Add files using upload-large-folder tool

preprint/preprint__480fe05899d1520cb4b6e72a4a2f2de5769529009fcec55abe3ac7fda0ce6eaa/preprint__480fe05899d1520cb4b6e72a4a2f2de5769529009fcec55abe3ac7fda0ce6eaa.mmd

@@ -0,0 +1,331 @@
+
+# Unveiling Mechanisms and Onset Threshold of Humping in High-Speed Laser Welding  
+
+Jingjing Li  ju1572@psu.edu  
+
+The Pennsylvania State University https://orcid.org/0000- 0001- 9337- 6178  
+
+Zhen- Hao Lai  
+
+The Pennsylvania State University https://orcid.org/0000- 0003- 3871- 6656  
+
+Siguang Xu General Motors  
+
+Samuel Clark Advanced Photon Source https://orcid.org/0000- 0002- 8678- 3020  
+
+Kamel Fezzaa  
+
+Argonne National Laboratory  
+
+## Article  
+
+Keywords: Laser welding, humping, synchrotron X- ray imaging, CFD simulation, scaling analyses  
+
+Posted Date: March 27th, 2024  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 4087583/v1  
+
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on November 5th, 2024. See the published version at https://doi.org/10.1038/s41467- 024- 53888- w.
+
+<--- Page Split --->
+
+## Unveiling Mechanisms and Onset Threshold of Humping in High-Speed Laser Welding  
+
+Zen- Hao Lai \(^{1}\) , Siguang \(\mathrm{Xu}^{3}\) , Samuel J. Clark \(^{4}\) , Kamel Fezzaa \(^{4}\) , Jingjing Li \(^{1,2}\) \* \(^{1}\) Department of Materials Science and Engineering, The Pennsylvania State University, University Park, PA 16802, USA \(^{2}\) Department of Industrial and Manufacturing Engineering, The Pennsylvania State University, University Park, PA 16802, USA \(^{3}\) General Motors LLC, Warren, MI 48092, USA \(^{4}\) X- ray Science Division, Argonne National Laboratory, Lemont, IL, 60439, USA \(^{*}\) jul572@psu.edu (Corresponding author) Keywords: Laser welding, humping, synchrotron X- ray imaging, CFD simulation, scaling analyses  
+
+## Abstract  
+
+The fabrication of fuel cells relies on a rapid laser welding process. However, challenges arise with the occurrence of humping when the welding speed surpasses a critical threshold, which poses difficulties in achieving a smooth surface finish and a consistent weld strength. This study aims to elucidate the mechanisms behind humping by analyzing the morphology of molten pool and the characteristics of melt flow at varying welding speeds via in situ synchrotron high- speed X- ray imaging and computational fluid dynamics simulations. Our findings indicate that the short keyhole rear wall, the high backward melt velocity, and the prolonged tail of molten pool are the primary factors contributing to the onset of humping. Furthermore, a dimensionless humping index \((\pi_{h})\) was introduced, which successfully captured the onset threshold of humping across different literatures. This index not only provides a quantitative description of the humping formation tendency but also serves as a valuable tool for optimizing the laser welding process.  
+
+## 1 Introduction  
+
+The laser welding process offers several key advantages over conventional welding methods, including a small heat- affected zone, high welding speed, and excellent flexibility in welding path
+
+<--- Page Split --->
+
+design 1. These advantages make it well- suited for the fabrication of fuel cells, which requires a long and narrow welding path between two bipolar plates 2,3. While increasing the laser welding speed enhances productivity, the formation of humps is particularly a major issue that limits the maximum welding speed 4- 6. Humps manifest as severe periodic undulations along the top surface of the weld seam. They not only pose difficulties in achieving a smooth surface finish but also significantly deteriorate the weld strength by reducing the effective bonded region. In laser welding, the occurrence of humping is influenced by laser parameters such as laser welding speed, power, and spot diameter 7. Interestingly, the issue of humping is not unique to laser welding. It is also encountered in other arc- based and laser- based processes, such as arc welding 8- 11, wire arc additive manufacturing 12- 14, and powder bed fusion 15,16, as the moving speed of heat source increases. It indicates that the mechanisms governing humping are associated with the heat and mass transfer in the molten pool (MP), regardless of its origin.  
+
+The current understanding of humping can be summarized as follows. First, it always occurs at the end of MP once a critical welding speed is exceeded. Second, the formation of humps is primarily linked to the high backward melt velocity created at a high welding speed, so lowering the heat input proves effective in eliminating humping 14. The shallower inclination angle of the MP boundary also reduces the deceleration rate of the backward melt velocity 4,6. Third, a high welding speed results in a long and narrow MP, which is more susceptible to the Rayleigh instability 17, where a cylinder liquid phase tends to fragment into multiple droplets as the length- to- width ratio increases. However, the above understandings primarily relied on simulations and assumptions, lacking real- time experimental validation. In addition, the existing understanding is predominantly qualitative rather than quantitative, posing challenges in predicting humping under diverse process conditions.
+
+<--- Page Split --->
+
+The analyses of keyhole and fluid dynamics during the laser welding process requires in situ characterization techniques or numerical simulations. Previous studies implemented a high- speed optical camera positioned from the top to capture optical images \(^{4,18}\) or vapor plume \(^{19}\) for studying keyhole dynamics. However, this method is limited in providing information solely on the tilting angle of the keyhole front wall and cannot capture the shape of keyhole rear wall or trailing MP. Another approach involved using a transparent glass on one side of the weld to observe the keyhole from a side view \(^{20}\) , which, however, does not fully replicate real welding conditions. In this research, in situ high- speed synchrotron X- ray imaging was adopted to resolve the limitations mentioned above. This technique has been successfully employed to analyze the keyhole instability \(^{21}\) , morphology \(^{22}\) , and the pore formation mechanisms \(^{23,24}\) in additive manufacturing.  
+
+In this study, in situ synchrotron X- ray imaging was employed to investigate the humping phenomenon in high- speed laser welding. The geometries of the keyhole and MP offered crucial understanding into the mechanisms of humping formation. Then, computational fluid dynamics simulations were conducted to analyze the characteristics of melt flow within MP, where the streamline and volumetric flow rate provide further insights into the effects of the MP tail on humping. Lastly, a dimensionless humping index for the laser welding process was developed. This index not only identified the onset threshold of humping across different references but also described the humping tendency. It can serve as a valuable tool to predict humping in laser welding and to optimize the laser welding process.
+
+<--- Page Split --->
+
+## 2 Results  
+
+2.1 In situ characterization of humping during high-speed laser welding  
+
+2.1 In situ characterization of humping during high- speed laser weldingThe in situ high- speed synchrotron X- ray observation of laser welding was carried out at beamline 32- ID- B at Argonne National Laboratory (ANL), using an ytterbium single- mode continuous wave laser (YLR- 500- AC) with a spot diameter of 43 \(\mu \mathrm{m}\) . The tested material was an \(85\mu \mathrm{m}\) - thick 439 stainless steel. Two sheets were lap- welded from the top while in situ X- ray images were acquired from the side. More details on the experimental setup can be found in the “Methods” section and Supplementary Fig. 1a. The corresponding power for each condition was determined beforehand at Edison Welding Institute (EWI) by iteratively increasing the laser power until full penetration was achieved, as shown in Fig. 1a. Full penetration is crucial in the fabrication of fuel cell to prevent leakage between channels. It is worth noting that due to the equipment setup at ANL, the fastest laser welding speed adopted was \(1.42\mathrm{m / s}\) , slightly lower than the fastest speed used at EWI (1.50 m/s). The appearances of the weld seams are presented in Fig. 1b, showing that the critical welding speed for the onset of humping is \(1.00\mathrm{m / s}\) . While the number of humps slightly decreased as the welding speed increased beyond this threshold, the surface topography of the weld seams (Fig. 1c) revealed that the average height of humps increased from \(56.08\pm 7.45\mu \mathrm{m}\) at \(1.00\mathrm{m / s}\) to \(81.31\pm 14.22\mu \mathrm{m}\) at \(1.42\mathrm{m / s}\) . This observation suggests that the momentum of the backward melt flow continued to intensify with the welding speed, exacerbating the humping phenomenon.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig. 1 a The corresponding laser power at each welding speed was determined beforehand at EWI by iterative increase of laser power until full penetration was achieved. b Top surfaces of weld seams at the welding speeds of 0.33, 1.00, and 1.42 m/s. Note that the fastest speed was adopted as 1.42 m/s instead of 1.50 m/s due to the equipment setup at ANL. All scale bars correspond to 500 μm. c Surface topography of weld seams shown in b at the welding speeds of 1.00, and 1.42 m/s. </center>  
+
+The results of in situ high- speed synchrotron X- ray observations are presented in Fig. 2a–c, where the keyhole and humping phenomena during laser welding were successfully captured. For a comprehensive view of all in situ observations for five different welding speeds, refer to Supplementary Movie 1. A keyhole is a less dense space filled with dilute metal vapor compared to liquid MP, therefore providing sufficient contrast for X- ray observation. However, because of the small difference in the attenuation rates between the liquid and solid phases of stainless steel, the MP dimensions can only be measured from the fluctuating wavy patterns on the top surface. Several key characteristics of the keyhole and MP geometries were quantified from the in situ X- ray observations.  
+
+First, the depth of keyhole rear wall indicates the extent of the geometrical barrier to the backward melt flow, where a shorter keyhole rear wall suggests a reduced barrier to the backward melt flow. As the laser welding speed increased from 0.33 to 1.42 m/s (Fig. 2d), the depth of
+
+<--- Page Split --->
+
+keyhole rear wall gradually declined from \(12.29 \pm 2.46 \mu \mathrm{m}\) to \(- 100.82 \pm 17.72 \mu \mathrm{m}\) , which was significantly beneath the top surface.  
+
+Second, a longer MP length reduces the effectiveness of the MP tail to decelerate the melt flow. The MP length was determined as the average of the minimum instantaneous MP length observed in each humping cycle under conditions where humping occurred, which will be discussed in the following section and Fig. 3a. It increased from approximately \(350 \mu \mathrm{m}\) at laser welding speed of \(0.33 \mathrm{m / s}\) to the highest value of \(800 \mu \mathrm{m}\) when the laser welding speed reaching \(1.25 \mathrm{m / s}\) , as shown in Fig. 2e. A prolonged MP not only suggests an increased propensity for humping based on the Rayleigh instability criterion \(^{17}\) but also indicates that it is ineffective in mitigating issues of backward volumetric flow. This phenomenon will be discussed in the section of computational fluid dynamics simulations.  
+
+Third, a greater melt velocity increases the volumetric flow rate of the backward melt flow. The maximum melt velocity occurs at the MP surrounding the keyhole and is predominantly backward as the melt flow is directed from the front to the rear of the keyhole \(^{25}\) . Literatures reported that the measurement of melt flow velocity can be achieved through in situ X- ray observation with additions of W or Ta particles as the tracer materials \(^{26 - 28}\) . However, this approach is limited by the particle size in relation to the MP dimensions. It has been reported in arc welding \(^{26}\) for its larger size of MP compared to laser welding and powder bed fusion \(^{27,28}\) where the tracer particles were premixed with the powder bed. In this study, an analytical approach was adopted, where the maximum melt velocity \((u_{max})\) is calculated by considering the equation of continuity proposed by Beck et al. \(^{25}\) as follow:
+
+<--- Page Split --->
+
+\[u_{m a x} = u_{w}\left\{1 + \left[\frac{c_{p}\rho u_{w}r}{k}\left(1 + \frac{2(c_{p}(T_{m} - T_{b}) + L_{m})}{c_{p}(T_{b} - T_{m})}\right)\right]^{0.5}\right\} \quad (1)\]  
+
+where \(u_{w}\) is the laser welding speed, \(c_{p}\) is the specific heat, \(\rho\) is the density, \(r\) is the spot radius, \(k\) is the thermal conductivity, and \(L_{m}\) is the latent heat of fusion. \(T_{b}\) , \(T_{m}\) , \(T_{0}\) are the boiling, melting, and room temperatures, respectively. Because \(u_{m a x}\) occurs at the MP right around the keyhole, the values of \(c_{p}\) , \(\rho\) , and \(k\) are obtained at the boiling temperature from the thermophysical databases \(^{29,30}\) . It is noted that among laser welding parameters, the maximum melt velocity is related to the spot diameter and welding speed, however, not affected by the laser power based on this equation. The calculated result is shown in Fig. 2f, where the maximum melt velocity significantly increases with the welding speed. Besides, the width of the keyhole, measured at the depth between two welded materials, was found to be correlated with the maximum melt velocity (Fig. 2f). This correlation is consistent with the assumption of this model, where the maximum melt velocity occurs in the MP surrounding the keyhole and is predominantly horizontal. Therefore, a greater melt velocity leads to a more elongated keyhole.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig. 2 a–c In situ high-speed synchrotron X-ray observations of laser welding at the welding speeds of 0.33, 1.00, and 1.42 m/s at the arbitrary times of \(t_0\) , \(t_0 + 0.3\) , and \(t_0 + 0.6\) ms. All scale bars are \(200 \mu \mathrm{m}\) . For a comprehensive view of all in situ observations for five different welding speeds, refer to Supplementary Movie 1. d Depth of keyhole rear wall. e MP length. f Keyhole width (measured) and maximum melt velocity (calculated by Equation 1). </center>  
+
+### 2.2 Periodic formation of humping  
+
+The occurrence of humps is always periodic. Berger et al. first introduced the concept of conservation of volume flow to explain this phenomenon \(^{5}\) , where the melt is incompressible, and the excessive melt is deflected upward at the end of MP to form hump. Xue et al. conducted simulation \(^{6}\) , proving that the melt accumulation and the Rayleigh instability collectively triggered the onset of humping. However, the above analyses were based on high- speed optical camera from the top view \(^{5}\) or simulation \(^{6}\) . Therefore, the in situ X- ray observation in this study provided clearer real- time observation of these phenomena.
+
+<--- Page Split --->
+
+The humping cycle was analyzed at the welding speed of 1.42 m/s with the instantaneous MP length on the top surface and the waviness of the front part of MP, as illustrated in Fig. 3a. The waviness is an indication of the gradient of volumetric flow rate, as a steeper gradient leads to a more excessive volume of backward melt and therefore a higher waviness. More details of their measurements can be found in the “Methods” section. After the instantaneous MP length reached a minimum, the formation of a new hump was initiated because the size of the MP tail became insufficient to accommodate all backward melt flow at this moment. The humping served as an approach to release the excessive melt, with the instantaneous MP length slightly extending on the top surface. This period was defined as the beginning stage of humping, marked in grey in Fig. 3a. In this period, the waviness of the front part of MP was generally small because the release of excessive melt decreases the gradient of volumetric melt flow. After the beginning stage of humping, the growth of hump gradually slowed down. Although the instantaneous MP length continued to increase, the release of excessive melt became less effective, therefore leading to a generally higher waviness of MP in these periods. Finally, when the instantaneous MP length reached another minimum, another humping cycle was triggered. The entire process repeated, leading to the periodical formation of humps as observed. The current observation offers direct evidence supporting the humping theory based on the conservation of volume.
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+  
+
+Fig. 3 Analysis of humping cycle at the welding speed of \(1.42 \mathrm{m / s}\) . a MP length on the top surface and waviness of the front part of MP. The periods of the beginning stage of humping are marked in grey. b Examples showing the change in MP waviness during humping. During the beginning stage of humping \((t_0 + 0.38 \mathrm{ms})\) , the MP waviness was averagely lower. When the humping was close to an end \((t_0 + 0.54 \mathrm{ms})\) , the MP surface started to show a higher waviness because of the steeper gradient of volumetric flow rate of backward melt flow. All scale bars \(= 200 \mu \mathrm{m}\) .  
+
+### 2.3 Computational fluid dynamics (CFD) simulation  
+
+In situ synchrotron X- ray imaging provided crucial insights into the morphologies of keyhole and MP and their effects on humping formation in the laser welding process. Then, CFD simulation was conducted to quantitatively analyze the melt flow within MP. The setup of CFD simulation can be found in the “Methods” section and Supplementary Fig. 2. The simulation was validated with the MP shape from the side view (Supplementary Fig. 3), and the linear number density of humps, defined as the number of humps per unit length (Supplementary Fig. 4). Fig. 4a–b presents the simulation results for a condition without (0.33 m/s, 102 W) and with humping (1.42 m/s, 348 W), respectively. It successfully captured the shorter keyhole rear wall, shallower inclination angle of MP boundary, and the more elongated keyhole along the welding direction at the higher welding speed, aligning well with the experimental findings from in situ synchrotron X- ray imaging.
+
+<--- Page Split --->
+
+Streamlines were then extracted from the simulation models using the Runge- Kutta integration method \(^{31}\) to visualize melt flow. At the welding speed of \(0.33 \mathrm{m / s}\) (Fig. 4a), the backward melt flow gradually diminished to zero at roughly the midpoint of MP and re- entered it. This finding suggested that the size of MP was sufficient to accommodate all backward melt, preventing the occurrence of humping. In contrast, at the welding speed of \(1.42 \mathrm{m / s}\) (Fig. 4b), the direction of melt flow was always backward and could not return to the MP, so it eventually traversed the entire MP length and be deflected upward to form hump. The net cross- sectional volumetric flow rate in the MP was further extracted (Fig. 4c) to analyze the accumulation of melt toward the MP tail. At the welding speed of \(0.33 \mathrm{m / s}\) , the net volumetric flow rate for the last 200 \(\mu \mathrm{m}\) of MP was close to net zero, indicating an insignificant melt accumulation. Conversely, at the welding speed of \(1.42 \mathrm{m / s}\) , throughout the entire MP length (1050 \(\mu \mathrm{m}\) ), the front \(1000 \mu \mathrm{m}\) exhibited a significant net backward volumetric flow, resulting in severe melt accumulation toward the MP tail. Moreover, the gradient of volumetric flow rate gradually decreases from the front to the end of MP, indicating that the MP tail cannot effectively decelerate the backward melt flow. This phenomenon can be attributed to the shallow inclination angle of the tail of MP boundary, as shown in Fig. 4b. Therefore, a long MP indicated a prolonged MP tail, increasing the humping tendency. It also suggests a strong humping tendency due to the Rayleigh instability, which has been studied by the simulation from Ai et al. \(^{32}\) in high- speed arc welding.  
+
+Based on the analyses of in situ synchrotron X- ray imaging and CFD simulation, the formation mechanisms of humping can be concluded as illustrated in Fig. 4d. First, the depth of keyhole rear wall increases with laser welding speed, indicating a reduced barrier for the backward melt flow. Second, the MP length also increases with the welding speed. A longer MP tail not only fails to effectively decelerate the melt flow but also is more susceptible to Rayleigh's instability.
+
+<--- Page Split --->
+
+Third, the maximum backward melt velocity increases significantly with the welding speed. All the aforementioned factors indicate an enhancement in backward melt velocity and a decrease in the barrier to melt flow as the laser welding speed increases. Therefore, the accumulation of melt towards the end of MP intensifies, ultimately triggering the onset of humping.  
+
+![](images/Figure_5.jpg)
+
+<center>Fig. 4 a–b Top and side views of streamlines extracted from the CFD simulation models at the welding speeds of 0.33 and 1.42 m/s. All scale bars correspond to 200 \(\mu \mathrm{m}\) . c Net cross-sectional volumetric flow rate extracted from different distances of y-z cross sections behind keyhole front wall. d Schematic diagram showing the formation mechanisms of humping. </center>  
+
+### 2.4 Development of index for humping tendency  
+
+The prediction of humping in laser welding usually relies on simulation. An alternative method involves developing a dimensionless index that calculates the humping tendency solely based on process parameters and materials properties. This approach is simpler and more efficient
+
+<--- Page Split --->
+
+since it eliminates the need for complex fluid dynamics calculations. While Meng et al. \(^{33}\) have introduced the dimensionless humping index for the arc welding process, its application to laser welding is limited due to inherent disparities between the two processes. In addition, laser welding lacks certain MP characteristics present in arc welding, such as temperature increase and gouging length.  
+
+The Buckingham \(\pi\) theorem was first employed to determine the number of dimensionless groups, where a system with \(m\) variables and \(n\) fundamental units can be described by \((m - n)\) dimensionless groups. In this study, there are 6 variables, i.e., maximum melt velocity \((u_{max},m\cdot\) \(s^{- 1}\) ), MP length \((l,m)\) , density \((\rho ,k g\cdot m^{- 3})\) , specific heat \((c_{p},m^{2}\cdot s^{- 2}\cdot K^{- 1})\) , thermal conductivity \((k,k g\cdot m\cdot s^{- 3}\cdot K^{- 1})\) , and surface tension coefficient \((\gamma ,k g\cdot s^{- 2})\) . It is noted that the depth of keyhole rear wall was not included as a variable, because it is correlated with the MP length (Fig. 2d- e). The above 6 variables involve 4 fundamental units, which are mass \((k g)\) , length \((m)\) , time \((s)\) , and temperature \((K)\) . Therefore, the theorem suggests two dimensionless groups representing the system as follows.  
+
+\[\begin{array}{l}\pi_{1} = \frac{u_{max}l\rho c_{p}}{k} \\ \pi_{2} = \frac{u_{max}^{2}l\rho}{\gamma} \end{array} \quad (2)\]  
+
+\(\pi_{1}\) is the Peclet number (Pe) \(^{34}\) , which indicates the relative importance of convection and conduction in heat transfer within the melt pool. \(\pi_{2}\) is the Weber number, representing the free surface deformation tendency \(^{33}\) . Note that all materials properties are extracted at the liquidus temperature from the thermophysical databases \(^{29,30}\) , given that the temperature at the end of MP is close to the liquidus temperature.
+
+<--- Page Split --->
+
+\(u_{max}\) and \(l\) can be calculated with only the process parameters ( \(r\) : spot radius, \(u_{w}\) : laser welding speed, \(P\) : power) and material properties \((\rho , c_{p}, k,\) and \(\gamma\) ). \(u_{max}\) is calculated by Equation 1, with the calculation results shown in Fig. 2f. For the MP dimensions, the scaling law is adopted. Studies have reported that the MP depth can be scaled with \(P / u_{w}r^{18,35}\) for a same material. The scaling of MP length has not been reported due to its requirement of in situ observation. Therefore, it was performed using the experimental results from the current study and other literatures \(^{5,36}\) , as shown in Fig. 5a. This relationship can be expressed by the following equation.  
+
+\[l\propto P / u_{w}^{2}\frac{3}{r^{4}} \quad (4)\]  
+
+It is noted that the MP length was determined as the average of the minimum instantaneous MP length observed in each humping cycle (Fig. 3a) under conditions where humping occurred. In addition, the scaling law for the MP length is slightly different from the case for the MP depth, where it is scaled with \(P / u_{w}r\) .  
+
+Multiple possibilities of the combination of \(\pi_{1}\) and \(\pi_{2}\) can be formed, but it was found that a simple product of them can be the dimensionless humping index \((\pi_{h})\) .  
+
+\[\pi_{h} = \pi_{1}\times \pi_{2} = \frac{u_{max}^{3}l^{2}\rho^{2}c_{p}}{k\gamma} \quad (5)\]  
+
+The calculated results using the data from the current study and other references \(^{5,6,18}\) are shown in Fig. 5b. This dimensionless humping index not only captures the onset threshold of humping, which occurs at approximately 10,000 of \(\pi_{h}\) , but also describes the humping tendency. It is also in a good agreement with the study from Kawahito et al., where among the process parameters \((P, u_{w},\) and \(r\) ), the humping tendency increases with a higher \(P\) , a greater \(u_{w}\) , and a finer \(r\) when the other two parameters are fixed \(^{7}\) . It also aligns well with the conclusions from the in situ
+
+<--- Page Split --->
+
+synchrotron X- ray imaging and the CFD simulation, where a longer \(l\) or a greater \(u_{max}\) increases the humping tendency.  
+
+Furthermore, the model allows for the prediction of the critical laser welding speed and the corresponding power where humping begins to occur. Fig. 5c shows the isolines of several penetration depths and the 10,000- isoline of \(\pi_{h}\) calculated under a spot radius of \(50 \mu \mathrm{m}\) . The intersection points between each penetration depth and the 10,000- isoline of \(\pi_{h}\) are the respective critical laser welding speeds and power. In addition, by reducing the spot radius to \(25 \mu \mathrm{m}\) as depicted in Fig. 5d, the critical laser welding speed for each penetration depth can be increased for the following reasons. First, the maximum melt velocity decreases with a finer spot radius according to Equation 1. Second, a smaller power is required for a same penetration depth with a finer spot radius, resulting in a shorter MP length based on the scaling law in Fig. 5a and Equation 4. Both factors reduce the \(\pi_{h}\) according to Equation 5, therefore yielding a higher critical laser welding speed.
+
+<--- Page Split --->
+![PLACEHOLDER_16_0]
+
+<center>Fig. 5 a Scaling law between the MP length and \(P / u_{w}^{2}r^{4}\) \((R^{2} = 0.9856)\) performed using the experimental results from the current study and other literatures \(^{5,36}\) . b Calculated dimensionless humping index \(\pi_{h}\) . The onset threshold occurs at approximately 10,000 across different published data \(^{5,6,18}\) . c–d Process windows with (red) and without humping (green) defined by the \(10^{4}\) -isoline of \(\pi_{h}\) and penetration depth \((\mu m)\) calculated by scaling law \(^{18,35}\) at the spot radii of 50 and \(25\mu \mathrm{m}\) . </center>  
+
+Humping is an issue that occurs not only in laser welding but also in arc welding and additive manufacturing under a high moving speed of heat source. This study identified the formation mechanisms of humping by analyzing the morphologies of keyhole and MP as well as the melt flow characteristics across various welding speeds via in situ high- speed synchrotron X- ray imaging and CFD simulation. The mechanisms are concluded as follows. First, a high welding speed leads to a shorter keyhole rear wall, therefore reducing the geometrical barrier to the
+
+<--- Page Split --->
+
+backward melt flow. Second, a prolonged MP tail created at a high welding speed is not only ineffective in decelerating the backward melt flow owing to the shallower inclination angle but also susceptible to Rayleigh's instability. Third, the maximum backward melt velocity increases substantially with the welding speed, leading to a significant increase in the backward volumetric flow rate of melt. All above factors collectively enhance the accumulation of melt towards the end of MP, ultimately causing the occurrence of humping. In addition, the humping cycle were analyzed with several instantaneous MP characteristics. The observation aligned well with the humping theory based on the conservation of volume.  
+
+Lastly, a dimensionless humping index \((\pi_h)\) was developed for the laser welding process using Buckingham's \(\pi\) theorem, which only requires the process parameters and materials properties. This index offers a quantitative depiction of humping formation tendency and serves as an essential tool for optimizing the laser welding process, as it enables the prediction of critical welding speed and power. It also concludes that a finer spot size can increase the critical welding speed with a reduced power when a same penetration depth is required.  
+
+## 3 Methods  
+
+3.1 In situ high-speed synchrotron X-ray imaging  
+
+In situ high- speed synchrotron X- ray imaging of laser welding was conducted at beamline 32- ID- B at ANL using an ytterbium single- mode continuous wave laser source (YLR- 500- AC). A pseudo pink X- ray beam with the 1st harmonic energy centered at \(24.7\mathrm{keV}\) was generated using a short- period (18 mm) undulator. This X- ray beam was directed through the sample from the side during the laser lap welding of two stainless steel sheets, as shown in Supplementary Fig. 1a. The propagated X- ray signal was recorded with a high- speed camera (Photron FastCam SA- Z) at a frame rate of 20,000 frames per second. The tested material was an \(85\mu \mathrm{m}\) - thick 439 stainless steel
+
+<--- Page Split --->
+
+(Cleveland Steel, USA). The specimens were cut by electrical discharge machining into the dimension in Supplementary Fig. 1b, with a gauge width of only \(500 \mu \mathrm{m}\) due to the high extent of attenuation on X- ray for steels. The specimens were clamped in an in- house designed fixture to avoid movement during laser welding.  
+
+### 3.2 Image processing and quantification  
+
+The in situ X- ray images were processed by using ImageJ. Two different approaches of image processing were employed for the following purposes. In the first approach, each image was subtracted by the average of the initial 50 images to reveal the change of contrast during the welding process compared to the initial condition. As for the second approach, each image was successively subtracted by the preceding time frame, revealing the contrast variations between two consecutive time frames, and providing a clearer visualization of the MP lengths on the top surfaces. Following both methods, the brightness and contrast of images were modified manually to enhance image contrast.  
+
+The boundaries of the keyhole and MP were first manually delineated, followed by interpolation using MATLAB to achieve smooth boundaries. The dividing point between the keyhole and MP was defined as the initial point on the keyhole rear wall with a slope of - 1. The z coordinate at this point was defined as the depth of keyhole rear wall. Then, the region between this point and the one \(350 \mu \mathrm{m}\) behind the keyhole front wall was defined as the front part of MP. This curve was then linearly fitted, with its root mean square error defined as its waviness.  
+
+### 3.3 Optical micrograph and surface topography  
+
+The top surfaces of laser weld seams were characterized by optical micrography and profilometry using Keyence VX- X3100. The laser confocal method was employed to measure the
+
+<--- Page Split --->
+
+surface topography of the weld seams. To calibrate the weld seam into a horizontal surface, the measurements were adjusted using quadratic correction based on the positional data from the base materials adjacent to the weld seam.  
+
+### 3.4 CFD simulation modeling  
+
+The simulation of thermal and mass transfer within the melt pool was performed using CFD modeling with Flow- 3D under the following assumptions: (1) The melt flow was laminar, incompressible, and Newtonian. (2) The plasma inside the keyhole and the spattering were not considered. (3) No heat or mass transfer occurred at any faces of the simulation domain. The dimensions of the simulation domain are depicted in Supplementary Fig. 2a, consisting of uniformly divided cubic grids with an edge length of \(10 \mu \mathrm{m}\) . The domain lengths were set to 3 and \(8 \mathrm{mm}\) for welding speeds of 0.33 and \(0.42 \mathrm{m / s}\) , resulting in a total number of 384,000 and 1,536,000 cells, respectively. Laser welding was conducted in the \(+x\) direction, with lengths of 2 and \(7 \mathrm{mm}\) corresponding to the respective welding speeds. The laser spot diameter was \(43 \mu \mathrm{m}\) , and its Gaussian beam distribution was modeled by the heat flux factor shown in Supplementary Fig. 2b.  
+
+The coupled governing equations in the CFD modeling included conservation of momentum, energy, and continuity \(^{37}\) . Laser absorption, thermal conduction, surface radiation and convection were incorporated for energy calculation. The temperature- dependent laser absorptivity data were acquired from reference \(^{38}\) , and the temperature- dependent physical properties were obtained from references \(^{29,30}\) . Primary forces considered in the model included recoil pressure, surface tension, viscosity, buoyancy, and gravity. Additional details regarding the force setup can be found elsewhere \(^{39}\) . Model validation was performed based on the MP length and depth (Supplementary Fig. 3). The difference of the MP length between simulation and experiment is \(8.0\%\) at \(0.33 \mathrm{m / s}\)
+
+<--- Page Split --->
+
+and \(2.9\%\) at \(1.42\mathrm{m / s}\) , respectively. In addition, for the condition where humping occurs (1.42 m/s), the linear number density of humps (defined as the number of humps per unit length), closely matched between the simulation (1.57 #/mm) and experimental observations (1.50 #/mm), as shown in Supplementary Fig. 4. The difference between simulation and experiment is \(4.7\%\) .  
+
+## Acknowledgement  
+
+This material is based upon work supported by the U.S. Department of Energy's Office of Energy Efficiency and Renewable Energy (EERE) under the Vehicles Technology Office Award Number DE- EE0009616. This research used resources of the Advanced Photon Source, a U.S. DOE Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE- AC02- 06CH11357. The authors would like to thank the experimental support from Alex Deriy at the 32- ID beamline. We also acknowledge Zixuan Wan's support for the discussions on CFD simulation. The views expressed herein do not necessarily represent the views of the U.S. Department of Energy or the United States Government.  
+
+## References  
+
+1. Katayama, S. Handbook of Laser Welding Technologies. (Elsevier, 2013).  
+2. Haddad, E. et al. Laser micro welding with fiber lasers for battery and fuel cell based electromobility. Journal of Advanced Joining Processes 5, 100085 (2022).  
+3. Yi, P., Du, X., Kan, Y., Peng, L. & Lai, X. Modeling and experimental study of laser welding distortion of thin metallic bipolar plates for PEM fuel cells. Int J Hydrogen Energy 40, 4850-4860 (2015).  
+4. Ai, Y. et al. Investigation of the humping formation in the high power and high speed laser welding. Opt Lasers Eng 107, 102-111 (2018).
+
+<--- Page Split --->
+
+5. Berger, P., Hügel, H., Hess, A., Weber, R. & Graf, T. Understanding of Humping Based on Conservation of Volume Flow. Phys Procedia 12, 232–240 (2011).  
+
+6. Xue, B. et al. Humping Formation and Suppression in High-Speed Laser Welding. Materials 15, (2022).  
+
+7. Kawahito, Y., Mizutani, M. & Katayama, S. High quality welding of stainless steel with 10 kW high power fibre laser. Science and Technology of Welding and Joining 14, 288–294 (2009).  
+
+8. Meng, X., Qin, G. & Zou, Z. Investigation of humping defect in high speed gas tungsten arc welding by numerical modelling. Mater Des 94, 69–78 (2016).  
+
+9. Nguyen, T. C., Weckman, D. C., Johnson, D. A. & Kerr, H. W. The humping phenomenon during high speed gas metal arc welding. Science and Technology of Welding and Joining 10, 447–459 (2005).  
+
+10. Wu, D., Hua, X., Ye, D. & Li, F. Understanding of humping formation and suppression mechanisms using the numerical simulation. Int J Heat Mass Transf 104, 634–643 (2017).  
+
+11. Soderstrom, E. & Mendez, P. Humping mechanisms present in high speed welding. Science and Technology of Welding and Joining 11, 572–579 (2006).  
+
+12. Yuan, L. et al. Investigation of humping phenomenon for the multi-directional robotic wire and arc additive manufacturing. Robot Comput Integr Manuf 63, 101916 (2020).  
+
+13. Yuan, L. et al. Fabrication of metallic parts with overhanging structures using the robotic wire arc additive manufacturing. J Manuf Process 63, 24–34 (2021).
+
+<--- Page Split --->
+
+14. Koli, Y., Yuvaraj, N., Sivanandam, A. & Vipin. Control of humping phenomenon and analyzing mechanical properties of Al-Si wire-arc additive manufacturing fabricated samples using cold metal transfer process. Proc Inst Mech Eng C J Mech Eng Sci 236, 984–996 (2021).  
+
+15. Tang, C., Le, K. Q. & Wong, C. H. Physics of humping formation in laser powder bed fusion. Int J Heat Mass Transf 149, 119172 (2020).  
+
+16. Bhatt, A. et al. In situ characterisation of surface roughness and its amplification during multilayer single-track laser powder bed fusion additive manufacturing. Addit Manuf 77, 103809 (2023).  
+
+17. U Gratzke, P D Kapadia, J Dowden, J Kroos & G Simon. Theoretical approach to the humping phenomenon in welding processes. J Phys D Appl Phys 25, 1640 (1992).  
+
+18. Fabbro, R. Melt pool and keyhole behaviour analysis for deep penetration laser welding. J Phys D Appl Phys 43, 445501 (2010).  
+
+19. Xue, B., Chang, B. & Du, D. Monitoring of high-speed laser welding process based on vapor plume. Opt Laser Technol 147, 107649 (2022).  
+
+20. Zhang, D. et al. Dynamic keyhole behavior and keyhole instability in high power fiber laser welding of stainless steel. Opt Laser Technol 114, 1–9 (2019).  
+
+21. Gan, Z. et al. Universal scaling laws of keyhole stability and porosity in 3D printing of metals. Nat Commun 12, 2379 (2021).  
+
+22. Cunningham, R. et al. Keyhole threshold and morphology in laser melting revealed by ultrahigh-speed x-ray imaging. Science (1979) 363, 849–852 (2019).
+
+<--- Page Split --->
+
+23. Huang, Y. et al. Keyhole fluctuation and pore formation mechanisms during laser powder bed fusion additive manufacturing. Nat Commun 13, 1170 (2022).  
+
+24. Zhao, C. et al. Critical instability at moving keyhole tip generates porosity in laser melting. Science (1979) 370, 1080–1086 (2020).  
+
+25. Beck, M., Berger, P., Dausinger, F. & Huegel, H. Aspects of keyhole/melt interaction in high-speed laser welding. in Proc.SPIE vol. 1397 769–774 (1991).  
+
+26. Aucott, L. et al. Revealing internal flow behaviour in arc welding and additive manufacturing of metals. Nat Commun 9, 5414 (2018).  
+
+27. Guo, Q. et al. Revealing melt flow instabilities in laser powder bed fusion additive manufacturing of aluminum alloy via in-situ high-speed X-ray imaging. Int J Mach Tools Manuf 175, 103861 (2022).  
+
+28. Hojjatzadeh, S. M. H. et al. Pore elimination mechanisms during 3D printing of metals. Nat Commun 10, 3088 (2019).  
+
+29. Chen, N. et al. Microstructural characteristics and crack formation in additively manufactured bimetal material of 316L stainless steel and Inconel 625. Addit Manuf 32, 101037 (2020).  
+
+30. Kim, C. S. Thermophysical Properties of Stainless Steels. https://www.osti.gov/biblio/4152287 (1975) doi:10.2172/4152287.  
+
+31. Wischgoll, T. & Scheuermann, G. Detection and visualization of closed streamlines in planar flows. IEEE Trans Vis Comput Graph 7, 165–172 (2001).
+
+<--- Page Split --->
+
+32. Ai, Y., Liu, X., Huang, Y. & Yu, L. Numerical analysis of the influence of molten pool instability on the weld formation during the high speed fiber laser welding. Int J Heat Mass Transf 160, 120103 (2020).  
+
+33. Meng, X., Qin, G. & Zou, Z. Characterization of molten pool behavior and humping formation tendency in high-speed gas tungsten arc welding. Int J Heat Mass Transf 117, 508-516 (2018).  
+
+34. van Elsen, M., Al-Bender, F. & Kruth, J. Application of dimensional analysis to selective laser melting. Rapid Prototyp J 14, 15-22 (2008).  
+
+35. Rubenchik, A. M., King, W. E. & Wu, S. S. Scaling laws for the additive manufacturing. J Mater Process Technol 257, 234-243 (2018).  
+
+36. Lei, T., Rong, Y., Xu, J. & Huang, Y. Experiment study and regression analysis of molten pool in laser welding. Opt Laser Technol 108, 534-541 (2018).  
+
+37. Dal, M. & Fabbro, R. [INVITED] An overview of the state of art in laser welding simulation. Opt Laser Technol 78, 2-14 (2016).  
+
+38. Kwon, H., Baek, W.-K., Kim, M.-S., Shin, W.-S. & Yoh, J. J. Temperature-dependent absorptance of painted aluminum, stainless steel 304, and titanium for \(1.07\mu \mathrm{m}\) and \(10.6\mu \mathrm{m}\) laser beams. Opt Lasers Eng 50, 114-121 (2012).  
+
+39. Wan, Z. et al. Novel measures for spatter prediction in laser welding of thin-gage zinc-coated steel. Int J Heat Mass Transf 167, 120830 (2021).
+
+<--- Page Split --->
+
+## 4 Supplementary Figures  
+
+![PLACEHOLDER_25_0]
+  
+
+Supplementary Fig. 1 a Experimental setup of in situ high- speed synchrotron X- ray imaging at ANL. b Geometry of the dog bone- shaped specimen. Narrow width (500 \(\mu \mathrm{m}\) ) in the gauge region is required due to the strong attenuation on X- ray for stainless steel.  
+
+![PLACEHOLDER_25_1]
+  
+
+Supplementary Fig. 2 a Setup of CFD simulation. b Heat flux factor which describes the laser profile with a Gaussian distribution.
+
+<--- Page Split --->
+![PLACEHOLDER_26_0]
+  
+
+Supplementary Fig. 3 Validation of CFD simulation models from the side view of molten pool performed at a \(0.33 \mathrm{m / s}\) , 102 W and b \(1.42 \mathrm{m / s}\) , 348 W. The difference between simulation and experiment is \(8.0\%\) for a and \(2.9\%\) for b, respectively. All scale bars correspond to \(200 \mu \mathrm{m}\) .  
+
+![PLACEHOLDER_26_1]
+  
+
+Supplementary Fig. 4 Side view of the weld geometry in CFD simulation performed at \(1.42 \mathrm{m / s}\) , 348 W. The linear number density of the hump at a welding speed of \(1.42 \mathrm{m / s}\) closely matched between the simulation \((1.57 \# / \mathrm{mm})\) and experimental observations \((1.50 \# / \mathrm{mm})\) , with a difference of \(4.7\%\) between simulation and experiment. Scale \(\mathrm{bar} = 100 \mu \mathrm{m}\) .
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- 20mfirst50.avi- 40mfirst50.avi- 60mfirst50.avi- 75mfirst50.avi- 85mfirst50.avi
+
+<--- Page Split --->

