTA-TATE)] -
For biodistribution studies HEPES (50 µmol, pH=3.01) buffer and AAZTA-C4-TATE (6.76 nmol) solution was added to a 100 µL solution of 205/206 Bi III . The radiolabelling experiments were performed at 95 °C in closed 1.5 mL Eppendorf tubes in a dry bath for 1 hour. Then mixture was passed through a Strata-X 33u Polymeric Reversed Phase that had previously been conditioned by purging with ethanol, u.p. water (5 mL) and 0.1 M HEPES (1 mL). After the adsorption, the cartridge was purged with water (1 mL). The labelled product was eluted with ethanol and evaporated to dryness under nitrogen gas flow at 95°C in a heating block. The product was dissolved in 1 mL PBS buffer for ex vivo measurements and in 50 µL PBS for the stability tests.
VII.1 Animals and Animal Housing
16 weeks old, 21.2±1.4 g weighted female CB17 SCID mice (n=9; Animalab Ltd, Budapest, Hungary) were used for the experiments. Animals were housed under sterile conditions in IVC cages (Techniplast, Italy) at 23±2°C with 50±10% humidity and artificial lighting with a circadian cycle of 12 h. Sterile semi-synthetic diet (VRF1; Akronom Ltd., Budapest, Hungary) and sterile drinking water were available ad libitum to all animals. The animal experiments were authorized by the Ethical Committee for Animal Research, University of Debrecen, Hungary (permission number: 19/2017/DEMÁB).
VII.2 Cellular uptake studies
The . Decay-corrected radiotracer uptake was expressed as counts/(min*(10 6 cells)) (cpm). The uptake of 205/206 Bi-labeled probes was expressed as percentage of the total radioactivity of radiotracers added to the cells (%ID/10 6 cells). Each experiment was performed in triplicate and the displayed data represents the means of at least three independent experiments (± SD). In our in vitro studies for the comparison of the cellular uptake of the two somatostatin specific radiotracer, it was found that the accumulation of [ 205/206 Bi(AAZTA-C4-TATE)]in AR42J cells was somewhat higher (29.76±10.84 %ID/10 6
VII.3 Ex vivo Biodistribution Studies of free 205/206 Bi III
For the determination of the localization of the dissociated bismuth isotope, ex vivo biodistribution studies were performed using 205/206 Bi radiotracer. Healthy control 12-week old female BALB/c mice (n=3) were injected intravenously with 2.5±0.2 MBq of 205/206 Bi(NO 3 ) 3 . After 30 and 90 min incubation time mice were euthanized with 5% Forane. Tissue samples were taken from selected organs and the weight and the radioactivity of the samples were measured with calibrated gamma counter and the uptake was expressed as %ID/g tissue ( Figure S29). The ex vivo (%ID/g) data indicate that relatively high radioactivity in the liver, spleen, and kidneys only. Moderate accumulation was observed in the lungs and gall bladder at 30 and 60 min time point ( Figure S29). In contrast, [ 205/206 Bi(AAZTA-C4-TATE)]was only accumulated in the AR42J tumour and kidneys due to its excretion through the urinary system (Figure 7). | 2022-09-09T06:18:07.603Z | 2022-09-08T00:00:00.000 | {
"year": 2022,
"sha1": "079c7281f5a3427d745f181751c0d31d225b79ad",
"oa_license": "CCBYNCND",
"oa_url": null,
"oa_status": "CLOSED",
"pdf_src": "PubMedCentral",
"pdf_hash": "7382bd5f9e17616cdf5d8f37b7c9ea6e4a5f3af2",
"s2fieldsofstudy": [
"Chemistry",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
267723221 | pes2o/s2orc | v3-fos-license | Teaching Teamwork in Engineering and Computer Science
—For a hundred years, there has been no change in the basic structure of the electrical power grid. Experiences have shown that the hierarchical, centrally-controlled grid of the twentieth century is ill-suited to the needs of the twenty-first. To address the challenges of the existing power grid
INTRODUCTION
Today's electrical infrastructure has remained unchanged for about a hundred years.The components of the hierarchical grid are near to the end of their lives.While the electrical grid has been ageing, the demand for electricity has gradu-ally increased.According to the U.S. Department of Energy report, the demand and consumption for electricity in the U.S. have increased by 2.5 % annually over the last twenty years [1].Today's electric power distribution network is very complex and ill-suited to the needs of the twenty-first century.Among the deficiencies are a lack of automated analysis, poor visibility, mechanical switches causing slow response times, lack of situational awareness, etc. [2].These have contributed to the blackouts happening over the past 40 years.Some additional inhibiting factors are the growing population and demand for energy, the global climate change, equipment failures, energy storage problems, the capacity limitations of electricity generation, one-way communication, decrease in fossil fuels and resilience problems [5].Also, the greenhouse gas emissions on Earth have been a significant threat that is caused by the electricity and transportation industries [6].Consequently, a new grid infrastructure is urgently needed to address these challenges.
To realize these capabilities, a new concept of next genera-tion electric power system, the smart grid, has emerged.The smart grid is a modern electric power grid infrastructure for improved efficiency, reliability and safety, with smooth integration of renewable and alternative energy sources, through automated control and modern communications technologies [1], Published by: Longman Publishers www.jst.org.inCommunity (EC) countries has started doing research and development on smart grid applications and technologies.For example, the U.S. Government has announced the largest power grid modernization investment in the U.S. history, i.e., $ 3.4 billion in grant awards, funding a broad range of smart grid technologies [2].Local Distribution Companies (LDCs) are integrating advanced metering and two-way communication, automation technologies to their distribution systems [15].In addition to research and development projects, many electric utilities are also taking incremental steps to make the smart grid technology a reality.Most of them are signing agreements with telecom operators or smart meter vendors to carry out smart grid projects.All these agreements define the main requirements and features of the necessary communications infrastructure to provide online communication between smart meters and the utility's back-haul system, i.e., the so-called advanced metering infrastructure (AMI).In general, the AMI is a two-way communications network and is the integration of advanced sensors; smart meters, monitoring systems, computer hardware, software and data management systems that enable the collection and distribution of information between meters and utilities [14].
In this paper, a comprehensive but brief review on smart grid communications technologies is presented.Section II describes smart grid communications technologies and their advantages and disadvantages.Section III mentions smart grid communications requirements in terms of security, system reliability, robustness, availability, scalability and the Quality of Service (QoS) mechanism.The standardization activities are reviewed in Section IV.Finally, the paper is concluded in Section V.
I. COMMUNICATIONS TECHNOLOGIES AVAILABLE FOR
SMART GRIDS A communications system is the key component of the smart grid infrastructure [1], [14], [16].With the integration of advanced technologies and applications for achieving a smarter electricity grid infrastructure, a huge amount of data from different applications will be generated for further analysis, control and real-time pricing methods.Hence, it is very critical for electric utilities to define the communications requirements and find the best communications infrastructure to handle the output data and deliver a reliable, secure and cost effective service throughout the total system.Electric utilities attempt to get customer's attention to participate in the smart grid system, in order to improve services and efficiency.Demand side management and customer participation for efficient electricity usage are well understood, furthermore, the outages after disasters in existing power structure also focus the attention on the importance of the relationship between electric grids and communications systems [1].
Different communications technologies supported by two main communications media, i.e., wired and wireless, can be used for data transmission between smart meters and electric utilities.In some instances, wireless communications have some advantages over wired technologies, such as low cost infrastructure and ease of connection to difficult or unreachable areas.However, the nature of the transmission path may cause the signal to attenuate.On the other hand, wired solutions do not have interference problems and their functions are not dependent on batteries, as wireless solutions do.
Basically, two types of information infrastructure are needed for information flow in a smart grid system.The first flow is from sensor and electrical appliances to smart meters, the second is between smart meters and the utility's data centers.As suggested in [17], the first data flow can be accomplished through power line communication or wireless communications, such as ZigBee, 6LowPAN, Z-wave and others.For the second information flow, cellular technologies or the Internet can be used.home owners, and home owners can reach the information of their real-time energy consumption.
1) Advantages: ZigBee has 16 channels in the 2.4 GHz band, each with 5 MHz of bandwidth.0 dBm (1 mW) is the maximum output power of the radios with a transmission range between 1 and 100 m with a 250 Kb/s data rate and OQPSK modulation [18].ZigBee is considered as a good option for metering and energy management and ideal for smart grid implementations along with its simplicity, mo-bility, robustness, low bandwidth requirements, low cost of deployment, its operation within an unlicensed spectrum, easy network implementation, being a standardized protocol based on the IEEE 802.15.4 standard [4].ZigBee SEP also has some advantages for gas, water and electricity utilities, such as load control and reduction, demand response, real-time pricing programs, real-time system monitoring and advanced metering support [18], [19].
2) Disadvantages: There are some constraints on ZigBee for practical implementations, such as low processing capa-bilities, small memory size, small delay requirements and being subject to interference with other appliances, which share the same transmission medium, license-free industrial, scientific and medical (ISM) frequency band ranging from IEEE 802.11 wireless local area networks (WLANs), WiFi, Bluetooth and Microwave [18].Hence, these concerns about the robustness of ZigBee under noise conditions increase the possibility of corrupting the entire communications channel due to the interference of 802.11/b/g in the vicinity of ZigBee [20].Interference detection schemes, interference avoidance schemes and energy-efficient routing protocols, should be implemented to extend the network life time and provide a reliable and energy-efficient network performance.
B. Wireless Mesh
A mesh network is a flexible network consisting of a group of nodes, where new nodes can join the group and each node can act as an independent router.The self-healing character-istic of the network enables the communication signals to find another route via the active nodes, if any node should drop out of the network.Especially, in North America, RF mesh based systems are very popular.In PG&E's SmartMeter system, every smart device is equipped with a radio module and each of them routes the metering data through nearby meters.Each meter acts as a signal repeater until the collected data reaches the electric network access point.Then, collected data is transferred to the utility via a communication network.A private company, SkyPilot Networks uses mesh networking for Published by: Longman Publishers www.jst.org.insmart grid applications due to the redundancy and high availability features of mesh technology [37].
1) Advantages: Mesh networking is a cost effective so-lution with dynamic self-organization, self-healing, selfconfiguration, high scalability services, which provide many advantages, such as improving the network performance, balancing the load on the network, extending the network coverage range [21].Good coverage can be provided in urban and suburban areas with the ability of multi-hop routing.Also, the nature of a mesh network allows meters to act as signal repeaters and adding more repeaters to the network can extend the coverage and capacity of the network.Advanced metering infrastructures and home energy management are some of the applications that wireless mesh technology can be used for.
2) Disadvantages: Network capacity, fading and interfer-ence can be counted as the major challenges of wireless mesh networking systems.In urban areas, mesh networks have been faced with a coverage challenge since the meter density cannot provide complete coverage of the communications network.Providing the balance between reliable and flexible routing, a sufficient number of smart nodes, taking into account node cost, are very critical for mesh networks.Furthermore, a third party company is required to manage the network, and since the metering information passes through every access point, some encryption techniques are applied to the data for security purposes.In addition, while data packets travel around many neighbors, there can be loop problems causing additional overheads in the communications channel that would result in a reduction of the available bandwidth [20].
C. Cellular Network Communication
Existing cellular networks can be a good option for com-municating between smart meters and the utility and between far nodes.The existing communications infrastructure avoids utilities from spending operational costs and additional time for building a dedicated communications infrastructure.Cel-lular network solutions also enable smart metering deploy-ments spreading to a wide area environment.2G, 2.5G, 3G, WiMAX and LTE are the cellular communication technologies available to utilities for smart metering deployments.When a data transfer interval between the meter and the utility of typically 15 minutes is used, a huge amount of data will be generated and a high data rate connection would be required to transfer the data to the utility.For example, T-Mobile's Global System for Mobile Communications (GSM) network is chosen for the deployment of Echelon's Networked Energy Services (NES) system.An embedded T-Mobile SIM within a cellular radio module will be integrated into Echelon's smart meters to enable the communication between the smart meters and the back-haul utility.Since T-Mobile's GSM network will handle all the communication requirements of the smart metering network, there is no need for an investment of a new dedicated communications network by utilities.Telenor, Telecom Italia, China Mobile, Vodafone have also agreed to put their GSM network into service for data flow of smart metering communications.Itron's SENITEL electricity meter is integrated with a GPRS module and communicates with a server running SmartSynch's Transaction Management System.CDMA, WCDMA and UMTS wireless technologies are also used in smart grid projects.A CDMA smart grid solution for the residential utility market has been introduced by Verizon, and Verizon's 3G CDMA network will be used as the backbone of the smart grid communications with the SmartSynch smart grid solutions [37].UMTS is IP-based and a packet oriented service that is suitable for metering applications [37].Telenor with Cinclus technology is offering UMTS technology for smart grid communications [37].
An Australian energy delivery company, SP AusNet, is building a dedicated communications network for smart grid applications and chose WiMAX technology for the commu-nications need of the smart meters.WiMAX chip sets are embedded into the smart meters and wireless communications is dedicated between smart meters and the central system in SP AusNet's system.A U.S. wireless carrier, Sprint Nextel, had signed a partnership with the smart grid software provider, Grid Net, on a project to provide communication between smart meters and smart routers over its 4G wireless network.General Electric (GE) is developing WiMAX based smart meters with CenterPoint Energy and had collaborated with Grid Net, Motorola and Intel to focus on WiMAX connectivity solutions.In GE's smart meter project with CenterPoint Energy, it will deploy WiMAX based MDS Mercury 3650 radios to connect the utility's back-haul system to collection points, which will collect data from smart meters that are installed by CenterPoint [37].Furthermore, some major companies, such as Cisco, Silver Springs Network and Verizon, also implement WiMAX smart grid applications.The world's largest WiMAX vendor, Alvarion, has announced its partnership with a U.S. utility company, National Grid, for a WiMAX based smart grid project.Lower deployment and operating costs, proper security protocols, smooth communications, high data speeds (up to 75 Mbps), an appropriate amount of bandwidth and scalability are the advantages of today's WiMAX technology.
1) Advantages: Cellular networks already exist.Therefore, utilities do not have to incur extra cost for building the communications infrastructure required for a smart grid.Wide-spread and cost-effective benefits make cellular communication one of the leading communications technologies in the market.Due to data gathering at smaller intervals, a huge amount of data will be generated and the cellular networks will provide sufficient bandwidth for such applications.When security comes into discussion, cellular networks are ready to secure the data transmissions with strong security controls.To manage healthy communications with smart meters in rural or urban areas, the wide area deployment capability of smart grid becomes a key component and since the cellular networks coverage has reached almost 100 %.In addition, GSM technology performs up to 14.4 Kbps, GPRS Published by: Longman Publishers www.jst.org.inperforms up to 170 Kbps and they both support AMI, Demand Response, Home Area Network (HAN) applications.Anonymity, authentication, signaling protection and user data protection security services are the security strengths of GSM technology [37].Lower cost, better cov-erage, lower maintenance costs and fast installation features highlight why cellular networks can be the best candidate as a smart grid communications technology for the applications, such as demand response management, advanced metering infrastructures, HAN, outage management, etc.
2) Disadvantages: Some power grid mission-critical ap-plications need continuous availability of communications.However, the services of cellular networks are shared by customer market and this may result in network congestion or decrease in network performance in emergency situations.Hence, these considerations can drive utilities to build their own private communications network.In abnormal situations, such as a wind storm, cellular network providers may not pro-vide guarantee service.Compared to public networks, private networks may handle these kinds of situations better due to the usage of a variety of technologies and spectrum bands.
D. Power Line Communication
Power line communication (PLC) is a technique that uses the existing power lines to transmit high speed (2 -3 Mbps) data signals from one device to the other.PLC has been the first choice for communication with the electricity meter due to the direct connection with the meter [20] and successful implementations of AMI in urban areas where other solutions struggle to meet the needs of utilities.PLC systems based on the LV distribution network have been one of the research top-ics for smart grid applications in China [22].In a typical PLC network, smart meters are connected to the data concentrator through power lines and data is transferred to the data center via cellular network technologies.For example, any electrical device, such as a power line smart transceiver-based meter, can be connected to the power line and used to transmit the metering data to a central location [37].France has launched the "Linky meter project" that includes updating 35 million traditional meters to Linky smart meters.PLC technology is chosen for data communication between the smart meters and the data concentrator, while GPRS technology is used for transferring the data from the data concentrator to the utility's data center [37].ENEL, the Italian electric utility, chose PLC technology to transfer smart meter data to the nearest data concentrator and GSM technology to send the data to data centers.
1) Advantages: PLC can be considered as a promising technology for smart grid applications due to the fact that the existing infrastructure decreases the installation cost of the communications infrastructure.The standardization efforts on PLC networks, the cost-effective, ubiquitous nature and widely available infrastructure of PLC, can be the reasons for its strength and popularity [23].Data transmissions are broadcast in nature for PLC, hence, the security aspects are critical.Confidentiality, authentication, integrity and user intervention are some of the critical issues in smart grid communications.HAN application is one of the biggest applications for PLC technology.Moreover, PLC technology can be well suited to urban areas for smart grid applications, such as smart metering, monitoring and control applications, since the PLC infrastructure is already covering the areas that are in the range of the service territory of utility companies.
2) Disadvantages: There are some technical challenges due to the nature of the power line networks.The power line transmission medium is a harsh and noisy environment that makes the channel difficult to be modeled.The low-bandwidth characteristic (20 kbps for neighborhood area net-works) restricts the PLC technology for applications that need higher bandwidth [37].Furthermore, the network topology, the number and type of the devices connected to the power lines, wiring distance between transmitter and receiver, all, adversely affect the quality of signal, that is transmitted over the power lines [37].The sensitivity of PLC to disturbances and dependency on the quality of signal are the disadvantages that make PLC technology not suited for data transmission.However, there have been some hybrid solutions in which PLC technology is combined with other technologies, i.e., GPRS or GSM, to provide fullconnectivity not possible by PLC technology.
E. Digital Subscriber Lines
Digital Subscriber Lines (DSL) is a high speed digital data transmission technology that uses the wires of the voice telephone network.It is common to see frequencies greater than 1 MHz through an ADSL enabled telephone line [16].The already existing infrastructure of DSL lines reduces installation cost.Hence, many companies chose DSL technology for their smart grid projects.The Current Group, a Smart Grid Solution Company, has collaborated with Qwest to implement a Smart Grid project.Qwest's existing low latency, secure, high capacity DSL network will be used for data transmissions.Xcel Energy's "SmartGridCity" project has also proved the interoperability of the technology by utilizing the Current's intelligent sensors and OpenGrid platform and Qwest's DSL network.A smart metering project has been car-ried out for Stadtwerke Emden-Municipal Utilities in Germany by Deutsche Telekom.In the project, Deutsche Telekom is responsible to provide the data communications for electric and gas meters.A communication box will be installed at the customer premises and the consumption information will be transmitted over DSL to Stadtwerke Emden [37].Deutsche Telekom offers many services in this project, such as reading consumption data, installation and operation, data transmis-sion, etc.However, the throughput of the DSL connection Published by: Longman Publishers www.jst.org.independs on how far away the subscriber is from the serving telephone exchange and this makes it difficult to characterize the performance of DSL technology [16].
1) Advantages: The widespread availability, low cost and high bandwidth data transmissions are the most important rea-sons for making the DSL technology the first communications candidate for electricity suppliers in implementing the smart grid concept with smart metering and data transmission smart grid applications.
2) Disadvantages : The reliability and potential down time of DSL technology may not be acceptable for mission critical applications.Distance dependence and lack of standardiza-tion may cause additional problems.The wired DSL-based communications systems require communications cables to be installed and regularly maintained, and thus, cannot be implemented in rural areas due to the high cost of installing fixed infrastructure for low-density areas.
To conclude, wired technologies, such as DSL, PLC, optical fiber, are costly for wide area deployments but they have the ability to increase the communications capacity, reliability and security.On the other hand, wireless technologies can reduce the installation costs, but provide constrained bandwidth and security options.
II. SMART GRID COMMUNICATIONS REQUIREMENTS The communication infrastructure between energy generation, transmission, and distribution and consumption requires two-way communications, inter-operability between advanced applications and end-to-end reliable and secure communica-tions with low-latencies and sufficient bandwidth [25]; More-over, the system security should be robust enough to prevent cyber-attacks and provide system stability and reliability with advanced controls.In the following, major smart grid commu-nication requirements are presented.
A. Security
Secure information storage and transportation are extremely vital for power utilities, especially for billing purposes and grid control [24].To avoid cyber-attacks, efficient security mechanisms should be developed and standardization efforts regarding the security of the power grid should be made.
B. System Reliability, Robustness and Availability
Providing the system reliability has become one of the most prioritized requirements for power utilities.Aging power infrastructure and increasing energy consumption and peak demand are some of the reasons that create unreliability issues for the power grid [26].Harnessing the modern and secure communication protocols, the communication and informa-tion technologies, faster and more robust control devices, embedded intelligent devices (IEDs) for the entire grid from substation and feeder to customer resources, will significantly strengthen the system reliability and robustness [26].The availability of the communication structure is based on pre-ferred communication technology.Wireless technologies with constrained bandwidth and security and reduced installation costs can be a good choice for large-scale smart grid de-ployments [24].On the other hand, wired technologies with increased capacity, reliability and security can be costly [24].To provide system reliability, robustness and availability at the same time with appropriate installation costs, a hybrid communication technology mixed with wired and wireless solutions can be used.
C. Scalability
A smart grid should be scalable enough to facilitate the operation of the power grid [3].Many smart meters, smart sensor nodes, smart data collectors, and renewable energy resources are joining the communications network.Hence, smart grid should handle the scalability with the integration of advanced web services, reliable protocols with advanced functionalities, such as self-configuration, security aspects.
D. Quality of Service (QoS)
The communication between the power supplier and power customers is a key issue of the smart grid.Performance degradation like delay or outage may compromise stability, therefore, a QoS mechanism must be provided to satisfy the communications requirements (for example high speed routing) and a QoS routing protocol must be applied in the communications network.This incurs two important questions unique to smart grid: • How to define the QoS requirement in the context of smart grid.
• How to ensure the QoS requirement from the home appliance in the communications network.
To answer the first question, the detailed mechanism of power price, based on the dynamics of the load, must be investigated.Then, a reward system is built for the home appliance based on the power price and the utility function of the appliance, thus obtaining the impact of delay and outage on the reward of the home appliance.Finally, the QoS requirement is derived by optimizing the reward [27].To answer the second question, routing methodologies meeting the derived QoS requirement are focused on.Due to the requirements of high computing and storage capabilities imposed by the heterogeneity of the smart Published by: Longman Publishers www.jst.org.ingrid, multiple QoS-aware routing within multiple (more than 2) constraints must be considered (for example a greedy algorithm with K-approximation, where K is the number of constraints) [27].A QoS requirement usually includes specifications, like average delay, jitter and connection outage probability.To derive the QoS requirement, it is important to describe the probabilistic dynamics of the power system, to evaluate the impact of different QoS specifications on the smart grid system and to derive the QoS requirement from the corresponding impact.
The power price is typically determined by locational mar-gin price (LMP) [28] driven by the load that varies with time.A constrained optimization problem can be used to derive the LMP from the load and other parameters, where the Lagrange factors of the constraints are considered as prices.
To efficiently link together the large number of smart grid components, a powerful data communications infrastructure will be provided.It is expected that part of this infrastructure will make use of the power distribution lines themselves as communications carriers using PLC technology [29].It is also expected to have a combination of wireless technologies to establish a reliable communications infrastructure.Also, recent standardization efforts under the umbrella of IEEE (P1901.2),ITU (ITU-G.hnem),and others are dedicated to PLC technology for Smart Grid applications.
One of the challenges of employing PLC in power distri-bution grids is multi-hop transmission message routing.The basic idea is that network nodes, i.e., PLC enabled devices, act as repeaters of messages in order to achieve sufficient coverage [30], [31].The focus in these two previous studies is on reliable delivery of messages taking into account unpredictable and possible sudden changes of communications links and network topology.In this regard, for flooding of messages, the concept of single-frequency network (SFN) transmission is presented in [30], [31].In [32], the problem of routing in PLC networks is revised taking into account that network nodes are static and thus, their location is known a priori.In other words, the nodes know in which direction a message is intended to flow.More specifically, if a node receives a packet it can decide whether to forward it or not.Such routing algorithms are known as geographic routing in the wireless communications literature [33], where they have been applied mainly in the context of wireless sensor networks.These algorithms present high performance for the application at hand: they close the gap between flooding on the one hand and improved shortest path routing on the other.
In [34], the implementation of a smart monitoring system over a wireless sensor network is presented, with particular emphasis on the creation of a solid routing infrastructure through the routing protocol for low-power and lossy networks (RPL), whose definition is currently being discussed within the IETF ROLL working group.RPL was designed in order to match the requirements of networks characterized by low power supplies and by deployment in lossy environments.This involves both wired and wireless networks deployed in difficult environments, where the presence of high interference requires adaptive and reconfigurable network operations.In [35] a hybrid routing protocol that combines local agility with centralized control is presented.It meets the requirements of robust collection, point-to-point communication, and low footprint.It uses a distributed algorithm to form a Directed Acyclic Graph (DAG) for routing data from in-network nodes to border routers, allowing nodes to maintain multiple options that are ranked through data-driven link estimation.
III. SMART GRID STANDARDS
There are many applications, techniques and technological solutions for smart grid system that have been developed or are still in the development phase.However, the key challenge is that the overall smart grid system is lacking widely accepted standards and this situation prevents the inte-gration of advanced applications, smart meters, smart devices and renewable energy sources and limits the inter-operability between them.The adoption of inter-operability standards for the overall system is a critical prerequisite for making the smart grid system a reality.Seamless interoperability, robust information security, increased safety of new products and systems, compact set of protocols and communication exchange are some of the objectives that can be achieved with smart grid standardization efforts [37].There are many regional and national attempts towards achieving this goal; for example, the European Union Technology Platform orga-nization's strategic energy technology plan is all about the development of a smart electricity system over the next 30 years; Ontario Energy Board, Canada, has committed itself to-wards the completion of a smart meter installation [37].On the other hand, NIST, the American National Standards Institute (ANSI), the International Electro technical Commission (IEC), the Institute of Electrical and Electronics Engineers (IEEE), the International Organization for Standardization (ISO), the International Telecommunication Union (ITU), the 3rd Gener-ation Partnership Project (3GPP) and on the regional level, the Korean Agency for Technology and Standards (KATS) and Joint Inannounced formation Systems Committee(JISC) are the recognized standard development organizations that are worth to mention.In addition, the CEN, CENELEC and ETSI has formed a Joint working group for smart grid standardiza-tion efforts and aim to achieve the European Commission's policy objectives regarding the smart grid [37].Their efforts focus on smart metering functionalities and communication interfaces for electric, water and heat sectors in Europe.An overview of smart grid standards are given in Table II.In the following, the details of these standards are explained.Published by: Longman Publishers www.jst.org.in
A. Revenue Metering Information Model
• ANSI C12.19: ANSI C12.19 is an ANSI standard for utility industry end device data tables.This standard is defining a table structure for data transmissions between an end device and a computer for utility applications using binary codes and XML content.ANSI C12.19 is not interested in defining device design criteria or specifying the language or protocol used to transport that data.• M-Bus: M-Bus is a European standard and provides the requirements for remotely reading all kinds of utility meters.The utility meters are connected to a common master that periodically reads the meters via M-Bus.The wireless version, Wireless M-Bus, is also specified recently.• ANSI C12.18: ANSI C12.18 is an American National Standard (ANSI) standard that is specifically designed for meter communications and responsible for two way com-munications between smart electricity meters (C12.18device) and a C12.18 client via an optical port.
B. Building Automation
• BACnet: BACnet is a standard communication proto-col that was developed by the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) for building automation and control networks and support the implementation of intelligent buildings with full integration of computer-based building automa-tion and control systems from multiple manufacturers.
C. Substation Automation
• IEC 61850: IEC 61850 is a flexible, open standard that defines the communication between devices in transmis-sion, distribution and substation automation systems.To enable seamless data communications and information exchange between the overall distribution networks, it is aimed to increase the scope of IEC 6180 to whole electric network and provide its compatibility with Common Information Model (CIM) for monitoring, control and protection applications [25].This technology is imple-mented by modern manufacturers in their latest power en-gineering products like distribution automation nodes/grid measurement and diagnostics devices [13].
D. Powerline Networking
• HomePlug: HomePlug is a power line technology and the existing home electricity is used to connect the smart appliances to HAN; HomePlug Command and Control (HPCC) version is designed for low-cost applications.HomePlug is a promising technology to create a reliable HAN between electric appliances and a smart meter.• HomePlug Green PHY: HomePlug Green PHY specifi-cation is developed as a low power, cost-optimized power line networking specification standard for smart grid applications used in home area networking by the Smart Energy Technical Working Group within the HomePlug Powerline Alliance.The inputs for optimization of speci-fications for field tests were gathered from many utilities, i.e., Consumers Energy, Duke Energy, Pacific Gas and Electric, and Southern California Edison.Backwards in-teroperability, lower data rate and IP networking support, low power consumption, full interoperability with both HomePlug devices are the leading features of HomePlug Green PHY specification.• PRIME: PRIME is an open, global power line stan-dard that provides multi-vendor interoperability and wel-comes several entities to its body.Advanced Digital Design, CURRENT Group, Landis+Gyr, STMicroelec-tronics, uSyscom and ZIV Medida are some of the current companies that have extensive experience in PLC technology and smart metering.• G3-PLC: G3-PLC is a power line communications speci-fication launched by ERDF and Maxim that aims to pro-vide interoperability, cyber security and robustness and reduce infrastructure costs in smart grid implementations worldwide.
E. Home Area Network Device Communication Measurement and Control
• U-SNAP: There have been a variety of incompatible standards for HAN.This lack of standardization in HAN Utility has driven major AMI suppliers and product manufacturers to develop a solution, namely Utility Smart Network Access Port (U-SNAP).The main requirement is the existence of an interface to connect any type of product to a HAN.U-SNAP basically enables the standardization of a connector and serial interface and identifies the hardware interface, physical dimensions, data transfer, message contents and protocol specifics for HAN devices to provide many communication protocols to connect HAN devices to smart meters.• IEEE P1901: The IEEE P1901 Working Group (WG) un-der the sponsorship of the IEEE Communications Society developed the IEEE P1901 standard for high speed power line communications to meet in-home multimedia, utility and smart grid application requirements [36].Access control and physical layer specifications for broadband over power line networks are analyzed in detail and the access system with cell structure is defined by the IEEE P1901Working Group [36].The IEEE P1901 standard has an important effect on communications technology by integrating power line communications into wireless networks with extensive features, such as high-speed, walls-penetration, etc.
• Z-Wave: Z-Wave is an alternative solution to ZigBee that handles the interference with 802.11/b/g since it operates in the 800MHz range [20].Z-Wave is not an open standard and developed by The Z-Wave Alliance, an international consortium of manufacturers.The simple, modular and low-cost features make Z-Wave one of the leading wireless technologies in home automation.Z-Wave can be easily embedded to consumer electronic appliances, such as lighting, remote control, security systems that require low-bandwidth data operations.
F. Application-Level Energy Management Systems
• IEC 61970 and IEC 61968: Two defined standards, i.e., IEC 61970 and IEC 61969, provide what is called a Common Information Model (CIM), which is necessary for exchanging data between devices and networks.IEC 61970 works in the transmission domain, while IEC 61968 works in the distribution domain.CIM standards are integral to the deployment of a smart grid scenario, in which many devices connect to a single network.• OpenADR : OpenADR is a research and standards devel-opment effort that is defined as a fully-automated demand response using open standard, platform-independent and transparent end-to-end technologies or software systems.OpenADR was originally developed at Lawrence Berke-ley Labs, a U.S. government research laboratory.The adoption of OpenADR to the smart grid is very impor-tant to provide effective deployment of dynamic pricing, demand response and grid reliability.
G. Inter-control and Inter-operability Center Communications
• IEEE P2030: IEEE P2030 is a guide for smart grid inter-operability of energy technology and information technol-ogy operation with the electric power system (EPS), and customer-side applications.Basically, IEEE P2030 can be understood as a single document for the inter-operability of smart grid.IEEE P2030 is responsible for seamless data transfer in a two-way communications manner for electric generation, reliable power delivery and customer-side applications.The IEEE Standards Board, as spon-sored by the IEEE Standards Coordinating Committee, confirmed the development of the IEEE P2030 smart grid inter-operability standard.Understanding and defin-ing smart grid inter-operability of the power system with customer applications was the initial goal of IEEE P2030.Integration of energy technology and ICT, seam-less communication between electric generation, delivery, and end-use benefits and increasing knowledge about grid architecture to provide reliable, flexible and secure electric delivery are some of the benefits of IEEE P2030.• ANSI C12.22: ANSI C12.22 is defining a protocol for transporting ANSI C12.19 table data over networks to achieve the inter-operability among communications modules and smart meters by using AES encryption for enabling strong, secure communications, including confidentiality and data integrity .• ISA100.11a:ISA100.11a is an open standard for wireless systems for industrial automation developed by the Inter-national Society of Automation (ISA).The ISA100.11a focuses on robustness, security and network management requirements of wireless infrastructure and low power consuming devices that provides large-scale installations.The ISA100.11a standard is simple to use and deploy and provides multi-vendor device interoperability.
• ITU-T G.9955 and G.9956: The two G.hnem stan-dards ITU-T G.9955 and G.9956 contain the physical layer specification and the data link layer specification, respectively, for narrowband OFDM power line com-Published by: Longman Publishers www.jst.org.inmunications transceivers for communications via alter-nating current and direct current electric power lines over frequencies below 500 kHz.These ITU-T standards support indoor and outdoor communications over low voltage lines, medium voltage lines, through transformer low-voltage to medium-voltage, and through transformer medium-voltage to low-voltage power lines in both urban and long distance rural communications.
H. Cyber Security
• IEC 62351: IEC 62351 defines cyber security for the communication protocols defined by the previous four sets.Security is a major concern with smart grids, which are especially vulnerable to attack because of the two-way communication between devices and the utility grid.
I. Electric Vehicles
• SAE J2293: This standard was developed by the Hybrid Committee that is a part of SAE international and pro-vides requirements for Electric Vehicles (EV) and Electric Vehicle Supply Equipment (EVSE).It standardizes the electrical energy transfer from electric utility to EVs. • SAE J2836: SAE J2836 standard supports use cases for communication between plug-in electric vehicles and the power grid for energy transfer and other applications.• SAE J2847: SAE J2847 supports communication mes-sages between PEVs and grid components.
IV. CONCLUSION
The smart grid has been conceived as an evolution of electric power systems due to the increasing diffusion of distributed generation by renewable sources, but with the additional aim to enhance efficiency, reliability and safety of the existing power grid.To this end, remote and timely information gathering about equipment failures, capacity limi-tations, and natural accidents is extremely critical for ensuring proactive and real-time and reliable diagnosis of possible failures in the smart grid.This makes cost-effective remote sensing technologies vital for safe, seamless and efficient power delivery in the smart grid.
In this paper, communications technologies and require-ments for smart grids have been discussed.The QoS mecha-nism is introduced and standards are presented.Clearly, there are many important open research issues for the realization of smart grid communications and applications.Future work includes discussion of grid characteristics, architectures, key players, pilot projects, applications and research challenges on ICT issues, in order to give a complete overview on the subject.
Figure 1 .
Figure 1.Smart grid architecture increases the capacity and flexibility of the network and provides advanced sensing and control through modern communications technologies. | 2024-02-18T16:03:17.327Z | 2022-12-21T00:00:00.000 | {
"year": 2022,
"sha1": "85deaafcb45ad558010afaeb1a046553cfbcfc68",
"oa_license": null,
"oa_url": "https://doi.org/10.46243/jst.2022.v7.i10.pp136-148",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "be633d4f648e42e2b08ad620dee4ab8e2e6aa6d2",
"s2fieldsofstudy": [
"Engineering",
"Computer Science"
],
"extfieldsofstudy": []
} |
14032074 | pes2o/s2orc | v3-fos-license | Intermittent Screening and Treatment versus Intermittent Preventive Treatment of Malaria in Pregnancy: A Randomised Controlled Non-Inferiority Trial
Background The effectiveness of intermittent preventive treatment of malaria in pregnancy (IPTp) may be compromised by the spread of resistance to sulphadoxine/pyrimethamine (SP) across Africa. But little informtion exists on alternative drugs for IPTp or alternative strategies for the prevention of malaria in pregnancy. Therefore, we have investigated whether screening with a rapid diagnostic test and treatment of those who are positive (IST) at routine antenatal clinic attendances is as effective and as safe as SP-IPTp in pregnant women. Methods and Findings During antenatal clinic sessions in six health facilities in Ghana held between March 2007 and September 2007, 3333 pregnant women who satisfied inclusion criteria were randomised into three intervention arms (1) standard SP-IPTp, (2) IST and treatment with SP or (3) IST and treatment with amodiaquine+artesunate (AQ+AS). All women received a long-lasting insecticide treated net. Study women had a maximum of three scheduled follow-up visits following enrolment. Haemoglobin concentration and peripheral parasitaemia were assessed between 36 and 40 weeks of gestation. Birth weight was measured at delivery or within 72 hours for babies delivered at home. Parasite prevalence at enrolment in primigravidae and in multigravidae was 29.6% and 10.2% respectively. At 36–40 weeks of gestation the prevalence of asymptomatic parasitaemia was 12.1% in study women overall and was very similar in all treatment groups. The risk of third trimester severe anaemia or low birth weight did not differ significantly between the three treatment groups regardless of gravidity. IST with AQ+AS or SP was not inferior to SP-IPTp in reducing the risk of low birth weight (RD = -1.17[95%CI; -4.39-1.02] for IST-SP vs. SP-IPTp and RD = 0.78[95%CI; -2.11-3.68] for IST-AQAS vs. SP-IPTp); third trimester severe anaemia (RD = 0.29[95%CI; -0.69-1.30] for IST-SP vs. SP-IPTp and RD = -0.36[95%CI;-1.12-0.44] for IST-AQAS vs. SP-IPTp). Conclusion The results of this study suggest that in an area of moderately high malaria transmission, IST with SP or AS+AQ may be a safe and effective strategy for the control of malaria in pregnancy. However, it is important that these encouraging findings are confirmed in other geographical areas and that the impact of IST on placental malaria is investigated. Trial Registration ClinicalTrials.gov NCT00432367 [NCT00432367]
Introduction
Plasmodium falciparum infection in pregnancy causes maternal anaemia and low birth weight associated with parasitisation of the placenta [1]. The World Health Organization recommends intermittent preventive treatment (IPTp) with sulphadoxinepyrimethamine (SP), insecticide-treated bed nets (ITNs), and effective case management as measures to protect against these outcomes [2].
Intermittent preventive treatment linked to antenatal care will remain an effective and sustainable strategy for the prevention of malaria in pregnancy provided that the antimalarial drug used is efficacious, safe, tolerable, cheap, and easy to administer, preferably as a single dose [3]. So far SP is the only drug which has these attributes. Thus, as resistance to SP increases and spreads across Africa, [4,5,6] the effectiveness of the IPT strategy may be compromised. Trials are under way to evaluate alternative drugs that could be used for IPTp in place of SP but none of the potential candidates have the favourable characteristics of SP when used in areas where parasites are sensitive to this drug.
In the absence of an effective drug to replace SP for IPTp alternative strategies for the prevention of malaria in pregnancy need to be evaluated. Insecticide treated bed nets (ITNs) used during pregnancy are beneficial to both mother and her newborn baby [7]and so strenuous efforts are being made to increase ITNs accessibility to pregnant women across Africa [8]. It is not clear whether women who are protected by an ITN also need IPTp. In a study conducted in western Kenya, the combination of IPTp and ITNs was slightly more effective than ITNs alone (PE 56% vs. 42%) but only in primigravidae. [9]. In the Gambia, where the prevalence of HIV infection and SP resistance are low, no beneficial effect from SP-IPTp on anaemia or birth weight was seen in multigravidae, with the exception of a small sub-group of women who did not use a bed net [10].
Screening for malaria infection using a malaria rapid diagnostic test (RDT) at scheduled antenatal clinic visits and treatment of women who are positive with an effective antimalarial drug (IST), combined with effective vector control, provides a potential, alternative strategy to SP-IPTp. The WHO's current recommendation of four scheduled antenatal clinic visits, one at booking and three subsequent visits 4 to 8 weeks apart, provides a potential framework for implementation of an IST strategy. This strategy may be considered in parts of Africa where resistance to SP is increasing at alarming rates and in situations of low exposure to malaria as in The Gambia and Zanzibar where the incidence of malaria has gone down [11,12,13] or in Asia where transmission is generally low [14].
Intermittent screening and treatment combined with vector control may be an especially attractive option in such communities where continuing use of IPTp with SP will result in a large proportion of pregnant women receiving SP unnecessarily. This is also the case for communities where the incidence of malaria is highly seasonal. In these areas, many women receive IPTp during months of the year when the risk of malaria is minimal.
Whether a screening and treatment strategy would prove to be as effective as SP-IPTp is not known. Therefore, we have undertaken a randomised, controlled trial in an area of Ghana with moderate malaria transmission to determine whether IST using SP or amodiaquine+artesunate (AQ+AS) is as effective in preventing maternal anaemia and low birth weight as SP-IPTp. The costs and acceptability to women and providers of such a strategy have also been assessed.
Overall study design
The protocol for this trial and supporting CONSORT checklist are available as supporting information; see Checklist S1 and Protocol S1. It was an individually randomised, open, controlled trial was undertaken to investigate whether screening for malaria with an RDT and treatment of women with a positive test with either SP or AS+AQ was not inferior to SP-IPTp in the prevention of anaemia in pregnancy or low birth weight.
Study population
The study was conducted in the Ejisu-Juaben and Afigya-Sekyere East districts of the Ashanti Region of Ghana from March 2007 to September 2008. Malaria transmission in this area is perennial but with a peak in the rainy season. The predominant parasite is P. falciparum. The entomological inoculation rate in the neighbouring area of Kintampo is about 250 infectious bites per year. HIV prevalence in Ashanti region is reported to be 3.0% and 2.2% in the general population and in pregnant women respectively (Ghana Health Service 2007 Annual Report).
Enrolment was undertaken at the antenatal clinic of three district hospitals and three health centres in the study area. The study population comprised pregnant women of all parities who presented at the antenatal clinics with a gestational age of 16 to 24 weeks at their first booking. Women who were temporary residents, had had a prior dose of SP-IPTp, had a haemoglobin concentration ,5 g/dl, gave a history of sensitivity to SP, amodiaquine or an artemisinin, had an illness requiring hospital admission or declined to join the trial were excluded (Figure 1).
Ethics
The study protocol was approved by the ethics committee of the London School of Hygiene and Tropical Medicine and the Committee on Human Research and Publications Ethics of the School of Medical Sciences, Kumasi -Ghana. A Data, Safety and Monitoring Board (DSMB) approved the protocol, standard operating procedures and analysis plan.
Sample size
Sample size was calculated on the assumption that the prevalence of moderately severe anaemia (Hb,8 g/dl) in the third trimester of pregnancy and of low birth weight in women in the SP-IPTp arm of the study would be at least 12% and 6% respectively, figures based on findings from a study undertaken in The Gambia [15]. To establish that IST was not inferior to SP-IPTp we decided that it was necessary to show that the differences in the percentages of women with moderately severe third trimester anaemia or low birth weight between the IST groups and the SP-IPTp group should not be more than 5% or 4% respectively, differences that would be clinically important. To meet these criteria with 90% statistical power and allowing for 20% loss to follow-up it was calculated that 1,110 women were needed in each study arm to give a total sample size of 3,330.
Procedures
After written informed consent had been obtained, eligible women of all parities were randomised to one of the three treatment groups described above. Women who declined to participate were given SP-IPTp according to the national guidelines. The following process was followed to randomise women into the study arms. A list of random numbers was computer generated as identification numbers, randomly allocated to treatment groups and grouped in blocks of 15 by an IT specialist who did not participate further in the study. The list of identification numbers with their corresponding treatment groups was printed and cut into slips. Fifteen slips each with an identification number and an allocated treatment group were sealed in an envelope. During enrolment an eligible pregnant woman was asked by the recruitment team to pick a slip from the sealed envelope. The allocated treatment group printed on each slip indicated which treatment arm the women belonged to and the corresponding identification number was used to identify her. This constituted entering the study, and treatment group allocation was binding on the recruiting team and the woman from this point. Another envelope was opened only when the contents of the previous one had been exhausted. The PI and all other project staff were blinded to the randomisation process and treatment allocation.
At enrolment, a finger prick blood sample was obtained for measurement of haemoglobin concentration, preparation of thin and thick blood films for malaria parasite counts and preparation of a filter paper blood spot. During the last 3 months of the trial, women in treatment group 1 who had a positive blood film and received SP-IPTp were seen again on day 14 and day 28 after treatment and a further blood film and blood spot obtained to check on the response to treatment with SP. HIV screening was offered to all pregnant women study women as part of the routine antenatal services recommended in Ghana with an option for treatment but the results of HIV screening were not available to the investigators. Women in the SP-IPTp arm received an initial dose of SP (1500 mg sulphadoxine/75 mg pyrimethamine) as a single dose. Pregnant women in treatment groups 2 and 3 were screened for malaria infection with an OptiMALH dipstick, a lactate dehydrogenase (LDH) based RDT. The OptiMALH dipsticks were purchased from DiaMed AG, Cressier, Switzerland who supplied and organised transportation to the study site in batches as and when needed,. The test kits were kept at room temperature always at the study site. The viability of the kits were tested monthly throughout the study period using positive test controls obtained from the manufacturer. The tests were performed and interpreted by the study team following the manufacturer's instructions. Women in treatment group 2 were treated with a single dose of SP (1500 mg sulphadoxine/75 mg pyrimethamine) if the RDT was positive. Women in treatment group 3 were treated with AQ+AS (AQ-300 mg + AS-100 mg twice daily for 3 days) if the RDT was positive. All treatments with SP and the first dose of AQ+AS treatment were administered by a member of the study team but doses 2 and 3 of AQ+AS were given to study women to take at home unsupervised. Women in groups 2 and 3 received no antimalarial treatment or IPT if their RDT results were negative. All women received a daily supplement of ferrous sulphate (200 mg) and folic acid (4 mg) tablets as part of the routine antenatal care services provided in Ghana throughout the study period. SP was purchased from Kinapharma Ltd, Ghana and AQ+AS was provided by the Ministry of Health, Ghana. All study women received a long-lasting insecticide treated bed net (LLIN) and instructions on how to use this at enrolment.
Study women were asked to attend for follow-up antenatal care and IPTp or screening at 24, 32 and 36 weeks of gestation. At the 24 and 32 week visits, women in treatment group 1 received SP-IPTp whilst women in treatment groups 2 and 3 were screened with the RDT and, if positive, treated with SP or AS+AQ according to the allocation arm. As part of safety assessment, all women were visited at home by a trained community health worker to record any complaints that the women might have seven days after each scheduled antenatal visit. Blood samples were obtained from all study women between 36 and 40 weeks of gestation (before delivery) for determination of haemoglobin concentration and for preparation of thin and thick blood films. However, the smears were read retrospectively and so the results were not available for the point of care. Any study woman who presented with a history of fever or other features suggestive of malaria between scheduled antenatal visits was screened for malaria using an RDT and treated with quinine (30 mg/kg in divided doses daily for 5 days) regardless of treatment group.
Information on the outcome of pregnancy was obtained for 2706 of the 3333 study women (81.2%), 2144 of whom delivered at a health centre or hospital and 562 at home. If a woman delivered at a health facility, birth weight was recorded by a midwife who was unaware of the treatment group of the woman who she was attending. Women who delivered at home were traced through a network of trained community health workers within 72 hours of delivery and the infant's weight was measured at home. The occurrence of miscarriages, still births, neonatal deaths and the presence of congenital abnormalities was recorded by midwives. If any congenital abnormality was suspected, a full examination, including a neurological assessment, was undertaken by a qualified medical doctor. Field teams visited all women and babies at approximately 6 weeks post delivery to obtain reports of any neonatal adverse events and a blood film was obtained from mothers at this time. We could not determine the prevalence of placenta malaria and efficacy of SP in postpartum study women as proposed due respectively to inadequate funding and difficulty in finding postpartum women with parasitaemia.
An experienced microscopist, unaware of treatment group assignment, read all the blood films and quantified parasitaemia against 200 leucocytes in thick blood smears. A thick blood film was declared negative only after examination of 100 high power fields (HPF). An independent expert microscopist from the Noguchi Memorial Institute of Medical Research read ten percent of all blood slides obtained at enrolment and on follow up visits for quality assurance. Agreement between the study microscopist and the reference microscopist was 91.7% and a kappa of 0.83. Haemoglobin (Hb) concentration was measured using Hb 301 Hemocues (HemoCue AB, Angelholm, Sweden). Paired samples from women who failed treatment with SP were tested for molecular markers to differentiate reinfections from recrudescences using genetic markers as described previously [16]. HIV testing was not part of the study but it is likely that some HIV positive may have been included in the sample.
Statistical methods
Stata version 10 (StataCorp, College Station, Texas) was used for data analyses. The primary objective of the study was to demonstrate that the risk of third trimester moderately severe anaemia (Hb,8 g/dl) in the IST groups was no more than 5% greater than in the SP-IPTp arm. Secondary objectives were to demonstrate that the risks of low birth weight (BW,2500 g), spontaneous abortions, intrauterine deaths/stillbirths, neonatal and maternal mortality were not more than 4% higher in women in the IST groups than in women who received SP-IPTp. The principal analysis of primary and secondary outcomes was per protocol but an "intention-to-treat" analysis was also undertaken according to a statistical analysis plan approved by the Data and Safety Monitoring Board. In the per protocol analysis, only data from women who remained within their randomization group, received two courses of SP-IPTp (Group 1) or were screened twice using an RDT at scheduled visits (groups 2 and 3) and in whom the primary outcome had been recorded were considered for analysis. In the intent-to-treat analysis, women were included if they had received an initial treatment of IPTp or had had an initial screening test done and provided that an outcome had been recorded.
The proportion of the per protocol and intention-to-treat populations experiencing each primary and secondary outcome for the treatment groups, and the associated 2-sided 95% CI for the difference, was estimated using the generalized linear model. To declare non-inferiority with a significance level of 0.05%, the upper boundary of the 2-sided 95% CI for the estimated treatment effect (risk difference) had to be below the pre-defined noninferiority margins (D) of 5% and 4% for third trimester severe anaemia and low birth weight respectively. We controlled for gestational age at enrolment, gravidity, baseline parasitaemia and anaemia using binomial regression.
Baseline demographic and clinical characteristics
A total of 3333 of the 3472 potentially eligible pregnant women who were screened (96%) were enrolled into one of the three treatment groups (Figure 1). At the end of follow up, 2674 (80.2%) evaluable records for third trimester haemoglobin concentration and 2675 (80.2%) records for birth weight were available.
Baseline characteristics of women in the three treatment groups were very similar at enrolment (Table 1). In all three treatment groups, more than half of the women were aged between 20 and 30 years with an overall mean age of 26.6 years. About ninety percent of the women had received a formal education, mainly up to secondary school level (70%). Fifty percent of the study population households already owned a bed net. Twenty-three percent of women were primigravidae and twenty-three percent were secundigravidae. About 4% of women had a haemoglobin concentration below 8 g/dl; the mean Hb overall was 10.97 g/dl. Anaemia at enrolment was associated significantly with asymptomatic malaria parasitaemia (p,0.05). The prevalence of malaria infection on presentation at the antenatal clinic in women in the SP-IST and AQAS-IST groups as determined by the OptiMALH RDT was 22.9% (RDTs were not done in women in the SP-IPTp group). Asymptomatic parasitaemia as determined by microscopy was present in 16.3% of study women overall with parasite densities less than 1000/mL in 99% of them. P. falciparum was the most prevalent malaria species. P. malariae and P.ovale were also detected but only few women (,3%) carried these species. Additional blood samples were obtained 14 and 28 days after administration of a first dose of SPI-IPTp from 71 women who were parasitaemic on presentation. Eleven (15.5%; 95% CI 7.9%-26.0%) were positive by day 28. PCR corrected parasitological failure on day 28 was 5.6% (95% CI 1.6%-13.8%). The overall prevalence of HIV in pregnant women who agreed to screening in the study clinics during the period of the trial was approximately 1.5%.
Study outcomes -per protocol analysis
At 36-40 weeks of gestation (before delivery) the prevalence of severe anaemia (Hb,8 g/dl) and moderate (8, = Hb,11 g/dl) anaemia was respectively 1.7% and 45.9% overall and was similar in all treatment groups; similarly asymptomatic parasitaemia was 12.0% in study women overall and was very similar in all treatment groups. The mean Hb concentration of all women between 36 and 40 weeks of gestation was 11.0 g/dl and was similar in all the groups. Generally, there was a significant increase in mean Hb concentration of 0.29 g/dl at 36-40 weeks gestation over the baseline Hb concentration (Figure 2). The increase was higher in women with malaria parasitaemia at baseline (0.87 g/dl) compared to non-parasitic women (0.21 g/dl) but there was no significant difference in the increase in haemoglobin concentration during the course of pregnancy between the three treatment groups. The overall prevalence of low birth weight was 11.0% in the study women and did not differ significantly among the study groups ( Table 2).
The risk of anaemia (severe or moderate) did not differ significantly between the three treatment groups across gravidity (Table 3). It was no higher in the IST groups than in women who . The upper boundaries of the 2-sided 95% CI for the risk differences estimated between SP-IPTp and the IST groups were below the non-inferiority margin of 5% for third trimester severe anaemia, the main trial endpoint ( Figure 3A). Third trimester severe anaemia, the primary trial end-point, was associated significantly in a univariate analysis with baseline parasitaemia baseline anaemia and young maternal age (p,0.05) ( Table 3). In a multivariate analysis, which adjusted for the number of visits made by a woman, weeks of gestation at enrolment, gravidity and treatment group, third trimester severe anaemia remained significantly associated with baseline anaemia (p,0.0001) and young maternal age (p = 0.02) but only marginally with baseline parasitaemia (p = 0.07) (Table 3). Similarly, the risk of low birth weight did not differ significantly between the three treatment groups regardless of gravidity; (RD = -1.17 [95% CI; -4.39-1.02] for IST-SP vs. SP-IPTp and RD = 0.78 [95% CI; -2.11-3.68] for IST-AQAS vs. SP-IPTp). The upper boundaries of the 2-sided 95% CI for the risk differences estimated between women in the SP-IPTp group and the IST lie below the non-inferiority margin of 4% set for low birth weight ( Figure 3B). In a univariate analysis, low birth weight was associated significantly with baseline parasitaemia, low gravidity and young maternal age (p,0.05). However, in a multivariate analysis, which adjusted for the number of visits made by a woman, weeks of gestation at enrolment, gravidity and treatment group, only gravidity remained significantly associated with low birth weight. Primigravidae (RR = 2.10; 95%CI 1.49-2.95; p,0.0001) and secundigravidae (RR = 1.76; 95% CI 1.28-2.43; p = 0.001) were more likely to deliver low birth weight babies compared to multigravidae regardless of intervention arm. But the risk of low birth weight did not differ significantly between the three treatment groups across gravidity (Table 4).
One hundred and forty-seven episodes of illness associated with malaria parasitaemia were recorded among study women between scheduled antenatal clinic visits; 44, 51 and 52 of these were in SP-IPTp, IST with SP, and IST with AS+ AQ groups respectively. The differences in the illness episodes are not statistically significant.
In all 2211 RDT tests were done in women in treatment groups 2 and 3 respectively of which 23% and 23% were positive and led to treatment. The risk of third trimester severe anaemia and low birth weight in those who were RDT negative throughout pregnancy and did not receive antimalarial treatment were no higher than in the SP-IPT arm. But the risk of parasitaemia at 36-40 weeks was higher in the SP-IPT arm than in the screening and treatment arms (Supplementary Table S1).
There were no statistically significant differences in the risk of preterm deliveries, abortions or perinatal births between the treatment groups (Supplementary Table S2). There were no maternal deaths. The proportion of women who reported adverse events during the 7 days following administration of SP-IPTp or following treatment with SP or AS+AQ did not differ significantly between the treatment groups with the exception of a complaint of general weakness (Supplementary Table S3) which was reported slightly more frequent in the group who received treatment with AS+AQ compared to the SP-IPT group (33.8% vs. 29.1%; p = 0.04).
Study outcomes -intention to treat analysis
All primary and secondary outcomes were also analysed by intention to treat analysis and the results were very similar to those presented above (supplementary tables provided). Supplementary Table S4 show the comparison of key outcomes in women
Discussion
This study has shown that IST, using either SP or AS + AQ, was not inferior to IPTp with SP in preventing maternal anaemia and low birth weight, according to the non-inferiority criteria that were set prior to the trial, in women who used an LLIN in an area of moderately high malaria transmission in Ghana. There are a number of possible reasons why IST performed as well as SP-IPTp in our study. Firstly, it is possible that malaria is not an important cause of anaemia or low birth weight in the study area, in which case it would be difficult to detect an impact of different control strategies on the prevalence of these complications of malaria in pregnancy unless a much larger trial than the one we did was undertaken. We think that this is an unlikely explanation for our findings for several reasons. The study area is one of derived forest with perennial malaria transmission, although with a pronounced seasonal peak. The entomological inoculation rate in the neighbouring area of Kintampo was recently estimated to be 267 infective bites per year [17] and the prevalence of parasitaemia on presentation at the antenatal clinic among study women was 29.6% and 10.2% in primigravidae and in multigravidae respectively. These findings suggest that pregnant women in the study area are still being moderately exposed to malaria and that the level of transmission in the study area is comparable to that seen in communities where SP-IPT has been shown previously to have a significant impact [18,19]. Furthermore we found an association between anaemia and malaria parasitaemia on presentation. A second possible explanation for our findings is that SP is no longer effective in the study area which would again make any comparisons between IST with SP and SP-IPTp unhelpful. In these circumstances, IST with AS+AQ would have been expected to show a better response than either IST or IPTp with SP and this was not the case. Furthermore, the adequate parasite clearance rate at day 28 in women who received a single dose of SP-IPTp was 85% (uncorrected) indicating only a moderate level of SP resistance. In an additional study carried out in the study area six months after completion of the study, the adequate parasite clearance rate by day 42 in 109 women was also 85% (unpublished data). Thus, it seems likely that IST with SP or AS+AQ can be as effective as SP-IPTp with the same drug in an area with moderate malaria transmission, marked seasonality and moderate SP resistance.
Demonstration of efficacy is only one aspect of the evaluation of a potentially new control tool; other important aspects are the ease with which it could be implemented, its acceptability and its cost. These aspects of IST have been evaluated in the current study and compared with those of SP-IPTp. The results of these studies will be presented in detail elsewhere. However, preliminary analysis of the findings studies indicates that IST can be introduced into a busy, antenatal clinic without disrupting its function. The Comparison was restricted to only women who remained within their randomization group, received two courses of SP-IPTp (IPT group) or were screened twice using an RDT at scheduled visits (IST groups) and in whom the primary outcome had been recorded. b p,0.05 means observed differences between comparison groups is statistically significant or not significant if P.0.05. approach appears acceptable to women who are used to receiving IPT with SP provided that the rationale for the intervention is explained to them [20]. Available evidence suggests that RDTs detect circulating parasite antigen and so compared to microscopy are better at detecting sub microscopic parasitaemia and may be a reliable indicator of placental infections [21]. An IST strategy draws on this and is heavily dependent upon the sensitivity and specificity of the RDT used to detect malaria infection. However, it is uncertain which kind of RDT would be most appropriate for diagnosing malaria in pregnancy. In general, assays based on HRP2 are considered more sensitive and more thermostable than those based on LDH and a sensitive test may be needed to detect parasites in the placenta. On the other hand HRP2 may persist in the circulation for several weeks after parasites have disappeared so that there is a danger that using an HRP2 test for repeated screening may result in women being treated unnecessarily when they attend for their next scheduled screening if they have been treated for an infection a month previously. Conversely a relatively insensitive test may not be able to detect an infection confined largely to the placenta. Few studies [21,22,23,24] have reported good accuracy of RDTs used in the diagnoses of malaria in peripheral and placenta blood at delivery. However, studies of diagnostic accuracy in pregnant women attending antenatal clinics [25,26,27] are scarce and report a range of sentivities. In this study we found that the sensitivity and specificity of the OptiMAL test were 93.8% and 81% respectively compared to microscopy. A large study conducted by the WHO, with collaboration from FIND and CDC, has recently assessed the performance of about 40 commercially available RDTs in terms of their sensitivity, thermostability and ease of use [28]. This study is timely and will guide the choice of malaria RDTs in varying situations but more work is needed on defining the optimum test for detection of malaria in pregnancy.
No difference was found between the prevalence of peripheral blood parasitaemia between the treatment groups at 36-40 weeks of pregnancy or at six weeks post partum but the possibility that IST with SP or AS+AQ is not as effective as IPT with SP in preventing infection of the placenta cannot be excluded. This study had only limited financial resources which precluded analysis of the impact of each of the study interventions on the prevalence of placental malaria as measured by histology. How important such a difference might be in the absence of any apparent clinical adverse effects is uncertain but this needs to be investigated. Other weaknesses of this study are the absence of linked data on HIV positivity, although it is known that the overall prevalence of HIV positivity in the study population is relatively low, approximately 1.5% in the study clinics.
We are confident that the major finding of this study that IST is not inferior to SP-IPTp in the prevention of anaemia in pregnancy and low birth weight is sound and that IST is a potentially promising strategy for the control of malaria in pregnancy, especially in areas where the incidence of malaria is decreasing or where malaria is very seasonal. However, it is important that these encouraging findings including birth outcomes and maternal morbidity findings are confirmed in other geographical areas including parts of eastern and southern Africa, where SP is now failing since we have no alternative drug yet for SP and because the next generation of potential drugs are less likely to fulfill the same favourable profile of SP. As part of these it is also important that the impact of IST on placental malaria is investigated. A large multicentre trial of IST (clinicaltrials.gov identifier: NCT01084213) has started in four countries in Africa where malaria transmission is markedly seasonal will address these issues.
Supporting Information
Checklist S1 | 2014-10-01T00:00:00.000Z | 2010-12-28T00:00:00.000 | {
"year": 2010,
"sha1": "c3158e3a004dae559d039867f19c848ad75ba7d9",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0014425&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "004661a3ad07f8533e2dc881e60cf5a2266c1e4c",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
233715067 | pes2o/s2orc | v3-fos-license | Scalar Adjective Identification and Multilingual Ranking
The intensity relationship that holds between scalar adjectives (e.g., nice<great<wonderful) is highly relevant for natural language inference and common-sense reasoning. Previous research on scalar adjective ranking has focused on English, mainly due to the availability of datasets for evaluation. We introduce a new multilingual dataset in order to promote research on scalar adjectives in new languages. We perform a series of experiments and set performance baselines on this dataset, using monolingual and multilingual contextual language models. Additionally, we introduce a new binary classification task for English scalar adjective identification which examines the models' ability to distinguish scalar from relational adjectives. We probe contextualised representations and report baseline results for future comparison on this task.
Introduction
Scalar adjectives relate the entities they modify to specific positions on the evoked scale (e.g., GOOD-NESS, TEMPERATURE, SIZE): A wonderful view is nicer than a good view, and one would probably prefer a delicious to a tasty meal. But not all adjectives express intensity or degree. Relational adjectives are derived from nouns (e.g., wood → wooden, chemistry → chemical), have no antonyms and serve to classify nouns (e.g., a wooden table, a chemical substance) (McNally and Boleda, 2004). The distinction between scalar and relational adjectives is an important one. Identifying adjectives that express intensity can serve to assess the emotional tone of a given text, as opposed to words that mostly contribute to its descriptive content. Additionally, estimating the intensity of a scalar adjective is useful for textual entailment (wonderful |= good but good |= wonderful), product review analysis and recommendation systems, emotional chatbots and question answering (de Marneffe et al., 2010). DEMELO EN dim < gloomy < dark < black FR terne < sombre < foncé < noir ES sombrío < tenebroso < oscuro < negro EL αμυδρός || αχνός < μουντός < σκοτεινός< μαύρος WILKINSON EN bad < awful < terrible < horrible FR mauvais < affreux < terrible < horrible ES malo < terrible < horrible < horroroso EL κακός < απαίσιος < τρομερός < φρικτός Table 1: Example translations from each dataset. "||" indicates adjectives at the same intensity level (ties).
Work on scalar adjectives has until now evolved around pre-compiled datasets (de Melo and Bansal, 2013;Taboada et al., 2011;Wilkinson and Oates, 2016;Cocos et al., 2018). Reliance on external resources has also restricted research to English, and has led to the prevalence of pattern-based and lexicon-based approaches. Recently, Garí Soler and Apidianaki (2020) showed that BERT representations (Devlin et al., 2019) encode intensity relationships between English scalar adjectives, paving the way for applying contextualised representations to intensity detection in other languages. 1 In our work, we explicitly address the scalar adjective identification task, overlooked until now due to the focus on pre-compiled resources. We furthermore propose to extend scalar adjective ranking to new languages. We make available two new benchmark datasets for scalar adjective identification and multilingual ranking: (a) SCAL-REL, a balanced dataset of relational and scalar adjectives which can serve to probe model representations for scalar adjective identification; and (b) MULTI-SCALE, a scalar adjective dataset in French, Spanish and Greek. In order to test contextual models on these two tasks, the adjectives need to be seen in sentential context. We thus provide, alongside the datasets, sets of sentences that can be used to extract contextualised representations in order to promote model comparability. We conduct experiments and report results obtained with simple baselines and state-of-the-art monolingual and multilingual models on these new benchmarks, opening up avenues for research on sentiment analysis and emotion detection in different languages. 2 2 The Datasets
The MULTI-SCALE Dataset
We translate two English scalar adjective datasets into French, Spanish and Greek: DEMELO consists of 87 hand crafted half-scales 3 (de Melo and Bansal, 2013) and WILKINSON contains 12 full scales (Wilkinson and Oates, 2016). We use the partitioning of WILKINSON into 21 half-scales proposed by Cocos et al. (2018). In what follows, we use the term "scale" to refer to half-scales.
The two translators have (near-)native proficiency in each language. They were shown the adjectives in the context of a scale. This context narrows down the possible translations for polysemous adjectives to the ones that express the meaning described inside the scale. For example, the Spanish translations proposed for the adjective hot in the scales {warm < hot} and {flavorful < zesty < hot || spicy} are caliente and picante, respectively. Additionally, the translators were instructed to preserve the number of words in the original scales when possible. In some cases, however, they proposed alternative translations for English words, or none if an adequate translation could not be found. As a result, the translated datasets have a different number of words and ties. modify. In order to keep the method resource-light, we gather sentences that contain the adjectives in their unmarked form.
For each scale s, we randomly select ten sentences from OSCAR where adjectives from s occur. Then, we generate additional sentences through lexical substitution. Specifically, for every sentence (context) c that contains an adjective a i from scale s, we replace a i with ∀ a j ∈ s where j = 1...|s| and j = i. This process results in a total of |s| * 10 sentences per scale and ensures that ∀ a ∈ s is seen in the same ten contexts. For English, we use the ukWaC-Random set of sentences compiled by Garí Soler and Apidianaki (2020) which contains sentences randomly collected from the ukWaC corpus (Baroni et al., 2009).
The SCAL-REL Dataset
SCAL-REL contains scalar adjectives from the DEMELO, WILKINSON and CROWD (Cocos et al., 2018) datasets (i.e. 79 additional half-scales compared to MULTI-SCALE). We use all unique scalar adjectives in the datasets (443 in total), and subsample the same number of relational adjectives, which are labelled with the pertainym relationship in WordNet (Fellbaum, 1998). There are 4,316 unique such adjectives in WordNet, including many rare or highly technical terms (e.g., birefringent, anaphylactic). 4 Scalar adjectives in our datasets are much more frequent than these relational adjectives; their average frequency in Google Ngrams (Brants and Franz, 2006) is 27M and 1.6M, respectively. We balance the relational adjectives set by frequency, by subsampling 222 frequent and 221 rare adjectives. We use the mean frequency of the 4,316 relational adjectives in Google Ngrams as a threshold. 5 We propose a train/dev/test split of the SCAL-REL dataset (65/10/25%), observing a balance between the two classes (scalar and relational) in each set. To obtain contextualised representations, we collect for each relational adjective ten random sentences from ukWaC. For scalar adjectives, we use the ukWaC-Random set of sentences (cf. Section 2.1).
Methodology
Models We conduct experiments with state-ofthe-art contextual language models and several baselines on the MULTI-SCALE dataset. We use the pre-trained cased and uncased multilingual BERT model (Devlin et al., 2019) and report results of the best variant for each language. We also report results obtained with four monolingual models: bert-base-uncased (Devlin et al., 2019), flaubert_base_uncased (Le et al., 2020), bert-base-spanish-wwmuncased (Cañete et al., 2020), and bert-basegreek-uncased-v1 (Koutsikakis et al., 2020). We compare to results obtained using fastText static embeddings in each language (Grave et al., 2018).
For a scale s, we feed the corresponding set of sentences to a model and extract the contextualised representations for ∀ a ∈ s from every layer. When an adjective is split into multiple BPE units, we average the representations of all wordpieces (we call this approach "WP") or all pieces but the last one ("WP-1"). The intuition behind excluding the last WP is that the ending of a word often corresponds to a suffix with morphological information.
The DIFFVEC method We apply the adjective ranking method proposed by Garí Soler and Apidianaki (2020) to our dataset, which relies on an intensity vector (called − −− → dV ec) built from BERT representations. The method yields state-of-the art results with very little data; this makes it easily adaptable to new languages. We build a sentence specific intensity representation ( − −− → dV ec) by subtracting the vector of a mild intensity adjective, a mild (e.g., smart), from that of a ext , an extreme adjective on the same scale (e.g., brilliant) in the same context.
We create a dV ec representation from every sentence available for these two reference adjectives, and average them to obtain the global − −− → dV ec for that pair. Garí Soler and Apidianaki (2020) showed that a single positive adjective pair (DIFFVEC-1 (+)) is enough for obtaining highly competitive results in English. We apply this method to the other languages using the translations of a positive English (a mild , a ext ) pair from the CROWD dataset: perfect-good. 6 Additionally, we learn two dataset specific representations: one by averaging the − −− → dV ec's of all (a ext , a mild ) pairs in WILKINSON that do not appear in DEMELO (DIFFVEC-WK), and another one from pairs in DEMELO that are not in WILKINSON (DIFFVEC-DM). We rank adjectives in a scale by their cosine similarity to each − −− → dV ec: The higher the similarity, the more intense the adjective is.
Baselines We compare our results to a frequency and a polysemy baseline (FREQ and SENSE). These baselines rely on the assumption that low intensity words (e.g., nice, old) are more frequent and polysemous than their extreme counterparts (e.g., awesome, ancient). Extreme adjectives often limit the denotation of a noun to a smaller class of referents than mild intensity adjectives (Geurts, 2010). For example, an "awesome view" is more rare than a "nice view". This assumption has been confirmed for English in Garí Soler and Apidianaki (2020). FREQ orders words in a scale according to their frequency: Words with higher frequency have lower intensity. Given the strong correlation between word frequency and number of senses (Zipf, 1945), we also expect highly polysemous words (which are generally more frequent) to have lower intensity. This is captured by the SENSE baseline which orders the words according to their number of senses: Words with more senses have lower intensity.
Frequency is taken from Google Ngrams for English, and from OSCAR for the other three languages. The number of senses is retrieved from WordNet for English, and from BabelNet (Navigli and Ponzetto, 2012) for Spanish and French. 7 For adjectives that are not present in BabelNet, we use a default value which corresponds to the average number of senses for adjectives in the dataset (DEMELO or WILKINSON) for which this information is available. We omit the SENSE Greek due to low coverage. 8
Evaluation
We use evaluation metrics traditionally used for ranking evaluation (de Melo and Bansal, 2013;Cocos et al., 2018): Pairwise accuracy (P-ACC), Kendall's τ and Spearman's ρ. Results on this task are given in Table 3. Monolingual models perform consistently better than the multilingual model, except for French. We report the best wordpiece approach for each model: WP-1 works better with all monolingual models and the multilingual model for English. Using all wordpieces (WP) is a better choice for the multilingual model in other languages. We believe the lower performance of WP-1 in these settings to be due to the fact that the multilingual BPE vocabulary is mostly English-driven; this naturally results in highly arbitrary partitionings in these languages (e.g., ES: fantástico → fantástico; EL: γιγάντιος (gigantic)→γ-ι-γ-άν-τιος). Tokenisers of the monolingual models instead tend to split words in a way that more closely reflects the morphology of the language (e.g., ES: fantástico → fantás-tico; EL: γιγάντιος→γιγά-ντι-ος. Detailed results are found in Appendix A. We observe that DIFFVEC-1 (+) yields comparable and sometimes better results than DIFFVEC-DM and DIFFVEC-WK, which are built from multiple pairs. This is important especially in the multilingual setting, since it shows that just one pair of adjectives is enough for obtaining good results in a new language. The best layer varies across models and configurations. The monolingual French and Greek models generally obtain best results in earlier layers. A similar behaviour is observed for the multilingual model for English to some extent, whereas for the other models performance improves in the upper half of the Transformer network (layers 6-12). This shows that the semantic information relevant for adjective ranking is not situated at the same level of the Transformer in different languages. We plan to investigate this finding further in future work. The lower results in French can be due to the higher amount of ties present in the datasets compared to other languages. 9 The baselines obtain competitive results showing that the underlying linguistic intuitions hold across languages. The best models beat the baselines in all configurations except for Greek on the DEMELO dataset, where FREQ and static embeddings obtain higher results. Overall, results are lower than those Figure 1: Illustration of two scalar adjectives that are close to − −− → dV ec and to its opposite (which represents low intensity). The red vector describes a relational adjective that is perpendicular to − −− → dV ec.
reported for English, which shows that there is room for improvement in new languages.
Scalar Adjective Identification
For each English adjective in the SCAL-REL dataset, we generate a representation from the available ten sentences (cf. Section 2.2) using the bert-base-uncased model (with WP and WP-1). We experiment with a simple logistic regression classifier that uses the averaged representation for an adjective (ADJ-REP) as input and predicts whether it is scalar or relational. We also apply the DIFFVEC-1 (+) method to this task and measure how intense an adjective is by calculating its cosine with − −− → dV ec. The absolute value of the cosine indicates how clearly an adjective encodes the notion of intensity. In Figure 1, we show two scalar adjective vectors with negative and positive cosine similarity to − −− → dV ec, and another vector that is perpendicular to − −− → dV ec, i.e. describing a relational adjective for which the notion of intensity does not apply. 10 We train a logistic regression model to find a cosine threshold separating scalar from relational adjectives (DV-1 (+)). Finally, we also use as a feature the cosine similarity of the adjective representation to the vector of "good", which we consider as a prototypical scalar adjective (PROTO-SIM).
The best BERT layer is selected based on the accuracy obtained on the development set. We report accuracy on the test set. The baseline classifiers only use frequency (FREQ) and polysemy (SENSE) as features. We use these baselines on SCAL-REL because the WordNet pertainyms included in the dataset are rarer than the scalar adjectives. The intuition behind the SENSE baseline explained in Section 3.1 also applies here. Results on this task are given in Table 4. The classifier that relies on ADJ-REP BERT representations can distinguish the two types of adjectives with very high accuracy (0.946), closely followed by fastText embeddings (0.929). The DV-1 (+) method does not perform as well as the classifier based on ADJ-REP, which is not surprising since it relies on a single feature (the absolute value of the cosine between − −− → dV ec and ADJ-REP). Comparing ADJ-REP to a typical scalar word (PROTO-SIM) yields better results than DV-1 (+). The SENSE and FREQ baselines can capture the distinction to some extent. Relational adjectives in our training set are less frequent and have fewer senses on average (2.59) than scalar adjectives (5.30). A closer look at the errors of the best model reveals that these concern tricky cases: One of the four misclassified scalar adjectives is derived from a noun (microscopic), whilst five out of eight wrongly classified relational adjectives can have a scalar interpretation (e.g., sympathetic, imperative). Overall, supervised models obtain very good results on this task. SCAL-REL will enable research on unsupervised methods that could be used in other languages.
Conclusion
We propose a new multilingual benchmark for scalar adjective ranking, and set performance baselines on it using monolingual and multilingual contextual language model representations. Our results show that adjective intensity information is present in the contextualised representations in the studied languages. We also propose a new classification task and a dataset that can serve as a benchmark to estimate the models' capability to identify scalar adjectives when relevant datasets are not available. We make our datasets and sentence contexts available to promote future research on scalar adjectives detection and analysis in different languages. Table 3 of the main paper contains results of the DIFFVEC method with the best approach for selecting wordpieces (WPs) for each model. In Table 5, we present results obtained using the alternative approach for each model and language:
A Comparison of Wordpiece Selection Methods
• for all monolingual models and the multilingual model for English, Table 5 contains results obtained with the WP approach; • for the multilingual models in the other languages, we show results with WP-1.
The best approach was determined by comparing their average scores across the different methods. Some configurations improve, but they yield overall worse results per model, especially in Spanish. Differences between WP and WP-1 are generally more pronounced in the multilingual models than in the monolingual models. Table 5: Results of DIFFVEC (DV) methods with contextualised representations derived from monolingual and multilingual models for each language, using an alternative approach to selecting wordpieces (WP, WP-1) than the one used for the results reported in Table 3. For all languages but Greek, the multilingual model is cased. | 2021-05-05T01:31:36.354Z | 2021-05-03T00:00:00.000 | {
"year": 2021,
"sha1": "6498f3a9101bc8728c9013e75cfbbefbf5b24e5e",
"oa_license": "CCBY",
"oa_url": "https://www.aclweb.org/anthology/2021.naacl-main.370.pdf",
"oa_status": "HYBRID",
"pdf_src": "Arxiv",
"pdf_hash": "6498f3a9101bc8728c9013e75cfbbefbf5b24e5e",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
15013447 | pes2o/s2orc | v3-fos-license | The Use of Parsimonious Questionnaires in Occupational Health Surveillance: Psychometric Properties of the Short Italian Version of the Effort/Reward Imbalance Questionnaire
Purpose. To perform a parsimonious measurement of workplace psychosocial stress in routine occupational health surveillance, this study tests the psychometric properties of a short version of the original Italian effort-reward imbalance (ERI) questionnaire. Methods. 1,803 employees (63 percent women) from 19 service companies in the Italian region of Latium participated in a cross-sectional survey containing the short version of the ERI questionnaire (16 items) and questions related to self-reported health, musculoskeletal complaints and job satisfaction. Exploratory factor analysis, internal consistency of scales and criterion validity were utilized. Results. The internal consistency of scales was satisfactory. Principal component analysis enabled to identify the model's main factors. Significant associations with health and job satisfaction in the majority of cases support the notion of criterion validity. A high score on the effort-reward ratio was associated with an elevated odds ratio (OR = 2.71; 95% CI 1.86–3.95) of musculoskeletal complaints in the upper arm. Conclusions. The short form of the Italian ERI questionnaire provides a psychometrically useful tool for routine occupational health surveillance, although further validation is recommended.
Background
Work-related stress is one of the leading causes of workers' ill health in developed countries [1], where it has considerable effects on sickness absence and disability [2]. Epidemiological investigations document significant associations of stressful work with coronary heart disease [3][4][5], depression [6][7][8], musculoskeletal disorders [9,10], and other stressrelated health problems [11]. In recent years, specific theoretical concepts and their measurement in terms of psychometrically validated standardized questionnaires has led to progress in identifying psychosocial working conditions that produce adverse health effects.
Occupational epidemiological research has focused on two such concepts: the "demand-control" model and the "effort-reward imbalance" model. The former assumes that job task profiles characterized by high psychological demands and a low level of control and decision latitude are stressful [12,13], whereas the latter is concerned with stressful features of the work contract resulting from failed reciprocity between (high) efforts spent and (low) rewards received [14,15]. In this model, three types of rewards are important: money, esteem, and career, including job security. Moreover, in addition to extrinsic effort this model assesses a distinct personal pattern of coping with job demands, termed overcommitment. According to this theoretical approach, workers who experience high effort, low reward, and a mismatch between them (high cost-low gain), and workers who exhibit a high level of overcommitment are susceptible to an elevated risk of stress-related disorders. 2 The Scientific World Journal The two models complement each other, and some evidence indicates that the former concept is of particular use in industrial workers, whereas the latter may be more appropriate in the tertiary sector [16,17]. In several countries, including Italy, routine assessment of psychosocial stress at work has become compulsory for occupational health services. Therefore, short, validated, and easily applicable questionnaires are needed to assess the prevalence of workrelated stress as a basis for potential preventive efforts. A few years ago, a short version of the original questionnaire used to measure the effort-reward imbalance (ERI) model [14,15] was developed and tested in a German [18] and a Swedish [19] sample of male and female workers. We set out to conduct a psychometric test of the short version of the original Italian questionnaire [20] in a large sample of male and female employees in Italy. More specifically we aim at replicating its construct validity and analyzing its association with job satisfaction, musculoskeletal complaints, and selfrated health.
Methods
In 2010, workers undergoing regular health surveillance in the workplace under the responsibility of the first author were asked to complete an anonymous questionnaire composed of three sections; (1) basic socio-demographic information (restricted to gender and age to ensure anonymity); (2) musculoskeletal disorders, general health and job satisfaction; (3) the short version of the ERI questionnaire. As mentioned, health surveillance is mandatory in Italy for workers exposed to occupational hazards, yet assessment procedures are flexible, thus, allowing some innovation-in our case the administration of the short ERI questionnaire. Workers who had been employed for at least one year in the same workplace were eligible (see sample description below). The study was approved by the Ethics Committee of the Catholic University of Sacred Heart. Musculoskeletal complaints were assessed by the Nordic questionnaire [21]. Participants were asked to assess (1) if they had had problems (pain, ache, discomfort) at any time during the previous 12 months (recurrent symptoms); (2) if they had had problems during the previous 7 days (persistent symptoms); (3) if they were experiencing problems at the time of the medical examination (active symptoms). Those who gave a positive response were requested to specify in which of 9 parts of the body (neck, shoulders, elbows, wrist and hands, upper back, low back, hips and thighs, knees, ankles, and feet) they had experienced health problems. All items are binary, with the category "yes" defining presence and "no" defining absence of musculoskeletal symptoms.
The questionnaire provides information about one-year prevalence and the point prevalence of musculoskeletal complaints.
General self-rated health was measured by the one-item question: "How would you rate your general state of health?". Respondents answered on a scale from 1 to 5 ranging from 1 "I've serious health problems" to 5 "I've very good health." Job satisfaction was measured by a single item of the scale proposed by Warr et al. [22]: "How do you feel about your job as a whole?". Respondent answered on a scale ranging from 1 "I'm extremely dissatisfied" to 7 "I'm extremely satisfied." Effort, reward, and overcommitment were measured by the short version of the ERI questionnaire where effort is measured by three questions, reward by seven questions, and overcommitment by six questions [18]. All respective items were answered on a four-point Likert scale (ranging from 1 "strongly agree" to 4 "strongly disagree"). Total scores of these ratings were calculated with appropriate recoding, so that high scores reflect high effort, high reward and high overcommitment. Thus, the range of effort scale is 3-12, of the reward scale is 7-28, and of the overcommitment scale is 6-24. Additionally, an effort-reward ratio was calculated by dividing the score of "effort" by the score of reward, adjusted for unequal number of items. This was done in order to quantify the degree of mismatch between effort and reward at individual level.
Sample.
This voluntary survey was offered to all male and female workers from 19 different service companies who had been employed there for more than one year. Overall, 1,803 subjects (669 male, 37%, 1,134 female workers, 63%) completed the questionnaire and were included in the analysis. The response rate ranged from 89% to 100% (average: 95.3%). Given the high response rate there was no indication of selection bias. Workers were employed in health care services (n = 998, 55.4%), social services (n = 395, 21.9%), retail (n = 281, 15.6%), and financial services (n = 129, 7.1%) ( Table 1).
Statistical
Analysis. An exploratory factor analysis, with principal component extraction and varimax rotation method with Kaiser normalization was used to test the psychometric characteristics of the scales of the ERI short questionnaire. To identify the scales, we adopted the criteria that item loading should be greater than .50 on the loading factor and at least .20 above that for any other factors, thus, indicating that the item was conceptually distinct [23]. The resulting factor structure was compared to the original construct of the German version. Reliability of each subscale was calculated with the Cronbach's alpha statistic.
We tested some aspects of criterion validity of the short ERI version assuming that employees who scored high on the scales of the construct were at elevated risk of experiencing poor self-rated health and poor job satisfaction, compared to those with lower scores. The associations between exposure variables and outcomes were examined by multiple linear regression analysis. After first calculating an unadjusted model, we subsequently adjusted for age and gender, entering effort, reward and overcommitment as independent variables. We then repeated the same steps entering effort-reward ratio as an independent variable.
Finally, the association between exposure variables and musculoskeletal complaints was investigated by logistic regression analysis. First we calculated an adjusted model, and then we adjusted it for age and gender. Odds ratios and confidence intervals at 95% were calculated. We used version 15.0 of SPSS for Windows for the statistical analyses.
Psychometric Properties.
The factor analysis identified four factors that accounted for 53% of the total variance. After rotation, each factor was seen to correspond to a specific construct, as in the original version ( Table 2). The first factor corresponded to the "overcommitment" scale, the second one to the "effort" scale, and the last two factors to the "reward" construct. The subdivision of "reward" into two factors, although already observed in a previous Italian study based on the original 23-item questionnaire [20], did not conform to the theoretical assumption of three subcomponents. Although "job security" was well replicated, there was no clear distinction between "esteem" and "salary, career prospects." It seems that the items measuring non-monetary rewards (esteem) and those measuring monetary and statusrelated rewards "status" did not cluster in two separate factors. Almost all items showed significant factor loadings (>0.50) on the factor corresponding to their construct, and negligible loading (<0.20) on other factors; only two items pertaining to the "overcommitment" scale (oc1 and oc4) had loadings inferior to .50 on their factor, but their loadings on other scales were negligible, thus, providing evidence of their specificity.
Using factor analysis, we composed 4 scales ("effort," "job security," "esteem and status", "overcommitment") and calculated their internal consistency. Cronbach's alpha coefficients were 0.72 for "overcommitment", 0.69 for "effort", 0.68 for "esteem and status", and 0.63 for "job security", suggesting satisfactory internal consistency in view of the small number of items (Table 2). Item-total correlations coefficients (corrected) varied between 0.34 and 0.61 and were all above the threshold of 0.30 [24], indicating considerable consistency between items defining respective scales. Using the theoretical model, we constructed a general "reward" scale based on the two subscales, and a ratio of "effort" and "reward" was defined according to established procedures [15]. Thus, in further analyses that tested associations with health and job satisfaction, four explanatory variables were included ("effort", "reward", "effort-reward ratio", "overcommitment").
Associations with Health and Job Satisfaction.
Multiple regression analyses showed that the "effort", "reward" and "overcommitment" scales' were significantly related to selfrated general health. "Effort" had an inverse relationship, while "reward" and also "overcommitment" were positively related to the level of self-rated health. Correction for age and gender weakened the association between "effort" and health to some extent, while the relationships between "reward", "overcommitment" and health remained unchanged. Low levels of "effort" and high levels of "reward" were significantly associated with job satisfaction, whereas there was no evident association with "overcommitment". Significant relationships were observed between the "effort-reward ratio" and self-rated health as well as job satisfaction. As illustrated in Table 3, scoring low on the "effort-reward ratio" was associated with good self-rated health and high job satisfaction (Table 3).
Discussion
This study was set up to test the psychometric properties of the short version of the original questionnaire that measured occupational effort-reward imbalance in a convenience sample of Italian workers. Data show satisfactory internal consistency of these shortened scales, and exploratory factor analysis supported the conceptual distinction between the "effort" and "over commitment" scales. Contrary to most other studies [15,18,19,25], but in keeping with a previous Italian investigation based on the original questionnaire [20], we observed two (instead of three) subcomponents of the construct "reward" as there was no clear distinction between nonmaterial and material components of jobrelated rewards. Although this result calls for further investigation, it may well be that it reflects the specific sample composition of this study which was largely composed of women working in health and social services where the material and nonmaterial rewards of professional work are often not separated as clearly as is the case in other branches and sectors of the labor market. Model fits derived from confirmatory factor analysis were not tested, but in keeping with previous studies, a summary "reward" scale and a ratio quantifying the mismatch between effort and reward at individual level were constructed for further analyses of criterion validity. With this aim in mind, we analyzed associations with some health indicators (selfrated health and two types of musculoskeletal complaints). While results in general support the hypothesis, there were some noticeable exceptions.
First, "overcommitment" was positively associated with self-rated health and musculoskeletal complaints. This contradicts previous findings demonstrating reduced rather than enhanced self-reported health among overcommitted workers [18,25,26]. In the Italian sample, mean scores of "overcommitment" were not very high, and this fact may prevent the detection of adverse health effects. Secondly, while all associations of "reward" with the criterion variables are significant, there is an exception to this trend as regards the "effort-reward ratio" (low back pain). On the other hand, in accordance with the theoretical assumption, associations with the ratio are generally stronger than those observed with single scales of the model.
There are obvious limitations to this study. Firstly, data are drawn from a limited sample of tertiary sector companies, located in the Latium region of Italy and under the responsibility of a single physician. It is, therefore, not clear to what extent findings can be extended to other geographic areas or economic sectors in Italy. In this study, no measure of negative affectivity was available, so reporting bias cannot be excluded. We cannot exclude bias due to common method variance, given the fact that both workrelated and health-related questions are based on subjective evaluations reported within the same assessment. The crosssectional study design, with restriction to a single exposure assessment, precludes any reference to temporal or causal directions of observed statistical associations. Finally, we provide only overall results, without specifying them according to occupational categories, hazard exposure, socioeconomic conditions, education, or type and duration of employment.
Yet, this study confirms the usefulness of a newly developed short version of the ERI questionnaire in the Italian context, as its basic findings are in agreement with the results of two previously published reports from Germany and Sweden. Moreover, it is one of the first studies applying a 4-point Likert scale format for answering the items instead of the previous 5-point Likert scale which has been the subject of methodological criticism [27]. Finally, given the exceptionally high response rate of this survey, selection bias is virtually ruled out.
In conclusion, by balancing the limitations and strengths of this study, our results indicate that the short form of ERI provides a psychometrically useful tool for epidemiological studies at the workplace. Given the model's success in Table 4: Associations of the scales of the short effort-reward imbalance questionnaire with acute low back pain and musculoskeletal problems at the upper arm (logistic regression analyses) (Odds ratios and 95% confidence intervals (CI)).
Unadjusted
Adjusted explaining work-related health risks it may be particularly useful for application in routine medical surveillance where parsimonious assessment has high priority. | 2016-05-04T20:20:58.661Z | 2012-08-13T00:00:00.000 | {
"year": 2012,
"sha1": "c0e24989ff153ac791a71ffe8d67e57817679ee4",
"oa_license": "CCBY",
"oa_url": "http://downloads.hindawi.com/journals/tswj/2012/372852.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "d4b2d44d9fbd0133c418ecd39f45b73b353aa41a",
"s2fieldsofstudy": [
"Medicine",
"Psychology"
],
"extfieldsofstudy": [
"Psychology",
"Medicine"
]
} |
53393251 | pes2o/s2orc | v3-fos-license | Geometrical Analysis of Warp Knit Auxetic Fabrics
In recent years, considerable interests have been shown on the formation, properties and characteristics of auxetic knit structures [17]. Auxetic structures can enable an article to exhibit an expansion in a lateral direction, upon subjecting the article to a longitudinal stress or strain. Conversely, auxetic structures also exhibit a contraction in the lateral direction upon subjecting such an article to longitudinal compression. Such materials are understood to exhibit a negative Poisson’s ratio. However, very limited studies have been produced on the fundamental geometrical analytics of warp knit auxetic fabrics. It is recognized that the geometry of fabrics has significant effects on their behavior. Hence Ugbolue et al. [4] focused on the geometry and structural properties of warp knit auxetic fabrics in one of their papers. Generally, the complexity of warp knitted structures have posed major challenges to researchers but several studies on the geometry of warp knitted structures have been published in the literature [8]. Early attempts on the study of the subject were experimental studies in which the dimensional properties of the warp-knitted structure were investigated by Fletcher and Roberts [9,10] and by Stimmel [11,12]. Later, a geometrical model for warp-knitted structures was developed by Allison [13] and then Grosberg [14] presented a full picture of the warp-knitted-loop configuration for two-bar warp-knitted fabrics. Subsequently, Grosberg [15] improved the previously suggested geometrical model wherein he assumed that the root end of the loop laid at the widest section of the previous loop and the underlap was part of a circle. The third geometrical model was developed by Shinn and El-Araf [16]. Wheatley [17] and Jacobsen [18] also proposed some geometrical loop models for warp-knitted structures produced from wool yarns. These researchers were keenly interested in finding some relationships between fabric parameters and fabric dimensions that were akin to those found for weft-knitted structures in order to facilitate some production calculations in industrial settings and encourage efforts in predicting some fabric properties after [19]. Our thrust in prior research combined our knowledge of geometry and fabric structural characteristics to engineer auxetic textiles and to determine the properties of such auxetic textile fabrics. Auxetic warp knit fabrics have great potentials in many areas of application such as protective clothing, blast resistant uniforms and other industrial textiles.
Introduction
In recent years, considerable interests have been shown on the formation, properties and characteristics of auxetic knit structures [1][2][3][4][5][6][7]. Auxetic structures can enable an article to exhibit an expansion in a lateral direction, upon subjecting the article to a longitudinal stress or strain. Conversely, auxetic structures also exhibit a contraction in the lateral direction upon subjecting such an article to longitudinal compression. Such materials are understood to exhibit a negative Poisson's ratio. However, very limited studies have been produced on the fundamental geometrical analytics of warp knit auxetic fabrics. It is recognized that the geometry of fabrics has significant effects on their behavior. Hence Ugbolue et al. [4] focused on the geometry and structural properties of warp knit auxetic fabrics in one of their papers. Generally, the complexity of warp knitted structures have posed major challenges to researchers but several studies on the geometry of warp knitted structures have been published in the literature [8]. Early attempts on the study of the subject were experimental studies in which the dimensional properties of the warp-knitted structure were investigated by Fletcher and Roberts [9,10] and by Stimmel [11,12]. Later, a geometrical model for warp-knitted structures was developed by Allison [13] and then Grosberg [14] presented a full picture of the warp-knitted-loop configuration for two-bar warp-knitted fabrics. Subsequently, Grosberg [15] improved the previously suggested geometrical model wherein he assumed that the root end of the loop laid at the widest section of the previous loop and the underlap was part of a circle. The third geometrical model was developed by Shinn and El-Araf [16]. Wheatley [17] and Jacobsen [18] also proposed some geometrical loop models for warp-knitted structures produced from wool yarns. These researchers were keenly interested in finding some relationships between fabric parameters and fabric dimensions that were akin to those found for weft-knitted structures in order to facilitate some production calculations in industrial settings and encourage efforts in predicting some fabric properties after [19]. Our thrust in prior research combined our knowledge of geometry and fabric structural characteristics to engineer auxetic textiles and to determine the properties of such auxetic textile fabrics. Auxetic warp knit fabrics have great potentials in many areas of application such as protective clothing, blast resistant uniforms and other industrial textiles.
Our continued efforts to produce auxetic knit structures from non-auxetic yarns are described here. Specifically, this paper examines our geometrical model of auxetic warp knit structure and validates its characteristics with data obtained from experimental analysis of nine recently produced warp knit auxetic fabrics.
Theoretical analysis
A typical warp knit structure under strain is shown in Figure 1. The length of interlooping repeat that is formed by alternation of tricot (n t ) and chain (n c ) courses during drawing off at knitting machine ( Figure 1) could be calculated as follows: Where B t max and B c max are maximum height of tricot and chain courses under a strain respectively mm: Where lt and lc are loop length of tricot and chain, mm respectively, dmin is the diameter of yarn under strain, mm; After relaxation as shown in Figure 2, Where a 1 is length of vertical rib, mm h is distance between two vertical ribs in wale wise direction, mm.
It is obviously clear that: Where B t is height of tricot course mm, Where B c -height of chain courses, mm 3( 1) 3 it is required to employ a high elastic yarn in the basic filet warp knit structure. This yarn must to be placed between the stitch wale in the knitting direction to insure that the fabric structure will retain necessary configuration after relaxation. The filling yarn makes contact with the ground yarn just at the vertical ribs from tricot courses and is inlaid between vertical and diagonal ribs inside the hole at the other part of net. Also in their previous paper, Kyzymchuk and Ugbolue [20] affirmed that the positioning of inlaid yarns in filet warp knit structure is determined by the amount and direction of inlaying and ground guide bars' shifting behind the needles as well as disposition of guide bars with ground and filling yarns. It was determined that the in-laid yarn could be positioned inside the structure in one, two, three or five courses of vertical rib and might also be laid between the tricot's junctures from different ground yarns or wrapped junctures from one or both ground guide bars. Thus, two variants of inlaid yarn positioning are showed at Figure 3.
Consequently, the length of interloping repeat of such structure could be calculated as: Where L i is length of elastomeric yarn that is fixed in the knitted structure, L f is length of elastomeric yarn that is not fixed in the knitted structure For our modelling we adopt the following assumptions: (i) The relaxation of elastomeric yarn, that is fixed in the structure, is insignificant and does not affect the size and shape of the loops; (ii) A full relaxation of elastomeric yarn occurs only in the area where the yarn is not fixed in the structure.
Thus, lengths of elastomeric yarn that are fixed in knitted structure are determined as follows: during drawing off at knitting machine: after relaxation: Where i is the number of tricot courses in which the elastomeric yarn is fixed. Where d is the diameter of yarn, mm, and α is angle between diagonal rib and horizontal.
The geometrical model depicted in Figure 2b shows the configuration of our Auxetic warp knit structure. Thus, the length of interlooping repeat of conventional warp knit structure is given by: Also, Figure 2b depicts the auxetic warp knit geometrical structure with hexagonal mesh and its auxetic configuration is realized when the following condition is satisfied: Therefore, the length of interlooping repeat of auxetic warp knit structure is: As was proposed by Ugbolue et al. [3,4] to achieve auxetic property, Based on prior presented dependences, the lengths of elastomeric yarn, which are not fixed in the structure, are given as follows: length the knitting machine: length needed to achieve auxetic effect after relaxation: Thus, to achieve the auxetic effect in warp knitted filet structure the elastomeric in-laid yarn must be fed with the prior stretching, , which is determined as follows: It is evident that the strain value decreases with increasing number of tricot courses and decreasing number of chain courses within a repeat of filet interlooping as well as the decreasing number of courses in which the in-laid yarn is fixed within the warp knit structure.
Typical results of calculated theoretical strain value of elastomeric yarn that predict auxetic property of filet net structure are shown at Table 1. The theoretical Poisson's ratio has been calculated with reference to the relationships derived from Figure 2 as follows: It is observed that the value of negative Poisson's ratio increases with the number of tricot courses as well as chain courses of the interlooping repeat. Thus, the best auxetic property of the filet warp knit is expected for the structure formed by alternation of 7 tricot and 3 chain courses at repeat, but prior stretching of elastomeric yarn should be increased.
Production of Auxetic Fabrics
In order to verify the auxetic behavior of the model, nine different filet warp knit fabrics with in-lay yarn that is fixed in two courses of repeat as shown in Table 2 were produced. These fabrics were made on a 10 gauge Jakob Muller crochet knitting machine with one needle bed. The warp knit fabrics were made from 250 denier polyester yarn as ground. The 150 denier (96 filaments) polyester yarn covered with one end of 70 denier spandex was used to achieve a high elastic inlay component. Elastomeric yarn was fed into the knitting zone with elongation of 150% [21]. This elastomeric yarn is in-laid between the stitch wale in the knitting direction to ensure that the fabric structure retains the necessary configuration after relaxation.
Measurement of some fabric physical properties
A description of each measured property is given below:
Courses per unit length and Wales per unit length:
A course is a predominantly horizontal row of loops produced by adjacent needles during the same knitting cycle. In warp knitting each loop in a course normally is composed of a separate yarn. Courses per inch (cpi) or courses per cm are used to evaluate the loops along the axial or walewise direction. A wale is a predominantly vertical column of needle loops produced by the same needle knitting at successive knitting cycles and thus intermeshing each new loop through the previous loop. In warp knitting a wale can be produced from the same yarn. Wales per inch (wpi) or wales per cm are used to evaluate the loops along the transverse or coursewise direction. The number of courses and the number of wales per unit length are obtained by using a counting glass. The results are reported using SI units namely, number of courses or wales in 10 mm. The mean of ten readings is recorded for each direction of the fabric.
Stitch density, S: The term loop or stitch density, S, is the total number of needle loops in a square area measurement. It is obtained by multiplying, for instance, the number of courses and wales, per square unit length together. The unit is loops/in 2 or loops per cm 2 .
Thickness and basis weight:
The thickness (with the unit of mm) of each sample is tested using a Thickness Testing Instrument according to ASTM D1777-64.8. Each sample is also weighed on an electronic balance to determine its basis weight or areal density (g/m 2 ).
Measurement of Poisson's ratio:
To measure the Poisson's ratio of the identified fabrics, video-extensometry along with micro-tensile testing techniques are employed by using Instron 5569 Mechanical Tester ASTMD5034-95 (2001). All samples are tested by straining the entire fabric strip evenly and each duly marked 2 cm × 2 cm square of the sample is measured to obtain the Poisson's ratio by using the equation, ν xy = -ε x /ε y where ε x is the strain in the x-direction, or transverse strain, and ε y is the strain in the y-direction, or the axial strain. Initially all samples of 10 cm long are strained at a rate of 5.08 cm/min, in the walewise and coursewise directions. The test process is observed with a Canon Pro 300 camera. The strain of the sample is measured using the camera to capture an image of the sample at different strain levels, totaling 16 pictures per sample. The width of each sample is measured in three locations to insure that the measured Poisson's ratio is as accurate of measurement as possible [22]. Each fabric structure is tested three times, using different samples of the fabric structure each test. The Poisson's ratio is obtained after all samples had been photographed and strain values obtained using appropriate image analysis software. It should be noted in the photographs that the relationship of interest (for determining if auxetic) is not necessarily the relationship of the strain in the y direction to the x direction. The axial strain can still be greater than the transverse strain (as pictured), yet the auxetic nature is a reflection of the transverse strain increasing under axial load (as measured, the width of the box increased relative to the initial width). All the samples were tested along the wale direction. Then the Poisson's ratio was calculated using the equation: where vxy is Poison's ratio, ε x is transverse strain, and ε y is axial strain.
Poisson's ratio is the ratio of transverse contraction strain to longitudinal extension strain in the direction of stretching force.
Results and Discussion
The structural parameters of the auxetic warp knit fabrics are shown in Table 2. Details of the experimental measurements of unit cell geometry and Poisson's ratio are presented in Table 3. Also, the values of the Poisson's ratio at different strain levels are presented in Figure 4. The measurement of Poisson's ratio is an important fundamental tool for determining the auxetic property of materials. If the Poisson's ratio is negative then the structure is auxetic. Ugbolue et al. [4] have shown that the factor which influences the Poisson's ratio is identified as intrinsic unit size displacement, which depends on chain course numbers. It is surmised that the Poisson's ratio values decrease as the number of tricot courses increase. Indeed, the auxetic properties depend on the interaction of vertical and horizontal ribs in the knitted structure as shown in the model. Figure 4, the Poison's ratio of warp knit structures is increasing with increasing number of tricot and chain courses per interlooping repeat. However, such tendency is observed at low strain levels, up to 10% of fabric stretching. At 30% fabric extension, only the knit structure with 7 tricot courses as its interlooping repeat has negative Poisson's ratio whose value is shown to decrease with increasing number of chain courses at repeat. The observed low Poisson's ratio effect is attributed to the poor prior (pre-knitting) stretching of the elastomeric in-laid yarn.
Conclusion
A geometrical analysis of warp knit auxetic fabrics has been Poisson's ratios primarily because of the low pre-tensioning of the inlay spandex yarns employed. On the basis of these results, future fabrication of auxetic warp knit structures must employ higher pretensioned 100% elastomeric inlay yarn in order to improve on the auxetic properties and characteristics of the warp knit auxetic fabrics. | 2018-10-17T16:41:52.422Z | 2015-07-15T00:00:00.000 | {
"year": 2015,
"sha1": "e04f05c1dd7afe14cea9c5de7a974a7af7010c79",
"oa_license": "CCBY",
"oa_url": "https://www.omicsonline.org/open-access/geometrical-analysis-of-warp-knit-auxetic-fabrics-2165-8064-1000201.pdf",
"oa_status": "HYBRID",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "e04f05c1dd7afe14cea9c5de7a974a7af7010c79",
"s2fieldsofstudy": [
"Geology"
],
"extfieldsofstudy": [
"Materials Science"
]
} |
106112406 | pes2o/s2orc | v3-fos-license | Reduction of hospital wastewater through ozone-air micro-nanobubbles
The reduction of the organic matter concentration of hospital wastewater is expressed in BOD5 and COD. The treatment was developed using a micro-nanobubble generator with continuous system. Micro-nanobubbles of 0.024μm average diameter were obtained. For the treatment, two samples were taken; Sample 1 (at 8:00 a.m.) and show 2 (at 11:00 a.m.). After applying the treatment with ozone-air micro-nanobubbles, it was possible to reduce the initial concentrations of sample 1: BOD5 from 132mg/L to 14mg/L, COD from 374mg/L to 30mg/L and from sample 2: BOD5 from 127mg/L to 21mg/L and COD from 297mg/L to 36mg/L. The efficiency obtained at 15 minutes of sample 1 of BOD5 was 90.2% and of COD was 92.51% and of sample 2 of BOD5 was 83.5% and of COD was 87.9%.
Introduction
In Peru, only 32% of wastewater sent to the sewerage system receives some type of treatment, and several treatment plants are added, which are not adequate.(OEFA, 2014).The discharge of effluents without treatment is one of the main causes of the loss of water quality.Hospital effluents represent a particular problem, encompassing a mixture of medicines, detergents, antiseptics, solvents and organic matter, to which are added excreta and secretions from patients contaminated by different pathogens.It is also known that large quantities of disinfectant solutions are used in hospitals for the elimination of pathogens on surfaces (floors, walls, etc.), equipment and biomedical instruments and on the skin.Having alcohol, aldehydes and different chlorinated compounds as active ingredients of the solutions mentioned (Núñez, 2006).The use of ozone (dose of 187mg O3/h) in hospital wastewater at pH = 10, increases biodegradability by 70% and acute toxicity is reduced by 62% (Grisales et al., 2012).The Micro-Nano bubbles are sub-micronic cavities containing gas in aqueous solution.The Microbubbles (MBs) have a diameter of more than 100 μm, the micro-nano bubbles (MNBs) have a diameter between 1 to 100 μm and the nano-bubbles (NBs) have a diameter less than 1 μm inside the fluid.The micro-nanobubbles generation technology in water is applied in: sea water, water bodies, groundwater, domestic wastewater and industrial wastewater (Valverde, 2017).There are reductions of thermotolelant coliforms presents in marine water from 1400 CFU /100 mL, until 56 CFU/100 mL, (96%) after applying the air-ozone micro-nanobubbles (Abate and Valverde, 2017).The best treatment reduction Efficiency of BOD in river's water was applying ozone micro-nanobubbles (Salguero and Valverde, 2017).The chemical oxygen demand of 0.5 g Amoxicillin per Liter of water were reduced until 76.9%, organic matter until 65.8% and turbidity until 19.35% in 45 minutes applying air micro-nanobubbles (Mendez and Valverde, 2017).Total coliforms presents in domestic wastewater were reduced from 14500 CFU /100 mL, until 4900 CFU / 100 mL.applying air micro-nanobubbles (Reyes and Valverde, 2017).Sanguaza treated applying air micro-nanobubbles in 60 minutes achieved an average turbidity of 81.5 NTU (79.61%), average BOD5 of 134 mg/L (69.65%) and COD of 150.43 mg/L.(69.24%) (Ventura and Valverde, 2017).
Materials and Methods
A sample of 40 liters of hospital wastewater was used to develop the experimental process and 4L of residual water for the initial analysis.For the development of this research, four stages were considered:
Collection and initial analysis of the residual water sample
In this stage, 40 liters of hospital wastewater were collected.Sample 1 was taken at 8:00 a.m. when there was little activity and sample 2 was taken at 11:00 a.m. when the activities increased.Subsequently, the field parameters (temperature, pH, electrical conductivity and Turbidity) and physical-chemical parameters (TSS, BOD5, and COD) were determined.
Treatment with Ozone-air MNBs
In this stage, the continuous-system ozone-air micro-nanobubble generator was conditioned
A
To treat both sample 1 and sample 2, time intervals of 5, 10 and 15 minutes were used.
Analysis after treatment
The field parameters (T, pH, electrical conductivity and Turbidity) and the physical-chemical parameters (BOD5, COD and TSS) of the samples were analyzed.
Treatment efficiency on parameters.
To measure the MNBs treatment's efficiency on EC will be used the equation 1: Also this equation is used to measure treatment's efficiency on: Turbidity, TSS, BOD5 and COD.
Collection of the residual water sample
40 Liters of hospital wastewater were collected at the main drainage point, which were filled in plastic containers for later treatment.Initial samples required 1 L of wastewater for BOD5, 0.5 L for TSS and 0.25 L for COD preserved with 10 drops of H2SO4.
Initial analysis of the sample
For the initial analysis of the field parameters, the HQ40d multiparameter and the OAKTON T-100 turbidimeter were used.The results are shown below.The initial analysis of the field parameters for sample 1 was carried out: temperature (21.4 °C), pH (7.48), electrical conductivity (743 μS/cm) and turbidity (44.9 NTU) of the wastewater.In addition, the initial analysis was carried out for sample 2: temperature (21.5 °C), pH (7.65), electrical conductivity (775 μS/cm) and turbidity (79.5 NTU) of waste water.The initial analysis of the physical-chemical parameters for sample 1 was carried out: BOD5 (132 mg/L), COD (374 mg/L) and TSS (43 mg/L).In addition, the initial analysis of the physicalchemical parameters for sample 2 was carried out: BOD5 (127 mg/L), COD (297 mg/L) and TSS (110 mg/L) of waste water.
Treatment with Ozone-air MNBs
The micro-nano bubbles were generated at an air pressure of 30 PSI, water flow of 8L/min and 1000mg of O3/h.
The micro-nanobubbles were characterized through a trinocular microscope MOD BM-120T-LED light with a 5 MP camera obtaining as a result micro-nano bubbles of 0.024 μm average diameter.
Internal pressure
The internal pressure was determined from the Young's equation -Laplace Therefore, it is concluded that the internal pressure of the micro-nanobubbles is 22.01 atm.
Ascent speed pf air micro-nanobubbles.
Ascent speed of the micro-nanobubble was obtained by replacing equation (3).
V = pgd 2 /18 ᶯ (3) Where: V: ascending speed = X p: density of the liquid = 998.2g/L = 998.2kg/m 3 g: gravitational acceleration = 9.8 m/s 2 d: diameter of the bubble= 0.024x10 -6 m n: viscosity of the liquid = 1.003 x 10 -3 m 2 /s 18( 1.003x10 −3 ) = 3.12 x 10 −10 / Therefore, it is concluded that the ascent speed of the air micro-nanobubbles is 3.12 x 10 -10 m/s.Both for the treatment of sample 1 (8:00 a.m.) and sample 2 (11:00 a.m.), 12 L of residual water were used in a period of 5, 10 and 15 minutes.The results obtained from the analysis of the field parameters of sample 1 (8:00 a.m.) and sample 2 (11:00 a.m.) at 0, 5, 10 and 15 minutes with the ozone-air micro-nanobubbles are observed.The results obtained from the analysis of the physical-chemical parameters of sample 1 (8:00 a.m.) and sample 2 (11:00 a.m.) at 0, 5, 10 and 15 minutes with the ozone-air micro-nanobubbles are observed.The results obtained before and after the treatment of the field and physicalchemical parameters are detailed below.It is observed that Turbidity was decreasing with longer treatment time.Turbidity in sample 1 achieved 4 NTU, and in sample 2 achieved 5.9 NTU in 15 minutes.It is observed that BOD5 was decreasing with longer treatment time.BOD5 in sample 1 achieved 13 mg/L, and in sample 2 achieved 21 mg/L in 15 minutes.It is observed that COD was decreasing with longer treatment time.COD in sample 1 achieved 28 mg/L, and in sample 2 achieved 36 mg/L in 15 minutes.
Treatment's efficiency
To calculate treatment's efficiency with ozone-air MNBs on Electrical conductivity in Sample 1 as % Remotion was used the equation 3: Then was calculated to the rest of Sample 1 and Sample 2. As a resume the efficiency is seen in figure 9. To calculate treatment's efficiency with ozone-air MNBs on Turbidity in Sample 1 as % Remotion was used the equation 1: % remotion (Sample1)15 = (44.9-4)*100/ 44.9 = 91.1 % Then was calculated to the rest of Sample 1 and Sample 2. As a resume the efficiency is seen in figure 10.Then was calculated to the rest of Sample 1 and Sample 2. As a resume the efficiency is seen in figure 11.
Figure 11.TSS removal efficiency of sample 1 and sample 2.
To calculate treatment's efficiency with ozone-air MNBs on BOD5 in Sample 1 as % Remotion was used the equation 1: % remotion (Sample1)15 = (132 -13)*100/ 132 = 90.2% Then was calculated to the rest of Sample 1 and Sample 2. As a resume the efficiency is seen in figure 12.To calculate treatment's efficiency with ozone-air MNBs on COD in Sample 1 as % Remotion was used the equation 1: % remotion (Sample1)15 = (374 -28)*100/ 374 = 92.51% Then was calculated to the rest of Sample 1 and Sample 2. As a resume the efficiency is seen in figure 13.
Figure 13.Efficiency of COD removal from sample 1 and sample 2
Conclusions
• It is concluded that the initial concentrations of the field parameters of the wastewater in sample 1 as: EC (743 μS/cm), and Turbidity (44.9 NTU) applying ozone-air micronanobubbles in 15 minutes decreased their concentrations at 405 μS/cm EC, and 4 NTU Turbidity.In addition, the initial concentrations of the field parameters of the wastewater in sample 2 as: EC (775 μS/cm), and Turbidity (79.5 NTU) applying ozone-air micronanobubbles in 15 minutes decreased their concentrations to 530 μS/cm of EC, and 5.9 NTU of Turbidity.
• It is concluded that the initial concentrations of the physical-chemical parameters of the wastewater in sample 1 such as: BOD5 (132mg/L), COD (374mg/L) and TSS (43mg/L) applying ozone-air micro-nanobubbles in 15 minutes decreased their concentrations to 13mg/L of BOD5, 28mg/L of COD and 9 mg/L of TSS.In addition, the initial concentrations of the parameters of wastewater in sample 2 such as: BOD5 (127mg/L), COD (297mg/L) and TSS (110mg/L) applying ozone-air micro-nanobubbles in 15 minutes decreased its concentrations at 21mg/L of BOD5, 36mg/L of COD and 13 mg/L of TSS.
• According to the results obtained, the percentage of the removal efficiency of the electrical conductivity was 45.5% for sample 1, turbidity was 92.6 for sample 2, Total Suspended Solids (TSS) was 88.2% for sample 2, BOD5 was 90.2% for sample 1 and COD was 92.51% for sample 1.
Figure 8 .
Figure 8. COD vs Time of Sample 1 and Sample 2.
Figure 9 .
Figure 9. Efficiency of removal of the electrical conductivity of sample 1 and sample 2
Figure 10 .
Figure 10.Efficiency of turbidity removal of sample 1 and sample 2
Figure 12 .
Figure 12.Efficiency of removal of BOD5 from sample 1 and sample 2
Table 1 .
Results of the initial measurement of the field parameters.
Table 2 .
Results of the initial measurement of the physical-chemical parameters
Table 3 .
Results of the characterization of the micro-nano bubbles
Table 4 .
Results of the field parameters of Sample 1 and sample 2.
Table 5 .
Results of the physical-chemical parameters of Sample 1 and sample 2. | 2018-12-05T17:19:17.215Z | 2018-01-13T00:00:00.000 | {
"year": 2018,
"sha1": "8b75c92300e831b0ae95ae6abe359f18016528a1",
"oa_license": "CCBY",
"oa_url": "http://journals.cincader.org/index.php/nanoj/article/download/40/41",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "8b75c92300e831b0ae95ae6abe359f18016528a1",
"s2fieldsofstudy": [
"Chemistry",
"Engineering"
],
"extfieldsofstudy": [
"Materials Science"
]
} |
18875896 | pes2o/s2orc | v3-fos-license | On the support of the Ashtekar-Lewandowski measure
We show that the Ashtekar-Isham extension of the classical configuration space of Yang-Mills theories (i.e. the moduli space of connections) is (topologically and measure-theoretically) the projective limit of a family of finite dimensional spaces associated with arbitrary finite lattices. These results are then used to prove that the classical configuration space is contained in a zero measure subset of this extension with respect to the diffeomorphism invariant Ashtekar-Lewandowski measure. Much as in scalar field theory, this implies that states in the quantum theory associated with this measure can be realized as functions on the ``extended"configuration space.
Introduction
The usual canonical approach to quantization of a (finite dimensional) system defines states as functions on a configuration space and defines an inner product of two such functions ψ and φ through where µ is some measure on the configuration space Q. Naively applying this procedure to Yang-Mills theories produces a "connection representation" with states that are functions of the Yang-Mills connection. In particular, these states are functions on the quotient space A/G, where A is the space of (C 1 -)connections and G is the group of (C 2 -)gauge transformations. The same is true for gravity formulated in terms of Ashtekar variables before one imposes the diffeomorphism and hamiltonian constraints [1,2].
A more sophisticated analysis of examples, such as scalar field theory [3][4][5], shows that the domain space of the wave functions may not be exactly the classical configuration space. Instead, some extension of Q is required.
In order to define an inner product for a connection representation, one expects to give A/G, or some suitable extension, the structure of a measurable space (by choosing the measurable sets) and to define appropriate measures. Ashtekar and Isham described an algebraic program to construct such measures in [2]. They proposed, for a compact gauge group G, a compact extension A/G of A/G on which regular Borel measures are well defined and are in one-to-one correspondence with positive continuous linear functionals on a certain C * -algebra of connection observables known as the holonomy algebra HA. In [6], Ashtekar and Lewandowski constructed such a Borel measure µ AL on A/G that is both diffeomorphism invariant and strictly positive on continuous cylindrical functions. To do so they, and independently Baez in [7], introduced the concepts of "cylindrical sets" and "cylindrical functions" on A/G. Baez then generalized the Ashtekar-Lewandowski measure by finding an infinite dimensional space of diffeomorphism invariant measures. In [6] it was also shown that the Ashtekar-Isham space A/G is in one-to-one correspondence with the set of homomorphisms from the group of piecewise analytic hoops (i.e. based loops modulo an equivalence relation defined by the holonomies) HG x0 to the gauge group K, modulo conjugation.
In what follows, we reinterpret some of the results of [2,6] in terms of the theory of projective limits. In particular, we consider projective limits of infinite families of finite dimensional topological and measurable spaces associated with arbitrary finite lattices. This theory provides an appropriate framework for studying different properties of A/G, both from the topological and measure theoretical points of view. Our main result is the use of this formalism to prove that the space A/G is contained in a zero measure subset of A/G (with respect to the Ashtekar-Lewandowski measure).
The present work is organized as follows. In Sect. 2 we recall (mainly from [8]) some aspects of the theory of projective limits of infinite families of measurable spaces. Sect. 3 is devoted to reinterpreting some results of [2,6] in the language of projective limits. In particular we show that A/G is a projective limit of a family of finite-dimensional spaces and that the Gel'fand topology on the spectrum A/G coincides with the Tychonov topology on the projective limit. While the Ashtekar-Isham space A/G is defined only for compact gauge groups G, the projective limit is defined for the noncompact case as well. On the measure theoretical side we show that the measurable space (A/G, B(C)) (where B(C) denotes the minimal σ-algebra containing the cylindrical sets C) is isomorphic to the projective limit. In Sect. 4 we prove the main result of the paper stated above. Sect. 5 is devoted to the study of the additive, but not σ-additive, measureμ AL induced by µ AL on the (finite) algebra C of cylindrical sets of A/G C = {C ∩ A/G ;C ⊂ C} where C denotes the algebra of cylindrical sets on A/G. We show thatμ AL cannot be extended to a σadditive measure on A/G and that the space of square integrable (cylindrical) functions on A/G is not complete. We also prove that the Cauchy completion of this space is L 2 A/G, µ AL , B(C) , justifying the use of the "generalized connections" in A/G.
Projective limit measurable spaces
In the present section we recall, mainly from [8], the relevant aspects of a class of measures on infinite dimensional spaces which are obtained as rigorously defined limits of measures on finite dimensional spaces. This class contains the direct product measures (on IR ∞ for example) and the projective limit measures. First, however, we introduce some more terminology and notation that will prove useful.
The pair (X, B) (or (X, F )), where X is a set and B (F ) is a σ-algebra (algebra) of subsets of X, will be called a σ-measurable (measurable) space. In the mathematical literature, definitions of a measurable space have been given both that require B to be a σ-algebra and that require only that B be closed under finite operations. As we will be interested in a comparison of these two cases it will be convenient to use the above terminology to distinguish between them.
We will be interested in σ − additive probability measures on B, which are, by definition non-negative, normalized and σ-additive functions on the σ-algebra B. That is, such a measure µ satisfies: Additive measures on an algebra F satisfy (2.1) with B replaced by F and with only finite unions and sums in (2.1c). For a given measure µ on F , an important question is whether or not it can be extended to a σ-additive measure on B(F ), the minimal σ-algebra that contains F . A necessary and sufficient condition for extendibility is given by the Hopf theorem [8]: A measure µ on F can be extended to a σ-additive measure on B(F ) if and only if for every decreasing Essentially, the condition (2.2) allows an extensionμ to be consistently defined on elements of B(F ) as limits of µ-measures of sets in F . The triplet {(X, B), µ} ({(X, F ), µ}), where B (F ) is a σ-algebra (algebra) and µ is σ-additive (additive) is called a σ-measure (measure) space. The possibility of extending a measure µ on F to a σ-additive measureμ on B(F ) is in particular relevant to physical applications in quantum mechanics. Recall that quantum mechanical systems are often defined by first giving a linear pre-Hilbert space and then completing this space with respect to an inner product. In general, if µ is cylindrical but not σ-additive, the space H of µ-square integrable cylindrical functions on X (denoted through CL 2 (X, F , µ)) is only a pre-Hilbert space. Such spaces will be discussed in section 5. However, if µ is extendible to a σ-additive measureμ on (X, B(F )) then the Cauchy completion of H leads to the spaceH = L 2 (X, B(F ),μ) (see section 5). On the other hand if µ is not extendible then the Cauchy completion of CL 2 (X, F , µ) leads in general to a space with state-vectors which cannot be expressed as functions on the initial space X. This is the case in scalar field theory if one considers X = S(IR 3 ) (the Schwarz space of rapidly decreasing smooth C ∞ functions on IR 3 ) and µ is a cylindrical measure defined with the help of a positive definite function on S(IR 3 ), continuous in the nuclear space topology (see [3,5,8]). As we shall see in Sect. 5 this is also the case in Yang-Mills theory if we take H = CL 2 (A/G, F = C,μ AL ), whereμ AL is the Ashtekar-Lewandowski measure on A/G. In the scalar field case the Cauchy completion of CL 2 (S(IR 3 ), F , µ) gives the space of square integrable functions on S ′ (IR 3 ) (the space of tempered distributions), while in the Yang-Mills case the completion of CL 2 (A/G, C,μ AL ) gives the space L 2 (A/G, B(C), µ AL ) of square integrable functions on the Ashtekar-Isham space A/G of generalized "distributional" connections modulo gauge transformations. Let so that if we are concerned only with such spaces we can restrict ourselves to Y and µ Y . This is particularly convenient when the set Y has advantages (for instance from the "differentiable" point of view) over X. When a set Y is µ-thick in X we say that the support of the measure µ is contained in Y . An illustrative example is the one given by the Wiener measure on IR [0,1] used in the (euclidean) path integral formulation of quantum mechanics. In this case the support of the measure is contained in the space Y = C 0 ([0, 1]) of continuous functions on the interval [3]. The inclusion map from Y to X above is referred to as measurable. In general, a map between σmeasurable (measurable) spaces φ : is called an isomorphism of σ-measurable (measurable) spaces if it is bijective and if both φ and φ −1 are measurable. Let us now briefly review (see [8]) the construction of infinite products of σ-measurable spaces and of (projective) limits of infinite projective families of σ-measurable spaces. Let be an indexed family of σ-measurable spaces. The product σ-measurable space (X (Λ) , B (Λ) ) is, by definition, given by with B (Λ) being the minimal σ-algebra for which all the projections are measurable. That is, B (Λ) is the σ-algebra generated by the inverse images of measurable sets in X (λ) under the projections p λ . If all the X (λ) , λ ∈ Λ are different copies of the same set Y with the same σ-algebras B (λ) = B then the points of X (Λ) = Y Λ , x ∈ Y Λ are (arbitrary) maps from Λ to Y : Examples are the set of all sequences of real numbers and the set of all real valued functions on the interval [0, 1] Suppose we have a σ-additive measure µ on X (Λ) in (2.6). Let L be the family of all finite subsets of Λ and for L ∈ L let (X (L) , B (L) ) be the partial products of σ-measurable spaces with (2.11) and the corresponding B (L) . Then all the projections for B ∈ B L , which, in the notation of measure theory is written as This family satisfies a self consistency condition: where p LL ′ denote the measurable projections from X (L ′ ) to X (L) . In [8] (see corollary to Th. 10.1) are found conditions for which the converse is also true: For a family of σ-compact or complete and separable, metric spaces every family of Borel measures that is consistent in the sense of (2.14) can be extended to a σ-additive measure on the product σ-measurable space.
Such a measure is in fact defined by (2.13), i.e. for B L ∈ B L , µ(p −1 L (B L )) is defined to be just µ L (B L ). Recall that a topological space (X, τ ) is said to be σ-compact if X can be represented as a countable union of compact sets.
Notice ( [8]) that a measure µ satisfying (2.13) and given on the algebra always exists. The only question is whether µ can be extended to a σ-additive measureμ on B (L) , which is the minimal σ-algebra that contains (2.15) defines a σ-additive measure on the infinite dimensional space Quite remarkably, in this example and in many others relevant to quantum field theory the answer is affirmative, as indicated by Proposition 2.
2. An infinite product σ-measure space can also be realized as a "projective limit" (which we will define next). However, the product space X (Λ) is a projective limit not of the family of spaces X (λ) labelled by λ ∈ Λ but rather of the family of spaces X L = X (L) labelled by L ∈ L, the set of all finite subsets of Λ. In general, a projective limit space can be defined for any "projective family" of σ-measurable spaces; that is, for any family of the following form. The set L is taken to be directed, i.e. partially ordered and such that for any two elements L 1 , L 2 ∈ L, there is some L such that L 1 ≤ L and L 2 ≤ L. We will also assume that L does not have a maximum. Here p LL ′ are measurable projections, i.e. surjective mappings Then the projective limit of the family (2.19) is by definition the σ-measurable space (X L , B L ) , where That is, X L is the subset of X (L) that is consistent with the projections p LL ′ . Note that a direct product space can also be thought of as a projective limit of the spaces formed by taking arbitrary finite products of the factors. A family of measures (µ L ) L∈L is said to be self-consistent if it satisfies (2.14) with L ⊂ L ′ replaced by L < L ′ . A measure µ on B L always defines a self consistent family of measures (µ L ) L∈L through (2.13) and a consistent family (µ L ) L∈L defines a finitely additive measure on X L through (2.13) as well. A measure on X L defined by such a family is called cylindrical. An important result is (see [8] Corollary to Th. 10.1):
Proposition 2.3
Under the same conditions as in Proposition 2.2, a self-consistent family of Borel measures on a projective family (2.19) defines a cylindrical measure that can be extended to a σ-additive measure in the projective limit σ-measurable space (2.21) if for every increasing sequence
Ashtekar-Isham space A/G as a projective limit
Let A/G denote the space A of smooth C 1 G-connections modulo the group G of gauge transformations on a three dimensional analytic manifold Σ where, as in [6], the gauge group G is assumed to be U (N ) or SU (N ). Following [6], we consider the G-hoop group HG x0 = LΣ x0 / ∼ where LΣ x0 is the space of piecewise analytic loops based at x 0 (see [6]) and the equivalence relation ∼ is Here, H(α, A) denotes the holonomy corresponding to the connection A and the loop α. The Ashtekar-Isham space A/G is a "compactification" of A/G obtained as follows (see [2]). Let T α , α ∈ LΣ x0 denote the Wilson loop function on A/G defined by where [α] denotes the equivalence class of α in HG x0 , [A] denotes the equivalence class of A in A/G and the trace is taken in the fundamental representation of the gauge group. In the following, for simplicity, α, β will denote hoops. The holonomy algebra HA is the commutative C * -algebra generated by the Wilson loop functions. The Ashtekar-Isham space A/G is the compact Hausdorff space that is the spectrum [2] of HA in which A/G is densely embedded [2,6,8]. Ashtekar and Lewandowski [6] obtained a useful algebraic characterization of the space A/G. They proved that there is a one-to-one correspondence between A/G and the space of all homomorphisms from the hoop group HG x0 to the gauge group G, modulo conjugation. We will therefore identify these two sets and writeh where g above does not depend on the hoop α. Notice that no continuity condition has been imposed on the homomorphisms h in (3.3). This will allow us to interpret A/G (both topologically and measure theoretically) as a projective limit of finite dimensional spaces. Let L denote the set of all subgroups of HG x0 generated by a finite number of hoops β 1 , . . . , β n that are strongly independent in the sense of [6], i.e. such that loop representatives of the hoop equivalence classes β i can be chosen in such a way that each contains an open segment which is traced exactly once and which intersects any of the other representative loops at most at a finite number of points. Then and we write S * = S * [β 1 , . . . , β n ]. Now let H S * = Hom(S * , G)/Ad be the set of equivalence classes of homomorphisms from S * to G under conjugation. If S * = S * [β 1 , . . . , β n ] then, as shown in [6], a homomorphism from S * to G is known if and only if we know it on the hoops β 1 , . . . , β n so that we have the one-to-one correspondence Consider now the following projective family of finite dimensional spaces and h S * ′ | S * denotes the restriction of h S * ′ to the subgroup S * of S * ′ . From [6] we see that these projections are surjective. According to (2.21) the projective limit H L of the family (3.6) is given by (3.8) We will now show that this is just the Ashtekar-Isham space A/G.
Proposition 3.1
There is a bijective map φ
Proof
Consider the space Hom(HG x0 , G) of all homomorphisms from HG x0 to G and the projective family to Hom(S * , G). Let K (L) be the infinite product space and K L be the projective limit space of this family (3.10) We will need the following lemmas.
Proof of Lemma 3.2 The injectivity ofφ is trivial. Let us prove thatφ is surjective. Fix an arbitrary element (h 0 S * ) S * ∈L ∈ K L . Let us construct the homomorphism h 0 which is the pre-image of this element. Let α be an arbitrary hoop and S * 1 ∈ L such that α ∈ S * 1 (S * 1 always exists for a piecewise analytic hoop α [2]). Then choose h 0 (α) = h 0 S * 1 (α). To see that h 0 (α) does not depend on the choice of the finitely generated group S * 1 ∋ α let α ∈S * 1 and S * 2 be a subgroup which contains both S * 1 andS * 1 . Then, according to the definition of K L , we have h 0 We can easily show that h 0 constructed in this way is an homomorphism and that the mapφ is equivariant The map is bijective.
Proof of Lemma 3.3
We will first show that φ 2 is surjective. To do so, recall that any element of H L is a family ([h S * ]) S * of consistent equivalence classes in the sense of (3.7b). Now, choose a representative h 0 S * from each [h S * ] and construct the subgroup C 0 S * of G that commutes with h 0 S * ; that is, let Note that C 0 S * is closed in G. Any closed subgroup of a Lie group is a Lie group and any closed subset of a compact space is compact, so that C 0 S * is again a compact Lie group. Thus, C 0 S * has some dimension d S * ≥ 0 and, by compactness, some finite number m S * ≥ 1 of connected components. There is then some least value d 0 of d S * (d 0 = min S * ∈L d S * ) and some m 0 that is the least value of m S * for which the and construct the corresponding C 1 S * : are of the same dimension. It follows that they agree in some neighborhood of the identity and thus on the entire component connected to the identity. Since C 0 . This means that h 1 S * is unique, since any g that commutes with h 0 S * and commutes with h 1 S * (α) for all α ∈ S * . Thus, no other representative of [h S * ] satisfies (3.15). It now follows that for any S * ′ ⊃ S * ⊃ S * 0 , Finally, for any S * that does not contain S * 0 , let S * ′ be any subgroup of HG x0 generated by a finite number of independent hoops that contains S * and S * 0 (we see from [6] that such a group exists) and let h 1 Then the representatives (h 1 S * ) S * ∈L ∈ ([h S * ]) S * ∈L form a consistent family of homomorphisms in K (L) and the equivalence class of this family under the adjoint action is a member of K L /Ad that maps to ([h S * ]) S * ∈L under the map φ 2 . We conclude that φ 2 is surjective. Now, injectivity of φ 2 follows in a straightforward fashion. Consider any other equivalence class of families [(h ′ S * ) S * ∈L ] ∈ K L /Ad that maps to the family ([h S * ]) S * ∈L chosen above under φ 2 . As with the family constructed above, h 1 We have just seen that (h 1 S * 0 ) S * ∈L is the unique self-consistent family of homomorphisms that includes h 1 S * and satisfies [h 1 showing that φ 2 is also injective.
Q.E.D.
We complete the proof of the proposition by noticing that the bijective map φ is given by Endowed with the natural topology, the spaces H S * are compact topological spaces (see (3.6)). The Tychonov topology τ T on the product space H (L) is the minimal topology for which all the projections where the last convergence is with respect to the topology on H S * = G n /Ad. In this topology, the space H (L) is compact (see [8,Tychonov theorem]). Let us also refer to the topology induced on the projective limit H L ⊂ H (L) from H (L) as the Tychonov topology τ T . Then from the continuity of the projections p S * S * ′ , H L is closed in H (L) and therefore is also a compact topological space. Since H L is compact in the Tychonov topology and A/G is compact in the Gel'fand topology τ Gd it is natural to expect that the bijective map φ in (3.9) is actually a homeomorphism. Indeed we have Proposition 3.4 The bijective map in (3.9) is a homeomorphism where τ Gd and τ T denote the Gel'fand and Tychonov topologies respectively.
Proof
First let us obtain a more convenient characterization of the topology on the spaces H S * . As mentioned above, H S * endowed with the standard topology induced from G n is a compact Hausdorff space. Consider on H S * the continuous functions They separate the points in H S * for the same reason that the T α , α ∈ HG x0 , separate the points in A/G [2,6]. Therefore, according to the Stone-Weierstrass theorem [9] the algebra HA S * obtained by taking finite linear combinations (with complex coefficients) and products of T S * α is dense in the C * -algebra C(H S * ) of all continuous functions on H S * i.e. (3.26) Using the first Gel'fand-Naimark theorem [2,9,10] we then conclude that the spectrum of HA S * , endowed with the Gel'fand topology (see below) is homeomorphic to H S * . An equivalent description of the initial topology in H S * is therefore given by the Gel'fand topology, which is, by definition, the weakest for which all the functions T S * α , α ∈ S * are continuous. Returning to (3.24) we see that, in accordance with (3.21), the Tychonov topology on H L is the weakest for which all the functions T S * α • π S * : H L → C α ∈ S * , S * ∈ L are continuous. On the other hand the Gel'fand topology on A/G is the weakest for which all the functions T α , α ∈ HG x0 are continuous. Since for all α ∈ HG x0 we conclude that φ in (3.9) is a homeomorphism.
Q.E.D.
We now proceed to derive a measure theoretic analog of proposition 3.4. Let B S * denote the Borel σ-algebra on H S * so that, since the projections p S * S * ′ are measurable, is a projective family of σ-measurable spaces (see (2.19)). Let denote the projective limit σ-measurable space. In A/G we take the measurable sets to be generated by the class C of "cylindrical sets" used in [6,7], i.e. the inverse images C B of Borel sets B in G n /Ad with respect to π S * • φ where, as in (3.5), we have identified H S * with G n /Ad with the help of the independent hoops β 1 , . . . , β n ∈ S * .
Note that the complement of a cylindrical set is cylindrical, as are finite unions and intersections of cylindrical sets so that C is in fact a (finite) algebra. Denoting the minimal σ-algebra algebra containing the cylindrical sets by B(C), the space (A/G, B(C)) (3.31) becomes a σ-measurable space. From the definition of B L and B(C), we see that Proposition 3.5 The map (3.9) is an isomorphism of σ-measurable spaces.
wheref is a measurable function on H S * . The Wilson loop functions ) are continuous cylindrical functions ( [6]). (iii) The projective limit H L provides a generalization of the Ashtekar-Isham space A/G to the case where the gauge group G is not compact. (iv) There is a one-to-one correspondence between cylindrical measures µ on C (i.e. additive on C but σadditive on the σ-subalgebras (π S * • φ) −1 (B S * )) and families of measures {(µ S * ) S * ∈L } (µ S * are Borel measures on the finite dimensional spaces H S * ) satisfying the self-consistency condition The correspondence is given by Recall [11] that a Borel measure µ is called regular if for every Borel set E
Proposition 3.7
There is a one-to-one correspondence between regular Borel measures µ on A/G and self-consistent families of measures {(µ S * ) S * ∈L }.
This functional is bounded with respect to the sup-norm Since the space of continuous cylindrical functions is dense in the C * -algebra C(A/G) of all continuous functions on A/G (see [6]) the functional Γ µ0 can be extended in a unique way to a continuous positive (norm 1) functional on C(A/G) (see [9]). But in accordance with the Riesz representation theorem (see [11]) there is then a unique regular Borel measure µ on A/G such that for every f ∈ C(A/G), where we denoted the extension of Γ µ0 to C(A/G) with the same letter. Regular Borel measures are completely determined if the integral of continuous functions is known (see [11], p.41), which implies that µ and µ 0 coincide on C. Therefore µ is the unique (see [12]) extension of µ 0 to B(C) and (as we have showed) the unique regular extension to a Borel measure. Q.E.D.
4.
A/G is contained in a zero measure subset of A/G.
The present section contains the main result of this paper. For simplicity we will use (3.32) to identify the σ-measurable spaces (A/G, B(C)) and (H L , B L ) so that we will consider A/G to be the projective limit of the projective family of finite dimensional spaces (3.6).
In [6] Ashtekar and Lewandowski introduced the following measure µ AL on (A/G, B(C)). Let µ H be the normalized Haar measure on G and µ H n and µ H S * the corresponding measures on G n /Ad and H S * (µ H S * is obtained from µ H n using (3.5)). Then the (uncountable) family (µ H S * ) S * ∈L satisfies the self-consistency conditions (2.20). The Ashtekar-Lewandowski measure µ AL is the corresponding (unique) measure on (A/G, B(C)) satisfying The measure µ AL is σ-additive, Dif f (Σ)-invariant, and strictly positive as a functional on the space continuous cylindrical functions on A/G (see [6]). The space A/G is canonically embedded in A/G [2] and is topologically dense there [6,10]. It is interesting to find out whether A/G is also µ AL -thick in A/G; that is, whether A/G supports the measure µ AL . We will in fact prove that this is far from being the case: There exists a measurable set such that µ AL (Z) = 0 (4.2b) and A/G ⊂ Z . (4.2c)
Proof
We need the following lemma
Proof of Lemma
The complement Q (q) c of Q (q) will be constructed essentially (i.e. modulo dividing by Ad) by taking an infinite product of sets consisting of copies of G with holes cut out around the identity such that the "diameter" of the holes decreases to zero. These copies of G are chosen to correspond to a certain "convergent" sequence of hoops. In order to do this explicitly, choose r 0 such that the exponential map is one-to-one in the subset U r0 (0) of Lie(G) where and denote by the same letter d e the following extension to the whole group G: The Ad-invariance of · on Lie(G) implies that d e (· ) is Ad-invariant on G. Consider now the basic sets The function given by is continuous, monotonically decreasing and s(0) = 1. Now let ∆ {ǫi} n i=1 n be the subset of G n given by Clearly we have Notice that the set ∆ {ǫi} n is an Ad-invariant subset of G n . It is the inverse image of the set under the quotient map π : G n → G n /Ad. By the definition of the measure µ H n on G n /Ad we thus have Now, for each q ∈ (0, 1] choose a sequence be an arbitrary sequence of independent hoops. Then the sets were we used (3.5) to identify H S * and G n /Ad, form a decreasing sequence such that we conclude from (4.13) and the σ-additivity of µ AL that Let us now turn to the second part of the lemma namely the choice of Q (q) satisfying (4.3) and (4.4). Take forβ i the hoops corresponding to coordinate squares (all parallel to a fixed coordinate plane) with a corner at x 0 and fix a metric. Chooseβ i to have areas such that is the same as in (4.14) and Then, for every A ∈ A we have (from the smoothness of A) and, since δ n → 0, for n large enough we have Thus, for every We have therefore proved that with our choice (4.18) ofβ i we have Let us now prove the theorem. From (4.4) we conclude that for every q > 0 A/G ⊂ Q (q) ⊂ A/G (4.21a) and µ AL (Q (q) ) = q . (4.21b) Considering now the decreasing sequence Q (1/n) . We have while the σ-additivity of µ AL implies that Q.E.D.
Completion of the space of square integrable functions on A/G
Although A/G is not a projective limit of the family (3.6) a procedure similar to that of (2.14), (2.19)-(2.21) can be used to define a measureμ AL on A/G as was noted in [6]. This is done by returning to the notion of a cylindrical set (3.17) but now in A/G. That is, we introduce (surjective) projectionŝ where again we are identifying G n /Ad with H S * , and take as measurable sets for some B ∈ B S * . Let C be the collection of such cylindrical sets in A/G. Note that C is closed under union, intersection and complementation (i.e. forms an algebra) so that the pair is a measurable space. The measureμ AL is then defined bŷ The additivity of µ H S * for every S * implies additivity ofμ AL . However, the σ-additivity of the µ H S * does not imply σ-additivity ofμ AL . Indeed, we have the following The measure (5.4) on A/G cannot be extended to a σ-additive measure on B(C).
Proof
This theorem follows easily from lemma 4.2. Indeed consider the same sets ∆ n ⊂ G n /Ad as in (4.12)-(4.14) and define analogously to (4.15) the decreasing sequencė Therefore, choosing q : 0 < q < 1 we conclude from the Hopf theorem 2.1 thatμ AL is not extendible to a σ-additive measure on B(C). Q.E.D.
Let us recall aspects of integration theory for the so called (non-σ) measurable spaces with limit structure (see [13] def. 1.5). The measurable space (X, F X ) is said to be a space with limit structure if where for all L ∈ L, B L is a σ-algebra and for every L 1 , L 2 ∈ L, there exists a L 3 such that B L1 ∪ B L2 ⊂ B L3 . If the family {B L } L∈L does not have a maximal element then F X is not a σ-algebra. Obviously every projective limit defined as in (2.19)-(2.21) is a measurable space with limit structure. The converse is also true as we can see by taking as projective family of σ-measurable spaces (see [8] p. 20) Though this makes the class of projective limit spaces equivalent to that of measurable spaces with limit structure the latter is more "natural" for integration theory. In a measurable space with limit structure (X, F X ) the sets F ∈ F X are called cylindrical sets and the map f to a σ-measurable space (Y, B) is called cylindrical if there is a L ∈ L such that is measurable. A measure µ on F X is called a quasi-σ-measure (quasi-measure in [13]) if its restriction µ L = µ | BL to every B L ⊂ F X is σ-additive. The triple {(X, F X ), µ X }, where (X, F X ) is a measurable space with limit structure and µ X is a quasi-measure is called a quasi-measure space. Let L 0 ∈ L be such that the (complex-valued) cylindrical function where B denotes the σ-algebra of the complex plane, is measurable. Then a function f on the quasi-measure space {(X, F X ), µ} is said to be µ-integrable if it is µ L0 integrable in the usual sense (5.10)
Definition 5.2
The set of square-integrable cylindrical functions on the quasi-measure space {(X, F X ), µ} will be denoted through CL 2 (X, F X , µ).
It is easy to see that CL 2 (X, F X , µ) is a pre-Hilbert space with inner product given by where L 0 is such that both f : (X, B L0 ) → (C, B) and g : (X, B L0 ) → (C, B) are measurable.
Proposition 5.3
Suppose that we are given two quasi-measure spaces and that Y ⊂ X. Let χ : F X → F Y be an isomorphism of set algebras given by χ(B) = B ∩ Y for B ∈ F X and such that the restriction to every B L (X) is an isomorphism of σ-algebras B L (X) :
Proof
Note that the map χ : F X → F Y induces a one-to-one correspondence between the sets X Y of characteristic functions of sets in F Y and X X of characteristic functions of sets in F X . Further, since χ is an isomorphism of finite set algebras, this correspondence extends to an isomorphism over the linear spans of X Y and X X . Finally, since χ preserves the measure of sets, this correspondence preserves the inner product in these linear spaces. We need the following lemma.
Lemma 5.4
(i) The completion of X Y (X X ) is equal to the completion of CL 2 (Y, F Y , µ Y ) (CL 2 (X, F X , µ X )).
(ii) The space X X is dense in L 2 (X, B(F X ),μ X ).
Proof of Lemma
(i) Obviously X Y is a subset of CL 2 (Y, F Y , µ Y ). It is sufficient to show that any f ∈ CL 2 (Y, F Y , µ Y ) can be represented as where φ n ∈ X Y and the sequence converges in the norm of CL 2 (Y, F Y , µ Y ). But for f ∈ CL 2 (Y, F Y , µ Y ) there exists a L 0 ∈ L such that f belongs to the (complete) space [12]) f can be represented in the form (5.10).
(ii) For a quasi-measure space {(X, F X ), µ X } satisfying the conditions of proposition 5.2 we have X X ⊂ CL 2 (X, F X , µ X ) ⊂ L 2 (X, B(F X ),μ X ) , (5.13) where clearly all the inclusions are isometric. It will be sufficient to prove that for every set B ∈ B(F X ) its characteristic function χ B is in the L 2 -closure of X X . But this result follows easily from Theorem 3.3 in [8]. Q.E.D.
Proof of Proposition
We have an isometric isomorphism (i.e. one which preserves the inner product) between the spaces X Y and X X , which are dense in X Y = CL 2 (Y, F Y , µ Y ) and L 2 (X, B(F X ),μ X ) respectively. The isomorphism therefore extends to a natural isometric isomorphism η : CL 2 (Y, F Y , µ Y ) → L 2 (X, B(F X ),μ X ) (5.14) Q.E.D.
In the case of A/G since the projectionsπ S * are surjective we havê if and only if there is some S * ⊂ S * 1 ∩ S * 2 and some B ⊂ H S * such that Since the same is true for the algebra C of cylindrical sets in A/G, there is a one-to-one correspondence between C and C given byπ where φ and π S * have been defined in (3.9) and (3.11) respectively. Note that the map χ is an isomorphism of set algebras, and that it preserves measures in the sense that which implies that f is itself a cylindrical function, f =f •π S * 0 for some functionf on some H S * 0 . | 2014-10-01T00:00:00.000Z | 1994-03-18T00:00:00.000 | {
"year": 1994,
"sha1": "c00176e87c664c3f8024602775ea89f8286c1aea",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/hep-th/9403112",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "359fc3f5ccab6f82569098577b77556696aa3a4c",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics",
"Physics"
]
} |
85665628 | pes2o/s2orc | v3-fos-license | The North American ostracod Eusarsiella zostericola (Cushman, 1906) arrives in mainland Europe
In sediment samples collected in the Oosterschelde, a marine embayment in the southwest of The Netherlands (southern North Sea), nine specimens of a non-native myodocopid ostracod were found. The ostracods were identifed as the North American species Eusarsiella zostericola (Cushman, 1906), previously introduced to southeastern England, probably with imported American oysters.
Introduction
At least three species introduced to southeast England with oysters from North America have subsequently been recorded from The Netherlands.The American slipper limpet Crepidula fornicata (L., 1758) and the American piddock Petricolaria pholadiformis (Lamarck, 1818) arrived in The Netherlands, possibly with shellfish imports from England (Wolff 2005), and the American oyster drill Urosalpinx cinerea (Say, 1822) almost certainly did so (Faasse and Ligthart 2009).In view of the proximity of southeast England to the southwest of The Netherlands, other American species introduced to southeast England with oysters were likely to follow.
In the course of a survey of the marine ostracod fauna of the southwest of The Netherlands, samples of different kinds of substrates are taken.Several samples of algae, hydroids, sand bottom, and shell bottom have been studied in recent years.The first samples of a muddy-sand sediment mixed with shells and shell grit, taken in 2012, yielded the first ostracods of the order Myodocopida from The Netherlands.They do not belong to any of the native species known from the North Sea area.Their similarity to a species introduced to England with oysters from North America was immediately evident and this was checked with the pertinent literature.
Material and methods
Samples were taken in the southwest delta area of The Netherlands, on the eastern shore of the Southern Bight of the North Sea.The samples described in this article were taken from the upper 1.5 cm of sediment during SCUBA dives in the Oosterschelde near Zierikzee.On 7 September 2012, a sample was taken approximately 1 km west of the "Zeelandbrug" (51°37'46"N, 03°54'27"E) at a depth of 12 m below the low water mark and consisted of muddy-sand mixed with shell grit.On 22 September 2012, a sample was taken at the "Zeelandbrug" (51°37'44"N, 03°54'44"E) at a depth of 15 m below the low water mark and consisted of sandy-mud mixed with shells.At the latter location, additional samples were taken on 6 and 13 October and 10 November.An additional sample was collected from mud at a depth of 6.5 m on 18 October (see Appendix 1).For identification of podocopid ostracods, Athersuch et al. (1989) was used.For the myodocopid species, Angel (1993), Cushman (1906), Kornicker (1986Kornicker ( , 1991)), Hall (1987) and Bamber (1987b) were consulted.Only living specimens were recorded.
The myodocopid ostracods belonged to the family Sarsiellidae, subfamily Sarsiellinae (for details of samples see Appendix 1).The only sarsielline ostracod native to the North Sea area is Sarsiella capsula Norman, 1869, described from a locality near Shetland (Angel 1993).The species found in the samples was Eusarsiella zostericola (Cushman, 1906) (Figure 1).This species (females) differs from congeners by a combination of the following characters: 3 lateral ridges join in the centre of the convex valves and extend almost to the margin, valve margin without processes, and valve surface not covered with numerous rounded processes bearing abundant hairs (Cushman 1906;Kornicker 1986;Bamber 1987b).Two specimens collected on 7 September 2012 were deposited in the collection of Naturalis Biodiversity Center (RMNH.Crus.B.2152).
The total length of the specimens collected ranged between 0.58 and slightly longer than 1 mm.
Discussion
According to Kornicker (1975), E. zostericola arrived in southeast England with imports of Atlantic oysters Crassostrea virginica (Gmelin, 1791) from North America.Additional unpublished records suggest that E. zostericola persists in the Thames Estuary and has spread as far as the Isle of Wight on the southern coast of England (D.J. Horne, personal communication, November 2012).The most likely introduction vector of Eusarsiella zostericola to The Netherlands is the importating of shellfish from abroad.Because the ostracod is associated with soft sediments, transportation on ships' hulls is unlikely.Transport in the sediment of ballast water tanks cannot be ruled out, although the location where E. zostericola was found in The Netherlands suggests otherwise.The Oosterschelde is only visited by inland freight carriers and recreational vessels, whereas shellfish culture and trade are an important characteristic of the area.Several non-native organisms have been detected along with imported mussels in recent years (Wijsman and Smaal 2006).
The occurrence of ostracods on oysters from the Oosterschelde was documented decades ago (e.g., Korringa 1951), and includes: Loxoconcha rhomboidea (Fischer, 1855) (as L. impressa), Leptocythere castanea (Sars, 1866), Heterocythereis albomaculata (Baird, 1838), Hirschmannia viridis, Semicytherura nigrescens (Baird, 1838) (as Cytherura nigrescens), Cytherois fischeri (Sars, 1866), and Hemicythere villosa (Sars, 1866).Kornicker (1975) suggests that E. zostericola has also been introduced with transports of Atlantic oysters Crassostrea virginica to the Pacific coast of North America between 1870 and 1910, together with the American slipper limpet Crepidula fornicata and the American oyster drill Urosalpinx cinerea.Eusarsiella zostericola was described from the Pacific coast of North America erroneously as Eusarsiella tricostata (see Jones 1958aJones , 1958b;;Kornicker 1975).The arrival of Eusarsiella zostericola in the Oosterschelde is not unexpected.At least five species of North American origin are known to have been introduced to southeastern England along with American oysters: the polychaete worm Clymenella torquata (Leidy, 1855), the ostracod E. zostericola, the American slipper limpet Crepidula fornicata, the American oyster drill Urosalpinx cinerea (see Kornicker 1975), and the American piddock Petricolaria pholadiformis (see Wolff 2005).The latter three species already have been introduced to The Netherlands (Wolff 2005;Faasse and Ligthart 2009).Eusarsiella zostericola is small and thus can hide in crevices of oyster shells or between byssus threads of mussels and will also be harder to detect during inspection of imported shellfish.Southeastern England is very close to the southwest of The Netherlands (Figure 2), which ensures short transportation times for imported shellfish.
It is hard to estimate the year of introduction of Eusarsiella zostericola to The Netherlands.Korringa (1951) was the last person to study the ostracods of the region, and he only studied a small number of samples of the epifauna of oysters.In standard benthos sampling, it is easy to miss a species that is slightly more than a millimeter long, only present as adults part of the year, and that occurs in low densities.
Eusarsiella zostericola has wide physiological tolerances.On the Atlantic coast of North America, it occurs from Maine to Texas (Kornicker 1986) and is exposed to temperature ranges and fluctuations comparable to those in coastal waters of The Netherlands.Kornicker (1967) mentions temperatures in Texas up to 25˚C.The same author mentions occurrence in salinities from 22-36, and that the species occupies substrates comprised of black mud, grey mud, sand, and shells as well as occurrences on sea grasses and hydrozoans.Bamber (1987a) found E. zostericola in southeast England in salinities down to 26 on bottoms with an average silt content of 77%.
According to Bamber (1987a) E. zostericola can occur in high densities.In view of its predatory lifestyle (Kornicker 1967), a significant influence of high densities of this species on ecological processes cannot be ruled out.