preprint/preprint__4819fa60e2ce209b34a157450ed28888546a35fcfa4cdfb7ed96c2ae1ed60b19/images_list.json

@@ -0,0 +1,122 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1 Principle of Lorentz imaging of optical fields. a A plane-wave electron beam probes the near field of a nanostructure, here a nanosphere with a dipolar field induced by external illumination. Stimulated inelastic scattering populates sidebands in the electron spectrum and imprints a spatial phase profile. b Simulated magnitude of the first electron sideband after energy-filtering and c, associated phase profile with lineouts shown below. d,e Simulated Lorentz images under overfocus conditions for loss \\((-h\\omega)\\) and gain \\((h\\omega)\\) filtered sidebands, exhibiting sensitivity to the phase profile.",
+    "footnote": [],
+    "bbox": [
+      [
+        70,
+        44,
+        485,
+        510
+      ]
+    ],
+    "page_idx": 2
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2 Polarization-dependent near-field strength. a Scanning TEM imaging of inelastic scattering at a laser-illuminated gold nanotip, recording an electron spectrum at every scanned position. An SEM image shows the nanostructure with the measurement area indicated as a dashed rectangle. b-e Maps of the magnitude of the near-field coupling coefficient \\(|g|\\) for different polarizations, extracted from the electron spectrum at each position. The shadow of the tip is overlaid in dark grey. Arrows indicate the polarization of the incident beam. f Total electron scattering probability into gain orders \\((N > 0)\\) , obtained from the map shown in e.",
+    "footnote": [],
+    "bbox": [
+      [
+        70,
+        45,
+        490,
+        303
+      ]
+    ],
+    "page_idx": 3
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3 Optical near-field Lorentz contrast at a gold nanotip. a Schematic of Lorentz-PINEM at the gold nanotip. Inelastic scattering creates a superposition of gain- and loss-scattered sideband wave functions with order-dependent optical phase profiles imprinted. b Gain-filtered in-focus TEM image. c Corresponding boundary-element method simulation. d-g Measured (d,f) and simulated (e,g) defocus Lorentz images of gain (blue frame) and loss (red frame) filtered sidebands, illustrating distinct phase-contrast features at the tip apex and shaft.",
+    "footnote": [],
+    "bbox": [
+      [
+        520,
+        46,
+        920,
+        458
+      ]
+    ],
+    "page_idx": 3
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4 Retrieval of optically induced electron phase profile. a The regularized reconstruction takes the measured coupling magnitude \\(|g|\\) and the transmission \\(T\\) of the nanostructure as an input, and retrieves the phase \\(\\arg \\{g\\}\\) by comparison of the resulting propagated intensity images with the experimental data, in a regularized iteration. b Resulting complex amplitude of the coupling parameter \\(g\\) . c Lineout (along the dashed line in b) of the phase, together with fit to a model of counter-propagating waves of unequal amplitude. The sketch indicates the main interfering components of the standing-wave pattern (see text).",
+    "footnote": [],
+    "bbox": [
+      [
+        512,
+        45,
+        930,
+        567
+      ]
+    ],
+    "page_idx": 4
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig. 5 Defocus series of the nanotip without inelastic electron light scattering. A defocus series of unfiltered bright field images taken before time-zero at \\(-10\\) ps with otherwise identical microscopy settings including the illumination.",
+    "footnote": [],
+    "bbox": [
+      [
+        348,
+        450,
+        650,
+        890
+      ]
+    ],
+    "page_idx": 9
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig. 6 Polarization control of the optical near field. A controlled change of the linear polarization of the incident laser excitation changes the optical near field at the nanotip as imaged with energy-filtered TEM. Simulated images are shown for comparison showing a similar behaviour.",
+    "footnote": [],
+    "bbox": [
+      [
+        315,
+        42,
+        683,
+        787
+      ]
+    ],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Fig. 7 Simulated image contrast formation. From a given \\(g(x,y)\\) (a) and transmission (b) maps of the specimen, the in-focus sideband magnitudes (c) can be calculated using Eq. 2. By summing up loss (d), all (e) or gain (f) sidebands, we can simulate the in-focus EFTEM images. Propagating the individual sidebands with the Fresnel propagator results in the defocused image contrast (g-j).",
+    "footnote": [],
+    "bbox": [
+      [
+        108,
+        44,
+        876,
+        468
+      ]
+    ],
+    "page_idx": 11
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_8.jpg",
+    "caption": "Fig. 8 Contribution of each plasmonic mode to the magnitude of the interaction coefficient \\(|g|\\) . Due to the rotational symmetry of the tip, the coupling coefficient can be rigorously decomposed as a sum over azimuthal numbers \\(m\\) (i.e., \\(g = \\sum_{m} g_{m}\\) ). These plots show that the coupling coefficient is dominated by the \\(m = 0\\) and \\(m = \\pm 1\\) components. Incidentally, the \\(m = \\pm 1\\) plots are asymmetric with respect to the tip axis because electron-beam positions in which the plasmons circulate along the same direction as the electron velocity are favored.",
+    "footnote": [],
+    "bbox": [
+      [
+        308,
+        45,
+        688,
+        395
+      ]
+    ],
+    "page_idx": 12
+  }
+]

preprint/preprint__481e74eddbc811fda69ed68e86d7c0484896cd9e4768cbe36b0ddb638b669048/images_list.json