Once a species has established a founder population on a different continent or in a different ocean, in this case after introduction with imported shellfish, further spread is likely to result as shellfish are transported to additional continents or oceans.Three additional species known to have been introduced with Pacific oysters Crassostrea gigas (Thunberg, 1793) to the Pacific coast of North America (Kornicker 1975), have also been introduced to the Oosterschelde: Ocenebra inornata (Récluz, 1851) (as Tritonalia japonica), Tapes philippinarum (Adams and Reeve, 1850) (as Paphia philippinarum) and Mytilicola orientalis Mori, 1935.Therefore, in risk assessments, e.g. of shellfish transports, species known as transoceanic or transcontinental invaders should receive special attention.Eusarsiella zostericola is one of those species. | 2018-12-05T11:32:49.004Z | 2013-02-01T00:00:00.000 | {
"year": 2013,
"sha1": "4e9c5e717c9bc84ab5579cac990f01bb16741c88",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.3391/bir.2013.2.1.08",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "4e9c5e717c9bc84ab5579cac990f01bb16741c88",
"s2fieldsofstudy": [
"Environmental Science",
"Biology"
],
"extfieldsofstudy": [
"Biology"
]
} |
256607379 | pes2o/s2orc | v3-fos-license | Bcl-2 regulates store-operated Ca2+ entry to modulate ER stress-induced apoptosis
Ca2+ plays a significant role in linking the induction of apoptosis. The key anti-apoptotic protein, Bcl-2, has been reported to regulate the movement of Ca2+ across the ER membrane, but the exact effect of Bcl-2 on Ca2+ levels remains controversial. Store-operated Ca2+ entry (SOCE), a major mode of Ca2+ uptake in non-excitable cells, is activated by depletion of Ca2+ in the ER. Depletion of Ca2+ in the ER causes translocation of the SOC channel activator, STIM1, to the plasma membrane. Thereafter, STIM1 binds to Orai1 or/and TRPC1 channels, forcing them to open and thereby allow Ca2+ entry. In addition, several anti-cancer drugs have been reported to induce apoptosis of cancer cells via the SOCE pathway. However, the detailed mechanism underlying the regulation of SOCE by Bcl-2 is not well understood. In this study, a three-amino acid mutation within the Bcl-2 BH1 domain was generated to verify the role of Bcl-2 in Ca2+ handling during ER stress. The subcellular localization of the Bcl-2 mutant (mt) is similar to that in the wild-type Bcl-2 (WT) in the ER and mitochondria. We found that mt enhanced thapsigargin and tunicamycin-induced apoptosis through ER stress-mediated apoptosis but not through the death receptor- and mitochondria-dependent apoptosis, while WT prevented thapsigargin- and tunicamycin-induced apoptosis. In addition, mt depleted Ca2+ in the ER lumen and also increased the expression of SOCE-related molecules. Therefore, a massive Ca2+ influx via SOCE contributed to caspase activation and apoptosis. Furthermore, inhibiting SOCE or chelating either extracellular or intracellular Ca2+ inhibited mt-mediated apoptosis. In brief, our results explored the critical role of Bcl-2 in Ca2+ homeostasis and the modulation of ER stress.
Introduction
Deregulation of apoptosis can lead to cancer and to autoimmune and degenerative diseases 1 . The first identified apoptotic regulator was Bcl-2. The Bcl-2 family of proteins decide the fate of cells with response to survival and death. The proteins of the Bcl-2 family are characterized by homology domains BH1-4 (for Bcl-2 homology domain 1-4). The family can be subdivided in two major groups: the anti-apoptotic subgroup (for example, Bcl-2 and Bcl-xL) and the pro-apoptotic subgroup comprising Bax-like proteins (for example, Bax and Bak), which contain the BH1-3 domains, and the BH3-only proteins (for example, Bid and Bad) 2 . Bcl-2 plays an important role in mitochondria and endoplasmic reticulum (ER) [3][4][5][6] . Most of the Bcl-2 family proteins contain a hydrophobic C-terminal domain, required for their specific localization to different subcellular compartments, such as the ER, mitochondria, and perinuclear membranes 7,8 . In the ER, Bcl-2 interferes with the induction of apoptosis by Bax 9 , ceramides, ionizing radiation 10 , serum withdrawal, and c-myc expression 11 . Recently, the focus of researchers has shifted toward finding the possible association between the effects of the Bcl-2 family on Ca 2+ homeostasis and their role in the control of apoptosis 12,13 . In addition, the specific localization of Bcl-2 in the ER membrane indicates that Bcl-2 regulates filling of ER intracellular Ca 2+ store 14,15 , suggesting that Ca 2+ signaling might be a target of the Bcl-2 oncoprotein.
The anti-apoptotic activity of Bcl-2 is mediated by its regulation of handling Ca 2+ level in the ER and mitochondria. However, determining whether Bcl-2 increases or decreases the ER luminal Ca 2+ will reveal its true role in the ER Ca 2+ handling 16 . One hypothesis states that Bcl-2 decreases the Ca 2+ concentration within the ER such that less Ca 2+ is available for release into the cytosol, thereby leading to a more modest mitochondrial Ca 2+ uptake. Many studies show that Bcl-2 can protect cells from stress-induced Ca 2+ release from ER and lead to subsequent cell death by lowering the loading of Ca 2+ in the ER 17,18 . In contrast, it has been argued that Bcl-2 does not diminish the content of Ca 2+ pool, instead it inhibits the opening of inositol 1,4,5-trisphosphate receptors (IP3Rs) on the ER, thereby reducing the extent of Ca 2+ mobilization for a given magnitude of cell stimulation 19 . He et al. reported that Bcl-2 mediated Ca 2+ uptake and preserved the Ca 2+ pool of the ER to prevent depletion of the pool 14 . Furthermore, Bcl-2 overexpression is associated with the reduction in the transient elevation of cytosolic Ca 2+ induced by thapsigargin (TG) 20 . A proposal common to many of these studies is the proposal that Bcl-2 functions to reduce the magnitude of increase in cytosolic Ca 2+ concentration in response to apoptotic stimuli. However, the two different functions of ERresident Bcl-2 may possibly depend on different cell contexts; this relation needs to be clarified in detail.
Ca 2+ can serve as an apoptotic signaling factor when delivered at the wrong time and to the wrong place 21,22 . Disruption of intracellular Ca 2+ homeostasis by influx of extracellular Ca 2+ is always lethal to cells 23 . It was believed that Ca 2+ -related cell death could be triggered by large, sustained increases in cytosolic Ca 2+ . More specifically, both persistent Ca 2+ release from the ER and store-operated Ca 2+ entry (SOCE) through Ca 2+ releaseactivated Ca 2+ channels are apoptogenic 24,25 . Several anticancer drugs that are used to induce cancer cell apoptosis function not only through the dysregulation of Ca 2+ signaling but also via the activation of other apoptotic modulators 26,27 , such as death receptor-and/or mitochondria-dependent pathways 28,29 . Ca 2+ is consdered to be a link between ER stress and mitochondrial apoptotic pathways 30,31 . SOCE, by definition, is activated by Ca 2+ efflux from the internal store. Stromal-interaction molecule 1 (STIM1) is a Ca 2+ sensor in the ER that triggers SOCE pathway activation. Once Ca 2+ is depleted in the ER, STIM1 proteins aggregate into multiple puncta and translocate to regions with close proximity to the plasma membranes 32 . Activated STIM1 binds to Orai1 (CRACM1, calcium release-activated calcium modulator 1) or/and TRPC1 (transient receptor potential canonical 1), two Ca 2+ channels, located in the plasma membrane, which allow Ca 2+ entry and are therefore involved in SOCE [33][34][35] . TG, a sarco/endoplasmic reticulum Ca 2 + -ATPase (SERCA) inhibitor that induces SOCE drastically following the depletion of Ca 2+ stores of the ER, has been shown to induce apoptosis 36 . However, the underlying mechanism of Bcl-2-mediated regulation of TGinduced apoptosis has not been clearly elucidated.
Interaction of Bcl-2 with Bax appears to be important for its activity, and the BH1 and BH2 domains of Bcl-2 are required for the inhibition of apoptosis and heterodimerization with Bax 37 . The α5-helix within the BH1 domain of Bcl-2 is essential for its cytoprotective function. Bcl-2, which lacks a BH1 domain (Bcl-2/ΔBH1) or bears alanine as a substitute for glycine 145 in the BH1 domain (Bcl-2/G145A), fails to interact with either Bax or Bak, accelerating Bax-or Bak-induced apoptosis. Bcl-2/ ΔBH1 or Bcl-2/G145A acts as a dominant-negative mutation of endogenous anti-apoptotic proteins such as Bcl-2 and Bcl-xL 38 . Furthermore, Bcl-2, but not the inactive point mutant Bcl-2/G145A, undergoes a conformational change in response to apoptotic agonists 39 . Bcl-2/G145A completely abrogated the death-repressor activity of Bcl-2 and disrupted its heterodimerization with Bax. Therefore, the structure of this region of Bcl-2 is important for its biological function 40 . In this study, in order to focus specifically on the effect of Bcl-2 on Ca 2 + -mediated apoptosis, we chose the apoptosis-inducing agent TG as an ER stress inducer to disrupt Ca 2+ homeostasis in the ER 41 . We observed that Bcl-2 downregulated the basal Ca 2+ levels in cytosol and mitochondria and prohibited Ca 2+ elevation under apoptotic stimulation. We also observed that SOCE contributed to the TG-induced Ca 2+ elevation and cytotoxicity when the α5-helix of Bcl-2 was mutated.
Results
Both wild-type and Bcl-2 mutants are localized in the ER and mitochondria A three-amino-acid mutation ( 144 WGR 146 to 144 AAA 146 ) was generated within the α5-helix of the Bcl-2 BH1 domain (Fig. 1a). We established Bcl-2 overexpressing stable clones of endogenous Bcl-2-free cells using MDCK and two other cervical cancer cell lines (SiHa and HeLa). Organelle fragmentation for Western blotting (Fig. 1b) and immunofluorescence staining (Fig. 1c) was performed to prove that both wild-type Bcl-2 (WT) and the Bcl-2 mutant (mt) were localized in the ER and mitochondria. In addition, overexpression of the isopropyl-beta-D-thiogalactoside (IPTG)-inducible WT in MDCK cells exhibited similar distribution of noninducible WT (Supplementary Fig. 1a).
Mutation of Bcl-2 enhances TG-induced apoptosis
The cells that overexpressed the WT resisted TGinduced apoptosis, whereas cells that overexpressed mt enhanced TG-induced apoptosis (Fig. 2a, b). Furthermore, overexpression of IPTG-inducible WT decreased TGinduced apoptosis in a dose-dependent manner (Supplementary Fig. 1b). In contrast, overexpression of mt enhanced TG-induced apoptosis of cervical cancer cells in a time-dependent manner ( Supplementary Fig. 1c, d).
These results indicate that the Bcl-2 α5-helix can modulate TG-induced Ca 2+ cytotoxicity.
TG-induced ER stress-mediated apoptosis
Overexpression of glucose-regulated protein 78 (Grp78), an ER-resident chaperone protein and a hallmark of ER stress, can prevent cell death by protecting against many apoptotic stimulations. The results revealed that after TG treatment, Grp78 was upregulated in the cells that overexpressed WT, and downregulated in the cells that overexpressed mt to that in MDCK cells (C) (Fig. 2c). Cleavage of procaspase-12, procaspase-8, procaspase-9, and procaspase-3 are indicators of activation of ER stress-, death receptor-, mitochondria-dependent, or common apoptotic pathway, respectively. Herein, cleavage of procaspase-12 and induction of the ER stress, which specifically activated caspase-12, were found in cells that overexpressed mt, whereas overexpression of WT inhibited the activation of caspase-12 ( Fig. 2d) and the effector caspase-3 (Fig. 2e). Neither WT nor mt activated caspase-8 ( Fig. 3a) or caspase-9 ( Fig. 3b) or caused the loss of mitochondrial membrane potential ( Fig. 3c) after TG treatment. These results implied that ER stress, but not mitochondria-dependent or death receptor-dependent pathways, contributed to the TGinduced apoptosis.
Bcl-2 and Bax interaction is not involved in TG-induced apoptosis
Immunoprecipitation results revealed no interaction between Bcl-2 and Bax under TG treatment, regardless of the overexpression of WT or mt (Fig. 3d). Conversely, HA14-1, an inhibitor of Bcl-2, prevents Bcl-2 from interacting with Bax, thereby inhibiting the anti-apoptotic effect of Bcl-2 and sensitizes cells to induction of apoptosis. As shown in this study, cells that overexpressed WT or mt, HA14-1 did not change the level of TG-induced apoptosis (Fig. 3e). In addition, we also tested the effect of Bcl-2 on tunicamycin (TUN)-induced apoptosis. TUN, a nucleoside antibiotic and an ER stress inducer, inhibits protein glycosylation, resulting in the accumulation of proteins in the ER and causing ER stress. TUN has been reported to increase cytosolic Ca 2+ levels by Ca 2+ efflux from the Ca 2+ pool in the ER, and influx of extracellular Ca 2+ across the plasma membrane 42 . As shown in Fig. 4, WGR 146 to 144 AAA 146 , in the Bcl-2 α5-helix motif within the BH1 domain. b Western immunoblotting of Bcl-2, β-actin, calnexin, and porin. SDS-PAGE for the whole-cell lysate, ER lysate, and mitochondrial lysates from MDCK cells that overexpressed control vector (C), wild-type Bcl-2 (WT), and Bcl-2 mutant (mt). β-actin, calnexin, and porin were used as the internal control for whole-cell, ER, and mitochondrial lysates, respectively. c Immunofluorescence staining was performed to label Bcl-2, ER, and mitochondria, and the fluorescence images were obtained using confocal microscopy (scale bar, 20 μm). Calreticulin and MitoTracker Orange were used as the ER and mitochondrial markers, respectively mt enhanced TUN-induced caspase-12 activation and apoptosis ( Fig. 4a-c), but downregulated Grp78 (Fig. 4b).
In contrast, WT exerted opposite effects on TUN-induced cell death, compared to mt. Similar to the results observed for TG-induced apoptosis treatment, inhibition of the interaction of Bcl-2 with Bax by HA14-1 did not affect TUN-induced apoptosis (Fig. 4d).
Bcl-2 α5-helix regulates intracellular Ca 2+ compartmentation
Depletion or overload of Ca 2+ in cytosol or organelles may cause stress and Ca 2+ cytotoxicity, leading to cell death. At first, we evaluated Ca 2+ levels in the Ca 2+ pool of ER stores by passive TG-induced Ca 2+ leakage and active ATP-induced Ca 2+ release from the ER into the Overexpression of control vector (C), wild-type Bcl-2 (WT), and Bcl-2 mutant (mt) in MDCK cells that were treated with 2 μM TG for 48 h and 72 h. a Representative images of cells as observed under a bright-field microscope (scale bar, 100 μm). b Quantitative analysis of the apoptosis ratio that was assessed from the hypodiploid DNA peak of propidium iodide (PI)-stained cells by flow cytometry from five independent experiments. The data were found to be statistically significant at p < 0.01 (indicated by **) and p < 0.001 (indicated by ***) compared to untreated cells (0 h) of the same cell line, and p < 0.05 (indicated by #) and p < 0.01 (indicated by ##) compared to control vector (C) overexpressing cells at the same time point (Student's t-test). c, d Cells were treated with 2 μM TG, and the cell lysates were harvested at the indicated time points. The lysates were analyzed by SDS-PAGE and Western blotting for Grp78, caspase-12, and the internal control β-actin. Arrow and arrowhead indicate the inactive and active caspase-12, respectively. Representative pictures of three independent experiments. c Lower panel, densitometry was performed to quantify the relative level of Grp78 expression from three independent Western blots. The data were found to be statistically significant at p < 0.001 (indicated by ***) compared to control vector (C)-overexpressing cells at the same time point (Student's t-test). e Caspase-3 activity was measured using the fluorogenic substrate DEVD-AFC cells treated with 2 μM TG for 6 h or 12 h. The data were found to be statistically significant at p < 0.01 (indicated by **) compared to control vector (C)-overexpressing cells at the same time point (Student's t-test) cytosol. We found that overexpression of WT and mt increased and decreased the Ca 2+ concentration within the ER, respectively (Fig. 5a, b). IP3Rs and SERCAs are major molecules acting in response to the release of Ca 2+ from the ER and uptake of Ca 2+ into the ER, respectively. As shown in Fig. 5c, overexpression of WT downregulated IP3R3 and upregulated SERCA2 and SERCA3.
In contrast, overexpression of mt upregulated IP3R3 and downregulated SERCA3 (Fig. 5c). Additionally, organellespecific fluorescent Ca 2+ indicators, mag-fura-2/acetoxymethyl ester (mag-fura-2/AM), fura-2/acetoxymethyl ester (fura-2/AM), and rhod-2/acetoxymethyl ester (rhod-2/AM), were used to identify ER ([Ca 2+ ] ER the resting state, respectively. Cells that overexpressed mt showed a lower Ca 2+ level in the ER but a higher Ca 2+ level in the cytosol and mitochondria (Supplementary Fig. S2 and Fig. 5d-f). However, cells that overexpressed WT showed a higher Ca 2+ level in the ER but lower Ca 2+ level in the cytosol and mitochondria. High cytosolic Ca 2+ concentration can activate calpains, which can regulate cellular functions and cause cell death by proteolyzing its cellular substrates. The fluorogenic substrate, t-Boc-LM-CMAC, was used to evaluate the basal activity of calpains. Our data showed that WT decreased calpain activity, whereas mt increased calpain activity (Fig. S3a, b). In addition, changes in the levels of μ-calpain expression also corresponded with the calpain activity as shown by cleavage of the calpain substrate, α-spectrin (Fig. S3c).
Bcl-2 α5-helix is important in regulating Ca 2+ homeostasis TG-mediated Ca 2+ -ATPase inhibition allows Ca 2+ to flow from the ER lumen into the cytoplasm, producing a transient elevation of cytosolic Ca 2+ level, and triggering a sustained elevation of cytosolic Ca 2+ level through SOCE. The genetically encoded FRET-based fluorescent protein, cameleon, and the ratiometric chemical dye, fura-2/AM, were used as indicators in Ca 2+ measurement (Fig. 6). The ratio images (Fig. 6a) and ratio values (Fig. 6b) demonstrated that during TG treatment, overexpression of WT was negatively correlated with cytosolic Ca 2+ elevation, whereas overexpression of mt was positively correlated with cytosolic Ca 2+ elevation during TG treatment (Supplementary Video S1, S2). The cytosolic Ca 2+ level in TG-treated cells that overexpressed WT returned to the original level rapidly. In contrast, cells that overexpressed mt showed dramatic elevation in the cytosolic Ca 2+ level, and sustained this level for a long period of time. In addition, similar results were obtained with the ratiometric Ca 2+ probe, fura-2/AM (Fig. 6c).
Bcl-2 α-helix regulates SOCE pathway activation during TG treatment
We found that mt highly enhanced and sustained the cytosolic Ca 2+ elevation during TG treatment. Therefore, we investigated whether Bcl-2 can regulate STIM1 translocation to plasma membrane and cause SOCE pathway activation. Confocal microscopy images showed that WT reduced STIM1, which acts as Ca 2+ sensors in the ER, that led to their translocation to the juxtaplasma membrane regions during TG-induced Ca 2+ depletion. In contrast, mt significantly promoted the aggregation of most of the STIM1 proteins into multiple puncta and translocated them to the juxtaplasma membrane region (Fig. 7a). As shown in Fig. 7b, c, the increase in SOCE activation, as observed by Ca 2+ measurement, is consistent with the translocation ability of STIM1 shown in Fig. 7a. In addition to SOCE pathway activation, the expression of SOCE-related molecules was examined. Western blotting results also demonstrated that WT decreased the level of STIM1, Orai1, and Orai2, resulting in reduction of SOCE. In contrast, Bcl-2 increased the level of STIM1, Orai1, Orai2, Orai3, and TRPC1 expression (Fig. 7d), leading to the enhancement of SOCE and Ca 2+ burst. The extracellular Ca 2+ chelator, EGTA, inhibited TG-induced apoptosis significantly (Fig. 7e). Furthermore, TG-induced apoptosis of the cells overexpressing mt could be reduced by intracellular Ca 2+ chelator, BAPTA/AM, extracellular Ca 2 + chelator, EGTA, and the SOCE inhibitor, 2-APB (Fig. 7f).
Discussion
Ca 2+ is an important secondary messenger that regulates multiple cellular processes 21 . Bcl-2 exerts a direct effect on the Ca 2+ handling in the ER by regulating the movement of Ca 2+ through the ER membrane 14,25,43 . However, the effect of Bcl-2 on the intracellular and intraorganellar Ca 2+ level remains controversial as it is (see figure on previous page) Fig. 3 Death receptor-or mitochondria-dependent pathways, and the interaction of Bcl-2 with Bax are not involved in thapsigragin (TG)induced apoptosis. a Control vector (C)-, wild-type Bcl-2 (WT)-, and Bcl-2 mutant (mt)-overexpressing MDCK cells were treated with 2 μM TG, and the cell lysates were harvested at the indicated time points. The lysates were analyzed by SDS-PAGE and Western blotting for caspase-9, caspase-8, and the internal control β-actin. Representative pictures of three independent experiments. b Caspase-9 activity was measured using the fluorogenic substrate LEHD-AFC in cells treated with 2 μM TG for 6 or 12 h. (n = 3; Student's t-test). c Cells were treated with 2 μM TG or 500 nM staurosporine (STS) for 12 h and subsequently stained with 500 nM TMRM for 30 min. Mitochondrial membrane potential was analyzed based on the TMRM intensity observed by performing flow cytometry. STS-induced decrease of TMRM fluorescence was used as the positive control. The x-axis indicates TMRM fluorescence intensity, and the y-axis indicates cell number (n = 3; Student's t-test). d Cells were treated with 2 μM TG for 12 or 24 h, and cell lysates were harvested at the indicated time points. The lysates were subjected to immunoprecipitation using anti-Bcl-2 antibody (upper-left) or an IgG isotype control (lower-left), followed by analysis of the precipitated proteins by SDS-PAGE and immunoblotting with Bcl-2, Bax, and Bak antibodies. The whole-cell lysate of wild-type Bcl-2 (WT)-overexpressing MDCK cells was used as positive control (indicated by + ). Western blotting of Bcl-2, Bax, Bak, and the internal control β-actin were performed in whole-cell lysates to represent input control (upper-right). e Cells were preincubated with or without 5 μM HA14-1 for 1 h and subsequently treated with or without 2 μM TG for 36 h. Quantitative analysis of the apoptosis ratio was assessed from the hypodiploid DNA peak of propidium iodide (PI)-stained cells from three independent experiments by flow cytometry (Student's t- Fig. S2a, b, and Fig. 5), which may be due to increased Ca 2+ uptake into the ER or decreased Ca 2+ leakage from the ER. In contrast, mt decreased [Ca 2+ ] ER and increased [Ca 2+ ] i ( Supplementary Fig. S2a, b, and Fig. 5), which may be due to decreased Ca 2+ uptake into the ER or increased Ca 2+ leakage from the ER. In addition, the mitochondrial Ca 2+ level appeared to be responding to the cytosolic Ca 2+ level ( Supplementary Fig. S2c and Fig. 5f). Consequently, the α5-helix of Bcl-2 is involved in Bcl-2-mediated Ca 2+ distribution. These or functional interaction between Bcl-2 and IP3Rs or SERCAs (Fig. 5c).
In this study, we found that overexpression of mt enhanced TG-induced apoptosis which selectively activated ER-stress-related apoptosis. This conclusion was supported by the following findings: (i) Most cells that overexpressed mt rounded up with the appearance of apoptotic bodies after TG treatment (Fig. 2a), which was confirmed by propidium iodide (PI) staining ( Fig. 2b and Supplementary Fig. S1). (ii) Several ER-stress-related molecules were activated and procaspase-12 was cleaved into the active caspase-12 fragment, suggesting the activation of caspase-12 (Fig. 2d); procaspase-3, the downstream target of caspase-12, was successively activated (Fig. 2e); and Grp78, the ER chaperone and ER stress responder, was overexpressed after TG treatment. (iii) The death receptor-and mitochondria-dependent signal pathways were not involved in TG-induced apoptosis because there were no significant changes in caspase-8 and caspase-9 activation and in the mitochondrial membrane potential (Fig. 3). Previous studies show that Grp78 prevents apoptosis induced by the disruption of Ca 2+ homeostasis in the ER 47 . Conversely, suppression of Grp78 expression causes an increase in cell death induced by Ca 2+ depletion in the ER 48 . In the present study, after treatment with ER stress inducer (TG and TUN), the level of Grp78 expression in the cells that overexpressed WT was higher than that in cells that overexpressed mt ( Fig. 2c and Fig. 4b). Our data indicated the effect of the Bcl-2 α5-helix on Grp78 expression in relation to the regulation of ER stress-induced apoptosis. Furthermore, intracellular Ca 2+ chelator (BAPTA-AM), extracellular Ca 2+ chelator (EGTA), and SOCE inhibitor (2-APB) attenuated TG-induced apoptosis (Fig. 7f). These results indicated that mt depleted the intracellular Ca 2+ stores to enhance the ER stress-induced apoptosis which involved SOCE-mediated sustained elevation of intracellular Ca 2+ . ] mito ), respectively. Quantitative analysis of the relative Ca 2+ levels was performed by measurement of fluorescence intensity of these Ca 2+ indicators. All values are represented as mean ± SEM (where, n ≥ 150 cells). The data were found to be statistically significant at p < 0.05 (indicated by *) and p < 0.01 (indicated by **) compared to control vector (C)-overexpressing cells (Student's t-test) The BH1 and BH2 domains of Bcl-2 are required for inhibition of apoptosis and heterodimerization with Bax 37 . Therefore, mutation of the Bcl-2 α5-helix within the BH1 domain may disrupt the interaction between Bcl-2 and Bax. Herein, overexpression of WT led to the downregulation of the pro-apoptotic Bcl-2 proteins, Bax and Bak, and increased the anti-apoptotic Bcl-2 activity, which was confirmed using anti-active Bcl-2 antibody. In contrast, overexpression of mt increased Bax activity, which was confirmed using anti-active Bax antibody (Fig. S4). Immunoprecipitation and HA14-1 pretreatment showed that the interaction between Bcl-2 and Bax is not involved in TG and TUN-induced apoptosis (Figs. 3 and 4). SOCE helps to replenish the Ca 2+ pool in the ER, where the filling state of the intracellular Ca 2+ stores regulates the entry of Ca 2+ across the plasma membrane. We found that overexpression of WT increased [Ca 2+ ] ER and slightly decreased SOCE. The slight decrease in SOCE could be a direct consequence of WT overexpression or an adaptive mechanism in response to the long-term elevation of Ca 2+ content in the ER 49 . Additionally, we observed a decrease in [Ca 2+ ] ER and significant increase in SOCE in cells that overexpressed mt (Fig. 5a, b, and Fig. 7b). In principle, a reduction in the steady state [Ca 2+ ] ER level due to mt overexpression should activate SOCE. It is indicated that overexpression of mt placed the cells at a [Ca 2+ ] ER -depleted status. TG-induced [Ca 2+ ] i increase via emptying of the Ca 2+ stores in the ER and activation of SOCE pathway has been shown to induce apoptosis in a wide variety of cell types 21,42,50 . TG-mediated Ca 2+ -ATPase inhibition allows Ca 2+ transport from the ER lumen into the cytoplasm, leading to a transient elevation of cytosolic Ca 2+ concentration, followed by sustained elevation of cytosolic Ca 2+ due to SOCE 51 . The results of this study showed that overexpression of WT could reduce sustained elevation of cytosolic Ca 2+ induced by TG, and nullify the Ca 2+ disturbance rapidly (Fig. 6). However, overexpression of mt helped to maintain the sustained elevation of cytosolic Ca 2+ (Fig. 6 Fig. 6 The putative Bcl-2 pore-forming domain regulates thapsigargin (TG)-induced Ca 2+ elevations. a, b Monitoring changes in cytosolic Ca 2 + level by cameleon, a FRET-based Ca 2+ indicator, expressed in control vector (C)-, wild-type Bcl-2 (WT)-, and Bcl-2 mutant (mt)-overexpressing MDCK cells. The relative emission ratios (EYFP em /ECFP em ) from EYFP and ECFP under excitation of ECFP were measured in 2 mM Ca 2+ buffer every 1.5 s for 140 frames using confocal microscopy. TG was added at the time point of 30 s during time-lapse recording. a Representative ratio images of EYFP em / ECFP em were shown every 30 s and presented in pseudocolor (scale bars, 20 μm). b Quantitative analysis of the relative emission ratios (EYFP em / ECFP em ) before and after TG treatment. Each curve represents the mean values from at least 10 cells. c Ratiometric dye, fura-2/AM, was used as a probe for cytosolic Ca 2+ measurement using a single cell fluorimeter. Cells were loaded with 2 μM fura-2/AM at 37°C for 30 min before TG stimulation. The arrow indicates that cells were stimulated with 2 µM TG at 100 s in 2 mM Ca 2+ buffer. Representative curves data are represented as mean ± SEM (where, n ≥ 60 cells) from three independent experiments leading to ER stress and cell apoptosis ( Fig. 2 and Supplementary Fig. S1).
As the Ca 2+ -ATPase SERCA is responsible for the Ca 2+ uptake in the ER, it is possible that this protein is modulated by Bcl-2. In fact, increasing direct interaction between Bcl-2 and SERCA is associated with Bcl-2 expression, which has been demonstrated by immunoprecipitation studies 44 . Recently, strong experimental evidence indicated that Bcl-2 regulates the phosphorylation of IP3R, which in turn regulates the leak rate through the channel 52,53 . Another study showed that Bcl-2 reduces the probability opening of IP3Rs in the lipid bilayers 19 . Further studies are required to clarify the relationship between the Bcl-2 BH1 domain and other Ca 2+ -regulating molecules, such as IP3R3, SERCA2, or SERCA3. In conclusion, these results highlight the importance of Ca 2+ regulation by the Bcl-2 BH1 domain in response to ER stress-mediated apoptosis. Bcl-2 with an α5-helix mutation was involved in SOCE pathway activation, which may be the cause of TG-induced apoptosis. In contrast, WT showed an anti-apoptotic effect by reducing SOCE pathway activation during TG-induced Ca 2+ disturbance (Fig. 8). The DNA constructs and transfection
Materials and methods
The vector pCΔj contained G418 resistance gene and EBV-derived replication origin. The human bcl-2 cDNA sequences inserted into pCΔj vector (pCΔj-bcl-2) were expressed by simian virus 40 (SV40) enhancer/promoter regulatory elements 54 . A DNA construct without bcl-2 cDNA sequences (pCAj-SV2) was used as a control. Transfection of cells was done using Lipofectamine™ 3000 Reagent (Thermo Fisher Scientific, Waltham, MA). Fotyeight hours after transfection, cells were passaged by 1:10 dilution into G418 (500 μg/ml) selective medium. The IPTG-inducible expression of Bcl-2 by the lac operator/ repressor system was established 55 . The stable transfection of a constitutively expressed lacI gene, encoding lac repressor, and the human bcl-2 gene that had been inserted downstream of a SV40 promoter containing the lac operator sequence in MDCK cells. The expression of the bcl-2 gene could be specifically activated by administration of the lactose analog IPTG.
Fractionation of organelles
Cell homogenates were obtained using Dounce homogenizer along with a non-detergent lysis buffer (50 mM Tris-HCl at pH 7.4, 4 mM ETDA, 2 mM EGTA, 20 μg/ml leupeptine, 50 μM PMSF) on ice. The homogenate was centrifuged at a low-speed of 1000×g for 10 min, following which the supernatant was subjected to a medium-speed centrifugation at 20,000×g for 30 min. The pellets, which contained the mitochondria, were dissolved using 1% Triton X-100 lysis buffer. Meanwhile, the supernatant from the medium-speed centrifugation step was subjected to high-speed centrifugation at 40,000×g for 60 min, and the pellets containing the ER were dissolved using 1% Triton X-100 lysis buffer.
Flow cytometric analysis
Cells were stained with 500 nM tetramethyl rhodamine methyl ester (TMRM; Molecular Probes, Eugene, OR), an indicator of mitochondrial membrane potential. The TMRM-stained cells were incubated in the dark at room temperature for 30 min and subsequently analyzed by flow cytometry (BD FACSCalibur, San Jose, CA) with excitation wavelength at 543 nm. In order to analyze the apoptotic ratio, the cells were fixed in 70% alcohol, followed by treatment with RNAse (100 mg/ml), and stained with PI (40 mg/ml) (Sigma, Saint Louis, MO). The PI- Wild-type Bcl-2 (WT) maintains high Ca 2+ levels in the ER but low Ca 2+ levels in cytosol and mitochondria. The three-amino acid mutation ( 144 WGR 146 to 144 AAA 146 ) in the Bcl-2 α5-helix (mutant Bcl-2, mt) depletes Ca 2+ in the ER store but causes Ca 2+ overload in both cytosol and mitochondria. WT and mt cause different expression of SOCE-related molecules. During thapsigargin stimulation, mtoverexpressing cells presented a significant Ca 2+ influx through SOCE and subsequent activation of ER-stress-mediated caspase activation and apoptosis. In contrast, WT-overexpressing cells decreases thapsigargin-induced SOCE and intracellular Ca 2+ elevation which protects cells against apoptosis stained cells were incubated in the dark at room temperature for 30 min and analyzed by flow cytometry (BD FACSCalibur, San Jose, CA) with excitation wavelength at 543 nm. The apoptotic ratio was assessed from the hypodiploid DNA peak of apoptotic cells (sub G0/G1 phase) using Cell Quest software.
Intracellular Ca 2+ measurement
Cytosolic Ca 2+ was measured at 37°C using the fura-2 fluorescence ratio method on a single cell fluorimeter. Cells were loaded with 2 μM fura-2/AM in DMEM culture medium at 37°C for 30 min. The excitation wavelength was alternated between 340 nm (I 340 ) and 380 nm (I 380 ) using the Polychrome IV monochromator (Till Photonics, Grafelfing, Germany). The fluorescence intensity was monitored at 510 nm, stored digitally, and analyzed by the program of TILLvisION 4.0 (Till Photonics).
Evaluation of cytosolic Ca 2+ levels by the fluorescent protein cameleon
We performed the stable transfection of the cells with the cameleon plasmid, followed by live-cell imaging using a confocal imaging system (Olympus FV-1000, Tokyo, Japan). An optional 440 nm LD laser was used to directly excite the enhanced cyan fluorescent protein (ECFP) without exciting the enhanced yellow fluorescent protein (EYFP). Conversely, emission spectra of ECFP and EYFP were 470-500 and 535-565 nm, respectively. The analysis program in FV-1000 imaging system allowed ratio calculation and ratio image acquisition after subtracting the background fluorescence intensity.
Immunofluorescence staining and imaging
Cells were fixed with 4% buffered paraformaldehyde, permeabilized using 0.5% Triton X-100 for 15 min, and stained with mouse anti-Bcl-2 antibody (DAKO, Grostrup, Denmark), mouse anti-STIM1 antibody (BD, Franklin Lakes, NJ), or rabbit anti-calreticulin antibody (Upstate, Charlottesville, VA) for 12 h at 4°C. In addition, cells were stained with goat anti-mouse IgG conjugated with Alexa 488 or goat anti-rabbit IgG conjugated with Alexa 594 (Molecular Probes, Eugene, OR) for 1 h. The fluorophore was excited by laser at 488 or 543 nm, respectively, and detected using a scanning confocal microscope (Olympus FV1000, Tokyo, Japan).
Measurement of μ-calpain activity
The μ-calpain activity was assessed by generation of the fluorescent product, 7-amino-4-methoxy coumarin (AMC), from hydrolysis of an artificial μ-calpain fluorescent substrate t-Boc-LM-CMAC (Molecular Probes, Eugene, OR) 56 . Cells were pretreated with 10 μM t-Boc-LM-CMAC for 30 min. Intracellular fluorescence was sequentially imaged under a confocal imaging system (Olympus FV1000, Tokyo, Japan) and the mean fluorescence intensity (excitation by 405 nm LD laser) of individual cells was quantitatively analyzed.
Quantitative analysis of caspase enzymatic activity
Activation of caspase-3 and caspase-9 was measured by a fluorimetric assay based on the specific hydrolysis of DEVD-AFC and LEHD-AFC (MBL, Nagoya, Japan), respectively. The cells were resuspended in chilled celllysis buffer kept on ice for 10 min followed by the addition of reaction buffer. Finally, 50 μM of DEVD-and LEHD-AFC was added, and the mixture was incubated at 37°C for 2 h. The cleavage of DEVD-and LEHD-AFC was analyzed by measuring the release of AFC from DEVD-or LEHD-containing peptides. Fluorescence emission of AFC (excitation by 405 nm LD laser) was measured at 505 nm using a Fluoroskan AsCent FL fluorometer (Thermo Electron Corporation, Waltham, MA).
Statistics
All data were represented as mean ± SEM (standard error of the mean) and Student's t-test was performed for statistical analysis. A p value < 0.05 was considered significantly different.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. | 2023-02-06T15:09:16.201Z | 2018-02-26T00:00:00.000 | {
"year": 2018,
"sha1": "d7b72c34aa804fb8884a57d3fd0455f14832e7ab",
"oa_license": "CCBY",
"oa_url": "https://www.nature.com/articles/s41420-018-0039-4.pdf",
"oa_status": "GOLD",
"pdf_src": "SpringerNature",
"pdf_hash": "d7b72c34aa804fb8884a57d3fd0455f14832e7ab",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": []
} |
17811610 | pes2o/s2orc | v3-fos-license | Sex Factors in the Metabolic Syndrome as a Predictor of Cardiovascular Disease
Background Metabolic syndrome (MetS) is a condition characterized by a cluster of metabolic disorders and is associated with increased risk of cardiovascular disease (CVD). This study analyzed data from the Korean Health and Genome Study to examine the impact of MetS on CVD. Methods A total of 8,898 subjects (4,241 males and 4,657 females), 40 to 69 years of age, were enrolled and evaluated for the development of new onset CVD from 2001 to 2012 (median 8.1 years of follow-up). Results The prevalence of MetS at baseline was 22.0% (932/4,241) and 29.7% (1,383/4,657) in males and females, respectively. MetS was associated with increased risk of coronary heart disease (CHD; hazard ratio [HR], 1.818; 95% confidence interval [CI], 1.312 to 2.520 in males; HR, 1.789; 95% CI, 1.332 to 2.404 in females) and CVD (HR, 1.689; 95% CI, 1.295 to 2.204 in males; HR, 1.686; 95% CI, 1.007 to 2.192 in females). Specifically, MetS was associated with risk of future stroke in females only (HR, 1.486; 95% CI, 1.007 to 2.192). Among MetS components, abdominal obesity and hypertension were independent predictors of both CHD and CVD. In addition, a higher number of MetS components correlated with higher CVD risk. Conclusion MetS is a significant risk factor for the development of CVD although its impact varies between sexes.
INTRODUCTION
Metabolic syndrome (MetS) is a condition characterized by a cluster of metabolic disorders including abdominal obesity, insulin resistance/glucose intolerance, dyslipidemia and hypertension. The concept of MetS was introduced by Reaven [1] in 1988, as a clustering of cardiovascular risks, and MetS prevalence is increasing worldwide. The significance of MetS has unraveled gradually in recent years. According to data from the National Health and Nutrition Examination Survey (NHANES) III and NHANESs 1999 to 2006, the age-adjusted prevalence of MetS increased from 29.2% to 34.2% in the United States [2], and a similar trend has been observed in Asian countries [3]. The age-adjusted prevalence of MetS in Korea increased rapidly from 24.9% in 1998 to 31.3% in 2007 [4]. MetS also increases the risk of developing diabetes mellitus (DM) because its components represent major risk factors for impaired glucose metabolism [5]. Accordingly, the prevalence of DM and obesity in Korea has steadily increased [6,7] becoming one of the main targets of public health intervention as well as a social and economic problem in the near future. It is well documented that MetS is associated with increased risk of all-cause mortality and cardiovascular disease (CVD) [8][9][10]. However, the role of MetS in the development of CVD in Korea has only been shown in studies with small sample sizes or cross-sectional designs [11][12][13]. Therefore, this study analyzed data from the largest-scale Korean community based prospective cohort study to examine the impact of MetS on CVD over a 10-year follow-up period.
Study population
The design and baseline characteristics of the Ansung-Ansan cohort study have been published previously [14]. Briefly, it is an ongoing prospective, community-based cohort study that is part of the Korean Health and Genome Study (KHGS), a community-based epidemiological survey to investigate trends in DM and associated risk factors. The baseline examination was performed in 2001 to 2002, and biennial follow-up examinations were continued through 2012. To be eligible, subjects must be 40 to 69 years old, have resided within the borders of the survey area for at least 6 months before testing, and be mentally and physically able to participate. Participants were recruited from two Korean communities within 60 km of Seoul. Ansung is a representative rural farming community that had a population of 132,906 in 2000 [15]. Ansan is a representative urban community that had a population of 554,998 in 2000 [15]. A total of 10,038 subjects (5,018 from Ansung and 5,020 from Ansan, respectively) were enrolled in KHGS. Throughout the study, the same trained researchers and instruments were used to collect data. Anthropometric parameters and blood pressure were measured by standard methods. Fasting plasma glucose, lipid profiles, insulin and proteinuria were measured in a central laboratory. Social factors (smoking, exercise, and alcohol intake) were assessed by questionnaire. Current smokers were defined as those who smoked at least one cigarette per day for at least the prior year. Excessive alcohol intake was defined as consumption of ≥30 g of alcohol per day. Informed written consent was obtained from all participants. The study protocol was approved by the ethics committee of the Korean Center for Disease Control and the Institutional Review Board of Ajou University School of Medicine.
Definition of the MetS
Subjects were diagnosed with MetS if they met at least three of the following revised National Cholesterol Education Program Adult Treatment Panel III criteria [16]: (1) abdominal obesity; (2) triglycerides 150 mg/dL or greater; (3) high density lipoprotein cholesterol (HDL-C) less than 40 mg/dL in males or less than 50 mg/dL in females; (4) blood pressure 130/85 mm Hg or greater, or current use of antihypertensive medications; or (5) fasting plasma glucose 100 mg/dL or greater, or previously diagnosed type 2 DM or on oral antidiabetic agents or insulin. Abdominal obesity was defined as a waist circumference ≥90 cm for males and ≥85 cm for females, which are the proposed ethnically appropriate abdominal obesity criteria for Korean people by the Korean Society for the Study of Obesity [17].
Determination of incident CVD
Coronary heart disease (CHD) was defined as definite myocardial infarction confirmed by electrocardiogram and/or enzyme changes or any angina diagnosis that required intervention after confirmation of coronary artery stenosis by coronary angiography. Stroke included cerebral infarction, hemorrhagic stroke, transient ischemic attack, and vertebrobasilar insufficiency as demonstrated by diagnostic work-ups, such as, computed tomography, magnetic resonance imaging studies and accompanying neurologic symptoms and/or signs. CVD was defined as the occurrence of CHD and/or stroke. Persons with medical events reported by the patient himself or found during routine follow-up were asked to bring their medical records. Each event was confirmed and documented by the relevant clinical information (death certificates, postmortem reports, medical records, hospital discharge summaries, laboratory results, etc.). Initial data were obtained from 10,038 subjects who participated in KHGS. Among these subjects, 1,140 were excluded for the following reasons: (1) previous history of CVD (n=230); Copyright © 2014 Korean Endocrine Society (2) lack of follow-up examinations after baseline examination (n=908); and (3) missing data (n=2). After applying the above exclusion criteria, a total of 8,898 subjects were eligible for the study. Among them, 692 developed CVD events during the 10year follow-up period (mean follow-up of 96.8 months). The CVD group comprised all those who developed CVD during the follow-up period.
Statistical analyses
Statistical analyses were carried out using PASW version 21.0 (IBM Co., Armonk, NY, USA). Descriptive statistics for continuous data are expressed as the mean±standard deviation if normally distributed. Discrete data were summarized as numbers with percentages. Independent t test and Mann-Whitney U test were used to compare differences in continuous variables between the two groups. Chi-squared tests were used to analyze dichotomous variables between the groups. Age-and multivariate-adjusted hazard ratios (HR) and 95% confidence intervals (CIs) were estimated with the use of the Cox propor-tional hazards model to examine MetS baseline status and determine whether the variables played a role in the prediction of incident CHD, stroke, or CVD. For all statistical analyses, a two-sided P<0.05 was considered statistically significant.
RESULTS
Baseline characteristics of the study subjects are shown in Table 1. Mean follow-up duration was 8.1 years. The overall prevalence of MetS at baseline was 26.0% (22.0% in males and 29.7% in females). Males had significantly higher waist circumference, fasting plasma glucose, diastolic blood pressure, and triglycerides. A greater proportion of males were also current smokers and reported excessive alcohol intake. Compared to males, females were older and had higher body mass index, HDL-C level and homeostasis model of assessment-insulin resistance (HOMA-IR). Females also had a greater number of MetS components (1.8 vs. 1.5) compared to males. MetS was associated with CVD in both males and females despite con- Table 2). In addition to MetS itself, some individual components of the MetS were also significantly associated with the risk of developing CHD, stroke or CVD. This was independent of known cardiovascular risk fac-tors such as age, current smoking, low density lipoprotein cholesterol (LDL-C), and HOMA-IR (Table 3). Abdominal obesity was an independent predictor of both CHD and CVD in both sexes. Hypertension was a significant risk factor in all outcome parameters including CHD, stroke, and CVD in both males and females. Hyperglycemia was significantly associated with CHD and CVD only in males. However, lipid parameters (i.e., triglycerides and HDL-C) were not significantly correlated with the development of CVD. The HR for developing CVD increased with increasing numbers of MetS components, after adjusting for confounding factors (Table 4).
DISCUSSION
In the present study, MetS was a significant risk factor for the development of CVD although its impact varies between sexes. More specifically, MetS was correlated with the development of stroke only in females. In addition, the number of MetS components was linearly associated with risk of CVD or CHD. The overall prevalence rate of CVD in KHGS during the 10year follow-up period was 7.7% (692/8,898). Overall, MetS prevalence was 26.0% (2,315/8,898) and a higher proportion of females had MetS compared to men, which is similar to previously reported values [4]. A meta-analysis by Gami et al. [18] that included 36 different prospective studies found that the overall relative risk for incident CVD for individuals with MetS was 1.78 (95% CI, 1.58 to 2.00). In this study, the HR for development of CVD was 1.689 (95% CI, 1.295 to 2.204) in males and 1.686 (95% CI, 1.319 to 2.154) in females. This HR is also close to the estimate obtained from Asia Pacific Cohort Study Collaboration, which gathered individual data from 35 cohorts from the Republic of Korea, China, Hong Kong, Japan, Taiwan, Thailand, and Singapore [19]. In the present study, MetS was associated with risk of stroke in females, which has been reported previously [11,20]. Northern Manhattan Study [21] and The Atherosclerosis Risk in Communities data [22] also found that the effect of MetS on stroke risk was greater among women. They suggested that there may be risk differentials according to sex for the MetS due to disparities in the prevalence and potency of vascular risk factors. Sex differences in MetS components (higher proportion of abdominal obesity and lower HDL-C levels in females; data not shown) in this study could explain this phenomenon. Alternative explanations for possible sex differences include a greater impact of the MetS among postmenopausal women. CVD development rate was much higher in females age ≥50 compared to the others (data not shown). However, we do not have data on menopause and hormone use in KHGS. Longitudinal studies based on stroke registries suggested that women are more likely than men to have a stroke and the natural course of stroke is worse in women having a higher probability to be functionally dependent and institutionalized [23]. Recently, meta-analysis found that the excess risk of stroke associated with DM is significantly higher in women than men [24]. These data support the idea that that women's metabolic and vascular risk factor profile has to deteriorate to a greater extent than men to transition from normal to a deranged metabolic state. Moreover, it has been suggested that gender difference in arterial structure and function may lead to gender difference in the relationship of risk factors with stroke [25].
Among the five MetS components, blood pressure and abdominal obesity components were the key predictors of CVD in this study. Hypertension itself is a known cardiovascular risk factor [26,27] and data from 57,237 subjects who came to regular health examinations at a single center in Korea also confirmed this association [28]. Moreover, abdominal obesity is significantly associated with risk of CHD [29] and insulin resistance, which is the underlying etiology of MetS [1]. Abdominal obesity is a major factor in the increased prevalence of MetS in Koreans over the past 10 years [4]. It was previously reported that MetS is associated with abnormal left ventricular structure and function [30,31]. This cardiac dysfunction induced by insulin resistance and MetS could explain the development of CVD [32]. Hyperglycemia was associated with CHD and CVD only in males, although components of hyperglycemia were predictors of CHD and CVD in total subjects including both males and females (data not shown). Lipid parameters did not predict the development of CVD, though previous reports are inconsistent regarding which MetS components predict CVD [10,11,13,28]. These discrepancies may be www.e-enm.org 527 explained in part by the different study populations and different follow-up periods of each study. Aside from MetS components, age, smoking status, LDL-C, and HOMA-IR are independent risk factors of CVD (data not shown) which are wellknown risk factors of CVD [27].
The present study showed that the risk of incident CVD was found to increase with the number of MetS components, and increased by 2.6-fold in those with four or more MetS components compared to those without any MetS components. This synergistic association is noteworthy because some individual components such as hypertriglyceridemia and low HDL-C levels were not significant predictors of CVD. Frequency of CHD, stroke, and CVD were significantly correlated with an increasing number of MetS components (data not shown). This linear synergistic correlation has been reported previously [13,20] and increases the likelihood of CVD to an extent greater than the likelihood conferred by any individual MetS component [33]. This study also found that the risk of CVD increased incrementally as the number of MetS components increased. A similar relationship was found when CVD was sub-classified as CHD or stroke (data not shown). Two or more MetS components were predictive of CVD. This suggests that subjects with two MetS components (especially those with abdominal obesity and hypertension) who do not yet have MetS, should be alert for the development of CVD. Furthermore, it is important to determine ethnicity-specific cutoff points for individual components of MetS based on cardiovascular outcomes.
To the best of the authors' knowledge, this study is the largest prospective cohort study of a Korean population with a long follow-up period that reports the association of MetS with incident CVD. The strengths of this study include its longitudinal population-based study design, long duration of follow-up and sufficient number of CVD events, data analysis according to sex and exclusion of patients with CVD at baseline. One limitation of this study is that MetS diagnosis was based on a single measurement of MetS components at baseline. During followup, risk factors may have changed due to medication or lifestyle modifications, and thus MetS diagnosis may have changed during the course of the study. This would weaken the associations found in this study, biasing the results toward the null hypothesis. Therefore, the associations reported here may be stronger than the results indicate.
In summary, MetS was found to be associated with future risk of CHD and CVD in a prospective study with 10 years of follow-up. More specifically, MetS predicted stroke in females and CHD/CVD in both sexes. Risk of incident CVD increased with the number of MetS components. Therefore, individualized lifestyle modifications according to sex and number of MetS components, with or without pharmacological interventions, should be conducted to reduce the burden and negative consequences of MetS.
CONFLICTS OF INTEREST
No potential conflict of interest relevant to this article was reported. | 2016-05-04T20:20:58.661Z | 2014-12-01T00:00:00.000 | {
"year": 2014,
"sha1": "a08244ce7ab92aa5cd37e895555acbab4852ff04",
"oa_license": "CCBYNC",
"oa_url": "https://doi.org/10.3803/enm.2014.29.4.522",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a08244ce7ab92aa5cd37e895555acbab4852ff04",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
220891532 | pes2o/s2orc | v3-fos-license | Salycilic Acid Induces Exudation of Crocin and Phenolics in Saffron Suspension-Cultured Cells
The production of crocin, an uncommon and valuable apocarotenoid with strong biological activity, was obtained in a cell suspension culture of saffron (Crocus sativus L.) established from style-derived calli to obtain an in-vitro system for metabolite production. Salycilic acid (SA) was used at different concentrations to elicit metabolite production, and its effect was analyzed after a 4 days of treatment. HPLC-DAD analysis was used for total crocin quantification while the Folin-Ciocâlteu method was applied for phenolic compounds (PC) content. Interestingly, despite cell growth inhibition, a considerable exudation was observed when the highest SA concentration was applied, leading to a 7-fold enhanced production of crocin and a 4-fold increase of phenolics compared to mock cells. The maximum antioxidant activity of cell extracts was evidenced after SA 0.1 mM elicitation. Water-soluble extracts of saffron cells at concentrations of 1, 0.5, and 0.1 µg mL−1 showed significant inhibitory effects on MDA-MB-231 cancer cell viability. The heterologous vacuolar markers RFP-SYP51, GFPgl133Chi, and AleuRFP, were transiently expressed in protoplasts derived from the saffron cell suspensions, revealing that SA application caused a rapid stress effect, leading to cell death. Cell suspension elicitation with SA on the 7th day of the cell growth cycle and 24 h harvest time was optimized to exploit these cells for the highest increase of metabolite production in saffron cells.
Introduction
Saffron, Crocus sativus L., representing the most expensive spice in the world [1], is mostly used in medicine because of its wide pharmaceutical properties such as anti-depressive, antitussive, antigenototoxic, anticonvulsant, antihypertensive, anti-Alzheimer's, antioxidant, anti-diabetic, anti-Parkinson's, antinociceptive, and anti-inflammatory activities [2]. Saffron is valorized by crocin (crocetin ester), picrocrocin, and safranal bioactive compounds, the three main bioactive compounds of saffron s stigmas [3]. Crocin, the first most abundant compound of saffron, is responsible for the red color and is the highest and only water-soluble carotenoid, serving as a natural food colorant and an antioxidant even stronger than α-tocopherol [4,5]. Recently, crocin effects have proven to be responsible for most of the saffron medicinal effects [6]. Among other effects, it had anti-depressive effects on subjects with metabolic syndrome [7]; its application on rat hippocampus also improved memory function and learning [8]. Unfortunately, the supply of crocin remains expensive. Plant secondary metabolite production by cell cultures is gaining attention as a reliable, safe, and continuous supply strategy alternative to chemical synthesis [9]. The natural variability occurring in plant in-vitro cultures, although often unwanted, could be exploited for identifying interesting biosynthetic variants [10]. Moreover, the production of bioactive metabolites through cell cultures may be advantageous when the metabolites are biosynthesized in specific plant tissues [11]. Elicitor application to plant cell cultures has been proved to successfully enhance the biosynthesis of the desired bioactive compounds. In particular, it was shown with phytosterol and tocopherol in mung beans, safflower, and sunflower [12,13], artemisinin in Artemisia annua [14] or taxol in Taxus baccata suspension cultures [15]. Establishing such a bioproduction strategy for crocin, using a cell suspension culture and applying an elicited induction of biosynthesis, can be considered as a relevant improvement in satisfying the great demand for this valuable pigment in the food industry.
Salicylic acid (SA), known to be a plant defense signaling compound, has a pivotal role in alleviating injuries due to abiotic and biotic stress in plants [16,17]. SA was also shown to act as an elicitor to induce secondary metabolites accumulation. The effects of SA on the biosynthesis of several metabolites have been extensively clarified in many plants [18]. Phenolic compounds (PC) were elicited in cell suspension cultures of Thevetia peruviana [19] and Polygonum multiflorum [20]; similarly, alkanes and fatty acids were elicited in cell cultures of Jatropha curcas [21].
In our recent work, callus cultures were established from two different corm and style explants of saffron [22]. Applying different plant growth regulators (PGR), we optimized crocin production in style-derived calli, using thidiazuron in combination with naphthalene acetic acid on MS medium. In this study, style-derived calli with the best crocin production were used to establish a saffron cell suspension culture. The influence of different concentrations of SA was studied on cell growth, and on the accumulation of crocin and PC content in saffron cell extracts or their corresponding spent media. The extracts were further tested for their antioxidant activity and potential effect on cancer cells. To better interpret the effects of SA on cells, their compartmentalization was investigated microscopically in vivo using saffron protoplasts transiently transformed with different vacuolar markers such as RFP-SYP51, GFPgl133Chi, and AleuRFP. Altogether, our results provide a wide characterization to support a simple but effective optimization of the crocin bioproduction process.
Cell Suspension Culture and SA Effects on Cell Growth
The saffron cell suspension was obtained from optimized friable calli [22] and a 10-day subculture cycle was optimized in a few months. Methanol-dissolved SA at different concentrations (0.1, 0.5, and 1 mM) was administered to saffron cells. The same amount of methanol used for the preparation of each SA concentration was applied to saffron control cells and referred to as 0.1, 0.5, and 1 mock, while cells without methanol or SA represented controls. The effects of these treatments on cell growth were evaluated by measuring the fresh weight and shown in Figure 1. Three days after subculturing, the fresh weight of untreated control cells began to increase until day seven when the stationary phase started. SA negatively affected the growth rate of treated cells compared to the mock cells; the inhibitory effect of SA 1 mM on the cell growth rate was remarkable, showing just 0.2 g increased biomass after 10 days of subcultivation. The doses of 0.1 and 0.5 mM SA also showed an inhibitory effect on growth, but no significant difference was evidenced between them. Further, mock 1 showed a significant decrease in growth rate with a shorter log phase and a slight increase in fresh weight, while mock 0.5 and 0.1 showed a pattern similar to the untreated cells.
Analysis of Crocetin Esters and PC Profile Following the Imposition of SA
An increase of the characteristic pigmentation of in-vitro saffron cell cultures in the spent medium was observed soon after treatment with SA 0.1, 0.5, and 1 mM ( Figure 2). It is generally considered that a few days are required for a convenient metabolite accumulation but the color intensity is not a reliable quantitative measurement when dealing with metabolites with different oxidation forms and optical characteristics. For this reason, we measured molecular species more precisely. The total crocetin esters (glycosylated pigments) were evaluated in saffron cell cultures as previously published by our team [22]. In the present study, the production of total crocin (from 2 g FW incubated cells were quantified as mg mL −1 ) was measured, quantifying concentration in cell extracts as well as in the spent medium. Along with increasing SA concentrations, total intracellular crocin content decreased, but crocin content in the spent medium increased, reaching a maximum value of 0.22 mg mL −1 at SA 1 mM, approximately 5-fold higher than the mock 1 (0.032 mg mL −1 ). Moreover, 0.1 and 0.5 mM SA induced a significant increase of crocin production compared to the mock 0.1 and 0.5, while no significant difference was found between them ( Figure 3a). Similarly, by increasing SA concentration, the intracellular PC decreased significantly compared to the mock cells, showing the highest contents at SA 0.1 mM mock cells (1.18 mg mL −1 ), while the extracellular PC increased progressively to a maximum level of 3.91 mg mL −1 at SA 1 mM (Figure 3b), nearly 2-fold more than mock 1. In fact, SA extremely and surprisingly affected the exudation of nearly the whole amounts of crocin and phenolics biosynthesized inside saffron cells into the spent media. No significant differences were observed between crocin and PC amounts obtained from mock 0.1 and untreated control cells (data not shown).
SA Effect on Subcellular Compartmentalization
The cellular suspension can be reduced to protoplasts by digesting the wall using the same enzymatic mixture classically used for tobacco or Arabidopsis leaf tissues [23]. The resulting protoplasts are autofluorescent due to the high content of complex metabolites but can also be transiently transformed with fluorescent reporter proteins to characterize their compartmentalization. The vacuolar marker GFPgl133Chi was transiently expressed and evidenced its usual pattern [24]. Within 20 h GFP fluorescence was visible in the central vacuole ( Figure 4) as well as in the Endoplasmic Reticulum (ER), revealed by the observation of the nuclear envelope ( Figure 4A,C, evidenced by arrow 3). With time the ER labeling decreased. This marker was co-expressed with the Trans Golgi Network (TGN) and tonoplast marker RFP-SYP51 [25]. It showed that TGN is well visible ( Figure 4B, arrow 1) and, in addition to the central vacuole, several small compartments in close contact but separated from the tonoplast are also visible ( Figure 4B, arrow 2). GFPgl133Chi was also co-expressed with the lytic vacuole marker AleuRFP [26]. When GFPgl133Chi is completely exported from the ER and labels the central vacuole, AleuRFP still has a granulous distribution, evidencing that it s sorting to the vacuole is not yet complete ( Figure 4D-G). We observed that protoplasts had compartmentalization similar to parenchimatic cells with a unique central vacuole, but the population was anyhow heterogeneous in relation to autofluorescence. The autofluorescence generated by the excitation of 488 nm, emitted in the entire spectrum with relevant fluctuations. To visualize these fluctuations (not quantified), we show in Figure 5 that different emission patterns can be observed in control conditions ( Figure 5A,B). Differences were probably originated from the different age of cells. Since cell suspension was elicited with SA to stimulate crocin production, 1 mM SA treatment was also applied to protoplasts. The treatment caused tremendous stress to protoplasts ( Figure 5C,F) and induced death within a few hours. This observation suggested that SA treatment was determining the accumulation of metabolites in the first hours after treatment since cells were not increasing in number and were probably dying. We observed that protoplasts had compartmentalization similar to parenchimatic cells with a unique central vacuole, but the population was anyhow heterogeneous in relation to autofluorescence. The autofluorescence generated by the excitation of 488 nm, emitted in the entire spectrum with relevant fluctuations. To visualize these fluctuations (not quantified), we show in Figure 5 that different emission patterns can be observed in control conditions ( Figure 5A,B). Differences were probably originated from the different age of cells. Since cell suspension was elicited with SA to stimulate crocin production, 1 mM SA treatment was also applied to protoplasts. The treatment caused tremendous stress to protoplasts ( Figure 5C,F) and induced death within a few hours. This observation suggested that SA treatment was determining the accumulation of metabolites in the first hours after treatment since cells were not increasing in number and were probably dying.
Optimization of the Elicitation Time and Harvest Time for Metabolite Productions
Once we had observed the SA (1 mM) lethal effects on saffron protoplast within 24 h, we also evaluated total phenolics and crocin production 24 h after treatment. At first, the SA application was done conventionally at the time of the beginning of the exponential phase of the growth cycle of Saffron cells on day 3. The results evidenced that nearly the entire metabolite exudation into the spent medium was obtained within 24 h (Figure 6a). The amount of nearly 0.25 mg mL −1 total crocin and 5 mg mL −1 phenolics were obtained after SA 1 mM elicitation at day 3 and 24 h post elicitation time in the spent medium of saffron cells. Then, to obtain as much biomass as could be achieved, we also applied SA elicitor to saffron cell suspensions at the end of the exponential phase on day 7. No significant differences were observed between the metabolite yields (amount/biomass) from cells treated on day 7 compared to those obtained from cells treated on day 3 ( Figure 6b). The advantage of having larger biomass was then evident.
Optimization of the Elicitation Time and Harvest Time for Metabolite Productions
Once we had observed the SA (1 mM) lethal effects on saffron protoplast within 24 h, we also evaluated total phenolics and crocin production 24 h after treatment. At first, the SA application was done conventionally at the time of the beginning of the exponential phase of the growth cycle of Saffron cells on day 3. The results evidenced that nearly the entire metabolite exudation into the spent medium was obtained within 24 h (Figure 6a). The amount of nearly 0.25 mg mL −1 total crocin and 5 mg mL −1 phenolics were obtained after SA 1 mM elicitation at day 3 and 24 h post elicitation time in the spent medium of saffron cells. Then, to obtain as much biomass as could be achieved, we also applied SA elicitor to saffron cell suspensions at the end of the exponential phase on day 7. No significant differences were observed between the metabolite yields (amount/biomass) from cells treated on day 7 compared to those obtained from cells treated on day 3 (Figure 6b). The advantage of having larger biomass was then evident.
Antioxidant Activity of Saffron Cell Culture Extracts
To measure antioxidant activity in cell extract and spent medium of saffron cell cultures, in mock and SA-treated, three different assays were applied: oxygen radical absorbance capacity (ORAC), trolox equivalent antioxidant capacity (TEAC), and 1,1-diphenyl-2-picrylhydrazyl (DPPH). ORAC assay measures antioxidant activity by hydrogen atom transfer and provides a comprehensive analysis of a test sample's antioxidant activity in biological fluids, cells, tissues, and natural extracts [27]; TEAC assay is based on a single electron transfer activity [28]. These two assays could be normalized to Trolox equivalent units to quantify the composite antioxidant activity of the samples. Further, DPPH activities in cell extracts and their spent media were represented as % Inhibition (calculated by the electron-donation ability of the samples) and IC 50 value (the sample concentration required to inhibit DPPH free radical by 50%) [29]. In Figure 7, the antioxidant activity is reported as µmol of Trolox equivalents (TE) per mL (µmol TE mL −1 ) of saffron. The treatment with 0.1 mM SA showed the highest antioxidant activity in the spent medium of saffron cell cultures being 44. 22 and 159.88 µmol TE mL −1 for TEAC and ORAC assay, respectively. Moreover, ORAC and TEAC assays showed enhanced values of antioxidant activity in the spent medium of SA-treated saffron cells which correlated with increased amounts of metabolites like crocin and phenolics. In the case of DPPH assay, 20, 30, 40, and 50 µg mL −1 of saffron cell extracts and their spent media were evaluated as the optimum concentration for antioxidant activity, bleaching the purple color to yellow color. Inhibition percentage (I %) and IC50 are reported in Table 1. In the present study, the three applied assays showed a similar trend. In the case of DPPH assay, the lowest value of 19.47 IC 50 radical scavenging activity was obtained from cell extracts treated with 0.1 SA and their spent media indicating the most effective antioxidant activity, compared to the positive control ascorbic acid having IC50 6.79.
Effects of Saffron Cell Suspension Extract on Breast Cancer Cells
Water-soluble extract of saffron cells was examined for its effect on the viability of a well-characterized human breast cancer cell line MDA-MB-231. As shown in Figure 8, almost all used concentrations of saffron cell extract (1, 0.5, and 0.1 µg mL −1 ) showed significant inhibitory effects on cell viability after 72 h from their application, compared to untreated cells (control, CTR). Furthermore, this activity seems to be directly proportional to the concentration used, showing a stronger inhibitory effect at higher concentrations (1 and 0.5 µg mL −1 , p < 0.001) and a slight but significant activity at 0.1 µg mL −1 (p < 0.01).
Effects of Saffron Cell Suspension Extract on Breast Cancer Cells
Water-soluble extract of saffron cells was examined for its effect on the viability of a well-characterized human breast cancer cell line MDA-MB-231. As shown in Figure 8, almost all used concentrations of saffron cell extract (1, 0.5, and 0.1 µg mL −1 ) showed significant inhibitory effects on cell viability after 72 h from their application, compared to untreated cells (control, CTR). Furthermore, this activity seems to be directly proportional to the concentration used, showing a stronger inhibitory effect at higher concentrations (1 and 0.5 µg mL −1 , p < 0.001) and a slight but significant activity at 0.1 µg mL −1 (p < 0.01).
Discussion
A saffron cell suspension culture was established from calli optimized for their production of crocin and phenolic compounds. Intending to set an appropriate strategy to produce extracts with antioxidant and cell inhibitory activity, we here characterized the effect of cell elicitation by SA. Saffron is a sterile triploid plant of the Iridaceae family which is propagated only through cormlet production as the multiplication rate is quite slow, so concerns have risen due to the decreased production of saffron during the last two decades [30]. Many researchers studied the propagation of saffron through in vitro cultures such as shoot regeneration [31,32] and cormlet production [33,34], or the production of saffron typical metabolites through the induction of callus cultures [35] and stigma like structures [36] from various saffron explants; however, there are no valid reports on saffron cell suspension culture establishment and characterization so far. Since saffron metabolites mostly crocin are very precious, therefore the availability of plant suspension cultures, for the production of saffron valuable bioactive compounds is an important improvement in satisfying the great demand for this valuable product for the food industry. The availability of plant cell cultures for the production of specific metabolites is beneficial for providing a continuous supply of phytochemicals in controlled and axenic conditions, as a valuable alternative to the extraction from plants grown in the field under environmental and seasonal strain [9].
In this study, a saffron suspension cell culture was established and characterized as for the in vitro production of crocin and phenolics. In our system, SA was confirmed to be a valuable elicitor not only in terms of significant metabolite induction but also for the induction of metabolite exudation into the spent medium. At any applied concentration, SA significantly increased the extracellular amounts of apocarotenoid crocin and phenolic compounds which decreased in cell extracts. At the same time, it was evident that the reduced intracellular content of these metabolites could also be explained by the dramatic changes in cell viability due to the SA application. The growth rate of saffron cell suspension was significantly affected, decreasing along with the increase of SA concentration within a 10-day growth period, compared to the mock cells. In this regard, our results are in agreement with those collected with other cell cultures such as Taxus baccata [37], Salvia miltiorrhiza [38], Hypericum perforatum [39], and Polygonum minus [40], where cell growth was also inhibited after elicitation by SA. Indeed, SA elicitor application at the highest concentration dramatically stressed the saffron cell culture, causing cell death within 24 h. In this time, the elicitor also increased the production of saffron natural metabolites especially crocin, and caused their exudation into the culture medium. This behavior makes SA a promising elicitor to be used in a possible scaling-up procedure for bioreactor production of saffron bioactive compounds [41,42].
In plant cell cultures, phenylpropanoid compounds are those metabolites that are mostly affected by elicitors such as SA, commonly associated to a rapid and temporarily increased activity of the key biosynthetic enzymes phenylalanine ammonia-lyase (PAL) and chalcone isomerase (CHI) [43][44][45]. Enhanced levels of PC after SA treatment were reported in suspension cultures of Vitis vinifera and Polygonum multiflorum [20,46], as well as in Thevetia peruviana [19]. It seems that exogenous SA uptake into the cell induces resistance responses through the activation of anti-oxidative enzymes and eventually produces ROS species like H 2 O 2 . These play an important role as a second messenger in SA signaling pathways leading to the production of secondary metabolites in plant cell cultures [38,47]. In the established saffron suspension cell culture, we evaluated that 1 mM SA was the most effective treatment inducing crocin and phenolic biosynthesis. Besides, it stimulated the exudation of the bioactive compounds. In protoplasts from the same suspension culture, we observed that the elicitor SA induced strong cellular stress leading to cellular death within 24 h post elicitation time.
The protoplasts derived from saffron cells provided interesting information about the compartmentalization of cells in suspension culture, being representative of the original cells [24]. We transiently transformed the protoplasts with the vacuolar marker GFPgl133Chi. This marker consists of a GFP carrying the vacuolar sorting determinant of tobacco ChitinaseA and characterizes vacuoles with the less acidic lumen and less pronounced lytic activity [46]. Its sorting pathway was recently shown to bypass the Golgi apparatus [24]. It showed that protoplasts from cells always had a unique well-labeled central vacuole. GFPgl133Chi was co-expressed with the TGN and tonoplast marker RFP-SYP51, based on the fusion of RFP with the Arabidopsis SNARE SYP51 [25,48]. It revealed that in addition to TGNs and the central vacuole tonoplast, several small compartments could be observed around the central vacuole. GFPgl133Chi labeling did not appear evident in these small compartments and, since it follows a Golgi-independent pathway alternative to SYP51 sorting, we consider this observation as a confirmation that these are separated from the central vacuole. Then we co-expressed GFPgl133Chi with a second soluble vacuolar marker, AleuRFP [25,26] that follows the same sorting pathway of SYP51 [25]. This marker is considered to label lytic vacuoles following the classic sorting pathway dependent on the Golgi apparatus. The marker is the product of the fusion between the N-terminal vacuolar sorting determinant of the protease Aleurain from barley and RFP [49]. Aleurain is recognized by the well-known vacuolar sorting receptors VRS1 and its sorting is usually more efficient than GFPgl133Chi's sorting [24] and all previous literature comparing the two markers). We also observed that protoplasts from cells are rich of autofluorescence and that SA triggers cell death through the formation of a multitude of small autofluorescent compartments, often observed during stress and cell death. These observations suggested that protoplasts from cells had a vacuolar system with a central vacuole characterized by a marker of neutral vacuoles [24,50] and several small peripheric compartments characterized instead by the marker usually associated with proteases sorted through the Golgi apparatus. These peripheric compartments may be associated with those elicited by stress and SA.
We then demonstrated, by metabolite quantification after 24 h SA application, that also the metabolite production was induced immediately after elicitation. In agreement with our study, Mendoza et al. [19] also evidenced that SA affected cell suspension cultures of Thevetia peruviana obtaining an increased value by 21.41% of phenolic compounds during the first hours after elicitation. No significant differences were observed between the metabolite yields, in terms of amount/biomass, from cells treated on day 7 compared to those obtained from cells treated on day 3, thus the application of SA to saffron cells on day 3 of the growth cycle could be considered more efficient for biotechnological purposes in terms of time saved but the attainment of the maximum biomass after 7 days of cell growth is preferable to optimize the bioproduction effort. SA 1 mM led to the highest production of saffron bioactive compounds, likely as stress and defense response, and eventually their exudation into the spent media, while SA 0.1 mM showed the highest effects on antioxidant activity inside the cells. The latter can be evaluated as the optimal concentration in terms of induction of intracellular bioactive antioxidants in saffron cells. Investigation of saffron antioxidant and free radical scavenging activities has been conducted by many researchers but differing results were obtained due to the different origins, methods, and solvents used for extraction [51,52]. In Indian Crocus sativus L. the free radical scavenging activities were tested at concentrations ranging from 100 to 500 µg mL −1 with the highest activity in stigma extract of 70% followed by 79.6% in callus extracts [53]. Karimi et al. [54] reported a maximum of 210.79 µg mL −1 IC 50 of methanolic stigma extracts as compared with 255.44, and 299.44 µg mL −1 for boiling water and ethanol extract, respectively. They identified 6.5 mg gallic acid equivalent g −1 DW Phenolic Content and 5.8 ± 0.12 mg rutin equivalent g −1 DW flavonoid content in stigma extract as well. Further, the IC50 value of ethanolic extract of stigma was reported at 207.16 µg mL −1 by Baba et al. [55], while aqueous extract of Italian saffron showed the low IC50 value of 3.76 mg DW [56]. Since the most abundant metabolites of saffron are crocin and safranal, DPPH radical scavenging activity has been proved to be attributed to these two main bioactive compounds [57,58]. In the present study, saffron cells evidenced a considerable antioxidant activity under SA treatment which could be related to the high production of phenolic content and crocin. Phenolic compounds of plants have been identified to directly contribute to their antioxidant value [59], as well as the correlation between them and antioxidant activity have also been corroborated [60][61][62].
Saffron crocin and safranal have been studied for their ability to reduce cellular proliferation in different models including lung (A549), colon (HCT-116), breast (T47d), and prostate (PC-3) cancer cell lines [63]. For this reason, to test whether the saffron extract obtained from our cell suspension can reproduce this biological action in vitro, we determined the effects on a breast cancer human cell line. We showed that the aqueous extract of saffron cell suspensions showed a significant inhibitory effect on the growth of the aggressive breast cancer MDA-MB-231 cells [64], in a dose-dependent manner, even at the concentration of 0.1 µg mL −1 . This is an issue of great importance since we clearly described the approach used to obtain a fresh extract to be used on cells.
Establishment of Cell Suspension Cultures and SA Treatment
Cell suspension cultures were established by adding 2 g of friable calli to 100 mL Erlenmeyer flasks containing 3-5 mL of liquid MS medium with the addition of NAA 3 mg L −1 and TDZ 1 mg L −1 ), incubating on a rotary shaker (120 rpm) in the dark at 25 • C. The final volume of 25 mL was reached gradually in one month. Once the suspension culture was established, subcultivation was performed every 10 days. For subcultivation, 5 mL of grown suspension was transferred into 50 mL of fresh MS medium. The biomass accumulation was determined by measuring the fresh weight of the cultures at every 2 days. Each point of the growth curve represents the mean of three independent replications (flasks). Staining of the cells with fluorescein diacetate was done for controlling the cell viability [65].
SA at three different concentrations of 0.1, 0.5, and 1 mM was applied to examine the elicitation effect on biomass accumulation and metabolite production. SA was first dissolved in methanol, diluted with deionized water, and filter-sterilized before adding to the culture medium. The same amount of filter-sterilized methanol as used for each concentration of SA was added to the cell suspension cultures representing mock 0.1, 0.5, and 1. Ten-day old cells were filtered and 2 g fresh weight (FW) of them was transferred to 50-mL flasks containing 30 mL MS medium with 3 mg L −1 NAA + 1 mg L −1 TDZ. Flasks were incubated on a rotary shaker at 120 rpm and maintained at 23-25 • C in the dark. Cells were elicited with SA and mock solutions, unless differently specified, after 3 days of culture and maintained in the presence of treatment for further 4 days. Then, cells were centrifuged at 4000× g for 10 min (Beckman Allegra 22R, Beckman Coulter, Brea CA, USA), lyophilized (Labconco, Kansas City, MO, USA), and kept at −20 for metabolite analysis. The supernatant was used as the spent medium for further analyses.
Extraction of Crocins from Cells
The extraction of crocins was accomplished according to Hoshyar and Bathaie [66]. The lyophilized cells (1 g FW) were extracted with 5 mL of methanol-water (50:50% v/v) for 24 h under magnetic stirring in dark at 4 • C and centrifuged at 4000× g. The supernatant was collected and filtered through a nylon membrane (Acrodisc 13, 0.45 µm pore size, and 13 mm diameter), and injected as the cell extracts into the HPLC device (Agilent-1290 infinity Agilent Technologies, Waldbronn, Germany).
Extraction of Crocins and Phenolic Compounds from the Spent Medium of Suspension Cultures
The extraction of crocetin esters (crocins) from the spent medium was performed using the method of Schweiggert et al. [67], used for the extraction of free carotenoid and carotenoid esters with some modifications. Briefly, the 30 mL of discarded medium was washed with a mixture of 50 mL acetone: hexane (1:1, v/v) containing ascorbic acid as an antioxidant by a glass separatory funnel. After shaking, 50 mL of sodium chloride solution (10%, w/v) was used to wash the solution, followed by twice with 50 mL of water to remove acetone. The aqueous phase was re-extracted with ethyl acetate until it was colorless then evaporated with a vacuum pump. The residue was dissolved in a 5 mL mixture of methanol-water (1:1, v/v), membrane-filtered (0.2 mm), and used for crocin quantification by HPLC analyses as well as total phenolic content spectrophotometrically (Infinite M200 Tecan, Männedorf, Switzerland).
Extraction of Phenolic Compounds from Cells
The extraction of phenolic compounds from saffron cells was done according to Ranganna [68]. The cells were centrifuged at 4000× g for 10 min and then filtrated through nylon filters (30-µm pore size). Fresh cells (0.5 g) were homogenized with methanol 80% and kept for 3 h in a water bath at 70 • C. The extract was then centrifuged at 9000× g for 15 min at room temperature. The methanolic extract was discarded and brought up to volume for the analysis.
Total Phenolic Content
The total phenolic content of the samples was quantified according to the Folin-Ciocalteu method [69]. For quantification of intra and extracellular phenols, 1 mL of each sample extracted from cells and spent media were mixed with 2.5 mL Folin-Ciocalteu reagent (diluted with distilled water 1:10 v/v), 2 mL of 7.5% sodium carbonate and left in the dark at room temperature for 2 h. The absorbance of samples was measured at 725 nm against a blank. Total phenolic content was expressed as gallic acid equivalent (mg g −1 fresh weight) based on a gallic acid standard curve.
HPLC-DAD Analysis
Crocin was quantified by an auto-sampler HPLC device (Agilent-1290 infinity Agilent Technologies, Waldbronn, Germany) equipped with a C18 column (250 mm × 4.6 mm) and a multiple ultraviolet (UV) wavelength photodiode array detector. The elution was performed with a linear gradient of methanol: water: acetonitrile (50:42.5:7.5 v/v) as a mobile phase, a flow-rate of 1.0 mL min -1 and a maximum elution time of 30 min at room temperature and 440 nm. We used 2 crocin standards one including all identified crocetin esters (Sigma-Aldrich, Taufkirchen, Germany) for identification, and the other for pure α-crocin for the quantification of the crocetin esters (PhytoLab, Darmstadt, Germany).
Protoplasts Preparation, Transformation, and Analysis
C. sativus protoplasts were prepared as previously described for Nicotiana tabacum leaves and other plant material such as cell suspension cultures [23,70] from cells harvested from 3 or 7 day-old suspension culture. The observations evidenced no differences among protoplasts at the end of the process. Since the experimental procedure was strictly dependent on a good amount of cells, more experimental replicas were performed with cells grown 7 days available with higher biomass. The three vacuolar markers were previously characterized and used in several studies: GFPgl133Chi [24], RFP-SYP51 [25,49], AleuRFP [26,50]. Protoplasts were examined with a confocal laser microscope (LSM 710 Zeiss, ZEN software, GmbH, Jena, Germany). GFP was detected within the short 505-530 nm wavelength range, assigning the green color, RFP within 560-615 nm assigning the red color. Excitation wavelengths of 488 and 543 nm were used. Green autofluorescence was detected within the 505-560 nm wavelength range, red autofluorescence was detected above 650 nm. The laser power was set to a minimum and appropriate controls were made to ensure there was no bleed-through from one channel to the other. Images were processed using Adobe Photoshop 7.0 software (Mountain View, CA, USA).
Trolox Equivalent Antioxidant Capacity (TEAC) Assay
TEAC assay was done according to Re et al. [28]. TEAC assay is based on the scavenging of radical cation ABTS •+ (2,2 -azino-bis (3-ethylbenzthiazoline-6-sulfonic acid) by antioxidants, ABTS •+ is green in color and can be measured by absorbance at 734 nm. For the preparation of ABTS •+ radical cation stock solution, we used the concentration of 7 mM ABTS and 2.45 mM potassium persulfate, which was mixed and kept in dark at room temperature for 12-16 h before use. The fresh ABTS radical cation working solution was obtained with a dilution of 7 mM ABTS •+ stock solution with PBS (pH 7.4) to an absorbance of 0.70 (±0.02) at 734 nm on the plate reader equilibrated at 30 • C. The assay was adjusted to a 96-well format (Costar, 96-well clear round bottom plate, Corning). 210 µL final volume (10 µL of the sample or Trolox standard solutions (from 0 to 25 µM) with 200 µL diluted ABTS •+ solution) was prepared to each well of the microplate, reading the absorbance at 734 nm after 10 min using a microplate reader (Infinite M200 Tecan, Männedorf, Switzerland). For the Trolox reference standard, the percentage inhibition of absorbance at 734 nm was plotted and used along with the solvent blanks in every assay.
Oxygen Radical Absorbance Capacity (ORAC) Assay
ORAC assay was performed according to Davalos et al. [27] method using a 96-well format (Costar, 96-well clear round bottom plate, Corning). Assay final volume of 200µL was carried out in 75 mM phosphate buffer (pH 7.4). The solutions of the antioxidant (20 µL) and fluorescein (120 µL; 70 nm, final concentration) were prepared in the well of the microplate and preincubated for 15 min at 37 • C 60 µL (12 mM, final concentration) of 2,2 -Azobis-(2-methylpropionamidine) dihydrochloride (AAPH) solution was added and placed immediately in a microplate reader (Infinite M200 Tecan, Männedorf, Switzerland), and the fluorescein was recorded (excitation and emission wavelengths of 485 and 527 nm, respectively) every minute for 1 h. Three dilutions of samples were used in each assay and for every sample, three replications were used. We used a blank of fluorescein and AAPH in each assay. Diminution curves of fluorescence (intensity vs. time) were registered, followed by the calculation of area between the two Diminution curves (with or without antioxidant). Subtracting the blank amount from that of the sample or standard was acquired for the pure area under the curve. A standard curve of Trolox solution using different concentrations (1-6 µM) was used in every assay.
Final ORAC values were represented as µmol Trolox equivalents (TE) mL −1 of extract.
1-Diphenyl-2-picrylhydrazyl (DPPH) Radical Scavenging Assay
DPPH assay was performed according to Blois [29] method. The antioxidant ability of saffron cell extracts treated with SA and their spent media were measured by their scavenging activity in terms of hydrogen atoms or electron-donation ability to the stable free radicals of DPPH solution. Three mL DPPH methanolic solution was added to 1 mL samples containing different concentrations of saffron cell extracts and their spent media (20,30,40, and 50 µg mL −1 ) and kept in the dark at room temperature. After 30 min, the absorbance was read at 517 nm against the reagent blank. Ascorbic acid was used as the positive control in the same concentrations as the samples.
The free radical scavenging activities of the samples were determined as inhibition (%) and were calculated according to the following formula: A0 = DPPH solution absorption, A1 = DPPH solution absorption after sample addition. IC 50 value was calculated by plotting % inhibition versus concentration of samples curve to determine the optimal concentration of the antioxidants for decreasing the initial DPPH concentration by 50%, using the GraphPad Prism software.
MTT Assay
The concentrations of 1, 0.5, 0.1, and 0.01 µg mL −1 saffron extracts were prepared from 0.5 g lyophilized saffron cell suspensions which were grown for 10 days beforehand. The extraction was conducted as described above, but extracts were dissolved in only distilled water at 25 • C and kept at 4 • C until the analysis. Saffron aqueous extract effects were tested against the breast cell line MDA-MB-231. Cells were cultured in DMEM medium (4500 mg L −1 glucose) supplemented with 10% Fetal Bovine Serum (FBS), 100 U/mL penicillin, and 100 µg mL −1 streptomycin at 37 • C in an atmosphere of 5% CO 2 . For determining cell viability, the MTT (3-[4,5-Dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide) assay was performed as previously described [71]. Cells were plated at 5 × 10 3 cells per well in a 96-well plate and allowed to adhere to the plate overnight under the growth conditions described above. After 72 h of treatment, the culture medium was aspirated and the MTT solution was added, according to the manufacturer's instructions (SIGMA). After one hour of incubation, the MTT solution was removed and 100 µL of DMSO was added to dissolve MTT-formazan crystals. The absorbance of the converted dye was measured at a wavelength of 570 nm using an iMark microplate reader (BIORAD, Hercules, CA, USA). Data were normalized to the control value and expressed as mean ± standard deviation.
Data Acquisition and Statistical Analysis
All experiments were repeated 3 times and analyzed by one-way analysis of variance (ANOVA) using the SPSS-22 software. Differences among the means were tested at the significance level of p < 0.05 with the Duncan test. Statistical analyses on MTT data was performed by GraphPad software (Prism 5.03). Analysis of variance (ANOVA) and Dennett's post-hoc-test was applied to compare control with all treatments, Differences were considered statistically significant for values p < 0.05. Data are represented as mean ± standard deviation (SD).
Conclusions
We obtained from style-derived calli a cell suspension culture of saffron (Crocus sativus L.) producing relevant quantities of crocin together with total phenolics. Interestingly, saffron cell extracts also showed significant inhibitory effect on the viability of a breast cancer cell line. Both crocin and phenolics accumulation patterns after SA elicitation indicated their contribution to the increase of antioxidant activity. We investigated at a subcellular level the status of suspension cells, using protoplast transient transformation with heterologous vacuolar markers followed by confocal microscopy. We learned that vacuolar morphology was similar in all the cells, but endogenous autofluorescence revealed heterogeneity when considering metabolite production. Since a remarkable increase of total crocin and phenolic compounds were quantified in the spent medium of cell suspensions under SA elicitation, we performed microscopic observation of the effect induced by the highest SA concentration (1 mM). We observed that it caused stress and death within 24 h after elicitation when the exudation of high amounts of bioactive compounds occurred. Most of the crocin and phenolics production occurred during the SA elicited cell apoptosis within the first 24 h so that the biotechnological exploitation of SA elicitation may take place at any moment, but at the end of the growth curve, at day 7, the optimal biomass accumulation is obtained. On the other hand, reducing the elicitation time to 24 h before harvesting could be considered a high-valued innovation in terms of saving time in scaling up the process for metabolite production independently from saffron cell growth rate. | 2020-08-01T13:06:02.936Z | 2020-07-28T00:00:00.000 | {
"year": 2020,
"sha1": "be496628e514a22983424b5c51a22e9df75f0214",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2223-7747/9/8/949/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "993390cb05d61674ff56f4f47844fb42b1d83f5d",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Chemistry",
"Medicine"
]
} |
127051446 | pes2o/s2orc | v3-fos-license | Analysis of growth response of cool season cereals "wheat vs. rye" grown in organic and inorganic soils.
Plant growth response of wheat (Triticum aestivum L.) vs. rye (Secale cereale L.) to organic and inorganic soils was performed at 30, 60 and 90 days after emergence (DAE). A pot experiment was conducted at The Dryland Agriculture Institute, West Texas A&M University Texas, USA, during winter 2009-2010 using three organic soil mixtures (potting soil/compost) [Miracle Grow (MG), Sunshine Peat Moss (SPM), and Garden Basic Peat Humus (GBPM)], and two inorganic soils [Canyon Soil (CS) and Amarillo Soil (AS)]. The following measurements were performed to quantify plant growth: leaf area plant-1 (LAPP), leaf area expansion rate (LER), specific leaf area (SLA) , leaf area ratio (LAR), plant height, stem elongation rate (SER) , root length (RL), number of roots plant-1 (NRPP) number of tillers plant-1 (NTPP), stem girth, carbon exchange rate (CER), absolute growth rate (AGR), crop growth rate (CGR), and net assimilation rate (NAR) The experiment was performed in completely randomized design with three replicates. The objective of this experiment was to check whether growth response of wheat vs. rye differ under organic and inorganic soils? The results revealed that both crops responded differently in terms of growth at different crop growth stages under organic and inorganic soils. The planned mean comparison indicated that both wheat and rye grown under organic soils had higher LAPP, LER, SLA, LAR, plant height, SER, root length, NRPP, NTPP, AGR and CGR than crops grown under inorganic soils. Interestingly, the stem girth, CER and NAR declined under organic soils as compared with inorganic soils. The leaf characteristics (LAPP, LER, SLA and LAR) was greater under GBPM > MG > SPM > AS > CS, and rye performed better than wheat at different growth stages. Plant height, SER and root lengths was higher under MG > GBPM > SPM > AS > CS, however no significant differences were found between the two crops at different growth stages. The NRPP and NTPP were more under MG > GBPM > SPM > AS > CS, and rye performed better than wheat. The AGR and CGR values were more under MG > GBPM > SPM > AS > CS, rye performance was good at 60 DAE, and wheat ranked first at 30 and 90 DAE. Both AGR and CGR values were higher at 90 > 60 > 30 DAE indicated increase with the passage of time. The AGR and CGR showed positive relationship with increase in LAPP, LER, plant height, root length, NRPP and NTPP. The shoot to root ratio decreased with the passage of time and showed negative relationship with increase in root length.
INTRODUCTION
Organic agriculture is important for the improvement of the environmental conditions and human health (Kurtar and Ayan, 2004;Zengin, 2007). According to Plaster (2009), organic soils contain more than 20 to 30 % organic matter is mostly classifi ed as Histosols. Organic soils having high water holding capacity; hold water about 4 to 5 times of its weight and dries out very slowly than mineral soils. Organic soils are light weight because of the lower bulk density (0.2 to 0.3) than mineral soils (1.2 to 1.5). Organic soils generally have very good physical conditions than mineral soils. The organic soils are more porous, open, and easy to cultivate than mineral soils (Cowan, online). Amanullah (2014) reported that organic soils had more positive effects on the shoot and root development, dry matter portioning and accumulation, and water use effi ciency of wheat and rye than inorganic soils.
There is lack of research on crops growth response under organic and inorganic soils. The objective of this study was to study whether different soil types "organic vs. inorganic" infl uence the growth analysis [leaf area plant -1 (LAPP), leaf area expansion rate (LER), specifi c leaf area (SLA), leaf area ratio (LAR), plant height, stem elongation rate (SER), root length (RL), number of roots plant -1 (NRPP) number of tillers plant -1 (NTPP), stem girth, carbon exchange rate (CER), absolute growth rate (AGR), crop growth rate (CGR), and net assimilation rate (NAR)] of the two winter cereals "wheat vs. rye" at different growth stages or not?
After one week of emergence 15 plants were maintained per pot, and then fi ve plants were uprooted at 30, 60 and 90 days after emergence (DAE). The root were washed with tap water, and the plants were then divided into three parts i.e. roots, leaves and stems. The materials was put in paper bags and then in oven at 80°C for about 20 hours. The samples were weighing by electronic balance (Sartorius Basic, BA2105) and the average data on dry weight of root, leaf, and stem per plant was worked out. Shoot dry weight per plant was obtained by adding leaf dry weight with stem dry weight per plant. The sum of the shoot and root dry weight was called as total dry weight per plant. Plant height (shoot length) and root length was measured in cm with help of measuring tape. Plant height was divided by root length to get data on shoot-root ratio (by length). The SER (stem elongation rate), LER (leaf area expansion rate), mean single leaf area (MSLA), leaf area plant -1 (LAPP), specifi c leaf area (SLA) and leaf area ratio (LAR) were determined using the following formulae: Where: H 1 = Plant height (cm) at the beginning of interval H 2 = Plant height (cm) at the end of interval LAPP 1 = Leaf area per plant (cm 2 ) at the beginning of interval LAPP 2 = Leaf area per plant (cm 2 ) at the end of interval t 2 -t 1 = The time interval between the two consecutive samplings Absolute growth rate (AGR): dry matter accumulation per plant per unit time; crop growth rate (CGR): dry matter accumulation per unit ground area per unit time; and net assimilation rate (NAR): dry matter accumulation per unit leaf area per unit time was determined using the following formulae: Where: W 1 = Dry weight per plant at the beginning of interval W 2 = Dry weight per plant at the end of interval t 2 -t 1 = The time interval between the two consecutive samplings GA = Ground area occupied by plants at each sampling LAI = Leaf area index (leaf area per plant divided by ground area per plant)
Statistical analysis
Data were subjected to analysis of variance (ANOVA) according to the methods described in Steel and Torrie (1980) and treatment means were compared using the least signifi cant difference (LSD) at P ≤ 0.05.
Leaf area plant -1
Leaf area plant -1 (LAPP) was calculated as the product of mean single leaf area and number of leaves plant -1 . Rye produced signifi cantly higher LAPP of 43.8 cm 2 than wheat (29.3 cm 2 ) at 30 DAE (Table 1). The highest LAPP (87.2 cm 2 ) was found when crops were grown in GBPH, followed by 81.1 cm 2 in MG, and the LAPP of rye was higher than wheat in both GBPH and MG. The minimum LAPP of 1.7 cm 2 was recorded when crops were grown in Canyon soil. At 60 DAE, rye had higher LAPP (151.2 cm 2 ) than wheat (101.4 cm 2 ). The highest LAPP (324.1 cm 2 ) was obtained when crops were grown in GBPH, followed by MG (285.5 cm 2 ). In both GBPH and MG, rye had signifi cantly higher LAPP than wheat. The LAPP reduced to minimum in Canyon soil (3.7 cm 2 ), but no signifi cant differences were observed in the LAPP of the two crops. At 90 DAE, rye had relatively higher LAPP (251.7 cm 2 ) than wheat (198.1 cm 2 ). The highest LAPP (470.8 cm 2 ) was noted when crops were grown in GBPH, being at par with MG (468.9 cm 2 ), and the LAPP of rye was higher than wheat in both soils. The planned mean comparison indicated that crops grown under organic soils produced higher LAPP than crops grown under inorganic soils. The LAPP was higher at 90 DAE than at 30 and 60 DAE (90 > 60 > 30 DAE).
Leaf area expansion rate
Leaf area expansion rate (LER) was calculated as LAPP divided by number of days taken. The LER of rye (1.459 cm 2 plant -1 day -1 ) was signifi cantly (P ≤ 0.05) higher than wheat (0.977 cm 2 plant -1 day -1 ) at 30 DAE ( Table 2). The highest LER (2.905 cm 2 plant -1 day -1 ) was calculated when crops were grown in GBPH, being at par with MG (2.702 cm 2 plant -1 day -1 ), and the LER of rye was higher than wheat in both GBPH and MG. The minimum LER (0.055 cm 2 plant -1 day -1 ) was recorded when crops were grown in Canyon soil. At 60 DAE, rye had higher LER (2.52 cm 2 plant -1 day -1 ) than wheat (1.69 m 2 plant -1 day -1 ). The highest LER (5.40 cm 2 plant -1 day -1 ) was noted when crops were grown in GBPH, being at par with MG (4.76 cm 2 plant -1 day -1 ). In both GBPH and MG, rye had signifi cantly higher LER than wheat. The LER reduced to minimum in Canyon soil (0.06 cm 2 plant -1 day -1 ), but no signifi cant differences were observed in the LER of the two crops. At 90 DAE, rye had relatively higher LER (2.80 cm 2 plant -1 day -1 ) than wheat (2.20 cm 2 plant -1 day -1 ). The highest LER (5.23 cm 2 plant -1 day -1 ) was noted when crops were grown in GBPH, being at par with MG (5.21 cm 2 plant -1 day -1 ), and the LER of rye was higher than wheat in both soils. The LER reduced to minimum in Canyon soil (0.48 cm 2 plant -1 day -1 ); however, there was no difference in LER of wheat and rye. The planned mean comparison indicated that crops grown under organic soils produced higher LER than crops grown under inorganic soils. The LER was higher at 90 DAE than at 30 and 60 DAE (90 > 60 > 30 DAE).
Specifi c leaf area
Specifi c leaf area (SLA) was calculated as leaf area plant -1 divided by leaf dry weight plant -1 . According to Craufurd et al. (1999), plants with higher SLA (thicker leaves) usually have higher densities of chlorophyll per unit area, and hence have greater photosynthetic capacities than thinner leaves. The SLA of rye was higher (1.28 cm 2 mg -1 ) than wheat (0.76 cm 2 mg -1 ) at 30 DAE (Table 3). The highest SLA (2.26 cm 2 mg -1 ) was calculated when crops were grown in GBPH, followed by MG (1.63 cm 2 mg -1 ), and the SLA of rye was higher than wheat in both GBPH and MG. The minimum SLA of 0.19 cm 2 mg -1 was recorded when crops were grown in Canyon soil. At 60 DAE, rye had relatively higher SLA (0.51 cm 2 mg -1 ) than wheat (0.44 cm 2 mg -1 ). The highest SLA (0.85 cm 2 mg -1 ) was obtained when crops were grown in GBPH, followed by MG (0.68 cm 2 mg -1 ). In GBPH and MG, rye had relatively higher SLA than wheat. The SLA reduced to minimum of 0.15 cm 2 mg -1 in Canyon soil, no signifi cant differences were observed in the SLA of the two crops. At 90 DAE, rye had higher SLA (0.69 cm 2 mg -1 ) than wheat (0.40 cm 2 mg -1 ). The highest SLA (0.98 cm 2 mg -1 ) was calculated when crops were grown in SPM, followed by GBPH (0.55 cm 2 mg -1 ), and the SLA of rye was relatively higher than wheat in both soils. The SLA reduced to minimum (0.33 cm 2 mg -1 ) when crops were grown in Canyon soil; however, there was no difference in SLA of wheat and rye when grown under Canyon soil. The planned mean comparison indicated that crops grown under organic soils produced higher SLA than crops grown under inorganic soils. The SLA was higher at 30 DAE than at 60 and 90 DAE.
Leaf area ratio
Leaf area ratio (LAR) was calculated as leaf area plant -1 divided by total dry weight plant -1 . The LAR of rye was higher (0.74 cm 2 mg -1 ) than wheat (0.43 cm 2 mg -1 ) at 30 DAE (Table 4). The highest LAR (1.38 cm 2 mg -1 ) was calculated when crops were grown in GBPH, followed by MG (0.93 cm 2 mg -1 ), and the LAR of rye was higher than wheat in both GBPH and MG. The minimum LAR (0.07 cm 2 mg -1 ) was recorded when crops were grown in Canyon soil. At 60 DAE, rye had higher LAR (0.28 cm 2 mg -1 ) than wheat (0.23 cm 2 g -1 ). The highest LAR (0.50 cm 2 mg -1 ) was obtained when crops were grown in GBPH, followed by MG (0.40 cm 2 mg -1 ). In GBPH and MG, rye had relatively higher LAR than wheat. The LAR of crops reduced to minimum (0.05 cm 2 mg -1 ) when grown in Canyon soil, but no signifi cant differences were observed in the LAR of the two crops. At 90 DAE, rye had higher LAR (0.31 cm 2 mg -1 ) than wheat (0.19 cm 2 mg -1 ). The highest LAR (0.42 cm 2 mg -1 ) was calculated when crops were grown in SPM, and the LAR of rye was relatively higher than wheat crop. The LAR reduced to minimum (0.17 cm 2 mg -1 ) each in Canyon soil and MG; and the LAR of rye was higher than wheat in the two soils. The planned mean comparison indicated that crops grown under organic soils produced higher LAR than crops grown under inorganic soils. The LAR was higher at 30 DAE than at 60 and 90 DAE.
Plant height
There was no signifi cant (P ≤ 0.05) difference in heights of wheat and rye at 30 days after emergence (DAE) as shown in Table 5. Among the soil types, tallest plants (34.8 cm) were produced when crops were grown in miracle grow (MG) being at par with 33.7 cm heights recorded in garden basic peat humus (GBPH), and the heights of rye was higher than wheat in both soil. At second cut (60 DAE), there was no signifi cant difference in heights of wheat and rye. Among the soil types, tallest plants (41.0 cm) were produced when crops were grown in MG, followed by 38.8 cm in GBPH, and the heights of rye was higher than wheat in both soils. At third cut (90 DAE), there was no signifi cant difference in heights of wheat and rye. Tallest plants (51.2 cm) were produced when crops were grown in GBPH, being at par with MG (50.0 cm), and the height of rye was higher than wheat in both soils. The planned mean comparison indicated
Stem elongation rate
Stem elongation rate (SER) was calculated as plant height divided by number days taken. There were no signifi cant differences in stem elongation rate (SER) of wheat (0.77 cm day -1 ) and rye (0.75 cm day -1 ) at 30 DAE ( Table 6). The highest SER (1.16 cm day -1 ) was calculated when crops were grown in MG, being at par with GBPH (1.12 cm day -1 ), and the SER of rye was higher than wheat in both MG and GPM. At 60 DAE, there were no signifi cant differences in SER of wheat and rye (0.47 cm day -1 each). The highest SER (0.68 cm day -1 ) was recorded when crops were grown in MG, being at par with GBPH (0.65 cm day -1 ), and the SER of rye was higher than wheat in both soils. At 90 DAE, there was also no signifi cant difference in SER of wheat and rye. The highest SER (0.57 cm day -1 cm ) was noted when crops were grown in GBPH, being at par with MG (0.56 cm day -1 ), and the SER of rye was relatively higher than wheat in both GBPH and MG. The planned mean comparison indicated that crops grown under organic soils had higher SER than crops grown under inorganic soils. The SER in contrast to plant height, decreased with passage of time (30 > 60 > 90 DAE).
Root length
There were no signifi cant differences in the root length of wheat and rye at 30, 60 and 90 DAE (Table 7). At 30 DAE, longer roots (15.3 cm) were produced when crops were grown in MG, being at par with SPM (14.2 cm), however, in MG, rye roots were signifi cantly longer than that of wheat; incontrast, wheat roots were signifi cantly longer than rye when grown in SPM. At 60 DAE, longer roots (18.5 cm) were recorded when crops were grown in MG, being at par with SPM (17.5 cm, and the rye roots were longer (19.3 cm) than wheat (17.7 cm) when grown in MG, in contrast, wheat roots were longer (18.7 cm) than rye (17.7 cm) when grown in SPM. At 90 DAE, longer roots (34.0 cm) were produced by the crops when grown in SPM, followed by MG (26.7 cm), and rye roots were relatively longer than wheat in both SPM and MG. The planned mean comparison indicated that crops grown under organic soils produced longer roots than crops grown under inorganic soils. The root lengths increased with the passage of time (90 > 60 > 30 DAE).
Shoot root ratio (plant height ÷ root length)
There was no signifi cant difference in shoot: root of wheat (3.2) and rye (3.0) at 30 DAE (P ≤ 0.05) shown in Table 8. Among soil types, the highest shoot: root (4.0) was calculated when crops were grown in Canyon soil, followed by 3.7 in Amarillo soils, and the shoot: root of wheat was relatively higher than rye in both soils. The minimum shoot: root (2.1) was recorded when crops were grown in SPM, being at par with MG (2.3). At second cut, there was no signifi cant difference in shoot: root of wheat (2.6) and rye (2.4). Among soil types, the highest shoot: root (3.5) was obtained when crops were grown in Amarillo soil, followed by GBPH (2.7). In Amarillo soil wheat had higher shoot: root of 4.0 than rye (2.9), but in GBPG rye had higher shoot: root (3.0) than wheat (2.5). The shoot: root reduced to minimum (2.0) when crops were grown in SPM, but no signifi cant differences were observed in the shoot: root of the two crops under SPM. At third cut (90 DAE), rye had relatively higher shoot: root (2.1) than rye (1.9). Among soil types, the highest shoot: root (2.6) was obtained when crops were grown in GBPH, and the shoot: root of rye (2.9) was signifi cantly higher than wheat (2.3) under GBPH. The shoot: root reduced to minimum (1.2) when crops were grown in SPM; however, there was no difference in shoot: root of wheat and rye when grown under SPM.
Number of roots plant -1
Number of roots plant -1 (NRPP) of rye were more (6.1) than wheat (5.2) at 30 DAE (Table 9). The highest NRPP (10.8) was calculated when crops were grown in GBPH, followed by 7.7 in MG, and the NRPP of rye (11.7) were higher than wheat (10.0) in GBPH. The minimum NRPP was reported in Canyon soil (2.5). At 60 DAE, rye had higher NRPP (9.6) than wheat (7.6). The highest NRPP (15.8) was obtained when crops were grown in GBPH, followed by MG (14.3). In both GBPH and MG, rye had higher NRPP than wheat. The NRPP reduced to minimum of 3.0 in Canyon soil, but no signifi cant differences were observed in the NRPP of the two crops. At 90 DAE, wheat had relatively higher NRPP (15.0) than rye (11.6). The highest NRPP (22.9) were noted when crops were grown in MG, followed by GBPH (20.6), and the NRPP of wheat were higher than rye both soils. The planned mean comparison indicated that crops grown under organic soils produced more NRPP than crops grown under inorganic soils. The NRPP increased with the passage of time (90 > 60 > 30 DAE).
Number of tillers plant -1
Rye had signifi cantly higher (6.78) number of tillers plant -1 (NTPP) than wheat (3.50) at 90 DAE (Table 10). The highest NTPP (9.00) was noted when crops were grown in MG, being at par with GBPH (8.67 cm 2 plant -1 day -1 ), and the NTPP of rye was signifi cantly higher than wheat in both soils. The NTPP reduced to minimum (2.33) in Canyon soil; however, there was no difference in NTPP of wheat and rye when grown in Canyon soil. The planned mean comparison indicated that crops grown under organic soils produced more NTPP than crops grown under inorganic soils.
Stem girth
Wheat had signifi cantly more stem girth (2.38 mm) than rye (1.79 mm) at 90 DAE (Table 10). The highest stem girth (2.68 mm) was noted when crops were grown in AMAS, being at par with MG (2.55 mm), and the stem girth of wheat was signifi cantly higher than rye in both soils. Interestingly, the planned mean comparison indicated that crops grown under organic soils had less stem girth (2.18 mm) than crops grown under inorganic soils (2.39 mm).
Absolute growth rate
Absolute growth rate (AGR) was calculated as the ratio of dry matter accumulation per plant per day. The AGR of wheat (1.92 mg plant -1 day -1 ) was higher than rye (1.67 mg plant -1 day -1 ) at 30 DAE (Table 11). The highest AGR (3.48 mg plant -1 day -1 ) was produced when crops were grown in MG, followed GBPH (2.45 mg plant -1 day -1 ), and the AGR of wheat was higher than rye in both MG and GBPH. At 60 DAE, rye (10.27 mg plant -1 day -1 ) had higher AGR than wheat (8.51 mg plant -1 day -1 ). The highest AGR (20.16 mg plant -1 day -1 ) was obtained when crops were grown in MG, being at par with GBPH (19.43 mg plant -1 day -1 ), and the AGR of rye at this stage was higher than wheat in both MG and GBPM. At 90 DAE, wheat had higher AGR (30.67 mg plant -1 day -1 ) than rye (19.62 mg plant -1 day -1 ). The highest AGR (69.82 mg plant -1 day -1 ) was obtained when crops were grown in MG, followed by GBPH (45.90 mg plant -1 day -1 . The planned mean comparison indicated that crops grown under organic soils had higher AGR than crops grown under inorganic soils. The AGR increased with the passage of time (90 > 60 > 30 DAE).
Crop growth rate
Crop growth rate (CGR) was calculated as the ratio of dry matter accumulation per unit ground area per day. The CGR of wheat (0.86 g m -2 day -1 ) was higher than rye (0.75 g m -2 day -1 ) at 30 DAE (Table 12). The highest CGR (1.56 g m -2 day -1 ) was produced when crops were grown in MG, followed by GBPH (1.10 g m -2 day -1 ), and the CGR of wheat was higher than rye in both MG and GBPH. At 60 DAE, rye (2.85 g m -2 day -1 ) had higher CGR than wheat (2.29 g m -2 day -1 ). The highest CGR (5.59 g m -2 day -1 ) was obtained when crops were grown in MG, being at par with GBPH (5.50 g m -2 day -1 ), and the CGR of rye was higher than wheat in both MG and GBPM. At 90 DAE, wheat had higher CGR (9.20 g m -2 day -1 ) than rye (5.89 g m -2 day -1 ). The highest CGR (20.95 g m -2 day -1 ) was obtained when crops were grown in MG, followed by GBPH (13.77 g m -2 day -1) The planned mean comparison indicated that crops grown under organic soils had higher CGR than crops grown under inorganic soils. The CGR increased with the passage of time (90 > 60 > 30 DAE).
Net assimilation rate
Net assimilation rate (NAR) was calculated as the ratio of dry matter accumulation per unit leaf area per day. The NAR of wheat (42.42 g m -2 day -1 ) and rye (42.02 g m -2 day -1 ) was almost the same at 30 DAE (Table 13). The highest NAR (116.95 g m -2 day -1 ) was produced when crops were grown in Canyon soil, and there were no signifi cant differences in the NAR of wheat and rye. The minimum NAR (6.76 g m -2 day -1 ) was calculated when crops were grown in GBPH, and wheat (9.41 g m -2 day -1 ) had higher NAR than rye (4.10 g m -2 day -1 ). At 60 DAE, rye (13.74 g m -2 day -1 ) had higher NAR than wheat (12.44 g m -2 day -1 ), although the differences in the crops were not signifi cant. The highest NAR (29.84 g m -2 day -1 ) was obtained when crops were grown in Canyon soil, and the NAR of rye was higher than wheat. The NAR reduced to minimum (5.54 g m -2 day -1 ) when crops were grown in GBPH, and no significant differences was observed in the NAR of two crops when grown in GBPH. At 90 DAE, wheat had higher NAR (27.59 g m -2 day -1 ) than rye (23.12 g m -2 day -1 ), but the differences were not signifi cant. The highest NAR (58.62 g m -2 day -1 ) was obtained when crops were grown in MG, followed by GBPH (37.03 g m -2 day -1 ), and the NAR of rye at this stage was higher than wheat in both MG and GBPM. The NAR reduced to minimum (7.62 g m -2 day -1 ) when crops were grown in SPM, and no signifi cant differences was observed in the NAR of wheat (8.33 g m -2 day -1 ) and rye (6.91 g m -2 day -1 ) when grown in SPM. The planned mean comparison indicated that crops grown under organic soils had higher NAR than crops grown under inorganic soils at 90 DAE, while organic soils had less NAR than crops grown under inorganic soils at 30 and 60 DAE.
Wheat vs. Rye
No signifi cant differences were found between wheat and rye for plant height, SER, root length, root to shoot ratio, CER and NAR. Although, earlier research (Amanullah et al., 2011), signifi cant differences were found in plant height of wheat and barley cultivars. The LAPP and LER values were higher for rye than wheat at the two early growth stages (30 and 60 DAE), and the increase was higher under organic soils. The increase in LAPP and LER of rye over wheat was due to the increase in number of tillers and leaves per plant of rye over wheat. Van den According to Bultynck et al. (2004), crop species with more rapidly elongating leaves showed a faster increase in LER have more biomass allocation to leaf sheaths and less to roots. Boogaard et al. (1996) suggested that a fast LER in wheat was positively correlated with above-ground biomass and grain yield. The SLA and LAR values were also higher for rye than wheat and the increase was more under organic soils. The increase in SLA and LAR of rye over wheat was due to the increase in LAPP of rye over wheat. The NRPP at 30 and 60 DAE, and NTPP at 90 DAE was signifi cantly higher for rye than wheat. The differences in the NRPP and NTPP of the two crops may be due to the difference in their genetic makeup. In our earlier research (Amanullah et al., 2011), we also found signifi cant differences in the number of tillers m -2 of wheat and barley cultivars. Wheat had signifi cantly thicker stem (tiller) girth than rye at 90 DAE. The differences in the stem girths of the two crops may also be due to the difference in the genetic makeup of both crops. Stem girth showed negative relationship with increase in NTPP. Wheat had higher AGR and CGR values than rye at 30 and 90 DAE. However, the AGR and CGR values were higher for rye than wheat at 60 DAE. The differences in the AGR and CGR values at different growth stages of both crops was due to the differences in the total dry matter accumulation (Amanullah, 2014).
Organic vs. Inorganic soils
The four parameters under study viz. shoot to root ratio, stem girth, CER and NAR had higher mean values under inorganic soils as compared to organic soils. Indicating the shoot to root ratio, stem girth, CER and NAR had negative relationship with LAPP, plant height, root length, NRPP, and NTPP. The higher shoot to root ratio (plant height divided by root length) under inorganic soil was attributed to the shorter root lengths under inorganic soils. In contrast, the longer roots of the crops under organic soils at different growth stages, on the other hand reduced the shoot to root ratio (by length) in both crops. Plants under inorganic soils were in stress than plants under organic soil. The stress condition under inorganic soil reduced the root lengths and thereby increased shoot to root ratio under inorganic soil. These results are in conformity with the results obtained by Eghball and Maranville (1993) and Amanullah et al. (2015). The increase in the stem girth under inorganic soils showed negative relationship with increase in NTPP and NRPP. In contrast, both crops under organic soils had signifi cantly more NTPP having thinner stem girths. Crops with higher CER under inorganic soil than organic soils was not too much clear and we recommend further research to confi rm the difference in CER of crops under different soil types. The increase in NAR under inorganic soils was due to the decrease in LAPP, leaf area index, CGR and total dry matter accumulation under inorganic soils (Amanullah, 2014). Amanullah and Stewart (2013) reported that NAR of oats had negative relationship with increase in leaf area index and positive relationship with increase in CGR.
The LAPP, LER, SLA, LAR, plant height, SER, root length, NRPP, NTPP, AGR and CGR were signifi cantly better (higher) under organic soils than inorganic soils. Improvement in all these parameters under organic soils probably may be due to high water holding capacity; lower bulk density and better physical soil conditions (Cowan, online) as well as carry more nitrogen (Pedosphere, 2001) in organic soils. Moreover, the published research from this study confi rmed that water use effi ciency, dry matter accumulation in roots and shoots was signifi cantly higher in both crops under organic soils (Amanullah, 2014) that had positive impact on crop growth analysis. On the other hand, the restricted root and shoot development of both crops and less water use effi ciency under inorganic soils (Amanullah, 2014) had adverse effects on the growth analysis of both crops. Amanullah and Stewart (2013) suggested that increase in total dry matter accumulation per plant had positive relationship with AGR, CGR and NAR in oats.
CONCLUSIONS
Wheat and rye responded differently in growth under different soil types. Both crops had better performance in terms of higher leaf area per plant, leaf area expansion rate, specifi c leaf area, leaf area ratio, plant height, stem elongation rate, root length, number of roots per plant, number of tillers per plant, absolute growth rate and crop growth rate under organic soils as compared with inorganic soils at different growth stages. Interestingly, the stem girth, carbon exchange rate and net assimilation rate in both crops was declined under organic soils as compared with inorganic soils. Further research is needed to confi rm the reduction in stem girth, carbon exchange rate and net assimilation rate under inorganic soils than organic soils.
ACKNOWLEDGMENTS
Prof. Dr. Paigham Shah, The University of Agriculture Peshawar, Pakistan for the statistical analysis of the data, and Dryland Agriculture Institute, West Texas A&M University, Texas, United States for technical support are greatly acknowledged.
Funding
Financial support from the Higher Education Commission, Islamabad, Pakistan is gratefully acknowledged. | 2019-04-23T13:28:45.270Z | 2015-01-01T00:00:00.000 | {
"year": 2015,
"sha1": "6122638c2c909756bd327eeaeb55c471751e4982",
"oa_license": "CCBY",
"oa_url": "http://www.ejfa.me/index.php/journal/article/download/702/511",
"oa_status": "HYBRID",
"pdf_src": "Adhoc",
"pdf_hash": "f14ab4fe4d8a0b65b7c1e109112fe5d391e39b94",
"s2fieldsofstudy": [
"Agricultural and Food Sciences"
],
"extfieldsofstudy": [
"Biology"
]
} |
248001196 | pes2o/s2orc | v3-fos-license | Two adolescent cases of acute tubulointerstitial nephritis after second dose of COVID-19 mRNA vaccine
ABSTRACT The Food and Drug Administration (FDA) expanded the emergency use authorization for the BNT162b2 messenger RNA (mRNA) vaccine (Pfizer–BioNTech) for children aged 12–15 years on 10 May 2021. To date, less than a year has passed since vaccination against COVID-19 has been used in children and adolescents, and the overall effects and safety of these vaccines are still being assessed. The BNT162b2 vaccine originally had a favorable profile in 12–17-year-old recipients compared with older ages, and no serious adverse events had previously been reported. Despite various adverse events, the benefit of reducing the infection rate or the frequency of severe COVID-19 has been evaluated to outweigh the harm caused by COVID-19 vaccination. Additionally, several cases of sudden development of new-onset or relapsing glomerular diseases, including acute kidney injury (AKI), have been reported in adults following the BNT162b2 SARS-CoV-2 mRNA vaccine. Herein, we present two cases of adolescents who developed AKI following the second administration of the BNT162b2. These are the first pediatric cases of acute tubulointerstitial nephritis temporarily linked to SARS-CoV-2 vaccination.
Introduction
Coronavirus disease 2019 (COVID-19)-associated hospitalization rates increased in March and April 2021. Nearly one-third of them required pediatric intensive care unit admission, and 5% needed invasive mechanical ventilation. 1 There is potential for serious disease or death in adolescents, hence vaccination of adolescents is warranted. Accordingly, on 10 May 2021, the Food and Drug Administration (FDA) expanded the emergency use authorization (EUA) for the BNT162b2 messenger RNA (mRNA) vaccine (Pfizer-BioNTech) for children aged 12-15 years. 2 The BNT162b2 vaccine originally had a favorable profile in 12-17-year-old recipients compared with older ages, and no serious adverse events had previously been reported. 3,4 However, various COVID-19 vaccine associated adverse events, including myocarditis in adolescents, have been reported since these vaccinations have become more common worldwide. 5,6 Among the solicited systemic adverse events of BNT162b2 vaccination, headache, and fatigue are common, while gastrointestinal symptoms, such as vomiting (1-3%) and diarrhea (4-11%), are uncommon, regardless of age. 3,4 A similar distribution of adverse events occurred in Korean adolescents. 7 Additionally, several cases of sudden development of newonset or relapsing glomerular diseases, including acute kidney injury (AKI) have been reported in adults following the BNT162b2 SARS-CoV-2 mRNA vaccine. 8 Herein, we present two cases of adolescents who developed AKI following the second administration of the BNT162b2. These are the first pediatric cases of acute tubulointerstitial nephritis temporarily linked to SARS-CoV-2 vaccination.
Case 1
A previously healthy 17-year-old man presented to our hospital with epigastric pain and poor oral intake for 3 days, and abnormal creatinine (Cr) serum levels following the second dose of the BNT162b2 SARS-CoV-2 vaccine.
The patient received the second dose of COVID-19 vaccine on 8 November 2021. The next day, he had myalgia and a fever of 38℃; therefore, he underwent a polymerase chain reaction test for COVID-19 on 10 November 2021. The result was negative. He denied taking any medication, including herbal teas, except for one dose of acetaminophen for fever that occurred after vaccination, in the past few months. He denied any cardiorespiratory symptoms.
From 11 November 2021, the patient had problems eating because of epigastric pain and nausea, and he visited a nearby hospital emergency room. He was noted to have renal insufficiency, with a serum Cr level of 3 mg/dL determined by a blood test.
On 13 November 2021, the patient was transferred to our hospital because of renal insufficiency and persistent gastrointestinal problems. Upon physical examination, he was acutely ill-looking with a blood pressure of 150/85 mmHg, heart rate of 75 beats/min, respiratory rate of 20 breaths/min, and body temperature of 36.9℃ with a peripheral oxygen saturation of 100% at room air. His weight was 95 kg (>99th percentile), height 178 cm (75-90th percentile) and BMI 30.0 kg/m 2 (>99th percentile). Epigastric tenderness and costovertebral angle tenderness were positive on both sides. Table 1 shows the initial laboratory findings on day 1. Additional blood tests were performed. The erythrocyte sedimentation rate (ESR) was 43 mm/h. Complement 3 (C3) and 4 (C4) serum levels were 136 mg/dL and 31 mg/dL, respectively, both within the normal range. Results for antinuclear antibody (ANA) and antineutrophil cytoplasmic antibody (ANCA) tests were negative, and immunoglobulin G (IgG), A (IgA), and M (IgM) levels were 846, 166, and 78 mg/dL, respectively. D-dimer level was 2.19 mg/L FEU (normal range, ≤0.5 mg/L FEU). Prothrombin time (PT) and activated partial thromboplastin time (aPTT) were 1.09 INR (normal range, 0.88-1.20 INR) and 29 s (normal range, 20.0-36.0 s), respectively Urine microscopic findings were negative, and the spot urine protein:Cr ratio (PCR) was 0.10 g/g·Cr. Urine N-acetylβ-D-glucosaminidase (NAG):Cr was 2.34 IU/g·Cr (normal range, ≤5.61 IU/g·Cr) and β2-microglobulin (β2MG) level was 0.29 mg/L (normal range, ≤0.30 mg/L) ( Table 1). Chest and abdominal radiography revealed no specific findings. Kidney ultrasonography on hospital day 3 revealed normalsized kidneys with normal echogenicity and normal range of renal arterial resistive index values.
Electrocardiography showed marked sinus arrhythmia with normal findings. Gastroduodenoscopy revealed no mucosal abnormality with a small amount of bile juice in the stomach and no ulcerative lesions in the duodenum.
A renal biopsy was performed on day 3 of hospitalization. The glomerulus appeared slightly larger and segmentally hypercellular, involving mesangial cells. Interstitial infiltrates were mainly mononuclear. Focal and moderate interstitial fibrosis and tubular atrophy were noted in approximately 20% of the renal cortices. His blood vessels are unremarkable. Immunofluorescence was negative for IgG, IgM, IgA, C3, C1q, C4, fibrinogen, and kappa and lambda light chains. Electron microscopy revealed a non-significant focal epithelial foot process effacement with no electron-dense deposits. The glomerular basement membrane showed focal wrinkling with partly irregular inner contours ( Figure 1). These findings were consistent with those of acute tubulointerstitial nephritis. Renal insufficiency gradually improved with supportive care, oral intake increased, and he was discharged after 1 week.
Case 2
A previously healthy 12-year-old man presented to our hospital with 3-week persistent anorexia, nausea, vomiting, and weight loss of 7 kg, followed by renal insufficiency with a Cr level of 2.28 mg/dL.
The patient received a second dose of COVID-19 vaccine on 24 December 2021. The next day, he began to experience nausea, vomiting with gastric juice approximately four times per day, and poor oral intake. He visited a nearby clinic because of persistent anorexia and vomiting. A blood test was performed, and an abnormal Cr level of 2.28 mg/dL was noted on 11 January 2022.
On 17 January 2022, the patient was admitted to our hospital because of renal insufficiency and dehydration with a weight loss of 7 kg. He denied taking any medication, including herbal teas, except for 3 days of medication with trimebutine and domperidone for vomiting that occurred after vaccination, in the past few months. He denied any cardiorespiratory symptoms. There was no family history of chronic kidney disease or hypertension. Upon physical examination, he was acutely illlooking and dehydrated with a blood pressure of 139/87 mmHg, heart rate of 127 beats/min, respiratory rate of 20 breaths/min, and body temperature of 36.7℃ with a peripheral oxygen saturation of 97% at room air. His weight was 56.3 kg (50-75th percentile), height 148.9 cm (10-25th percentile) and BMI 25.4 kg/m 2 (90-95th percentile). Epigastric tenderness was positive, without costovertebral angle tenderness. Table 1 shows the initial laboratory findings. Additional blood tests were performed. Venous blood gas analysis showed metabolic acidosis (pH 7.261 and HCO 3 − 16.3 mmol/L) with an anion gap of 17.7. C3 and C4 serum levels were 120 mg/dL and 13 mg/ dL, respectively, both within normal range. ANA, c-ANCA and p-ANCA were all negative, and the IgG, IgA, and IgM levels were 2,834, 557, and 250 mg/dL, respectively. PT and aPTT were 1.29 INR and 31s, respectively. D-dimer level was 0.91 mg/L FEU and lactate dehydrogenase (LDH) was 157 IU/L, respectively. Cardiac enzymes, including creatine kinase MB isoenzyme and troponin-I, were within the normal range, except for the pro-brain natriuretic peptide level of 237.6 pg/ml (normal range, 0-125 pg/ml).
Urine microscopic findings noted pyuria with white blood cells of 10-19/high power field (HPF) and casts of >21/low power field (LPF). The spot urine PCR was 1.95 g/ g·Cr, and spot urine calcium:Cr ratio was 0.19. Proteinuria during the 24-h urine collection was 861 mg/day. Urine NAG: Cr was 32.61 IU/g·Cr, and β2MG level was 2.85 mg/L (Table 1). Chest and abdominal radiography revealed no specific findings. Kidney ultrasonography on hospital day 2 revealed normal-sized kidneys with slightly increased echogenicity and a normal range of renal arterial resistive index values. Electrocardiography revealed sinus arrhythmia with a nonspecific ST abnormality. Gastroduodenoscopy revealed no mucosal abnormality in the stomach or duodenum.
A renal biopsy was performed on day 5 of hospitalization. Immunofluorescence was negative for IgG, IgM, IgA, C3, C1q, C4, fibrinogen, and kappa and lambda light chains. The glomeruli were normal in size and cellularity. Tubules revealed severe necrosis, tubulorrhexis, and loss, with heavy infiltration of neutrophils, eosinophils, and mononuclear cells in the interstitium. Blood vessels were unremarkable. Ultrastructurally, the glomerular basement membrane thickness was normal with relatively smooth contours. No electron-dense deposits were observed. Epithelial cell foot processes showed slight focal effacement (Figure 2).
After receiving oral steroid treatment on day 10 of hospitalization, the patient showed remarkable improvement in renal insufficiency (Figure 3).
Discussion
In the two cases described, Cr elevation was confirmed, starting with nonspecific digestive symptoms including vomiting, after the second dose of vaccine against COVID-19. In the first case, Cr elevation recovered rapidly, but in the second case, the Cr did not return to normal, and tubulopathy was continuously noted by urinalysis (Figure 3). The common feature of both cases was that they were both male, had high BMIs, and their initial symptom presentation appeared after the second dose of BNT162b2 mRNA vaccine.
Despite various adverse events, the benefit of reducing the infection rate or the frequency of severe COVID-19 has been evaluated to outweigh the harm caused by COVID-19 vaccination. In adults, the population vaccinated against COVID-19 is large and sufficient information about reported adverse reactions has been provided. Associations between kidney disease and adult receipt of BNT162b2 vaccination are varied. These include minimal change in disease with AKI, 9 and IgA nephropathy. 10 A case with acute interstitial nephritis in a 45-year-old female patient was recently reported. 11 Similar to our cases, symptoms developed after inoculation with the second dose of the BNT162b2 SARS-CoV-2 vaccine. The patient was treated with corticosteroids and hemodialysis, which led to progressive recovery of renal function. Mira et al. suggested that patient might develop an acute reaction to the BNT162b2 based on immunophenotyping results, which could be useful in diagnosing hypersensitivity to the vaccine and polyethylene glycol excipient. 11 However, we have not been able to prove drug hypersensitivity in our cases. Based on that study, we concluded that the acute interstitial nephritis in our cases was related to BNT162b2 vaccination because of the coincidence of symptom onset and similar pathological findings.
Since December 2020, the Pfizer-BioNTech COVID-19 vaccine has been available under EUA for vaccination of individuals aged ≥16 years, and the authorization was expanded to include those 12-15 years of age in May 2021. It was the first COVID-19 vaccine to be approved by the FDA on 23 August 2021. To date, less than a year has passed since vaccination against COVID-19 has been used in children and adolescents, the overall effects and safety of these vaccines are still being assessed. Information regarding adverse reactions is therefore limited. However, the new development of nephrotic syndrome after the first dose of BNT162b2 in 15-year-old Japanese boy has been reported. 12 In addition, two pediatric patients with IgA nephropathy presented with macroscopic hematuria within 24 h following the second dose of the Pfizer-BioNTech COVID-19 vaccine. In both cases, new-onset gross hematuria was self-resolved; renal function was recovered spontaneously in one patient, while the other recovered only after receiving methylprednisolone pulses. 13 Evidence for drug-induced interstitial nephritis caused by the COVID-19 vaccine is limited. However, other adverse events, including interstitial nephritis caused by influenza vaccination, have also been reported. Moreover, the COVID-19 pandemic may impede patients' access to general medical care and delay diagnosis of undiagnosed underlying diseases. It is also possible that our cases had a previously undiagnosed kidney disease rather than new-onset glomerular disease resulting from COVID-19 vaccination. Monitoring of renal function is therefore necessary after inoculation with a COVID-19 vaccine, and further research and discussion is necessary. Future reports of similar cases will facilitate the discussion regarding possible major adverse events associated with SARS-CoV-2 vaccines. | 2022-04-08T06:22:42.949Z | 2022-04-06T00:00:00.000 | {
"year": 2022,
"sha1": "48537e6a2f8fb78e9018fa7f583f8d5455b884d0",
"oa_license": "CCBYNCND",
"oa_url": "https://www.tandfonline.com/doi/pdf/10.1080/21645515.2022.2059308?needAccess=true",
"oa_status": "GOLD",
"pdf_src": "TaylorAndFrancis",
"pdf_hash": "e690ea6c03b9f200029fd919e3a657f1e838f4bd",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
244779579 | pes2o/s2orc | v3-fos-license | Effect of Fear of Falling on Mobility Measured During Lab and Daily Activity Assessments in Parkinson’s Disease
In chronic disorders such as Parkinson’s disease (PD), fear of falling (FOF) is associated with falls and reduced quality of life. With inertial measurement units (IMUs) and dedicated algorithms, different aspects of mobility can be obtained during supervised tests in the lab and also during daily activities. To our best knowledge, the effect of FOF on mobility has not been investigated in both of these settings simultaneously. Our goal was to evaluate the effect of FOF on the mobility of 26 patients with PD during clinical assessments and 14 days of daily activity monitoring. Parameters related to gait, sit-to-stand transitions, and turns were extracted from IMU signals on the lower back. Fear of falling was assessed using the Falls Efficacy Scale-International (FES-I) and the patients were grouped as with (PD-FOF+) and without FOF (PD-FOF−). Mobility parameters between groups were compared using logistic regression as well as the effect size values obtained using the Wilcoxon rank-sum test. The peak angular velocity of the turn-to-sit transition of the timed-up-and-go (TUG) test had the highest discriminative power between PD-FOF+ and PD-FOF− (r-value of effect size = 0.61). Moreover, PD-FOF+ had a tendency toward lower gait speed at home and a lower amount of walking bouts, especially for shorter walking bouts. The combination of lab and daily activity parameters reached a higher discriminative power [area under the curve (AUC) = 0.75] than each setting alone (AUC = 0.68 in the lab, AUC = 0.54 at home). Comparing the gait speed between the two assessments, the PD-FOF+ showed higher gait speeds in the capacity area compared with their TUG test in the lab. The mobility parameters extracted from both lab and home-based assessments contribute to the detection of FOF in PD. This study adds further evidence to the usefulness of mobility assessments that include different environments and assessment strategies. Although this study was limited in the sample size, it still provides a helpful method to consider the daily activity measurement of the patients with PD into clinical evaluation. The obtained results can help the clinicians with a more accurate prevention and treatment strategy.
INTRODUCTION
Parkinson's disease (PD) is a neurodegenerative disease that is associated with the degeneration of the dopaminergic nerve cells in the substantia nigra (Rai and Singh, 2020). Although currently, there is no cure for PD, treatment options such as Levodopa focus on alleviating PD symptoms. Fear of falling (FOF) is one of the most stressful symptoms for patients with PD (Frazier, 2000;Jonasson et al., 2018), leading to reduced quality of life and social isolation (Howcroft et al., 2013). Moreover, it is the strongest predictor currently known for future falls in this population (Lindholm et al., 2015), which indirectly but strongly associates FOF with the consequence of falls, such as fractures and other injuries (Bloem et al., 2001;Allen et al., 2013).
Fear of falling can be assessed by several scales of which the Falls Efficacy Scale-International (FES-I) is the most widely used to evaluate the concerns of patients about falling during various daily activities (Delbaere et al., 2010). These activities include walking, postural transitions, and turnings during daily activities. Being subjective in nature, FOF can have impacts on mobility that can be measured objectively (Rochat et al., 2010). Therefore, by the assessment of mobility, future falls can be predicted (Delbaere et al., 2004). Inertial measurement units (IMUs) enable the objective evaluation of mobility performance, both during functional tests in the lab and during daily activities. Instrumenting functional tests such as the timed-up-and-go (TUG) and five-time sit-to-stand (5xSTS) with IMUs provide a more in-depth analysis of gait and balance performance (Salarian et al., 2010;Van Lummel et al., 2016). Furthermore, IMUs can also help clinicians to evaluate the performance of patients during daily activities that are often very different from the supervised assessment in the lab and the clinic .
The potential of IMUs to distinguish patients with falls from those without has already been shown (Howcroft et al., 2013). These studies suggest that the most promising mobility parameters to detect an increased risk of falling are in the area of gait (Marschollek et al., 2009;Greene et al., 2010;Weiss et al., 2011Weiss et al., , 2013, postural transition (Najafi et al., 2002;Narayanan et al., 2008;Doheny et al., 2011;Weiss et al., 2011), and turning (Haertner et al., 2018). However, none of these studies investigated the contribution of FOF to these associations in detail.
In community-dwelling older adults, it has been shown that IMU-derived TUG parameters, such as total duration, turning velocity, and sit-to-stand duration, have a significant association with the FES-I total score (Williams and Nyman, 2018). Moreover, it has been shown in patients with PD that FOF affects their turning performance during the TUG test (Haertner et al., 2018). Patients with PD with FOF had significantly lower turning peak angular velocity, and PD fallers had significantly lower gait speed, compared to non-fallers (Latt et al., 2009). A drawback of the previous studies is that the performance of the participants has been studied mostly during assessments performed in the clinic while the association between FOF and the performance of the investigated cohorts during daily activities remains unknown. This is an enormous disadvantage, as daily activity and mobility are influenced by psychological and environmental factors that cannot be effectively investigated in a supervised environment (Owsley and McGwin, 2004;Feltz and Payment, 2005;Rudman et al., 2006;Kaspar et al., 2015;Evers et al., 2020;Shah et al., 2020b;Del Din et al., 2021;Maetzler et al., 2021).
Based on these findings, the first goal of this study was to determine whether there exist mobility differences between patients with PD with (PD-FOF+) and without FOF (PD-FOF−). For this purpose, we compared the IMU-derived gait, sit-tostand, and turning parameters from the respective lab and daily activity assessments. The second goal was to determine whether daily activity assessment can complement lab assessment in differentiating PD-FOF+ from PD-FOF−. The third goal was to investigate the associations between the same parameters obtained during these two assessment settings and study their differences in PD-FOF+ and PD-FOF−.
Participants and Study Cohort
Twenty-six participants with PD were included in the analysis. The inclusion criteria were age between 50 and 85 years, PD based on the United Kingdom Brain-Bank Society criteria, and the ability to understand and communicate well with the investigator. Patients with dementia were excluded from the study (Emre et al., 2007). All participants gave their written informed consent and the study was approved by the ethics committee of the Medical Faculty of the University of Tübingen (protocol no. 686/2013BO1) (Haertner et al., 2018).
Lab Assessments
Lab assessments were performed during ON medication state and included the Unified Parkinson's Disease Rating Scale (UPDRS-III) (Goetz et al., 2008) and the Hoehn and Yahr (H&Y) score (Hoehn and Yahr, 2001). Fear of falling was assessed with the FES-I (Yardley et al., 2005). An FES-I score > 19 was defined as the presence of FOF (Delbaere et al., 2010). The patients were also evaluated for depressive symptoms (Beck's Depression Inventory, BDI), the amount of Levodopa equivalent dose, and quality of life (Parkinson's Disease Questionnaire, PDQ-39).
For the mobility assessments, the participants were equipped with Mobility Lab R (APDM, Portland, United States) IMUs on the lower back and the two feet. The sampling frequency was set at 128 Hz. For the analysis, accelerometer and gyroscope data were used. All participants performed first a 7-m TUG test at their convenient speed. The TUG test includes a sit-to-stand movement, a walking phase, a 180 • turn, a walking back phase, and a turn-to-sit movement. The turn-to-sit transition consists of a simultaneously performed stand-to-sit transition and a 180 • turn. Then, the participants performed the 5xSTS test once with their preferred speed and once as fast as possible. Rest periods were given between these three lab mobility tests.
For the analysis of the TUG test, the lower back IMU was used to analyze the sit-to-stand and stand-to-sit postural transitions with a previously validated algorithm . The beginning of the sit-to-stand (t b,SiSt ) and the end of the standto-sit (t e,StSi ) times, as well as the sit-to-stand peak power (P TUG ) were extracted. The two turns within the TUG were analyzed by another validated algorithm, using the data from the lower back IMU (Salarian et al., 2010). The end of the second turn (t e,Turn2 ), as well as the maximum angular velocities around the vertical axis of each of the two turns (ω TUG,1 and ω TUG,2 ) were extracted. The total time of the TUG was calculated by subtracting the start of the sit-to-stand from the maximum value between the end of the second turn and the end of stand-to-sit: T TUG = Max t e,StSi , t e,Turn2 − t b,SiSt .
The IMUs on the lower back and feet were used to extract the instantaneous gait speed during the TUG test based on the algorithm introduced in Atrsaei et al. (2021b). The mean gait speed of the whole test was calculated (V TUG,avg ).
The 5xSTS tests were analyzed by the algorithm given in Atrsaei et al. (2020), using the data obtained from the lower back IMU. The following parameters were calculated: total time and mean sit-to-stand peak power of the normal (T 5xSTS,N , P 5xSTS,N ) and the fast 5xSTS (T 5xSTS,F , P 5xSTS,F ).
Mobility Assessment During Daily Activities
The participants were equipped with a RehaGait R IMU (Hasomed, Magdeburg, Germany) in an elastic belt on the lower back and were asked to wear the system for 14 days. The patients were instructed to plug the sensor into a personal computer to charge during the night. The following morning, the patients were asked to unplug the sensor and wear it. The data recording was started automatically right after the sensor is unplugged. Measurement phases of less than 6 h/day were discarded from the analysis. The following mobility parameters were extracted for each patient.
Gait
Walking bouts were detected by the algorithm introduced in Atrsaei et al. (2021b). Instantaneous gait speed, i.e., gait speed at each second was calculated (Atrsaei et al., 2021b). Instances in which the gait speed was less than 0.2 m/s were not included in the walking bouts as these instances can be considered as "nongait" periods (Atrsaei et al., 2021a). Walking bouts of less than 15 s were excluded from the analysis, to have a more steady-state gait and prevent non-locomotion movements to be detected (Atrsaei et al., 2021a). The total duration of walking for each day was obtained and was expressed as the percentage of the measurement duration of the respective day. Over all the days of measurement, the minimum (Gait ALL,min ), average (Gait ALL,avg ), and maximum (Gait ALL,max ) values of the walking percent were calculated. For instance, when a participant was assessed over a period of 5 days, and walked 5, 10, 15, 20, and 25% of the entire daily assessment periods, respectively, the Gait ALL,min , Gait ALL,avg , and Gait ALL,max would be 5, 15, and 25%, respectively.
The walking bouts were divided into short (between 15 and 30 s), medium (between 30 and 60 s), and long ones (longer than 60 s). Again, the minimum, average, and maximum values of the walking percentage per day for each type of walking bout were calculated. The indices SWB, MWB, and LWB were used to describe short, medium, and long walking bouts.
Over all the days of measurement stacked together, the gait speed distribution during all the walking bouts (V ALL ), as well as during the short (V SWB ), medium (V MWB ), and long (V LWB ) walking bouts were obtained separately. For each of these four distributions, the median, and the 95th percentile values were calculated.
There is evidence in the literature that gait speed often has a bimodal distribution during daily activities (Van Ancum et al., 2019;Atrsaei et al., 2021a). The first mode represents the lower preferred gait speed of the participants while the second mode represents the higher preferred gait speed of the participants (Van Ancum et al., 2019). Therefore, we also extracted the first and second modes of V ALL distribution as V µ 1 and V µ 2 , respectively.
Sit-to-Stand Transitions
Sit-to-stand transitions were detected during daily activities with a validated algorithm . For each day, the number of sit-to-stands per hour was obtained. The minimum (SiSt min ), average (SiSt avg ), and maximum (SiSt max ) number of sit-to-stands per hour were calculated over all days of measurement. Furthermore, for each sit-to-stand, the vertical peak power was determined as this parameter is a predictor of prospective falls (Regterschot et al., 2014). The distribution of all the peak power values over all the days of measurement stacked together was obtained as P H . The median of this distribution (P H,P50 ) and its 95th percentile (P H,P95 ) were calculated.
Turns
Turns were detected during daily activities with a validated algorithm (El-Gohary et al., 2014). The number of turns per hour was determined for each day. The minimum (Turns min ), average (Turns avg ), and maximum (Turns max ) number of turns per hour was also calculated over all the days of measurement. For each turn, the peak angular velocity around the vertical direction was obtained. The distribution of all the peak angular velocity values over all the days of measurement stacked together were obtained as ω H . The median (ω H,P50 ) and 95th percentile (ω H,P95 ) of this distribution were calculated.
Comparison Between PD-FOF+ and PD-FOF−
All the mobility parameters extracted from the lab and daily activity assessments were compared between PD-FOF+ and PD-FOF−. To exclude the potential differences due to gender and PD stage, the values were adjusted for gender and UPDRS-III with a multivariable logistic regression model. This analysis determines the odds of being PD-FOF+ considering gender (binary value, 0 for male, 1 for female), UPDRS-III (real-valued), and one of the mobility parameters (real-valued) explained in the previous section as independent variables. Moreover, the effect size (ES) (obtained by the r-value) was obtained by dividing the Wilcoxon rank-sum test statistics by the square root of the population (Ivarsson et al., 2013). An r value of about 0.1 indicates a small, 0.3 a medium, and 0.5 a large effect size, respectively (Cohen, 1992).
Fear of Falling Classification
To determine the predictive power of the extracted parameters in classifying PD-FOF+ and PD-FOF−, three classifiers based on a decision tree were used. Each classifier was trained based on one of the three sets of features mentioned below: • F1, Lab and daily activity (selected features): From all the parameters extracted from the lab and daily activity measurements, we selected those with an absolute r value of higher than 0.2. A backward elimination method was further applied to select the optimal features (Dadashi et al., 2014). • F2, Lab: From the set F1, the parameters from the lab assessment were used. • F3, Daily activity: From the set F1, the parameters from daily activity assessment were used.
The decision tree approach was used due to its proven performance in classifying patient populations based on mobility biomarkers (Millor et al., 2017;Rehman et al., 2019). For all the three sets mentioned above, cross-validation was performed based on the leave-one subject-out approach. The classification performance was evaluated by sensitivity, specificity, precision, accuracy, and area under the receiver operating characteristic curve (AUC) metrics.
Lab Versus Daily Activity Assessment
For each of the two groups, the gait speed, sit-to-stand peak power, and peak angular velocity were compared between lab and daily activities. For each parameter, a paired comparison was performed with the Wilcoxon sign rank test, and the significance level was set at p = 0.05. Pearson's correlation coefficient (ρ) was also obtained. A correlation coefficient < 0.5 was considered as low, between 0.5 and 0.7 as moderate, and >0.7 as high (Mukaka, 2012).
Moreover, each parameter obtained during the daily activities was divided by the same parameter obtained during the lab assessment. The new unitless parameters were compared between PD-FOF+ and PD-FOF− by the Wilcoxon rank-sum test.
Comparison Between PD-FOF+ and PD-FOF−
The characteristics of the participants are shown in Table 1. Of the 26 participants, nine had an FES-I score > 19. The PD-FOF+ showed a trend toward higher UPDRS-III scores in comparison with the PD-FOF−. The Levodopa equivalent dose, as well as the BDI score, was not significantly different between the two groups. However, the PDQ scale, as well as its mobility subpart, were significantly different between PD-FOF+ and PD-FOF− (p < 0.001 and = 0.0014, respectively). Table 2 presents results from the comparison of the lab and daily activity mobility parameters between PD-FOF+ and PD-FOF−. After the adjustment for UPDRS-III and gender, PD-FOF+ participants had significantly longer T TUG accompanied by slower ω TUG,1 , ω TUG,2 , and a slower V TUG,avg which, however, did not reach significance. Several parameters were slightly different between the two populations although the logistic regressions showed no statistical significance. For instance, compared with PD-FOF−, PD-FOF+ had, on average, lower gait speeds during the TUG (V TUG,avg ) and daily activities (V µ 1 ), longer T 5xSTS,F , lower percentages of walking bouts (i.e., Gait ALL,min , Gait ALL,avg , and Gait ALL,max ), and lower numbers of sit-to-stands (SiSt max ) and turns (Turns min ) per hour during daily activities.
No significant differences were found between the two groups when dividing the walking bouts based on their duration (Table 3). However, PD-FOF+ tended to have a lower percentage of short (e.g., Gait SWB,max ) and long (e.g., Gait LWB,max ) walking bouts, compared with PD-FOF− ( Table 3).
The effect sizes of the parameters are shown in Table 2 and Figure 1 in descending order. As a general note, labextracted parameters showed higher effect sizes than those extracted from the daily activity assessment. ω TUG,2 had the highest effect size, followed by other parameters extracted from the TUG test (except P TUG which had a very small effect size, see also Figure 1). ω TUG,1 had a lower effect size than ω TUG,2 . Directly after the TUG test, parameters ranked the T 5xSTS,F and P 5xSTS,F from the 5xSTS test with fast speed. The effect sizes of the parameters from the 5xSTS with normal speed (T 5xSTS,N and P 5xSTS,N ) were lower than those from the fast version. T 5xSTS,N had a smaller effect size compared with P 5xSTS,N . Gait ALL,max , SiSt max , Turns min , and V µ 1 had the highest effect sizes among the daily activity parameters, and the median gait speed (V ALL,P50 ) the lowest.
Fear of Falling Classification
Out of the 41 mobility parameters, 23 had an effect size > 0.2 (Figure 1). From these parameters, 19 features (used for machine learning-based classifier; marked with x in Figure 1) were selected by the backward elimination method and used for the F1 set.
Based on the three sets of features mentioned in the "Fear of Falling Classification" section, the results of the classification are shown in Table 4. The best performance was achieved based on set F1 which was a combination of features obtained from the lab and daily activity assessments (AUC = 0.75). The accuracy of this set was higher than when using lab (F2, AUC = 0.68) or daily activity (F3, AUC = 0.54) features alone. The p-value shows the significance of the coefficient of the inertial measurement unit (IMU)-based parameter in the logistic regression. *P < 0.05 was considered significant. The values of IMU-based parameters are shown by Median [IQR]. The effect size obtained by the r-value is ES. The sensitivity of the classification based on the features from the lab (F2) was higher than that obtained from the daily activity features, while the specificity of the classification based on daily activity features (F3) was higher. Moreover, F2 features achieved higher accuracy and AUC values, than the F3 features. Figure 1. F2: 7 lab features from F1. F3: 12 daily activity features from F1.
Lab Versus Daily Activity Assessment
The results of the paired comparison between lab and daily activity assessments for gait speed, sit-to-stand peak power, and turning peak angular velocity are shown in Table 5.
In the PD-FOF+ group, no significant correlations were found between lab and daily activity assessments concerning gait speed. Moreover, PD-FOF+ had significantly higher gait speeds at the 95th percentile of their walking speed distributions compared with the lab (V ALL,P95 , V SWB,P95 , V MWB,P95 , and V LWB,P95 ). In the PD-FOF− group, V TUG,avg had a significant but low correlation with V H,P95 (ρ = 0.48). A high correlation was also observed between V TUG,avg and V µ 2 (ρ = 0.70). Moderate correlations were observed between V TUG,avg and a gait speed of medium (V MWB,P95 ) and long (V LWB,P50 ) walking bouts (ρ = 0.59 and ρ = 0.57, respectively). The PD-FOF− group walked significantly faster during the TUG than during their daily activities.
Regarding the sit-to-stand peak power, a high and significant correlation was found between P TUG and P H,P95 for both groups (PD-FOF+, ρ = 0.77; PD-FOF−, ρ = 0.79). In both groups, P 5xSTS,N had a high and significant correlation with P H,P95 (PD-FOF+, ρ = 0.83; PD-FOF−, ρ = 0.70). No significant correlations were found between the 5xSTS with fast speed and daily activity assessment. Both groups had significantly higher peak power during the 5xSTS tests compared with P H,P50 during daily activities. However, P H,P95 values were not significantly different from the 5xSTS tests in the lab.
Finally, for turning peak angular velocity, no significant correlations were found between the lab and daily activities in any group. For PD-FOF+, there were no significant differences between ω H,P95 and both turns of the TUG. However, PD-FOF− had faster turns in the lab, compared with the home environment.
For a better representation of lab versus daily activity parameters, the gait speed, sit-to-stand peak power, and turning peak angular velocity are presented in Figure 2 as unitless ratios (daily activity parameter divided by the respective lab parameter). Most of the ratios were less than 1 (i.e., lower value of a parameter in the daily life environment). However, a few parameters, e.g., ω TUG,2 , with higher ratios in the PD-FOF+ group.
DISCUSSION
Most of the previous studies on this topic that have shown mobility-associated differences between PD-FOF+ and PD-FOF− have investigated their participants only in the lab. In this study, thanks to IMUs and dedicated algorithms, several mobility parameters were collected from patients with PD with and without FOF, when performing functional tests in the lab and living in their usual environment. The effect of FOF was investigated by quantifying the changes in mobility parameters between lab and daily life. The discriminative power between PD-FOF+ and PD-FOF− was shown by a logistic regression model considering each setting separately and in combination. And finally, the association between the lab and daily activity setting was studied by considering their correlation. Regarding the effect of FOF on mobility, PD-FOF+ needed more time to perform the TUG test than the PD-FOF−, which was -at least partly-explained by the slower performance of the two turns included in this test ( Table 2). This supports previous findings (Bryant et al., 2014;Haertner et al., 2018;Abou et al., 2021) and suggests that PD-FOF+ suffer from increased fear, especially during turns. This fear may be justified, e.g., through increased dysbalance or other constraints associated with FOF (Pourghayoomi et al., 2020). The larger difference between the two groups in the second turn, which also includes a stand-or walk-to-sit movement, may also argue for the different balance capacities between the groups. This argument is further supported by the slower peak angular velocity during the second turn compared with the first turn in the PD-FOF+ group.
In contrast to the evidence in the literature (Bryant et al., 2014), we did not observe a significant difference in gait speed between PD-FOF+ and PD-FOF− during the TUG test (V TUG,avg ). As the r-value showed a large effect size for this parameter in both groups, we hypothesize that PD severity rather than FOF has a particular influence on this parameter. We performed a Wilcoxon rank-sum test on V TUG,avg without adjusting for the aforementioned confounders, and obtained a significant difference between the PD-FOF+ and PD-FOF− (p = 0.021). Therefore, more evidence with a larger dataset is required to confirm this hypothesis as most of the previous studies did not adjust the statistical analysis for potential confounders.
Although none of the 5xSTS tests could sufficiently discriminate between PD-FOF+ and PD-FOF−, the fast 5xSTS test presented larger effect sizes than the preferred speed 5xSTS test (Table 2 and Figure 1). This is an argument for including the fast version rather than the preferred speed version (Goldberg et al., 2012;Staartjes and Schröder, 2018) in the assessment panel of clinical protocols. For the 5xSTS with preferred speed, the mean peak power of sit-to-stands (P 5xSTS,N ) had a medium effect size while the effect size for the total duration of the test (T 5xSTS,N ) was low ( Table 2 and Figure 1). This again highlights the usefulness of an instrumented 5xSTS test with IMUs to extract biomechanical parameters beyond the conventionally measured duration of the test (Van Lummel et al., 2016). Nevertheless, the IMU-derived sit-to-stand peak power did not differentiate PD-FOF+ from PD-FOF−. Also, the sit-to-stand peak power derived from the TUG test (P TUG ) was not significantly different between the groups. An explanation can be that the PD-FOF+ group might not have particular difficulties in performing postural transitions. However, numerous studies showed the predictive power of the 5xSTS test for future falls (Buatois et al., 2008;Duncan et al., 2011;Doheny et al., 2013;Qiu et al., 2018). Therefore, our results, together with previous results, suggest that the 5xSTS test is associated with aspects of falls that are independent of FOF.
None of the parameters derived from the daily activity assessment could significantly differentiate PD-FOF+ from PD-FOF−. However, medium effect size values were observed for several parameters. Interestingly, the effect size for the lower preferred gait speed (V µ 1 ) was higher than the median or 95th percentile values of gait speed distribution. This FIGURE 2 | Unitless daily activity divided by lab parameter ratios of (A) gait speed, (B) sit-to-stand peak power, and (C) turning peak angular velocity in PD-FOF+ and PD-FOF−. Differences between the groups were analyzed by the Wilcoxon rank-sum test. Only significant differences were shown on the plots * p < 0.05.
shows the importance of more precise modeling of gait speed distribution, rather than assuming a simple normal distribution of the obviously complex movements that occur in the usual environment (which was done in most of the previous studies, e.g., Toosizadeh et al., 2015;Takayanagi et al., 2019;Shah et al., 2020b). Interestingly, V µ 1 showed a higher effect size than V µ 2 . It should be noted that V µ 1 is assumed to correspond more to shorter walking bouts and V µ 2 represents mostly longer walking bouts that are more likely to occur outdoors (Van Ancum et al., 2019). Thus, our results regarding the higher effect size of V µ 1 vs. V µ 2 suggest that shorter walking bouts are more meaningful to describe mobility performance (limitations) of PD-FOF+, and maybe an interesting therapeutic target for future trials. It could also be speculated that PD-FOF+ have more problems than PD-FOF− during multitask-walking, as shorter walking bouts have obviously a higher probability to be associated with additional tasks, compared with long walking bouts which have a high probability for reflection, e.g., walks without relevant dual-task claim. Therefore, according to Figure 1, it is not surprising that the features that remained for the classification included more parameters from short walking bouts (Gait SWB,min , Gait SWB,avg , and Gait SWB,max ) than from medium and long walking bouts (Gait MWB,avg and Gait MWB,max ). In addition to Gait ALL,max , SiSt max and Turn min were among the daily activity parameters with the highest effect sizes. Thus, the number of various types of activities should also be considered in addition to parameters such as gait speed, sitto-stand peak power, and turning peak angular velocity that characterizes these activities. Moreover, there was a tendency toward a lower amount of activity in PD-FOF+. The PDQ score was significantly lower in PD-FOF+, showing that the quality of life of the patients was highly affected by their FOF. This can explain the lower amount of daily activities in this group of patients.
After feature selection in section "Lab Versus Daily Activity Assessment, " several parameters from the lab and daily activity assessments remained in the selected features (Figure 1). Training three classifiers based on three sets of features, i.e., F1, F2, and F3, revealed that set F1 led to the most accurate classifier to distinguish the PD-FOF+ from the PD-FOF− group (Table 4). This selection, including features from both the lab and daily activity assessments, further supports the usefulness of including daily activity assessments in clinical practice as they have complementary information to the assessments performed in the lab (Maetzler et al., 2021). The more accurate classification of FOF with lab features (F2), compared with daily activity features (F3, Table 4), suggests that capacity aspects play an important role for the definition of FOF (Maetzler et al., 2021) and functional tests in the lab should always be performed for the evaluation in FOF. Still, the inclusion of environmental context and psychological factors from daily life is a valuable addition and can contribute to increased specificity.
Comparing the gait speed between the lab and daily activity assessments, significant correlations were found for PD-FOF− but not PD-FOF+ (Table 5). Interestingly, PD-FOF+ had higher gait speed values in the "capacity" area of their daily activity assessment compared with the lab. For these participants, V TUG,avg , and V LWB,P95 V TUG,avg had values greater than 1 ( Table 5). One explanation can be that PD-FOF+ might be more cautious in non-familiar environments such as the lab. Moreover, and potentially more relevant for future management strategies, they might have been less cautious in their daily life especially when it comes to fast (and therefore more dangerous) gait episodes (Salkovic et al., 2017).