@@ -0,0 +1,130 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1: The \\(c_{\\mathrm{sat}}\\) of FUS-SNAP is similar in KCl versus KGlu buffers. The timescales for forming micron-scale condensates shows differences influenced by clustering at concentrations below \\(c_{\\mathrm{sat}}\\) . (a) Schematic of chloride anion showing its electron distribution in \\(2s\\) and \\(2p\\) orbitals as well as its \\(\\mathrm{pK_a}\\) . (b) Schematic of glutamate. (c) Sample data for absorbance-based spin-down assays. Data are shown here for FUS-SNAP in \\(20\\mathrm{mM}\\) Tris.HCl \\(\\mathrm{pH}7.4\\) , with \\(100\\mathrm{mM}\\) KCl and \\(20\\mathrm{mM}\\) Tris.Glu \\(\\mathrm{pH}7.4\\) and \\(100\\mathrm{mM}\\) Kglu at \\(\\approx 25^{\\circ}\\mathrm{C}\\) . Panels (d)-(h) show bright field microscopy images collected at the 1-hour time point for solutions containing different concentrations of FUS-SNAP in \\(20\\mathrm{mM}\\) TRIS.HCl, \\(\\mathrm{pH}7.4\\) , with a final KCl concentration of \\(100\\mathrm{mM}\\) . Panels (i)-(m) show microscopy images collected at the 1-hour time point for solutions containing different concentrations of FUS-SNAP in \\(20\\mathrm{mM}\\) Tris.Glu, \\(\\mathrm{pH}7.4\\) , with \\(100\\mathrm{mM}\\) KGlu, and panels (n)-I show microscopy images collected at the 1-hour time point for solutions containing different concentrations of FUS-SNAP in \\(20\\mathrm{mM}\\) TRIS.HCl, \\(\\mathrm{pH}7.4\\) , with \\(100\\mathrm{mM}\\) KCl. In both KGlu and KCl-control buffer, the residual KCl \\((< 30\\mathrm{mM})\\) from FUS-SNAP stock was added. The total concentration of KCl is marked on the panels (i-r). For imaging purposes, \\(5\\%\\) of the total mixture in each sample is made up of Alexa 488 labeled FUS-SNAP. The scale bar in each panel corresponds to \\(10\\mu \\mathrm{m}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        205,
+        87,
+        800,
+        508
+      ]
+    ],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2: The abundance of mesoscale clusters in FUS and FUS family proteins increases in KGlu buffers compared to KCl buffers. (a) NTA data shows the volume fraction of mesoscale clusters for FUS-SNAP at different sub-saturation concentrations in \\(100\\mathrm{mM}\\) KCl and KGlu buffers. (b) Relative abundance of mesoscale clusters for FUS-SNAP at different sub-saturation concentrations in \\(100\\mathrm{mM}\\) KCl and KGlu buffers. (c) NTA data shows the volume fraction of mesoscale clusters for FUS at different sub-saturation concentrations in \\(100\\mathrm{mM}\\) KCl and KGlu buffers. (d) Single-molecule analysis by microfluidic confocal spectroscopy (MCS) shows that FUS-EGFP cluster formation increases with increasing concentration from \\(125\\mathrm{mM}\\) to \\(1000\\mathrm{mM}\\) in KGlu buffer. (e) The intensity distributions from the multiparameter fluorescence detection (MFD) experiment in the presence of \\(15\\mathrm{mM}\\) Nile red titrated with various concentrations of FUS-SNAP shows that the amount and size of cluster increase in \\(100\\mathrm{mM}\\) KGlu visibly by tailing towards higher count rates up to \\(10^{4}\\mathrm{kHz}\\) per burst compared to \\(100\\) mM KCl buffer. (f) Single-molecule FRET measurements with \\(200\\mathrm{pM}\\) of FUS-SNAP-AF488 as donor and FUS-SNAP-AF647 as acceptor in KGlu buffer (blue) compared to KCl buffer (red) show that the number of FRET events are significantly higher for KGlu (1D-histograms). The FRET populations are broadened beyond shot noise due to intermolecular dynamics (dynamic FRET line, dashed light green), resulting in a deviation from the static FRET line (solid black). DLS data shows the derived count rate of FUS-SNAP (d), TAF15-SNAP (e), and EWSR1-SNAP (f) in different buffers.",
+    "footnote": [],
+    "bbox": [
+      [
+        234,
+        88,
+        750,
+        500
+      ]
+    ],
+    "page_idx": 7
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3: Mutations modulate the extent of clustering in sub-saturated solutions and this can be influenced by glutamate. DLS data show the derived count rate of FUS(Y-S) (a), FUS(R-G) (b), FUS(R-K) (c), FUS(F/Y-G/S) (d), and FUS(10D/4E-G) (e) in different buffers.",
+    "footnote": [],
+    "bbox": [
+      [
+        180,
+        270,
+        820,
+        620
+      ]
+    ],
+    "page_idx": 9
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4: FUS-SNAP clusters form reversibly in the \\(100\\mathrm{mM}\\) KGlu buffer. (a) A schematic representation of the reversibility experiments of FUS-SNAP clusters in KGlu buffer. (b) DLS data were collected at different time points for FUS-SNAP, showing the changes in intensity vs. size distribution profiles at \\(3\\mu \\mathrm{M}\\) , \\(1\\mu \\mathrm{M}\\) , \\(2.75\\mu \\mathrm{M}\\) , and \\(1.65\\mu \\mathrm{M}\\) . (c) The confocal point measurements with multiparameter fluorescence detection (MFD) display the number of burst vs. fluorescence lifetime of Nile Red at the same concentrations shown in (b), revealing repeatedly shifts to lower lifetimes upon dilution.",
+    "footnote": [],
+    "bbox": [
+      [
+        220,
+        240,
+        780,
+        586
+      ]
+    ],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig. 5: FUS-SNAP clusters create a distinct microenvironment compared to proteins in solution. (a) The fluorescence lifetimes of Nile Red with various concentrations of FUS-SNAP equilibrated for 30 minutes in KGlu and KCl buffer. (b)-(d) The different concentrations of FUS-SNAP solutions and buffers are mixed with \\(2\\mu \\mathrm{M}\\) bis-ANS in \\(100\\mathrm{mM}\\) KGlu buffer (b), \\(100\\mathrm{mM}\\) KCl buffer (c), and \\(200\\mathrm{mM}\\) KCl buffer (d). For bis-ANS studies (b)-(d), the mixture solutions were excited using a \\(355\\mathrm{nm}\\) laser, and the emission spectra were measured from \\(425\\mathrm{nm}\\) to \\(650\\mathrm{nm}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        245,
+        81,
+        752,
+        472
+      ]
+    ],
+    "page_idx": 11
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Figure 6: Preferential exclusion of glutamate from peptide amide sites drives protein associations and increased partitioning into dense phases. (a) Schematic for the calculation of preferential interaction coefficients as a function of distance \\(r\\) from the center-of-mass of the peptide of interest. (b) Preferential interaction coefficients for different amino acids as a function radial distance \\(r\\) . The computations are of the anion-specific preferential interaction coefficients in \\(500~\\mathrm{mM}\\) KCl and \\(500~\\mathrm{mM}\\) KGIu, respectively. Error bars denote the standard error of the mean. (c) The difference in the preferential interaction coefficient for KGIu and KCl \\((\\Gamma_{\\mathrm{KGIu}} - \\Gamma_{\\mathrm{KGI}})\\) at \\(r \\approx 30 \\mathrm{\\AA}\\) for different amino acids. (d) Representative intensity time trace of FUS-EGFP samples post induction of phase separation using a single photon counting detection unit. (e) The intensity histogram of the time trace is shown in (d). (f) The \\(c_{\\mathrm{sat}}\\) of FUS-EGFP decreases when KGIu concentrations are increased at a constant KCl concentration of \\(40~\\mathrm{mM}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        140,
+        866,
+        480
+      ]
+    ],
+    "page_idx": 12
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_1.jpg",
+    "caption": "Extended Data Fig. 1: Potassium Glutamate (KGlu) buffer minimally affects the driving forces for macrophage separation of FUS-SNAP although the evolution of condensates is discernibly different. (a)-(e) shows microscopy images collected at the 4-hour time point for solutions containing different concentrations of FUS-SNAP in \\(20\\mathrm{mM}\\) Tris.HCl, pH 7.4, with a final concentration of \\(100\\mathrm{mM}\\) KCl. Panels (f)-(j) show microscopy images collected at the 4-hour time point for solutions containing different concentrations of FUS-SNAP in \\(20\\mathrm{mM}\\) TRIS.Glu, pH 7.4, with \\(100\\mathrm{mM}\\) KGlu, and panels (k)-(o) show microscopy images collected at the 4-hour time point for solutions containing different concentrations of FUS-SNAP in \\(20\\mathrm{mM}\\) Tris.HCl, pH 7.4, with \\(100\\mathrm{mM}\\) KCl. In both KGlu and KCl buffer, the residual KCl \\((< 30\\mathrm{mM})\\) from FUS-SNAP stock was added. For imaging purposes, \\(5\\%\\) of the total mixture in each sample is made up of Alexa-488 labelled FUS-SNAP. The total KCl concentration in the solution is marked on the panels. The scale bar in each panel corresponds to \\(10\\mu \\mathrm{m}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        125,
+        135,
+        870,
+        452
+      ]
+    ],
+    "page_idx": 27
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_2.jpg",
+    "caption": "Extended Data Fig. 2: Potassium Glutamate (KGlu) buffer minimally affects the driving forces for macrophage separation of FUS-SNAP, although the evolution of condensates is discernibly different. (a)-(c) The correlation function from the dynamic light scattering of solutions containing different concentrations of FUS-SNAP, \\(1 \\mu \\mathrm{M}\\) (a), \\(2 \\mu \\mathrm{M}\\) (b), and \\(3 \\mu \\mathrm{M}\\) (c) in \\(20 \\mathrm{mM}\\) Tris.HCl, pH 7.4, with a final concentration of \\(100 \\mathrm{mM}\\) KCl. Panels (d), (e), and (f) show the correlation functions of solutions containing \\(1 \\mu \\mathrm{M}\\) , \\(2 \\mu \\mathrm{M}\\) , and \\(3 \\mu \\mathrm{M}\\) concentrations of FUS-SNAP, respectively, in \\(20 \\mathrm{mM}\\) TRIS.Glu, pH 7.4, with \\(100 \\mathrm{mM}\\) KGlu. Panels (g), (h), and (i) show the correlation functions of solutions containing \\(1 \\mu \\mathrm{M}\\) , \\(2 \\mu \\mathrm{M}\\) , and \\(3 \\mu \\mathrm{M}\\) concentrations of FUS-SNAP, respectively, in \\(20 \\mathrm{mM}\\) TRIS.HCl, pH 7.4, with \\(100 \\mathrm{mM}\\) KCl. The total concentration of KCl is marked on the panels. The correlation coefficient value indicates the abundance of clusters in the solutions. The time axis correlates with the size of the species, larger the size requires more time to decay and vice versa.",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        90,
+        863,
+        659
+      ]
+    ],
+    "page_idx": 28
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_6.jpg",
+    "caption": "Extended Data Fig. 6: Site-site radial distribution functions, g(r) around backbone carbonyl oxygen atoms. These quantify the relative probability, with respect to an ideal gas prior, of finding Cl- atoms (solid curve) or the \\(sp^2\\) oxygen OE1 of glutamate around the carbonyl oxygen of the backbone for different capped amino acids. Data are shown for the g(r) of Cl- (solid curve) and OE1 atom of glutamate (dashed curve) around the backbone carbonyl oxygen of (a) Gly, (b) Ser, (c) Gln, (d) Lys, (e) Arg, and (f) Asp.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 29
+  }
+]

preprint/preprint__483a7c25f93b46e958c7647bac54d6284eb0e0588af3ad7439c754c5ccd549c1/images_list.json

@@ -0,0 +1,62 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1. (A) Frequency of children (translucent) and adults (solid) with each antimicrobial resistance gene (ARG), stratified by ARG class. (B) Number of ARGs detected in children and adults by age subgroups. Two outliers were omitted for visualization purposes; one 11-18 year-old patient with 18 ARGs detected and another 70-79 year-old patient with 12 ARGs detected. (C) Number of ARG classes detected in children and adults by age subgroups. For Figures B and C, p-values were calculated using Wilcoxon-rank sum test and adjusted for multiple comparisons with False Discovery Rate (FDR) correction. The asterisks indicate statistically significant comparisons; all had a p-value <0.01. (D) Proportion of patients with ARGs by ARG class, stratified by pediatric and adult cohorts. The 95% confidence intervals were calculated by the Clopper-Pearson exact binomial method. P-values were obtained by Pearson's Chi-square test and Fisher's exact test for samples with <5 total ARGs. (E) Beta diversity of antimicrobial resistance children and adults. P-value calculated based on the Bray-Curtis dissimilarity index and the PERMANOVA test with 1000 permutations. Abbreviation: TMP-SMX, trimethoprim-sulfamethoxazole; NMDS, nonmetric multidimensional scaling.",
+    "footnote": [],
+    "bbox": [
+      [
+        140,
+        80,
+        870,
+        690
+      ]
+    ],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. Multivariable logistic regression model evaluating the association of (A) binary age and (B) age subgroups with the presence of ARGs, accounting for sex, race/ethnicity, and lower respiratory tract infection (LRTI) status.",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        160,
+        833,
+        840
+      ]
+    ],
+    "page_idx": 12
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3. (A) Bacterial abundance in the lung microbiome measured in total bacterial alignments to the NCBI NT database per million reads sequenced (NT rpm) in children and adults by age subgroups. (B) Alpha diversity, calculated by the Shannon diversity index, of the bacterial lung microbiome of children and adults by age subgroups. (C) Beta diversity of the bacterial lung microbiome of children and adults. P-value calculated based on the Bray-Curtis dissimilarity index and the PERMANOVA test with 1000 permutations. (D) Statistically significant (p-value <0.05) differential abundant bacterial genera, by log2 fold change of bacterial counts, detected in children and adults. Bar colors indicate whether the species was more abundant in children (blue) or adults (red). (E) Frequency of the bacterial species detected in ≥5% of children (translucent) and adults (solid) among the differentially abundant bacterial genera. For patients with multiple species detected per genus, only the most abundant species was included in this analysis. Abbreviations: NT rpm, sequencing alignments to the NCBI NT database per per million reads sequenced; NMDS, nonmetric multidimensional scaling.",
+    "footnote": [],
+    "bbox": [
+      [
+        112,
+        75,
+        870,
+        700
+      ]
+    ],
+    "page_idx": 14
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4. (A) Multivariable logistic regression model evaluating the association of binary age with the presence of ARGs, accounting for total bacterial abundance (NT rpm) per patient sample, bacterial alpha diversity. (B) Statistically significant \\((p< 0.05)\\) differentially abundant bacterial genera, by \\(\\log 2\\) fold change of bacterial counts, detected in patients with ARGs compared with patients without ARGs. All detected bacterial genera were more prevalent in patients with ARGs compared with patients without ARGs.",
+    "footnote": [],
+    "bbox": [
+      [
+        128,
+        266,
+        875,
+        784
+      ]
+    ],
+    "page_idx": 15
+  }
+]

preprint/preprint__483a7c25f93b46e958c7647bac54d6284eb0e0588af3ad7439c754c5ccd549c1/preprint__483a7c25f93b46e958c7647bac54d6284eb0e0588af3ad7439c754c5ccd549c1.mmd

@@ -0,0 +1,262 @@
+
+# The antibiotic resistance reservoir of the lung microbiome expands with age  
+
+Charles Langelier ( \(\boxed{\infty}\) Chaz.Langelier@ucsf.edu) University of California, San Francisco https://orcid.org/0000- 0002- 6708- 4646  
+
+Victoria Chu UCSF https://orcid.org/0000- 0002- 7480- 965X  
+
+Alexandra Tsitsiklis UCSF  
+
+Eran Mick University of California, San Francisco https://orcid.org/0000- 0002- 7299- 808X  
+
+Lilliam Ambroggio University of Colorado and Children's Hospital Colorado  
+
+Katrina Kalantar Chan Zuckerberg Initiative  
+
+Abigail Glascock Chan Zuckerberg Biohub  
+
+Christina Osborne University of Colorado and Children's Hospital Colorado  
+
+Brandie Wagner University of Colorado and Children's Hospital Colorado  
+
+Michael Matthay University of California, San Francisco https://orcid.org/0000- 0003- 3039- 8155  
+
+Joseph DeRisi University of California, San Francisco  
+
+Carolyn Calfee University of California San Francisco  
+
+Peter Mourani Arkansas Children's Hospital  
+
+## Article  
+
+Keywords: antimicrobial resistance, antibiotic resistance, resistome, lung microbiome, metatranscriptomics, aging  
+
+Posted Date: September 18th, 2023
+
+<--- Page Split --->
+
+DOI: https://doi.org/10.21203/rs.3.rs- 3283415/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on January 2nd, 2024. See the published version at https://doi.org/10.1038/s41467- 023- 44353- 1.
+
+<--- Page Split --->
+
+## The antibiotic resistance reservoir of the lung microbiome expands with age  
+
+2 Victoria T. Chu \(^{1,2}\) , Alexandra Tsitsiklis \(^{1}\) , Eran Mick \(^{1,2,4}\) , Lilliam Ambroggio \(^{5}\) , Katrina L. Kalantar \(^{3}\) , Abigail Glascock \(^{2}\) , Christina M. Osborne \(^{5}\) , Brandie D. Wagner \(^{5,6}\) , Michael A. Matthay \(^{4}\) , Joseph L. DeRisi \(^{2,7}\) , Carolyn S. Calfee \(^{4}\) , Peter M. Mourani \(^{8}\) , Charles R. Langelier \(^{1,2*}\)  
+
+## 6 Affiliations  
+
+7 1 Division of Infectious Diseases, University of California, San Francisco, CA, USA  8 2 Chan Zuckerberg Biohub, San Francisco, CA, USA  9 3 Chan Zuckerberg Initiative, San Francisco, CA, USA  10 4 Division of Pulmonary and Critical Care Medicine, Cardiovascular Research Institute, University of California, San Francisco, CA, USA  11 5 Department of Pediatrics, University of Colorado and Children's Hospital Colorado, Aurora, CO, USA  12 6 Department of Biostatistics and Informatics, Colorado School of Public Health, University of Colorado, Aurora, CO, USA  13 7 Department of Biochemistry and Biophysics, University of California, San Francisco, CA, USA  14 8 Arkansas Children's Research Institute, Arkansas Children's Hospital, Little Rock, AR, USA  
+
+18 \*Correspondence to: chaz.langelier@ucsf.edu  
+
+## 19 Keywords  
+
+20 antimicrobial resistance, antibiotic resistance, resistome, lung microbiome, metatranscriptomics, 21 aging
+
+<--- Page Split --->
+
+Antimicrobial resistant lower respiratory tract infections (LRTI) are an increasing public health threat, and an important cause of global mortality. The lung microbiome influences LRTI susceptibility and represents an important reservoir for exchange of antimicrobial resistance genes (ARGs). Studies of the gut microbiome have found an association between age and increasing antimicrobial resistance gene (ARG) burden, however corollary studies in the lung microbiome remain absent, despite the respiratory tract representing one of the most clinically significant sites for drug resistant infections. We performed a prospective, multicenter observational study of 261 children and 88 adults with acute respiratory failure, ranging in age from 31 days to \(\geq 89\) years, admitted to intensive care units in the United States. We performed RNA sequencing on tracheal aspirates collected within 72 hours of intubation, and evaluated age-related differences in detectable ARG expression in the lung microbiome as a primary outcome. Secondary outcomes included number and classes of ARGs detected, proportion of patients with an ARG class, and composition of the lung microbiome. Multivariable logistic regression models (adults vs children) or continuous age (years) were adjusted for sex, race/ethnicity, LRTI status, and days from intubation to specimen collection. Detection of ARGs was significantly higher in adults compared with children after adjusting for sex, race/ethnicity, LRTI diagnosis, and days from intubation to specimen collection (adjusted odds ratio (aOR): 2.16, 95% confidence interval (CI): 1.10- 4.22). A greater proportion of adults compared with children had beta-lactam ARGs (31% (CI: 21- 41%) vs 13% (CI: 10- 18%)), aminoglycoside ARGs (20% (CI: 13- 30%) vs 2% (CI: 0.6- 4%)), and tetracycline ARGs (14% (CI: 7- 23%) vs 3% (CI: 1- 5%)). Adults \(\geq 70\) years old had the highest proportion of these three ARG classes. The total bacterial abundance of the lung microbiome increased with age, and microbiome alpha diversity varied with age. Taxonomic composition of the lung microbiome, measured by Bray Curtis dissimilarity index, differed between adults and children (p = 0.003). The association between age and increased ARG detection remained significant after additionally including lung microbiome total bacterial abundance and alpha diversity in the multivariable logistic regression model (aOR: 2.38, (CI: 1.25- 4.54)). Furthermore, this association remained robust when modeling age as a continuous variable (aOR: 1.02, (CI: 1.01- 1.03) per year of age). Taken together, our results demonstrate that age is an independent risk factor for ARG detection in the lower respiratory tract microbiome. These data shape our understanding of the lung resistance in critically ill patients across the lifespan, which may have implications for clinical management and global public health.
+
+<--- Page Split --->
+
+## Introduction  
+
+Antimicrobial resistance (AMR) is one of the top global health threats facing humanity<sup>1</sup>. Lower respiratory tract infections (LRTI) are a leading cause of death worldwide<sup>1,2</sup>, and account for a disproportionate burden of global AMR- related mortality, with an estimated 1.5 million deaths in 2019 attributable to resistant microbes<sup>2</sup>.  
+
+Despite the rise in AMR respiratory infections, the antimicrobial resistance genes (ARG) within the lung microbiome remain understudied and incompletely defined<sup>3</sup>. As with the gastrointestinal tract, the respiratory tract harbors diverse microbial communities acquired early during life<sup>4- 6</sup> that are continually influenced over the lifespan by exposures to organisms from the environment and other humans, as well as to antimicrobials. The gut, respiratory tract, and other human anatomical microbiomes serve as reservoirs for ARGs, or antimicrobial resistomes, and act as potential sites of ARG acquisition and transmission<sup>7</sup>.  
+
+An understanding of the epidemiological, biological, and clinical factors associated with AMR acquisition is crucial to halting the spread of resistant infections. Prior studies of the gut microbiome have demonstrated an association between age and the composition and burden of ARGs<sup>8,9</sup>, suggesting that cumulative exposures might shape the resistance landscape of endogenous microbial communities. Other factors influencing the gut resistance include travel<sup>10</sup>, hospital exposure<sup>11</sup>, and antibiotic use<sup>12</sup>. Despite these findings, corollary studies in the respiratory microbiome have not yet been performed, a key gap given the global magnitude of drug resistant LRTI. Furthermore, few studies have used metatranscriptomic RNA sequencing (RNA- seq) to both profile lower respiratory microbial ecology and detect ARG expression in the airway microbiome<sup>3,13</sup>.  
+
+Here, we sought to test the hypothesis that older age is associated with an increased prevalence of ARGs in the lung microbiome, using metatranscriptomics and multivariable logistic regression modeling. We find that age is indeed an independent risk factor for detecting ARGs in the lower airway microbiome, even after adjusting for multiple covariates including sex,
+
+<--- Page Split --->
+
+race/ethnicity, LRTI diagnosis, community- versus hospital- acquired infection, days from intubation to specimen collection, and composition of the lung microbiome.  
+
+## Methods  
+
+Study Design and Clinical Cohorts  
+
+We leveraged data from prospective pediatric<sup>14- 16</sup> and adult<sup>17</sup> cohorts of patients with acute respiratory failure admitted to intensive care units (ICUs) in the United States (USA). Pediatric patients (n=261), aged 31 days to 18 years, were enrolled from eight tertiary care hospitals in the Collaborative Pediatric Critical Care Research Network (CPCCRN) between February 2015 and December 2017. Adults (n=88), aged >18 years, were enrolled from a single tertiary care center in California, USA between July 2013 to October 2017. From each enrolled patient, tracheal aspirates (TA) were collected within 72 hours of intubation, mixed with DNA/RNA shield, and stored at - 80°C.  
+
+Electronic medical records were reviewed to obtain demographics and clinical data. LRTI status was retrospectively adjudicated by study physicians based on a previously described algorithm<sup>16,17</sup>, grouping patients as follows: 1) LRTI defined clinically, with or without a clinical microbiological diagnosis (LRTI); 2) No evidence of respiratory infection and a clear alternative etiology for the acute respiratory failure (No LRTI); or 3) patients who did not meet either above criteria (Indeterminate). LRTI was further separated into community- acquired LRTI (CA- LRTI; LRTI diagnosed within 48 hours of hospital admission), and hospital- acquired LRTI (HA- LRTI; LRTI diagnosed ≥48 hours after hospital admission).  
+
+Metatranscriptomic RNA Sequencing, Taxonomic Alignment, and Detection of ARGs RNA extracted from TA specimens underwent library preparation and paired- end Illumina sequencing, as previously described<sup>16</sup>. Quantification of microbial taxa from raw sequencing reads was carried out using the CZ- ID bioinformatics pipeline<sup>18</sup>, which performs
+
+<--- Page Split --->
+
+reference- based alignment against microbial genomes from the National Center for Biotechnology Information (NCBI) nucleotide (NT) database, as previously described<sup>18</sup>. ARGs annotated in the Antibiotic Resistance Gene- ANNOTation (ARG- ANNOT) database<sup>19</sup> were detected using the Short Read Sequence Typing (SRST2) algorithm<sup>20</sup>. Negative control water samples were processed in parallel, and a previously described negative binomial model was used to filter out microbial contaminants from the laboratory environment<sup>16</sup>. ARGs with \(< 5\%\) coverage or found in \(\geq 10\%\) of negative control water samples (TEM- 1D, TetC, Sull, OXA- 22, Aph3'la, CatA1) were excluded from the analysis.  
+
+## Statistical Analysis Framework  
+
+Age was defined in three ways: (1) a binary variable of children (31 days to 18 years) or adults (over 18 years); (2) nine subgroups of 0- 2 years, 3- 10 years, 11- 18 years, 19- 39 years, 40- 49 years, 50- 59 years, 60- 69 years, 70- 79 years, and \(\geq 80\) years; or (3) continuous age in years. We used Pearson's Chi- square test for comparison of categorical variables. P- values \(< 0.05\) were considered statistically significant. All analyses were conducted in R (v4.2.1).  
+
+## Resitome Analyses  
+
+The number of ARGs detectably expressed in the lower respiratory tract microbiome of children and adults were compared at the individual gene and ARG class (e.g., beta- lactamase) levels. P- values were calculated using the Wilcoxon rank- sum test for nonparametric continuous variables and false discovery rate (FDR) correction was applied for multiple comparisons. We compared the proportion of detected ARG classes by binary age (pediatric versus adult) and by age subgroups. 95% confidence intervals [CI] for population proportions were obtained using the Clopper- Pearson exact binomial method.  
+
+ARG abundance was calculated based on the average sequencing read depth across each gene, normalized by gene length and total reads, reported as depth per million (dpm)<sup>20,21</sup>.
+
+<--- Page Split --->
+
+Resistome alpha diversity was calculated using the Shannon Diversity Index (SDI) and ARG dpm. Beta diversity was calculated on patients with ARGs detected using the Bray- Curtis method with 1000 permutations using the PERMANOVA test and displayed via nonmetric multidimensional scaling (NMDS). Alpha and beta diversity calculations were performed using the R package vegan<sup>22</sup>.  
+
+A multivariable logistic regression model incorporating demographic and clinical characteristics (sex, race/ethnicity, LRTI status, days from intubation to specimen collection) was used to determine associations between binary age (adults vs children) and detection of ARGs Additional regression models were performed using: 1) age years as a continuous variable, and 2) the previously defined nine age subgroups. To assess for potential geographic differences in ARGs, an additional analysis was performed within the pediatric cohort only and included adjustment for U.S. census region and presence of complex chronic conditions; the latter was defined by a previously validated pediatric medical complexity algorithm<sup>23</sup>. A sensitivity analysis limited to pediatric and adult patients from the same U.S. census region was also performed. 95% confidence intervals (CI) for the multivariable logistic regression models were calculated using the Wald CI.  
+
+## Microbiome Analyses  
+
+We assessed the respiratory tract microbiome of children and adults to evaluate age- related differences in taxonomic composition and diversity, which we considered as possible confounders or mediators of the relationship between age and detectably expressed ARGs. We assessed microbiota at the genus level, calculated total bacterial abundance (measured in reads per million, rpm), and calculated bacterial alpha diversity across age subgroups. We further stratified by LRTI status (CA- LRTI, HA- LRTI, No LRTI). Lung microbiome beta diversity calculations were carried out using the Bray- Curtis dissimilarity index and PERMANOVA to assess statistical significance. Differential abundance analysis was performed using the R
+
+<--- Page Split --->
+
+package DESeq \(2^{24}\) by assessing bacterial genera in the lung microbiome present in \(\geq 20\%\) of patients. We also described the prevalence of the most abundant species within each differentially expressed genus.  
+
+Associations Between the Microbiome and Resistome Analyses  
+
+To test whether age- related differences in the lung microbiome might influence ARG results, we carried out additional analyses adjusting for bacterial abundance and alpha diversity. To test whether specific taxa might influence age- related changes in ARG detection, we performed a differential abundance analysis of bacterial genera detected in patients with or without detectable expression of ARGs, using DESeq \(2^{24}\) . Subsequently, for each differentially abundant genus, we fit individual regression models for the outcome of having ARGs detected, accounting for bacterial abundance, alpha diversity, LRTI status, and presence of one of the differentially abundant genera. Lastly, additional sensitivity analyses were performed for these models using age as a continuous variable.  
+
+## Ethics  
+
+The pediatric cohort study was approved by a single Institutional Review Board (IRB) at the University of Utah (protocol #00088656). The adult cohort study was approved by the UCSF IRB (protocol #10- 02701). Informed consent was obtained from parents or other legal guardians (pediatric patients) and from patients or their surrogates (adult patients), which included permission for collected respiratory specimens and data to be used in future studies. For the adult cohort, the IRB approved of an initial waiver consent for obtaining excess respiratory samples, and informed consent was subsequently obtained for continued study participation according to CHR protocol 10- 02701 and as previously described \(^{25}\) .
+
+<--- Page Split --->
+
+## Results  
+
+Patient Cohorts  
+
+We studied 261 children (median age: 1 year, interquartile range (IQR): 0- 15 years, range: 0- 17 years), and 88 adults (median age: 63 years, IQR: 54- 72 years, range: 21- 94 years) (Supplemental Table 1). Of the 349 patients, 231 (66%) were adjudicated as LRTI- positive, 67 (19%) had no evidence of LRTI, and 51 (15%) of patients had indeterminate LRTI status. The proportion of patients in each LRTI adjudication group did not differ between the two cohorts. Adults had a higher proportion of HA- LRTI than children (25% vs 6%, respectively), emphasizing the need to include this as a covariate in our subsequent logistic regression model. In both cohorts, 90% of the patients received antibiotics prior to tracheal aspirate collection. All four U.S. census regions (Midwest, Northeast, South, West) in the U.S. were represented among the 261 pediatric patients; adult patients were from one enrollment site located in the regional West.  
+
+## Lower Respiratory Tract Resistence  
+
+ARGs were detectably expressed in the lower respiratory tract microbiome of 40 (45%) adults compared with 53 (20%) children (Pearson's Chi- square \(\mathsf{p}< 0.01\) ). Across all patients, 74 distinct ARGs representing nine ARG classes were detected (Figure 1A). The number of detectably expressed ARGs (Figure 1B) and the number of ARG classes (Supplemental Figure 1) significantly differed between the youngest age subgroups (0- 2 years and 3- 10 years) and the oldest age subgroups (60- 69 years, 70- 79 years, and \(\geq 80\) years age groups), respectively. A significant increase was also noted between the 3- 10 and the 11- 18 years of age subgroups.  
+
+The most frequently detected ARG classes across all patients conferred resistance to beta- lactams (n=85 patients), macrolides (n=41), and aminoglycosides (n=37). A greater proportion of adults compared with children had beta- lactam ARGs (31% (CI: 21- 41%) vs 13%
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Figure 1. (A) Frequency of children (translucent) and adults (solid) with each antimicrobial resistance gene (ARG), stratified by ARG class. (B) Number of ARGs detected in children and adults by age subgroups. Two outliers were omitted for visualization purposes; one 11-18 year-old patient with 18 ARGs detected and another 70-79 year-old patient with 12 ARGs detected. (C) Number of ARG classes detected in children and adults by age subgroups. For Figures B and C, p-values were calculated using Wilcoxon-rank sum test and adjusted for multiple comparisons with False Discovery Rate (FDR) correction. The asterisks indicate statistically significant comparisons; all had a p-value <0.01. (D) Proportion of patients with ARGs by ARG class, stratified by pediatric and adult cohorts. The 95% confidence intervals were calculated by the Clopper-Pearson exact binomial method. P-values were obtained by Pearson's Chi-square test and Fisher's exact test for samples with <5 total ARGs. (E) Beta diversity of antimicrobial resistance children and adults. P-value calculated based on the Bray-Curtis dissimilarity index and the PERMANOVA test with 1000 permutations. Abbreviation: TMP-SMX, trimethoprim-sulfamethoxazole; NMDS, nonmetric multidimensional scaling. </center>
+
+<--- Page Split --->
+
+(CI: 10- 18%)), aminoglycoside ARGs (20% (CI: 13- 30%) vs 2% (CI: 0.6- 4%)), and tetracycline ARGs (14% (CI: 7- 23%) vs 3% (CI: 1- 5%)) (Figure 1C, Supplemental Table 2). When evaluated by age subgroup, the proportion of patients with beta-lactam, macrolide, or tetracycline ARGs was highest in patients \(\geq 70\) years of age (Supplemental Figure 2). Among young children, 13% (95% CI: 8- 19%) of patients aged 0- 2 years had a beta-lactam ARG compared with 2% (95% CI: 0.05- 10%) of patients aged 3- 10 years; this pattern was not seen for the other ARG classes. Among the beta-lactam ARGs, we detected six AmpC beta-lactamase genes, five extended- spectrum beta-lactamase genes, and 2 carbapenemase genes (Supplemental Figure 3).  
+
+ARG alpha diversity as measured by the Shannon Diversity Index increased primarily in patients \(\geq 60\) years of age (Figure 1D). The composition of the lung resistome significantly differed between children and adults, as measured by the Bray Curtis dissimilarity index (p = 0.003 by PERMANOVA) (Figure 1E).  
+
+In a logistic regression model assessing the association of binary age group with detection of any ARGs, accounting for sex, race/ethnicity, LRTI status (CA- LRTI, HA- LRTI, No LRTI), and days from intubation to specimen collection, the risk of ARG detection was increased in adults compared with children (adjusted odds ratio [aOR]: 2.16, 95% CI: 1.10- 4.22) (Figure 2A). Age remained significant in a sensitivity analysis of the same logistic regression model using age as a continuous variable (Supplemental Table 3). In a second sensitivity analysis using a regression model based on age subgroups (Figure 2B), children aged 3- 10 years had a lower risk (aOR: 0.32, 95% CI: 0.10- 0.97) and adults \(\geq 80\) years of age had a higher risk of detectably expressed ARGs (aOR: 6.21, 95% CI: 1.33- 28.99) compared with children aged 0- 2 years.  
+
+In an analysis restricted exclusively to children and accounting for enrollment U.S. census region and presence of a complex chronic condition, children 3- 10 years of age continued to have the lowest risk of having detectably expressed ARGs, however enrollment
+
+<--- Page Split --->
+
+237 site was a significant risk factor (Supplemental Table 4). Given this, we performed a sensitivity  
+
+238 analysis limited to pediatric and adult patients from the same U.S. census region (West), and  
+
+![](images/Figure_2.jpg)
+
+<center>Figure 2. Multivariable logistic regression model evaluating the association of (A) binary age and (B) age subgroups with the presence of ARGs, accounting for sex, race/ethnicity, and lower respiratory tract infection (LRTI) status. </center>
+
+<--- Page Split --->
+
+found that binary age group remained associated with increased risk of ARG detection (uOR: 2.33, 95% CI: 1.2- 4.52).  
+
+Lower Respiratory Tract Microbiome  
+
+The total bacterial abundance of the lung microbiome increased with age (Figure 3A). Bacterial microbiome alpha diversity initially increased during childhood and adulthood, peaked in the 40- 49 year- old age group, and then decreased in older adults (Figure 3B). The community composition of the bacterial respiratory microbiome differed between children and adults, based on Bray Curtis dissimilarity index ( \(p < 0.01\) by PERMANOVA) (Figure 3C). Differential abundance analysis revealed eight bacterial genera with statistically significant differences in abundance between children and adults (Enterococcus, Pseudomonas, Staphylococcus, Bacteroides, Prevotella, Mannheimia, Haemophilus and Moraxella). The most abundant bacterial species within each of these genera also differed between age groups (Figure 3E).  
+
+Interactions Between the Lower Respiratory Tract Microbiome and Resistome  
+
+In a logistic regression model accounting for total bacterial abundance, bacterial alpha diversity, and LRTI status, binary age group remained associated with an increased risk of ARG detection (aOR: 2.38, 95% CI: 1.25- 4.54) (Figure 4A). Differential abundance analysis revealed seven bacterial genera with statistically significant differences between patients with or without detectable ARGs (Figure 4B). In individual fitted logistic regression models where the outcome was presence of ARGs, and independent variables included age group, total bacterial abundance, bacterial alpha diversity, LRTI status, and one of the seven differentially abundant bacterial genera, adults remained associated with an increased risk of having ARGs detected compared with children (Figure 4A, Supplemental Table 5).
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3. (A) Bacterial abundance in the lung microbiome measured in total bacterial alignments to the NCBI NT database per million reads sequenced (NT rpm) in children and adults by age subgroups. (B) Alpha diversity, calculated by the Shannon diversity index, of the bacterial lung microbiome of children and adults by age subgroups. (C) Beta diversity of the bacterial lung microbiome of children and adults. P-value calculated based on the Bray-Curtis dissimilarity index and the PERMANOVA test with 1000 permutations. (D) Statistically significant (p-value <0.05) differential abundant bacterial genera, by log2 fold change of bacterial counts, detected in children and adults. Bar colors indicate whether the species was more abundant in children (blue) or adults (red). (E) Frequency of the bacterial species detected in ≥5% of children (translucent) and adults (solid) among the differentially abundant bacterial genera. For patients with multiple species detected per genus, only the most abundant species was included in this analysis. Abbreviations: NT rpm, sequencing alignments to the NCBI NT database per per million reads sequenced; NMDS, nonmetric multidimensional scaling. </center>
+
+<--- Page Split --->
+
+264 In these models, Fusobacterium spp. (aOR: 2.54, 95% CI: 1.32- 4.90), Bacteroides spp. (aOR: 2.23, 95% CI: 1.15- 4.33), Enterococcus spp. (aOR: 2.17, 95% CI: 1.04- 4.51), Staphylococcus spp. (aOR: 2.09, 95% CI: 1.09- 4.02), and Prevotella spp. (aOR: 1.98, 95% CI: 1.05- 3.72) were statistically significant risk factors. Sensitivity analyses of all these models using age as a continuous variable demonstrated similar results (Supplemental Table 6).  
+
+![](images/Figure_4.jpg)
+
+<center>Figure 4. (A) Multivariable logistic regression model evaluating the association of binary age with the presence of ARGs, accounting for total bacterial abundance (NT rpm) per patient sample, bacterial alpha diversity. (B) Statistically significant \((p< 0.05)\) differentially abundant bacterial genera, by \(\log 2\) fold change of bacterial counts, detected in patients with ARGs compared with patients without ARGs. All detected bacterial genera were more prevalent in patients with ARGs compared with patients without ARGs. </center>
+
+<--- Page Split --->
+
+## Discussion  
+
+Utilizing metatranscriptomics, we identify age as an independent risk factor for ARG detection in the lung microbiome among critically ill, recently intubated patients. We find that detection of ARGs in the lower respiratory tract increases across most of the age spectrum, with the oldest patients harboring the highest number of ARGs detectably expressed at the individual gene and class levels. These findings advance our understanding of the lung microbiome as a potential antimicrobial resistance reservoir and highlight its potential contribution to drug- resistant respiratory infections.  
+
+Across every ARG class examined, adults had a greater number of ARGs detected compared to children. The majority of detected ARGs in the pediatric cohort conferred resistance to beta- lactams and macrolides, while in adults, beta- lactam and aminoglycoside ARGs were most prevalent. Studies of nasal samples from healthy neonates<sup>26</sup> and neonates with cystic fibrosis<sup>27</sup> found a similar beta- lactam ARG dominance, while oral flora samples from children<sup>28</sup> and sputum samples from adults<sup>29</sup> found that macrolide resistance genes were most prevalent in the oropharynx. Intriguingly, we found that children 0- 2 years of age had a higher proportion of beta- lactam ARGs detected than those 3- 10 years of age. This may reflect maternally- derived microbial communities and associated ARGs acquired at birth, which go on to comprise the lung microbiome during the first years of life<sup>30</sup>.  
+
+Our findings differ from gut resistome studies which have found that tetracycline ARGs are most prevalent, followed by macrolides and beta- lactam ARGs<sup>31</sup>. These differences may reflect differences in the routes of ARG and antibiotic exposure (e.g., inhaled, ingested, intravenous) and highlight potential AMR transmission differences from the lungs compared with the gut. Indeed, recent work demonstrates the presence of ARGs and microbiota in urban air samples<sup>32,33</sup>, suggesting that environmental exposures may be a relevant route of lung resistome exchange particularly for patients with prolonged or repeated exposure to healthcare facilities where resistant microbes are more prevalent.
+
+<--- Page Split --->
+
+In the U.S., the most prescribed outpatient antibiotics are beta- lactams, fluoroquinolones, and macrolides<sup>34</sup>, while beta- lactams, macrolides, and glycopeptides are the most frequently prescribed inpatient antibiotics<sup>35</sup>. While beta- lactam ARGs were the most prevalent ARG class detected in both children and adults, adults had a greater proportion of aminoglycoside and tetracycline ARGs. Given that these antimicrobial classes have been widely used in the agriculture and livestock industries<sup>36</sup>, exposures to environmental bacteria harboring these ARGs over the lifespan could be one possible explanation. Other possible explanations include community exposures to other individuals, co- selection of ARGs on mobile genetic elements carrying multiple ARGs, or cross- resistance due to multi- drug efflux pumps<sup>37</sup>.  
+
+We also observed differences in the lower respiratory tract microbiome with age, including bacterial abundance, diversity, and taxonomic composition. Our findings are in line with a prior study demonstrating that bacterial abundance in the lung microbiome of CF patients increases with age<sup>38</sup>. We also found that bacterial alpha diversity increased in childhood, peaked in middle age, and decreased with older age. The role of endogenous respiratory microbiota in both the pathophysiology and diagnosis of critical illness syndromes in increasingly recognized, and our results suggest that adjusting for age should be considered in clinical and translational studies of the lung microbiome.  
+
+Even when accounting for demographic, clinical, and microbiome differences, age remained an independent risk factor for ARG detection. Our findings raise the possibility that selective environmental pressures driving AMR acquisition from the environment may be continuous over the lifespan, shaping the airway microbiome and associated resistome. While the detection of ARG expression in the airway microbiome does not equate to clinically relevant resistance, it suggests the potential for development of phenotypic resistance<sup>3</sup> with possible implications for patient care. Commensal bacteria within the lung microbiome can exchange ARGs via horizontal gene transfer to pathogens or pathobionts, leading to the emergence of
+
+<--- Page Split --->
+
+drug resistant LRTI<sup>39- 41</sup>. Further research is needed to better characterize and understand the prevalence, acquisition, and transmission dynamics of ARGs within the lung microbiome.  
+
+This study has limitations. First, there was not an even distribution of patients across all ages, with a greater number of young children and older adults, reflecting the distribution of the critically ill, mechanically ventilated patient populations. To account for this, we performed sensitivity analyses using age as a continuous variable or using age subgroups. Second, there were differences in geographic location and timing of tracheal aspirate collection between age groups, which we accounted for by incorporating the variables into the multivariable logistic regression models. Third, the study included only patients from the U.S. and may not be representative of the global population. Fourth, detection of ARGs is biased towards the most abundant taxa in the lung microbiome and we are likely missing detection of ARGs from less abundant taxa. Finally, enrollment occurred prior to the COVID- 19 pandemic. Thus, ARG abundance and class profiles may be different than in the current population given the increase in antibiotic use and AMR infections since \(2020^{42,43}\) .  
+
+In summary, we demonstrate that age is independently associated with the detection of ARGs in the lung microbiome in a population of critically ill patients soon after intubation. Our results suggest that healthcare, community, and environmental exposures throughout life may contribute to the reservoir of ARGs in the respiratory tract. Taken together, these findings advance our understanding of AMR in the context of the human microbiome and have implications for the management of infectious diseases, antimicrobial stewardship programs, and public health policies.
+
+<--- Page Split --->
+
+## Data and Code Availability  
+
+FASTQ files containing non- host reads identified by the CZ- ID pipeline, following subtraction of reads aligning to the human genome, are available in the NCBI Sequence Read Archive (SRA) database under BioProject accessions PRJNA875913 and PRJNA450137. All data, code, and results are available at: https://github.com/victoriatchu/agingAMR.  
+
+## Acknowledgements  
+
+This work was supported in part by the following grants from NHLBI: R01HL155418 (CRL, PMM), R01HL124103 (PMM), K23HL138461 (CRL), and R53HL140026 (CSC), as well as grants from the Chan Zuckerberg Biohub (VC, CRL, JLD). The investigators thank all patients and their families for participating in this project. We also would like to acknowledge the contributions from principal investigators, co- investigators, research coordinators, and allied research personnel at the following sites: University of California San Francisco, San Francisco, CA; Children's Hospital of Colorado, Aurora, CO; Chan Zuckerberg Biohub; Children's Hospital of Michigan, Detroit, MI; Children's Hospital of Philadelphia, Philadelphia, PA; Children's National Medical Center, Washington, DC; Nationwide Children's Hospital, Columbus, OH; Mattel Children's Hospital, University of California Los Angeles, Los Angeles, CA; Children's Hospital of Pittsburgh, University of Pittsburgh Medical Center, Pittsburgh, PA; University of Utah; and Data Coordinating Center, Salt Lake City, Utah.  
+
+Funding: NHLBI R01HL155418, NHLBI K23HL138461, Chan Zuckerberg Biohub, NHLBI R53HL140026 (CSC)
+
+<--- Page Split --->
+
+365 References:366 1 EclinicalMedicine. Antimicrobial resistance: a top ten global public health threat. 367 eClinicalMedicine 41 (2021). https://doi.org:10.1016/j.eclinm.2021.101221368 2 Murray, C. J. L. et al. Global burden of bacterial antimicrobial resistance in 2019: a 369 systematic analysis. The Lancet 399, 629- 655 (2022). https://doi.org:10.1016/S0140- 3706(21)02724- 0371 3 Serpa, P. H. et al. Metagenomic prediction of antimicrobial resistance in critically ill 372 patients with lower respiratory tract infections. Genome Medicine 14, 74 (2022). https://doi.org:10.1186/s13073- 022- 01072- 4374 4 Charlson, E. S. et al. Topographical continuity of bacterial populations in the healthy 375 human respiratory tract. American journal of respiratory and critical care medicine 184, 957- 963 (2011).377 5 Muhlebach, M. S. et al. Initial acquisition and succession of the cystic fibrosis lung 378 microbiome is associated with disease progression in infants and preschool children. 379 PLOS Pathogens 14, e1006798 (2018). https://doi.org:10.1371/journal.ppat.1006798 380 6 Tracy, M., Cogen, J. & Hoffman, L. R. The pediatric microbiome and the lung. Curr 381 Opin Pediatr 27, 348- 355 (2015). https://doi.org:10.1097/mop.0000000000000212 382 7 Pereira- Dias, J. et al. The Gut Microbiome of Healthy Vietnamese Adults and Children Is 383 a Major Reservoir for Resistance Genes Against Critical Antimicrobials. J Infect Dis 224, 5840- s847 (2021). https://doi.org:10.1093/infdis/jiab398 384 8 Lu, N. et al. DNA microarray analysis reveals that antibiotic resistance- gene diversity in 385 human gut microbiota is age related. Sci Rep 4, 4302 (2014). https://doi.org:10.1038/srep04302 386 9 Wu, L. et al. Metagenomics- Based Analysis of the Age- Related Cumulative Effect of 387 Antibiotic Resistance Genes in Gut Microbiota. Antibiotics (Basel) 10 (2021). 388 https://doi.org:10.3390/antibiotics1008106390 1 10 Boolchandani, M. et al. Impact of international travel and diarrhea on gut microbiome 392 and resistome dynamics. Nature Communications 13, 7485 (2022). 393 https://doi.org:10.1038/s41467- 022- 34862- w394 11 Kohler, P. P. et al. Emergence of Carbapenemase- Producing Enterobacteriaceae, South- 395 Central Ontario, Canada(1). Emerg Infect Dis 24, 1674- 1682 (2018). 396 https://doi.org:10.3201/eid2409.180164397 12 Reyman, M. et al. Effects of early- life antibiotics on the developing infant gut 398 microbiome and resistome: a randomized trial. Nat Commun 13, 893 (2022). 399 https://doi.org:10.1038/s41467- 022- 28525- z400 13 Spottiswoode, N. et al. Pneumonia surveillance with culture- independent 401 metatranscriptomics in HIV- positive adults in Uganda: a cross- sectional study. The 402 Lancet Microbe 3, e357- e365 (2022). https://doi.org:10.1016/S2666- 5247(21)00357- 8403 14 Tsitsiklis, A. et al. Lower respiratory tract infections in children requiring mechanical 404 ventilation: a multicentre prospective surveillance study incorporating airway 405 metagenomics. The Lancet Microbe 3, e284- e293 (2022). https://doi.org:10.1016/S2666- 5247(21)00304- 9407 15 Mourani, P. M. et al. Temporal airway microbiome changes related to ventilator- 408 associated pneumonia in children. Eur Respir J 57 (2021). 409 https://doi.org:10.1183/13993003.01829- 2020
+
+<--- Page Split --->
+
+410 16 Mick, E. et al. Integrated host/microbe metagenomics enables accurate lower respiratory tract infection diagnosis in critically ill children. J Clin Invest 133 (2023). 411 https://doi.org:10.1172/jci165904 412 Langelier, C. et al. Integrating host response and unbiased microbe detection for lower respiratory tract infection diagnosis in critically ill adults. Proc Natl Acad Sci U S A 115, E12353-e12362 (2018). https://doi.org:10.1073/pnas.1809700115 413 18 Kalantar, K. L. et al. IDseq—An open source cloud-based pipeline and analysis service for metagenomic pathogen detection and monitoring. GigaScience 9 (2020). https://doi.org:10.1093/gigascience/giaa111 419 Gupta, S. K. et al. ARG- ANNOT, a new bioinformatic tool to discover antibiotic resistance genes in bacterial genomes. Antimicrob Agents Chemother 58, 212- 220 (2014). https://doi.org:10.1128/aac.01310- 13 422 20 Inouye, M. et al. SRST2: Rapid genomic surveillance for public health and hospital microbiology labs. Genome Medicine 6, 90 (2014). https://doi.org:10.1186/s13073- 014- 0090- 6 425 21 Tong, X. et al. Alterations to the Lung Microbiome in Idiopathic Pulmonary Fibrosis Patients. Front Cell Infect Microbiol 9, 149 (2019). https://doi.org:10.3389/fcimb.2019.00149 428 22 vegan: Community Ecology Package_v. R package version 2.6- 4 (2022). 429 23 Simon, T. D. et al. Pediatric Medical Complexity Algorithm: A New Method to Stratify Children by Medical Complexity. Pediatrics 133, e1647- e1654 (2014). https://doi.org:10.1542/peds.2013- 3875 432 24 Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA- seq data with DESeq2. Genome biology 15, 1- 21 (2014). 433 Sarma, A. et al. Tracheal aspirate RNA sequencing identifies distinct immunological features of COVID- 19 ARDS. Nat Commun 12, 5152 (2021). https://doi.org:10.1038/s41467- 021- 25040- 5 436 26 Manenzhe, R. I. et al. Longitudinal changes in the nasopharyngeal resistome of South African infants using shotgun metagenomic sequencing. PLoS One 15, e0231887 (2020). https://doi.org:10.1371/journal.pone.0231887 440 27 Allemann, A. et al. Nasal resistome development in infants with cystic fibrosis in the first year of life. Frontiers in microbiology 10, 212 (2019). 441 28 Sukumar, S. et al. Development of the oral resistome during the first decade of life. Nat Commun 14, 1291 (2023). https://doi.org:10.1038/s41467- 023- 36781- w 444 29 Aogain, M. M. et al. Metagenomics Reveals a Core Macrolide Resistome Related to Microbiota in Chronic Respiratory Disease. American Journal of Respiratory and Critical Care Medicine 202, 433- 447 (2020). https://doi.org:10.1164/rccm.201911- 22020C 448 30 Yagi, K., Asai, N., Huffnagle, G. B., Lukacs, N. W. & Fonseca, W. Early- Life Lung and Gut Microbiota Development and Respiratory Syncytial Virus Infection. Front Immunol 13, 877771 (2022). https://doi.org:10.3389/fimmu.2022.877771 451 31 Hu, Y. et al. Metagenome- wide analysis of antibiotic resistance genes in a large cohort of human gut microbiota. Nat Commun 4, 2151 (2013). https://doi.org:10.1038/ncomms3151 454 32 Li, J. et al. Global Survey of Antibiotic Resistance Genes in Air. Environmental Science & Technology 52, 10975- 10984 (2018). https://doi.org:10.1021/acs.est.8b02204
+
+<--- Page Split --->
+
+456 33 Zhu, G. et al. Air pollution could drive global dissemination of antibiotic resistance genes. Isme j 15, 270- 281 (2021). https://doi.org:10.1038/s41396- 020- 00780- 2
+457 34 Durkin, M. J. et al. Outpatient Antibiotic Prescription Trends in the United States: A National Cohort Study. Infect Control Hosp Epidemiol 39, 584- 589 (2018).
+458 35 Baggs, J., Fridkin, S. K., Pollack, L. A., Srinivasan, A. & Jernigan, J. A. Estimating National Trends in Inpatient Antibiotic Use Among US Hospitals From 2006 to 2012.
+459 36 JAMA Intern Med 176, 1639- 1648 (2016).
+460 37 https://doi.org:10.1001/jamainternmed.2016.5651
+461 38 Mulchandani, R., Wang, Y., Gilbert, M. & Van Boeckel, T. P. Global trends in antimicrobial use in food- producing animals: 2020 to 2030. PLOS Global Public Health 3, e0001305 (2023). https://doi.org:10.1371/journal.pgph.0001305
+462 37 Partridge, S. R., Kwong, S. M., Firth, N. & Jensen, S. O. Mobile Genetic Elements Associated with Antimicrobial Resistance. Clinical Microbiology Reviews 31, e00088- 00017 (2018). https://doi.org:10.1128/CMR.00088- 17
+463 38 Coburn, B. et al. Lung microbiota across age and disease stage in cystic fibrosis. Sci Rep 5, 10241 (2015). https://doi.org:10.1038/srep10241
+464 39 Vinayamohan, P. G., Pellissery, A. J. & Venkitanarayanan, K. Role of horizontal gene transfer in the dissemination of antimicrobial resistance in food animal production. Current Opinion in Food Science 47, 100882 (2022).
+465 40 Launay, A., Ballard, S. A., Johnson, P. D. R., Grayson, M. L. & Lambert, T. Transfer of Vancomycin Resistance Transposon Tn1549 from Clostridium symbiosum to Enterococcus spp. in the Gut of Gnotobiotic Mice. Antimicrobial Agents and Chemotherapy 50, 1054- 1062 (2006). https://doi.org:doi:10.1128/AAC.50.3.1054- 1062.2006
+466 41 Bengtsson- Palme, J. et al. The Human Gut Microbiome as a Transporter of Antibiotic Resistance Genes between Continents. Antimicrob Agents Chemother 59, 6551- 6560 (2015). https://doi.org:10.1128/aac.00933- 15
+467 42 Nandi, A., Pecetta, S. & Bloom, D. E. Global antibiotic use during the COVID- 19 pandemic: analysis of pharmaceutical sales data from 71 countries, 2020&#36;x2013;2022. eClinicalMedicine 57 (2023). https://doi.org:10.1016/j.eclinm.2023.101848
+468 43 Malik, S. S. & Mundra, S. Increasing Consumption of Antibiotics during the COVID- 19 Pandemic: Implications for Patient Health and Emerging Anti- Microbial Resistance. Antibiotics (Basel) 12 (2022). https://doi.org:10.3390/antibiotics12010045
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+Supplementalfinal.pdf
+
+<--- Page Split --->