Another interesting observation, in our view, was that in PD-FOF−, V TUG,avg was significantly correlated with parameters during daily activity assessments that represent mostly the capacity aspects, i.e., V H,P95 , V H,µ 2 , V MWB,P95 , and V LWB,P50 . Moreover, the correlation between V TUG,avg and V µ 2 was high (ρ = 0.70). These findings firstly confirm the relevant association of lab parameters with daily activity parameters that are near the capacity area (Van Ancum et al., 2019;Warmerdam et al., 2020). These results suggest that capacity-associated values obtained during daily activities can indeed predict the capacity of a participant in the lab. Furthermore, the high association between V TUG,avg and V µ 2 is again in favor of considering a bimodal gait speed distribution during daily activities (Atrsaei et al., 2021a).
Regarding the sit-to-stand peak power, P H,P95 had high correlations with P TUG and P 5xSTS,N but not with P 5xSTS,F ( Table 5). This indicates that the 5xSTS test with preferred speed and the sit-to-stand part of the TUG test is most representative of the sit-to-stands performed during daily activities. In fact, in the TUG test, it is more accurate to name the initial postural transition as sit-to-walk rather than sit-to-stand. Since in daily life, there is often more sitting-to-walking than sittingto-standing, the high correlation between P TUG and P H,P95 seems reasonable. Therefore, to have a better understanding of the sitto-stand performance of patients during daily activities, clinicians should consider the 5xSTS test with preferred speed and the TUG sit-to-stand movement, rather than the fast 5xSTS test. The high association of sit-to-stand peak power between the lab and daily activity assessments was also observed in a study in communitydwelling older adults (Zhang et al., 2017). Nevertheless, as we demonstrated earlier, the 5xSTS test with fast speed had higher discriminative power for differentiating PD-FOF+ from PD-FOF−.
Our results are comparable to a very recent study on the impact of FOF on mobility parameters in a relatively large population of community-dwelling older adults (Wang et al., 2021). In that study, FOF led to a poorer mobility performance during both lab and daily activity assessments. Moreover, and comparable to this study, the consideration of both assessments showed the best discriminatory power between the presence and absence of FOF (lab assessment, AUC = 0.64; lab and daily activity assessment, AUC = 0.77). The strengths of our study, compared with the aforementioned study, are that we included postural transition and turning in addition to walking, and we assessed the daily activity over an average period of 12 (and not only 2) days (Wang et al., 2021).
Our study faces some limitations. First, our sample size could be small for statistical analyses. The observations and results could be supported more strongly with a larger population. This could explain why the parameters obtained during the daily activities did not differ significantly between PD-FOF+ and PD-FOF−. For instance, V µ 1 was at the edge of a statistically significant difference. However, it should be noted that finding participants with a specific impairment that are willing to participate in several clinical assessments, as well as 2 weeks of activity monitoring, can be challenging. While in this study, we explored the difference between participants with low and moderate FOF, the difference between participants with low and high FOF might be more evident with mobility parameters obtained during daily activities. Using other questionnaires in addition to FES-I can also be investigated. For example, participants can be asked whether their FOF restricts their activities or not (Rochat et al., 2010).
The duration of daily activities measurements could still be increased to have a more accurate estimation of the daily routines of the patients. Nevertheless, considering the current usability of IMUs, especially for older adults, it is a bit challenging to engage the participants for more than 2 weeks of measurements. It is not surprising that in the literature, a lot of studies consider only 1 week of daily activity measurements (Storm et al., 2018;Galperin et al., 2019;Van Ancum et al., 2019;Shah et al., 2020a).
Another point of limitation can be the turning assessment during daily activities. The turning algorithm considered turns with durations of 0.5-10 s and angles > 45 • (El-Gohary et al., 2014). This is a broad range, and future studies should investigate whether more specific definitions for turns that are performed in daily life have higher discriminatory power. Furthermore, the employed algorithm detected turns regardless of their occurrence during walking or sedentary behavior. Although it might be rare, participants might have been in a sitting position in a moving vehicle that had similar turning to those of a human that walks and turns at the same time. Therefore, further work is required to adapt the algorithm to detect turnings that occur during locomotion.
Another point was that since some of the unitless parameters showed significant differences between the two groups (Figure 2), we were curious if adding them to the feature set F1 will improve the classification results in Table 4. However, no improvement was observed. The reason might be that the ratio of home-derived mobility parameter divided by the same parameter obtained in the lab did not bring additional information as the information regarding both of the assessment settings was already there. Finally, to keep the data accuracy as high as possible, we excluded walking bouts < 15 s from the analysis to prevent other activities from being wrongly detected as a walking bout. However, these walking bouts contribute to a relevant portion of daily walking (Del Din et al., 2016;Shah et al., 2020c), and removing them might affect the meaningfulness of walking parameters with respect to the actual research question.
To conclude, the use of the IMU along with the dedicated algorithms allowed an unobtrusive assessment of mobility during daily activities. Although lab-based mobility parameters had generally higher discriminative power in differentiating PD-FOF+ and PD-FOF−, integrating daily activity assessments provided a more accurate classification of these patients. By comparing the same parameters from both settings, we could show for the first time that (i) considering lab and daily activity mobility parameters can lead to more accurate classification of PD-FOF+ and PD-FOF− compared with each lab and daily activity assessments alone (ii) the PD-FOF+ group performs the lab assessments with a rather cautious gait but used a rather incautious gait pattern in the usual environment; (iii) the sitto-stand peak power of the 5xSTS test with preferred speed and of the TUG was more closely associated with sit-to-stand movement in daily life, than was the same parameter obtained from the fast 5xSTS, and (iv) the 5xSTS test with fast speed mostly measured the capacity aspects of daily activities. These results provide further insight into the daily life behavior of PD patients with FOF, can stimulate prevention and treatment strategies, and can serve as a template for further studies using these novel techniques and assessment strategies.
DATA AVAILABILITY STATEMENT
The datasets generated and/or analyzed during this study are not publicly available but might be accessible from the corresponding author on reasonable request, upon joint approval from AA, CH, WM, and KA.
ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Medical Faculty of the University of Tübingen (protocol no. 686/2013BO1). The patients/participants provided their written informed consent to participate in this study.
AUTHOR CONTRIBUTIONS
AA designed the technical research and scientific question, extracted and analyzed the mobility parameters from clinical and home-based measurements, and wrote the manuscript. CH, ME, SS, DB, and IL-S performed the clinical measurements and extracted and analyzed the clinical questionnaires. AA, WM, DB, and KA discussed the results. WM, DB, and KA supervised and led the study. All the authors read, revised, and approved the final manuscript.
FUNDING
This study was supported by an unrestricted grant from Janssen Research and Development, a division of Janssen Pharmaceutica N.V. The funding of the study was pre-competitive and the funders did not have any influence on the design, conduction, and analysis of the results presented in this study. We acknowledge financial support by Land Schleswig-Holstein within the funding programme Open Access Publikationsfonds. | 2021-12-02T14:34:34.953Z | 2021-11-30T00:00:00.000 | {
"year": 2021,
"sha1": "a839be6a6c23863bdad2da9b9a6daa4507f4eed7",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fnagi.2021.722830/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a839be6a6c23863bdad2da9b9a6daa4507f4eed7",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
218925653 | pes2o/s2orc | v3-fos-license | Patient’s Perspective on Disease Burden, Remission Definition, and Symptoms Associated With Treatment Seeking: A Qualitative Study in Adult and Adolescent Patients With Crohn’s Disease
Abstract Background Disease burden, a definition of remission, and symptoms that drive treatment seeking were explored in a Crohn’s disease (CD) population. Methods A qualitative semistructured interview guide was developed, informed by published literature. Clinicians identified adolescents and adult patients with CD. Face-to-face interviews were audio-recorded and transcribed. Two rounds of interviews were conducted with patients. Transcripts were analyzed using thematic methods facilitated by ATLAS.ti. Results Twenty-four patients participated in the first round of interviews (n = 16 adults, mean age 50.3 years; n = 8 adolescents, mean age 15.6 years). Abdominal pain (n = 24), urgent bowel movements (n = 24), diarrhea (n = 23), and frequent bowel movements (n = 21) were the most frequently reported symptoms. CD affected patients’ physical functioning, daily activities, emotional wellbeing, social functioning, work/education, and relationships. No major difference in disease burden was observed between adolescents and adults. Twenty-three patients (96%) reported they would seek or had sought medical treatment for at least one symptom including abdominal pain (n = 19), diarrhea (n = 12), and blood in stools/rectal bleeding (n = 9). On a 0–10 scale (0 = no symptom and 10 = symptom at its worst possible), most patients (87%, 20/23) answered they would seek/had sought treatment when the symptom’s severity was at least 7. In the second round of interviews (n = 6 adults, mean age 51.5 years), 5/6 patients described that they did not require a complete absence of abdominal pain or loose/watery stools to consider their CD to be in remission. Conclusions CD is associated with substantial disease burden. Worsening of some symptoms drives treatment seeking. To some patients, remission is not defined as a complete absence of symptoms.
INTRODUCTION not well described from the patient's perspective. Identification of the factors that drive treatment seeking is important because of the recognized high healthcare burden 6 of CD. An understanding of such drivers may better inform healthcare resource use, aid identification of unmet treatment needs, and lead to improved patient outcomes. An understanding of what patients expect in order to consider a treatment to be successful is also required. Few studies have specifically explored patients' perceptions of a successful treatment response and if/how they define CD remission.
In this study, we aimed to understand adult and adolescent patients' experiences of CD, including CD-related symptoms, the burden of living with CD, as well as the symptoms that drive patients to seek medical treatment. Patients' understanding and definitions of CD remission were also explored.
MATERIALS AND METHODS
This study was designed as a cross-sectional, qualitative, noninterventional interview study. Two rounds of qualitative interview were conducted in patients with a confirmed diagnosis of CD.
Round 1
Patients were recruited from 2 sites in the United States by gastroenterologists from various private group medical practices that treat and manage patients with CD. The objectives of Round 1 were to understand (1) patients' CD symptom experience, including and beyond abdominal pain and diarrhea, (2) the impact of CD on patients' daily living, functioning, and wellbeing, and (3) symptoms that drive patients to seek medical treatment. Current symptom experience and symptoms reported during flares were explored. All patients were recruited according to predefined eligibility criteria. Gastroenterologists were required to complete and return case report forms to a specialist patient recruitment agency for all referred patients. Patients were eligible for inclusion if they were aged at least 12 years with a confirmed diagnosis of CD for at least 12 months and fluent in English. Patients were excluded if they had serious conditions other than CD that could impair their ability to participate in the study or affect the results of the study (eg, ulcerative colitis, abdominal abscesses, colonic mucosal dysplasia, intestinal obstruction, and irritable bowel syndrome) or had undergone recent abdominal surgery.
Diagnosis was based on a previous laboratory examination of blood and/or stool matter, X-ray, or endoscopic examination.
All clinical information was provided by referring clinicians (eg, disease severity, treatment regimens, and comorbidities). Purposive sampling is a nonrandom method of ensuring that particular characteristics within a sample are represented 9 and was used to recruit a diverse and representative sample, including male and female patients, patients of non-white ethnicities, and patients educated to high school diploma or less.
Round 2
An additional sample of patients, who did not participate in Round 1, aged at least 18 years was recruited using the same referring strategy and eligibility criteria as in Round 1. The objective of these follow-on interviews was to explore adult patients' understanding, perceptions, and definitions of disease remission; this was not a specific objective of the Round 1 interviews. Purposive sampling aimed to include patients who had experienced current or previous severe levels of CD as assessed by their clinician.
Round 1 interviews
A semistructured interview guide was developed which was informed by a conceptual model (CM) of patientreported symptoms and impacts of CD. The CM had been developed following a review of published qualitative literature. 4,7,[10][11][12][13][14][15][16][17][18][19] Patient-reported CD symptoms were split between 3 domains: "GI pain," "GI symptoms," and "other CD symptoms." A number of symptoms that patients associated with CD were placed under "Extraintestinal manifestations." Patient-reported impacts of living with CD were split between 3 domains: "GI-specific impacts," "Impacts on wellbeing," and "Overall impacts on HRQoL." The CM and interview guide were reviewed by a clinician with expertise in the management of patients with CD to verify the clinical relevance and correct domain placement of the CM's symptoms and impacts in addition to the appropriateness of the interview questions.
Patients participated in 90-minute face-to-face, one-toone interviews conducted between December 9 and 16, 2018. A concept elicitation discussion included open-ended questions to explore patients' experiences of CD, including the symptoms they experienced, and the associated impact of CD on their daily lives, functioning, and wellbeing. Next, patients were asked which of the symptom experiences that they had reported would drive/have driven them to seek medical treatment. Finally, patients were presented with a 0-10 scale (0 = no symptom and 10 = symptom at its worst possible) and asked to select a score at which they would seek medical treatment for each of those symptoms that they had reported during the discussion.
Round 2 interviews
The semistructured interview guide used in Round 1 was adapted to explore adult patients' understanding and definitions of disease remission and expectations of treatment response. Patients participated in 30-minute telephone/WebEx interviews between February 25 and 27, 2019. An initial concept elicitation discussion involved open-ended questions to explore patients' understanding and perceptions of CD remission. Additionally, patients were asked about their symptoms that would be present or absent during remission and asked to describe the severity, frequency, and extent of symptoms' bother that would represent remission, with a particular focus on abdominal pain and total number of liquid/watery stool bowel movements (BMs).
All interviews (Round 1 and Round 2) were conducted by an experienced qualitative researcher and were audio-recorded and transcribed verbatim.
Analysis
Demographic and clinical information collected during recruitment were descriptively summarized. Interview transcripts were analyzed using qualitative thematic techniques 20, 21 aided by ATLAS.ti v7.5 software which facilitated the coding and organization of data. First, the transcripts were read by the lead analysts and overarching ideas were identified. Next, descriptive codes were assigned to quotes within the transcripts to summarize the core meaning of patients' narratives, from which concepts and themes were identified. Codes were collated into potential themes and these themes were then compared and contrasted in order to interpret structure among them, both within patients and between patients according to clinical or demographic characteristics. Where new codes or themes emerged, the previous transcripts were reviewed to ensure that the code or theme was not missed during the first review. Through this method, a list of key concepts and themes were iteratively identified within each interview and across the sample. The first 2 transcripts were coded by several of the research team to develop a preliminary codebook for the interview. Using this codebook, the research analysts then coded the remaining interview transcripts.
Adult and adolescent analysis was initially conducted simultaneously as no prominent difference between the 2 groups was noted during the interviews. However, further subgroup analysis was conducted to identify any subtle differences.
Saturation analysis of CD symptoms reported by both the adult and adolescent groups was conducted following Round 1 (n = 24) interviews to determine if conceptual saturation was achieved. Conceptual saturation is defined as the point at which no new concept-relevant information emerges 22 and is applied to spontaneously elicited concept elicitation data only. Patients transcripts were divided into 3 sets (n = 8) in chronological order. Set 1 was compared to Set 2 to identify any new symptom concepts that had arisen in Set 2. Similarly, Set 3 transcripts were compared with Set 1 and Set 2.
Updated CM
The draft CM was updated following the Round 1 interviews (n = 24).
Ethical Considerations
Ethical approval for this study was granted by the New England Independent Review Board (#120180264) on November 12, 2018. Informed consent was obtained from all of the adult patients and from the adolescents and their parent/ guardian/caregiver.
Round 1
Twenty-four patients participated in Round 1 of the interview study. and both adult and adolescent groups contained an equal proportion of male and female patients. The majority of patients were Caucasian or white (n = 16, 67%). Eighteen (75%) patients had moderate or severe CD as rated by the referring clinician and 18 (75%) patients received treatment for their CD; 17 (71%) were in receipt of a biologic therapy. Two patients were prescribed corticosteroids.
Round 2
Six adult patients participated in Round 2 of the interview study. The sample included male (n = 2) and female (n = 4) patients with a mean age of 51.5 years (range 41-74). Most patients (83%; n = 5) were Caucasian or white and 1 patient was Hispanic. Most patients (66%; n = 4) were educated to a graduate level. Patients had been diagnosed with CD for a mean 7.1 years (range 1-17). Three (50%) of the patients had severe CD in the opinion of their clinician while the rest of the sample (50%) had moderate CD. All 6 patients received biologic therapy and 3 (50%) were also prescribed corticosteroids.
CD symptoms experienced by patients
Twenty-five symptoms were reported by patients in Round 1 of the interview study. Symptoms reported by more than 1 patient are presented in Table 2 by adolescent and adult subgroups. Symptoms that were reported spontaneously or reported following interview probing techniques are noted.
Abdominal pain/cramps and GI symptoms, including urgent BMs, diarrhea, frequent BMs, blood in BMs/rectal bleeding, bloating, and nausea/vomiting, were the symptoms most frequently reported by the patients.
All 24 patients (n = 16 adults, n = 8 adolescents) reported experiencing abdominal pain and urgent BMs. The severity of abdominal pain varied and patients described the most severe pain as occurring during a flare. The frequency at which patient experienced abdominal pain also varied; some patients experienced abdominal pain 2 to 3 times a day or every other day, while others only experienced abdominal pain in the context of less frequent flares which occurred every 2 to 3 months.
Most patients described urgency to have a BM as an immediate need to go to the toilet. Frequency of urgency was different among patients, with some patients experiencing urgency every day. However, some patients reported experiencing urgency only a few times a month. Severity of urgency also varied among patients, with some patients describing instances when they found it hard to "hold in" their BMs. Many patients also reported that urgency of BM was not specifically related to one type of BM such as diarrhea; patients reported that food could trigger urgent BMs and this would often be experienced at the same time as abdominal pain and cramping. Additionally, the majority of patients had experienced diarrhea (96%; n = 16 adults, n = 7 adolescents) and frequent BMs (88%; n = 15 adults, n = 6 adolescents).
Patients described diarrhea as loose or liquid BMs, with the frequency and consistency of BMs varying between patients. Some patients experienced diarrhea every week, while other patients experienced diarrhea occasionally over the course of a few months. Diarrhea could co-occur with abdominal pain, abdominal cramping, and blood in the stool.
The frequency of BMs varied between patients, but all patients acknowledged experiencing an increased frequency of bowel moments during a flare and following abdominal pain, cramping, and discomfort. Several patients felt that eating certain foods could increase the frequency of their BMs.
Symptoms that were reported by one patient each included peri-anal discharge, leg pain, back pain, dizziness, and shortness of breath. Extraintestinal manifestations such as joint pain, hair loss, skin and eye problems were also reported by a small number of patients.
Abdominal pain appeared to be almost equally bothersome for both adult and adolescent patients, while diarrhea appeared to be more bothersome for adults than for adolescents even though diarrhea was reported in similar proportions among both groups (n = 16/16 adults; 7/8 adolescents). Interestingly most adolescent patients who described experiencing fatigue/tiredness (3/4) also reported it as one of their most bothersome symptoms, compared with only 1/8 adult patients who had experienced fatigue/tiredness.
Flares
The concept of "flares" was discussed with patients; most (92%; n = 22) were familiar with and used the term flares or flare-ups to describe times when their usual symptoms (eg, abdominal pain, frequent BMs) substantially worsened in severity or times when symptoms that were not usually present occurred (eg, nausea, vomiting, and blood in stools). Only 2 of the adolescent patients were unfamiliar with the term "flares." All 24 patients reported having experienced worsening and/or the reoccurrence of one or more of their symptoms during the course of their CD, for example, "Just like all of those symptoms really worse, if they're at their worst and maybe needing to see the Doctor or get another medication for that time" (02-04-A), "When I have a flare-up I get some, sometimes diarrhea, very uncomfortable. I: So, could you tell me the
difference. in a day when you're flaring and a day when you're not flaring? I feel okay…I don't have any symptoms" (02-14-A).
The most frequently reported flare symptoms were abdominal pain/cramping (83%; n = 20) and changes in BMs, including more frequent BMs and/or diarrhea (71%; n = 17), for example, "I'd say, like, a flare-up is, like, your most severe symptoms, like cramping and bloating, all at once and you have, like, an urgency and it lasts pretty long" (02-05-P). In general, there were no major differences reported between adult and adolescent flare experiences.
Impacts of living with CD
All patients in Round 1 of the interview study reported impacts they had experienced as a consequence of living with CD. CD affected patients' daily activities and ability to fully participate in their life activities.
Impacts on physical functioning and daily activities
The most prominent impact was on diet and eating. Most patients (n = 15 adults, n = 6 adolescents) described the need to change their diet and avoid certain foods to prevent triggering symptoms such as pain, cramping, urgent BMs, diarrhea, and frequent BMs.
Patients' physical functioning was affected by pain. Adolescent patients may have been more affected due to scheduled school physical activities which they sometimes could not participate in, for example, "I played lacrosse and after I got sick it was a lot harder to move around. Like, if-I was going to go back on the swim team this year…but I couldn't because with what I, what I usually swam, it would cause me very, very bad pain in my, my abdominal area" (01-06-P).
GI-specific impacts
Impacts that were specific to GI symptoms included the need to go to the toilet frequently (n = 13 adults, n = 7 adolescents), dealing with unpredictable GI pain, urgent BMs, and vomiting (n = 9 adults, n = 3 adolescents), for example, "I'm asking two or three, four times, to go to the bathroom in one period [lesson] by, by myself " (01-05-P).
Patients were constantly seeking the nearest toilet (n = 6 adults, n = 2 adolescents), for example, "… always checking wherever I am in unfamiliar places, where the bathroom is and how long it takes to get there" (02-01-A) and were also worried about not reaching the toilet in time due to their urgent need to have a BM (n = 7 adults), for example, "I'm worried that I might even go in my pants" (02-13-A).
Impacts on social functioning, hobbies, and leisure pursuits
Sixteen adult and 5 adolescent patients reported that their social lives had been affected due to CD. The unpredictable nature of symptoms made it difficult for patients to make plans as they could not be sure they would be well enough to attend, for example, "In some ways I, kind of, socially pull back" (02-01-A).
Impacts on work and education
Twenty patients (n = 13 adults, n = 7 adolescents) reported that CD had impacted their work or school life. Most patients (n = 7 adults, n = 6 adolescents) had needed to miss days at work or school due to abdominal pain and BMs, for example, "Like about a month before for work and then recently, I was out, like, a week, well, every month. I miss work probably about two or three days" (01-01-A).
Impacts on emotional and psychological wellbeing
Many patients (n = 13 adults, n = 2 adolescents) were anxious about urgent BMs, locating and getting to a bathroom in time, experiencing flares, and the future implications of having CD.
Patients also reported feeling embarrassed and/or self-conscious (n = 10 adults; n = 4 adolescents), especially with relation to frequent bathroom use, for example, "Kind of embarrassing when you're at school, you know? You don't really tell people that. You just have to tell the Teacher… 'I have an emergency, I got to use the bathroom'" (02-08-P). Feelings of sadness, upset, and depression were often described by patients (n = 9 adults, n = 5 adolescents), either as a generic feeling accompanying their experience of CD symptoms or more specifically when they reflected on the impact of CD on their work, family, and social life, for example, "It does affect me on a daily basis… it's not only physically, it's mentally you get a little bit down, because, …you're incapacitated" (02-11-A).
Impacts on relationships
Relationships, including friendships, were affected by CD; adolescents, in particular, described feeling left out of social situations and losing friends due to being unable to participate in activities, for example, "I haven't been on any sports teams, so I've lost a couple of friends through that" (01-06-P).
Updated CM
The findings from this study, specifically the concepts arising from the open-ended discussion in Round 1, were compared to the CM developed from the literature review to provide support for existing concepts and identify any potential new concepts relating to CD symptoms and impacts.
Most symptom and impact concepts within the CM were also identified in this study. At a domain level, GI pain symptoms, GI symptoms, and other CD symptoms were all reported by patients within this study. Three new GI symptoms arose as follows: constipation, heartburn/ indigestion, and gurgling noises. In addition, 2 new concepts unrelated to pain or GI symptoms also arose as follows: dizziness and shortness of breath. No new impact concepts were identified from the interviews. Figure 2 represents the updated CM.
Treatment Seeking
Twenty-three patients (96%; n = 16 adults, n = 7 adolescents) confirmed that they would seek or have sought medical treatment for at least one of their symptoms. Seventeen (n = 13 adults, n = 4 adolescents) reported that they have sought or would seek help for more than one symptom. Patients stated that they would either call or go to their primary/specialist doctor or attend a hospital emergency room (ER).
Only one adolescent patient reported that they would not seek additional medical treatment for their symptoms. This may be due to their mild disease severity as confirmed by their clinician. This patient reported that they were able to manage their symptoms by diet, rest, and nonprescription pain relief as needed.
Abdominal pain was the most frequently reported symptom that patients would seek medical treatment for (n = 14 adults, n = 5 adolescents). Patients described constant or prolonged "uncontrollable" pain that could not be alleviated by nonprescription medication that rendered them immobile and "doubled-up," "on the floor," or "in tears" from the pain.
Diarrhea (n = 10 adults, n = 2 adolescents) that was either constant, frequent, occurred both during the day and night, not controlled by nonprescription medication or that was accompanied by abdominal pain or nausea were all cited as reasons for seeking treatment. Dehydration due to constant diarrhea was a concern for a few patients.
Blood in stools/rectal bleeding was reported by 9 adults. Most patients would only seek treatment if they experienced what they considered large amounts of blood seen when using the toilet.
Fatigue/tiredness (n = 1 adult, n = 3 adolescents) that was described as "excessive" or "unbearable" and that caused excessive sleepiness during the day despite adequate nighttime sleep was also a driver of treatment seeking.
Nine patients (n = 8 adults, n = 1 adolescent) explained that they would seek medical help, for example, attend hospital or go to their doctor, if they experienced a flare-up or a combination of symptoms or developed new symptoms. Figure 3 represents the 4 most frequently reported symptoms that have driven or would drive patients to seek medical treatment. Most patients (20/23) reported that they would seek/ had sought treatment when the symptom's severity was at least 7 on a 0-10 numerical rating scale. The severity score ranges that would prompt patients to seek medical treatment for diarrhea and abdominal pain were 6-10. Patients reported a range of 5-10 for frequent BMs and blood in stools/rectal bleeding.
A few patients explained they were pain-tolerant therefore selected higher scores for abdominal pain, while a few other patients selected lower scores for symptoms such as constant diarrhea or blood in stools/rectal bleeding than their other symptoms, as they were considered more disruptive, important, or more worrisome.
Patient terminology
All 6 adult patients in Round 2 of the interview study were asked if they had ever heard the term "remission" in relation to their CD and what they called "remission" or times when their CD had "got better." Three patients had heard this term in relation to CD, however, one of these patients had never used the term "remission" in relation to an improvement in their CD symptoms. Two patients had not heard this term used in relation to CD and another patient had only heard the term used in relation to cancer. The patients who had heard the term "remission" in relation to CD (or another condition) had graduate-level education whereas the patients who had not heard this term had lower educational attainment.
Definition and experiences/perceptions of remission
The 3 patients who had heard the term "remission" in relation to CD were asked how they would define "remission" in CD. All 3 considered symptoms that were "not severe" or had "eased" to define remission, for example, "It's when your symptoms are not as severe, and you're in a calm spot" (02-17).
The 3 patients who had not heard the term remission previously in relation to CD were asked how they would define times when their CD had got better (n = 3). All 3 cited a reduction in abdominal pain, diarrhea, and frequency of BMs as defining the times when their CD had got better, for example, "When I'm not experiencing pains in my stomach and probably not going to the bathroom quite as often" (02-19). One patient also cited fewer bathroom visits (BMs) and one cited a lack of fatigue. One patient felt that they never had experienced remission because their symptoms always returned. This patient associated remission with a permanent absence of symptoms, for example, "I consider remission where they're permanently gone or they're not going to come back, and I've never been through that, so I can't really consider myself in remission or anything" (02-16).
All 6 patients were asked if all of their CD symptoms would have to go away for them to consider themselves to be in "remission." Five patients reported that not all their symptoms would need to go away to consider themselves in remission, for example, "I just wanted just a drastic improvement. And, you know, and just issues, maybe, few and far between" (02-18). ; only the lower score is reported in the figure. ‡ These symptoms were reported by some patients during the general discussion but were not additionally discussed during the NRS task: frequent bowel movements (n = 4), abdominal pain (n = 3), and diarrhea (n = 2).
Only one patient required that all of their symptoms, including nausea, vomiting, fatigue, and urgent BMs, would have to go away to consider themselves in remission.
Six patients confirmed that both the number of liquid/ watery BMs and the severity of their abdominal pain needed to improve to consider their CD in "remission," for example, "I think they go together, they're part and parcel… whenever you have diarrhoea you have discomfort … the less I have to deal with the pain, the less I have to go to the bathroom… the better it makes my life" (02-19).
All 6 patients were asked the number of liquid/watery (type 6/7) BMs they would have in a typical 24-hour period to consider their CD in remission. Additionally patients were asked to specify the smallest number of liquid/watery (type 6/7) BMs that was acceptable for them to consider themselves to be in remission. The number of BMs ranged from 0 to 5 and 0 to 4, respectively (Table 3). Patients were also asked which level of abdominal pain would indicate CD remission. Most patients identified "mild" as the level of abdominal pain that would indicate CD remission (Table 4).
SATURATION ANALYSIS
Saturation analysis showed that almost all symptom concepts were identified in the first 2 sets of the Round 1 interviews. Only 2 new concepts emerged in Set 3: leg pain and back pain/ discomfort. However, "general pain" in multiple locations was reported in previous sets; therefore, saturation was considered achieved. Saturation analysis was not conducted for the Round 2 interviews due to the difference in study objectives, the exploratory nature of this second round of interviews, and the small sample size of 6 patients.
DISCUSSION
Several symptoms and impacts on wellbeing were described by the patients in this qualitative study which demonstrated the substantial disease burden that adult and adolescent patients with CD are faced with. Most of the symptoms and impacts reported by patients were also identified in the targeted literature review conducted before the interviews as shown in the CM (Fig. 3). The symptoms and impacts reported in this study are in-line with those included in the CM developed by Gater et al. 23 Abdominal pain/cramps and GI symptoms including diarrhea, frequent BMs, bloating, and blood in BMs were the most frequently reported and most bothersome symptoms for patients. This result is in-line with other qualitative studies that identified these symptoms as common and important symptoms experienced by patients with CD. 7,12,14 There were few differences between symptom reports from males and females although notably more females than males reported bloating. Urgent BMs were also notably problematic for those who experienced them because they were a cause of substantial anxiety. Patients always had to know where the closest toilet was and feared losing control of their BMs. This finding is also noted in other CD studies. 14,19 The non-GI symptoms which were considered bothersome included fatigue, weakness, and joint pain; indeed, joint pain may be one of the most severe and difficult symptoms for patients with CD to manage. 7,12 This study provided an insight into patients' experience of flares which were found to be disruptive to patients' daily lives and functioning; a number of patients experienced variation in the severity of their flares which has been noted in previous research. 11 Several patients sought additional medical help when they considered themselves to be having a flare, including going to their primary care doctors, seeking different medication or even visiting the ER.
A small sample of adolescent patients were included in this study and few differences in the type of symptoms reported were identified between adults and adolescents. The impact of frequent BMs appeared to be slightly more prominent in adolescents, mainly due to the disruption of lessons at school. Additionally, nearly half of the adolescent sample cited fatigue as one of their most bothersome symptoms and also reported that they would seek treatment for their fatigue. However, comparisons based on this small sample should be considered exploratory only. Indeed, there appears to be a paucity of published qualitative studies that explore the lived experiences of adolescents living with CD, despite the common incidence and increasing prevalence of CD in adolescents. 24 Worsening of the symptoms that generally characterize CD, for example, abdominal pain, frequent BMs, and diarrhea were the main drivers of treatment seeking among patients. Interestingly, when patients do seek treatment they do not necessarily expect all of their symptoms to be completely alleviated. Instead a reduction in severity in abdominal pain and loose/watery stools may be accepted as a satisfactory treatment outcome. This finding potentially reflects patients' expectations for living with a chronic condition; Norton et al 7 found that patients did not expect to regain their prediagnosis health status and many reported being in remission or feeling good despite still experiencing CD symptoms. However, other studies have noted that some patients defined remission as no symptoms present and not requiring steroid treatment. 25 Further exploration of patient perceptions of CD remission and their perceptions of treatment expectations would help align patient and healthcare professional treatment outcome goals and aid the evaluation of treatment success in patient-focused drug development.
Limitations of this study are acknowledged. This study was completed in the United States only with English-speaking patients and therefore the results may not be applicable to other countries and cultures without further investigation. While the sample in Round 1 included adolescent patients, no younger adolescents aged 12-13 years were included. However, diagnosis of CD in American patients is most common between the ages of 15 and 35 26 and therefore the lack of participation among 12-to 13-year olds is reflective of the population. Additionally, the sample overall included patients with a age range of 14-75 years who shared their experiences. Subgroup analyses to identify differences and similarities between adult and adolescent patients were somewhat limited by the sample size and therefore any differences highlighted in the data should be treated with caution. The Round 2 interviews included only short discussions with a small sample of adult patients who were mostly educated to graduate level and therefore the findings should be considered exploratory; further confirmation in a larger and more diverse sample is recommended. There is an inherent risk of sampling bias in that those who were interested in participating in the study may be those who are more likely to seek care to begin with. In addition, no objective disease severity/activity assessment was collected at recruitment. Finally, qualitative data interpretation can be influenced by the interviewers' and analysts' personal biases. This was somewhat mitigated by using the same interviewer in all interviews so that interviewer effects were consistent. Additionally, the findings of this study were reviewed by a multidisciplinary team including clinical experts.
CONCLUSIONS
This qualitative study explored the lived experience and disease burden of CD which was primarily associated with GI symptoms and abdominal pain which significantly impacted patients' daily activities, physical functioning, sleep, work/ school, and emotional wellbeing. Treatment seeking was generally driven by worsening in severity of GI symptoms, while CD disease remission was primarily considered by patients to be associated with an improvement or reduction in symptoms including abdominal pain and BMs.
DATA AVAILABILITY
The data that support the findings of this study are available on reasonable request from Eli Lilly [April Naegeli]. The data are not publicly available due to protect the privacy of research participants. | 2020-05-07T09:10:36.828Z | 2020-04-01T00:00:00.000 | {
"year": 2020,
"sha1": "2717dba089711b545b486a5be255113b8f4b0a35",
"oa_license": "CCBY",
"oa_url": "https://academic.oup.com/crohnscolitis360/article-pdf/2/2/otaa033/33378588/otaa033.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "50728de0ed4799b694aea32c2731325ac58dae65",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
18158664 | pes2o/s2orc | v3-fos-license | Indirect Reciprocity with Optional Interactions and Private Information
We consider indirect reciprocity with optional interactions and private information. A game is offered between two players and accepted unless it is known that the other person is a defector. Whenever a defector manages to exploit a cooperator, his or her reputation is revealed to others in the population with some probability. Therefore, people have different private information about the reputation of others, which is a setting that is difficult to analyze in the theory of indirect reciprocity. Since a defector loses a fraction of his social ties each time he exploits a cooperator, he is less efficient at exploiting cooperators in subsequent rounds. We analytically calculate the critical benefit-to-cost ratio above which cooperation is successful in various settings. We demonstrate quantitative agreement with simulation results of a corresponding Wright–Fisher process with optional interactions and private information. We also deduce a simple necessary condition for the critical benefit-to-cost ratio.
Introduction
The evolution of human cooperation is an intensely researched topic in the biological and economic sciences [1][2][3][4].In the classic prisoner's dilemma game, the rational choice for either player is to defect regardless of the other player's choice of strategy.Even in the repeated prisoner's dilemma with exactly m rounds, defection on every round is the only strict Nash equilibrium strategy and the only evolutionarily stable strategy.Indirect reciprocity is an important mechanism for the evolution of cooperation [1].The basic setting of indirect reciprocity is repeated interactions in a group of players [5][6][7].My behavior toward you depends on what you have done to me and to others [8][9][10].
Indirect reciprocity is a generalization of direct reciprocity [1].Direct reciprocity is the phenomenon in which a person behaves toward another individual based on prior personal experience with that individual.For example, if Individual A had a positive interaction with Individual B, then Individual A would be more inclined to cooperate with Individual B in subsequent encounters.Indirect reciprocity works via reputation [11][12][13].For example, if Individual A had a positive interaction with Individual B, then Individual B would achieve a good reputation also in the eyes of an observer, Individual C. Individual C would remember Individual B's good score and be more likely to cooperate with Individual B in subsequent encounters.In this way, people help those who help others, and helpful people have a higher payoff in the end, as is shown in experimental studies [14][15][16][17][18][19][20][21][22].One sees that indirect reciprocity can foster cooperation in the long run because cooperators are able to channel benefits primarily toward other cooperators.
Many theoretical studies have focused primarily on the case of public information [23][24][25].In this setting, all individuals in the population have the same opinion about the reputation of any particular individual.Various subtopics have been researched, including the effects of image scoring and good standing strategies [23], the reputation dynamics that lead to evolution of indirect reciprocity [24,25], involuntary defection [26], games among more than two players [27,28], the ability of cheaters to disrupt stable strategies [29], costly information transfer [30], trinary (instead of binary) reputation models [31], mixing of social norms [32], and others.Public information is a nice modeling simplification, and it can serve as a powerful promoter for the evolution and maintenance of cooperation.
But the original and more general formulation of indirect reciprocity allows for private information, where players differ in their opinion concerning the reputation of others [33][34][35][36][37][38][39].Some of the original studies of indirect reciprocity via image scoring have explored the evolutionary dynamics with private information [38,39].More recent studies have discussed topics of indirect reciprocity with private information, such as the effects of private information on assessment rules [40,41], incomplete observation [42], and assessment errors [43], among others.Importantly, the distribution of reputations in the population, which specifies how individuals perceive each other, can be evolving over time.Basic analytical results with private information-and with an evolving distribution of reputations-are typically difficult to achieve.
Model
We investigate a simple model of indirect reciprocity with private reputation.The identity of a defector may be known to some people, but not to others.We consider a prisoner's dilemma with two strategies: cooperation, C, and defection, D. The payoff matrix is given by Consider a population of N = N C +N D individuals, in which N C individuals are cooperators, and N D individuals are defectors.In a single round, two individuals are chosen randomly, and they are offered a game.A cooperator always plays with a cooperator, and each receives payoff b − c.If a cooperator and a defector are chosen in a round, then the cooperator accepts to play the game only if he or she is unaware of the defector's identity.If the cooperator is cognizant of the defector's identity, then the potential game is rejected.Thus, our model features optional interactions between players [44].If a game is played between an unknowing cooperator and a defector, then the cooperator receives payoff −c, while the defector receives payoff b.If two defectors are chosen in a round, then they play, and each receives payoff 0. (Alternately, one could consider that two defectors simply do not play a game together; since a game between two defectors has no effect on payoffs and results in no information transfer, such a distinction is inconsequential for our model.)Each time a defector plays a game with a cooperator, his or her identity is revealed to each of the N C cooperators in the population with probability p.We assume that this information transfer about the identity of the defector between cooperators is not costly [30].
A total of M rounds occur in a single generation.We use k to denote one of the N individuals in the population.The payoffs from all games of individual k over M rounds are added to obtain that individual's total payoff, S k .Individual k's fitness is computed as F k = exp(βS k ), where β represents the intensity of selection.After M rounds, Wright-Fisher updating is performed.For each individual in the next generation, a parent is chosen from the current generation proportional to fitness.Mutation occurs with probability µ and leads equiprobably either to a cooperator or to a defector.Then M rounds are played in the new generation.Payoff does not accumulate from one generation to the next, but is always reset to zero.
What is the critical value of b/c above which cooperators are more abundant than defectors when their numbers, N C and N D , are averaged over many successive generations of the mutation-selection dynamics?What is the critical value of b/c above which, with no mutation, defection fixes in the population with probability less than 1/N when starting with a single defector?Can these questions be answered analytically?
Single Defector
To make progress analytically, imagine the simpler case of a single defector in an infinitely large population of cooperators.In this simplified setting, the defector participates in m rounds and therefore is offered to play a game m times, each time with a random cooperator.(Note that in the dynamical model, M rounds are run in total between randomly selected pairs of individuals, but the number of rounds that a particular defector in the population participates in is a random variable.In the simplified static model, the sole defector participates in exactly m rounds.)After each exploitation of a cooperator, the exploited cooperator shares the identity of the solitary defector with each of the other uninformed cooperators with probability p = 1 − q.Thus, with probability q, each defector-cooperator link remains active after an exploitation of a cooperator by a defector.Informed cooperators always refuse a potential game with the defector.The fraction of cooperators in the population that are aware of the defector's identity increases with the number of games played.Since informed cooperators always reject a game with the defector, cooperators can achieve a higher total payoff than the defector if many rounds are offered.A sample sequence of rounds is shown in Figure 1.Changes in the fraction of active links to cooperators as games are played are shown schematically in Figure 2.
Figure 1.Schematic illustrating indirect reciprocity with optional interactions and private information in a population with eight cooperators (blue) and a single defector (red).The population size is N = 9.The defector is at the center of each image of the population.Each solid black line connecting the defector with a cooperator represents an active link.In the first round, the defector is offered a game with a random cooperator (light blue with a "?"), and the game is played.The defector gains payoff b.Each of the defector's links is removed with probability p = 0.5.After the first round, four of the defector's links with cooperators are eliminated (thin dotted lines).In the second round, the defector is offered a game with a cooperator that knows the defector's identity, so the potential game is rejected.In the third round, the defector again exploits an ignorant cooperator, and the defector again receives payoff b.Two of the remaining active links are eliminated.In the fourth round, the defector is offered a game with an informed cooperator, and the potential game is rejected.After this sequence of four rounds, the defector gains total payoff 2b.Schematic showing how the fraction of active links between the defector and cooperators decreases as additional exploitations occur.At the start of a generation, all links to cooperators are active.On the first round, the defector plays a game, and a fraction q of the links remain active after the game.On each round thereafter, the defector plays another game with probability q i , where i is the number of games that have been played to that point.
We denote the average number of games played in m rounds by a solitary defector in a population of infinitely many cooperators by F m (q).We must find an expression for F m (q).First, let us derive an expression for x i,m -the probability that the defector plays i games in m rounds.Suppose that the defector plays i times in m rounds.There are two possible cases: (i) the defector plays i times in the first m − 1 rounds and is rejected for play in round m; and (ii) the defector plays i − 1 times in the first m − 1 rounds and is accepted for play in round m.The probability for the first case is (1 The probability for the second case is q i−1 x i−1,m−1 .Therefore, the probability that the defector plays i times in m rounds is This expression holds for any m ≥ i ≥ 2 with the convention that x m,m−1 = 0.For i = 1, the second case does not occur and we have the recurrence formula
Probability of i Games in m Rounds
To solve the recurrence, Equation (1), for x i,m , it is helpful to first solve for its z-transform, which we write as From Equations ( 3) and ( 2), the z-transform of x 1,m is We can consider the summation as a geometric series (with region of convergence |z| > 1 − q), and we obtain From Equations ( 3) and ( 1), the z-transform of We see that xi (z) can be expressed recursively in terms of xi−1 (z): Iterating Equation ( 5) and substituting Equation ( 4), we find Thus, the z-transformed quantity xi (z) has a simple, closed-form expression in terms of q, i, and z.All that remains is to invert the z-transform to obtain x i,m .Intuitively, Equation ( 6) has exactly i simple poles, and each simple pole corresponds to a separate term in the evaluation of x i,m .We have (I denotes the imaginary unit.)Here, C represents a counterclockwise closed contour encircling the origin and enclosing all poles of the integrand.For example, we can take a unit circle for C as all the zeros and poles have modulus less than 1.From the residue theorem, we have We evaluate this as Simplifying Equation (7), we obtain This can be rearranged slightly as Collecting factors of q, we have We make the substitution i = j + k.We also define Equation ( 8) can then be simplified as Equation ( 9) is proven using an alternative method in Appendix A.
Average Number of Games
The average number of games played by the defector, F m (q), is given by From Equations ( 9) and ( 10), and after rewriting the summations, we have In Figure 3, we plot Equation ( 11) in several ways.
Fixed Number of Rounds per Generation
More generally, consider that there are N C cooperators and N D defectors.Two players are selected randomly in each round and are offered a game.Each individual is selected, on average, m times in M total rounds.The total number of rounds, M , is a constant for each generation.
Average Number of Games
Define by f = N C /N the fraction of individuals in the population that are cooperators.A total of M rounds are run in each generation in a population of size N .Here, M can be expressed in terms of m (the average number of rounds that each individual engages in) as M = mN/2.The probability that a given defector engages in a particular round is 2/N .In the limit N → ∞, the binomial distribution for the number of rounds that a given defector engages in becomes Poisson-distributed with expected value (mN/2)(2/N ) = m.Moreover, the average number of times that a defector is selected for a possible game with a cooperator is f m, and this quantity is Poisson-distributed with expected value f m.Thus, we can write the average number of games played by a defector with cooperators in a generation as Here, p = 1 − q is the probability that a link between a cooperator and a defector is eliminated after the defector exploits a cooperator.Figure 3. Average number of games that the defector plays.(a) Expected number of exploitations, F m (q), against the probability, q, to maintain each link when m = 5, 10, and 20 (bottom to top); (b) F m (q) against the number of rounds, m, when q = 0.2, 0.5, and 0.8 (bottom to top); (c) Expected number of exploitations normalized by the number of rounds, F m (q)/m, against the probability, q, to maintain each link when m = 5, 10, and 20 (top to bottom); (d) F m (q)/m against the number of rounds, m, when q = 0.2, 0.5, and 0.8 (bottom to top).
Critical Benefit-to-Cost Ratio
What is the critical benefit-to-cost ratio, b/c, needed to ensure that cooperators receive a higher payoff, on average, than defectors?We reason as follows: Each cooperator plays an average of f m games with other cooperators in a generation, so each cooperator receives a contribution to its payoff in the amount f m(b − c).Each defector plays, on average, G(f, m; p) games in a generation, so each defector receives average payoff bG(f, m; p).Each exploitation by a defector corresponds to a cost incurred by a cooperator without any compensating benefit.The total number of games that are played by defectors, on average, is N (1−f )G(f, m; p).Therefore, the average cost incurred by a cooperator due to exploitations by defectors is (−c)((N (1−f ))/(N f ))G(f, m; p).Putting these pieces together, we find that the average cooperator payoff in a generation exceeds the average defector payoff in a generation if Rearranging Equation ( 12), we find that cooperators achieve a higher payoff that defectors, on average, if
Simulation Results
We next perform simulations for which the total number of rounds per generation, M , is fixed.The success of cooperation depends on the rate, p, at which information about the defectors is transferred among cooperators, on the fraction, f , of cooperators in the population, and on the average number of rounds per generation, m, that an individual engages in.We have that m = 2M/N .
Simulation results for a fixed total number of rounds per generation are shown in Figure 4a,b.In Figure 4a, the total number of rounds per generation, M , is 400.Therefore, the average number of rounds per individual in a generation, m, is equal to 8. In Figure 4b, the total number of rounds per generation, M , is 800.Therefore, the average number of rounds per individual in a generation, m, is equal to 16.
We plot the critical value of b/c needed for cooperators to be more abundant than defectors with nonzero mutation rate, µ, and with weak intensity of selection, β ("+" symbols), for m = 8 and m = 16.We also plot the critical value of b/c above which cooperators achieve a higher payoff in M total rounds, on average, than do defectors if f = 1/2 (green circles) for m = 8 and m = 16.The two sets of data points agree well, as we would expect.In addition, we show the functions A(1/2, 8; p) and A(1/2, 16; p).The agreement between theory and simulation is excellent.
In another measurement, we start with a single defector and allow the dynamics to progress with no mutation (µ = 0) and with weak intensity of selection.We plot the critical value of b/c needed for the fixation probability at the all-defector state to be less than 1/N ("X" symbols) for m = 8 and m = 16 when starting with a single defector.We also plot the critical value of b/c above which cooperators achieve a higher payoff in M total rounds, on average, than do defectors if f = 2/3 (red squares) for m = 8 and m = 16.The two sets of data points agree well, as we would expect from the one-third law of evolutionary dynamics [2,[47][48][49][50][51].In addition, we show the functions A(2/3, 8; p) and A(2/3, 16; p).The agreement between theory and simulation is again excellent.
In yet another measurement, we plot the critical value of b/c above which a solitary defector in a population of cooperators achieves a lower payoff in M total rounds, on average, than does a random cooperator (black triangles).This is the condition for cooperation to be a strict Nash equilibrium.We also show the functions A(1, 8; p) and A(1, 16; p), again demonstrating outstanding agreement between theory and simulation.
Variable Number of Rounds per Generation
The only strict Nash equilibrium strategy and the only evolutionarily stable strategy in the classic prisoner's dilemma with exactly m repeated interactions is to defect in every round [2,12,52].
In realistic settings, an individual may not know at the outset how many rounds it will participate in with potential interaction partners.It is therefore worthwhile to also consider a variable number of rounds per generation.In a variation of our dynamical model, we postulate that another round is run between two randomly chosen individuals with probability W .The average total number of rounds that are run is M = 1/(1 − W ). The average number of rounds that each defector engages in, averaged over generations, is m.
Average Number of Games
The probability that a total of M rounds are run is (1 − W )W M −1 .Analogously to G(f, m; p), we can write the average number of games played by a defector, H(f, m; p), as Intuitively, we are just multiplying the probability that a total of M rounds are offered in a generation with the average number of rounds offered to an individual, G(f, m = 2M/N ; p).We consider the limit of large population size, N → ∞, which is realistic and also admits simplified results.Denote the average number of rounds that an individual engages in by m.The value of W for which the average number of rounds engaged in per individual equals m is W = 1 − 2/(mN ).We have We can write this as In the limit of large population size, N → ∞, the factor in square brackets becomes e −1 .Thus, we have This can be rewritten more suggestively as Here, note that ∆M = 1.(The summation increments M by 1 at each iteration.)Recall that m = 2M/N .Since 2∆M/N = ∆m, we can consider Equation ( 14) as an integration over m.We have Next, we define We have Notice that the integration over ζ is just Γ(k + 1) = k!.Thus, we have Analogously to Equation ( 12), the average cooperator payoff in a generation exceeds the average defector payoff in a generation if Rearranging Equation ( 15), we have We next perform simulations for which the total number of rounds per generation, M , is a random variable.The success of cooperation depends on the rate, p, at which information about the defectors is transferred among cooperators, on the fraction, f , of cooperators in the population, and on the average number of rounds per generation, m, that an individual engages in.The average number of rounds that an individual engages in during any single generation, m, is a random variable and is given by m = 2M/N .
Simulation results for a variable total number of rounds per generation are shown in Figure 4c,d.In Figure 4c, the probability of another round, W , is 399/400.Therefore, the average number of rounds per individual in a generation, m, is equal to 8. We plot B(1/2, 8; p), B(2/3, 8; p), and B(1, 8; p).The agreement between theory and simulation is excellent.In Figure 4d, the probability of another round, W , is 799/800.Therefore, the average number of rounds per individual in a generation, m, is equal to 16.We plot B(1/2, 16; p), B(2/3, 16; p), and B(1, 16; p).The agreement between theory and simulation is again excellent.
Note the subtle distinction between the models investigated in Figure 4a,b and Figure 4c,d.In Figure 4a,b, the total number of rounds per generation is fixed and is equal to M , and the average number of rounds per individual in a generation is equal to m = 2M/N .In Figure 4c,d We expect Equation (17) to yield a necessary condition on b/c for cooperation to be successful.Rearranging, we arrive at the condition If the total number of rounds, M , in a generation is a random variable, then we propose a variation of Equation ( 18 The simple condition, Equation (18), is plotted in Figure 4a,b (solid black curves).Notice that this form has the intuitively correct limiting behavior: The critical value of b/c increases as the rate of information transfer, p, decreases, and becomes infinite in the limit p → 0. The critical value of b/c decreases as the number of rounds, m, increases.As m → ∞, cooperators are able to accumulate an infinitely larger payoff than defectors in a single generation (provided that b − c > 0), and the critical value of b/c approaches 1.The simple condition, Equation (19), is also plotted in Figure 4c,d (solid black curves).
Discussion
We have studied a simple model for evolution of indirect reciprocity with private information.As information about the identity of defectors spreads among cooperators, defectors lose their fitness advantage.Interactions are optional; a cooperator can reject a possible game with a defector if it is cognizant of that defector's identity.Thus, cooperation can eventually prevail.
We have derived exact conditions on the benefit-to-cost ratio for cooperation to dominate.We have also derived exact conditions on the benefit-to-cost ratio for natural selection to oppose the fixation of defectors starting with a single defector.Agreement of our theoretical predictions for either a fixed or a variable total number of rounds (Equations ( 13) and ( 16), respectively) with simulation data is excellent.We have also deduced simple conditions on the benefit-to-cost ratio, Equations ( 18) and (19), which ensure the success of cooperation.
Our calculations and simulations lend insight into indirect reciprocity with an evolving distribution of reputations.Increasing the rate of information transfer among cooperators and increasing the number of games offered to each individual in a generation both facilitate the transition to a population of cooperators.Intriguingly, the distribution of the total number of rounds in a generation significantly affects the critical benefit-to-cost ratio needed for cooperators to be favored.This is evident in Figure 4, where one sees that having a constant probability of another round in a generation (as opposed to having an exactly fixed total number of rounds per generation) is more favorable to cooperation.
Variations of our model are also possible.For example, a defector's identity may be revealed in any round in which it is paired with a cooperator, even if the cooperator rejects the potential game.Notice that this modified model is significantly more conducive to the evolution of cooperation since information about defectors is transferred more rapidly.In this case, our critical benefit-to-cost ratios, Equations ( 13) and ( 16), would represent sufficient conditions for cooperators to succeed.We derive results for the modified model in Appendix B.
In another variation, there is a cost for entering a game.In this case, even defectors would want to avoid playing with other defectors.It is now plausible that also defectors reveal the identities of other defectors that they have met.This modification also allows for easier evolution of cooperation, because the identity of defectors is revealed faster.
There are also possibilities for implementation errors.For example, when a cooperator engages in a round with a defector, the cooperator could mistake the defector for being another cooperator.Such recognition errors may involve new information transfer regarding the defector's identity, which would reduce future exploitations of cooperators by that defector.Implementation errors could also result in erroneous information transfer if cooperators are mistakenly perceived as defectors.In this case, cooperators that are wrongly perceived as defectors might have their links with other cooperators severed for all subsequent rounds, which would make it more difficult for cooperators to thrive.However, a key point is that we endow cooperators with optional interactions.Therefore, on rare occasions, a cooperator might wrongfully view another cooperator as a defector and reject a potential game, but both individuals subsequently retain their strategies as cooperators.For this reason, we expect our derivations of critical benefit-to-cost ratios to be fairly robust provided that implementation errors are rare.Each of these modifications to our model is a potential topic for future study.
B. Modification to the Model: Loss of Links on Every Round
Here, we consider a modification to our model in which there is information transfer every time a cooperator and a defector are paired.A fraction p of the defector's remaining links with cooperators are broken on every round, even if the cooperator that is paired with the defector rejects the potential game.The formula analogous to Equation (1) for describing this modified model is x i,m = (1 − q m−1 )x i,m−1 + q m−1 x i−1,m−1 (B1) Here, x i,m is the probability of the solitary defector playing i games in m rounds when a fraction p of links are lost on every round.Equation (B1) holds for any m ≥ i ≥ 2 with the convention that x m,m−1 = 0.For i = 1, we have the recurrence formula To solve Equation (B1), we introduce a z-transform on the index i: Notice that the only nonzero value of x i,m for m = 1 is x 1,1 = 1.Using Equation (B2), we have From Equations (B2) and (B1), the z-transform of x i,m with m ≥ 2 is xm (z) = (1 − q m−1 )x m−1 (z) + q m−1 z −1 xm−1 (z) We see that xm (z) can be expressed recursively in terms of xm−1 (z): Iterating Equation (B4) and substituting Equation (B3), we find (1 − q n−1 )z + q n−1 From the above formula, we have the following by direct computation: (1 − q n−1 )z + q n−1 | z=0 (B5) Equation (B5) can be simplified: We can compute the average number of games played by the defector by substituting Equation (B6) into Equation (10).In Figure B1, F m (q)/m is plotted for the original model and for the modified model.Notice that F m (q) is larger for the original model than for the modified model for all m ≥ 3.
Figure 2 .
Figure2.Schematic showing how the fraction of active links between the defector and cooperators decreases as additional exploitations occur.At the start of a generation, all links to cooperators are active.On the first round, the defector plays a game, and a fraction q of the links remain active after the game.On each round thereafter, the defector plays another game with probability q i , where i is the number of games that have been played to that point.
Figure 4 .
Figure 4.The critical value of b/c for cooperation to be favored over defection in various settings.The green circles represent measured critical values of b/c for the average cooperator payoff to exceed the average defector payoff when f = 1/2.The "+" symbols represent measured critical values of b/c for cooperators to outnumber defectors with weak selection and with weak mutation when their numbers are averaged over many successive generations.The red squares represent measured critical values of b/c for the average cooperator payoff to exceed the average defector payoff when f = 2/3.The "×" symbols represent measured critical values of b/c for defectors to fix with probability less than 1/N when starting with a single defector with weak selection and with no mutation.The black triangles represent measured critical values of b/c for the average cooperator payoff to exceed the average defector payoff when f = 1 (i.e., a single defector in a large population of cooperators).Panels (a,b) are for a fixed total number of rounds per generation, M , while panels (c,d) are for a variable total number of rounds per generation, M , where the average of M over generations is M = 1/(1 − W ). Error bars are roughly the size of the data points.
, the probability of another round between two random individuals in a generation is equal to W .The total number of rounds per generation, M , for Figure4c,d is therefore a random variable, with average given by M = 1/(1 − W ), and the average number of rounds per individual in a generation is equal to m = 2M /N = 2/(N (1 − W )). Notice that the critical values of b/c are lower in Figure4cthan in Figure4aand are also lower in Figure4dthan in Figure4b.
3. 4 .
Simple Lower Bound on b/c We now seek a simple condition on b/c for cooperation to be successful.Consider an individual in an infinitely large population of cooperators.The individual participates in m + 1 rounds with the infinite population of cooperators and has two choices of strategy: The focal individual may cooperate in all m + 1 rounds, receiving total payoff (b − c)(m + 1).Or the focal individual may defect on the first round but cooperate on all m subsequent rounds, receiving total payoff b + qm(b − c).The strategy for the focal individual to always cooperate delivers a higher payoff after m + 1 rounds if
Figure B1 .
Figure B1.Average number of games that the defector plays in the original model (black curves) and the modified model (red curves).The expected number of exploitations normalized by the number of rounds, F m (q)/m, is plotted against the probability, q, to maintain each link when (a) m = 3 and (b) m = 4.The inset in each panel shows the difference, ∆, between the black and red curves. | 2018-04-03T00:46:52.786Z | 2015-09-30T00:00:00.000 | {
"year": 2015,
"sha1": "52d4f3b50d4341cf00f19069e68ca93a0a622aa3",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2073-4336/6/4/438/pdf?version=1443617699",
"oa_status": "GOLD",
"pdf_src": "Crawler",
"pdf_hash": "52d4f3b50d4341cf00f19069e68ca93a0a622aa3",
"s2fieldsofstudy": [
"Economics"
],
"extfieldsofstudy": [
"Economics",
"Computer Science"
]
} |
253278715 | pes2o/s2orc | v3-fos-license | Prognostic Factor of Lower Limb Amputation among Diabetic Foot Ulcer Patients in North-East Peninsular Malaysia
Lower limb amputation (LLA) is a common complication of diabetic foot ulcer (DFU), which can lead to a higher 5-year mortality rate compared to all cancers combined. This study aimed to determine the prognostic factors of LLA among DFU patients in Kelantan from 2014 to 2018. A population-based study was conducted using secondary data obtained from the National Diabetic Registry (NDR). There were 362 cases that fulfilled the study criteria and were further analysed. The prognostic factors were determined by Multiple Cox Proportional Hazards Regression. There were 66 (18.2%) DFU patients who underwent LLA in this study, while 296 (81.8%) were censored. The results revealed that the factor leading to a higher risk of LLA was abnormal HDL-cholesterol levels (Adj. HR 2.18; 95% CI: 1.21, 3.92). Factors that led to a lower risk of LLA include DFU in patients aged 60 or more (Adj. HR 0.48; 95% CI: 0.27, 0.89) and obesity (Adj. HR 0.45; 95% CI: 0.22, 0.89). In conclusion, our model showed that abnormal HDL cholesterol was associated with a 2 times higher risk of LLA when adjusted for age and BMI. Any paradoxical phenomena should be addressed carefully to avoid wrong clinical decision making that can harm the patient.
Introduction
The global trend of Diabetes Mellitus (DM) is worrying. The total number of people with DM is projected to continue increasing. By 2030, the overall incidence of DM is expected to rise to 578 million (10.2%), and by 2045, it will rise to 700 million (10.9%). DM also contributes to global NCD deaths, having been identified as the fourth most common cause of NCD death at around 1.6 million [1]. DFU is one of the commonest diabetic complications. It is characterised as the ulceration or destruction of foot tissues caused by neuropathy or peripheral arterial disease (PAD) among people with DM [2].
The International Diabetic Federation reported the global prevalence of DFU to be 6.3%, but it could range up to 14% [3][4][5]. The prevalence of DFU in Malaysia has generally ranged from 5-10%, but in a study conducted in 2018, the prevalence of DFU reached as high as 42% among diabetic patients in Kuala Lumpur [6,7]. D. G. Armstrong et al. found that the most significant risk factor for a DFU was a healed DFU [8]. DFU was the leading cause of LLA, and 85% of poorly healed DFU cases ended in LLA because the neuropathy and PAD persisted and readily caused another ulcer if proper foot care was not established [8,9]. Furthermore, Zhang et al. reported that 25% of diabetic patients would develop diabetic foot ulcers in their lifetime [4].
In their study in 2007, D. G. Armstrong et al. reemphasised the significant impact of DFU and LLA on the mortality rate. Their 5-year mortality data revealed that overall LLA led to a higher 5-year mortality rate compared to all cancers combined. The 5-year mortality with major LLA was 56.6%, and minor LLA led to 46.2% mortality. Hence, they concluded that both DFU and LLA were independent risk factors for premature mortality [10].
The increasing number of undiagnosed DM cases is a worrying trend, as it is expected that more patients will have already established complications, such as DFU, upon diagnosis. The rising prevalence of diabetes, especially undiagnosed diabetes, will lead to an increasing number of diabetic feet at risk of amputation. By identifying the prognostic factors of LLA, a better strategy can be formulated to manage diabetic foot ulcers and prevent LLA from taking place.
Materials and Methods
Study design: This study was a retrospective cohort study that was conducted from 1 October 2020 to 31 May 2021. The sampling frame was all DFU patients in Kelantan registered in the NDR from 2014 to 2018 who fulfilled the study criteria. Patients with DFU diagnosed from 1 January 2014 to 31 December 2018 who were registered in NDR and were treated in the Government Health Facility in Kelantan were included in this study. Patients with an error in the diagnosis date for DFU/LLA or patients with >30% incomplete data were excluded from the study sample.
Sample Size: The sample size to determine the prognostic factors of LLA among DFU patients was calculated using two median time options in PS Software: Power and Sample Size Calculation version 3.1.6. The two median times entered were 14 months and 20 months [11]. The accrual time was 60 months, and the follow-up time was 12 months, which led to a calculated sample size of 360. After adding a 10% dropout rate, the total sample size was 400 patients.
Data collection: This study was conducted using secondary data, whereby the data were extracted from the National Diabetic Registry (NDR), which consists of registry datasets and clinical audit data compiled by the Kelantan State Health Department. The NDR is an innovation of the Ministry of Health (MOH) for diabetic surveillance. It is a web-based registry that can enable a systematic approach to collecting data and monitoring the quality of care and type 2 DM patients managed in the MOH facility [12]. Names and identification numbers were made anonymous and coded into numbers for analysis purposes. Proforma was used as a checklist of data required to be analysed in this study while ensuring confidentiality. The independent variables were extracted from the NDR and DM Audit Data. The data from DM Audit Data were obtained at or after the diagnosis of DFU was made and before amputation took place to determine the hazard relationship. Statistical Analysis: These data were downloaded into Microsoft Excel and later exported to SPSS Software version 26 for further analysis. Descriptive statistics were used to summarise the characteristics of each sample. Finally, the categorical data were presented as frequency (n) and proportion (%). Normality tests were conducted using SPSS and manual calculations. Given that non-Malay cases represented very small numbers in the sample compared to Malay cases, ethnicities other than Malay, such as Chinese, Indian, Siamese, and others, were recoded into one category: "non-Malay". HDL cholesterol was categorised as normal and abnormal based on the cut-off point stated in the Clinical Practice Guideline on Management of Dyslipidaemia 2017, where less than 1.0 mmol for males and less than 1.2 mmol/L for females are considered normal [13]. This study used the survival analysis method. The event of interest was LLA. Hence, the duration from the diagnosis of DFU to the date when LLA occurred was called amputation-free time. This cohort was followed up from the date of diagnosis until the date of the event of interest, LLA. Patients who did not experience amputation, who were lost to follow-up, or who died before the event took place until 31 December 2019 were categorised as censored data. Univariable (Simple Cox Proportional Hazards Regression) and multivariable analyses (Multiple Cox Proportional Hazards Regression) were performed to determine the prognostic factors of LLA among the independent variables. Univariable analyses were used to determine which factors are significant at the univariate level, where the variables of choice to proceed with Multiple Cox Proportional Hazards Regression were those with a p-value less than 0.25 or those that were clinically significant. The variables of choice that were obtained from Simple Cox Proportional Hazards Regression were analysed further using Multiple Cox Proportional Hazards Regression. In this multivariate analysis, the variables of choice were those with p-value < 0.25 or those that were clinically significant and underwent variable selection using Backward and Forward LR methods. The best model selected was checked further for any two-way interactions. Once no interaction was identified, the preliminary main effect model was obtained. There were two assumptions to be fulfilled in the proportional hazards analysis. The first assumption was that the hazard function is constant over time. This assumption would compare the effect of two or more factors that were superior or were more hazardous to the outcome over time. The second assumption was that the hazard is proportional over time. This assumption is expressed as the hazard ratio, which is the ratio between two expected hazards.
Results
There were 598 DFU cases registered in the NDR from 2014 to 2018. After 236 cases were removed because they were duplicates or fulfilled the exclusion criteria, 362 remained for further analysis (Figure 1).
Hazards Regression were those with a p-value less than 0.25 or those that were significant. The variables of choice that were obtained from Simple Cox Pro Hazards Regression were analysed further using Multiple Cox Proportional Regression. In this multivariate analysis, the variables of choice were those with 0.25 or those that were clinically significant and underwent variable selecti Backward and Forward LR methods. The best model selected was checked furth two-way interactions. Once no interaction was identified, the preliminary m model was obtained. There were two assumptions to be fulfilled in the pro hazards analysis. The first assumption was that the hazard function is constant o This assumption would compare the effect of two or more factors that were su were more hazardous to the outcome over time. The second assumption was hazard is proportional over time. This assumption is expressed as the hazard rat is the ratio between two expected hazards.
Results
There were 598 DFU cases registered in the NDR from 2014 to 2018. After were removed because they were duplicates or fulfilled the exclusion crit remained for further analysis (Figure 1).
Descriptive Statistics of DFU Patients
The description of patients' profiles with DFU in Kelantan from 2014 to 2018 in Table 1. Most of the patients were of Malay ethnicity (96.7%), with a mean (S 59.0 (10.0) years old, and most were female (61.6%), obese according to BMI (55 non-smokers (92.8%). The ratio of males to females in this study was 1:1.6.
Descriptive Statistics of DFU Patients
The description of patients' profiles with DFU in Kelantan from 2014 to 2018 is shown in Table 1. Most of the patients were of Malay ethnicity (96.7%), with a mean (SD) age of 59.0 (10.0) years old, and most were female (61.6%), obese according to BMI (55.9%), and non-smokers (92.8%). The ratio of males to females in this study was 1:1.6. In terms of comorbidities, most DFU patients in this study had hypertension (81.1%) and dyslipidaemia (77.6%). On the other hand, only 26.9% had nephropathy, and 12.1% had ischaemic heart disease. For the biochemical parameters, 87.5% of DFU patients in this study had Hba1c levels of 6.5% or more, but only 55.5% had triglyceride levels of more than 1.7. In addition, 75.3% had an LDL-cholesterol level of more than 2.6, but only 37.3% had an abnormal HDL-cholesterol level. From the treatment perspective, most patients were on Statins (76.2%), Insulin (63.5%), and Metformin (65.7%), whereas only 29.6% were on Sulphonylurea.
Univariate and Multivariate Analysis
The univariate analysis results revealed that patients who were on Sulphonylurea (HR 0.50; 95% CI: 0.27, 0.94; p = 0.019) showed a statistically significant protective hazard ratio. Older age at diagnosis of DFU showed a significantly lower hazard ratio (HR 0.54; 95% CI: 0.33, 0.89; p = 0.016). Table 2 showed the overall result of univariate analysis. The results obtained from the univariate analysis were further analysed using multivariate analysis, which was Multiple Cox Proportional Hazards Regression. Variables with a p-value less than 0.25 were included in the variable selection. The preliminary final model was further analysed to ensure that there were no interactions and that all assumptions were met.
The final model for the LLA prognostic factors is shown in Table 3. After adjusting for the other variables, patients with DFU aged 60 or more had a 52% lower risk of LLA compared to DFU patients aged less than 60 (Adj. HR 0.48; 95% CI: 0.27, 0.89). Patients with DFU who were obese had a 55% lower risk of LLA than patients with DFU who had a normal BMI (Adj. HR 0.45; 95% CI: 0.22, 0.89). Patients with abnormal HDL cholesterol had a 2.18 times higher risk of LLA than patients with normal HDL-cholesterol levels (Adj. HR 2.18; 95% CI: 1.21, 3.92).
Discussion
Data from the National Diabetic Registry, a well-established public health monitoring system, were used in this retrospective cohort analysis. This registry's data include all DFU cases and their amputation status, including the date of diagnosis, and are maintained regularly. However, for this review, only DFU patients who were followed up by the Kelantan Government Health Facility and were diagnosed between 1 January 2014 and 31 December 2018 were chosen. As a result, the data used in this study are accurate for and descriptive of Kelantan's DFU patients.
Since the study's emphasis was on the prognostic factors of LLA, survival analysis was the chosen method. These analyses considered the unmeasured absolute amputation-free time among patients who had not had their limbs amputated at the end of the analysis. Some researchers have studied all types of amputation, major amputation, and survival after amputation [11,[14][15][16]. Therefore, LLA of all types was valid for use as the event of interest in this study. However, our model includes two results that seem contradictory to the current knowledge, where older age and obese BMI appear to be protective against LLA.
Literature searches mostly reveal that ageing can lead to a higher risk of LLA [15,16]. Established evidence suggests that older age leads to delayed wound healing. Ageing has been associated with delayed inflammation, angiogenesis, and epithelialisation, leading to delayed wound healing [17]. The first paradox in this study was that older age appeared to be protective against LLA. For healing to occur, biological capacity is not the only consideration, as compliance with follow-up for wound treatment is also crucial in determining the outcome of DFU. Older people have been shown in various literature studies to have a better compliance rate with follow-up and medications [18][19][20]. Moreover, in Malaysia, wound clinics have been established in most primary care clinics that these older patients can easily access. These clinics provide standardised wound care according to the national manual at an affordable fee [21]. Further studies can be conducted to prove the confounding effect of clinical follow-up on the outcome of DFU.
The second paradox found in this study was that obesity appeared to be protective against LLA. This phenomenon is popularly termed the obesity paradox, where obesity appears to have a positive effect on health outcomes. This observation has appeared in multiple observational studies involving study populations with existing chronic diseases [22,23]. However, the low-BMI status in chronic disease patients cannot be compared to having a low BMI in the general population. Having a low BMI for patients with existing chronic diseases can mean that the disease is poorly controlled, and hence, they are more likely to develop poor health outcomes, such as amputation [24]. As shown in this study, most had uncontrolled diabetes, where those with HbA1c of 6.5% or more were more likely to undergo LLA (87.5%) compared to those with HbA1c less than 6.5% (12.5%). Hence, this could be the explanation for the "obesity paradox" seen in this study.
HDL plays a key role in reverse cholesterol transfer (RCT). RCT is a process in which excess cholesterol is extracted from peripheral vessels and transferred back to the liver for disposal [25]. On the other hand, HDL has several other biological properties that help it defend against CVD. The antioxidative properties of HDL particles in the circulation, especially small, dense, protein-rich HDL3, can protect LDL from oxidative damage by free radicals in the arterial intima [26]. In addition, HDL can exert a direct cytoprotective effect on endothelial cells by inhibiting the suicide pathway that leads to endothelial cell apoptosis by lowering cysteine protease P32 (CPP32)-like protease activity. As a result, HDL protects against "injury", as defined in the atherogenesis "response-to-injury" hypothesis [27,28]. The protective hazard ratio in this study is consistent with other research in Japan, in which patients with lower HDL cholesterol were reported to have an association with a higher risk of minor and major amputations [29]. Kaneko et al. also reported that HDL cholesterol was associated with a 33% lower risk of LLA, but the result was not statistically significant [30]. This study complements these findings by proving that normal HDL cholesterol significantly lowered the LLA risk, even after controlling for age and BMI status.
This study used data from the NDR, a registry of diabetic patients who were followed up by government health clinics in Malaysia, specifically in Kelantan. The majority of DM patients in Malaysia are followed up by government health clinics. So, the NDR is a reliable tool for use as a surveillance system for diabetic patients, their complications, and their comorbidities in Malaysia. The registry is continuously maintained and updated by a well-trained staff in every health clinic in Malaysia. The Kelantan State Health Department conducts annual training in Kelantan to ensure that every clinic has at least one health worker capable of using the system and regularly updating necessary information. Every district in Kelantan also has its own medical and health officer in charge of the diabetic program. They supervise and ensure that the system is well-maintained at the district level. However, like other secondary data, the data quality was not free from uncertainty, as errors may still arise when staff key in their clinic data. Another limitation is that this study was unable to include diabetic patients who had their follow-ups in private clinics, as they are not registered in the NDR. However, the number is small compared to the 74.3% of diabetic patients in Malaysia who seek treatment in government health clinics [12].
There are other variables that were significant in other studies but were found to be non-significant in this study, such as smoking, the duration of diabetes, gender, comorbidities, and treatments such as Insulin and Oral Anti-Diabetic Medication. The non-significance could be due to the result of confounding effects, and perhaps a larger sample size is required.
Conclusions
The prognostic factor of LLA found in this study was abnormal HDL cholesterol, which was associated with a 2 times higher risk of LLA when adjusted for age and BMI. This prognostic factor can be used by healthcare professionals to give more attention to high-risk cases and to suggest appropriate action in their plan and strategy to prevent LLA among DFU patients. The prevention of DFU will not only improve the patient's quality of life but also help in increasing the patient's survival. Any paradoxical findings should be doubted and carefully examined to ensure that they do not lead to wrong clinical decisions that can harm patients. The confidentiality of the data was maintained at the highest level possible, as all forms were made anonymous. Data are presented as grouped data and do not identify the responders individually.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data from the National Diabetic Registry used to support this study have not been made available because of the privacy and confidentiality of the data. | 2022-11-04T17:52:21.719Z | 2022-10-31T00:00:00.000 | {
"year": 2022,
"sha1": "aef114365d2d8cf2823a1d9572d3be41a191df0a",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1660-4601/19/21/14212/pdf?version=1667195511",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "1355a306b1fb2f342cf946595f4cff528e083850",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
142942756 | pes2o/s2orc | v3-fos-license | A Cross Cultural Analysis of Textual and Interpersonal Metadiscourse Markers : The Case of Economic Articles in English and Persian Newspapers
This study was an attempt to investigate the functional role of textual and interpersonal metadiscourse markers in English and Persian Economic news reports. To this end, 10 news articles, 5 in each language, were randomly selected from the Economic sections of the leading newspapers published in 2013-2014 in Iran and the United States. Based on Kopple’s (1985) taxonomy, the type and frequency of metadiscourse markers used in the texts were analyzed to find out their functions in the text. The findings revealed that the textual markers used by Persian authors were considerably more frequent than those employed by the American writers. Interestingly, unlike the Persian writers, the American authors enlisted a larger number of interpersonal markers, which made their angle of the subject treatment different. It is evident that the differential use of metadiscourse markers by nationally different authors could be attributed to culturespecific norms governing the development and organization of discourse.
Introduction
Metadiscourse, a relatively new concept in the area of discourse analysis, refers to the ways speakers and writers address and communicate with their audience.It embodies the idea that writers and speakers should go beyond the ideational dimension, or propositional content, of text and speech in order to communicate their message effectively.As such, metadiscourse is a term used by most practitioners of the field so as to refer to the textual resources beyond sentence or pragmatic levels.Rather than only representing their ideas and information through language, writers need to consider the expectations and requirements of their receivers so as to engage them in the reading process and affect their understanding of the discourse produced.This view, as Dafouze-Milne (2008) maintains, is based on the assumption that writing is a social and communicative process and, in this regard, metadiscourse is used to organize and create a given text by involving the reader and expressing the author's inputs and stances.According to Halliday (1994), language can be functionally divided into three categories; namely, ideational, interpersonal, and textual.While ideational function refers to the information the writer or speaker communicates, interpersonal function is concerned with the way language establishes, maintains, and signals relations among people.Finally, textual function of language aims at creating coherent written and spoken texts related to its audience as well as its context.However, the focus of the present study is on the textual and interpersonal functions of the language known as metadiscourse.The concept of metadiscourse has been defined differently by various researchers.Vande Kopple (1985), as an example, defines metadiscourse as "discourse that people [writers] utilize to expand referential material and help their readers connect, organize, interpret, evaluate, and develop attitudes towards that material" (P.83).As Kopple puts it, writers work on two levels.On the primary level, the propositional content or the information about subject matter is satisfied; on the metadiscourse level, nothing is added to the content but the readers are engaged in finding out the message and the writer's views; in fact, writers focus on how they are communicating with the readers.Clearly, according to Vande Kopple, primary discourse is primarily concerned with the ideational function, whereas metadiscourse is related to the textual and interpersonal functions of language.Focusing on the definition provided by Vande Kopple (1985), in 1998, Hyland organized metadiscourse markers into two main categories, viz, textual and interpersonal.As Hyland maintains, the textual metadiscourse elements can be, based on the functions they serve in the text, further divided into five subtypes; namely, Endophoric Markers, Frame Markers, Logical Connectives, Code Glosses, and Evidentials.The first category, endophoric markers, can be described as the linguistic elements referring to the earlier materials in the text so as to support the argument and help readers understand the text better (e.g., as noted above).Frame markers are words or phrases that sequence parts of a text (e.g., first, then, at the same time) or change the topic (e.g., lets return to, now) or label the text stages (e.g., finally, to summarize).Logical connectives, as the name suggests, semantically connect the main ideas that are internal to text.They are mostly conjunctives and adverbial phrases such as in addition, but, therefore, and likewise.Code glosses are the next category of metadiscourse markers.As Hyland (2005, P. 52) states, code glosses are "textual devices that supply additional information by rephrasing, explaining or elaborating what has been said, to ensure the reader recovers the writer's intended meaning" (e.g., namely, in other words, such as).And finally, evidentials refer to the sources of information from other texts.The utility of evidentials by authors or researchers is mostly to provide evidence for their work by citing the works or ideas of other authors (e.g., according to X, as X puts it, X states that).The interpersonal metadiscourse markers, on the other hand, provide the writers with linguistic elements so as for them to express their attitudes and perspectives toward the propositional content of the text.These linguistic signals help writers to engage the readers in the text by addressing them directly.Along the same line with textual metadiscourse markers, Hyland (1998) also classified the interpersonal metadiscourse markers into five major categories: Emphatics, Hedges, Person Markers, Relational Markers, and Attitude Markers.Emphatics can be described as metadiscourse markers writers use in order to express their certainty concerning an idea, or to emphasize their claims in the text by such linguistic elements as certainly, definitely, and it is obvious.However, when writers are uncertain about the truth of their claims they employ linguistic entities like might, perhaps, it is possible, etc.These subtypes of metadiscourse markers are referred to as "hedges".There are also situations in which writers desire to convey their presence in the text.In such cases, they enlist linguistic items called "Person Markers"; examples of such markers include I, we, my, and mine.The other type of interpersonal metadiscourse markers is relational markers by which writers directly refer to or build relationship with the readers.Phrases like Dear reader, please consider, and note that, are but a few examples of relational markers.And the last category subsumed within the interpersonal metadiscourse markers classification includes "attitude markers".They are employed when writers are in need of communicating their perspectives and attitudes towards the propositional content of the text.For example, words or phrases such as surprisingly, I agree, and I hope, fall into such category.Accordingly, metadiscourse markers, as Hyland (2005) believes, are linguistic elements writers (or speakers) utilize to not only exchange the information, but also express their attitudes, personalities, and assumptions by addressing and interacting with the receivers of the message.Hyland further argue that in this way, "the writer is not simply presenting information about the suggested route by just listing changes of direction, but taking the trouble to see the walk from the reader's perspective" (P.3).Using metadiscourse markers in the text, writers would be able to instantiate the intended propositional content and their ideas both coherently and intelligibly for revealing the maze of their units of thoughts to the readers.Furthermore, metadiscourse markers would build an interaction between the reader and writer and account for the atmosphere and reader-friendliness of the text (Hyland & Tse, 2004).For instance, writers intelligibly communicate their own ideas when they employ illocution markers such as I recommend that, or I believe that; or they inform readers about the degree of their certainty in making a proposition with regard to a given idea by applying hedges like perhaps, might, and apparently as well as emphatics such as clearly, undoubtedly, and surely.
Review of Relevant Literature
Indubitably, metadiscourse is claimed to be an important area in discourse analysis in that it helps writers to convey their intended message effectively by creating a social and communicative interaction with the reader.Through using metadiscourse markers, writers would be capable of creating a coherent text and thus increasing the efficiency of the text.As a result of these and other merits of the utility of metadiscourse markers in the text, metadiscourse analysis has recently captured the attention of many practitioners of the field and a wide range of empirical studies have been conducted in various genres and contexts including, but not limited to, academic writing (Hyland & Tse, 2004;Simin & Tavangar, 2009;VahidDastjerdi & Shirzad, 2010), research articles (Abdi, 2002;Zarei & Mansouri, 2011), and newspaper discourse (Abdollahzadeh, 2007;Hashemi & Golparvar, 2012).
The first, and most important, area of investigation regarding the role of metadiscourse markers in the text has focused on the persuasive function of these linguistic elements.Metadiscourse markers play an important role in persuasive writing and act as persuasive tools which writers utilize in their texts to influence the readers.Hyland (2005) addresses this issue stating that metadiscourse markers, if used properly, might contribute to the art of persuasion by virtue of the fact that they foster logical appeal once they directly associate ideas with arguments, and indicate approval if they are in line with the reader's way of thinking.In this regard, Dafouz-Milne (2003) examined the use of interpersonal and textual metadiscourse markers in the opinion columns of two elite newspapers: the Spanish El Pais and the British The Times.Using 40 opinion articles, 20 in each newspaper, as her corpus materials, Dafouz concluded that the frequency of textual metadiscourse markers (e.g., logical connectives, code glosses) used by Spanish writers was more than that of English reporters, whereas the British writers used more instances of interpersonal markers (e.g., hedges, attitude markers) compared with the Spanish news reporters.Abdollahzadeh (2007) also tried to investigate the ways writers with different cultures organize their written texts by using metadiscourse markers.To this end, he analyzed 53 Persian and English (British and U.S) newspaper editorials.The results indicated no significant difference between Persian and English editorials for the metadiscourse subtypes of text connectives, attitude markers, and person markers.Concerning the interpersonal metadiscourse markers, however, it was found that English editorials used more hedges while Persian editorials used more emphatics.The researcher concluded that the frequent use of emphatics by the Persian editorial writers was most likely due to an Iranian tradition of valuing and abiding by the rules of those in power without questioning them or without expressing uncertainty about social and, especially, religious issues.On the other hand, the heavy use of hedges by the English editorial writers was related to their being more considerate and polite to their readers.
Along the same line, Noorian and Biria (2010) investigated the frequency and degree of the use of interpersonal metadiscourse markers in persuasive discourse.In this study, the metadiscourse markers used in English opinion articles written by American and Iranian columnists were compared.The results revealed that interpersonal metadiscourse markers were present in both sets of corpora, but there were significant differences between the two groups regarding the occurrences of interpersonal markers, especially in the case of Commentaries.The findings also suggested that different factors such as culture-driven preferences, genre-driven conventions, and Iranian EFL writers' extent of foreign language experience interacted in choosing the interpersonal metadiscourse markers by the columnists.
Regarding the frequency of the use of different metadiscourse markers in newspapers discourse, in 2012, Hashemi and Golparvar investigated the textual and interpersonal metadiscourse markers used in Persian news reports.The results indicated that metadiscourse markers were frequently utilized in Persian news reports and also the number of textual metadiscourse markers was much higher compared with interpersonal metadiscourse markers.Along the same line, Yazdani, Sharifi, and Elyassi (2014) tried to examine the role interpersonal metadiscourse markers play in Political English and Persian news articles.Choosing 30 news articles extracted from both languages and using Hyland's (2005) classification of interpersonal metadiscourse markers, they tried to discover the existing differences between the two languages.The results indicated that there was a statistically significant difference between two sets of data in terms of the frequency of interpersonal markers.The findings reflected that American journalists tended to use these linguistic items more frequently in their news articles.Moreover, it was concluded that Iranian writers, in formal contexts, did not show a tendency towards using personal markers such as I, we, and our, in writing news articles; instead, they preferred to apply third person pronouns and passive structures to address the reader about their ideas.
Obviously, metadiscourse analysis has been dealt with in a number of empirical studies focusing on different genres and contexts, as was mentioned above.However, few studies, if any, have considered the role of metadiscursive elements in Economic newspaper reports.As Crismore and Abdollahzadeh (2010) claim, little attention has been paid to the newspaper discourse which is considered as an important genre in the field.As such, this study set out to compare the frequency of different types of textual and interpersonal metadiscourse markers in English and Persian Economic news articles.The logic behind considering this genre was that Economic news reports are undoubtedly among the most widely read newspaper articles by Iranian people nowadays because of the current economic situation in Iran.
Materials
This study was an attempt to comparatively investigate the type and frequency of metadiscourse markers employed in Economic articles in English and Persian newspapers.Accordingly, the corpus data were collected from the online archive (2013-2014) of the leading newspapers published periodically in Iran and United State.Using random sampling, from among 50 articles, a total of 10 news articles, 5 in each language, were selected.The logic behind using random sampling relied on the fact that it would help the researchers bring the problem of particularity of writers' styles under control.To have an equal amount of data in both languages, the first 1000 words from each text were analyzed.The texts were also selected from the same field; i.e.Economy.This would ensure comparability of the texts because, as some practitioners of the field (e.g., Thompson, 2001;Dafouz-Milne, 2003) maintain, the type and frequency of the linguistic elements such as metadiscourse markers in a given text may be considerably influenced by the topic of the text.The most reliable national news agencies like IRNA, IPNA, and ISNA were considered as the sources for selecting the Persian news articles.The English articles were also selected from the leading American newspapers such as The New York Times, Washington Times, and USA Today.
Data Collection Procedures
To reach the purpose of the study, 10 news articles (5 from each language) were randomly selected from among 50 articles in the Economic sections of the newspapers.The analysis of the type and frequency of the metadiscourse markers used in the selected texts was based on the Kopple's (1985) taxonomy, which would follow.' Taxonomy Kopple (1985) introduced the first sophisticated classification of the metadiscourse markers based on their functions in the text.In his classification system, Kopple divided two main categories of textual and interpersonal items into seven different classes, which would follow.The first four categories are considered as textual metadiscourse markers and the last three categories as interpersonal metadiscourse markers.
Vande Koppel's Metadiscourse Markers
1.Text Connectives: these are linguistic items employed by writers in order to link different parts of a text or different ideas presented in the text.they include phrases or words that show the sequence of the ideas (e.g., first, then, after that), expressions that indicate logical or temporal relationships (e.g., as a consequence, at the same time, however), reminders of the previous ideas or materials (e.g., as mentioned before, as we saw earlier), expressions that point to the upcoming materials (e.g., details will be discussed in the forthcoming chapter), and words or phrases that point to the main topic of a sentence or text as a whole (e.g., with regard to, considering, regarding).2. Code Glosses: these items are used by writers to ensure the readers understand the meanings of specific elements, phrases, or idioms (e.g. that is, it means that, in other words).Code glosses are further divided into three main subcategories of Defining (e.g., it is defined as), Explaining (e.g., it means that, namely), and Delimiting (e.g., somewhat, to a certain degree) 3. Illocution Markers: They can be defined as linguistic elements used in a given text so as to signal the readers what specific speech or discourse act the writer is performing in a given point in the text (e.g., to sum up, I hypothesize that, the purpose is that,).4. Narrators: They are expressions that inform readers who has said or written a given idea or comment (e.g. as X states, according to X, X and Y claimed that). 5. Validity/modality markers: as a category of interpersonal markers, validity markers are elements that are employed by writers in order to express their certainty or uncertainty about a given idea or comment presented in the text.By applying such linguistic elements in their texts, authors would be able to indicate the extent to which the text's content is valid.Kopple classified them into three subcategories; namely, Hedges (e.g., perhaps, probably, it is possible that), Emphatics (e.g., indubitably, it is obvious that, clearly), and attributors that are used to refer explicitly to the source of information presented in a text in order to persuade readers (e.g. according to Prime Minister, as X remarked).6. Attitude Markers: They help the readers to grasp the author's attitude toward a specific idea or a given material in the text (e.g. it is interesting, surprisingly, considerably).7. Commentary: They are applied by authors so as to establish a relationship with the readers/audience (e.g.you may not agree that, dear reader).
To analyze the type and frequency of the metadiscourse markers utilized by Persian authors in their texts, the researchers made an attempt to take into consideration the proper Persian equivalents for each subcategory introduced in Kopple's taxonomy.Some instances of such equivalents and their corresponding English items for textual metadiscourse markers are represented in Table 1.For the purpose of specifying the type and frequency of occurrence of metadiscourse markers in each text, the researchers employed the opinions of two specialists in the field.Their opinions were of great help in not only identifying the metadiscourse markers but also consistently coding the related texts.The inter-rater reliability of the coders was estimated, which turned out to be 0.80.Having identified and categorized the metadiscourse markers, the researchers conducted a quantitative analysis in order to determine the frequency of different types of textual and interpersonal metadiscourse markers employed in each set of texts.It should be noted here that although a given metadiscourse marker may functionally play different roles in different contexts, the primary function of each marker in the related context was determined as the basis for the analysis of the metadiscourse markers used in these texts.
Finally the collected data were analyzed using non-parametrical means, viz, Mann-Whitney U test, to see whether or not the differences between the two sets of data in terms of frequency of metadiscourse items were significant.The reason for selecting Mann-Whitney U test was that the metadiscourse elements employed in the sample news articles did not enjoy a normal distribution.
To conclude, in this study, the metadiscourse element used in both English and their corresponding Persian texts were first qualitatively analyzed based on their function in the related context so as to specify and classify them into different categories.In the next stage, the collected data were quantitatively analyzed in order to determine their frequency of occurrence in a given text and to realize whether there was a statistically significant difference between two sets of corpus data in this respect.
Results and Discussion
As mentioned, the collected data were analyzed via non-parametrical means (Mann-Whitney U test).Regarding textual metadiscourse markers, as it can be induced from Table 3, the findings demonstrate that there is a statistically significant difference in the Announcements frequency (p=.034) between American and Persian economic news reports.Moreover, both sets of texts enjoyed a high frequency of occurrence of Logical Connectors, with the Persian writers using more such elements in their texts than American writers (70 vs. 50 items, respectively).Within the categories of textual markers, illocution markers were used the least in both sets of data.For further analysis of each category and sub-category of textual metadiscourse markers, see Table 3.With respect to interpersonal metadiscourse markers' categories and sub categories, the results of statistical analysis, as shown in Table 4, reveal that most of the markers are used to the same extent in both sets of data, with the exception of Attributors.Although American authors indicated more tendency towards using attributors in reporting the Economic ALLS 5(2):59-66, 2014 64 news, but the difference is not too much to be statistically significant.The only statistically significant difference in terms of frequency of interpersonal metadiscourse markers is found in Emphatics (p=.041).Accordingly, when it comes to employing emphatic items (e.g., clearly, undoubtedly) in their texts, American and Persian writers act differently.Figures in the Table 4 show that Persian authors are more in favour of using such linguistic elements in their texts than are their corresponding American writers.Finally, Commentaries were the least frequent category of interpersonal markers both groups of writers tended to employ in writing Economic news reports.5 below illustrates the overall percentages of the textual and interpersonal metadiscourse markers' macrocategories used in Economic texts written by both American and Persian authors.In the vein of textual macrocategories, in the texts written by Iranian writers, text connectives are the most numerous markers (50.30%), followed by narrators (7.57%), code glosses (6.96%), and finally illocution markers (.6%).In the American group's texts, text connectives are also more frequently used (38.32%)than other markers, followed by code glosses (8.71%), narrators (6.96%), and finally illocution markers (1.4%).Following the interpersonal metadiscourse markers' macro-categories, in texts written by Iranian writers, validity markers capture the most proportion (26.06%), followed by commentary (4.54%), and at last attitude markers (3.93%).Similarly, in the texts written by corresponding American authors, validity markers are the most numerous interpersonal marker employed (33.10%), followed by commentary (6.27%), and finally attitude markers that are situated in the last place (5.57%).To conclude, the results of the statistical analysis represented in Tables 3, 4, and 5 clearly reveal that there exist many similarities between two sets of data concerning the frequency of different metadiscourse markers.Regarding the textual markers in the macro-category of Text Connectives, the logical connectors appear to be the most frequent items utilized in the Economic texts written by both American and Persian authors.Alternatively, neither American nor Persian writers seem to favour using illocution markers in their texts.The only statistically significant difference between the two groups is found in the frequency of the announcements (p=.034) which are a subcategory of text connectives.In fact, the frequency of announcements employed by Persian authors was twice more than that of American writers.
It is clearly observed that American authors utilized all subcategories of interpersonal markers slightly more frequently in their texts except for emphatics.In this case, as Table 4 illustrates, the difference between two groups is statistically significant (p=.041).In fact, Persian writers made use of emphatics in their texts far more frequently than the American group.Contrary to Persian group, American writers appear to use hedges more frequently in their texts in order to show uncertainty about a given idea.
As Hyland (2005) maintains, metadiscourse markers are universal features of texture by which authors would be able to explicitly organize and evaluate their texts to communicate with the readers.This claim is evidently substantiated by the results in this study as both groups of writers made a logical use of these linguistic elements.Furthermore, the findings also indicate that both American and Persian authors used textual metadiscourse markers far more frequently in their texts than interpersonal markers.This finding is also in line with the claim made by Hempel and Degand (2008) on the importance of textual markers used in various texts.In this regard, Hempel and Degand believe that textual metadiscourse resources are the authors' conscious stratagem in constructing the propositional content which they aim to convey to the addressee.In the vein of textual metadiscourse markers, the results of the analyses reveal that text connectives are the most frequent items used by Persian writers compared to the American group.This is acknowledged by the results found in the study conducted by Simin and Tavangar (2009).Based on their findings, they concluded that all Iranian EFL writers with different proficiency levels used text connectives as the most frequently used textual markers in their English texts.
By contrast, the findings of the present study suggest that interpersonal markers are used more frequently in the Economic texts written by American writers than those written by Persian authors (see Table 5).This is quite consistent with the results of the studies conducted by Abdollahzadeh (2003) and Yazdani, Sharifi, and Elyassi (2014).In both projects, researchers made a comparative investigation of interpersonal metadiscourse markers employed by both American and Iranian authors in their texts.The results indicated that American writers tended to use these items in writing academic texts more than that of Iranian authors.One possible reason for this tendency, according to Leki (2002), would be that the writers' cultural background has a great bearing on their writing styles.In fact, American authors are so concerned about the reader-writer interaction that they prefer to use more interpersonal markers in their texts in order to establish a strong solidarity with their target audience.This idea is supported by the results of this study in that American writers appeared to use hedges more frequently in their texts compared to Persian writers.As Hyland (2005) claims, hedges play a pivotal role in creating rapport between authors and their respective audiences.Contrary to American group, Persian writers appear to use more emphatics (e.g., certainly, undoubtedly) in their texts in order to show they are certain about a given idea.Such a difference is clearly indicative of the fact that cultural differences are certainly at work in text creation.This reality is appreciated in the study carried out by Noorian and Biria (2010) in which they concluded that while Persian writers favoured the use of emphatics in their texts, the American group showed a great tendency towards using hedges.Based on their findings, they claimed that Persian writers tend to be more assertive in their writings, whereas the American authors tried to be more polite to their readers by limiting the use of emphatics in their texts.
Conclusion
This study set out to explore the similarities and differences between English and Persian Economic news reports in terms of the frequency of textual and interpersonal metadiscourse markers used.Using 10 news articles, 5 from each language, the researchers analyzed the data based on the Kopple's (1985) taxonomy of metadiscourse markers.
Regarding the textual metadiscourse markers, the results of the analyses revealed that both groups of writers utilized textual markers, especially text connectives, far more frequently in their texts compared to interpersonal markers.Surprisingly, the only statistically significant difference between two sets of data was found in the frequency of Announcements (p=.034).
The results of interpersonal markers, on the other hand, indicated that the American group used these items slightly more than Persian writers.However, there was a difference between two groups in the case of certainty markers.In fact, as it has been acknowledged in similar studies (Abdollahzadeh, 2007;Noorian & Biria, 2010), while the Persian authors appeared to use more emphatics in their texts, the American group tended to make use of hedges more in writing the Economic news reports.Accordingly, it can be concluded that contrary to Persian writers, American Authors may be less assertive, more conservative, and more inclined to express their affective values in their writings.In the case of American writers, the dominance of the use of interpersonal markers can be attributed to the cultural background of the writers in that they show more tendencies towards establishing reader-writer rapport in their texts.
Due to the fact that metadiscourse is a branch of pragmatics, teachers and practitioners of the field should pay more attention to this aspect of language because, as Crismore, Markkanen, and Steffensen (1993) maintain, gaining knowledge in this area is rather difficult.As such, metadiscourse studies such as this may be of great help to both foreign language teachers and learners by revealing the possible problematic areas in the utility of metadiscourse markers in the texts.Students need to become familiar with the concepts of cohesion and coherence in the text and the only way to reach this end is through learning the functional roles of textual and interpersonal metadiscourse markers in different contexts and genres.
Table 1 .
Sample Persian equivalents for textual metadiscourse markers
Table 5 .
Results for textual and interpersonal metadiscourse markers' macro-categories | 2019-05-03T13:06:23.927Z | 2014-04-01T00:00:00.000 | {
"year": 2014,
"sha1": "60bd39041831f6d43a5e9156f3fe4f60ada125e3",
"oa_license": "CCBY",
"oa_url": "http://www.journals.aiac.org.au/index.php/alls/article/download/305/262",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "60bd39041831f6d43a5e9156f3fe4f60ada125e3",
"s2fieldsofstudy": [
"Linguistics"
],
"extfieldsofstudy": [
"Psychology"
]
} |
40048950 | pes2o/s2orc | v3-fos-license | Efficacy and Safety of IDegLira in Participants with Type 2 Diabetes in India Uncontrolled on Oral Antidiabetic Drugs and Basal Insulin: Data from the DUAL Clinical Trial Program
Introduction The efficacy and safety of insulin degludec/liraglutide (IDegLira) has been evaluated in the Dual Action of Liraglutide and Insulin Degludec in Type 2 Diabetes (DUAL) phase 3 clinical trial program. In this post hoc analysis, we compared the efficacy and safety of IDegLira in the Indian subpopulation with the results from the global trial population of DUAL trials. The analysis includes participants uncontrolled on oral antidiabetic drugs (OADs) in DUAL I and DUAL IV and participants uncontrolled on basal insulin and OADs in DUAL II. Methods Three phase 3 trials were included in the analysis: DUAL I extension (IDegLira vs. insulin degludec or liraglutide 1.8 mg in participants uncontrolled on metformin ± pioglitazone; 52 weeks; n = 1663), DUAL IV (IDegLira vs. placebo as an add-on to a regimen of sulfonylurea ± metformin; 26 weeks; n = 435) and DUAL II (IDegLira vs. insulin degludec in participants uncontrolled on basal insulin + OADs; 26 weeks; n = 398). There were 251, 64 and 64 participants, respectively, at the Indian sites. Results In the Indian subpopulations, the reductions in glycated hemoglobin (HbA1c) with IDegLira were substantial [DUAL I: 1.96% (−21 mmol/mol); DUAL IV: −1.40% (−15 mmol/mol); DUAL II: −2.20% (−24 mmol/mol)] and significantly greater than those in the comparators in each trial. IDegLira was generally weight-neutral after the administration of OADs (−0.3 and +0.6 kg in DUAL I and DUAL IV) and resulted in weight loss after the administration of basal insulin (−2.1 kg in DUAL II). Hypoglycemia rates were 1.98, 1.08 and 0.37 events/patient-years of exposure (PYE) for IDegLira, insulin degludec and liraglutide in DUAL I, 4.06 and 0.36 events/PYE for IDegLira and placebo in DUAL IV and 1.16 and 0.83 events/PYE with IDegLira and insulin degludec in DUAL II. Conclusions Results from the Indian subpopulations reflect those of the global study populations, supporting IDegLira as an effective and safe treatment option for people with type 2 diabetes inadequately controlled on OADs or basal insulin + OADs in the South Asian population. Trial Registration ClinicalTrials.gov identifier, NCT01336023 (DUAL I), NCT01392573 (DUAL II), NCT01618162 (DUAL IV). Funding Novo Nordisk A/S, Bagsvaerd, Denmark.
results from the global trial population of DUAL trials. The analysis includes participants uncontrolled on oral antidiabetic drugs (OADs) in DUAL I and DUAL IV and participants uncontrolled on basal insulin and OADs in DUAL II. Methods: Three phase 3 trials were included in the analysis: DUAL I extension (IDegLira vs. insulin degludec or liraglutide 1.8 mg in participants uncontrolled on metformin ± pioglitazone; 52 weeks; n = 1663), DUAL IV (IDegLira vs. placebo as an add-on to a regimen of sulfonylurea ± metformin; 26 weeks; n = 435) and DUAL II (IDegLira vs. insulin degludec in participants uncontrolled on basal insulin ? OADs; 26 weeks; n = 398). There were 251, 64 and 64 participants, respectively, at the Indian sites. Results: In the Indian subpopulations, the reductions in glycated hemoglobin (HbA 1c ) with IDegLira were substantial [DUAL I: 1.96% (-21 mmol/mol); DUAL IV: -1.40% (-15 mmol/mol); DUAL II: -2.20% (-24 mmol/ mol)] and significantly greater than those in the comparators in each trial. IDegLira was generally weight-neutral after the administration of OADs (-0.3 and ?0.6 kg in DUAL I and DUAL IV) and resulted in weight loss after the administration of basal insulin (-2.1 kg in DUAL II). Hypoglycemia rates were 1.98, 1.08 and 0.37 events/patient-years of exposure (PYE) for IDegLira, insulin degludec and liraglutide in DUAL I, 4.06 and 0.36 events/PYE for IDegLira and placebo in DUAL IV and 1.16 and 0.83 events/PYE with IDegLira and insulin degludec in DUAL II. Conclusions: Results from the Indian subpopulations reflect those of the global study populations, supporting IDegLira as an effective and safe treatment option for people with type 2 diabetes inadequately controlled on OADs or basal insulin ? OADs in the South Asian population. Trial Registration: ClinicalTrials.gov identifier, NCT01336023 (DUAL I), NCT01392573 (DUAL II), NCT01618162 (DUAL IV).
INTRODUCTION
India has the second highest number of people with diabetes worldwide, with approximately 8.6% of the population reported to have this disease [1]. Glycemic control is poor in Indian diabetic patients, with a mean glycated hemoglobin (HbA 1c ) of 8.9 ± 2.1% (74 ± 23 mmol/mol) and only 19.7% of the Indian diabetic population achieving the American Diabetes Association/European Association for the Study of Diabetes HbA 1c target of \7% (\53 mmol/mol) [2,3]. In addition to the increased prevalence of diabetes, higher rates of diabetes-related complications and mortality are also observed in South Asian people [4]. The factors contributing to this situation are multifactorial and include a genetic predisposition to diabetes, increased visceral adiposity and insulin resistance in this population, which are compounded by lifestyle factors [5]. Despite these numbers, South Asian participants in trials of glucose-lowering therapies are under-represented [6]; hence, it is important to report trial data from diverse populations.
IDegLira is a novel, once-daily, titratable, fixed-ratio combination of insulin degludec (IDeg) and the glucagon-like peptide-1 receptor agonist (GLP-1RA) liraglutide (Lira) that has been developed for the treatment of people with type 2 diabetes [7][8][9]. Importantly, GLP-1RAs act in a glucose-dependent manner and are also commonly associated with a low risk of hypoglycemia and weight loss [10]. The efficacy and safety of IDegLira and the benefits of its complementary mode of action have been examined in the large global DUAL clinical trial program [7][8][9]11]. Here we describe a post hoc analysis of the efficacy and safety of IDegLira in participants from Indian trial sites included in the DUAL program. Our hypothesis was that the safety and efficacy of IDegLira in the Indian subpopulation would be similar to that demonstrated for the global population.
METHODS
This post hoc analysis used data from the IDe-gLira (Novo Nordisk, Bagsvaerd, Denmark) phase 3 trials which included participants with type 2 diabetes from Indian trial sites, namely, the DUAL I extension (52 weeks of data), DUAL IV and DUAL II trials, which included Indian sites [8,9,11].
Study Designs
The study designs of DUAL I extension, DUAL II and DUAL IV have been described previously [8,9,11] and include participants uncontrolled on oral antidiabetic drugs (OADs; DUAL I extension and DUAL IV) and participants uncontrolled on basal insulin ? OADs (DUAL II). The aim of DUAL I extension was to investigate the efficacy and safety of IDegLira versus its individual components IDeg or Lira alone in participants inadequately controlled by metformin ± pioglitazone over 52 weeks [8]. In DUAL IV IDegLira was compared with placebo, both added on to a therapeutic regimen of sulphonylurea ± metformin [9], and in DUAL II, IDegLira was compared to IDeg (maximum dose of 50 U) in participants inadequately controlled on 20-40 U of basal insulin ? one to two OADs [11]. Owing to differences in trial design, individual trial data were analyzed separately, rather than pooling the IDe-gLira data.
Endpoints and Statistical Analyses
The primary endpoint in each study was change from baseline in HbA 1c . Secondary efficacy endpoints included participants achieving a HbA 1c of\7% (\53 mmol/mol), change in body weight, insulin dose and laboratory-measured fasting plasma glucose (FPG). Safety endpoints included adverse events (AEs) and confirmed hypoglycemia [severe (unable to self-treat) and/ or plasma glucose level of \3.1 1 mmol/L]. Change from baseline in HbA 1c , FPG and body weight were analyzed using an analysis of covariance (ANCOVA) model with treatment and pre-trial medication as fixed factors, and baseline HbA 1c stratum (DUAL I only), substudy (DUAL I only) and baseline value as covariates-all performed on the full analysis set using last observation carried forward (LOCF) method to impute missing values. Attainment of a HbA 1c level of \7% was analyzed using a logistic regression model (with LOCF), with the same explanatory variables as used for the primary endpoint. The results of the analyses were also summarized descriptively for other endpoints (insulin dose, rates of hypoglycemia and other safety endpoints).
As a post hoc analysis, the original individual trials were not powered to perform statistical analyses in the Indian subpopulations. The study reported here is based on a post hoc analysis of previously conducted studies and does not involve any new studies of human or animal subjects performed by any of the authors.
RESULTS
The Indian subpopulations comprised 15, 16 and 15% of the global trial populations of the DUAL I extension, DUAL II and DUAL IV trials, corresponding to 251, 64 and 64 participants, respectively. Baseline characteristics of the global trial populations and Indian subpopulations are shown in Table 1. Overall, the baseline characteristics of the Indian subpopulations were largely comparable to those of the global trial populations, although the Indian participants had lower FPG and body mass index (BMI).
Insulin Dose
End-of-trial insulin dose was 35 U with IDegLira and 59 U with IDeg in the DUAL I extension trial, 25 U with IDegLira in the DUAL IV trial and 42 and 46 U with IDegLira and IDeg in the DUAL II trial.
Safety Endpoints
Adverse events for the global populations and Indian subpopulations are summarized in Table 2 [8,9,11,12].
Mean observed values are based on the full analysis set and LOCF inputted data. The odds ratio (OR) is estimated from a logistic regression analysis 9.4 (IDegLira) and 36.2 (placebo) events/100 PYE in DUAL IV. In DUAL II, the rates were 82.1 (IDegLira) and 20.7 (IDeg) events/100 PYE.
There were no severe hypoglycemic episodes in the Indian subpopulations of the three trials. Confirmed hypoglycemia rates were 1.98, 1.08 and 0.37 events/PYE in the IDegLira, IDeg and Lira groups, respectively, in the DUAL I extension trial (compared with 1.77, 2.79 and 0.19 events/PYE in the global trial population) and 4.06 and 0.36 events/ PYE with IDegLira and placebo, respectively, in DUAL IV (compared with 3.52 and 1.35 events/PYE in the global trial population). In DUAL II, confirmed hypoglycemia rates were 1.16 and 0.83 events/PYE with IDegLira and IDeg, respectively (compared with 1.53 and 2.63 events/PYE in the global trial population).
DISCUSSION
In the Indian subpopulations, IDegLira resulted in substantial HbA 1c reductions that were significantly greater than those achieved by all comparators. Mean end-of-trial HbA 1c with IDegLira was \7% (\53 mmol/mol) in all trials and more participants achieved a HbA 1c target of\7% (\53 mmol/mol) versus all comparators, consistent with the global trial populations [8,9,11]. [8,9,11]. Maximum dose of IDeg was 50 U in DUAL II. Mean observed values are based on full analysis set and LOCF inputted data. Treatment difference is estimated from an ANCOVA analysis IDegLira was generally weight-neutral when used as an add-on to oral therapy (DUAL I extension/DUAL IV) and resulted in weight loss when used after basal insulin (DUAL II).
Considering IDegLira alone, rates of confirmed hypoglycemia were consistent with those observed with IDegLira in the global trial populations. However, the rate of confirmed hypoglycemia with IDeg was substantially lower in the Indian subpopulations (by around 1.7/ 1.8 events/PYE) versus the global trial populations of DUAL I and II. As a result, the rate of confirmed hypoglycemia was numerically lower with IDeg versus IDegLira in the Indian subpopulations of DUAL I and II, while in the global trial population IDegLira resulted in a lower rate of hypoglycemia versus IDeg, reaching statistical significance in the DUAL I trial [8,9]. This difference might be explained by the small number of participants-and therefore hypoglycemic events-in the Indian subpopulations and by baseline differences. The hypoglycemia rates should also be considered in the context of the significantly lower mean end-of-trial HbA 1c achieved with IDegLira compared to comparators. IDegLira was generally well tolerated, with a lower rate of gastrointestinal AEs in the Indian subpopulation versus the global trial populations in DUAL I and IV, and a higher rate in DUAL II. Across all trials, the proportion of participants experiencing AEs was higher in the Indian subpopulation than in the global trial population, but the pattern of AEs did not differ [8,9,11].
An inherent limitation of this study is that it is a post hoc analysis of three randomized controlled DUAL trials that were not designed to compare the safety and efficacy of IDegLira in the Indian subpopulation versus the global population. As such, there were differences at baseline in terms of FPG and BMI with the Indian subpopulation compared with the global population. Furthermore, owing to trial design differences, individual trial data were analyzed separately rather than the IDegLira data being pooled, and therefore sample sizes were small. Differences between Indian and non-Indian populations were not analyzed as the Indian population constituted a relatively low proportion of the global population.
CONCLUSION
In conclusion, results from the Indian subpopulations largely reflected those of the global study populations. These results provide evidence to support the efficacy and safety of IDegLira as a treatment modality for participants with type 2 diabetes uncontrolled on OADs and/or basal insulin of the South Asian population, offering better glucose control than either component alone, improved weight profile versus the basal insulin component, and with fewer gastrointestinal side effects than a GLP-1 analogue.
Compliance with Ethics Guidelines. This article is based on a post hoc analysis of previously conducted studies and does not involve any new studies of human or animal subjects performed by any of the authors.
Data Availability. The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
Open Access. This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/ by-nc/4.0/), which permits any noncommercial use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. | 2018-04-03T04:51:45.617Z | 2017-03-22T00:00:00.000 | {
"year": 2017,
"sha1": "7b1e0e680b78dabdba5726d969812b01fe0c23b1",
"oa_license": "CCBYNC",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s13300-017-0252-9.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7b1e0e680b78dabdba5726d969812b01fe0c23b1",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
257396609 | pes2o/s2orc | v3-fos-license | Antecedents Predicting Students’ Active Use of Learning Strategies in Schools of Low SES Context within the Framework of Self-Determination Theory
Meeting a student’s autonomy, competence, and relatedness needs is one of the conditions to help him/her learn effectively. In this study, we aim to understand how teacher support (relationship with students, their autonomy support) and general classroom atmosphere (equity, relationships between students) predict students’ learning; that is, the learning strategies they use. Data were collected from 24 secondary schools in 9 municipalities in Lithuania with low SES (socioeconomic status) contexts (N = 632 students; 330 girls and 302 boys). The following instruments were used in the research: What Is Happening in this Class? (WIHIC) questionnaire, a short form of the Learning Climate Questionnaire (LCQ), and the Learning Strategies scale. The results showed that girls use learning strategies statistically significantly more actively than boys. Students’ use of learning strategies in a sample of both boys and girls are predicted by perceived teacher support, student cohesiveness, and perceived autonomy support. Therefore, responding to the relatedness and autonomy needs of students from low SES is very significant because it can increase their engagement in the learning process. The difference found, that equity is a significant predictor of learning strategies in the sample of girls, but not in boys, encourages further research and interpretation of such research results.
Introduction
The results of international achievement studies [1] show that students' social, economic, and cultural environment has clear connections with their achievements. Students' disadvantaged social and economic situation is a predictor of low achievement [2,3]. However, some of these students attain high academic achievement. In Australia, Canada, Estonia, Hong-Kong (China), Ireland, Macao (China), and the United Kingdom, more than 13% of disadvantaged students were academically resilient; that is, they achieved the highest quarter of learning results [1]. These results show that the school community can help students from low socioeconomic status (SES) overcome difficulties and achieve academic success. Factors related to academic resilience are parents' and teachers' support, school climate, and the student's beliefs in one's own abilities [1]. Recognizing the importance of these factors, in this study we aim to understand how teacher support (relationship with students, their autonomy support) and general classroom atmosphere (equity, relationships between students) predict students' learning; that is, the learning strategies they use. According to Self-Determination Theory [4], students are characterized by three basic psychological needs-autonomy, competence, and relatedness. When they are answered, students are likely to be engaged in the learning process, curious, active, and high-achieving [5]. However, it should be noted that this process is not automatic and requires support [4], so the teacher's professionalism and understanding of the student's needs are very important in this case. Appropriate teacher behaviour can enable effective learning for all students, including students from low SES.
Relatedness is one of the three psychological needs of students, which facilitates the process of internalization: students who feel accepted at school and experience a sense of belonging to it tend to adopt the example of school members and teachers' behaviour as their own; that is, internalize their values [5]. The significance of the relationship between the teacher and the student is emphasized by many other scientists, highlighting their influence in the educational process. For example, Hargreaves [6] states that students learn best when there is a safe, stable school environment and healthy relationships. Positive teacher-student relationships increase students' engagement in the learning process [7], their learning motivation [8,9], and self-esteem [10,11]. Greater perceived support from teachers protects the student from the negative consequences of self-isolation [12] and helps the teacher to notice gaps in the student's learning [13]. Finally, researchers [14] found that students who maintain positive relationships with teachers are more likely to have positive expectations and values related to academic success. Perhaps in this case, higher student achievement can be attributed to teacher support in reducing student stress because, according to Hughes [15], a positive relationship with the teacher allows students to direct their energy toward classroom tasks and constructive interactions with peers and other teachers.
It is clear that for students from low SES, a harmonious relationship with teachers is essential. Such a relationship can be one of the conditions that help to engage these students in the learning process [16,17]. It is likely that harmonious relationships as a sign of relatedness can also help to overcome uncertainties and tensions, which are certainly not lacking in the everyday life of students from low SES. On the other hand, it is necessary to note that not only relationships with the teacher, but also interpersonal relationships with peers are significant. A common social context that supports students' relatedness promotes their internal learning motivation [18] and also predicts a positive experience and well-being [19]. Researchers [20] point out that the created relationship with peers can help or hinder the student's learning success, so it is important for the teacher to see their expression. It appears that joint academic activities with a friend or friendly relationships in the classroom are associated with a better learning process and higher results [21].
From our point of view, meeting the student's need for relatedness is also related to equity. Positive teacher-student relationships promote equity [22], mitigate the negative effect of poor performance [23], and can even protect students from delinquent behaviour in the presence of less than favourable classroom environmental conditions, e.g., not having close relationships with peers [22]. Consequently, teacher support is also a significant element of equity, and the lack of it can be especially dangerous for students who demonstrate lower learning results [24]. By purposefully creating a collaborative atmosphere in the classroom and inviting all students to be active participants in the learning process, teachers can ensure equity [25]. Namely, the promotion of equity in education is less about the introduction of particular techniques or new organizational arrangements, and much more about processes of social learning [26]. However, it is important to understand that equality is not self-evident for teachers, so their professional development is necessary in this regard [27]. General school policies and actions must be directed to improve teaching and the school learning environment [28] if equity is to be ensured for students from low SES [29].
Autonomy is another psychological need of students, the response of which is very important in the learning process [5]. Research results [30] show that autonomy support and relatedness were equally important for student achievement in both Western and Eastern cultures. Both of these needs (autonomy and relatedness) are also related and manifest together in the social, i.e., teaching and learning, context [19]. Therefore, teachers must use a motivating style that supports student autonomy rather than a controlling one [31,32]. In this way, they will meet the basic psychological needs of students and create conditions for them to feel good, be active, and strive for excellence. Researchers claim that teacher behaviour that supports student autonomy strengthens and develops students' internal motivational resources [32], as it also provides structure and options to choose [33]. The conscious behaviour of students learning in this way is compatible with internal learning motivation, high volition, and a sense of choice over their actions [34]. In this case, it is likely that even students from low SES will be empowered for effective learning; that is, self-regulated use of learning strategies.
The quality of students' learning is affected by learning strategies applied by themselves [35]. Learning strategies reveal how students process and assimilate information, how they manage their learning process, and what learning tools they use to achieve their goals. As stated by Melvina, Lengkanawati, and Wirza [36] (p. 63), "Learning strategies are steps taken by learners to accelerate the attainment of knowledge, the storage of that data, and retrieval of information when they are needed". These strategies are often named as techniques used by learners to help them in the learning process [37]. The analysis of learning strategies allows to answer the questions of what makes learners successful in the learning process and why some students are more effective at learning than others [38]. Conducted studies show that successfully used learning strategies develop students' learning autonomy, competence, and self-confidence, and increase students' motivation and activity in the process of learning [36]. Thus, the use of learning strategies promotes students' self-regulated learning: students become more aware of their learning process; learn to regulate their learning efforts in order to achieve their final goals; and, thus, become more and more independent [38]. It is interesting that researchers obtain different results regarding the use of learning strategies in groups of girls and boys. Some researchers claim that girls use more learning strategies [39]; others indicate that the use of strategies does not differ between groups of girls and boys [36,38]; and others find that the number of learning strategies used in groups of boys and girls do not differ, but the nature of strategies used does [40]. Thus, we tend to look at the use of learning strategies as a certain student's competence, like a third psychological need's expression. That is, if the student uses effective learning methods more often and more actively, this is an Important condition for mastering the skills necessary for learning and achieving high learning results.
Thus, it becomes clear that all the listed aspects-interpersonal relations with the teacher and peers, equity, autonomy-supportive behaviour, and learning strategies-are significant for an effective student learning process and can be justified by the three basic psychological needs identified in the Self-Determination Theory [4]-autonomy, competence, and relatedness. Understanding that a student's learning strategies are his/her activities that directly describe the student's autonomy, i.e., self-regulated learning, and can guarantee good learning results, in this study, we propose a hypothesis: the student's mutual relations with the teacher and peers, equity, and perceived autonomy support predict more active use of learning strategies applied by students.
Sample and Procedure
An invitation to participate in the study was sent to 54 Lithuanian general education schools located in small towns or rural areas [41] (p. 33). These are schools with a small number of students and a low social, economic, and cultural context. Between 30% and 40% of pupils in these schools receive free school meals (the use of eligibility for a free lunch is a measure of a student's low socioeconomic status). Twenty-four general education schools accepted the invitation. After informed consent was gained from school principals, data collection took place in May 2021. Only students who received parental permission participated in the study and voluntarily completed the self-report anonymous questionnaire on the online platform, https://apklausa.lt (accessed on 31 May 2021). The research sample was composed of 632 students (330 girls and 302 boys). The students were enrolled from seventh (21.8%), eighth (20.1%), ninth (23.7%), and tenth (34.3%) grades. The pupils' ages ranged from 13 to 16 years; hence, these students were in a formal education programme (lower secondary education), which is compulsory in Lithuania until the age of 16.
Instruments
In this study, we will measure the expression of the student's relationship with teachers, peers, and equity with three subscales from What Is Happening in this Class? (WIHIC) [42]. In the Student Cohesiveness subscale, we will measure students' relationships with their peers because it evaluates the extent to which students are friendly and supportive of each other (e.g., this teacher talks with me). The Teacher Support subscale (e.g., the teacher considers my feelings) describes students' perception to which extent the teacher helps, befriends, and is interested in students, so using it we will see the expression of personal student-teacher relationships. The Equity subscale (e.g., I get the same opportunity to answer questions as other students) identifies the extent to which the teacher treats students equally, including distributing praise, questions, and opportunities to be included in discussions. Each item employs a 5-point Likert response format (from 1-almost never to 5-almost always). The KMO index (0.956) and Bartlett's test of sphericity (χ 2 = 11645.544, p < 0.001) indicated that the data were suitable for factor analysis. Factor analysis indicated three factors with eigenvalues greater than 1.00, which accounted for 66.9% of the total variance. Factor loadings ranged from 0.414 to 0.807. Cronbach's alpha for the three subscales were 0.927, 0.913, and 0.928, respectively. McDonald's omega was 0.928, 0.914, and 0.928, respectively.
The Learning Climate Questionnaire [43] is chosen to measure students' response to the need for autonomy. The short form of this questionnaire measured how students perceive the autonomy support provided by their teachers. This questionnaire consisted of 6 items (e.g., I feel that my teacher provides me choices and options) answered on a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree). The KMO index (0.875) and Bartlett's test of sphericity (χ 2 = 3092.001, p < 0.001) indicated that the data were suitable for factor analysis. The one-factor measurement model explained 73.2% of the total variance, with factor loadings ranging from 0.813 to 0.894. The overall Cronbach's alpha and McDonald's omega were 0.926 and 0.924, respectively. Students' learning strategies were measured by the Learning Strategies scale. Statements for the scale were formulated by the first authors of the article. The scale consists of nine self-report Likert scale items (e.g., I use a variety of techniques (such as repeating orally, doing diagrams) to memorize the information I need), ranging from 1 (almost never) to 5 (almost always). The internal consistency reliability of the scale as estimated by Cronbach's alpha and McDonald's omega was reasonably high at 0.905. The KMO index (0.928) and Bartlett's test of sphericity (χ 2 = 2904.941, p < 0.001) indicated that the data were suitable for factor analysis. The one-factor measurement model explained 57.3% of the total variance. Item loadings from a principal factor analysis with Varimax rotation are presented in Table 1.
Data Analysis
The statistical analyses in the study were conducted with the IBM SPSS Statistics 26.0. The degree of normality of continuous variables' distributions was checked by skewness and kurtosis. Parametric statistics were calculated for variables with skewness <−1 or >1 and kurtosis <−3 or >3. To examine the factorial validity of the research instruments, an exploratory factor analysis was carried out using the principal factor extraction method and Varimax rotation. We used the Cronbach's alpha and McDonald's omega to test the internal consistency and reliability of the questionnaire. These coefficients of 0.70 or higher for a set of items were considered acceptable [44].
First, descriptive statistics were run on the independent (student cohesiveness, perceived teacher support, equity, and perceived autonomy support) and dependent variables (learning strategies applied by students) used in the study to determine means and standard deviations. Additionally, bivariate correlation analysis of all variables was performed using Pearson's correlation coefficients, wherein r = 0.10-0.29 was a small correlation, r = 0.30-0.49 a moderate correlation, and r ≥ 0.5 a strong correlation. Secondly, the multivariate analysis of variance (MANOVA) was calculated to analyse the differences of variables by gender. Effect sizes (Partial Eta Squared Coefficient) were calculated for the interaction effects, with an effect size of η 2 = 0.01 representing a small effect, η 2 = 0.06 representing a medium effect, and ηp 2 = 0.14 representing a large effect [44]. Next, a multiple linear regression was performed separately for the girls' sample and boys' sample. Pearson's correlation coefficient and variance inflation factor (VIF) helps to check the multicollinearity of independent variables. High correlation values (greater than 0.8) and a VIF score of four or above indicate multicollinearity [45]. In regression analysis, the effect size of the predictor variables is given by the beta loadings. In interpreting the effect, size gives the following guidance: 0-0.1 = weak effect, 0.1-0.3 = modest effect, 0.3-0.5 = moderate effect, and >0.5 = strong effect [44]. An alpha level of 0.05 was used as a significance level for all the statistical analyses.
Results
The descriptive statistics and intercorrelations between all variables in the study are outlined in Table 2. From the analysis of the skewness (−0.608 to −0.030) and kurtosis (−0.590 to 0.243) values, it is established that the research data were close to a normal distribution. An examination of the means of the variables revealed that the highest scores were students' perceived autonomy support (M = 4.39), and the lowest average score was ascribed to perceived teacher support (M = 3.06). The results of bivariate Pearson's correlations indicated that the dependent variable (learning strategies applied by students) correlated positively and significantly to the independent variables (r = 0.483, p < 0.01 for the student cohesiveness; r = 0.572, p < 0.01 for the perceived teacher support; r = 0.552, p < 0.01 for the equity; and r = 0.494, p < 0.01 for the perceived autonomy support). The correlations between independent variables revealed that absolute values of Pearson r coefficients are less than 0.8; this indicates that multicollinearity is very unlikely to exist. We then ran MANOVA with student gender as an independent variable and the five dependent variables, namely student cohesiveness, perceived teacher support, equity, perceived autonomy support, and learning strategies applied by students. The Box's M test (Box M = 10.64, p = 0,784) was not significant; thus, the observed covariance matrices of the dependent variables are equal across groups. The results of the MANOVA showed that there was a statistically significant difference between the girls and boys on the combined de-pendent variables F = 4.295 b, p ≤ 0.001, Wilk's Lambda = 0.033. The obtained results (Table 3) show that the averages of the three independent variables of the study (student cohesiveness, perceived teacher support, and perceived autonomy support) do not differ statistically significantly in the sample of boys and girls. However, the mean of the fourth independent variable (equity) in the sample of girls (M = 3.75, SD = 0.90) is higher than that of boys (M = 3.55, SD = 0.92). A statistically significant difference was obtained (F = 7.488, p = 0.006), although partial eta squared (η 2 = 0.012) shows a small effect.
It was also found that the averages of the dependent variable (learning strategies applied by students) are statistically significantly different (F = 16.128, p = 0.0001): in the sample of girls (M = 3.22, SD = 0.79), it is higher than that of boys (M = 2.97, SD = 0.77). Although there is a difference, η 2 = 0.025 shows a small effect.
A multiple linear regression was conducted to determine if the independent variables (student cohesiveness, perceived teacher support, equity, and perceived autonomy support) collectively predict the dependent variable (learning strategies applied by students). Since the means of the dependent variable are statistically significantly different, the regression analysis was performed separately for the sample of boys and the sample of girls.
The first multiple linear regression analysis included the boys sample (n= 302). All four independent variables were initially included in the model. However, when included in multiple regression analyses, equity as predictor was marked as not significant (p = 0.269). Therefore, the multiple regression analysis was repeated. The results of the improved model are presented in Table 4. The R 2 = 0.409 shows that 40.9% of the dependent variable can be predicted by the independent variables (F = 68.722, p < 0.0001). A detailed analysis of the β coefficients showed that the perceived teacher support (β = 0.335) was the best predictor of learning strategies applied by students (t = 5.665, p = 0.0001). The other two predictors are weaker. The second multiple linear regression analysis included the girls sample (n = 330). All four independent variables were statistically significant (Table 5). Perceived teacher support had a stronger predictive effect for girls' when predicting learning strategies (β = 0.318), compared to equity (β = 0.173), student cohesiveness (β = 0.145), and perceived autonomy support (β = 0.119). A significant regression model was found (F = 53.704, p = 0.0001), with an R 2 of 0.398. This means that 39.8% of learning strategies applied by girls can be explained by all four predictors.
Discussion
Learning strategies applied by the students are an important antecedent of students' learning performance and satisfaction [35,46], so it is important to study the factors of the learning environment that determine the more active use of learning strategies. The results of the conducted research allow us to provide several important insights. It becomes clear that girls and boys do not equally actively apply learning strategies: girls use them more actively than boys and this difference is statistically significant. Since the use of learning strategies can be associated with engagement in learning [47], this research result is corroborated by the data of other researchers. For example, a large group of researchers [48] see a consistent trend of higher engagement in learning for girls than boys in as many as 12 countries. In 2022, the published UNESCO report "Leave no child behind: global report on boys' disengagement from education" [49] also shows that, according to data from many countries, boys are more at risk than girls of not being active learners and achieving poorer academic results at school. Therefore, it is obvious that it is necessary to look for factors that could encourage all students (and boys in particular) to become more and more actively involved in the learning process, and the results of our research provide certain answers.
The factors that determine the learning process and, accordingly, the achievement of boys and girls are many and varied. According to Cascella et al. [50], they can be individual, social, and cultural, as well as factors related to the school context, such as curriculum, teaching practices in the classroom, and teacher evaluation methods [51]. It seems that even the gender of the teacher can make a difference. For example, some researchers [52] claim that male and female teachers perceive and evaluate male and female students differently. Theoretically, teachers may favour students who are more similar to themselves and, consequently, give them higher evaluations [53]. On the other hand, a teacher of the same gender can serve as a role model [54], and, thus, a teacher of the same gender can influence a student's effort [55]. Lowe and colleagues [56] state that the gender of the teacher is an important factor and that girls respond more strongly than boys to same-gender role models. Therefore, the result of our study is quite clear: that in the studied sample, girls were more involved in the learning process; that is, they used learning strategies more actively. However, it is necessary to mention that research does not provide unequivocal answers as to whether there is a causal relationship between a same-gender teacher and student achievement because, as already mentioned earlier, the student's learning process can be influenced by various factors [53].
The regression analysis carried out in our study reveals that the satisfaction of relatedness and autonomy needs is very important for students of both genders. In the sample of boys, even 40.9% of the dependent variable (that is, the use of learning strategies) predicts student cohesiveness, perceived teacher support, and perceived autonomy support. In the sample of girls, 39.8% of learning strategies applied by girls are explained by four predictors: perceived teacher support, equity, student cohesiveness, and perceived autonomy support. Thus, the results of this study confirm the idea of other researchers that factors promoting relatedness such as peer learning, working in peers' groups, good relationship with the teacher, and the teacher's support encourage more active use of learning strategies [35]. On the other hand, it is important for teachers to support students' autonomy because thanks to it, students' self-regulated learning is developed, which, in turn, promotes more active use of learning strategies, and can support students to learn independently inside and outside the classroom [36].
It is noteworthy that perceived teacher support is the best predictor of learning strategies applied by boys, when the other two predictors (student cohesiveness and perceived autonomy support) are weaker. Thus, it is clear that a close relationship between the teacher and a boy-student is necessary, which is confirmed by other researchers [57]. It is evident that perceived teacher support also has the strongest predictive effect for girls, whereas the other three variables (equity, student cohesiveness, and perceived autonomy support) have less. Hargreaves [6] states that healthy relationships between teachers and students, when attention is paid to the social-emotional aspects of interaction, promote student learning. Research by other researchers [3,48] also show that teacher support is significant: it predicts class and school-related interest, greater motivation for learning, students' engagement in learning, and better achievement. It is claimed that teachers' support is particularly important in adolescence and can act as a protective factor in the learning process [58].
We would like to point out one more difference that our research results show. In the sample of girls, equity is a significant predictor of learning strategies; that is, it seems that for girls' more active use of learning strategies, it is important how teachers create an atmosphere of cooperation in the classroom by generating equal learning opportunities for every student, involving all students in an active learning process [25]. However, in the sample of boys, equity is not a significant factor in the active use of learning strategies. It must be admitted that such a result of the study is somewhat unexpected. Perhaps one of the possible explanations for such an unexpected result could be the claims of other scientists [59] that teachers are important starting points for promoting gender equality in education, as their attitudes and instructional practices influence students' performance. A certain favouritism of the teacher towards the student of the same gender is expected [53] because according to the data of our study, girls experience a higher expression of equity and this is a significant predictor of learning strategies; hence, the respect to diversity is one of the most significant values in the relationship between teachers and students [25], and teachers have to reflect on their own gender stereotypes [59].
Thus, the results of the research show that boys are less active in the learning process than girls, and the use of learning strategies by students of both genders predicts perceived teacher support, student cohesiveness, and perceived autonomy support; hence, meeting students' relatedness and autonomy needs is very significant, as it can increase students' engagement in the learning process [4,34,60]. Relatedness contributed strongly to the autonomous motivation [61], and relatedness and autonomy not only do not contradict each other or are opposites, but are closely related [19]. Finally, we want to emphasize once again the significance of teacher support as a personal emotional relationship between teacher and student for the successful learning process of a student from low SES. It becomes evident that the school community can help students from low SES to be more involved in learning and academical resilience if teachers provide support and create a favourable school learning environment [1,28].
It is also necessary to discuss the limitations of this study. First, our research design was cross-sectional. This makes it possible to evaluate the correlations between different factors, but does not allow a deeper analysis of how the inclination to more actively use learning strategies is formed. Therefore, the longitudinal research design for future studies would be preferable. Secondly, the self-report method was used in this study: students themselves evaluated teacher's support, autonomy support, and equity. In order to obtain more objective data about the significance of the researched factors for learning strategies applied by the students in future studies, it would be useful to use more diverse research data collection sources (e.g., not only from the students, but also teachers, evaluations of independent observers, etc.).
Conclusions
Understanding students' needs and professional teacher behaviour can enable effective learning for all students, including students from low SES. The results of a study conducted in small towns or rural areas of Lithuania with a low social, economic, and cultural context show that boys in grades 7-10 use learning strategies less actively than girls. This use of learning strategies by students of both genders predicts perceived teacher support, student cohesiveness, and perceived autonomy support. Therefore, responding to the relatedness and autonomy needs of students from low SES is very significant because it can increase their involvement in the learning process. The difference found, that equity is a significant predictor of learning strategies in the sample of girls, but not in boys, encourages further research and interpretation of such research results. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement:
The datasets generated and analysed during the current study are not publicly available due to privacy and ethical concerns, but are available from the corresponding author on reasonable request. | 2023-03-08T16:03:49.283Z | 2023-03-01T00:00:00.000 | {
"year": 2023,
"sha1": "0d0bf81183d26babca4857368d1978f0fac95544",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2254-9625/13/3/44/pdf?version=1678086865",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "4df5eed7b689cda0111288fb05c72f125e6e68e1",
"s2fieldsofstudy": [
"Education",
"Psychology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
118424367 | pes2o/s2orc | v3-fos-license | Solar flares as harbinger of new physics
This work provides additional evidence on the involvement of exotic particles like axions and/or other WISPs, following recent measurements during the quietest Sun and flaring Sun. Thus, SPHINX mission observed a minimum basal soft X-rays emission in the extreme solar minimum in 2009. The same scenario (with ~17 meV axions) fits also the dynamical behaviour of white-light solar flares, like the measured spectral components in the visible and in soft X-rays, and, the timing between them. Solar chameleons remain a viable candidate, since they may preferentially convert to photons in outer space.
Introduction
The solar coronal heating mechanism remains an unsolved problem since 1939, being therefore one of the prominent challenges in solar physics and astrophysics. In addition, the so called White-Light (WL) solar flares were observed for the first time in 1859, but their trigger remains elusive, and "flares might be unpredictable for more fundamental reasons" [1]. In this work we provide additional evidence that the working principle of CAST can be at work at the Sun, where outstreaming solar axions can convert to the otherwise unexpectedly measured X-rays [2]. If CAST's working principle has been fine tuned by nature at the Sun, then the most promising places to look out for similar effects, should be preferentially above Active Regions (ARs), i.e., sunspots. Note that these regions are also the birth places of flares, while the corona above nonflaring/quiet ARs is hotter than the rest of the quiet Corona. The temperature reaches ∼10 MK and the unexpectedly emitted photons have a characteristic spectral shape (power law) that resembles down-comptonized hard X-rays appearing otherwise unexpectedly [2]. Indeed, the "Sun's intense X-ray emission is a remarkable and fascinating...mystery" [3]. The idea of the down-comptonization underneath magnetized solar surface has been presented in [2]. Below, we update some of the conclusions in [2], following recent X-ray measurements with SPHINX in the ∼1-5 keV. The X-ray mission SPHINX has measured a minimum basal level of solar X-ray emission even during the most quiet solar minimum without flares or sunspots in 2009, below which solar activity never dropped. The measured quiet solar X-ray luminosity is L x ≈ 6.7×10 20 erg/s ≈ 1.5×10 −13 L ⊙ [4]. This fraction of the total solar luminosity is even 10 times below the estimated axion-photon conversion of about 10 −12 (see Footnotes 9 and 14 of [2]), making thus a quantitative X-ray intensity reconstruction reasonable, assuming that axions escape from the solar hot core with a rest mass of about 17 meV and do convert to X-rays near the magnetized solar surface. In addition, the spectral shape follows a perfect power-law as it is shown in Figure 1, pointing at a depth of 1000 km below the surface as the initiation place of the following down-comptonization. The spectral agreement between observation (SPHINX) and simulation extends over 3-4 orders of magnitude (see Figure 1). The soft X-ray component below ∼2.5 keV could be explained in general with light axions or other particles with similar couplings [2]. An alternative scenario is that with the solar chameleons [7,8]. In fact, X-ray photons from regenerated outstreaming chameleons could emerge in the outer solar magnetic fields, resembling that of the 2-3 MK quiet Sun X-ray emission (see Figure 1). If the rest of the spectrum above 2.5 keV is real [5], such a second component might come, in principle, from regenerated axions in the upper photosphere / lower atmosphere, with suppressed down-comptonization. Furthermore, other exotica like radiatively decaying massive particles [6] could also be at the origin. Four degraded spectra due to multiple Compton scattering from 4 different depths into the photosphere are also shown. The spectrum measured by the SPHINX mission (green dashed line) during the extremely quiet Sun in 2009 [5] agrees with that of the axion or axion-like scenario, assuming the conversion place is 1000 km underneath the photosphere [2]. Figure 2 shows the quiet Sun analog spectrum (in blue) [9] along with a few points (stars in red) of the white-light (WL) emission during a flare. One notes at the high-energy end the one flare-related point, which is far above that expected from the ∼9000 K WL flare black-body emission. Thus, comparing with the quiet Sun spectral distribution with its striking photon excess at the high energy end (> 10 15 Hz), which reflects the solar corona problem, it seems that a flare develops its own and relatively more intense "corona". This flare related "corona" is even better seen in Figure 2 of ref. [10], where the EUV intensity resembles also a flare's intense "corona", being far above that of a black body WL flare distribution of ∼ 10 4 K. The appearance of a flare "corona" fits our axion or axion-like picture, since it is widely accepted that a flare is magnetic in origin, though with several open questions remaining, like: where, when and how electrons are accelerated [11]? In the axion scenario an intervening magnetic field is just the catalyst, which transforms outstreaming axions or other particles like chameleons [7,8] to photons near and/or far from the Sun. As it was noticed in [12,13], the magnetic field is a viable flare forecasting tool, while what powers/triggers a solar eruption is not known. But, "it might be possible that an unknown mechanism produces the black body white-light flare spectrum near 9000 K " [12,13] (Figure 2). Moreover, the total energy radiated by flares exceeds by ∼100× the flare soft X-ray energy emission, with a major contribution in the visible and near-UV. Then, the required axion conversion efficiency is not unnaturally large. Figure 2: The continuum flare spectrum (red stars), with the dashed line representing a blackbody at ∼9000 K For comparison, it is also given the spectrum of the quiet Sun (blue points). The overall spectrum of the quiet Sun has a near blackbody shape at visible and near-IR. Excesses over a fitted blackbody appear at EUV and far-IR and beyond. The striking EUV excess of the quiet Sun is the manifestation of the solar corona problem. Courtesy H.S. Hudson [9].
The White-Light flares: analog spectrum and time correlations
As it is shown in Figure 3, the visible light (TSI) appears first and later takes place the soft X-ray emission. The axion scenario [2] for the WL flares is this: the energy deposition from converted axions streaming out of the inner Sun occurs at some deep photospheric layer. This gives rise to local ionization and thermalization, once a sufficiently strong magnetic field is combined with the appropriate density (which fits an axion rest mass of about 17 meV [2]), maximizing thus the coherence length for the axion-to-photon conversion to happen efficiently. The so heated-up environment does not thermalize completely to ∼6000 K, and photons are escaping from the photosphere being a little hotter (∼ 9000 K). In addition, thanks to the horizontal magnetic field component, an outwards moving compressed front is possible, which pushes the resonant axion-conversion place (ρ ≈ m axion ) upwards. This results in a decreasing column density above the actual conversion place. With time, the newly back-converted hard X-rays 1) do propagate in an ionized environment, and 2) suffer a more and more limited down-comptonization until they can leave the Sun as soft X-rays [2].
In other words, within the axion scenario, the sequence of events, is this: first starts the energy deposition by axion conversion in a relatively deep photospheric layer; this gives rise to the ambient ionization associated with a rather incomplete thermalization of about 10 4 K with the escaping WL. The initial hard X-rays make their way outwards in a random walk, once the ambient plasma is complete, which allows multiple Compton scatterings to occur. Following this reasoning, the soft X-ray emission can come only later, and this is what has been observed recently, but only (!) when analyzing many flares [12,13]. An estimate of this time delay is interesting: the outward propagation speed of any energy deposition [14], e.g., by converted axions or other exotica, is about 1 km/s; this determines the compression time of the moving heated layer by some 500 km upwards inside the photosphere, which takes some 10 minutes. Interestingly, this order of magnitude estimate fits the observed delay of ∼5 minutes (see Figure 3). In summary: a. the WL can come from not completely thermalized ∼ 4 keV photons from converted axions (inverse Primakoff effect) at large photospheric depths, and b. this additional radiation pressure combined with the magnetic field in sub-photospheric layers [15] pushes the ionized axion conversion place upwards; the decreasing depth of the axion conversion layer (relative to that of the WL source origin), allows the X-rays to undergo a limited down-comptonization before escaping as soft X-rays with a characteristic power-law spectral shape, and this occurs 5 minutes later, matching also observation. Thus, a chain of processes observed during WL flares fit the axion scenario. However, this does not exclude synergism with conventional physics reactions and/or the involvement of other WISPs; axions [2] and chameleons [7,8] are rather two generic examples, which can be at work underneath and above the solar surface, respectively. Most probably, the conventionally mysterious multifaceted Sun cannot be understood only by a single global reaction mechanism.
Discussion
Outstreaming solar axions with a rest mass of ∼17 meV can explain solar X-ray activity being enhanced above the magnetized photosphere [2]. Here we have elaborated the underlying processes, following the recent findings of WL flares. After the same reasoning, the quiet and the flaring Sun appear as two extreme cases. Within the suggested axion scenario, and assuming that the same chain of processes happens in both cases, the photosphere dynamics determines the spectral outcome. Because, the initial axion conversion layer (i.e., the origin of the WL) can move with the compressed hot front at less deep layers upwards, being thus shielded with a decreased column density above (i.e., the origin of the X-ray emission). This defines also the degree of the down-comptonization of the escaping radiation, -as it was pointed out in Figure 10 of [2]-the solar magnetic fields appear deeper in the quiet Sun than in the ARs, and therefore the quiet Sun is less X-ray active as a decreasing number of X-rays can reach the surface from deeper layers. This fits the fact that the solar corona above non-flaring ARs is hotter than that of the near quiet Sun (∼10 MK vs. ∼2 MK) [2]. While any approach must explain first why the quiet corona is hot at all, before asking: why is it hotter there? Though, corona's dynamical behaviour might point at its workings.
As a second example in support of potential exotic particle involvement we mention solar chameleons with an energy of about 600 eV, which can be created near the strongly magnetized tachocline [8]. Since their conversion efficiency is maximum in vacuum, the magnetized solar outer space appears as their favourable place to get back-converted to soft X-rays by the inverse Primakoff effect. For example, the strong dipole magnetic field component between Sun and Earth (BL ≈10 5 Tm) is of relevance only for particles like chameleons. Similar or even stronger fields appear in Corona Mass Ejections (BL ≈ 10 6 Tm), since the magnetic field strength is up to 200 Gauss [16]. Note, CAST has BL=84 Tm, while the signal strength goes with (BL) 2 . Then, the Sun, its near and/or far transverse magnetic field components as seen from an orbiting observatory, being X-ray sensitive up to 10 keV, resemble an Earth bound axion helioscope like CAST or Sumico. Therefore, such a "natural" helioscope might have occasionally an enormous built-in axion or chameleon parameter fine tuning, which man-made equipment did not yet have the time to reach. | 2011-12-19T13:08:23.000Z | 2011-12-19T00:00:00.000 | {
"year": 2011,
"sha1": "e0eabe96338f398fd406995a3c74c22057bee03c",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "5f8a846b716b39a9b42aed152ceec5a11125e5b7",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
247255437 | pes2o/s2orc | v3-fos-license | Regulation of paternal 5mC oxidation and H3K9me2 asymmetry by ERK1/2 in mouse zygotes
Background Extracellular-signal-regulated kinase (ERK) direct cell fate determination during the early development. The intricate interaction between the deposition of H3K9me2, de novo 5mC, and its oxides affects the remodeling of zygotic epigenetic modification. However, the role of fertilization-dependent ERK in the first cell cycle during zygotic reprogramming remains elusive. Methods In the present study, we used the small molecule inhibitor to construct the rapid ERK1/2 inactivation system in early zygotes in mice. The pronuclear H3K9me2 deposition assay and the pre-implantation embryonic development ability were assessed to investigate the effect of fertilization-dependent ERK1/2 on zygotic reprogramming and developmental potential. Immunofluorescence and RT-PCR were performed to measure the 5mC or its oxides and H3K9me2 deposition, and the expression of related genes. Results We reported that zygotic ERK1/2 inhibition impaired the development competence of pre-implantation embryos. Following the ERK1/2 inhibition, H3K9me2, as well as 5mC and its oxides, were all accumulated abnormally, and the excess accumulation of paternal H3K9me2 and 5mC resulted in reduced asymmetry between parental pronuclei. Furthermore, ERK1/2 inhibition triggered paternal pronuclear localization of the H3K9 methyltransferase G9a and Tet methylcytosine dioxygenase 3 (Tet3). Moreover, the excess localization of G9a antagonized the tight binding of Tet3 to paternal chromatin when ERK1/2 was inhibited. Conclusions In conclusion, we propose that zygotic H3K9me2 and 5mC are regulated by fertilization-dependent ERK1/2, which contributes to the development competence of pre-implantation embryos in mice. Supplementary Information The online version contains supplementary material available at 10.1186/s13578-022-00758-x.
During the zygotic reprogramming, the paternal genome initiates widespread deposition of de novo epigenetic marks, making it an excellent model to understand how epigenetic marks are loaded on the genome orderly [1,2,7].
In mammals, mitogen-activated protein kinases (MAPKs) regulate the stability of the maternal and zygotic transcripts [8] and are essential for the early development of mouse zygotes [9]. It was reported that the MII-phase oocytes are stabilized by appropriate levels of MPF [10] and MAPKs [11]. After sperm-oocyte fusion, the decrease of MPF activity (within 10 min) precedes that of MAPKs activities [12][13][14]. Recently, ERK signal was reported to direct fate specification of embryonic stem cells and the inhibition of ERK promotes enhanced stabilization of Nanog protein after mitosis [15]. However, the mechanisms by which zygotic ERK controls gene expression to influence developmental pattern formation and the quantitative understanding of these mechanisms remain unknown [16].
H3K9me2 is associated with the compact chromatin and the transcriptionally repressive state [17,18]. It was reported that the zygotic genome exhibits an asymmetric H3K9me2 pattern. The maternal pronuclei remain considerable H3K9me2, while paternal pronuclei bear no H3K9me2 [19][20][21]. Subsequently, the paternal genome progressively gains H3K9me2 from the late zygote to the cleavage stage [19,21], suggesting fine regulation of H3K9me2 during zygotic reprogramming. G9a (also known as EHMT2 and KMT1C) is responsible for this progressive deposition of paternal H3K9me2 [21], and it protects the maternal pronucleus from 5mC oxidation [22]. During embryogenesis, G9a controls transcription levels of zygotic genes [23]. These studies suggest that G9a plays pivotal roles during zygotic reprogramming [21,22]. 5mC is one of the well-documented epigenetic factors associated with gene silencing and plays an important role in facilitating the propagation of cellular identity through cell divisions [24,25]. Meanwhile, 5-hydroxymethylcytosine (5hmC), the oxide of 5mC [26], is positively correlated with gene expression and plays an important role in epigenetic regulation and genome reprogramming during mammalian development [1,7,25,27]. It was reported that the sperm-derived genome undergoes actively demethylated through Tet3-mediated 5mC oxidation, whereas the maternal genome is passive dilution by DNA replication [28,29]. Importantly, Tet3 also contributes to paternal demethylation by counteracting the de novo 5mC [30]. While immunostaining results support a role for H3K9me2 in protecting 5mC [31,32], many bisulfite sequencing (BS-seq) results show that 5mC in mammals is largely independent of H3K9me2 [33,34], suggesting unknown factors regulated the crosstalk between H3K9me2 and 5mC.
Therefore, in the present study, we examined the role of zygotic ERK1/2 during preimplantation development and reported the fine regulation of paternal 5mC oxidation and H3K9me2 asymmetry by ERK1/2, which provide insights into the role of ERK1/2 during zygotic reprogramming.
Zygotic ERK1/2 is essential for preimplantation development
We first analyzed the expression of MAPKs during early embryo development by re-analysis the RNA-seq datasets of mouse embryos. As shown in Fig. 1A, B, MAPKs, specifically ERK1/2, were highly expressed during zygotic genome activation (ZGA), suggesting that MAPKs might play pivotal roles during ZGA in mice. To confirm the hypothesis, U0126, GDC-0994, SB203580, and SP600125 were used to inhibit the protein of MEK1/2, ERK1/2, p38, and JNK, respectively. We selected the maximum concentration before the morphology of zygotes becomes abnormal as the working concentration for short-term treatment (Additional file 1: Fig. S1), and the suppression effect of these inhibitors at the working concentration has been verified (Additional file 1: Fig. S2). As H3K9me2 deposition is critical for preimplantation [34], we analyzed the H3K9me2 signal in zygotes. As shown in Fig. 1C, the signal intensity of H3K9me2 was increased in ERK1/2-inhibited paternal and maternal pronuclei compared to the controls (p < 0.01). Similar results were found in both MEK1/2 and p38-inhibited paternal and maternal pronuclei (p < 0.01). Meanwhile, the paternal/ maternal signal ratio of H3K9me2 was greatly increased in ERK1/2-inhibited zygotes (p < 0.01), but decreased in both MEK1/2 and p38-inhibited zygotes (p < 0.01; Fig. 1C, D). Of note, H3K9me2 signal intensity was also increased in ERK1/2-inhibited paternal pronuclei when compared to that of MEK1/2 and p38-inhibited paternal pronuclei (p < 0.01).
ERK1/2 regulates the H3K9me2 and DNA methylation reprogramming in paternal genomes
We next investigated whether the dynamic change of H3K9me2 is ERK1/2 dose-dependent. As shown in Fig. 2A, B, the signal intensity of H3K9me2 was not statistically changed in 50 μM GDC-0994 treatment zygotes compared to the controls. However, in 100 μM GDC-0994 treatment zygotes, the level of H3K9me2 signal was markedly increased in both the paternal and maternal pronuclei (p < 0.01), and the paternal to . B Values are represented as the mean of H3K9me2 signal intensity in paternal and maternal pronuclei after background subtraction (left axis) or a ratio between signals of parental pronuclei (pat./mat., right axis). Number of zygotes analysed for each group: control n = 10; 50 μM GDC-0094-treated n = 15; 100 μM GDC-0994-treated n = 15. C 5mC (green) and 5caC (red) staining of control, 50 μM GDC-0994-treated, and 100 μM GDC-0994-treated zygotes at PN4-5 stage (10 hpf ). D, E Quantification of 5mC and 5caC is represented as signal intensity in paternal and maternal pronuclei (left axis) or as a ratio between parental signals (pat./mat., right axis). Number of zygotes analysed for each group: control n = 23; 50 μM GDC-0994-treated n = 34; 100 μM GDC-0994-treated n = 28. Statistical analysis was carried out using Student's t-test (two-sided). P values are indicated. Error bars indicate SD. ♀, maternal pronucleus; ♂, paternal pronucleus. Scale bar, 20 µm maternal signal ratio of H3K9me2 was also significantly increased (p < 0.01), suggesting that regulation of paternal H3K9me2 is ERK1/2 dose-dependent. We further examined the level of 5mC and 5caC signals in ERK1/2inhibited zygotes. As shown in Fig. 2C, D, the level of paternal 5mC was increased following treatment with 50 μM GDC-0994 (p < 0.01). The product of the Tet3 oxidation chain, 5caC, was also accumulated more than the controls (p < 0.01) (Fig. 2C, E). Compared with the 50 μM GDC-0994 treated zygotes, the paternal 5mC remained stable in 100 μM GDC-0994-treated zygotes, but the accumulation of paternal 5caC was significantly reduced (p < 0.01), indicating that ERK1/2 might play pivotal roles in the oxidation of paternal 5mC. Notably, the asymmetries of 5mC, 5caC, and H3K9me2 between the paternal and maternal genomes were all decreased following 100 μM GDC-0994 treatment (p < 0.01, p < 0.05, p < 0.01, respectively) ( Fig. 2A-E), suggesting that ERK1/2 is involved in the asymmetric regulation of these parental epigenetic marks in zygotes.
We further investigated the distribution of parental H3K9me2 at 17 hpf and 24 hpf. As shown in Fig. 3E, H3K9me2 was asymmetrically distributed at the 17 hpf and the 24 hpf in the controls. However, the asymmetry of parental H3K9me2 disappeared following treatment with GDC-0994 at the 17 and 24 hpf (Fig. 3E). H3K9me2 is important for protecting the methylation of paternally imprinted gene H19 against active DNA demethylation [31]. The aberrant H3K9me2 asymmetry might impair the expression of H19. As expected, H19 was significantly down-regulated whereas Igf2 was highly expressed in blastocysts of the ERK1/2-inhibited group (Fig. 3F, p < 0.01, p < 0.01, respectively). Our data suggest that zygotic ERK1/2 is responsible for the asymmetry of parental H3K9me2 during ZGA by impeding the deposition of paternal H3K9me2.