preprint/preprint__483a7c25f93b46e958c7647bac54d6284eb0e0588af3ad7439c754c5ccd549c1/preprint__483a7c25f93b46e958c7647bac54d6284eb0e0588af3ad7439c754c5ccd549c1_det.mmd

@@ -0,0 +1,334 @@
+<|ref|>title<|/ref|><|det|>[[44, 107, 839, 177]]<|/det|>
+# The antibiotic resistance reservoir of the lung microbiome expands with age  
+
+<|ref|>text<|/ref|><|det|>[[44, 195, 746, 238]]<|/det|>
+Charles Langelier ( \(\boxed{\infty}\) Chaz.Langelier@ucsf.edu) University of California, San Francisco https://orcid.org/0000- 0002- 6708- 4646  
+
+<|ref|>text<|/ref|><|det|>[[44, 243, 465, 284]]<|/det|>
+Victoria Chu UCSF https://orcid.org/0000- 0002- 7480- 965X  
+
+<|ref|>text<|/ref|><|det|>[[44, 290, 216, 330]]<|/det|>
+Alexandra Tsitsiklis UCSF  
+
+<|ref|>text<|/ref|><|det|>[[44, 336, 748, 378]]<|/det|>
+Eran Mick University of California, San Francisco https://orcid.org/0000- 0002- 7299- 808X  
+
+<|ref|>text<|/ref|><|det|>[[44, 382, 536, 424]]<|/det|>
+Lilliam Ambroggio University of Colorado and Children's Hospital Colorado  
+
+<|ref|>text<|/ref|><|det|>[[44, 429, 285, 469]]<|/det|>
+Katrina Kalantar Chan Zuckerberg Initiative  
+
+<|ref|>text<|/ref|><|det|>[[44, 475, 273, 515]]<|/det|>
+Abigail Glascock Chan Zuckerberg Biohub  
+
+<|ref|>text<|/ref|><|det|>[[44, 521, 536, 562]]<|/det|>
+Christina Osborne University of Colorado and Children's Hospital Colorado  
+
+<|ref|>text<|/ref|><|det|>[[44, 568, 536, 609]]<|/det|>
+Brandie Wagner University of Colorado and Children's Hospital Colorado  
+
+<|ref|>text<|/ref|><|det|>[[44, 614, 746, 655]]<|/det|>
+Michael Matthay University of California, San Francisco https://orcid.org/0000- 0003- 3039- 8155  
+
+<|ref|>text<|/ref|><|det|>[[44, 660, 388, 701]]<|/det|>
+Joseph DeRisi University of California, San Francisco  
+
+<|ref|>text<|/ref|><|det|>[[44, 707, 386, 747]]<|/det|>
+Carolyn Calfee University of California San Francisco  
+
+<|ref|>text<|/ref|><|det|>[[44, 753, 304, 793]]<|/det|>
+Peter Mourani Arkansas Children's Hospital  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 835, 101, 852]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 872, 783, 915]]<|/det|>
+Keywords: antimicrobial resistance, antibiotic resistance, resistome, lung microbiome, metatranscriptomics, aging  
+
+<|ref|>text<|/ref|><|det|>[[44, 933, 352, 952]]<|/det|>
+Posted Date: September 18th, 2023
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 45, 474, 64]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 3283415/v1  
+
+<|ref|>text<|/ref|><|det|>[[42, 82, 911, 125]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 143, 531, 163]]<|/det|>
+Additional Declarations: There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[42, 199, 935, 242]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on January 2nd, 2024. See the published version at https://doi.org/10.1038/s41467- 023- 44353- 1.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[70, 89, 828, 110]]<|/det|>
+## The antibiotic resistance reservoir of the lung microbiome expands with age  
+
+<|ref|>text<|/ref|><|det|>[[70, 137, 880, 207]]<|/det|>
+2 Victoria T. Chu \(^{1,2}\) , Alexandra Tsitsiklis \(^{1}\) , Eran Mick \(^{1,2,4}\) , Lilliam Ambroggio \(^{5}\) , Katrina L. Kalantar \(^{3}\) , Abigail Glascock \(^{2}\) , Christina M. Osborne \(^{5}\) , Brandie D. Wagner \(^{5,6}\) , Michael A. Matthay \(^{4}\) , Joseph L. DeRisi \(^{2,7}\) , Carolyn S. Calfee \(^{4}\) , Peter M. Mourani \(^{8}\) , Charles R. Langelier \(^{1,2*}\)  
+
+<|ref|>sub_title<|/ref|><|det|>[[70, 227, 210, 245]]<|/det|>
+## 6 Affiliations  
+
+<|ref|>text<|/ref|><|det|>[[70, 258, 886, 460]]<|/det|>
+7 1 Division of Infectious Diseases, University of California, San Francisco, CA, USA  8 2 Chan Zuckerberg Biohub, San Francisco, CA, USA  9 3 Chan Zuckerberg Initiative, San Francisco, CA, USA  10 4 Division of Pulmonary and Critical Care Medicine, Cardiovascular Research Institute, University of California, San Francisco, CA, USA  11 5 Department of Pediatrics, University of Colorado and Children's Hospital Colorado, Aurora, CO, USA  12 6 Department of Biostatistics and Informatics, Colorado School of Public Health, University of Colorado, Aurora, CO, USA  13 7 Department of Biochemistry and Biophysics, University of California, San Francisco, CA, USA  14 8 Arkansas Children's Research Institute, Arkansas Children's Hospital, Little Rock, AR, USA  
+
+<|ref|>text<|/ref|><|det|>[[68, 492, 485, 510]]<|/det|>
+18 \*Correspondence to: chaz.langelier@ucsf.edu  
+
+<|ref|>sub_title<|/ref|><|det|>[[68, 538, 203, 556]]<|/det|>
+## 19 Keywords  
+
+<|ref|>text<|/ref|><|det|>[[67, 570, 880, 607]]<|/det|>
+20 antimicrobial resistance, antibiotic resistance, resistome, lung microbiome, metatranscriptomics, 21 aging
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 120, 879, 884]]<|/det|>
+Antimicrobial resistant lower respiratory tract infections (LRTI) are an increasing public health threat, and an important cause of global mortality. The lung microbiome influences LRTI susceptibility and represents an important reservoir for exchange of antimicrobial resistance genes (ARGs). Studies of the gut microbiome have found an association between age and increasing antimicrobial resistance gene (ARG) burden, however corollary studies in the lung microbiome remain absent, despite the respiratory tract representing one of the most clinically significant sites for drug resistant infections. We performed a prospective, multicenter observational study of 261 children and 88 adults with acute respiratory failure, ranging in age from 31 days to \(\geq 89\) years, admitted to intensive care units in the United States. We performed RNA sequencing on tracheal aspirates collected within 72 hours of intubation, and evaluated age-related differences in detectable ARG expression in the lung microbiome as a primary outcome. Secondary outcomes included number and classes of ARGs detected, proportion of patients with an ARG class, and composition of the lung microbiome. Multivariable logistic regression models (adults vs children) or continuous age (years) were adjusted for sex, race/ethnicity, LRTI status, and days from intubation to specimen collection. Detection of ARGs was significantly higher in adults compared with children after adjusting for sex, race/ethnicity, LRTI diagnosis, and days from intubation to specimen collection (adjusted odds ratio (aOR): 2.16, 95% confidence interval (CI): 1.10- 4.22). A greater proportion of adults compared with children had beta-lactam ARGs (31% (CI: 21- 41%) vs 13% (CI: 10- 18%)), aminoglycoside ARGs (20% (CI: 13- 30%) vs 2% (CI: 0.6- 4%)), and tetracycline ARGs (14% (CI: 7- 23%) vs 3% (CI: 1- 5%)). Adults \(\geq 70\) years old had the highest proportion of these three ARG classes. The total bacterial abundance of the lung microbiome increased with age, and microbiome alpha diversity varied with age. Taxonomic composition of the lung microbiome, measured by Bray Curtis dissimilarity index, differed between adults and children (p = 0.003). The association between age and increased ARG detection remained significant after additionally including lung microbiome total bacterial abundance and alpha diversity in the multivariable logistic regression model (aOR: 2.38, (CI: 1.25- 4.54)). Furthermore, this association remained robust when modeling age as a continuous variable (aOR: 1.02, (CI: 1.01- 1.03) per year of age). Taken together, our results demonstrate that age is an independent risk factor for ARG detection in the lower respiratory tract microbiome. These data shape our understanding of the lung resistance in critically ill patients across the lifespan, which may have implications for clinical management and global public health.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 90, 223, 107]]<|/det|>
+## Introduction  
+
+<|ref|>text<|/ref|><|det|>[[112, 119, 876, 234]]<|/det|>
+Antimicrobial resistance (AMR) is one of the top global health threats facing humanity<sup>1</sup>. Lower respiratory tract infections (LRTI) are a leading cause of death worldwide<sup>1,2</sup>, and account for a disproportionate burden of global AMR- related mortality, with an estimated 1.5 million deaths in 2019 attributable to resistant microbes<sup>2</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[112, 247, 880, 459]]<|/det|>
+Despite the rise in AMR respiratory infections, the antimicrobial resistance genes (ARG) within the lung microbiome remain understudied and incompletely defined<sup>3</sup>. As with the gastrointestinal tract, the respiratory tract harbors diverse microbial communities acquired early during life<sup>4- 6</sup> that are continually influenced over the lifespan by exposures to organisms from the environment and other humans, as well as to antimicrobials. The gut, respiratory tract, and other human anatomical microbiomes serve as reservoirs for ARGs, or antimicrobial resistomes, and act as potential sites of ARG acquisition and transmission<sup>7</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[112, 470, 881, 779]]<|/det|>
+An understanding of the epidemiological, biological, and clinical factors associated with AMR acquisition is crucial to halting the spread of resistant infections. Prior studies of the gut microbiome have demonstrated an association between age and the composition and burden of ARGs<sup>8,9</sup>, suggesting that cumulative exposures might shape the resistance landscape of endogenous microbial communities. Other factors influencing the gut resistance include travel<sup>10</sup>, hospital exposure<sup>11</sup>, and antibiotic use<sup>12</sup>. Despite these findings, corollary studies in the respiratory microbiome have not yet been performed, a key gap given the global magnitude of drug resistant LRTI. Furthermore, few studies have used metatranscriptomic RNA sequencing (RNA- seq) to both profile lower respiratory microbial ecology and detect ARG expression in the airway microbiome<sup>3,13</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[112, 790, 877, 905]]<|/det|>
+Here, we sought to test the hypothesis that older age is associated with an increased prevalence of ARGs in the lung microbiome, using metatranscriptomics and multivariable logistic regression modeling. We find that age is indeed an independent risk factor for detecting ARGs in the lower airway microbiome, even after adjusting for multiple covariates including sex,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 820, 140]]<|/det|>
+race/ethnicity, LRTI diagnosis, community- versus hospital- acquired infection, days from intubation to specimen collection, and composition of the lung microbiome.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 185, 192, 202]]<|/det|>
+## Methods  
+
+<|ref|>text<|/ref|><|det|>[[115, 216, 393, 234]]<|/det|>
+Study Design and Clinical Cohorts  
+
+<|ref|>text<|/ref|><|det|>[[112, 247, 883, 492]]<|/det|>
+We leveraged data from prospective pediatric<sup>14- 16</sup> and adult<sup>17</sup> cohorts of patients with acute respiratory failure admitted to intensive care units (ICUs) in the United States (USA). Pediatric patients (n=261), aged 31 days to 18 years, were enrolled from eight tertiary care hospitals in the Collaborative Pediatric Critical Care Research Network (CPCCRN) between February 2015 and December 2017. Adults (n=88), aged >18 years, were enrolled from a single tertiary care center in California, USA between July 2013 to October 2017. From each enrolled patient, tracheal aspirates (TA) were collected within 72 hours of intubation, mixed with DNA/RNA shield, and stored at - 80°C.  
+
+<|ref|>text<|/ref|><|det|>[[112, 505, 881, 749]]<|/det|>
+Electronic medical records were reviewed to obtain demographics and clinical data. LRTI status was retrospectively adjudicated by study physicians based on a previously described algorithm<sup>16,17</sup>, grouping patients as follows: 1) LRTI defined clinically, with or without a clinical microbiological diagnosis (LRTI); 2) No evidence of respiratory infection and a clear alternative etiology for the acute respiratory failure (No LRTI); or 3) patients who did not meet either above criteria (Indeterminate). LRTI was further separated into community- acquired LRTI (CA- LRTI; LRTI diagnosed within 48 hours of hospital admission), and hospital- acquired LRTI (HA- LRTI; LRTI diagnosed ≥48 hours after hospital admission).  
+
+<|ref|>text<|/ref|><|det|>[[112, 792, 848, 909]]<|/det|>
+Metatranscriptomic RNA Sequencing, Taxonomic Alignment, and Detection of ARGs RNA extracted from TA specimens underwent library preparation and paired- end Illumina sequencing, as previously described<sup>16</sup>. Quantification of microbial taxa from raw sequencing reads was carried out using the CZ- ID bioinformatics pipeline<sup>18</sup>, which performs
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 864, 330]]<|/det|>
+reference- based alignment against microbial genomes from the National Center for Biotechnology Information (NCBI) nucleotide (NT) database, as previously described<sup>18</sup>. ARGs annotated in the Antibiotic Resistance Gene- ANNOTation (ARG- ANNOT) database<sup>19</sup> were detected using the Short Read Sequence Typing (SRST2) algorithm<sup>20</sup>. Negative control water samples were processed in parallel, and a previously described negative binomial model was used to filter out microbial contaminants from the laboratory environment<sup>16</sup>. ARGs with \(< 5\%\) coverage or found in \(\geq 10\%\) of negative control water samples (TEM- 1D, TetC, Sull, OXA- 22, Aph3'la, CatA1) were excluded from the analysis.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 376, 364, 394]]<|/det|>
+## Statistical Analysis Framework  
+
+<|ref|>text<|/ref|><|det|>[[111, 407, 872, 555]]<|/det|>
+Age was defined in three ways: (1) a binary variable of children (31 days to 18 years) or adults (over 18 years); (2) nine subgroups of 0- 2 years, 3- 10 years, 11- 18 years, 19- 39 years, 40- 49 years, 50- 59 years, 60- 69 years, 70- 79 years, and \(\geq 80\) years; or (3) continuous age in years. We used Pearson's Chi- square test for comparison of categorical variables. P- values \(< 0.05\) were considered statistically significant. All analyses were conducted in R (v4.2.1).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 600, 282, 617]]<|/det|>
+## Resitome Analyses  
+
+<|ref|>text<|/ref|><|det|>[[111, 630, 883, 842]]<|/det|>
+The number of ARGs detectably expressed in the lower respiratory tract microbiome of children and adults were compared at the individual gene and ARG class (e.g., beta- lactamase) levels. P- values were calculated using the Wilcoxon rank- sum test for nonparametric continuous variables and false discovery rate (FDR) correction was applied for multiple comparisons. We compared the proportion of detected ARG classes by binary age (pediatric versus adult) and by age subgroups. 95% confidence intervals [CI] for population proportions were obtained using the Clopper- Pearson exact binomial method.  
+
+<|ref|>text<|/ref|><|det|>[[111, 855, 870, 905]]<|/det|>
+ARG abundance was calculated based on the average sequencing read depth across each gene, normalized by gene length and total reads, reported as depth per million (dpm)<sup>20,21</sup>.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 87, 864, 235]]<|/det|>
+Resistome alpha diversity was calculated using the Shannon Diversity Index (SDI) and ARG dpm. Beta diversity was calculated on patients with ARGs detected using the Bray- Curtis method with 1000 permutations using the PERMANOVA test and displayed via nonmetric multidimensional scaling (NMDS). Alpha and beta diversity calculations were performed using the R package vegan<sup>22</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[111, 247, 876, 587]]<|/det|>
+A multivariable logistic regression model incorporating demographic and clinical characteristics (sex, race/ethnicity, LRTI status, days from intubation to specimen collection) was used to determine associations between binary age (adults vs children) and detection of ARGs Additional regression models were performed using: 1) age years as a continuous variable, and 2) the previously defined nine age subgroups. To assess for potential geographic differences in ARGs, an additional analysis was performed within the pediatric cohort only and included adjustment for U.S. census region and presence of complex chronic conditions; the latter was defined by a previously validated pediatric medical complexity algorithm<sup>23</sup>. A sensitivity analysis limited to pediatric and adult patients from the same U.S. census region was also performed. 95% confidence intervals (CI) for the multivariable logistic regression models were calculated using the Wald CI.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 632, 290, 650]]<|/det|>
+## Microbiome Analyses  
+
+<|ref|>text<|/ref|><|det|>[[111, 662, 878, 905]]<|/det|>
+We assessed the respiratory tract microbiome of children and adults to evaluate age- related differences in taxonomic composition and diversity, which we considered as possible confounders or mediators of the relationship between age and detectably expressed ARGs. We assessed microbiota at the genus level, calculated total bacterial abundance (measured in reads per million, rpm), and calculated bacterial alpha diversity across age subgroups. We further stratified by LRTI status (CA- LRTI, HA- LRTI, No LRTI). Lung microbiome beta diversity calculations were carried out using the Bray- Curtis dissimilarity index and PERMANOVA to assess statistical significance. Differential abundance analysis was performed using the R
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 857, 172]]<|/det|>
+package DESeq \(2^{24}\) by assessing bacterial genera in the lung microbiome present in \(\geq 20\%\) of patients. We also described the prevalence of the most abundant species within each differentially expressed genus.  
+
+<|ref|>text<|/ref|><|det|>[[112, 217, 625, 235]]<|/det|>
+Associations Between the Microbiome and Resistome Analyses  
+
+<|ref|>text<|/ref|><|det|>[[111, 248, 880, 523]]<|/det|>
+To test whether age- related differences in the lung microbiome might influence ARG results, we carried out additional analyses adjusting for bacterial abundance and alpha diversity. To test whether specific taxa might influence age- related changes in ARG detection, we performed a differential abundance analysis of bacterial genera detected in patients with or without detectable expression of ARGs, using DESeq \(2^{24}\) . Subsequently, for each differentially abundant genus, we fit individual regression models for the outcome of having ARGs detected, accounting for bacterial abundance, alpha diversity, LRTI status, and presence of one of the differentially abundant genera. Lastly, additional sensitivity analyses were performed for these models using age as a continuous variable.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 569, 168, 585]]<|/det|>
+## Ethics  
+
+<|ref|>text<|/ref|><|det|>[[111, 599, 881, 842]]<|/det|>
+The pediatric cohort study was approved by a single Institutional Review Board (IRB) at the University of Utah (protocol #00088656). The adult cohort study was approved by the UCSF IRB (protocol #10- 02701). Informed consent was obtained from parents or other legal guardians (pediatric patients) and from patients or their surrogates (adult patients), which included permission for collected respiratory specimens and data to be used in future studies. For the adult cohort, the IRB approved of an initial waiver consent for obtaining excess respiratory samples, and informed consent was subsequently obtained for continued study participation according to CHR protocol 10- 02701 and as previously described \(^{25}\) .
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 90, 183, 107]]<|/det|>
+## Results  
+
+<|ref|>text<|/ref|><|det|>[[113, 120, 245, 138]]<|/det|>
+Patient Cohorts  
+
+<|ref|>text<|/ref|><|det|>[[111, 150, 883, 492]]<|/det|>
+We studied 261 children (median age: 1 year, interquartile range (IQR): 0- 15 years, range: 0- 17 years), and 88 adults (median age: 63 years, IQR: 54- 72 years, range: 21- 94 years) (Supplemental Table 1). Of the 349 patients, 231 (66%) were adjudicated as LRTI- positive, 67 (19%) had no evidence of LRTI, and 51 (15%) of patients had indeterminate LRTI status. The proportion of patients in each LRTI adjudication group did not differ between the two cohorts. Adults had a higher proportion of HA- LRTI than children (25% vs 6%, respectively), emphasizing the need to include this as a covariate in our subsequent logistic regression model. In both cohorts, 90% of the patients received antibiotics prior to tracheal aspirate collection. All four U.S. census regions (Midwest, Northeast, South, West) in the U.S. were represented among the 261 pediatric patients; adult patients were from one enrollment site located in the regional West.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 536, 400, 555]]<|/det|>
+## Lower Respiratory Tract Resistence  
+
+<|ref|>text<|/ref|><|det|>[[111, 567, 884, 778]]<|/det|>
+ARGs were detectably expressed in the lower respiratory tract microbiome of 40 (45%) adults compared with 53 (20%) children (Pearson's Chi- square \(\mathsf{p}< 0.01\) ). Across all patients, 74 distinct ARGs representing nine ARG classes were detected (Figure 1A). The number of detectably expressed ARGs (Figure 1B) and the number of ARG classes (Supplemental Figure 1) significantly differed between the youngest age subgroups (0- 2 years and 3- 10 years) and the oldest age subgroups (60- 69 years, 70- 79 years, and \(\geq 80\) years age groups), respectively. A significant increase was also noted between the 3- 10 and the 11- 18 years of age subgroups.  
+
+<|ref|>text<|/ref|><|det|>[[111, 790, 868, 875]]<|/det|>
+The most frequently detected ARG classes across all patients conferred resistance to beta- lactams (n=85 patients), macrolides (n=41), and aminoglycosides (n=37). A greater proportion of adults compared with children had beta- lactam ARGs (31% (CI: 21- 41%) vs 13%
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[140, 80, 870, 690]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[131, 700, 901, 890]]<|/det|>
+<center>Figure 1. (A) Frequency of children (translucent) and adults (solid) with each antimicrobial resistance gene (ARG), stratified by ARG class. (B) Number of ARGs detected in children and adults by age subgroups. Two outliers were omitted for visualization purposes; one 11-18 year-old patient with 18 ARGs detected and another 70-79 year-old patient with 12 ARGs detected. (C) Number of ARG classes detected in children and adults by age subgroups. For Figures B and C, p-values were calculated using Wilcoxon-rank sum test and adjusted for multiple comparisons with False Discovery Rate (FDR) correction. The asterisks indicate statistically significant comparisons; all had a p-value <0.01. (D) Proportion of patients with ARGs by ARG class, stratified by pediatric and adult cohorts. The 95% confidence intervals were calculated by the Clopper-Pearson exact binomial method. P-values were obtained by Pearson's Chi-square test and Fisher's exact test for samples with <5 total ARGs. (E) Beta diversity of antimicrobial resistance children and adults. P-value calculated based on the Bray-Curtis dissimilarity index and the PERMANOVA test with 1000 permutations. Abbreviation: TMP-SMX, trimethoprim-sulfamethoxazole; NMDS, nonmetric multidimensional scaling. </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 87, 877, 325]]<|/det|>
+(CI: 10- 18%)), aminoglycoside ARGs (20% (CI: 13- 30%) vs 2% (CI: 0.6- 4%)), and tetracycline ARGs (14% (CI: 7- 23%) vs 3% (CI: 1- 5%)) (Figure 1C, Supplemental Table 2). When evaluated by age subgroup, the proportion of patients with beta-lactam, macrolide, or tetracycline ARGs was highest in patients \(\geq 70\) years of age (Supplemental Figure 2). Among young children, 13% (95% CI: 8- 19%) of patients aged 0- 2 years had a beta-lactam ARG compared with 2% (95% CI: 0.05- 10%) of patients aged 3- 10 years; this pattern was not seen for the other ARG classes. Among the beta-lactam ARGs, we detected six AmpC beta-lactamase genes, five extended- spectrum beta-lactamase genes, and 2 carbapenemase genes (Supplemental Figure 3).  
+
+<|ref|>text<|/ref|><|det|>[[111, 343, 875, 460]]<|/det|>
+ARG alpha diversity as measured by the Shannon Diversity Index increased primarily in patients \(\geq 60\) years of age (Figure 1D). The composition of the lung resistome significantly differed between children and adults, as measured by the Bray Curtis dissimilarity index (p = 0.003 by PERMANOVA) (Figure 1E).  
+
+<|ref|>text<|/ref|><|det|>[[111, 470, 875, 725]]<|/det|>