Our previous study revealed that G9a is required for paternal H3K9me2 deposition in zygotes [21]. As shown in Fig. 4A, G9a was enriched in both paternal and maternal pronuclei (p < 0.01), and paternal to maternal signal ratio of G9a was increased following 100 μM (p < 0.05), but not 50 μM GDC-0994 treatment compared to the controls (Fig. 4A, B). To confirm that ERK1/2 regulates H3K9me2 asymmetry through G9a, we carried out IVF in the presence of BIX-01294, an inhibitor that competed for the substrate of G9a [41,42]. Inhibition of G9a showed no statistical change of H3K9me2 in paternal pronuclei. However, the deposition of paternal H3K9me2 induced by ERK inhibition was barely detected in both ERK1/2 and G9a-inhibited zygotes, similar to the controls (Fig. 4C, D). In addition, the paternal pronuclear localization of G9a was not only increased in ERK1/2inhibited zygotes, but also increased in both ERK1/2 and G9a-inhibited (GDC-0994 + BIX-01294-treated) zygotes (p < 0.01 Fig. 4E, F). These data demonstrate that ERK1/2 impedes paternal localization of G9a and contributes to the formation of H3K9me2 asymmetry between parental pronuclei.
Taken together, our results demonstrate that ERK1/2 inhibition promotes Tet3-driven oxidation of excess de novo 5mC in paternal pronuclei, and subsequently leads to accumulation of paternal 5hmC.
Excess localization of G9a impedes oxidation of 5mC in paternal pronuclei
Since the pronuclear localization of G9a and Tet3 was both increased after ERK1/2 inhibition, and the GDC-0994 concentration required for accumulation of paternal Tet3 was lower than that of paternal G9a, we further assessed the interaction between G9a and Tet3. Under the TP conditions, we found reduced tight binding of Tet3 to chromatin following DMOG treatment, especially in paternal pronuclei (p < 0.01) (Fig. 7A, B). Conversely, the G9a localization was increased significantly in paternal pronuclei (p < 0.01), but not in maternal pronuclei with DMOG treatment (Fig. 7C, D). Moreover, paternal Tet3 was decreased in G9a-inhibited (BIX-01294-treated) zygotes (p < 0.01) (Fig. 7E, F). These data suggest that the accumulation of pronuclear G9a resulted in the reduction of Tet3 localization in paternal pronuclei.
To validate the interaction between Tet3 and G9a, we detected the level of paternal 5mC and 5hmC signals in the presence of both ERK1/2 and Tet3 inhibitors. As shown in Fig. 7G, H, both inhibition of ERK1/2 and G9a catalyst activity (50 μM GDC-0994 and BIX-01294, respectively) resulted in the accumulation of paternal 5mC (p < 0.05) and the loss of paternal 5hmC (p < 0.01) compared to the ERK1/2-inhibited zygotes. Since inhibition of both ERK1/2 and G9a led to a loss of H3K9me2 in paternal pronuclei, these data suggest that excess localization of G9a impedes oxidation of 5mC in paternal pronuclei by inhibition of Tet3. Moreover, paternal H3K9me2 was increased in both ERK1/2 and Tet3-inhibited (50 μM GDC-0994 and DMOG, respectively) zygotes (Fig. 7I, J), suggesting that Tet3 impedes the paternal H3K9me2 deposition in zygotes as well.
Discussion
In the present study, we found that fertilization-dependent ERK1/2 is the critical regulator for early embryo development in mice. We further reported reduced asymmetries of 5mC, 5hmC, 5caC, and H3K9me2 at the PN stage after ERK1/2 inhibition. In addition, ERK1/2 inhibition led to excess pronuclear localization of G9a, which subsequently antagonizes the tight binding of Tet3 to paternal chromatin, providing crucial insights into the regulation of DNA methylation and histone modifications crosstalk by ERK1/2.
Previous studies revealed that ERK1/2 protein levels increase after fertilization [14,43,44], which is the minor ZGA. Consistently, we found that ERK1/2 was highly expressed during ZGA, as revealed by RNA sequencing data. We further observed that multiple MAPKs pathways are involved in the regulation of the maternal H3K9me2 deposition. As for paternal genomes, H3K9me2 deposition showed the most significant increase, and the asymmetry of the parental H3K9me2 was altered following zygotic ERK1/2 inhibition, indicating that ERK1/2 plays an important role in the regulation of paternal H3K9me2 and the asymmetry between the parental genomes.
It was demonstrated that mouse minor ZGA occurs at the PN3 stage, during which, the first zygotic transcription is promiscuous [45] and the epigenetic modifications are rapidly reprogrammed [5,30,31,46]. In the present study, we found decreased levels of 5mC and H3K9me2 in maternal pronuclei from the PN3 stage, consistent with previous studies [21,47,48]. Considering that 5mC and H3K9me2 are transcriptional repressive markers [24,25,49,50], the removal of 5mC and H3K9me2 would be important for the initiation of minor ZGA. Indeed, studies reported the symmetry distribution of 5mC between paternal and maternal pronuclei at PN stage impairs the embryonic development [32] and that of H3K9me2 is embryonic lethal [22], suggesting the asymmetries distribution of 5mC and H3K9me2 are critical for the early embryo development.
A recent study revealed that ERK is related to transcription activity and the fluctuation or persistence of ERK signal would irreversibly change the expression of pluripotency genes [51]. In the present study, the expression of two genes Oct4 and Nanog, which serve as hubs in the core pluripotency network, was increased in the expanded and hatched blastocysts after zygotic ERK1/2 inhibition, suggesting that the regulation of ERK on the Fig. 7 Excess pronuclear localization of G9a impedes 5mC oxidation in paternal pronuclei. A Tet3 staining under TP conditions of control and DMOG-treated zygotes at the PN4-5 stage (10 hpf ). B Quantification is represented as Tet3 signal intensity or a ratio between parental signals (pat./ mat.). (n = 16 for control, n = 18 for iTet3.). C G9a staining of control, and DMOG-treated zygotes at the PN4-5 stage (10 hpf ). D Quantification is represented as G9a signal intensity or a ratio (pat/mat). (n = 13 for control, n = 19 for iTet3). E Tet3 staining under TP conditions of control, and BIX-01294-treated zygotes at the PN4-5 stage (10 hpf ). F Quantification is represented as Tet3 signal intensity or a ratio (pat./mat.) (n = 18 for control, n = 19 for iG9a.) G Competitive inhibition of G9a activity by BIX-01294 affects the accumulation of paternal 5hmC ERK1/2 inhibition (50 μM GDC-0994) produced as assessed by 5mC and 5hmC staining at the PN5 stage (11 hpf ). H Quantification of 5mC and 5hmC is represented as signal intensity in paternal pronuclei. (n = 23 for control, n = 20 for iERK1/2, n = 17 for iERK1/2 + iG9a; n = 16 for iG9a.) I Tet3 inhibition by DMOG promotes the deposition of paternal H3K9me2 as assessed by H3K9me2 staining at PN4-5 stage (10hpf ). J Quantification of the parental H3K9me2 signal intensity. Number of zygotes analysed for each group: n = 14 for control, n = 13 for iERK1/2, n = 19 for iERK1/2 + iTet3; n = 17 for iTet3. Statistical analysis was carried out using Student's t-test (two-sided). P values are indicated. Error bars indicate SD. ♀, maternal pronucleus; ♂, paternal pronucleus. PB, polar body. Scale bar, 20 µm determination of the embryonic fate occurs in the first cell cycle, consistent with the results of directing fate specification in the preimplantation embryo by ERK [15]. The level of the 5hmC signal was increased in both paternal and maternal pronuclei at the PN3 stage following the treatment with a low concentration of ERK1/2 inhibitor. A previous study reported the suppression of Tet1 by the ERK pathway in cells [52]. However, Tet1 was barely expressed at the PN stage. Therefore, we focused on Tet3, which has previously been reported as a key regulator of the 5hmC asymmetry between paternal and maternal pronuclei [30,31,53]. As expected, the Tet3 localization in paternal pronuclei was increased following the treatment with a low concentration of ERK1/2 inhibitor, suggesting that ERK1/2 controls the process of paternal 5mC oxidation by preventing pronuclear localization of Tet3 in the zygotic genome.
G9a is essential for preimplantation development [34,54] and limits the range of the promiscuous transcriptions during exposure to stress [55]. Following the treatment with a higher concentration of ERK1/2 inhibitor, the H3K9me2 deposition and G9a localization in paternal pronuclei were both increased. Meanwhile, the number of blastocysts from zygotes with the same treatment was decreased sharply, suggesting that the limitation of G9a on the promiscuous transcription of the zygotic genome and the lethality of excess H3K9me2 deposition in paternal pronuclei during early development.
Our study shows that G9a plays a vital role in 5mC oxidation during paternal genomic reprogramming. The excess pronuclear localization of G9a antagonized the tight binding of Tet3 to paternal chromatin and effectively blocked the accumulation of paternal 5hmC in both ERK1/2 and G9a-inhibited zygotes, suggesting that fertilization-dependent ERK1/2 promotes rapid oxidation of paternal 5mC by inhibiting G9a pronuclear localization. The catalytic activity of G9a has previously been reported that essential for protecting the maternal genomic methylation from Tet3-mediated 5mC oxidation [22]. Interestingly, in both ERK1/2 and Tet3-inhibited zygotes where G9a pronuclear localization increased and catalytic activity was maintained, the paternal genomes showed a significant H3K9me2 deposition and a downward trend in de novo 5mC accumulation that ERK1/2 inhibition produced. Furthermore, compared with paternal genomes, the accumulation of de novo 5mC and 5hmC in maternal genomes both showed a lag after ERK1/2 inhibition. These data are consistent with the notion that H3K9me2 protects the zygotic genome against the de novo 5mC [56,57]. Therefore, our findings support the notion that G9a pronuclear localization impedes the oxidation of paternal 5mC during zygotic epigenetic reprogramming, regardless of whether G9a retains its catalytic performance or not.
Our study also reveals the crosstalk between 5mC and H3K9me2 in the zygotic genome. In our ERK1/2 inhibition system, Tet3 converts de novo 5mC into 5hmC, while the deposition of H3K9me2 prevents the accumulation of de novo 5mC in the zygotic genome. Recently, the particular epigenetic landscapes show that 5mC is largely independent of H3K9me2 in differentiated [33] and diseased cells [58,59]. H3K9me2 protects 5mC inherited from gametes, but it might prevent the accumulation of de novo 5mC at some loci in zygotes. Therefore, this independence between genomic 5mC and H3K9me2 can be explained by DNA replication and the inheritance 5mC dilution.
Conclusions
We have demonstrated that the inhibition of zygotic ERK1/2 causes excess deposition of H3K9me2 and accumulation of 5mC and its oxides in paternal pronuclei by triggering the pronuclear localization of G9a and Tet3, respectively. In conclusion, we propose that zygotic reprogramming is regulated by fertilization-dependent ERK1/2, which contributes to the development competence of pre-implantation embryos in mice.
Sperm collection
Sperm was obtained from ICR males aged 10-20 weeks. The cauda epididymis was cut open with the tip of syringes to allow sperm swimming out. A mass of sperm was put into Human Tubal Fluid (HTF) fertilization medium supplemented with 4 mg/ml bovine serum albumin (BSA, Sigma-Aldrich) and incubated for 1-1.5 h at 37 °C in 5% CO 2 .
In vitro fertilization
ICR females aged 4-6 weeks were superovulated by intraperitoneal injection of 10 U pregnant mare's serum (PMS) and 10 U of human chorionic gonadotropin (HCG) 48 h later. Cumulus oocyte complexes collected 14 h post HCG injection were incubated with capacitated sperm for 2.5 h. Tet3 inhibition was performed by supplementing with 1 mM dimethyloxallyl glycine (DMOG, Sigma-Aldrich) in fertilization medium; oocytes were incubated with DMOG for at least 40 min before addition of sperm.
Gene expression analysis
RNA-seq data of mice (GSE98150) were downloaded from Gene Expression Omnibus. Gene expression of MAPKs during early embryo development was normalized with log2(count + 1).
Blastocysts (30 for each treatment) were collected at day E4, and mRNA was purified using Dynabeads mRNA DIRECT ™ KIT (invitrogen) following the manufacturer's instructions. Random primed reverse transcription was performed using HiScript III RT SuperMix with gDNA wiper (R323-01, Vazyme, Nanjing, China). cDNA was added to 10 µl of quantitative PCR mix (Q111-02, Vazyme, Nanjing, China). RT-PCR reactions were performed on a Step-One Plus Real-Time PCR system (Applied Biosystems, Carlsbad, CA, USA). The primers for quantitative analysis are shown in Additional file 2: Table S1. Gene expression was calculated using the 2 ΔΔct method, and GAPDH was used for normalization as endogenous reference genes.
Data analysis
Images were analyzed using ImageJ software. The midsection of each pronucleus was identified using Hoe-chst33342 staining and determined by the maximal area. The midsection was used to quantify the total intensity following the subtraction of the signal corresponding to the cytoplasmic area (representing staining background). Statistical analysis was carried out using two-tailed Student's t-test with Welch's correction when required, using GraphPad Prism software. For gene expression analysis, statistical analysis was performed using two-tailed unpaired t-test. At least three biological replicates were performed for each analysis. Each replicate was conducted by an independent experiment at different times. | 2022-03-08T14:19:31.174Z | 2022-03-07T00:00:00.000 | {
"year": 2022,
"sha1": "ffed55fe0a16e2d3cce5ddfc90c07648741eb25f",
"oa_license": "CCBY",
"oa_url": "https://cellandbioscience.biomedcentral.com/track/pdf/10.1186/s13578-022-00758-x",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "8134332421a69c5191893a40cf277cb7270ef28b",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
49864463 | pes2o/s2orc | v3-fos-license | The Toronto Cognitive Assessment (TorCA): normative data and validation to detect amnestic mild cognitive impairment
Background A need exists for easily administered assessment tools to detect mild cognitive changes that are more comprehensive than screening tests but shorter than a neuropsychological battery and that can be administered by physicians, as well as any health care professional or trained assistant in any medical setting. The Toronto Cognitive Assessment (TorCA) was developed to achieve these goals. Methods We obtained normative data on the TorCA (n = 303), determined test reliability, developed an iPad version, and validated the TorCA against neuropsychological assessment for detecting amnestic mild cognitive impairment (aMCI) (n = 50/57, aMCI/normal cognition). For the normative study, healthy volunteers were recruited from the Rotman Research Institute registry. For the validation study, the sample was comprised of participants with aMCI or normal cognition based on neuropsychological assessment. Cognitively normal participants were recruited from both healthy volunteers in the normative study sample and the community. Results The TorCA provides a stable assessment of multiple cognitive domains. The total score correctly classified 79% of participants (sensitivity 80%; specificity 79%). In an exploratory logistic regression analysis, indices of Immediate Verbal Recall, Delayed Verbal and Visual Recall, Visuospatial Function, and Working Memory/Attention/Executive Control, a subset of the domains assessed by the TorCA, correctly classified 92% of participants (sensitivity 92%; specificity 91%). Paper and iPad version scores were equivalent. Conclusions The TorCA can improve resource utilization by identifying patients with aMCI who may not require more resource-intensive neuropsychological assessment. Future studies will focus on cross-validating the TorCA for aMCI, and validation for disorders other than aMCI.
Background
Brief tests such as the Mini-Mental State Examination (MMSE) [1] and the Montreal Cognitive Assessment (MoCA) [2] are popular screens for cognitive function. Neuropsychological assessments facilitate better understanding of cognitive performance for diagnosis but are time consuming, resource intensive, and suited for administration only by neuropsychologists-a resource that is often not readily available. Consequently, given the growing emphasis on early detection of cognitive impairment, there is a need for assessment tools that are intermediate between brief screening tests and neuropsychological batteries, can be administered by physicians as well as any health care professional or trained assistant in any medical setting, and can accurately identify mild cognitive decline. To accomplish this goal, the psychometric properties of the Behavioural Neurology Assessment [3], a screening test covering a broad spectrum of cognitive functions for diagnosing mild to moderate dementia, were significantly enhanced to detect mild cognitive deficits by development of the Toronto Cognitive Assessment (TorCA). This was done through the addition of more robust verbal learning and delayed recall, a complex figure copy with delayed recall, semantic knowledge items, a version of Trails A and B, and revision of the subset of language tests.
Our objectives were to obtain normative data on the TorCA and to validate this test for detection of amnestic mild cognitive impairment (aMCI). In addition to the paper version, we developed an electronic application for the iPad and assessed equivalency between the two versions. The advantages of an electronic application include automatic scoring, automatic point-of-care data collection for potential data entry into a clinical or research registry, a printable summary of results, and graphical representation of percentile performance on each cognitive domain.
Test description
The TorCA consists of 27 subtests within seven cognitive domains-Orientation, Immediate Recall, Delayed Recall, Delayed Recognition, Visuospatial Function, Working Memory/Attention/Executive Control, and Language (Table 1)-and can be administered by any health care professional or trained assistant and is suitable for use in any medical setting. Domain index scores represent addition of subtest scores within each domain. The Sum Index represents addition of all subtest scores.
Orientation
There are 12 items included: year, month, day, date, season, place/building, floor, city, province, country, Prime Minister, and Premier of the province.
Immediate Verbal Recall
The CERAD 10-Word list [4] is presented over three trials.
Delayed Verbal and Visual Recall
Delayed recall of the CERAD Word List and the Benson Figure Copy [5] are assessed after at least 10 min.
Delayed Verbal and Visual Recognition
Recognition of whether words appeared in the CERAD list and which one of four complex figures was copied are assessed.
Visuospatial Function
This scale consists of Clock Drawing [6] and the Benson Figure Copy [5].
Working Memory/Attention/Executive Control
Working memory and attention are assessed by Digit Span and Serial Subtractions. Executive control [7] is assessed by drawing Alternating Sequences, Verbal Letter Fluency, and Trail Making A and B [8]. A left-right reversed version of Trail Making is used to reduce practice effects on the standard version.
Language
There are eight subtests included: Verbal Fluency (animal names), confrontation naming of 15 items from the Multilingual Naming Test (MINT) [9], Sentence Repetition, Sentence Comprehension, Single Word Reading and Comprehension (auditory and reading), and Semantic Knowledge.
TorCA Sum Index
Consistent with standard practice in neuropsychology, there is no upper limit on Verbal Fluency for "F" words and animals. Therefore, there is no maximum on the Sum Index.
Standardization and normative sample
The study was approved by the Research Ethics Board at Baycrest Health Sciences. Healthy volunteers (n = 303) were recruited from the Rotman Research Institute (RRI) registry. There were four age groups: 50-59, 60-69, 70-79, and 80-89 years. Exclusion criteria were history of neurological disease, drug abuse, head injury with loss of consciousness, attention deficit hyperactivity disorder, active psychiatric illness, or use of medication containing any opioid. Non-native English speakers were included if they could understand all instructions. For test items, and administration and scoring instructions, see the Toronto Dementia Research Alliance website (www.tdra.ca). Figure 1 shows a flow chart of the participants analyzed in the normative study.
Reliability
To assess test stability, the TorCA was readministered to 29 participants after a median interval of 73 days (range 28-120) with mean difference, percentage score change, and stability coefficients (Pearson r) calculated between the first and second tests. Internal consistency was determined by calculating Cronbach's α for domain and Sum Index scores from the normative data study. Figure 2 shows a flow chart of the participants analyzed in the validation study.
As it proved difficult to find individuals with normal cognition in memory clinics, the remaining 31 normal participants were recruited from the current normative study sample and the community. The paper version of the TorCA was administered prior to neuropsychological assessment in all but three instances. The interval between neuropsychological assessment and TorCA was within six months.
As assessments were conducted in a clinical context, the neuropsychologists were aware of the TorCA scores and differential diagnoses. The majority of neuropsychological assessments were conducted by trained assistants not directly involved in the diagnostic process, although one of the neuropsychologists tested 42 participants. The TorCA was conducted by trained nurses, medical trainees, or research assistants who were blinded to the neuropsychological assessment results.
Exclusion criteria for the validation study were medical or neurological disorders that could cause cognitive deficits including untreated sleep apnea, traumatic brain injury with loss of consciousness greater than 30 min, history of stroke, attention deficit hyperactivity disorder requiring medication, substance abuse, or other significant psychiatric disorders.
All participants with aMCI met published criteria [19]. Objective memory impairment was defined as deficits on three of four memory tests relative to expectations based on age, education, and intellectual status. Memory tests were WMS-R Logical Memory, KBNA Word List [10], KBNA Complex Figure, and WAIS-III Digit Symbol incidental recall [11]. Deficit was defined as 1.5 standard deviations below estimated IQ based on the two-subtest IQ estimate of the WASI. Memory deficits had to occur at encoding or retention stages. Isolated retrieval deficits were not sufficient for diagnosis of aMCI.
Concurrent validity was determined by the ability of the TorCA to discriminate between aMCI and NC participants. Construct validity was determined by correlations between TorCA subtests and neuropsychological tests in the aMCI and NC groups and by testing for expected group differences on TorCA indices and subtests.
The results of the test-retest study using the paper version in normal participants are presented in Table 7. The scores remained remarkably stable across the retest intervals. Only the Memory-Immediate Recall (MIR), Memory-Delayed Recall (MDR), and Sum Index scores demonstrated significant increases and the increase in the latter was due to increase in the MIR and MDR indices. This indicates that there was a practice effect on the memory tests. Stability coefficients ranged from low The intratest reliabilities of the TorCA indices are presented in Table 8. Reliability estimates ranged from low to good. The low coefficients of Orientation, Memory-Delayed Recognition, and Visuospatial Indices again are attributable to the restricted range of scores noted earlier. The Delayed Recall Index reliability coefficient was calculated by comparing the results of the Memory-Delayed Verbal Recall and the Memory-Delayed Visual Recall subtests and therefore did not represent a homogeneous construct. The Visuospatial Index reliability coefficient was calculated by comparing the results of the Benson Figure Copy and Clock Drawing subtests. Although both Benson Figure Copy and Clock Drawing measure visuospatial function, Clock Drawing is also a measure of planning, monitoring, and abstraction. Thus, these subtests are not homogeneous. Likewise, the Working Memory/Attention/Executive Control Index is not homogeneous in construct as it consists of measures of attention, working memory, conceptualization, and reasoning. Table 9 presents demographic features of the aMCI and NC groups. The groups did not differ in mean age, education, or Full-Scale IQ. The NC group had a higher proportion of females (67%) to males (33%) (χ 2 = 6.33, p < 0.02), whereas the aMCI group had an approximately equal gender balance (54% male; 46% female). Effect sizes based on difference between group means and standard deviations for neuropsychological tests used to determine group membership are provided in Overall, the aMCI group scored lower on neuropsychological testing but the largest effect sizes, in excess of 1.5 SD, were obtained on learning and episodic memory, thereby substantiating group classification as aMCI. Table 9 presents between-group differences on TorCA indices. The aMCI group achieved a significantly lower TorCA Sum Index than did the NC group (F(1,105) = 36.86, p < 0.001). A MANOVA on the remaining seven domain indices revealed a significant effect for group (Wilk's λ = 0.37, F(1,99) = 23.78, p < 0.001). Pairwise comparisons, with Bonferroni correction for seven multiple comparisons at p ≤ 0.05/7 (0.007), revealed significant differences for orientation, immediate memory recall, delayed memory recall, and delayed memory recognition indices.
Validation in aMCI
Prior to analyzing TorCA subtest scores for group differences, boxplots for each subtest were inspected. Distribution of scores on Trail Making (completed trials measure, total correct minus incorrect lines), Alternating Sequences, Similarities, Sentence Repetition and Comprehension, Single Word Reading and Comprehension, and Semantic Knowledge showed a marked negative skew with a ceiling effect for both groups. Kolmogorov-Smirnov tests on these subtests revealed no differences in distribution of scores between the two groups. Therefore, these subtests were dropped from further between-group analyses. Scores on Verbal Learning, Verbal Recall, Verbal Recognition, Visual Recall, Serial Subtractions, Digit Span, Trail Making A and B completed times measure, Benson Figure Copy, Clock Drawing, Verbal Fluency-F Words, Verbal Fluency-Animals, and MINT Naming were analyzed with a MANOVA for between-group differences (Table 10). There was a significant group effect (Wilk's λ = 0.36, F(13,93), p < 0.001). Table 10
Construct validity
The neuropsychological tests were grouped into nine domains: Immediate Recall, Delayed Recall, Delayed Recognition, Visuospatial, Cognitive Flexibility, Attention/ Concentration, Executive Control, Verbal Fluency, and Language. Correlations between TorCA and neuropsychological domains are presented in
Equivalency of paper and iPad versions
There was a strong correlation between paper and iPad versions (r(43) = 0.86, p < 0.001) and no difference be- Test-retest reliability between first and second administration was good (r(43) = 0.87, p < 0.001). There was no association between test-retest interval and
Discussion
The TorCA was administered to 303 healthy volunteers between ages 50 and 89 years, yielding a relatively brief assessment of multiple cognitive domains with median administration time of 34 min. Test-retest results remained relatively stable over a median of 73 days (range 28-120) with mean increase of only 3.3 points. Age and education accounted for only 5% of the variance in total score. Although age-adjusted norms are available for each decade from 50 to 89 years, the TorCA can be administered across this range with minimal need for age correction. Paper and iPad version scores were not significantly different. The iPad version provides easier administration with near automation of scoring and graphical representation of percentile scores (Fig. 4).
Overall stability was good with only modest increase in the Sum Index on retesting. Stability coefficients were low for Orientation, Delayed Recognition, Visuospatial Function, and Working Memory/Attention/Executive Control due to the restricted range of scores. Nevertheless, these scores demonstrated a very small percentage change in scores. The change in the Sum Index (1.1%) reflected increases in the immediate and delayed memory indices (14.3% and 10.7% respectively) with no other index exceeding an increase of 1.5% (Language).
Internal consistency of the Sum Index was adequate and reflected the heterogeneous nature of individual tests. Low internal consistency reflected the diverse nature of cognitive abilities on Delayed Recall and Working Memory/Attention/Executive Control. The former combines verbal and visual memory, whereas the latter combines heterogeneous measures related to frontal system function. Low internal consistency also reflected restricted range in scores on Orientation, Delayed Recognition, and Visuospatial Function. We validated the TorCA for detection of aMCI based on a need for cognitive assessment tools that can identify early decline, that are much shorter than typical neuropsychological batteries, and that can be administered by any health professional or trained assistant. A combination of TorCA subscores yielded correct classification, sensitivity, and specificity of over 90%. Logistic regression revealed that scores in four domains-Immediate Recall, Delayed Verbal and Visual Recall, Visuospatial Function, and Working Memory/ Attention/Executive Control-correctly classified 92% of participants, and yielded an easily applied formula to calculate the probability of aMCI (www.tdra.ca). This is automatically calculated with the iPad version of the TorCA. It should be emphasized that the correct classification of 92% arises from four domains of the TorCA rather than the total score on the entire test. In contrast, correct classification was 79% based on the Sum Index (total score).
Although the logistic regression probability of 0.55 for aMCI is the optimal cutoff value, this may not always represent the best decision value for determining positive or negative cases. If sensitivity and specificity are held constant, PPV decreases as pretest disease probability (prevalence) decreases and increases as pretest probability increases. Conversely, NPV increases with decrease in pretest probability and decreases as pretest probability increases. PPVs and NPVs listed earlier for the optimal value relate only to the pretest probability of aMCI in our sample (50/107 = 0.47). Table 13 presents the range of PPV and NPV values for a cutoff value of 0.55 for pretest probabilities ranging from 0.05 to 0.90. PPVs and NPVs for a cutoff value of 0.90 are also provided. If a logistic regression value of 0.55 or higher is obtained for individuals with pretest probability of 0.20, then 72% will be correctly classified as aMCI. However, 28% will be misclassified, which is unacceptable. At the same level of pretest probability, a logistic regression value less than 0.55 results in correctly ruling out aMCI in 98% of negative cases. At a pretest probability of 0.20, raising the "rule-in" predicted value to 0.90 results in 88% of positive cases being true aMCI with only 12% false positives. A level of 0.20 was chosen in these examples because this is approximately the estimated prevalence of aMCI in community samples [24]. Based on the validation data for TorCA Sum Index reported in this article, the TorCA is comparable to published data on the MoCA for detection of MCI. A meta-analysis of 20 studies conducted by Ciesielska et al. [25] reported that a MoCA cutoff value of 25/30 correctly yielded a sensitivity of 80% and specificity of 81%. A meta-analysis of nine studies [26] evaluating the MoCA's ability to discriminate aMCI from normal controls found that a cutoff value of 23/30 yielded a correct classification of 86% (95% CI 83-90%) with a sensitivity of 83% (95% CI 76-89%) and specificity of 88% (95% CI 84-92%), while the original cutoff value of 26/30, as suggested by Nasreddine et al. [2], yielded correct classification of only 78% (95% CI 75-82%) with sensitivity of 94% (95% CI 91-97%) and specificity of 66% (95% CI 60-71%). This compares to correct classification of 79% for the TorCA with a sensitivity and specificity of 80% and 79% using the Sum Index. The TorCA is also comparable to the Addenbrooke's Cognitive Examination (ACE-R and ACE III) based on published data [27,28]. Ahmed et al. [27] reported that the ACE-R correctly classified 74% (95% CI 56-87%) of MCI and normal controls with a sensitivity of 90% (95% CI 58-98%) and specificity of 67% (95% CI 41-84%). Matias-Guiu et al. [28] reported that the ACE-III correctly classified 75% (95% CI 66-82%) of MCI and normal controls with a sensitivity of 77% (95% CI 62-87%) and specificity of 75% (95% CI 62-83%). Although confidence intervals were not provided in the reports by Ahmed et al. and Matias-Guiu et al. [27,28], we calculated them for comparison to our data.
The TorCA has potential resource allocation implications in centers with neuropsychology resources by identifying patients who do not require neuropsychological assessment due to a high probability of aMCI or because this disorder is effectively ruled out. Although the logistic regression was exploratory, a reasonable strategy might be to rule out aMCI if probability, based on the logistic regression formula, is below 0.55. Due to the likelihood that the logistic regression formula overestimates classification [29], we recommend a value of 0.90 or higher to rule in aMCI. For values between 0.55 and 0.90, referral should preferably be made for neuropsychological assessment to confirm diagnosis. In the absence of available neuropsychology resources, these patients should be followed to establish diagnosis. Study limitations should be acknowledged. First is the need for cross-validation. Whereas the validation study revealed that the use of the logistic regression formula would refine the identification of aMCI, this represents an initial, exploratory result and further cross-validation of the formula is needed to confirm critical values and stability of constituent indices. A second limitation is that the logistic regression formula for probability of aMCI applies only to differential diagnosis of aMCI vs normal aging. Future studies are needed to validate the TorCA for differentiating aMCI from other cognitive disorders, and to determine whether it performs equally well for identifying single vs multiple domain aMCI. A third limitation is that participants in the validation study had relatively high IQs. Studies are needed to determine validity of the TorCA for diagnosing aMCI in participants with lower IQs. In addition, a caution is that interpretation of positive or negative cases must take into account differences between patients' estimated pretest probabilities of a condition and prevalence of the condition in validation studies. A fourth limitation is that the orientation items consisting of Prime Minister, Premier, and season are country specific. This will be addressed in future by translating the TorCA into languages other than English and carrying out normative and validation studies using the translated tests. Ideally, normative and validation studies should also be carried out in English-speaking countries other than Canada. Finally, this study focused only on aMCI from a diagnostic perspective. Future studies will be needed to validate the TorCA for diagnosis of other forms of mild cognitive decline. It is likely that the discriminating indices on the TorCA will differ from those that predict aMCI. Chair in Curative Approaches to Alzheimer's Disease. GN receives support from the George, Margaret and Gary Hunt Family Chair in Geriatric Medicine, University of Toronto. SCS, RS, and TG received partial grant support from CIHR MOP 201403, the Ontario Brain Institute, and Brain Canada. The funding sources had no role in the study design, in the collection, analysis and interpretation of data, in writing of the report, and in the decision to submit data for publication.
Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Ethics approval and consent to participate The study was approved by the Research Ethics Board at Baycrest Health Sciences. Written informed consent was obtained from all participants.
Competing interests MF received financial support for a Behavioural Neurology fellow from Eli Lilly Canada, served on an advisory board for Eli Lilly Canada, receives royalties for a book on Clock Drawing from Oxford University Press, is listed on a provisional patent related to methods and kits for differential diagnosis of Alzheimer's disease vs frontotemporal dementia using blood biomarkers, and may be listed on the planned patent application, and serves on the editorial board of Brain and Cognition. | 2018-07-21T03:32:53.943Z | 2018-07-18T00:00:00.000 | {
"year": 2018,
"sha1": "ba3d10f44aa26cbcf4ceec68f88d1d2590d4788e",
"oa_license": "CCBY",
"oa_url": "https://alzres.biomedcentral.com/track/pdf/10.1186/s13195-018-0382-y",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "975529cc847af556ceef49aed28dbf66db7dc1a4",
"s2fieldsofstudy": [
"Psychology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine",
"Psychology"
]
} |
5971418 | pes2o/s2orc | v3-fos-license | Evolution of Antifreeze Protein Genes in the Diatom Genus Fragilariopsis: Evidence for Horizontal Gene Transfer, Gene Duplication and Episodic Diversifying Selection
Hypotheses about horizontal transfer of antifreeze protein genes to ice-living diatoms were addressed using two different statistical methods available in the program Prunier. The role of diversifying selection in driving the differentiation of a set of antifreeze protein genes in the diatom genus Fragilariopsis was also investigated. Four horizontal gene transfer events were identified. Two of these took place between two major eukaryote lineages, that is from the diatom Chaetoceros neogracile to the copepod Stephos longipes and from a basidiomycete clade to a monophyletic group, consisting of the diatom species Fragilariopsis curta and Fragilariopsis cylindrus. The remaining two events included transfers from an ascomycete lineage to the proteobacterium Stigmatella aurantiaca and from the proteobacterium Polaribacter irgensii to a group composed of 4 proteobacterium species. After the Fragilariopsis lineage acquired the antifreeze protein gene from the basidiomycetes, it duplicated and went through episodic evolution, characterized by strong positive selection acting on short segments of the branches in the tree. This selection pattern suggests that the paralogs differentiated functionally over relatively short time periods. Taken together, the results obtained here indicate that the group of antifreeze protein genes considered here have a complex evolutionary history.
Introduction
Genes coding for antifreeze proteins (AFPs) have been discovered in various taxa that occur in cold temperatures; these include unicellular eukaryotes, plants, bacteria, fungi, fish, crustaceans and insects. [1][2][3][4][5][6][7][8][9] Two major groups of AFPs are recognized. One type prevents cell/body fluids from freezing while the other kind consists of proteins that make it possible for organisms to survive cell/body fluid freezing. 1,5 It is thought that both AFPs and protein ice nucleators play an important role in both freeze avoidance and tolerance. 1,5 The former prevents freezing while the latter limits supercooling and induce freezing. 1 In some prokaryotes and eukaryotes that survive cell/ body fluid freezing, the AFPs act as cryoprotectants. Even though this process is not fully understood, it may take place through ice recrystallization inhibition and possibly also by cell membrane stabilization. 5,14 Those organisms that survive through freeze avoidance have AFPs that lower the freezing temperature of body fluids noncolligatively without affecting the melting temperature. 1,5 Genes coding for proteins with antifreeze properties have probably evolved independently in a number of lineages 10 while some organisms have most likely acquired the genes through horizontal gene transfer (HGT). 4,11,12 The evolution of prokaryote genomes is thought to have been profoundly influenced by lateral gene transfer. 13 The effect of this process on eukaryote genome evolution appear to be much more prevalent than previously thought. This is evidenced by an increasing number of documented HGTs in eukaryotes, particularly in unicellular eukaryotes. 13,14 For instance, the nuclear genomes of the diatoms Phaeodactylum tricornutum and Thalassiosira pseudonana appear to contain a large number of prokaryote genes. [15][16][17] This has probably also contributed to the great success of the diatoms in different ecosystems. 16 All the currently known AFPs in ice-living diatoms appear to act as cryoprotectants. 4,11 AFPs, which show similar amino acid sequences to those found in diatoms, have also been described for different prokaryote, fungal and crustacean lineages. 4,11 This observation suggests that diatoms may have acquired the AFP genes from distantly related taxa through HGTs. 4,11 However, another possibility, which cannot completely be ruled out, is convergent evolution of AFP genes in diatoms and certain prokaryote, fungal and crustacean lineages. 4,11 In this study, hypotheses about horizontal transfer of AFP genes to sea-ice living diatoms from distantly related taxa were tested using the program Prunier-an algorithm developed by 18 for inferring HGT events based on statistical criteria. The role of diversifying selection in driving the differentiation of a set of duplicated AFP genes in the diatom genus Fragilariopsis was also investigated. 19 The results indicated that at least four statistically significant HGT events, involving AFP genes and genes coding for antifreeze-like proteins (AFLPs) took place. Two of these events included diatoms. It was also shown that episodic diversifying selection has most likely been instrumental in driving the differentiation of the paralogs in the two Fragilariopsis species, suggesting that the duplicated AFP genes have changed functionally over relatively short time periods.
Material and Methods Sequences
The choice of taxa in the current study was largely influenced by the paper of, 11 that is one of the major goals was to test hypotheses of HGT using the taxa shown in Figure 2 in 11 Moreover, amino acid sequences similar to the various AFPs of the two Fragilariopsis species were used. The degree of similarity was determined through multiple NCBI blasts. Another aspect considered when choosing taxa was to include representatives of different higher level taxa (especially eukaryotes), that is as long as they were reasonably similar to the different AFP genes of the two Fragilariopsis species. The SSU rRNA data and the amino acid sequences (+CDS) of the AFPs considered here were downloaded from the GenBank. The accession numbers of all the downloaded SSU rRNA sequences and proteins are shown in Table 1 11 and references therein and Chaetoceros neogracile. 20 The remaining amino
Data analyses
The software package DAMBE (Version 5.0.5), 21 was implemented to manage the data and to match the codons against the aligned amino acid sequences. The alignments of the amino acid and SSU rRNA sequences, which are available upon request, were generated using the MAFFT (Version 6 22 ) Web Server available at http://mafft.cbrc.jp/alignment/server/. The alignment mode G-INS-i 23 with an offset value of 0.1 and the MAFFT homolog function "turned on" (all the other parameter settings were default values), was used for the proteins. When aligning the SSU rRNA sequences option Q-INS-i 24 was implemented.
Only the highly conserved regions in the SSU rRNA alignment were used in the phylogenetic analysis. A species phylogeny was created based on the NCBI taxonomic classification (http://itol.embl.de/ other_trees.shtml). Since the relationships among some of the eubacteria lineages ( Fig. 1) were unresolved, a SSU rRNA phylogeny was inferred using BayesPhylogenies (available from http://www. evolution.rdg.ac.uk/BayesPhy.html). This program implemented a joint model that accommodated shifts in site-specific substitution rates over time (ie, heterotachy) and among-site rate heterogeneity (ie, "pattern heterogeneity") 25,26 ; these phenomena are expected to occur in SSU rRNA-based phylogenetic analyses of prokaryote/eukaryote lineages. If heterotachy and "pattern heterogeneity" are not accounted for in phylogeny reconstruction, the resulting relationships are likely to be distorted. 26 A reversible-jump Markov chain Monte Carlo (rjMCMC) algorithm was used to determine how many distinct among-site rate-variation patterns, and branch length parameters (with a maximum of two parameters for each branch) were required to optimally describe the empirical data matrix. This approach is appealing because it requires far fewer parameters than conventional mixture models to describe heterotachy and "pattern heterogeneity". 25 Monte Carlo) analyses (each with 3 chains running for 2 × 10 7 generations, sampling every 10 3 generations) were carried out to estimate the posterior distribution of phylogenetic trees, and post-burnin samples (with burnin set to 10%) from all analyses were combined. Convergence of the MCMC runs was determined by visually examining the cumulative posterior and between-run variation in split frequencies 27 using the on-line tool AWTY ("Are We There Yet"). 28 The software package FigTree (available from http://tree.bio. ed.ac.uk/software/ figtree/) was used to generate the trees.
The program Prunier (version 2.0) 18 which works in conjunction with Treefinder 29 was used to detect potential HGTs between the species included in this investigation. The "slow" Prunier method uses Kishino-Hasegawa, 30 Shimodaira-Hasegawa 31 expected likelihood weights 32 and the approximately unbiased 33 tests to infer whether topological differences between the gene and species trees are statistically significant. The "fast" Prunier method uses LR-ELW (Expected-Likelihood Weights applied to Local Rearrangements) edge support values 32 to identify such discrepancies. HGT events are inferred when statistically significant topological conflicts between the species and gene trees were identified. 18 Both the "fast" and "slow" methods, which in simulations have been shown to perform equally well 18 were used in the current study. The aligned amino acid sequences and the species phylogeny were provided as input for the Prunier runs. In each Prunier run, the gene trees were inferred by the Treefinder program. 29 For the "slow" method the default settings were used. The following settings were implemented for the "fast" method: boot.thresh.conflict = 95 (ie, support value threshold for topological conflict); fwd.depth = 1 (ie, maximal depth at which Prunier looks forward to find a significant HGT when the current HGT is not significant). The "boot.thresh.conflict" number is the minimum LR-ELW edge support value for a given node in the gene tree used for recognizing topological conflict between the gene and species trees. The "fwd.depth" (for further details see http://pbil.univ-lyon1.fr/software/prunier/) is the maximal depth at which the "fast" method Prunier looks forward to find a significant HGT when the current event is not significant. A depth value of 1 implies that if no significantly supported conflict can be removed with one HGT event the algorithm looks one step "forward" to see if the next HGT in the list will remove a significant conflict. If Prunier finds a better solution with a depth value of 2, it provides this solution in the output.
A newly developed unrestricted random effects branch-site model 19 was implemented for the purpose of detecting episodic diversifying selection on codons in the paralogs of the two Fragilariopsis species. This method, which is available in a free public web implementation, 34 has been shown to have low error rates and to be powerful in identifying episodic adaptive evolution restricted to a few amino acid sites in a gene tree. 19 Without prior assumptions of which lineages have gone through episodic adaptive evolution, this approach can, through the use of likelihood ratio tests, identify all the branches in a tree which have a certain proportion of sites evolving with dN/ dS values significantly higher than 1. 19 The gene tree required for this branch-site analysis was inferred through a partitioned maximum likelihood analysis in Treefinder. 29 The following optimal models, elucidated through a testing procedure in Treefinder,
Species and gene trees
The initial species phylogeny was generated based on the NCBI taxonomic classification information. This tree showed unresolved relationships for Candidatus Methanoregula boonei, Colwellia sp., Psychromonas ingrahamii, Stigmatella aurantiaca and Rhodoferax ferrireducens. Since the methods in Prunier require a fully resolved species tree, a Bayesian phylogenetic analysis of a sample of SSU rRNA sequences (Table 1) was conducted. In the absence of other data, SSU rRNA sequences can serve as a first approximation to a species phylogeny because these genes seem to undergo lateral transfers at relatively low frequencies. The resulting consensus tree resolved the branching orders of Candidatus Methanoregula boonei, Colwellia sp., Psychromonas ingrahamii, Stigmatella aurantiaca and Rhodoferax ferrireducens (Fig. 1) and did not show any conflicts with the NCBI taxonomic classification information. The resolved species tree, with posterior clade probabilities for the taxa included in the Bayesian phylogenetic analysis are shown in Figure 1. The only poorly supported relationship is the one between Cytophaga hutchinsonii and Marivirga tractuosa. The species phylogeny was rooted on the branch between Eubacteria and Archaea/Eukaryota (Fig. 1) and the phylogenetic relationships are as expected, that is the Bacillariophyta (ie, the diatoms) form a well supported clade among the eukaryotes and so do the Ascomycota and Basidiomycota clades (Fig. 1). Among Eubacteria two major groups were inferred, namely Bacterioidetes and Proteobacteria (Fig. 1). Figure 2 shows the optimal unrooted maximum likelihood tree derived from the aligned amino acid sequences of the AFPs/AFLPs. A number of major conflicts are apparent between the species and gene trees (compare Fig.1 with Fig. 2). The prokaryotes Rhodoferax ferrireducens, Methanoregula boonei, Stigmatella aurantica are found within a eukaryote group that consists of Navicula glaciei, Chaetoceros neogracile, Stephos longipes, Phaeosphaeria nodorum and Talaromyces stipitatus (Fig. 2). Furthermore, within this eukaryote clade, the proteobacterium Stigmatella aurantica form a sister relationship with the ascomycete Talaromyces stipitatus and the copepod Stephos longipes share a common node with the diatom Chaetoceros neogracile. Rather than being grouped together with the other eukaryotes, the Fragilariopsis and basidiomycete lineages are found among a number of proteobacteria and bacterioidetes species (Fig. 2).
horizontal gene transfer
The outcome of the Prunier analyses, based on the "slow" and the "fast" methods, were the same, namely the same HGT events were inferred (see Fig. 1). From here on the results of the "fast" method is described and discussed. The arrows in Figure 1 show the direction of HGTs. For these events to be inferred the topological conflicts between the gene and species phylogenies had to have a minimum LR-ELW edge support values of 95% in the former tree (Fig. 2). Thus, two transfers occurred among eukaryotes, that is from Chaetoceros neogracile to Stephos longipes and from the ancestor of the basidiomycete group, composed of Flammulina populicola, Lentinula edodes, Typhula ishikariensis, Leucosporidium sp., to the ancestral lineage of the Fragilariopsis clade. In the other two transmissions, the genes moved from the ascomycete lineage (ie, Phaeosphaeria nodorum and Talaromyces stipitatus) to a proteobacterium (ie, Stigmatella autantiaca) and from Polaribacter irgensii to the ancestral lineage of a group, consisting of Shewanella denitrificans, Shewanella frigidamarina, Colwellia sp., Psychromonas ingrahamii. The result of the "fast" HGT analysis was not affected by the position of the root, that is whether it was assumed to be positioned between the Eubacteria and the Archaea/eukaryote clade or between the Eubacteria/Archaea clade and the eukaryotes. Lowering the minimum support value for recognizing topological conflict between the gene and species trees to 90% (ie, the default setting in Prunier for the "fast" method) did not affect the results.
Episodic diversifying selection analysis
Episodic evolution took place in 32% (15/47) of the branches in the Fragilariopsis gene tree (Fig. 3 and Table 2). The strength of diversifying selection (ω + ) and the length of the branch segments (q + ) affected by this selection regime showed substantial variation across the topology (Fig. 3 and Table 2). The former parameter (ω + ) was in most cases much larger than 8.0 whereas the latter parameter (q + ) ranged from 0.004 to 0.16. Thus, in the majority of the cases, "strong" positive selection acted on very "short" segments of the branches (Fig. 3 and Table 2).
Discussion horizontal gene transfer
Conflicts between species and gene trees can be due to lateral gene transfers between distantly related species. 18 The program Prunier (see Material and Methods) has been demonstrated to perform well when it comes to identifying incongruences between species and gene trees caused by horizontal gene transfers. 18 This method allowed me to further investigate hypotheses about horizontal transfers of AFP and AFLP genes between distantly related taxa.
The Prunier analyses of the amino acid alignment rejected the idea that the prokaryote genera Shewanella or Cytophaga have been instrumental in transferring genes coding for AFPs or AFLPs to the diatoms 4 -at least this appears to be the case with regard to the taxa considered here. The only diatom lineages that have "picked up" the AFP or AFLP gene from another taxon are Fragilariopsis curta and Fragilariopsis cylindrus (Fig. 1). The most likely donor are the basidiomycetes (Fig. 1) which also is consistent with the hypothesis proposed by. 4 The AFP genes of Chaetoceros neogracile and Navicula glaciei do not appear to have been acquired from other lineages. Instead the AFPs in these taxa may have evolved from ancestral genes with different functions. 4,11 This scenario suggests that the AFPs found in diatoms could have a complex origin 11 which may or may not involve HGT events. To further address this issue a wider taxonomic sampling of the AFP and AFLP genes needs to be considered. The AFP gene in Stephos longipes, a sea-ice copepod, has been transferred to it from Chaetoceros neogracile, 9 also see 11 (Fig. 1). This is not unexpected since C. neogracile occur in the same habitat as S. longipes and copepods are known to "graze" on diatoms.
Few studies, that provide evidence of eukaryoticto-prokaryotic HGTs, have been reported. 13 However, the acquisition of an AFLP gene by the proteobacterium Stigmatella aurantiaca from the ascomycetes Talaromyces stipitatus and Phaeosphaeria nodorum constitutes an example of such a rare event (Fig. 1). Gene transfers from eukaryotes to prokaryotes are disfavored due to the presence of introns in the former and because of conjugation/transduction processes in the latter. 13 Other reasons that have been mentioned are the eukaryote genes being of less adaptive significance to prokaryotes and possibly the lower frequency of interaction between representatives of the eukaryotic and prokaryotic domains relative to prokaryote-prokaryote contacts. 13
Chaetoceros neogracile
Stephos One prokaryote-to-prokaryote transfer took place, that is between Polaribacter irgensii and a lineage composed of Shewanella denitrificans, Shewanella frigidamarina, Colwellia sp., Psychromonas ingrahamii (Fig. 1). Perhaps what is the most surprising result in this study is that not a single prokaryoteto-eukaryote transfer was detected. These are events that are thought to have been common throughout the history of life. 13 This unexpected outcome suggests that the group of AFPs and AFLPs examined Table 2 for the strength of positive selection (ω + ) on each of the affected branches and for the proportion of the total branch length (q + ) influenced by this selective regime. The scale bar shows the expected number of substitutions per nucleotide.
here may have evolved independently in some prokaryotes and eukaryotes. Since the Prunier analysis did not identify the "clustering" of Methanoregula boonei and Rhodoferax ferrireducens with N. glaciei, C. neogracile, S. longipes in the gene tree as being the result of HGTs (Fig. 2), it is quite possible that this close association between distantly related taxa can be explained by convergent evolution. Due to the strong selection pressures for the evolution of proteins with antifreeze properties in ice-living organisms, it is conceivable that some of the cryoprotectants have originated independently in different polar taxa.
Not all the proteins included here have been shown to have antifreeze properties (see Material and Methods). In database searches a number of genes show a high degree of similarity to those with known antifreeze functions-some of those sequences were included in the current investigation (also see discussion in 11 ). Even though information about their distribution is scanty, many of the organisms that have AFLP genes are not known to occur in polar habitats nor in ice, suggesting that some of these proteins do not have a function related to cold tolerance. This is probably the case with Talaromyces stipitatus, Phaeosphaeria nodorum and Stigmatella aurantiaca. The former two lineages appear to have donated a gene coding for an AFLP to the S. aurantiaca (Fig. 1) but none of these species are known occur in the polar regions. The antifreeze activity of the proteins, found in life forms that are known to exist in sub-zero environments, may have evolved from genes with a different function. 11 Bayer-Giraldi and colleagues (2010) noted that many of the AFPs show a high degree of similarity to adhesins which are surface proteins with functions that could easily be modified, through molecular evolution, to have antifreeze properties (also see discussion in 11 ). Natural selection is expected to readily favor transfers of AFP or AFLP genes to organism living in sub-zero environments if they increase the chance of becoming better adapted to these extreme conditions over time (ie, evolution of AFLP genes to AFP genes) 13 for a general discussion on HGT and adaptations.
Episodic diversifying selection analysis
The results of the Prunier analysis suggested that the Fragilariopsis lineage acquired the AFP gene from the basidiomycetes after which the transferred gene(s) have gone through a number of duplication events (Fig. 3). The interesting question here is whether the duplicated genes in Fragilariopsis changed Fig. 3 for branch numbers) that were inferred to be under episodic diversifying selection; b mean ω is the average dn/dS estimated for each branch under the free-ratio Mg94 × rEV model (no site-to-site rate variation); c ωand d qvalues reflect the strength of negative selection (ω -) and the proportion of the total branch length affected by negative selection (q -); e ω n and f q n -values reflect (nearly) neutral evolution (ω n ) and the proportion of the total branch length affected by (nearly) neutral evolution (q n ); g ω + and h q + -values reflect the strength of positive selection (ω + ) and the proportion of the total branch length affected by positive selection (q + ); i LrT is the likelihood ratio test statistic; j p is the uncorrected P-value; k corrected p is the probability based on holm's multiple testing correction.
functionally over time. Functional modification of a gene is expected to be associated with shifts in the selective regime acting on it. 35,36 In fact, it has been demonstrated that positive selection can play a role in the early functional evolution of duplicate genes. 37 If positive selection is instrumental in driving functional changes one should observe rapid amino acid replacements in the encoded proteins, that is rates much higher than expected under neutral evolution and negative selection. 35,36 The current analysis showed that "strong" positive selection acted on "short" segments of the branches (Fig. 3 and Table 2). This scenario provides support for the Fragilariopsis paralogs having differentiated functionally over short time periods. Duplication of AFP genes in association with functional differentiation is thought to be an adaptive strategy for coping with cold stress 10,37,38 because it can expand the range of functions in AFPs. 10 Acknowledegments I thank James Raymond for introducing me to questions associated with horizontal transfer of antifreeze protein genes in diatoms and Vincent Daubin for advice on using Prunier. Three anonymous reviewers provided comments that improved the content of the paper.
Disclosures
Author(s) have provided signed confirmations to the publisher of their compliance with all applicable legal and ethical obligations in respect to declaration of conflicts of interest, funding, authorship and contributorship, and compliance with ethical requirements in respect to treatment of human and animal test subjects. If this article contains identifiable human subject(s) author(s) were required to supply signed patient consent prior to publication. Author(s) have confirmed that the published article is unique and not under consideration nor published by any other publication and that they have consent to reproduce any copyrighted material. The peer reviewers declared no conflicts of interest. | 2014-10-01T00:00:00.000Z | 2011-01-01T00:00:00.000 | {
"year": 2011,
"sha1": "764914644cfc82d117c2656af524b346d1c2b9be",
"oa_license": "CCBYNC",
"oa_url": "https://doi.org/10.4137/ebo.s8321",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "764914644cfc82d117c2656af524b346d1c2b9be",
"s2fieldsofstudy": [
"Biology",
"Environmental Science"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
248810981 | pes2o/s2orc | v3-fos-license | Opinion Polarization in Human Communities Can Emerge as a Natural Consequence of Beliefs Being Interrelated
The emergence of opinion polarization within human communities—the phenomenon that individuals within a society tend to develop conflicting attitudes related to the greatest diversity of topics—has been a focus of interest for decades, both from theoretical and modelling points of view. Regarding modelling attempts, an entire scientific field—opinion dynamics—has emerged in order to study this and related phenomena. Within this framework, agents’ opinions are usually represented by a scalar value which undergoes modification due to interaction with other agents. Under certain conditions, these models are able to reproduce polarization—a state increasingly familiar to our everyday experience. In the present paper, an alternative explanation is suggested along with its corresponding model. More specifically, we demonstrate that by incorporating the following two well-known human characteristics into the representation of agents: (1) in the human brain beliefs are interconnected, and (2) people strive to maintain a coherent belief system; polarization immediately occurs under exposure to news and information. Furthermore, the model accounts for the proliferation of fake news, and shows how opinion polarization is related to various cognitive biases.
Introduction
As Evan Williams, co-founder of Twitter, famously said in a 2017 interview, "I thought once everybody could speak freely and exchange information and ideas, the world is automatically going to be a better place. I was wrong about that" [1]. This is a good overview of the surprise that the unprecedented connectivity among people-primarily driven by various internet-based social media platforms in the early 21st century-brought us unprecedented factions, dissension and fake news [2,3], instead of agreement and conciliation [4,5]. The reasons behind these phenomena are diverse and manifold and, accordingly, are the subject of the most diverse scientific fields from history [6] through sociology [7] to computational social science [8][9][10].
The quest for finding a valid explanation-and a practicable model-for the phenomena of the above mentioned polarization and fragmentation has been underway for decades. Specifically, regarding computational models, an entire field, opinion dynamics , has emerged in order to study the way opinions, information, views and beliefs propagate in human communities [11][12][13][14][15][16]. In general, these models study the dynamics of attitudes related to a certain topic, such as issues related to climate change, abortion, immigration, vaccination, a certain politician or political party, etc. Typically, the attitudes of the agents towards the given issue are described by scalar values which are assumed to be altered due to communication with peers [11,12,17]. Within this scientific field, consensus refers to the state in which all social actors share the same opinion, polarization to the condition when During the last decades, a vast amount of knowledge has accumulated in scientific fields on the ways humans perceive the world, make decisions and structure their beliefs. Despite the fact that obtaining a detailed understanding of these processes still require further work, some scientific fields, such as neurobiology [32][33][34][35][36] or various human sciences [37,38], such as psychology [39][40][41][42], anthropology [43], economics [44][45][46] and political science [4,5,[47][48][49][50][51] have progressed considerably. From our point of view, the key finding is that in humans, opinions and beliefs never occur alone, that is, no concept or belief can exist in isolation. (Actually, humans are not even able to memorize anything without connecting it to something meaningful [32].) Rather, concepts and beliefs are organized into a structure, a system ("belief system") which has well-defined features [32,47,48,52]:
1.
First and fore-most, it seek to be consistent. This means that people try to maintain a belief system in which the elements mutually support each other, or are independent [48,52]. In the case of holding conflicting beliefs, people experience discomfort called cognitive dissonance [53,54] which people will try to reduce. By this time, cognitive dissonance has become one of the most influential and well-researched theories in social psychology [54][55][56].
2.
Secondly, beliefs are not equally important: those that are more personal, closer to the "self", "identity" or "ego", trigger more intense feelings and are more difficult to change [43,46,54,57]. In this sense, beliefs have a hierarchical property: the ones that are higher in rank define or constrain the ones that are lower in rank [58]. For example, the belief or disbelief in God is a central (high-ranking) element in one's belief system, while the belief that "an egg should be boiled for seven minutes in order to get the best soft-boiled egg" is a low ranking one, and, accordingly, can be changed more easily [43]. Furthermore, beliefs that people hold in high regard tend to cause greater dissonance in case of contradiction with other beliefs [54].
3.
Finally, beliefs belonging to the same broader topic (e.g., health, art-related topics, religion, political issues, etc.) are more strongly interrelated than beliefs belonging to different topics. For example, attitudes towards "freedom of speech", "religious freedom" and "freedom to choose spouse" are more closely related than beliefs regarding "freedom of speech" and, say, homeopathic treatments. In mathematical (graphtheoretical) terms, belief systems are "modular" or "compartmentalized" [43,50].
As a first approximation, such a structure can be represented as a modular, hierarchical network (graph) in which the nodes are connected by supportive (positive) or invalidating (negative) relations [48][49][50]59] (Figure 1). For example, according to a widespread belief, "Pathogens can cause diseases". According to another wide-spread belief-primarily in historical societies-"Diseases are caused by evil spirits" [60,61]. These two concepts are in negative relation: somebody believing in one of these opinions will probably disagree with the other. In contrast, the beliefs "Pathogens can cause diseases" and "Contagion is due to the spread of pathogens" support each other (positive relationship), since accepting one of them renders the acceptance of the other more probable.
This last requirement ensures that the level of consistency can be defined [62]. In this description, nodes are beliefs, and edges represent functional relationships between them [52,59,63]. This approach has already been applied by sociologists and economists as well in order to model political belief system dynamics [49,50,59,63,64]. Within this framework, the focus is on the relatedness of beliefs-captured by graph representationwhile the hierarchical and modular characteristics do not gain special importance (see the Supplementary Material for the results incorporating the hierarchical characteristics as well). The graph representation of belief systems. (a) As a first approximation, human belief systems can be represented by networks in which nodes are beliefs ("elements of a belief system" [59]) and edges represent relationships. (b) Links (relationships) can be either supportive (positive) or contradictory (negative) [62].
In these models, a node is "an element of a person's belief system" [59], which the related literature names slightly differently, such as "opinions" [49], "concepts" [62] "attitudes" [49,59,64], "beliefs" [50,59,62], or "positions" [63,64]. In the present article, the terms "belief", "attitude" and "concept" are mostly used, under the condition that we consider all sorts of human thoughts as an "element of a belief system" (that is: a "belief"), whether they be simple or complex, that can be transmitted with the use of language from one person's mind to another's [38,43]. Accordingly, a node can be a thought or attitude towards public issues such as abortion, climate change, immigration, vaccination, gun control; it can be information regarding a public figure or a political party, an idea related to "proper behaviour", "justice"; or the belief that somebody did or said something, etc. Furthermore, this definition implies that beliefs can be transmitted via communication. Communication, in its most basic form, can be discussion (talking) between two or more individuals, but it can also be news/beliefs/information spread by a single agent or organization to many people at the same time, for example via public and social media, news channels, journals, etc. In short, communication is the circulation of news, information and beliefs within a certain community.
The fact that beliefs (attitudes/concepts) are in functional relation with each other (that is, if they are connected, they either support or contradict each other) is crucial, because it implies that some "new" belief will fit into an already existing system-the ones that increase the system's consistency-while others-the ones decreasing the system's consistency-will not [52]. These latter gives rise to the disturbing feeling of cognitive dissonance and people will apply various strategies in order to avoid them. Among other strategies, they will try to keep contact only with those from whom they expect reassuring information (homophily), they will ignore certain information and focus on other information (attentional bias), while greater credence will be given to evidence that fits with the existing beliefs (confirmation bias). These strategies are known as various biases in the field of psychology and sociology [40,[65][66][67]. Different people apply different strategies to various extents; however, to some level, all the strategies are applied by all of us [40,41,65,66].
Furthermore, these cognitive dissonance avoiding mechanisms are in close relation with the proliferation of fake news and the circulation of various types of questionable information as well. In case the well-fitting of a piece of information into the already existing belief system weights more than its credibility, people will adopt it-simply because it provides the pleasant feeling of reassurance. This mechanism is applied in all aspects of life, not only in case of political issues. For example, in the field of economics, it has been observed that managers, whose sales fall short of expectations, rather than rethinking the qualities of the product, tend to identify the cause of the failure elsewhere, for example in the marketing campaign. In such cases, they state that the marketing campaign failed, so it is actually a miracle that the product was sold at all [68]. By finding this new explanation, the sales results show directly the merits of the product, not its failure, and as such, serves as a basis for the pleasant feeling of reassurance. This mechanism is analyzed in Section 4.2.
The Model
We assume a population of N agents. At this point, only their attitudes towards two concepts are important (the attitudes towards, say, vaccination (Concept 1) and, say, a certain public figure (Concept 2)). People can hold any kind of attitudes towards these concepts, from total condemnation (marked by −1) to total support (denoted by +1). Neutrality or indifference is indicated by zero or near-zero values. We are interested in how the agents' attitudes evolve due to being exposed to some news (piece of information) that creates a relation between two originally independent concepts (see Figure 1b).
The relation K 0 can be positive or negative. Using the above example, a trivial positive connection can be that "XY public figure (concept 2) has spoken out in favor of vaccination (concept 1)" (K 0 = +1), while a negative connection can be that "XY public figure has spoken out against it" (K 0 = −1).
In case an agent holds positive attitudes towards both concepts, the positive message (support of vaccination) will give rise to the comforting feeling of reassurance [41]. In this case, the original attitudes are reinforced, since both concepts become better connected and further embedded into the belief system. In contrast, in case the XY trusted and respected politician takes a position against vaccination, a supported matter, the agent will experience cognitive dissonance with an intensity proportional to the original attitude values, and will apply a strategy in order to reduce it [53,54].
Turning back to the basic scenarios, an agent can hold negative attitude towards one of the concepts, and a positive one towards the other-say, for example, a negative attitude towards the public figure and positive attitude towards vaccination. In this case, a negative relation will give rise to reassurance, i.e., "XY politician, whom I anyway hold very low, talked out against vaccination, a cause so important for me ... No surprise here, a fool is known by his conversation" and so on. All scenarios can be analyzed with the same train of thought.
Accordingly, from a mathematical point of view, the cognitive dissonance (or reassurance), C i (t), that agent i will experience at time-step t, can be formulated as: where a i,1 (t) and a i,2 (t) are the original attitudes of agent i towards concept 1 and 2, respectively, at time-step t, and K 0 is the type of connection, which can take two values, +1 or −1, according to the supportive or opposing nature of the connection between the concepts (see also Figure 1b). In case C is positive, it is called reassurance, while in case it is negative, it is usually referred to as cognitive dissonance. Anyhow, in both cases, C denotes the value by which the information alters the coherence or consistency level of agent i's belief system. Note that According to the literature [40,41,65,66], in case of facing information inducing cognitive dissonance, people attempt to relieve the discomfort in different ways, among which the most common ones are: (i) Rejecting new information that conflicts with the already existing ones; (ii) Re-evaluating the attitudes; (iii) A tendency of "explaining things away", that is, finding alternative explanations (developing new beliefs) which supplement the original information in a way that the primordial contradiction is dissolved.
From a modeling point of view, the first strategy-rejecting the information-simply leaves the belief system unaltered. In this case, in the framework of the model, the networknodes, edges and weights-remain unchanged. The second and third strategies do modify the belief system, due to the new connection between the originally unconnected concepts. In the following section, we will focus on modelling these strategies.
Modelling the Re-Evaluation of Beliefs
The constant re-evaluation of our already existing beliefs is an inevitable part of the process of learning and development [32]. New information often comes in the form of creating connection among concepts and beliefs that were originally disconnected. As a matter of fact, this is a basic form of learning. Furthermore, people tend to evaluate most information, beliefs and concepts according to some personal narrative, a personal "frame of mind", which is different from person to person. Simply put, this unique narrative is our personality [37], which defines the very way we perceive the world and make decisions [45,46]. This variety entails individual differences in evaluating the most diverse topics around us, whether it be the judgement of a public figure, a movie or the question of immigration.
In the context of a formal model, the most simple and plausible way to grasp these attitudes is to use numbers between −1 and +1 in a way that negative values represent negative attitudes and positive ones refer to positive stances. The two extreme values, −1 and +1, refer to complete condemnation/approval, respectively.
In order to see how these values might change, consider for example the following case: Paul believes that, say, genetically modified food is harmful. He has already heard it from his friends, and now he reads it in his favorite blog as well. This gives him a feeling of reassurance, due to which he will be convinced about the verity of this belief even more, and will be more attached to his favorite blog as well. In other words, the "embeddedness" of the original attitudes will increase. Mathematically speaking, his already positive attitudes towards these concepts (his belief and the blog) will increase even more due to the positive connection. Now consider a situation where he learns the opposite from his favorite blog, namely that there is nothing at all that could be harmful in genetically modified food (that is, a negative association appears among the two positive concepts: the belief and the blog). In this case, he will experience some level of cognitive dissonance, whose extent depends on his original commitments towards the two concepts [42,53,56]. This experience will make him less convinced, either of the reliability of the blog or of the belief itself-or both. Mathematically speaking, the originally positive values (attached to the two concepts) will decrease somewhat. In other words, cognitive dissonance (negative C i (t) values) decreases the absolute value of the affected attitude (k), while reassurance (positive C values) increases it. Consider the following formula: where a i,k (t) is the original attitude of agent i at time-step t towards attitude k, sign(a i,k (t)) is its signal (+ or −1), ρ is a random value ("noise") taken from the [0, 1] interval with uniform distribution, effecting the extent to which the attitude changes, and C i (t) (defined by Equation (1)), is the level of "coherence" (commonly known as cognitive dissonance, in case it is negative, and reassurance in case it is positive). Finally, the Z A noise comprises the effects of other factors influencing the change of attitudes. It can be either positive or negative with equal probability. In case the updated attitude value a i,k (t + 1) falls outside the predefined [−1, 1] interval, it is set to the nearest threshold (+1 or−1). Attitudes do not vary with the same probability and to the same extent in case of different people and topics; for some, environmental issues are extremely important (and "nothing can change" this stance), some people are detached, while others are convinced that they are just evil-minded hoaxes. The more extreme an attitude is, the more difficult is to change it [43,46,54,57]. (See also Section 2, 2nd bulleted point, "hierarchical property" of belief systems).
Mathematically speaking, the feature "more difficult to change" can be introduced into the model in two ways:
1.
The more extreme an attitude value a is, the lower the probability that it will change. Equation (3) expresses the most simple mathematical formulation of this relation.
2.
The more extreme an attitude value a is, the smaller the magnitude with which it can change.
For the results presented in the main text, the above mentioned hierarchical property was introduced into the model according to the first way, that is, by setting the probability p(AttCh i,k (t)) of attitude-change according to Equation (3), and setting ρ-the parameter controlling the maximal extent with which the attitude values alter due to the experienced cognitive dissonance or reassurance (C)-to 1. In other words, in Equation (2), ρ = 1 for the results presented in the main text. In the Supplementary Material, a detailed analysis is provided on how the parameter ρ effects the simulations (leading to the conclusions that the main claims remain valid, independently of the maximal extent of the alterations, see Figure S7).
Note that in all the equations, the updated attitude values depend only on the agents' previous attitude values (a i,k (t)), the type of the news (K 0 ), and on the cognitive dissonance (or reassurance) values that the news creates in the agents (C i (t)). This means that agents develop their attitudes independently from each other. This originates from the fact that the source of the information does not matter in the present model. Accordingly, if the assumption is that it is the agents who circulate the news among themselves (for example in the form of "gossiping" either in person or on social media), then they interact with each other. In contrast, if the source of the information is something else (for example, state media or some kind of propaganda) then agents do not interact directly with each other. In reality, information usually circulates in both ways. The reason why entire populations are considered is twofold. Firstly, because one single agent cannot "polarize"; they can develop extreme attitudes under certain circumstances. Polarization is an emergent, statistical property of communities, a phenomenon which does not have an interpretation on the level of individuals. The larger the statistics, the more apparent the phenomenon. The second reason is that after studying the elementary process of attitude-update in detail (which is the topic of the present paper), an immediate next step is to study the way by which agents manipulate and organize their social ties (links) assuming similar motivations (avoiding cognitive dissonance and enjoying reassurance).
Modelling the Inclusion of New Beliefs in Order to Relieve Cognitive Dissonance
In case a social actor experiences the upsetting feeling of cognitive dissonance due to a certain piece of information, a commonly applied strategy is to adopt-or create-an even newer belief that changes the context of the original one in a way that it does not serve as a basis of cognitive dissonance any longer; rather, it becomes neutral or even gives rise to the pleasant feeling of reassurance [40,41,68]. An example of this maneuver is mentioned at the end of Section 2, related to managers whose sales data lag behind the expectations tendentiously conceive of various explanations, e.g., ones related to "awfully managed" marketing campaigns. By adopting this new belief (namely that the marketing campaign was awfully managed), the cognitive dissonance caused by the negative sales results (linking a failure to their "self") is eliminated; furthermore, in this light, the sales-results could be seen as an achievement rather than a failure.
The most simple assumption is that the probability p(NB i (t)) of adopting such a new belief (by agent i at time-step t) is proportional to the relief its adoption provides. Since only positive C NB i (t) values represent reassurance, the most simple mathematical formula is the following: where, As shown before, a i,1 (t) and a i,3 (t) are the attitudes of agent i towards concept 1 and the new belief at time-step t, respectively, and K NB is the (positive or negative) connection type between them. Note that in case C NB i (t) is negative-marking cognitive dissonance, instead of reassurance-the agent is highly unlikely to adopt the new belief.
Re-Evaluating Beliefs
Let us consider a population in which the agents' initial attitudes towards two arbitrarily chosen concepts are distributed uniformly, taking values from the [−1, 1] interval. In other words, at the beginning of the simulation, all sorts of attitudes are present in the population with equal probability, from complete condemnation to complete support and everything in between, with an average of zero. Let us now assume that this population is exposed to some kind of news, connecting the two originally unconnected concepts.
Assuming the most general setup, at each time-step t, a randomly chosen agent i acquires the information, and updates his/her attitudes according to Equation (2). (For the flowchart of the algorithm, see Figure A1a. The source of information can be anything, such as public or social media, propaganda, government information, etc. As it can be seen in Figure 2a,b,d,e, proportionally to the level of exposure (iteration number t), the attitudes tend to move towards the two extreme values, +1 and −1, either due to the experienced reassurance or due to the attempt to reduce cognitive dissonance (Equation (2)). The distribution of the attitude values within the population evolves very similarly in case of the two attitudes, since both are governed by Equation (2). (See Figure 2a,b,d,e). At high iteration numbers (indicating strong exposure to the news), around half of the population fully supports Concept 1-marked by attitude values close to +1-while the other crowdcomposed of those whose attitude values are close to −1-fully rejects it. In other words, the population is polarized with respect to Concept 1 (Figure 2a,d). The same applies to Concept 2 (Figure 2b,e). In case the type of connection (K 0 ) is negative (Figure 2 bottom row), the two "stable points"-adopted by the vast majority of the population-are (+1, −1) and (−1, +1), that is, where the two attitudes are reversed, either complete rejection of concept 1 and complete acceptance of concept 2 occurs, or vice versa. These are the two peaks in Figure 2f. In a symmetric manner, in case the connection type, K 0 is positive, the vast majority of the population will either completely support both concepts (one of the peaks will be at (+1, +1)), or will completely reject both of them (the other peak will be at (−1, −1)), as in Figure 2c. That is, independently of the type of connection, the originally uniformly distributed attitudes will tend towards the extremities, meaning that the mere attempt to maintain a consistent belief system alone promotes the processing of attitudes tending towards extremities in case of being exposed to persistent information. Of course, in reality, it is not only one type of news that circulates within a community, but many types, often with different messages and connotations, but it is certainly an important-and so far overlooked-point, that this human drive (the urge to maintain consistent beliefs) alone has the capacity to push attitudes towards extremities-a phenomenon increasingly experienced in our increasingly connected world. (1, 1)) or rejects it (marked by the peak at (−1, −1)). Bottom row (d-f): K 0 = −1 (conflicting relation). (f) The major difference in this case is that at the end of the simulation most agents support one of the beliefs and disagree with the other (marked by the sharp peaks at the (+1, −1) and (−1, +1) points). The parameters are: population size N = 100, number of iterations T = 50,000, and connection type K 0 = −1/ + 1, and the noise value is Z A = 0.01.
In Figure 3c "extremity" is defined as "being closer to +1 or −1 than a certain threshold value ". Accordingly, if = 0.01, then attitudes between 0.99 and 1, and attitudes between −0.99 and −1 will be considered as "extreme". Similarly, if = 0.1, then attitudes between 0.9 and 1, and the ones between −0.9 and −1 will be considered as "extreme". Apparently, as can be seen in Figure 3c, the exact value of does not matter. In this case, since in each time step 1 individual learns the news (out of the N = 1000), on average, each agent will have heard it 10 times at the end of the simulation. As can be seen, for such a level of exposure, developing a neutral standpoint (adopting attitude values close to zero) is a good "strategy" as well. However, this neutrality vanishes in case of more enduring circulation of the news. (c) Proportional to the level of exposure (iteration number t), the ratio of the population holding "extreme" attitudes monotonically grows, independently of how "extremity" is defined (by the parameter ). The parameters are: population size N = 1000, Number of iterations T = 150,000 (except for sub-figure (b), on which T = 10,000), connection type between the concepts K 0 = −1 and the noise value is Z A = 0.01.
Note the small peaks around near-zero values in Figure 2a,b,d,e, at small t values. According to the simulations, in case of limited exposure to the news, agents might also adopt neutral standpoints (marked by near-zero attitude values) in order to avoid cognitive dissonance. This phenomenon is highlighted in Figure 3b. However, this is an unstable equilibrium point, since any further information regarding the given concept (appearing as noise Z A in Equation (2)) pushes the attitude value away from zero. (See also Supplementary Information, Figure S3).
Finding Relief in New Ideas
As has already been mentioned, the other "basic strategy" applied by people in order to reduce the unpleasant feeling of cognitive dissonance is to reinterpret the incoming information by placing it into a context in which the contradiction vanishes, or even better, serves as a basis for reassurance [53]. For example, a doctor in his blog recollected memorable moments of the first year of the COVID-19 pandemic [69]. He remembers that when he tried to convince his family members to take the vaccine, he received vehement rejection, which was settled by receiving the comment that "You have good intentions, we know it. But you do not see the reality, because the "big players" leave you out from the party". As it turned out, by this they meant that the "big players" know perfectly well that the pandemic is a hoax, but they use the everyday doctors-such as the one writing the blog-for their purposes, i.e., to force "everyday people" into take the unnecessary and harmful vaccine. In this example, the doctor is a positive concept in the eye of his relatives, but the epidemic is negative (considered to be a hoax). When it turned out that the doctor considered the epidemic real (hence they should take the vaccine), he created a positive (supportive) relation between himself and the pandemic. This resulted in cognitive dissonance in the relatives, which was dissolved by adopting the new belief (about the "party" of the "big players"), which allowed the original attitudes to remain unchanged.
In the context of the present framework, this scenario can be represented by supplementing the original graph (including two nodes and an edge between them, as in Figure 1) with a new node, representing the new belief (see Figure 4a. The new belief can be related to either of the original concepts, or to both of them. As an example, in Figure 4a, the new belief is connected to Concept 1. The type of connection, K NB , can be either supportive or contradictory, similarly to the connection relating the two original concepts, K 0 . The "stable configuration" toward which the dynamics tends to (after 150,000 simulation steps). (c) The ratio of individuals holding "extreme beliefs", and adopting the new belief (black semi-dotted line). "Extreme attitudes" are those closer to +1 or −1 than a certain threshold value , such as 0.01, 0.05 and 0.1. As it can be seen, the ratio is largely independent of the exact value of . The parameters are: N = 1000, T = 150,000, K 0 = −1, K NB = −1 and Z A = 0.01. Figure 4b depicts the "stable configuration" which the dynamics tends towards. As has been shown already, in the case of K 0 = −1, the attitudes towards concept 1 and 2 tend to be antagonistic and extreme (marked by the attitude values accumulating in the (−1, +1) and (+1, −1) points on the x − y plain), while in the case of K 0 = +1 (positive relation), the attitudes towards concept 1 and 2 tend to be coincidental and also extreme (marked by the attitude values accumulating in the (+1, +1) and (−1, −1) points on the x − y plain, see Supplementary Figure S5). The vertical, z axis depicts the attitude values towards the new, cognitive-dissonance-relieving belief; those who adopt it tend to develop an extreme relation towards this belief as well (in case of unceasing exposure). In contrast, those for whom the approval of the new belief would create cognitive dissonance, simply reject its adoption. In terms of the model, in their case, the edge K NB will simply not exist, and hence the node representing this belief will not be connected to the belief network. The two solid columns belong to these agents, depicting their original attitudes, which simply do not change throughout the simulation. (In case we stipulate that only those values are shown in the figure which participate in the belief system of an agent, these values could be omitted as well, but for sake of clarity, in Figure 4b, we have kept them.) Furthermore, a small vertical "cloud" can be seen in the middle of the chart, representing those who are neutral towards the original concepts, i.e., their attitude towards the new (cognitive-dissonance-relieving) belief can take any value. Importantly, as is apparent from Figure 4, this mechanism pushes the attitudes towards extremities as well.
Discussion
The detailed methods by which humans perceive and make sense of the worlddespite some eminent achievements [33,34,38,39]-is still to be understood. However, some basic characteristics have been elucidated by now, and have become part of mainstream science as well [32,35,36]. One such characteristic is that in the human mind, beliefs are strongly interconnected, and as such, no belief, concept or "piece of information" can exist on its own. Furthermore, in case of new information, humans immediately attempt to interlock it in a coherent way, seeking for connections and support with already existing beliefs.
Accordingly, the novelty of the present model lies not in "assuming" the abovementioned two human characteristics-since they are well-studied and widely accepted by main-stream science [43,53,55]-rather, it lies in their mathematical formulation and incorporation into agent-based models.
There are two more further points worthy of consideration related to the model: (i) Real belief systems have a tremendous amount of elements (instead of two or three), that are interconnected and embedded into each other in a complicated manner [39,43], and, accordingly, the "optimization process"-the attempt to minimize the contradictions among the components-refers to the entire system. From a physicist's point of view, this process is in close relation to physical structures aiming to reach an energy minimum. In this approach, "different realities" [57] can be different local energy minimums of similar systems. However, it is imperative to understand the elementary relation between two elements of the system before considering the entire structure. The present manuscript focuses on this elementary relation. Graph representation is important because, and only because, it serves as a mathematical tool for handling interrelated entities (which are the "beliefs" or "concepts" in our case). Since in the human mind a vast amount of concepts and beliefs are interrelated densely and intricately, any of its graph representations must also assume a vast amount of intricately interrelated (linked) nodes. However, from the viewpoint of the present study, the specific type of the graph does not play any role, because we focus on the elementary process altering the characteristics of two nodes (namely the "attitude values") due to a newly appearing link between them. (If a link appears, it is due to a certain piece of information connecting the two, originally unconnected beliefs/concepts). The nodes whose values alter are selected by the link (representing a piece of information). (ii) The present model does not assume that the repeated information is exactly the same, only that the type of connection between two concepts (say a political party and a public issue, such as immigration or environmental topics) is tenaciously either positive or negative. Hence, it also explains how attitudes can become extreme due to the continuous repetition of information, and as such, it serves as a complementary explanation [70] for the reason why, throughout history, the most diverse regimes found it useful to repeat the same messages over and over again (despite the fact that everybody had already heard them many times).
Furthermore, the present model has some additional results as well, which are yet to be studied. Specifically, according to the results, the attitudes a certain type of news or information triggers depend on the intensity of the exposure. More precisely, in case of limited exposure, people tend to develop a centralist attitude first (which is an unstable equilibrium point), which, in case of persistent news-circulation, give way to extreme stances. The dynamics under limited exposure to news was not studied extensively in the present manuscript.
The ambition of the paper was to call attention to certain human traits that have not yet been incorporated into current computational models aiming to simulate opinion dynamics in human communities. From this perspective, the main point is the naturalness by which polarization can emerge, despite the fact that the model incorporates only minimal assumptions which are considered to be part of well-established, main-stream scientific results.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/e24091320/s1. Figure S1: Histogram of the attitude values, along with their standard deviations, for four time-steps: t = 0, 1000, 5000 and 30,000. Figure S2: Simulation results for N = 100 agents, that is, for one scale smaller population than the one studied in the main text. Figure S3: The evolution of attitudes within a population, due to limited exposure to news, achieved by limiting the number of iterations in T = 3000. Figure S4: Simulation results with four different noise values. Figure S5: The effect of the choice of K 0 , the type of connection. Figure S6: Simulation result with four different initial attitude value distributions. Figure S7: Incorporating the hierarchical nature of beliefs.
Funding:
The research was partially supported by the Hungarian National Research, Development and Innovation Office (grant no. K 128780).
Institutional Review Board Statement: Not applicable.
Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.
Appendix A. Flowchart and Parameters
The simulation was written in Python. Figure A1 shows the flowchart of the algorithm which is enough to replicate the results. However, the source code of the simulation can be found on CoMSES, a public computational model library as well [71]. Figure A1a shows the algorithm for the results explained in Section 4.1, and Figure A1b depicts the algorithm for the case when people might accept a new, cognitive-dissonance-relieving belief as well, the case explained in Section 4.2. Figure A1. Flowchart of the algorithms. (a) is the algorithm of the case discussed in the Section 4.1. (b) depicts the algorithm detailed in the Section 4.2. The main difference-highlighted with purple-is that in the latter case, agents might adopt a new belief as well, in case it decreases their cognitive dissonance.
Since this is a minimalist model, altogether 5 parameters were used, which are overviewed in Table A1. The robustness of the results have been validated for a wide range of parameters, for which the results can be found in the Supplementary Material. Table A1. Summary of the model parameters. Left column: Nomination used in the manuscript. Middle column: description, and right column: values used in the simulation.
Notation
Meaning Values K O Positive or negative: Type of the connection between the two original concepts +1 or −1 K NB Type of the connection between the newly accepted concept and the one it is connected to. +1 or −1 | 2022-05-17T01:16:18.257Z | 2022-05-16T00:00:00.000 | {
"year": 2022,
"sha1": "b16e992440a77937699c48d6a9ae20f84fa2aa8d",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1099-4300/24/9/1320/pdf?version=1663597258",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "3dfbf876a113d6c68da64335ea9aebe92399c182",
"s2fieldsofstudy": [
"Sociology",
"Psychology"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine",
"Physics"
]
} |
86216429 | pes2o/s2orc | v3-fos-license | Transference of microsatellite markers from Eucalyptus spp to Acca sellowiana and the successful use of this technique in genetic characterization
The pineapple guava (Acca sellowiana), known in portuguese as the goiabeira-serrana or “Feijoa”, is a native fruit tree from southern Brazil and northern Uruguay that has commercial potential due to the quality and unique flavor of its fruits. Knowledge of genetic variability is an important tool in various steps of a breeding program, which can be facilitated by the use of molecular markers. The conservation of repeated sequences among related species permits the transferability of microsatellite markers from Eucalyptus spp. to A. sellowiana for testing. We used primers developed for Eucalyptus to characterize A. sellowiana accessions. Out of 404 primers tested, 180 amplified visible products and 38 were polymorphic. A total of 48 alleles were detected with ten Eucalyptus primer pairs against DNA from 119 A. sellowiana accessions. The mean expected heterozygosity among accessions was 0.64 and the mean observed heterozygosity 0.55. A high level of genetic diversity was also observed in the dendrogram, where the degree of genetic dissimilarity ranged from 0 to 65% among the 119 genotypes tested. This study demonstrates the possibility of transferring microsatellite markers between species of different genera in addition to evaluating the extent of genetic variability among plant accessions.
Introduction
The pineapple guava (Acca sellowiana, synonym Feijoa sellowiana), known in portuguese as the goiabeiraserrana or "Feijoa", is a native of the Brazilian southern plateau with secondary dispersion in Uruguay (Mattos, 1990;Thorp and Bieleski, 2002).Due to the uniqueness of its flavor, the economic importance of the pineapple guava is steadily increasing on the world market (Thorp and Bieleski, 2002) and it is an attractive commercial alternative for farmers in southern Brazil (Mattos, 1990).
Although the pineapple guava can be found on the European market or in the countries in which adapted cultivars are active (e.g.New Zealand, Colombia and the USA) as yet there are no improved cultivars in Brazil, its greatest center of diversity.However, there is an A. sellowiana Active Germplasm Bank (AGB) located at the São Joaquim Experimental Station (Estação Experimental de São Joaquim (EPAGRI), São Joaquim-SC, Brazil) in the town of São Joaquim in the Brazilian state of Santa Catarina.This germplasm bank contains 119 A. sellowiana accessions from several regions of Brazil and other countries, and it is possible to use directly an accession as a clone or to develop a cultivar by means of genetic breeding methods in order to scale up commercial production The genetic variability of this species is normally high at the center of origin, and information on such variability is essential for A. sellowiana conservation, breeding and commercial production.In general, specific phenotypes of discreet variation are used as morphological markers.However, a limited number of morphological markers have been identified for this species (Nodari et al., 1997), which are frequently affected by dominance and epistatic gene interactions, environmental effects and pleiotropy.To overcome such problems, molecular markers can be used to help genetic characterization and breeding (Nodari et al., 1997;Brondani et al., 1998Brondani et al., , 1997)).
Among the classes of molecular markers available to identify variation at DNA level, the microsatellites, or simple sequence repeats (SSRs), are considered ideal markers for genetic studies because they combine several suitable features: (i) co-dominance; (ii) multiallelism; (iii) high polymorphism, allowing precise discrimination even of closely related individuals; (iv) abundance and uniform dispersion in plant genomes; and (v) the possibility of efficient analysis by a rapid and simple polymerase chain reaction (PCR) assay (Morgante and Olivieri, 1993;Rafalski and Tingey, 1993;Sharma et al., 1995;Brondani et al., 1998).In addition, for the amplification of microsatellite loci, a knowledge of their DNA sequence is required, and this is an expensive and time consuming process (Zucchi et al., 2003).However, the approach of using enriched libraries with repetitive sequences has been very successful in developing SSRs at a reasonable cost (Zane et al., 2002).
The ability to use the same microsatellite primers in different plant species, called transferability, depends on the extent of sequence conservation in the primer sites flanking the microsatellite loci and the stability of those sequences during evolution (Choumane et al., 2000;Decroocq et al., 2003;Zucchi et al., 2003).It has been shown that closely related species are more likely to share microsatellite priming sites than more distantly related ones, but it is possible to transfer functional microsatellite primers even from more distantly related species (Lorieux et al., 2000).
Because there are no microsatellites available for A. sellowiana, the Eucalyptus spp.primers of microsatellite loci (Brondani et al., 1998) can be used as an alternative to find similar regions on the A. sellowiana genome, since they belong to the same family (Zucchi et al., 2003).
Thus, the objectives of the work described in this paper were to evaluate the transferability of microsatellite markers from Eucalyptus to A. sellowiana (both members of the Myrtaceae) and to characterize the genetic variability present in the Active Germplasm Bank (AGB) of this species.
Genetic material
The 119 accessions tested shown in Table 1 were obtained from the pineapple guava Active Germplasm Bank (AGB) located at the São Joaquim Experimental Station (Estação Experimental de São Joaquim -EPAGRI, São Joaquim-SC, Brazil).Most of the accessions came from the Brazilian state of Santa Catarina, although a few accessions came from other countries (Table 1).Samples of DNA were obtained following the protocol developed by Doyle and Doyle (1987).The extracted DNA was quantified in agarose gel (Sambrook et al., 1989) and diluted to 3 ng μL -1 for further use in the amplification reactions.Leaf DNA from Eucalyptus grandis was used as a control.
The screening of the pairs of primers was done in two steps.The first step employed the DNA of two A. sellowiana plants and one control E. grandis plant, the amplification products being visualized on 1.5% (w/v) agarose gel.In the second step, the pairs of primers showing positive amplification were confronted with an additional group of eight A. sellowiana genotypes to detect polymorphism and 10 selected primers were then utilized to analyze the genetic variability in the 119 A. sellowiana accessions, the amplification products being separated on 6% (w/v) denaturing polyacrylamide gel.A 100 bp DNA ladder was used as a molecular weight reference to estimate the sizes of the amplification products.The gels were stained with silver nitrate, as described by Creste et al. (2001).
Data analysis
The genetic diversity characterization potential of the primers was based on allele frequency estimates of the mean observed heterozygosity (Ho), mean expected heterozygosity (He) (Nei, 1978) and the number of alleles per locus for the AGB accessions.These estimates were obtained using the BIOSYS-1 program (Swofford and Selander, 1989).In addition, a dendrogram was plotted from an unbiased genetic similarity matrix (Nei, 1978) grouped by the unweighted pair group method with arithmetic mean (UPGMA) of Sneath and Sokal (1973).
Results
Of the 404 primer pairs tested we found that 180 (44.5%) amplified visible products in A. sellowiana.Furthermore, 38 (9.4%) primer pairs allowed the detection of clear bands and easy fragment recognition for eight A. sellowiana genotypes, generating an average of 1.5 alleles per locus.Satisfactory amplification products were obtained using an annealing temperature of 56 °C.
When we screened the 119 accessions with ten selected polymorphic primer pairs we detected 49 alleles, varying from 120 bp to 320 bp.The quality of the amplification products is shown in Figure 1.
The number of detected alleles per locus ranged from 2 to 9, averaging 4.9 alleles per locus, with the EMBRA 26 marker being the most polymorphic one (Table 2).The ten pairs of primers used were able to detect low frequency alleles (≤ 0.05), which were distributed in accessions of different origins (Table 3).The mean expected heterozygosity among loci was He = 0.640, while the mean observed heterozygosity was Ho = 0.551 (Table 2).In the dendrogram (Figure 2), the 119 accessions were distributed in different groups and the degree of genetic similarity ranged from 35% to 100%.Two sub-groups contained two accessions with 100% similarity, one sub-group consisting of accessions 228 and 331 from Lages and the other sub-group consisting of the New Zealand accessions 456 and 457.On the other hand, three accessions formed two sub-groups, one of which contained accession 80 from Curitibanos (37.5% similarity with the other 117 accessions) and the other accession 247 from Lages plus accession 501 from Fraiburgo (35% similarity with the other 117 accessions), these two sub groups being very different from the other sub-groups.Besides this main feature of grouping, the other sub-groups did not reveal any special structure, except one, which included 7 out of 9 accessions from outside Brazil.Interestingly, the two genotypes from Israel were located outside the group that included accessions from the USA, New Zealand and Brazil.
Discussion
Considering the time-consuming and expensive process of microsatellites isolation (Powell et al., 1996), we took advantage of the availability of Eucalyptus primer sequences and used them in Acca sellowiana.Our study demonstrated the transferability of microsatellite markers from Eucalyptus to Acca across different genera belonging to the same family (Myrtaceae).This demonstration of transferability means that future genetic studies can be carried out, this marker type being extremely useful due to its ease of use and high amount of information generated.Because of these features, microsatellites are considered as useful molecular markers in plant breeding, and are widely used for cultivar fingerprinting, paternity testing and genome mapping.
However, a cautious approach is required when comparing similar PCR products obtained across different species, since various factors can cause size homoplasy.Over long periods of evolution, the interspecific allelic differences at one locus are often more complex than simple changes in repeat number.Products amplified in different species might include mutation, rearrangements and duplications in the flanking region and/or changes in the number of repeats (Peakall et al., 1998).
Microsatellite transferability has also been confirmed to occur between species from different genera.In the work described in this paper we have demonstrated that primer pairs developed for Eucalyptus were able to amplify in the A. sellowiana genome.Zucchi et al. (2003) used a sample from the same set of Eucalyptus complex primer pairs to test their transferability to other Myrtaceae species such as Eugenia dysenteria and found that of the 356 microsatellite primer pairs tested it was possible to transfer 10, representing a transferability of 2.8%.Interestingly, none of the microsatellite primer pairs transferred to Eugenia dysenteria by Zucchi et al. (2003) coincided with the primers transferred to A. sellowiana in our study.Although it is premature to make inferences about relatedness before obtaining further evidence, by considering these results together it can be hypothesized that there is more similarity between Acca and Eucalyptus than between Eugenia and Eucalyptus.According to Palop et al. (2000), microsatellites loci are more likely to be amplified in closely related species.
The number of primer pairs transferred from Eucalyptus to A. sellowiana (44.5%) can be considered very high in comparison to other studies (Padian et al., 2000).In addition, at least 26% of the primer pairs transferred were able to detect polymorphism among the 119 A. sellowiana genotypes screened, which can be considered to be a high level in light of the fact that in the study by Zucchi et al. (2003) cited above with 10 pairs of Eucalyptus primers transferred to E. dysenteria the level of polymorphism was only 2%.
It is relevant to mention the existence of a large amount of genetic variability among the A. sellowiana accessions.Besides supporting such a conclusion, the presence of alleles with a low frequency in accessions of different origins suggests that the genetic variability is dispersed across locations.Genotypes with a low level of similarity in comparison with others were also found in this study.This high amount of genetic variability is no surprise, since the A. sellowiana germplasm bank contains a collection of 119 representatives from 14 locations distributed across southern Brazil.Most of the accessions have agronomic importance, since they express one or more agronomic traits that can be further integrated into breeding programs.
The high value of expected heterozygosity in comparison with observed heterozygosity among the accessions in the A. sellowiana germplasm bank indicates that heterozygote deficit is present in the germplasm bank accessions.However, this deficit is relatively low and the heterozy-Transference of microsatellite markers from Eucalyptus to Acca sellowiana gosity is high, which suggests the existence of high genetic diversity in the natural populations from which the A. sellowiana accessions were collected.
The dendrogram which we obtained based on 10 loci showed two sub-groups consisting of two accessions each, with 100% of similarity.At the other extreme, three accessions showed a low degree of similarity (from 35.0 to 37.5%) with the other 117 plant accessions.These results agree with the high values of heterozygosity and the 48 alleles detected in the Active Germplasm Bank.It is important to mention that the accessions from New Zealand and the United States were included in a sub-group, indicating the narrowing of the genetic base for these accessions and a substantial degree of relationship among them.
This study has demonstrated the transferability of Eucalyptus spp.microsatellite markers to Acca sellowiana, which belong to the same family (Myrtaceae) but are distinct genera.Because the Eucalyptus microsatellite loci were able to detect the existence of a large amount of genetic variability among the A. sellowiana accessions they can be used for genetic characterization of both accessions and natural populations, knowledge of which helps to accelerate not only the establishment of appropriate conservation strategies but also marker assisted selection, the selection of parents for controlled crosses and the monitoring of the segregation of genomic regions of agronomic interest in segregating progenies.In addition, the Eucalyptus primers can be used for genetic studies in other Myrtaceae species for which there are no species-specific microsatellites available.
Figure 1 -
Figure 1 -Profile of the PCR products of 21 Acca sellowiana accessions from the Active Germoplasm Bank.The products were amplified by the Eucalyptus primer EMBRAPA 123 and resolved on silver stained polyacrylamide gel.Left line: 1 kb Plus DNA Ladder (Invitrogen).
Figure 2 -
Figure 2 -Unweighted pair group method with arithmetic mean (UPGMA) dendrogram reflecting the genetic similarities, based on 10 microsatellite loci, among 119 accessions from the Acca sellowiana Active Germplasm Bank.
Table 2 -
Transference of microsatellite markers from Eucalyptus to Acca sellowiana 75 Sequence of the 10 used pairs of primers developed for Eucalyptus genera, allele size range, number of alleles per locus (A), observed heterozygosity (H o ) and expected heterozygosity (H e ) of amplified microsatellite loci in Acca sellowiana.
Brondani et al. (1998)e reported byBrondani et al. (1998)and were obtained from the Genetics and Biotechnology unit of the Brazilian agricultural company Embrapa (Empresa Brasileira de Pesquisa Agropecuária-Recursos Genéticos e Biotecnologia, Brasilia, DF, Brazil).Santos et al.
Table 3 -
Origin of the plant accessions presenting alleles with a low frequency (f < 0.05).All Brazilian cities are located in the state of Santa Catarina.
Brondani et al. (1998)e reported byBrondani et al. (1998)and were obtained from the Genetics and Biotechnology unit of the Brazilian agricultural company Embrapa (Empresa Brasileira de Pesquisa de Agropecuária-Recursos Genéticos e Biotecnologia, Brasilia, DF, Brazil).Local: Country, city of origin and accession number. | 2019-03-30T13:10:19.206Z | 2007-01-01T00:00:00.000 | {
"year": 2007,
"sha1": "a9a95d1e6b82b1f965dcf046f550477e06234afc",
"oa_license": "CCBY",
"oa_url": "https://www.scielo.br/j/gmb/a/ppTD4LgJ7QtkYKYLRYWmcXw/?format=pdf&lang=en",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "a9a95d1e6b82b1f965dcf046f550477e06234afc",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology"
]
} |
118464922 | pes2o/s2orc | v3-fos-license | Internal One-Particle Density Matrix for Bose-Einstein Condensates with Finite Number of Particles in a Harmonic Potential
Investigations on the internal one-particle density matrix in the case of Bose-Einstein condensates with a finite number ($N$) of particles in a harmonic potential are performed. We solve the eigenvalue problem of the Pethick-Pitaevskii-type internal density matrix and find a fragmented condensate. On the contrary the condensate Jacobi-type internal density matrix gives complete condensation into a single state. The internal one-particle density matrix is, therefore, shown to be different in general for different choices of the internal coordinate system. We propose two physically motivated criteria for the choice of the adequate coordinate systems which give us a unique answer for the internal one-particle density matrix. One criterion is that in the infinite particle number limit ($N=\infty$) the internal one-particle density matrix should have the same eigenvalues and eigenfunctions as those of the corresponding ideal Bose-Einstein condensate in the laboratory frame. The other criterion is that the coordinate of the internal one-particle density matrix should be orthogonal to the remaining $(N - 2)$ internal coordinates. This second criterion is shown to imply the first criterion. It is shown that the internal Jacobi coordinate system satisfies these two criteria while the internal coordinate system adopted by Pethick and Pitaevskii for the construction of the internal one-particle density matrix does not. It is demonstrated that these two criteria uniquely determine the internal one-particle density matrix which is identical to that calculated with the Jacobi coordinates. The relevance of this work concerning $\alpha$-particle condensates in nuclei, as well as bosonic atoms in traps, is pointed out.
I. INTRODUCTION
The second 0 + state of 12 C which is located near the 3α breakup threshold and is called the Hoyle state [1] is well known as one of the mysterious 0 + states in light nuclei. The understanding of its structure has been actually one of the most difficult and challenging problems of nuclear structure. Its small excitation energy of 7.66 MeV is clearly not easy to explain by the shell model and, in fact, even the most advanced modern shell model approach known as the no-core shell model, fails by far to reproduce its position [2].
About 40 years ago Morinaga proposed to assign the 3α linear-chain structure to this Hoyle state [3]. However the observed reduced α decay width of this state which is larger than the Wigner limit value was pointed out to be contradictory to the linear-chain structure which can give at most only one third of the Wigner limit value [4]. The large α-decay reduced width of the Hoyle state in the 8 Be(0 + 1 )+α channel was reproduced by a full threebody calculation with a semi-microscopic α-α interaction [5], namely an OCM ( orthogonality condition model ) [6] calculation for the 3α system. This 3α calculation contradicted the 3α chain structure of the Hoyle state and gave as the dominant component of the Hoyle state the 8 Be(0 + 1 ) + α structure with relative S wave between 8 Be(0 + 1 ) and α. Since 8 Be(0 + 1 ) consists of two α clusters weakly coupled in relative S wave, the Hoyle state was concluded to have a weakly coupled 3α structure in relative S waves with large spatial extent, which is a gas-like structure of α clusters. A few years later, this understanding of the structure of the Hoyle state was reported to be reproduced by full microscopic 3α calculations by two groups, namely Kamimura and his collaborators [7] and Uegaki and his collaborators [8].
These calculations nicely reproduced not only the position of the Hoyle state but also other experimental properties including inelastic electron form factor and E0 and E2 transition properties. Other states of 12 C below and around the Hoyle state were also well described.
The microscopic 3α model treatments of 12 C of Refs. [7,8] were later extended to reaction theory in Ref. [9]. Furthermore investigation by antisymmetrized molecular dynamics [10] and that by fermionic molecular dynamics [11] starting from a realistic bare N-N force, both of which do not assume alpha clustering, have now reproduced all the salient features of the Hoyle state [1]. Let us also mention that for a product state of only a few bosons, the phase fluctuates. However, we use the word 'condensate' in the same 'loose' sense as nuclear physicists are used to say that 'nuclei are superfluid', in spite of the fact that there is not a macroscopic number of Cooper pairs in nucleus.
Almost 30 years after the first proposal of the 8 Be(0 + 1 ) + α structure for the Hoyle state, this state was reconsidered in a new light in Ref. [12] involving four members of the present authors. They presented the following new model wave function Φ B (3α), called the THSR wave function: where φ(α j ) and X j stand for the internal wave function and the center-of-mass coordinate of the j-th α cluster, respectively, and A is the antisymmetrizer of the nucleons. As shown in Eq. (2), the THSR wave function can be regarded as expressing the cluster structure where a 8 Be(0 + 1 )-like cluster A{exp(−(1/B 2 )ξ 2 2 )φ(α 2 )φ(α 3 )} and the α 1 cluster couple via an S-wave inter-cluster wave function exp(−(4/3B 2 )ξ 2 1 ). On the other hand, Equation (1) shows that the THSR wave function represents the state where three α clusters occupy the same single 0S-orbit exp(−(2/B 2 )X 2 ), namely a 3α condensate state which is the finite size counterpart of the macroscopic α-particle condensation in infinite nuclear matter at low density [13].
What the authors of Ref. [12] proposed was that the 8 Be(0 + 1 ) + α structure of the Hoyle state can be regarded as being a 3α condensate state and that one can expect in general the existence of nα condensate states in the vicinity of the nα breakup threshold in 4n selfconjugate nuclei [14]. It was soon discovered [15] that the microscopic 3α wave functions of both of Refs. [7,8] have overlaps of more than 95% with a single THSR wave function with a large size parameter B, implying small overlaps between individual α clusters.
This above-mentioned striking fact reported in Ref. [15] means without doubt that the Hoyle state structure has a strong relation with the α condensation physics in dilute infinite nuclear matter. One of the important tasks for the clarification of this relation is to study the magnitude of the component of the ideal Bose-Einstein condensation of structureless α particles which is contained in the Hoyle state wave function. This was investigated by the authors of Refs. [16,17] who solved the eigenvalue problem of the internal one-particle density matrix of the Hoyle state wave functions where the center-of-mass coordinate is eliminated. As the internal coordinates for calculating the internal one-particle density matrix ρ they used the Jacobi coordinates, namely ξ 1 and ξ 2 in Eq. (3). They obtained a maximum eigenvalue larger than 0.7 for the normalized density matrix ρ ( Tr ρ = 1 ), which means that the corresponding eigenfunction which is a 0S-orbit is occupied to more than 70% by the three α particles. This is a large percentage corroborating the almost ideal αparticle condensation nature of the Hoyle state, that is the Hoyle state is describable to very good approximation by a product state of these bosons, each in the same 0S orbit [16,17].
For the discussion of the Bose-Einstein condensation of a confined macroscopic system, one uses in general the one-particle density matrix in the laboratory frame where the coordinate system consisting of individual particle coordinates is adopted. However, Pethick and Pitaevskii [18] (PP) proposed to use the internal one-particle density matrix by eliminating the center-of-mass coordinate when one discusses the condensate fraction of the system where only the center-of-mass degree of freedom is excited but all relative degrees of freedom are kept unchanged. Their proposal was to point out that the conclusion of the paper by Wilkin et al. [19] does not appropriately reflect the physics. Wilkin et al. discussed the lowest excitation of a condensate of cold bosonic atoms with attractive interactions which rotates with its center of mass in a harmonic potential but keeps all the internal degrees of freedom as in its ground state which shows complete condensation. It was found that the corresponding one-body density matrix in the laboratory frame has many eigenvalues of comparable size, and thus the system should be characterized as a fragmented condensate [19]. On the other hand PP claimed that if one uses an internal one-particle density matrix by eliminating the center-of-mass coordinate it should give a single eigenvalue of the order of the number of particles indicating the non-fragmented condensate character of the internal part of the system. Unfortunately PP presented only their idea and did not demonstrate explicit results of the diagonalization of their definition of the internal one-particle density matrix.
In this paper, following the definition of PP, we construct the internal one-particle density matrix of a many-boson system in a harmonic trap and then give the explicit analytical form of the eigenvalues and eigenfunctions of this density matrix. We will see that the eigenvalues are fragmented, which is contrary to PP's initial objective. We will discuss that this result does not mean that the original idea of PP is incorrect but it means that the internal coordinate system which PP adopted is inadequate. Actually when we construct the internal one-particle density matrix by using the internal Jacobi coordinates, the resulting eigenvalues show complete condensation. We will give two criteria for the choice of the adequate coordinate systems which will give us a unique answer for the definition of the internal one-particle density matrix. One criterion is that in the infinite particle number limit ( N = ∞ ) the internal one-particle density matrix should have the same eigenvalues and eigenfunctions as those of the corresponding ideal Bose-Einstein condensate in the laboratory frame. This is in line with the general wisdom that in the thermodynamic limit all macroscopic quantities shall be the same, regardless whether considered in the laboratory frame or in the internal frame [20]. The other criterion is that the coordinate used in the internal one-particle density matrix should be orthogonal to (or maximally independent from) the remaining (N − 2) internal coordinates, though the (N − 2) coordinates, in general, do not need to be mutually orthogonal. This second criterion is shown to imply the first criterion. We show that the internal Jacobi coordinate system satisfies these two criteria, while the internal coordinate system adopted by PP for the construction of the internal one-particle density matrix does not. Furthermore we argue that these two criteria uniquely determine the internal one-particle density matrix which is the same as that calculated with the Jacobi coordinates. The results of this paper justify the use of the Jacobi coordinates in Refs. [16,17] and, hence, corroborate the almost ideal α-particle condensation nature of the Hoyle state.
The present paper is organized as follows. In Sec. II, we formulate the internal one-particle density matrix with respect to the internal wave function of a Bose-Einstein condensate with finite particle number in a harmonic trap. Then, the analytical form of the eigenvalues and eigenfunctions is presented for the internal density matrix and numerical eigenvalues are discussed. In Sec. III. we propose two criteria for the choice of proper coordinate systems for internal one-particle density matrix. Finally, the summary is given in Sec. IV. Appendix A serves to present the analytical solution of the eigenvalue problem of the density matrix and Appendix B is given for the explanation of the bosonic symmetry of the Jacobi-type internal one-particle density matrix. The original version of this paper was given in an article on the arXiv [21] and a short version of this paper is reported in a letter paper [22] with special attention to the cold atom community.
II. INTERNAL ONE-PARTICLE DENSITY MATRIX
First we consider the one-particle density matrix in the laboratory frame for an ideal Bose-Einstein condensate with N spinless bosons in a harmonic potential. The result is trivial but instructive for studying the nature of the internal one-particle density matrix, as will be discussed later.
The N-particle Hamiltonian in laboratory frame is presented as The normalized ground-state wave function of this system is expressed as a product of identical Gaussian single-particle wave functions, i.e.
where ν = mω/2h, and {r i } N i=1 denotes the set of the coordinates r i (i = 1 · · · , N). The one particle density matrix in the laboratory frame is defined as ρ (1) Lab (r, r ′ ) = dr 2 dr 3 · · · dr N Φ * (r, r 2 , r 3 , · · · , r N )Φ(r ′ , r 2 , r 3 , · · · , r N ), It is noted that the density matrix is independent of the number of particles N and is separable with respect to r and r ′ . The separability originates from the fact that the Hamiltonian is separable with respect to different particle operators, r i and p i (i = 1, · · · , N), in Eq. (4), or equivalently from the fact that the wave function of Eq. (6) is separable with respect to different particle coordinates, r i (i = 1, · · · , N).
The nature of the single particle orbits and their occupation probabilities in the relevant system can be obtained by solving the eigenvalue problem of the density matrix, where φ(r) is the single particle orbit, and λ is its occupation probability. This equation can easily be solved, and we find that the density matrix has only one non zero eigenvalue λ = 1 with one eigenfunction, namely, the zero-node S-wave Gaussian φ(r) = (2ν/π) 3/4 exp(−νr 2 ) (or 0S harmonic oscillator wave function φ 000 (r, ν), which will be defined later) with 100 % occupancy (λ = 1). This means that all particles are condensed in that single orbit, i.e.
an ideal Bose-Einstein condensation is realized in the laboratory frame. This feature is independent of the number of particles N.
Next we consider the internal one-particle density matrix for the N-particle Bose-Einstein condensation in a harmonic trap described by the wave function Eq. (6) with the total Hamiltonian Eq. (4). Internal means that the density is free from the center-of-mass coordinate of the system, as it must be the case, when dealing with, e.g., selfbound systems. In the present paper, two kinds of internal coordinate sets are introduced; 1) coordinates with respect to the center of mass of the total system and 2) Jacobi coordinates. The former set was first considered by Pethick and Pitaevskii [18] to define the internal one-particle density matrix. We call it Pethick-Pitaevskii-type (PP-type) internal one-particle density matrix in the present paper. For the latter set, we call it Jacobi-type density matrix.
A. Pethick-Pitaevskii-type internal one-particle density matrix As already mentioned in the introduction, Pethick and Pitaevskii proposed to consider the internal single particle density matrix for a Bose condensed system when only the center-ofmass but no internal degree of freedom is excited. In order to define an internal one-particle density matrix, Pethick and Pitaevskii adopted internal coordinates defined with respect to the center of mass of the total system [18]. The center-of-mass coordinate R and the coordinate q i of particle i relative to the center of mass are given, respectively, by Hereafter, as the independent internal coordinates, we adopt q i (i = 1, · · · , N − 1). We introduce the conjugate momenta π i (i = 1, · · · , N − 1) and P for the coordinates q i (i = 1, · · · , N − 1) and R, respectively.
Then, the total Hamiltonian in Eq. (4) is rewritten as where H int and H cm denote the internal and center-of-mass Hamiltonians, respectively. It should be mentioned that the above Hamiltonian includes the cross terms, q i ·q i ′ and π i ·π i ′ , which, in general, appear so far as one takes non-orthonormal coordinate systems [23][24][25][26].
By using the relation the total wave function in Eq. (6) is expressed as where Φ int and Φ cm denote the internal and center-of-mass wave functions, respectively.
The separability of Φ int and Φ cm comes from the fact that the total Hamiltonian is separable with respect to the internal and center-of-mass parts [see Eq. (11)] or directly from Eq. (14) inserted into Eq. (6). The factor 1/N 3/2 in Eq. (15) drops out when taking into account , coming from the coordinate transformation from the space-fixed system to the center-of-mass system. It should be noted that both wave functions Φ int and Φ cm are normalized and satisfy H int Φ int = (3/2)(N − 1)hωΦ int and H cm Φ cm = (3/2)hωΦ cm , respectively.
According to Pethick and Pitaevskii [18], the internal one-particle density matrix is defined as, where the Jacobian factor N 3 is inserted in the above equation, which is missing in Ref. [18].
Using the wave function Φ({r i } N i=1 ) in Eq. (6), the internal one-particle density matrix (18) is expressed as It is noted that this density matrix depends on the number of particles N and contains the cross term q · q ′ . The origin of the cross term comes from the nonseparability of the coordinate q 1 from the other (N − 2) coordinates q i (i = 2, · · · , N − 1) in the internal wave function of Eq. (16), or equivalently from the fact that the internal Hamiltonian of Eq. (12) is non-separable between the first particle operators π 1 and q 1 and the other (N −2) particle operators π i and q i (i = 2, · · · , N − 1). The result of Eq. (20) can also be verified with the use of the internal density matrix in Eq. (19) presented as follows: is the internal wave function defined in Eq. (16). Let us discuss the nature of the internal one-particle density ρ (1) int (q, q ′ ). First we study the single-particle orbits and their eigenvalues obtained by solving the eigenvalue problem for the density matrix, This equation can be solved analytically. There are several ways to solve it. In Appendix A we give one method. We can also analyze the description given in Ref. [27]. The singleparticle orbits φ are expressed by the harmonic oscillator wave functions φ nLM (q, β N ) with the orbital angular momentum L, magnetic quantum number M and harmonic oscillator quanta Q = 2n + L (n = 0, 1, 2, · · ·), The eigenvalues or occupation probabilities λ are given as and satisfy the following completeness relation, Then, the internal one-particle density matrix for the N particles in Eq. (20) can be presented in terms of the wave functions (23) The occupation probability with respect to the partial wave with quantum number L is defined as In the macroscopic limit (N → ∞), the internal one-particle density matrix ρ (1) int,PP,N =∞ , its eigenfunctions φ int,PP,N =∞ , and eigenvalues λ (LM ) n,N =∞ are given by We remark that the summed eigenvalues Λ = 0.002 for F , and so on. Increasing the particle number, the S-wave occupation probability is decreasing, while the higher partialwave ones are increasing (see Fig. 1 in Refs. [21,22]). These results show that the PP-type one-particle density matrix leads to a fragmented condensate. The reason why the PP-type internal density matrix shows the fragmented condensate is due to the existence of the cross term q · q ′ in Eq. (20).
B. Jacobi-type internal one-particle density matrix For the N-particle system, we define the (N − 1) internal Jacobi coordinates {ξ i , (i = 1, · · · , N − 1)} and the center-of-mass coordinate R as follows: where ξ 1 denotes the relative coordinate between the first particle and the remaining (N −1) particles, and other Jacobi coordinates are self-evident. Then, the N-particle Hamiltonian in Eq. (4) can be separated into the internal and center-of-mass Hamiltonian, where π i and P denote the conjugate momenta corresponding to the coordinates ξ i and R, respectively.
Since the total Hamiltonian is a sum of decoupled N harmonic oscillator Hamiltonians ( for R and (N − 1) internal Jacobi coordinates ), the total wave function in Eq. (6) is This expression of Φ({r i } N i=1 ) can also be derived by inserting the relation into the original expression given in Eq. (6). The total wave function is the product of the internal and center-of-mass wave functions, Φ int and Φ cm , which are normalized and satisfy the relations, H int Φ int = (3/2)(N −1)hωΦ int and H cm Φ cm = (3/2)hωΦ cm . It is noted that the internal wave function is given as a product of harmonic oscillator wave functions φ 000 (ξ i , ν i ) (i = 1, · · · , N − 1).
The Jacobi-type internal density matrix writes Then, the Jacobi-type one-particle density matrix is defined with respect to ξ 1 and ξ ′ 1 as This choice of the coordinate ξ 1 for the internal density matrix is natural, because the single particle orbit should be defined with respect to the relative coordinate between one particle and the other remaining N − 1 particles in the Jacobi coordinate system. The compatibility between the bosonic symmetry of the system and the above definition of the internal one-particle density matrix, where it may seem that one special coordinate is singled out, is explained in Appendix B. The eigenvalue equation of the one-particle density matrix can easily be solved analytically. We find that the density matrix has only one nonzero eigenvalue λ = 1 with corresponding eigenfunction φ int,J,N which is the 0S harmonic oscillator wave function with 100 % occupancy (λ = 1), This means that all particles are condensed in the single 0S particle state, although the size parameter in the state [Eq. (46)] depends on N and is slightly different from that in the eigenfunction φ 000 (r, ν) in laboratory frame, discussed in the beginning of Sec. II. In the macroscopic limit (N → ∞), the internal one-particle density matrix ρ (1) int,PP,N =∞ , its eigenfunction φ int,PP,N =∞ , and eigenvalues λ (LM ) n,N =∞ are given by The Jacobi-type one-particle density matrix for finite particle number is separable with respect to the coordinates ξ and ξ ′ . This leads to the only one non-zero eigenvalue λ = 1 which is the same as for an ideal Bose-Einstein condensate system in the laboratory system. Also the internal density matrix (49) with respect to different Jacobi coordinates. This feature has it origin in the fact that the Jacobi coordinates form an orthogonal coordinate system.
A. Convergence to Bose-Einstein Condensation in the Macroscopic Limit
In the previous section we learned that the outcome of the diagonalization of the internal density matrix depends on the choice of the internal coordinates. This is a serious problem for treating condensates in internal self-bound systems such as α-particle condensates in nuclear systems or small droplets of superfluid 4 He, because only internal degrees of freedom are relevant in these systems.
In order to overcome the difficulty, we gave a criterion for the choice of the internal coordinates: In the macroscopic limit (N → ∞) the internal density matrix should have the same eigenvalues and eigenfunctions as those of the ideal Bose-Einstein condensate in the laboratory frame. This is a very physical boundary condition. For understanding this physical condition, consideration of the following situation in the macroscopic limit may be helpful: In the laboratory frame, the center-of-mass motion of the present system is described by the wave function Φ cm (R) of Eq. (17). In the macroscopic limit, the centerof-mass coordinate should be at the coordinate origin of the laboratory frame, because the probability of finding the center-of-mass coordinate at position R is given by |Φ cm (R)| 2 = (2Nν/π) 3/2 exp(−2NνR 2 ) which becomes a delta function δ(R) in the limit of N → ∞.
Thus the internal coordinate q i should have the same meaning as the position of the i-th particle coordinate r i in the laboratory frame. Also the first Jacobi internal coordinate ξ 1 should have the same meaning as the position of the first particle coordinate r 1 because As shown in Eqs. (30) and (31), the PP-type internal density matrix in the macroscopic limit does not satisfy the condition, while the density matrix of the Jacobi-type fulfills this condition as given in Eqs. (50) and (51). As already mentioned, the reason why the PP-type internal density matrix does not satisfy the physical boundary condition is due to the fact that it exhibits a nonvanishing cross term or correlation term q · q ′ in the case of N = ∞, originating from the pseudo two-body interaction terms in the internal Hamiltonian. On the other hand, in the Jacobi-type internal density matrix, the pseudo two-body interaction terms disappear.
These results mean that one should take internal coordinates which do not produce any correlation in the internal one-particle density matrix in the macroscopic limit. Otherwise, unphysical situations occur like for the PP-type internal density matrix. One choice fulfilling the physical condition is the one of the internal Jacobi coordinates. Of course, there are many sets of internal coordinates which satisfy the physical condition. For example, they may be those satisfying the following two conditions: (1) The relative coordinate ξ between one particle and the other remaining particles should be used as the coordinate of the internal one-particle density matrix ρ (1) int , and (2) the internal one-particle density matrix be separable with respect to ξ and ξ ′ in the macroscopic limit, ρ Among the coordinates satisfying the two conditions, Jacobi coordinates are very convenient and useful to describe the internal Hamiltonian and thus the internal one-particle density matrix. We will see in the next section that this is in fact the only choice. In the previous section, a criterion was discussed for the choice of the adequate coordinate systems of the internal density matrix. Here we present the other criterion that the coordinate of the internal one-body density matrix, ξ 1 , should be orthogonal to the remaining (N − 2) internal coordinates, where the (N − 2) coordinates do not need to be mutually orthogonal. This is called "maximal independence of the coordinate ξ 1 from the rest of (N − 2) internal coordinates".
The coordinate of the internal one-particle density matrix should be either the one particle coordinate measured from the total center-of-mass or the one particle coordinate measured from the center-of-mass of the other (N − 1) particles. The PP-type internal one-particle density matrix adopts the former type coordinate q 1 while the Jacobi-type internal oneparticle density matrix adopts the latter type coordinate ξ 1 , coinciding in the macroscopic number. These two types of coordinates are essentially the same since they are related by q 1 = ((N − 1)/N)ξ 1 . In constructing the internal one-particle density matrix, we integrate the internal density matrix over the other (N − 2) internal coordinates. In order for the internal one-particle density matrix to have maximum information on the one-particle degree-of-freedom of the system, these (N − 2) internal coordinates should be maximally independent from the coordinate q 1 or ξ 1 . Since this requirement seems natural, we will study in this subsection its consequence. We will see below that this requirement implies the criterion proposed in the preceding subsection.
The requirement that the coordinate q 1 or ξ 1 is maximally independent from the other (N − 2) internal coordinates means mathematically that coordinate q 1 or ξ 1 should be orthogonal to the other (N − 2) internal coordinates. The Jacobi coordinates are just such coordinates, since they constitute an orthogonal coordinate system. On the other hand the internal coordinates {q i , i = 1 ∼ (N − 1)} do not satisfy this requirement since they constitute a non-orthogonal coordinate system. For the sake of self-containedness we recall here the meaning of orthogonality between coordinates. Two coordinates β and γ are defined to be mutually orthogonal when their expansion coefficients of linear combination with respect to the N particle coordinates are mutually orthogonal, N i=1 C i (β)C i (γ) = 0. This definition can be stated as follows. To any coordinate δ we associate an N-dimensional number vector C(δ) like C(β) = {C i (β), (i = 1 ∼ N)} for β and C(γ) = {C i (γ), (i = 1 ∼ N)} for γ. If β = γ, C(β) = C(γ). β and γ are said to be orthogonal when C(β) · C(γ) = 0. Needless to say, the total center-of-mass coordinate R is orthogonal to any kind of internal coordinates δ int , i.e. C(R) · C(δ int ) = 0.
Let an internal coordinate system {η i , (i = 1 ∼ (N − 1))} be such a system satisfying the above-mentioned requirement. The coordinate η 1 is either q 1 or ξ 1 and hence we here fix it as η 1 = ξ 1 . The orthogonality of ξ 1 to the other (N − 2) coordinates {η i , (i = 2 ∼ (N − 1))} means C(ξ 1 ) · C(η i ) = 0, (i = 2 ∼ (N − 1)). Since the internal Jacobi coordinate system {ξ i , (i = 1 ∼ (N − 1))} is an orthogonal coordinate system, there the relations C(ξ 1 ) · C(ξ i ) = 0, (i = 2 ∼ (N − 1)) holds. Therefore the subspace spanned by the N-dimensional vectors {C(η i ), (i = 2 ∼ (N − 1))} is identical to the subspace spanned by N-dimensional vectors {C(ξ i ), (i = 2 ∼ (N − 1))}. We, therefore, have This relation is equivalent to the relation It is to be noticed that unlike the Jacobi coordinates the mutual orthogonality within the By using Eq. (57) we can calculate the internal one-particle density matrix ρ (1) int as follows This result shows that ρ (1) int (η, η ′ ) is just the same as the Jacobi-type internal one-particle density matrix. Thus we see that our above requirement concerning the internal coordinates, gives us a unique result for the internal one-particle density matrix. It is just identical to the Jacobi-type internal one-particle density matrix.
In the previous section III A we have proposed also another requirement, namely that the internal one-particle density matrix should converge to the one-particle density matrix in the laboratory frame in the macroscopic limit. As we discussed in Sec. III A, this requirement implies that there should not appear any cross terms of ξ 1 with the other internal coordinates in the internal wave function at least in the macroscopic limit. Clearly the absence of the cross terms of ξ 1 with the other internal coordinates is realized only when the coordinate ξ 1 is orthogonal to all the other (N − 2) internal coordinates. Thus the criterion in the preceding subsection results from the present requirement of the orthogonality of ξ 1 to the other (N − 2) internal coordinates, where the (N − 2) coordinates generally do not need to be mutually orthogonal.
One may argue that a physical quantity should not depend on the choice of the coordinate system. In the case of the one-particle density matrix this argument can be true under the condition that we have extracted maximum information on the one-particle degree of freedom of the system. As already noticed this condition is the same as the requirement that the coordinate of the one-particle density matrix is orthogonal to all the other coordinates of the system. This is also true in the macroscopic system where we usually adopt the coordinate system composed of individual particle coordinates {r i , (i = 1 ∼ N)}. The coordinate r 1 of the one-particle density matrix is of course orthogonal to all the other coordinates. If we adopt the coordinate system {R, q i , (i = 1 ∼ (N − 1))} and calculate the one-particle density matrix with respect to the coordinate q 1 which is practically the same as r 1 in the macroscopic system by integrating out the remaining coordinates {R, q i , (i = 2 ∼ (N −1))}, we get the result given in Eq. (29), namely (2ν/π) 3/2 exp[−(3/2)ν(q 2 1 + q ′ 1 2 ) + νq 1 · q ′ 1 ]. This inadequate result is, of course, due to the non-orthogonality of q 1 to the other coordinates {q i , (i = 2 ∼ (N − 1))}, as we have already seen earlier.
We have proved in this section that the internal one-body density matrix is uniquely determined for the 0S harmonic oscillator wave function. However, it is noted that this uniqueness holds in the case of a general wave function including the 0S harmonic oscillator one. In fact, Suzuki et al. [29,30] have already given the proof of the uniqueness for the more general wave function Ψ which is expanded in terms of the correlated Gaussian basis g. The explicit forms of Ψ and g are expressed as follows: where x = {x 1 , x 2 , · · · , x N −1 } and S are a set of internal coordinate of N-boson system and the symmetrization operator acting on the N bosons, respectively. C k , A (k) and s (k) are the expansion parameters, in which A (k) denotes a symmetric positive-defined (N − 1) × (N − 1) matrix, and s (k) represents s (k) = {s N −1 }. The correlated Gaussian basis g is often used in ab-initio calculations and has succeeded in describing structures of many few-body systems [31]. The uniqueness of the internal one-body density matrix ρ for the wave function Ψ in Eq. (70) is proved [29,30] under the condition that one takes the following set of coordinates y = {y 1 , y 2 , · · · , y N −1 } (obtained by a linear transformation from the coordinates x) as the coordinates of Ψ adopted in calculating ρ: q 1 is chosen as the coordinate (y 1 ) of the internal one-body density matrix and the remaining (N − 2) internal coordinates (y 2 , y 3 , · · · , y N −1 ) are orthogonal to q 1 , although the (N −2) coordinates do not need to be mutually orthogonal. This requirement for the coordinates y just corresponds to the "maximally independence of the coordinate ξ 1 (q 1 ) from the rest of (N − 2) internal coordinates" as mentioned above. The proof we gave in this section is specified to the simple system which is composed of N bosons in the harmonic oscillator potential and has been discussed by many authors (for example, see Refs. [18,19,21,22]).
IV. SUMMARY
We investigated the internal one-particle density matrix in the case of ideal Bose-Einstein condensates with a finite number ( N ) of particles in a harmonic trap. We calculated the explicit form of the internal one-particle density matrix following the definition of Pethick and Pitaevskii (PP) and solved its eigenvalue problem. The result was found to show a fragmented condensate, contrary to what PP expected. On the other hand the Jacobitype internal one-particle density matrix gives us complete condensation. It means that the internal one-particle density matrix is different in general for different choices of the internal coordinate system. In this paper we outlined two physically motivated criteria for the choice of the adequate coordinate system leading to a unique answer for the internal one-particle density matrix. One criterion is that in the infinite particle number ( N = ∞ ) limit the internal one-particle density matrix should have the same eigenvalues and eigenfunctions as those of the corresponding ideal Bose-Einstein condensate in the laboratory frame. The other criterion is that the coordinate of the internal one-particle density matrix which is either q 1 = r 1 − R or ξ 1 = (N/(N − 1))q 1 , should be maximally independent from the remaining (N − 2) internal coordinates. Mathematically this criterion means that q 1 (or ξ 1 ) is orthogonal to the remaining (N − 2) internal coordinates, though the (N − 2) coordinates, in general, do not need to be mutually orthogonal. This second criterion was shown to imply the first criterion. We saw that the internal Jacobi coordinate system satisfies these two criteria while the internal coordinate system adopted by Pethick and Pitaevskii for the construction of the internal one-particle density matrix does not. Furthermore we argued that these two criteria uniquely determine the internal one-particle density matrix which is the same as that calculated with the Jacobi coordinates. The results of this paper justify the use of the Jacobi coordinates in Refs. [16,17] where the Bose condensation of a few α-particles in extended states of self-conjugate light nuclei was considered. However, our results are of more general interest. For example the number of bosons captured in each site of an optical lattice is often very small [28] and, therefore, our analysis surely applies to that situation as well. We also believe that our present study has considerably clarified the somewhat controversial issue of how to define a Bose condensate of a finite number of particles in their internal coordinate system. This question is particularly, but not only, The one-particle density matrices discussed in this paper are symmetric Gaussian integral kernels of the form exp{−a(r 2 + r ′ 2 ) + br · r ′ }.
This kernel is normalized for which we need the condition 2a − b > 0.
The eigenvalue problem of the kernel Q(r, r ′ ) can be solved analytically. There are several ways to solve it. We can e.g. use the method given in Ref. [27]. Here we explain another procedure which gives us the following answer (see below) Here α † c and α c are creation and annihilation operators of harmonic oscillation of size parameter c, respectively. The eigenfunctions ofQ are just the same as those of the operator α † c · α c which are of course the harmonic oscillator functions of the size parameter c. Thus we haveQ The explicit form of φ nLM (r, c) is given in Eq. (23). It is to be noted that the eigenvalue The relation of Eq. (A3) comes directly from the following theorem [32] valid for a general operator P : where N op is the operator of normal ordering, and The state |A γ (Z) is the well-known coherent state of harmonic oscillator of size parameter γ, The proof of the general theorem of Eq. (A8) is quite easy. It is given by comparing the matrix elements of two operators, P and N op {p(α † γ , α γ )}, formed with the coherent states |A γ (Z) which constitute an overcomplete set of states. This is shown by the following operation: The equality of Eq. (A19) is due to Eq. (A14).
The expression of Eq. (B3) is the manifestation of the bosonic symmetry of the Jacobi-type internal one-particle density matrix. This equality is due to the following relations where O(k, ξ, ξ ′ ) is defined as (k = 1 ∼ N). | 2009-05-01T01:25:08.000Z | 2008-11-03T00:00:00.000 | {
"year": 2008,
"sha1": "77f699d26790eb0c04a4749c8f8fe418fc340998",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/0811.0288",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "77f699d26790eb0c04a4749c8f8fe418fc340998",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
117802692 | pes2o/s2orc | v3-fos-license | Shifted COCG method and its application to double orbital extended Hubbard model
We explains the shifted COCG method which can solve a series of the linear equations generated by numbers of scaler shifts, without time consuming matrix-vector operations, except at the only one reference energy. This is a family of the CG method and sharing the robustness and the capability of the accuracy estimation. Then shifted COCG is quite useful to calculate the Green's function of the many-electron Hamiltonian which have very large dimension. We applied it to the double orbital extended Hubbard model with twelve electrons on the periodic sqrt(8) x sqrt(8) site system, the dimension of the Hamiltonian equals to 64,128,064, and found the ground state is insulator. We also explained the crucial points of the shifted COCG algorithm for reducing the amount of required memory.
I. INTRODUCTION
Strongly interacting systems attract considerable attention because of fruitful phenomena in a new field of cross-correlation physics 1 and their potential applicability to developing field of spintronics. Theoretical study of strongly correlated systems, e.g. many-electron systems and interacting spin systems, becomes time-consuming and more difficult when one starts numerical investigation of larger systems.
One reason of this difficulty is, of course, the large dimension of the Hilbert space or the Hamiltonian matrix of many-electron systems. The dimension of the Hilbert space grows exponentially with increasing number of atoms linearly in a many-electron system and, on the contrary, that in a one electron problem (or the density functional theory, DFT) the size of the Hamiltonian matrix is proportional to the number of atoms. The second reason is the fact that the rigorousness or accuracy control becomes seriously difficult in a problem of the large Hamiltonian matrix. Because the width of the spectra is in proportion to the number of atoms in many cases, the energy interval between adjacent eigenenergies becomes small quite rapidly with increasing number of atoms. The short interval between adjacent eigenenergies causes the difficulty in separating of respective eigenvectors. Then, for example, it is very important to obtain the precise ground state, from which all the physical quantities are derived in the (zero-temperature) manyelectron theory. Thus, one needs higher energy resolution with increasing number of atoms, but, sometimes, we do not have fast, reliable and stable calculation algorithm for large Hamiltonian matrices.
Our main target is the calculation of the Green's func-tion matrix G(ω) in many-electron problems; where H and ω are a real Hamiltonian matrix and energy parameter, respectively. The suffices i and j denote arbitrary state such asĉ i | orĉ † i | , whereĉ is an annihilation operator and | is a ground state. Here, we should use a positive finite parameter η in a numerical calculation of a finite system, instead of infinitesimally small positive number. The spectral function is the important physical quantity derived from the Green's function Eq. (1).
There are two possibilities for calculating Eq. (1). One is to solve the eigenvalue problem with an eigenvalue ω, e.g. the Lanczos method. The other is to solve following linear equation and to take inner product between the solution and vector |i , with an arbitrary energy parameter ω, e.g. the shifted COCG (conjugate-orthogonal-conjugate-gradient) method, a family of the CG (conjugate-gradient) method. In both cases, we first restrict the space dimension of states to be finite. In other words, we assume the size of the Hamiltonian matrix to be finite. Then we construct the Krylov subspace defined as K n (A, |j ) = span{|j , A|j , A 2 |j , . . . , A n |j }.
In the Lanczos method, orthogonalized base vectors (Lanczos vectors) are successively generated in K n (A, |j ), and at the same time, the Hamiltonian matrix is tridiagonalized. In a large scale calculation, one can only use a small Krylov subspace, because of heavy load of computation and a corruption of the orthogonality of generated basis vectors. It is well known that the rounding error breaks down the orthogonality of the generated base vectors rapidly, when the dimension of the Krylov subspace exceeds several tens. The corruption of the orthogonality causes spurious eigenvalues and, more seriously, incorrect eigenvectors. Therefore, the size of the Krylov subspace should be limited usually to some tens or a hundred.
We developed the shifted COCG method, where the Eq. (4) is solved within the Krylov subspace, and applied it to the one-electron tight-binding Hamiltonian in the system with a large number of atoms. 3 A set of orthogonal base is created by the iterative process of the shifted COCG method, like Lanczos process, but the calculation is stable for large dimension of the Krylov subspace, in contrast to the Lanczos method. We must solve the Eq. (4), for every scalar shift σ of A corresponding to respective energy mesh point. The number of the σ's is as much as O(10 2 )∼O(10 4 ) generally, however, the most time-consuming matrix-vector operations are needed only at a single reference energy (σ = 0). Then the order of the total amount of calculation is just the same as Lanczos method. The reduction of the matrixvector operation at non-zero σ are based on the fact that a power of (A + σ) is decomposed into a linear combination of powers of A. Thus, Krylov subspace is invariant K n (A, |j ) = K n (A + σ, |j ) against σ.
In the application of this method to the many-electron theory, because the dimension of the vectors is huge, we must take care for the total amount of base vector storage for K n (A, |j ), in order to satisfy the memory constraint in modern computers. We explain the innermost loop index should be the iteration step n, for an extremely large size of the Hamiltonian matrix. This structure also give us following additional two merits. One is that a part of the program code can be used in the inverse iteration process to improve the ground state. Another is that the calculation with the different η can be done without time consuming matrix-vector operations.
The structure of the paper is as follows. In Sec. II, the basics of the shifted COCG method is explained briefly. Section III is devoted to explanation of how to obtain global convergence. Then we apply the shifted COCG method to an extended Hubbard Hamiltonian with orbital degeneracy and intra-and inter-site Coulomb interactions in Sec. IV, where the size of Hamiltonian matrix is equal to 64,128,064. We calculate one-electron excitation spectra and evaluate the insulating gap. In Sec. V, we will conclude that the essential difficulties of numerical investigation of many-electron problems, the accuracy control (or monitoring) and the robustness are achieved by the present method, within the moderate amount of memory space. We explain the two points to understand the mathematics in the back ground of the shifted COCG method in Appendix A. The practical design of storing the huge Hamiltonian matrix is discussed in Appendix B.
II. SHIFTED COCG METHOD
Assuming that the Hamiltonian is represented by using N -dimensional real matrix H and A is a complex symmetric matrix ω ref + iη ref − H, we should solve the linear simultaneous equation of and its shifted equation where σ = (ω+iη)−(ω ref +iη ref ). We represent quantities q in the shifted system as q σ . The right hand side b represents |j in Eq. (3). We assume that the vector b is a real and normalized.
In the family of CG method, here the shifted COCG method, it is important that the approximate solution of Eq. (6) is searched within the Krylov subspace K n (A, b). The subspace K n (A, b) becomes whole space at n = N − 1, and the solution becomes exact.
The accuracy of the approximate solution at n-th iteration x n is evaluated by using the residual vector, and the iteration is stopped as soon as the norm of the residual vector, ||r n ||, satisfies the criterion for the convergence. The residual vectors are "orthogonalized" with respect to the non-standard "inner-product" (u, v) = u T v. When η = 0, all the relevant vectors are real and the "inner-product" and "orthogonality" reduce to standard ones, respectively. Because r n 's are "orthogonalized", it is convenient to use them as base vectors of K n (A, b). In addition to that, owning to the "orthogonality", we obtain the important theorem of "collinear residual" (See appendix A).
Here, we must notice the fact, in the procedure of iteration, (v, v) = 0 can happen though v = 0. 5 This cannot happen in the CG method (η ref = 0, the matrix A is positive definite) and the other part is perfectly identical to the CG method. A set of residual vector r n forms the "orthogonalized" base. This "orthogonality" is very important for us to understand the theorem of collinear residual. We explain it in detail in Appendix A. We can choose an alternative set of the recurrence equations, as follows. Eliminating p's from Eqs. (10) and (11), we obtain the recurrence equation of r n , Taking "inner product" between r n and the Eq. (14), we obtain α n = (r n , r n ) (r n , Ar n ) − βn−1 αn−1 (r n , r n ) .
B. Shifted equations
The key to the reduction of the matrix-vector operations in solving the shifted system Eq. (7), is the theorem of collinear residual: where the π σ is a scalar function (actually polynomial) of σ. Then, once {r n } are given, the base set {r σ n } n for the arbitrarily shifted system can be obtained by using scalar multiplication. We obtain the recurrence equations that determines π σ n , α σ n , β σ n , x σ n , and p σ n , from Eqs. (9)∼(13), with replacing A by A + σ, with the same initial conditions: These recurrence equations can be solved without time consuming matrix-vector operation. In addition to that, each component of the vector Eqs. (17)∼(21) can be solved separately, due to the absence of the matrix operation.
C. Crucial remarks for extremely large matrix to save required memory space For the solution of the relatively small matrix (Dim 10 4 ), any loop structure of shifted COCG method can be applicable. However, in the many-electron theory, the dimension of the intermediate vectors is huge and then the number of intermediate vectors is restricted to some tens or hundreds. In the standard loop structure which Frommer showed, 2 the outermost loop index is the iteration step n, and, all the vectors p σ n , r σ n and x σ n for every energy mesh point σ, are required in order to start the calculation at the iteration step n + 1. Then all the energy mesh points must be fixed before the calculation starts. Because the calculation at the respective energy mesh points are independent to each other, the loop structure can be transformed such that the reference system is solved with the COCG method storing the {α n } n ,{β n } n and {r n } n , then the shifted systems are solved with stored information about reference system for each energy mesh points.
Here the innermost loop index is the iteration step n. Since the number of energy mesh points is larger than the number of iteration generally, the latter transformed loop structure requires smaller memory than the original one. In addition to that, we need not to prepare energy mesh points {σ} because all the required information related to the reference system are stored in the COCG process, the preceded part of the algorithm. Then, for example, we can change the smearing factor η freely without repeating COCG process that includes matrix-vector operations.
Further reduction of the required memory is possible, with further modification of the recurrence equations for the shifted system. Assuming that a real constant vector c is an adjoint vector and taking the inner product between c and the Eqs. (20) and (21), we obtain a set of self-contained equations for determining the (c, x n ), n-th approximate solution of the element of Green's function, due to the absence of matrix-vector operations.
In the applications in Sec. IV, we are interested in the case where c = b in order to calculate the trace of the Green's function. Therefore, we need to store (b, r n ), α n , β n and ||r n ||, in the COCG part, and later, solve only the b-component of Eqs. (20) and (21). Here, the norm of the residual vector ||r σ n || = 1 |π σ n | ||r n || is not necessary to solve the recurrence equations but is used to monitor the convergence of the approximate solution. Additionally, we store the full components of the last two r n 's in the COCG part, in order to extends the iteration number in the seed switching part (subsection III B).
Even if the full components of the Green's function are needed, we need to store just a few components of r n , because the suffix of the Green's function denotes the one-electron orbitals, the number of which is very small compared to the dimension of many-electron Hamiltonian H.
D. Preparation of ground state wavefunction
The transformation of the loop structure in the subsection II C increase the re-usability of the program code. The COCG part of the code can also be used in the process to improve the ground state wavefunction as follows.
First we use the Lanczos method in order to tridiagonalize the Krylov subspace, then, obtain the ground state diagonalizing it. The calculated lowest eigenenergy converges rapidly with increase of the dimension of the subspace, but the wavefunction does not, due to the unstable orthogonality against the inevitable rounding error. Next we improve the approximate eigenenergy and the wavefunction with the inverse iteration method. Because the COCG process with the real arithmetics is the same as CG process, here we can use the COCG part of the shifted COCG algorithm whose loop structure is changed as in the subsection II C. 7 Since the inverse iteration method works only when the approximate eigenvalue and eigenvector are given, the first Lanczos process can not be omitted. If the accuracy of the calculated wavefunction is not enough, the processes are repeated with replacing the initial Lanczos vector by the latest approximate wavefunction.
A. Estimating accuracy of Green's function
In this subsection, we explain the accuracy of the Green's function calculated by the shifted COCG method and give its estimation. Here G exact is the "exact" solution of Eqs. (3) and (4) for a given finite value of η. Then we say that "the calculated Green's function is accurate", when the | | (hereafter, "accuracy") is small. The "accuracy" and ||r n || are generally "truncation error" of G and x n , respectively. Because the shifted system is equivalent to the reference system, then we can estimate the accuracy for the shifted system, with replacing A by A + σ and any other quantities {q} by {q σ }.
We can derive following equation from Eqs. (3), (4) and (8), If the matrix A was positive definite real symmetric matrix and the vector b and r n were real vectors, the upper bound of the right hand side of Eq. (22) is equal to ||rn|| ||b|| = ||r n ||.
When the matrix A can be fully diagonalized numerically, we can estimate G exact within rounding errors, and then, obtain the "accuracy" of the approximate Green's function calculated by shifted COCG method.
The dotted and solid line in the Fig. 1 show the "accuracy" of [G sCOCG ] jj and the norm of the residual vector ||r n ||, respectively. The dashed line shows the excitation spectra. The figure shows that the "accuracy" is bounded by the norm of the residual vector. Therefore, we can estimate the "accuracy" of the calculated Green's function by using ||r n ||, without the knowledge of the "exact" solution G exact . We can also see from the figure that the Green's function calculated by the shifted COCG method is accurate more, near the bounds of the spectra.
B. Seed switching
Assuming that the approximate solution of the reference system, Eqs. In that occasion, we should change the seed ( where the norm of the residual vector is large and the approximate solution does not converge. Because the shifted system is equivalent to the reference system, we can change the seed as follows, without disposing the previous calculation at the old ω ref . 6 We define σ max so that r σmax n 's, β n 's, and ||r n ||'s (0 < n ≤ M ) without any matrixvector operation, though the matrix-vector operation is required to calculate the new ones at further n ′ -th iteration (M < n ′ ≤ M new ).
The same remark as the subsection II C is applicable to the implementation of the seed switching. The shifted COCG algorithm with seed switching of any loop structure can be applicable to relatively small matrices, but, we must change the loop structure of it from the previous one 6 , since the size of the intermediate vector is huge in the many-electron theory. We must even change the recurrence equation of r n to Eq. (14), so that the r n+1 is calculated only with the r n−1 and r n , in stead of all {r n } n=1,2,...,n , for extremely huge matrices. When we store {α n , β n } n=0,1,...,M−1 , r M−1 and r M in the COCG process, then we can calculate α σmax M−1 , r σmax M−1 and r σmax M , which are required for the following calculations. Then we calculate the new reference system up to M new -th iteration step, by using Eqs. (13), (15) and (14). Figure 2 shows the example of the seed switching. The system is the same one as in Fig. 1 except ω ref . Here the η = η ref = 0.05eV. At the ω ref = −1.40eV, ||r n || decrease exponentially, and satisfies the criterion ||r n || < 10 −10 at iteration step n = 330. However, there are many energies where the converging speed of the residual vector ||r σ n || is slower than that at ω ref . Then the σ max = 7.43eV is searched and the ω new ref is shifted to be 6.03eV. We need seed switching twice more at 1509-and 3297-th iteration in order to obtain the global convergence. The largest value of the "accuracy" is 1.5×10 −10 at the last iteration step n = 3455.
C. Robustness of shifted COCG method
In this subsection, we explain the robustness of the shifted COCG method, which is very important to obtain the converged approximate solution, especially in the case of the long iteration. We say that the calculation is robust, when the calculation is stable against the perturbation. For examples, the orthogonality of {r n } is not a robust property, because the inevitable rounding error perturb the calculation and the orthogonality is broken down quickly.
The robustness of the shifted COCG method consists of two parts. One is the robustness of COCG method at the reference energy ω ref . And the other is the robustness of the iterative solution of the shifted equations. Figure 2 shows the robustness of the COCG method at ω ref , because the norm of the residual vector ||r n || goes to 0 in spite of long iteration 3,540. The global convergence of the "accuracy" that is mentioned at the end of the subsection III B shows the robustness of the iterative solution of the shifted equations. In the shifted COCG method, the "orthogonality" of base vectors {r n } is not necessary for reducing ||r n ||, in contrast to the fact that the subspace diagonalization methods requires the unitarity of the base vectors.
IV. APPLICATION OF THE SHIFTED COCG METHOD TO THE MANY-ELECTRON PROBLEM
Here we apply the shifted COCG method, to the double orbital extended Hubbard Hamiltonian and calculate the excitation spectra. The experimental results show that the layered perovskite La 3 2 Sr 1 2 NiO 4 is an insulator with charge and spin stripe order, as depicted in Fig. 3. The charge and spin structures of the single layer of La 3 2 Sr 1 2 NiO 4 (pseudo twodimensional system), choosing Ni 3d e g orbitals as relevant ones, was studied with the extended Hubbard model recently. 8 Here we use the same Hamiltonian.
where the suffix {i, j} denote the site, {α, β} denote the orbital 3z 2 − 1 or x 2 − y 2 , and {σ, σ ′ } denote the spin co-ordinate. The annihilation and number operator areĉ andn, respectively. The symbol t, ε, U , J, V denote the Slater-Koster type hopping parameter, single electron energy, on-site Coulomb interaction, on-site exchange interaction, intersite Coulomb interaction, respectively. Hopping parameters are finite for nearest neighbor (n.n.) and second n.n. pair of sites. The braces · · · means that two sites enclosed by them are the n.n. sites. Though the anisotropy of the hopping parameters for the second n.n. pair stabilizes the spin structure, 8 we choose the isotropic (tetragonal) parameter set shown in Table I. The role of U and J in the present situation is stabilization of integral valency of Ni ions (Ni 3+ and Ni 2+ ) and spin polarization. Here we summarize the properties of the calculated ground state of the La 3 2 Sr 1 2 NiO 4 . 8 The calculated ground state shows the charge and spin stripe order consistent with experimental observation and the system is insulator. Diagonal hole stripes are separately localized on Ni 3+ site in order to reduce hole-hole interaction energy induced by inter-site Coulomb interaction V . Charge order and the inter-site Coulomb interaction V are directly related to the energy gap in the excitation spectra of the system. Spin stripe occurs only under the condition of the existence of multi-orbitals and the charge order with a help of anisotropy. The spin stripe is determined by the electronic structure with smaller energy scale than that of the charge stripe.
C. Computational details
Here we explain miscellaneous computational details.
The calculated system is two-dimensional square lattice. There are 12 electrons on the periodic √ 8 × √ 8 sites. Because the total S z of the system is preserved, we can reduce the number of relevant many-electron states to 64,128,064, by using the condition S z = 0.
The smearing factor η is also an arbitrary parameter in the present paper. Here we explain how we chose the value of η. The energy scale of the low energy excitations is t ∼ V ∼ O(10 −1 eV ), because the value of onsite Coulomb interaction U is much larger. Therefore we must set η lower than 10 −1 eV, so that η does not smear out the finer structure of the spectra than itself. There is another restriction that the interval of the energy mesh is small enough than η, in order to see the fine peak structure of the spectra. Then, because the calculation time increases with decreasing η, the value of η is roughly determined as O(10 −3 eV)∼O(10 −2 eV).
Next point is the criterion for the convergence of the ground state vector. Our calculations are of the double precision and the rounding error is inevitable (∼ 10 −16 ) in the each component of the eigenvector. Assuming the accumulated error is of O( √ N ) (N =64,128,064), the accuracy is expected to be 10 −16 × √ N ∼ 10 −12 . We set the allowance for the estimation by factor of 10 2 , and the criterion for the accuracy of calculated ground state energy E gs and eigenvector | is | (Ĥ − E gs ) 2 | < 10 −10 . ) and their "accuracy" (black line, see text). The spectra of affinity and ionization levels are calculated separately by using the shifted COCG method, for the given smearing factor η = 0.01eV (upper panel) and η = 0.10eV (lower panel). The highest occupied level is at 9.4eV and lowest unoccupied 10.3eV. Intersite Coulomb interaction V = 0.5eV. Energy zeroth is set at the ground state energy 36.755eV of 12 electron system.
D. Spectral function
We examined the spectral function This can be easily evaluated by the shifted COCG method. Figure 4 shows the spectral functions of the state D at V = 0.5eV. The upper and the lower panel show the case of η = 0.1eV and η = 0.01eV, respectively. Both of them are calculated from the same COCG calculations and the only difference between them is the imaginary part of the energy shift σ. The spectra of ionization and affinity levels are calculated separately and the ω ref 's for respective spectra are chosen to be (9.0+i0.01)eV and (10.4 + i0.01)eV. The highest occupied level is at 9.4eV and lowest unoccupied 10.3eV. The number of iterations equals to 800 for each spectra. If one attempt to obtain the profile of the spectra with smoothly connected curves as is in the bulk limit, one should set the value of η sufficiently larger than (width of spectra) (iteration number) , in order to smear out the excessive peaks caused by the finite system. Because iteration number equals to 800 in the present calculations, this criterion becomes η ≫ 0.03eV, and the gray curves in the upper panel of Fig. 4 shows the smooth profile of the spectral function. If one attempt to see whether the energy gap opens at the boundary between affinity and ionization levels, one must choose sufficiently smaller η than the width of energy gap. In the present calculation, this criterion becomes η ≪ 0.9eV, and, the gray curves in the lower panel of Fig. 4 show the energy gap around ω = 9.8eV. We can choose η independent with the reference energy, then the energetic resolution of the spectral function can be changed after all the time consuming matrix-vector operations have been finished.
The black curves in the upper and lower panels of the Fig. 4 show the "accuracy" of the respective spectral functions, and the spectra are extremely accurate near the boundary of the spectra (ω = 9.8eV), where the energy gap is open. Therefore, we conclude from the lower panel of Fig. 4 that the ground state of the present Hamiltonian is insulator.
Changing the value of V continuously from 0.5eV to 0.0eV, we find that the system becomes metal. 8 There-fore, the intersite Coulomb interaction makes the present system insulator, unlike the usual transition metal oxide where the large on-site Coulomb interaction makes the system insulator.
V. DISCUSSION AND SUMMARY
Once COCG method is applied to the reference system Eq. (6), the shifted system Eq. (7) is solved without time consuming matrix-vector operations, by shifted COCG method. This notable property is due to the mathematical structure of COCG method, such that the residual vector's are forming the "orthogonal" base set of vectors, whose direction does not change against σ. This reduction of the matrix-vector operation extremely accelerate the calculation speed of Green's function G(ω), keeping the robustness of COCG method. Simultaneously, the accuracy of the approximate G(ω) is easily estimated as the norm of the newly generated base vector (residual vector) at the latest iteration. The total accuracy of the shifted COCG method varies depending on ω, σ and ω ref 10 and generally very small near the bounds of the respective spectra. In the many-electron Green's function, we are usually interested in the low energy excitations, in other words, the spectra near the boundary between affinity and ionization levels. Therefore, we can calculate the Green's function accurately and quickly by the shifted COCG method, in the interesting energy range.
Another problem in the application of the COCG method to the many-electron theory is a memory constraint due to the extremely large size of vectors and matrices. We resolved this problem with separating the COCG part for the reference system and the part for the shifted equations, changing the loop structure. This change give us the following two merits. One is a usage of the former part for improving the ground state, as is mentioned in the subsection II D. The other is the fast calculation of changing smearing factor η, which is just a imaginary shift. When we do not know the proper energy scale a priori, the width of the energy gap in the present paper or the proper value of η, this merit is very important.
The seed switching is a very important idea for the shifted COCG method to give global convergence, which means that the calculated solution converges everywhere in the interested energy region. Because it takes much iteration steps to converge the solution especially in the middle of the spectra, sometimes we must discontinue the iteration step before obtaining global convergence. In that case, we must examine the accuracy of the result and check if the solutions in the required energy range satisfy the criterion.
The applicability of the above reconstruction and the seed switching are not specific to the many-electron problem. We can apply them to the general solution of the Green's function of extremely large dimension.
We applied the shifted COCG method to the charge and spin order in La 3 2 Sr 1 2 NiO 4 , where the intersite Coulomb interaction, relatively small compared with onsite one, plays an important role. Then we conclude the relatively small energy gap opens at the Fermi energy, and the system becomes insulator, due to the intersite Coulomb interaction.
Acknowledgments
Calculations were done at the Supercomputer Center, Institute for Solid State Physics, The University of Tokyo. This work was partially supported by a Grantin-Aid for Scientific Research in Priority Areas "Development of New Quantum Simulators and Quantum Design" (No.170640004) of The Ministry of Education, Culture, Sports, Science, and Technology, Japan.
APPENDIX A: MATHEMATICAL STRUCTURE OF SHIFTED COCG METHODS
Here we explain the two important points to understand the mathematical structure of the shifted COCG methods. 2,11 One is the "orthogonality" of the residual vectors {r k } with respect to a non-standard "inner-product" (u, v) = u T v, which the theorem of collinear residual is based on. On should be noticed that this "orthogonality" is different from the well-known "A-orthogonality" of the searching directions {p k }. Because {r k } k=0,1,...,n−1 is a base set of K n−1 (A, b), the "orthogonality" is also represented as follows r n ∈ K n (A, b) and r n ⊥ K n−1 (A, b), where the over line stands for taking conjugate. Therefore, the direction of r n is uniquely determined by above equation, for arbitrary A, as the 1-dimensional complementary space of K n−1 (A, b) within K n (A, b). Then the theorem of the collinear residual Eq. (16) is derived from the invariance of Krylov subspace against σ. Another point appears in the similarity between the two Eqs. (14) and (17). Representing x n and r n as X n−1 (A)b ∈ K n−1 (A, b) and R n (A)b ∈ K n (A, b), respectively, we can derive the relation between two polynomials R n (t) = 1 − tX n−1 (t) from Eq. (8). This relation and the theorem of the collinear residual lead to the relation between the two polynomials π σ n = R n (−σ). 2 This relation explains the similarity between the two Eqs. (14) and (17), and the plus sign in front of σ in the latter equation.
Thus the mathematical structure of the shifted COCG method consists of two structures, that of a vector space and that of a set of polynomials. The non-standard "inner product" in the shifted COCG method can be recognized as the conservation of the analytic property as a polynomial of σ.
APPENDIX B: REDUCING STORAGE SIZE OF HAMILTONIAN
In investigating the properties of the many-electron Hamiltonian, there is a trade-off between the speed of matrix-vector operation and the amount of the memory where the matrix elements of the Hamiltonian are stored. Assuming the all non-zero matrix elements are stored separately and the number of the non-zero matrix elements per each column is equal to 20, then about 10 GB is required to store the Hamiltonian in the case of present paper. That is too much for the most of modern computers. In stead, if the operation of the Hamiltonian on the vectors is implemented as the summation of the respective term in Eq. (23), then the number of operator equals to 640, and, the memory required to store 640 operators is negligibly small compared to 10GB. However, these 640 operators must be applied to the vectors, for single operation of the Hamiltonian. Therefore, the calculation time increases, compared to the case of the all elements of the Hamiltonian are stored.
For simplicity, the Fermion sign and the two particle operators in the Hamiltonian are neglected hereafter, then, the difference between above two manners of storing the Hamiltonian are described as follows, mathematically. We define V as a whole vector space of the single electron state and decompose it into the direct sum V = ⊕ α V α , α denotes orbital or spin or arbitrary combinations of relevant quantum number. Consequently, the whole space where n-body Hamiltonian acts is described as (B1) Decomposing single particle operator P as P = α,β P αβ , P αβ V β ∈ V α , then the action of P on the ⊗ n V is decomposed as P = ( α,β P αβ ⊗ 1 ⊗ . . . ⊗ 1) + . . .
since the single particle operator P acts on ⊗ n V as (P ⊗ 1 ⊗ . . . ⊗ 1) + . . . + (1 ⊗ . . . ⊗ 1 ⊗ P ⊗ 1 ⊗ . . . ⊗ 1) + . . . + (1 ⊗ . . . ⊗ 1 ⊗ P ). This decomposition is trivial but one should be noticed that the dimension of respective subspace decrease exponentially with n as dim(V α1 ⊗ . . . ⊗ V αn ) = ( dim V m ) n , where m is the number of the partition V = ⊕ α V α . In contrast, the number of the terms to be summed up increases in proportion to nm 2 Therefore, the total number of the matrix elements decrease by this decomposition in Eq. (B2). Extreme limit of the respective V α consists of only one dimension is corresponding to the above mentioned case of 640 operators.
In the present example of the spectral function, we choose the decomposition V = V ↑ ⊕ V ↓ , because total S z is preserved, and consequently, the numbers of ↑and ↓electrons are preserved, and more, the hopping part of the present Hamiltonian does not have the cross term with respect to spin. This partition of the vector space leads to more simplified partition of the operator than Eq. (B2), H = H ↑↑ ⊗1+1⊗H ↓↓ +(residual part), 12 where H ↑↑ and H ↓↓ are including the hopping term with respect to each spin and a part of on-site Coulomb interaction. As a result, the size of the memory area where the values of the matrix elements are stored is about 1GB. Actually, an extra 0.5GB is required for storing the indexes of the place of non-zero elements, then, totally 1.5GB is required for storing the whole Hamiltonian. That is not so big for a modern computer. | 2008-02-20T10:38:24.000Z | 2008-02-20T00:00:00.000 | {
"year": 2008,
"sha1": "0628b9e37644a5a0d0ecbd1fee63e1907c80e440",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "0628b9e37644a5a0d0ecbd1fee63e1907c80e440",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
237475207 | pes2o/s2orc | v3-fos-license | Emerging pollutants and antibiotics removed by conventional activated sludge followed by ultraviolet radiation in a municipal wastewater treatment plant in Mexico Ángeles
Different groups of emerging pollutants (EPs) were identi fi ed (drugs, pesticides, hormones, illicit drugs, and fi re retardant), and the removal of four antibiotics was determined (sulfamethoxazole 62.2 – 94.4%, cipro fl oxacin 71 – 83.1%, nor fl oxacin 82 – 89%, and o fl oxacin 78 – 97.9%) in a municipal wastewater treatment plant (WWTP) in Acapulco, Gro. Mexico. The compounds were extracted from in fl uent and ef fl uent samples by solid phase extraction (SPE). The identi fi cation of non-target EPs was performed by liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS). The antibiotic quanti fi cation was performed by liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS). Most of the non-target compounds were detected only in the in fl uent samples. Antibiotics levo fl oxacin and nalidixic acid, the anticonvulsants carbamazepine and oxycarbamazepine, the local anesthetic lidocaine, and the pesticide tridemorph were the main EPs not removed by the WWTP. In this study, it was shown that the Aguas Blancas WWTP does not manage to remove 100% of the various EPs identi fi ed in the ef fl uent, although the elimination degree is high in most cases, despite being one of the model plants in Mexico.
INTRODUCTION
There is great scientific interest worldwide due to the presence and effects of emerging pollutants (EPs) in aquatic ecosystems. EPs are compounds of different origin and chemical nature, which are recently introduced into the environment or which have been introduced into the environment for a long period of time and, however, they have only recently been detected (Barceló & López de Alda 2008). Most EPs have been studied around the world and are currently regulated and commonly monitored; however, in Latin America, the lack of information related to EPs in the aquatic environment has led to their limited regulation (Peña-Guzmán et al. 2019). EPs have the potential to cause known or suspected adverse ecological effects and represent a great risk for human health (Geissen et al. 2015). There is a great variety of EPs, among which pharmaceutical products, illicit drugs, personal care products, pesticides, surfactants, steroids, and hormones stand out. EPs have been identified in wastewater (WW; Tran et al. 2018), surface water (Peng et al. 2018), groundwater (Félix-Cañedo et al. 2013), and even in drinking water (Glassmeyer et al. 2017). Conventional wastewater treatment plants (WWTPs) are not designed for the removal of EPs and are considered as one of the main sources of entry of EPs into the aquatic environment (Tran et al. 2018;Zhou et al. 2019).
Pharmaceutical products are the EPs that cause the greatest concern, due to their variety, high consumption, degree of persistence, transformation into more toxic molecules, and their incidence in the aquatic environment (Rivera-Jaimes et al. 2018;López-Velázquez et al. 2021). The most studied drugs for their possible effects on aquatic organisms and their return to drinking water systems are antibiotics. A large number of these molecules and their metabolites are released into the aquatic environment. Their introduction is through excreta or urine (not completely metabolized within the body) and by the elimination of unused or expired products (Kümmerer 2009b). Antibiotics have a high degree of resistance to degradation in the environment some even tend to bioaccumulate or biotransform (Kümmerer 2009a). The development of new technological advances and more sensitive and precise analytical methods such as the use of liquid chromatography coupled to highresolution mass spectrometry (LC-HRMS) have allowed the detection and quantification of EPs in the aquatic environment at concentrations from ng L À1 to μg L À1 (Tran et al. 2018).
Fluoroquinolones (FQs) and sulfonamides (SAs) are two important groups of antibiotics that are continuously consumed in large quantities in the world. FQs and SAs are incompletely metabolized by organisms, and their resistance to microbiological degradation and negative effects in the aquatic environment make them the object of priority study in WWTPs (García-Galán Water Quality Research Journal Vol 56 No 3, 168 et al. 2012;Teglia et al. 2019). Removal efficiencies of FQs and SAs during wastewater treatment are still quite scarce, and elimination rates reported so far are usually low (García-Galán et al. 2012).
In Mexico, studies of EPs in aquatic environments are scarce (Félix-Cañedo et al. 2013;Estrada-Arriaga et al. 2016;Robledo-Zacarías et al. 2017;Rivera-Jaimes et al. 2018;Hernández-Tenorio 2021;López-Velázquez et al. 2021). Besides, only 57% of WW generated and collected in municipal sewage systems receive some type of treatment before being discharged to water bodies CONAGUA (2016). Taking into account that the technologies for the treatment of WW have not been fully developed and a large part of the WW only receives primary treatment or is carried out by means of stabilization ponds, exacerbates the problem of the discharge of EPs into the aquatic environment. Therefore, it is advisable to know the presence and removal of these types of contaminant in WWTPs including the secondary treatment with disinfection by UV radiation, as it is the case of Aguas Blancas WWTP. The objective of this study was to identify EPs and quantify FQs and SAs antibiotics present in the influent and effluent, as well as to determine the removal efficiency of a WWTP.
Description of the study area
Acapulco, Guerrero, is one of the most important tourist destinations in Mexico. Its population growth in recent years has promoted the development of the urban area in a significant way, generating a challenge in the supply of public services, especially for the treatment of WW. The study was carried out at the WWTP that treats the WW generated by an estimated population of 800,000 inhabitants. The WWTP treats 60% of the WW by discharging the treated water to the sea, at the point known as Playa Olvidada. The WWTP has an installed capacity of 1,350 L s À1 of flow in the effluent. The preliminary treatment consists of the elimination of floating matter, garbage, and coarse solids by means of mechanical treatments of coarse and fine slabs, mechanical desanding with aeration and with mechanical extraction of the granular material. The secondary treatment consists of the biological aerobic process by activated sludge in the following modality: carbonaceous oxidation, with a BOD 5 removal efficiency of 90.7%, hydraulic retention time of 5.8 h, cell retention time of 3.10 days, retention time in the 9-day aerobic digester, BOD 5 /BOD ratio is 0.68, TSS removal efficiency 0.52%, effluent organic nitrogen 12.0 mg/L, total effluent Kjelldal nitrogen 36.0 mg/L, and total phosphorus concentration 11.0 mg/L. Finally, it also has tertiary treatment based on disinfection by UV radiation, the type of lamps is low pressure/high intensity, the transmittance of 65% of 253.7 nm. The plant receives WW collected from the Bahía de Santa Lucía amphitheater in the influent, including 42 hospitals, some of which are first-class (e.g. General public sector hospital, Cancer hospital), as well as private hospitals and clinics in the city of Acapulco ( Figure 1). The WWTP is considered a model plant for removal of organic pollutants by the Comisión Nacional del Agua en México, based on its results for physico-chemical parameters, since it was rehabilitated in 2010 (CONAGUA 2016).
Sampling
Three sampling campaigns were carried out taking influent and effluent samples of the WWTP at different times of the year September 2017, February, and July 2018. Simple samples of 250 mL were collected every 4 h over a 24-h period to complete composite samples of 1 L in amber glass bottles. Each sample was acidified (pH 3) with hydrochloric acid to reduce microbial activity and improve solid phase extraction (SPE). Hereafter, this composite sample will be referred to as sample (S1, S2, S3; September 2017, February, and July 2018 sampling, respectively). The samples were transported and stored at 4°C in the dark until extraction within the next 48 h. In the laboratory, the samples were filtered (0.45 μm, GF/C, Whatman).
Preparation of standard solutions
Deuterated antibiotic solutions were prepared as internal standards, to compensate for random and systematic errors. The internal standard for sulfa drugs (sulfamethoxazole-d4) was prepared with Methanol (MeOH), and the internal standard for fluoroquinolones (enrofloxacin-d5 hydrochloride) was dissolved in a water-methanol (H 2 O/MeOH) mixture (1:1) that contains 0.2% (v/v) of hydrochloric acid (Golet et al. 2001), after preparation they were stored at À20°C. The individual standard solutions of each antibiotic were prepared and dissolved in methanol, except for the fluoroquinolones, which were dissolved in a mixture of water-methanol (H 2 O/MeOH) (1:1) with 0.2% (v/v) hydrochloric acid. All mixed working standards were prepared prior to analysis.
Solid phase extraction
A known amount of sulfamethoxazole-d4 and enrofloxacin-d5 were added to each influent and effluent sample, prior to extraction with internal standards, to quantify antibiotics from two of the most common antibiotic groups used in Mexico. The SPE method was adapted from Ngumba et al. (2016). SPE was carried out using a Phenomenex brand Strata X SPE column (500 mg/6 mL). The phases were conditioned with 6 mL of MeOH followed by 6 mL of ultrapure water and 6 mL of acidified water at pH 3. Subsequently, 500 mL of the effluent and 250 mL of the influent from each sample were passed through the columns at a speed flow rate of approximately 10 mL min À1 using vacuum. The cartridges were kept at a temperature of 4°C until elution. Elution was carried out with methanol at a flow rate of 6 mL min À1 . Subsequently, the samples were evaporated to dryness under a stream of nitrogen and then reconstituted to 1 mL with 90:10 (H 2 O:MeOH 3 mM ammonium formate, 0.1% formic acid solution). The samples were later stored at À20°C before analysis. Two reagent blanks were run with samples to evaluate analytical quality and the purity of reagents. Liquid chromatography coupled to mass spectrometry For the quantification of antibiotics, a liquid chromatography equipment (Accela 1250 HPLC) was used, coupled to a triple quadrupole mass spectrometer equipped with a heated electrospray ionization source (TSQ Quantum Ultra HESI-II; Thermo Scientific, San José, CA, USA).
In the selected reaction monitoring (SRM) acquisition mode, transitions and collision energies were selected for each compound, which were optimized by direct infusion into the source in a positive mode ( Table 1).
The optimal working conditions for the ESI source were: source voltage, 3,500 V; source gas pressure, 28 ua (arbitrary units); desolvation gas pressure, 2 ua; auxiliary gas pressure, 5 ua; vaporization temperature, 300°C and capillary temperature, 350°C and collision gas pressure, 1.5 mTorr. For the control of the equipment, the acquisition, and processing of the data, the Xcalibur 2.2 and Trace Finder TM 3.1 software was used. This analysis was carried out at the University of Santiago de Compostela, Spain, in the 'Laboratory of Research and Development of Analytical Solutions (LIDSA)' in the Department of Analytical Chemistry, Nutrition and Food Science of the Faculty of Chemistry.
The method exhibited suitable performance in terms of linearity with R 2 from 0.992 to 0.999, with the exception of ofloxacin, which resulted in the order of 0.987. The instrumental precision showed relative standard deviation (RSD) between 0.4 and 12%. SPE Recoveries were carried out loading fortified ultrapure water (1 μg L À1 ) at three different pHs between 3 and 5.5, obtaining acceptable values about 60% for the fluoroquinolones and higher than 83% for the SAs. The method precision was also satisfactory, with deviation values lower than 11%. In general, non-significant differences were observed and a pH value of 3 was selected. LODs of the overall SPE-LC-MS/MS were calculated giving a signal-to-noise ratio of three (S/N ¼ 3). They were between 0.7 and 7.1 ng L À1 allowing the detection of trace levels for the target antibiotics (see Supplementary Material, Table S1).
Liquid chromatography coupled to high-resolution mass spectrometry
For the identification of unknown EPs present in the WW samples, an HPLC system (Shimadzu Nexera X2) consisting of a low-pressure pump and two high-pressure pumps (LC-20AD and LC 30AD) and an automatic injector (SIL-30AC) was used. HRMS detection was performed using a TripleTOF 5600 þ SCIEX (Ontario, Canada) equipped with a DuoSpray ion source and electrospray ionization (ESI).
Chromatographic separation was performed with a Phenomenex Kinetex ECO C18 column (100 Â 2.1 mm, 2.6 μm). The column oven temperature was maintained at 40°C and the injection volume was 10 μL. The mobile phase consisted of water (aqueous phase A) and methanol (organic phase B), both buffered with 3 mM ammonium formate at a flow of 180 μL min À1 . An elution gradient of 0-8 min: 10-100% B was applied, with a total time of 15 min.
The conditions for HRMS were: source temperature of 550°C, ion source gas at 50 ua, sweep gas at 30 ua, and ionization voltage at 5,500 V (4,500 V negative). The HRMS workflow consisted of a complete scan, using 250 ms as the accumulation time and 80 V (À80 V in negative mode) as the decomposition potential in the ESI. Simultaneously, an independent acquisition approach to the data was performed based on SWATH (Sequential Window Acquisition of all Ion Mass Spectra of Theoretical Fragments). A wide mass range (80-850 Da) was divided into 30 mass windows with an accumulation time of 35 ms for each. The decay potential was adjusted to 80 V (À80 V in negative mode) and the collision energy was 40 V (À40 V in negative mode), with a power distribution of 20 V. It was operated with Analyst 1.7.1 control software, PeakView 2.2 processing software, and MultiQuant 3.0 quantification software. This study was carried out at the Centro de Investigaciones Agrarias (AGACAL-CIAM), Unidad de Contaminantes Orgánicos, in A Coruña, Spain.
Antibiotic removal efficiency
The antibiotic removal efficiency was used for the concentration of the target antibiotics in the influent and the effluent in each of the samplings, the equation proposed by Kovalova et al. (2012) was calculated. %Removal where C i is the concentration of the antibiotic in the influent and C e is the concentration of the antibiotic in the effluent.
Quality control
Procedural blanks were included in all analyses in order to avoid false-positive findings.
The linear relationship between target compounds contents and peak areas was daily checked as well as the intraday precision. Samples were analyzed in triplicate (n ¼ 3), with each replicate being produced from a subsample of the original sample.
Firstly, the quality criteria for potential compound identification were based on the exact masses of the unknown compounds, accepting only mass error below 5 ppm. However, for the unambiguous identification of these compounds, additional criteria besides exact mass were needed. Therefore, isotope profile scores and high-resolution MS/MS spectra were also taken into account for confirmation. Spectral libraries containing more than 2,000 high-resolution mass spectra were available.
The combination of the three above-mentioned criteria provided a final combined score. All results showing a combined score below 50 were excluded from the final list of identified compounds.
Antibiotics quantification by LC-MS/MS
This research provides pertinent data on the identification and quantification of various EPs present in effluent samples in an urban WWTP. Three samples were analyzed at different times of the year, and the concentrations in (ng L À1 ) of two groups of widely used antibiotics were determined sulfonamide drugs (sulfamethoxazole) and fluoroquinolones (ciprofloxacin, norfloxacin, and ofloxacin), as well as their respective removal percentage, is shown in Table 2, demonstrating that the WWTP fails to eliminate 100% of various types of contaminants, although the degree of elimination can be considered acceptable since the compounds were removed in more than 80% in most cases.
Sulfamethoxazole was detected in concentrations of 231.6-153.6 ng L À1 in the influent and 87.5-8.6 ng L À1 in the effluent with a removal efficiency of 62-94%. A similar concentration in the influent but lower removal efficiency (17%) was reported by Rosal et al. (2010), who studied an urban WWTP, located in Madrid, Spain, operated under A2O multistage configuration with nitrification-denitrification and enhanced phosphorous-accumulating microorganisms. In contrast, Estrada-Arriaga et al. (2016) found results with different concentrations (1,190 and 2,330 ng L À1 ) and determined that sulfamethoxazole was not removed, both in the rainy and the dry season, from an urban WWTP located in Guanajuato, Mexico. The WWTP is equipped with oxidation ditches in simultaneous nitrification-denitrification, designed for the removal of organic matter and nitrogen with disinfection module through UV. Estrada-Arriaga et al. (2016) also analyzed an activated sludge plant with simultaneous nitrification-denitrification, and phosphorus removal, with disinfection with chlorine dioxide and UV located in Valle de Bravo, Mexico, detecting high percentages of removal rates of sulfamethoxazole (99-96%). Another study by Rivera-Jaimes et al. (2018) showed a removal percentage of sulfamethoxazole with 36 and 53% in a WWTP of Water Quality Research Journal Vol 56 No 3,172 Cuernavaca, Mexico. The WWTP is equipped with a conventional activated sludge (CAS) treatment after aerobic and anaerobic digestion, and a tertiary treatment based on UV oxidation.
Fluoroquinolone ciprofloxacin was detected in concentrations of 2,733-1,717 ng L À1 in the influent and 494.6-444.7 ng L À1 in the effluent with a removal efficiency of 71-83%. The concentrations determined in this study are lower than those reported by Rosal et al. (2010), who archived removal of 57%. Estrada-Arriaga et al. (2016) obtained different results (21 and 89%), in the WWTP of Guanajuato and 85% to complete removal, in the WWTP of Valle de Bravo. Norfloxacin was detected in concentrations of 45.8-42.8 ng L À1 in the influent and 8.3-4.6 ng L À1 in the effluent with a removal efficiency of 82-89%. Norfloxacin was not detected in the two plants studied by Estrada-Arriaga et al. (2016). Ofloxacin was detected in concentrations of 338.9-291.5 ng L À1 in the influent and 64.4-7.1 ng L À1 in the effluent with a removal efficiency of 78-98%. The concentrations determined in this study are lower than those reported by Rosal et al. (2010), who archived removal of (64%). The percentages have different removal rates and effluent concentrations compared with the study by Estrada-Arriaga et al. (2016) that determined different removals in the rainy than the dry season, as well as different concentrations, they reported ofloxacin removal of (0 and 63%) in the WWTP of Guanajuato and complete removal in the WWTP of Valle de Bravo.
The difference in the concentration of the antibiotics studied with respect to other WWTPs in other regions shows the variation in the amount of antibiotics used in the different regions. Removal efficiencies may be mainly due to variations in activated sludge treatment. The Acapulco WWTP shows acceptable performance compared with the Madrid and Guanajuato plants, but is lower than the performance of the Valle de Bravo plant.
Identification of emerging contaminants by LC-HRMS
A non-target study of the samples was carried out to identify unknown compounds in the influent and effluent of the WWTP (Tables 3 and 4). These compounds were identified by LC-HRMS (SWATH), using libraries containing more than 2,000 highresolution mass spectra. Isotope profiles and mass accuracies were also taken into account and the combination of the three criteria provided a final combined score. All results showing a combined score below 50 were excluded from the final list of identified compounds.
Metaxalone
Muscle relaxant S1, S2, S3 S1, S3 Metoprolol β-blocker S2, S3 S1, S3 Alprenolol S2, S3 S2 S1, composite sample September 2017; S2, composite sample February 2018; S3, composite sample July 2018. Pyrethrin I Insecticide S1, S3 Promecarb Insecticide S1, S3 Pirimiphos-methyl Insecticide S1 Trimetacarb Insecticide S1 Tridemorph Fungicide S1, S2, S3 S1, S2, S3 Silthiofam Fungicide S1, S2, S3 Norflurazon Herbicide S1 Butralin Herbicide S3 Quinoclamine Algaecide S2 herbicides, and fungicides; in addition to a plasticizer and fire retardant (triphenyl phosphate). A high number of non-target compounds were identified in the effluent and especially in the influent samples, using the HRMS technique. Despite being one of the model plants in Mexico, it fails to remove 100% of various types of EPs, although in general, the degree of elimination is high, and many of the non-target compounds were only detected in the influent samples. It should be mentioned that these types of pollutants have been widely used and therefore detected in different parts of the world, and for this reason, they remain within the main lines of research by the scientific community.
Drugs removed by the WWTP Different groups of drugs were only detected in the influent samples. This is the case of caffeine and theophylline (stimulants), naproxen and ketorolac (anti-inflammatories), paracetamol (pain reliever), albendazole (dewormer), cocaine and its metabolites (illicit drugs), medroxyprogesterone acetate (hormone), erythromycin and trimethoprim (antibiotics), which could indicate they are removed by the WWTP (Figure 2). Removal percentages of some of these compounds in an activated sludge WWTP were reported by Rosal et al. (2010), who reported similar removal of paracetamol and caffeine (100 and 95%, respectively); different removals to naproxen and ketorolac (61 and 44%, respectively); and low removals of trimethoprim and erythromycin (5 and 4%, respectively). With the exception of ketorolac, albendazole, and the hormone, medroxyprogesterone acetate, all other drugs were reported in the two WWTPs studied by Estrada-Arriaga et al. (2016), who reported high removal rates of 96-100%, while the use of oxidation ditches without disinfection with chlorine dioxide showed a weakening in the elimination of antibiotics (erythromycin and trimethoprim) and cocaine and its metabolite (benzoylecgonine). Results of Rivera-Jaimes et al. (2018) showed a complete removal rate of paracetamol and high removal rates of 90-97%, in contrast, with incomplete removal of trimethoprim (0-50%).
Drugs not removed by the WWTP
Other drugs such as antibiotics, levofloxacin and nalidixic acid, anticonvulsants, carbamazepine and oxcarbazepine, lidocaine (local pain reliever), among others were detected in both influent and effluent samples, indicating that they are not
Pesticide identification
A high number of pesticides were identified in the WW samples. DEET and silthiofam were the compounds with the most intense signal. With the exception of DEET, tridemorph, and carbofuran, the other pesticides were below detection in the effluent samples, which means that they are removed in very high percentage by the WWTP (Table 4). In the case of DEET, the intensity in the influent samples is much higher compared with the effluent samples in the three sampling campaigns, which means that although it is not completely eliminated, its removal is a very high percentage (.90%) (Figure 4). Tridemorph was detected in the three sampling campaigns, and the intensities in the effluent samples were lower than in the influent samples with an average removal efficiency about 60%.
Carbofuran was detected in the influent in two campaigns (S2 and S3), but only in the effluent of campaign 3. These three pesticides are not completely removed by the WWTP.
Some of the pesticides identified in this study are included in the list of highly hazardous pesticides (HHPs) by the PAN (Pesticide Action Network International) (Neumeister & Weber 2015), due to their acute toxicity, long-term effects, environmental toxicity, and/or because their uses are regulated or prohibited by international agreements in some countries. Carbofuran and dichlorvos are classified as highly dangerous pesticides (Class Ib) according to the WHO and as deadly substances if inhaled (H330), according to the Globally Harmonized System of Classification and Labeling of Chemical Products (GHS), in addition to carbofuran being on the list of pesticides stipulated in the Rotterdam Convention. Malathion is considered a probably carcinogenic substance by the International Agency for Research on Cancer IARC (2006). Tridemorph is listed as a toxic substance for human reproduction in the European Union (EC 2008). The pesticide pirimiphos-methyl is considered HHPs according to the list of PAP due to its high degree of environmental toxicity, since it is very toxic to bees, essential in pollination.
Trimetacarb (trimethacarb, landrin) is an insecticide without EU's regulatory approval for use. It is moderately soluble in water, very volatile and there is also a slight risk of leaching to groundwater. In general, it is not persistent in soil systems; however, it can degrade rapidly by hydrolysis in aquatic systems. It should be mentioned that it is moderately toxic to mammals. The carbamate pesticides trimethacarb were detected in a non-target analysis of the dryer lint samples collected from two households in the Netherlands (Ouyang et al. 2017).
CONCLUSION
The data presented above show the presence of EPs in the effluent of the WWTP studied, which represents a priority problem at present in Mexico, as well as worldwide. Most of the research comes from developed countries, because they have the necessary economic resources and infrastructure. However, water contamination also exists in developing countries such as Mexico, which is why they are considered to be pollutants that must be continuously monitored even though they are not regulated by Mexican Standards.
Removal rates for the target compounds (sulfamethoxazole, ciprofloxacin, norfloxacin, and ofloxacin) were high, as well as for most of the non-target compounds. However, some of the identified compounds, mainly drugs, are not effectively eliminated by the urban WWTP, being discharged into the effluent (e.g. antibiotic levofloxacin, anticonvulsant carbamazepine, beta-blocker metoprolol). On the other hand, among the non-target compounds, pesticides were identified, some of which are not eliminated by the WWTP (e.g. tridemorph).
Considering that the discharge of the studied WWTP is located in a world-renowned tourist destination, and its discharge is done near the tourist beaches, the treatment conditions must be improved by applying efficient tertiary treatments in the removal of the different contaminants detected, to improve the quality of treated water and reduce contamination by EPs, which would allow it to be safely reused for recreational activities. | 2021-09-09T20:37:52.347Z | 2021-08-01T00:00:00.000 | {
"year": 2021,
"sha1": "6353f53b75d9c2a8185d538cb31aaf355429bd22",
"oa_license": "CCBY",
"oa_url": "https://iwaponline.com/wqrj/article-pdf/56/3/167/922141/wqrjc0560167.pdf",
"oa_status": "HYBRID",
"pdf_src": "Adhoc",
"pdf_hash": "117f43e71e571d28b2866d2daf246b95dfbda4f9",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": []
} |
86762360 | pes2o/s2orc | v3-fos-license | L-glutamine and L-glutamate in diets with different lactose levels for piglets weaned at 21 days of age
This study was developed to evaluate the effects of adding L-glutamine and L-glutamate in diets with no lactose or with 4.0 or 8.0% lactose inclusion for piglets weaned at 21 days old. One hundred and eight piglets with initial weight of 6.12±0.70 kg were allotted in a complete randomized block design, in a 3 × 2 factorial arrangement, with six treatments, six replicates, and three piglets per experimental unit. The experimental diets were supplied from 21 to 35 days. From 36 to 49 days, animals received the same diet with no lactose inclusion, but the animals in the treatments with L-glutamine + L-glutamate in the previous phase continued to receive diets containing these ingredients. There was no interaction between the level of lactose and the inclusion of L-glutamine + L-glutamate on the parameters evaluated. The levels of lactose did not affect the performance of piglets in either of the two periods. Adding L-glutamine and L-glutamate in the diet positively influenced the weight gain of pigs from 21 to 49 days of age and increased the villous height in the duodenum, jejunum and ileum. Inclusion of L-glutamine + L-glutamate in diets for piglets weaned at 21 days of age improves the performance and the intestinal mucosa morphology, regardless of lactose addition.
Introduction
The association of all the stressors related to early weaning of piglets results in reduction in daily feed intake, especially during the first week.The interruption of feed intake in this phase results in intestinal villous atrophy and increased crypt depth (Pierce et al, 2005).
According to Pinheiro & Machado (2007), the performance of piglets in the first week after weaning is strongly correlated with the weight at the end of the nursery phase, and the number of days required to reach slaughter weight.Some nutritional strategies for piglets at this stage have been taken to mitigate the harmful effects of weaning, including the use of lactose, whose benefit is mainly linked to the increase and maintenance of feed consumption due to the improved palatability of the feed (Bertol et al., 2000).However, diets containing lactose have their cost increased due to the high price of this ingredient, which is also used in human food.Thus, it is necessary to look for ingredients that, like lactose or by different mechanisms, can minimize the detrimental effects of early weaning.
The amino acid glutamine constitutes an ingredient with the potential to improve the performance of piglets through its action in the maintenance of the small intestine mucosal structure (Yi et al., 2005), in addition to its role as precursor of proteins, nucleotides and polyamines (Sources et al., 2007).The action of this amino acid, termed conditionally essential, is related to the multiplication of cells of intense development, such as enterocytes of the intestinal mucosa and immune cells (Abreu & Donzele, 2008;Wu et al., 1996).
Thus, there is a need to evaluate the effect of adding L-glutamine and L-glutamate in diets containing different levels of lactose on the performance and intestinal mucosa structure of piglets weaned at 21 days.
Material and Methods
The experiment was conducted in the Setor de Suinocultura do Departamento de Zootecnia da Universidade Federal de Viçosa.One hundred and eight hybrid piglets weaned at 21±1 day of age and weighing 6.12±0.70kg were allotted in a completely randomized block design with six treatments, six replicates, and three piglets per experimental unit.Weaning weight and piglets' kinship were considered as criteria in the blocks formation.Treatments consisted of diets containing 0.0, 4.0, and 8.0% lactose, supplemented with 0.8% L-glutamine + L-glutamate (L-L-Glu + Gln), from 21 to 35 days.From 36 to 49 days, all piglets were fed the same diet free of lactose, and those that were fed 0.80% L-glutamine + L-glutamate from 21 to 35 days received 0.60% L-glutamine + L-glutamate.The experimental diets from 21 to 35 days (Table 1) and from 35 to 49 days (Table 2) were formulated to meet or exceed the requirements of pigs at the pre-starter phase (7 to 15 kg) for all nutrients (Rostagno et al., 2005).
Piglets were housed in suspended metal cages with slatted floors, semi-automatic feeders and nipple drinkers, located in a facility with concrete floor, wood ceiling and covered with ceramic tile.The room temperature was kept at 28 °C, using hoods and/or heaters when needed.
For evaluation of performance, piglets were weighed at the beginning (35 days) and end of the experiment (49 days), and feed leftovers were collected and weighed daily when the animalsreached 35 and 49 days, i.e., at the end of each period, for calculation of average daily feed intake (ADFI), average daily gain (ADG) and feed conversion (FC).
At 49 days of age, one piglet from each experimental unit, with body weight closest to the average weight of the cage was slaughtered.Samples of 2 cm in length were taken from the regions corresponding to 4, 50 and 96% of the small intestine, corresponding respectively to the duodenum, jejunum and ileum.The collected material was sent to the Laboratório de Histologia do Departamento de Medicina Veterinária da Universidade Federal de Viçosa.Histological cuts were washed in physiologic solution, fixed in bouin for a 24-h period, dehydrated in ethyl alcohol, diaphanized in xylol, and included in paraffin.Each slide contained two 5 μm thick cuts.Slides were put back in xylol solution for removal of paraffin excess and then rehydrated.The pigments 1 Lac 8 = 8% lactose, Lac 4 = 4% lactose, Lac 0 = 0% lactose, L-Gln + L-Glu = L-glutamine + L-glutamate.
On the day of slaughter, samples of intestinal content of pigs in the colon proximal were collected in sterile containers and frozen immediately.Samples were subsequently sent to the Laboratório de Microbiologia Industrial do Departamento de Microbiologia da Universidade Federal de Viçosa.The total DNA extraction of the digesta was done by following the protocol described by Lu et al. (2003) with some modifications made in the Laboratório de Microbiologia Industrial do Departamento de Microbiologia da Universidade Federal de Viçosa.To amplify the rDNA 16S gene, corresponding to total Eubacteria, universal primers of Eubacterias with the following sequences of nucleotides were used: primer 2: 5'ATTACCGCGGCTGCTGG3' and p r i m e r : (Muyzer et al., 1993).To amplify the rDNA 16S region corresponding to Lactobacillus spp.specific primers with the sequences were used as follows: The intestinal microbiota genetic profile was assessed by the electrophoresis technique in gel with gradient denaturant -DGGE (Muyzer et al., 1993).The data obtained by using the program Bionumerics, version 5.10, were used to calculate the richness, measured in operational taxonomic units (OUT), the Shannon-Weaver diversity (H') and equitably (E') indexes, and with similarity corresponding to the diets containing different lactose levels.
The variable R was generated using a binary matrix in which the absence of DNA band corresponding to each OUT was coded as zero (0) and the presence as one (1).For each level of lactose, the frequency of occurrence of each OUT was obtained by counting the number of presences in the treatments.The index H 'was calculated based on the abundance, expressed as optical density of the band in the lane of the gel, using the formula: , where N = optical density in the lane, obtained by the sum of the bands optical densities in the lane; ni = optical density of each band in the lane.
E' was calculated as: The performance and morphological intestinal mucosa data obtained, corresponding to the pigs fed diets containing lactose supplemented or not with L-glutamine + L-glutamate, were subjected to analysis of variance and F test at 5% probability, using the program SAEG (Statistical Analysis System and Genetics, version 8.1), developed at the Universidade Federal de Viçosa.In the absence of interaction between the factors studied, the study proceeded to the analysis of the factors alone.
The evaluation of the composition of the intestinal microbiota of piglets was carried out using the program Bionumerics, version 5.10.
Results and Discussion
There was no interaction (P>0.05) between the factors studied on any parameters evaluated.
Adding L-glutamine + L-glutamate in the diets did not affect (P>0.05) the performance of pigs from 21 to 35 days of age (Table 3).However, in absolute values, the inclusion of L-glutamine + L-glutamate resulted in an increase of 4.6, 11.1, and 7.9% in the body weight at 35 days, average daily gain (ADG), and average daily feed intake (ADFI) of piglets, respectively, followed by a reduction of 3.2% in feed conversion (FC), also in absolute values.
From 21 to 49 days of age (Table 3), there was effect (P>0.05) of L-glutamine + L-glutamate on body weight at 49 days and on the body weight gain of piglets which increased 7.1 and 10.2%, respectively, compared with those fed diets without L-glutamine + L-glutamate.The ADFI and FC in this phase were not influenced by adding of L-glutamine + L-glutamate in the diets.
Beneficial effect of adding L-glutamine + L-glutamate in the diet on piglet performance was observed by Wu et al. (1996), Kitt et al. (2003) and Zou et al. (2006).Assessing four levels of glutamine supplementation (0, 0.2, 0.6, and 1.0%) on the performance of weaned piglets analyzed at weaning and in the first and second subsequent weeks, Wu et al. (1996) found a significant improvement in the order of 25% in the feed efficiency of piglets fed glutamine compared with those fed the control diet from 7 to 14 days after weaning.
In diets for weaned piglets challenged with LPS (E. coli lipopolysaccharide), Kitt et al. (2003) found that the inclusion of 5% glutamine maintained satisfactory the performance of the animals, which was not seen in those challenged and non-supplemented.In studies with piglets weaned at 21 days, Zou et al. (2006) found improvement in the FC within 10 days after weaning and for ADG within 20 days after weaning in response to the inclusion of 1% glutamine in the diet, followed by reduced rate of diarrhea and increased serum levels of growth hormone.
The inclusion of L-glutamine + L-glutamate resulted in an increase (P>0.05) in the villi height of the three segments of the small intestine evaluated, which were 7.7, 10.4 and 6.6% higher, respectively, in the duodenum, jejunum and ileum than those presented by piglets that were not supplemented with L-glutamine + L-glutamate (Table 4).However, the inclusion of L-glutamine + L-glutamate did not affect (P>0.05)crypt depth in any of the intestinal segments, and the values found in both groups of animals were very close to each other (Table 4).
The villous:crypt ratio, which is the result of the division between the measurement of villus height and the depth of its respective crypt, in the ileum was positively affected (P>0.05) by inclusion of L-Glu + L-Gln in the diet, evidently by the increasing in villi height (Table 4).Piglets fed diets containing L-glutamine + L-glutamate in their composition showed yet, in absolute values, villous:crypt ratio in the duodenum and jejunum 2.1 and 7.3% higher compared with those whose diets did not contain L-glutamine + L-glutamate.These results confirmed those found by Wu et al. (1996) and Yi et al. (2005).
Studying the effect of glutamine on intestinal villus height of piglets after weaning, Wu et al. (1996) reported that in addition to substantial improvement in performance, the inclusion of 1.0% glutamine in the diets of the piglets was able to prevent villus height atrophy of the jejunum in both first and the second week after weaning.Comparing the effect of adding glutamine or plasma in diets for piglets weaned at 17 days of age and challenged with Escherichia coli, Yi et al. (2005) found that the inclusion of 2% glutamine or 7% plasma significantly increased villus height and crypt depth of piglets with regard to the non-supplemented challenged treatment, although the inclusion of glutamine provided better performance.The authors attributed the beneficial effect of glutamine, mainly to the maintenance of the intestinal mucosa structure, electrolyte absorption, and intestinal homeostasis preventing diarrhea.
Comparing the effect of adding glutamine or plasma in diets for piglets weaned at 17 days of age and challenged with Escherichia coli, Yi et al. (2005) found that the inclusion of 2% glutamine or 7% plasma significantly increased villus height and crypt depth of piglets concerning the non-supplemented challenged treatment, although the inclusion of glutamine provided better performance.The authors attributed the beneficial effect of glutamine mainly to the maintenance of the intestinal mucosa structure, electrolyte absorption, and intestinal homeostasis preventing diarrhea.
The nutrition of piglets is closely linked to the immune system.It is known that early weaning is related to immunosuppression, reduced feed intake, and susceptibility to intestinal disorders.According to Newsholme (2001), the increased demand for glutamine in conditions of stress increases the release of glutamine by the muscle, resulting in decreased concentration of this amino acid in the muscle, and elevated rates of protein degradation, which is not desirable in growing pigs.
In a study with piglets infected or not by Escherichia coli, Yoo et al. (1997) found that supplementation with glutamine was able to maintain the concentration of glutamine in the muscle, and the concentrations of lymphocytes and leukocytes.However, the authors reported no effect of supplementation with glutamine on the nonchallenged piglets, suggesting that supplementation with glutamine is important, among other factors, for the maintenance of its extracellular concentration, preventing the mobilization of muscle glutamine to meet the needs of the immune system.
According to Karinch et al. (2001), moderate or severe infections change the flow of glutamine in the body.The complexity of glutamine metabolism in different sanitary conditions of piglet production may explain why, in this research, even no effect (P> 0.05) of lactose was detected on the performance (Table 5) nor on the intestinal mucosa structure (Table 6); the inclusion of glutamine in the diet provided significant improvement in the ADG from 21 to 49 days of age and in the villi height.In general, the animals have good health, indicating low level of health challenge.Under these conditions, the small intestine is the organ of highest uptake of glutamine in the body, and it is provided primarily by the muscle (Karinch et al., 2001), which would explain the increase in villus height and, consequently, the increase in the ADG.
Under conditions of systemic inflammatory response, such as induction of immune challenge observed in the study of Kitt et al. (2003), Yi et al. (2005), and Yoo et al. (1997) the liver becomes the most important organ of glutamine uptake and exportation of this amino acid for the synthesis of immune cells due to the increase in the production of these specific cells (Karinch et al., 2001).
The effect of L-glutamine + L-glutamate in this study was self-evident although no inflammatory response was induced in the piglets.In addition to the low health challenge reflected by the low number of piglets suffering from diarrhea during the experiment and the overall health of the piglets, the environmental condition of sanitary challenge typically found on the first weeks after early weaning was attenuated by the good sanitary conditions of the rooms and by adding blood plasma and antibiotics in the experimental diets.The results obtained by Abreu et al. (2010), in a study evaluating the combination of glutamine, nucleotides and plasma in diets for weanling pigs showed that the response of piglets to plasma, the association of plasma + nucleotides or plasma + glutamine supplanted the response in performance afforded by glutamine alone, suggesting, furthermore, that the inclusion of plasma may attenuate the inflammatory response of piglets at weaning.
The levels of lactose did not influence (P>0.05)performance (Table 5) or intestinal mucosa structure of piglets (Table 6).
These results do not confirm the findings of Hauptili et al. (2005), who observed improved performance of piglets weaned at 21 days of age during the nursery phase by increasing the level of inclusion of whey powder in the diets, which was attributed to the lactose present in the food.
The lactose level in diets for piglets weaned at 21 days recommended by Bertol (1999) is 12 to 14% on the first week.In addition, Ludke et al. (1998) point out that in feeding programs for young piglets the need of milk powder inclusion in the diet rises as early as weaning is carried out, which is confirmed by Mahan et al. (2004), who worked with piglets weaned at different ages.
However, the effect of the addition of lactose in the diet of piglets may vary depending on the environment in which they are housed, as we can see in the study of Cromwell et al. (2008), who found, in three different experimental stations, 1 Not significant at 5% probability.However, part of dietary lactose may escape enzymatic digestion in the stomach and serve as a substrate for intestinal microbial fermentation.Delcenserie et al. (2008) accurately described the variety of physiological effects brought about by some genera of microorganisms, particularly regarding the production of inflammatory cytokines.Therefore, the existence of a relationship between the level of lactose inclusion in the diet of piglets and the intensity of sanitary challenge are feasible to be verified, given the importance of the study in the intestinal microbial diversity.
Operational Taxonomic Units (OTUs) quantitatively represent the microbial population of a given environment (Figure 1).
It was observed that the level of 4% inclusion of lactose in the diet provided the highest counts of UTOs, both in non-supplemented (29 UTOs) and supplemented diets (27 UTOs) with L-glutamine + L-glutamate.Regarding the community of Lactobacillus spp. of piglets fed diets with no addition of L-glutamine + L-glutamatethe increase in the lactose level from 0 to 8% propitiated a crescent increase in the UTOs count.
However, the pattern of response to the inclusion of these amino acids in the diet for the amount of Lactobacillus spp. was not maintained as the diet with 4% lactose and L-glutamine + L-glutamate resulted in the lowest value of UTO's (11).
Similarly, the Shannon-Weaver Diversity Index (H') (Figure 1) followed the results of Richness (R).Considering only the community of Eubacteria, the level of 4% lactose resulted in higher values of UTOs being verified by H' equal to 3.01 and 2.92, respectively, for piglets fed both supplemented and non-supplemented diets with L-glutamine + L-glutamate.
The Equitability (E') is indicative of uniformity in the distribution of groups of microorganisms found in an environment, demonstrating a possible existence of dominance (Figure 1).The higher the E', the higher the balance between populations.The results concerning Eubacteria consistently indicated that 0% lactose inclusion resulted in the lowest values of E' in both non-supplemented (1.98) and supplemented (2.01) diets with L-glutamine + L-glutamate, which allows to infer about a likely dominance exerted by a particular microorganism in relation to others.Regarding the groups of Lactobacillus spp., there was an increase in E' when the level of lactose was increased in the piglets fed both non-supplemented and supplemented diets, which was 2.22 and 2.18, respectively.
With regard to the Richness and intestinal microbiota diversity of piglets, although the association of L-glutamine + L-glutamate with 4 and 8% lactose did not follow the same pattern of response for Lactobacillus spp. of piglets fed non-supplemented diets, it is not possible to say that the inclusion of L-glutamine + L-glutamate influenced the composition of the microflora.
Industrial amino acids, in general, have high digestibility, being fully used until the ileum.Stool et al. (1998) postulate that glutamine is degraded along the entire small intestine, which would not allow its fermentation by bacteria in the large intestine, where samples of digesta were collected.Collectively, the results suggest that the inclusion of lactose had a beneficial effect on the microbiota and the supplementation with L-Glu + L-Gln had no influence on the microbiota.The electrophoresis in denaturing gradient gel (DGGE) shows the profiles of similarity among treatments contributing to the consolidation of the results of H', E' and R (Figure 2).The occurrence of a high similarity in the intestinal microbiota of piglets that were fed diets containing the same levels of lactose (T1 and T2 = 8%, T3 and T4 and T5 = 4%, and T6 = 0% lactose) was verified for the community of Eubacteria, regardless of supplementation with L-glutamine + L-glutamate.
The similarity observed between treatments with the same level of lactose was 68.0, 76.4, and 63.8, respectively, for the levels 0.4 and 8% lactose.However, it was observed that the treatments with 0 and 4% lactose, with or without L-glutamine + L-glutamate inclusion presented average similarity, lower than that seen between treatments with the same level of lactose.The similarity is indicative of the resemblance between microbial communities in microecosystems, in which case the digesta of animals is strongly influenced by the diet (Williams et al. 2001).
The analysis of intestinal microbiota consistently indicated that the level of lactose exerted influence on the total bacterial community in the colon of piglets, regardless of the presence of L-glutamine + L-glutamate in the diets.
Conclusions
The inclusion of L-glutamine + L-glutamate in diets for piglets weaned at 21 days of age improves performance and morphology of the intestinal mucosa, regardless of the addition of lactose.The level of 8% lactose improves the composition of the intestinal microbiota of piglets.
Table 1 -
Composition and nutritional values of experimental diets fed to piglets from 21 to 35 days of age
Table 2 -
Composition and nutritional values of experimental diets fed to piglets from 35 to 49 days of age
Table 3 -
Effect of L-glutamine + L-glutamate (L-Gln + L-Glu) on performance of piglets weaned at 21 days of age, assessed from 21 to 35 and from 21 to 49 days of age 1 Significant at 5% probability.CV = coefficient of variation.
Table 4 -
Effect of L-glutamine + L-glutamate (L-Gln + L-Glu) on the intestinal mucosa structure of piglets weaned at 21 days of age, assessed at 49 days of age
Table 6 -
Intestinal mucosa structure of piglets fed diets with lactose, assessed at 49 days of age 1 2 0% lactose from 21 to 35, and 0% lactose from 36 to 49 days of age. 3 4% lactose from 21 to 35, and 0% lactose from 36 to 49 days of age. 4 8% lactose from 21 to 35, and 0% lactose from 36 to 49 days of age.CV = coefficient of variation; ADG = average daily gain; DFI = daily feed intake; FC = feed conversion.
Table 5 -
Performance of piglets fed diets with lactose, assessed from 21 to 35 and from 21 to 49 days of age 1 | 2019-03-28T13:42:11.508Z | 2012-01-01T00:00:00.000 | {
"year": 2012,
"sha1": "1543bd264719e86bf3ec2c4499473cce9ca7a4d1",
"oa_license": "CCBY",
"oa_url": "https://www.scielo.br/j/rbz/a/7RmDX6Mc54B3vChGWbJFkSS/?format=pdf&lang=en",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "1543bd264719e86bf3ec2c4499473cce9ca7a4d1",
"s2fieldsofstudy": [
"Agricultural and Food Sciences"
],
"extfieldsofstudy": [
"Biology"
]
} |
203299913 | pes2o/s2orc | v3-fos-license | The Pseudo-DF Approach for Learning Huge-scale Data
As the advent of the big data era, huge-scale data continuously appears in various fields of science, commerce, industry and society. More algorithms/methods/approaches are urgently required to learn huge-scale data collected from different applications/backgrounds. Therefore, the Pseudo Data Flow (Pseudo-DF) approach with ensemble ReOS-ELMs is proposed in this paper. The Pseudo-DF approach randomly divides a huge-scale data set into K (K>1) non-overlapping data chucks, and a Pseudo-DF is constructed by these data chucks. The computation of a huge-scale data is changed into that of a Pseudo-DF with smaller-scale chucks, the computational burden will be much reduced. Then, the ensemble Regularized OS-ELMs (ReOS-ELMs) based on Different random Hidden-node Parameters (DiffHPs) is presented to learn a Pseudo-DF, which is a recursive leaning algorithm possessing the advantages of low computational burden, high accuracy, well generalization and stability, and strong robustness. Lastly, experiments are performed to validate the effectiveness of the proposed approach.
Introduction
As the advent of the big data era, the scale of data enlarges rapidly and dramatically with the advancement of data acquisition and storage techniques. Large-scale data becomes commonly and huge-scale data continuously appears in various fields of science, commerce, industry and society [1]. Huge-scale data contains abundant information which is helpful to explain phenomena, discove laws, predict trends, and so on. By now, many algorithms/methods/ approaches aiming at large-scale data has emerged [2], and reversely those oriente to huge-scale data is much fewer. Moreover, huge-scale data collected from different applications/backgrounds inevitably contain unlike features. More algorithms/methods/approaches are urgently required to mine and learn huge-scale data. Therefore, this paper focuses on the study of learning approach for huge-scale data.
A natural idea to handle a huge-scale data set is to divide it into many smaller-scale chucks (sets), and then the computational burden of learning a huge-scale data set will be much reduced by learing these smaller-scale chunks. However, how to guarantee and even boost the utilization ratio of the information hidden in huge-scale data? That is, how to guarantee and even improve the generalization of the model learned on smaller-scale chunks is the most basic challenge. In order to meet this challenge, two key steps should be well considered, which are the division criterion and the information fusion of smaller-scale chucks.
On the basis of this idea, this paper proposes the Pseudo Data Flow (Pseudo-DF) approach to learn huge-scale data. In Pseudo-DF approach, a huge-scale data set is randomly divided into K (K>1) nonoverlapping data chucks, and a Pseudo-DF is constructed by these data chucks. Then, a recursive leaning algorithm is adopted to learn the Pseudo-DF, by which the information of data chucks can be transferred from the original generation of model into the (K-1)-th generation of model via updating. The last generation of model is the final model learning from a Pseudo-DF of a huge-scale data set.
Because a Pseudo-DF has the time-insensitive characteristic, and recursive leaning algorithms without forgetting mechanism should be adopted to learn it. Online Sequential Extreme Learning Machine (OS-ELM) [3] is firstly considere. With the advantages of low computational burden and strong nonlinear-fitting ability, OS-ELM can fast and accurately learn large-scale data flows [4,5]. However, OS-ELM is perplexed by the instability from random Hidden-node Parameters (HPs), becomes failed by singular and ill-posed problems, and yields bad generalization when learning on data flows with noise. In order to improve the stability of OS-ELM from random HPs, Lan et al. presented Ensemble OS-ELMs (EOS-ELMs) [6] which builds several OS-ELMs with the ensemble mode of Different random Hidden-node Parameters (DiffHPs) and averages/votes their outputs to obtain ensemble outputs. Liu et al. indicates that ensemble learning also can boost the robustness of OS-ELM from noise, and they combine particle swarm optimization selective ensemble method with OS-ELM [4]. Moreover, many scholars hold that the weak robustness of OS-ELM learning on noisy data flows is derived from its basic criterion of empirical risk, and then they added the consideration of structural risk to OS-ELM [7]. Regularized OS-ELM (ReOS-ELM) is a typical exemplar, which minimizes both the norm of the matrix of empirical error and the norm of the matrix of hidden-layer weights [7]. ReOS-ELM can also avoid singular and ill-posed problems.
Therefore, an ensemble ReOS-ELMs method is presented to learn the Pseudo-DF of a huge-scale data set. The ensemble mode of DiffHPs is applied to build diverse ReOS-ELM sub-models, and a weight average (or a vote) method is used to integrate sub-models. To validate the effectiveness (i.e., generalization and interpretability) of the proposed Pseudo-DF approach with ensemble ReOS-ELMs, experiments are carried out on one huge-scale data sets from UCI machine learning repository, and it related to a regression task.
ReOS-ELM
Suppose that there are N training samples (x n ,t n ), n=1,..,N, where t n is the target corresponding to the input vector x n , and x n =[x 1 ,x 2 ,…,x M ]. M is the dimension of input attributes. The mathematical description of an ELM model is as follows [8], where β is the matrix of hidden-layer weights, and β=[β 1 , , J is the number of hidden nodes. h is the vector of hidden-layer outputs, h j = G(a j ,b j ,x), j=1,…,J where G(.) is the activation function, and (a j ,b j ), are randomly generated Hidden-node Parameters (HPs). Given the activation function (e.g, "RBF", "Sigmoid", or "Sine"), and random HPs (i.e., a and b.), the goal of training is to estimate β. The estimation of β from ReOS-ELM is the solution to minimize 22 Hβ T β , where H is the matrix of hidden-layer outputs, and H=[h 1 ;h2;…,h N ]. T is the target vector, and T=[t 1 , t 2 ,…, t N ] T . λ is the Regularization Factor (RF) and λ≥0. The solution for β is given by Given an initial chunk of data 0 Ω . The solution to minimizing where When the k-th chunk of data k Ω is received, we have where
Text The Pseudo-DF Approach with Ensemble ReOS-ELMs
The mathematical description of a regression ensemble model is given by (6 ) , where f i (x), i=1,2,…,P, represents P members of ensemble, i.e., P sub-models, and ω i is the weight corresponding to f i (x). The mathematical description of a classification ensemble modelis given by (7 ) , where sign(.) is the sign function.
Algorihtm 1 The Ensemble ReOS-ELMs Algorithm
λ (λ≥0) is the RF of ensemble ReOS-ELMs. P is the number of ReOS-ELMsub-models and K is the number of data chunks. Ω is a huge-scale data set with N samples. The N samples are randomly ordered and then divided into K (K>1) non-overlapping data chucks 0 1 1 , ,..., K Ω Ω Ω which constructed a Pseudo-DF. N 0 , N 1 ,…,N K-1 denote the number of samples in 0 1 1 , ,..., K Ω Ω Ω , respectively. An ensemble ReOS-ELMs model will be built on a Pseudo-DF, the algorithm of which is presented in Algorithm 1. In line 5, the ensemble mode of DiffHPs is applied to build diverse ReOS-ELM sub-models, which enable ensemble ReOS-ELMs with parallel structure and sub-models to learn and update on different computers. That is, line 5~7 (and line 15~16) can be performed independently. In line 9 and line 18, the combination rule defined by Perrone & Cooper [9] is used to calculated the weighs of sub-models, and the calculation process of it is rewrite in [10], pages 771-772.
Experiments
The main goal of experiments was to validate the generalization and interpretability of the proposed Pseudo-DF approach with ensemble ReOS-ELMs for the learning of huge-scale data. Experiments involved two parts. One was to investigate if the model learning on a huge-scale data set by the Pseudo-DF approach was with high generalization and interpretability. The other was to investigate if ensemble ReOS-ELMs outperformed OS-ELM, ReOS-ELM and EOS-ELMs when learning a Pseudo-DF of huge data.
SGEMM GPU kernel performance data set related to a regression task was used in this sub-section [11]. This huge-scale data set has 241600 samples with 10 independent input attributes and 1 target attributes. There were no missing values. The input attributes of the data set were in integer values. The range of target values was between 13.25ms and 3397.08ms, and a target value was transformed into its logarithm form. A training set was random divided into 120 data chunks which constructed a Pseudo-DF. The huge-scale data set would be randomly divided into a training set and a testing set with the proportion of 9:1. A training set was to construct a Pseudo-DF and to learn a model, and a testing set was to validate the model. The data would be normalized in the interval [0,1] before entering a model, and the outputs would be reversed after existing the model. The "Sigmoid" activation function was used by all of four recursive leaning algorithm. The evaluating criterion of performance is Root Mean Square Error (RMSE). The random division of the huge-scale data set into a training set and a testing set was repeated 10 times and the reported RMSE was the averages over the 10 iterations, i.e., statistical results.
In any ReOS-ELM model, J was also set as 1300. To select a suitable value of λ for an ensemble ReOS-ELMs model, five nominated values of λ were investigated. The curves of the testing RMSEs of ensemble ReOS-ELMs models when λ=0, 0.1, 0.5, 1 or 2 in each ReOS-ELM sub-model were illustrated in Fig.1. The first point of each curve showed the testing RMSE of a single ReOS-ELM model. As the number of ReOS-ELM sub-models increasing, the testing RMSE gradually reduced and converged to a fixed RMSE. It was indicated that the generalization of an ensemble ReOS-ELMs model were higher than those of a single ReOS-ELM model.
The curves showed that when λ=0.5, the ensemble ReOS-ELM model obtained the lowest RMSE. Thus, it is useful to set the value of λ as 0.5 for an EOS-ELM model of the Pseudo-DF derived from SGEMM GPU kernel performance data set. Moreover, EOS-ELMs was a special case of ensemble ReOS-ELMs when λ=0. It was indicated that ensemble ReOS-ELMs outperform EOS-ELMs for building a model of the Pseudo-DF derived from a huge-scale data set. The comparisons of ensemble ReOS-ELMs, OS-ELM, ReOS-ELM, EOS-ELMs and on the testing set of SGEMM GPU kernel performance data set were displayed in Table 1. The results show that the model of a huge-scale data set learning by the Pseudo-DF approach had high generalization and interpretability. Among the four recursive leaning algorithms, the testing RMSE of the ensemble ReOS-ELMs model were lowest. The performance of an ensemble (i.e., EOS-ELMs or ensemble ReOS-ELMs) model outperformed that of single models (i.e., ReOS-ELM and OS-ELM).
Conclusion
The main contribution of this paper is that the Pseudo-DF approach with ensemble ReOS-ELMs for learning huge-scale data is proposed. The random and non-overlapping division criterion is applied to a huge-scale data set for generating a Pseudo-DF. As a more stable and robustness recursive leaning algorithm, ensemble ReOS-ELMs is presented to learn the Pseudo-DF, which can realize the information fusion of smaller-scale chucks via updating models, that is, the information of data chucks can be transferred from the original generation of model into the last generation of model. The results showed that the model of a huge-scale data set learning by the Pseudo-DF approach had high generalization and interpretability, which was satisfactory. The results also showed that the generalization of ensemble ReOS-ELMs outperforms that of EOS-ELMs, ReOS-ELM and OS-ELM for a regression. | 2019-09-17T02:47:02.484Z | 2019-08-01T00:00:00.000 | {
"year": 2019,
"sha1": "c00968d1959182a8252c1e9fe202f93428a3fdc6",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1742-6596/1302/2/022088",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "6acfa85b9789f9580fc459116945e9bad63614e4",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
235821194 | pes2o/s2orc | v3-fos-license | Compressor fault diagnosis system based on PCA-PSO-LSSVM algorithm
On the basis of the principal components analysis-particle swarm optimization-least squares support vector machine (PCA-PSO-LSSVM) algorithm, a fault diagnosis system is proposed for the compressor system. The relationship between the working principle of a compressor system, the fault phenomenon, and the root cause is analyzed. A fault diagnosis model is established based on the LSSVM optimized using PSO, the compressor fault diagnosis test experimental platform is used to obtain the fault signal of various fault occurrence states, and the PCA algorithm is employed to extract the characteristic data in the fault signal as input to the fault diagnosis model. The back-propagation neural network, the LSSVM algorithm, and the PSO-LSSVM algorithm are analyzed and compared with the proposed fault diagnosis model. Results show that the PCA-PSO-LSSVM fault diagnosis model has a maximum fault recognition efficiency that is 10.4% higher than the other three models, the test sample classification time is reduced by 0.025 s, the PCA algorithm can effectively reduce the input dimension, and the PSO-LSSVM fault diagnosis model based on the PCA algorithm for extracting features has a high recognition rate and accuracy. Therefore, the proposed fault diagnosis system can effectively identify the compressor fault and improve the efficiency of the compressor.
Introduction
Air compressors are the key equipment for coal mining engineering to provide power for pneumatic tools. 1 However, under full or over working loads and the harsh working environment, various failures will occur under the influence of numerous influencing factors during the work process. Moreover, the relationship between failures and the cause show a strong nonlinearity, and mathematical models cannot be used to express the faults in the diagnosis of mining air compressors. Thus, early warning of failures must be achieved based on the full analysis of huge data and information to discover existing risks in a timely manner. Only in this way can we ensure the stable and healthy development of intelligent coal mining enterprises. [2][3][4] Zhou et al. 5 proposed an online compressor liquid floodback fault diagnosis method for the VRF system based on the back-propagation neural network (BPNN), which fills the online compressor liquid floodback fault diagnosis knowledge gap. Van Tung et al. 6 proposed a hybrid deep belief network to diagnose the single and combined faults of suction and discharge valves in a reciprocating compressor. Zhang et al. 7 proposed a novel convolutional deep belief network-based method and employed a novel framework that fuses multi-source information to improve the performance of fault diagnosis. Data were collected from an industrial plant to verify the proposed method. De Cooman et al. 8 proposed a transformer fault diagnosis method based on the modified gray wolf optimization algorithm and support vector machine (SVM), realizing the optimization of the penalty factor, and the kernel parameter in SVM. Li et al. 9 proposed a multiscale local feature learning method based on BPNN for rolling bearings fault diagnosis. Based on the local characteristics of the fault features in the time and frequency domains, the BPNN is used to locally learn meaningful and dissimilar features from signals of different scales, thus the accuracy of fault diagnosis is improved obviously. Zhang et al. 10 proposed a fault diagnosis method based on multi-classification relevance vector machine for high voltage circuit breakers. On the basis of the ASHRAE RP-1043 data, Fan et al. 11 extracted information from three of the factory-installed (FI) sensors along with information from all eight of the FI sensors in order to establish the SVM-3 and SVM-8 diagnostic models based on grid search and cross validation parameter optimization for seven typical faults. Wang et al. 12 presented the results of a research on hybrid fault diagnosis techniques that utilize SVM and improved particle swarm optimization (PSO) to perform further diagnosis based on qualitative reasoning through knowledge-based preliminary diagnosis and the sample data provided by an on-line simulation model.
The above fault diagnosis model for compressors is mainly for equipment on the ground. Many machinery and equipment underground in coal mines will also use compressors for auxiliary operations. The underground environment makes the compressors used differ from conventional ground compressors, and the occurrence of accidents in the event of failure will cause major safety hazards. Now, more advanced deep learning and other fault diagnosis models require numerous experimental test samples. Given the complexity of the working environment of mining machinery and the equipment failure rate, the amount of collected data fails to meet the large sample data requirements, so some advanced fault diagnosis models cannot be used for fault diagnosis in this condition. Therefore, this study aims to effectively reduce the latitude of the sample input of the fault diagnosis model based on the principal components analysis (PCA) algorithm, increase the recognition efficiency and sample classification time of the PSO-LSSVM algorithm, establish fault diagnosis based on the least squares SVM (LSSVM) model optimized using PSO, use the PCA algorithm to extract characteristic data from the fault signal as input to the fault diagnosis model, use the compressor fault diagnosis test experimental platform to obtain the fault signal when various faults occur, and use four different fault diagnosis models. The signals are compared and analyzed by training to verify the high-efficiency recognition rate and accuracy of the proposed PCA-PSO-LSSVM fault diagnosis model, which provides a new fault diagnosis method for mining compressors.
Analysis of compressor fault diagnosis system
The establishment of a diagnostic model is the key to the fault diagnosis system for air compressors. Therefore, the air compressor fault diagnosis model should be established based on a full understanding of the air compressor instructions and related documents, and on-site monitoring records and references should be fully analyzed. The relevant experience of maintenance personnel is useful for analyzing and classifying the faults that occur during the operation of air compressors and finding suitable methods to solve these faults. [13][14][15] The work of the air compressor is divided into three processes, the suction process, the compression and injection process, and the exhaust process. Through these three processes, the air is compressed and transmitted to the machinery that needs to provide power. According to compressor field experience and the literature, the main four failure modes of compressors are as follows: compressor cooling device failure (S2), compressor lubrication device failure (S3), compressor bearing failure (S4), and compressor power device failure (S5).
The following features appear in the failure mode: insufficient compressor discharge (A), extremely low exhaust pressure (B), extremely high compressor discharge temperature (C), extremely high compressor cooling water temperature (D), extremely low compressor cooling water pressure (E), extremely low main engine speed (F), extremely large compressor vibration (G), extremely high lubricating oil temperature (H), extremely low lubricating oil pressure (I), and extremely high bearing temperature (J).
According to the different types of faults, the corresponding fault states are as follows: S 1 is the normal state, and the set vector is (1,0,0,0,0). S 2 is the state when the compressor cooling water system fails, and the set vector is (0,1,0,0,0). S 3 is the state when the compressor lubrication system fails, and the set vector is (0,0,1,0,0). S 4 is the state when the compressor bearing fails, and the set vector is (0,0,0,1,0). S 5 is the compressor power supply system failure, the set vector is (0,0,0,0,1).
The compressor's failure status, type, and symptoms are shown in Table 1. The fault diagnosis model studied in this work mainly includes the cooling temperature, the cooling pressure, the exhaust pressure, the bearing temperature, the lubricating oil pressure, the lubricating oil temperature, the exhaust volume, the exhaust temperature obtained by the compressor under normal and various fault conditions, the host speed, and the host vibration. The mapping relationship between the fault state and the characteristic data established by the LSSVM algorithm realizes the fault diagnosis of the compressor system.
Least squares support vector machine
The training samples of the fault diagnosis model are set as where l is the number of training samples of the fault diagnosis model, x i is the input vector of the fault diagnosis model, and y i is the output of the fault diagnosis model. In high-dimensional feature space H, the sample set of fault diagnosis can be expressed as the following mapping relationship: Relaxation factors j and j * ensure that the fitting error obtained by the linear fitting of the fault diagnosis model is minimized as follows: The constraints are expressed as follows: The LSSVM adopts the principle of regularization and transforms the above two formulas as follows through the least squares function and equality constraints: The partial derivation of constraints is expressed as follows: The regression function for the constraints after the partial derivative is solved as follows:
PSO algorithm
As an improved SVM algorithm, the inequality constraints in the SVM algorithm can be replaced with the equality constraints in the LSSVM algorithm, the error square sum loss function is the training set loss, and the linear equations solution method can be used to solve the quadratic programming problem. The solution is simplified and the convergence accuracy and speed are improved by reducing the solution requirements. However, the network search method selected by the conventional LSSVM recognition model takes a long time to determine the parameters and can only be performed on the grid points during the search, so the parameters cannot be guaranteed when there is no more appropriate grid size effectiveness. [16][17][18] The following steps are taken to optimize the algorithm. First, the parameters, the maximum iterations, the weight end value, the weight initial value, the learning factor, and the population size in the LSSVM algorithm are initialized.
Second, the optimal position of the entire group and the individual optimal position in the PSO algorithm are determined. The current position of each particle is used to determine the current parameters of LSSVM, and the error after training determines the fitness of each particle. If the value of the current particle itself is optimal, then the current position is regarded as the optimal position. If the fitness value is the same as the fitness value of the entire group, then it can be replaced with the optimal position in the group.
The precocious problem in the optimization algorithm is handled by the following formula: Third, the fitness value of the progeny particles is solved. If the adaption value of the parent particle is greater than the adaption value of the particle, then the maximum value is the speed of the particle, so that the local minimum value can be prevented from appearing in the particle.
The inertial weights are solved by the following formula, so that the optimization algorithm guarantees better convergence efficiency and accuracy: Finally, the optimization algorithm is ended according to the termination condition.
PCA principle steps
The mean and variance of the original data matrix are solved as follows: The normalized matrix of the original data matrix is solved by the following formula: The covariance matrix of the standardized matrix is solved as follows: The variance contribution rate of the principal component is solved by the following formula, the number of principal components is set according to the judgment threshold of 85%, and the remaining principal components are ignored: Using the PCA algorithm to extract data features as input to the fault diagnosis model, the fault diagnosis model based on the LSSVM optimized using PSO is shown in Figure 1.
Fault diagnosis data collection
The compressor fault diagnosis test experimental platform is adopted to collect various fault characteristic signals as shown in Figure 2. Compressor data are collected in the normal state, the compressor cooling water system failure state, the compressor lubrication device failure state, the compressor bearing failure state, and the compressor power device failure state to obtain the compressor data of the compressor system in different states.
In the test, an Embraco NJ9226GK compressor is used for the experimental test. An Audi Speed SH0002-000 Hall speed sensor is utilized to collect the speed signal of the compressor, an LD250 ultra-low frequency two-axis vibration acceleration sensor is used to collect the vibration signal of the compressor, and MIK-P300 series pressure and temperature sensors are employed to collect signals of exhaust pressure, exhaust temperature, cooling pressure, cooling temperature, bearing temperature, lubricant pressure, and lubricant temperature. The collected signals are transmitted to the signal storage computer via wired transmission. Tables 2 and 3 list the technical parameters of the main centralized signal acquisition sensors. Table 4.
Fault diagnosis model construction
Increasing the input dimension will increase the calculation time of the fault diagnosis model based on the LSSVM algorithm. Therefore, the MATLAB software is used to perform PCA on the data collected for compressor fault diagnosis analysis to obtain the characteristic values of the contribution and cumulative contribution rates of the components shown in Table 5. This processing reduces the dimensionality of the input data and the influence of redundant information on the analysis efficiency of the fault diagnosis model based on the LSSVM algorithm. The principal component and cumulative contribution rates obtained from the analysis of the PCA algorithm show that the cumulative contribution rate of the first two principal component components reached 94.476%, which can basically express the feature matrix of the original data. Therefore, the first two principal components are selected here to form a new feature for the fault diagnosis and analysis of the compressor, reducing the input data dimension.
The principal components (1 and 2) extracted by the PCA algorithm are projected onto a two-dimensional plane, and the two-dimensional distribution of the first two principal components is obtained as shown in Figure 3.
In Figure 3, ''*'' indicates the normal state of the compressor, ''h'' indicates the compressor cooling water system failure, ''o'' indicates the compressor lubrication The two-dimensional distribution of the first two principal components indicates that the clustering of the principal components is pretty well, and the distribution interval of each feature is obvious, facilitating the diagnosis and analysis of the compressor system faults through the pattern recognition method.
The basic parameters of the LSSVM and PSO algorithms are set as follows: parameter c in the LSSVM model is set to 50, and parameter s is set to 10. The population size in the PSO algorithm is set to 30; the learning factor is 1.5; the initial and final weight values are 0.9 and 0.3, respectively; and the maximum number of optimizations is 200.
Analysis of fault diagnosis results
Compressor fault diagnosis systems are established using conventional BPNN and LSSVM algorithms and the LSSVM algorithm optimized using PSO without feature extraction by the PCA algorithm.
The same training and test data are used to carry out model training and generalization ability test on the four established fault diagnosis models. The training and test results of the four fault diagnosis models are shown in Figures 4 to 7. Table 6 shows the comparison of model training time, the test sample classification time, and the fault recognition accuracy rate using four fault diagnosis models for compressor system fault diagnosis.
The comparison results of the four fault diagnosis models show that the fault diagnosis model based on the BPNN has low convergence speed when few training samples are used, requires long training time, and has low fault recognition accuracy. The LSSVM-based fault diagnosis model can use few training samples to obtain a diagnostic model with a certain generalization ability, but the random performance of parameter selection during the training process leads to a large difference in the performance of the fault diagnosis model and low overall recognition efficiency and accuracy.
The PSO-LSSVM fault diagnosis module without feature extraction through the PCA algorithm can optimize the selection of parameters in the LSSVM-based fault diagnosis model through the PSO algorithm, but redundant data and high input dimensions increase the fault diagnosis training time of the model and the accuracy of the results. In the PSO-LSSVM fault diagnosis model based on the features extracted through the PCA algorithm, PCA is first performed on the data collected for compressor fault diagnosis and analysis, which reduces the input data dimension and the redundant information to the fault based on the LSSVM algorithm diagnostic models analyze the impact of efficiency. Afterward, the blindness of selecting the parameters of the LSSVM fault diagnosis model is reduced through the PSO algorithm.
The comparison results indicate that the PSO-LSSVM fault diagnosis model based on the feature extraction through the PCA algorithm has a diagnosis accuracy rate of up to 10.4% higher than the other three models, and the test sample classification time is reduced by up to 0.025 s. Therefore, the proposed PSO-LSSVM fault diagnosis model diagnosis system based on PCA algorithm for extracting features has high recognition efficiency and accuracy, which can effectively identify compressor faults and improve compressor efficiency.
Conclusion
This work studies the compressor system fault diagnosis system. First, the compressor fault diagnosis system is analyzed, a fault diagnosis algorithm model based on the type of fault is established, and an example analysis is conducted to verify the reliability of the proposed fault diagnosis method. The conclusions are as follows: (1) According to the working principle and characteristics of the compressor system, the connection between the fault phenomenon and the root cause is analyzed, and the characteristic signal is extracted for fault diagnosis. (2) A fault diagnosis model based on the LSSVM optimized using PSO is established, and the PCA algorithm is used to extract data features as input to the fault diagnosis model and reduce the input dimension. (3) The comparative analysis of various fault diagnosis models indicates that the proposed PSO-LSSVM fault diagnosis model based on the feature extraction through the PCA algorithm has the highest fault recognition accuracy rate that is 10.4% higher than the that of the other three fault diagnosis models, and the test sample classification time is greatly reduced. Only 0.0162 s is consumed to verify the identification efficiency and accuracy of the proposed fault diagnosis model, providing a new method for the fault diagnosis of mine air compressors.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. | 2021-07-15T06:16:38.333Z | 2021-07-01T00:00:00.000 | {
"year": 2021,
"sha1": "a41556406d87fe0ceb62576bc81653ef195a224e",
"oa_license": "CCBYNC",
"oa_url": "https://journals.sagepub.com/doi/pdf/10.1177/00368504211026110",
"oa_status": "GOLD",
"pdf_src": "Sage",
"pdf_hash": "23ffdd71547463e2aaf1895036c1f332f6b2c9ad",
"s2fieldsofstudy": [
"Engineering",
"Computer Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
252615094 | pes2o/s2orc | v3-fos-license | Cost analysis of a Patient-Centred Medical Home for community-dwelling older adults with complex needs in Singapore
Introduction: The Patient-Centred Medical Home (PCMH) demonstration in Singapore, launched in November 2016, aimed to deliver integrated and patient-centred care for patients with biopsychosocial needs. Implementation was based on principles of comprehensiveness, coordinated care and shared decision-making. Method: We conducted a prospective single-arm pre-post study design, which aimed to perform cost analysis of PCMH from the perspectives of patients, healthcare providers and society. We assessed short-to-intermediate-term health-related costs by analysing data on resource use and unit costs of resources. Results: We analysed 165 participants enrolled in PCMH from November 2017 to April 2020, with mean age of 77 years. Compared to the 3-month period before enrolment, mean total direct and indirect participant costs and total health system costs increased, but these were not statistically significant. There was a significant decrease in mean cost for primary care (government primary care and private general practice) in the first 3-month and second 3-month periods after enrolment, accompanied by a significant decrease in service utilisation and mean costs for PCMH services in the second 3-month period post-enrolment. This suggested a shift in resource costs from primary care to community-based care provided by PCMH, which had added benefits of both clinic-based primary care and home-based care management. Findings were consistent with a lower longer-term cost trajectory for PCMH after the initial onboarding period. Indirect caregiving costs remained stable. Conclusion: The PCMH care model was associated with reduced costs to the health system and patients for usual primary care, and did not significantly change societal costs.
INTRODUCTION
The Patient-Centred Medical Home (PCMH) is a model of chronic care that replaces episodic primary care, with the delivery of primary care to patients, families and communities.It is guided by the principles of first-contact accessibility, comprehensiveness and whole-person orientation, integration and care coordination, sustained clinician-patient relationships, and quality and safety. 1,2e PCMH care model shifts from a disease-specific approach to a comprehensive biopsychosocial model that recognises the interplay of physical illnesses, mental disorders, and social and home environmental problems. 3his biopsychosocial perspective allows for management options that consider tailored preferences and aims for the healthcare of each patient, clinician-patient relationships, patient engagement and quality of life (QoL). 4
What is New
• To the best of our knowledge, this is the first cost analysis of a Patient-Centred Medical Home (PCMH) care model in Singapore from the patient, healthcare provider and societal perspectives.
Clinical Implications
• Early evidence suggested that the PCMH care model reduced the out-of-pocket expenses to patients and resource cost to providers for usual primary care, by shifting costs towards community-based primary care delivery.
• The PCMH had added benefits of comprehensive and individualised care for community-dwelling adults.
A few knowledge gaps on the integrated PCMH model still exist.While integrated PCMH models have been shown to improve patient outcomes, 5,6 the impact on overall costs and the relative burden on various stakeholders (patients, healthcare providers and society at large) remain less well-understood.Existing literature has focused on costs to the healthcare system and providers, rather than costs from a patient-centric perspective. 7,8Studies have also focused on the PCMH model for specific subgroups of patients, including older adults with multimorbidity or diabetic patients only. 9,10hile these studies have merit, their findings may not be generalisable to community-dwelling older adults with a combination of physical, psychological and social care needs.Additionally, it is challenging to draw conclusions on cost outcomes from current studies, due to the mixed findings across different target populations and settings.To illustrate, a pre-post study with controls in the US compared 8 practices that adopted the PCMH model with a group of 24 non-PCMH practices in 2010 to 2011, and reported no significant difference in costs. 11Other studies showed that the PCMH programme was cost-effective when provided to patients with more chronic conditions or whose conditions were poorly managed at baseline. 9,10Lastly, there is a dearth of literature on PCMH models in the Asian context.Current findings may not be generalisable to Asia due to differences in health systems and cultural backgrounds. 12e integrated PCMH care model in this study is part of a community-wide initiative (Community for Successful Ageing) on an integrated care system of comprehensive programmes and services promoting the well-being and health of older adults in Singapore. 6,13,14ur integrated PCMH model differs from current advanced primary care models in Singapore that centres on a multidisciplinary team to deliver chronic disease management, 15 or Family Medicine Clinics that provide a "one-stop platform" for individualised chronic condition prevention and management by general practitioners. 16The integrated PCMH model in this article targets community-dwelling older adults with complex biopsychosocial needs, assessed via social health, psychological and medical evaluations. 6,13,14ur model follows overarching principles on ageing in place (growing old at home), life-course approach (promoting the earlier implementation of interventions for health), socioecological model of care (health as an outcome of an individual's interaction with family, caregivers and community), and population health management. 6,13,14r study aimed to address gaps in the literature on the evaluation of costs associated with the implementation of an integrated PCMH model for community-dwelling older adults with complex needs.Our study examined costs not only from the health system perspective, but also from the perspectives of patients and the wider society.Findings will also be more generalisable to older populations across Asia.
Study design and participants
This study was part of an evaluation of a PCMH demonstration in Singapore that involved concurrent quantitative and qualitative components. 6,14We applied a prospective single-arm pre-post design that examined 2 time-points of the intervention compared to baseline.First, a true experimental design with randomisation of patients into intervention and control arms was impractical, due to the PCMH being a complex intervention with multiple care components.It would have been challenging to implement, as well as timeconsuming and resource-intensive.Second, having a non-randomised parallel control arm was also challenging due to limited resources to involve a control study site that provided usual care, and high refusal rates and low recruitment were expected from patients of the control site.In addition, there were challenges in Briefly, the intervention involved a multidisciplinary care team (doctors, registered nurses, programme coordinators and care managers), comprehensive needs assessment, and individualised care plans. 6The initial clinic visit involved comprehensive biopsychosocial assessment and the development of a preliminary individualised care plan with the patient and family members.Subsequent clinic visits reviewed individualised care plans and treated acute conditions. 6The development of care plans and reviews were discussed at interdisciplinary team meetings.PCMH primary care providers also partnered geriatric specialists from the tertiary acute hospital to provide shared care.Patients determined to have complex biopsychosocial needs at the first or subsequent clinic visits were given home-based care management services. 6ome-based care management provided extended care in the home setting, and addressed the physical home environment, financial needs, behavioural needs, and support systems by caregivers. 6,13
Outcome measures
We conducted the cost analysis from the perspectives of patients (out-of-pocket expenditures [OOPE] by patients), healthcare provider (resource cost to the health system) and society: Patient S1 (online Supplementary Material) shows the categories of direct and indirect costs under each perspective.
Cost outcomes included the direct costs of PCMH services, and direct medical and non-medical costs.Healthcare utilisation data in this study was selfreported.Specifically, an adapted Client Service Receipt Inventory (CSRI) survey was administered to participants to collect self-reported data on healthcare utilisation in the past 3-month period from the date of surveys. 18Surveys were conducted at baseline and repeated at 3 months and 6 months post-enrolment.
Cost to patients (out-of-pocket expenditures by patients)
Table 1 displays how unit prices were derived and unit price of each item.identifying a suitable control study site that would capture older adults with similar characteristics and health status profiles as the intervention group, but would only deliver usual care.Lastly, the option of utilising historical datasets to act as controls had possible limitations, such as systematic differences between groups due to the time difference in data collection, poor performance of matching techniques as the dataset may not contain suitable controls, and lack of clarity on what constituted usual care for controls.
Study participants with recruited from 1 November 2017 to 30 April 2019.Informed consent was taken from participants or proxies.As mentioned in our previous publications, these were the eligibility criteria included: 6,13 (1) Patients with high biopsychosocial health risk defined by the 37-item BioPsychoSocial Risk Screener validated in the Singapore setting, 17 pre-existing risk stratification criteria used by referring healthcare institutes, and/or clinical assessment.Functional ability and frailty were not part of the inclusion and exclusion criteria.
(2) Patients aged ≥40 years. 6,13Cut-off age was chosen to reflect the life-course approach.PCMH services could be provided to patients with complex needs from their fourth decade of life prior to entering old age, for early intervention to prevent further adverse health outcomes. 6,133) Patients who resided in Whampoa, a geographically defined district in Singapore (total population of 41,000) where the PCMH was located.6,13 Intervention There are advanced primary care models in Singapore, such as the Family Medicine Clinics where family medicine physicians provide individualised care for patients and the Teamlet Care Model where a multidisciplinary team manages non-communicable diseases.15,16 Our PCMH model differed by targeting community-dwelling older adults with biopsychosocial needs, and has 2 integrated parts: medical care in primary care and psychosocial care in home-based care management.
The intervention was described in detail in our earlier publication. 6Implementation was based on PCMH values on patient-centredness, comprehensive and coordinated care, accessible services, shared decisionmaking, and quality and safety. 1 The PCMH was an integrated care intervention comprising physician-led primary care clinics, and home-based care management services led by medical social workers and nurses. 12
Cost to healthcare providers (resource cost to health system)
Table 2 shows the derivation of unit resource costs and unit resource cost for each item, drawing on previous work by Abdin et al. 18 and Graham and Bilger. 19
Costs to informal caregivers
Table 2 shows the derivation of unit costs for work productivity loss and leisure time loss frm caregiving, based on a local study by Woo et al. 20 and the labour statistics from the 2017 Singapore Yearbook of Manpower Statistics. 21
Data analysis
To assess the short to intermediate-term health-related cost from different perspectives, we analysed cumulative health-related costs incurred over the quarter (3-month period) immediately before enrolment, from the first quarter post-enrolment (i.e.first 3-month period postenrolment), and the second quarter post-enrolment from 3 months post-enrolment to 6 months post-enrolment (i.e.second 3-month period post-enrolment).We estimated the change in quarterly costs in the latter 2 periods, with the quarter immediately before enrolment.
Costs were converted to 2017's Singapore Dollar amount using the consumer price index for Singapore from the Monetary Authority of Singapore.We reported mean and median costs per-person per-quarter.
Statistical analysis
Sample characteristics of the 165 participants analysed were presented, and data were summarised descriptively as mean and standard deviation.We used multivariable linear regression models with random intercepts and fixed slopes to compare the difference between mean costs per-participant per-quarter, during the first 3-month period post-enrolment and second 3-month period post-enrolment, compared to the 3-month period prior to enrolment.The multivariable regression models had random intercepts to allow for variation by individuals (i.e.between-participant variation), but slopes were fixed as time was not a continuous variable in this study.We recognised that each participant may act as its own control in this single-arm pre-test post-test study design.However, this study still aimed to adjust for observed and unobserved time-invariant differences between participants.The random intercept accounted for such time-invariant between-participant differences by adjusting for a list of covariates.The estimates and statistical significance of estimates would be similar in the unadjusted (not presented) and adjusted models, and we presented results that adjusted for between-participants differences.These methods were consistent with an earlier article that conducted a before-after study without controls to examine the changes in quality of life and patient activation (knowledge, skills, and confidence for self-management) of older adults, and applied multivariable regression modelling to adjust for between-participant differences. 6pecifically, this study adjusted for age at enrolment, sex, weighted Charlson Comorbidity Index (CCI) at baseline, having received any formal education (yes/ no), housing types in Singapore (1-, 2-and 3-room Housing and Development Board [HDB] apartments; 4-room or larger HDB apartments and Housing and Urban Development Company apartments and executive condominiums) and baseline 13-item Patient Activation Measure (PAM-13) score.Covariates were selected based on plausible relationships with healthcare utilisation.Ethnicity was not included as the sample was predominantly Chinese.Baseline PAM-13 measure was adjusted to account for differences in study participants' underlying knowledge, skills and confidence integral to managing one's own health and healthcare, 22 and was also applied as a covariate in our previous publication. 6tatistical significance was determined at P<0.05.All analyses were performed on Stata version 14.0 (StataCorp, College Station, US).
Participant characteristics
A total of 238 patients were enrolled into PCMH from 1 November 2017 to 30 April 2019, of which 16 did not fulfil study eligibility criteria.After excluding patients who did not consent to the study (n=34, 14.3%) and were uncontactable (n=3), this study recruited 184 study participants.The final sample analysed was 165 study participants after loss to follow-up at 3 months post-enrolment (n=11, 6.0%) and 6 months postenrolment (n=8, 4.3%).There were 6 deaths; 1 patient who was retrospectively found to be ineligible; and 12 withdrawals from PCMH due to being housebound, admitted to a long-term care facility, or relocated to be out of the PCMH service boundary.Participant flow diagram has been published previously. 12able 3 displays the sociodemographic characteristics of study participants (n=165).The mean age of study participants was 77 years, with 93.9% aged 60 years and above. calculated the unit cost of PCMH clinic visits based on attending professionals and recorded durations of each visit.
For home-based care management services, the duration of each home visit was assumed to be 1 hour.
Non-PCMH services
Government primary care Unit costs were based on the study in Singapore by Abdin et al. 18 The data were presented in manpower of physician cost per minute.To compute per-visit cost estimates, we assigned durations of visit for patients that were representative of our study population from consultations with local experts (A&E visit: 15 minutes; SOC visit: 15 minutes; outpatient allied health visit: 25 minutes; non-PCMH primary care visit: 15 minutes).
Sensitivity analysis
We conducted sensitivity analyses for resource cost from the healthcare provider perspective, by applying a longer duration of 30 minutes for non-PCMH primary care (government primary care, private general practice), SOC and A&E visits.The main objective of conducting the sensitivity analysis was to present alternative scenarios with higher resource cost estimations from the healthcare provider perspective, to reflect the additional resources needed to provide care to more complex patients.Note: This sensitivity analysis was not applied for price calculations from the patient perspective (Table 1), as these services are likely at a fixed price to patients.
Cost to patients
Table 4 presents the 3-month cumulative cost perparticipant per-quarter from a patient perspective.Compared to the 3-month period prior to enrolment, there was a statistically significant decrease in mean cost for non-PCMH primary care (government primary care, private general practice) by SGD9.40 (41.4%) in the first 3-month period post-enrolment and by SGD3.60 (15.9%) in the second 3-month period postenrolment based on the multivariable regression model.Compared to the 3-month period prior to enrolment, the mean total cost to study participants was SGD123.10 (13.8%) higher in the first 3-month period postenrolment and SGD180.20 (20.3%) higher in the second 3-month period post-enrolment, but these were not statistically significant based on the multivariable regression model.There were no statistically significant changes in mean costs from outpatient services (accident & emergency, specialist outpatient clinic, outpatient allied health, day surgery), inpatient admissions, work productivity loss and paid caregiving.
Compared to the first 3 months post-enrolment, there was a statistically significant decrease by SGD12.20 (35.0%) for PCMH services in the second 3 months postenrolment based on the multivariable regression model.
Cost to healthcare providers (resource cost)
Table 5 presents the 3-month cumulative cost per-participant per-quarter from a healthcare provider perspective.Compared to the 3-month period prior to enrolment, there was a statistically significant decrease in mean cost for non-PCMH primary care (government primary care, private general practice) by SGD50.70 (38.8%) in the first 3-month period post-enrolment and by SGD29.30(22.4%) in the second 3-month period post-enrolment, which remained significant (but with a larger quantum) in the sensitivity analysis.
Compared to the 3-month period prior to enrolment, mean total cost increased by SGD414.50 (19.0%) at the first 3-month period post-enrolment and by SGD194.70 (8.9%) at the second 3-month period post-enrolment, but these were not significant based on the multivariable regression model.There were no statistically significant changes in mean cost for outpatient services, inpatient admissions, and community care.
Compared to the first 3 months post-enrolment, there was a statistically significant decrease in mean cost by SGD427.30(51.2%) for PCMH services in the second 3-month period post-enrolment based on the multivariable regression model.This study assigned durations of visit for patients that were representative of our study population from consultations with local experts-A&E: 15 min; SOC: 15 min; outpatient allied health visit: 25 mins; non-PCMH primary care visit: 15 min.We conducted sensitivity analyses using a longer duration of 30 mins for A&E, SOC and non-PCMH primary care visits (see Although the PCMH has the added benefits of both clinic and home-based care management relative to existing primary care, this study found no evidence that implementing the PCMH model resulted in increased costs from the healthcare system perspective.An initial non-significant increase in mean total per-participant healthcare resource cost was recorded in the first quarter after enrolment, which may be primarily attributed to the higher frequency and longer duration of visits for PCMH providers to conduct initial comprehensive geriatric assessments.It was not unexpected for resource costs to be high during the initial implementation stages of PCMH. 8,9Importantly, we found that mean total per-participant resource costs subsequently dropped by approximately half in the subsequent quarter, consistent with a fall in resource cost to deliver PCMH services cost per-participant in the second quarter after enrolment.
Cost to society
The estimated mean resource costs associated with non-PCMH primary care consultations (government primary care, private general practice) fell from the first post-enrolment quarter.Given that the estimation only included consultation fees, it was plausible that unmeasured savings were even larger.There were no other significant changes in resource costs associated with other outpatient services (A&E, SOC, outpatient allied health, day surgery), inpatient admissions or community care.This suggested that at the health system level, change was driven primarily by patients' substitution from usual primary care (government primary care/polyclinics, private GP) to communitybased primary care by the PCMH.The PCMH model shifted OOPE of patients from non-PCMH primary care towards primary care in the community (i.e.PCMH).Other studies also suggested that implementing PCMH may not increase overall costs to patients and health systems. 8,23n the existing literature, findings on costs have been mixed and challenging to interpret due to the different contexts and target populations for PCMH. 8,9,24Our findings suggest potential for cost savings for patients and their families from longer-term reductions in other formal and informal care.Over our study period, we found no significant changes in the mean cost to patients for other outpatient services and inpatient admissions.At the same time, costs to patients and society associated with formal and informal caregiving remained stable with non-significant changes.From the societal perspective, therefore, the total mean cost per-participant initially rises due to the initial increase in resource use, but it subsequently decreased.We furthermore expect these to reduce with a longer follow-up, as PCMH management further reduces utilisation of services like A&E and hospitalisations. 8,9,24iven the evidence of better QoL and patient activation from our previous publication, 6 our findings show that the PCMH model dominates the current standard of care, namely, intervention was both more effective and not more costly.As such, the results are presented in the form of a detailed cost analysis rather than a costeffectiveness analysis.
A strength of PCMH is its ability to meet the multidimensional needs of patients, such as aspects of comprehensive assessments in the clinic and the home setting, a multidisciplinary care team, individualised care plans, shared decision-making, and empanelment (assignment of patients to primary care providers and care teams, taking into account patient and family preferences).We previously reported improved QoL and patient activation, and this study found no increase in societal cost, and a decrease in cost for usual primary care.We had low loss to follow-up and examined cost analysis holistically from the perspectives of patients, healthcare providers and society.When interpreted together, these findings suggest positive overall systemlevel outcomes for PCMH.
However, this study had a few limitations.We recognised that the participants recruited were older persons even though persons aged ≥40 years were eligible, which would affect the generalisability of our study.However, most PCMH care models have also focused on older populations. 25Next, unit prices of PCMH services were based on PCMH administrative data that consisted of fees for consultation, procedures and medications, whereas prices of non-PCMH services included consultation fees.The omitted category of procedures and medications was approximately 30-40% of the total bill.Our results likely understated costs of non-PCMH services to patients.This made PCMH appear relatively expensive by understating cost savings from reducing non-PCMH services.Calculation of resource cost to providers for non-PCMH health services used manpower cost of only doctors due to the lack of data availability, whereas the cost of PCMH included manpower cost of doctors, nurses and care managers.Hence, our findings likely underestimated the resource costs of PCMH services and resource costsavings from reducing non-PCMH services.Utilisation counts and durations of caregiving were self-reported, but the CSRI survey has been validated and used in Singapore. 18,26Importantly, we included relevant cost components with unit costs based on the literature.Utility bills, indirect costs (e.g.transportation) and other overhead costs were omitted due to lack of data.
Our study has important implications for practice and policy.Our analysis shows that the PCMH model effectively resulted in an increase in utilisation of community-based primary care services and suggested potential reductions in hospital-based care, consistent with Singapore's national healthcare policy directives to move beyond acute to preventive care, and from hospital to community settings.The model supports multiple mechanisms by which this may occur including more access to higher-quality care.This results in better care outcomes or increased care coordination at the community level, enabling stronger care networks and management of care in the lived environment.These remain to be investigated further.Second, our study recognises that the implementation of a PCMH care model may require a significant amount of initial investment but suggests that offsetting cost savings to the system and patients may be realised in a relatively short period.This makes the PCMH model a potentially more sustainable paradigm for patients with complex care needs.Finally, analysing estimates from different payers' and stakeholders' perspectives shows that the PCMH model is consistently aligned in economic impact and incentives to change.This is especially important in a multipayer health system to support decision-making on benefits and risks, so as to enable a comprehensive practice transformation towards effective team-based PCMH care. 27
CONCLUSION
Evidence suggested that the PCMH care model shifted the OOPE of patients and resource costs of providers from usual primary care and outpatient services, towards community-based primary care with the added benefits of comprehensive and individualised care for community-dwelling older adults.PCMH reduced costs to the health system and patients for usual primary care and did not increase informal caregiving costs.There may be potential for sustainability and scalability of PCMH.
Ethics approval and consent to participate
This study was registered with ClinicalTrials.gov(Protocol ID: 2017/00352) and was approved by the National Healthcare Group Domain Specific Review Board (NHG DSRB) Singapore (Reference: DSRB 2017/00352).Informed consent was taken from all participants or their proxies.All methods were performed in accordance with the relevant guidelines and regulations.
perspective = (a) OOPE on PCMH and non-PCMH health services + (b) OOPE of paid care services from domestic helpers or other professional carers + (c) work productivity loss by participants due to ill health Healthcare provider perspective = (d) Resource cost of providing PCMH and non-PCMH health services Societal perspective = (b) + (c) + (d) + work productivity loss and leisure time loss from providing caregiving to participants by family members Supplementary Table
0A&E:: 3 -: 3 -: 3 -
accident and emergency; PCMH: Patient-Centred Medical Home; SOC: specialist outpatient clinic T-3months_enrolment month period immediately before enrolment T enrolment_3months month period from enrolment to 3 months post-enrolment T 3months_6months month period from 3 months post-enrolment to 6 months post-enrolment a P value <0.05; b P value<0.001Compared to mean cost at T -3months_enrolment , the change in mean cost using multivariable linear regression modelling that adjusted for age at enrolment, sex, weighted Charlson Comorbidity Index at baseline, having received any formal education (yes/no), housing types in Singapore (1-2-and 3-room Housing and Development Board [HDB] apartments; 4-room or larger HDB apartments and Housing and Urban Development Company [HUDC] apartments and executive condominiums) and baseline 13-item Patient Activation Measure score.c Reference group was mean cost at T enrolment_3months Ann Acad Med Singap Vol 51 No 9 September 2022 | annals.edu.sgCost analysis of community-dwelling older adults-Grace Sum et al.
Table 1 .
Cost from the patient perspective: Items and unit
prices Items How this study derived price Unit price (SGD) PCMH services
19patient admissionBased on inpatient charges for Singapore citizens per bed day, according to hospital class wards.35 to 43 for C Class wardsNon-PCMH community careSocial day care Out-of-pocket expenses for community care services were based on the total costs given by the Agency for Integrated Care under the Ministry of Health of Singapore, and assumed to be 40% of the total costs based on Graham & Bilger.1917.20pervisit Hiring domestic helpers for caregiving We collected self-reported estimated time contributions from informal household-based caregivers (both paid domestic helpers and family caregivers) related to assisting with Activities of Daily Living and Instrumental Activities of Daily Living, supervision of study participants, and/or providing child and elderly care on behalf of study participants.The hourly cost of care provided by domestic helpers at home was based on figures reported by Woo et al. 20 3.45 per hour Work productivity loss due to ill health by study participants Work productivity loss was valued at the median gross monthly income of a full-time employed worker in Singapore, converted to hourly wages based on national average weekly total paid hours worked per employee.20.22 per hour PCMH: Patient-Centred Medical Home a The National Healthcare Group is one of the 3 healthcare clusters in Singapore, and the PCMH was under this healthcare cluster.Superscript numbers: Refer to REFERENCES Cost analysis of community-dwelling older adults-Grace Sum et al.Ann Acad Med Singap Vol 51 No 9 September 2022 | annals.edu.sg
Table 2 .
Cost from the healthcare provider and societal perspectives: Health services and unit resource costs
Table 2 .
Cost from the healthcare provider and societal perspectives: Health services and unit resource costs (Cont'd) Ann Acad Med Singap Vol 51 No 9 September 2022 | annals.edu.sgA&E: accident and emergency; PCMH: Patient-Centred Medical Home; SOC: specialist outpatient clinic Superscript numbers: Refer to REFERENCES
Table 3 .
Sociodemographic characteristics of study participants
Table 3 .
Sociodemographic characteristics of study participants(Cont'd) bThe weighted CCI was used as the summary measure for adjusting for comorbidities in our multivariable linear regression model.The CCI was based on the number of chronic conditions that are each assigned an integer weight from 1 to 6, with a weight of 6 representing the most severe morbidity.The summation of the weighted comorbidity scores resulted in a summary score.In this study, the International Classification of Diseases 10th Revision (ICD-10) codes of study participants were based on a national healthcare administrative database and the PCMH clinic administrative database.Subsequently, we compute the weighted CCIs based on ICD-10 codes.
Table 4 .
Cost from the patient perspective: 3-month cumulative cost per study participant
Table 5 .
Cost from the healthcare provider perspective: 3-month cumulative cost per study participant
Table 3
the change in mean cost using multivariable linear regression modelling that adjusted for age at enrolment, sex, weighted Charlson Comorbidity Index at baseline, having received any formal education (yes/no), housing types in Singapore (1-2-and 3-room Housing and Development Board [HDB] apartments; 4-room or larger HDB apartments and Housing and Urban Development Company [HUDC] apartments and executive condominiums) and baseline 13-item Patient Activation Measure score. ,
Table 6 .
Cost from the societal perspective: 3-month cumulative cost per study participant the change in mean cost using multivariable linear regression modelling that adjusted for age at enrolment, sex, weighted Charlson Comorbidity Index at baseline, having received any formal education (yes/no), housing types in Singapore(1-2-and 3-room Housing and Development Board [HDB] apartments; 4-room or larger HDB apartments and Housing and Urban Development Company [HUDC] apartments and executive condominiums) and baseline 13-item Patient Activation Measure score.Abdin E, Subramaniam M, Achilla E, et al.The Societal Cost of Dementia in Singapore: Results from the WiSE Study.J Alzheimers Dis 2016; 51:439-49.Ann Acad Med Singap Vol 51 No 9 September 2022 | annals.edu.sgCost analysis of community-dwelling older adults-Grace Sum et al.
, c Reference group was mean cost at T enrolment_3months Reference: | 2022-09-30T15:04:49.858Z | 2022-09-01T00:00:00.000 | {
"year": 2022,
"sha1": "2c7450c896fe431d5a26eda63de4fe9171d65afc",
"oa_license": null,
"oa_url": "https://annals.edu.sg/download/25316/",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "e77afd57f04247f75518291b4c059d76ab5db191",
"s2fieldsofstudy": [
"Medicine",
"Economics"
],
"extfieldsofstudy": [
"Medicine"
]
} |
1520725 | pes2o/s2orc | v3-fos-license | Quality of life of parents with children living at home: when one parent has cancer
Goals of work This study examined the quality of life (QoL) of cancer patients diagnosed 1–5 years previously and their spouses, with children 4–18 years living at home. Relationships between parents’ QoL and the children’s functioning were explored. Patients and methods 166 cancer patients and their spouses provided information on their QoL (RAND-36) and on their children’s functioning (Child Behavior Checklist). Main results Male and female patients scored similarly to a norm population on five domains. Patients’ QoL was clinically relevantly and/or statistically lower on social functioning, role limitations because of physical problems, and vitality than the norm. Male spouses’ QoL was comparable to the norm. However, female spouses reported better physical functioning but more social problems. QoL varied according to type of cancer, treatment intensity, and recurrence. Using the QoL composite scores, a significant relationship was found between patients’ psychosocial and physical functioning and spouses’ psychosocial functioning. Patients’ psychosocial functioning correlated moderately strongly to weakly with their reports of their younger children’s and adolescents’ functioning; physical functioning correlated only weakly with adolescents’ functioning. The patients’ functioning related weakly to moderately strongly to adolescents’ self-reports of functioning. Spouses’ psychosocial functioning weakly related to their and adolescents’ reports of adolescents’ functioning. Conclusions Cancer patients’ QoL 1–5 years after diagnosis was decreased in three of eight domains; their spouses seem to be doing well. Parents’ physical and psychosocial functioning related weakly to moderately strongly to their children’s functioning, depending on the child’s age and information source. The patients’ functioning related more strongly to the children’s functioning than the spouses’ did.
Introduction
The impact of cancer on a patient's psychosocial functioning has received much attention in the literature over the past decades. It is generally acknowledged that a patient's quality of life (QoL) decreases while they are battling their illness. Problems in QoL tend to diminish over time as a patient responds to treatment. However, a significant percentage of patients continues to experience clinically elevated levels of problems requiring professional treatment [40]. Furthermore, a growing number of studies provides evidence that cancer patients' spouses develop problems affecting their QoL [14,17,21,23,25,27]. Spouses have reported psychological and physical distress, which have been found to correlate with the patient's dysfunction [17].
However, the majority of studies examine patients and spouses in later adulthood [5,6,13,25,29]. Most incidences of cancer occur in older adults. According to the Dutch Comprehensive Cancer Center's database, 70% of all new incidences of cancer in The Netherlands in 2003 occurred in patients more than the age of 60 [11]. Parental cancer in younger families may be a more serious stressor than in later phases of life, as illness and death in later adulthood could be considered more natural or normative [37,39]. It has been shown that older cancer patients report less anxiety and depression and better QoL than younger adult patients [26]. Couples in an earlier life phase with young children and adolescents at home lead busy lives. They juggle childrearing responsibilities and the demands of sustaining a healthy marriage with trying to meet the individual (career) needs of each partner [28]. In the case of families with parental cancer, parents have the added strain of their loved one being ill. Additionally, their stress may be confounded when the patient may not be able to work or take care of the children as well as before the onset of the disease.
Only a few studies have examined the impact of cancer on the functioning of all members in families in this phase of life. In our previous investigations of families confronted with parental cancer, the main focus has been on the children's psychological well being [20,36]. Results have shown clinically elevated levels of distress in 35% of adolescent daughters and 21% of adolescent sons [20] and higher levels of emotional problems in adolescent daughters and elementary school-aged sons of cancer patients [36]. It is possible that the children's emotional and behavioral functioning, which may be affected by their parent's cancer, affects how the parents function.
The current study had four main aims. The first aim was to examine health-related QoL of patients and spouses in families with young children 1-5 years after diagnosis. We hypothesized that cancer patients and their spouses' QoL would be lower than that of a norm group and that QoL would differ as a function of health status and gender (i.e., which parent was ill: the mother or father). Based on literature [30], we expected to find that female patients reported the lowest QoL, followed by female partners, male patients, and male partners. The second aim was to examine the effect of illness-related variables (time since diagnosis, recurrence, treatment intensity, and type of cancer) on parents' QoL. We expected that patients and spouses would report more problems more shortly after diagnosis, when confronted with recurrence of cancer and when treatment was intense and, also, that QoL would differ depending on type of cancer. Our third aim was to examine relationships between patient's and spouse's QoL; we hypothesized that they would be significantly related. Finally, the fourth aim was to explore possible relationships between the parents' QoL and the children's emotional and behavioral functioning.
Procedure
Both hospitalized cancer patients and those being treated or seen for follow-up visits at the outpatient clinic of the University Medical Center Groningen between January 2001 and February 2003 were approached by surgical, medical, and radiation oncologists and oncology nurses. Patients were eligible for the study if they were diagnosed with cancer 1-5 years before study entry and had children between the ages of 4 and 18 years at the time of parent's diagnosis who resided or had frequent contact with the diagnosed parent. Furthermore, both parents and children needed to be fluent in Dutch. A family could participate in the study if the patient and at least one child agreed to participate. Physicians and oncology nurses offered all eligible patients and spouses written information plus an information brochure adapted for the children. Informed consent was obtained from each family member separately, as regulated by the Medical Ethical Committee of the University Medical Center Groningen. After obtaining informed consent, researchers mailed packages with questionnaires and prepaid return envelopes to each family member. Cancer patients, spouses, and children were instructed to fill in questionnaires independently of each other and to not discuss their answers.
Instruments
Demographic data was gathered on: age, gender, level of education, length of relationship, number of children living at home, and who the primary caregiver in the family is. Patients provided information about when they were diagnosed, their type of cancer, treatment intensity, and recurrence. Education was measured on a seven-point scale from (1) elementary school only to (7) university degree.
The eight subscales and two component summary scores of the Dutch translation of the RAND-36 were used to measure QoL [8,31,38]. Scores on the subscales can range from 0 to 100, with higher scores indicating better functioning. When comparing patients and spouses to a reference group, norm scores were used from the Dutch manual for the RAND-36 [31]. The manual provides mean scores from a random sample of 1,063 people between the ages of 18 and 89 (65% women, age M=44.1) from the population register of a municipality in The Netherlands. Norm scores are provided for each subscale for the entire sample and for men and women separately. Summary scores report physical functioning and psychosocial functioning and are standardized with a mean of 50 and a standard deviation of 10. The reliability and validity of the RAND-36 has been supported in a wide number of international and national studies [8,32,33]. Cronbach's alphas in this study on the eight subscales ranged between 0.72 and 0.90 for patients and 0.78 and 0.90 for partners.
To investigate how parents' QoL related to the children's functioning, parents were asked to complete the 120-item Child Behavior Checklist (CBCL) [1,34]. Adolescents also completed the 102-item Youth Self-Report (YSR) [2,35]. In this study, the internalizing (emotional functioning), externalizing (behavioral functioning), and total problem (total of internalizing, externalizing, and cognitive problems) scales were used to provide a picture of the problems occurring in the children of parents diagnosed with cancer. The CBCL's and YSR's reliability and validity have been supported in a great number of studies. In this study, Cronbach's alphas for the internalizing, externalizing, and total problem scales ranged from 0.84 to 0.94 for reports from patients, spouses, and adolescent children.
Analysis
In our analyses, a single variable was created to define treatment intensity. Patients were grouped into two categories based on the clinical expectation that surgical treatment alone (nonintense treatment) would be less distressing to the family because of less time away from home and fewer visible side effects. Other single-modal (chemo-, radiotherapy) and multimodal treatments (combination of surgery, chemotherapy, radiotherapy, hormone, or immunotherapy) were defined as intense. Time since diagnosis was calculated by subtracting the date on which the patient filled in the questionnaires from the date they were diagnosed. Descriptive analyses were performed on demographic information. Comparisons between patients, spouses, and the reference group were investigated using independent t tests. Results were corrected for the number of tests performed with a Bonferroni correction for multiple comparisons; p values less than or equal to 0.0015 were considered significant (0.05 of 32 tests). To assess clinical relevance, effect sizes (ESs) were calculated by dividing the difference between means by the square root of the average of the squared standard deviations [10]. ESs between 0.20 and 0.49 were considered small, 0.50 and 0.79 medium, and ≥0.80 large [10]. We considered that ESs greater than or equal to 0.50 indicated a clinically relevant difference [24]. A single variable was created to describe parents' gender and health status (ill mothers, ill fathers, healthy mothers, and healthy fathers). One-way analyses of variance (ANOVAs) with a Bonferroni post-hoc test were used to compare QoL of patient and spouse as a function of gender and health status. After the correction for multiple tests was applied, p values less than or equal to 0.005 were considered significant (p=0.05 of ten tests). In analyses of QoL as a function of type of cancer, only groups of ten or more patients were included. An ANOVA and a Kruskal-Wallis test (some patient groups were small) were performed to compare patients' and spouses' QoL as a function of type of cancer. After the correction for multiple tests was applied, p values less than or equal to 0.005 were considered significant (p=0.05 of ten tests). The two QoL composite scores (physical summary score and psychosocial summary score) were used in correlational analyses to explore relationships between parents' QoL and between each parent's QoL and the children's problems, with separate analyses for elementary school-aged (4-10 years) and adolescent (11-18 years) children. Pearson's productmoment correlation coefficients were calculated; correlations with a coefficient less than 0.30 were considered weak, 0.30-0.50, moderately strong, and greater than 0.50, strong [10].
Participants
A total of 476 families were approached for the study. Two hundred and nine agreed to participate (44%), including 336 children, with an average of 2.3 children per family. Because this study focused on couples, we selected data from families where both patient and spouse filled in questionnaires, resulting in a database of 166 couples and 304 children. Table 1 summarizes the demographic information. The majority of cancer patients in this study were women (78%). Patients were diagnosed with various types of cancer (Table 3), including breast (52%), gynecological (10%), hematological (9%), skin (9%), urological (6%), soft tissue and bone tumors (6%), head/neck (4%), gastrointestinal (3%), and central nervous system (1%). Fourteen percent of patients had undergone nonintense treatment (N=23), and 86% had undergone intense treatment (N=143).
There were no significant differences between participating and nonparticipating parents regarding patient's gender, type of cancer, or time since diagnosis. Of the nonparticipants, 22% (N=59) declined because of reasons directly related to the parents (e.g., parents had moved on with their lives or were too emotionally distressed). Reasons related to the children (e.g., children were not interested, children had not been informed of parent's illness) were given as the explanation for not wanting to be included in 20% (N=53) of nonparticipating families. Twenty-five percent (N = 67) mentioned a variety of reasons, including another illness in the family or the parents' or children's busy-ness. The remaining 33% (N= 88) did not provide an explanation.
Patients' and spouses' QoL in comparison with the norm Results are summarized in Table 2. Clinically relevant ESs were found for male patients' social functioning, role limitations because of physical problems, and vitality. Both male and female patients scored statistically significantly lower QoL than the norm on one subscale: vitality. Female patients additionally scored lower on two subscales: social functioning and role limitations because of physical problems.
Clinically relevant ESs were found for female spouses' physical functioning and social functioning. Female spouses reported statistically significantly higher scores than the norm on physical functioning. They also scored statistically significantly lower on social functioning than the norm.
Effects of health status and gender, time since diagnosis, recurrence, treatment intensity, and type of cancer on QoL Parents' health status and gender had a statistically significant effect on parents' physical summary score. A statistically significant effect was also found on parents' physical functioning, role limitations because of physical problems, vitality, and general health perception ( Table 2).
A Bonferroni post-hoc test revealed the following. On the physical summary score, female patients scored lower than male spouses (p<0.001), and male patients scored lower than female and male spouses (p<0.001 for both). On physical functioning, female patients scored lower than female and male spouses (p<0.001 for both). On role limitations because of physical problems, female and male patients scored significantly lower than male spouses (p= 0.003 and p=0.005, respectively). On vitality, female patients scored significantly lower than male spouses (p< 0.001). Finally, on general health perception, female patients scored significantly lower than male spouses (p= 0.003), and male patients scored significantly lower than female and male spouses (p=0.004, p<0.001, respectively).
Time since diagnosis did not correlate significantly with patients' QoL. Time since diagnosis correlated weakly with spouses' physical summary score (r=0.19, p =0.031). Patients with a recurrence scored clinically relevantly lower than patients without recurrence on the physical summary score (t=3.44, ES=0.82, p=0.001), social functioning (t= 3.12, ES=0.79, p=0.003), role limitations because of physical problems (t = 3.72, ES = 0.72, p < 0.001), and general health perception (t=3.69, ES=0.82, p=0.001). Spouses of patients with a recurrence scored clinically relevantly lower than spouses of patients without a recurrence on the psychosocial summary score (t=2. 23 p=0.01). Treatment intensity was not significantly related to spouses' QoL. Type of cancer had a statistically significant effect on patients' QoL as summarized in Table 3. Statistically significant differences between patient groups were found on the physical summary scale and five subscales. Patients with skin cancer reported the highest QoL. One third of skin cancer patients received intense treatment. Patient with hematological cancer reported the lowest QoL levels. All 16 hematological cancer patients received intense treat- Relationship between patient and spouse's functioning and emotional and behavioral functioning of the children Patients' physical functioning was not significantly related to their elementary school-aged children's functioning but was weakly related to the adolescents' total problems and internalizing as reported by the patients (Table 4). Patients' psychosocial functioning was moderately strongly to weakly related to their elementary school-aged children's internalizing, externalizing, and total problems and weakly related to the adolescents' internalizing, externalizing, and total problems as reported by the patient. The patients' physical and psychosocial functioning was moderately strongly to weakly related to the adolescents' internalizing and total problems according to adolescents' self-reports; their physical functioning was weakly related to adolescents' externalizing. Spouses' psychosocial and physical functioning was not found to be significantly related to their elementary school-aged children's functioning as reported by the spouses. Additionally, their physical functioning was not significantly related to the adolescents' functioning. Spouses' psychosocial functioning was weakly related to the adolescents' internalization and total problems as reported by the spouses. Finally, the spouses' psychosocial functioning was weakly related to the adolescents' internalizing as reported by the adolescents (Table 4).
Discussion
The aim of this study was to gain insight into the QoL of cancer patients in the child-rearing stage and their spouses. We found that cancer patients 1-5 years after diagnosis evaluate their QoL as clinically relevantly and/or statistically significantly lower than the normal population on three of the eight domains, partially supporting our first hypothesis. This is in line with studies that have reported a decrease in patients' social and physical domains [7,9]. However, our study did not find a decrease in patient's emotional functioning, in contrast to some reports [3,4,16]. Our study focused on a subgroup of cancer patients, namely those who are relatively young. The finding that our patient group scored similarly to the norm seems to indicate that these patients' lives have gotten fairly back to normal. These parents seem to be handling the unexpectedness of a cancer diagnosis during this life phase, the treatment, or the confrontation with a possible death at an early age, coupled with the responsibility of raising children fairly well. However, they still seem to experience problems in some areas, specifically social functioning, role limitations because of physical problems, and vitality.
Spouses of cancer patients reported a QoL comparable to the norm group, and female spouses reported even better Table 4 Correlations between parents QoL, as measured by RAND-36 composite scores, and children's functioning by age group as reported by parents (CBCL) and adolescents themselves ( physical functioning than the norm. Only on one subscale (social functioning) did spouses (and then only women) report decreased functioning. Our findings largely negate our hypothesis and are in contrast with other studies where spouses have reported a decreased QoL [27]. An explanation for our findings may be that spouses in other studies, where the average age is higher, may have had a lower QoL simply because of their older age. As age increases, QoL scores decrease [33]. A second explanation could be that spouses viewed their QoL relative to their ill partner and therefore regarded their own health as good as or better than people generally may. It is also possible that we did not find a decreased QoL in the spouses because our patient group ranged from 1 to 5 years after diagnosis; time since diagnosis correlated significantly positively with spouses' physical functioning. The threat that the patient might not survive may be less prominent for spouses at this point in time.
Our study found significant effects of gender and health status on the physical summary score and four of the eight subscales. To summarize our findings, patients scored lower than spouses, male spouses reported the highest QoL, and, on some scales, the female patient reported the lowest QoL and, on others, the male patient. In a study on gastrointestinal cancer, female patients and female spouses were both reported to suffer overall more distress than male patients [30], which our study did not find. They also reported that female spouses reported lower QoL than male patients and spouses. However, that study examined patients within 6 months after surgery. They also focused specifically on patients with gastrointestinal cancer, which is equally prevalent in men and women. That allowed them to more easily generalize that the differences they found were due to gender. As our study sample was diagnostically heterogeneous, it is difficult to differentiate whether the observed differences were due to gender or cancer site and consequent treatment received. Our study sample consisted of 78% women, which could seem skewed. However, according to the Dutch Comprehensive Cancer Center's database, cancer occurs more frequently in women during this age range; approximately 65% of cancer patients in this region of The Netherlands with invasive tumors are women [12].
To further explore our second aim, we analyzed QoL as a function of type of cancer and found differences depending on the type of tumor the patient had, supporting our hypothesis. Skin cancer patients in our study reported the highest QoL, while patients with hematological cancer reported the lowest QoL. These differences may be related to the kind of treatment the patient received. We found that patients who received nonintense treatment reported a better QoL than those who received intense treatment. The removal of a localized melanoma that only requires outpatient excision may affect QoL less than frequent and long hospital stays for courses of chemotherapy for hematological malignancies. Our findings are similar to another study reporting patient distress levels [40] that found that distress levels varied depending on cancer site. With regard to how spouses function, no significant QoL differences were found because of the type of cancer or treatment intensity. It would seem that tumor type or treatment intensity do not affect spouses' functioning. Whether the patient had a recurrence does seem to affect spouses' QoL; the ES of the psychosocial summary score was strikingly large. This may be due to, for those spouses, the still current threat that the patient may not survive. It is interesting to note that recurrence seems to have affected patients' physical and social functioning but not their mental functioning. Patients with recurrence do not report more mental distress than patients who have not had a recurrence.
Thirdly, we investigated relationships between patient and spouse QoL and found weak to moderate positive relationships between cancer patient's physical and psychosocial functioning and spouse's psychosocial functioning. These findings are in line with two studies reporting a moderate positive relationship between patients' and spouses' psychological distress [9,18].
Finally, the hypothesis that we would find a significant relationship between the parents' and the children's' functioning was supported, although the relationships were not strong. The patient's QoL related more often significantly with the children's functioning than the spouse's. Given that four of five of the patients were women, it could be that mothers were more often alert to possible problems in the children than fathers. This may be due to the large number of families in our sample where the mother, sick or healthy, is the primary caregiver. Mothers tend to orient themselves more toward others, whereas fathers tend be more self-oriented [18]; this could enable mothers to judge problems better. It is possible that illness plays a role in the patients' reports as distressed parents are likely to rate more behavioral and emotional problems in their children [22]. However, we found significant relationships between the adolescents' self-reports and parents' functioning. Unfortunately, the cross-sectional design of this study limits us in understanding the causal nature of this finding.
We found more significant, negative correlations with the children's internalizing and total problem scores than externalizing. It would seem that parents' QoL negatively relates to the children's emotional functioning, more than to their acting up. It is interesting to note that the parents' functioning related differently to adolescents and schoolaged children. The patient's physical functioning correlated weakly with the adolescents' functioning but not significantly with the younger children's functioning. This was not found for spouses-their physical functioning did not relate significantly to how children from either age group function. Patient's psychosocial functioning correlated weakly to moderately strongly with children in both age groups, while spouses' psychosocial functioning correlated weakly with the adolescents' functioning. Adolescent selfreports provided a similar picture; we found weak to moderately strong correlations between patients' physical and psychosocial functioning but only one weak correlation between spouses' psychosocial functioning and adolescents' internalization. Our finding that parents' functioning related more often significantly with the adolescents may be due to adolescents' stage of cognitive development. Adolescents are more able to understand the patient's illness and may pick up on physical problems more than elementary school-aged children [15].
Research considerations
In our investigation, being the first to evaluate family functioning in families in the child-rearing age, we purposely did not choose a homogenous cancer group (i.e., only breast cancer). Our results are statements for the general group of child-rearing families with cancer. Comparisons between patients in our research group with varying diagnoses showed differences between groups; however, the small subgroups make it difficult to generalize our findings. Future research will be required to investigate the effect of varying forms of cancer, the stage of cancer, its prognosis, and treatment on the functioning of individual family members.
Additionally, our response rate of 44% could mean that despite the fact that no differences were found between participants and nonparticipants in gender, type of cancer, or time since diagnosis, a sample bias may exist. We cannot be sure whether psychological problems were over-or underreported; some nonrespondents stated still being overwhelmed by the illness as a reason for not participating, while others indicated they had moved on. Additionally, the cross-sectional nature of this study makes it impossible to accurately capture the dynamic processes present in family relationships or determine whether parents' QoL effects the children's functioning or vice versa. Furthermore, a longitudinal study could provide insight into the QoL of cancer patients and their family members over a period of time. This study found significant relationships between patient and spouse functioning and between parent's functioning and children's behavior. However, these relationships appeared to have modest predictive power. The patients' and spouses' functioning is likely more influenced by other factors not measured in this study, such as personality, social support, or family environment [19,26].
In conclusion, cancer patients' QoL 1-5 years after diagnosis seems to be returning to normal, except in three domains. Their spouses seem to be doing well. The patients' QoL varied according to the type of cancer, how intense their treatment had been, and whether they had experienced a recurrence. Spouses' QoL seemed to be unaffected by the type of cancer and treatment intensity but did vary depending on whether the patient experienced a recurrence. A moderate positive relationship was found between the patient's functioning and his/her spouse's. Parents' physical and psychosocial functioning was weakly to moderately strongly related to their children's functioning. The patients' functioning related more strongly to the children's functioning than the spouses' did. How cancer patients' families function may have an impact on the patient's functioning, up to 5 years after diagnosis. This is something that should be taken into account by clinicians. | 2014-10-01T00:00:00.000Z | 2007-07-03T00:00:00.000 | {
"year": 2007,
"sha1": "63359c30f511e6dd9dd1977a4411b54001666110",
"oa_license": "CCBYNC",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s00520-007-0299-7.pdf",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "446b5cac8c85d96af409299180a941e3c45a10b0",
"s2fieldsofstudy": [
"Psychology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
234457245 | pes2o/s2orc | v3-fos-license | Selective Removal of Penicillin G from Environmental Water Samples by Using Molecularly Imprinted Membranes
Penicillin G imprinted membranes were prepared by utilizing UV induced in-situ polymerization. The characterization of membranes was conducted by using ATR-FTIR, XPS, SEM and AFM. The binding properties of imprinted membranes were evaluated against concentration and time. The binding capacity of the membranes was tested for real water samples. The size of the binding cavities of membranes was determined by using PALS. The specific selectivity of the membranes was investigated by using similar s-lactams, penicillin V and amoxycillin. The specific binding of the membranes was de-termined as 3.27 µg/g for penicillin G while this value was obtained as 0.83 and 0.51 µg/g for pencillin V and amoxycillin, respectively. The binding capacity of the membrane was determined as 5.03 µg.g-1 for ultra pure water while this value was obtained as 4.01 µg.g-1 and ~3.50 µg.g-1 for tap water and natural water samples from different sources, respectively.
the number of researches on the removal of Penicillin G is not many. Molecularly imprinted polymers are functional materials prepared by polymerizing functional monomers and a crosslinking agent around a template molecule. Removing the template molecule from the network provides recognition sites that can accurately fit the template molecule concerning its affinity, size and shape. In this study, Penicillin G imprinted membranes were prepared by utilizing UV induced polymerization method for removal of the template molecule Penicillin G from real water samples. Photopolymerization is a good alternative to thermal polymerization by providing polymerization and crosslinking at room temperature. It is possible to say that the synthesis at room temperature is an important factor to get molecularly imprinted polymers with high specific selectivity depending on the stability of pre-polymerization complexes [10].
Keywords:
Membrane; Molecularly imprinted polymer; Penicillin G. SEM membrane images were taken using an FEI Quanta 200FEG instrument with an accelerating voltage of 2.00 kV. Until screening, samples had been sputter-coated with gold.
Bruker Nanoscope 9 atomic force microscopy (AFM) was employed in tapping mode to make the morphological investigation of the membranes. PALS experiments were performed using a standard system with detectors whose scintillators are plastic. The position of the detectors was face-to-face. The positron source was 22 Na with an activity of 11 μCi, in the form of dried NaCl solution between two thin Kapton foils with glued together. The instrument time resolution was 258 ps (FWHM). The source was placed between two identical membranes in a typical 'sandwich' configuration. Each spectrum was saved in the air at room temperature every 3 h for a total count of 3×106. The sum of 5 spectra was used to produce 1.5×107 counts. The spectra obtained were examined using the program LT [12].
Investigation of binding properties of membranes
Batch binding experiments were performed to determine the binding capacity of imprinted and non-imprinted membranes. The membranes were incubated in 2 mL Penicillin G solutions with a various concentration in the range of 0.5-50 ppm. The concentration of Penicillin G before and after incubation was analyzed by employing a UV-vis Varian Cary100 spectrophotometer at a maximum absorption wavelength of 216 nm [11]. The number of repetition of the analyses was three. The binding behaviour of membranes was analysed by using adsorption isotherms. The adsorbed amount of penicillin G (qe) was determined by using the following equation: where C i and C f are the initial and equilibrium concentrations of template molecule, penicillin G (µmol/L), respectively, V is the volume of solution (L), W is mass of the dry membrane (g).
The kinetic behaviour of adsorption was investigated by incubation for various periods in 3 ppm Penicillin G solution.
The reusability of the membranes was investigated by an adsorption-desorption cycle which was repeated 7 times. The concentration of Penicillin G was 3 ppm and the incubation time was 20 min. After each adsorption step, the membrane was washed as described before and then the incubation was repeated.
Aldrich (Milwaukee, USA). All the reagents and solvents were used as received without any further purification step.
Preparation of Penicillin G imprinted membranes
Penicillin G (0.503 mmol) was dissolved in 2 mL DMF and then 2.01 mmol MAA was added into the solution. The solution had been kept overnight at 4 o C. EGDMA (5.00 mmol) and 8 mL of DMF were added. The ratio of MAA to penicillin G was kept as 4:1 [11]. The formation of pre-polymerization complex between template and functional monomers and the synthesis of penicillin G imprinted membranes were illustrated in Fig. S1. Finally, 20.0 mg of benzophenone was added. The solution was purged with nitrogen and then placed between two identical silanized glass slides which separated by a rubber O-ring. The solution filled glass slides was placed into a zip-lock plastic bag under a nitrogen blanket. The polymerization was carried out by using a Philips type HB 171/A self-tanning UV lamp, adapted with four fluorescent tubes of Philips CLEO 15 W UVA. The irradiation time was kept as 90 min. The distance from the glass slides to the UV lamp was held as 10 cm. Subsequently, Penicillin G imprinted membranes (MIM) were carefully removed from the surface of the glass slides. The template molecule and unpolymerised monomers were washed out of the membranes with a mixture of acetic acid/methanol/water (1/5/1, by volume) until no Penicillin G was detected through a UV spectrometer. The non-imprinted membranes (NIM) was prepared using the same technique except for the addition of penicillin G.
Characterization of membranes
ATR-FTIR, XPS, SEM and AFM were used in the chemical and physical characterisation of membranes. The size of the recognition sites in the membranes was investigated by employing positron annihilation lifetime spectrometer (PALS) to analyze the effect of the presence of the template molecule.
The FTIR experiments were performed using Thermo Nicolet iS10 Thermo Nicolet iS10 model spectrometer in attenuated total reflection (ATR) mode. Spectra were recorded by accumulating 32 scans with a resolution of 4 cm -1 .
Experiments with X-ray photoelectron spectroscopy (XPS) were carried out using a Thermo spectrometer with a monochromated Al K α X-ray source. Survey and core scans were reported with a pass energy of 30 eV and 200 eV, respectively. All analyzes were performed at a take-off angle of 90 o . Elemental composition of the surface of the membranes was examined with a 400 µm X-ray spot size and binding energy ranged from 0-1000 eV.
The specific selectivity of imprinted membranes was investigated by incubation in the presence of similar compounds, Penicillin V and amoxicillin (Fig. 1). The concentrations of antibiotics were kept as 3 ppm and the incubation time was 20 min.
The binding performance of the membranes was investigated in real water samples collected from Pazar Creek (in Ankara) and Kızılırmak (in İskilip/Çorum region). And the binding performance of the membranes was also tested in tap water. Water samples were firstly filtered by using a 0.25 µm Nylon filter and stored at +4 o C until analysis. Before experiments 3 ppm Penicillin G was spiked into the water samples. The incubation time was kept as 20 min. All the experiments were repeated three times.
Chemical characterization of membranes
As can be seen in Fig. 2 the structural characterization of imprinted and non-imprinted membranes was carried out by ATR-FTIR spectroscopy. The peak at 1721 cm -1 is attributed to stretching of C=O groups of methacrylic acid and EGDMA. The stretching vibration of sp 3 hybridized C-H bonds can be seen at 2943 cm -1 while the bending vibration of the same group is at 1454 cm -1 . The peak at 1140 cm -1 corresponds to the stretching of C-O groups in dialkyl ethers [13]. All the peaks can confirm the chemical structure of poly(ethylene glycol dimethacrylateco-methacrylic acid). Unfortunately, it is not possible to see any evidence of the presence of the template molecule due to its trace amount in the polymeric network by using FTIR. However, the presence of the template molecule can be proven by using more sensitive instrumental analysis methods such as XPS. XPS is a useful tool which provides detailed information about the chemical composition of the surfaces after modification for the approximately 10 nm depth from the surface. Therefore, the surface elemental composition of membranes was investigated by employing XPS. The survey scans of MIM and NIM can be seen in Fig 3. The chemical compositions of the imprinted and non-imprinted membranes show similarities for C1s (285 eV) and O1s (532 eV) peaks with an exception of the N1s peak at 400 eV for MIM [14]. The appearance of N1s peak proves the presence of the template in the crosslinked polymer.
The C1s core level spectra of MIM, pencillin G removed MIM and NIM can be seen in Fig 4. The common and the major component for all is the C-C and C-H peak which is detected at 284. eV. The component at 286.0 eV is attributed to the C-O [15]. The highest binding energy value (~288.3 eV) can be ascribed to C=O groups of copolymeric matrix [14]. There is one more component in MIM spectrum at 287.4 eV which corresponds to C-N type carbon atoms in amide (399.9 eV) and lactam groups (400.2 eV) of pencillin G (Fig. S2) [15]. This is a direct evidence of the presence of the template molecule in the imprinted membrane. It can be said that the procedure for the removal of template molecule is sufficient as can be noticed with the disappearance of C-N in the spectrum of penicillin G removed MIM. The binding ability of the MIM was also investigated by XPS analysis. The C-N peak at 287.4 eV appears again for the penicillin G removed MIM after penicillin G incubation in 3 ppm for 20 min.
Physical and morphological characterization of membranes
The presence of the template molecule in the imprinted polymers can cause various morphological differences [16]. The difference in the morphological structures of the surfaces of NIM and MIM can be seen from SEM images (Fig. 5). The interaction between the monomer and the template can cause more porosity and roughness on the surface of the membranes [17].
The roughness of the surfaces of the membranes was also investigated by employing AFM (Fig 6). The value of Rq is determined as 234 and 98 for MIM and NIM, respectively. The roughness of the surface of MIM provides higher binding capacity for penicillin G due to the increasing in surface area and permeability of the membranes that provides to enhance in the possibility of the interaction between MIM and penicillin G [18].
The size of the free volume holes in the membranes was investigated by employing PALS. It is an absolute and highly sensitive method which provides detailed and certain information on the size and size distribution of free volume holes in polymeric structures [19,20] and defects in metals [21]. The selectivity of the molecularly imprinted polymers can be controlled mainly by two factors: 1) chemical affinity of the recognition sites and 2) binding cavity size and size distribution [13]. The second factor gains more importance when the binding of compounds with a similar molecular size is considered. Therefore, it is important to have information about the size of the binding cavities of imprinted polymers. In previous work from our laboratory, the average diameter of the free volume holes of imprinted polymers of penicillin G prepared by grafting on the non-woven fabric was determined as 0.570 nm while this value was obtained as 0.469 nm for non-imprinted polymer grafts [13]. In this study, the diameter of the free volume holes in MIM was obtained as 0.558±0.003 nm that matches to previously obtained data. The pore size of the NIM (0.434±0.002 nm) was also similar to the previous data for the control polymer. The obtained data also correlates to the molecular diameter of the penicillin G which was reported as 0.500 nm by Wu and coworkers [22].
Adsorption isotherms
To examine the binding characteristics of the membranes binding isotherms were constructed. Adsorption isotherms are created by using the equilibrium concentration of an analyte onto an adsorbent (q e ) and the concentration of the analyte in the solution (c e ) with which it is in contact. This relationship is explained by using several equations. The most common models used to investigate the binding behaviours of the molecularly imprinted polymers are Langmuir, Bi-Langmuir and Freundlich isotherms (Table S1) [23]. The binding isotherms Langmuir and Freundlich were used to determine the binding efficiency of the membranes (Fig. 7). As it is apparent the binding behaviour of MIM fits Freundlich model with an R 2 value of 0.96. Heterogenous binding models are more suitable to evaluate the binding behaviours of imprinted polymers due to their heterogeneous nature of the recognition sites [23]. The heterogeneity index, n which varies in the range of 0-1 was determined as 0. 13. The value of n shows an increase with an increase in homogeneity of the system and it equals to the 1 for an ideal homogeneous system [24]. As can be noticed the MIM prepared by conventional free radical polymerization method has a heterogeneous nature due to the polymerization method [25].
Kinetics of adsorption of penicillin G imprinted membranes
The kinetic evaluation of the adsorption is important as the type of the adsorption gives information about the nature of the binding. The adsorption rate-limiting step can be managed by diffusing the template to the adsorbent surface, or by the chemical interaction between the template and the adsorbent. Generally, the binding in molecularly imprinted polymers is controlled by chemical affinity between template and host. The kinetics of adsorption was evaluated by using Lagergren's first-order rate equation and pseudo-second-order rate equation (Table S2). The kinetic behaviour of the prepared membranes matches the pseudo-second-order kinetic model with an R 2 value of 0.99 (Fig. 8), which suggests that the rate-determining step of penicillin G adsorption is regulated by the chemical affinity of penicillin G to MIM [16,26].
Reusability of the membranes
The reusability of the membranes was investigated by repeating the adsorption-desorption cycle 7 times (Fig. 9). After each binding step, the removal of the template molecule was achieved by solvent extraction which was described previously. The performances of the imprinted membranes did not show a significant decrease upon repeated use. The crosslinked imprinted membranes can be used for many times without a significant binding capability loss.
The binding capability of imprinted and non-imprinted membranes
As is evident from Fig. 10 the adsorption capacity of MIM is greater than the binding capacity of NIM. The morphological differences and the data about the size of the recognition sites support the higher binding capability of the MIM. The roughness of the surfaces of the MIM enhances the chemical interaction between penicillin G and the imprinted membranes. It can be assumed that the rise in surface area due to the roughness makes the availability of the binding sites easier [18]. Non-imprinted membranes (NIM) were produced without the addition of the template molecule using the same procedure. The absence of a template molecule produces smoother surface due to the lack of any specific chemical interaction [16].
It still has a porous structure because of the presence of a porogen solvent.
Specific selectivity of membranes
Specific selectivity of the membranes was investigated by using structurally similar β-lactam antibiotics such as amoxicillin and penicillin V (Fig. 11). The concentration of the antibiotics was kept as 3 ppm for 20 min incubation. The specific binding (SB) value is a criterion to investigate the specific selectivity of the imprinted polymers against the target molecule. SB is determined by substracting of the binding onto non-imprinted polymers due to the non-specific interactions from the binding of imprinted polymers. Thus, the obtained binding data is the result of the specific interaction between the template and adsorbent [27]. SB of the MIM was determined as 3.27 µg/g polymer for penicillin G while this value obtained as 0.83 and 0.51 µg/g polymer for penicillin V and amoxicillin, respectively. Because of the complete complement of the binding sites in terms of shape, size and unique interaction between the template molecule penicillin G and the imprinted membrane, the highest SB value was obtained for the template molecule [11].
The selectivity (k) and relative selectivity coefficients (k') were determined by using equations summarized in Table S3. The selectivity coefficient (k) is used to investigate the degree of selectivity of molecularly imprinted and non-imprinted polymers towards the target molecule. The selectivity of the MIM for penicillin G is 2.33 and 3.48 times higher than the selectivity towards penicillin V and amoxycillin, respectively (Table S4). The relative selectivity coefficients were determined as 1.38 and 1.46 for penicillin V and amoxycillin, respectively. The data prove that the specific selectivity of MIM is higher than the NIM as expected due to the formation of binding sites by templating of penicillin G. Figure 9. Binding of penicillin G which was determined by using the same imprinted membrane after each adsorption for seven cycles (incubation concentration of penicillin G= 3 ppm, incubation time= 20 min, n= 3). Figure 11. Selectivity of penicillin G imprinted (MIM) and nonimprinted membranes (NIM) against penicillin V and amoxicillin (concentration of each antibiotic= 3ppm, incubation time=20 min, n=3).
Investigation of the binding capacity in real samples
To analyze the binding capabilities of the imprinted membranes in real samples tap water and natural surface water samples from different regions were used. As it is apparent in Fig. 12 the binding capacities of the MIM decreases due to the complexity of the real samples [16]. The binding capacity of the MIM was determined as 4.01 µg.g -1 for tap water and ~ 3.50 µg.g -1 for natural surface waters which contains 3 ppm spiked penicillin G, respectively. It is possible to say that the imprinted membranes can be useful to remove penicillin G from real water samples.
CONCLUSION
Penicillin G imprinted membranes were produced by UV induced in-situ polymerization of methacrylic acid in the presence of a crosslinking agent. The detailed structural and physical characterization of the membranes was carried out using various methods of spectroscopy and microscopy. XPS was used to prove the existence of the template inside the imprinted membranes. The change in morphology of the membranes was investigated by SEM and AFM. The interaction of the template molecule and the network results in the roughness of the imprinted membrane surface. The size of the free volume holes of the imprinted membranes was investigated by PALS. The kinetic behaviour of the imprinted membranes supports the presence of the chemical affinity between penicillin G and the membranes. The binding capability of the membranes was investigated for real water samples. The binding capacity was determined as approximately 3.50 µg.g -1 . As a result, the prepared membranes can be useful to remove penicillin G from natural surface water samples. The use of the molecularly imprinted membranes as a filtration material for the selective removal of penicillin G from environmental water sources can be considered as the main novelty of this work.
ACKNOWLEDGEMENT
I would like to thank Prof. Dr. Olgun Güven for his continuing support and encouragement throughout my studies. Figure S1. Synthesis of penicillin G imprinted polymers by employing methacrylic acid and ethylene glycol dimethacrylate as functional monomer and crosslinking agent, respectively. Table S4. Dissociation constant (K d ) and selectivity selectivity (k) and relative selectivity coefficient (k').
I s o t h e r m M o d e l E q u a t i o n P a r a m e t e r
L a n g m u i r Lagergren's first-order rate equation and pseudo second-order rate equation [1] were employed to evaluate the kinetics of the binding by using the equations which summarized in Table S2. The binding capacity of the membranes changes with time as apparent in Fig. S3. Table S1. Adsorption isotherm models and related parameters. Periasamy K, Namasvayam C. Process development for removal and recovery of cadmium from waste water by a low cost adsorbent: adsorption rate and equilibrium studies. Ind. Eng. Chem. Res. 1994, 33, 317-320. Table S2. Kinetic models and related parameters. | 2021-01-07T09:05:01.389Z | 2020-12-01T00:00:00.000 | {
"year": 2020,
"sha1": "d7ac8b8a2a2f0a8dcb0fc777524ec01503d445d0",
"oa_license": null,
"oa_url": "https://dergipark.org.tr/en/download/article-file/1526962",
"oa_status": "GOLD",
"pdf_src": "Adhoc",
"pdf_hash": "444a5a1e5c1fd513c64a16669cb1142684df04f9",
"s2fieldsofstudy": [
"Chemistry"
],
"extfieldsofstudy": [
"Chemistry"
]
} |
59410322 | pes2o/s2orc | v3-fos-license | Mechanisms underpinning the permanent muscle damage induced by snake venom metalloprotease
Snakebite is a major neglected tropical health issue that affects over 5 million people worldwide resulting in around 1.8 million envenomations and 100,000 deaths each year. Snakebite envenomation also causes innumerable morbidities, specifically loss of limbs as a result of excessive tissue/muscle damage. Snake venom metalloproteases (SVMPs) are a predominant component of viper venoms, and are involved in the degradation of basement membrane proteins (particularly collagen) surrounding the tissues around the bite site. Although their collagenolytic properties have been established, the molecular mechanisms through which SVMPs induce permanent muscle damage are poorly understood. Here, we demonstrate the purification and characterisation of an SVMP from a viper (Crotalus atrox) venom. Mass spectrometry analysis confirmed that this protein is most likely to be a group III metalloprotease (showing high similarity to VAP2A) and has been referred to as CAMP (Crotalus atrox metalloprotease). CAMP displays both collagenolytic and fibrinogenolytic activities and inhibits CRP-XL-induced platelet aggregation. To determine its effects on muscle damage, CAMP was administered into the tibialis anterior muscle of mice and its actions were compared with cardiotoxin I (a three-finger toxin) from an elapid snake (Naja pallida) venom. Extensive immunohistochemistry analyses revealed that CAMP significantly damages skeletal muscles by attacking the collagen scaffold and other important basement membrane proteins, and prevents their regeneration through disrupting the functions of satellite cells. In contrast, cardiotoxin I destroys skeletal muscle by damaging the plasma membrane, but does not impact regeneration due to its inability to affect the extracellular matrix. Overall, this study provides novel insights into the mechanisms through which SVMPs induce permanent muscle damage.
Introduction
Snakebite envenomation is a recently reinstated neglected tropical disease [1] that causes around 100,000 deaths annually [2,3] and innumerable permanent disabilities predominantly on the rural population living in the lower income regions of the world [4][5][6]. The significant rate of mortality and morbidity occurs due to the difficulties associated with the treatment of snakebites [7], which vary depending on the species [8], geographical location [9], age of the offending snake [10,11], the quantity of venom injected, correct diagnosis and mode of treatment [12]. Snake venoms contain proteins and small peptides with diverse functional effects [12]. Medically important snakes are generally found in two main families; Elapidae, a family with venoms mainly composed of smaller, neurotoxic proteins such as phospholipase A2 (PLA2) and three finger toxins, and Viperidae, a family with generally larger proteins such as serine and metalloproteases that primarily affect the cardiovascular and musculoskeletal systems. Snake venom serine proteases (SVSPs) mainly cause systemic envenomation effects such as the alteration of blood pressure, activation or inhibition of coagulation factors and degradation of fibrinogen [13,14]. However, snake venom metalloproteases (SVMPs) primarily induce local envenomation effects such as swelling, necrosis and extensive tissue/muscle damage as well as the activation of certain coagulation factors and degradation of fibrinogen. SVMP-induced muscle damage is often difficult to treat due to the delay in obtaining appropriate medical treatment and poor outcome of anti-snake venom (ASV) treatment in the local tissues [15,16]. Hence, extensive tissue damage is frequently treated by fasciotomy, a surgical procedure to remove the damaged tissues, cleaning the affected areas followed by skin graft or amputation of affected limbs or fingers when fasciotomy fails to suffice [7]. This results in permanent disabilities for victims, and significantly affects their socio-economic status following snakebites. For example, long term (persisting for over 13 years) musculoskeletal disabilities were found in over 3% of snakebite victims in a rural population of Sri Lanka and of these over 15% had to undergo amputations [17].
Skeletal muscle is composed of myofibres surrounded by the collagen-rich basement membrane. This tissue is imbued with a resident stem cell population called satellite cells (SCs), located under that basement membrane (BM), which are able to regenerate a functional tissue even after extensive damage [18]. The BM plays a key role in muscle repair by orientating the regenerating myofibres, a process mediated by SCs and acting as a scaffold for fibres to grow parallel to the existing fibres [19]. The majority of the direct myotoxic effects of venoms are attributed to PLA2 [20]. They can induce either local or systemic effects depending on their specificity to muscle cells (systemic effects) or a broader range of cells (local effects) through hydrolysis of phospholipids in plasma membrane. Other myotoxic venom components include sodium channel-blocking myotoxins [21] and muscle fibre depolarising cardiotoxins [22]. SVMPs are enzymatic proteins that primarily attack the collagenous structures and various other important components of BM to induce muscle damage. It has recently been reported that SVMPs induce haemorrhage by cleaving components of the BM and extracellular matrix surrounding the smaller blood vessels [23] although as multi-domain proteins, they are capable of binding to and cleaving a range of different proteins [24,25].
SVMPs are generally classified into four groups based on the additional domains present in their structure: PI/Group I-contains only a metalloprotease domain; PII/Group II-contains a metalloprotease and disintegrin domain, and in some cases the disintegrin domain has been reported to be processed and liberated as a free disintegrin; PIII/Group III-contains a metalloprotease, a disintegrin-like and cysteine-rich domains; PIV/Group IV-contains two lectinlike domains connected by disulphide bonds to the other domains that are found in PIII SVMPs [26]. Although disintegrin-like domains show high sequence identity to disintegrins, they lack the typical RGD motif found in the venom disintegrins, which inhibit platelet aggregation via selectively blocking integrins. Both disintegrin-like and cysteine-rich domains have been found to inhibit collagen-induced platelet aggregation and induce early events of acute inflammation [27]. Notably, disintegrin-like domains were reported to contain an ECD motif that interacts with integrins and block their functions [28]. The non-proteinase domains play key roles in determining the diverse pharmacological effects of PII, PIII and PIV classes of SVMPs including the activation of coagulation factor X [29] and prothrombin [30] amongst others. These domains have also been found to co-localise in muscles, facilitating the hydrolysis of collagen and other BM components by the metalloprotease domain and promoting its accumulation in the BM [24], exerting haemotoxic activities. Moreover, SVMPs are also known to cause ischaemia in the local tissues due to poor blood supply as a result of their haemotoxic effects [31], which may prevent phagocytic removal of necrotic debris and reduce the supply of oxygen and nutrients needed for regeneration [32]. Given the complexity of their actions, a better understanding of the molecular mechanisms through which SVMPs induce permanent muscle damage may pave the way to the development of improved therapeutic strategies for snakebites. In this study, we demonstrate novel insights into the mechanisms by which a PIII/group III metalloprotease isolated from the venom of a North American viper, the western diamondback rattlesnake, Crotalus atrox triggers permanent muscle damage. Our results establish that this SVMP induces muscle damage and also prevents muscle regeneration by acting on the BM, myofibres, blood supply and SCs.
Materials
Lyophilised C. atrox venom was purchased from Sigma Aldrich (UK) and the purified Cardiotoxin 1 (CTX), a three-finger toxin from the venom of Naja pallida was obtained from Latoxan (France).
Protein purification
C. atrox venom (10mg) was dissolved in 1mL of 20mM Tris.HCl buffer (pH 7.6) and centrifuged at 5000g for 5 minutes before applying to a pre-made 1mL HiTrap™ Q HP Sepharose anion exchange column. Protein elution was performed at a rate of 1mL/min using 1M NaCl/ 20mM Tris.HCl gradient (up to 60%) by an Ä KTA purifier system (GE Healthcare, UK) over 20 minutes. The collected fractions were analysed by SDS-PAGE using standard protocols as described previously [33] and fractions with the protein of interest were pooled. The pooled fractions were then concentrated using a Vivaspin centrifugal filter and applied to a gel filtration column (Superdex 75, 1.6cm x 70cm). Protein elution was performed at a rate of 1mL/ min using 20mM Tris.HCl (pH 7.6). Following SDS-PAGE analysis, the fractions containing the protein of interest were pooled and concentrated before running through the same gel filtration column again for further purification. Finally, the fractions containing the pure protein were pooled, concentrated and stored at -80˚C until further use. Protein estimation was performed using Coomassie plus protein assay reagent (ThermoFisher Scientific, UK) and bovine serum albumin as standards.
Mass spectrometry
The purified protein was subjected to SDS-PAGE, and a gel section containing the pure protein was subjected to tryptic digestion and analysed by mass spectrometry at AltaBioscience (Birmingham, UK). The extracted protein (10μg) from the gel slice was added to 100mM ammonium bicarbonate (pH 8). This was then incubated with dithiothreitol (10mM) at 56˚C for 30 minutes. After cooling to room temperature, the cysteine residues were alkylated using iodoacetamide (50mM). Trypsin gold (Promega, UK) was subsequently added and the samples were incubated overnight at 37˚C. The digested peptides were concentrated and separated using an Ultimate 3000 HPLC series (Dionex, USA). Samples were then trapped on an Acclaim PepMap 100 C18 LC column, 5um, 100A 300um i.d. x 5mm (Dionex, USA), then further separated in Nano Series Standard Columns 75μm i.d. x 15 cm. This was packed with C18 PepMap100 (Dionex, USA) and a gradient from 3.2% -44% (v/v) solvent B (0.1% formic acid in acetonitrile) over 30 minutes was used to separate the peptides. The digested peptides were eluted (300nL/min) using a triversa nanomate nanospray source (Advion Biosciences, USA) into a LTQ Orbitrap Elite Mass Spectrometer (ThermoFisher Scientific, Germany). The MS and MS/MS data were then searched against Uniprot using Sequest algorithm and the partial sequence was then compared to the other similar protein sequences available in the protein database.
Fibrinogenolytic assay
Human plasma fibrinogen (100μg/mL) was incubated with different concentrations of the whole venom or the purified protein, and a small volume of digested samples were removed at 30, 60, 90 and 120 minutes and mixed with reducing sample treatment buffer [4% (w/v) SDS, 10% (v/v) β-mercaptoethanol, 20% (v/v) Glycerol and 50mM Tris.HCl, pH 6.8]. The samples were then analysed by 10% SDS-PAGE and stained with Coomassie brilliant blue to determine the fibrinogenolytic activity of venom and the purified protein.
Enzymatic assays
The metalloprotease activity of both C. atrox whole venom and the purified protein was assessed using a fluorogenic substrate, DQ-gelatin (ThermoFisher Scientific, UK). Briefly, the whole venom or purified protein (10μg/mL) was mixed in phosphate buffered saline (PBS, pH 7.4) with DQ gelatin (10μg/mL). The reaction mix was incubated at 37˚C and the level of fluorescence was measured at 60 minutes using an excitation wavelength of 485nm and emission wavelength of 520nm by spectrofluorimetry (FLUOstar OPTIMA, Germany).
Similarly, the serine protease activity was measured using a selective substrate, Nα-Benzoyl-L-Arginine-7-Amido-4-methylcoumarin hydrochloride (BAAMC) (Sigma Aldrich, UK). The whole venom or the purified protein (10μg/mL) was incubated with BAAMC (2μM) at 37˚C and the level of fluorescence was measured at an excitation wavelength of 380nm and emission wavelength of 440nm by spectrofluorimetry.
Ethical statement
The University of Reading Research Ethics Committee has approved the procedures for blood collection from healthy human volunteers and the consent forms used to obtain written consent. Experiments with mice were performed in line with the principles and guidelines of the British Home Office and the Animals (Scientific Procedures) Act 1986 (PPL70/7516). All the procedures were approved by the University Research Ethics Committee (License number: UREC 17/17).
Platelet aggregation
Human blood was obtained from healthy volunteers in vacutainers with 3.2% (w/v) sodium citrate as an anti-coagulant and the platelet-rich plasma (PRP) was prepared as described previously [34][35][36]. Platelet aggregation assays were performed by optical aggregometry using 0.5μg/mL cross-linked collagen related peptide (CRP-XL) as an agonist in the presence and absence of different concentrations of the purified protein.
Administration of CTX or the purified protein in mice
The C57BL/6 mice (8 weeks old) were obtained from Envigo, UK. Mice were anaesthetised with 3.5% (v/v) isofluorane in oxygen before maintaining at 2% for the procedure. They were then injected intramuscularly with 30μL of either PBS (undamaged control), 50μM CTX, and 8 or 16 μM of the purified protein into their right tibialis anterior muscle. Mice were then allowed to recuperate for either 5 or 10 days before sacrificing by carbon dioxide asphyxiation and cervical dislocation.
Dissection of tibialis anterior (TA) and extensor digitorum longus (EDL)
The TA muscles from mice were dissected, weighed and frozen on liquid nitrogen cooled isopentane prior to storage at -80˚C. The EDL muscle was dissected from the undamaged contralateral hind limb of experimental mice and immediately placed in a 2mg/mL collagenase solution (Sigma Aldrich, UK) and incubated at 37˚C with 5% CO 2 for 2 hours to isolate the single fibres as previously described [37].
Proliferation and migration of satellite cells
To determine the proliferation of SCs and myogenic differentiation, isolated single fibres were cultured for up to 48 hours at 37˚C with 5% CO 2 in single fibre culture medium (SFCM-DMEM, 10% (v/v) horse serum, 1% (v/v) penicillin-streptomycin and 0.5% (v/v) chick embryo extract) supplemented with either 0, 0.3, 1 or 3 μM of the purified protein prior to fixation in 2% (w/v) paraformaldehyde in PBS and maintained in PBS prior to immunocytochemistry.
The migration of muscle fibre SCs was analysed as described previously [37]. Briefly, the isolated single fibres were cultured for 24 hours in SFCM before transferring to SFCM containing 0, 0.3, 1 or 3μM of the purified protein and monitoring by a phase contrast microscope at 37˚C with 5% CO 2 using a 10X objective. A time-lapse video was captured at a rate of 1 frame every 15 minutes for a 24-hour period and analysed to determine the rate of migration.
Immunohistochemistry
The collected TA muscles were mounted in Tissue-TEK 1 OCT compound in an orientation allowing the transverse sections of 13μm thickness to be obtained using a cryo microtome. The tissue sections were incubated in permeabilisation buffer [20mM HEPES, 3mM MgCl 2 , 50mM NaCl, 0.05% (w/v) sodium azide, 300mM sucrose and 0.5% (v/v) Triton X-100] for 15 minutes at room temperature. To remove the excess permeabilisation buffer, 3 x 5 minute washes were performed using PBS before the application of wash buffer [PBS with 5% fetal bovine serum (v/v), 0.05% (v/v) Triton X-100] for 30 minutes at room temperature.