+In a logistic regression model assessing the association of binary age group with detection of any ARGs, accounting for sex, race/ethnicity, LRTI status (CA- LRTI, HA- LRTI, No LRTI), and days from intubation to specimen collection, the risk of ARG detection was increased in adults compared with children (adjusted odds ratio [aOR]: 2.16, 95% CI: 1.10- 4.22) (Figure 2A). Age remained significant in a sensitivity analysis of the same logistic regression model using age as a continuous variable (Supplemental Table 3). In a second sensitivity analysis using a regression model based on age subgroups (Figure 2B), children aged 3- 10 years had a lower risk (aOR: 0.32, 95% CI: 0.10- 0.97) and adults \(\geq 80\) years of age had a higher risk of detectably expressed ARGs (aOR: 6.21, 95% CI: 1.33- 28.99) compared with children aged 0- 2 years.  
+
+<|ref|>text<|/ref|><|det|>[[111, 760, 852, 844]]<|/det|>
+In an analysis restricted exclusively to children and accounting for enrollment U.S. census region and presence of a complex chronic condition, children 3- 10 years of age continued to have the lowest risk of having detectably expressed ARGs, however enrollment
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 90, 866, 110]]<|/det|>
+237 site was a significant risk factor (Supplemental Table 4). Given this, we performed a sensitivity  
+
+<|ref|>text<|/ref|><|det|>[[55, 122, 857, 140]]<|/det|>
+238 analysis limited to pediatric and adult patients from the same U.S. census region (West), and  
+
+<|ref|>image<|/ref|><|det|>[[130, 160, 833, 840]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[142, 852, 840, 897]]<|/det|>
+<center>Figure 2. Multivariable logistic regression model evaluating the association of (A) binary age and (B) age subgroups with the presence of ARGs, accounting for sex, race/ethnicity, and lower respiratory tract infection (LRTI) status. </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 857, 140]]<|/det|>
+found that binary age group remained associated with increased risk of ARG detection (uOR: 2.33, 95% CI: 1.2- 4.52).  
+
+<|ref|>text<|/ref|><|det|>[[112, 184, 410, 202]]<|/det|>
+Lower Respiratory Tract Microbiome  
+
+<|ref|>text<|/ref|><|det|>[[111, 216, 884, 492]]<|/det|>
+The total bacterial abundance of the lung microbiome increased with age (Figure 3A). Bacterial microbiome alpha diversity initially increased during childhood and adulthood, peaked in the 40- 49 year- old age group, and then decreased in older adults (Figure 3B). The community composition of the bacterial respiratory microbiome differed between children and adults, based on Bray Curtis dissimilarity index ( \(p < 0.01\) by PERMANOVA) (Figure 3C). Differential abundance analysis revealed eight bacterial genera with statistically significant differences in abundance between children and adults (Enterococcus, Pseudomonas, Staphylococcus, Bacteroides, Prevotella, Mannheimia, Haemophilus and Moraxella). The most abundant bacterial species within each of these genera also differed between age groups (Figure 3E).  
+
+<|ref|>text<|/ref|><|det|>[[112, 535, 736, 554]]<|/det|>
+Interactions Between the Lower Respiratory Tract Microbiome and Resistome  
+
+<|ref|>text<|/ref|><|det|>[[111, 567, 877, 842]]<|/det|>
+In a logistic regression model accounting for total bacterial abundance, bacterial alpha diversity, and LRTI status, binary age group remained associated with an increased risk of ARG detection (aOR: 2.38, 95% CI: 1.25- 4.54) (Figure 4A). Differential abundance analysis revealed seven bacterial genera with statistically significant differences between patients with or without detectable ARGs (Figure 4B). In individual fitted logistic regression models where the outcome was presence of ARGs, and independent variables included age group, total bacterial abundance, bacterial alpha diversity, LRTI status, and one of the seven differentially abundant bacterial genera, adults remained associated with an increased risk of having ARGs detected compared with children (Figure 4A, Supplemental Table 5).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[112, 75, 870, 700]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[130, 713, 878, 890]]<|/det|>
+<center>Figure 3. (A) Bacterial abundance in the lung microbiome measured in total bacterial alignments to the NCBI NT database per million reads sequenced (NT rpm) in children and adults by age subgroups. (B) Alpha diversity, calculated by the Shannon diversity index, of the bacterial lung microbiome of children and adults by age subgroups. (C) Beta diversity of the bacterial lung microbiome of children and adults. P-value calculated based on the Bray-Curtis dissimilarity index and the PERMANOVA test with 1000 permutations. (D) Statistically significant (p-value <0.05) differential abundant bacterial genera, by log2 fold change of bacterial counts, detected in children and adults. Bar colors indicate whether the species was more abundant in children (blue) or adults (red). (E) Frequency of the bacterial species detected in ≥5% of children (translucent) and adults (solid) among the differentially abundant bacterial genera. For patients with multiple species detected per genus, only the most abundant species was included in this analysis. Abbreviations: NT rpm, sequencing alignments to the NCBI NT database per per million reads sequenced; NMDS, nonmetric multidimensional scaling. </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 88, 866, 234]]<|/det|>
+264 In these models, Fusobacterium spp. (aOR: 2.54, 95% CI: 1.32- 4.90), Bacteroides spp. (aOR: 2.23, 95% CI: 1.15- 4.33), Enterococcus spp. (aOR: 2.17, 95% CI: 1.04- 4.51), Staphylococcus spp. (aOR: 2.09, 95% CI: 1.09- 4.02), and Prevotella spp. (aOR: 1.98, 95% CI: 1.05- 3.72) were statistically significant risk factors. Sensitivity analyses of all these models using age as a continuous variable demonstrated similar results (Supplemental Table 6).  
+
+<|ref|>image<|/ref|><|det|>[[128, 266, 875, 784]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 796, 884, 872]]<|/det|>
+<center>Figure 4. (A) Multivariable logistic regression model evaluating the association of binary age with the presence of ARGs, accounting for total bacterial abundance (NT rpm) per patient sample, bacterial alpha diversity. (B) Statistically significant \((p< 0.05)\) differentially abundant bacterial genera, by \(\log 2\) fold change of bacterial counts, detected in patients with ARGs compared with patients without ARGs. All detected bacterial genera were more prevalent in patients with ARGs compared with patients without ARGs. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 90, 214, 107]]<|/det|>
+## Discussion  
+
+<|ref|>text<|/ref|><|det|>[[112, 120, 884, 330]]<|/det|>
+Utilizing metatranscriptomics, we identify age as an independent risk factor for ARG detection in the lung microbiome among critically ill, recently intubated patients. We find that detection of ARGs in the lower respiratory tract increases across most of the age spectrum, with the oldest patients harboring the highest number of ARGs detectably expressed at the individual gene and class levels. These findings advance our understanding of the lung microbiome as a potential antimicrobial resistance reservoir and highlight its potential contribution to drug- resistant respiratory infections.  
+
+<|ref|>text<|/ref|><|det|>[[112, 343, 874, 650]]<|/det|>
+Across every ARG class examined, adults had a greater number of ARGs detected compared to children. The majority of detected ARGs in the pediatric cohort conferred resistance to beta- lactams and macrolides, while in adults, beta- lactam and aminoglycoside ARGs were most prevalent. Studies of nasal samples from healthy neonates<sup>26</sup> and neonates with cystic fibrosis<sup>27</sup> found a similar beta- lactam ARG dominance, while oral flora samples from children<sup>28</sup> and sputum samples from adults<sup>29</sup> found that macrolide resistance genes were most prevalent in the oropharynx. Intriguingly, we found that children 0- 2 years of age had a higher proportion of beta- lactam ARGs detected than those 3- 10 years of age. This may reflect maternally- derived microbial communities and associated ARGs acquired at birth, which go on to comprise the lung microbiome during the first years of life<sup>30</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[112, 662, 882, 906]]<|/det|>
+Our findings differ from gut resistome studies which have found that tetracycline ARGs are most prevalent, followed by macrolides and beta- lactam ARGs<sup>31</sup>. These differences may reflect differences in the routes of ARG and antibiotic exposure (e.g., inhaled, ingested, intravenous) and highlight potential AMR transmission differences from the lungs compared with the gut. Indeed, recent work demonstrates the presence of ARGs and microbiota in urban air samples<sup>32,33</sup>, suggesting that environmental exposures may be a relevant route of lung resistome exchange particularly for patients with prolonged or repeated exposure to healthcare facilities where resistant microbes are more prevalent.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 884, 350]]<|/det|>
+In the U.S., the most prescribed outpatient antibiotics are beta- lactams, fluoroquinolones, and macrolides<sup>34</sup>, while beta- lactams, macrolides, and glycopeptides are the most frequently prescribed inpatient antibiotics<sup>35</sup>. While beta- lactam ARGs were the most prevalent ARG class detected in both children and adults, adults had a greater proportion of aminoglycoside and tetracycline ARGs. Given that these antimicrobial classes have been widely used in the agriculture and livestock industries<sup>36</sup>, exposures to environmental bacteria harboring these ARGs over the lifespan could be one possible explanation. Other possible explanations include community exposures to other individuals, co- selection of ARGs on mobile genetic elements carrying multiple ARGs, or cross- resistance due to multi- drug efflux pumps<sup>37</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[111, 377, 880, 618]]<|/det|>
+We also observed differences in the lower respiratory tract microbiome with age, including bacterial abundance, diversity, and taxonomic composition. Our findings are in line with a prior study demonstrating that bacterial abundance in the lung microbiome of CF patients increases with age<sup>38</sup>. We also found that bacterial alpha diversity increased in childhood, peaked in middle age, and decreased with older age. The role of endogenous respiratory microbiota in both the pathophysiology and diagnosis of critical illness syndromes in increasingly recognized, and our results suggest that adjusting for age should be considered in clinical and translational studies of the lung microbiome.  
+
+<|ref|>text<|/ref|><|det|>[[111, 632, 876, 875]]<|/det|>
+Even when accounting for demographic, clinical, and microbiome differences, age remained an independent risk factor for ARG detection. Our findings raise the possibility that selective environmental pressures driving AMR acquisition from the environment may be continuous over the lifespan, shaping the airway microbiome and associated resistome. While the detection of ARG expression in the airway microbiome does not equate to clinically relevant resistance, it suggests the potential for development of phenotypic resistance<sup>3</sup> with possible implications for patient care. Commensal bacteria within the lung microbiome can exchange ARGs via horizontal gene transfer to pathogens or pathobionts, leading to the emergence of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 864, 140]]<|/det|>
+drug resistant LRTI<sup>39- 41</sup>. Further research is needed to better characterize and understand the prevalence, acquisition, and transmission dynamics of ARGs within the lung microbiome.  
+
+<|ref|>text<|/ref|><|det|>[[111, 152, 875, 523]]<|/det|>
+This study has limitations. First, there was not an even distribution of patients across all ages, with a greater number of young children and older adults, reflecting the distribution of the critically ill, mechanically ventilated patient populations. To account for this, we performed sensitivity analyses using age as a continuous variable or using age subgroups. Second, there were differences in geographic location and timing of tracheal aspirate collection between age groups, which we accounted for by incorporating the variables into the multivariable logistic regression models. Third, the study included only patients from the U.S. and may not be representative of the global population. Fourth, detection of ARGs is biased towards the most abundant taxa in the lung microbiome and we are likely missing detection of ARGs from less abundant taxa. Finally, enrollment occurred prior to the COVID- 19 pandemic. Thus, ARG abundance and class profiles may be different than in the current population given the increase in antibiotic use and AMR infections since \(2020^{42,43}\) .  
+
+<|ref|>text<|/ref|><|det|>[[112, 535, 870, 745]]<|/det|>
+In summary, we demonstrate that age is independently associated with the detection of ARGs in the lung microbiome in a population of critically ill patients soon after intubation. Our results suggest that healthcare, community, and environmental exposures throughout life may contribute to the reservoir of ARGs in the respiratory tract. Taken together, these findings advance our understanding of AMR in the context of the human microbiome and have implications for the management of infectious diseases, antimicrobial stewardship programs, and public health policies.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 90, 344, 108]]<|/det|>
+## Data and Code Availability  
+
+<|ref|>text<|/ref|><|det|>[[112, 120, 867, 238]]<|/det|>
+FASTQ files containing non- host reads identified by the CZ- ID pipeline, following subtraction of reads aligning to the human genome, are available in the NCBI Sequence Read Archive (SRA) database under BioProject accessions PRJNA875913 and PRJNA450137. All data, code, and results are available at: https://github.com/victoriatchu/agingAMR.  
+
+<|ref|>sub_title<|/ref|><|det|>[[113, 281, 286, 299]]<|/det|>
+## Acknowledgements  
+
+<|ref|>text<|/ref|><|det|>[[112, 310, 881, 682]]<|/det|>
+This work was supported in part by the following grants from NHLBI: R01HL155418 (CRL, PMM), R01HL124103 (PMM), K23HL138461 (CRL), and R53HL140026 (CSC), as well as grants from the Chan Zuckerberg Biohub (VC, CRL, JLD). The investigators thank all patients and their families for participating in this project. We also would like to acknowledge the contributions from principal investigators, co- investigators, research coordinators, and allied research personnel at the following sites: University of California San Francisco, San Francisco, CA; Children's Hospital of Colorado, Aurora, CO; Chan Zuckerberg Biohub; Children's Hospital of Michigan, Detroit, MI; Children's Hospital of Philadelphia, Philadelphia, PA; Children's National Medical Center, Washington, DC; Nationwide Children's Hospital, Columbus, OH; Mattel Children's Hospital, University of California Los Angeles, Los Angeles, CA; Children's Hospital of Pittsburgh, University of Pittsburgh Medical Center, Pittsburgh, PA; University of Utah; and Data Coordinating Center, Salt Lake City, Utah.  
+
+<|ref|>text<|/ref|><|det|>[[112, 726, 830, 777]]<|/det|>
+Funding: NHLBI R01HL155418, NHLBI K23HL138461, Chan Zuckerberg Biohub, NHLBI R53HL140026 (CSC)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[50, 90, 880, 893]]<|/det|>
+365 References:366 1 EclinicalMedicine. Antimicrobial resistance: a top ten global public health threat. 367 eClinicalMedicine 41 (2021). https://doi.org:10.1016/j.eclinm.2021.101221368 2 Murray, C. J. L. et al. Global burden of bacterial antimicrobial resistance in 2019: a 369 systematic analysis. The Lancet 399, 629- 655 (2022). https://doi.org:10.1016/S0140- 3706(21)02724- 0371 3 Serpa, P. H. et al. Metagenomic prediction of antimicrobial resistance in critically ill 372 patients with lower respiratory tract infections. Genome Medicine 14, 74 (2022). https://doi.org:10.1186/s13073- 022- 01072- 4374 4 Charlson, E. S. et al. Topographical continuity of bacterial populations in the healthy 375 human respiratory tract. American journal of respiratory and critical care medicine 184, 957- 963 (2011).377 5 Muhlebach, M. S. et al. Initial acquisition and succession of the cystic fibrosis lung 378 microbiome is associated with disease progression in infants and preschool children. 379 PLOS Pathogens 14, e1006798 (2018). https://doi.org:10.1371/journal.ppat.1006798 380 6 Tracy, M., Cogen, J. & Hoffman, L. R. The pediatric microbiome and the lung. Curr 381 Opin Pediatr 27, 348- 355 (2015). https://doi.org:10.1097/mop.0000000000000212 382 7 Pereira- Dias, J. et al. The Gut Microbiome of Healthy Vietnamese Adults and Children Is 383 a Major Reservoir for Resistance Genes Against Critical Antimicrobials. J Infect Dis 224, 5840- s847 (2021). https://doi.org:10.1093/infdis/jiab398 384 8 Lu, N. et al. DNA microarray analysis reveals that antibiotic resistance- gene diversity in 385 human gut microbiota is age related. Sci Rep 4, 4302 (2014). https://doi.org:10.1038/srep04302 386 9 Wu, L. et al. Metagenomics- Based Analysis of the Age- Related Cumulative Effect of 387 Antibiotic Resistance Genes in Gut Microbiota. Antibiotics (Basel) 10 (2021). 388 https://doi.org:10.3390/antibiotics1008106390 1 10 Boolchandani, M. et al. Impact of international travel and diarrhea on gut microbiome 392 and resistome dynamics. Nature Communications 13, 7485 (2022). 393 https://doi.org:10.1038/s41467- 022- 34862- w394 11 Kohler, P. P. et al. Emergence of Carbapenemase- Producing Enterobacteriaceae, South- 395 Central Ontario, Canada(1). Emerg Infect Dis 24, 1674- 1682 (2018). 396 https://doi.org:10.3201/eid2409.180164397 12 Reyman, M. et al. Effects of early- life antibiotics on the developing infant gut 398 microbiome and resistome: a randomized trial. Nat Commun 13, 893 (2022). 399 https://doi.org:10.1038/s41467- 022- 28525- z400 13 Spottiswoode, N. et al. Pneumonia surveillance with culture- independent 401 metatranscriptomics in HIV- positive adults in Uganda: a cross- sectional study. The 402 Lancet Microbe 3, e357- e365 (2022). https://doi.org:10.1016/S2666- 5247(21)00357- 8403 14 Tsitsiklis, A. et al. Lower respiratory tract infections in children requiring mechanical 404 ventilation: a multicentre prospective surveillance study incorporating airway 405 metagenomics. The Lancet Microbe 3, e284- e293 (2022). https://doi.org:10.1016/S2666- 5247(21)00304- 9407 15 Mourani, P. M. et al. Temporal airway microbiome changes related to ventilator- 408 associated pneumonia in children. Eur Respir J 57 (2021). 409 https://doi.org:10.1183/13993003.01829- 2020
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 90, 884, 895]]<|/det|>
+410 16 Mick, E. et al. Integrated host/microbe metagenomics enables accurate lower respiratory tract infection diagnosis in critically ill children. J Clin Invest 133 (2023). 411 https://doi.org:10.1172/jci165904 412 Langelier, C. et al. Integrating host response and unbiased microbe detection for lower respiratory tract infection diagnosis in critically ill adults. Proc Natl Acad Sci U S A 115, E12353-e12362 (2018). https://doi.org:10.1073/pnas.1809700115 413 18 Kalantar, K. L. et al. IDseq—An open source cloud-based pipeline and analysis service for metagenomic pathogen detection and monitoring. GigaScience 9 (2020). https://doi.org:10.1093/gigascience/giaa111 419 Gupta, S. K. et al. ARG- ANNOT, a new bioinformatic tool to discover antibiotic resistance genes in bacterial genomes. Antimicrob Agents Chemother 58, 212- 220 (2014). https://doi.org:10.1128/aac.01310- 13 422 20 Inouye, M. et al. SRST2: Rapid genomic surveillance for public health and hospital microbiology labs. Genome Medicine 6, 90 (2014). https://doi.org:10.1186/s13073- 014- 0090- 6 425 21 Tong, X. et al. Alterations to the Lung Microbiome in Idiopathic Pulmonary Fibrosis Patients. Front Cell Infect Microbiol 9, 149 (2019). https://doi.org:10.3389/fcimb.2019.00149 428 22 vegan: Community Ecology Package_v. R package version 2.6- 4 (2022). 429 23 Simon, T. D. et al. Pediatric Medical Complexity Algorithm: A New Method to Stratify Children by Medical Complexity. Pediatrics 133, e1647- e1654 (2014). https://doi.org:10.1542/peds.2013- 3875 432 24 Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA- seq data with DESeq2. Genome biology 15, 1- 21 (2014). 433 Sarma, A. et al. Tracheal aspirate RNA sequencing identifies distinct immunological features of COVID- 19 ARDS. Nat Commun 12, 5152 (2021). https://doi.org:10.1038/s41467- 021- 25040- 5 436 26 Manenzhe, R. I. et al. Longitudinal changes in the nasopharyngeal resistome of South African infants using shotgun metagenomic sequencing. PLoS One 15, e0231887 (2020). https://doi.org:10.1371/journal.pone.0231887 440 27 Allemann, A. et al. Nasal resistome development in infants with cystic fibrosis in the first year of life. Frontiers in microbiology 10, 212 (2019). 441 28 Sukumar, S. et al. Development of the oral resistome during the first decade of life. Nat Commun 14, 1291 (2023). https://doi.org:10.1038/s41467- 023- 36781- w 444 29 Aogain, M. M. et al. Metagenomics Reveals a Core Macrolide Resistome Related to Microbiota in Chronic Respiratory Disease. American Journal of Respiratory and Critical Care Medicine 202, 433- 447 (2020). https://doi.org:10.1164/rccm.201911- 22020C 448 30 Yagi, K., Asai, N., Huffnagle, G. B., Lukacs, N. W. & Fonseca, W. Early- Life Lung and Gut Microbiota Development and Respiratory Syncytial Virus Infection. Front Immunol 13, 877771 (2022). https://doi.org:10.3389/fimmu.2022.877771 451 31 Hu, Y. et al. Metagenome- wide analysis of antibiotic resistance genes in a large cohort of human gut microbiota. Nat Commun 4, 2151 (2013). https://doi.org:10.1038/ncomms3151 454 32 Li, J. et al. Global Survey of Antibiotic Resistance Genes in Air. Environmental Science & Technology 52, 10975- 10984 (2018). https://doi.org:10.1021/acs.est.8b02204
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 88, 875, 696]]<|/det|>
+456 33 Zhu, G. et al. Air pollution could drive global dissemination of antibiotic resistance genes. Isme j 15, 270- 281 (2021). https://doi.org:10.1038/s41396- 020- 00780- 2
+457 34 Durkin, M. J. et al. Outpatient Antibiotic Prescription Trends in the United States: A National Cohort Study. Infect Control Hosp Epidemiol 39, 584- 589 (2018).
+458 35 Baggs, J., Fridkin, S. K., Pollack, L. A., Srinivasan, A. & Jernigan, J. A. Estimating National Trends in Inpatient Antibiotic Use Among US Hospitals From 2006 to 2012.
+459 36 JAMA Intern Med 176, 1639- 1648 (2016).
+460 37 https://doi.org:10.1001/jamainternmed.2016.5651
+461 38 Mulchandani, R., Wang, Y., Gilbert, M. & Van Boeckel, T. P. Global trends in antimicrobial use in food- producing animals: 2020 to 2030. PLOS Global Public Health 3, e0001305 (2023). https://doi.org:10.1371/journal.pgph.0001305
+462 37 Partridge, S. R., Kwong, S. M., Firth, N. & Jensen, S. O. Mobile Genetic Elements Associated with Antimicrobial Resistance. Clinical Microbiology Reviews 31, e00088- 00017 (2018). https://doi.org:10.1128/CMR.00088- 17
+463 38 Coburn, B. et al. Lung microbiota across age and disease stage in cystic fibrosis. Sci Rep 5, 10241 (2015). https://doi.org:10.1038/srep10241
+464 39 Vinayamohan, P. G., Pellissery, A. J. & Venkitanarayanan, K. Role of horizontal gene transfer in the dissemination of antimicrobial resistance in food animal production. Current Opinion in Food Science 47, 100882 (2022).
+465 40 Launay, A., Ballard, S. A., Johnson, P. D. R., Grayson, M. L. & Lambert, T. Transfer of Vancomycin Resistance Transposon Tn1549 from Clostridium symbiosum to Enterococcus spp. in the Gut of Gnotobiotic Mice. Antimicrobial Agents and Chemotherapy 50, 1054- 1062 (2006). https://doi.org:doi:10.1128/AAC.50.3.1054- 1062.2006
+466 41 Bengtsson- Palme, J. et al. The Human Gut Microbiome as a Transporter of Antibiotic Resistance Genes between Continents. Antimicrob Agents Chemother 59, 6551- 6560 (2015). https://doi.org:10.1128/aac.00933- 15
+467 42 Nandi, A., Pecetta, S. & Bloom, D. E. Global antibiotic use during the COVID- 19 pandemic: analysis of pharmaceutical sales data from 71 countries, 2020&#36;x2013;2022. eClinicalMedicine 57 (2023). https://doi.org:10.1016/j.eclinm.2023.101848
+468 43 Malik, S. S. & Mundra, S. Increasing Consumption of Antibiotics during the COVID- 19 Pandemic: Implications for Patient Health and Emerging Anti- Microbial Resistance. Antibiotics (Basel) 12 (2022). https://doi.org:10.3390/antibiotics12010045
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 311, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[61, 130, 279, 150]]<|/det|>
+Supplementalfinal.pdf
+
+<--- Page Split --->