Primary antibodies were pre-blocked in wash buffer for 30 minutes prior to application onto muscle sections overnight at 4˚C. In order to remove the primary antibodies, muscle sections were washed three times (5 minutes each) in wash buffer. The sections were then incubated with species-specific secondary antibodies that were conjugated with Alexa Fluor 488 or 594. The secondary antibodies were pre-blocked in wash buffer for minimum of 30 minutes before their application onto the slides and incubated for 1 hour in the dark at room temperature. Thereafter, the muscle sections were washed 3 x 5 minutes in PBS to remove the unbound secondary antibodies. Finally, the slides were mounted in fluorescent mounting medium, and the myonuclei were visualised using 4, 6-diamidino-2-phenylindole (DAPI) (2.5μg/mL). The images of sections were obtained using a fluorescence microscope (Zeiss AxioImager) and analysed using ImageJ. Macrophages were detected by F4.80 staining using the Vector Laboratories ImmPRESS Excel Staining Kit. A list of antibodies used in this study is provided in S1 Table.
Statistical analysis
All the statistical analyses were performed using GraphPad Prism 7 and the P-values were calculated using one-way ANOVA followed by Dunnett's post hoc multiple comparisons test.
Protein purification and identification
In order to purify a protein with a molecular weight of around 50kDa (as predicted for group III SVMPs) from the venom of C. atrox, a two-dimensional chromatography approach was employed. Following the initial fractionation of venom via anion exchange chromatography (Fig 1A and 1B), the selected fractions (14-18) with a highly abundant protein at approximately 50kDa were pooled and run through a gel filtration chromatography column (Fig 1C and 1D). The fractions (62-67) were pooled and run through the same gel filtration column again to refine the purification (Fig 1E and 1F). Finally, a pure protein with a molecular weight of around 50kDa was isolated. Mass spectrometry characterisation of the tryptic digested peptides of this protein and further Mascot analysis confirmed it to be a similar or identical protein to vascular apoptosis inducing proteins (VAP) such as VAP2, a protein with a molecular weight of 55kDa (an identical molecular weight to the purified protein) [38], which is a group III metalloprotease (Fig 1G). The identified peptide sequences of the purified protein covered around 43% of the sequence of VAP2A (highlighted in red in Fig 1G). The purified protein has been referred to as 'CAMP' to denote C. atrox metalloprotease throughout this article.
Fibrinogenolytic and collagenolytic activities of CAMP
By using fluorogenic substrates, the protease activity of CAMP was analysed in comparison to the whole venom. CAMP displayed no serine protease activity as it failed to cleave a serine protease selective fluorogenic substrate, BAAMC although the whole venom displayed significant serine protease activity (Fig 2A). However, it showed high levels (similar to the whole venom) of collagenolytic activity (Fig 2B). Furthermore, the ability of CAMP to digest fibrinogen was analysed by incubating it with human plasma fibrinogen. The SDS-PAGE analysis of samples that were taken at different points of incubation confirmed that CAMP is capable of cleaving Aα and Bβ chains of fibrinogen although it was unable to cleave the γ chain ( Fig 2C). The digestion of fibrinogen with CAMP appears to be rapid as the levels of Aα and Bβ chains of fibrinogen were reduced significantly as early as 30 minutes of incubation. These results corroborate CAMP as an SVMP with collagenolytic and fibrinogenolytic activities, which may affect the collagen in the BM around the local tissues at the bite site and fibrinogen in the blood.
CAMP inhibits human platelet aggregation
The ability of CAMP to inhibit agonist-induced platelet activation was analysed using human platelet-rich plasma (PRP) by optical aggregometry. The pre-treatment of human platelets (PRP) with CAMP (50μg/mL) has significantly inhibited 0.5μg/mL CRP-XL-induced platelet aggregation (Fig 2D and 2E). This data confirms the ability of CAMP to affect human platelet activation.
CAMP induces haemorrhage and fluctuations in muscle size in mice
In order to determine the mechanisms through which SVMPs induce permanent muscle damage, CAMP was used as a tool to determine its pathological effects in TA muscle of mice in comparison with CTX. The intramuscular injection of CAMP induced haemorrhage in the damaged muscles and thereby, caused swelling and increase in muscle weight after five days of administration (Fig 3A and 3B). However, CTX did not induce haemorrhage or swelling although the muscle weight was reduced compared to the controls at the same time point. In contrast, after ten days of administration, muscle weight in CAMP-treated mice was decreased similar to CTX-treated muscle (Fig 3A and 3C). These data demonstrate that CAMP is capable of inducing haemorrhage and swelling and thereby, increases in muscle weight initially although it decreases at a later time point.
Attenuated regeneration in CAMP-damaged muscle
We examined the cellular processes underpinning the morphology of skeletal muscle and assessed muscle regeneration after damage induced by CAMP. Haematoxylin (H) and eosin (E) staining facilitates the identification of cellular organisation within a tissue and also the presence of fibres containing centrally located nuclei (CLN), which is an indicator of muscle regeneration. Five days after tissue damage, muscles treated with CTX contained many large fibres with CLN ( Fig 4A). Furthermore, there were regions of high cell density between fibres displaying CLN. In contrast, 5 days after CAMP damage the number of fibres with CLN was less abundant and smaller than in CTX damaged muscle (Fig 4A and 4B). Additionally, there were areas of sparsely populated regions between fibres. Ten days after CTX damage, large a Coomassie stained gel shows the purified protein at approximately 50kDa. G, the tryptic digested samples of the purified protein were analysed by mass spectrometry and the identified peptide sequences match (via Mascot search) with the known sequence of VAP2A (coverage 43%; the mass spectrometry-identified peptide sequences of the purified protein are shown in red) and confirms that the purified protein is most likely to be VAP, a group III metalloprotease. The purified protein was named as CAMP to represent C. atrox metalloprotease. fibres with CLN were evident with very little space between muscle fibres (Fig 4C and 4D). The fibres appeared to be regular in terms of shape and size, evidencing robust muscle regeneration. Whereas, at the same time point, muscle damaged with CAMP displayed smaller fibres with CLN and inter-fibre regions populated with cells were prominent (Fig 4C and 4D). Next, we documented the profile of dying muscle fibres, facilitating the infiltration of circulating immunoglobulins (Ig) into the damaged fibres. Five days after CTX injection, low density of small calibre fibres displayed the infiltration by IgG (Fig 4E-4G). In contrast, at the same time point, CAMP treatment resulted in not only a higher density of fibres with infiltrated IgG, but they were also of larger size (Fig 4E-4G). By day 10, very few dying fibres were present in CTX treated muscle, however dying fibres were prominent in CAMP treated muscles ( Fig 4H and 4I). We then examined the presence of regenerating muscle fibres, facilitated through the expression of embryonic myosin heavy chain protein (MYH3). Muscle regeneration was clearly evident in muscles damaged by CTX at day 5 (Fig 4J and 4K). Large numbers of evenly sized fibres expressing MYH3 featured in CTX-damaged tissue (Fig 4J and 4L). In contrast to CTX treatment, the number of regenerating fibres in CAMP-treated muscle was lower and when present were of heterogeneous size (Fig 4J-4L). By Day 10, the expression of MYH3 has been cleared in CTX-damaged muscle and when present was in very large fibres (Fig 4M-4O). In contrast, MYH3 expression was clearly evident at day 10 in CAMP-damaged muscle but in smaller, non-uniform fibres (Fig 4M-4O). Next, we examined the impact of CAMP and CTX on blood vessels through immunostaining with the endothelial cell specific antibody, CD31. Snake venom metalloprotease-induced permanent muscle damage At both 5 and 10 days, the number of capillaries serving each regenerating fibre was greater in the CTX treated sample compared to CAMP (Fig 4P-4S). Importantly, the number of capillaries serving each regenerating fibre in the CTX treated sample was identical to the undamaged sample. Moreover, the degree of macrophage infiltration into the damaged area was analysed, as these cells are key to effective muscle regeneration. The density of macrophages in damaged muscle was greater in the CTX treated muscle compared to CAMP at day 5 (Fig 4T and 4U). However, by day 10, the situation was reversed; there was a greater density of macrophages in the CAMP treated samples compared to CTX (Fig 4V and 4W).
CAMP extensively damages the extracellular matrix (ECM) surrounding the myofibres
Efficient regeneration of skeletal muscle following acute damage is contingent on stem cells capable of replacing damaged tissue and their highly ordered formation into myotubes/fibres, a process orchestrated by the ECM. The organisation of collagen IV, a major BM component of muscle fibres was analysed as described previously [39]. A thin circle of collagen IV surrounding muscle fibre was evident 5 days after CTX treatment (Fig 5A). In contrast, at an identical time after CAMP treatment, muscle displayed large irregular, thick depositions of collagen IV (Fig 5A and 5B) but by day 10, the picture was even more polarised, as CTX-damaged muscle showed a relatively normal distribution of collagen IV (Fig 5C and 5D). In contrast, very few fibres from CAMP-treated muscle (at day 10) displayed a ring of collagen IV, and instead this protein was localised at thick foci (Fig 5C and 5D). A near-identical pattern was documented for the distribution of laminin, another major component of the muscle fibre ECM (Fig 5E-5H).
Furthermore, the impact of CTX and CAMP on molecules that are associated with linking the contractile apparatus to the ECM was investigated. Dystrophin is normally localised under the sarcolemma of mature muscle fibres. Its expression was evident around some of the larger Snake venom metalloprotease-induced permanent muscle damage regenerating muscle fibres 5 days after CTX damage (Fig 5I). Whereas, very few fibres expressing dystrophin were detected at a similar time point in CAMP-treated muscles (Fig 5I). However, when present, the thickness of the dystrophin expression domain was similarly reduced by the two treatments (Fig 5J). At day 10, most of the fibres from CTX-treated muscle displayed a continuum of dystrophin expression, although at a lower thickness compared to undamaged tissue (Fig 5K and 5L). However, very few fibres with a ring of dystrophin were present in CAMP-treated muscles at day 10 ( Fig 5K). Furthermore, the domain, when present was thinner than both control as well as CTX treated muscles (Fig 5L). Then, the distribution of nNOS, a protein that localises to a sub-sacrolemmal position which is dependent on its binding to dystrophin was assessed. At 5 days after treatment, very little nNOS was present in either CTX or CAMP-damaged muscles (Fig 5M and 5N). By day 10, a thin band of nNOS was evident in CTX-treated muscle but not in the muscle damaged by CAMP (Fig 5O and 5P). Lastly, the muscles were analysed to determine the presence of remaining CAMP in damaged tissues. The immunohistological profiling showed that CAMP was clearly present at both 5 and 10 days after its administration (Fig 5Q and 5R). These results show that CAMP treatment damages not only the ECM of muscle fibres but also affects intracellular components that link it to the contractile machinery.
CAMP affects the functions of satellite cells (SCs)
The role of SCs adjacent to the muscle fibres is critical for muscle regeneration. In order to determine the impact of CAMP on SCs, we have isolated myofibres from intact EDL muscles and exposed them to a range of concentrations of CAMP. As early as 24 hours after CAMP treatment, it was evident that there was a concentration dependent disturbance to the collagen component of the ECM around muscle fibres. A uniform layer of collagen expression was detected in untreated fibres (Fig 6A). At the lower concentration, CAMP caused a localised denuding of the myofibre (Fig 6B), whereas the higher concentration resulted in the absence of collagen from most parts of the fibres and caused it to concentrate in specific locations ( Fig 6C). The cell growth, proliferation and migration were monitored on the isolated single muscle myofibres over a 48 hour time period. SCs were immunostained using the myogenic transcription factors, Pax7 (uncommitted cells) and MyoD (activated cells) in order to monitor the progression of cells through myogenesis. The concentrations of above 0.3μM of CAMP induced hypercontraction, which is indicative of extensive fibre damage. At 0.3μM, viable fibres were present, and revealed that CAMP significantly decreased the number of associated SCs (Fig 6D). Furthermore, the number of SC clusters was reduced per fibre (Fig 6E), although each cluster at the lower concentration had more cells than untreated fibres (Fig 6F). Analysis of differentiation was only possible at the lowest concentration of CAMP (Fig 6G) as at higher concentrations, hypercontraction prevented this analysis. The migration speed was calculated between 24 and 48 hours and was found to decrease significantly as the concentration of CAMP increased (Fig 6H). These data demonstrate that CAMP is able to affect both the proliferation and migration of SCs but not the differentiation.
Discussion
The swelling and necrosis at the bite site as well as permanent muscle damage are common effects of snakebite envenomation (particularly viper bites). These effects frequently lead to amputation and therefore disable victims, which adds to their inability to earn money, and exacerbates the poverty that is already experienced by the vast majority of snakebite victims [4]. Here we have purified a metalloprotease from one of the most studied venomous snake species, C. atrox. Although deaths from this snake are now uncommon, disfigurement is still a prevalent side effect for survivors. SVMPs are a predominant component in viper venoms that are involved in inducing the local envenomation effects including muscle damage. The ability of SVMPs to degrade collagen has been established, but its impact on permanent muscle damage under in vivo settings has not been previously demonstrated in sufficient detail. Therefore, we deployed a metalloprotease from the venom of C. atrox and analysed its impact on skeletal muscle damage in comparison to a three-finger toxin, CTX from the venom of Naja pallida. Mass spectrometry analysis of the purified protein suggests it to be a group III metalloprotease, which possess a metalloprotease domain as well as a disintegrin-like and cysteine-rich domains [40]. Based on the peptide sequences identified by the mass spectrometry, the purified protein is likely to be VAP2 or one of its heterodimers; VAP2A or VAP2B, both of which are vascular apoptosis inducing proteins (38) that are known to be haemorrhagic [41] and in the case of VAP2B to inhibit collagen-induced platelet activation [42]. Due to the limited peptide sequences identified by mass spectrometry for the purified protein, we are unable to conclude whether the purified protein (CAMP) is identical to VAP2 or either of its heterodimers. CAMP was characterised to be a collagenolytic and fibrinogenolytic enzyme. It also inhibited CRP-XL-induced platelet aggregation; group III metalloproteases are known to interact with the integrin α 2 β 1 , binding to the α 2 subunit and causing the shedding of β 1 subunits [43]. However, VAP2B (a protein described from C. atrox) has been reported to inhibit collagen induced platelet aggregation by binding to collagen [44], although whether the SECD sequence found in disintegrin-like domains is able to bind CRP-XL in the same way, is unknown.
In order to determine the impact of SVMPs in stimulating permanent muscle damage, different concentrations of CAMP were administered in mice along with CTX and control groups and the effects were analysed at five and ten days after the administration. We suggest that this occurs at two levels; by breaking down the ECM which normally acts as a scaffold for the formation of new muscle fibres and around existing blood vessels and secondly attenuating properties of resident stem cells that are essential to effective tissue repair. SVMPs are known for their collagenolytic activities and for targeting various components of the BM in the vasculature and inducing haemorrhage [45]. In line with previous studies, here we demonstrate that CAMP induces haemorrhage and affects the architecture of collagen and laminin. The destruction of the collagen based ECM may be the key to long-term tissue destruction wrought by CAMP. The disintegrin-like and cysteine-rich domains have already been identified as essential to the haemorrhagic activity and ECM degradation attributed to the PIII metalloproteases [46]. This is in contrast to myotoxic PLA2 and three-finger toxins that are well documented in causing membrane permeabilisation and consequentially myonecrosis via the hydrolysis of membrane phospholipids or imbedding directly into the membrane respectively [47][48][49]. Our data emphasise that CAMP in comparison to CTX significantly hindered the regeneration of skeletal muscle fibres most probably by disturbing the organisation of the ECM. The elevated levels of necrosis seen five days after administration with CAMP improved after ten days, although it was still evident. However, it must be noted that at day 10, the CTX treated muscles had almost completely regenerated with healthy fibres. Moreover, very low levels of MYH3 were detected five days after CAMP treatment. In CTX-treated muscles, this marker of regenerating fibres was clearly evident and present at a high level. This indicates that the initiation of the regeneration process was attenuated by CAMP in comparison to CTX. Muscle regeneration is dependent on blood supply and clearance of damaged fibres. We show here that both these cellular compartments are affected in a detrimental manner by CAMP. We found that the number of capillaries serving each regenerating muscle fibre was smaller in CAMP treated muscle compared to CTX. Importantly the number of capillaries serving each fibre in damaged CTX muscle was the same as in undamaged regions. These results show that capillaries as well as muscle fibres are damaged by CAMP whereas it is only the latter in CTX treated tissue. Additionally, we show that there was a greater influx of macrophages into the CTX damaged muscle compared to regions affected by CAMP. Furthermore, the density of macrophages decreased in CTX treated muscle over time, attesting to regeneration. In contrast, the density of macrophages in CAMP treated muscle was lower at day 5 compared to CTX, possibly indicating an attenuated clearance process. Importantly the density of macrophages did not change in the CAMP treated muscle over 10 days suggesting on-going muscle damage. Although the abundance of MYH3 increased in CAMP-treated muscles by day 10, its expression in CTX-injured muscles was almost undetectable, signifying advanced regeneration. This was reflected in the appearance of dystrophin and nNOS at their normal sub-sarcolemmal position. In keeping with the notion that CAMP treatment not only affects the degree of regeneration but also its timing, we showed that very few fibres expressed dystrophin in its normal position and a significantly reduced expression of nNOS was observed even at day 10. Most importantly we show that CAMP is still present at the site of injury even 10 days after its administration and that it profoundly disorganises the ECM.
The single fibre experiments highlight another aspect to explain the attenuated muscle regeneration following CAMP-mediated muscle damage. We demonstrate that the proliferation and migration of SCs was significantly reduced by CAMP treatment. Both of these factors are key in promoting muscle regeneration. Therefore, CAMP may bring about permanent impairment of muscle organisation and function by firstly destroying muscle fibres, secondly breaking down the organisation of the ECM. This is required by the SCs in order to align and fuse in a coordinated manner and lastly by diminishing the ability for SCs to expand their numbers and migrate to the site of injury to enact efficient regeneration. It is clear that current ASV treatment is not effective at preventing muscle damage. Although translating the results of this study into therapeutics might be difficult, these will improve the understanding of SVMP-induced permanent muscle damage. The ability of group III metalloproteases to bind components of the BM and prolong muscle exposure to their myotoxic effects suggests a therapeutic agent that is capable of interacting with these enzymes and non-enzymatic domains and preventing the longevity of these proteins in the area surrounding fibres may be able to speed up the rate of regeneration considerably. Moreover, any drugs aimed at treating this aspect of snakebite envenomation may struggle to reach it intravenously, and therefore, they may have to be administered via multiple local injections considering the widespread damage to microvasculature [32] and consequential lack of blood supply to affected tissues.
ASV is the only effective treatment for systemic envenoming, however local venom pathology is largely unaffected by ASV when treatment is not immediately administered [50]. ASV is composed of large immunoglobulins that appear to struggle to reach the areas affected by SVMPs. The combination of small vessel destruction combined with BM cleavage results in a poor blood supply and therefore weak neutralisation by intravenously administered ASV. Local injections of ASV have also been found to be of no benefit to the snakebite victims [51]. However, there are a range of matrix metalloprotease inhibitors that have undergone testing for their specificity to SVMPs and some promising compounds have been identified [52]. The small molecule inhibitors aimed at the metalloprotease domain such as batimastat [53] have been tested extensively and they were found to abrogate the haemorrhagic effects of venom if administered immediately after envenoming. Given their haemorrhagic effects are largely dependent upon collagen degradation, it is reasonable to postulate that this prevention of haemorrhagic effects may also apply to muscle damage. Moreover, metal chelating agents such as EDTA have also been tested in vivo at non-toxic doses and found to prevent venom-induced lethality [54].
The need for immediate administration is of course unrealistic with conventional ASV but small stable inhibitors have the potential to be spread and made available to those in areas with a high density of snakebites. Multiple local injections do bring the potential for delivery directly to the bite site and administering to multiple sites may overcome the problematic spread of drug through a site of damaged muscles and vessels. Future experiments should aim to investigate the effect of these drugs on BM components, using both pre-incubation with drugs and post envenomation delivery models. Overall, the complete destruction or loss of a range of BM and dystrophin-glycoprotein complex components as well as the effect of this SVMP on muscle regeneration highlights the significant difficulties involved in treating the necrosis and muscle damage associated with snakebite envenomation. Hence, this study provides greater insights into the understanding of SVMP-induced permanent muscle damage and local snakebite envenomation effects.
Supporting information S1 | 2019-01-31T14:03:01.522Z | 2019-01-01T00:00:00.000 | {
"year": 2019,
"sha1": "f22b6dcb112770b4ae590a7b42386586d738f818",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosntds/article/file?id=10.1371/journal.pntd.0007041&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "467f2fc3763a4d8331d87bf94efd3e6eef3110e7",
"s2fieldsofstudy": [
"Biology",
"Chemistry"
],
"extfieldsofstudy": [
"Medicine",
"Biology"
]
} |
235665952 | pes2o/s2orc | v3-fos-license | A Morphometric Study of the Mandibular Foramen in Dry Adult Human Mandibles in a Black Malawian Population
The inferior alveolar nerve block(IANB) technique is a common technique performed on patients in dental surgery, placement of mandibular implants and other procedures involving the mandible. Precise identification of the mandibular foramen (MF) is essential for dental surgeons to accurately administer local anesthetics. Inaccurate localization of the mandibular foramen may result in IANB failure and injury to neurovascular tissues. Therefore, this study aimed at investigating the precise location of the MF from various anatomical land marks in dry adult human mandibles of Malawian population. The study was conducted on 29 dry adult human mandibles of unknown sex of Malawian origin from the Anatomy Division collection of human skeletons housed in the Biomedical Sciences Department, College of Medicine, University of Malawi. To determine the position of the mandibular foramen, distances from mandibular foramen to anterior margin, posterior margin, mandibular notch, gonial angle and mandibular base using a Vernier caliper were measured. The mean distance of the MF from posterior margin of mandibular ramus was 11.26± 1.22 mm (right side) and 11.47± 1.35 mm (left side), from the anterior margin 20.85± 3.12 mm (right side) and 20.85± 3.22 mm (left side) mandibles. The mean distance between mandibular notch (MN) and MF was 23.87± 2.61 mm (right side) and 23.53 ± 2.65 mm (left side). The mean distance between mandibular base(MB) and MF for the right and left were 28.47 ± 2.90 mm and 27.85 ± 2.99 mm respectively. The inferior limit of the mandibular foramen was located at 24.69 ± 3.65 mm (right side) and 24.25 ± 2.77 mm (left side) to the angle (AG) of the mandible. The findings of this study show that the anterior margin mean distance from the MF for both right and left mandibles seem to be bilateral symmetrical suggesting the interpretation that the needle for IANB could be inserted at about 21 mm from the anterior margin to the MF in an adult of Malawian origin during surgery.
INTRODUCTION
In dentistry, the inferior alveolar nerve block (IANB) is the most common technique used to administer local anesthetics to the mandibular teeth, gingiva and lower lip (Shalini et al., 2016). However, there are reports of frequent technique failure and the failure rate is as high as 20 % to 25 % (Shalini et al.). According to Sastya & Preeti (2016) the failure rate in this technique could be as high as 45 %.The causes of the IANB technique failure are alluded to variations in the location of the mandibular foramen (Russa & Fabian, 2014) and presence of accessory mandibular foramen through which additional branches of inferior alveolar nerves and vessels course in some individuals (Samanta & Kharb, 2013). The MF gives the entry point of the inferior alveolar nerve, which is accompanied by inferior alveolar artery and vein. This neurovascular bundle courses through the mandibular canal supplying sensation to the lower teeth, associated gingivae, the mucosa and skin of lower lip and chin. The inferior alveolar nerve exists the mandibular canal through the mental foramen (Drake et al., 2015;Rajkumari et al., 2017;Moore, 2018).
The location of MF is clinically important in osteotomy, orthognathic reconstructive surgery of the mandible, dental implant procedures and anthropometric assessment (Juodzbalys & Wang, 2010;Lalitha et al., 2016;Rajkumari et al.). The proximity of the anesthetic injection needle to the MF determines the success of the IANB technique. If the hypodermic needle is improperly placed, the anesthetic may not be properly deposited and will not induce the required anaesthetizing effect on the alveolar nerve. Besides, it may lead to some complications such as hemorrhage, injury to the neurovascular bundle, and necrosis of the mandibular ramus (Sghaireen et al., 2020). The accuracy of the IANB technique is based on thorough anatomical knowledge of the precise location of the mandibular foramen.
Anatomical variations in the location of the MF have been reported in a number of studies worldwide. The variations in the MF location are due to race and growth rate among populations (Russa & Fabian;Correa et al., 2019). Some studies relate the location of the MF to the occlusal plane (Mbajiorgu, 2000;Russa & Fabian). Mwaniki & Hassanali (1992) in their study on localization of mandibular foramen among adult Kenyan Africans have reported that the MF is below the occlusal plane in about 64.6 % and at the level of the occlusal plane in about 31.1 %. This differs from a study by Nicholson (1985) that notes that 75 % of mandibles studied on an East Indian population have the MF below the occlusal plane and 22.5 % are above the occlusal plane. In Tanzanian mandibles, the MF is above the occlusal plane at the mandibular first molar and second premolar reference points (Russa & Fabian). Moreover, most of these studies on the mandibular foramen relate its location to the anatomical landmarks such as anterior and posterior margins of mandibular ramus, gonial angle and the lowest point on the mandibular notch (Samanta & Kharb; Sastya & Preeti; Shalini et al.). A morphometric study among South African dry mandibles has reported that the MF is always more located on the posterior margin of ramus of mandible than on the anterior region in both males and females (Tshite, 2017). The MF is more situated in the superior part than the inferior part of the mandible, more towards the sigmoid notch than the inferior margin of ramus of the mandible (Tshite). In the light of the foregoing, it is clear that the MF dimensions and location might be population specific even within the same ethnic grouping and region. It was because of this that this study was designed to obtain baseline morphometric data for this population to determine the precise location of the MF from the anterior margin, posterior margin, mandibular notch, mandibular base and gonial angle in adult dry mandibles of black Malawians.
MATERIAL AND METHOD
This study was conducted in accordance with the Government of Malawi Anatomy Act No.14 of 1990 and was approved by the University of Malawi's College of Medicine Research and Ethics Committee (COMREC) with a clearance number P02/10/872. The study was conducted on adult cadavericdry mandibles obtained from the Anatomy Division Collection of Human Skeletons housed in the Biomedical Sciences Department, College of Medicine, University of Malawi. The sample included only dry mandibles prepared from adult black Malawian cadavers. A total of 29 dry adult human mandibles of unknown sex, aged 21-82 years, were studied from the skeletal collection. Malformed and broken bones were excluded from the sample. To precisely locate the mandibular foramen, the following parameters were taken using a sliding Vernier caliper ( Fig. 1
RESULTS
The mean and standard deviation (SD) values of distance of the MF from different landmarks on the right and left side of the mandibles are presented in Tables I and II Data analysis. The mean and standard deviation for each distance used to locate the mandibular foramen were calculated separately for both right and left mandibles using Stata. The mean distances of this population were compared with other ethnic groups from previous similar studies. Differences were considered statistically significant when p<0.05. The mean distance of the MF from posterior margin of the mandibular ramus was 11.26± 1.22 mm(right side) and 11.47± 1.35 mm(left side), from the anterior margin 20.85± 3.12 mm (right side) and 20.85± 3.22 mm(left side) mandibles. The mean distance between MN and MF was 23.87± 2.61 mm (right side) and 23.53 ± 2.65 mm (left side). The mean distance between MB and MF for the right and left sides were 28.47 ± 2.90 mm and 27.85 ± 2.99 mm respectively. The inferior limit of the mandibular foramen (MF) was located at 24.69 ± 3.65 mm (right side) and 24.25 ± 2.77 mm (left side) from the angle (AG) of the mandible. Figure 3 presents the mean distances of MF from the anterior and posterior margins of mandibular ramus, mandibular notch, mandibular base and angle of mandible.
DISCUSSION
Antero-posterior dimensions. In this study,a significant difference is observed between the posterior and anterior mean distances (PB-MF) for the right and left mandibles(P = 0.002). The posterior mean distances (PB-MF) for the right and left mandibles are less than the anterior mean distances (AB-MF). This suggests that the MF appears to be located more towards the posterior margin than the anterior margin. Additionally, there is no significant difference between the right and left mean distances of PB-MFand AB-MF (p>0.05). Similarly, Tshite, Samanta & Kharb; Sastya & Preeti; Shalini et al. have asserted that no significant differencein the mean distances of PB-MF and AB-MFis noted between the right and left mandibles. A study by Tshite on black South African dry mandibles established that the mean of AB-MF is 18.8 mm on the right and 18.9 mm on the left whereas the mean of PB-MF is 13.7 mm on the right and 13.4 mm on the left suggesting that the position of MF on dry bones is more towards the posterior margin of the ramus of the mandible than on the anterior margin. The results from a study of dry mandibles by Samanta & Kharb, among an Indian population reports that the averagedistance of the MF from the anterior margin of mandibular ramus is 15.72 ± 2.92 mm and 16.23 ± 2.88 mm right and left side respectively. The posterior margin is 13.29 ± 1.74 mm and 12.73 ± 2.04 mm from right and left side respectively. Thus the MF is located more towards the posterior margin than the anterior margin. Furthermore, Shalini et al. note that the mean distance of mandibular foramen from the anterior margin of ramus of mandible is 17.11±2.74 mm on the right side and 17.41±3.05 mm on the left side; from the posterior margin, it is 10.47±2.11 mm on the right side and 9.68±2.03 mm on the left side showing that the MF is located more towards the posterior margin.
However, other studies have shown a significant difference in the AB-MF between the right and left mandibles (p<0.05) (Russa & Fabian). Moreover, the AB-MF mean distance for the present study is generally wider than and in variance with other studies done in Indian populations whose results are several millimeters smaller (Samanta & Kharb; Shalini et al.).
Comparing the PB-MF mean distances of this study with those of other populations, there is a statistical significant difference (p<0.05) between the right and left sides. For this reason, observations of the present study suggest that the MF location and morphometrics variation could be population specific. This supports earlier studies that have reported that mandibular dimensions are race and population specific (Ashkenazi et al., 2011). These variations could not only be due to geographical, nutritional and genetic factors, but also different techniques and methodologies used to assess the MF morphometrics (Ashkenazi et al.) and the limitation of the present study due to a small sample size.
Supero-inferior and gonial angle -mandibular foramen dimensions. With regard tothe superior (MN-MF) and inferior (MB-MF) dimensions, it has been established in the present study that the mean distance between MN and MF for the right and left is less than the mean distance between MB and MF for the right and left mandibles suggesting the interpretation that the MF is located more towards the mandibular notch than the mandibular base. Likewise, Shalini et al., Mbajiorgu, and Tshite found similar results in that the MF is located more towards the mandibular notch than the mandibular base.
Also, a significant difference (p<0.05) is observed in the MN-MF mean distance between the present study and other studies (Shalini et al.; Mbajiorgu; Tshite). Shalini et al. found that the MN-MF mean distance to be 21.74 ± 2.74 mm and 21.92 ± 3.33 mm for the right and left sides respectively. They also reported the MB-MF mean distances of 22.33 ± 3.32 mm for right and 25.35 ± 4.5 mm for the left side. Tshite also found that the MF is situated more towards the superior margin of the ramus of the mandible than the inferior margin.
According to Tshiste (2017) the MN-MF was shown to be 20.1 3.0 mm and 20.0 3.1 mm on the right and left sides respectively. He recorded 22.9mm on the right and 22.7mm on the left as mean distances from MF to the Gonial angle which he regarded as the most inferior point of the mandible. Whereas for adult black dry Zimbabwean mandibles, a mean distance of 22.50 ± 0.50 mm and 28.44 ± 0.65 mm from MN and AG to MF were recorded respectively (Mbajiorgu). For a Tanzanian population, Russa & Fabian reports that the mean distance from MF to the MN to be 21.54 ± 3.02 mm and 20.70 ± 2.98 mm on the right and left side respectively. They established 26.23 ± 3.95 mm and 25.68 ± 4.03 mm as AG-MF mean distance for the right and left side respectively. The current study found 24.69 ± 3.65 mm and 24.25 ± 2.77 mm as the average distances from MF to the mandibular angle for right and left side; with the right demonstrating a higher reading. An independent samples t-test reveals that there is no significant difference the right and left sides for AG-MF (p=0.1651).
Besides, there is no significant difference between the right and the left sides for MN-MF (p=0.1477) and MB-MF(p=0.065) in the current study.
Comparing to previous studies done in Tanzania and South African black populations an Independent samples ttest showed that there is a no statistically significant (Table II). However, there is a statistical significant differencein AG-MF distance for both right and left sides to that of an Indian population studied by Samanta & Kharb (Table III). In the present study, if we consider the AG-MN and MB-MN dimensions, still the MF is situated more towards the MN. However, the finding of the present study differs from the study by Samanta & Kharb, which shows that the MF is located more towards the inferior margin than the mandibular notch.
Angle of mandible.
A study by Agrawal & Kanwar (2018) found that angle of mandible in 29 male and 23 female mandibles of right side, mean and standard deviation is 122.240 ± 6.71 and 124.610 ± 5.33 respectively and "P" value is 0.1734. Similarly, in case of left side mean and SD is 121.590 ± 6.65 and 124.170 ± 5.87 respectively and "P" value is 0.1489. Shalini et al. found that the angle of the mandible was 117.470 ±4.95 on the right side and 117.470±5.88 on the left side. There was no statistically significant difference between the angles of the mandible on the right and left sides (P>0.05). This is in agreement with the results of the present study in which there is no significant difference between the right and left angle measurements (p=0.5154).The mean Gonial angle measurements in the present study were 116.070± 4.02 and 115.750 ± 4.25 for right and left sides respectively, with an overall range of 1050 -1260. | 2021-06-29T01:25:41.582Z | 2021-01-01T00:00:00.000 | {
"year": 2021,
"sha1": "d4c365e9d087a3493532fdc2aa9ae4eea28b7706",
"oa_license": "CCBYNC",
"oa_url": "http://www.scielo.cl/pdf/ijmorphol/v39n2/0717-9502-ijmorphol-39-02-390.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "d4c365e9d087a3493532fdc2aa9ae4eea28b7706",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
216172756 | pes2o/s2orc | v3-fos-license | Metformin treatment in heart failure with preserved ejection fraction: a systematic review and meta-regression analysis
Background Observational series suggest a mortality benefit from metformin in the heart failure (HF) population. However, the benefit of metformin in HF with preserved ejection fraction (HFpEF) has yet to be explored. We performed a systematic review and meta-analysis to identify whether variation in EF impacts mortality outcomes in HF patients treated with metformin. Methods MEDLINE and EMBASE were searched up to October 2019. Observational studies and randomised trials reporting mortality in HF patients and the proportion of patients with an EF > 50% at baseline were included. Other baseline variables were used to assess for heterogeneity in treatment outcomes between groups. Regression models were used to determine the interaction between metformin and subgroups on mortality. Results Four studies reported the proportion of patients with a preserved EF and were analysed. Metformin reduced mortality in both preserved or reduced EF after adjustment with HF therapies such as angiotensin converting enzyme inhibitors (ACEi) and beta-blockers (β = − 0.2 [95% CI − 0.3 to − 0.1], p = 0.02). Significantly greater protective effects were seen with EF > 50% (p = 0.003). Metformin treatment with insulin, ACEi and beta-blocker therapy were also shown to have a reduction in mortality (insulin p = 0.002; ACEi p < 0.001; beta-blocker p = 0.017), whereas female gender was associated with worse outcomes (p < 0.001). Conclusions Metformin treatment is associated with a reduction in mortality in patients with HFpEF.
Background
Heart failure (HF) with preserved ejection fraction (HFpEF) is a distinct phenotype hallmarked by clinical signs and symptoms of HF coupled with a normal ejection faction (EF ≥ 50%) and evidence of increased left ventricular (LV) pressures and impaired LV filling on echocardiography [1][2][3]. HFpEF accounts for almost half of the cases of HF and carries an equally poor prognosis to those with HF with reduced ejection fraction (HFrEF), with an estimated 4-year mortality rate of 32% [4]. In contrast to HFrEF, where several therapies have shown good long-term morbidity and mortality outcomes, despite multiple aetiologies leading to the same pathophysiological end-point [1][2][3]5], effective therapy options for HFpEF have yet to be established. In part, this is because HFpEF is a heterogenous condition, with phenotypic clusters based on age, gender and comorbid illnesses such as obesity, type 2 diabetes mellitus (T2DM) and hypertension [6,7]. This ultimately leads to dysfunctional metabolic pathways and mechanics within the myocardium resulting in the condition [6,8]. Therefore, establishing therapy that targets these phenotypes may be the means by which HFpEF therapy evolves [6].
Metformin is a common anti-diabetic drug with both systemic and cardioprotective benefits in addition to its hypoglycaemic effect [9,10]. At the cellular level metformin activates adenosine monophosphate-activated protein kinase (AMPK) an important regulator of several metabolic pathways resulting in enhanced glucose utilisation, reduction of protein synthesis and improvement of mitochondrial function [11][12][13]. Furthermore, metformin has been shown to reduce collagen accumulation and potentially reduce LV hypertrophy and improve diastolic function in the diabetic myocardium [14]. Several observational series have shown a reduction in mortality in the HF population [15,16]. Its mortality benefit in the HFpEF population however has yet to be explored. We performed a systematic review and meta-regression analysis to identify whether variations in ejection fraction (EF) impact mortality outcomes in HF patients treated with metformin.
Methods
This systematic review was undertaken in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement [17]. The review was registered with the PROSPERO International prospective register of systematic reviews (ID CRD42019133780) in September 2019.
Literature search
MEDLINE (1946 to October 2019) and EMBASE (1947 to October 2019) electronic databases using Ovid ® were searched for randomised controlled trials and observational studies that assessed the impact of metformin therapy on mortality outcomes in adult HF patients (aged over 18 years). Common search terms included ('heart failure' or 'cardiomyopathies'), ('diastolic heart failure' or 'preserved ejection fraction'), ('metformin' or 'biguanide') and ('mortality' or 'death'). The full MEDLINE and EMBASE search strategies are detailed in the Additional file 1: Appendix S1. Reference mining of articles in the full-text review was undertaken as well as grey literature searching. Searches were restricted to human studies and those reported in the English language.
Study selection
Two reviewers (A.H. and J.S.) independently undertook abstract screening and included studies reporting mortality outcomes in HF patients treated with metformin. Studies were divided as preserved or reduced EF; with the preserved group further subdivided into those that reported proportion of patients with EF ≥ 50% or EF ≥ 40%. and reported the proportion of patients with an EF ≥ 50%. Studies were excluded if [1] diagnosis of HF was based purely on hospital discharge codes and no information specifically pertaining to EF and mortality outcomes were recorded, and [2] quality of methodology was not able to be critically appraised, for example in conference abstracts and unpublished studies. After exclusions based on title and abstract review, two investigators (A.H. and J.S.) independently undertook full text reviews for eligibility. Reference searching of review articles was undertaken to search for additional studies, however review articles were not formally included in the systematic review. Covidence ® (Melbourne, Australia) software was used to track articles in the systematic review process. Conflicts were resolved by a third reviewer (T.H.M.).
Data extraction
Data extraction from eligible studies was undertaken independently by two researchers (A.H. and J.S.). Data was extracted on study design and characteristics, including year of publication, number of subjects, gender, duration of follow-up, medical history (including history of coronary artery disease (CAD), hypertension and peripheral vascular disease (PVD)), baseline treatment with cardio-protective (angiotensin converting enzyme inhibitor (ACEi)/angiotensin receptor blocker (ARB), beta-blocker), and anti-diabetic medications (insulin and sulfonylurea therapy) and EF ≥ 50% on echocardiography. Hazard ratios (HR) with associated 95% confidence intervals (95% CI) on mortality outcomes stratified by the presence or absence of metformin treatment were extracted.
Quality and risk of bias assessment
Quality and risk of bias was assessed using the Newcastle-Ottawa quality assessment scale for cohort studies. This scale assesses the quality of a study based on patient selection, comparability and outcome. Included studies were ranked as good, fair or low quality as outlined in Additional file 1: Appendix S2.
Statistical analysis
Proportion of individuals within each subgroup are expressed as a percentage (%). The meta-analysis was performed using maximally adjusted HR and 95% CI to obtain an overall effect size using a random-effects model. Heterogeneity between studies was tested for mortality outcomes using the Chi square test with a p value < 0.05 being statistically significant. An I 2 statistic was generated with > 20% heterogeneity considered significant. A sampling weight was assigned to adjust for differences in study size contributions between each study by calculating the proportion of patients within each subgroup based on the total number of individuals included in the final meta-analysis. Meta-regression models were then performed to determine the interaction between the presence or absence of metformin therapy on mortality outcomes in each sub-group. All statistical analyses were performed using STATA software (StataCorp LLC 2019, v.16.0, College Station, TX, USA).
Study selection
There were 836 studies identified in the search strategy on mortality outcomes in HF patients treated with metformin, with 10 undergoing full-text review (Fig. 1).
Of these, 4 reported the total proportion of patients at baseline with an EF ≥ 50% (Table 1) and were included in the final analysis [18][19][20][21]. Of the 6 studies excluded from the final analysis, 1 study excluded HF patients with an LVEF ≥ 40% [22] and 5 studies diagnosed HF at baseline based on hospital discharge diagnosis or International Classification of Diseases (ICD) or Medicare codes and did not report EF data [23][24][25][26][27] as outlined in the Additional file 1: Appendix S2. Full text articles reporting mortality in heart failure population n = 10
Study characteristics
Study characteristics, baseline patient data and mortality outcomes are reported in Tables 1, 2 and 3. Populations studied were primarily in the United States of America and Spain. A total of 22, 469 individuals with HF were analysed with 7655 mortality events identified. Of the total number of individuals 10,168 (45%) had a preserved EF, however, one study reported the total proportion of participants with an EF ≥40% [18]. Follow-up ranged from 1 to 4.7 years. All studies used multiple covariates in adjustments of the effect size (Table 3) with two studies performing propensity matching in the analysis [18,21].
Risk of bias and quality assessments
All studies included were observational, retrospective cohort studies (Table 1). Based on Newcastle-Ottawa quality assessment, all studies were deemed high quality (Additional file 1: Appendix S3).
Patient characteristics
Overall, patients in the metformin group tended to be younger than those in the non-metformin group (age 71.2 ± 4 years vs. 73.3 ± 3 years, respectively; Table 2). In the metformin group, 1796 (39%) of patients were female, 3499 (76%) were on an ACEi or ARB and 1351 (29%) were on insulin therapy. In contrast, in the non-metformin group 7890 (44%) were female, 11,546 (65%) were on an ACEi or ARB and 9762 (55%) were on insulin therapy.
Discussion
This review demonstrates that metformin is associated with an 18% mortality reduction in all HF patients and that this benefit is observed in patients treated with concurrent cardio-protective medications, as seen in other clinical trials. However, this meta-analysis is the first to examine a mortality benefit of metformin therapy specifically in patients with a preserved EF.
Diabetic cardiomyopathy
T2DM is a complex metabolic disorder, with the initial hallmarks of insulin resistance and progressive impairment in insulin secretion from the pancreas [28]. Over time a pro-inflammatory state develops, potentiated by
Metformin treatment
Masoudi [20] ( alterations in gut microbiota and excess adiposity [29]. Ultimately, end-organ failure ensues. Diabetic cardiomyopathy is a major adverse outcome of the disease. Where the pathophysiology of DM and atherosclerotic CAD is well understood [30], there is an emergence of data on the cellular and metabolic mechanisms of non-ischaemic driven diabetic cardiomyopathy. Alterations in the AMPK pathway and mitochondrial dysfunction are major components in the development of myocardial impairment [10].
Table 3 Development of HF in patients with and without preserved EF
HF heart failure, ECG electrocardiogram, eGFR estimated glomerular filtration rate, ACR albumin-creatinine ration, Hb haemoglobin, AHF acute heart failure, PVD peripheral vascular disease, HTN hypertension, LVEF left ventricular ejection fraction, NT pro-BNP N-terminal pro b-type natriuretic peptide, BMI body mass index, SBP systolic blood pressure, DBP diastolic blood pressure, CVD cerebrovascular disease, AF atrial fibrillation, MI myocardial infarction, COPD chronic obstructive pulmonary disease, ACEi angiotensin converting enzyme, ARB angiotensin receptor blocker
Metformin systemic and myocardial mechanism of action
Metformin is the most commonly prescribed anti-diabetic drug [31]. It has negligible hypoglycaemic risk, has beneficial effects on HbA1c and weight reduction and is relatively inexpensive [31]. In recent years its position in guideline-directed management of new-onset T2DM has somewhat changed. The American Diabetes Association still recommends metformin as first-line therapy in all newly diagnosed T2DM patients [31]; however, the European Society of Cardiology now recommends metformin as first-line therapy only in patients who are deemed not at high-risk or do not have established cardiovascular disease, instead recommending a sodium glucose cotransporter-2 inhibitor (SGLT-2i) or glucagon-like protein-1 receptor agonist (GLP-1 RA) for these patients [32]. In recent years our understanding of metformin's mechanism of action has evolved. At the cellular level, metformin accumulates in the mitochondrial matrix ultimately causing a reduction in the synthesis of adenosine triphosphate (ATP) and an increase in the level of AMP resulting in the activation of the AMPK pathway [10]. In the liver, decreased ATP availability and inhibition of enzymes involved in lactate uptake results in inhibition of gluconeogenesis [9,10,33,34]. Additionally, by activating AMPK metformin modifies lipid production and breakdown [10,34]. In the intestinal tract, metformin inhibits glucose absorption and improves insulin production by the incretin affect [35]. In adipocyte tissue metformin reduces free-fatty acid release [36], further improving glucose uptake in other tissues such as skeletal muscle [37].
In the myocardium, metformin also activates AMPK resulting in increased glucose uptake, reduction in protein synthesis and improved mitochondrial function [10].
Furthermore, metformin decreases nitric oxide (NO) production via inhibition of inducible NO synthetase (iNOS) [12]. Finally, metformin has been shown to reduce collagen synthesis and fibrosis in myocardial tissue [14].
Metformin in T2DM and cardiovascular disease
In the UK Prospective Diabetes Study, obese DM-individuals treated with metformin had a 42% (p = 0.017) risk reduction in diabetes-related deaths and a 36% (p = 0.011) reduction in all-cause mortality [38]. Furthermore, there was a 30% (p = 0.020) reduction of all macrovascular complications (MI, sudden death, angina, stroke and PVD) with metformin treatment [38].
In heart transplant recipients, cardiac allograft vasculopathy (CAV) is a major cause of morbidity and mortality with limited treatment options [42]. However, in heart [42]. Furthermore, metformin was independently associated with a 91% reduction (p = 0.003) in the combined end-point of CAV and cardiovascular mortality in these patients [42].
Metformin in HF
Historically, the use of metformin has been restricted in HF owing to concerns regarding the development of lifethreatening lactic acidosis [43]. This adverse side-effect was largely extrapolated from data regarding phenformin, a biguanide that was ultimately withdrawn from the market [43]. However, the development of metformin-associated lactic acidosis in HF patients has since been refuted. In a large observational study spanning over 10 years, 27% of HF patients were on metformin therapy and there were no observed hospitalisations or deaths due to lactic acidosis [23]. Nonetheless, several observational studies have shown both morbidity and mortality benefits of HF patients treated with metformin. In a previous meta-analysis, metformin in HF patients was associated with a 7% reduction in HF hospitalisations (adjusted RR 0.93 [95% CI 0.86, 0.98], p = 0.01) and a 20% reduction in all-cause mortality (RR 0.80 [95% CI 0.74, 0.87], p < 0.001) [16]. Furthermore, this benefit extends to subgroups of HF such that in patients with CAD, all-cause mortality was reduced by 16% (HR 0.84 [0.81, 0.87], p = 0.03) [39]. In patients with hypertension, long-term metformin treatment was associated with a reduction in LV filling pressures and LV mass over-time [44]. In these patients the incidence of symptomatic HFpEF was also reduced with metformin therapy compared to non-metformin therapy (4.6% vs. 11.9% respectively, p = 0.020) [44]. The results of the current meta-analysis supporting these findings in the HF population with an 18% reduction in mortality associated with metformin therapy. Furthermore, this meta-analysis has shown a mortality benefit associated with metformin in patients with HFpEF. This is of clinical relevance as there are limited therapeutic options with mortality benefits in this population of HF patients.
Limitations
Despite a thorough literature search, there is a potential risk of not identifying all studies that have evaluated metformin use in HF patients, particularly those that were performed as a sub-group analysis. However, our use of reference mining and grey-literature searches are likely to have minimised this. Furthermore, due to the nature of observational studies, unaccounted confounding variables may have influenced individual study results. No randomised control trial of the use of metformin in HF patients exists and so this issue cannot be mitigated. The use of aggregate data rather than individual patient data may have limited the analysis.
Finally, our analysis was limited to observational studies of metformin in all HF patients, rather than HFpEF as a single entity. Due to the lack of reporting of outcomes in HFpEF patients our analysis was limited to a metaregression in this sub-group. Furthermore, due to the low number of studies included we encountered a significant amount of heterogeneity, ultimately reflecting the paucity of evidence in this area and the requirement for further research.
Conclusions
This meta-analysis is the first to highlight a mortality benefit for metformin therapy in HF patients with a preserved EF. | 2020-03-19T10:37:08.897Z | 2020-03-01T00:00:00.000 | {
"year": 2020,
"sha1": "29a44a9c3e627d775f8096321a87ccab158d8a71",
"oa_license": "CCBY",
"oa_url": "https://cardiab.biomedcentral.com/track/pdf/10.1186/s12933-020-01100-w",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "bcaddb89bb97d8f2a815c90f890f28128ddb4f5f",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
244931971 | pes2o/s2orc | v3-fos-license | Risk of COVID-19 Infection in Public Transportation: The Development of a Model
South Korea’s social distancing policies on public transportation only involve mandatory wearing of masks and prohibition of food intake, similar to policies on other indoor spaces. This is not because public transportation is safe from coronavirus disease 2019 (COVID-19), but because no suitable policies based on accurate data have been implemented. To relieve fears regarding contracting COVID-19 infection through public transportation, the government should provide accurate information and take appropriate measures to lower the risk of COVID-19. This study aimed to develop a model for determining the risk of COVID-19 infection on public transportation considering exposure time, mask efficiency, ventilation rate, and distance. The risk of COVID-19 infection on public transportation was estimated, and the effectiveness of measures to reduce the risk was assessed. The correlation between the risk of infection and various factors was identified through sensitivity analysis of major factors. The analysis shows that, in addition to the general indoor space social distancing policy, ventilation system installation, passenger number reduction in a vehicle, and seat distribution strategies were effective. Based on these results, the government should provide accurate guidelines and implement appropriate policies.
Introduction
A novel coronavirus, the causative agent for coronavirus disease 2019 (COVID- 19), was first identified in Wuhan City, Hubei Province, China, in December 2019, and it has since spread rapidly worldwide. The World Health Organization (WHO) declared it an international public health emergency in January 2020 and a global pandemic two months thereafter [1]. Since the first confirmed case at the beginning of 2020, the cumulative number of cases has exceeded 170,000 in South Korea as of July 2021 [2].
COVID-19 is a severe acute respiratory syndrome that is mainly transmitted through droplets from an infected person and can spread easily in daily life. The unprecedented rate of transmission and mortality of this infectious disease resulted in panic. Many countries and cities worldwide quarantined infected people to prevent the transmission of COVID-19 and implemented local lockdowns and facility closures, as well as movement restrictions.
Unlike other countries, South Korea did not impose such measures. Instead, the policies focused on preventing the spread of the disease through early detection by expanding preemptive testing, isolation, and treatment of patients with confirmed COVID-19, tracking and managing close contacts; and limiting the size of gatherings. Residents were able to move freely with no restrictions; however, movement was still greatly reduced due to fear of infection. Many individuals started avoiding the use of public transportation and used personal vehicles instead.
In the Seoul metropolitan area, more than 15 million people use public transportation daily. Although the Korea Disease Control and Prevention Agency insisted that there were no confirmed cases of COVID-19 infection due to the use of public transportation, this seemed insufficient to relieve the fear. Transportation is crucial for daily life, and those without access to private vehicles have no choice but to rely on public means, despite the anxiety of contracting COVID-19. Because of the prolonged spread of COVID-19, the anxiety and precaution of people gradually faded, which led to the third and fourth waves of infection.
Experts now believe that the COVID-19 pandemic will not be resolved in the near future and that people will have to live with it as with the common cold. This calls into question the future of the use of public transportation, and whether the government needs to continue with these public transportation services without any clear countermeasures. The introduction of any policy regarding public transportation services should enable the prevention of COVID-19. However, the government needs to provide accurate information to relieve fear, while making efforts to reduce the spread of COVID-19.
To date, many studies have been conducted on COVID-19 transmission to effectively contain its spread. Representative infection transmission models, such as the susceptibleinfected-removed model [3][4][5][6][7] and data-driven time series [8,9], are macroscopic models. These models have the advantage of including macro-parameters and data. However, they cannot reflect the detailed factors affecting transmission, such as mask wearing, ventilation, and the environment.
To address the limitations of these macroscopic models, several studies have been conducted to estimate the risk of infection [10,11]. The Rail Safety and Standards Board (RSSB) estimated the probability of COVID-19 transmission on rail transit and calculated the risk per journey by multiplying the 'infection risk per passenger contact' and 'contacts per journey' with the 'impact of mitigating factors', such as mask-wearing rate and ventilation rate. The probability of infection was estimated using three types of trains (44, 30, and 31 seaters), and the risk was 0.0045-0.0051%. They assumed the number of infected persons to be 0.0225, which is 0.05% of 45 passengers on a train. Lelieveld et al. estimated the risk of COVID-19 transmission in indoor environments using the infection risk of a single viral RNA copy and the density of the viral copies in the air considering exposure time, mask efficiency, and ventilation rate [11]. That study hypothesized that the virus spreads uniformly in one space, making it impossible to analyze the effect of the distance between people on the risk of COVID-19 transmission.
Therefore, our study aimed to develop a model that can estimate the risk of COVID-19 infection and the number of infected individuals on public transportation, and to identify the influencing factors, including distance measures. The WHO estimates that the transmission routes of COVID-19 infection include contact, droplets, airborne, fomite, fecal-oral, bloodborne, mother-to-child, and animalto-human transmission [12]. Among them, the transmission routes of infection on public transportation are possibly limited to droplets and airborne transmission.
Materials and Methods
Respiratory tract secretions discharged during conversations or coughing are generally classified into droplets and aerosols depending on the particle size. The WHO defines droplets as particles > 5-10 µm in diameter and aerosols as particles of <5 µm [13]. Meanwhile, Hinds defined aerosols as particles of 0.001-100 µm in size that are suspended in a gas phase [14].
There is still a lack of consensus among experts on the standard for distinguishing between droplets and aerosols. Droplets are relatively large and cannot spread far under the influence of gravity, whereas aerosols are small and can spread farther. When aerosols float in the air and water evaporates, they form droplet nuclei that can penetrate the alveoli [15].
Effectiveness of Masks
According to the SGS (Société Générale de Surveillance) [16], mask certification standards vary by country. The South Korean government is implementing a Korea Filter (KF) system for health mask certification, as shown in Table 1. A mask that has received KF80 certification means that it can block more than 80% of fine particles with an average particle size of 0.6 µm. Those certified as KF94 and KF99 can block 94% and >99% of particles, at an average particle size of 0.4 µm, respectively [17]. KF-certified masks have the advantage of lowering the risk of infection caused by external droplets due to their high dust-collection efficiency. However, this can result in breathing difficulty. Therefore, in June 2020, the Ministry of Food and Drug Safety added a Korea Filter Anti-Droplet (KF-AD) grade that prevents infection by blocking droplets and improves breathability. These masks are known to block 55-80% of fine particles with an average particle size of 0.4-0.6 µm.
Nevertheless, mask-wearing does not guarantee complete protection. Although the previously mentioned KF94 certified masks can block 94% of droplets and aerosols with a diameter > 0.4 µm, the remaining 6% can still penetrate. Moreover, since the inward leakage (IR) of KF94 is 11.0% or less, approximately 11% of air can leak through the gaps between the mask and the face. Thus, the possibility of infection caused by aerosol still exists. In addition, the risk of aerosol leakage can be higher with dental or cotton masks that are not KF certified.
Upon reviewing studies on the propagation of droplets and aerosols by mask type, Jayaweera et al. reported that dental masks allowed a significant leakage of 20-30% compared with the 5% allowed by elastomeric respirators and N95 masks certified in the US; they emphasized the importance of indoor social distancing, given that no mask can completely block the transmission of COVID-19 [18].
Although wearing a mask while using public transportation is the most basic means to reducing the spread of infection, there is still a risk of transmission, as mask-wearing does not completely block all droplets and aerosols. However, the risk of leakage may vary depending on the mask efficiency and the method of wearing it.
1-1 Infection Risk Model
Lelieveld et al. calculated the risk of COVID-19 transmission via indoor aerosols [11]. As shown in Equation (1), this risk was calculated using the infection probability of a single virus and the number of transferred viruses. In our study, the same model was used to calculate the risk of subject B being infected by the virus discharged by infected subject A.
In this equation, p represents the risk of infection by a single virus, and n(B←A) represents the number of viruses discharged from infected person A to infect person B.
As shown in Figure 1, the risk of subject B being infected by infected person A increases with an increasing risk of infection by a single virus and an increasing number of viral populations that can cause infection.
In this equation, p represents the risk of infection by a single virus, and n(B←A) represents the number of viruses discharged from infected person A to infect person B. As shown in Figure 1, the risk of subject B being infected by infected person A increases with an increasing risk of infection by a single virus and an increasing number of viral populations that can cause infection.
Virus Transmission Model
The number of viral populations transferred from person A to person B is unknown. In general, owing to the large particle size, the number of transmitted droplets is affected by gravity and decreases with an increasing distance. Nevertheless, aerosols, which are smaller and less affected by gravity, remain suspended in the air and travel farther. According to Wang et al., unlike other existing viruses, coronavirus is reported to spread through both droplets and aerosols [19].
Given that studies on the transmission pattern of individual coronaviruses are difficult to find in the literature, our study intends to present a distribution model based on two assumptions. The first assumption is that the risk of infection is the same for all people in the same space, as the virus floats in the air in the form of droplet nuclei and spreads
Virus Transmission Model
The number of viral populations transferred from person A to person B is unknown. In general, owing to the large particle size, the number of transmitted droplets is affected by gravity and decreases with an increasing distance. Nevertheless, aerosols, which are smaller and less affected by gravity, remain suspended in the air and travel farther. According to Wang et al., unlike other existing viruses, coronavirus is reported to spread through both droplets and aerosols [19].
Given that studies on the transmission pattern of individual coronaviruses are difficult to find in the literature, our study intends to present a distribution model based on two assumptions. The first assumption is that the risk of infection is the same for all people in the same space, as the virus floats in the air in the form of droplet nuclei and spreads uniformly regardless of the distance between infected and non-infected persons. The second assumption is that the risk of infection decreases with a decreasing number of viruses in the air and an increasing distance from the infected person, given that the density of the virus discharged from the infected person is lowered, while being spread by mixing with the surrounding air.
Uniform Distribution Model
In the model showing the risk of infection for subject B infected by infected person A, n(B←A) represents the number of viruses discharged from A to B. The number of transferred viruses is proportional to the number of viruses discharged by infected person A per hour and exposure time. The uniform distribution model assumes that the virus floats in the air in the form of a droplet nucleus and spreads uniformly in the same space. As the interior volume of the public transportation vehicle increases, the density of viral particles floating in the air decreases, exhibiting an inverse relationship, as shown in Equation (2): where N denotes the number of aerosol particles discharged per unit time, h denotes the exposure time, b denotes the respiratory rate per hour ( l h ), w denotes the interior volume of the vehicle ( l = 1 1000 × m 3 ), and R denotes the number of viruses per aerosol particle. Since viruses are much smaller than aerosols in size, there are several individual viruses per aerosol particle. The number of viruses R per aerosol particle can be calculated as the product of the volume of aerosol particles v and the number of viruses per unit volume c.
The viruses discharged from infected person A are filtered to a certain extent by the mask of subject B, depending on the mask effectiveness, which had a value of 0-1. Higher effectiveness is demonstrated by a value closer to 1, and lower effectiveness by a value closer to 0.
There are differences in the method and degree of ventilation depending on the operating characteristics of the public transportation mode. City buses and subways have a certain degree of ventilation achieved by the opening and closing of the doors at each station. At such times, a part of the aerosol discharged from infected person A moves outside the vehicle. Thus, the number of aerosol particles discharged from infected person A is reduced to a certain extent, depending on the level of ventilation.
Moreover, since the virus discharged from infected person A must reach the respiratory tract of subject B to cause infection, the probability of virus deposition must be considered. Thus, when infected persons A and B stay together for h duration, the number of viruses that can be released by infected person A to cause infection in subject B can be summarized as shown in Equation (3): where f denotes the ventilation rate ( h −1 ), m denotes the mask effectiveness, and t denotes the probability that the virus is deposited in the respiratory tract. The number of aerosol particles discharged by infected person A per unit time can be calculated from their respiratory rate. Thus, it can be calculated by differentiating between aerosol discharges when speaking and during respiration without conversation and then subtracting the amount of discharge filtered by the mask of infected person A.
The number of aerosol particles N discharged by infected person A per unit time was calculated using Equation (4): where c b denotes the number of aerosol particles per unit volume discharged during respiration ( Number of aerosol particles/cm 3 ), c t denotes the number of aerosol particles per unit volume discharged during the conversation ( Number of aerosol particle/cm 3 ), s denotes the ratio of speaking (0 ≤ s < 1), and b denotes the respiratory rate per unit time
. Distance Reduction Model
Sun and Zhai conducted a study on the effectiveness of social distancing and ventilation in preventing the spread of COVID-19 and presented a model for estimating the risk of exposure to droplets, based on distance from an infected person, as shown in Equation 5 [20]. Based on this model, a subject 1 m away from an infected person would be exposed to about 43% of the total discharged droplet mass, and a subject 2 m away would be exposed to about 30%.
, with a value of 0-100%, denotes the risk of exposure to a droplet at a distance d.
This study aimed to establish a viral transmission model using the risk of exposure to droplets established by Sun and Zhai [20].
In the model showing the risk that subject B is infected by person A, n(B←A) represents the number of viruses discharged from infected person A to infect subject B. The number of viruses transferred is proportional to the number discharged and exposure time, as in the uniform distribution model. The number of viruses is assumed to decrease as the distance from the infected person increases. Therefore, n2(B←A) decreases with an increase in the distance between infected persons A and B.
When infected person A and subject B are separated by a distance of d and stay together for h hours, subject B is exposed to the aerosol present in the space within the radius of the infected person. Because the number of viruses decreases with increasing distance, only E(d)% of the initial discharge of infected person A is delivered to the subject, and the number of discharged viruses can be expressed as in Equation (6): where N denotes the number of aerosol particles discharged by infected person A per unit time (Equation 4) used in the uniform distribution model, E(d) denotes the risk of exposure according to the distance, and W(d) denotes the volume of the space between the infected person and the subject ( l = 1 1000 × m 3 ). As the distance d between infected persons A and B increases, the volume W(d) of the space between the two people increases, and the number of transmitted viruses decreases by W(d)%.
If the aerosol discharged from the infected person spreads uniformly and threedimensionally, it will spread in a spherical shape centered on the infected person. To simplify, a one-dimensional spread model is applied in this study, as shown in Figure 2. In other words, the volume of the space is proportional onlyto the straight-line distance between the two individuals by fixing the two axes of height and width. Because the aerosol discharged from the infected person spreads uniformly back and forth, the volume of space affecting the infected person can be calculated by simplifying it as in Equation (7).
Model for Calculating Infection Risk and the Number of Infected Individuals on Public Transportation
If there are M people in a public transportation vehicle, of which m f are infected, the risk of passenger i getting infected is calculated using Equation (8).
Model for Calculating Infection Risk and the Number of Infected Individuals on Public Transportation
If there are M people in a public transportation vehicle, of which are infected, the risk of passenger i getting infected is calculated using Equation (8).
The number of new infections caused by infected individuals in a public transportation vehicle with M passengers is calculated using Equation (9).
The risk of one or more new infections caused by infected people in a public transportation vehicle is calculated using Equation (10).
Probability of Infection by a Single Virus p
Lelieveld et al. calculated the probability of infection by a single virus using the infective dose (ID50) concept used in the medical community; it refers to the amount of virus required to infect 50% of the subjects [11].
If the probability of infection by a single virus is p, the probability of not being infected is (1 − p). Therefore, the probability of n individuals not being infected is (1 − ) . If the ID50 concept is applied, Equation 8 indicates that the probability of not being infected by ID50 viruses is 0.5.
Equation (9) can be obtained by rearranging for p. The number of new infections caused by m f infected individuals in a public transportation vehicle with M passengers is calculated using Equation (9).
The risk of one or more new infections caused by m f infected people in a public transportation vehicle is calculated using Equation (10). [11].
If the probability of infection by a single virus is p, the probability of not being infected is (1 − p). Therefore, the probability of n individuals not being infected is (1 − p) n . If the ID50 concept is applied, Equation (8) indicates that the probability of not being infected by ID50 viruses is 0.5.
Although there is no clear data about the ID50 value of COVID-19, Lelieveld et al. defined it as 316 based on information on similar viruses [11]. Substituting this into Equation 9, the probability of infection p by a single coronavirus becomes approximately 0.0022 (0.22%), which was the value used in this study.
Probability of Deposition t
The probability of deposition, t, is the likelihood that the coronavirus reaches and deposits in the respiratory tract of the exposed person. As described above, not all viruses that a subject is exposed to are deposited, and only a portion reaches the respiratory tract. Lelieveld et al. stated that the probability of coronavirus deposition would be approximately 0.5 (50%) [11]; thus, it was set accordingly in our study.
Number of Viruses per Unit Volume c
The specific amount and concentration of the virus in a person infected with COVID- 19 have not yet been determined. Lelieveld et al. investigated these in the throat of an infected person under the assumption that respiratory droplets and aerosols discharged from an infected person contain the same number of viruses as the fluid in the respiratory tract that produces droplets and aerosols [11]. The number of viruses detected in the throat of an infected person was approximately 10 4 ∼ 10 11 per 1 mL; 5 × 10 8 viruses per 1 mL were detected on average in those with symptoms of infection, accounting for one-fifth of those testing positive. Accordingly, the number of coronaviruses per unit volume was set to 5 × 10 8 per 1 mL (cm 3 ).
Volume of Aerosol Particle v
To set the volume of the aerosol particles used in the model, the diameter must be known. Although the size of aerosol particles discharged through breathing and speaking varies, as mentioned above, most of the particles have a diameter of 1 µm.
Meanwhile, small particles such as aerosols evaporate as they are discharged from the mouth. Lelieveld et al. assumed that the diameter was 5 µm but reduced to 1 µm due to evaporation while falling [11]. Therefore, in this study, the average volume of aerosol particles was set to 6.54 × 10 −11 cm 3 under the assumption that the average volume of the aerosol particles was 5 µm.
Number of Aerosol Particles per Unit Volume Discharged during Respiration and Conversation
Morawska et al. stated that the concentration of aerosol particles 0.3~20 µm generated during respiration and conversation is at the level of 0.1~1.1 cm −3 [21]. Lelieveld et al. assumed that 0.06 aerosol particles per 1 cm 3 were discharged during respiration and 0.6 per 1 cm 3 during conversation, considering that the effective average diameter of fine particles was 5 µm [11]. The same values were used in this study.
Respiratory Rate per Unit Time b
While the respiratory rate varies depending on individual characteristics, in adults it ranges from 6-8 L per minute. Accordingly, in this study, the respiratory rate per hour was set to 0.12 L, assuming that the rate per minute was 7 L.
Results of Analyzing Changes in the Risk of Infection on Public Transportation
The exposure time, ventilation rate, mask effectiveness, and conversation rate may vary depending on the operation and usage characteristics of the public transportation that is analyzed. These constitute the parameters in the model, and the risk of infection varies accordingly. To properly analyze the results of the scenario, it is necessary to first examine the changes in the risk of infection according to each parameter. Therefore, in the most basic scenario of a bus, the default parameters were set to 45 passengers, a conversation rate of 0.1, a ventilation rate of 3.3, mask effectiveness of 0.7, and an exposure time of 1 h. Changes in the number and risk of new infections were examined according to the changes in each parameter.
Exposure Time
The risk of infection is proportional to the number of viruses discharged by an infected person, which is also proportional to the respiratory rate. Since the respiratory rate is proportional to the unit time, the longer the contact time with an infected person, the higher the risk of infection. The uniform distribution model was named Model 1, and the model of reduction with distance was named Model 2. With an exposure time of 1 h, the number of new infections was calculated as 0.0157 for Model 1 and 0.0156 for Model 2. This is interpreted as the probability of infecting 1.57 and 1.56 people, respectively, when an infected person traveling for 1 h uses a bus 100 times. In addition, the probability of one or more new infections in a vehicle is 1.56% for Model 1 and 1.55% for Model 2, which are interpreted as the probability of one or more new infections when an infected person uses the vehicle approximately 64 times.
As shown in Figure 3, when the exposure time was extended from 1 h to 2 h, the probability doubled to 0.0315 and 0.0311, respectively, and when it was extended to 3 h, the risk was three times greater at 0.0472 and 0.467, respectively. The difference between Model 1 and Model 2 was insignificant, at approximately 1%, with the probability based on Model 1 being higher than that based on Model 2.
most basic scenario of a bus, the default parameters were set to 45 passengers, a conversation rate of 0.1, a ventilation rate of 3.3, mask effectiveness of 0.7, and an exposure time of 1 h. Changes in the number and risk of new infections were examined according to the changes in each parameter.
Exposure Time
The risk of infection is proportional to the number of viruses discharged by an infected person, which is also proportional to the respiratory rate. Since the respiratory rate is proportional to the unit time, the longer the contact time with an infected person, the higher the risk of infection. The uniform distribution model was named Model 1, and the model of reduction with distance was named Model 2. With an exposure time of 1 h, the number of new infections was calculated as 0.0157 for Model 1 and 0.0156 for Model 2. This is interpreted as the probability of infecting 1.57 and 1.56 people, respectively, when an infected person traveling for 1 h uses a bus 100 times. In addition, the probability of one or more new infections in a vehicle is 1.56% for Model 1 and 1.55% for Model 2, which are interpreted as the probability of one or more new infections when an infected person uses the vehicle approximately 64 times.
As shown in Figure 3, when the exposure time was extended from 1 h to 2 h, the probability doubled to 0.0315 and 0.0311, respectively, and when it was extended to 3 h, the risk was three times greater at 0.0472 and 0.467, respectively. The difference between Model 1 and Model 2 was insignificant, at approximately 1%, with the probability based on Model 1 being higher than that based on Model 2.
Ventilation Rate
As mentioned above, ventilating the air inside a public transportation vehicle decreases the number of viruses discharged by an infected person as well as the risk of infection. In terms of changes in the risk of infection, when ventilating once per hour, the number of new infections was calculated as 0.0519 for Model 1 and 0.0514 for Model 2, as shown in Figure 4. When ventilation was increased to twice per hour, the number of new infections was reduced to 0.026 and 0.0257, respectively, and to 0.0173 and 0.0171, respectively, when ventilating three times per hour.
Ventilation Rate
As mentioned above, ventilating the air inside a public transportation vehicle decreases the number of viruses discharged by an infected person as well as the risk of infection. In terms of changes in the risk of infection, when ventilating once per hour, the number of new infections was calculated as 0.0519 for Model 1 and 0.0514 for Model 2, as shown in Figure 4. When ventilation was increased to twice per hour, the number of new infections was reduced to 0.026 and 0.0257, respectively, and to 0.0173 and 0.0171, respectively, when ventilating three times per hour.
Mask Effectiveness
The higher the mask effectiveness, the lower the number of viruses discharged by an infected person. The number of viruses transferred to the respiratory tracts of the subjects would also decrease, thereby lowering the probability of infection. In terms of changes in the risk of infection, with mask effectiveness of 0 (not wearing a mask), the number of new infections was calculated as 0.1746 for Model 1 and 0.1723 for Model 2. KF-AD is known to block 55-80% of fine particles with an average size of 0.4-0.6 µm, as shown in Figure 5.
Mask Effectiveness
The higher the mask effectiveness, the lower the number of viruses discharged by an infected person. The number of viruses transferred to the respiratory tracts of the subjects would also decrease, thereby lowering the probability of infection. In terms of changes in the risk of infection, with mask effectiveness of 0 (not wearing a mask), the number of new infections was calculated as 0.1746 for Model 1 and 0.1723 for Model 2. KF-AD is known to block 55-80% of fine particles with an average size of 0.4-0.6 μm, as shown in Figure 5. If all passengers in a transportation vehicle are wearing a mask with KF-AD or higher grade, the average effectiveness is approximately 0.7. The number of new infections was calculated as 0.0157 for Model 1 and 0.0156 for Model 2, with the number of new infections reduced to 1/10 or less of the number when no masks were worn.
Frequency of Conversation
Since a higher number of viruses are released when conversing versus not, the more an infected person talks, the higher the risk of infection. When the infected person does not speak at all for 1 h, the aerosol is discharged only through respiration. The number of new infections, in that case, is calculated as 0.0083 for Model 1 and 0.0082 for Model 2, as shown in Figure 6. The number of new infections increased by approximately 0.9 times for each 10% increase in the conversation rate.
Frequency of Conversation
Since a higher number of viruses are released when conversing versus not, the more an infected person talks, the higher the risk of infection. When the infected person does not speak at all for 1 h, the aerosol is discharged only through respiration. The number of new infections, in that case, is calculated as 0.0083 for Model 1 and 0.0082 for Model 2, as shown in Figure 6. The number of new infections increased by approximately 0.9 times for each 10% increase in the conversation rate.
Frequency of Conversation
Since a higher number of viruses are released when conversing versus not, the more an infected person talks, the higher the risk of infection. When the infected person does not speak at all for 1 h, the aerosol is discharged only through respiration. The number of new infections, in that case, is calculated as 0.0083 for Model 1 and 0.0082 for Model 2, as shown in Figure 6. The number of new infections increased by approximately 0.9 times for each 10% increase in the conversation rate.
Analysis of Changes in the Risk of Infection on Public Transportation Due to Passenger Number Reduction
The greater the number of passengers in a vehicle, the higher the probability of the presence of an infected person and the risk of transmission. A policy to reduce the number of passengers in a vehicle may be considered to decrease the spread of COVID-19.
Compared to short-distance travel of about 1 h, the probability of having a conversation with the person in the next seat or on the phone, as well as the probability of a mask falling off while talking or removing it because of stuffiness, increases during longdistance travel of about 3 h. In other words, it can be assumed that the conversation rate increases and the mask efficiency decreases during long-distance travel. In this study, the short-distance travel conditions were assumed to be a travel time of 1 h, conversation rate of 0.1, mask effectiveness of 0.7, and ventilation rate of 3.3, whereas long-distance travel conditions were 3 h, 0.2, 0.5, and 3.3, respectively. Changes in the risk of infection due to the reduction in the number of passengers were analyzed under short-distance and long-distance travel conditions.
Analysis of Changes in the Risk of Infection on Public Transportation Due to Passenger Number Reduction
The greater the number of passengers in a vehicle, the higher the probability of the presence of an infected person and the risk of transmission. A policy to reduce the number of passengers in a vehicle may be considered to decrease the spread of COVID-19.
Compared to short-distance travel of about 1 h, the probability of having a conversation with the person in the next seat or on the phone, as well as the probability of a mask falling off while talking or removing it because of stuffiness, increases during long-distance travel of about 3 h. In other words, it can be assumed that the conversation rate increases and the mask efficiency decreases during long-distance travel. In this study, the short-distance travel conditions were assumed to be a travel time of 1 h, conversation rate of 0.1, mask effectiveness of 0.7, and ventilation rate of 3.3, whereas long-distance travel conditions were 3 h, 0.2, 0.5, and 3.3, respectively. Changes in the risk of infection due to the reduction in the number of passengers were analyzed under short-distance and longdistance travel conditions. As shown in Table 2 and Figure 7, under short-distance travel conditions with 49 passengers in the vehicle, the number of new infections was 0.0172. When the number of passengers was reduced to 22, the number decreased from 0.00966 to 0.0075. In comparison, under long-distance travel conditions with 49 passengers, the number of new infections was 0.2105. When the number of passengers was reduced to 22, the number decreased from 0.11841 to 0.0921. Under both conditions, the reduction in the number of passengers and the risk of infection were analyzed. The number of new infections decreased more than 12 times under long-compared to short-distance travel. In other words, the reduction in the number of passengers was 12 times more effective in decreasing the risk of infection during long-distance travel. Table 2. Analysis of the risk of infection with passenger number reduction under short-and long-distance travel conditions. As shown in Table 2 and Figure 7, under short-distance travel conditions with 49 passengers in the vehicle, the number of new infections was 0.0172. When the number of passengers was reduced to 22, the number decreased from 0.00966 to 0.0075. In comparison, under long-distance travel conditions with 49 passengers, the number of new infections was 0.2105. When the number of passengers was reduced to 22, the number decreased from 0.11841 to 0.0921. Under both conditions, the reduction in the number of passengers and the risk of infection were analyzed. The number of new infections decreased more than 12 times under long-compared to short-distance travel. In other words, the reduction in the number of passengers was 12 times more effective in decreasing the risk of infection during long-distance travel. Table 2. Analysis of the risk of infection with passenger number reduction under short-and long-distance travel conditions.
Analysis of Changes in the Risk of Infection on Public Transportation Based on Distancing Strategy
Since the risk of COVID-19 infection depends on distance, it should also change based on the seating arrangement. We measured the difference in the risk of infection according to the seating arrangement by applying Model 2, assuming that virus transmission decreased with increasing distance.
There must be empty seats in the arrangement. In this study, strategies were established for 22 seats on a bus that could accommodate a maximum of 45 passengers under short-distance travel conditions with a travel time of 1 h, conversation rate of 0.1, mask effectiveness of 0.7, and ventilation rate of 3.3.
Analysis of Changes in the Risk of Infection on Public Transportation Based on Distancing Strategy
Since the risk of COVID-19 infection depends on distance, it should also change based on the seating arrangement. We measured the difference in the risk of infection according to the seating arrangement by applying Model 2, assuming that virus transmission decreased with increasing distance.
There must be empty seats in the arrangement. In this study, strategies were established for 22 seats on a bus that could accommodate a maximum of 45 passengers under short-distance travel conditions with a travel time of 1 h, conversation rate of 0.1, mask effectiveness of 0.7, and ventilation rate of 3.3.
As shown in Figure 8, five arrangement strategies were set as comparative alternatives. The first was to arrange the seats one space apart in a zigzag, and the second was similar to the first but with the seats on both sides of the same row in a symmetrical fashion. The third strategy was to skip a row between the seats, and the fourth was to only use window seats. The last strategy was to maximize the total relative distance between 22 people by arranging the seats in the front and rear of the vehicle. As shown in Figure 8, five arrangement strategies were set as comparative alternatives. The first was to arrange the seats one space apart in a zigzag, and the second was similar to the first but with the seats on both sides of the same row in a symmetrical fashion. The third strategy was to skip a row between the seats, and the fourth was to only use window seats. The last strategy was to maximize the total relative distance between 22 people by arranging the seats in the front and rear of the vehicle. Figure 9 show the number and risk of new infections based on each of the five seating strategies. The risk was lowest for the window-seat-only strategy and highest for the maximum space apart. Separating all passengers as far apart as possible from each other seems to be the best to minimize the risk of infection. Figure 9 show the number and risk of new infections based on each of the five seating strategies. The risk was lowest for the window-seat-only strategy and highest for the maximum space apart. Separating all passengers as far apart as possible from each other seems to be the best to minimize the risk of infection. Table 3 and Figure 9 show the number and risk of new infections based on each of the five seating strategies. The risk was lowest for the window-seat-only strategy and highest for the maximum space apart. Separating all passengers as far apart as possible from each other seems to be the best to minimize the risk of infection. Table 3. Number and risk of new infections by seat arrangement strategy.
Discussion
COVID-19 is spread by droplets and aerosols discharged through the human respiratory tract; the probability of contracting it is proportional to the number of viruses a subject is exposed to. The higher the respiratory rate of an infected person, the higher the risk. Therefore, the risk of infection in a public transportation vehicle increases in proportion to the exposure time with an infected person, implying higher risk for long-distance travel. The risk is also higher for intercity compared to city buses, inter-regional compared to intercity buses, and the Korea Train eXpress(KTX) compared to the metro.
More droplets and aerosols are discharged during conversation compared to respiration, which leads to a higher number of discharged viruses and an increased risk of infection on public transportation. The risk is doubled at a conversation rate of 10% and tripled at 20%. Therefore, when using public transportation, it is necessary to minimize phone calls and verbal communication with other passengers.
Most of the droplets are blocked by masks due to their large particle size, but since aerosols are smaller, some leaks can occur. The higher the mask effectiveness, the more viruses are filtered, and the more the transmission to the respiratory tract of the subject is reduced, resulting in an overall infection risk reduction on public transportation. As previous studies indicated [22,23], masks are undoubtedly crucial, and wearing any type is better than not wearing one at all. A KF-AD mask that blocks droplets reduces the risk of infection by 10 times when compared to not wearing a mask, and a KF94 or a higher-grade mask reduces the risk by 100 times.
Aerosols are small particles that float in the air instead of droplets that fall immediately. Therefore, the number of viruses in the air can increase over time in unventilated indoor spaces. When there is good ventilation, indoor air is replaced by outdoor air, decreasing the number of viruses in the room and decreasing the risk of infection. When ventilation is performed twice per hour, the risk of infection is reduced by half, and when it is performed three times per hour, it is reduced to a third. Unlike city buses and metros, where ventilation occurs naturally while stopping every 2-3 min to open and close the vehicle doors, natural ventilation is likely to occur less frequently on intercity buses that make fewer stops. Inter-regional buses traveling between regions are even less likely to be ventilated. An air conditioner may be considered as an in-vehicle ventilation system, but circulating the air inside without a virus filter will only spread it more, which does not help to lower the risk of infection. Therefore, a ventilation system equipped with a virus filter is needed to reduce the risk of infection on public transportation.
Long-distance travel is associated with longer in-vehicle travel time, which is likely to increase the probability of talking and taking off one's mask. These factors significantly increase the risk of infection. To counteract this, the number of passengers on public transportation must be limited; this can reduce the risk of infection by 12 times for longdistance travel.
Since aerosols float in the air, the risk of infection may be perceived as equal for everyone within that space. However, those close to an infected person are more likely to be affected than those far away. If the risk of infection varies according to distance from the infected person, it should vary according to the seat arrangement in the vehicle as well. We found that the window-seat-only strategy lowered the risk of infection the most, while the maximum-space-apart strategy exhibited the highest risk of infection. Compared to the maximum-space-apart strategy, the window-seat-only strategy showed a 0.0033 lower number of new infections and a 0.328% lower risk. This is similar to the effect of reducing the number of passengers by 20% of the maximum capacity or by 9-10 people for short-distance travel. When the number of passengers per vehicle approaches the maximum capacity due to high traffic demand, it is necessary to consider a plan to reduce the number of passengers in order to improve the service level by reducing congestion in the vehicle while lowering the probability of infection. If not, applying the strategy of spacing passengers as far apart as possible is expected to have a similar effect.
In this study, the spread of COVID-19 was modeled considering exposure time, mask efficiency, the ratio of speaking, and ventilation rate. Using the model, the probability of and the number of new infections were estimated according to the number of viruses spreading in public transportation vehicles. Since there are no significant statistical data on the probability of COVID-19 infection from public transportation use, and such a study cannot be conducted in real life, we could not verify the accuracy of the results in this study.
It is difficult to directly compare the results of previous studies because of the different assumptions and conditions. Under the conditions of 1 h travel time, conversation rate of 0.1, mask effectiveness of 0.7, and ventilation rate of 3.3, the infection risk in this study was 1.56% in the uniform distribution and 1.55% in the distance reduction model. The RSSB results were 0.2% [10], which is much lower than our study because they assumed a very low probability of contact with an infected person. Lelieveld et al. had a result of 1.60% [11], which is almost the same result as ours, and their model had the same structure as the uniform distribution model in our study.
As noted to date, the results may vary depending on the assumptions of the analysis. Therefore, rather than the absolute values of the probability of infection and the number of infected individuals, it is more important to decide on policies by analyzing variations according to each variable and scenario.
Conclusions
Public transportation is essential to guarantee mobility in daily life. Nevertheless, people worldwide are reducing travel and avoiding public transportation due to COVID-19 [23][24][25][26]. Since public transportation appears to play an important role in the spread of COVID-19, it faces a major crisis. Therefore, more measures are needed to prevent the spread of COVID-19 on public transportation [27][28][29][30]. As Przybylowske et al. suggested, to prevent people from avoiding public transportation due to the risk of COVID-19 infection, the government should provide accurate guidelines and implement appropriate policies to help people perceive public transportation as safe [25].
Wang et al. emphasized that precautionary measures, such as assessing ventilation, airflows, air filtration, UV disinfection, and mask fit, must be implemented to mitigate aerosol transmission at both short and long ranges [19]. In this study, two infection risk estimation models were proposed to examine the correlation between these various factors and the infection risk in public transportation through sensitivity analysis of major factors. The analysis showed that the risk of infection in public transportation vehicles can be minimized through mask-wearing, prohibiting food intake, limiting conversations, and ventilation, similar to the findings noted in previous studies. Among these factors, maskwearing and ventilation were the most important factors. In addition, reducing the number of passengers and distributing the seating arrangement were found to significantly reduce the risk of infection in public transportation.
However, the proposed models have a limitation, in that they do not reflect all external environmental factors, such as temperature. Therefore, future studies should focus on developing new models (considering, for example, the effects of temperature) and modifying existing ones to include the factors addressed in this study. In addition, the models developed in this study can be used for evaluating the effectiveness of measures to reduce the number of passengers and distribute the seating arrangement; it may also be used for infectious diseases other than COVID-19.
Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article. | 2021-12-08T16:05:30.084Z | 2021-12-01T00:00:00.000 | {
"year": 2021,
"sha1": "4016c433a593dee600fb227188b1098f12ee3272",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1660-4601/18/23/12790/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "1104239ff92759f2e9c7cb778a226837b195411a",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
18851201 | pes2o/s2orc | v3-fos-license | Protective Effect of Curcumin against the Liver Toxicity Caused by Propanil in Rats
We investigated the protective effects of curcumin on propanil-induced alterations in biochemical indices in blood and liver of male Wistar rats. The study consisted of four treatment groups, with six animals each, designated as control, propanil (20mg/kg), curcumin(50 mg/kg), and curcumin (50 mg/kg) + propanil (20 mg/kg). Rats were administered their respective doses orally, every other day, for 28 days. Propanil administration elicited significant (P < 0.001) increases in plasma aspartate aminotransferase and alkaline phosphatase activities, by 24% and 56%, respectively, compared to the control. Treatment with propanil elevated bilirubin, creatinine, and total cholesterol levels in rats, but these were not significant relative to controls. Administration of propanil to rats significantly (P < 0.001) increased lipid peroxidation levels. However, catalase activity, vitamin C, and reduced glutathione levels were significantly reduced. Exposure to propanil did not produce any significant changes in packed cell volume, neutrophils, and leukocyte counts. The supplementation of curcumin attenuated the adverse effects of propanil intoxication by reducing lipid peroxidation levels and restored the levels of serum enzymes and reduced glutathione. The present study showed that propanil increased oxidative stress and altered some biochemical parameters in the rats but curcumin could afford some protection to attenuate propanil-induced toxicity in the liver.
Introduction
Curcumin (diferuloylmethane) is a yellow colouring ingredient of the spice turmeric obtained from the rhizome of Curcuma longa Linn (Zingiberaceae). It is a perennial herb distributed mainly throughout tropical and subtropical regions of the world [1]. Curcumin possesses anti-inflammatory, immunomodulatory, and antiatherogenic activities and is a potent inhibitor of various reactive oxygen-generating enzymes [2,3]. It has been used in indigenous herbal medicine for the treatment of inflammatory and liver disorders. Antioxidative properties of curcumin are well documented. Curcumin is a potent scavenger of reactive oxygen species including superoxide anion radicals and hydroxyl radicals. It has also been reported to inhibit erythrocyte lipid peroxidation [4]. Curcumin administration attenuated the arsenic, gentamicin, and acetaminophen-induced oxidative stress in rats [5][6][7]. Curcumin also prevented free radical formation-induced myocardial ischemia and paraquatinduced lung injury in rats [8,9]. Additionally, curcumin protected against diazinon-induced toxicity in blood, liver, and erythrocyte of male Wistar rats [10]. Furthermore, Canales-Aguirre and coworkers [11] had also reported the protective effects of curcumin against the oxidative damage in the hippocampus of rats after exposure to parathion.
Propanil (3, 4-dichloropropioanilide) is an acylanilide herbicide used to control barnyard grass, broadleaf weeds, and for the postemergent treatment of rice [12]. The widespread use of the herbicide on rice and wheat crops means that humans could be at risk of high level exposure. This necessitated the World Health Organization to recognize propanil as slightly hazardous in terms of human risk [13]. 2 International Scholarly Research Notices However, in a retrospective study assessing the clinical features of sixteen patients admitted for acute propanil poisoning, the authors advocated that the mild-to-moderate toxicity label of propanil should be revisited by the WHO [14]. Earlier reports by McMillan and coworkers [15] indicated that exposure to propanil had been associated with toxicity in humans and other mammals. Furthermore, it had been demonstrated that propanil could potentially induce cytotoxicity and nephrotoxicity in vitro [16]. A recent study reported that propanil induced dose-dependent histopathological changes in the liver and kidney tissues of exposed mice [17].
Antioxidants had been proven to play an important role in the regulation of a vast array of physiological and pathological processes. They principally contribute to the protection of cells and tissues against deleterious effects of reactive oxygen species and other free radicals.
The aim of this study was to determine the effect of subacute propanil exposure in the plasma of male rats and to assess whether these effects could be ameliorated by cotreatment with curcumin. To achieve this aim, rats were given propanil and curcumin by oral gavage for 28 days, after which malondialdehyde (MDA) and GSH levels and GST and CAT activities as well as other biochemical endpoints were evaluated.
Animals and Treatments.
Albino rats, weighing between 150 g and 200 g, purchased from covenant farm animal house located at Ibadan, Oyo State, Nigeria, were used in this study. The animals were kept in well-ventilated cages at room temperature and under controlled conditions of ambient temperature (25 ∘ C) at the Redeemer's University Animal House Facility, Mowe, Ogun State, Nigeria. They were maintained on normal laboratory chow and water ad libitum. The experiment was approved by the Animal Ethics Committee of the Redeemer's University.
Animal Treatment and Sample
Collection. Animals were divided into 4 experimental groups ( = 6). The first group received olive oil at a dose of 2 mL/kg body weight. The second group received an oral administration of propanil at a dose of 20 mg/kg/bodyweight equivalent to one-fifth of the oral dose used previously in our laboratories [18]. The third group received an oral administration of curcumin at a dose of 50 mg/kg/bodyweight. This dose was selected on the basis of a previous study [19]. The fourth group received both curcumin (50 mg/kg/bodyweight) + propanil (20 mg/kg/body weight). Curcumin and propanil were dissolved in olive oil at a dose of 2 mg/kg bw. The treatments (using intragastric feeding needle) were three times a week for a total of 28 days. Rats were sacrificed, after the last dose of administration and an overnight fast, by cervical dislocation and blood was obtained by cardiac puncture. Ethylenediaminetetraacetic acid was used as an anticoagulant. Plasma was obtained by centrifugation of samples at 4000 rpm for 5 minutes and stored at −20 ∘ C till analysis. Liver of rats was excised, homogenized in phosphate buffer (0.1 M, pH 7.4), and centrifuged at 4,500 rpm and the supernatants were used for the various biochemical measurements.
Biochemical Assays.
Alkaline phosphatase (ALP), aspartate aminotransferase (AST), alanine aminotransferase (ALT), total cholesterol, creatinine, and bilirubin were estimated in plasma using Randox commercial kits. Lipid peroxidation (LPO) was determined by measuring the formation of thiobarbituric acid reactive substances (TBARS) according to the method of Varshney and Kale [20] while glutathione S-transferase (GST) activity was determined according to Habig et al. [21]. Catalase activity was determined according to the method of Sinha [22]. Vitamin E was estimated by the method of Desai [23]. Briefly, 0.3 mL of ethanol and 0.4 mL diethyl ether were added to 1 mL plasma and centrifuged for 5 minutes at 4000 rpm. The ascorbic acid (Vitamin C) present in the liver tissue was determined by the spectrophotometric method of Omaye et al. [24]. The method of Beutler et al. [25] was followed in estimating the level of glutathione (GSH). This method is based upon the development of a relatively stable (yellow) color when 5 , 5 -dithiobis-(2nitrobenzoic acid) (Ellman's reagent) is added to sulfhydryl compounds. The chromophoric product resulting from the reaction of Ellman's reagent with the reduced glutathione, 2-nitro-5-thiobenzoic acid had a molar absorption at 412 nm.
Hematological
Analyses. Noncoagulated blood was tested, after collection, for packed cell volume (PCV) and differential white blood cells count according to the procedure outlined by Cheesbrough and McArthur [26]. Haematocrit centrifuge, reader, and capillary tubes were used to determine the PCV. A blood smear was prepared by dragging the blood over the slide with the help of another slide. The smear was then diluted with Leishman's stain and allowed to dry. The stained smears were microscopically observed under oil immersion to determine the number of each cell type per cubic millimeter of blood. The stain used allowed exact identification of neutrophils and leukocytes. Protein level was estimated according to the method described by Lowry et al. [27], using bovine serum albumin as standard.
2.6. Statistical Analysis. All values were expressed as mean ± standard deviation (SD) of six observations. Differences between the groups were determined by one-way analysis of variance (ANOVA) and post hoc testing was performed for intergroup comparisons using Tukey's test using Graph Pad Prism 3 software. Values were regarded as significantly different at < 0.001.
International Scholarly Research Notices 3
Effect of Curcumin on Plasma Biochemical Parameters in
Propanil-Treated Rats. Evaluation of plasma marker indices in rats exposed to propanil (20 mg/kg/bodyweight) showed that there were significant ( < 0.001) increases in AST and ALP activities by 24% and 56%, respectively, when compared with controls ( Figure 1). The result presented in Figure 1 also showed insignificant differences in the activity of ALT and total bilirubin in propanil-challenged rats when compared with the controls. Curcumin, administered alone, had similar effects on plasma biochemical parameters when compared with controls. However, treatment with both curcumin and propanil resulted in a significant reduction in AST and ALP activities.
Effect of Curcumin on Some Hematological Parameters and
Cholesterol Levels in Propanil-Treated Rats. The hematological properties of rats exposed to propanil are shown in Table 1. Treatment with propanil alone did not significantly decrease PCV and neutrophil levels relative to the control. Treatment with propanil elevated the total leukocyte counts but this was not significant relative to the control. However, the curcumin + propanil group showed a significant ( < 0.05) increase in the total leukocyte counts when compared to the control. Furthermore, there were no significant differences in the level of total cholesterol in propanil-treated rats when compared with the control. Although the curcumin + propanil group elicited a significant increase in the cholesterol levels when compared with control animals, however, there was no significant change when compared with the propanil-treated group (Table 1).
Effect of Curcumin on the Antioxidant Status of Propanil-Treated Rats.
The ability of a cell or tissue to maintain its integrity is a function of the levels of the antioxidants present when compared with the level of the oxidants. The balance between these two determines the susceptibility of the cell or tissue to free radicals attack or oxidative stress. Results indicated that lipid peroxidation level was significantly ( < 0.001) increased by 103%, while GSH levels and catalase activity were significantly ( < 0.001) decreased in livers of rats treated with propanil (Figures 2 and 3) when compared with controls. However, there were no significant ( < 0.05) changes in the activity of GST and levels of Vit C and Vit E in liver of rats treated with propanil in comparison with the control group ( Figure 3). Curcumin, administered alone, did not alter the lipid peroxidation and GSH levels when compared with the control. However, supplementation of curcumin to propanil-intoxicated rats significantly ( < 0.05) attenuated the increase in lipid peroxidation levels and restored the GSH levels to normal.
Discussion
The protective role of curcumin in propanil-induced toxicity in Wistar rats was examined in this study. Hematological and biochemical profiles of blood are known to provide important information about the internal environment of the organism. Curcumin, an antioxidant and anticarcinogenic substance, was reported to have a protective effect against liver damage and ferric nitrilotriacetate induced peroxidation of microsomal membrane lipids [28]. The protective action of curcumin against adverse effects of cisplatin had also been reported [29]. Curcumin could exert antioxidative effects either directly as a chemical antioxidant due to its ability to scavenge reactive oxygen and nitrogen free radicals or by modulating cellular defenses which themselves exert antioxidant effects [30,31].
Due to its role in the transformation of environmental xenobiotics, the liver is at great risk of injury. Lipid peroxidation represents one of the most frequent reactions resulting from free radicals' attack on biological structures. In our study, the liver MDA level, index of lipid peroxidation, significantly increased in the propanil-treated rats. Supplementation with curcumin significantly attenuated the increase in lipid peroxidation levels. It is plausible that the mechanism of this protection could be the inhibition of reactive oxygen species generation by suppressing cytochrome P450 isozymes which are involved in the bioactivation of propanil to toxic reactive metabolites such as 3, 4-dichloroaniline [15,32]. Reports are not available regarding Vitamins C and E concentrations in propanil-induced organ toxicity. A previous study had reported decrease in ascorbic acid levels in plasma and liver of rats treated with dimethylnitrosamine [33]. However, in the present investigation, a nonsignificant decrease in Vitamin C was observed in the liver of propaniltreated rats. This decrease may be indicative of increased oxidative stress, free radical formation, and simultaneous damage of the liver plasma membrane lipid bilayer arising from propanil intoxication. Reduction in Vitamin E content following treatment with 1, 2-dibromoethane had also been previously reported [34]. Data obtained for Vitamin E in liver of propanil exposed rats did not reveal any significant change in the vitamin content. Compared with the control, GSH concentrations in treated Wistar rats significantly decreased after exposure to propanil. This corroborated the findings that GSH concentrations in treated Crucian carp significantly decreased dose-dependently after acute exposure to dichloroaniline [35]. The depletion of GSH is suggestive of a deleterious effect of propanil on the antioxidant defense in the liver of rats. It is also consistent with the generation of oxidative stress, thereby reenforcing the role of GSH in molecular protective mechanisms that modulate cellular responses to toxic chemicals. The present study showed that administration of curcumin improved the GSH levels in rats. This finding is similar to the data reported by Piper et al. [36], who indicated an increase of the GSH level corresponding to curcumin dosage in rats fed curcumin, at doses up to 500 mg/kg body weight daily for 14 days.
GST detoxifies a variety of electrophilic compounds to less toxic forms by conjugation with-SH groups from GSH. In the present study, GST activity was not changed by any of the exposure schemes. This correlated with the findings that exposure of rats to cypermethrin in single and repeated doses had no relevant effect on GST activities [37]. In contrast, Saxena et al. [38] reported an increase in GST activity in rat Values are mean ±SD of six rats in each group. a < 0.001 compared to control group. red blood cells after exposure to chlorpyrifos and endosulfan. In addition, on the basis of our observations, a dose as low as 20 mg/kg propanil seemed not to induce such a level of oxidant stress and alteration of the antioxidant system in rats. In consonance with our earlier studies there was a significant decline in the activities of catalase after propanil administration [18], which may be due to oxidative stress; however, treatment with curcumin did not ameliorate this decline.
An increase in plasma cholesterol as a result of pesticide exposure may indicate loss of membrane integrity. Carlson and Kolmodin-Hedman [39] reported that the accumulation of pesticide in the liver was associated with the disturbance of lipid metabolism and elevation of serum cholesterol. In our study, cholesterol levels were mildly increased after the administration of propanil, but this was not statistically significant. Furthermore, there were no significant differences in the level of total cholesterol in propanil-treated rats when compared with the control. Although the curcumin + propanil group elicited a significant increase in the cholesterol levels when compared with control animals, however, there was no significant change when compared with the propanil-treated group. A plausible explanation for this observed effect may be due to synergetic prooxidant International Scholarly Research Notices effects of propanil and curcumin combination. This is in agreement with a report that combining flavonoids with low level pesticides could modify their antioxidant potential and also trigger molecular events that could increase their health risks [40]. This study confirmed the liver damage in the propaniltreated group by the increase in AST and ALP levels in plasma. Exposure of propanil-intoxicated rats with curcumin normalized the activities of AST and ALP to their control values. The decrease in AST and ALP activities supports the hepatoprotective effects of curcumin, consistent with the findings that curcumin modulated the increased activity of marker enzymes and plasma lipid levels in nicotinetreated rats [41]. However, curcumin, when given alone, produced an increase in ALP activity, suggesting that the subacute administration of curcumin at the dose of 50 mg/kg used in this study may not be completely harmless. This agreed with an earlier report that curcumin, administered to human subjects at doses ranging from 0.9 to 3.6 g day −1 for 4 months, could cause some adverse effects including nausea, diarrhea, and increase in serum alkaline phosphatase and lactate dehydrogenase activities [42]. Clinical chemistry analysis of urea and creatinine could serve as prognostic indicators of renal dysfunction in the following exposure to toxicants. In the present study, we observed a nonsignificant increase in the level of creatinine in propanil-treated rats when compared to the control. This may possibly be due to increased catabolic state in the rats from prolonged appetite as a result of manifestation of acute oral toxicity [43]. However, coadminstration of curcumin did not mitigate the increased creatinine levels.
Hematological and biochemical profiles of blood can provide important information about the internal environment of the organism [44]. In this study, there were no prominent changes in the PCV and neutrophils. These findings are supported by the results of Garg et al. [45] who reported that treatment of broiler chicks with monocrotophos did not cause any significant change in erythrocyte count and PCV values. Furthermore, treatment of rats with curcumin + propanil increased total leukocyte count compared to rats in the control group. Our results are in accordance with the findings that treatment of rats with propetamphos plus propolis increased total leukocyte count. The increase observed in total leukocyte count might indicate an activation of the animal's immune system due to either stimulated lymphopoiesis or disturbance of the nonspecific immune system leading to the increased production of leukocytes [46,47].
Conclusions
In summary, the present results suggested that curcumin protected against the liver toxicity induced by propanil treatment in rats. In this study, propanil elicited increase in AST and ALP activities. Furthermore, it had been suggested that lipid peroxidation might be a contributing factor to the development of liver toxicity. The significantly decreased activities of hepatic plasma markers (AST and ALP) and lipid peroxidation marker (malondialdehyde) along with normalizing of the endogenous GSH level suggest that curcumin is a strong antioxidant. We also observed that the curcumin + propanil combination resulted in a significant increase in cholesterol levels when compared with the control group suggesting a possible synergetic hyperlipidemic effect when curcumin combines with propanil. | 2018-04-03T05:58:07.224Z | 2014-10-29T00:00:00.000 | {
"year": 2014,
"sha1": "cd68c7f6e875ffe1f0ffa88f2aec14b2bf264521",
"oa_license": "CCBY",
"oa_url": "http://downloads.hindawi.com/archive/2014/853697.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "9dd9cdbfa08665d57fa9c260bcaf1a8365d66d26",
"s2fieldsofstudy": [
"Biology",
"Chemistry"
],
"extfieldsofstudy": [
"Chemistry",
"Medicine"
]
} |
256808304 | pes2o/s2orc | v3-fos-license | On Abstract Spectral Constants
We prove bounds for a class of unital homomorphisms arising in the study of spectral sets, by involving extremal functions and vectors. These are used to recover three celebrated results on spectral constants by Crouzeix--Palencia, Okubo--Ando and von Neumann in a unified way and to refine a recent result by Crouzeix--Greenbaum.
Introduction
Let M be a bounded linear operator on a complex Hilbert space H and W a bounded subset of C. Estimates of the form for some positive constant κ and all complex polynomials p, are classical, and appear for instance in von Neumann's inequality or Crouzeix's conjecture. Such W is called a κ-spectral set 1 for M and κ is referred to as a spectral constant for M . In this article we present an abstract framework to derive spectral constants in a rather direct way. In particular, this approach allows for unified proofs of several well-known results in this context. More precisely, we study norms of unital bounded homomorphisms where A is unital uniform algebra on some compact Hausdorff space Kthat is, a closed unital subalgebra of the continuous complex-valued functions C(K)-and L (H) denotes the space of bounded linear operators on H. Our interest in norms of such mappings stems from spectral constants appearing in functional calculi for holomorphic functions.
Let Ω ⊂ C be a bounded open subset with smooth boundary. If we write A(Ω) for the space of all continuous functions from the closure Ω − to C that are holomorphic on Ω, then A(Ω) can be viewed as a unital uniform algebra acting on the boundary Ω ∂ by the maximum modulus principle. Suppose that M is a bounded linear operator on H whose spectrum is contained in Ω. The latter guarantees that for each f ∈ A(Ω) the integral is well-defined. For the rest of the paper we will refer to the linear map γ : A(Ω) → L (H) as the (Dunford-Riesz) functional calculus of M on Ω. It is well-known that γ is a unital bounded homomorphism such that γ(p) = p(M ) for polynomials p, see e.g. [DS58], where p is interpreted as a function on Ω in the obvious way. Clearly, bounds on the norm of γ provide spectral constants for M . We note in passing that the above homomorphism actually extends to the algebra of holomorphic functions on Ω, equipped with the topology of uniform convergence on compact sets, see e.g. [Haa18].
Let us now return to the general setting. Note that we always have the lower bound γ ≥ 1 as γ(1) = 1. Our starting point for finding an upper bound for the norm of γ is the assumption that the norm of γ is attained. More precisely, a couple (f 0 , x 0 ) ∈ A × H with f 0 K = 1 and x 0 = 1 is called an extremal pair for γ if γ = γ(f 0 )x 0 . While the existence of extremal pairs may seem restrictive, the fact that their existence is e.g. guaranteed for functional calculi γ on finite-dimensional spaces H if Ω is convex, see [Cro04, Theorem 2.1] is typically sufficient to find spectral constants for general H. The main advantage of extremal pairs (f 0 , x 0 ) is that they translate the operator norm γ into a Hilbert space norm γ(f 0 )x 0 , which is a fruitful tool that we shall heavily exploit. Moreover, we use recently introduced extremal measures associated with an extremal pair, [BGG + 20]. We emphasize that we are merely working with the existence of an extremal pair rather than assuming explicit knowledge of (f 0 , x 0 ) which would trivially yield the norm of γ.
Our research builds on techniques for deriving spectral constants by employing contour integral representations of the functional calculus. These were developed in the last decades, including the influential paper by Delyon-Delyon [DD99], and Crouzeix's celebrated result showing that the numerical range of any bounded linear operator is an 11.08-spectral set, see [Cro04] as well as [Cro07]. At the same time Putinar-Sandberg [PS05] also studied properties of the operator-valued double-layer potential, which is crucial in all these references to derive an norm bound on a related "symmetrized functional calculus" 2 where Φ : A(Ω) → A(Ω) refers to a suitable antilinear transformation. See also the recent work on finite-dimensional dilations in [HL21, Theorem 1.6]. The more recent refinement of Crouzeix's theorem due to Crouzeix-Palencia [CP17], see also [RS18], exploits more properties of the double-layer potential, and of the specific Φ, yielding that the numerical range is even a (1 + √ 2)spectral set. As shown in [RS18], their proof technique cannot yield any better constant, which implies that yet more structure of the problem has to be used to possibly show Crouzeix's conjecture stating that the numerical range is a 2-spectral set.
Under additional assumptions, the bound 1 + √ 2 can indeed be lowered. Indeed, if Ω is a disk containing the numerical range, then the optimal spectral constant 2 can be proved by slightly diverging from Crouzeix-Palencia's proof and instead exploiting properties of extremal pairs, see [CGL18]. The result, however, is much older as it traces back to Okubo-Ando and dilation theory, see [OA75]. Yet, the fact that such contour integral methods can be used to prove sharp spectral constants seems remarkable. Another instance of this is Delyon-Delyon's reasoning for von Neumann's inequality [DD99], see also [CG19, Section 4].
One finding of our current work is that all of the mentioned results indeed follow directly from one abstract theorem. Thereby we aim to provide a very clean mutual relation of existing results around the conjecture. Surely, this is not the first attempt to study the approach to Crouzeix's conjecture in an abstract fashion. In [RS18] and [COR20] an abstract framework was used to discuss limitations of the proof ingredients used by Crouzeix-Palencia and how to possibly overcome those. On the other hand, the links between von Neumann's inequality and statements about the numerical range have also been studied in the past, see [OA75] and [PPW20].
Finally, we remark that spectral constants naturally appear in numerical linear algebra, as they can be used to bound residuals appearing in Krylov subspace approximation methods such as the GMRES algorithm, see e.g. [CG19, Section 7.1].
Main Result
First we discuss some properties of extremal pairs. In [CGL18, Theorem 5.1] the following proposition was proved for γ associated to functional calculi. We slightly adapt their argument, and thereby showing that the result also holds for general γ.
Proposition 1. Let A be a unital uniform algebra acting on a compact Hausdorff space K, H a Hilbert space and γ : n γ(f 0 ))x 0 . Squaring both sides and using that (f 0 , x 0 ) is extremal for γ yields and therefore, after rearranging, The desired result follows as the last inequality holds for all n ∈ N.
We will use the following elementary fact in the proof of the subsequent proposition.
Lemma 2. Let M be a bounded linear operator on a Hilbert space
Proof. One readily verifies that and this completes the proof.
The next proposition can be found in [BGG + 20, Theorem 4.5] for functional calculi γ. Using elementary operator theory from the book [Pau02], we present a new proof that also works for general γ.
Proposition 3. Let A be a unital uniform algebra acting on a compact Hausdorff space K, H a Hilbert space and γ : A → L (H) a unital bounded homomorphism. If (f 0 , x 0 ) is extremal for γ, then there is a Radon measure µ on K such that for all f ∈ A. Moreover, µ is a probability measure.
Proof. Consider the functional ω : and therefore |ω(f )| ≤ 1 as desired. By [Pau02, Proposition 2.12] the functionalω : A + A * → C given byω(f + g * ) : = ω(f ) + ω(g) * for all f, g ∈ A is well-defined and is the unique positive extension of ω to the operator subsystem A + A * of the C*-algebra C(K). By a Hahn-Banach argument, see also [Pau02, Exercise 2.10], we can extendω to a positive functional on C(K), which we also denote byω. So there exists a unique Radon measure µ on K such thatω for all h ∈ C(K). In particular, which shows that µ is a probability measure.
A Radon measure µ on K satisfying (3) is called an extremal measure for the extremal pair (f 0 , x 0 ). Any such µ is linked to f 0 by the relation which follows directly from Proposition 1. In particular, if γ > 1, then f 0 and 1 are mutually orthogonal in the Hilbert space L 2 (K, µ).
Remark 4. The proof of Proposition 3 tells us that an extremal measure µ . The authors of [GG71] determined exactly the topological conditions on Ω that ensure that A(Ω) is Dirichlet.
For any bounded antilinear map Φ : A → A we define a bounded linear map γ Φ : A → L (H) by the formula We are now ready to discuss the main theorem.
Theorem 5. Let A be a unital uniform algebra acting on a compact Hausdorff space K, H a Hilbert space, γ : A → L (H) a unital bounded homomorphism and Φ : A → A a bounded antilinear map. If γ > 1 and (f 0 , x 0 ) is extremal for γ, then Proof. It follows from Proposition 1 that γ(f 0 )x 0 , x 0 = 0 and therefore This proves the desired inequality.
Remark 6. The infimum appearing in Theorem 5 is by definition the shortest distance from γ Φ to the subspace A ′ in the Banach space of bounded linear maps from A to L (H). Moreover, for each ω ∈ A ′ the map γ Φ − ω is a rank-one perturbation of γ Φ .
The following example is taken from [RS18], where it was used to prove that κ := 1 + √ 2 is a sharp absolute upper bound for the operator norms of functional calculi γ : A(Ω) → L (H) for which there exists a contractive antilinear map Φ : In addition to the extremal function f 0 given in [RS18], we provide an extremal vector x 0 and an expression for the associated extremal measure µ. The example also shows that the estimate in Theorem 5 cannot be improved.
for all f ∈ A(Ω). It is not difficult to see that the pair (f 0 , x 0 ) defined by is extremal for γ and that If µ is an extremal measure for (f 0 , x 0 ), then a direct computation yields for all f ∈ A(Ω). The antilinear map Φ : A(Ω) → A(Ω) given by 1 by equation (4). So Theorem 5 gives which implies that the infimum is attained at ω = 0 and is equal to γ Φ = 2. 2πi
Applications to Functional Calculi
Let v : Ω ∂ → C be the map that sends any boundary point in Ω ∂ to the outward unit normal vector of Ω at that point. We define the double-layer potential P : Ω ∂ → L (H) of M on Ω by for all f ∈ A(Ω). Already in Delyon-Delyon's paper [DD99] as well as in Putinar-Sandberg's [PS05] and Crouzeix's paper [Cro07], the fact that the values of P are non-negative was used. Since c⊂Ω ∂ connected component ℓc 0 P (σ c (s)) ds = 2, this readily implies that γ Φ ≤ 2 and thus which we shall use to reprove Crouzeix-Palencia's result (Theorem 11) result and Okubo-Ando's result (Theorem 14) as an application of Theorem 5. It turns out that von Neumann's inequality (see Theorem 16 below) also follows from Theorem 5. For this we need the following proposition due to Caldwell-Greenbaum-Li [CGL18, Lemma 2.1]. For the reader's convenience, we sketch the proof. For any Hermitian R we write λ min (R) for the smallest element of the (real) spectrum of R. for f ∈ A(Ω) is bounded. Using that P (z) − λ min (P (z)) ≥ 0 for all z ∈ Ω ∂ , one can prove that ω ∈ A(Ω) ′ satisfies the desired estimate.
Note that Theorem 5 relies on the existence of extremal pairs. For finitedimensional Hilbert spaces it is well-known that functional calculi admit extremal pairs, see [ Since x ∈ H was chosen arbitrarily, we obtain γ(p) ≤ κ p Ω ∂ . Since we assumed that the polynomials are dense in A(Ω), we obtain γ ≤ κ.
Remark 10. Due to the use of the holomorphic functional calculus, the results on spectral sets always comes with the obstruction that the set Ω has to be open, smoothly bounded and contain the spectrum of M . In light of von Neumann's inequality and Crouzeix's conjecture this is unnatural in the sense that one would like to allow for spectral points on the boundary of the spectral set. However, if the spectral constants are universal and thus not depending on Ω such stronger statements follow automatically. More precisely, given a fixed bounded set W ⊂ C and a constant κ ≥ 1, the statement for all polynomials p, and Ω ⊃ W open, convex, smoothly bounded, which contain the spectrum of M , implies that W is a κ-spectral set for M .
Crouzeix-Palencia's Result
We apply our main result to recover that, if Ω is convex and contains the closure of the numerical range of M , then 1 + √ 2 is a spectral constant for M on Ω. This is was first shown by Crouzeix-Palencia [CP17] by combining γ Φ ≤ 2 with the fact that Φ ≤ 1 in a rather technical fashion. As observed in [RS18], this final step can be simplified by a trick involving the formula together with the multiplicativity of the homomorphism. The following proof is different to that approach, but instead exploits the Hilbert space structure and the homomorphism through the extremal measure.
Theorem 11 (Crouzeix-Palencia). Suppose that Ω ⊂ C is a convex bounded open subset with smooth boundary. If M is a bounded linear operator on a Hilbert space H such that W (M ) − ⊂ Ω, then 1 + √ 2 is a spectral constant for M on Ω.
Proof. If M ↓ is a finite-dimensional compression of M , then W (M ↓ ) ⊂ W (M ) and therefore W (M ↓ ) − ⊂ Ω. Hence by Proposition 9 we can reduce to the case where H is finite-dimensional.
Let γ : A(Ω) → L (H) be the functional calculus of M on Ω. Let Φ : A(Ω) → A(Ω) be the conjugate Cauchy transform on Ω. The convexity of Ω and the inclusion W (M ) ⊂ Ω imply that (5) holds. If we now use Proposition 3 to find an extremal measure µ for an extremal pair (f 0 , x 0 ) of γ, then we can estimate where we used in the last step that Φ is contractive by convexity of Ω. Thus an application of Theorem 5 yields the desired result.
Remark 12. In fact, the technique used in [RS18] and later in [COR20] shows a slightly stronger statement of Crouzeix-Palencia's result. In fact, it is proved that 1 + √ 2 is still a spectral constant for M on Ω after replacing the convexity assumption on Ω and the inclusion W (M ) ⊂ Ω by the weaker condition that there is a contractive antilinear map Φ : A(Ω) → A(Ω) for which γ Φ ≤ 2 holds. This more general result can also be recovered with a proof technique similar to that of Theorem 11 under the additional assumption that an extremal pair exists.
Remark 13. Suppose that Ω is convex and contains W (M ) − . Note that, if it were true that then γ ≤ 2 would follow by Theorem 5. This would establish Crouzeix's conjecture [Cro04].
Okubo-Ando's Result
The original proof of the following result, due to Okubo-Ando [OA75], was based on dilation theory. A more recent proof strategy involved the estimate γ Φ ≤ 2 and properties of extremal pairs, see [CGL18]. Following the latter approach, we demonstrate how the result can be obtained from Theorem 5. Remark 15. Drury [Dru08] was able to refine Okubo-Ando's result in the case where Ω is the open unit disk centered at the origin by showing the following: Under the conditions of Theorem 14, it holds that for all polynomials p with p Ω ∂ ≤ 1, where ν : [0, 1] → [1, 2] is an explicitly given strictly decreasing function with ν(0) = 2 and ν(1) = 1. We point out that our methods do not seem to allow for reproving this stronger statement. This may not be surprising as Dury's technique strongly rests on a theorem by Berger-Stampfli [BS67, Theorem 4] stating that, if the numerical range of M is included in the closed unit disk, then the same is true for p(M ) provided that p Ω ∂ ≤ 1 and p(0) = 0. The latter result was originally proved via dilation theory, see also [PS05, Theorem 3] for a proof relating to the doublelayer potential. There also is a more recent proof of Berger-Stampfli's result using Blaschke products, [KMR16, Theorem 2.1].
Von Neumann's Inequality
The following well-known result, due to von Neumann [Neu50], has its origins already in 1950. The proof presented below combines Theorem 5 with an idea from [CG19, Lemma 6], which in turn can be traced back to [DD99]. In what follows we write T for the unit circle in C.
Theorem 16 (von Neumann). Suppose that Ω ⊂ C is an open disk centered at the origin. If M is a bounded linear operator on a Hilbert space H such that M T ⊂ Ω, then 1 is a spectral constant for M on Ω.
Proof. If M ↓ is a finite-dimensional compression of M , then M ↓ ≤ M and therefore M ↓ T ⊂ Ω. Hence by Proposition 9 we can reduce to the case where H is finite-dimensional.
Let γ : A(Ω) → L (H) be the functional calculus of M on Ω and assume to the contrary that γ > 1. Let r denote the radius of the disk Ω and consider the arc length parametrization σ : which contradicts the assumption γ > 1. Hence we must have γ = 1.
Crouzeix-Greenbaum's Result
Using an extremal measure µ for an extremal pair (f 0 , x 0 ) of γ we find that see the proof of Theorem 11 for a similar reasoning. This directly yields the following weaker version of Theorem 5 to reprove a statement from Crouzeix-Greenbaum [CG19].
Apart from the assumption that γ admits an extremal pair, Corollary 17 is precisely [CG19, Theorem 2]. Our main result shows that the term ω can be dropped and seems to be an artifact of the proof technique used in [CG19]. However, in most applications presented in [CG19] the functional ω is chosen to be 0 anyway.
Relation to Numerical Radii
The numerical radius of a bounded linear operator on M on H is given by |z| and defines a norm on L (H) that is known to be equivalent to the operator norm on L (H). Recall our general framework where A is a unital uniform algebra acting on a compact Hausdorff space K and γ : A → L (H) a unital bounded homomorphism. Badea-Crouzeix-Klaja [BCK18], see also [DPW18], it was shown that the induced numerical radius operator norm of γ, which we denote by γ w , is explicitly linked to γ via 2 γ w = 1 γ + γ .
Hence bounds on γ w give bounds on γ and vice versa. Also note that an analogous notion of extremal pairs and associated measure can be defined in this setting, see [BGG + 20]. A variant of Proposition 3 using this different norm can be proved easily in a similar fashion. | 2023-02-13T06:41:50.934Z | 2023-02-10T00:00:00.000 | {
"year": 2023,
"sha1": "4e7a59b0ab2bafc0d52e48b9407233e26cd394bf",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "4e7a59b0ab2bafc0d52e48b9407233e26cd394bf",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Computer Science",
"Mathematics"
]
} |
255034783 | pes2o/s2orc | v3-fos-license | Fatty-Acid-Based Membrane Lipidome Profile of Peanut Allergy Patients: An Exploratory Study of a Lifelong Health Condition
Peanut allergy is a lifelong, increasingly prevalent, and potentially life-threatening disease burdening families and communities. Dietary, particularly polyunsaturated fatty acids (PUFAs), intakes can exert positive effects on immune and inflammatory responses, and the red blood cell (RBC) membrane lipidome contains stabilized metabolic and nutritional information connected with such responses. The fatty-acid-based membrane lipidome profile has been exploratorily evaluated in a small cohort of patients (eight males and one female, age range 4.1–21.7 years old, body mass index BMI < 25) with angioedema and/or anaphylaxis after peanut ingestion. This analysis was performed according to an ISO 17025 certified robotic protocol, isolating mature RBCs, extracting membrane lipids, and transforming them to fatty acid methyl esters for gas chromatography recognition and quantification. Comparison with a group of age- and BMI-matched healthy individuals and with benchmark interval values of a healthy population evidenced significant differences, such as higher levels of ω-6 (arachidonic acid), lower values of ω-3 eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), together with an increased ω-6/ω-3 ratio in allergic patients. A significant inverse correlation was also found between specific immunoglobulin E (IgE) levels and ω-6 di-homo-gamma-linolenic acid (DGLA) and total PUFAs. Results of this preliminary study encourage screenings in larger cohorts, also in view of precision nutrition and nutraceuticals strategies, and stimulate interest to expand basic and applied research for unveiling molecular mechanisms that are still missing and individuating treatments in chronic allergic disorders.
Introduction
Peanut allergy (PA) is one of the most common food allergies in the pediatric age, affecting approximately 1-3% of children in Western countries, and the prevalence has been increasing in the last decades worldwide [1,2]. Compared with other food allergies, PA is usually lifelong and associated with higher rates of accidental exposure, severe reactions, and potentially fatal anaphylaxis. This, in turn, has a significant impact on the management of the patient, also from a nutritional point of view. The immune system and impaired immune maturation are involved in tolerance breakdown and the development of immunemediated diseases such as food allergies [3]. The involvement of lipids is progressively recognized both as basic constituents of cellular structures involved in immune regulation and components of diets [4].
Fatty Acids as Structural and Functional Constituents of Cell Membranes and Relationship with Diet
Fatty acids are distinguished by their structures into saturated, monounsaturated, and polyunsaturated families (SFA, MUFA, and PUFA), with an important dietary dependence on PUFAs, which are not directly produced by human metabolism. The omega-6 (ω-6) and omega-3 (ω-3) precursors, linoleic and alpha-linolenic acids, are processed after intake by desaturase and elongase enzymes to form long-chain PUFAs (LC-PUFAs), as shown in Figure 1 [5]. management of the patient, also from a nutritional point of view. The immune system and impaired immune maturation are involved in tolerance breakdown and the development of immune-mediated diseases such as food allergies [3]. The involvement of lipids is progressively recognized both as basic constituents of cellular structures involved in immune regulation and components of diets [4].
Fatty Acids as Structural and Functional Constituents of Cell Membranes and Relationship with Diet
Fatty acids are distinguished by their structures into saturated, monounsaturated, and polyunsaturated families (SFA, MUFA, and PUFA), with an important dietary dependence on PUFAs, which are not directly produced by human metabolism. The omega-6 (ω-6) and omega-3 (ω-3) precursors, linoleic and alpha-linolenic acids, are processed after intake by desaturase and elongase enzymes to form long-chain PUFAs (LC-PUFAs), as shown in Figure 1 [5]. LC-PUFAs incorporated into membrane phospholipids are crucial for their subsequent enzymatic release and, consequently, the formation of bioactive lipid mediators (prostaglandins, leukotrienes, etc.), playing important and well-known biological roles in the development and regulation of the immune system and inflammatory process [6,7]. It is worth recalling that allergic patients start to exclude food(s) from their diets, and this is an initial step for introducing fewer essential fatty acids (EFAs) or breaking the balance between ω-6 and ω-3 intakes, i.e., between pro-and antiinflammatory components. Indeed, an optimal ω-6/ω-3 balance realizes the ordered sequence of ω-6 LCPUFAs derived mediators, which initiate the "reactivity" response, Figure 1. The omega-6 (ω-6) and omega-3 (ω-3) pathways of main long-chain polyunsaturated fatty acid (LCPUFA) biosynthesis, with the interplay of desaturase and elongase enzymes starting from essential (dietary) fatty acid precursors.
LC-PUFAs incorporated into membrane phospholipids are crucial for their subsequent enzymatic release and, consequently, the formation of bioactive lipid mediators (prostaglandins, leukotrienes, etc.), playing important and well-known biological roles in the development and regulation of the immune system and inflammatory process [6,7]. It is worth recalling that allergic patients start to exclude food(s) from their diets, and this is an initial step for introducing fewer essential fatty acids (EFAs) or breaking the balance between ω-6 and ω-3 intakes, i.e., between pro-and anti-inflammatory components. Indeed, an optimal ω-6/ω-3 balance realizes the ordered sequence of ω-6 LCPUFAs derived mediators, which initiate the "reactivity" response, followed by mediators derived from ω-3 LCPUFAs called "specialized proresolving mediators" (SPMs), which bring resolution and tissue repair [8,9]. Only with the balanced production of these mediators can the overall cell response occur in a physiological manner. Dietary unbalances are recognized to be risk factors for atopy, asthma, and allergy [2,10], and fatty acid supplementations are proposed for the treatment of allergic conditions [11,12]. On the other hand, the association between ω-3 and ω-6 intakes and clinical improvements in food allergy patients [13] are under discussion and, more importantly, are not combined yet with the examination of the patient's fatty acids status, per se or during supplementation, in particular utilizing erythrocyte membrane lipidome information, which is a reliable mirror of fatty acid intake and metabolism in the body tissues [14].
Fatty-Acid-Based Membrane Lipidome Analysis
Fatty acid detection can be performed in plasma lipids or blood cell lipids, in particular red blood cells (RBCs), providing different types of information. Diet composition of the days before blood withdrawal strongly affects fatty acid plasma levels, in contrast to RBC membranes, where fatty-acid-based phospholipid composition reflects a balance between nutritional and metabolic factors (i.e., fatty acids transformation into phospholipids). Easiness of the sampling and work-up procedures leads to be more prone to choose plasma or whole blood specimens, but knowledge of the importance of membranes for cell signaling and precision medicine applications strongly indicates RBC membrane isolation and analysis to obtain information on the stabilized metabolic and nutritional status of patients [14]. Indeed, the fatty-acid-based membrane lipidome profile of mature RBCs (mean lifetime of 120 days) was developed and used as a nutritional, homeostatic, and metabolic biomarker in several human physiopathological states [15,16]. Certified laboratory protocol with fatty acid identification and quantitation following ISO 17025 international requirements ensures the reliability and repeatability of the results, as well as the quality of the data. With such a procedure, observational clinical studies were performed identifying membrane profiles in different health conditions [17][18][19][20][21][22]. This information opens the way to personalize the approach of membrane lipid therapy, a natural medicine tool effective in several health conditions [23,24].
Results
With the aim of exploring the relevance of the membrane lipidome profile in food allergy, a cohort of peanut-allergic patients was selected among subjects followed along years at the Pediatric Immunopathology and Allergology Unit of the university hospital. As shown in the diagram of Figure 2, we recruited n = 29 subjects with a clear history of angioedema and/or anaphylaxis, and exclusion criteria were applied (body mass index (BMI) > 25, other food allergies, autoimmune diseases, and allergic reactions in the last 18 months). The final cohort presented n = 9 patients (age 12 ± 5.6 years old). At the same time, 15 healthy individuals (8 females and 7 males) with BMI < 25 and age = 17 ± 4 years old were selected from the anonymous database available at the Lipidomic Laboratory, for which informed consent for research use was gathered at the moment of blood withdrawal, and respect of EU general data protection regulation 2016/679 (GDPR) guaranteed by the ISO 17025 certification. The controls did not have allergic problems. Mature RBC membrane fatty acid values compared with those of the patients were obtained by the same procedure. The observational study focused on the fatty acid composition of RBC membranes addressing the following key points: (a) the status of ω-6 vs. ω-3 fatty acid residues of the membrane phospholipids as the expression of proinflammatory tendency in allergic patients, and (b) the correlation of membrane lipidome profile features with specific immunoglobulin E (IgE) levels.
Clinical and laboratory data from peanut-allergic patients are described in Table 1. All patients had a history of angioedema and/or anaphylaxis after ingestion of peanuts; however, the recruited patients did not have any allergic reactions in the last 18 months. Clinical and laboratory data from peanut-allergic patients are described in Table 1. All patients had a history of angioedema and/or anaphylaxis after ingestion of peanuts; however, the recruited patients did not have any allergic reactions in the last 18 months. In Table 2, the results of the mature RBC fatty-acid-based membrane lipidome analysis are shown; as in previous work, the main fatty acids of the RBC membrane phospholipids are grouped to give a cohort of 10 SFA, MUFA, and PUFA components, the latter ones representing the ω-6 and ω-3 metabolic cascades shown in Figure 1 [15]. Each unsaturated fatty acid of the cohort is recognized by appropriate standards, ensuring that it is not superimposed to geometrical or positional isomers. Calibration with standard references allows to calculate the quantity of each fatty acid in the sample and express it as a relative quantitative amount (% rel. quant.) over the total of the 10 fatty acid quantities. In Table 2, the results of the mature RBC fatty-acid-based membrane lipidome analysis are shown; as in previous work, the main fatty acids of the RBC membrane phospholipids are grouped to give a cohort of 10 SFA, MUFA, and PUFA components, the latter ones representing the ω-6 and ω-3 metabolic cascades shown in Figure 1 [15]. Each unsaturated fatty acid of the cohort is recognized by appropriate standards, ensuring that it is not superimposed to geometrical or positional isomers. Calibration with standard references allows to calculate the quantity of each fatty acid in the sample and express it as a relative quantitative amount (% rel. quant.) over the total of the 10 fatty acid quantities. From these values, the ω-6/ω-3 ratio, related to pro-and anti-inflammatory balance [25], and peroxidation index (PI), as a measure of the PUFA contribution to elevating membrane peroxidizability [26,27], were calculated. Each of the 10 fatty acids and related indexes were compared with the 15 age-and BMI-matched healthy controls and with the benchmark of interval values referred to a healthy population described in previous works [15][16][17]. In Table 2, significant changes observed between patients and controls are reported with their p-values. 129.95 ± 12.04 * † Fatty acids are reported from the gas chromatographic analysis (GC) after transformation of membrane phospholipids of mature RBCs into fatty acid methyl esters (FAMEs); ‡ values are expressed as relative percentages (mean ± SD) of the quantitative values of each fatty acid obtained by calibration curves of the standards and referred to the representative 10 fatty acids cohort. Statistical analyses comparing patients with healthy controls gave the following p-values: * ≤0.026; ** ≤0.0022; *** ≤0.0001. Table 2, in allergic patients, we found a marked unbalance mainly concerning the PUFA residues of cell membrane phospholipids. Compared with the cohort of healthy controls, the patients showed a significant increase in ω-6 20:4 (arachidonic acid, ARA), PUFA ω-6 and the ratio ω-6/ω-3, whereas ω-3 20:5 (eicosapentaenoic acid, EPA), 22:6 (docosahexaenoic acid, DHA), total PUFA ω-3, and the PI value were significantly diminished. In addition, total saturated fatty acids (SFAs) and stearic acid (18:0) were significantly lower than controls. The distribution of PUFA values and ratio as scattered dot plots showing each patient is presented in Figure 3 in comparison with those found in healthy controls, reporting also the benchmark of interval values in the healthy population. Although the cohort of patients is small, in Figure 3, it is possible to appreciate that their membrane fatty acid values are all positioned outside (lower or higher) both healthy controls and benchmark interval values. It is worth underlining that levels of trans fatty acids isomers of oleic and arachidonic acids were also evaluated since these compounds are available as standard references [17,18] and found to be significantly increased, although within the values reported for the benchmark (≤0.4). fatty acids isomers of oleic and arachidonic acids were also evaluated since these compounds are available as standard references [17,18] and found to be significantly increased, although within the values reported for the benchmark (≤0.4). Table 2. The benchmark of interval values in healthy population is highlighted in light gray in the graphs.
Patients' clinical and laboratory data (Table 1) were then correlated to membrane FA values (Table 2), resulting in the heatmap shown in Figure 4. (Table 1) and fatty acid values ( Table 2). Table 3, significant inverse correlations were found for peanut-specific IgE with ω-6 20:3 di-homo-gamma-linolenic acid DGLA and total PUFA (p-value = 0.033 and p-value = 0.038, respectively). Table 2. The benchmark of interval values in healthy population is highlighted in light gray in the graphs. p-values, *** ≤ 0.0001.
As detailed in
Patients' clinical and laboratory data (Table 1) were then correlated to membrane FA values (Table 2), resulting in the heatmap shown in Figure 4. fatty acids isomers of oleic and arachidonic acids were also evaluated since these compounds are available as standard references [17,18] and found to be significantly increased, although within the values reported for the benchmark (≤0.4). Table 2. The benchmark of interval values in healthy population is highlighted in light gray in the graphs. p-values, *** ≤ 0.0001.
Patients' clinical and laboratory data (Table 1) were then correlated to membrane FA values (Table 2), resulting in the heatmap shown in Figure 4. (Table 1) and fatty acid values ( Table 2). Table 3, significant inverse correlations were found for peanut-specific IgE with ω-6 20:3 di-homo-gamma-linolenic acid DGLA and total PUFA (p-value = 0.033 and p-value = 0.038, respectively). (Table 1) and fatty acid values ( Table 2). Table 3, significant inverse correlations were found for peanut-specific IgE with ω-6 20:3 di-homo-gamma-linolenic acid DGLA and total PUFA (p-value = 0.033 and p-value = 0.038, respectively). Table 3. Correlation (expressed by the Spearman correlation coefficient, r) of the fatty acid values of Table 2 with the peanut-specific IgE shown in Table 1. See also heatmap in Figure 4.
Discussion
This is the first study to explore the relevance of the membrane lipidome profile in patients with peanut allergy. We observed a clear unbalance between ω-6 and ω-3 fatty acids in this cohort of patients, mainly due to an increase in ω-6 20:4 (ARA) and a decrease in ω-3 (EPA and DHA). Considering the role of ARA as a precursor of proinflammatory eicosanoids [9], but also as a result of the metabolism of omega-6 linoleic acid taken from the diet (see Figure 1), our data support the synergy of metabolism and diet to build up a specific profile of the cell membranes of peanut-allergic patients [28]. The role of omega-3, both as essential components of the diet and as modulators of the immune and anti-inflammatory processes, is well known [28], and in our cohort, a significant reduction in ω-3 fatty acids involves EPA and DHA. Examining in detail the patients' values in Figure 3, in nine patients, the DHA values resulted to be in the lower ranges compared with both healthy controls and the benchmark of the healthy population. As shown in Figure 1, EPA and DHA are the members of the ω-3 cascade containing the highest number of double bonds among all PUFAs (see Figure 1). Besides their transformation into bioactive lipid mediators with immune-stimulating and anti-inflammatory activities [12], DHA exerts a greater influence on membrane biophysical properties such as flexibility, fluidity, and thickness [29,30]. It is also clear that the quantity and quality of fat intake directly from seafood (fish, algae) can influence the levels of these fatty acids in cells; however, all patients apparently had regular fish consumption in their nut-free diet, not different from the healthy controls. Another cause of EPA and DHA diminutions can be the efficiency of the metabolic transformations in these patients, as shown in Figure 1. Indeed, humans have a limited capability to synthesize LC PUFA from the precursor alpha-linolenic acid [31], so that daily intakes of 250 mg EPA and DHA are indicated by the most relevant food safety agencies, such as EFSA (European Food Safety Authority) [32]. These preliminary data on ω-3 deficiency must be deepened in a study with larger cohorts.
We also consider it relevant that the ω-6 components came into the scenario of allergic patients, in particular with an increase in arachidonic acid, with known roles in the propagation of inflammatory responses and cellular reactivity. Moreover, ω-6 DGLA, which is a precursor of arachidonic acid and prostaglandins (PG series 1) [33], showed an inverse correlation with specific IgE levels. Literature data connect DGLA supplementation with an increase in PGD1 and improvement in atopic dermatitis, although in an animal model [34]. Moreover, prostaglandin E1 (PGE1) originates from DGLA, and a synthetic analog misoprostol has been reported to modulate histamine release from basophils [35]. It must be underlined that clinical trials on gamma-linolenic acid (GLA) supplementation as a precursor of DGLA (see Figure 1) gave very heterogeneous results, but no data are yet reported on the follow-up of the GLA-DGLA metabolic transformation and incorporation of DGLA in membrane phospholipids. Evidently, the possibility of monitoring DGLA levels in the RBC membrane offers the best strategy to follow up subjects during treatments. Moreover, total PUFA levels were negatively correlated with specific IgE, and both correlations underline the importance of deepening the PUFA status in patients in view of its importance for cell signaling. On the other hand, despite multiple evidence for PUFA relevance in different health conditions [6,7,9,13,33], the punctual follow-up of patients is still missing in clinical approaches.
The exclusion of subjects with a BMI > 25 from our cohort, as well as from the healthy controls, takes into account that overweight and obesity are known to increase arachidonic acid levels in RBC membranes [22,36], and we wanted to exclude interferences from known proinflammatory conditions. In addition, the distance of 18 months from the last allergic episode in patients and the exclusion of allergic conditions in the healthy cohort were carefully checked to eliminate interferences from immune reactivity. It is interesting to observe that in obesity, the membrane fatty acid asset had some common features with that of allergic patients (such as low levels of DHA and increased levels of ARA and omega-6/omega-3 ratio) [22,36], whereas an increase in the SFA/MUFA ratio was not found, which is known to influence membrane properties through an increased rigidity of RBC membranes [37].
The overall oxidative reactivity estimated with the peroxidizability of RBC membranes (peroxidation index, PI; cfr., Table 2) in allergic patients was lower than in controls. However, it must be underlined that we did not perform direct measurements of peroxidation processes through a measure of oxidation metabolites, and this is a limitation of the present study. In our analysis, we also measured trans fatty acid (TFA) isomers of oleic and arachidonic acids, which are known markers of endogenous cis-trans isomerization of the corresponding cis MUFA and PUFA caused by increased free radical production [38,39]. Previously, we reported TFA to increase in the blood cell membranes of children affected by atopic eczema/dermatitis syndrome [40]. In our allergic cohort, we found TFA significantly increased compared with healthy controls (0.35 ± 0.07 vs. 0.18 ± 0.05, p-value ≤ 0.0001) [15]; however, the threshold value of the benchmark (0.4%) was not overcome.
The detection of fatty acids and isomers is one of the various aspects of the analytical protocol that starts from the choice of mature RBC membranes for sampling, as detailed in the experimental part and discussed in previous research papers and reviews [16,[20][21][22]. Precision medicine must rely on analytical data not only obtained by protocols unified among laboratories but also certified for the results and reliability by a competent auditing process, usually provided by national bodies of accreditation through compliance with the ISO 17025 regulation. The use of gas chromatographic methodology, with cis and trans fatty acid references and calibration procedures, allows to examine fatty acid levels with the highest molecular identification and sensitive quantification, and this is needed when membrane lipidome profiles are developed to identify disease onset [14,16,18,20]. Moreover, the inclusion of a robotic platform in the certified protocol reduces errors due to manual operations. Our small cohort certainly took advantage of such precise measures to evidence significant results, but we are aware that the sample size must be increased to achieve greater statistical power. Here, we preliminarily showed results of membranebased diagnostics in peanut allergy patients, supporting the importance of this tool that has reached maturity and technological advancement to serve larger screenings.
Further studies will be able to address an integrated vision in allergic patients in which specific alterations found in RBC membrane lipidome profiles mirror crucial changes in molecular components and related signaling to develop allergic reactivity [41]. Some of these changes are indicated in Figure 5, starting from the formation of the fatty acid pool, influenced by nutritional and metabolic contributions specific to each individual, and the consequences on the composition of membrane phospholipids. Once membranes are formed, they present different qualities and quantities of PUFAs that, in turn, can bring a balance/unbalance of bioactive lipids created in cells after the release of PUFA residues from phospholipids. An increase in ARA and a diminution in DGLA, EPA, and DHA in allergic patients can contribute to an unbalance of signaling with an influence on inflammatory and anti-inflammatory cytokine productions, as well as on a diminution in the protection from antigen-induced activation given by ω-3 proresolving mediators (epoxins, resolvins, protectins) [41], resulting in a general cellular response toward hypersensitivity and augmented IgE production [42]. DHA in allergic patients can contribute to an unbalance of signaling with an influence on inflammatory and anti-inflammatory cytokine productions, as well as on a diminution in the protection from antigen-induced activation given by ω-3 proresolving mediators (epoxins, resolvins, protectins) [41], resulting in a general cellular response toward hyper sensitivity and augmented IgE production [42]. Alteration of the PUFA composition in membranes is basic information to acquire and translate into precision medicine and nutrition strategies. In fact, the type and dosage of fatty acids able to rebalance the molecular status of patients can be personalized for each distinct profile and regularly monitored. This is particularly important in children in which PUFAs are used to sustain the exponential growth of cells for all tissues, and their exact levels must be determined [43] in order to promptly individuate deficiency or excess with impact on normal tissue functioning [44,45]. Our exploratory study highlights cell membranes for effecting PUFA detection exactly in the active site of their immuno modulatory effects, evidencing molecular mechanisms that are still missing in the evalu ation of PUFA treatments in chronic allergic disorders.
Study Subjects
This study was conducted in accordance with the ethical principles of the Declaration of Helsinki and obtained ethical clearance from the ethics committee at Policlinico Tor Vergata, University of Rome Tor Vergata (n.82.21). Informed consent was obtained from each patient or patient's legal guardians and each of the healthy controls. As reported in Figure 2, a total of 29 Italian native patients with peanut allergy (18 M and 11 F, age range 4.1-21.7 years) were identified at the Pediatric Immunopathology and Allergology Unit Policlinico Tor Vergata, University of Rome Tor Vergata. Peanut allergy was defined by the following criteria: (a) history of significant peanut-related clinical symptoms, (b) pos Alteration of the PUFA composition in membranes is basic information to acquire and translate into precision medicine and nutrition strategies. In fact, the type and dosage of fatty acids able to rebalance the molecular status of patients can be personalized for each distinct profile and regularly monitored. This is particularly important in children, in which PUFAs are used to sustain the exponential growth of cells for all tissues, and their exact levels must be determined [43] in order to promptly individuate deficiency or excess with impact on normal tissue functioning [44,45]. Our exploratory study highlights cell membranes for effecting PUFA detection exactly in the active site of their immunomodulatory effects, evidencing molecular mechanisms that are still missing in the evaluation of PUFA treatments in chronic allergic disorders.
Study Subjects
This study was conducted in accordance with the ethical principles of the Declaration of Helsinki and obtained ethical clearance from the ethics committee at Policlinico Tor Vergata, University of Rome Tor Vergata (n.82.21). Informed consent was obtained from each patient or patient's legal guardians and each of the healthy controls. As reported in Figure 2, a total of 29 Italian native patients with peanut allergy (18 M and 11 F, age range 4.1-21.7 years) were identified at the Pediatric Immunopathology and Allergology Unit, Policlinico Tor Vergata, University of Rome Tor Vergata. Peanut allergy was defined by the following criteria: (a) history of significant peanut-related clinical symptoms, (b) positive skin prick test to peanut allergen (wheal ≥ 3 mm larger than the saline control), and (c) positive in vitro serum peanut IgE (CAP-FEIA) > 0.1 Ku/L. Exclusion criteria were as follows: (a) BMI > 25 Kg/mq, (b) other concomitant food allergies, and (c) autoimmune diseases. A final number of 9 peanut-allergic patients (8 males), age range 4.1-21.7 years, were included in the study. These criteria allowed appropriateness of patient enrollment in the study. Further, blood samples for the lipidome profile of red blood cell membranes were obtained ≥18 months since last allergic reaction to standardize time of testing and reduce the risk of interfering factors. Component-resolved diagnostics (CRDs) for Arah2 and Arah9 were available in 4 patients. As healthy controls, 15 healthy individuals (8 females and 7 males) with BMI < 25 and age 17 ± 4 years old were selected from an anonymous database available in the Lipidomic Laboratory, for which informed consent was gathered at the moment of blood withdrawal, and respect of EU general data protection regulation 2016/679 (GDPR) was ensured by the ISO 17025 certification. All controls did not have any history of allergic reactivity. The benchmark of reference interval values for mature RBC membranes served for observing controls and patients using the data reported for populations [15][16][17].
Isolation of Fatty Acids from RBC Membrane Phospholipids and Gas Chromatographic Analysis
Fatty-acid-based membrane lipidome analyses were performed by the Lipidomic Laboratory of Lipinutragen (Bologna, Italy). Blood samples (0.5 mL) collected in vacutainer tubes with ethylenediaminetetraacetic acid (EDTA) were treated according to the ISO17025 certified procedure (accredited Lab. #1836L) by robotic equipment and processed as described in previous studies [21,22,36,46]. Briefly, the mature cell fraction was isolated based on the higher density of the aged cells with control of diameter controlled by cell counter (Scepter 2.0 with Scepter Software Pro, EMD Millipore, Darmstadt, Germany) [47]. After phospholipid extraction, derivatization to fatty acid methyl esters (FAME) was performed, transforming membrane glycerophospholipids (mainly phosphatidylcholine, phosphatidylethanolamine, phosphatidylserine, phosphatidyl inositol, and plasmalogens) to examine up to 80% of the RBC membrane lipidome [48]. Fatty acids analysis was performed by gas chromatography (GC), and percentages are given as % relative quantitative (% rel. quant.), as previously described [21,22,36,46], comparing with the benchmark of the interval values of each fatty acid and index [15].
Statistical Analysis
Statistics were performed using GraphPad Prism 8.0 software (GraphPad Software, Inc., San Diego, CA, USA), applying unpaired t-test and Spearman correlation with a 95% confidence interval. Informed Consent Statement: Informed consent was obtained from all subjects or from their parents in case of <18-year-old patients involved in the study. | 2022-12-24T16:17:29.692Z | 2022-12-21T00:00:00.000 | {
"year": 2022,
"sha1": "2bd9f97030a4ba268d07c058151dd6a1cdfb656c",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1422-0067/24/1/120/pdf?version=1671673463",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "542d4536d61e8c9bdeaebdc47c7473a4e6556e9d",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
220407288 | pes2o/s2orc | v3-fos-license | miR-138-5p inhibits the malignant progression of prostate cancer by targeting FOXC1
Background This study aimed to uncover the effect of miR-138-5p on the proliferation and metastasis of PCa cell lines, and further explore the potential regulatory mechanisms via regulating FOXC1. Methods 60 pairs cancer tissues and corresponding paracancerous ones from PCa patients were collected to assess the expression level of miR-138-5p by qRT-PCR. Subsequently, over-expression of miR-138-5p were established to explore the proliferation and metastasis of miR-138-5p in PCa cell lines was analyzed by CCK-8, Transwell assay and Wounding healing assay, respectively. Bioinformatics analysis and luciferase reporter gene assay were performed to search for the target genes of miR-138-5p, and FOXC1 was selected. Finally, the biological role of miR-138-5p and FOXC1 in the progression of PCa was clarified by a series of rescue experiments. Results The results of qRT-PCR revealed that miR-138-5p was lowly expressed in PCa tissues and cell lines. Besides, the PCa patients with low-miR-138-5p had a high Gleason score, lymph node metastasis and poor prognosis of PCa, compared with these patients with high-miR-138-5p. Over-expression of miR-138-5p inhibited the proliferative, migratory and invasive capacities of PC-3 and DU-145 cells. Bioinformatics analysis and luciferase reporter gene assay suggested that FOXC1 was predicted to be the target gene of miR-138-5p. Moreover, FOXC1 expression level was negatively correlated to that of miR-138-5p in PCa tissues. Importantly, over-expression of FOXC1 could reverse miR-138-5p mimic induced-inhibition of PCa malignant progression. Conclusions Downregulated miR-138-5p was closely associated with high Gleason score, more lymph node metastasis and poor prognosis of PCa patients. In addition, miR-138-5p alleviated the malignant progression of PCa by targeting and downregulating FOXC1.
Background
Prostate cancer (PCa) is one of the most common malignancies all over the world, which is the leading cause of cancer-related deaths in the United States and European countries [1][2][3]. In recent years, the incidence and mortality rate of PCa in China has rapidly increased year by year, which has become a serious threat to human healthy [4,5]. When prostate specific antigen (PSA) test was primarily used to screen for PCa before symptoms appear, the detection rate of PCa peaked in the early 1990s [6,7]. So far, approximately 85% new diagnosed PCa cases were limited to early-stage cancer [8]. Although PSA test greatly improved the early-stage diagnostic rate of PCa, its benefit in decreasing the mortality of PCa remained controversial [9,10]. Like other malignancies, the malignant progress of PCa is a multi-step and multi-stage process, including inactivation of anti-oncogenes and/or activation of proto-oncogenes [11,12]. Currently, the target therapy based on cancer-related miRNAs presented a promising application, and the results of these studies showed that miRNA had a good application prospect in the diagnosis, treatment, prognosis and other aspects of cancer, to provide new ideas for the pathogenesis of PCa [13,14].
MiRNAs are small, endogenous non-coding RNAs that negatively regulate expressions of protein-coding genes at translational level [15,16]. MiRNAs exert the biological function by degrading or inhibiting translation of mRNAs [16]. It is reported that miRNAs are extensively involved in affecting cellular behaviors and progression of disease [17,18]. Accumulating evidences have demonstrated the effects of abnormally expressed miRNAs on the malignant progression of tumors [17,19]. These certain miRNAs might be utilized for developing anticancer drugs or biological hallmarks [20]. About 30% human genome could be regulated by miRNAs, and most of human miRNAs locate on cancer-associated genomic regions or gene fragile sites [21,22]. MiR-138-5p is a newly discovered cancer-related miRNA, which has been confirmed to be down-regulated in pancreatic cancer, colorectal cancer and other malignant tumors [21,22]. However, its expression and function of PCa are rarely reported.
Bioinformatics analysis has been widely applied for analyzing the target gene of miR-138-5p to further uncover the expression pattern of genes. By analyzing miR-138-5p profiling microarray, Forkhead box C1 (FOXC1) was selected to be the target gene of miR-138-5p. FOXC1 is an essential member of the forkhead box transcription factors and has been highlighted as an important transcriptional regulator of crucial proteins associated with a wide variety of carcinomas [23,24]. Based on the above series of researches, this study aimed to elaborate the possible roles of miR-138-5p and FOXC1 in the progression of PCa, as well as the association with its clinical characteristics and prognosis of PCa patients, so as to bring a new idea for clinical diagnosis and treatment of PCa.
Patients and PCa tissue samples
A total of 60 PCa patients undergoing radical prostatectomy were enrolled in this study from Department of Urology Surgery, Chifengshi hospital. PCa cancer tissues and corresponding paracancerous ones (5 cm away from tumor edge) were surgically resected and preserved within 5 min ex vivo. The clinical and pathological characteristics of PCa patients were collected for further analyses. The clinical stages of PCa were graded in accordance with International Federation of PCa staging criteria. In addition, the longest diameter of the tumor is measured and used to assess the tumor size. All patients in this study had been fully signed the informed consent. In addition, this study has been approved by the Ethics Committee of Chifengshi hospital.
Transfection
Negative control (NC mimic) and miR-138-5p overexpression sequence vectors (miR-138-5p mimic) were purchased from GenePharma (GenePharma, Shanghai, China). After PCa cell lines were plated in 6-well plates and grown to a cell density of 40-60%, transfection was performed using Lipofectamine 3000 (Invitrogen, CA, USA) according to the manufacturer's instructions. These cells were collected for verification of transfection efficacy and subsequent experiments after 48 h.
CCK-8 assay
After 48 h of transfection, the cells were seeded in the 96-well plates with 2 × 10 3 cells per well. After cultured for 24 h, 48 h, 72 h and 96 h, these cells were added with CCK-8 kit (Dojindo Laboratories, Japan). After incubation for 2 h, the fluorescent absorbance at the optical density (OD) value of 450 nm of each sample was recorded for plotting the proliferative curves in the microscope.
Transwell assay
After 48 h of transfection, the cells were adjusted to a dose of 2.0 × 10 5 /mL. 200 μL suspension was applied in the upper side of Transwell chamber (Millipore, MA, USA) inserted in a 24-well plate. In the bottom side, 700 μL of medium containing 10% FBS was applied. After 48 h of incubation, these cells penetrated to the bottom side were fixed in 4% paraformaldehyde for 15 min, dyed with crystal violet for 20 min and counted using the microscope. The number of migratory cells was counted in five randomly selected fields per sample.
Wound healing assay
After 48 h of transfection, these cells were inoculated in 6-well plates and grown to 90% confluence. After the creation of an artificial wound in cell monolayer, the medium with 1% FBS was replaced. After 24 h, the wound closure was captured by the microscope.
Quantitative real-time PCR (qRT-PCR)
Total RNA was extracted from PCa cell lines and tissues using TRIzol reagent (Invitrogen, Carlsbad, CA, USA), purified by DNase I treatment, and reversely transcribed into cDNA by Primescript RT Reagent (Takara, Otsu, Japan). The obtained cDNA was subjected to qRT-PCR by SYBR ® Premix Ex Taq ™ (Takara, Japan). Data were normalized to GAPDH and U6. The following primers for qRT-PCR were used as followed: miR
Western Blotting
The transfected PCa cell lines were lysed using PRO-PREPTM lysis buffer, shaken on ice for 30 min, and centrifuged at 14,000×g for 15 min at 4 °C. Total protein concentration was calculated by the BCA Protein Assay Kit (Pierce, Rockford, Il, USA). Rabbit anti-human monoclonal antibody against FOXC1 was purchased from Santa Cruz, USA; horseradish peroxidase-labeled goat anti-rabbit secondary antibody was purchased from Genscript. Data were normalized to GAPDH. Protein samples were separated by SDS-PAGE, transferred to PVDF membrane, and blocked with 5% skim milk powder for 1 h at room temperature. Primary antibody was added for incubation overnight at 4 °C shaker. In the next day, the membrane was rinsed 3 times with TBST and incubated with second antibody for 1 h at room temperature. After that, the protein samples on the membrane were finally semi-quantitatively analyzed by alpha SP image analysis software.
Dual-luciferase reporter assay
3′-UTR of wild-type (WT) human FOXC1, which contains a putative miR-138-5p binding DNA sequence, was amplified by PCR and inserted into a p-miR-reporter (Ambion, USA) to create a firefly FOXC1-WT luciferase vector. The mutant (MUT) 3′-UTR was also inserted into p-miR-reporter to create a firefly FOXC1-MUT luciferase vector. Human HEK293T cells were transduced with NC mimic or miR-138-5p mimic, then cross-transfected with FOXC1-WT or FOXC1-MUT for 48 h. After that, the relative luciferase activities were measured using a Dual-luciferase reporter assay (Promega, USA) according to the manufacturer's protocol.
In vivo xenograft vectors
The Animal Ethics and Use Committee of Chifengshi hospital approved the cancer-forming experiment in nude mice. 8-week-old male nude mice were purchased from the animal center and randomly divided into two groups (5 in each group). The PC-3 cells with miR-138-5p mimic were injected subcutaneously into the axilla of mice. Tumor size was monitored every 7 days; Then, after 6 weeks, the mice were sacrificed. The tumor volumes were calculated using the following formula: tumor volume = (width 2 × length)/2.
Statistically analysis
GraphPad Prism 6 V6.01 was used for data analyses. Data were expressed as mean ± standard deviation, and p < 0.05 was considered as statistically significant. Intergroup differences were analyzed by the t-test. Kaplan-Meier curves were introduced for survival analysis.
Chi-square test was performed to evaluate the correlation between miR-138-5p level and the pathological indexes of PCa patients.
miR-138-5p was down-regulated in PCa tissues and cell lines
Data from PCa patients of TCGA were complied for investigating the potential relevant miRNAs associated with the progression of PCa. We firstly focused insight into the expression level of miRNAs form TCGA database, and miR-138-5p was finally selected and was significant statistical difference in PCa tissues (Fig. 1a). qRT-PCR was performed to evaluate the expression of miR-138-5p in PCa tissues and cell lines. As showed in Fig. 1b, miR-138-5p was down-regulated in PCa tissues, compared with paracancerous tissues. Similarly, miR-138-5p was also down-regulated in PCa cell lines, compared with that of Prostate epithelial cell line (RWPE-1) (Fig. 1e).
miR-138-5p expression was correlated with clinicopathologic characteristics and overall survival in PCa patients
The clinicopathology characteristics and follow-up data of enrolled PCa patients were collected for further analyses. According to the median level of miR-138-5p, PCa patients were assigned into two groups, including high-miR-138-5p level group and low-miR-138-5p level group. As shown in Table 1, the associations between the expression level of miR-138-5p and age, tumor size, Gleason score, lymph node metastasis and bone metastasis of PCa patients were analyzed. The results found that these PCa patients with low-miR-138-5p expression had a higher Gleason score and more lymph node metastasis in PCa patients, compared with these patients with high-miR-138-5p expression (Fig. 1c). In addition, Kaplan-Meier methods revealed the poor prognosis in PCa patients of low-miR-138-5p level group than that of high-miR-138-5p level group (Fig. 1d).
miR-138-5p alleviated the proliferation and metastasis of PCa cell lines
To explore the biological function of miR-138-5p in PCa cell lines, the proliferation and metastasis of PCa was analyzed by CCK-8 assay, Transwell assay and Wounding healing assay, respectively. miR-138-5p mimic and NC mimic were successfully constructed in PC-3 and DU-145 cell lines, respectively (Fig. 2a). It was found by the CCK-8 assay that the cell proliferation ability of miR-138-5p mimic was significantly decreased in PCa cell lines, compared with that of NC mimic (Fig. 2b). Transwell assay revealed that the metastasis ability of PCa cells was significantly decreased in miR-138-5p mimic, compared with NC mimic (Fig. 2c). In addition, Wound healing assay also showed that the overexpression of miR-138-5p could hinder the invasion and crawling ability of PCa cell lines (Fig. 2d). These results suggested that miR-138-5p could inhibited the proliferation and metastasis in PCa cell lines.
Interaction of miR-138-5p and FOXC1
Potential target genes of miR-138-5p were predicted in the miRDB, TargetScan and StarBase (Fig. 3a). At last, the intersection contained five potential targets (FOXC1, SYT13, SIN3A, FOXP4 and KLF11). Among them, FOXC1 was one of the most differentially expressed gene after the transfection of miR-138-5p mimic in PC-3 cells (Fig. 3b). Western Blotting showed that the expression level of FOXC1 was significantly down-regulated in PC-3 and DU-145 cells in miR-138-5p mimic, compared to NC mimic (Fig. 3c). In addition, the results of qRT-PCR also revealed the same trend (Fig. 3d). To further uncover the biological role of FOXC1 in PCa cell lines, we constructed FOXC1 overexpressing plasmid (pcDNA3.1-FOXC1) and empty overexpressing plasmid NC (pcDNA3.1-NC). qRT-PCR found that miR-138-5p level was found to be significantly down-regulated in PCa cell lines transfected with pcDNA-FOXC1, compared to that transfected with pcDNA-NC (Fig. 3e). Luciferase reporter assay verified that miR-138-5p could indeed combine with FOXC1 through specific sequences (Fig. 3f, g). Additionally, a significant negative correlation was identified to detect the expression levels of miR-138-5p and FOXC1 in PCa tissues (Fig. 3h).
miR-138-5p negatively regulated FOXC1 to inhibit the malignant progression of PCa
To further explore the regulatory mechanisms in which miR-138-5p exactly regulated FOXC1 to inhibit malignant progression of PCa. Firstly, the overexpressed endogenous FOXC1 was established with pcDNA3.1-FOXC1, and pcDNA3.1-NC to transfect PCa cell lines with miR-138-5p mimic. qRT-PCR demonstrated that PC-3 and DU-145 cells transfected with pcDNA3.1-FOXC1 had significantly lower miR-138-5p expression level than these cells transfected with pcDNA3.1-NC (Fig. 4a). In addition, Western Blotting also demonstrated that PC-3 and DU-145 cells transfected with pcDNA3.1-FOXC1 had significantly higher FOXC1 expression level than these cells transfected with pcDNA3.1-NC (Fig. 4b). Subsequently, overexpression of FOXC1 was demonstrated to be able to counteract the carcinostasis of miR-138-5p mimic on the occurrence of PCa by Transwell assay and Wound healing assay (Fig. 4c, d). Therefore, these results revealed that miR-138-5p could inhibit the malignant progression of PCa through modulating FOXC1.
Over-expression of miR-138-5p suppressed the PCa in vivo tumorigenicity
In an in vivo tumorigenicity assay, NC mimic or miR-138-5p mimic transduced PC-3 cells were subcutaneously inoculated into the abdominal compartments of athymic nu/nu mice for 6 weeks. The volumes of PC-3 xenografts were calculated weekly. It showed that, in vivo tumor growth was significantly suppressed by miR-138-5p mimic, compared to NC mimic (p < 0.05; Fig. 5a, b). Subsequently, we validated the reduction of weight in the tissues of nude mice injected with miR-138-5p mimic (p < 0.05; Fig. 5c). The results of qRT-PCR revealed that miR-138-5p mimic in the tissues of nude mice could decrease the expression level of miR-138-5p (Fig. 5d). In addition, compared with NC mimic, FOXC1 expression significantly decreased in the tissues of nude mice with miR-138-5p mimic by Western Blotting (Fig. 5e). Immunohistochemistry showed that FOXC1 expression level of miR-138-5p mimic-transduced PC-3 xenografts significantly decreased than NC mimic-transduced xenografts (Fig. 5f ).
Discussion
The etiology of PCa is complex, and no clear results about the pathogenesis of PCa is found [1,7,9]. Nowadays, the understanding of the pathogenesis and biological behavior of PCa still has great limitations [8,9]. The development of PCa is a multi-factors process, influenced by a variety of biomolecules and regulated by signaling pathways [11,12]. With the progress of the Human Genome Project, most researches on molecular and cellular pathobiology have been extensively conducted to detect the differential expression profile of tumor genes, which is of great significance for exploring the molecular mechanism of PCa, and finding the molecular biomarkers in the early diagnosis and prognosis of PCa [13,14].
MiRNAs not only participate in normal physical biological processes, but also regulate cancer progression at transcriptional and post-transcriptional levels [15,16]. It is estimated that over 30% of genes in cellular processes are regulated or controlled by miRNAs [17,20]. In the past years, tumor-promotor or tumor-suppressor miR-NAs were used to be as drug targets for the treatment of PCa [15,18,20]. Nowadays, cancer-related miRNAs have been well concerned, exerting more crucial application in early-stage diagnosis, treatment and prognosis of cancer [13,14,19]. As a member of cancer-associated miRNAs family, miR-138-5p is located on chromosome Xq38.13 [25]. Previous researches showed that miR-138-5p could inhibit the malignant progression of some human cancer, such as pancreatic cancer and colorectal cancer [25,26]. In this study, miR-138-5p was down-regulated in PCa tissues and cell lines. Besides, these PCa patients with low-miR-138-5p expression had a high Gleason score, more lymph node metastasis, and poor prognosis of PCa patients, compared with these patients with high-miR-138-5p expression. Thus, the above results suggested that miR-138-5p might act as anti-cancer effect in the proliferation and metastasis of PCa. In order to further investigate the biological function of miR-138-5p in PCa cell lines, CCK-8, Transwell assay and Wounding healing assay were used to introduce that miR-138-5p mimic could inhibit the proliferation and metastasis of PCa. The above results provided a theoretical basis for revealing the mechanism of the development of PCa. Of course, the specific molecular mechanism of signal transduction in PCa need to further study in the subsequent studies.
Regulatory mechanism of miRNAs depends on the expression and function of the related target genes [15,20]. A miRNA degrades target mRNA or suppress its translation by base pairing with 3′UTR of the mRNA [15,16]. The degree of base pairing decides the degradative or translation inhibitory effect of a miRNA, that is, complete base pairing leads to mRNA degradation; otherwise, translation inhibition is achieved [27,28]. MiRNAs only account for only 1% of the whole human genome, but are able to regulate more than 30% proteinencoding genes [29]. Multiple miRNAs could precisely regulate a single target gene [16,22]. Bioinformatics analysis and luciferase reporter gene assay showed that FOXC1 was the target gene of miR-138-5p through predicting in the miRDB, TargetScan and StarBase. FOXC1 expression level was up-regulated in PCa tissue and cell lines. Based on these findings, we hypothesized that miR-138-5p might act as a ceRNA in the malignant progression of PCa and found that miR-138-5p negatively regulated FOXC1 expression in PCa tissues. In addition, qRT-PCR showed that compared with NC mimic, miR-138-5p mimic could decreased the expression level of FOXC1. In order to explore the associations between miR-138-5p and FOXC1 in the development of PCa, the (See figure on next page.) Fig. 3 Interaction of miR-138-5p and FOXC1. a Bioinformatics analysis websites (miRDB, TargetScan and StarBase) showed the potential downstream target gene of miR-138-5p; b qRT-PCR was used to detect the differential expression of the potential downstream target gene of miR-138-5p in NC mimic and miR-138-5p mimic, respectively; c Western Blotting verified the expression level of FOXC1 after transfection of NC mimic and miR-138-5p mimic PCa cell lines, respectively; d qRT-PCR verified the expression level of FOXC1 after transfection of NC mimic and miR-138-5p mimic PCa cell lines, respectively; e qRT-PCR verified the expression level of miR-138-5p after transfection of pcDNA3.1-NC and pcDNA3.1-FOXC1 in PC-3 and DU-145 cell lines, respectively; f, g Dual luciferase reporter assays demonstrated direct targeting of miR-138-5p to FOXC1; h A significant negative correlation between miR-138-5p and FOXC1 expression in PCa tissues. Data are mean ± SD, *p < 0.05, **p < 0.01 Zhang et al. Cancer Cell Int (2020) 20:297
Conclusions
In summary, the down-regulated miR-138-5p was closely associated with high Gleason score, more distant metastasis and poor prognosis of PCa patients. In addition, Tumor volume growth curves were calculated for different nude mice after injection of NC mimic and miR-138-5p mimic, respectively; c Tumor weight growth curves were calculated after injection of NC mimic and miR-138-5p mimic, respectively; d qRT-PCR was used to detect the expression level of miR-138-5p in the tumor-forming tissues of nude mice; e Western Blotting was used to detect the expression level of FZD4 in the tumor-forming tissues of nude mice; f Immunohistochemistry was used to detect the expression level of FZD4 in the tumor-forming tissues of nude mice with Hep3B cell line (magnification: 40×). Data are mean ± SD, *p < 0.05, **p < 0.01 | 2020-07-09T15:04:28.352Z | 2020-07-09T00:00:00.000 | {
"year": 2020,
"sha1": "31cb420d14f0c27a232db7556a6c0e68fb086ee9",
"oa_license": "CCBY",
"oa_url": "https://cancerci.biomedcentral.com/track/pdf/10.1186/s12935-020-01386-6",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7bf4bbfba8c4d65e89299d0ff003d0a4dd95bede",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine",
"Biology"
]
} |
257557669 | pes2o/s2orc | v3-fos-license | LCS-TF: Multi-Agent Deep Reinforcement Learning-Based Intelligent Lane-Change System for Improving Traffic Flow
Discretionary lane-change is one of the critical challenges for autonomous vehicle (AV) design due to its significant impact on traffic efficiency. Existing intelligent lane-change solutions have primarily focused on optimizing the performance of the ego-vehicle, thereby suffering from limited generalization performance. Recent research has seen an increased interest in multi-agent reinforcement learning (MARL)-based approaches to address the limitation of the ego vehicle-based solutions through close coordination of multiple agents. Although MARL-based approaches have shown promising results, the potential impact of lane-change decisions on the overall traffic flow of a road segment has not been fully considered. In this paper, we present a novel hybrid MARL-based intelligent lane-change system for AVs designed to jointly optimize the local performance for the ego vehicle, along with the global performance focused on the overall traffic flow of a given road segment. With a careful review of the relevant transportation literature, a novel state space is designed to integrate both the critical local traffic information pertaining to the surrounding vehicles of the ego vehicle, as well as the global traffic information obtained from a road-side unit (RSU) responsible for managing a road segment. We create a reward function to ensure that the agents make effective lane-change decisions by considering the performance of the ego vehicle and the overall improvement of traffic flow. A multi-agent deep Q-network (DQN) algorithm is designed to determine the optimal policy for each agent to effectively cooperate in performing lane-change maneuvers. LCS-TF's performance was evaluated through extensive simulations in comparison with state-of-the-art MARL models. In all aspects of traffic efficiency, driving safety, and driver comfort, the results indicate that LCS-TF exhibits superior performance.
I. INTRODUCTION
Autonomous driving has garnered significant attention from both the research community and industry, owing to its high potential for reducing traffic congestion and enhancing driving safety [1], [2]. Discretionary lane change for autonomous vehicles (AVs) is a key research focus in autonomous driving due to its substantial impact on both traffic efficiency and driving safety [3], [4], [5]. Especially in dense or congested traffic environments, relying solely on the longitudinal motion control of AVs is insufficient for enhancing traffic flow. Timely and agile lane changes are critical to improve traffic efficiency [6]. However, making effective lane-change decisions for AVs can be particularly challenging, especially in complex and dynamic traffic environments where AVs and human-driven vehicles (HVs) coexist [7].
A substantial body of research has been conducted on lanechanging systems for AVs. The existing research works on lane-changing systems for AVs can be broadly classified into rule-based and machine learning-based approaches. Rulebased approaches rely on hand-crafted rules designed to emulate the behaviors of human drivers [8], [9]. However, despite their usefulness for specific scenarios, rule-based approaches suffer from limited generalization performance [10]. In response to the limitations of rule-based approaches, machine learning-based approaches have been the subject of active research especially concentrating on reinforcement learning (RL)-based methods [11], [7], [12].
We classify the RL-based solutions into two main categories: the ego vehicle-based approaches [13], [14], [10], [15] and multi agent-based approaches [16], [17], [2], [18], [19]. Ego vehicle-based approaches aim to enhance the efficiency and safety of the ego vehicle by incorporating surrounding vehicles as part of the state of the RL model. While ego vehicle-based approaches are highly scalable, they may result in poor performance since the complex interaction between vehicles is not effectively accounted for. In contrast, the MARL-based solutions exploit the close coordination of multiple agents to reach better lane-changing decisions for multiple autonomous vehicles, leading to higher traffic efficiency and safety. However, the majority of existing MARL-based solutions rely on onboard sensor data to make lane-changing decisions, thereby restricting the perception capability of the agents to the surrounding area [19], [1]. More importantly, the potential of intelligent lane-changing systems for improving the overall traffic flow of a road segment has not been fully leveraged.
In this paper, we propose a novel MARL-based intelligent lane-change system for AVs, namely LCS-TF, that simultaneously optimizes the efficiency, safety, and driving comfort of the ego vehicle, along with the traffic flow of the road segment to which the agents belong. To the best of our knowledge, this is the first study to present a hybrid intelligent lane-change approach that maximizes both local and global performance through the incorporation of a roadside unit (RSU) that manages a road segment, and vehicle-toeverything (V2X) capabilities for the agents. Specifically, the proposed intelligent lane-change problem is formulated as a decentralized partially observable Markov decision process. We then develop a MARL framework to solve the problem. Our state space design incorporates both local traffic infor-mation obtained from surrounding vehicles of an ego vehicle, as well as global traffic information collected from the RSU responsible for managing the corresponding road segment. A novel reward function is created to effectively carry out the joint optimization for the performance of the ego vehicle and the overall improvement of traffic flow. A multi-agent DQN algorithm is designed to identify the optimal policy that each agent can use to collaboratively execute lanechange maneuvers. Extensive simulations were conducted to evaluate the performance of LCS-TF in comparison with two main types of existing MARL models [19], [2]. The results demonstrate the superior performance of LCS-TF in all aspects of traffic efficiency, driving safety, and driver comfort.
This paper is organized as follows. We present a review of the latest efforts for intelligent lane-change solutions for AVs in Section II. We then present the details of the proposed intelligent lane-change solution in Section III. In Section IV, we present the simulation results. We then conclude in Section V.
II. RELATED WORK
This section presents a review of machine learning-based intelligent lane-changing systems for autonomous vehicles. Most of the existing works are concentrated on improving the safety and efficiency of the ego vehicle [13], [14], [10], [15]. These ego vehicle-based approaches aim to train the lane-changing behavior of the ego vehicle in a singe-AV setting where the surrounding vehicles are considered as the environment for decision making [11], [7], [12]. Wang et al. proposed an RL framework to train the ego vehicle for intelligent lane-changing in diverse traffic environments [13]. In particular, a Q-function based on a closed form greedy policy and a deep Q-learning algorithm based on continuous state and action space were designed to develop the RL framework. Ye et al.'s work was focused on improving the learning efficiency [14]. They created a lanechanging strategy based on proximal policy optimizationbased deep RL and demonstrated that stable performance could be achieved with improved learning efficiency. Zhang et al. took into account human factors for training the ego vehicle [10]. The driving styles of both the ego vehicle and surrounding vehicles were incorporated into their decision making model. He et al. concentrated on improving the robustness of their machine learning model for intelligent lane-changing systems [15]. They proposed an observation adversarial RL approach with an aim to manage observation uncertainties caused by sensor noises, measurement errors, and adversarial perturbations. Although these approaches focused on improving the performance for the ego vehicle can achieve high scalability, they do not perform well in complex traffic environments that require coordination of multiple AVs [20].
To address the performance limitation of the ego vehicle-based approaches, multi-agent reinforcement learning (MARL)-based solutions were proposed [16], [17], [2], [18], [19]. The key focus of the MARL-based approaches is to exploit the close collaboration among multiple agents to make more efficient lane-changing decisions to improve traffic efficiency and safety.
Hou and Graf proposed a multi-agent deep RL approach that is designed to control both the lane-changing behavior and longitudinal movement of the ego vehicle [18]. The authors demonstrated that the dual control approach improved the roadway capacity, especially for highway merging and weaving areas. However, the traffic efficiency for their reward function is defined based only on the speed of the ego vehicle without considering the overall impact of a lane-changing decision on the surrounding traffic.
Zhou et al. created a MARL-based approach using the multi-agent advantage actor-critic network (MA2C) and a novel parameter-sharing scheme to facilitate collaboration among agents [2]. Noting that the reward functions of existing MARL-based approaches do not fully consider the passenger's comfort, the authors focused on designing a MARL-based approach that simultaneously improves traffic efficiency, safety, and driving comfort. However, similar to [18], their reward function does not account for the overall impact of lane-changing behavior on surrounding traffic. Furthermore, the representation of the driving comfort in their reward function is simplistic, i.e., determined based only on a specific threshold.
Wang et al. designed a MARL-based lane-changing solution without relying on vehicle-to-everything (V2X) communication [1]. The proposed approach is targeted to optimize both the individual and overall efficiency based only on the limited sensing results of individual vehicles. However, since no V2X is adopted, a vehicle can only have a narrow view, which makes it difficult, if not possible, to make a decision to optimize the overall traffic flow of a road segment while ensuring traffic safety and driving comfort.
Zhang et al. proposed a bi-level approach where the upper level is the MARL model for making lane-change decisions, and the lower level manages potential conflicts during the implementation of the lane-change decision [19]. The proposed approach improves efficiency by encoding the driving intentions of surrounding vehicles into the state space. Their reward function considers both the efficiency of the ego vehicle and the impact on the overall traffic efficiency. However, this paper is based on a critical assumption that there does not exist vehicular communication and uses only the onboard sensors, therefore limiting the perception capability of the ego vehicle only to its surrounding area.
Unlike recent MARL-based approaches that overlook the impact of lane-change decisions on traffic flow in a road segment, LCS-TF aims to utilize the collaborative lanechange decisions of multiple agents to enhance the overall traffic flow while simultaneously optimizing the individual efficiency, safety, and driving comfort of the agents.
III. MULTI-AGENT DEEP REINFORCEMENT LEARNING FOR INTELLIGENT LANE-CHANGING
A. Overview Fig. 1 illustrates an operational overview of LCS-TF. LCF-TF is deployed to cover a roadway segment with mixed traffic of AVs and HVs. We assume that AVs are equipped with the V2X capability. We also assume that there exists a roadside unit (RSU) that is assigned to monitor and manage traffic on a particular roadway segment. More specifically, the RSU gathers traffic data from the roadway segment and disseminates the aggregated information to the agents within the same segment.
LCS-TF allows the RL agents to make lane-change decisions based on both the local and global traffic information, aiming to optimize the traffic flow of the roadway segment as well as the efficiency, safety, and driving comfort of the ego vehicle. More specifically, the local information is obtained from the surrounding vehicles using the ego vehicle's onboard sensors. The global information is directly received from the RSU via V2X. More details on the different kinds of local and global traffic information used by the agents to make lane-change decisions, as well as the corresponding machine learning algorithm are discussed in Section III-C.
B. Problem Formulation
The problem of making a lane-change decision for each RL agent to jointly optimize the traffic flow and the performance of the ego vehicle can be represented as a decentralized partially observable Markov decision process (decPOMDP). The decPOMDP is described by the tuple T is a set of transitional probabilities which is unknown. The goal is to find a decentralized policy π i : O i × S → [0, 1] for each agent i to choose an action a t at time t such that the expected cumulative reward is maximized. LCS-TF solves the problem based on a MARL framework for which the details are presented in subsequent sections.
C. State Space
The state space O i of agent i consists of both local and global traffic information. The local traffic information refers to the traffic state of the ego vehicle and its surrounding vehicles. Specifically, vehicles that fall within a range of 100m from the ego vehicle are referred to as surrounding vehicles. Therefore, the traffic state for surrounding vehicles is given as a 2D vector N × F , where N is the number of surrounding vehicles, and F is the feature set. The feature set F of a surrounding vehicle consists of the longitudinal/lateral position, longitudinal/lateral speed, acceleration, and driver imperfection of the vehicle. In particular, driver imperfection is incorporated to take into account the driving behavior of HVs with an aim for the agents to make more efficient lane-change decisions. The traffic state for the ego vehicle includes the ego vehicle's longitudinal/lateral position, longitudinal/lateral speed, and acceleration.
The state space O i of agent i also integrates global traffic information, enabling the agent to make lane-change decisions while considering the effect on traffic flow for the present road segment. The global traffic information includes the vehicle density, average vehicle speed, maximum allowed speed, and the number of lanes of the roadway segment. It also includes the mean speed and vehicle density for each lane. Additionally, the key road-specific parameters used by the agent to make lane-change decisions, including the lateral safety distance, longitudinal safety distance, and vehicle action update interval, are also included in the global traffic information. The parameter values used in this study are presented in the description of the proposed reward function.
D. Action Space
The action space A i of agent i consists of five different discrete lane-change and/or speed adjustment decisions as shown below.
More specifically, an agent i can undertake the following actions, namely, (i) switching to the left lane, (ii) switching to the right lane, (iii) maintaining the current speed, (iv) accelerating, and (v) decelerating. In this study, the action update interval was set to 0.1s.
E. Reward Function
The reward function R i for agent i is comprised of four subfunctions, which represent the traffic efficiency r e , safety r s , and driving comfort r c , and lane change utility r u for the ego vehicle, respectively, i.e., where w 1 , w 2 , w 3 and w 4 are weighting parameters.
Efficiency Reward: The efficiency reward function r e not only incentivizes RL agents to improve their own speed but also promotes the flow of traffic within the road segment. More specifically, the efficiency reward function r e is defined as follows.
where g e denotes the global efficiency, and l e denotes the local efficiency. Global efficiency represents the effect of the ego vehicle's action on the traffic flow for the road segment. More specifically, the global efficiency g e is defined as follows.
where v e is the average speed of vehicles in the road segment. v min and v max are the minimum average speed and maximum speed limit defined for the road segment, respectively. A positive reward is granted when the ego vehicle's lane-change decision leads to an improvement in v e , provided that it lies between v min and v max . Otherwise, a negative reward is issued. When computing the reward, we otherwise action a i t = max a Q * (s i t , a; θ i ) 10: apply action a i t , observe reward r i t and next observation s i end for 13: sample random minibatch of transitions (s k , a k , r k , s k+1 ) from B 14: compute y k = r k if episode ends r k + γmax a ′ (s k+1 , a ′ ; θ) otherwise 15: compute loss (y k − Q(s j , a; θ)) 2 16: adjust online network weights using eq (1) 17: update target network weights at every t ′ 18: end for 19: end for ensure that it is proportional to the difference between v e and v min (or v max ).
The local efficiency l e is determined based on the speed of the ego vehicle in comparison with v min and v max , i.e., where v ego is the speed of the ego vehicle. The way how the local efficiency reward is computed is similar to that of the global efficiency reward. Safety Reward: The safety reward function r s is designed to promote the safety of both the ego vehicle and the surrounding vehicles. More specifically, the lateral and longitudinal safety of the ego vehicle and the potential collision with the surrounding vehicles are incorporated in the safety reward function, i.e., r s = l lon + l lat + l col , where the longitudinal safety component l lon of the reward function ensures that the agent maintains the minimum safety distance while executing lane changes, i.e., where d long is the longitudinal distance, and T minGap is the longitudinal safety threshold. The lateral safety component l lat ensures that a safe lateral gap is maintained while executing lane changes, i.e., where d lat is the lateral distance, and T lat is the lateral safety threshold.
The collision component l col imposes a significant penalty on the agent in the event of a collision with surrounding vehicles, i.e., l col = −5, if collide with other vehicle 0, otherwise Driving Comfort Reward: We employ an existing methodology to quantify driving comfort [21], [22], [23]. Specifically, the driving comfort reward function is designed based on the change rate of acceleration as follows.
where jerk is calculated based on the acceleration range. For example, in our simulation, an agent makes a decision every 0.1s, and the acceleration range is between 2.6 m/s 2 to -2.6 m/s 2 . Hence, the greatest possible jerk jerk max is (2.6-(-2.6))/0.1=52.
Lane Change Utility Reward: Although the reward function R i is designed to maximize traffic efficiency, safety, and comfort, we observe that agents at times make invalid lane-change decisions, especially during the initial period of training. To facilitate the training process and achieve better performance, we incorporate the lane change utility
F. Multi-Agent DQN Algorithm
We propose a multi-agent DQN algorithm designed to identify the optimal policy that facilitates the efficient coordination of multiple agents in maximizing the cumulative reward, based on their partial observations of the environment. We adopt the centralized training and decentralized execution paradigm where each RL agent shares the same network architecture and parameters [24], [25]. In this shared learning architecture, each RL agent senses surrounding traffic conditions as observations and performs optimal longitudinal or lateral actions based on the current observations. Furthermore, each agent stores its experiences in shared memory, which contributes to updating policy network weights by minimizing the loss function at each gradient step based on Eq. III-F; as a result, other agents benefit from performing optimal decisions using updated policy networks.
where Q(s, a; θ i )(s ∈ S, a ∈ A) denotes the Q-network with weights θ i for agent i, which is used to estimate the action-value function. θ i ′ is the parameters in the previous iteration for agent i. ρ(s, a) is a probability distribution over observations s and actions a. Moreover, each agent stores its experiences in the shared memory so that other agents can benefit. The algorithm 1 outlines the pseudocode for our multi-agent lane change decision-making strategy.
At every simulation step, all agents perform actions simultaneously following the ǫ−greedy policy [26] and store their transitions in the shared replay buffer. We trained the policy network by taking a mini-batch of uniform random samples of transitions from the replay buffer at every simulation second to optimize the network parameters so that agents can make better decisions.
IV. SIMULATION RESULTS
In this section, we present the simulation results. Section IV-A introduces the simulation settings. Section IV-B discusses the details of training the model. We then analyze the traffic efficiency, traffic safety, and driver comfort of the proposed solution in Sections IV-C, IV-D, and IV-E, respectively.
A. Simulation Settings
We used a traffic simulator called SUMO [27] to implement LCS-TF and two main types of approaches, namely NP-MARL, i.e., a MARL-based approach with no V2X support [19], and LSC-MARL which uses V2X but has limited functionalities for improving traffic safety and driving comfort [2]. The MARL models were created based on Keras and Tensorflow [28] and were interfaced with SUMO via Traffic Control Interface (TraCI) [29]. A workstation equipped with the Intel Xeon Gold 5222 Processor, NVIDIA® RTX™ A4000, and 48GB RAM running on Windows 11 OS was used to train and test the MARL models.
We consider a 3.5km-long, 5-lane road segment. The initial 500m of the road segment was allocated for vehicle injection and acceleration for both AVs and HVs. The following 3km of the road segment was where RL agents perform lane-changes. The speed limit for the road segment was set to 75mph. Vehicles were randomly injected at a rate of 2,160 veh/h for which the proportion of AVs was varied for evaluation purposes. The initial speed of the vehicles was 45mph, and the range of acceleration for each vehicle was set to -2.6m/s 2 to 2.6m/s 2 . The intelligent driver model (IDM) [30], [27] was used to simulate the driving behavior of HVs, and a widely-used lane-change model [31] was employed to simulate the lane-changing behavior of HVs. For a more realistic simulation, four different types of HVs with different body lengths and driving imperfections were used.
The average speed of all vehicles in the road segment was used as the main metric to represent traffic flow. The rate of collisions that occur while executing lane changes was utilized as a key metric for evaluating driving safety. Moreover, we assessed driver comfort by quantifying changes in the vehicle's acceleration and deceleration.
B. Training MARL Model
In this section, we present the details regarding training the MARL model. The optimal values for the hyperparameters used for training the model were identified based on the grid search method [32] (See Table 1). The MARL model was trained until the network converges. As depicted in Fig. 2, our MARL model (as well as that for NP-MARL and LSC-MARL) converged within 500 epochs. In our simulation setting, the simulation update interval was 0.1s, and the duration of an epoch was 5 minutes of simulation steps. Agents updated the model parameters every 10 simulation steps which are equivalent to 1s.
C. Traffic Efficiency
This section presents simulation results related to traffic efficiency. A space-time diagram depicted in Fig. 3 provides a general view of how NP-MARL and LCS-TF perform in terms of traffic efficiency. As illustrated, NP-MARL led We then examined how the average speed changes over time for the three MARL models. Fig. 4 depicts the results. It was observed that despite the random driving behavior of HVs, overall, LCS-TF achieved a significant improvement in the average speed compared with the two other solutions. More specifically, LCS-TF improved the average speed by up to 12% and 25.7% compared with NP-MARL and LSC-MARL, respectively. Interestingly, compared to the other solutions, we observed agents refraining from executing lane changes even though immediate speed improvement was expected. Instead, those agents allowed surrounding vehicles to increase their speed, leading to the overall improvement in traffic flow for the road segment. We also evaluated traffic efficiency by varying the agent density. Here the agent density refers to the percentage of RL agents out of all vehicles injected into the road segment. Fig. 5 shows the results. As depicted, traffic flow improved for all models, as the agent density increased. We made two key observations. First, LCS-TF consistently outperformed the other two models in terms of traffic efficiency, regardless of the agent density. More specifically, on average, LCS-TF achieved 5.3% and 13.2% higher average speed compared with that for NP-MARL and LSC-MARL, respectively. Notably, LCS-TF exhibited a significantly greater performance improvement at higher agent densities, as compared to the other two models. Specifically, for an agent density of 60%, LCS-TF achieved a higher performance gain by 14.55% and 17.78% compared to NP-MARL and LSC-MARL. In contrast, for an agent density of 10%, the performance gain was 0.9% and 13.7% higher than that of NP-MARL and LSC-MARL, respectively.
D. Driving Safety
This section evaluates the performance of LCS-TF focusing on driving safety in comparison with NP-MARL and LSC-MARL. In particular, we measured the rate of collisions to represent driving safety. Fig. 6 depicts the results. As shown, in general, driving safety for all three models improved as the agent density increased. A noteworthy observation was that driving safety for LCS-TF more sharply improved compared to the other two models. More specifically, for LCS-TF, the rate of collisions decreased by 62.5% as the agent density increased from 10% to 60%; in contrast, NP-MARL and LSC-MARL achieved improvements of only 38.9% and 12.2%, respectively. We also noted that under low agent density, all MARL models exhibited unstable performance. On the other hand, more stable performance was achieved as the agent density increased.
E. Driver Comfort
This section presents simulation results on driving comfort. The driving comfort was measured based on the jerk as defined in the driving comfort reward. Note that the value of jerk is always negative, and a higher jerk value means better driving comfort. Fig. 7 depicts the results. In general, as the agent density increased, there were more vehicles executing lane changes with greater care. Therefore, with the higher agent density, we observed the increasing trend of driving comfort for all three models. While improvement in driving comfort was achieved for all models, a notable observation was that LCS-TF achieved the highest driving comfort compared to the other two solutions. Specifically, LCS-TF outperformed by up to 19.4% and 60.2% in terms of driving comfort in comparison with that for NP-MARL and LSC-MARL, respectively.
V. CONCLUSION
We have presented LCS-TF: a multi-agent deep reinforcement learning-based intelligent lane-change system for autonomous vehicles. The design of LCS-TF aims to achieve joint optimization of both the ego vehicle's performance and the overall traffic flow for a given road segment. To achieve the goal, a novel state space design was introduced that incorporates the local traffic information obtained from surrounding vehicles along with the global traffic information collected from the RSU assigned to manage the road segment. We also presented a novel reward function to ensure effective lane-change decisions that improve traffic flow and optimize the ego vehicle's performance. Additionally, a multi-agent deep Q-network (DQN) algorithm was developed to identify the optimal policy for each agent to effectively collaborate in performing lane-change maneuvers. LCS-TF's effectiveness was assessed via extensive simulations and compared against the two main types of existing MARL models. The results indicated that LCS-TF outperforms existing MARL models in all aspects of traffic efficiency, driving safety, and driver comfort. Our future work is to extend the scalability of LCS-TF. Although the current version concentrates on optimizing the performance for a specific road segment, we aim to accomplish multiple road segmentlevel coordination, examine a more large-scale impact of LCS-TF, and develop the next-generation intelligent lanechange system for AVs. | 2023-03-17T01:16:19.336Z | 2023-03-16T00:00:00.000 | {
"year": 2023,
"sha1": "035eaafcdf38f965414522a8f0386f4e90163966",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "035eaafcdf38f965414522a8f0386f4e90163966",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
43291758 | pes2o/s2orc | v3-fos-license | Chitin-binding proteins in invertebrates and plants comprise a common chitin-binding structural motif.
Tachycitin, a 73-residue polypeptide having antimicrobial activity is present in the hemocyte of horseshoe crab (Tachypleus tridentatus). The first three-dimensional structure of invertebrate chitin-binding protein was determined for tachycitin using two-dimensional nuclear magnetic resonance spectroscopy. The measurements indicate that the structure of tachycitin is largely divided into N- and C-terminal domains; the former comprises a three-stranded beta-sheet and the latter a two-stranded beta-sheet following a short helical turn. The latter structural motif shares a significant tertiary structural similarity with the chitin-binding domain of plant chitin-binding protein. This result is thought to provide faithful experimental evidence to the recent hypothesis that chitin-binding proteins of invertebrates and plants are correlated by a convergent evolution process.
dues, and glycines and is frequently referred to as a hevein domain (8). It has been well demonstrated that this domain is indispensable for the antimicrobial activity and exhibits a significant conservation in primary sequence (Ͼ40%) and in three-dimensional (3D) 1 structure (9 -12). Although this advanced knowledge has been provided for the plant chitin-binding proteins, less is known for the invertebrate chitin-binding proteins including tachycitin (1,(13)(14)(15)(16). Kawabata et al. (1) identified that tachycitin is a 73-residues chitin-binding protein having antimicrobial activity. They also revealed that tachycitin consists of five intramolecular disulfide bridges; the connected Cys pairs are 6 -33, 12-30, 24 -61, 25-68, and 40 -53. For invertebrates, the chitin-binding domain was assumed to comprise about 65 residues (17) involving a high percentage of cysteine and aromatic residues in a similar manner to the plant chitin-binding domain. On the basis of such similarity between plant and invertebrate chitin-binding proteins, Shen and Jacobs-Lorena (17) proposed a hypothesis that they are correlated by a rare evolutional process, convergent evolution, i.e. proteins from different origins develop to construct the same active site structure to acquire the same function. However, complete lack of 3D-structural information of invertebrate chitin-binding protein obscures the evolutional relationship between invertebrate and plant chitin-binding proteins. The present study determines the solution structure of tachycitin using NMR spectroscopy, which provides the first 3D structural information of invertebrate chitin-binding protein.
EXPERIMENTAL PROCEDURES
An invertebrate chitin-binding protein, tachycitin, was isolated from hemocyte debris of horseshoe crab (Tachypleus tridentatus) as described previously (1) and used without further purification. The NMR samples were prepared by dissolving tachycitin in either 0.3 ml of D 2 O or H 2 O containing 10% D 2 O to give a final concentration of 1-2 mM, whose pH values were adjusted to be 4.0 -6.5 by addition of DCl and/or NaOD. The NMR experiments were performed on JEOL JNM-Alpha 500 and 600 spectrometers operating at temperatures of 15, 20, 30, and 40°C. The two-dimensional experiments, DQF-COSY, TOCSY (mixing time ϭ 75, 85 ms), and NOESY (mixing time ϭ 75, 250 ms), were acquired with low-power (20 Hz) presaturation on the water. The temperature coefficient (Ϫ⌬␦/⌬T, ppb K Ϫ1 ) was estimated from the temperature dependence (15-40°C) of the chemical shift of the H N resonance. The chemical shifts were referenced to the internal standard, TSP (0.00 ppm). Interproton distance restraints were derived from NOE crosspeaks in the NOESY spectra (mixing time ϭ 75 ms), calibrated the peak intensities with known distances (2.2 Å for H ␣ (i)-H N (i ϩ 1) of -sheet and 1.75 Å for H  -H  Ј), and were used as inputs for 3D structural calculations of tachycitin. The NOEs were classified into strong, medium, and weak, corresponding to three distance restraints with an upper limit of 2.7, 3.5, and 5.0 Å, respectively. The upper distance limit was corrected for methyl and methylene protons that were not assigned stereospecifically. The 35 dihedral angle restraints were obtained by measuring 3 J NH-H␣ coupling constants; the angle restraint of Ϫ60 Ϯ 30°was used for the residues having 3 J HN-H␣ coupling constants less than 6 Hz, and that of Ϫ120 Ϯ 30°was used for the residues having 3 J HN-H␣ constants larger than 8 Hz. Hydrogen bond distance restraints were applied between nitrogen and oxygen atoms (2.4 -3.5 Å) and Hn and oxygen atoms (1.5-2.5 Å) for regular secondary structures. The hydrogen bonding was assumed for the residues 18, 27-29, 31, 34, 36, 38, 45, 47, 52, 54, and 59, which show low temperature coefficients * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The atomic coordinates and structure restraints ( (Ͻ4.5 ppb K Ϫ1 ). Our previously determined disulfide bond formations were also used as the distance restraints (1). All NMR data were processed using NMRPipe (18) and PIPP (19), and the structure calculations were performed using simulated annealing (SA) protocol in X-PLOR 3.851 (20). An extended structure of tachycitin was used as the starting structure for calculation, whose C-terminal end is patched (1) with an amide group, CONH 2 . The initial set of restraints included only NOE restraints, no dihedral restraints, and no hydrogen bond restraints. From this initial set of restraints, 100 structures were generated using the SA protocol with heating for 60 ps and cooling for 30 ps. Of those 100 structures, 50 structures with lower total energy were selected as starting structures for the next refinement cycles. The refinement cycles were performed using the SA protocol with heating for 30 ps and cooling for 20 ps. The 25 lowest-energy structures presented in this paper were selected from the 50 structures calculated from the final round of refinement. The program PROCHECK (21) reveals that for all residues, 97.3% of the backbone and dihedral angles fell into core and allowed regions of the Ramachandran map.
RESULTS AND DISCUSSION
The structural determination was performed (Table I) using NMR-derived 1,070 experimental restraints, on the basis of the whole assignments of the 1 H resonances of tachycitin at pH 4.2 and at 30°C, which were deposited to BioMagResBank with the accession number of 4290. Structure of tachycitin ( Fig. 1) appears to comprise a three-stranded -sheet (1, 2, and 3; residues 17-19, 26 -31, and 34 -39) in the N-terminal region and a two-stranded -sheet (4 and 5; residues 45-47 and 52-54) following a short helical turn (␣1; residues 56 -59) in the C-terminal region. Such arrangements of the secondary structures of tachycitin are not similar to those of any other known antimicrobial peptides in invertebrates; for example, the insect defensin family consists of one long loop, one ␣-helix, and one -sheet from the N terminus (22). As shown in Fig. 1B, a distorted -sandwich structure is constructed by the threestranded and two-stranded -sheets connected through a bending loop (Cys-40 -Leu-44). It appears that this bending loop involves a type III' -turn contributed by the residues Pro-41-Leu-44, for which the formation of a hydrogen bond between Leu-44 H N and Pro-41 OЈ is evidenced by low temperature coefficient of Leu-44 (3.0 ppb K Ϫ1 ). A short segment (residues His-31-Leu-34) flanked between the strands 2 and 3 constructs a -turn conformation. For this -turn, molecular motional restraint contributed by a disulfide bond (Cys-6 -Cys-33) is suggested by observations of significant line-broadening of the H N resonances for Lys-32 and Cys-33. Another segment comprising six residues (Asn-47-Val-52) flanked by strands 4 and 5 adapt a -hairpin structure. In this -hairpin, Asn-47 OЈ presumably forms a hydrogen bond with the Lys-51 H N , which is supported by the extremely low temperature coefficient (2.0 ppb K Ϫ1 ) obtained for Lys-51. Overall, the structure of tachycitin is characterized by -sheets flanking short loops and turns, which is typical for most of the small disulfide-rich polypeptides (23).
It was revealed that tachycitin shares a remarkable local structural similarity with a plant chitin-binding protein named hevein. Comparison between our determined structure of tachycitin ( Fig. 2A) and a previously reported structure of hevein (9) (Fig. 2B) clearly shows that an antiparallel -sheet (colored in blue) and a helical turn (colored in red) are constructed in both proteins in highly similar manners. In addition, formation of a disulfide bridge (between Cys-40 and Cys-53) connecting the middle of 5 and the C terminus of 3 for tachycitin (colored in green, Fig. 2A) is similarly identified in hevein (Fig. 2B). The structural similarity further includes the loop regions, e.g. a hairpin loop structure involved in the antiparallel -sheet (colored in orange). It should be noted that the hairpin loop of tachycitin (Asn-47-Val-52) comprises six residues with ␣␣␥␣ L  conformation whereas the corresponding loop of hevein comprises five residues with ␣␥␣ L  conformation.
Kawabata et al. (1) reported that the N-terminal 5-28 region of tachycitin shows sequence similarity with the N-terminal 2-21 region of hevein. However, such similarity is not identified by the present study; the secondary structural arrangement, as well as the disulfide-bond patterns, appears to be quite different for the suggested regions.
The structural similarity between segment Cys-40 -Gly-60 of tachycitin and segment Cys-12-Ser-32 of hevein, both comprising the antiparallel -sheet (4 and 5), was examined by looking at the superimpositions of the segments (Fig. 3). The structural motif shown in Fig. 3 has been found in several plant chitin-binding proteins (10 -12) (Fig. 4). For hevein, segment Cys-12-Ser-32 was identified as an essential chitin-binding domain (24). It appears that arrangements of the two structural motifs shown in Fig. 3 are significantly consistent with each other (backbone RMSD ϭ ϳ1.5 Å). The aromatic sidechain groups of Trp-21 and Trp-23 of hevein (Fig. 3) are known to bind specifically to chitin-derived oligosaccharides through hydrophobic interactions (24,25). This binding is further strengthened by a hydrogen bonding with Ser-19 of hevein (25). As shown in Fig. 3, the residues of Asn-47, Tyr-49, and Val-52 of tachycitin are located at perfectly corresponding positions to the residues of Ser-19, Trp-21, and Trp-23 of hevein. Therefore, one could assume that the region shown in Fig. 3 comprising an antiparallel -sheet and a helical turn (4, 5, and ␣1; Fig. 2A) in the C-domain of tachycitin serves as an essential chitinbinding site, which protrudes the side-chains of the putative functional residues, Asn-47, Tyr-49, and Val-52. Overall, it could be assumed that the N-terminal region comprising 1-3 Fig. 2A) behaves as a stable domain so as to locate the C-terminal domain chitin-binding site proper for its function.
Conservation of the chitin-binding structural motif among the chitin-binding proteins in invertebrates and plants was further examined by alignment tests of the proteins with regard to their amino acid sequences corresponding to Cys-40 -Gly-60 of tachycitin (Fig. 4). The 3D structural information has been available for the plant chitin-binding proteins (9 -11). The information is now available for only tachycitin among the invertebrate chitin-binding proteins. It appears that the residues of Cys, Pro, and Gly, all of which have significant influence on the structural constructions, are well conserved in the chitin-binding proteins listed in Fig. 4. Conservation of polar and hydrophobic residues is further identified for the putative chitin-binding residues (e.g. Asn-47, Tyr-49, and Val-52 for tachycitin). For all plant chitin-binding proteins, the 21-residues segments listed in Fig. 4 appear to construct a closely similar 3D structure (9 -11) to the putative chitin-binding site of tachycitin. Further similarity in primary sequence identified between tachycitin, Ag-chit, Pj-chit1, Ch-chit, Peritrophin-44, and Tn-IM (nomenclatures described in the figure legend) assumes that these segments of the invertebrate chitin-binding proteins commonly comprise the chitin-binding structural motif as identified in tachycitin. In 1999, Shen and Jacobs-Lorena (17) proposed a hypothesis that chitin-binding proteins in invertebrates and plants are correlated by a rare evolutional process, convergent evolution. Our present structural determination of tachycitin and the 3D structure-based sequence alignment are thought to provide faithful evidences for the proposed idea of the convergent evolution relationship between invertebrate and plant chitin-binding proteins. (29). Plants are as follows: hevein from rubber tree (Hevein) (9), Amaranthus caudatus antimicrobial protein 2 (Ac-AMP2) (10), and four homologous domains of wheat germ agglutinin (WGA A, -B, -C, and -D) (11). Residue numbers for each segment are indicated in parentheses. Amino acids conserved between invertebrate and plant proteins are indicated with bold letters. The chitin-binding residues in plants and the corresponding residues in invertebrates are found to be aligned (indicated by asterisks at the bottom), for which polar and hydrophobic residues are colored in red and blue, respectively.
NMR Structure of the Invertebrate Chitin-binding Protein 17931 | 2018-04-03T05:32:33.890Z | 2000-06-16T00:00:00.000 | {
"year": 2000,
"sha1": "e43c38e24624c7c64d9d89008c95f7d2239a94ae",
"oa_license": "CCBY",
"oa_url": "http://www.jbc.org/content/275/24/17929.full.pdf",
"oa_status": "HYBRID",
"pdf_src": "Adhoc",
"pdf_hash": "26e16f497cee62f4671a43dcfbc83766271928cf",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine",
"Biology"
]
} |
257445484 | pes2o/s2orc | v3-fos-license | Multivariate bias correction of regional climate model boundary conditions
Improving modeling capacities requires a better understanding of both the physical relationship between the variables and climate models with a higher degree of skill than is currently achieved by Global Climate Models (GCMs). Although Regional Climate Models (RCMs) are commonly used to resolve finer scales, their application is restricted by the inherent systematic biases within the GCM datasets that can be propagated into the RCM simulation through the model input boundaries. Hence, it is advisable to remove the systematic biases in the GCM simulations prior to downscaling, forming improved input boundary conditions for the RCMs. Various mathematical approaches have been formulated to correct such biases. Most of the techniques, however, correct each variable independently leading to physical inconsistencies across the variables in dynamically linked fields. Here, we investigate bias corrections ranging from simple to more complex techniques to correct biases of RCM input boundary conditions. The results show that substantial improvements in model performance are achieved after applying bias correction to the boundaries of RCM. This work identifies that the effectiveness of increasingly sophisticated techniques is able to improve the simulated rainfall characteristics. An RCM with multivariate bias correction, which corrects temporal persistence and inter-variable relationships, better represents extreme events relative to univariate bias correction techniques, which do not account for the physical relationship between the variables.
Introduction
The direct use of Global Climate Models (GCMs) is limited at regional or hydrological catchment scales because of the coarse spatial and temporal scales used (Caldwell et al. 2009). Regional Climate Models (RCMs) forced with GCM simulations through initial, lateral, and lower boundary conditions are more commonly used to simulate detailed information at finer scales.
Although the GCM-driven RCM simulations show better performance than the raw GCM simulations (Diffenbaugh et al. 2005;Leung and Qian 2009;Di Luca et al. 2016;Li et al. 2018), systematic biases that can be propagated from GCM data into the input boundary conditions of the RCM still remain. These improper boundary conditions can affect model outputs, leading to biases when compared to observation (Caldwell et al. 2009;Xu and Yang 2012;Rocheta et al. 2017;Kim et al. 2020). Applying bias correction, sometimes called bias adjustment but herein as bias correction, to climate model outputs is therefore an important step in improving model performance. Different approaches for correcting and characterizing systematic bias have been applied to either surface variables of the RCM outputs before use for impact assessment, or the GCM simulations prior to downscaling to remove biases in the RCM input (lateral and lower) boundary conditions. The second approach above forms the focus of this study.
For the case where bias is removed as a post-processing step following the climate model simulation, past studies have evaluated various bias-correction approaches that vary in complexity in terms of the mathematical operations used, as well as the nature of the attributes that are corrected. Most common approaches use quantile mapping as the basis for adjusting the distribution of a simulated variable to match 1 3 that of the observations to correct both current and future climate simulations (Wood et al. 2004;Li et al. 2010;Piani et al. 2010). Alternatives that address the model biases in temporal persistence are also widely used given their relevance in hydrological applications (Johnson and Sharma 2012;Ojha et al. 2013). Recently, bias correction methods have evolved toward more advanced techniques that correct inter-variable relationships among the atmospheric variables of the model outputs, such as precipitation and temperature, which possess a strong co-dependence (Mehrotra and Sharma 2015;Sharma and Mehrotra 2016;Cannon 2016Cannon , 2017François et al. 2020). The abovementioned studies, however, were specifically designed for correcting the surface fields of the climate model output and have not been applied to the full atmospheric fields required for RCM lateral boundary inputs.
The method for correcting the RCM input boundary conditions has evolved from simple to more complex techniques that attempt to mimic observed multi-scale relationships in simulations. Xu and Yang (2012) investigated mean and variance bias correction of the RCM boundary conditions, and showed that the bias corrections improved the model performance for climatological means and extremes. Bruyere et al. (2014) corrected the RCM boundary conditions to investigate simulations of hurricanes, which can produce extreme precipitation. The results showed that simple mean bias correction produced the greatest improvement.
The bias correction approaches that have been used to correct the RCM input boundaries in previous studies, however, often overlooked the biases in the persistence-related attributes in the model outcomes, and corrected the biases at a single time scale (Bates et al. 2008;Nguyen et al. 2016). To overcome this limitation, Rocheta et al. (2017) investigated correcting low-frequency rainfall variability representation using nested bias correction, which includes correcting lag1 auto-correlation correction at multiple time scales. They showed that this method produced an added improvement compared with the simple scaling approach. Kim et al. (2020) also used nested bias correction to improve RCM simulations of rainfall extremes. They showed that even simple bias correction techniques could significantly reduce bias in terms of extreme rainfall events. They also showed that nested bias correction performed better than simple corrections for seasonal extremes, suggesting that correcting persistence characteristics is essential for the simulations.
While bias corrections of RCM input boundaries have shown consistent improvement in mean and extreme events in previous studies, these approaches have addressed only the correction of individual variables under the assumption that inter-variable bias is not of key importance. However, if the physical relationships between the variables in dynamically linked fields are not considered, the errors introduced can influence the outputs, including precipitation, temperature, and humidity (Chen et al. 2011;Rocheta et al. 2014). The misconception of interacting physical processes across multiple temporal and spatial scales can potentially lead to the underestimation of extreme events, such as heavy rainfall and droughts (Zscheischler et al. 2018).
In a recent study, Kim et al. (2021) investigated whether the RCM could reproduce spatio-temporal and inter-variable dependence with respect to observed variables. They showed that the univariate bias correction techniques improved temporal and spatial dependence but not multivariate dependence. They highlighted that the mismatch in the physical relationships among the atmospheric variables was not sufficiently adjusted through the relaxation zone, suggesting further investigation to address all systematic biases in the boundary conditions.
In brief, RCM outputs with univariate bias-corrected boundaries still contain large biases in the multivariate dependence even though they exhibit adequate performance for simple statistics, leading to substantial anomalies in simulated extremes. (Kim et al. 2020).
Considering this, the focus of this study is the impact that multivariate bias correction can have on lateral boundary conditions (LBCs) and the lower boundary condition, which represents the sea surface temperature (SSTs) field, particularly for rainfall characteristics. Along with addressing biases in the cross-dependence attributes of an inter-variable field, we addressed biases across multiple time scales using the nesting approach. We also used mean, mean and variance, and univariate nested bias correction methods to compare the model performance between the simple scaling and the more complex approaches.
The paper is organized as follows. In Sect. 2, the datasets and methods which include bias correction approaches and the Weather Research and Forecasting (WRF) model setup are described. Results are presented in Sect. 3. The discussion and limitations of this work are presented in Sect. 4. Conclusions are described in Sect. 5.
Models and data
This study used the Australian Community Climate and Earth System Simulator Earth System Model Version 1.5 (ACCESS-ESM1.5) GCM simulation made available by Commonwealth Scientific and Industrial Research Organization (CSIRO) for the purpose of contributing to the internationally coordinated Coupled Model Intercomparison Project Phase 6 (CMIP6) (Ziehn et al. 2020). For RCM simulations, the Weather Research and Forecasting model (WRF) with dynamical core (ARW), version 4.2.1 (Skamarock et al. 2019) was used. The resolution of the ACCESS-ESM1.5 is based on the N96 Gaussian grid (approximately 1.875°EW × 1.25°NS) with 38 vertical levels extending from the surface to 40 km.
The reanalysis data used for correcting GCM biases was ERA5 which is the fifth-generation model reanalysis of the global climate from the European Centre for Medium-range Weather Forecasts (ECMWF) (Hersbach et al. 2020). It has a resolution of 31 km with 37 pressure levels. The ERA5driven RCM simulation was considered a "perfect" regional climate simulation in this study.
The atmospheric variables [specific humidity q (g/kg), temperature T (K), zonal wind u (m/s), and meridional wind v (m/s)] and the sea surface temperature (SST) in the lateral and lower boundary conditions of RCM were corrected towards those of ERA5. The ERA5 boundary variables were first regridded using both the conservative (for the specific humidity) and bilinear (for the other four variables) remapping methods and then linearly interpolated to match the ACCESS-ESM1.5 horizontal and vertical resolution. The pressures from the surface to the top were recalculated using the bias-corrected fields based on the hypsometric equation. Therefore, the pressure field, including the surface pressure, is also adjusted during the bias correction procedure. All other variables remained identical, and the boundary conditions were built after bias correction. The model was simulated over 31 years from 1 January 1982 to 31 December 2012. The first year of the model simulation was ignored as a spin-up period in order to remove issues associated with the equilibrium state for various soil types (Cosgrove et al. 2003;Chen et al. 2007).
Bias correction approaches
In this study, four different types of bias correction methods were implemented to correct the atmospheric variable (specific humidity, temperature, zonal and meridional winds) and surface fields (sea surface temperature) at the boundaries. Simple scaling methods, which correct climatological mean and mean and variance with respect to those of reanalysis data, were used to correct the magnitude of events independently for each variable at a single time scale. In this study, these corrections were on a daily basis, and the daily statistics were calculated using the data falling within a centered moving window of 31 days (Sharma and Lall 1999). A more sophisticated alternative that also corrected lag1 autoand lag0 cross-correlation across variables with respect to those of reanalysis data, was applied at multiple time scales using a nesting approach which can correct the biases in the persistence related attributes in the model outcomes (Johnson and Sharma 2012). The bias corrections were implemented at 6-hourly time scale over the 31 years GCM dataset. Here we denote X g () is the GCM data, X o () is the reanalysis data, and X g () is the transformed GCM variable. The subscripts 6h , d , m , s , y indicate the time scale for 6-h, day, month, season, and year, respectively. The superscripts g , o indicate the model and observation, respectively.
Mean bias correction
This correction was implemented to calculate climatological means in which each of the GCM simulations was transformed toward that of the reanalysis data. The transformation function can be expressed as follows: where X g d and X o d are the climatological means of the original GCM and the reanalysis data at the daily time scales, respectively. The daily statistics were calculated using a centered moving window of 31 days.
Mean and variance bias correction
Another simple approach for correcting biases in boundary variables was the mean and variance bias correction. This approach is similar to the mean bias correction, but it is extended to include the variability of the atmospheric variables and surface fields. This can be expressed as below by applying a standard deviation correction to the mean correction: where s g d and s o d represent the standard deviation of the original GCM and the reanalysis variables at the daily time scale, respectively. After this bias correction, the mean and variance of bias-corrected GCM simulations ( X g 6h ) are similar to those of reanalysis data. These statistics were estimated on 6-hourly GCM data in a daily group over 31 years, and the centered 31 days moving window was applied.
Nested bias correction (NBC)
A more comprehensive alternative extended to include lag1 auto-correlation attributes on multiple nested timescales was used. This approach has shown an improvement in the rainfall mean and variability at multiple time scales (monthly, seasonal, and annual time scale) in the previous studies (Johnson and Sharma 2012;Kim et al. 2020). Based on a standard autoregressive lag1 model, lag1 auto-correlation of GCM was replaced by the observed lag1 auto-correlation as follows: where z g t is standardized GCM variable for time step t, r o t and r g t are lag1 auto-correlation of the reanalysis and GCM variables, respectively. The corrections were applied at daily, monthly, seasonal, and annual time scales, which are extended to the daily time scale with respect to the previous studies. These different sets of bias-corrected time series are then combined with the nesting approach. This means that the bias-corrected daily GCM data is aggregated to a longer timescale (month, season, and year) and follows the same procedures for other timescales as well. The daily GCM data thus incorporates the effect of bias correction at longer time scales, which include nested bias-corrected monthly, seasonal, and annual time scales. The general process of the nesting approach is described in (Johnson and Sharma 2012;Mehrotra and Sharma 2012). From Srikanthan and Pegram (2009), the corrected variables at different time scales can be applied to the raw 6-hourly GCM data as a form of weighting factor: where X g 6h is the 6-hourly corrected value and X g 6h is the original 6-hourly value, X d,m,s,y is the daily GCM data for day d, month m, season s, and year y. X m,s,y is the aggregated monthly value, X s,y is the aggregated seasonal value, and X y is the aggregated yearly value. Finally, X g h exhibits the same persistence attributes as X o () across the nesting time scale. These bias corrections were carried out at each grid cell over the 31 years GCM dataset. The full details of the nested bias correction can be found in Johnson and Sharma (2012).
Multivariate bias correction (MBC)
The univariate bias correction methods as described above are designed to correct variables independently. Although univariate distribution features can be adjusted toward reference data, they can generate inappropriate cross-dependence structures among multiple variables simulated by GCMs, leading to substantial anomalies in the simulation.
To correct the inter-variable relationships, we investigate the multivariate recursive nesting bias correction method which is extended to also include lag0 and lag1 auto-and cross-correlations among the variables. This approach was presented by Mehrotra and Sharma (2015) and has shown that it can correct for variability and persistence biases in the atmospheric variables on a range of timescales. A large number of parameters, however, may cause large estimation errors and possibly hide (overestimate) the true quality of the model (Ehret et al. 2012;Mehrotra and Sharma 2021). In this study thus some simplifications in the model structure were applied and lag1 cross-correlations are assumed to be unimportant and are ignored. We investigate the distributional attributesmean and variance and the dependence attributes-the lag1 auto-and lag0 cross-correlation coefficients were calculated at four selected time scales, daily, monthly, seasonal, and annual. Three iterations (for the recursive scheme) were adopted to reduce biases in the nesting procedure across all timescales (Mehrotra and Sharma 2012). We denote the vectors Z g () and Z o () are the standardized GCM and reanalysis variables with zero mean and unit variance, respectively.
A multivariate transformation for the day t is based on a standard Multivariate AutoRegressive order 1 (MAR1) model and can be expressed as follows (Salas 1980): and where , and , are coefficient matrices of the lag1 and lag0 cross-correlations of the reanalysis and GCM data, respectively. The random vector ε t is assumed to be mutually independent, indicating that where is the identity matrix, and T denotes the transpose of the matrix.
Since the elements of ε t are mutually independent having a zero mean and unit variance, the coefficient matrices and or and can be expressed as follows (Matalas 1967): which are the lag0 and lag1 cross-correlation matrices. The elements of these matrices corresponding to variables i and j are obtained from: where N is the total number of samples. The coefficient matrices and are obtained by singular value decomposition, which is a method of decomposing a matrix into three component matrices that are easy to manipulate and analyze.
These procedures are applied by rearranging the Eq. (5b) for ε t : where ε t represents a standardized vector that has removed the lag0 and lag1 auto-and cross-correlations obtained from the Z g t series. This vector is then used to modify Z g t to Ź g t that has the lag0 and lag1 attributes of reanalysis data.
Finally, by adding back the means and standard deviations of reanalysis data, the bias-corrected output provides appropriate attributes in means, standard deviations, lag1 auto-, and lag0 and lag1 cross-correlations.
Similarly, MAR1 with periodic parameters can be calculated as: where Z g y, represents the vectors of the standardized GCM periodic series in time interval , and year y , indicating that the series all contain the seasonal cycle. The matrix parameters of this model can be obtained following as: where 0, and 1, are the periodic lag0 and lag1 matrix correlation matrices. The elements of these matrices are obtained as similar to those with constant parameters. As abovementioned, some simplifications in the MAR1 model structure are applied in this study to avoid a large number of parameters, which may hide the true quality of the model simulations. The lag1 cross-correlations are thus assumed to be unimportant, and the parameters and are considered as diagonal matrices in which the elements outside the main diagonal are all zero. This simplification is also applied to the periodic model. The elements of these matrices corresponding to variables i and j can be obtained as: These different sets of bias-corrected time series are then combined with the nesting approach as abovementioned in the Eq. (4).
Performance assessment
Three statistics were used to quantify the impact of bias correction: mean absolute error (MAE), Pearson correlation coefficient (R), and bias. The MAE is defined as where N is the total number of grid cells, X mod n and X obs n represent the climatological WRF outputs and observation data at each grid cell, respectively.
The differences of the means between the model simulation and observation can be defined by mean bias as: The biases are calculated for each vertical level at each grid cell.
To assess the ability of the model simulations in terms of multivariate aspect, Fisher z-transformation was used here to convert the correlation coefficient to a transformed variable (Fisher 1915(Fisher , 1921. The test statistic ( z ) can be represented as a standard normal deviate where z M r and z O r are Fisher's z transformation of the model and observed correlation coefficient ( z r = 1 2 ln 1+r 1−r ), and s M and s O are sample sizes of the model outputs and observation, respectively. This study assumed a 5% significance is a criterion for rejecting the null hypothesis, which means that there is a significant difference in the correlation between the two simulations if the computed z-value is not in the z -value distribution table. Full details of this formulation can be found in Kim et al. (2021).
Results
This section will first evaluate the statistics of the biascorrected atmospheric variables in the boundary conditions with respect to the ERA5 datasets. RCM simulations are then compared to the ERA5-driven RCM outputs over the Australasia CORDEX domain to reveal the effectiveness of bias correction in terms of the climatological mean, standard deviation, lag1 auto-correlation, and multivariate dependence measures. The ERA5-driven RCM simulation is treated as a simulation with "perfect" boundary conditions. To evaluate the model performance, the outermost five grid cells from the boundary were trimmed off as the relaxation zone to avoid a bias due to boundary effects.
Evaluation of bias-corrected GCM datasets
This section evaluates the impact of bias correction on the GCM outputs that have been subsetted along the lateral boundary conditions. Four different bias correction approaches introduced in Sect. 2 were applied to the atmospheric variables (u, v, T, q) and sea surface temperature (SST) to the raw GCM datasets. From the results presented here, we can evaluate whether the bias correction approaches reduce a bias before downscaling. Figure 1 shows a scatterplot for specific humidity q (g/kg) comparing ERA5 to the four bias correction techniques: GCM(M), GCM(MSD), GCM(NBC), and GCM(MBC). These indicate the raw GCM datasets, GCM with mean bias-corrected boundary conditions, GCM with mean and standard deviation bias-corrected boundary conditions, GCM with nested bias correction, and GCM with multivariate bias-corrected boundary conditions, respectively. The results show the effect of each bias correction approach at daily, monthly, seasonal, and annual time scales for the three statistics over 38 vertical levels along the western boundary. Details regarding the other variables, T and w, along the eastern, northern, and southern boundaries can be found in the online supplemental material: Fig. A1 to A10. Each statistic presented in the figure shows a separate point for each day, month, season, and year; hence there is a single point for each grid cell at each time scale.
It is clear that the bias corrections produce an improvement as expected by the construction of each formulation that is specified to adjust in the mean, standard deviation, and lag1 auto-correlation fields. We see that there is a large difference between before and after bias correction for lag1 auto-correlation, particularly for the annual time scale, which can affect the model performance in simulating persistence characteristics and the temporal variability of the rainfall. Despite the GCM being different in all statistics, the biases for the simple statistics, mean and standard deviation, are noticeably reduced compared to the old versions of simulations (Rocheta et al. 2017;Kim et al. 2020). This indicates that the GCM performance has been improved in CMIP6 as it represents state-of-the-art climate modeling. Figure 2 shows a scatter plot comparing the bias-corrected and uncorrected atmospheric variables to ERA5 in terms of multivariate relationships among the three atmospheric variables over all vertical levels. Specific humidity above the 24th level has been ignored during multivariate bias correction as it is always close to zero which can lead to an unstable output. Unlike the model performance shown in Fig. 1, there are large biases even after the bias corrections, GCM(M), GCM(MSD), and GCM(NBC), particularly in the annual time scale. This indicates that the univariate bias corrections are not capable of correcting physical relationships, which may cause unrealistic behavior of the atmospheric variables. On the other hand, GCM(MBC) performs well, showing most points are spread around the 45-degree line, meaning that bias in the relationships among the variables is well corrected. Details along the eastern, northern, and southern boundaries can be found in the online supplemental material: Fig. B1 to B3. Figure 3 shows a bias map of seasonal SST over 31 years comparing the bias-corrected and uncorrected SST to ERA5 covering the research domain for three statistics. The results show that the raw GCM contains a large positive bias for mean and standard deviation over the domain and a negative bias in the tropics. The results show that each bias correction performs well with regard to the statistics as they are constructed.
Evaluation of RCM simulations over the Australasia CORDEX domain
This section presents the RCM outputs to assess whether the influence of the bias-corrected boundary conditions is preserved even after the RCM simulations in terms of four aspects: mean, standard deviation, extremes, and the multivariate relationship.
Climatological mean
We first evaluate the RCM simulations with regard to the climatological mean. Figure 4 shows a bias map of RCM simulations compared with ERA5-driven RCM outputs for three surface variables for monthly mean statistics over 30 years.
The results show that RCM(GCM) produces a large positive bias for the specific humidity and temperature over the domain. It can also be seen that the RCM output with the raw GCM boundary conditions are significantly different for precipitation showing a significant negative (dry) bias over the Western Pacific Warm Pool and positive (wet) bias along the South Pacific Convergence Zone. In contrast, the RCM with bias-corrected boundary conditions shows a noticeable improvement across the variables. While negative biases are shown in western and southern Australia for the specific humidity, a bias is substantially reduced with the 0.1 mean absolute error over the domain. We see that the specific humidity and temperature are quite insensitive to the bias correction approaches with respect to a monthly climatological mean. It is clear that correcting variability of the atmospheric variables along the boundaries can better represent the monthly mean precipitation in the tropics compared to the simple mean bias correction technique.
Inter-annual variability
We then evaluate the model performance in terms of the standard deviation aspect. Figure 5 shows a bias map of the coefficient of variation (CV) of annual precipitation. The higher the CV, the more variable the inter-annual precipitation of a location is. The mean absolute error over the domain is represented in the bottom right of each map. We see that RCM(GCM) produces a large positive bias compared to the ERA5-driven RCM outputs in the tropics, Indian Ocean, and South Australia. This rainfall variability can be influenced by several global and local climate drivers, such as the Indian Ocean Dipole, El Niño Southern Oscillation, and northwest cloudbands (Risbey et al. 2009). Whereas the RCM outputs with bias-corrected boundary conditions show improvement, biases still remain in the tropics and on Australia's northwest coast. The results show that RCM(M) can improve rainfall variability even though the boundary variables were corrected without regard to the standard deviation. From the results we see that RCM(MBC) performs well with the lowest mean absolute error, 4.5, over the domain. Figure 6 compares the models and RCM(ERA5) for standard deviation of the three surface variables at the presented here shows some degradations in MAM for precipitation, suggesting that correction on mean statistics only may increase the model uncertainty in capturing output variability. The figure shows that RCM(MSD), RCM(NBC), and RCM(MBC) improve most time scales for the mean absolute error compared to RCM(GCM). While the complex approaches often show a marginal improvement, and can be worse than the simple approaches for some variables in some seasons (DJF for specific humidity and SON for precipitation), we see that they generally perform better over the domain. This means a corresponding improvement in the statistics of RCM outputs over the domain, indicating that the rainfall simulation can be changed as a result of changes in the atmospheric variables in the boundary conditions.
Extreme events
In order to evaluate the effect of bias-corrected boundary conditions on extreme events, we used the 99th percentile determined for each grid cell and each variable at a daily time scale individually. Figure 7 shows a bar plot similar Fig. 6, we broadly see that the better representation of relationships between the atmospheric variables in the boundaries translates effectively to the fine-scale extreme events. RCM(NBC) shows worse performance for precipitation in DJF than the RCM(MSD), even though it includes the lag1 auto-correlation attribute, indicating that using more complex techniques without regard to the physical relationship between the variables does not guarantee the better performance for rainfall extremes.
Multivariate relationship
This section investigates the effect of bias correction with regard to multivariate dependence in the RCM outputs to assess whether or not the relationship among the variables is preserved even after passing through the boundary relaxation zone.
The impact of lateral boundary conditions is linearly reduced through the relaxation zone (from the outermost specified zone to the inner zone). One can expect that the mismatch in the multivariate relationship will be increased when compared to that of ERA5. Here pairwise correlation of the atmospheric variables was computed for each RCM simulation and then compared against RCM(ERA5) with a 5% significance level. Grid cells in the specified zone and the sixth grid cells, which are just inside the relaxation zone, were assessed along the boundaries as suggested in Kim et al. (2021). Table 1 shows the results of the multivariate cross-correlation between the RCM simulations and RCM(ERA5) at the 5% significance level as a criterion for rejecting the null hypothesis that the correlation is equal. For example, one can test the hypothesis that the correlation between temperature 'a' and specific humidity 'b' of an RCM simulation is the same as that for temperature 'c' and specific humidity 'd' of observation (i.e., H 0 ∶ ab = cd ). The percentage indicates a number of grid cells showing a significant difference in multivariate relationship compared to RCM(ERA5). As mentioned above, specific humidity close to zero was ignored during calculation. The results show that 79.9% of RCM(GCM) grid cells are significantly different compared to RCM(ERA5) for the variable wind speed (w) and temperature (T) at a daily time scale along the specified zone at western boundary. Although the RCM(M), RCM(MSD), and RCM(NBC) marginally improve the multivariate aspects of the variable combination, large differences for all three relationships are still shown ranging from 62.7 to 78.3% along the western boundary. This indicates that those bias corrections lack the ability to capture the inter-variable relationship presented in RCM(ERA5). On the other hand, RCM(MBC) reduces the bias by showing 21.6-72.0%, and substantial improvements are shown in correlations with wind speed. While the improvement falls into 50.0-63.5% after passing through the relaxation zone as expected, RCM(MBC) still shows better correlations between the variables. Although there are some exceptions where the effect of multivariate bias correction is reduced in the interior model domain, w and q at 6th grid line in the north and T and q at 6th grid
Discussion
The work presented in this study demonstrates the influence of bias correction techniques on the input boundary conditions of an RCM. The focus here is whether multivariate bias correction can reduce bias with respect to the intervariable relationships and, if so, whether this improvement is preserved through the relaxation zone where the model is nudged or relaxed towards the driving model. This section is divided into three parts. Section 4.1 discusses the performance of the different versions of the raw GCM simulations. Section 4.2 evaluates the effect of the relaxation zone, Sect. 4.3 discusses the implications of the bias correction techniques used in this study, and Sect. 4.4 evaluates the advantages of the multivariate bias correction approach. Finally, Sect. 4.5 discusses the limitations and future work.
Comparison with previous research
This study used state-of-the-art GCM simulations and the WRF model to assess the performance of RCM simulations using various bias-correction alternatives. The ACCESS-ESM1.5 model used here represents Australia's contribution in CMIP6 and has shown significant improvements over the historical period against observations over the previous version, ACCESS-ESM1 (Ziehn et al. 2020). One can expect that the systematic bias will decrease as the model is updated; the newer the model, the lower the bias. The results of this study also agree with this expectation by showing that the RCM simulation without bias correction over the domain has improved when compared to previous studies that have used older generation GCMs (Rocheta et al. 2017;Kim et al. 2020Kim et al. , 2021.
However, it is also clear that the newer model still contains biases even though it has been updated. Although the GCM used here shows a relatively good performance in the climatological mean, there are significant differences in other statistics: standard deviation, lag1 auto-, and lag0 cross-correlation, leading to considerable anomalies in simulation of extremes.
It should be noted that correcting RCM boundary conditions before downscaling is important to reduce the biases in the outputs, and more complex techniques, which extend to lag1 auto-and lag0 cross-correlation, appear to be improving details that are important in capturing persistence and maintaining physical consistency (Fig. C1 to C4). This is expected to provide a complete picture of the efficacy of the ACCESS models in simulating the historical climate, as well as a better estimation of future climate change as noted in .
Can the impact of the bias correction along the outermost zone be preserved inside the RCM domain?
The results in Table 1 showed that improvements are certainly evident, and MBC reduces the bias related to physical inconsistency by up to 53.9% in the western boundary compared to those obtained from RCM(GCM). We see that MBC is capable of preserving the inter-variable relationships even after passing through the relaxation zone, showing up to 24.5% improvement. This indicates that the bias-corrected dependent structure in the boundaries of the RCM can be preserved inside the domain despite the influence of the internal dynamics of WRF that propagate into the atmospheric fields in complex ways. This result is in line with a previous study ) that has shown that a large portion of bias-corrected information infiltrating into the model interior is lost in the process of generating the lateral boundary conditions and through relaxation zones where GCM data is passed into the RCM. We see that although MBC still shows the greatest improvement at the specified zone and adjacent grid cells to the relaxation zone, the effect of bias correction, which was almost perfectly corrected toward ERA5, immediately deteriorates after generation of the boundary conditions. This indicates that the multiple interpolations for the purpose of bias correction and the model configuration may introduce additional bias in the creation of the lateral and lower boundary conditions. These are generally unavoidable limitations of a regional numerical model that has boundaries where the model could suffer from changes in resolution with differing parameterizations and physics schemes derived from the GCM, and further assessment is needed to address such issues (Warner et al. 1997;Wu et al. 2005).
Can the sophisticated bias correction of the GCM outputs effectively improve the RCM rainfall simulation?
Improvements in the model performance relative to simple bias correction techniques, which do not account for the dependence structure, are shown in the RCM(NBC) and RCM(MBC) outputs over the Australian CORDEX domain, particularly for precipitation. We see that the RCM performs better in capturing the rainfall coefficient of variation and standard deviation (Figs. 5 and 6) with the increasing complexity of the bias correction techniques. In contrast, RCM with uncorrected boundary conditions introduces a large bias across the variables used in this study. Although RCM with complex bias-corrected boundary conditions, RCM(NBC) and RCM(MBC), show degradation in comparison with RCM(MSD) for precipitation in SON, they broadly perform better at different time scales. We see that the RCM simulations tend to produce more bias in the ocean than inland areas. RCMs with simple bias correction, RCM(M) and RCM(MSD), present larger bias in rainfall climatological mean and variability in the ocean, indicating that they may be weak to capture the seasonal SST variation affected by a large-scale circulation. (Table A1 in the online supplemental material). Previous studies have shown that complex correction techniques affect the model's internal process and can correct details for rainfall extremes as well as its magnitude (Kim et al. 2020). This work represents that the effectiveness of increasingly sophisticated techniques, which aim to correct temporal persistence and physical consistency, adds improvement to rainfall characteristics, particularly in the simulation of rainfall variability and extremes.
What are the advantages of multivariate bias correction over the univariate bias correction?
This work focuses on the inter-variable relationships in RCM input variables and the expected improvement in simulations by correcting multivariate dependence at the boundaries. A combination of interacting physical processes across multiple climate variables, such as relative humidity, temperature, and wind speed, often leads to floods, wildfires, heatwaves, droughts, and even compound events (Zscheischler et al. 2018;Kim et al. 2023). Poor representation of these dependencies makes them challenging to foresee, leading to a significant impact on the risk assessment. These implications could be understood through the results presented in this study. Although simple techniques show comparable performance to the complex ones in the mean-field, they produce large biases in the tropics and the Indian Ocean.
Although RCM(NBC) shows good performance in the attributes used here (Fig. 1), it should be noted that more complex techniques that do not take into account inter-variable correlations do not guarantee better performance for extreme events. This identifies the importance of multivariate bias correction, especially when persistence biases are accounted for. It is observed that RCM(MBC) better preserves the physical relationships between the atmospheric variables (Table 1) in the specified zone and after passing through the relaxation zone, generally showing a better representation of rainfall variability and extreme events (as shown in Figs. 5, 6, and 7).
Limitations and future work
Several limitations may influence the results presented here. While this work has used newer models to assess model performance with several bias correction approaches, we only focused on one RCM (WRF) driven by one GCM (ACCESS-ESM1.5). Hence, coupling different GCM and RCM may change the results. Based on previous work that investigated the models' performance for CORDEX dynamical downscaling, even the updated models still contain biases (Di Virgilio et al. 2022). This indicates that the impact of the bias correction on the RCM boundaries is not significantly different if alternate GCMs are used.
The RCM used in this study was WRF4.2.1 which has complex internal dynamics, alternative options and an elaborate setup. Although the parameterizations for the WRF simulations were chosen following past evaluations, different setups would affect the model outputs.
In this study, the bias corrections were applied towards ERA5 fields, meaning that the ERA5-driven WRF outputs were used as ideal cases to assess the model performance against. Although reanalysis produces the most accurate outputs overall compared to the in-situ data (Kim et al. 2020), the model performance can be limited by the bias that comes from the reanalysis data, in particular for precipitation (Moalafhi et al. 2016(Moalafhi et al. , 2017. It should also be noted that specific humidity at higher levels where it is always close to zero was ignored during the multivariate bias correction as it makes the models unstable and creates physically unrealistic values. The parameters in the correlation structure, for example, including the lag1 auto-and lag0 cross-correlation in a contemporaneous model, are applied to a normalized variable. These changes, even if minor, influence the original value that was very close to zero to become unrealistic. The sea surface temperature was also corrected by nested bias correction only for RCM(MBC) as it shows high seasonal and annual variability but relatively much lower daily variability than the atmospheric variables. This large difference in variability between the daily and the longer time scales can cause unrealistic monthly variation during the multivariate correction.
Furthermore, the structure of the bias correction model makes implicit assumptions where biases are significant. For instance, correcting bias at daily and longer time scales implicitly assumes that diurnal patterns are simulated accurately. While these assumptions may not impact the overall results, there can be situations where their implications become important.
As mentioned above, for a better understanding of the underlying physical consistency and interaction between the variables under climate change, this study can be extended to future simulations. We note that the performance of MBC applied to future climate projections will be affected by any non-stationarity in the inter-variable correlation Guo et al. 2019Guo et al. , 2020Nguyen et al. 2020). Future work should investigate the impact of this effect. Additionally, it would be of interest to extend the bias correction approaches presented here to consider the spatial dependence across grid cells to reduce the uncertainty in the bias estimates (Kim et al. 2021;Switanek et al. 2022).
Conclusions
Despite the efforts to understand single drivers of extremes, most major weather and climate-related catastrophes are caused by a joint occurrence of different types of extreme events (Zscheischler et al. 2018). Correcting single variables independently, as many previous studies have done under the assumption that there is no (or negligible) bias in the dependence structure, limits the model performance in the simulation of extreme events. Although several studies have used sophisticated approaches that also corrected lag1 auto-and lag0 cross-correlation across variables to deal with inter-variable relationships, only the surface variables of the model outputs, such as precipitation and temperature, have been addressed.
This study investigated whether correcting the RCM input boundary conditions could reduce bias inside the domain. Several bias correction techniques, from simple scaling, which has been used in previous studies, to sophisticated techniques, which can correct persistence and physical relationships between the variables compared to the reanalysis data, have been applied. The corrections were applied to all the vertical levels of GCM based on four statistics: mean, standard deviation, lag1 auto-correlation, and lag0 cross-correlation. Substantial improvements in model performance are shown after applying bias correction to the boundaries of the RCM in the statistics used here. This work shows that the effectiveness of increasingly sophisticated techniques substantially improves rainfall characteristics. The RCM with multivariate bias-corrected input boundary conditions represents extreme events better than univariate bias correction techniques, which do not account for the physical relationship between the variables, and considerably better than the approach of using uncorrected GCM simulations to serve as the lateral or lower boundary inputs into an RCM. | 2023-03-12T15:07:57.640Z | 2023-03-10T00:00:00.000 | {
"year": 2023,
"sha1": "6f24def9873d7917575f96ccc3e22641fd057251",
"oa_license": "CCBY",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s00382-023-06718-6.pdf",
"oa_status": "HYBRID",
"pdf_src": "Springer",
"pdf_hash": "39330563ab5cf12794965cb8ad894408f84d0e49",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": []
} |
53473582 | pes2o/s2orc | v3-fos-license | Non-surgical Management of Endodontic Periradicular Pathosis-A Clinical Research
Background: In most cases the aetiological factors of periradicular diseases are oral contaminants through the root canal or degenerating pulpal tissues. Therefore, the mere surgical removal of the periapical lesions without proper root canal disinfection and obturation will not result in the healing of the periradicular tissues. On the other hand, traditional surgical technique rather fearful and troublesome job due to various reasons. So successful apical and periapical repair depends on conventional root canal treatment-a non surgical procedure. Crucial to this management and ultimate success is the complete debridement of the root canal system, followed by three-dimensional obturation to seal both the apical foramen and coronal orifice. Objective: Thus, the purpose of this study was to clinically verify the possibility of management of periradicular pathosis by non-surgical conventional root canal therapy. Methods: The present study was a prospective observational study carried out in the department of Conservative Dentistry and Endodontics, Faculty of Dentistry, Bangabandhu Sheikh Mujib Medical University (BSMMU), Shahbag, Dhaka. Here, 75 cases of eondodontically involved symptomatic non-vital teeth having periradicular pathosis were managed by conventional root canal therapy. Result: After 18 months follow up with post operative clinical and radiological evaluations, in this study the final outcome was favourable-88%, doubtful-08 %, and failure-04%.
by plasma cells, lymphocytes , mononuclear phagocytes and occasional neutrophils.Occasionally cholesterol clefting is seen, as are foreign-body giant cells 3 .If, in addition, adjacent strands of epithelium or rests of Malassez have been stimulated by the inflammatory response to form a stratified squamous epithelium-lined cavity filled with fluid or semisolid material, a cyst may be considered to be developed.There are two distinct categories of radicular cysts-one containing cavities completely enclosed in epithelial lining does not communicate with the root canal known as true radicular cysts and another containing epithelium-lined cavities that are open to the root canals 4,5 known as "bay" 4 or "pocket" 5 cysts.
Previously, teeth having periradicular pathology were managed by extraction or by surgical means (apisectomy) 6 .But recent development of endodontics indicate that teeth having periradicular pathosis can be managed by conventional root canal therapy.As long as there is passage of irritants from the root canal system, or a failure of the phagocytic macrophage system to control this irritation, simultaneous destruction and repair of periradicular tissues continued 3 .This concept relates to the rationale for non surgical root canal treatment as opposed to surgical intervention.The mere surgical removal, via curettage, of the reactive periradicular tissues will not address the source of the patient's problem, i.e. the removal of irritants from within the canal.As bacteria present in the root canal system play a fundamental role on the development and persistence of the periradicular lesions, successful treatment depends on effective removal of the irritants (inflamed or necrotic tissue) from the root canal system, followed by development of a fluid tight seal at the apical foramen, and total obliteration of the root canal system so healing and repair can take place 2,7,8 .The healing of periradicular tissues after root canal treatment is often associated with granulation tissue formation and maturation 2 .The granulation tissue is infiltrated by neutrophils, lymphocytes and plasma cells, which are responsible for removal endogenous and exogenous irritants, followed by fibroblastic proliferation and collagen deposition.On the periphery of the granulation tissue, osteoblasts and osteoclasts abound.With maturation, the number of cells decreases, whereas collagen increases.Ultimately, mature bone forms from the periphery toward the center 9 and cemental apposition if root resorption was present.Integrated within these events are complex cellular and immunological phenomena designed to eradicate all debris and foreign material that is counterproductive to tissue repair.When granuloma is present the course of healing is uneventful and generally follows the events described 10 .The healing of abscess may be slower; the exudates and bacteria must be cleared from the tissues before regeneration occurs.A variation of the abscess, the sinus tract (intraoral and extraoral) will heal following root canal treatment 11 .As the aim of root canal treatment is the elimination of infectious agents from the root canal and prevention of reinfection by obturation, a periapical pocket cyst is, therefore, likely to heal after this conventional endodontic therapy 12 .As tissue dynamics of a true cysts is selfsustaining by virtue of its independence of the presence or absence of irritants in the root canal.Therefore, the true cysts, particularly the larger ones containing cholesterol crystal, are less likely to resolve following root canal treatment 12,13 .True radicular cysts is less than 10% of all periapical lesions.In fact, most of the cases in which apical surgery have been performed based on radiographic diagnosis of the presence of cysts might have resolved by conventional root canal therapy.
Thus, the purpose of this study was to clinically verify the possibility of management of periradicular pathosis by non-surgical conventional root canal therapy.
Materials and method:
This was a prospective observational study carried out in the department of Conservative Dentistry and Endodontics, Faculty of Dentistry, Bangabandhu Sheikh Mujib Medical University (BSMMU), Shahbag, Dhaka.A total 75 patients of endodontically involved infected teeth, with periradicular pathosis irrespective of age, sex and tooth numbers who fulfilled the inclusion criterias and gave consent regarding the study were selected for this study.Patients included have endodontically involved tooth with periradicular radiolucency upto 10 mm, chronic periapical periodontitis, acute exacerbation of chronic periapical lesions, intra oral swelling with or without discharging sinus in nonvital tooth with periapical pathosis, discolored non vital tooth with radiological evidence of periradicular pathosis.Patients with endodontically infected teeth with periradicular pathosis but not included in the study when periapical radiolucency greater than 10 mm, nonrestorable tooth, patients having degenerative bone diseases such as diabetes mellitus, hyperparathyroidism.All the patients were thoroughly interrogated about their occupation, socio economic condition, general health status, drug and past dental history.Detailed clinical, radiological examinations and pulp vitality tests were done to confirm the diagnosis.Patients symptoms, clinical signs and radiographic evidences of periapical pathosis before treatment were recorded.The presence of periradicular lesion was determined by examination of x-ray with standard magnification.Widening of the periodontal ligament space by more than 1 mm on at least one root was necessary to categorize the tooth having periradicular lesion.All teeth having straight-canals were instrumented with H-files, Kfiles by conventional method and curved canal by crown down technique.Then after proper biomechanical preparation of the root canal, irrigation was done with 1% NaOCI solution for both the groups.All cases were treated using calcium hydroxide paste as the intracanal medicament and gutta-percha points and calcium hydroxide containing root canal sealer (Sealapex) were used for obturation of root canals.All cases were evaluated after completion of treatment at 6 month, 12 month and at 18 month.Results of the investigations were recorded and analyzed.Success, failure and doubtful were considered on evaluation criteria.A case was considered successful clinically by absence of pain, no tenderness on percussion and disappearance of swelling, sinus tract.Radiologically success was considered by disappearance or reduction of periapical lesion.On the other hand clinically failure was determined when there was no remission of symptoms that 65 means persistence of pain, tenderness on percussion, swelling and sinus tract.Whereas radologically failure was considered when there was increase size of periapical lesion and development of external resorption.A case was evaluated doubtful when all other sign symptoms disappeared but size of radiolucency was stagnant.
Results:
The study involved 75 cases of endodontically involved teeth with periradicular pathosis irrespective of age, sex and tooth numbers.Among these, 70(93.33%)cases were symptomatic presenting pain, swelling, discharging sinus and tenderness to percussion while 05(6.67%)cases were asymptomatic.(Table I)
Case-2
Pre-operative X-ray Working length measuring X-ray
Case-3
Pre-operative X-ray Working length measuring X-ray
Post operative X-ray with periradicular lesion
Disappearance of peri-radicular lesion after 18 th months
Fig-3: Peri-radicular pathosis managed by RCT (case-3).
Non-surgical Management of Endodontic Periradicular Pathosis Tahmeena Ishrat Ahmed & Mujibur Rahman Howlader 67 Discussion: The present study was undertaken to evaluate postoperative clinical and radiological presentations after 6,12,18 months in the management of periradicular pathosis by non-surgical conventional root canal therapy, which have been compared to those in other similar type study reports from the different parts of the world.
Removal of irritants from the root canal system and its total obturation result in repair of inflamed periradicular tissue 8 , Depending on the extent of tissue damage, repair varies from a simple reduction and resolution of inflammation to a more complex regeneration, involving remodeling of bone, periodontal ligament and cementum.Repair of the lesion, thefore, may take days to years.
Case-4
Pre-operative X-ray Working length measuring X-ray
Case-5
Pre-operative X-ray Working length measuring X-ray
Post operative X-ray with periradicular lesion
Disappearance of peri-radicular lesion after18 th months
Case-6
Pre-operative X-ray Working length measuring X-ray
In this study after 18 months clinical and radiological evaluation showed endodontic treatment outcome was favorable in 88%, doubtful in 08% and a failure in 04%.
In a similar type of study Suchina JA et al 6 found in clinical and radiological evaluation endodontic treatment outcome was successful in 88% and 80%, questionable in 10% and 15%, failure in 2% and 5% respectively.In another study by Caliskan 14 and Sen BH 15 , a total of 172 mature teeth with periapical radiolucency with or without symptoms were treated endodontically using calcium hydroxide paste as the intracanal medicament and a calcium hydroxide containing root canal sealer was used.The long term clinical and radiological outcome of this non-surgical root canal therapy, the complete healing rate was 80.8% while incomplete healing had taken place in 7.6% of the cases.
The change in size of periapical radiolucency in the followup visits was taken as the criteria for success or failure of root canal treatment.In this study, radiological evaluation of periradicular pathosis after 18 months in 52% cases radiolucency was disappeared, in 36% cases had reduced radiolucency, in 04% cases had increased radiolucency and 08% cases had no change in radiolucency.Similar type results were also observed by Katebzadeh N et al 16 .in a study of radiographic evaluation of periradicular healing after obturation of infected root canals.This report was supported by Peter LI & Wesselink PR 17 .and Kalaskar R et al 18 .In a study by Oztan 19 it was found that large periapical lesions can respond favourably to non-surgical root canal treatment.This result agreed with those reported by Soares et al 20 In most cases, the aetiological factors of periradicular diseases are oral contaminants through the root canal or degenerating pulpal tissues.Therefore, the mere surgical removal of the lesions of pulpal origin without removing the causes through proper cleaning, shaping and obturation, will not result in the healing of the periradicular tissues.In addition to removing effectively the causative factors, enviromental changes in the periradicular tissues caused by the process of canal debridement may inherently aid in the demise of epithelium, when present.So a success rate of 80% to 90% for management of periradicular pathosis by nonsurgical root canal therapy has been recorded by many practitioners and endodontic investigators 1,22,23 .Success, failure and doubtful were considered on evaluation criteria.The present study showed doubtful outcome may possibly be due to lack of proper biomechanical preparation or failure to achieve hermatic seal of root canal system.Also may possibly be due to the comparatively short observation time used.The 18 months follow-up may not have been enough for large lesion to heal.Histologically some areas of periradicular periodontitis need 4 to 5 years to heal 24 .
Conclusion
On the basis of the results obtained, it may be concluded that endodontically involved teeth with periradicular pathosis (e.g.granuloma, cysts etc) can be managed by non-surgical conventional root canal therapy.And it is the proper biomechanical preparation and hermetic obturation which decide the success of the root canal treatment.
Though the study performed clinically and radiologically, a histological evaluation with a larger sample size and longer follow up period is necessary to reach a sound conclusions.
Table - I
Preoperative clinical findings of the study subjects (n=75) III ).
Overall outcome of the study subjects after 18 months, based on clinical and radiological evaluation, favourable outcome was 66(88%), doubtful 06(08%) and failure 03(04%), indicate that non-surgical procedure by conventional root canal therapy is effective in the management of periradicular pathosis.(Table-V).
21Maalouf and Gutmann21.According to Maalouf and Gutmann21, Differential radiographic criteria are unreliable for the histological diagnosis of periradicular lesions of pulpal origin.It should not, therefore, be assumed that large lesions or lesions with an opaque border are cystic. Evn if cystic, or if strands of epithelium are present, repair following the removal of the source of the pathosis. | 2018-10-19T06:09:34.935Z | 2016-11-11T00:00:00.000 | {
"year": 2016,
"sha1": "20a48dc2323d56de3f791f378e93169556686863",
"oa_license": "CCBY",
"oa_url": "https://www.banglajol.info/index.php/BSMMUJ/article/download/3694/3041",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "20a48dc2323d56de3f791f378e93169556686863",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
219471487 | pes2o/s2orc | v3-fos-license | Building a model for recognition of morphostructure pathologies in animal tissues
Microscopic images of histological analysis have a complex flat structure. The result of the analysis and interpretation of the diagnosis strongly depends on the qualification of the veterinarian; therefore, the development of methods and models for automated histological diagnosis of diseases is an urgent task. The article considers the known approaches for building expert systems of histological diagnostics. Expert systems of histological analysis can be used in clinical veterinary medicine for advanced training of veterinary pathologists. The possibility of using the theory of finite automata for construction of the device - histological analyzer is shown. As an example of histological analysis the task of recognition of images of morphofunctional changes in rodent tissues on the example of the soft brain of a rat damaged by chlamydia infection is considered. Pathological indicators of morphostructural changes in tissues in case of chlamydia infection have been revealed and explained: erythrocyte slide with clot formation, nucleihyperchrome, tissue edema, hemorrhage and desquamation of endotheliocytes. For image recognition it is suggested to use computer neural network technologies based on multilayer perceptron. The number of synoptic scales for perceptron design has been calculated. Examples are given and construction of training sets on the basis of which the table of truth for neural network training constructed is shown. Computer experiments have shown the possibility of using neural network technologies for recognition of pathology indicators during histological analysis of morphostructural changes in animal tissues.
Introduction
The work is devoted to automating the recognition of histological analysis images, widely used in veterinary medicine, to detect pathologies at the cellular level. Pattern recognition requires preliminary processing of the examined tissue, obtaining samples and their analysis. Histological diagnostics is a complex of labor-intensive operations, including analysis of clinical data, analysis of micro-and macropreparation, and setting up histological analysis. Development of an automated workplace of a histologist veterinarian will help to solve the problem of rapid diagnosis of the disease and take timely measures for treatment [1].
Equipment and devices used in studies
Theoretical studies in the field of pathology recognition have been performed using approaches of general systems theory, system analysis and intelligent mathematical models. The following equipment was used for laboratory histological studies: technical and analytical scales, pH-meter, microtomes (sledge, rotary, freezing), cryostat, water bath, table for melting wax slices, set of automatic pipettes, thermostat, refrigerator and microscope.
General scheme for finding pathology indicators
To automate the process of determining the diagnosis of the disease it is necessary to classify the histological analysis images. The task is to define the resulting image, represented by a set of features, to one or more predefined classes. Currently, for these purposes, wavelet transformation, which converts analogue image into digital frequency domain [2,3], micro and macro diagnostics of histological analysis, realized in diagnostic complex "Atlant" [4,5], computer neural network technologies, end machines [6] and others are widely used. Let us consider the task of recognizing images of morphofunctional changes in rodent tissues on the example of the soft brain of a rat damaged by chlamydia infection (figure 1). Let us introduce a set of indicators Y(y1, y2, … , yn) diagnosing the disease D(d1, d2, …, dm). A simplified set of indicators specific to the pathology of the rat pia materdamaged by chlamydia infection is presented as an array described in table 1. Functional connection of Y indicators and preliminary diagnoses D can be connected with the table of truth with the construction of logical functions and relay-contact circuits, as, for example, was described in the construction of the histological analyzer in the work [6]. The task of pattern recognition of pathological changes in tissues finds its application in oncology [7], histology [8] and other fields of medicine and veterinary medicine.
To detect indicators and the possibility of machine search it is necessary to form a topological map, for example, in the form of a mesh area. An important element in solving the problem is to determine the mesh topology. To form the mesh topology, we will take orthonormalized vectors, through which the input signals retina will be formed. Optimization of the retina size is also important for finding the required indicators. The general algorithm of setting and solving the task of recognizing indicators of pathological changes and making a diagnosis can be presented in the following diagram (figure 2).
Neural network construction and training
The use of neural network technologies can be offered for pathology indicators recognition. We will use the technique of drawing a neural network based on perceptron. We will determine the required number of synoptic weights A using the formula for elements with sigmoid nonlinearity: Input signal in case of 5×5 field will be a vector of 25 elements. Thus, the first layer of the retina will consist of 25 elements -neurons. The last YN layer will consist of five elementspathology indicators. We'll take a training PM set of 50 samples.
Then the number of synoptic weights A must satisfy the condition: A Let's take the number of neuro-like elements in the hidden layer equal to two, then the number of synoptic links between the first and the second layers will be: , 50 2 25 = and the number of links between the second and the third layers will be equal: .
The total number of synoptic links will be A=60, which satisfies the calculated condition. The constructed structure of a neural network with calculated synoptic links on the perceptron is shown in figure 4. Once the network has been trained, it can be used to recognize the pathology indicator. For a running network, it is possible to calculate the correlation coefficient and the mean square error of diagnosis recognition. As a result of the search the most probable indicator of the investigated pathology sample is determined by the coefficient of reliability of the obtained result.
Pathology recognition
Identification of pathology images was carried out on the developed computer program that implements the construction and training of the perceptron. Let us consider the recognition of indicator y1 (Sludge of erythrocytes with the formation of blood clots). To simplify the solution of the problem we will take a white background, and the indicators will be marked by dark areas. An example of a training set according to table 2 is shown in figure 5. The image shown in figure 6 has been created to check the program operation. The result of the image recognition was recognized as set а. ICMSIT Figure 6. The result of image recognition of the pathology.
The conducted experiments showed the possibility of using neural networks for recognition of pathology indicators in histological analysis.
Conclusion
The use of a neural network for the development of a module for automated recognition of indicators of morphological and structural changes in animal tissues is proposed. The module of automated recognition of pathology indicators is supposed to be introduced into the designed histological analyzer of disease diagnosis. For image recognition it is proposed to use neural networks on perceptron. The conducted experiments have shown the possibility of using neural network technologies for recognition of pathologies in histological analysis of morphological changes in animal tissues. | 2020-05-21T00:09:12.524Z | 2020-04-01T00:00:00.000 | {
"year": 2020,
"sha1": "f1b8a78049e7186963e868078142999c63642482",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1742-6596/1515/5/052005",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "08bbfbf6d8bff94066508864c4ddeade55c4ceb4",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Physics",
"Computer Science"
]
} |
229454243 | pes2o/s2orc | v3-fos-license | Features of urban greening for people with visual impairment in Vladivostok
The creation of an urban environment designed taking into account the needs of people with disabilities is currently one of the priority areas of state, social and urban planning policy, the practical results of which are aimed at ensuring equal opportunities with other citizens in various fields of work, life, and leisure. The objective is to identify the features of greening for blind and visually impaired people in the conditions of the city of Vladivostok. To achieve the goal, it is necessary to solve the following problems: to determine the assortment of ornamental plants and to identify compositional techniques of landscape design that ensure the possible use of non-visual receptors (auditory, olfactory, tactile, etc.) by blind people, and also, enhance the impact on the visual perception of the visually impaired when they orient in the environment. As a result of the study, the assortments of ornamental plants and compositional methods of greening for the conditions of the city of Vladivostok were identified, which facilitate the orientation of blind and visually impaired people in an urban environment and allow the formation of an accessible environment for people with visual impairments.
Introduction
At present, the importance of an environment accessible to various segments of the population, including people with different needs and opportunities, is growing. In this regard, issues of the formation of public urban spaces using the principles of universal design are becoming relevant. Most often, such approaches are considered from the perspective of architecture, however, greening also plays an important role in the formation of a favorable environment.
In Russian regulatory documents, the term "low-mobility groups of people" (close to English term "people with disabilities (PWD)") is used -people who have difficulty in independently moving, getting a service, necessary information, or orienting in the environment (Set of Rules 59.13330.2016 Accessibility of buildings and structures for people with disabilities. The updated version of Building Norms and Rules 35-01-2001. Available online: https://tiflocentre.ru/download/SP59-2016.pdf, paragraph 3.21). This definition covers all people, including those with the motor, sensory and mental disorders, as well as people of different ages, height, weight, with structural features of the body, etc. (Lazovskaya N. [1]). Most often, the literature highlights the features of creating a universal ICCATS 2020 IOP Conf. Series: Materials Science and Engineering 962 (2020) 032050 IOP Publishing doi:10.1088/1757-899X/962/3/032050 2 (accessible) environment for the category of people with mobility impairments, and much less oftenfor people with sensory impairments -with visual and hearing impairments.
The purpose of the study is to identify the features of greening for blind and visually impaired people using the example of Vladivostok. To achieve this goal, it is necessary to solve the following problems: to determine the assortment of ornamental plants and to identify compositional techniques of landscape design that ensure the possible use of non-visual receptors (auditory, olfactory, tactile, etc.) by blind people, and also enhance the impact on the visual perception of the visually impaired when they orient in the environment of Vladivostok.
The authors have collected, summarized and analyzed information from literary and online sources. The main information on theoretical developments was obtained from domestic and foreign scientific journals, including those presented on the electronic library portal and the ScienceDirect multidisciplinary platform from Elsevier. A review of information sources revealed the state of scientific development of the problem. In this work, the authors relied on studies that examined the issues: the impact of urban public greening on quality of life, health, social interaction and inclusion (Lindemann-Matthies P. and H Brieger H. [2,3], Ward Thompson C. [4,5]); shaping the environment of urban public spaces based on the principles of universal design (Lazovskaya N. [1], Leontyeva E. [5], Kopeva A., Ivanova O. and Zaitseva T. [6]); creating an accessible environment for people with disabilities by landscape design of public urban spaces (Sholukh N [10]).
At the stage of posing the problem and determining its relationship with important scientific and social tasks related to the formation of public urban spaces for visually impaired groups of the population, an analysis of studies was conducted on the impact of the availability of urban green public spaces on sustainable planning and the well-being of various population groups (Norgate SH [11], Gupta K. and Roy A. [12], Fan P. and Xu L. [13], De la Barrera F. and Reyes-Paecke S. [14], Kabisch N. and Strohbach M. [15], Wüstemann H. and Kalisch D. [16]); the formation of sensory gardens and other recreational spaces for the visually impaired ( [20], Panova N. [21], Maidanov A. [22]. A review of contemporary literary sources showed that an integrated approach of creating an accessible environment for visually impaired people is used in some countries of the world (Hitter T. and Cantor M. (eds) [23]). Other researchers (Perry MA and Devan H. (eds) [24]; Hassan S. and Soltani K. (eds) (et al. [25]) note insufficient attention to this issue in their countries. Many studies show that living in green areas and caring for plants helps people with limited mobility, including the blind and visually impaired, develop various skills, regulate emotions and maintain self-confidence (Ruiz-Gallardo J. and Verde A. (eds) [26]; Skinner EA and Chi U. [27]; Passy R. [28]).
In order to determine the assortment of ornamental plants and to identify compositional techniques of landscape design that ensure the possible use of non-visual receptors (auditory, olfactory, tactile, etc.) by blind people, and also, enhance the impact on the visual perception of the visually impaired when they orient in the environment, a preliminary analysis was made for the assortment of ornamental plants and the results of the introduction of ornamental plants of the natural flora of the south of Primorsky Krai, as well as the principles of landscape design for the territories with limited use (
Methods
At the stage of determining the degree of knowledge on the problem, a method of systematization of theoretical sources (domestic and foreign scientific publications, including journals presented on the E-library portal and Elsevier's multidisciplinary ScienceDirect platform) was used. The main approaches to landscaping the territories based on the needs of the visually impaired were preceded by the study of ornamental plants in the collections of the Botanical Garden-Institute of the Far Eastern Branch of the Russian Academy of Sciences (Unique scientific setting (collection of living plants of the open ground). At the stage of collecting and studying the initial data, such methods as observation, photo-fixing, description of the functional and aesthetic characteristics of plants were used.
Results
As a result of the study, assortments of ornamental plants and compositional techniques of landscape design that ensure the possible use of non-visual receptors (auditory, olfactory, tactile, etc.) by blind people, as well as enhancing the impact on visual perception of visually impaired people were determined for the conditions of the city of Vladivostok.
Discussion
Of all the categories of people with disabilities, people with sensory disabilities -those with visual and hearing impairments -are closest to "standard" indicators by their anthropometric data. Creating an accessible environment for this category of PWD does not require making such significant changes in the main landscape elements, such as for people with a violation of the musculoskeletal system. Nevertheless, when developing compositional techniques of landscape design for the environment accessible for people with visual impairments, many features should be taken into account. Since visual impairments lead to significant disturbances in the perception of the world, the urban environment should maximize compensate for visual deprivation by the interaction of auditory, tactile, olfactory and other sensations, creating opportunities for the effective use of nonvisual receptors.
Disabled people with visual impairments belong to two types: blind and visually impaired. For blind people, tactile information is especially important, visually impaired people can navigate in space both with the help of tactile sensations and with the help of color and light contrasts (Leontyeva E. [5]). That is, an important aspect of the organization of landscape objects for people with visual impairments is the translation of visual images into images and a form that is accessible for tactile and/or sound perception (Dovganyuk A. [19]). It is noted in the literature (Sholukh N. and Nad'iarna A. (eds) [7]) that measures to facilitate the spatial orientation of the blind can be based on more active use of their unique compensatory abilities involving some effective natural landmarks from among certain breeds trees and other ornamental plants that are used in urban greening.
The use of plant material in the landscape design of an accessible environment should be based on the general principles of using plant material: taking into account the biological and environmental characteristics of plants, their decorative qualities, climatic and microclimatic conditions of sites for plant placement. Additionally, one should take into account the qualities of plants, which are of particular importance in creating a comfortable environment for people with visual impairments: the texture of branches, trunks, and leaves, the smell of plants, the color of the foliage and the color of flowers, etc. The use of planting types also has limitations. While the plant assortments used can be diverse, it is necessary to place decorative plants in homogeneous groups, dividing the compositions according to their decorative properties, not to mix their aromas, etc.
The basis of any landscape planting is trees and shrubs. When creating an accessible environment for visually impaired and blind people, the main selection criteria of the species will be the texture of the trunk and branches, a pleasant smell, and also the shape of the crown and the color of the leaves. Primorsky Krai is characterized by mixed coniferous-deciduous forests, the leading species of which are coniferous. A number of them can be used in landscape planting, giving them a peculiar natural flavor. Conifers are distinguished by a variety of forms and architectonics of crowns, the color of needles, and the texture of the bark. In addition, many conifers secrete volatile, inhibiting the growth of pathogenic bacteria. One of the most decorative Far Eastern breeds is Abies holophylla, it is characterized by a dense conical or oval-cone-shaped crown. Fir can be used for row plantings, creating groups, and as individual trees. It goes well with flat-leaved birch, maples, from shrubs -with Rhododendron, Philadelphus, Weigela. However, it should be remembered that in urban conditions, evergreen conifers often suffer from pollution, Dahurian larch (Larix gmelinii), the needle-like leaves of which change every year, can be used more widely. It is most suitable for creating compositional groups, it is especially good in autumn with bright yellow needle-like leaves. The brightly colored maple crowns will beautifully stand out against its background.
Of the variety of perspectives for creating green plantings from deciduous trees, when forming a favorable environment for the blind and visually impaired people with disabilities, beautifully flowering, with a characteristic color of foliage, with flowers with a pleasant smell, should be used. It can be Betula platyphylla, soft green foliage and a white-painted trunk of which will create a contrast against the dark crown of conifers. Brightly colored berries of the Sorbus amurensis (Sorbus aucuparia or Mountain-ash) help to complement the colorfulness of this composition and create a color accent.
Disabled people with visual impairments will get a variety of tactile sensations when they get acquainted with the characteristic bark of Phellodendron amurense (Amur cork tree), and from conifers -Abies holophylla and Pinus koraiensis (Korean pine).
Planting plants with a pronounced smell will help the blind and visually impaired people to navigate the space. So, groups of blooming Syringa vulgaris and Tilia amurensis will designate a specific area, however, it must be remembered that this sign will "work" only during the flowering of these plants.
The planting of shrubs will allow us to indicate the direction of movement and to highlight any territory. A hedge of flowering shrubs (Syringa amurensis, Forsythia, etc.) will mark the path, and the shrubs planted in a group with brightly colored foliage (Berberises, Spiraeas) will create an emphasis on the turn of the path.
One of the main factors affecting the orientation of the visually impaired in space is color; for them, brightness and color contrast are very important (Panova N. [21]). As a contrasting color, visually impaired people prefer bright yellow, bright orange and bright red, because they see this gamut of colors best. Yellow is the last color that people with minimal residual vision can see (Leontyeva E. [5]). Yellow color causes a feeling of light, the sun. As accents on a uniform green background of lawns, various compositions of annuals with yellow and orange-colored flowers (Eschscholzia, various species of Calendula and Tagetes) can be placed.
The red color is the brightest and most active color of the color gamut, plants with such a color of flowers can be used to create accents that attract attention. In the spring, you can create such an accent using bright red Tulips, in the summer -with Heucheras, Monards, in autumn the time comes for Salvia splendens. The pink color is softer; in summer, flower beds with the Astilbes, Phloxes, Paeonies against the background of the lawn will attract the attention. Expressive spots will be created by white flower beds; the Alyssum planted in such a flower bed will create not only white but also fragrant carpet.
When creating floral arrangements, it must be borne in mind that the colors of one gamut or alternating colors are viewed by visually impaired better than patterns and drawings. Therefore, choosing assortments for flower beds or flowers, preference should be given to plants with contrasting colors of flowers or foliage (Maidanov A. [22]). In one composition, plants with different leaf colors can be combined (Hostas with a white border of a leaf, yellow, bluish leaves, etc.).
In addition to color dominants, smells play a large role in perceiving the environment by people with visual impairments. There are many grassy ornamental plants for use in landscape planting, that possess not only decorative leaves, expressive flowers, but also their characteristic aroma. In the spring it will be Hyacinthuses and Narcissuses, which in the summer will be replaced by Tageteses and Petunias. The aromas of fragrant plants also have healing properties. So, the smell of Mentha (mint), for example, normalizes metabolic processes in humans.
Conclusions
A theoretical study allows us to conclude that when designing an accessible urban environment and, in particular, when choosing greening elements and compositional methods for combining them, it is ICCATS 2020 IOP Conf. Series: Materials Science and Engineering 962 (2020) 032050 IOP Publishing doi:10.1088/1757-899X/962/3/032050 5 necessary to take into account not only the visual qualities of the plants but also their possible impact on hearing, smell, tactile and other sensations to ensure the orientation of the blind and visually impaired people in urban space. Our assessment revealed the assortment of ornamental plants and compositional methods of greening for the conditions of the city of Vladivostok, contributing to the orientation of blind and visually impaired people in urban space and allowing the formation of an accessible environment for people with visual impairments. The assumption is made that the use of the identified assortment of plants and compositional methods of combining them in urban landscape design practices can significantly improve the orientation of people with visual impairments in the city space and, thereby, significantly increase the degree of accessibility of the urban environment for blind and visually impaired peoples. | 2020-11-19T09:13:41.001Z | 2020-11-18T00:00:00.000 | {
"year": 2020,
"sha1": "f88aa7d92427ab7b52dfb5d427ebf7723ec35a93",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1757-899x/962/3/032050",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "1d14b0414e75d5b19956286a70ea63bc508662fe",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Physics",
"Geography"
]
} |
258298114 | pes2o/s2orc | v3-fos-license | An Unsupervised STDP-based Spiking Neural Network Inspired By Biologically Plausible Learning Rules and Connections
The backpropagation algorithm has promoted the rapid development of deep learning, but it relies on a large amount of labeled data and still has a large gap with how humans learn. The human brain can quickly learn various conceptual knowledge in a self-organized and unsupervised manner, accomplished through coordinating various learning rules and structures in the human brain. Spike-timing-dependent plasticity (STDP) is a general learning rule in the brain, but spiking neural networks (SNNs) trained with STDP alone is inefficient and perform poorly. In this paper, taking inspiration from short-term synaptic plasticity, we design an adaptive synaptic filter and introduce the adaptive spiking threshold as the neuron plasticity to enrich the representation ability of SNNs. We also introduce an adaptive lateral inhibitory connection to adjust the spikes balance dynamically to help the network learn richer features. To speed up and stabilize the training of unsupervised spiking neural networks, we design a samples temporal batch STDP (STB-STDP), which updates weights based on multiple samples and moments. By integrating the above three adaptive mechanisms and STB-STDP, our model greatly accelerates the training of unsupervised spiking neural networks and improves the performance of unsupervised SNNs on complex tasks. Our model achieves the current state-of-the-art performance of unsupervised STDP-based SNNs in the MNIST and FashionMNIST datasets. Further, we tested on the more complex CIFAR10 dataset, and the results fully illustrate the superiority of our algorithm. Our model is also the first work to apply unsupervised STDP-based SNNs to CIFAR10. At the same time, in the small-sample learning scenario, it will far exceed the supervised ANN using the same structure.
Introduction
Simulating and designing a machine that thinks like a human is the ultimate goal of artificial intelligence. The vast majority of deep learning models rely on backpropagation algorithms, which require a large amount of labeled data to adjust parameters. However, obtaining labeled data is expensive. The backpropagation algorithm has a series of constraints, such as weight transport problem Lillicrap et al. (2016), and requires accurate gradient derivation, which is quite different from the learning process in the human brain. The human brain learns rapidly by relying on unsupervised local learning rules. Meanwhile, the traditional artificial neurons are far from the real spiking neurons which are rich in spatiotemporal dynamics Maass (1997). Spiking neurons receive input current and accumulate membrane potential, transmitting information through discrete spike sequences when the membrane potential exceeds the threshold. The spiking neural networks (SNNs) are more biologically plausible and energy efficient and have been widely used in various fields Fang et al. (2021); Zhao et al. (2022Zhao et al. ( , 2021.
Training an efficient and robust spiking neural network is a critical problem many researchers have been paying attention to. Due to the non-differentiable characteristics of the spiking neural network, it is challenging to directly use the backpropagation algorithm for training, which significantly restricts the development of the SNNs. Many researchers take inspiration from the learning process in the human brain and design some biologically plausible learning rules to train SNNs. The synaptic plasticity of neurons is the neurological Figure 1: The backbone of our model, which introduces the adaptive synaptic filter, the adaptive threshold balance, and the adaptive lateral inhibitory connection to improve the information transmission and feature extraction of STDP-based SNNs.
basis of learning and memory in the brain Bi and Poo (1998). Spike Timing Dependent Plasticity (STDP) is a common learning rule that exists in multiple areas of the brain and plays a vital role in the brain's perception and learning process. STDP influences the strength of synapses through the temporal relationship of pre-and postsynaptic spikes.
SNNs trained based on STDP still perform poorly due to the local optimization rule without global guided error compared with the backpropagation algorithm. This will lead to a lack of coordination and self-organization within and between layers of the model. Different parameter settings can easily lead to disordered spikes, making it challenging to transfer useful information. The human brain is not regulated by a single learning ruleAbbott and Regehr (2004). The brain dynamically coordinates multiple learning rules and connections for rapid learning and inference. In mammals, short-term synaptic plasticity (STP) is another essential learning rule. It lasts for a short time and adaptively controls the activity of different firing frequencies to regulate the information transmission better in a different layerZucker and Regehr (2002); Citri and Malenka (2008); Tauffer and Kumar (2021). Inspired by this, this paper designs an adaptive synaptic filter to help amplify the difference of the input current for better information transmission. Also, the adaptive spiking threshold is designed as the neuron plasticity to reduce the information loss during transmission. The adaptive lateral inhibition connection of different input samples is introduced to the spiking neurons of the same layer, which improves the self-organization ability of the model and enables the network to learn more abundant representations. Also, this paper extends the original STDP with sample temporal batched processing, which significantly accelerates the training process. To summarize, our key contributions are: 1. We propose the adaptive synaptic filter, adaptive lateral inhibitory connection, and the adaptive threshold balance to assist the training of unsupervised STDP-based SNNs, which significantly improves the representation ability of SNNs, alleviates the problem of repetitive features, and compensates for the input-output mismatch between layers. 2. We extend the original STDP by integrating multiple samples and different moments into a batch (STB-STDP), which significantly speeds up and stabilizes the training process. 3. Experimental results on MNIST and FashionMNIST show that our algorithm achieves the state-of-the-art performance of unsupervised spiking neural networks. At the same time, the performance of SNNs in complex scenarios is improved, allowing unsupervised SNNs to show excellent performance in CIFAR10 and small-sample training scenarios.
Related Work
The training of spiking neural networks is currently divided into three categories, conversion-based, backpropagation-based, and brain-inspired algorithm learning rules based.
By exploring the relationship between the spike activation and the artificial activation function, the real value of the artificial neural networks (ANNs) can be approximated to the average firing rates of SNNs. As a result, an alternative way is explored to add constraints on the weights of well-trained ANNs to convert them to SNNs ; Li et al. (2021); Han and Roy (2020). Although these converted methods make SNNs show excellent performance in more complex network structures and tasks, it does not fundamentally solve the training problems of SNNs. Other researchers introduce the surrogate gradient to make the backpropagation algorithm can be directed used in the training of SNN Lee et al. (2016); Wu et al. (2018Wu et al. ( , 2019; Shen et al. (2021). However, as said before, the backpropagation algorithm is implausible and far from how the brain learns.
Since STDP is a ubiquitous learning rule in the brain, many researchers trained spiking neural networks based on STDP. Querlioz et al. (2013) tried a two-layer fully connected SNN using a simplified unsupervised approach of STDP. used an unsupervised STDP method with two layers of activation and inhibition. Notably, despite two layers, only one has trainable parameters. Kheradpisheh et al. (2018) used hand-designed DoG filters for feature extraction, STDP to train convolutional layers, and SVM as the classifier. The convolution kernels of each layer are designed individually. Only the training in the intermediate convolutional layers is unsupervised. However, due to the local optimization property of STDP, it tends to perform poorly on deep networks, so many researchers have tried to introduce supervisory signals to guide STDP tuning based on the global feedback connections Zhao et al. (2020), equilibrium propagation Zhang et al. (2018), backpropagation Liu et al. (2021), and the dopamine-modulated Hao et al. (2020). Some methods combine STDP with backpropagation for hybrid training. Such as Liu et al. (2021) and Lee et al. (2018a), they both first performed STDP training to extract weights with better generalization. The training of supervised backpropagation is then performed to obtain better performance.
Lateral inhibition is usually used to help neurons achieve mutual competition mechanism Heitzler and Simpson (1991); Amari (1977); Blakemore et al. (1970). The lateral inhibition mechanism has been tried to be added to the training of spiking neural networks. tried to use a static lateral inhibition mechanism, so that the firing neurons can inhibit other non-spiking neurons by reducing the membrane potential. Cheng et al. (2020) help the spiking neural network to have stronger noise-robustness by introducing lateral inhibitory connections.
Neuron Model
The leaky integral-and-fire (LIF) neurons Dayan and Abbott (2005) are the most commonly used computational model in the SNNs. LIF neurons receive the pre-synaptic spikes as the input currents and accumulate them on the decayed membrane potential. When the membrane potential reaches the threshold, the neuron releases a spike with the membrane potential reset to the resting potential u reset . Here we set u reset = 0. The details are shown in Equation 1: where u is the membrane potential. u thresh is the threshold for this neuron. τ is the time constant. i is the input current. We denote i = j w ij s j . s is the spikes from pre-synaptic neuron. w ij is the strength of synapses. R is resistance.
In order to facilitate the calculation, we convert the differential formula into a discrete representation as shown in Equation 2, where C is capacitance, which we set equal to 1.
where u (t) is the membrane potential at the time t. u (t) thresh is the threshold for this neuron at the time t. i (t) is the input current at the time t.
STDP Algorithm
In this paper, we improved the commonly used unilateral STDP. For conventional STDP, as seen in Equation 3, the modification of weights is determined by the time interval of the pre-and post-synaptic spikes. The larger the time gap, the less correlated the two spikes are and the less affected the synaptic weights.
where ∆w j is the modification of the synapse j, W (∆t) is the STDP function. For unilateral STDP, focus only on the presynaptic spikes before the firing of the postsynaptic spikes, which makes the synapse strength continue to grow Lee et al. (2018a). So x offset is to determine whether the modification are potentiated or depressed.
For the efficient implementation, we take another form of STDP using eligibility traces Lee et al. (2018a); Izhikevich (2007); Zenke et al. (2015). As shown in the Equation 4 , x (t) trace accumulates presynaptic spikes and gradually decays over time.
Proposed Algorithms
To be more explicit about the problem we focus on, we built a onelayer convolutional spiking neural network and used the STDP algorithm, described in Equation 4 , to train in MNIST dataset until convergence. This convolutional layer consists of a 5x5 convolution kernel with 20 channels. As shown in Figure 2(a), we show the synaptic weights of the convolutional layers after training. It can be observed that a large number of repeated convolution kernels appear. The same convolution kernel features are boxed in the same color. Repeated convolution kernels will affect the effective feature representation Glorot and Bengio (2010).
On the other hand, neurons from different layers may also work in a disordered way. Since the algorithm lacks a global guided signal, the neuron can not judge whether it is a suitable firing rate. Therefore This will cause the adjacent next layer of neurons to receive too high or too low input current, which makes the last layer's neurons fire unstably. As shown in Figure 2 The membrane potential will take longer to build up if the input current is too small, leading to delays in the transmission of information. When the firing frequency is significant, the neuron fires nearly all the time, which will damage the effective information representation. Different parameter settings can easily lead to disordered spikes, making it challenging to transfer useful information.
To alleviate the above problems, we propose three adaptive algorithms. To address the problem of repetition of neuron features within a layer, we propose an Adaptive Lateral Inhibitory Connection (ALIC), different from static lateral inhibition , which provides a way to coordinate neurons by automatically selecting those that need to be inhibited. Next, we use Adaptive Threshold Balance (ATB) to solve the mismatching of input and output ranges between adjacent layers. In order to make the spikes firing more stable, we designed Adaptive Synaptic Filter (ASF) inspired by STP. Finally, we propose STB-STDP, which combines spatial and temporal information into a single batch.
Adaptive Lateral Inhibitory Connection
We introduce lateral inhibition to solve the problem of neurons in the same layer tending to have the same weights. Interaction between neurons is enabled by lateral inhibition, which allows firing neurons to dominate this input by inhibiting the firing of other neurons. The dominant neuron is more likely to experience learning according to STDP, and sensitivity to its input gradually increases. Conversely, non-dominant neurons will not be sensitive to this input. This prevents neurons from convergent towards the same weight. static lateral inhibition, however, is usually a constant set manually.
ij denotes the inhibition received by the neuron i, j. α is a parameter defining the degree of inhibition. The same inhibiting degree is maintained across all neurons and inputs. As a part of the network, the lateral inhibitory connections are static, with the same weights assigned to each neuron. However, this is not reasonable. Evidence from neuroscience suggests that for better inhibition, it does not exert inhibition on all neurons, but on those with relevant activity Linster et al. (2005); Kuffler (1953); Arevian et al. (2008). Thus, we expect a dynamic structure of lateral inhibitory connections to produce various structures for different inputs. Such a dynamic structure would help enhance coordination between neurons.
To this end, we introduce Adaptive Lateral Inhibitory Connection (ALIC), as shown in Figure 3 . We designed a dynamic structure of lateral inhibition. we determine which neurons may fire by setting a threshold and inhibit the membrane potential for these neurons. As shown in the Equation 6, we choose the maximum input current as the reference. Inhibition depends on maximum current for different inputs. Threshold is set with ij is the inhibition for the neuron at the position of i, j at the time t. α inh is a coefficient that adjusts the degree of inhibition. i (t) denotes the input current, which obtained by synaptic weight w ij and spikesŝ ij ) allows the inhibition to act only on neurons that are not firing at this moment. We adopt the winner-take-all strategy, randomly take out a firing neuron, and set the remaining spikes to 0,ŝ ij ). A more detailed analysis of section 5.3 shows that ALIC actually improves the performance of the network.
Adaptive Threshold Balance
By introducing an adaptive threshold method, we are able to eliminate the mismatch between input and output between layers. The input current varies from different inputs, leading to spikes firing variability. The current may be too small to reach the threshold, delaying the transmission of information and increasing the network delay. A large value may also be well above the threshold, increasing firing frequency. The portion of current above the threshold will be lost due to the reset of the membrane potential. Therefore, a dynamic threshold method is needed so that the threshold can be adaptively changed according to the magnitude of the current. Due to neurons' inherent plasticity, the adaptive threshold can reduce the loss during transmission and facilitate the expression of more precise informationWilent and Contreras (2005); Huang et al. (2016). We introduce an adaptive threshold balancing (ATB) mechanism.
The convolutional and fully connected layers play different roles in the network. The convolutional layer extracts features, while the fully connected layer exhibits feature selectivity sensitive to different features. Therefore, we employ variable threshold methods at layers. For convolutional layers, ATB set threshold positively related to the maximum input current, as shown in Equation 7. where u (t) thresh denotes the threshold of neuron at time t. β thresh is a parameter controlling the threshold scale. The maximum current is selected from the maximum input value of all neurons in the convolutional layer. c, w, h represents the number of channels in the convolutional kernel and the size of the output, respectively. The input current of each neuron is obtained from the synaptic weights and the input spikes. ATB ensures that no information is lost due to excessive current. Meanwhile, it allows spikes to be transmitted for a limited time.
For fully connected layers, ATB improves the method in . Where u t,j thresh denotes the threshold of the neuron j at time t. When a neuron fires a spike, we increase the threshold θ t,j plus , making the next firing more difficult. When θ t,j plus reaches γ, all thresholds reduce the difference between maximum threshold and γ. As shown in Equation 8 . Where θ init is the initial value. θ plus is the increment of u t,j thresh . α plus is a coefficient controlling the growth rate.
In this manner, a dynamic balance is established, where each neuron is measured for relative sensitivity. When γ is reached, the whole shifts downward. All thresholds are limited within a range. Thus, the difference between the thresholds is indicative of the neuron's sensitivity. Neurons with higher thresholds have lower sensitivity. This dynamic balance prevents a single neuron from dominating and ensures that each neuron has an opportunity to fire.
Adaptive Synaptic Filter
With adaptive threshold balancing (ATB), the threshold is adjusted within the convolutional layers in order to match it to the current magnitude. However, current regulation for individual neurons will still assist the network in achieving a more stable performance.
The short-term plasticity (STP) of synapses affects short-term information processing at synapses, as shown in Figure 3 δ asf is the function of ASF, where u (t) thresh denotes the threshold at time t. i (t) is the input current at time t. α asf and β asf are coefficients and control the function of the filter. The ASF adjusts the current through a nonlinear function, making the current more likely to concentrate near the threshold or resting potential. Concentrating current towards the threshold will result in more competition for the neuron. More competition will help avoid the dominance of neurons. And the current approach to resting potential will decrease the noise generated by low current strength neurons firing. Since ASF is calculated according to the threshold, ASF must be performed concurrently with ATB. We verify the effectiveness of ASF in section 5.3.
Adaptive Synaptic Filter
Adaptive Threshold Balance
Sample 1 Sample 2 Sample 3 Sample 4 (a) Adaptive Synaptic Filter (ASF) and Adpative Threshold Balance (ATB) (b) Short-tetrm Plasticity (c) Adaptive Lateral Inhibitory Connection (ALIC)
spikes affect strength of the synapse Figure 3: (a) The adaptive synaptic filter helps to regulate the inputs, and (b) the adaptive threshold balance helps to regulate the outputs. (c) The adaptive lateral inhibitory connection helps to suppress the same state to avoid learning repeat features.
STB-STDP
A majority of STDP-based SNN algorithms are trained with one sample at a time, with the weights being updated at every step of the training process. However, as with the backpropagation algorithm, combining a set of samples into batches for training can reduce the convergence instability caused by deviations in the distribution of each sample. Due to the batch processing of the samples, the convergence speed of the model is also significantly increased. In addition, SNNs exhibit multiple forward propagation processes in the temporal dimension, in contrast to artificial neural networks. By aggregating multiple time steps, the network can converge more smoothly instead of updating at every step.
From this, we propose a sample-temporal batches STDP algorithm (STB-STDP). A weight update is done by integrating the information of samples and temporal information simultaneously. As the Equation 10 shows: where N batch is the batchsize of the input, T batch is the batchsize of time step. Weights are normalized after updating, preventing them from diverging or shifting. As shown in Equation 11. Different normalization methods are used for convolutional layers and fully connected layers. A − f c , A − conv is scale factor to control normalization. Additionally, a spike normalization module is added between each layer to stabilize the inputs and outputs, which limits the input range of the spike. This module captures spikes and outputs a firing frequency of up to one.
Experiments
To demonstrate the effectiveness of our model, we conduct experiments on the commonly used datasets, MNIST LeCun et al. (1998) and Fash-ionMNIST Xiao et al. (2017), CIFAR10 Krizhevsky et al. (2009). All the experiments are based on the structure consisting of a convolutional layer followed by a 2*2 max pooling layer, a spiking normalization layer, and a fully connected layer. Since our model is an unsupervised network, we adopt the same voting strategy as in of the output of the final layer for category prediction. And the parameters of the network are trained layer-wisely. Where A − f c = 0.01, A − conv = 1,α inh = 1.625, θ init = 10, α sfa = 0.4, β sfa = 8, α plus = 0.001, λ = 0.99, β thresh = 1, x offset = 0.3.
Experimental Results
MNIST is a digital handwriting recognition dataset widely used as a benchmark for evaluating model performance in recognition and classification tasks. The dataset contains a total of 60,000 training set samples and 10,000 test set samples. The size of each sample is 28*28 pixels. In the experiments, the examples in MNIST were normalized using direct encoding and without any form of data augmentation and preprocessing. For the MNIST dataset, we set the kernel size of the convolutional layer to 5 with 12 channels, and 6400 neurons in the fully connected layer. Timestep is 300 in our experiments. To verify the superiority of our model, we compare the results with other famous STDP-based SNN models. Un-&Supervised denotes the former layer is trained unsupervised, while the final decision layer is trained with supervised information. As shown in Table 1, Our model achieves 97.9% accuracy. Compared with , which only uses the STDP, our model improves by nearly 3%. Our model has surpassed all the unsupervised STDP-based SNNs and even some SNNs with supervised information. FashionMNIST is more complex than MNIST within cloths and shoes as the samples. Both the shape and the size of the data are the same as those obtained by MNIST. The kernel size for the convolutional layer is set at 3 with 64 channels, and 6400 neurons in the fully connected layer. The timestep is the same as that used in MNIST. As seen in Table 2, our model succeeded in achieving 87.0% accuracy and outperformed most STDP-based SNNs. Although the performance of GLSNN is higher than ours, it introduces global supervised connections, while our network has no supervision information.
Result on CIFAR10
Only a few STDP-based SNN algorithms are capable of classifying the FashionMNIST dataset and achieving impressive results. As far as we know, few unsupervised STDP-based models run experiments on more complex datasets, such as CIFAR10. Trying to maintain high performance on more complex datasets will be our exploration direction. First, we used RGB data and the image size is 32x32. We use a 5x5 convolution kernel with 3 input channels and 64 output channels in the convolutional layer. Meanwhile, 3200 neurons are used in the fully connected layer. After training, our model achieved 42.36% accuracy on the CIFAR10 dataset. We visualize the accuracy curves of our model with 3200, 6400, and 10240 neurons in the fully connected layer. The highest accuracy was obtained for the test set at the number of neurons of 3200.
Discussion
In order to determine what causes misclassified samples, we visualized the classification confusion matrix of our model on the MNIST and FashionMNIST datasets, respectively. As shown in the Figure 5. There is not much distinction between the categories on MNIST's matrix. However, FashionMNIST's matrix showed lower accuracy on four categories of 0,2,4,6, corresponding to 't-shirt', 'pullover', 'coat', and 'shirt', respectively. There is a high degree of similarity between images in these four categories, which makes it more difficult for the model to classify them accurately. We explore the impact of different parameter settings on network performance. In convolutional layers, we changed the number of kernels, while in fully connected layers, we changed the number of neurons. As shown in Figure 5. The different number of channels and the number of neurons in the fully connected layer are combined, respectively. The height of the histogram represents the final accuracy of the model. There is a direct correlation between performance and the number of voting neurons set. Because more voting neurons will provide more predictions, combining all predictions, the network will produce more accurate results. However, this correlation does not exist for the setting of the number of convolution kernels. The best performance for MNIST is achieved when the number of kernels is set to 12.
In contrast, FashionMNIST requires more convolution kernels due to its complexity, and it works best when the number of kernels is set to 64. More convolution kernels cannot extract more features, because repeated features will appear in large numbers. In addition, the fully connected layer also needs more connections due to the increased number of kernels because more connections increase the complexity of the network and the difficulty of convergence.
To fully illustrate the impact of each module of our model on the results, we conduct ablation experiments on two datasets separately. We conducted experiments with the settings "w/o ASF", "w/o ASF and ALIC", and "w/o ASF, ALIC, and ATB", respectively. "w/o" denotes that we remove the relevant module. We show the convergence curve of the model on the test set, as shown in Figure 5. All experiments included STB-STDP to ensure the efficiency of the experiment. It would take a much longer time for the model to run without it. Adding each module can assist in improving the model's learning ability and its ability to achieve higher accuracy on the MNIST and FashionMNIST datasets.
To better illustrate the role of lateral inhibition, we added "softmax" for comparison. "softmax" is a commonly used function that can serve as a lateral inhibition. For the experiments, we removed the lateral inhibition module and performed the experiments on the MNIST dataset with the same model structure. The convolutional layer contains 12 5x5 convolutional kernels and 6400 neurons in the fully connected layer. We have trained the model with the same experimental setup. The model reached 28.26%. Then we added the softmax layer after the input currents in the convolutional layer and the fully connected layer. Also, to ensure that the currents are of the same scale, we multiplied by the maximum value of the current. After training, the model obtained 78.16%. This shows that the softmax has a certain effect. However, lateral inhibition can help the model to achieve the inhibition in a better way. In contrast to the backpropagation algorithm, our network trained with the STDP unsupervised algorithm is more adaptable to different number of samples, especially in small-sample task, and shows superior performance compared to supervised algorithms with the same structure. There are only a very small number of samples in the whole train set in small-samples task, which differs from few-shot learning. To this end, to illustrate the ability of our model to train on small samples tasks, a very small number of samples are randomly selected from the MNIST dataset. The number of samples per class ranges from 20 to 10 to 5, in the most extreme case, to 1. At the same time, we also designed an artificial neural network model with the same structure for comparison. The model consists of 5x5 convolutions with 12 channels, a maxpool layer, a fully connected layer of 6400 neurons, and a final fully connected classifier. The backpropagation algorithm is used to train this ANN model.
Experiment for Small Samples
As shown in the Table 3, our model has better performance than the ANN model. As the number of training samples gradually decreases, the performance gap between our model and ANN gradually increases. When there is only 1 training sample per class, our model outperforms ANN by 4.43%. It fully illustrates that our model requires only a small number of samples to achieve high performance compared to artificial neural networks that require a large amount of labeled training data.
Visualization
To illustrate the feature extraction capability of our model, we visualize the weights of different layers. Figure 6 a and b shows the weight of the convolutional layer on the MNIST and FashionMNIST dataset respectively. We randomly selected 10 neurons from the corresponding categories. The corresponding weights are visualized. Each of these rows shows one category, (d) Fully connected layer on the FashionMNIST dataset. We visualized a portion of these weights, where each row represents the weight of neurons corresponding to a category.
The convolutional kernels capture simple features such as edges, lines. With the introduction of our adaptive lateral inhibitory connections, our network does not have a large number of repeated features. Figure 6 c and d show the weight of the fully connected layer. According to the label assigned to the neuron, we visualize the weight of ten categories, and each row represents a category. It can be seen that the fully connected layer automatically combines the features of the convolutional layers to form higher-level semantic representations. For MNIST, with the combination of simple features, the different numbers can be easily classified. While FashionMNIST is more complex, it is not easy to distinguish similar objects such as Shirts and T-Shirt. In future work, we will consider introducing more biologically plausible rules to improve the performance of our model.
Conclusion
Spiking neural networks (SNNs) trained with STDP alone are inefficient and hardly achieve a high performance of SNN. In this paper, we design an adaptive synaptic filter and introduce the adaptive threshold balance to enrich the representation ability of SNNs. We also introduce an adaptive lateral inhibitory connection to help the network learn richer features. We design a samples temporal batch STDP (STB-STDP), which updates weights based on multiple samples and moments. By integrating the above three adaptive mechanisms and STB-STDP, we have achieved current state-ofthe-art performance for unsupervised STDP-trained SNNs on MNIST and FashionMNIST. Further, we tested on the more complex CIFAR10 dataset, and the results fully illustrate the superiority of our algorithm. Since we consider more of the unsupervised learning rules, it does not obtain significant improvement when it is extended to deep networks. In future work, we intend to consider more learning methods of the human brain, such as dopaminebased regulation of the reward mechanism. Also, We will introduce more brain-like structures, such as global feedback connections. | 2022-07-07T01:16:13.554Z | 2022-07-06T00:00:00.000 | {
"year": 2022,
"sha1": "ffa70f147756c642669ab5b5298c48296c2ba1ca",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "ffa70f147756c642669ab5b5298c48296c2ba1ca",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
231606076 | pes2o/s2orc | v3-fos-license | Coping with COVID-19 in an international border region: health and economy*
Objective: to analyze how the social isolation measures and closed borders affected the health and economy in an international border region. Method: descriptive cross-sectional study conducted in the western region of Paraná, Brazil, using an electronic form created using Google® forms. A sample of 2,510 people was addressed. Descriptive analysis and the Chi-square test were performed, with a level of significance established at 5%. This public opinion survey, addressing unidentified participants, is in accordance with Resolutions 466/2012 and 510/2016. Results: the participants were 41.5 years old on average, most were women and worked in the education sector; 41.9% reported that the closing of borders/commercial businesses negatively influenced income; 17.7% reported the possibility of losing their jobs; 89.0% consider that a larger number of people would be sick if the borders/commercial had not been closed; 63.7% believe the health services are not prepared to deal with the pandemic; 74.9% realize that the Brazilian Unified Health System may not have sufficient service capacity; 63.4% reported anxiety; and 75.6% of commercial workers will experience changes in their income level. Conclusion: the closing of international borders and commercial businesses was related to a perception of physical and mental changes, job loss, and decreased income.
growing speed, with variations depending on the regions, states, and cities. There was a significant increase in the number of cases and deaths, despite social isolation measures intended to contain the virus (1,5) , revealing the current health context is of concern. Brazil exceeded 2 million cases up to July 2020, and lethality reached 3.8% with an incidence of 1,031.8 cases/100,000 inhabitants and 38.8 deaths/100,000 inhabitants. There is potential underreporting, though, considering there is a lack of tests for the Brazilian population. Thus, the number of cases confirmed thus far is expected to be six times higher (1) .
A systematic review reports that because there is no vaccine or medication with robust in vivo scientific evidence thus far, social isolation is the primary strategy to decrease the number of new cases and deaths (6) in the population in general. Scientific evidence confirms that social isolation is efficient; however, it also decreases economic activity in general (5) . A combination of social isolation and the release of essential activities only, associated with individual preventive and control measures, such as the use of face masks and alcohol at 70%, present a more significant effect (5)(6) .
If, on the one hand, social isolation practices have been proven to be a good measure to contain the spread of COVID-19, on the other hand, it directly affects the economy at a global and local level. It is important to note that decreased industry and commercial activities and services, impact the businesses' health.
Results
The participants were 41.5 years old on average (SD±12.3) and had 1.4 children on average (SD±1.4).
Most were women (67.7%), married (57.1%), with children (69.4%), worked in the education sector, in public or private schools (19.6%), with statistically significant differences between professions (p<0.001), as shown in Table 1. as shown in Table 3. In the questions assessing the impact of social isolation, Table 4 shows that 63.5% of the respondents reported that social isolation caused anxiety, and 32.3% related the onset of body pain they had not previously experienced to social isolation (p<0.001).
Additionally, 7.8% of the respondents reported that social isolation caused the emergence of diseases they had not previously experienced (p<0.001).
The results show that workers from different sectors report that border closure will influence their family income, especially those in commercial/tourism businesses (75.6%; p<0.001). Even workers with some job stability reported income losses ( Table 5).
Analysis of the impact of closing commercial businesses shows that workers in the commercial/ tourism businesses and those in the Others category (i.e., homemakers, students, other occupations) were the most frequently affected, that is, 87% and 60.6%, respectively, reported that the measure influenced or will influence their families' income (p<0.001).
The highest impact of the closure of commercial businesses was reported by those working in commercial/ tourism businesses (87%) followed by 60.6% of those in the others category (i.e., homemakers, students, and other occupations) reported that this measure influenced or will influence their family income (p<0.001).
Whether the closing of borders was an appropriate measure to prevent the spread of SARS/CoV2/ COVID-19, those in the commercial/tourism businesses more frequently disagreed, 18.5% (Table 5). This percentage virtually double folds (36.1%) when this question specifically refers to the closing of commercial businesses (p<0.001). Rev. Latino-Am. Enfermagem 2021;29:e3398. Regarding the onset of physical symptoms, as verified in this study, it is essential to consider that Even though it is clear that the stay-at-home orders and the closure of borders and businesses will cause families' financial problems, the respondents confirmed there was a need to adopt these sanitary measures.
Nonetheless, the task of balancing the effects of social isolation on the economy and social determinants of health is complex (12,29) .
High-income countries rapidly allocated financial resources to support citizens and businesses as soon as they envisaged the economic threat posed by the COVID-19. The European Union and the USA announced the release of funds, based on stimulus bills, to protect people against economic impacts (29) . The Brazilian government acted in the same direction when it implemented economic measures to decrease the impact of the COVID-19 pandemic (14) . or illegal ways to obtain subsistence or ensure survival, whether to access essential goods or health services. | 2021-01-15T06:16:23.313Z | 2021-01-08T00:00:00.000 | {
"year": 2021,
"sha1": "73ac28a3607e76ea15da6af1e6afdc83e4155af9",
"oa_license": "CCBY",
"oa_url": "https://www.scielo.br/j/rlae/a/H4fqkQHNBCkrVqdDxtxyvWj/?format=pdf&lang=en",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7fe65ac555a344c4229ab6f22843da14adde5d65",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
236705284 | pes2o/s2orc | v3-fos-license | United Kingdom Anti-crisis Alternative
The article deals with the factors that ensure the sustainability of the reproduction of the British socio-economic model in the long term. Some features of the conceptual role of the UK in the formation of the global world order were studied. The impact of the COVID-19 on the economy of the United Kingdom was shown and the reasons for the deepest decline in production in the history of the country were summarized. The factors of chronic decline in productivity on the British labor market and strategic approaches to solving this problem were presented. Some provisions of the Trade and Cooperation Agreement between the UK and the European Union are discussed. The concept of sovereign money is analyzed as a possible response of British economic experts to the challenges of modern crises.
INTRODUCTION
The UK is the fifth largest economy in the world in terms of GDP [1] and the volume of exports [2] and it holds the first position in the world in terms of trade surplus in financial services [3, p. 8]. The United Kingdom is one of the least regulated economies [4] and is among the top five countries in terms of innovation development, attracting more foreign investment in R&D than many of its major competitors, including Germany, France and China [5]. In spite of such high ranking demonstration, the year 2020 in the country was marked by a record drop in production, caused by the extremely negative reactions of market structures to coronavirus and Brexit. Now the UK government is using an ambitious industrial strategy the goal of which is to transform the country into the most innovative economy in the world by 2027. It is quite probable that implementing the planned structural transformations and holding one of the top positions in the the global economy will force British politicians to turn to the time-tested experience of solving internal development tasks by working out innovative approaches to financial sector regulation and their popularization.
A SYNTHESIS OF LIBERALISM
AND DIRIGISME According to President of the Adam Smith Institute M. Pirie, the key to the success of the Anglo-Saxon model are its unique principles-adversarial principle and evolutionary principle [6]. Given the permanent influence of Great Britain on the global world order, it is difficult to overestimate the practical implementation of these principles. It is the United Kingdom that made a significant contribution into the formation of the international political system as well as the formation of the economic, financial, scientific, technological, informational and cultural image of the modern world.
It should be noted that maintaining internal political stability and influence in international relations was done at the cost of a firm combination of liberal and dirigistic methods in the regulation of economic activity [7]. For example, to prevent the destabilization of the economy after the collapse of the financial pyramid of the South Sea Company a law prohibiting the creation of new public joint stock companies was in action in England for a hundred years (1720-1825). Actively promoting the liberal ideas of Adam Smith, Britain was seen as the country with the most uncompromising protectionist policy towards the rest of the world until the middle of the 19th century [8]. After World War II, to compensate for the loss of the colonial income, the main sectors of the economy (including the Bank of England) were nationalized and # Aleksei V. Kuznetsov, Dr. Sci. (Econ.), is a Professor and Leading Researcher at the Financial University under the Government of the Russian Federation. extraordinary tax rates (reaching 98%) on the income of the richest individuals were introduced [9, p. 501]. At the same time British politicians worked to create the most favorable regime for the influx of foreign capital into the country. The result of this painstaking work was the emergence of the first world monetary system with a key role of the pound sterling and the emergence of the largest international financial centre in the City of London. Let us emphasize that after the United States had surpassed Great Britain in terms of foreign trade in 1880, the pound sterling retained its status as the leading reserve currency for more than half a century. Thus, in 1947 it accounted for more than 80% of international foreign exchange reserves [10, p. 10]-in other words, politics sometimes plays a more important role than the size of the economy, business and finance.
The collapse of the British Empire, completed by the Suez Crisis, required new schemes for extracting financial profits from the rest of the world. In the time of the Bretton Woods system crisis and the oil price shocks of the 1970s the symbol of financial liberalization was the Eurodollar market in London, which allowed banks to accept deposits and issue loans in foreign currencies, bypassing national laws and regulations. In the 1980s the policy of deregulation was continued by the Thatcher government, which extended neoliberal principles to stock trading, insurance and investment business, thus eliminating all possible obstacles to a large-scale inflow of capital into the British economy from other countries, primarily from the United States. In the 1990s the neoliberal paradigm was further developed within the framework of the Third Way doctrine, which actually legalized the offshore financial model of the global market with the key role of tax havens, the most significant of which were in direct or indirect British jurisdiction [11]. Thanks to these innovations, London has regained its right to be the largest international financial center. And to this day the British capital tops the list of the global cities that are the most favorable for the inflow of international investment [12, p. 7].
COVID-19 and Brexit brought an end to a halfcentury of neoliberal experiments that had let Britain to draw its energy from the outside world with virtually no limits. Obviously, the time has come when, in order to restore the economic dynamics, England will again have to turn to the time-tested methods of tight regulation of domestic economic activity.
IMPACT OF COVID-19 ON THE UK ECONOMY
The UK economy suffered immensely from the consequences of the coronavirus pandemic, which led to the deepest drop in GDP among the G7 countries, amounting to 21.2% in the first half of 2020. Such negative market reaction to the consequences of the pandemic can be explained by a long period of the govern-ment's restrictive measures (lockdown), the extraordinary role of the service sector in the structure of the national economy and the deep integration of British business into the world economy.
During the first Covid wave, the isolation regime was introduced in the UK later (which is also associated with an increased number of deaths from the virus) and for a longer time than in some other European countries. The impact of the Covid shock on the economy has been uneven. The biggest decline was observed in tourism, transport and entertainment. Industries that were largely able to continue operation, observing social distancing rules, such as financial services, energy and agriculture proved to be relatively well protected from the effects of the pandemic.
It should be noted that British economy is extremely dependent on the service sector, which accounts for about 80% of GDP. In October 2020, in 45 out of 51 service sectors, the level of production was lower than in February 2020. The worst-hit areas were aviation (-83.5%), entertainment industry (-58.4%), railways (-51.8%) and the hotel business (-40.8%). A similar situation was observed in 27 out of 44 manufacturing industries. Among those, the largest drop was registered in the maintenance of aircraft (-36.3%) and their production (-22.4%), the production of weapons and ammunition (-24.8%), and the production of coking coals and petroleum products (-22.0%) [13].
The automobile sector plays an important role in the UK manufacturing industry, accounting for over 8% of total industrial production. The automobile industry in the United Kingdom is deeply integrated into the global economy: in 2017, 80% of the total number of cars produced in the country were exported (compared to 77% in 2015) [14, p. 7]. The UK's automobile production in September 2020 was 5.0% lower than in the previous year, mostly due to a decline in shipments to the major foreign importing countries including China, the EU and the US [15].
In spite of a sharp decline in economic activity in the second quarter, employment remained at a high enough level. In October 2020, the unemployment rate was 4.9%, just 1.1% higher than a year earlier. Government grants, loans, tax holidays and benefits to business have helped to retain employees and avoid bankruptcies. The rapid rise in unemployment (which took place in the United States) was prevented, in part, thanks to the Coronavirus Job Retention Scheme (CJRS). In order to preserve jobs under this scheme, the government offered to compensate 80% of wages (but not more than £2500) to employees of companies sent on forced leave under the lockdown conditions. Also, under the Self-Employment Income Support Scheme (SEISS), self-employed people have been able to receive a taxable subsidy in the amount of up to 80% of their previous average income over the past three years.
The government allowed deferred tax payment of 40 billion pounds to legal entities. More than a million firms took loans against government guarantees, and for hotel companies the VAT was reduced from 20 to 5%. The Bank of England has cut the interest rate twice since March from 0.75 to 0.1%, and increased the purchase of corporate and government bonds within the framework of the "quantitative easing" policy to 895 billion pounds. Together, these measures helped to reduce the number of bankruptcies by almost a third compared to 2019. In spite of this support, the number of employees fell by 750000 from March to September 2020.
The introduction of the abovementioned government measures caused severe and growing damage to the national budget. The government allocated huge amounts of money to treat the coronavirus patients, control the spread of the coronavirus and mitigate its financial impact on households and businesses. The government's proactive fiscal measures can increase the budget deficit to 394 billion pounds (19% of GDP) and debt growth to 105% of GDP-their highest indicators, respectively, from 1944-1945 and 1959-1960 fiscal years [16, p. 5].
The sharp rise in the ratio of public debt to GDP is favored by the fact that the government takes loans for up to three years at negative interest rates and is able to issue debt obligations with a maturity of 50 years and a service cost of less than 0.5% per annum. Besides, most of the UK government debt has a long maturity. In March 2020, 42% of the debt portfolio liabilities had maturities over 15 years, while the average maturity of government bonds and treasury bills was 15 years. According to OECD experts, one of the measures to reduce the sensitivity of public debt to higher interest rates in the medium and long terms could be fixing a low interest rate on long-term loans to finance emergency spending [17, pp. 28-29].
POOR PRODUCTIVITY PROBLEM
Apart from the need to cope with the negative consequences of the Covid crisis, another key problem for the UK's economic development has been the chronic decline in labour productivity over the past decade, which is considered to be the deepest one since the start of the industrial revolution 250 years ago. British scholars tie the slowdown in productivity growth to the prolonged impact of the 2008 crisis on the financial system, poor achievements in computer technology in the recent years after their boom in the late 1990s and early 2000s, as well as the uncertainty of post-Brexit trade relations which undermine investment into business [18].
The slowdown in productivity growth is also characteristic for the United States and other developed economies, but in the UK this problem manifests itself most visibly: in terms of the growth rate of production per hour in the years 2008-2017 the country ranked 31st among 35 OECD countries. Given that the UK has a leading position in the ICT employment, where productivity growth is the highest, the decline in overall productivity looks paradoxical.
Among the leading British economists, there are two broad approaches to explaining this phenomenon. The first one focuses on supply factors, including the professional qualifications of employees. The second explanatory approach focuses on the demand factors associated with the financial crisis, austerity and other reasons for the slow growth in the demand in the past decade [19].
To drive productivity, the UK government since 2017 has been implementing the ambitious Industrial Strategy which should turn the UK into the most innovative economy in the world by 2027. Accordingly, it is planned to increase the total investment in R&D from 1.7 to 2.4% of GDP by 2027. By 2024-2025 a gross capital investment of about 640 billion pounds is planned (an average of 5.8% of GDP per year) [20]. The government is committed to becoming the first zero carbon emission economy in the world by 2050. It should be emphasized that the ambitious Industrial Strategy was adopted the following year after the UK referendum on EU membership (2016).
THE POST-BREXIT FUTURE
After four and a half years of tiring negotiations, the UK signed the Trade and Cooperation Agreement with the EU on December 24, 2020, which signaled the beginning of a new phase of trade relations in connection with the country's official withdrawal from the Union on January 31, 2020. We should remember that the first referendum on maintaining the membership took place as early as in 1975. Since then various movements and organizations have held Eurosceptical views. It was for these purposes that the UK Independence Party (UKIP) was created in the early 1990s, which later received a significant part of the funds from the sources close to the City of London [21, p. 145].
Among the reasons for dissatisfaction with the UK's membership in the EU the ones most often mentioned were: common trade, fisheries and agricultural policy, the inefficiency of the pan-European fiscal and monetary systems, social democratic approaches to the employment market, excessive business regulation, negative trade balance [22, pp. 57-64].
Analyzing the Agreement, we can conclude that it includes the fundamental issues that have caused the greatest resentment of British Eurosceptics so far. On the other hand, most of the questions regarding the future relationship between the UK and the EU remain open, requiring additional negotiations and agreements.
Thus, by its exit from the EU, the UK liberated itself from numerous pan-European restrictions and regulatory practices. At the same time, the country ceased to be a part of the Single European Marketthe "four freedoms" that ensure the unhindered movement of people, goods, services and capital within the single European space. The concept of the Single European Market was actively supported by the Thatcher government. Since its practical introduction in 1993, leading multinational companies, motivated by the elimination of tariff and non-tariff barriers, began to move their production facilities to the British Isles in order to have a guaranteed free access to the world's largest market of goods and services. Views were expressed that after Brexit, the common market model will cease to work [23]. However, the opposite happened-after 2016, the EU accelerated the conclusion of intercontinental trade and investment agreements, gradually including such significant regional players as Canada, Mexico, Mercosur, Japan, Singapore, Vietnam, Australia, New Zealand, and even China into the sphere of its economic interests.