preprint/preprint__4841b1fcf83ab880bb3afb296fa2b3ae0c5a97728da9b5ceacdd382aa94e3ebd/images_list.json

@@ -0,0 +1,258 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        135,
+        840,
+        707
+      ]
+    ],
+    "page_idx": 37
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. PDXs expressing high p16 are resistant to ribociclib. A) Summary of genetic alterations in the PDX panel from Figure 1B, including the PDX subtype classification, based on IHC (Molecular subtype) or PAM50 analysis (intrinsic subtype), and the response to CDK4/6 inhibitors. Genes with similar function such as TSC1/TSC2 or CDKN2A/CDKN2B were considered as one single feature. B) Quantification of IHC staining for p16, pRb, cyclin E1 and cyclin D1 in 23-untreated PDX in relationship with ribociclib-response. Semiquantitative analysis was performed for pRb and p16, or the Allred scoring method for cyclin E1 and cyclin D1 in",
+    "footnote": [],
+    "bbox": [
+      [
+        147,
+        100,
+        860,
+        714
+      ]
+    ],
+    "page_idx": 39
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        103,
+        808,
+        690
+      ]
+    ],
+    "page_idx": 41
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        100,
+        840,
+        730
+      ]
+    ],
+    "page_idx": 43
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5",
+    "footnote": [],
+    "bbox": [
+      [
+        160,
+        110,
+        810,
+        720
+      ]
+    ],
+    "page_idx": 45
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Figure 6",
+    "footnote": [],
+    "bbox": [
+      [
+        155,
+        103,
+        810,
+        300
+      ]
+    ],
+    "page_idx": 47
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_0.jpg",
+    "caption": "Sensitive",
+    "footnote": [],
+    "bbox": [
+      [
+        160,
+        300,
+        808,
+        360
+      ]
+    ],
+    "page_idx": 47
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_1.jpg",
+    "caption": "B",
+    "footnote": [],
+    "bbox": [
+      [
+        160,
+        400,
+        380,
+        550
+      ]
+    ],
+    "page_idx": 47
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Figure 6. Acquisition of subclonal RB1 mutations in tumors with underlying RB1 heterozygous loss as mechanism of CDK4/6i acquired resistance in BC PDXs. A) On the",
+    "footnote": [],
+    "bbox": [
+      [
+        160,
+        616,
+        850,
+        715
+      ]
+    ],
+    "page_idx": 47
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Figure 7. PI3K inhibition sensitizes non-basal like BC PDX to CDK4/6i. A) Waterfall plot showing the growth of 23 PDX treated with ribociclib 75 mg/kg plus alpelisib 35mg/kg (bars) and vehicle (circles). The percentage change from the initial volume is shown at day 35 of treatment. Dashed lines indicate the range of PD (>20%), SD (20% to -30%) and PR/CR (< -30%). The molecular subtypes are indicated. Hashtags indicate models harboring mutations in PI3K3CA. Data represent mean and error bars ± SEM. Boxes underneath show the molecular and intrinsic tumor's subtypes as well as their responses to the indicated treatments. The preclinical benefit",
+    "footnote": [],
+    "bbox": [
+      [
+        144,
+        90,
+        848,
+        720
+      ]
+    ],
+    "page_idx": 49
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 51
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2",
+    "footnote": [],
+    "bbox": [
+      [
+        50,
+        50,
+        800,
+        767
+      ]
+    ],
+    "page_idx": 51
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3",
+    "footnote": [],
+    "bbox": [
+      [
+        60,
+        49,
+        800,
+        768
+      ]
+    ],
+    "page_idx": 53
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4",
+    "footnote": [],
+    "bbox": [
+      [
+        56,
+        48,
+        803,
+        777
+      ]
+    ],
+    "page_idx": 55
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5 High p16 levels associated with lack of response to CDK4/6i in ER+ BC patients. A) Representation of the percentage distribution of Luminal A/B tumors vs. sensitivity to abemaciclib after 15 days of treatment in the neoadjuvant setting in the ABC- POP trial. Tumors were classified as Luminal A if \\(\\% \\mathrm{Ki67} < 15\\) or as Luminal B if \\(\\% \\mathrm{Ki67} \\geq 15\\) . Tumors showing In Ki67 < 1 at day 15 were considered sensitive and those with In Ki67 \\(\\geq 1\\) were resistant to the studied drug. B) Logistic model to evaluate the effect of p16 on the",
+    "footnote": [],
+    "bbox": [
+      [
+        57,
+        60,
+        770,
+        777
+      ]
+    ],
+    "page_idx": 57
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_2.jpg",
+    "caption": "B",
+    "footnote": [],
+    "bbox": [
+      [
+        55,
+        45,
+        770,
+        370
+      ]
+    ],
+    "page_idx": 59
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Figure 6",
+    "footnote": [],
+    "bbox": [
+      [
+        55,
+        395,
+        796,
+        655
+      ]
+    ],
+    "page_idx": 61
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Figure 7",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 61
+  }
+]

preprint/preprint__4844999f9c48fab95f677e12ff57f18204d1027644dc64a7ffea5ccbbcee4859/images_list.json