It should be emphasized that the EU is the largest foreign economic partner of Great Britain-the volume of mutual trade between them in 2019 amounted to 668 billion pounds. At the same time, the UK has a negative trade balance with the EU, reaching a record 79 billion pounds. On the other hand, the UK registered a record positive balance of 51.7 billion pounds in trade with the United States though the trade turnover is almost three times less than with the EU (232 billion pounds) [24].
One of the main complaints of the United Kingdom at the time of the EU membership was that the latter was a customs union, not a free trade zone, which did not allow the UK to conclude profitable trade agreements with other countries, primarily the United States, on its own. As a Brexit supporter, Donald Trump promised that the United States would conclude a profitable trade deal with the UK after it leaves the European Union. However, President-elect J. Biden called Brexit a mistake and ruled out the possibility of new trade agreements with any country until, he said, the United States improves its competitiveness at home [25].
The UK is home to the largest number of headquarters and subsidiaries of the biggest multinational companies, including the leading European manufacturers of goods and services. In terms of the accumulated volume of incoming foreign direct investment (2.1 trillion dollars in 2019), the United Kingdom is ahead of all other EU countries, second only to the United States in this indicator [26]. Thus it is difficult to imagine that the formal withdrawal of the UK from the Single European Market would affect big business in the extremely negative way, given the enormous political influence of these companies in their countries.
As a result, neither party is interested in the worsening of trade relations. However, the Agreement does not look so unambiguous. For example, even though an agreement was reached on the tariff-free trade, but if one of the parties departs from common standards as of December 31, 2020, a dispute resolution mechanism (the Joint Partnership Council) in the form of new tariffs can be activated. At the same time, the European Court of Justice will no longer play a key role in monitoring the implementation of the Agreement. However, it will remain competent in Northern Ireland, which, in accordance with the terms of the Agreement, will continue to be subject to the rules of the EU's single market and customs union [27].
The agreement does not specify the future role of the City of London as the largest financial centre of Europe and one of the leading centers of the world. The UK plays a key role in the management of European capital, OTC derivatives and foreign exchange trading. Thus, the City of London clears 75% of eurodenominated financial derivatives equivalent to about 1 quadrillion dollars, and banks located in the UK manage 40% of European assets and 60% of European capital markets [28, pp. 82-83].
London provides global financial intermediation services to European investors and provides access to European markets to non-EU countries. In accordance with the Agreement, beginning from 2021 financial institutions located in the UK will lose the socalled "European passport," which gives the right to provide financial services anywhere in the European Union without any additional approvals from the host party. British companies' access to EU markets will now be governed by the principle of "equivalence" applied to third countries unless an additional agreement is signed between the United Kingdom and the European Union facilitating access to European financial markets for British companies. According to the British government's estimates, the loss of the European passport may increase the costs of British operators of financial services by 13% [29, p. 44]. Nevertheless, in the long run, the City of London will most probably remain attractive to European clients, given the advantages of the UK regulatory and institutional framework, including protection of the rights of creditors and shareholders, regulations on tax and on employment, as well as the historical volume and liquidity of the London market.
On the whole, it is now hard to predict the future of EU-UK relations judging by the Agreement alone. One thing is clear-Brexit pushed the EU to open its economic space to the rest the world, bringing closer the fulfilment of the British conservatives' dream -a global free trade zone [30, pp. 98-99]. It is beyond doubt that the year 2020 was a turning point not only for the UK, but for the entire world. Brexit and COVID-19 confirmed the complete obsolescence of the neoliberal paradigm of financial globalization.
The creation of a multipolar world on a new technological platform requires a new social contract, which seems impossible in the current configuration of the world monetary system based on credit money and American-centric offshore financial institutions.
As mentioned above, Britain has a unique experience in crisis management through the development and global distribution of innovative financial schemes that allow to stabilize the domestic economic situation with the resources of the rest of the world [31]. We cannot exclude that the post-neoliberal concept of financial sector regulation will also be based on the ideas of British experts. Let us look at one of these ideas in more detail.
SOVEREIGN MONEY CONCEPT
In Great Britain-the mother country of the modern central bank-there is now an active discussion going on various options for returning full state control over monetary circulation. Among others, the former Governor of the Bank of England M. King proposes to reform the modern banking system. According to M. King, in the future the risks created by the transformation of individuals' insured deposits into highrisk loans should be born by the financial institutions that gain profit from such "financial alchemy" [32].
M. King proposes to create two types of banks to prevent systemic risks: narrow banks and wide banks. Narrow banks must insure their deposits with 100% reserves in the form of highly liquid safe assets such as government securities and mandatory reserves held at the central bank. The income of narrow banks should come exclusively from the provision of payment services. Along with narrow banks, M. King proposes to create "wide" banks, which today include a variety of non-banking financial institutions. Wide banks can finance all other types of activities, and above allprovide high-risk loans to the private sector. However, unlike narrow banks, they cannot accept demand deposits or short-term deposits. Instead, wide banks must finance high-risk loans by issuing shares or longterm debt. This system should prevent financial abuse resulting from the severing of the link between money creation and credit creation.
The second aspect of the reform proposed by M. King is the change in the function of the central bank as lender of last resort. M. King proposes turning the central bank into a "pawnbroker for all seasons." According to this innovative approach, any bank (or other financial intermediary) must reserve a part of its assets from the central bank in advance for their subsequent use as collateral (with a discount if necessary) against the future lending of liquid funds. Accordingly, the bank's liquid assets must exceed its liquid liabilities. According to M. King, such an approach can eliminate the problem of moral hazard associated with the performance of the central bank as the lender of last resort, and significantly simplify financial regulation [33].
M. King's proposals are only a part of a broader concept of "sovereign money" which proposes to reform the banking system in such a way as to deprive private banks of the opportunity to create new money altogether. New money in the form of bank deposits is created when banks issue loans and is destroyed when the borrower repays the loans. The problem is that most of the new money created with the help of bank lending goes not to finance production or activities indirectly related to it, but to buy existing assets-for example, real estate or securities, in order to resell them at higher prices. The turn of the central banks to such unconventional monetary policy as "quantitative easing" and near-zero interest rates does not seem to be a fully adequate answer to this problem.
The idea of "sovereign money" originated in the work of the English scientist, Nobel laureate in chemistry F. Soddy, which was published in 1926. During the Great Depression, American economists F. Knight and H. Simons presented Soddy's ideas to US President Franklin D. Roosevelt (later they were popularized by I. Fischer). Since then, the idea of "sovereign money" has been developed in various ways by M. Friedman, D. Tobin, H. Minsky, D. Kay and L. Kotlikoff.
Quite recently the economists of the International Monetary Fund tested I.Fischer's initial model of sovereign money on the indicators of the modern US economy and found a "solid confirmation" of all the declared advantages and additional positive effects of this model [34]. Since then, the idea of sovereign money, in addition to the former Governor of the Bank of England, M. King, was supported and developed by the chief economics commentator of the Financial Times M. Wolf, the former chairman of the UK Financial Services Authority A. Turner, as well as Vice-President of the European Central Bank Vítor Constâncio [35].
Technically, the idea of sovereign money is extending the exclusive right to create money which is reserved by constitution to the central banks of many countries, also to the balance sheets of commercial banks, which would deprive the latter of the opportunity to create money through demand deposits and term bank deposits.
In the system of sovereign money, the creation of required reserves at central banks by commercial banks becomes redundant, because all money acquires the status of reserve money. Due to a significant simplification of the banking business and a bank's balance in the system of sovereign money, the proposal to create a system of 100% bank reserve seems inappropriate [36, pp. 45-47].
The main limitation of the sovereign money system is that the central bank, acting as the central planner for money creation, needs to have much more infor-mation than it has under present conditions. Another limitation is that commercial banks, due to a lack of central bank money, cannot always take advantage of attractive lending opportunities. Consequently, the elasticity of bank lending in the real economy is generally higher in the current system than in the sovereign money system [37].
Characteristically, the concept of sovereign money does not contradict the tendency of transferring the world economy to a digital platform, which implies an unprecedented centralization of the monetary system if money issuers gain unlimited access to Big Data. It should be noted that the UK is leading among the largest economies of Europe in the level of digitalization of the national economy [38]. In the years 2009-2019 the share of cash payments in the United Kingdom declined from 58 to 23% [39, p. 3]. These trends are in tune with the preparatory work of the Bank of England to transfer the national currency to a digital basis [40]. The digitalization of the pound sterling is seen as a tool that will increase the security and the efficiency of payments, simplify international transfers, reduce the risks of issuing private currencies, and activate the digitalization of the economy.
It can be assumed that even a partial implementation of the concept of sovereign money, in conjunction with the successful introduction of the digital pound, can not only provide Great Britain with financial support for large-scale national projects in the era of the fourth industrial revolution, but also serve as an example of anti-crisis regulation for other countries.
CONCLUSIONS
For centuries financial innovation has been a constant factor in restoring the UK's leading position in the international arena. Until now, the political and economic stability of the British state has largely depended on the consistent adherence to the key principles of the Anglo-Saxon socio-economic model. The future will show how much the adherence to these principles will help the United Kingdom keep its role as a mediator of the relations between the Old and the New Worlds in the conditions of the emerging technological paradigm of the polycentric world. | 2021-05-11T00:07:18.871Z | 2021-03-31T00:00:00.000 | {
"year": 2022,
"sha1": "69c4805b3b2daa3dfd5363b6a0279e51bce95e9b",
"oa_license": null,
"oa_url": "https://link.springer.com/content/pdf/10.1134/S1019331621060174.pdf",
"oa_status": "BRONZE",
"pdf_src": "PubMedCentral",
"pdf_hash": "86b6b194499413f3d717e809df02cf698d587db0",
"s2fieldsofstudy": [
"Economics"
],
"extfieldsofstudy": [
"Political Science"
]
} |
35015874 | pes2o/s2orc | v3-fos-license | Usefulness Lost: Aggregating Information with Differing Levels of Verifiability
In this paper, we study information asymmetries about verifiability between a principal and an agent. Our main result is that an information asymmetry about verifiability not only reduces the usefulness of a given performance measure for stewardship purposes, it can completely destroy that performance measure's usefulness.
Introduction
Many of the Financial Accounting Standards Board's (FASB's) recent standards and ongoing projects emphasize fair value measurements in an attempt to increase the relevance of financial reports. These same standards and projects have reduced or have the potential to reduce the reliability of accounting. 1 Critics of fair value and other subjective measurements often focus on the limited verifiability of such measurements and the concern that limited verifiability facilitates manipulation. 2 An expanded use of fair value measurements and the incorporation of those measurements in determining managerial rewards seem to have played a significant role in Enron's spectacular fraud and collapse. 3 One might have expected the fiction of Enron's fair value measurements to result in the FASB and the Securities and Exchange Commission (which pre-cleared Enron's expanded use of fair value measurements) reversing course, but such a reversal does not appear to have occurred or to be forthcoming anytime soon.
Limited verifiability produces numbers that are less reliable and, holding all other attributes constant, less useful. From a measurement perspective, aggregating reliable line items with unreliable line items to arrive at an overall measure of performance such as net income pollutes that measure. Arguably, an even worse problem in financial reporting today is that financial statement users cannot tell how verifiable a given line item is. Highly verifiable amounts from realized transactions are often included in the 1 In CON 2, the FASB defines reliability as consisting of verifiability, representational faithfulness, and neutrality. 2 See, for example, Carmichael (2004) and Watts (2003). Watts writes: "[m]anagers' limited tenures and limited liability give them incentives to introduce bias and noise into value estimates. The lack of verifiability of many valuation estimates gives managers the ability to do so." same financial statement line items as fair value remeasurements that are, at least sometimes, much less verifiable. The preparer knows which types of measurements are included in any given line item in any given reporting period, but the user does not.
In this paper, we use a principal-agent model to study information asymmetries about verifiability between an agent (manager) and a principal (owner). Our focus is not on measurement per se, but instead on information content and, in particular, on the stewardship role of accounting Ijiri, 1975), which we treat as one of the contracting roles of accounting (Sunder, 1996;Watts and Zimmerman, 1986). Our main result is that an information asymmetry about verifiability not only reduces the usefulness of a given performance measure (e.g., a financial statement line item) for contracting, it can completely destroy that performance measure's usefulness.
In contrast, if the principal and the agent have symmetric information about the level of verifiability of a performance measure, a less verifiable (but still informative) performance measure always has a positive value.
Our contribution lies in distinguishing between limited verifiability and information asymmetries about verifiability. Under existing financial reporting, limited verifiability and information asymmetries about verifiability seem to go hand in hand. If we instead disaggregated measures based on their level of verifiability, limited verifiability would, by design, be less associated with information asymmetries about verifiability. Glover, Ijiri, Levine, and Liang (2005) suggest one such verifiability-based disaggregation, focusing on the distinction between facts and forecasts. Barker (2004) proposes disaggregating initial measurements from remeasurements because remeasurements have less predictive value, but such a disaggregation can also be viewed as a verifiability-based disaggregation. In their final chapter, Paton and Littleton (1940) discuss the possibility of reporting historical cost measurements in a first column and market values in a second supplementary column.
This paper can be viewed as providing a stewardship/information economics foundation for such verifiability-based separations within financial statements. While it seems natural that the separation would better facilitate multiple uses of accounting information (e.g., valuation and stewardship), we find it interesting that a demand for such a separation arises in our model for stewardship purposes alone.
The FASB's Exposure Draft on Fair Value Measurements (FASB, 2004) suggests yet another approach, relegating information about verifiability to the footnotes.
Arguably, the most important information about verifiability is whether the measurement comes from a realized transaction or an unrealized remeasurement. The current working draft of the standard calls for a disaggregation of unrealized and realized amounts for only the worst level (Level 5) fair value measurements, which are fair value measurements arrived at from models that use entity-specific rather than market inputs.
We disagree with what seems to be the implicit view of the FASB that the distinction between realized and unrealized amounts is important only as a second-order consideration and only in the most extreme case (Level 5 measurements).
We now elaborate on our model and results. The model starts with underlying transactions that stochastically depend on the agent's effort. The transactions are then converted into accounting numbers via a mechanistic auditing/verification process. The agent's reward depends on only the final accounting measurements. Limited verifiability is modeled as a garbling of accounting measurements. The garbling reduces informativeness (in the sense of Holmstrom, 1979), which increases the expected cost of compensation but does not affect the form of the optimal contract.
Our equating verifiability with a garbling of measurements is different than the way the term verifiability is typically used in contract theory, where verifiability is used to mean a report can be contracted on because it can be verified by a court of law.
Instead, we attempt to stay close to the FASB's definition of verifiability, which is: "the dispersion of independent measurements of some particular phenomenon (FASB, CON 2)." The FASB's definition of verifiability corresponds to Ijiri and Jaedicke's (1966) definition of objectivity.
To our basic structure, we then add an information asymmetry about the verifiability of the accounting system. Specifically, the agent knows more about the verifiability of the second of two performance measures than does the principal. The contract now takes on a different form. If the information asymmetry about the verifiability of the second performance measure is large enough, that performance measure is rendered useless and is excluded from the contract.
The omitted performance measure satisfies Holmstrom's (1979) informativeness condition, which would lead to the measure being included in the optimal contract in the standard principal-agent model. Our model is different than the standard one in assuming (i) there is an information asymmetry about verifiability and (ii) the contract cannot depend on an (unaudited) report by the agent on the level of verifiability. (More on this second assumption later.) The intuition for the optimality of ignoring the second performance measure is that the agent's informational advantage requires the principal to design a contract that is robust to a variety of possible levels of verifiability. When the information asymmetry is large, the least expensive means of satisfying the robustness requirement is to eliminate the measure for which there is an information asymmetry about verifiability from performance evaluation.
We then add the possibility of performance measure manipulation by the agent.
As it turns out, under symmetric information about verifiability, the addition of manipulation actually benefits the principal. In our simple one-period model, the principal is better able to anticipate the agent's (equilibrium) attempt to overstate a performance measure than she can the two-sided noise introduced by limited verifiability alone. 4 Manipulation is facilitated by limited verifiability but results in overstatements only, since the agent's compensation is increasing in the reported performance measures.
Here we have a case of "the devil that you know is better than the devil that you don't." If instead there is asymmetric information about verifiability, the principal can prefer the agent not to have the manipulation option. The reward needed to motivate the agent is determined by the lowest level of manipulation. If the manager can manipulate more extensively, the contract can become extremely costly. When the information asymmetry about manipulability (induced by the underlying information asymmetry about verifiability) is large enough, it can be optimal to exclude from the contract the performance measure for which there is an information asymmetry about verifiability. 4 Importantly, our comparison is between manipulation and random noise. In Demski (1998), earnings management is better than no noise from the principal's perspective when the agent's ability to manipulate (smooth) is dependent on his productive effort. See also Arya, Glover, and Sunder (1998), Christensen, Demski, and Frimor (2002), and Sunder (1996) for related models and discussions of earnings management.
The remainder of the paper is organized as follows. Section 2 presents the basic model and its solution as a benchmark. Section 3 contrasts verifiability with information asymmetries about verifiability. Section 4 extends the model to include manipulability.
Section 5 incorporates verifiability-dependent effort. Section 6 concludes the paper.
The Basic Model
A risk-neutral principal contracts with a risk-neutral agent subject to moral hazard on the agent's action supply and limited liability on the payment to the agent. The principal makes a payment to the agent, not the other way around. This limited liability constraint gives rise to the possibility of the agent earning rents, which the principal tries to minimize. (See Laffont and Martimort, 2002, p. 155.) Agent risk neutrality makes the analysis particularly tractable, but the qualitative results of the paper continue to hold when the agent is risk averse (see Example 1B).
The principal would like to motivate the agent to supply an unobservable and personally costly action a = a H rather than a L < a H . The implicit assumption is the parameters are such that the high action generates a large enough marginal benefit to the principal that she always finds it optimal to motivate the high action instead of the low action. (See Section 5 for additional details.) The agent's action gives rise to two transactions t 1 and t 2 , each of which can take on realizations of t L or t H . Transactions t 1 and t 2 are then converted into accounting reports r 1 and r 2 through a combination of reporting and verification. We assume throughout the paper that the first accounting report, r 1 , is a perfectly verifiable performance measure: when t 1 is t j , it will be reported as r j , j € ∈ {L, H}. The second accounting report, r 2 , is less verifiable: with probability θ, when t 2 is t j , it will be reported as r k , k ≠ j; j,k € ∈ {L, H}, θ ∈ [0,1/2]. θ = 0 means the second transaction, t 2 , is perfectly verifiable-there is no accounting-induced noise in its measurement. θ = 1/2 means the report is perfectly unverifiable. Intermediate values of θ allow for differing levels of verifiability.
Throughout the paper, the agent knows the verifiability of the reporting system, θ. In some cases, the principal will also know θ. In others, the principal will not know θ, i.e., there will be an information asymmetry about verifiability. In these cases, the principal knows only that θ is uniformly distributed over the interval [0,θ max ], θ max ≤ 1/2. 5 If the agent chooses a H , the probabilities over the set of possible (t 1 ,t 2 ) transactions If the agent chooses a L , the corresponding probabilities are {q LL , q LH ,q HL ,q HH }.
The contract is allowed to depend on only the final reports r 1 and r 2 . The contract is simply a report-contingent payment s(r 1 ,r 2 ). Importantly, we have deliberately ruled out a report on θ. Ruling out a report on θ allows us to study the impact of information asymmetries about verifiability that persists (are not resolved by unaudited direct communication). We are interested in studying the implications of such a restriction because it seems to be characteristic of the current financial reporting environment. As we discussed in the introduction, we think current financial reporting would be well served by introducing ways of (partially) resolving the information asymmetry about verifiability. 6 5 The uniform distribution is important only in that, if the distribution is too concentrated around the mean, the information asymmetry can never be large enough to drive the second report out of the contract. 6 Ruling out a report on θ can also be viewed as a way of studying robust contracts vs. the standard approach of (highly) fine-tuning the contract to the environment (e.g., the exact probability distribution).
The agent (who knows θ) has the induced probability distribution over (r 1 ,r 2 ) realizations presented in Table 1.
The ordering of the payments is determined by the standard likelihood ratios. For tractability, we assume the likelihood ratios are ordered as follows.
The agent's payoff is the payment he receives less the cost of effort, also denoted by a H or a L . The principal's objective is to minimize the expected payment to the agent E[s(r 1 ,r 2 )] (Equation 1) subject to the constraints that choosing a H is incentive compatible for the agent (Equation 2) and the payments s(r 1 ,r 2 ) be nonnegative (Equation 3). The usual individual rationality constraint is omitted. Alternatively, the individual rationality constraint is dominated by the other constraints if we assume a L and the agent's reservation utility are both zero.
As a benchmark, we formulate and solve the principal's program assuming the problem is a standard moral hazard problem in which both performance measures are objectively determined (θ = 0): Standard contracts typically perform poorly if the environment is even slightly different than assumed and (thus) do not bear a close resemblance to real-world contracts. It seems inevitable, even desirable, that robust contracts would exclude (informative) performance measures that are not well understood by both parties to the contract. The study of robust mechanisms is known as the (Robert) Wilson Program in game theory. Bergemann and Morris (2005) The following result is well known (e.g., Laffont and Martimort, 2002, p. 164).
Observation. If both performance measures are perfectly verifiable (θ = 0), the solution is to make a single bonus payment of (a H -a L )/(p HH -q HH ) for the report (r H ,r H ) with the highest associated likelihood ratio p HH /q HH . All other payments are zero.
Verifiability vs. Information Asymmetries about Verifiability
In the case of limited verifiability (θ > 0), the solution is modified in a straightforward way to incorporate θ. The compensation contract still makes a single payment in the state when both of the reports are high, although the actual amount of compensation changes.
Proposition 1. If the second performance measures is subject to limited verifiability but there is no information asymmetry about θ, then the optimal contact uses both performance measures. The solution is to make a single bonus payment of: We now study the case in which there is an information asymmetry about verifiability (θ is not known by the principal). Here, we have in mind that the information asymmetry is created by exogenous accounting standards that allow for the aggregation of more and less verifiable information into a single line item. The preparer will know how verifiable the given line item (r 2 in our model) is, but the financial statement user will not.
See Banker and Datar (1989) for a model in which there is a direct connection between the statistical notions of precision and sensitivity and the optimal weight placed on a performance measure in aggregating it with other performance measures. In our model, the garbling produced by limited verifiability decreases both the second performance measure's precision and its sensitivity. In their model, unlike ours, the principal and the agent share symmetric beliefs about the stochastic properties of possible performance measures.
The principal's objective function (Equation 4) uses the expected level of verifiability (for the uniform case, θ max /2). Since the agent knows the actual level of verifiability when making his effort decision, the principal must find a contract that satisfies the agent's incentive compatibility constraint for every possible level of verifiability. That is, the incentive compatibility constraint (Equation 5) is actually a family of constraints, one for each θ . The form of the optimal contract is characterized in Proposition 2.
Proposition 2.
If there is an information asymmetry about the verifiability of the second performance measure, then the optimal contact uses only the first performance measure if and only if: In the case where the contract "drops" the second measure, the bonus state occurs with much higher probability (the sum of the top two states). This more muted contract seems roughly consistent with common managerial accounting wisdom that bonus targets should be difficult but achievable. Instead of a behavioral explanation for achievable targets, our paper suggests that rewarding agents for extreme outcomes can be hazardous because extreme outcomes are likely to be less well understood than more probable outcomes. The two cases correspond to the agent's marginal productivity in the performance measures, which determines the size of the required bonus if both performance measures are to be used. If p HH -q HH ≤ p HL -q LH , the size of the bonus payment is based on a θ of 0. If p HH -q HH > p HL -q LH , the size of the bonus payment is based on a θ of θ max . Example 1A illustrates the underlying intuition for the first case.
Manipulability
We now generalize the model to allow for both limited verifiability and manipulation by introducing a parameter µ ∈ {0,1}. µ = 0 represents the model in the previous section. µ = 1 denotes the presence of a manipulation option for the agent, which interacts with the accounting/auditing process. We offer two equivalent descriptions. One, the agent observes the underlying transaction, t 2 , and then prepares a report for a mechanistic auditor to verify. The auditor discovers any manipulation with probability (1-θ ). Two, a mechanistic accounting system first produces a preliminary report that correctly identifies t 2 with probability (1-θ ). The agent then selectively corrects any mistake the accounting system makes. Under this second interpretation, the agent is assumed to be able to provide further evidence to make corrections but not to be able to make false corrections. (As in the previous sections of the paper, we assume t 1 is perfectly verifiable.) The only way the principal can motivate the agent to choose the high action is to offer an incentive scheme that is non-decreasing in each report (and strictly increasing in at least one report). As a result, the agent will manipulate the report only when the underlying transaction is t L . Given t L , the final report will be r L with probability (1-θ ) and r H with probability θ. Given t H , the final report will be r H with probability (1-θ +µθ ) and r L with probability (1-µ)θ. The following table presents the induced probabilities under our more general model, taking the agent's equilibrium reporting behavior as given.
Suppose the agent has the ability to manipulate (µ = 1). When the verifiability parameter θ is common knowledge, the principal also knows what the amount of expected manipulation is. When the agent knows more about the verifiability parameter than does the principal, there is an induced information asymmetry about the level of manipulability, since limited verifiability facilitates manipulability. Assuming µ = 1 and that there is an induced information asymmetry about manipulability, the principal's program is given as: First, consider the case in which manipulation is possible but the probability of successful manipulation is common knowledge (because θ is common knowledge).
Proposition 3. If the second performance measure is subject to manipulation but there is no information asymmetry about the extent of manipulability, the optimal contact uses both performance measures. The solution is to make a single bonus payment of: (a H -a L )/(p HH + θ p HL -q HH − θ q HL ) for the report (r H ,r H ).
The intuition for Proposition 3 is the same as for Proposition 1. The known amount of manipulation is costly for the principal (relative to perfect verifiability) but can be incorporated into the contract in a straightforward manner. An important distinction between the optimal contract in Proposition 3 and the optimal contract in Proposition 1 is that the principal can take the agent's equilibrium manipulation into account and rule out the possibility of downward misstatements. Now, consider the case of asymmetric information about verifiability, which induces asymmetric information about manipulability.
Proposition 4. If the second performance measure is subject to manipulation, there is an information asymmetry about the extent of manipulability, and then the optimal contract ignores the second performance measure.
A difference between Propositions 2 and 4 is that the two cases in Proposition 2 (corresponding the agent's marginal productivity in the performance measures) become one case in Proposition 4. The reason is that manipulation eliminates downward misstatements. The following corollaries follow from Propositions 1-4.
Corollary 1 (to Propositions 1 and 3). Limited verifiability alone is worse for the principal than manipulability.
The intuition for Corollary 1 is the principal can better anticipate the agent's attempt to manipulate a performance measure than she can the noise introduced by a verifiability problem alone. This is a comparison of two standard models in which there is no information asymmetry about verifiability.
(i) If p HH -q HH < p HL -q HL , the principal prefers an information asymmetry about verifiability to an information symmetry about manipulability, and the latter drives a performance measure out of the optimal contract sooner than the former.
(ii) If p HH -q HH > p HL -q HL , the principal prefers an information asymmetry about manipulability to an information asymmetry about verifiability, and the latter drives a performance measure out of the optimal contract sooner than the former.
The intuition for Corollary 2 is the principal's informational disadvantage regarding the accounting treatment makes it more difficult to customize the contract to anticipate the agent's manipulation, since the principal does not know how much manipulation is possible. If p HH -q HH is small (Case (ii)), the problem is particularly acute since using both performance measures means the s(r H ,r H ) payment must be set equal to (a H -a L )/(p HH -q HH ), which is large. Arguably, Case (ii) is more "typical" than Case (i) in that extreme outcomes are associated with small probabilities.
In our model, opportunistic manipulation is costly to the principal (relative to limited verifiability alone) only when the agent knows more about the underlying accounting system, as shown in Corollaries 1 and 2. Examples 2 and 3 are illustrative.
Verifiability-Dependent Effort
Throughout the earlier sections of the paper, we assumed the marginal productivity of the agent's effort was sufficiently large that hiring the agent and motivating him to choose high effort was always optimal. In this section of the paper, we relax this assumption. Given the principal maximizes the expected value of the firm less payment to the agent, we derive conditions under which our earlier main result, Proposition 2, holds. Throughout this section, we assume there is an information asymmetry about variability.
Assume t 1 and t 2 are each cash inflows. The principal now maximizes E[t 1 +t 2 -s(r 1 ,r 2 )]. To keep the notation to a minimum, assume a L = t L = 0. a L = 0 also ensures that hiring the agent and motivating him to shirk is always preferred to not hiring the agent at all. To rule out the case that always motivating a L is optimal, assume Denote the value of t H that satisfies (10) with equality by t * .
If p HH -q HH < p HL -q HL , (10) by itself ensures that high effort is always optimal, whether one or both performance measures are used. If p HH -q HH > p HL -q HL , (10) leaves room for it to be optimal to use both performance measures and to motivate the agent to choose a H when θ is below some cutoff θ C and a L when θ is between θ C and θ max . That is, the agent is motivated to work if and only if the actual verifiability of the second performance measure is high. The principal's objective function under this new contract is: The optimal cutoff θ C is determined by setting the derivative of (11) with respect to θ C equal to zero. (The second derivative is negative, guaranteeing a maximum.) To ensure that motivating high effort is always optimal, we need to find a value of t H under which θ C = θ max . Denote this cutoff level by t ** . A closed form expression for t ** is given in the proof of the following proposition (a restatement of Proposition 2).
Proposition 5.
If there is an information asymmetry about the verifiability of the second performance measure, then the optimal contact uses only the objective performance measure if: € p HH − q HH ≤ p HL − q HL , t H ≥ t * , and θ max 2 >
Concluding Remarks
In this paper, we study the impact of information asymmetries about verifiability and find they can lead to performance measures becoming not only less valuable but valueless for stewardship purposes. An implication seems to be that, if the stewardship role of accounting is to be improved, verifiability-based disaggregations within the financial statements are likely to help as they will reduce information asymmetries about verifiability. Of course, limited verifiability itself raises many concerns (some of which we discussed in the introduction). We view our paper as complementing existing concerns about limited verifiability, which have become central in accounting as the FASB introduces more and more difficult to verify fair value measurements.
To be fair (no pun intended), our model is of contracting, while the FASB's focus is on capital market decisions. We suspect similar forces can be developed in a market model (for example, if information asymmetries about verifiability distort the firm's investment decisions) but also view the contracting and stewardship roles of accounting as important and often under-emphasized.
Proof of Proposition 1. The principal's problem is a linear program. Let λ be the multiplier on the incentive compatibility constraint and the choice variable in the dual program. Consider the following proposed solution to the primal and the dual. Under each proposed solution, the objective functions of the dual and the primal are equal. The condition given in the statement of the proposition determines which of these solutions is feasible.
Proof of Proposition 5. For the p HH -q HH < p HL -q HL case, the new possibly optimal contract has the agent working when θ is above a cutoff level θ C and shirking for all smaller θs. t H > t * , the solution to equation (10) in the text, ensures θ C = 0, i.e., the agent always works. So, the additional condition given in Proposition 2 (and restated in Proposition 5) is sufficient to ensure it is optimal to ignore the second performance measure.
For the p HH -q HH > p HL -q HL case, the new possibly optimal contract has the agent working when θ is below a cutoff level θ C (verifiability is high) and shirking for all larger θs (verifiability is low). Differentiating the principal's objective function, given as equation (11) in the text, with respect to θ C and setting it equal to zero yields the firstorder condition to be solved. Evaluating that first-order condition at θ C = θ max yields a cutoff value of t H , denoted by t ** , above which setting θ C = θ max is always optimal. | 2017-05-20T09:26:02.599Z | 2005-05-07T00:00:00.000 | {
"year": 2005,
"sha1": "b5ff19573aacb1a351e1a56002b9d4dbf7e431e8",
"oa_license": "CCBY",
"oa_url": "https://figshare.com/ndownloader/files/12238082",
"oa_status": "GREEN",
"pdf_src": "ElsevierPush",
"pdf_hash": "6518b428aa0435e8e2f13b809d5952c932fd1952",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Engineering"
]
} |
115752062 | pes2o/s2orc | v3-fos-license | Application of Macro-Synthetic-Fiber-Reinforced Sprayed Concrete for Primary Lining Support on Highway Tunnel Project in Bosnia and Herzegovina
The application of macro synthetic fiber reinforced sprayed concrete is seen by many design engineers as the alternative to steel reinforced sprayed concrete. The first applications of macro synthetic fiber reinforced sprayed concrete in Bosnia and Herzegovina were applied during the construction of the Sarajevo By-pass Motorway. The Project included a Motorway tunnel-system of three (3) double tubed, two-lane tunnels receptively 860m,
The carriageway width is 7.0m, and consists of two lanes of 3.50m.
Underground Conditions
The Underground conditions (geological formations) on the location of Tunnel T3 were diverse and complex. Three (3) basic types of rock were found (Sedimentary, Metamorphic and Magmatic rocks) [2]. The Metamorphic rocks were from Paleozoic and the Dolomites were from Devon. In-between two Magmatic layers of Quartz Porphyries were also found ( Figure 2).
Tunnel excavation and support elements (Main Design of the Project)
In accordance to the Main Design documentation in the Project the excavation works were divided into three (3) phases. The top heading, bench and the invert arch excavation phase. In better underground conditions the excavation works were divided in top heading and bench phases only [3]. The Main Design defined rock categories, predicting the round lengths and defining support elements (system) for each of the rock category respectively (Table 1).
Tunnel T3 excavation works
At the early beginning of the Tunnel T3 Project it was obvious that the predicted rock mass categories given in the Main Design are different compared to the actual underground conditions on site. At the entrance portal and the tunnel precut it was identified that the Main Design was predicting 4 th (IV) rock mass category and actually 5 th (V) rock mass category was present. Consequently it was necessary to apply other support elements, demanding more time (require longer installation time) then predicted. Due to the underestimation of underground conditions inevitably delays in the excavation works in Tunnel 3 were going to occur. Only undertaking appropriate measures it was possible to mitigate delays and finish the Project within the expected program of work.
According to the geological predictions approximately two thirds of the Tunnel T3 was located either in dolomitic limestone (approx. 900m) or quartz-porphyric metamorphic rock (approx. 900m). In such underground conditions it was possible to modify the conventional support elements (systems) and apply fiber reinforced sprayed concrete instead of mesh reinforced sprayed concrete. It has been estimated that the changes in these support elements would probably save 2 to 3hours per round length (step of excavation) [4]. The length of the tunnel rock mass categories where fiber reinforced sprayed concrete could be used was estimated to be 1800m in length. Average round length (step of excavation) was estimated to be 2.5m. 1800m/2.5m = 720 steps 720 (steps) x (2-3hours-shorter installation time) =makes up to 1440 to 2160hours
Calculation
In order to implement this idea a 60m test field in Tunnel T3 was formed to compare the behavior of the support elements in real underground conditions. The behavior of mesh reinforced sprayed concrete (cross section 24.300-24.330; 30m) and fiber reinforced sprayed concrete (cross section from 24.330m to 24.360m ; 30 m)lining had to be tested. Measurement profiles were installed (measuring the movements in the primary lining for a period of time until the settlements stabilize) at 24.315m (cross section with mesh reinforced sprayed concrete) and 24.345m (cross section with fiber reinforced sprayed concrete).
The updated Q-System
The relation between the rock-mass quality and the support elements (supporting system) using fiber reinforced sprayed concrete is defined by the updated Q-System by Grimstad and Barton (Figure 3). Identifying the rock-mass quality (Q) for certain rocks the supporting elements (system) can be predicted using the updated Q-System.
For the intended underground conditions in Tunnel T3 where fiber reinforced sprayed concrete could have been used, the rockmass quality index was ranging between 1 and 10 (D-Poor, C-Fair, B-Good). Considering the Span or Height ratio (for ESR =1) being10 the expected reinforcement categories 4, 5 and 6 would need energy absorption capacity between E=500-700J ( Figure 4).
The test field of macro-syntheticfiber reinforced sprayed concretein tunnel T3
The interaction of the rock-mass and the tunnel primary lining could be observed during the construction, making real scale measurements. Measuring the displacements of the rock-mass the impact on the tunnel support elements could be observed in real time. According to the NATM concept, measurement profiles are installed systematically at characteristic sections along the tunnel. The aim of the monitoring is to predict and to prove the adequacy of the supporting system. This approach was also applied in the test field with the aim to confirm the adequacy of the fiber reinforced sprayed concrete lining in Tunnel T3. The back calculation analysis was used to determine the relevant mechanical characteristics of the rock-mass and to refine the support measures. The test field of a 60m section of the both mesh and fiber-reinforced sprayed concrete lining was installed to examine if the fiber reinforced sprayed concrete primary lining could be equally efficient as the conventional primary lining when used in similar geological conditions [5]. In particular the bending capacity had to be assessed this way. To quantify the efficiency of the fiber reinforced sprayed concrete primary lining, the results from the test section had to be compared with the results of the measurement profiles featuring the conventional primary lining for the same overburden, and very similar geological conditions.
The test setup
The test setup was as follows: i.
In to the already used sprayed concrete mixture 4kg of polypropylene fibers were added v.
Measuring the profiles on cross sections at 24.315m (mesh reinforced sprayed concrete) and 24.345m (cross sectionfiber reinforced sprayed concrete) was based on the geological mapping and data collected by the geologist on the site (RMR=47). The geotechnical parameters were assumed using the Hoek & Marinos [6] method taking into account all the relevant indicators of the rock mass behavior including the favorableness of the rock layering. Relevant parameters (σci=40 MPa, GSI=42, mi=10, D=0, 5, MR=500, γ= 0, 025 MN/m3, Tunnel Depth=65m) where analyzed using the software ("Roc Lab") to acquire Mohr-Coulomb strength parameters for numerical analysis (c=0, 274MPa; phi=42, 86deg; E=1744, 22MPa). After the backcalculation the following results were computed (Figure 6 & 7): The bending moments (Md, max=8, 78kNm/m) and axial forces (Nd, max=489, 80N/m) are plotted on the interaction diagrams (Figure 7) of the mesh reinforced sprayed concrete for the installed sprayed concrete C25/30 and reinforcement steel mesh type-Q 188. The predicted bending capacity of the primary lining in the B1 rock mass category should not exceed 20kNm/m.
Conclusion
The macro-synthetic fiber reinforced sprayed concrete samples have shown less energy absorption capacity in comparison with the mesh reinforced sprayed concrete samples determined by the EFNARC plate test (EN 14485-5;2006). The required energy absorption capacity according to the updated Q-System rock classification by Grimstad & Barton [2] was between E=500-700J see ( Figure 3) and the tested energy absorption capacity of tested specimens using macro-synthetic fiber in the test field was of an average of E=740J see ( Figure 5). It could be concluded that the energy absorption capacity of the macro-synthetic fiber reinforced sprayed concrete was still of satisfactory capacity and could be applied for the support system in category B1.
The test filed results show a very good post cracking performance of the macro-synthetic fiber reinforced sprayed concrete at deflections of about 5mm. It was measured that the deflections of 5mm usually occur during the first hours after excavation.
Analyzing the energy absorption test results at the deflection of 5mm it can be concluded that both the mesh and fiber sprayed concrete capacity absorb almost the same force F=45kN see (Figure 4 & 5). Considering these results the fiber sprayed concrete provides more safety during the first hours of excavation due to simple installation.
Analyzing the results of the back calculation the Md, max=8,78kNm/m at the total displacement of 10mm is much more smaller compared to the predicted bending capacity of 20kNm/m in the primary lining for B1 rock mass according to ÖNORM B2203. The measurement profiles on cross sections at 24.315m and 24.345m showed almost the same deflections for both the mesh and fiber sprayed concrete primary lining. So it can be concluded that the fiber sprayed concrete lining can be equally efficient. The in-situ identified geological rock mass categories and the applied support systems are shown in Table 2. Comparing the geological rock mass categories from the Main Design (Table 1) and the categories identified during the excavation obviously shows that the share of 4th (IV) rock category was increased and on the other hand, the shares of 2nd (II) and 3th (III) geological rock mass categories were reduced.
After all, in Tunnel T3 almost 1600m of fiber reinforced sprayed concrete was applied. The net saving time using macro synthetic fiber in the sprayed concrete lining was 70days. This fully compensated the lost time for the increased length of 4th (IV) rock category that was not predicted by the Main Design. The Project was successful accomplished on time, with no delays. The application of fiber sprayed concrete using engineering know-how and experience played a dominant role in this successful construction project. Macro synthetic fiber sprayed concrete and fiber sprayed concrete in general should be considered as a powerful tool in tunnel engineering projects. | 2019-04-16T13:28:12.144Z | 2018-03-14T00:00:00.000 | {
"year": 2018,
"sha1": "24f74af8d0811f5198975bda2536b11e4f5d94d8",
"oa_license": "CCBY",
"oa_url": "https://crimsonpublishers.com/amms/pdf/AMMS.000513.pdf",
"oa_status": "GOLD",
"pdf_src": "Adhoc",
"pdf_hash": "3ecb49bd464cd6a59c29266006a81eca13f961de",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Engineering"
]
} |
49869604 | pes2o/s2orc | v3-fos-license | CT findings of adnexal torsion: A matched case-control study
Objective The purpose of our study was to assess computed tomographic (CT) findings of adnexal torsion through a matched case-control analysis. Materials and methods This retrospective, single-institution case-control study included 43 women with adnexal torsion and 43 age- and ovarian mass-matched control women. CT images were evaluated independently by two readers for the following: prominent peripheral follicles, uterine deviation, thickened pedicles, a whirl sign, and a navel sign. Comparisons of CT findings were performed using the Chi square test and receiver operating characteristic (ROC) curves were obtained to assess the diagnostic performance. Differences between the areas under the ROC curves (AUCs) were compared by using a Delong test. Results The CT findings significant for adnexal torsion were uterine deviation toward the side of the affected ovary (P = < .01 for reader 1 and P = .02 for reader 2) and thickened pedicles with ancillary findings including a whirl sign, a navel sign, and uterine deviation facing thickened pedicles (P < .01 for both readers). Thickened pedicles with ancillary findings had the highest diagnostic accuracy, as measured with ROC curves (AUC, 0.86 in reader 1 and 0.85 in reader 2). Combining uterine deviation toward the side of the affected ovary with thickened pedicles with ancillary findings did not increase the performance relative to that of thickened pedicles with ancillary findings alone. Conclusions Thickened pedicles with ancillary findings including a whirl sign, a navel sign, and uterine deviation facing thickened pedicles could be helpful for the diagnosis of adnexal torsion.
Results
The CT findings significant for adnexal torsion were uterine deviation toward the side of the affected ovary (P = < .01 for reader 1 and P = .02 for reader 2) and thickened pedicles with ancillary findings including a whirl sign, a navel sign, and uterine deviation facing thickened pedicles (P < .01 for both readers). Thickened pedicles with ancillary findings had the highest diagnostic accuracy, as measured with ROC curves (AUC, 0.86 in reader 1 and 0.85 in reader 2). Combining uterine deviation toward the side of the affected ovary with thickened pedicles with ancillary findings did not increase the performance relative to that of thickened pedicles with ancillary findings alone.
Conclusions
Thickened pedicles with ancillary findings including a whirl sign, a navel sign, and uterine deviation facing thickened pedicles could be helpful for the diagnosis of adnexal torsion. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111
Introduction
Adnexal torsion is a gynecologic emergency accounting for 2.7% of female acute pelvic pain [1]. Transvaginal ultrasound (US) has been considered to be the diagnostic modality of choice to assess torsion [2,3] but computed tomography (CT) may often be the initial imaging modality because the presentation of adnexal torsion is nonspecific and overlapping with other more commonly encountered acute abdominal conditions such as appendicitis or ureteral stones [4][5][6]. As prompt diagnosis of adnexal torsion may reduce the risk of irreversible ovarian ischemia, it is important to be familiar with the CT findings of adnexal torsion for a timely diagnosis and to improve clinical outcomes. Several previous studies have reported CT findings of adnexal torsion [4,5,7,8]. However, most of the previous studies were descriptive studies performed in a selected series of cases, carrying the risk of biased results and limited generalizability due to the lack of controls. Furthermore, previous studies did not consider the effect of an ovarian mass in the assessment of CT findings of adnexal torsion. Most cases of adnexal torsion are associated with ovarian masses [9,10]. However, given that most ovaries containing a mass are not torsed, an ovarian mass cannot be a specific finding of adnexal torsion but it can affect CT findings [11]. For this reason, we made additional efforts to eliminate the confounding caused by ovarian masses in this retrospective case-control study. Therefore, our objective was to move beyond an ovarian mass as a CT predictor of adnexal torsion and to identify additional CT findings that could be helpful in the diagnosis of adnexal torsion.
Materials and methods
This retrospective, single-institution, case-control study was approved by our institutional review board of Seoul Metropolitan Government Seoul National University Boramae Medical Center (16-2017-26), and informed consent was waived due to the retrospective nature of the study.
Study population
The diagnosis-based search engine of our electronic medical record (EMR) system revealed 96 consecutive women with a diagnosis of ovarian/adnexal torsion from July 2011 to June 2016. Among them, 66 women who underwent multiphase CT scans including both unenhanced phase and portal venous phase were included in the study. On the basis of retrospective medical record review, 23 women were excluded from the 66 women for the following reasons: no surgical exploration (n = 9), inconclusive for adnexal torsion on surgical records (n = 12), long interval between CT imaging and surgery (n = 1), and previous hysterectomy (n = 1). Finally, 43 women (mean age, 40.8 years; range, 13-84 years) compose the case women who had met study criteria of receiving multiphase CT scans, surgically confirmed diagnosis of ovarian/ adnexal torsion, and no hysterectomy. A total of 43 control women (mean age, 41.3 years; range, 16-87 years) were selected from 1,152 consecutive women who underwent oophorectomy or salphingo-oophrectomy during the same study period. The selection of control women was performed as follows: 1) 1,152 women were arranged in the order of age. 2) One author (H.W.) selected a candidate in order, who matched with a case woman for age (± 3 years), laterality of ovarian mass, and size of ovarian mass. The size of the ovarian mass was assessed on axial CT scans and less than 20% difference based on the smaller one was considered matched. 3) Selection as a control was finalized after confirming that the operation was performed due to a reason other than ovarian/adnexal torsion (Fig 1). The median delay between CT imaging and surgery was 1 day (range, 0-13 days) in case women and 10 days (range, 0-76 days) in control women.
Clinical and pathologic data for the study population were extracted from our EMR system, with additional retrospective medical record review. The following clinical findings related to the abdominal pain were recorded: (a) presence of abdominal pain, (b) duration of abdominal pain, and (c) intensity of abdominal pain. At our institution, pain intensity was quantified and recorded by using an 11-point numerical rating scale in which 0 means no pain and 10 means the worst pain possible [12,13].
CT acquisition technique
CT images were obtained using three different types of scanners: (a) LightSpeed Pro 16 (GE Medical Systems, Milwaukee, Wi, USA) with a detector configuration, 16 Ã 1.25 mm; tube voltage, 120 kVp; noise index, 12.35 with automatic exposure control (smart mA, GE Healthcare, Milwaukee, WI, USA); gantry rotation period, 0.6 second; pitch factor, 1.375; table speed, 27.5 mm per rotation, reconstructed section width, 3.75 mm; and reconstructed section interval, 3.75 mm [n = 55], (b) Brilliance 64 (Philips, Cleveland, OH, USA) with detector configuration, 64 Ã 0.625 mm; tube voltage, 120 kVp; 300 reference mAs with automatic exposure control (DoseRight ACS, Philips, Cleveland, Ohio); gantry rotation period, 0.75 second; pitch factor, 1; table speed, 40 mm per rotation, reconstructed section width, 3 mm; and reconstructed section interval, 3 mm [n = 19], and (c) Ingenuity 128 (Philips, Cleveland, OH, USA) with detector configuration, 64 Ã 0.625 mm; tube voltage, 120 kVp; dose index, 24 with automatic exposure control (DoseRight ACS, Philips, Cleveland, OH, USA); gantry rotation period/pitch factor, automatic scan time; reconstructed section width, 3 mm; and reconstructed section interval, 3 mm [n = 12]. No oral contrast material was given for gastrointestinal tract opacification. Nonionic low-osmolar iodine contrast material containing 350 mg of iodine per milliliter (iohexol, Omnihexol 350; Korea United Pharm Co, Seoul, Korea) was intravenously administered in a volume of 1.5 mL/kg by using a power injector (Optivantage DH; Mallinckrodt Imaging Solutions, Hazelwood, MO, USA) at a rate of 2-3 mL/sec. Women underwent multiphase CT scans including at least two phases (unenhanced and portal venous phase). The scanning delay for the portal venous phase was 90 seconds after initiation of contrast material injection and the phases were obtained craniocaudally from the diaphragmatic dome to the inferior margin of the symphysis pubis. Coronal images for the portal venous phase were reconstructed in 30 of 43 case women (69.8%) and 35 of 43 control women (81.4%), with a section width/interval of 3 mm / 3 mm.
Image analysis
All CT images were reviewed retrospectively and independently on a picture archiving and communication system workstation (Marosis M-view; infinitt, Seoul, Korea) by two radiologists (M. S.L. [reader 1] and M.H.M. [reader 2] with 3 years and 14 years of experience in genitourinary imaging, respectively), who were informed of the primary aim of the study but blind to the clinical and surgical data of the study population. Prior to the study, both readers underwent a training session to standardize their criteria for CT findings of adnexal torsion. Five CT examinations from another study period were used for the training session. Using a standardized questionnaire, the readers reported the presence of the following CT findings: a) prominent peripheral follicles, b) uterine deviation towards the ovarian mass, c) a thickened pedicle without an ancillary finding, and d) a thickened pedicle with an ancillary finding. A thickened pedicle was defined as an amorphous solid mass between the ovarian mass and the ipsilateral uterine cornu [8]. A whirl sign, a navel sign, and uterine deviation facing a thickened pedicle were categorized as ancillary findings that reinforce the association of thickened pedicles with adnexal torsion (Figs 2-4). The navel sign was defined as denting or flattening of the affected ovarian surface at the presumed attachment site of the thickened pedicles. The enhancement of the ovarian parenchyma of the affected ovary was also evaluated. The regions of interest for ovarian enhancement were drawn to encompass as much remnant ovarian parenchyma as possible, and care was given to avoid partial-volume artifacts. Each drawing was repeated twice, and measured values were averaged for later analysis. No ovarian enhancement was defined as an increase of less than 10 Hounsfield units between unenhanced and portal venous phase scans.
Statistical analysis
Inter-reader agreement was evaluated by using the Cohen kappa statistic: A kappa score of 0.20 or less than 0.20 was considered slight agreement; 0.21-0.40, fair agreement; 0.41-0.60, moderate agreement; 0.61-0.80, substantial agreement and 0.81-1.00, almost perfect agreement [14]. CT findings of adnexal torsion were compared between women with adnexal torsion and women without adnexal torsion by using the Chi square test. Receiver operating characteristics (ROC) analysis was used to evaluate the diagnostic performance of the identified CT findings of adnexal torsion. The areas under the ROC curves (AUCs) were evaluated as the measure of a diagnostic test's discriminatory power and compared by using a Delong test [15]. All statistical analyses were performed using MedCalc version 16.8.4 software (Med-Calc, Ostend, Belgium). A p value of less than 0.05 was considered to indicate a statistically significant difference.
Results
Conservative treatment was performed in 11 (25.6%) of the 43 women with adnexal torsion: detorsion with cystectomy in 10, and detorsion in only 1. Definitive treatment was performed in 32 women (74.4%): salphingo-oophorectomy in 31 and oophorectomy in 1. Surgery revealed ovarian masses in 42 of the 43 women with adnexal torsion and in 43 of the 43 women without adnexal torsion. One woman was proven to have torsion of the normal ovary. Final pathologies for the ovarian masses are summarized in Table 1. A specific diagnosis could not be made in 18.6% (8/43) of the women with adnexal torsion due to massive hemorrhagic necrosis. Abdominal pain was significantly more common in women with adnexal torsion than women without adnexal torsion (97.7% vs. 23.3%, P < .01), and the intensity of the abdominal pain was also more severe in women with adnexal torsion than that in women without adnexal torsion (mean scale; 5.4 vs. 3.0, P = .01). However, there was no significant difference between the groups in the duration of the abdominal pain (mean day; 5.3 days vs. 1.8 days, P = .56) ( Table 2).
Chi square analysis revealed that the following CT findings were significantly more common in women with adnexal torsion than in women without adnexal torsion: uterine deviation toward the side of the affected ovary (P = < .01 for reader 1 and P = .02 for reader 2) and thickened pedicles with ancillary findings (P < .01 for both readers). The evaluation of the parenchymal enhancement of the affected ovary was possible in 31 (72.1%) and 31 (72.1%) of 43 women with adnexal torsion and in 19 (44.2%) and in 15 (34.9%) of 43 women without adnexal torsion for reader 1 and reader 2, respectively. The lack of ovarian enhancement was also more frequently seen in women with adnexal torsion than in women without adnexal torsion (83.9% vs. 15.8%, P < .01 for reader 1 and 87.1% vs. 3%, P < .01 for reader 2). Thickened pedicles without ancillary findings and prominent peripheral follicles showed no statistically significant differences between the two groups (Table 3). Table 4 shows the diagnostic performance of the identified CT findings in the diagnosis of adnexal torsion. Comparison of the AUCs revealed that thickened pedicles with ancillary findings were significantly superior to other identified CT findings or their combinations in the diagnosis of adnexal torsion (uterine deviation, P < .01; uterine deviation OR thickened pedicles with ancillary findings, P < .01; uterine deviation AND thickened pedicles with ancillary findings, P < .01 for both readers). When thickened pedicles with ancillary findings were used as a diagnostic criterion for adnexal torsion, we could achieve a sensitivity, specificity, positive likelihood ratio, negative likelihood ratio and accuracy of 74.4%, 97.7%, 32.00, 0.26 and 86% for reader 1 and 72.1%, 97.7%, 31.00, 0.29 and 85% for reader 2 (Table 4). For reader 1, the addition of ovarian non-enhancement as a diagnostic criterion for adnexal torsion made additional diagnosis in 6 of 11 false-negative women possible but resulted in 3 false-positive women. For reader 2, the addition of ovarian non-enhancement made additional diagnosis in 7 of 12 false-negative women possible but resulted in 3 false-positive women. The histologic diagnosis for the false-positive women is mature cystic teratoma (n = 1), mucinous borderline tumor (n = 1) and endometriotic cyst (n = 1) in reader 1 and mature cystic teratoma (n = 2) and fibrothecoma (n = 1) in reader 2.
Discussion
Adnexal torsion is defined as a twisting of adnexal structures including the ovary, the fallopian tube and their supporting structures [16]. Torsion of adnexal structures generally involves both the ovary and the fallopian tube because the wide attachment of the mesovarium or mesosalphinx to the rest of the broad ligament acts as a fulcrum [16][17][18]. Twisting of adnexal structures leads to anatomical changes that can be depicted on imaging studies including twisted pedicles, uterine deviation towards the affected ovary, prominent peripheral follicles, and decreased ovarian enhancement [4][5][6][7][8]17].
Twisted pedicles may consist of various components including the mesovarium, fallopian tube, broad ligaments, suspensory ligament, ovarian vessels and uterine vessels. Previous studies dealing with CT findings of adnexal torsion described a twisted pedicle as a tubal thickening or a whirl sign [5,8]. In a study of 25 women with surgically proven adnexal torsion, Rha et al. [8] reported that tubal thickening in the form of an amorphous solid mass around the adnexal mass was noted in 75% (15/20) of the patients and a whirl sign with a target-like appearance in 10% (2/20) of the patients. In Hiller et al.'s study of 35 women with surgically proven adnexal torsion [5], they reported that tubal thickening was present in 17% (6/35) of the subjects and a whirl sign in 5.7% (2/35) of the subjects. Our understanding is that tubal thickening could be a thickened pedicle without a whirl sign or thickened fallopian tube itself. Because tubal thickening may also be seen in diseases other than adnexal torsion such as pelvic inflammatory disease [19,20], we think tubal thickening itself cannot be a specific finding of adnexal torsion (Fig 5) and needs ancillary findings to be a specific finding of adnexal torsion (Figs 2-4).
A whirl sign is a well-known ancillary finding and a thickened pedicle with a whirl sign has been recognized as a pathognomonic sign for adnexal torsion [4,21]. In addition to the whirl sign, we added a navel sign and uterine deviation facing the thickened pedicle as ancillary findings. The navel sign is denting or flattening of the affected ovarian surface at the presumed insertion site of the thickened pedicles. Because twisting of the pedicles induces such denting or flattening on the ovarian surface, we thought that the navel sign could be a useful ancillary finding suggestive of pedicle twisting. This supposition proved true with our data that showed significant differences between women with and without adnexal torsion (39.5% vs. 2.3% for reader 1 and 48.8% vs. 2.3% for reader 2) ( Table 3). Uterine deviation towards the affected ovary is a well-known CT finding of adnexal torsion. However, our experience has shown that ovarian masses may induce uterine deviation towards the affected ovary even in women without adnexal torsion (Fig 6). Uterine deviation needs additional findings to reduce false positive cases; thus, we adopted uterine deviation facing thickened pedicles as ancillary findings of thickened pedicles. In the present study, an ovarian mass could induce uterine deviation toward the affected ovary in the cases without adnexal torsion (6/43, 14.0% for reader 1 and 7/ 43, 16.3% for reader 2), and the application of uterine deviation facing thickened pedicles led to a significant reduction in the false positive rate (14.0% vs. 0.0% for reader 1 and 16.3% vs. 0.0% for reader 2) with little sensitivity loss (51.2% vs. 51.2% for reader 1 and 39.5% vs. 37.2% for reader 2) ( Table 3). A twisted pedicle is the pathognomonic sign of adnexal torsion but it is identified in less than one third of women with adnexal torsion on CT [4]. We extended such a pathognomonic sign into thickened pedicles with a navel sign and thickened pedicles with uterine deviation as well as thickened pedicles with a whirl sign. This effort has nearly doubled the sensitivity of CT in the diagnosis of adnexal torsion (from 48.8% to 74.4% for reader 1 and from 34.9% to 72.1% for reader 2), with little loss in the specificity (from 100% to 97.7% for both readers) ( Table 3). Uterine deviation towards the affected ovary was also proven to be a CT finding of adnexal torsion. However, the combination of uterine deviation towards the affected ovary with thickened pedicles with ancillary findings did not increase the diagnostic value of thickened pedicles with ancillary findings in the diagnosis of adnexal torsion. Axial contrast-enhanced CT scan shows a thickened pedicle (arrow) between a right ovarian cystic mass (asterisk) and the right uterine cornu. The thickened pedicle is not associated with a whirls sign or a navel sign. Uterine deviation is also not noted (not seen here). Surgery revealed a right ovarian cystic mass without torsion. https://doi.org/10.1371/journal.pone.0200190.g005 Lack of enhancement is also expected to be one of CT findings suggestive of adnexal torsion because ovarian torsion leads to obstruction to venous outflow and arterial inflow. Lack of ovarian enhancement sometimes can be difficult to distinguish from non-enhancing cysts and an enlarged ovary with a mass often has no measurable solid portion for evaluation of enhancement. However, to our knowledge, it is not known about how much ovarian enhancement can be assessed quantitatively on CT scans in women with suspicion of adnexal torsion. In the present study, the evaluation of ovarian enhancement was possible in 72.1% of women with adnexal torsion and in 34.9% -44.2% of women without adnexal torsion. When the evaluation of ovarian enhancement was possible, the lack of ovarian enhancement was more frequently seen in women with adnexal torsion than in women without adnexal torsion (83.9% vs. 15.8%, P < .01 for reader 1 and 87.1% vs.3%, P < .01 for reader 2). Although considerable missing values made it difficult to compare ovarian enhancement with other CT findings, we think that ovarian non-enhancement could be also helpful in the diagnosis of adnexal torsion when the evaluation of ovarian enhancement is possible. Prominent peripheral follicles, one of the CT findings of ovarian torsion, were not common findings in our study. Furthermore, there was no significant difference in the presence of prominent peripheral follicles between women with adnexal torsion and women without adnexal torsion (14.0% vs. 11.6%, P = .75 for reader 1 and 11.6% vs. 7.0%, P = .46 for reader 2). Prominent peripheral follicles are thought to result from peripheral displacement of follicles due to ovarian stromal edema [17]. Lee et al [18] found in a retrospective review of 116 women with surgically proven adnexal torsion that women without a detectable ovarian mass showed ovarian edema more frequently than women with a visible ovarian mass. They suggested smaller ovarian stroma reduced by an ovarian mass as the reason for low detectability of ovarian edema. The same is presumed to be true for the low prevalence of prominent peripheral follicles in the present study. The majority of our study population had ovarian masses; thus, a smaller stromal portion may be affected by outflow obstruction. This could be responsible for the low prevalence of prominent peripheral follicles in our study population. The relatively low tissue contrast of CT compared to MR or US may be another explanation for such a result.
Our study has several limitations. First, our study is limited by its retrospective nature, with the potential for selection and verifications biases. The choice of a consecutive sample of women with oophorectomy or salphingo-oophrectomy as control women is another limitation of our study because selection of an enriched sample of women with non-painful masses may have biased our results. A third limitation is that the interval between imaging and surgery was lengthy for some women and the ovary may torse or detorse over time. We think this is an unavoidable limitation given that the episodes of pain can occur for several days to months before admission in some women with adnexal torsion [16]. A fourth limitation is that coronal reformation of acquired images was not available in some women. Because the use of coronal reformation may improve the detection of adnexal torsion [22], this limitation could influence the delineation of the CT findings we analyzed. Although readers were blind to the clinical and surgical data of the study population, they were informed of the primary aim of the study, and the level of interrogation of CT findings of adnexal torsion may have been higher than it would be when reading a CT in a routine clinical situation. This is another limitation of our study.
In conclusion, a thickened pedicle with ancillary findings is the most accurate CT predictor of adnexal torsion. In addition to the previously recognized whirl sign as a strong ancillary predictor of torsion, we found the addition of a navel sign, and uterine deviation facing the thickened pedicles, when seen with thickened pedicles, can increase the sensitivity of CT in the diagnosis of adnexal torsion. We hope that our results will be helpful in the management of women with adnexal torsion in whom early recognition is important to preserve the affected ovary and prevent serious complications. | 2018-08-01T19:03:37.238Z | 2018-07-11T00:00:00.000 | {
"year": 2018,
"sha1": "7f397f9bb4b8ecc8958e7f27b4111994ab21b347",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0200190&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7f397f9bb4b8ecc8958e7f27b4111994ab21b347",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
203687483 | pes2o/s2orc | v3-fos-license | Get rid of inline assembly through verification-oriented lifting
Formal methods for software development have made great strides in the last two decades, to the point that their application in safety-critical embedded software is an undeniable success. Their extension to non-critical software is one of the notable forthcoming challenges. For example, C programmers regularly use inline assembly for low-level optimizations and system primitives. This usually results in driving state-of-the-art formal analyzers developed for C ineffective. We thus propose TInA, an automated, generic, trustable and verification-oriented lifting technique turning inline assembly into semantically equivalent C code, in order to take advantage of existing C analyzers. Extensive experiments on real-world C code with inline assembly (including GMP and ffmpeg) show the feasibility and benefits of TInA.
I. INTRODUCTION
Context. Formal methods for the development of high-safety software have made tremendous progress over the last two decades [1], [2], [3], [4], [5], [6], with notable success in regulated safety-critical industrial areas such as avionics, railway or energy. Yet, the application of formal methods to more usual (non-regulated) software, for safety or security, currently remains a scientific challenge. In particular, extending the applicability from a world with strict coding guidelines and disciplined mandatory validation processes to more liberal and diverse development and coding practices is a difficult task.
Problem. We consider here the issue of analyzing "mixed code", focusing on the use of inline assembly in C/C ++ code. This feature allows to embed assembly instructions in C/C ++ programs. It is supported by major C/C ++ compilers like GCC, clang or Visual Studio, and used quite regularly -usually for optimization or to access system-level features hidden by the host language. For example, we estimate that 11% of Debian packages written in C/C++ directly or indirectly depends on inline assembly, with chunks containing up to 500 instructions, while 28% of the top rated C projects on GitHub contains inline assembly according to Rigger et al. [7]. As a matter of fact, inline assembly is a common engineering practice in key areas such as cryptography, multimedia or drivers. However, it is not supported by current state-of-the-art C/C ++ program analyzers, like KLEE [4] or Frama-C [1], possibly leading to incorrect or incomplete results. This is a clear applicability issue for advanced code analysis techniques.
Given that developing dedicated analyzers from scratch is too costly, the usual way of dealing with assembly chunks is to write either equivalent host code (e.g, C/C ++ ) or equivalent logical specification when available. But this task is handled manually in both cases, precluding regular analysis of large code bases: manual translation is indeed time-consuming and error-prone. The bigger the assembly chunks are, the bigger these problems loom.
Goal and challenges. We address the challenge of designing and developing an automated and generic lifting technique turning inline assembly into semantically equivalent C code amenable to verification. The method should be: Verification-friendly The produced code should allow good enough analyses in practice (informally dubbed verifiability), independently of the underlying analysis techniques (e.g., symbolic execution [8], [9], deductive verification [10], [11] or abstract interpretation [12]); Widely applicable It should not be tied to a particular architecture, assembly dialect or compiler chain, and yet handle a significant subset of assembly chunks found in the wild; Trustworthy The translation process should be insertable in a formal verification context without endangering soundness: as such it should maintain exactly all behaviors of the mixed code, and provide a way to show this property. Verifiability alone is already challenging: indeed, straightforward lifting from assembly to C (keeping the untyped byte-level view) does not ensure it as standard C analyzers are not well equipped to deal with such low-level C code.
Scarce previous attempts do not fulfill all the objectives above. Vx86 [13] is tied to both the x86 architecture and deductive verification, while the recent work by Corteggiani et al. [14] focuses on symbolic execution. None of them addresses verifiability or trust. At first sight, decompilation techniques [15], [16], [17] may seem to fit the bill. Yet, as they mostly aim at helping reverse engineers, correctness is not their main concern. Actually, "existing decompilers frequently produce decompilation that fails to achieve full functional equivalence with the original program" [18]. Some recent works partially target this issue: Schwartz et al. [19] do not demonstrate correctness (they instead measure a certain degree of it via testing), while Schulte et al. [18] use a correct-bydesign but intractable (possibly non-terminating) search-based method. Again, none of them study verifiability.
Proposal. We propose TINA (Taming Inline Assembly), the first automated, generic, verification-friendly and trustworthy lifting technique for inline assembly. The main insight behind TINA is that by focusing on inline assembly rather than arbitrary decompilation, we tackle a problem both more restricted (simple control-flow, smaller size) and better defined (interfaces with C code, no dynamic jumps), paving the way to powerful targeted methods. TINA relies on the following key principles: • Recent binary-code lifters [20], [21], [22] translating binary opcodes to generic low-level intermediate representations (IR) provide minimalist architecture-agnostic and well-tested IRs adapted to our goal; • While direct byte-level lifting severely hinders current C analyzers, verifiability is enhanced by dedicated transformations refining the raw original IR with C-like abstractions such as explicit variables, arithmetic data manipulation, structured control-flow, etc.; • Trust relies on translation validation [23] (validating each translation), a more tractable option than full translator validation, which reduces the trust base to a (usually simpler) checker. Here, this checker requires to prove program equivalence -a notoriously hard problem 1 .
We propose a dedicated equivalence checking algorithm tailored to our processing chain.
Contributions. In summary, this paper makes the following contributions: • A new cooperating toolchain allowing formal verification of programs mixing inline assembly and C, based on an original combination of novel and existing components (Sec. IV), addressing verifiability and trust issues; • A new principled method lifting inline assembly to highlevel C amenable to further formal analysis built upon 4 simplification steps (Sec. V) countering clearly identified threats to verifiability (Sec. II); • The automated validation of said method to make the lifter trustworthy, via a new dedicated program equivalence checking algorithm taking advantage of our transformation process to achieve both efficiency and high success rate, with a limited trust base (Sec. VI) ; • Thorough experiments (Sec. VII) of a prototype implementation on real-world examples to show its wide applicability (all Debian GNU/Linux 8.11 x86 assembly chunks, some ARM, GCC and clang) and its substantial impact on 3 different verification techniques on samples from GMP, ffmpeg, ALSA and libyuv.
Discussion.
This work targets assembly chunks as found in real-world programs: we lift and validate 76% of all assembly chunks from Linux Debian 8.11 (Table I) and benefit a range of stateof-the-art verification tools and techniques (Sec. VII-B). Still, system and floating-point instructions are currently considered out-of-scope. Especially, floats are not tackled here since handling them well is a challenge in itself for the whole toolchain (lifter, solver, verifier) -see the extended discussion in Sec. VIII. Also, TINA's implementation targets C since this is the principal language used for low-level programs, but the method itself would work unchanged on similar imperative languages, like LLVM. Finally, though some prior work has addressed code lifting for verification, it is worth noting that verifiability has never been explicitly addressed so far.
II. CONTEXT AND MOTIVATION
Consider the code snippet of Fig. 1a extracted from UDP-Cast sources. It consists of the x86 assembly code itself (here: "cld ; rep stosl ") together with a specification linking C variables to registers and declaring inputs, outputs and clobbers (i.e., registers or memory cells possibly modified by the assembly chunk). The compiler, upon encounter of such an (extended) assembly chunk, may use this specification -for example during register allocation. However, it is fully blind to the rest of the information (e.g., mnemonics) and will forward the chunk as is until code emission.
This basically tells the compiler to flush its memory cache before entering the chunk.
Informal semantics. The code "cld ; rep stosl " has the following informal semantics (Fig. 1b): put the direction flag df to 0, then fill ecx double words from the edi pointer with the value from eax. Intel's manual [24] explains that df drives the sign of the increment: when df is 0, the sign is positive. TINA produces the code in Fig. 1c: the loop from the informal semantics is there, but the lifter optimized away (see Sec. V) elements like df, eax or edi.
Running the analyzers. If we try to run industrial-strength C code analyzers on this code, we observe erratic behaviors: KLEE [4] stops with an error message; Frama-C, on the other hand, warns that 'Clobber list contains "memory" argument.
Assuming no side-effect beyond those mentioned in output operands'. ("cld; rep; " "stosl" 19 : "=c" ( __ d0), "=D" ( __ d1) 20 : "a" (0), 21 "0" (sizeof(fd _ set) / sizeof( __ fd _ mask)), This message is clear but the behavior incorrect: the keyword "memory" stipulates that all memory may be assigned but Frama-C simply ignores it. This small example shows that a single line of assembly may throw off these tools. Of course, one may manually rewrite the chunks into semantically equivalent C code, then use C analyzers, but this is error-prone and not scalable. With TINA, we are able to automatically generate the code of Fig. 1c, illustrative of our code transformations (see Sec. V), and automatically validate it (trustworthy). We can then formally show with Frama-C [1], using abstract interpretation or deductive verification, that the code indeed verifies the informal semantics laid out before (verification-friendly).
Identified threats to verifiability. Straightforward lifting from assembly to C (keeping the untyped byte-level view) does not ensure verifiability, as standard C analyzers are not well equipped to deal with such low-level C code. For example we cannot prove with Frama-C that a basic lifting of Fig. 1a meets its specification (cf. Appendix A). We identify 3 main threats to verifiability: T1. Low-level data: explicit flags -including overflows or carry, bitwise operations (masks), low-level comparisons, byte-level memory; T2. Implicit variables: variables in the untyped byte-level stack, packing of separate logical variables inside large-enough registers; T3. Implicit loop counters/index: structures indexed by loop counters at high-level are split into multiple low-level computations where the link between the different logical elements is lost. Experiments in Sec. VII-B demonstrate that a straightforward encoding (BASIC) fails to get the best of any analysis -symbolic execution, abstract interpretation, or deductive verification.
Properties of inline assembly. TINA exploits the following properties, specific to inline assembly: P1. The control flow structure is limited: only a handful of conditionals and loops, hosting up to hundreds of instructions; P2. The interface of the chunk with the C code is usually given: programmers annotate chunks with the description of its inputs, outputs and clobbers with respect to its C context; P3. Furthermore, the chunk appears in a C context, where the types, and possibly more, are known: this kind of information is sought after in decompilers, using heuristics, whereas we only need to propagate it here. All in all, the above points show that lifting assembly chunks is actually an interesting sub-problem of general decompilation, both simpler and richer in information and thus significantly more amenable to overall success.
A. Inline Assembly
We focus here on inline assembly in C/C ++ code as supported by GCC and clang. MASM (Microsoft Macro Assembler) has a different syntax but works similarly. __ asm __ volatile ("movdqa b, %xmm0\n\t" "movdqa %xmm0, a\n\t"); (a) Basic version __ asm __ ("movdqa %1, %%xmm0\n\t" "movdqa %%xmm0, %0\n\t" : "=m" ( * a) : "m" (b) : "xmm0"); Assembly chunks in the GAS syntax of GCC have two flavors: basic and (recommended) extended (see Fig. 2). Basic assembly (Fig. 2a) allows the insertion of assembly instructions anywhere in the code. They will be emitted as is during the production of the assembly file. In this case, compilers assume the chunk has no effect on its C scope, preventing safe interactions between assembly and C code -yet that does not stop developers from using it when the implicit context looks safe. In Fig. 2a, it is implicitly assumed that no optimization will occur on global variables a and b and that xmm registers are not used by default.
Extended assembly. Extended assembly allows in addition the description of the interactions with C through its inputs, outputs and clobbers (i.e., registers or memory cells whose value is rewritten by the chunk).
Such annotations work like a printf string format, as shown in Fig. 2b: some assembly operands may be replaced by placeholders referring to a list of C operands. The syntax requires binding C operands to location constraints, as in Fig. 1a and 2b. Constraints may also specify more than one location, and let the compiler choose the best way to place this operand. Common placement constraints include r to bind to a general register; m to bind to a memory address; i to an immediate value; and g which means "r, m or i". Operands may be read-only (for inputs) or write-only (for outputs) with the = modifier. A read-write operand is created either by linking an input to the same location as an output ("0" ( sizeof ( fd_set ) / sizeof (__fd_mask)) in Fig. 1a) or by using the + modifier instead of =. Without special modifiers, compilers assume read-only operands are consumed before write-only operands are produced, so that these may share the same locations. The clobber list may also contain keywords like "memory" (arbitrary memory cells may be read and/or written) and "cc" (conditional flags will be changed).
The specification of inputs, outputs and clobbers stands as a contract between the chunk and the compiler. Compilers are totally blind to what actually happens inside the chunk, relying on the contract, and will not warn about mistakes inside the chunk. Forgetting to list an input or a clobber is an easy mistake that can result in code which does not behave as expected.
Adoption. The use of inline assembly is pretty widespreadwe estimate that 11% of Debian packages written in C/C++ directly or indirectly depends on inline assembly. It includes major projects like GMP and ffmpeg -Rigger et al. [7] actually reports that 28% of the top rated C projects on GitHub uses inline assembly. We further estimate that 75% of the chunks found in Debian Jessie 8.11 (used in Sec. VII-A) serve an optimizing purpose, with an average size of approximately 10 instructions, and up to 341. Inline assembly is often used in conjunction with C macros or inlineable functions to be specialized by the compiler at each location.
B. Binary-code lifters
Binary-code lifters are the cornerstones of modern binarylevel analyzers. They are used to abstract the different binary Instruction Set Architectures (ISA) and formats into a single intermediate representation (IR) [20], [21], [22]. We rely on the IR of BINSEC [25], called DBA -other IRs are similar. Its syntax is shown in Fig. 3 DBA is a minimalist language, comprising only two types of elements (bitvector values and memory) and three instructions: assignments, jumps and conditionals. Yet, this is enough to encode the functional semantics of major ISAs -including x86 and ARM.
Binary lifters provide specialized decoders for supported architectures, in the same spirit that a compiler has one code emitter per supported architecture. Lifters are then used in disassembly algorithms to (try to) recover the semantics of the binary program. We use them to disclose the semantics of compiled assembly chunks.
IV. TAMING INLINE ASSEMBLY: AN OVERVIEW
TINA lifts inline assembly to semantically equivalent C taking advantage of properties P1-P3. This original process consists mainly of two (new) phases: verification-friendly lifting and validation, detailed respectively in Sec. V and VI. First, let us discuss the overall approach, as schematized in Fig. 4.
Compilation. We compile the source code for the target architecture with debug information. Since we control code compilation, we also include all contextual data that can help to reconstruct C code, e.g., variable names and types.
Initial low-level IR lifting [genericity]. We now start the translation per se, by lifting the code back to the IR level. The use of binary code may seem gratuitous at first sight. This is however the best place to start working, since assembly chunks are totally instantiated and embedded in their context -register names and memory locations have been resolved by the compiler. Debug information here allows to locate the assembly chunk in the compiled code.
Transformation into high-level C [verifiability]. We then lift the IR back to C, through a combination of dedicated passes aiming at refining the low-level IR with high-level information Validation [trust]. The validation phase starts by recompiling the pure C code, without optimization in order to preserve the code structure -our validation technique depends on it.
We locate the binary code corresponding to the lifted code once more, and get back its IR representation. We now possess two distinct IR pieces: this one and the one from the first compilation. We will aim to prove their semantic equivalence in Sec. VI. This step is original.
V. FROM LOW-LEVEL IR TO HIGH-LEVEL C
The goal of this lifting phase is to recover verifiable C code preserving the semantics of the original assembly chunk. The transformations at IR level mitigate the identified threats to verifiability (Sec. II), and reinforce each other (Sec. VII).
Type verification & propagation.
To lift assembly code back to C, chunk operations on bitvectors and memory need to be mapped to C operations on integers (signed/unsigned) and pointers. To this end, we propagate types from the interface into the IR operations. IR types can either be addresses (typed pointers) or values (signed or unsigned, with an associated size). Type information is further synthesized using forward propagation and constraints imposed on operands by low-level operations. This step also guarantees that inputs' and outputs' types are respected. The lifter gives concrete C types using the type size information from DWARF.
High-level predicate recovery (threat T1). Low-level conditionals use flags -zero (zf), sign (sf), carry (cf) or overflow (of) -set by previous instructions. In most situations, they have little meaning on their own and the way they are computed hampers understanding the purpose of the condition. This pass applies Djoudi et al.'s recent technique [29] based on semantic equivalence proved by SMT solvers. It substitutes the low-level condition, built on flags, by a more readable arithmetic comparison. For example, this phase recovers if (ecx + 1 > 1) goto label ; instead of if ( zf == 0 && sf == of) goto label ; from the assembly snippet "decl ecx; jg label ; ".
Register unpacking (threat T2). Assembly chunks often contain optimizations exploiting data level parallelism in order to use the full capacity of the hardware by packing multiple value inside a bigger one fitting inside a machine register. For instance, loading 4 (byte) characters inside an integer is more efficient than doing four smaller loads. The concept has been exacerbated with Single Instruction Multiple Data extensions, providing vectorized registers up to 512-bits. The issue here is that such packed code has very low-level semantics (masks, shifts, etc.). Our novel register unpacking method uncovers the independent variables stored in a container, thus preventing packed arithmetic from destroying the abstractions of the analyzers. The method amounts to splitting registers into independent variables, whose size depends on the uncovered usage, rewrite the code accordingly and then clean up unused variables and code, and rebuild higher-level chunks through dedicated simplifications. The principle is the following: if a subpart of a variable is read in the code (e.g., extract0..15 eax), then this subpart is likely to correspond to a logical entity. So we generate a fresh variable for this entity, receiving the restricted value, and replace each such extraction by this new variable. To avoid the need for a fixpoint until every variable extraction is replaced, we perform the replacement eagerly, in 3 steps: 1) A forward pass where each assignment of 8 × 2 k bits is split into multiple fresh assignments of 8 × 2 i bits where i ≤ k (for instance, eax will be split into {{al, ah, eax_16_23, eax_24_31}, {ax, eax_16_31}, {eax}}; 2) At the same time, each variable restriction extract i..j var corresponding to one of the newly generated variables is replaced by this new variable; 3) A final pass of dead code elimination removes each unused freshly generated variable. Note that subparts may overlap with each other (for instance, al, ax and eax share common parts) but we found that most of the time, only one of them survives the final step. Thus, the size of the produced code does not increase much in the end.
Finally, we also rely on the fact that expression propagations together with concatenation-extraction simplification will automatically reconstruct bigger sized variables from concatenation of smaller sized ones (e.g., ax half-word from al and ah bytes).
In Fig. 5, the chunk loads two char in a register before adding them, using the h and l prefixes to access them. Without register unpacking, the lifter uses bitmasking (Fig. 5c), making the code more complex than its clear initial intent (Fig. 5b). extern const unsigned char src [2]; unsigned char sum; __ asm __ ("movzxw %1, %k0\n\t" "addb %h0, %b0\n\t" : "=&Q" (sum) : "m" (src)); (a) Source (c) Lifting without unpacking Expression propagation (threats T1 and T2). We draw inspiration from compiler optimization techniques to devise a novel dedicated simplification mechanism geared toward our needs. In particular, we can afford very aggressive simplifications (small code size w.r.t. standard compilation setting) but we have to address particular kinds of low-level instructions (coming from IR translation). Our method originally combines eager expression propagation coupled with dedicated (lowlevel) simplifications and a posteriori control to revert fruitless propagations -when no simplification rule has been triggered.
Eager expression propagation relies on the idea that more expression propagation raises more opportunity for further simplifications by dedicated rules. Yet, systematic propagation can yield an exponential blowup of the code under analysis rather than the desired simplifications. To mitigate this problem we propose eager propagation coupled with a posteriori control to revert fruitless propagation. The algorithm works as follows: • As a preliminary step, a data flow analysis collects all symbolic values (terms) associated to each pair (name, program point) used in the IR code; • First, we unconditionally propagate symbolic values in a first pass but save a reverse map for each propagated expression (in case the propagation is not fruitful); • Second, we expect simplification rules (described below) to simplify the whole expression; • Third, we identify expressions not yet simplified (by syntactically comparing the terms before and after sim-plification) and revert back the propagation on such case thanks to the reverse map (a posteriori control); • Finally, we cleanup the code by filtering out unused variables, dead branches and dead code.
Regarding simplification rules, we use a mixture of standard and dedicated simplification rules -standard for typical integerlevel properties and dedicated for more low-level aspects. Here is a representative (incomplete) subset of these rules -see Appendix E for a complete set -where |x| denotes the size of the expression x, any binary operator, C a condition (|C| = 1), k is a constant.
Loop normalization (threat T3). This pass aims at highlighting the relations between the current iteration of the loop and the variable values. We especially look for affine relations of the form a × x + b where x is the loop iteration counter. We indeed found out that tools much prefer to analyze for ( int i = 0; i < N; i++) T[i ] = C; instead of for (char * t = T; t < T + N; t++) * t = C;. Assembly code, though, is more likely to have the second form. cltd # sign extend eax in edx xor %edx, %eax # 1-complement eax if eax < 0 sub %edx, %eax # add one to eax if eax < 0 (a) Branchless absolute value implementation 1) rebasing replaces the initial value I by 0 and each occurrence of the variable v by I + v; 2) rescaling replaces the increment k by 1 and each occurrence of the variable v by k * v; 3) merging unifies the transformed variables with the loop iteration counter. For example, in Fig. 1c, the byte-level affine relation between the counter ecx, lifted as __tina_ecx, and the moving pointer edi, based at __tina_4, is edi ≡ __tina_4 + 4 * ( __tina_3 − ecx) -the code is lifted as __tina_4 + ( __tina_3 − __tina_ecx) to take pointer arithmetic into account (__tina_4 is an int * , pointing to 4 bytes long values in x86).
VI. VALIDATION
For our translation to be trustworthy, we use a two-pronged approach: 1) We try to prove the semantic equivalence of the code prior to lifting with the lifted C code; 2) If this fails, we rely on intensive random testing (fuzzing) to increase the level of trust in the lifted C code.
Block-based semantic equivalence. The lifting process of Sec. V strives to preserve the isomorphism of the controlflow graphs based on basic blocks between the initial assembly chunk and its lifted C representation over their DBA IR representation. This property allows us to tackle the equivalence proof at basic block level. The proof of equivalence proceeds as follows: S1. We check the isomorphism of the control-flow graphs extracted from the two lifted programs. Since we deal with deterministic labeled directed graphs, this check is immediate -and usually succeeds. TINA is actually very careful during simplifications and recompilation to preserve the control-flow structure (see details below). For the isomorphism check, we track the relation between the heads of IR basic blocks and the corresponding emitted C code thanks to C labels and debug line information. If the check succeeds, we go to S2, otherwise we [fallback] on fuzzing -in practice (Sec. VII-A), the latter has never happened.
S2. Once we know the two control-flow graphs are isomorphic, we try to demonstrate the pairwise equivalence of corresponding vertices. This allows to avoid directly dealing with loops. Each pairing of basic blocks is translated to logical formulas for which we ask SMT solvers: if inputs are identical, can outputs be different? If all queries are unsatisfiable then equivalence is proven [success], otherwise we use our [fallback].
Taming simplifications. In order to help the equivalence proof succeed, TINA passes were designed to preserve the controlflow graph structure and to be traceable. For the first goal, simplifications never modify jump instruction, except for trivial dead branch elimination and the lifter avoids inserting branches with lazy constructions such as &&, || or ternary operators. For the second goal, when a simplification changes the input-output relation of a basic block, it records the changes w.r.t the old ones and these properties will be added to the assumptions of S2. For instance, in Fig. 1, the expression propagation records that eax holds the value 0 for the entire chunk. It will then be used during S2 to prove the equivalence of the loop body where the register no longer exists in the generated part (Fig. 1c).
What could go wrong? While TINA uses simplifications and lifting passes tailored to make the block-based semantic equivalence algorithm possible, the recompilation step is blind to this requirement and may therefore threaten it.
The S1 check may fail if the compiler modifies the control flow graph, for example if some elements outside of the assembly chunk render a branch dead or a loop unrollable. In Fig. 1c, since sizeof is known at compile time, clang -O1 unrolls the loop, making the isomorphism check fail.
The S2 query may fail if the compiler moves parts of the computation across basic blocks, changing the relation between inputs and outputs. It may happen during code motions, like loop-invariant code motions. In this case, the graph isomorphism still holds but the relation between basic blocks is lost. GCC -funroll-loops partially unrolls (8 times) the loop body in Fig. 1c leading to a failed equivalence query.
To avoid such problems, we recompile the code without any optimization (-O0).
Note that SMT checks never time out in our experiments (Sec. VII), probably due to the naturally small size of blockbased queries. However, we can imagine that code showing hard-to-reverse behaviors, such as cryptographic hash functions, could make the S2 query fail.
Trust base. Validation allows to increase the confidence in the lifting process, using 3 components as the trust base: the binary-code lifter, the compiler and the solver. All are well tested software and the last two are part of the trust base of (most) modern source-level verification tools anyway. Furthermore, while we trust the compiler debug information, we argue that the compilation process itself is not part of the trusted base: assembly chunks are untouched by it and validation will very likely catch errors during re-compilation. Besides, further mitigation includes systematic testing of assembly chunks vs. their IR representation, and using multiple compilers and/or solvers. Assembly chunks 3039 100 25 103 237 4 0 85 308 1 Trivial 126 0 0 6 13 0 -1 28 0 Out-of-scope 449 40 0 17 0 3 -2 0 0 Rejected 138 11 0 12 1 0 -12 2
VII. EXPERIMENTAL EVALUATION
We evaluate our implementation of TINA on 3 research questions: RQ1) How applicable is it on assembly chunks found in the wild? RQ2) How do off-the-shelf program analyzers behave on lifted code? RQ3) What is the impact of each optimization?
A. Wide applicability (RQ1)
We run our prototype on all assembly chunks found in the Linux Debian 8.11 distribution (for x86), i.e. ≈ 3000 chunks distributed over 200 packages and 1000 functions. As chunk distribution is not smooth, we also fix 2 subsets of samples: one with the 100 biggest chunks, and another with all chunks from 4 key major projects exploiting low-level optimizations: GMP, ffmpeg, ALSA and libyuv. We exclude trivial (empty or unused), out-of-scope and rejected chunks. Out-of-scope chunks include those with floating point operations, OS-level hardware instructions or hardware-based crypto-primitives, like AES. Rejected chunks are those deemed unsafe because they do not respect their interface. Yet, we activate options in our tool to specifically regard accessing flags, xmm registers or memory as safeallowing to consider 150 extra chunks as relevant, notably in ffmpeg. The statistics of Table I report on the tool's behavior with these settings.
On in-scope chunks, TINA performs extremely well, with 100% chunks lifted and fully validated (no resort to testing) -this amounts to 76% of all chunks found -for a negligible cost (0.7s per chunk on average). The biggest 100 chunks are a little less successful as they have a fair amount of (unhandled) floating-point instructions. TINA works equally well on major projects for ARM or x86, and with GCC or clang on x86 (Table II), confirming its genericity.
Experiments on both symbolic execution and abstract interpretation use 58 functions (out of 366) from the 4 key projects in Sec. VII-A, selected due to the ease of automatically generating the initial contexts for both analyses. For all 3 tools, we also report the observed differences using a basic lifter and different optimization levels: O1 (high-level predicate recovery), O2 (O1 + register unpacking), O3 (O2 + expression propagation) and O4 (O3 + loop normalization). Note that O4 is TINA. Symbolic execution. We perform our experiments with KLEE [4] which at present does not handle inline assembly chunks and stops upon meeting one -except for a very few simple cases such as assembly-level rotations. This fact can sometimes prevent the adoption of symbolic execution [33]. Table III summarizes our findings. (RQ2) First, KLEE alone can analyze only few functions (3/58) as (almost any block of) assembly stops the analysis, and none of them is fully path-covered. Adding lifting allows to analyze all considered functions (58/58), to completely path-cover 43% of them (25/58) and to explore significantly more paths within the same analysis budget (×4.7).
The lifting strategy (RQ3) does not impact the functions that KLEE can fully cover, but TINA optimizations considerably speed up code exploration, enabling to cover significantly more paths (×4) than basic lifting in the same amount of time. This is explained by TINA-produced code being higherlevel, with fewer instructions and local variables, thereby accelerating SMT-solving. Note that control-flow structure, and thus total number of paths, does not change. Moreover, each optimization step brings some degree of improvement. The major improvement gaps here are brought by register unpacking (O2) and expression propagation (O3). As expected, loop normalization (O4) has no impact as symbolic execution simply unrolls loops. Additional experiments (Supplementary material, Table VII) demonstrates that high-level recovery (O1) has also a substantial impact on the analysis (removing it leads to 5.4M explored paths, vs 6.6M in full TINA).
Abstract interpretation. We use the Frama-C EVA [30] plugin. Frama-C has limited support for inline assembly based on interfaces, translating them into logical assigns annotations for modified variables -safely interpreted in EVA (and WP) as non-deterministic assignments. Table IV The lifted C code also contains new alarms (17/58) which we could not detect before and should be taken into account (usually out-of-bounds or other memory accesses). We also found some possibly buggy behaviors (Sec. VII-D).
For short, we observe positive impact from TINA w.r.t. nonlifted code on 77% (45/58) of the functions (more precision, reducing alarms from over-approximations of inline assembly, or new memory alarms in lifted code) . Table V additionally shows the impact of the lifting strategy (RQ3). Compared with basic lifting, each additional optimization increases the quality of the lifted code (fewer ASM and total alarms) and the precision of the analysis (more functions without alarms, fewer memory alarms, more errors) -including loop normalization which allows finer approximations of loop fixpoints (widening). TINA (O4) thus significantly improves all these aspects. Moreover, the produced alarms are more precise: possible buffer overflows (such as a ffmpeg -1 index access -see Appendix D) are now recognized as errors and not mere alarms. Additional experiments (Supplementary material, Table VII) demonstrates that removing any of the optimization steps leads us quite far from the whole chain result. Weakest precondition calculus. We use the deductive verification Frama-C plug-in WP [34], [32]. We take 12 assemblyoptimized functions (see details in Supplementary, Table IX): 6 excerpts from ffmpeg, GMP, libyuv, libgcrypt and UDPCast, 2 others adapted from optimized assembly snippets and 4 translated examples from ACSL by example [35]. Functional specifications and loop invariants are manually inserted before verification, as usual for WP-based methods -we do not insert any other annotation. Moreover, recall that without lifting, assembly chunks are correctly over-approximated by non-deterministic assignments to the modified C variables. Table VI details our results. The unlifted code does not require invariants (no C-level loops), while lifted codes all require identical invariants as they share the same controlflow structure. A quick glance at Table VI shows that (RQ2) while WP without lifting never succeeds and basic lifting is far from enough (1/12), TINA does allow to prove the functional correctness of all functions (12/12). The simple over-approximations of assembly chunks provided by Frama-C without lifting are not sufficient to prove properties as strong as functional correctness.
Regarding optimization steps (RQ3), it turns out that loop normalization (O4) has no direct impact since the user must provide manual loop invariants. On the other hand, all other steps are complementary (Table VI) and crucial: removing only one of them yields at best a 6/12 success rate (Supplementary material, Table VII).
C. Conclusion
Experiments show that our code lifting method is highly practical (100% Debian 8.11 in-scope blocks are lifted and validated), that it has a positive and significant impact on all 3 formal verification tools considered -allowing them to effectively handle code with inline assembly, and, finally, that full TINA (O4) is needed to facilitate further code analysesas less refined lifting yields poorer analyses.
Interestingly, all analyses do not behave the same w.r.t the optimization chain: symbolic execution mostly takes advantage of register unpacking and expression simplifications, abstract interpretation is sensitive to the 4 optimization steps and weakest precondition calculus strongly requires all of them but loop normalization -which is already granted by user-supplied loop invariants.
D. Epilogue: post-analysis considerations
We found 567 compliance issues during our experiments. Most have no impact with current compilers but may induce bugs out of compiler changes, maintenance or code reuse.
While evaluating verifiability, we ran into 6 potential buffer overflows hidden in assembly chunks. For example, a ffmpeg function accesses index −1 of its input buffer -this is actually reported in the comments (see details in Appendix D). All errors initially reported by Frama-C EVA were also reproduced with KLEE. After determining and adding relevant logical preconditions, we were able to show the absence of runtime errors in the reported "corrected" functions. Besides, we were able to prove (with Frama-C WP) the functional correctness of 6 functions from the Debian distribution code base, including SumSquareError (24 assembly instructions).
A. Threats to validity
Benchmark representativeness. The considered code base is quantitatively and qualitatively representative of the use of inline assembly: it is extensive and comprises highly popular and respected projects. We mainly experiment on GCC and x86, but our experiments on ARM and clang show our results also hold in these settings. Still, we obviously miss closed-source software and code which relies on Microsoft's C compiler (different assembly syntax). Yet, there is no reason to believe it would behave differently.
Verification methods. We consider three of the most popular verification techniques (symbolic execution, abstract interpretation, deductive verification), representative of the major classes of analysis, both in terms of goal (bug finding, runtime error checking and proving functional correctness) and underlying core technologies (domain propagation, constraint solving & path exploration, first-order reasoning). Also, we rely on wellestablished verification tools, each applied in several successful industrial case studies. Thus, we reckon that our experiments support our claim regarding the general verifiability of the codes TINA produces.
B. Limitations
Our lifting has two main limitations: hardware-related instructions and floating-point operations.
Since we aim to lift assembly chunks back to C, the support of hardware related instructions cannot be achieved outside of modeling hardware in C as well -for example, neither DBA IR nor C can make direct reference to hardware interrupts. Here we probably cannot do better than having two (approximated) C models of hardware instructions, one for over-and one for under-approximations. While not necessarily that difficult for reasonable analysis precision, this is clearly a manpowerintensive task.
The float limitation is primarily due to the lack of support in BINSEC. Adding such support is also manpower-intensive, but not that hard. Yet, the real issue is that efficient reasoning over floats is still ongoing scientific work in both program analysis and automated solvers (e.g., theory support is new in SMT-LIB [36], only 2 solvers in the relevant category of SMT-COMP 2018). As such, it threatens our validation part, and most program analyzers would not be able to correctly handle these lifted floats anyway. Despite these limits, we still lift and validate 76% of assembly chunks of a standard Linux distribution.
Finally, our technique is amenable, to a certain extent, to standalone assembly code or even binary code decompilation. However, this case can quickly deteriorate to the usual difficult problem of lifting an arbitrary program. Especially, dynamic jumps or large-size complicated CFG would probably yield serious issues.
IX. RELATED WORK
Though some prior work has addressed code lifting for verification, it is worth noting that verifiability has never been explicitly addressed so far. We hereafter review approaches (partly) related to our method.
Assembly code lifting and verification. Maus [13], [37] proposes a generic method simulating the behavior of assembly instructions in a virtual machine written in C. This work was used by the Verisoft project to verify the code of an hypervisor consisting of mixed low-level code. Maus' technique relies on VCC [38] to write and prove verification conditions regarding the state of its machine. While we strive to produced high-level code, Maus' virtual code contains all low-level code details, including flags.
Further work by Schmaltz and Shadrin [39] aims (only) at proving the ABI compliance of the assembly chunks. This method is however restricted to MASM and the Windows operating system. TINA, here applied to GCC inline assembly, is independent of the assembly dialect by leveraging binary level analyzers and is applicable to a wider range of architectures.
Fehnker et al. [40] tackle the analysis of inline assembly for ARM architecture, using a model-checking based syntactic analysis to integrate C/C ++ analyses with inline assembly. This solution is however limited by its purely syntactic basis: first, it is restricted to one single inline assembly dialect; second it loses the soundness properties we target. Losing soundness may be an appropriate practical trade-off but not when targeting sound formal analyses.
Corteggiani et al. [14] also use code lifting within their framework. However, their end goal is to perform dynamic symbolic analyses on the produced lifted code. Sec. VII-B shows that such very targeted lifting may not be enough for other formal analyses. Moreover, correctness of the translation is not addressed.
Myreen et al. [41] targets verification of pure assembly code. The translation corresponds to our basic lifter, yet the approach proves the initial lifted IR is semantically equivalent to a very detailed ISA model. This paper then targets verification at the level of assembly code but requires code annotations and interactive proving. Our proposal targets the lifting of inline assembly within C for (general) verification purposes, is geared at ensuring the verifiability of the produced code, and its validation establishes the correctness of the IR transformations producing the final extracted C code.
Decompilation. Decompilation [15], [16], [17] tackles the challenge of recovering the original source code (or a similar one) from an executable. This goal is very difficult and requires hard work to find back the information lost during compilation [42]. Despite significant recent progress [19], decompilation remains an open challenge. Still, it is used to enhance program understanding, e.g., during reverse engineering. As such, correctness is not the main concern -for example it does not always need produce compilable source code.
Soundness is addressed by two recent works. Schulte et al. [18] use search-based techniques to generate source-code producing byte-equivalent binaries to the original executable. This technique, when it succeeds, ensures soundness by design but it is only applied to small examples, with limited success. Brumley et al. [19] on the other hand use testing to increase trust in their lifted code.
We do draw inspiration from some decompilation techniques for type reconstruction [43], [44]. Even though we do not construct types that are not derived from inputs, it helps in strengthening our type system.
Recovering the instructions and CFG of the code under analysis is a big challenge in decompilation [45], [46], especially for adversarial codes like malware. The regularity and patterns of managed codes allow a very good recovery in practice [43] by unsound methods, yet without any guarantee. Inline assembly chunks have more limited behaviors (clear control-flow, no dynamic jumps) and the fact that we control compilation makes it a non-issue for us.
Binary-level program analysis. For more than a decade now, the program analysis community has spent significant efforts on binary-level codes [47], either to analyze source-less programs (malware, COTS) or to check the code that is really running. The efforts have mainly been concerned with safe high-level abstraction recovery [48], [49], [50], [29], [51] and invariant computation.
Several generic binary lifters have been produced [20], [21], [22], reducing complex ISAs to a small set of semantically well-defined primitives. Though well tested [22], more trust could be achieved if lifters were automatically derived from something akin to ARM's formal specifications [52].
Mixed code problems. Morrisett et al. [53] have proposed Typed Assembly Language to ensure memory and control flow integrity in low-level assembly. Patterson et al. [54] have exploited the idea to mix low-level code with functional languages. We borrow some elements to propagate types between C and inline assembly.
Translation validation and code equivalence. In order to achieve safe lifting, we use translation validation [55], [23], [56], a technique also used in CompCert register allocation [57]. Our formal needs thus rely on well-established and tested tools (here SMT solvers), usable as blackboxes, instead of a full formal proof of the whole lifting chain.
Program equivalence checking is considered a challenging verification task. Dedicated approaches start to emerge, like relational weakest precondition calculus [58] (for proof) or relational symbolic execution [59] (for bug finding).
X. CONCLUSION
We have presented TINA, a method enabling the analysis of C/C ++ code mixed with inline assembly, by lifting the assembly chunks to equivalent C code. This method is the first to generate well-structured C code amenable to formal analysis through a dedicated principled succession of transformations geared at improving the verifiability of the produced code. To boot, translation validation builds trust into the lifting process. Thorough experiments on real-world code show that TINA is widely applicable (100% of in-scope chunks from Linux Debian Jessie 8.11 are validated) and that its semantic transformations positively (and significantly) impact popular verification techniques. Tables III, V and VI demonstrated the positive impact of stacking optimizations (levels O1 to O4) on common analyzers -recall that TINA is O4. What happens, now, if we remove any of the 3 initial optimizations from TINA? Here, we extend previous experiments with a new set of optimization levels: • O1 : O4 -high-level predicate recovery; • O2 : O4 -register unpacking; • O3 : O4 -expression propagation. Note that O4 (i.e., O4 -loop normalization) is actually O3, hence this case is not discussed.
Symbolic execution. Conclusion. Extended experiments show that the 4 optimizations are complementary and necessary to obtain the best of the analyzers. Register unpacking has the greatest impact on KLEE while high-level predicate recovery is necessary for the integer abstractions used in Frama-C EVA. Removing any optimization outside of loop normalization actually kills the effectiveness of Frama-C WP. The size of the generated code w.r.t the size of the assembly chunk is indicative of what one can expect of the code produced by TINA. Since the size of the C code is also correlated with the quality of subsequent analyses when produced by TINA, this information becomes more notable. Broadly speaking, the smaller the code, the easier it will be to analyze and the better the analysis will be. Table XI shows the ratio between the number of generated C statements and the number of assembly instructions. As expected, two optimizations significantly change the ratio. First, register unpacking increases the number of generated statements as it splits single assignments into many smaller assignments of independent values. On the other hand, expression propagation greatly reduces the number of statements: on average, this number decreases by a factor of 6 when it comes from register unpacking and by a factor of 2.5 in other cases. In the end, TINA roughly generates one C statement per assembly instruction for the chunks of the Debian distribution. 18 "paddb %%mm1, %%mm2 \n\t" // L + T -LT 19 "movq %%mm4, %%mm5 \n\t" // L 20 "pmaxub %%mm1, %%mm4 \n\t" // max(T, L) 21 "pminub %%mm5, %%mm1 \n\t" // min(T, L) 22 "pminub %%mm2, %%mm4 \n\t" 23 "pmaxub %%mm1, %%mm4 \n\t" 24 "psubb %%mm4, %%mm3 \n\t" // dst -pred 25 "movq %%mm3, (%3, %0) \n\t" 26 "add $8, %0 \n\t" 27 "movq -1(%1, %0), %%mm0 \n\t" // LT 28 "cmp %4, %0 \n\t" 29 " jb 1b \n\t" 30 : "+r" (i) 31 : "r" (src1), "r" (src2), 32 "r" (dst), "r" ((x86 _ reg) w)); Sec. VII-D refers to a ffmpeg function accessing index −1 of its input buffer. It is part of the lossless video encryption provided by ffmpeg. More specifically, it is a field of struct LLVidEncDSPContext (Fig. 8a) which is initialized dynamically, depending on the hardware capabilities. Comments show that the developers know of this behavior. Fig. 8b and 8c show the original and generated version respectively of the function implementation using mmxext extensions. The latter has been sliced for readability reasons and the ellipsis stands for repetitive patterns. (each MMX instruction is translated to 8 independent C statements) The −1 access occurs during the first iteration of the loop, when %0 (eax) is 0, at line 14 ( Fig. 8b) in the assembly chunk. This bug is consequently found at line 27 in the lifted C version (Fig. 8c).
Since the situation is acknowledged in the code documentation, it is arguably not a bug sensu stricto. However, the function is exported (not static), does not contain defensive programming to avoid the bad behavior, and the documentation is not directly on the function itself but in a record containing a possible function pointer to this function. Thus, we consider this a serious programming flaw that could lead to issues down the line in several situations: code maintenance and refactoring, code reuse (in other projects), compiler upgrades (taking advantages of potential undefined behaviors to trigger more aggressive code optimizations), etc.
Compliance issues. Moreover, the chunk interface (Fig. 8b, lines 30-32) misses information about clobbering the mm registers 0 to 5 and accessing memory from src1, src2 and dst. It is actually not an issue here as caller functions assumes (according to the Application Binary Interface) that memory and mm registers are clobbered. However, it is not easy to know in advance how compilers will handle function inlining in terms of memory write barriers and register clobbering. The missing information at the interface of the chunk for clobbered entities could thus lead to serious issues when compiling in a different environment. Similar concerns arise in the cases of code reuse, especially in another project.
APPENDIX E TECHNICAL FOCUS ON SIMPLIFICATION RULES
We use a mixture of standard and dedicated simplification rules -standard for typical integer-level properties and dedicated for more low-level aspects. In the rules below, we use the following notations |x| is the size of the expression x, any binary operator, C a condition (|C| = 1), k is a constant. 1 |x| denotes a bitvector of size |x| with all bits set to 1.
• standard "lightweight" term normalization in order to ease further other simplifications, including common subexpression elimination (a.k.a. sharing), unused variables elimination, associativity-commutativity re-ordering. • constant propagation (modular arithmetic) • neutral elements: • absorbing(-like ) elements: x urem 1 → 0 x srem 1 → 0 • inverse elements: x − x → 0 x udiv x → 1 x shl k |x|≤k → 0 x ⊕ x → 0 x sdiv x → 1 x shr k |x|≤k → 0 uext k (uext k (x)) → uext k (x) • condition simplifications The ecx register stands as the loop counter, the edi register is rebased, rescaled and unified with ecx. Before merging edi with ecx, the following relation is recorded: edi ≡ __tina_4 + 4 × ( __tina_3 − ecx). Fig. 10 shows the translation validation of the running example (Fig. 1). Codes before and after lifting have the same form (S1), as shown in Fig. 10a. Original blocks B 0 , B 1 and B 2 have been paired respectively with their lifted counterparts B 0 , B 1 and B 2 for S2. Let us focus here on the equivalence check between B 2 and B 2 (Fig. 10b). The two blocks are obviously syntactically different, due to simplification and recompilation. Compilation splits complex expressions using general registers as temporary variables while the ones from B 2 have been lifted to C variables with a close but different name. For instance __tina_ecx is equal to ecx whereas eax is used differently in B 2 and B 2 . Due to simplification passes, the two blocks no more have the same number of inputs or outputs, because lifting inferred some constraints and removed unused variables. Thus, eax and df are actually 0 upon entering B 2 . As pointer edi is incremented by 4 while ecx is decremented by 1, the two variables are linked by the linear relation edi = tina_4 + 4 × (tina_3 − ecx). The formula is built to take into account these differences by adding logical assertions for each inferred constraint. In the end, the equivalence query is then discharged ; B 2 instructions sequence eax'0 = tina_3in eax'1 = eax'0 − __tina_ecxin edx'0 = 4 × eax'1 eax'2 = tina_4in eax'3 = eax'2 + edx'0 memory'out = store4 memoryin eax'3 0x00000000 __tina_ecxout = __tina_ecxin − 1 ; inputs restrictions eaxin = 0x00000000 dfin = false ediin = tina_4in + 4 × (tina_3in − ecxin) ecxin = __tina_ecxin ; outputs restrictions ecx'out = __tina_ecxout edi'out = tina_4in + 4 × (tina_3in − ecx'out) | 2019-03-15T08:39:27.000Z | 2019-03-15T00:00:00.000 | {
"year": 2019,
"sha1": "6b1e4961393c03c68d0799baf122836cd3eec4bf",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1903.06407",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "6b1e4961393c03c68d0799baf122836cd3eec4bf",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
119303309 | pes2o/s2orc | v3-fos-license | Analysis of the consistency of parity-odd nonbirefringent modified Maxwell theory
There exist two deformations of standard electrodynamics that describe Lorentz symmetry violation in the photon sector: CPT-odd Maxwell-Chern-Simons theory and CPT-even modified Maxwell theory. In this article, we focus on the parity-odd nonbirefringent sector of modified Maxwell theory. It is coupled to a standard Dirac theory of massive spin-1/2 fermions resulting in a modified quantum electrodynamics (QED). This theory is discussed with respect to properties such as microcausality and unitarity, where it turns out that these hold. Furthermore, a priori, the limit of the theory for vanishing Lorentz-violating parameters seems to be discontinuous. Since it is not clear, whether this is a gauge artifact, the cross section for a physical process - modified Compton scattering - is calculated numerically. Despite the numerical instabilities occurring for scattering of unpolarized electrons off polarized photons in the second physical polarization state, it is shown that for Lorentz-violating parameters much smaller than 1 the modified theory approaches standard QED, which is strengthened by analytical investigations. Hence, the theory proves to be consistent, at least with regard to the investigations performed. This leads to the interesting result of having a well-defined parity-odd extension of QED.
I. INTRODUCTION
Modern quantum field theories are based on fundamental symmetries. This holds for QED as well as for the standard model of elementary particle physics. Whenever physicists talk about symmetries they usually think of gauge invariance or the discrete symmetries charge conjugation C, parity P, and time reversal T. However, there is one symmetry that often takes a back seat: Lorentz invariance. This is not surprising, since until now there had been no convincing experimental evidence for a violation of Lorentz invariance. 1 However, a violation of other symmetries is part of the everyday life of any high energy physicist. For example, violations of P and CP were measured long ago [8,9] and a broken electroweak gauge symmetry with massive W ± , and Z 0 bosons is an experimental fact. Why then should Lorentz symmetry and its violation not be of interest?
There exist good theoretical arguments for Lorentz invariance being a symmetry that is restored at low energies [10]. At the Planck length the topology of spacetime may be dynamical, which could lead to it having a foamy structure. The existence of such a spacetime foam [11,12] may define a preferred reference frame -as is the case for a liquid in a glass of water -and thus violate Lorentz invariance. Since a fundamental quantum theory of spacetime is still not known, we have to rely on well-established theories such as the standard model or special relativity for a description of Lorentz violation. By introducing new parameters that deform these theories it is possible to parameterize Lorentz violation on the basis of standard physics. One approach is to modify dispersion relations of particles. However, such a procedure is very ad hoc and it is not evident where the modification comes from. Therefore, a more elementary possibility is to parameterize modifications on the level of Lagrange densities. A collection of all Lorentz-violating deformations of the standard model that are gauge invariant is known as the Lorentz-violating extension of the standard model [13]. The minimal version of this extension relies on power-counting renormalizable terms, whereas the nonminimal version also includes operators of mass dimension d > 4 (see e.g. the analyses performed in [14][15][16]).
The theoretical consistency of the standard model itself has been verified by investigations based on Lorentz-invariant quantum field theory that were performed over decades (see for example [17]). However, it is not entirely clear if a Lorentz-violating theory is consistent. Some results on certain sectors of the standard model extension already exist [18][19][20][21][22][23][24][25], but there still remains a lot what we can learn about Lorentz-violating quantum field theories. Because of this it is very important to check Lorentz-violating deformations with respect to fundamental properties such as microcausality and unitarity. Furthermore, it is of significance, whether the modified theory approaches the standard theory for arbitrarily small deformations. The purpose of this paper is to investigate these questions.
Especially in the case where Lorentz violation resides in the photon sector, it can lead to a variety of new effects, for example a birefringent vacuum [13], new particle decays [26,27], and "aether-like" deviations from special relativity, which are modulated with 1 At the end of September 2011 this seemed to change with the publication of the result by the OPERA collaboration, which claimed to have discovered Lorentz violation in the neutrino sector [1]. A large number of theoretical models emerged trying to explain the observed anomaly, for example by Fermi point splitting [2], spontaneous symmetry breaking caused by the existence of a fermionic condensate [3], or a multiple Lorentz group structure [4]. However, the physics community remained sceptical and articles were published trying to explain the result by an error source that had not been taken into account [5][6][7]. Unfortunately, at the 25th International Conference on Neutrino Physics and Astrophysics OPERA announced that their new measurement yields a deviation of the neutrino velocity from the speed of light, which is consistent with zero. Now again all laws of nature seem to obey Lorenz invariance. the rotation of the earth around the sun (e.g. [28,29]). From an experimental point of view, photons produce clean signals making the photon sector very important, in bounding Lorentz-violating parameters.
There exist two gauge-invariant and power-counting renormalizable deformations of the photon sector: Maxwell-Chern-Simons theory (MCS-theory) [30] and modified Maxwell theory [13,31]. Each Lagrangian contains additional terms besides the Maxwell term of standard electrodynamics. The consistency of the isotropic and one anisotropic sector of modified Maxwell theory was already shown in [24]. In this article a special sector, that violates parity and is supposed to show no birefringence, will be investigated.
The paper is organized as follows. In Sec. 2 modified Maxwell theory is presented and restricted to the parity-odd nonbirefringent case. Additionally, it is coupled to a standard Dirac theory of massive spin-1/2 fermions, which leads to a theory of modified QED. In Secs. 3 and 4, we review the nonstandard photon dispersion relations and the gauge propagator, which are determined from the field equations [22,23]. That completes the current status of research concerning this special sector of modified Maxwell theory. The successive parts of the article deal with the main issue, beginning with the deformed polarization vectors, which can also be obtained from the field equations. After setting up the building blocks we are ready to discuss unitarity in Sec. 6 and microcausality in Sec. 7. The subsequent two sections are devoted to the polarization vectors themselves. Since their form is rather uncommon -even when considering Lorentz-violating theories -we make comparisons with MCS-theory and other sectors of modified Maxwell theory. It will become evident that the polarization vectors have a property that distinguishes them from the polarization vectors of standard electrodynamics, even in the limit of vanishing Lorentz violation. To test, whether or not some residue of the deformation remains in this limit, in Sec. 9 we compute the cross section of the simplest tree-level process involving external modified photons that is also allowed by standard QED: Compton scattering. We conclude in the last section. Readers may skip Secs. 4 -8 on first reading.
A. Action and nonbirefringent Ansatz
In this article, we focus on modified Maxwell theory [10,13,31]. This particular Lorentzviolating theory is characterized by the action . The fields are defined on Minkowski spacetime with global Cartesian coordinates (x µ ) = (x 0 , x) = (c t, x 1 , x 2 , x 3 ) and metric g µν (x) = η µν ≡ diag (1, −1, −1, −1) . The first term in Eq. (2.1b) represents the standard Maxwell term and the second corresponds to a modification of the standard theory of photons. The fixed background field κ µν̺σ selects preferred directions in spacetime and, therefore, breaks Lorentz invariance.
The second term in Eq. (2.1b) is expected to have the same symmetries as the first. These correspond to the symmetries of the Riemann curvature tensor, which reduces the number of independent parameters to 20. Furthermore, a vanishing double trace, κ µν µν = 0, is imposed. A nonvanishing κ µν µν can be absorbed by a field redefinition [13] and does not contribute to physical observables. This additional condition leads to a remaining number of 19 independent parameters.
Modified Maxwell theory has two distinct parameter sectors that can be distinguished from each other by the property of birefringence. The first consists of 10 parameters and leads to birefringent photon modes at leading order Lorentz violation. The second is made up of 9 parameters and shows no birefringence, at least to first order with respect to the parameters. Since the 10 birefringent parameters are bounded by experiment at the 10 −32 level [32], we will restrict our considerations to the nonbirefringent sector, which can be parameterized by the following Ansatz [33]: with a constant symmetric and traceless 4 × 4 matrix κ µν . Here and in the following, natural units are used with = c = 1, where c corresponds to the maximal attainable velocity of the standard Dirac particles, whose action will be defined in Sec. II C.
There exists a premetric formulation of classical electrodynamics, that is solely based on the concept of a manifold and does not need a metric. In this context a tensor density F (electromagnetic field strength) and pseudo-tensor densities H, J (electromagnetic excitation and electric current) are introduced. Since the resulting field equations for these quantities are under-determined, an additional relation between F and H has to be imposed, which is governed by the so-called constitutive four-tensor χ. Modified Maxwell theory emerges as one special case of this description, namely as the principal part of the constitutive tensor previously mentioned [35,36]. In Eq. (D.1.80) of the book [36] the nonbirefringent Ansatz of Eq. (2.2) can be found, as well. Section D.1.6 gives a motivation for it as the simplest -but not the most general -decomposition of the principal part of χ.
Furthermore, note that a special sector of CPT-even modified Maxwell theory arises as a contribution of the one-loop effective action of a CPT-odd deformation involving a spinor field and the photon field [34].
B. Restriction to the parity-odd anisotropic case
The anisotropic case considered concerns the parity-odd sector of modified Maxwell theory (2.1) with the Ansatz from Eq. (2.2). This case is characterized by one purely timelike normalized four-vector ξ µ and one purely spacelike four-vector ζ µ containing three real pa-rameters κ 01 , κ 02 , and κ 03 : where (2.3a) is the most general Ansatz for a symmetric and traceless tensor constructed from two four-vectors. The second term on the right-hand side of (2.3a) vanishes for the special choice (2.3b).
With the replacement rules given in [37], we can express our parameters in terms of the Standard Model Extension (SME) parameters [31,33]: (12) . (2.4c) Hence, the case considered here includes only parity-violating coefficients. This parity-odd case may be of relevance, since it might reflect the parity-odd low-energy effective photon sector of a quantum theory of spacetime. Besides five parameters of the birefringent sector of modified Maxwell theory, whose coefficients are already strongly bounded, there is only one alternative parity-odd Lorentz-violating theory for the photon sector, which is gauge-invariant and power-counting renormalizable: MCS theory [30]. However, the MCS parameters are bounded to lie below 10 −42 GeV by CMB polarization measurements [38].
Since the bounds are not as strong for the parity-odd case of nonbirefringent modified Maxwell theory defined by Eqs. (2.3), a physical understanding of this case is of importance.
C. Coupling to matter: Parity-odd modified QED Modified photons are coupled to matter by the minimal coupling procedure to standard (Lorentz-invariant) spin- 1 2 Dirac particles with electric charge e and mass M. This results in a parity-odd deformation of (QED) [39][40][41], which is given by the action S parity-odd modQED κ 0m , e, M = S parity-odd modMax κ 0m + S Dirac e, M , Equation (2.6) is to be understood with standard Dirac matrices γ µ corresponding to the Minkowski metric η µν .
III. DISPERSION RELATIONS
The field equations [13,31,33] of modified Maxwell theory in momentum space, lead to the following dispersion relations [22] for the two physical degrees of freedom of electromagnetic waves (labeled λ = 1, 2): for wave vector k = (k 1 , k 2 , k 3 ) and with the terms linear in the components k m explicitly showing the parity violation. To first order in κ 0m , the dispersion relations are equal for both modes, but they differ at higher order. 2 With the modified Coulomb and Ampère law it can be shown that the dispersion relations (3.2) indeed belong to physical photon modes. The procedure given in [13] eliminates dispersion relations of unphysical i.e. scalar and longitudinal modes from the field equations. The two are given by where the index "0" refers to the scalar and the index "3" to the longitudinal degree of freedom of the photon field. The dispersion relations (3.2) can be cast in a more compact form by defining components of the wave-vector k which are parallel or orthogonal to the background "three-vector" ζ: where k ∈ (−∞, ∞) and k ⊥ ∈ [0, ∞). By doing so, it is possible to write the dispersion relations (3.2) as follows: where the three Lorentz-violating parameters κ 01 , κ 02 , and κ 03 are contained in the single parameter E that is defined as It is obvious that E ∈ [0, ∞), whereas each single parameter κ 01 , κ 02 , and κ 03 can be either positive or negative. From the first definition of Eq. (3.4) we see that negative parameters κ 01 , κ 02 , κ 03 are mimicked by a negative k .
The phase and group velocity [42] of the above two modes can be cast in the following form for small enough E: where θ is the angle between the three-momentum k and the unit vector ζ: cos θ = k · ζ.
To leading order in E, the velocities above are equal: Furthermore, Eqs. (3.6), (3.7) show that both phase and group velocity can be larger than 1. However, what matters physically is the velocity of signal propagation, which corresponds to the front velocity [42]: Equation (3.9) can be interpreted as the velocity of the highest-frequency forerunners of a signal. As can be seen from Eqs. (3.6), v ph and hence also v fr do not depend on the magnitude of the wave vector, but only on its direction. For E ≪ 1, we obtain v fr, 1 ≃ v fr, 2 ≡ v fr , where v fr < 1 for π/2 < θ < 3π/2 , Observe that, for small enough E, having v fr < 1 or v fr ≥ 1 does not depend on the Lorentzviolating parameters but only on the direction in which the classical wave propagates. For completeness, we also give the phase velocities for propagation parallel and orthogonal to ζ: Note that the latter results are in agreement with the inequalities of Eq. (3.10). We conclude that the front velocity can be larger than 1 for the wave vector pointing in certain directions. That leads us to the issue of microcausality, which will be discussed in Sec. VII.
IV. PROPAGATOR IN FEYNMAN GAUGE
So far, we have investigated the dispersion relations of the classical theory. For a further analysis, especially concerning the quantum theory, the gauge propagator will be needed. The propagator is the Green's function of the free field equations (3.1) in momentum space. In order to compute it the gauge has to be fixed. We decide to use Feynman gauge [41,43,44], which can be implemented by the gauge-fixing condition The following Ansatz for the propagator turns out to be useful: The propagator coefficients a = a(k 0 , k), . . ., g = g(k 0 , k) and the scalar propagator part K 1 = K 1 (k 0 , k) follow from the system of equations ( G −1 ) µν G νλ = i δ µ λ with the differential operator in Feynman gauge transformed to momentum space. Scalar products ξ µ ξ µ , ζ µ ζ µ , and ξ µ ζ µ will be kept in the result, in order to gain some insight in the covariant structure of the functions. However, we remark that, for the case considered, ξ 2 ≡ ξ µ ξ µ = 1, ζ 2 ≡ ζ µ ζ µ = 4E 2 , and ξ · ζ ≡ ξ µ ζ µ = 0. Specifically, the propagator coefficients and the scalar propagators K 1 and K 2 , where K 2 appears in some of these coefficients, are given by .
where definition (4.5g) enters (4.5b). The poles of K 1 and K 2 can be identified with the dispersion relations obtained in Sec. III.
the dispersion relation (3.2a) of the λ = 1 mode is recovered. Similarly, the dispersion relation (3.2b) of the λ = 2 mode follows from K 2 (ω 2 , k) −1 = 0, that is The third pole k 2 = 0 corresponds to the dispersion relation of scalar and longitudinal modes. This is clear from the fact that this pole appears only in the gauge-dependent coefficients b, d, and f . These are multiplied by at least one photon four-momentum and vanish by the Ward identity, 3 if they couple to a conserved current [44]. Since the Ward identity results from gauge invariance, it also holds for modified Maxwell theory, which is expected to be free of anomalies [13]. Because of parity violation the physical poles are asymmetric with respect to the imaginary k 0 -axis. The above result (4.2)-(4.5) equals the propagator given in [23]. Every propagator coefficient, which contains the scalar propagator K 2 , is also multiplied by K 1 . Hence, both modes appear together throughout the propagator and the question arises, whether they can be separated. It can be shown that the propagator can also be written in the following form: where the tensor structure Ξ µν is the same for both parts, hence with the coefficients a, . . . , g from Eq. (4.5). The scalar propagator functions are then given by: The first part D (1) (k) contains both polarization modes encoded in K 1 and K 2 , whereas the second part does not involve any mode. The denominator [(k · ξ) 2 − k 2 ]ζ 2 + (k · ζ) 2 that appears in both parts does not have a zero with respect to k 0 , hence it contains no dispersion relation. So it does not seem that the polarization modes can be separated, such that each propagator part contains exactly one of the modes.
Finally, we can state that the structure of the propagator of parity-odd nonbirefringent modified Maxwell theory is rather unusual. In the next section we will compute the polarization vectors.
V. POLARIZATION VECTORS
In what follows, the physical (transverse) degrees of freedom will be labeled with (1) and (2), respectively. For a fixed nonzero "three-vector" ζ and a generic wave vector k, the polarization vector of the λ = 1 mode reads where N ′ is a normalization factor to be given later. The polarization vector of the λ = 2 mode is orthogonal to (5.1) and has a longitudinal component. It is given by The polarization vector ε (1) is a solution of the field equations (3.1), when k 0 is replaced by ω 1 (k) from Eq. (3.2a). The polarization ε (2) is the corresponding solution for k 0 replaced by ω 2 (k) from Eq. (3.2b). The normalization factors N ′ in (5.1) and N ′′ in (5.2) can be computed from the 00-component of the energy-momentum tensor. Note that the above polarization vectors have been calculated in the Lorentz gauge, ∂ µ A µ = 0.
For the Lorentz-violating decay processes considered, both the λ = 1 and the λ = 2 polarization modes contribute.
where ω 1 = ω 1 (k ⊥ , k ) is given by (3.5a). The denominator Q vanishes only for κ 01 = κ 02 = κ 03 = 0 or k ⊥ = 0. If the polarization tensor of the λ = 1 mode is contracted with a gauge invariant expression using the Ward identity, 4 it can be replaced by Π µν | λ=1 : The polarization tensor of the λ = 2 mode is lengthy and is best written up in terms of k and k ⊥ defined in (3.4).
and ω 2 = ω 2 (k ⊥ , k ) is given by (3.5b). Again, if the tensor is contracted with a gaugeinvariant expression, it can be replaced by Π µν | λ=2 : Finally holds where the second contraction only vanishes for E → 0 due to the longitudinal part of ε (2) µ . The polarization vector (5.2) is normalized to unit length by N . This normalization factor cancels in Π µν | λ=2 . Note that the metric tensor η µν does not appear on the right-hand side of (5.9b), whereas it does on the right-hand side of (5.6b).
Furthermore, note that each truncated polarization tensor Π µν | λ=1 and Π µν | λ=2 can be written in a covariant form. This behavior is different from the polarization vectors of standard QED, 5 where only the whole polarization sum is covariant.
It is now evident that not only is the structure of the photon propagator uncommon, but the polarization vectors are unusual as well. In the next section we will analyze how both results are connected.
VI. THE OPTICAL THEOREM AND UNITARITY
In order to investigate unitarity, the simple test of reflection positivity used in [24] for the isotropic case of modified Maxwell theory cannot be adopted, because there are now essentially two different scalar propagators, namely K 1 and K 2 from (4.5). Hence, we could either examine reflection positivity of the full propagator or study the optical theorem for physical processes involving modified photons. As unitarity of the S-matrix results in the optical theorem and the latter is directly related to physical observables, we choose to proceed with the second approach.
The optical theorem will also show how the modified photon propagator in Sec. IV is linked to the photon polarizations from the previous section. The following computations will deal with the physical process that we already considered for isotropic modified Maxwell theory [24] in the context of unitarity: annihilation of a left-handed electron e − L and a righthanded positron e + R to a modified photon γ. The fermions are considered to be massless particles, which renders their helicity a physically well-defined state. Neglecting the axial anomaly, which is of higher order with respect to the electromagnetic coupling constant, the axial vector current j µ 5 = ψγ µ γ 5 ψ is conserved: ∂ µ j µ 5 = 0. This is the simplest tree-level process including a modified photon propagator. It has no threshold and is allowed for both photon modes. We assume a nonzero Lorentz-violating parameter E. Furthermore, the four-momenta of the initial electron and positron are not expected to be collinear.
If the optical theorem holds, the imaginary part of the forward scattering amplitude is related to the cross section for the production of a modified photon from a left-handed electron and a right-handed positron: Herein, dΠ 1 is the corresponding one-particle phase space element. By performing an integration over the four-momentum of the virtual photon, the forward scattering amplitude is given by with the propagator coefficients b, . . . , g from Eq. (4.5). Recall, that the physical poles have to be treated via Feynman's iǫ-prescription. Hence, the denominator K −1 2 from Eq. (4.4b), which appears in the coefficients b, c, d, e, f , and g also has to be replaced by K −1 2 + iǫ. The first contribution to the imaginary part of the matrix element M 1 comes from the physical pole of the scalar propagator function K 1 and corresponds to the dispersion relation (3.5a) of the λ = 1 polarization mode. Using the positive and negative photon frequency of the parity-odd case considered, the scalar part of the propagator is .
The pole with positive real part can be cast in the following form: Because of energy conservation only ω + 1 and not ω − 1 contributes to the imaginary part. We define M 1 ≡ M(e − L e + R → γ) and obtain: and k 0 replaced by the dispersion relation ω 1 from Eq. (6.3). Furthermore, Using the Ward identity, in the first step of Eq. (6.6) we could eliminate all propagator coefficients that are multiplied by at least one photon four-momentum. Then we employed the truncated λ = 1 polarization tensor from Eq. (5.6b).
The second contribution to the imaginary part of the matrix element comes from the λ = 2 mode given by the dispersion relation (3.5b). That mode is contained in K 2 from (4.4b), where Feynman's iǫ-prescription leads to: with The pole with the positive real part results in the following contribution to the imaginary part of the matrix element M 1 : Again, the pole ω − 2 with negative real part does not contribute because of energy conservation. Using the Ward identity leads to: and k 0 is to be replaced by ω + 2 from Eq. (6.10). Moreover, we have used that for with the right-hand side given by (5.9b). Adding the two contributions from Eqs. (6.6) and (6.12) leads to But the right-hand side of the previous equation is just the total cross section of the scattering process. Hence, the optical theorem is valid for the parity-odd sector of modified Maxwell theory. Furthermore, it reveals the connection between the modified photon propagator (cf. Eq. (6.2)) and the polarization tensors (cf. penultimate line of Eqs. (6.6) and (6.12)). The optical theorem thus provides a good cross check for the obtained results of Eqs. (4.4), (5.6b), and (5.9b). Since the process itself only plays a role at the level of providing a valid Ward identity, the obtained result is consistent with having a unitary theory, at least for a tree-level process involving conserved currents. As a final remark we state that the unphysical pole k 2 = 0, which appears in the propagator coefficients b, d, and f , is prevented from being reached by energy conservation. Hence it plays no role in the calculation.
VII. MICROCAUSALITY
In order to decide whether or not the particular case of parity-odd modified Maxwell theory considered satisfies the condition of microcausality, we have to compute the commutator of physical fields at different spacetime points y and z. The latter can be derived from the commutator of vector potentials: where the second step follows from translation invariance. The tensor structure of this expression is to be put into the function θ µν . The causal structure of the commutator is completely determined by the scalar commutator function D(x), which corresponds to the scalar part of the Feynman propagator (see, for instance, [19] and [24]). For this reason we will restrict our considerations solely to D(x) and forget about the tensor structure. Looking at the propagator (4.2) of Sec. IV it is clear that there are two scalar parts, K 1 from Eq. (4.4a) and K 2 from Eq. (4.4b), one for each photon polarization. We begin with K 1 : where positive and negative energies are defined in Eq. (6.3). These are the poles of the scalar propagator K 1 , where ω + 1 delivers the first contribution to the imaginary part of the forward scattering amplitude considered in the previous section.
The evaluation of the contour integral gives Hence, Eq. (7.19a) is of the same form as an integral that appears in the context of the standard propagator (see e.g. Eq. (26a) in [39]). This leads to the final result: Just as for the isotropic case of modified Maxwell theory, whose consistency was discussed in [24], the commutator function (7.20) vanishes everywhere except on the modified null-cone An analogous calculation for the scalar part K 2 from Eq. (4.4b) delivers the following final result for the commutator function D 2 (x): which corresponds to a second modified null-cone: Both null-cones coincide to linear order in E. This is not surprising, since the theory is birefringent to quadratic order in the Lorentz-violating parameters. Each of the Eqs. (7.21) and (7.23) corresponds to a null-cone, whose rotation axis is different for the past and future null-cone. Neither axes coincides with the time-axis, but each is rotated by a small angle, as shown in Fig. 1. Since there are two modes with two different dispersion relations, one may wonder, if this result is sufficient for taking a decision about microcausality. For this reason we tried to separate both modes in Sec. IV with the result (4.8) -(4.10). Therefore, we should investigate G (1) (k) and G (2) (k) from Eq. (4.10): Using will immediately give zero. Hence, the dispersion relation corresponding to the second mode does not seem to play any role here. The λ = 1 mode seems to be preferred compared to the λ = 2 mode, what follows from forcing a parity-odd theory to be nonbirefringent via the Ansatz (2.2). The transversal polarization vectors can be interpreted as two distinct polarization modes: left-and right-handed. In a parity-violating theory they are expected to behave differently, for example with respect to their phase velocity. This would automatically lead to birefringence, which is suppressed by using Eq. (2.2) as a basis. The result of Eq. (7.20) establishes microcausality for the following parameter domain: where ζ is defined in terms of the SME parameters by (2.3b) and (2.4). Hence, the parityodd "nonbirefringent" sector of modified Maxwell theory is unitary and microcausal for the full parameter range.
VIII. COMPARISON TO OTHER LORENTZ-VIOLATING THEORIES
In the previous sections we have seen that both the modified photon propagator and the polarization vectors have an uncommon structure. For this reason, we want to have a general look at the photon propagator and polarization vectors in other Lorentz-violating theories. We start with the photon polarizations of MCS theory. Besides modified Maxwell theory, MCS theory is another possible example of a gauge-invariant and power-counting renormalizable theory that violates Lorentz invariance in the photon sector. MCS theory is characterized by a mass scale m CS and a fixed spacelike 6 "four-vector" ζ µ , that plays the role of a background field. The Chern-Simons mass m CS gives the amount of Lorentz violation. MCS theory exhibits two photon modes, which we call '⊕' and '⊖'. They obey different dispersion relations [30], which results in birefringence. The polarization vectors follow from the field equations and, in temporal gauge A 0 = 0, they are given by with the normalization constants Using the temporal gauge fixing four-vector (n µ ) = (1, 0, 0, 0), the polarization tensor for each of the two modes can be cast in the following form (see [45] for the truncated versions): The polarization sum of standard QED is expected to be recovered for vanishing m CS . For the truncated polarization sum this is, indeed, the case: it is evident that both modes deliver equal contributions to the polarization sum. This even holds for nonvanishing m CS . Hence, the behavior of MCS theory with respect to the polarization modes is completely different compared to parity-odd nonbirefringent modified Maxwell theory. For m CS → 0, there is no residual dependence from the preferred spacetime direction ζ µ in the polarization tensors of the individual modes, which can be seen from Eq. (8.4). Furthermore, for MCS-theory the photon propagator in axial gauge has been shown to be of the following form [19]: where further terms with the index structure composed of the four-momentum, the preferred spacelike four-vector ζ µ , the axial gauge vector and the four-dimensional Levi-Civita symbol have been omitted. The denominator P(k) is a fourth-order polynomial in k 0 , with its zeros corresponding to the two different physical dispersion relations. For a special case of parity-odd 'birefringent' modified Maxwell theory 7 we could show that the propagator in Feynman gauge looks like where P 1 (k) is a second-order polynomial in k 0 , involving the Lorentz-violating parameters and P 2 (k) is of fourth order in k 0 . The two distinct physical dispersion relations of this birefringent theory follow from P 2 (k) = 0. Again, remaining propagator coefficients multiplied by combinations of the four-momentum and preferred four-vectors have been omitted. Hence, we see that our result for the propagator for parity-odd nonbirefringent modified Maxwell theory given by Eqs. (4.2) -(4.5) is rather unusual. For MCS-theory and birefringent modified Maxwell theory (at least for the special case examined), both physical modes emerge as poles of the coefficient before the metric tensor η µν . However, in the case of parity-odd nonbirefringent modified Maxwell theory, the dispersion relation for the λ = 2 polarization mode is not contained in the coefficient K 1 of Eq. (4.4a), which is multiplied with η µν . This peculiarity is also mirrored in the polarization tensors, where we have shown the interplay in the previous section.
IX. LIMIT OF THE POLARIZATION TENSORS FOR VANISHING LORENTZ VIOLATION
Taking the limit E → 0 followed by the limit k ⊥ → 0 (see the definition (3.4)) for the physical polarization vectors (5.1) and (5.2) leads to: (9.1) Taking into account the limit of the four-momentum, the physical polarization vectors reduce to the standard transversal QED results. Note that for both vectors in Eq. (9.1), the order in which the limits are taken does not play any role. As we will see below, this is not the case for the gauge-invariant parts of the polarization tensors from Eqs. (5.6b), (5.9b), that is, if the polarization vectors are coupled to conserved currents. For E → 0 these tensors result in: with ( ζ µ ) = (0, ζ) and |k| = (k 2 + k 2 ⊥ ) 1/2 . For completeness, after inserting the explicit four-vectors, we obtain the following matrices: Note that these matrix representations only hold for the special choice ζ = (0, 0, 1). It is evident that the additional limit k ⊥ → 0 does not exist for each contribution Π µν | λ=1 or Π µν | λ=2 separately, but only for the truncated polarization sum λ=1,2 Π µν | λ , which leads to the standard QED result. For this reason, the polarization vectors are not only deformed -unlike for the isotropic case that was examined in [24] -but their structure completely differs from standard QED. Besides that, no covariant expression exists for each polarization tensor in standard QED, where only the sum λ=1,2 Π µν | λ can be decomposed covariantly.
A. Description of the process
The results obtained for the polarization vectors in Sec. V together with the observations that followed forces us to think about the consistency of the modified theory. The form of the propagator, the polarization vectors and tensors observed in Secs. IV, V, and IX reveal the following uncommon properties: 1) one of the two physical photon modes seems to be preferred with respect to the other, 2) both polarization vectors are interweaved with the spacetime directions ξ µ and ζ µ , even for vanishing Lorentz-violating parameters, 3) each physical polarization tensor can be written in covariant form, 4) and one of the physical polarization vectors has a longitudinal part.
One the one hand, these peculiarities may emerge from the fact that a parity-odd QED is combined with the claim of being nonbirefringent. Two physical photon polarizations can be interpreted as two distinct polarization modes:"left-handed" and "right-handed". These are supposed to behave differently because of parity violation, for example with respect to the phase velocity of each mode. Hence, birefringence would result from this, which clashes with the nonbirefringent Ansatz of Eq. (2.2). On the other hand, the above properties may have emerged from a bad gauge choice and could possibly be removed by picking a more appropriate gauge. For this reason a physical process will be considered, whose cross section does not depend on the gauge. If the mentioned behavior of the polarization modes is not a gauge artifact, it will show up in the results for polarized cross sections. The simplest tree-level process involving external photons, which also occurs in standard QED, is Compton scattering. We consider an electron scattered off a photon in the λ = 1 polarization and in the λ = 2 polarization, respectively. Hence, we want to compute cross sections for the processes e − (p 1 ) γ 1 (k 1 ) → e − (p 2 ) γ 1 (k 2 ), e − (p 1 ) γ 1 (k 1 ) → e − (p 2 ) γ 2 (k 2 ), e − (p 1 ) γ 2 (k 1 ) → e − (p 2 ) γ 1 (k 2 ), and e − (p 1 ) γ 2 (k 1 ) → e − (p 2 ) γ 2 (k 2 ), where γ 1,2 denotes a modified photon in the λ = 1 or λ = 2 polarization state, respectively. The corresponding Feynman diagrams are shown in Fig. 2.
For a review of Compton scattering experiments it is referred to [46]. Furthermore, Ref. [47] gives a new bound on two of the three parameters of parity-odd nonbirefringent modified Maxwell theory from the study of Compton scattering kinematics at the GRAAL experiment 8 on the European Synchrotron Radiation Facility (ESRF) at Grenoble in France.
B. Numerical results for polarized Compton scattering cross sections
We choose special momenta p 1 , k 1 for the initial electron and photon. The outgoing photon momentum configuration is described in spherical coordinates with polar angle ϑ and azimuthal angle ϕ. We consider the initial momentum configuration, for which the electron is at rest: (p µ 1 ) = (m, 0, 0, 0), (k µ 1,λ ) ≡ (ω λ (k), k) with k = (0, 0, |k|) = (0, 0, k 1 ). Values for the modified polarized Compton scattering cross sections σ 11 , σ 12 , σ 21 , and σ 22 are obtained. These correspond to the processes 1 → 1, 1 → 2, 2 → 1, and 2 → 2, where the numbers give the initial and final photon polarization, respectively. To form gauge invariant expressions, the sum over final photon polarizations has to be performed: Our calculation is based on the assumption that only the initial photon state can be prepared, especially its polarization. However, the final photon polarization can only be measured, if the photon is observed or scattered at a second electron. Since we consider the final photon as an asymptotic particle according to the Feynman diagrams in Fig. 2, it is not observed and one has to sum over final photon polarizations [43,44]. Hence, what can be measured in this context are only the quantities σ 1X and σ 2X , so we also give them. Finally, we list the sum of all cross sections, which is averaged over the initial photon [43] or to Eq. (5.81) of [44]. The results are given in units of α 2 with the fine structure constant α ≡ e 2 /4π. The electron mass is set to m = 1. [40]. The cross sections are given in units of α 2 and the electron mass is set to m = 1. polarizations: For comparison with the modified Compton cross sections, the cross sections for unpolarized and polarized Compton scattering in standard QED are presented in Table I and Table II, respectively, for different initial photon momenta k 1 .
An important issue has to be mentioned first: the calculation of the modified cross section in the parity-odd theory can be performed in two different ways. The first possibility is to calculate the matrix element squaredà la section (11.1) of [40] by directly using the modified polarizations vectors from Eqs. (5.1), (5.2). For completeness, we give this equation in a compact form: where Here, λ denotes the initial and λ ′ the final photon polarization. For standard QED Eq. (10.3a) results in Eq. (11-13) of [40] (which we transform to fit our conventions): Alternatively, the computation can be performed with the matrix element squared that is obtained without the direct use of the polarization vectors, but with the polarization tensors from Eqs. (5.4), (5.7). This expression is lengthy and we will not give it in full detail. However, we will state it in a formal manner: where Π µν are photon polarization tensors and |M(k 1 , k 2 )| 2 µν̺σ includes all parts that do not directly involve the photon: traces of combinations of γ-matrices, electron propagators etc. The structure of X λλ ′ is similar to Eq. (5-81) of [44]. However, the latter equation gives the sum over all polarizations, whereas X λλ ′ is the amplitude square for a distinct polarization. For the configurations of Table III and IV the results are shown for different Lorentzviolating parameters κ, which is defined by κ ≡ E/ √ 3 with E from Eq. (3.5c). It suffices to give κ, since for both tables the three Lorentz-violating parameters κ 01 , κ 02 , and κ 03 are chosen to be equal. other. Compare the obtained results to the classical Thomson cross section, which follows from the standard QED result -first obtained by Klein and Nishina -in the limit of vanishing initial photon momentum [44]: Furthermore, the sum of all cross sections then corresponds to the Thomson limit given in Eq. (10.6). The results for σ 11 , σ 12 , and σ 21 do not depend on whether Eq. (10.3a) or Eq. (10.5) is used for the calculation. From Table IV it also follows that for vanishing Lorentz violation σ 1X = σ 2X and, furthermore, that the averaged sum over all cross sections corresponds to the standard Klein-Nishina results. Besides, all results are independent of the fact of whether the calculation is based on X λλ ′ or X λλ ′ . This is also the case for the first selection of parameters in Table V. Furthermore, this table shows that the individual cross sections σ 11 , σ 12 , σ 21 , and σ 22 depend on the direction of ζ, which is encoded in the choice of κ 01 , κ 02 , and κ 03 . However, it is evident that the gauge invariant expressions defined in Eq. (10.1) are independent of the direction of ζ. In the second row we also give the result that follows X λλ ′ . The photon momentum is k 1 = 10 −10 m.
C. Plots of the amplitude squares X λλ ′ and X λλ ′ By plotting the matrix element squares X λλ ′ from Eq. (10.3a) and X λλ ′ from Eq. (10.5) for each process 1 → 1, 1 → 2, 2 → 1, and 2 → 2 leads to a surprise. We first present graphs of both X λλ ′ and X λλ ′ for different sets of Lorentz-violating parameters, where the azimuthal angle ϕ is set to zero. In Fig. 3, for which κ 01 = κ 02 = κ 03 = 1/10 was inserted, we see that X λλ ′ corresponds to X λλ ′ for the processes 1 → 1, 1 → 2, 2 → 1, and 2 → 2. The graphs in Fig. 4 for Lorentz-violating parameters κ 01 = κ 02 = κ 03 = 10 −10 indicate that for the 2 → 2 process the amplitude square X λλ ′ approaches X λλ ′ , but there remains a residue, which appears for X λλ ′ as a narrow peak at an angle ϑ 0 ≈ 2.35 (given in arc measure). Finally, in Fig. 5 we depict X λλ ′ and X λλ ′ for the 1 → 1 and the 2 → 2 modified Compton scattering as a function of both the polar angle ϑ and the azimuthal angle ϕ. It is evident that X λλ ′ and X λλ ′ for 1 → 1 perfectly agree with each other. This is also the case for the processes 1 → 2 and 2 → 1, but we will not display the corresponding plots here. However, the 2 → 2 scattering behaves differently. The matrix element X λλ ′ looks smooth 9 , whereas for κ 01 = κ 02 = κ 03 = 1/10, the amplitude square X λλ ′ is characterized by a set of sharp peaks. For small Lorentz violation some of these peaks seem to remain. Whether or not the limit for vanishing Lorentz violation is influenced by such structures cannot be investigated numerically, but require analytical computations.
D. Interpretation
1. Discrepancies between X λλ ′ and X λλ ′ for 2 → 2 scattering We already know from Eq. (5.2) that the second polarization vector splits into two contributions: a transverse and a longitudinal part. For vanishing Lorentz violation it is explicitly true that (ε (2) µ ) = (ε 0 , ε transv + ε 0 k) = (0, ε transv ) + ε 0 (1, k) = (0, ε If ε (2) µ couples to a gauge-invariant quantity, its longitudinal part will vanish because of the Ward identity, since it is directly proportional to the four-momentum k µ . The weird structures appearing in the matrix element squared X 22 , which were discussed in the last section, originate from the longitudinal part of ε (2) µ . As mentioned, from (10.8) it follows that in the limit of zero Lorentz violation the longitudinal part vanishes by the Ward identity when contracted with physical quantities. However, this only holds if the prefactor ε 0 is not zero. Otherwise, we run into a "0/0" situation, which is mathematically not defined. Now, the physical phase space of the process contains a sector, for which |k | becomes arbitrarily small. This sector is characterized by two angles (ϕ 0 , ϑ 0 ), where for ϕ 0 = 0, ϑ 0 ≈ 2.35. This is depicted in Fig. 6. For this special case the normalization factor N ′′ and, therefore, the prefactor ε 0 can become arbitrarily small. This destroys the applicability of the Ward identity and shows up as peaks in X 22 of Figs. 4 and 5.
Now we would like to analytically investigate the limit E → 0 of the second polarization vector, with its transversal part subtracted. We distinguish between two cases, k ∼ Ek ⊥ and k ≫ Ek ⊥ . The first represents the phase space sector for Compton scattering, for which k becomes arbitrarily small. We begin with the zeroth component of Eq. (5.3): The longitudinal part ε long , which can be extracted from Eq. (5.3) as well, results in: The normalization factor from Eq. (5.8g) is (10.11) Respecting k ≫ Ek ⊥ , we obtain for the second polarization vector: which vanishes for E → 0. In contrast to the latter case, the result for k ∼ Ek ⊥ is as follows: The latter diverges in the limit E → 0. Hence, it becomes evident that for vanishing Lorentz-violating parameter E, when k runs into the phase space sector where it becomes of the order of Ek ⊥ , a peak emerges. Its width is then ∼ Ek ⊥ and its height is ∼ 1/E. This leads us undoubtedly to the following representation of a δ-function as the limit of a function sequence: g n (x) , g n (x) = 2/n for |x| ≤ n , 0 for |x| > n . (10.14) The role of the function sequence index n in Eq. (10.14) is taken by the Lorentz-violating parameter E in the polarization vector. As a result, we finally obtain in the limit E → 0: This analytic result shows, besides the numerically obtained plots in Figs. 4 and 5, that the longitudinal part of the second polarization vector may still play a role for vanishing Lorentz-violating parameter. Because of the δ-function, the Ward identity can perhaps not be applied any more. Now we want to look at the third term of Eq. (10.4a), which is enclosed by round brackets. It will be denoted as X
Limit of X λλ ′ for vanishing Lorentz violation
In section IX we have seen that preferred spacetime directions ξ µ and ζ µ appear in the polarization tensors Π µν even for vanishing Lorentz violation. However, since the limit of X λλ ′ for vanishing Lorentz-violating parameters seem to coincide with the standard QED result, they obviously do not play a role for physical quantities. The question then arises as to why this is the case.
We consider an amplitude M, to which one external photon with four-momentum k µ and polarization λ couples: In what follows, the term "matrix element squared" is understood in the sense of individual contributions |ε For a virtual state, 11 all polarization vectors, hence also the scalar and the longitudinal ones, contribute to the polarization-summed matrix element squared -denoted as |M| 2 : Evaluating |M| 2 for a real state means that the Ward identity is used. For standard QED, if (k µ ) = (k, 0, 0, k) is chosen, the Ward identity will result in Because of this, the unphysical degrees of freedom cancel each other and what remains are terms which involve the physical polarization vectors (λ = 1, 2). Since the latter can be chosen as (ε µ 1 ) = (0, 1, 0, 0) and (ε µ 2 ) = (0, 0, 1, 0), we obtain where 'phys' means that the Ward identity has been used. In order to understand the limits of the polarization tensors from Eq. (9.3) we will perform a similar analysis in the context of the modified theory. For (k µ ) = (|k|, k) with k = (k ⊥ , 0, k ) the Ward identity reads and therefore, M 1 can be expressed as follows: Using the result of Eq. (10.21b), the contribution of the matrix element squared involving the first polarization mode results in: where the Ward identity has been used in the second step. Hence, restricting the "matrix element squared" to the physical subspace with the Ward identity guarantees that the additional parts, that depend on the preferred directions ξ µ and ζ µ , cancel. Now consider the λ = 2 polarization mode. With Eq. (10.21b) we obtain: Hence, we see that by using the Ward identity all contributions depending on ξ µ and ζ µ also vanish for the second mode. Therefore, for vanishing Lorentz violation the standard result is recovered.
XI. DISCUSSION AND CONCLUSION
In this article, a special sector of a CPT-even Lorentz-violating modification of QED, with the characteristics of being parity-odd and nonbirefringent, was examined with respect to consistency. The deformation of QED is described by one fixed time-like "four-vector", one fixed space-like "four-vector", and three Lorentz-violating parameters.
The nonbirefringent Ansatz combined with the parity-violating parameter choice leads to two distinct physical photon polarization modes. These modes are characterized by dispersion relations, that differ to quadratic order in the Lorentz-violating parameters. Hence, the theory is only nonbirefringent to linear order. The dispersion relations coincide with the formulas previously obtained in [22]. The new most important results of this article are summarized in the subsequent items: • With the optical theorem, unitarity is verified for tree-level processes involving conserved currents.
• Microcausality is established for the full range of Lorentz-violating parameters. Information only propagates along the modified nullcones.
• It has turned out that covariant polarization tensors can be constructed for each photon mode. This is not possible in standard QED, where only the polarization tensor of the sum of both modes can be written covariantly.
• The gauge-invariant 12 polarization tensor of each mode depends on the background field directions. For vanishing Lorentz violation this dependence remains. It only cancels when considering the sum of both modes, which leads to the polarization sum of standard QED.
• The fact that the polarization tensors depend on the background field directions even for vanishing Lorentz violation, makes us think about the question of whether the limit of zero Lorentz violation is continuous. In other words, a priori it is not clear, whether or not the modified theory approaches standard QED for vanishing Lorentz violation. This is the motivation to test the theory via brute force by calculating one special process: Compton scattering for unpolarized electrons scattered by polarized photons.
• The cross sections can be computed either by using the modified polarization vectors or the modified polarization tensors. The upshot is that the results for 1 → 1, 1 → 2, and 2 → 1 coincide, but a numerical treatment reveals a discrepancy for 2 → 2 scattering. 13 The Ward identity is shown to cure the polarization vectors and tensors from their bad behavior for vanishing Lorentz violation, at least for the first three processes. However, if the matrix element squared is computed for the fourth process by using the modified polarization vectors, there exists a phase space sector, for which the longitudinal part of the second polarization vector is proportional to a δ-function. This could be shown by an analytic investigation. It could also be proven analytically that the Ward identity can cancel this contribution, nevertheless.
To conclude, the parity-odd "nonbirefringent" sector of modified Maxwell theory seems -with regard to the performed investigations -to be consistent. Further steps in the context of consistency of Lorentz-violating quantum field theories may involve the analysis of unitarity at one-loop level, where the Lorentz-violating structure is treated in an exact way. Especially for this parity-odd theory it would be interesting to know if its consistency is inherited to higher orders of perturbation theory. However, this is beyond the scope of that article.
In light of the consistency of this Lorentz-violating extension at tree-level, nature decides on the values of the Lorentz-violating parameters. Therefore, they have to be measured with experiments. For a summary of the current experimental status we refer to [52] and references therein. The latter article also gives new experimental bounds on the parity-odd parameters.
FIG. 7. Chosen coordinate system for the phase space integration, where the initial photon threemomentum k 1 lies along the third axis of the coordinate system. For the outgoing photon momentum k 2 spherical coordinates are chosen with the azimuthal angle ϑ corresponding to the angle between k 2 and the third axis. Cases are treated with the three-vector ζ having equal or unequal components.
(|k 2 |, ϑ, ϕ), where the initial photon momentum points along the third axis of the coordinate system. The general case is mimicked by ζ pointing in an arbitrary direction. As a special -but nevertheless very generic -case we can choose its components to be equal (however, computations were also done for different cases as shown in Table V): The integration over |k 2 | is eliminated at once with the energy conservation equation in the δ-function. Here, we have to keep in mind that where (k 2 ) 0 is the corresponding zero. The analytic solution (k 2 ) 0 is a complicated function of E and m, so we determine it numerically with Newton's method inside the C++ program. The integrations over ϑ and ϕ are performed with the Simpson rule, which is sufficient for our purpose. The integration domain, that includes all physical states, is determined automatically with (k 2 ) 0 . If no zero (k 2 ) 0 exists, then the corresponding angles ϑ and ϕ lie outside the domain.
Appendix B: Compton scattering and Thomson limit in standard (quantum) electrodynamics The low-energy limit of the Compton scattering cross section (Thomson limit) can be calculated classically via the following equation (see e.g. [51]): where ε i,λ is the polarization three-vector of the incoming and ε f,λ ′ that of the outgoing electromagnetic wave. For the initial wave traveling along the z-axis we can choose the transverse polarization vectors as In general, the propagation direction of the final wave can be described in spherical coordinates by the basis vector e r = (sin ϑ cos ϕ, sin ϑ sin ϕ, cos ϑ). Then we can pick the physical polarization vectors to point along the other two basis vectors e ϑ and e ϕ : If we rotate, for example, the set of initial polarization vectors by angle α and the final ones by angle β in their corresponding polarization planes, the single contributions σ Th λλ ′ will depend on β. However, this dependence cancels in σ Th 1X and σ Th 2X that are defined as follows: σ This is also the case for MCS-theory. The MCS polarization tensors of Eq. (8.4) even give equal results for each individual polarized scattering process in the limit m CS → 0: For parity-odd modified "nonbirefringent" modified Maxwell theory the individual contributions are not equal for E → 0. However, the above expressions from Eqs. (2.5a) and (2.5b) correspond to each other. With Eq. (11-13) of [40] and the standard polarization vectors from Eq. (2.3) we obtain the polarized Compton scattering values given in Table. II. | 2012-11-11T21:42:55.000Z | 2011-11-17T00:00:00.000 | {
"year": 2011,
"sha1": "e24fb426c88e499f3423ca33b843225925e6e2dc",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1111.4182",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "e24fb426c88e499f3423ca33b843225925e6e2dc",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.