@@ -0,0 +1,115 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_0.jpg",
+    "caption": "D: This work: Povarov reaction/aromatization",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 2
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2 |Process of enantiomerization for 1-(3-indol)-quino[5]helicene, the relative Gibbs free energy (kcal/mol) was calculated at the M06-2X/def2-TZVPP//B3LYP/def2-SVP lever.",
+    "footnote": [],
+    "bbox": [
+      [
+        220,
+        52,
+        777,
+        241
+      ]
+    ],
+    "page_idx": 2
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_1.jpg",
+    "caption": "Table 1 |Optimization of the reaction conditionsa,d",
+    "footnote": [],
+    "bbox": [
+      [
+        170,
+        330,
+        808,
+        760
+      ]
+    ],
+    "page_idx": 2
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3 | Scope of aromatic aldehydes and indole derivatives. Reaction conditions: 1) 1a (0.1 mmol) and 2 (0.4 mmol) in toluene (1.5 mL) at \\(110^{\\circ}\\mathrm{C}\\) for 12 h, then added 3a (0.2 mmol) and (S)-A5 (0.005 mmol) at rt for 12 h, purified by silica gel column chromatography to get 4.2) 4 and 1,2-Dichloro-4,5-Dicyanobenzoquinone (DDQ, 0.3 mmol) and DCM (2 mL) at rt for 3 h.",
+    "footnote": [],
+    "bbox": [
+      [
+        87,
+        170,
+        910,
+        680
+      ]
+    ],
+    "page_idx": 2
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4 |Scope of various amine derivative. Reaction conditions: 1) 1 (0.1 mmol) and 2a (0.4 mmol) in toluene (1.5 mL) at \\(110^{\\circ}C\\) for \\(12h\\) , then added 3a (0.2 mmol) and (S)-A5 (0.005 mmol) at rt for \\(12h\\) , purified by silica gel column chromatography to get 4. 2) 4 and 1,2-Dichloro-4,5-Dicyanobenzoquinone (DDQ, 0.3 mmol) and DCM (2 mL) at rt for \\(3h\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        95,
+        45,
+        907,
+        440
+      ]
+    ],
+    "page_idx": 3
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig. 5 |Reaction conditions: 4Large-scale reaction was conducted under optimized reaction conditions. 1a (0.1 mmol), 2a (0.4 mmol), 3a and (S)-A5 (0.005 mmol) in solvent (1.5 mL) at rt for \\(24h\\) then added 1,2-dichloro-4,5-dicyanobenzoquinone (DDQ 0.5 mmol) for another \\(3h\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        199,
+        512,
+        796,
+        821
+      ]
+    ],
+    "page_idx": 3
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig. 6 |Barriers to enantiomerization for quinohelicenes. aExperimental value. bThe relative Gibbs free energy (kcal/mol) was calculated at the M06-2X/def2-TZVP/P/B3LYP/def2-SVP lever.",
+    "footnote": [],
+    "bbox": [
+      [
+        246,
+        305,
+        740,
+        440
+      ]
+    ],
+    "page_idx": 4
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_2.jpg",
+    "caption": "Measured at \\(1\\times 10^{-5}\\mathrm{M}\\) in DCM. Maximum UV absorption. Maximum fluorescence wavelength. Absolute fluorescence quantum efficiency under absorbance lower than 0.1. Measured with \\(99\\%\\) ee.",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        240,
+        850,
+        640
+      ]
+    ],
+    "page_idx": 5
+  }
+]

preprint/preprint__4844999f9c48fab95f677e12ff57f18204d1027644dc64a7ffea5ccbbcee4859/preprint__4844999f9c48fab95f677e12ff57f18204d1027644dc64a7ffea5ccbbcee4859.mmd

@@ -0,0 +1,222 @@
+
+# Enantioselective Synthesis of Quinohelicenes through a Sequential Organocatalyzed Povarov Reaction/Oxidative Aromatization  
+
+Chengwen Li Nankai University  
+
+Xi Gao Nankai University  
+
+Zhiyuan Ren Nankai University  
+
+Chenhao Guo Institute of Chemistry, CAS  
+
+Meng Li chinese academy of sciences Xin Li (xin_li@nankai.edu.cn) Nankai University https://orcid.org/0000- 0001- 6020- 9170  
+
+Article  
+
+Keywords:  
+
+Posted Date: January 4th, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 2327586/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on June 8th, 2023. See the published version at https://doi.org/10.1038/s41467- 023- 39134- 9.
+
+<--- Page Split --->
+
+# Enantioselective Synthesis of Quinohelicenes through a Sequen- tial Organocatalyzed Povarov Reaction/Oxidative Aromatization  
+
+Chengwen \(\mathsf{L}\mathsf{i}^{1}\) , Xi Gao \(^{1}\) , Zhiyuan \(\mathsf{Re n}^{1}\) , Chenhao Guo \(^{2}\) , Meng \(\mathsf{L}\mathsf{i}^{2}\) & Xin \(\mathsf{L}\mathsf{i}^{1,3}\)  
+
+Heterohelicenes are of increasing importance to the field of material science, molecular recognition, and asymmetric catalysis. However, the enantioselective construction of heterohelicenes, especially by organocatalytic method, remains rare and challenging. Herein, enantioenriched 1- (3- indol)- quino[5]helicenes have been synthesized through chiral phosphoric acid catalyzed Povarov reaction, followed by an oxidative central- to- helical chirality conversion process. The reaction has a broad scope and offers rapid access to an array of chiral quinohelicene derivatives with pretty enantioselectivities (up to \(99\%\) ee). Furthermore, optical properties of the studied quinohelicenes have been disclosed.  
+
+Helicenes, with ortho- fused \(\pi\) - conjugated rigid polycyclic aromatic structures, represents an important class of stereogenic elements. Due to their distinctive electronic properties, helicenes have been widely investigated for their potential use in materials science and molecule recognition. \(^{1,6}\) In addition, they can be used as chiral ligands and catalysts, \(^{1,7 - 10}\) and even be used to increase the oxygen evolution reaction activity of catalysts. \(^{11}\) Thereafter, their extensive applications have triggered massive investigations on the synthesis of chiral helicenes with novel structures and functional groups. Conventional syntheses of these valuable chiral molecules mainly depended on the resolution of racemic helicenes using chiral resolution reagents or through chiral HPLC separation, the chiral auxiliary- and chiral substrate- enabled asymmetric synthesis. \(^{12 - 17}\)  
+
+However, in a sharp contrast with the well- developed central chirality and axial chirality, the catalytic enantioselective synthesis of helicenes was largely under explored (Fig. 1A). In 1999, Stara and Starý realized the first example of enantioselective synthesis of helicenes, through a chiral [Ni] complex catalyzed \([2 + 2 + 2]\) cycloaddition reaction of triple alkynes. \(^{18}\) Since then, the transition- metal- catalyzed enantioselective \([2 + 2 + 2]\) cycloaddition has been applied as a universal method in the asymmetric construction of multiple chiral helicenes. \(^{19 - 28}\) Furthermore, the [Aul- catalyzed enantioselective intramolecular hydroxylation of alkynes represents another effective approach for the synthesis of chiral helicenes. \(^{29 - 34}\) In addition, transition- metal- catalyzed other approaches have been reported \(^{35 - 38}\) also, for example, a [V] catalyzed oxidative coupling of polycyclic phenol by Sasai, Takizawa and co- workers \(^{35}\) and a [Rh]- catalyzed enantioselective C- H activation/annulation by You, \(^{37}\) are also favorable supplements to the synthesis of chiral helicenes. Despite these efforts so far, specific challenges that are associated with the relatively high catalyst loading, hash reaction conditions and limited substrate scope remain unsolved.  
+
+In comparison with the transition- metal- catalyzed enantioselective synthesis of helicenes, corresponding organocatalytic strategies are very limited (Fig. 1A). In 2014, List reported an enantioselective synthesis of azalehicenes via chiral phosphoric acid catalyzed asymmetric Fischer indole reaction. \(^{39}\) Yan presented highly enantioselective synthesis of helicenes which undergo VQM intermediates enabled by asymmetric bifunctional amide catalysis. \(^{40,41}\) In 2020, Bonne and Rodriguez described an asymmetric synthesis of dioxa[6]helicenes by cinchona alkaloid- derived bifunctional catalyst through a Michael/O- alkylation heteroannulation process. \(^{42}\) It is valuable to note that the increasing demand for enantiomeric helicene compounds in various fields has stimulated the development of the highly efficient stereoselective asymmetric synthesis of structurally diverse helicenes. Therefore, the development of efficient strategy, especially orgaocatalyzed reaction, for the synthesis of helicenes is very meaningful and highly desirable.   
+
+On the other hand, quinohelicenes is a noted subclass of the heterohelicenes family, that possesses potential applications of optoelectronic materials, \(^{9,43 - 50}\) asymmetric catalysis \(^{7,9,51,52}\) and molecular recognition. \(^{53}\) (Fig. 1B). To data, only two examples of transition- metal- catalyzed methodologies were reported to synthesize chiral quinohelicenes. In 2014, Tanaka \(^{29}\) reported a [Au]- catalyzed intramolecular hydroxylation of alkynes, which suffered from high catalyst loading (30 mol%), limited substrate scope (only two examples) and moderate enantioselectivity (74% ee) (Fig. 1C). Very recently, Zhu \(^{38}\) realized the synthesis of chiral quinohelicenes through a [Pd]- catalyzed asymmetric double imidoylative cyclization (Fig. 1C). It should be noted also that the above mentioned two methods are both the quinoline heterocycles in the middle of the polycyclic aromatic structures. Accordingly, there is urgently need to develop a highly enantioselective method for the synthesis of quinohelicenes with the quinoline heterocycle on the side of polycyclic aromatic structures.  
+
+The Povarov reaction, which is a commendable means to build ring system from simple and readily available substrates, has been demonstrated to be one of the most attractive approaches in the construction of chiral molecules containing quinoline units. \(^{54 - 62}\) We speculated that chiral Bronsted acid catalyzed Povarov reaction followed with the subsequent oxidative aromatization process can synthesize chiral quinohelicenes (Fig. 1D). \(^{63 - 65}\) This process is highly desirable because it not only avoids the use of transition metal, but also has very high step- economy, which can
+
+<--- Page Split --->
+
+construct the chiral quinohelicene skeleton through one pot reaction of simple three component raw materials. Herein, we report a combination strategy of a chiral phosphoric acid catalyzed Povarov reaction and DDQ oxidative aromatiza-tion to prepare a wide range of chiral quinohelicenes with excellent enantioselectivities (Fig. 1D). Notably, the obtained quinohelicenes have rich optical properties.  
+
+![](images/Figure_unknown_0.jpg)
+  
+
+### C: Enantioselective synthesis of quinohelicenes  
+
+a) Tanaka's work: hydroarylation of alkynes:  
+
+![](images/Figure_2.jpg)
+  
+
+![](images/Figure_unknown_1.jpg)
+  
+
+![](images/Figure_3.jpg)
+
+<center>D: This work: Povarov reaction/aromatization </center>  
+
+## RESULTS AND DISCUSSION  
+
+As is known that, the energetic barrier for the interconversion between the two enantiomers of helically chiral compounds is highly dependent on the number of ortho- fused benzene rings, in which helicenes with more than 5 rings are usually stable. To probe the feasibility of our strategy, we firstly carried out the theoretical calculations of the activation energy of racemization of the target quinohelicene.66  
+
+To our delight, the calculated energy barrier of 1- (3- indol)- quinol[5]helicene is 36.9 kcal/mol (Fig. 2), which is 11.4 kcal/mol higher than the energy barrier of corresponding hydrogen substituted quinol[5]helicene.67 This result indicated that the product our plan to synthesize has sufficient stability at room temperature. So benzo[c]phenanthren- 2- amine 1a, which can be synthesized efficiently from cheap and readily materials, has been designed and prepared.35, 68, 69
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 2 |Process of enantiomerization for 1-(3-indol)-quino[5]helicene, the relative Gibbs free energy (kcal/mol) was calculated at the M06-2X/def2-TZVPP//B3LYP/def2-SVP lever. </center>  
+
+![](images/Figure_5.jpg)
+
+<center>Table 1 |Optimization of the reaction conditionsa,d </center>  
+
+Reaction conditions: \(^a A\) : Optimization of Povarov reaction: 1a (0.05 mmol) and 2a (0.2 mmol) in toluene (1.5 mL) at \(110^{\circ}C\) for \(12\textrm{h}\) then added 3a (0.1 mmol) and CPA\\* (0.0025 mmol) at rt for \(12h\) all dr values of 4a were \(>20:1\) . \(^b\) solated yield. The ee values were determined by high- performance liquid chromatography (HPLC) analysis with a chiral stationary phase. \(^b B\) :Optimization of oxidative aromatization: 4a (0.05 mmol, with \(99\%\) ee) and 1,2- Dichloro- 4,5- Dicyanobenzoquinone (DDQ, 0.15 mmol) in solvent (2 mL) at rt for \(3\textrm{h}\) . Conversion percentage (cp) \(= \text{ee}_{4a} / \text{ee}_{5a}\times 100\%\)  
+
+The reaction conditions was then optimized (Table 1). We started by using benzo[c]phenanthren- 2- amine 1a, benzaldehyde 2a and 3- vinyl- 1H- indole 3a as the model substrates. Initially, the reaction was performed in toluene in the  
+
+presence of 5 mol% (S)- A1 as the catalyst for 12 hours. As a result, the desired Povarov reaction product tetrahydroquinoline 4a was obtained in \(36\%\) yield and \(92\%\) ee (Table 1A, entry 1). Lightly, the desired quinohilcene 5a can be
+
+<--- Page Split --->
+
+further synthesized in \(87\%\) yield and \(90\%\) ee by aromatization of 4a.67 Inspired by this result, then other phosphoric acids (S)- A2 to (S)- C8 were assessed, showing that the (S)- A5 was the optimal one, in which the enantioselectivity of 4a was increased to \(99\%\) ee (Table 1A, entry 5).67 However, we found that the results of enantiomeric excess of 5a were highly irregular in toluene.67 In order to obtain stable and higher enantioselective conversion percentage values, we then carried out the solvent screening for the aromatization step (Table 1B). To our delight, when dichloromethane was used as the solvent for DDQ oxidation, 5a can be obtained in \(91\%\) yield and \(99\%\) ee with \(100\%\) cp value (Table 1B). To sum up, the optimal reaction condition is locked in that the enantioselective Povarov reaction is first carried out in toluene to generate 4a, and then the oxidation of 4a by DDQ in dichloromethane to generate the final product 5a.  
+
+![](images/Figure_6.jpg)
+
+<center>Fig. 3 | Scope of aromatic aldehydes and indole derivatives. Reaction conditions: 1) 1a (0.1 mmol) and 2 (0.4 mmol) in toluene (1.5 mL) at \(110^{\circ}\mathrm{C}\) for 12 h, then added 3a (0.2 mmol) and (S)-A5 (0.005 mmol) at rt for 12 h, purified by silica gel column chromatography to get 4.2) 4 and 1,2-Dichloro-4,5-Dicyanobenzoquinone (DDQ, 0.3 mmol) and DCM (2 mL) at rt for 3 h. </center>  
+
+With optimized reaction conditions in hand, we explored the substrate generality of this reaction. First, various aromatic aldehydes 2 were tested and the reaction results were summarized in Fig. 3A. The results showed that when the substituent group was present at p- position of benzaldehydes 2b- 2j, the corresponding quinohelicene products 5b- 5j can be obtained in moderate yields (40- 67%) with very good enantioselectivities (90- 98% ee), no matter whether the substituent is an electron- withdrawing (F, Cl, Br, NO2, CF3) or an electron- donating (Me, OMe, SMe, Ph) group. We also examined the reaction outcome with substituents at the m- and o- position on the benzaldehyde. As a result, the variations of either electron- donating or withdrawing character could be well- tolerated to deliver the corre  
+
+sponding products 5k- 5o in good to excellent enantioselectivities (84- 99% ee). Naphthaldehydes have also been investigated, affording the quinohelicenes 5p and 5q in 92% ee and 98% ee, respectively. Furthermore, furfural and 2- thenaldehyde were also compatible with the standard reaction conditions and delivered 5r and 5s in 95% ee and 98% ee, respectively.  
+
+To further extend the substrate scope, we evaluated the reaction of 1a and benzaldehyde 2a with various of indole substrate (Fig. 3B). To our delight, indoles with different substituents at the C- 5 position all tolerated this Povarov/oxidation strategy to give the corresponding quinohelicenes 5t- 5w in excellent 98- 99% ee values.
+
+<--- Page Split --->
+![](images/Figure_unknown_2.jpg)
+
+<center>Fig. 4 |Scope of various amine derivative. Reaction conditions: 1) 1 (0.1 mmol) and 2a (0.4 mmol) in toluene (1.5 mL) at \(110^{\circ}C\) for \(12h\) , then added 3a (0.2 mmol) and (S)-A5 (0.005 mmol) at rt for \(12h\) , purified by silica gel column chromatography to get 4. 2) 4 and 1,2-Dichloro-4,5-Dicyanobenzoquinone (DDQ, 0.3 mmol) and DCM (2 mL) at rt for \(3h\) . </center>  
+
+![PLACEHOLDER_5_1]
+
+<center>Fig. 5 |Reaction conditions: 4Large-scale reaction was conducted under optimized reaction conditions. 1a (0.1 mmol), 2a (0.4 mmol), 3a and (S)-A5 (0.005 mmol) in solvent (1.5 mL) at rt for \(24h\) then added 1,2-dichloro-4,5-dicyanobenzoquinone (DDQ 0.5 mmol) for another \(3h\) . </center>  
+
+We next investigated the aromatic amine substrates. And various substituted benzo[c]phenanthren- 2- amines 1b-  
+
+1i were designed and synthesized (Fig. 4A). Gratifyingly, all of the substituted benzo[c]phenanthren- 2- amines could
+
+<--- Page Split --->
+
+work well under the optimal conditions, giving the desired products in \(48 - 62\%\) yields with good to excellent enantioselectivities \((86 - 99\%)\) ee). In addition, 5- amino- 15H- benzo[c]indeno[2,1- a]phenanthren- 15- one was also apply to the current studied reaction, giving 5ai in \(45\%\) yield and \(95\%\) ee.  
+
+Further exploration of the substrate scope was focused on the skeleton of helicene. As shown in Scheme 4B, Chromen- containing quinohelicenes 5aj- 5al have been formed from chromen- containing naphthalen- 2- amines 1j- 11 in moderate to good yields \((46 - 81\%)\) and very good enantioselectivities \((86 - 97\%)\) ee). Furthermore, dinaphtho[2,1- b:1',2'- d]furan- 2- amine 1m with five rings was also attempted, which gave the quinohelicene product 5am in \(46\%\) yield and \(93\%\) ee. The absolute configuration of 5a and 5al were both assigned as \(M\) based on an X- ray diffraction analysis, and those of other products were assigned by analogy.67  
+
+To confirm the efficiency of our strategy in preparative synthesis, 1.0 mmol scale reaction was conducted with 1a, 2a and 3a (Fig. 5A). To our delight, the corresponding product 5a was obtained in \(57\%\) yield with almost maintained \(95\%\) ee. To illustrate the synthesis applicability of this protocol, some transformations of the allylation products were conducted. As shown in Fig. 5A, alkynyl- and cyanoquinol[5]helicenes 6a and 6b could been quantitative formed from 5a by simple operation without a loss in optical purity. These chiral products are prone to click chemical reactions and potential applications in biometrics. In order to further explore the practicability of this methodology, this three- step reaction was carried out in a one- pot procedure (Fig. 5B). As a result, the target chiral quinohelicene products can be obtained with satisfied enantioselectivities in the attempted three different solvents respectively. To our delight, when ethyl acetate was used as the solvent, 5a can be obtained in \(54\%\) yield and \(93\%\) ee.  
+
+![PLACEHOLDER_6_0]
+
+<center>Fig. 6 |Barriers to enantiomerization for quinohelicenes. aExperimental value. bThe relative Gibbs free energy (kcal/mol) was calculated at the M06-2X/def2-TZVP/P/B3LYP/def2-SVP lever. </center>  
+
+In order to investigate the stereochemical stability of the studied chiral quinohelicenes, we carried out the racemization experiment.67 As shown in Fig. 6, the racemic barrier of 5a and 5aj are \(34.9\mathrm{kcal / mol}\) and \(29.1\mathrm{kcal / mol}\) respectively, agreeing well with the calculated \(36.4\mathrm{kcal / mol}\) and \(30.6\mathrm{kcal / mol}\) .66 The reason for the observed 5.8 kcal/mol energy gap between 5a and 5aj is maybe due to the presence of C- sp in 5aj that makes the molecule more flexible. In addition, quinol[6]helicenes 5am exhibited good configurational stability even at \(170^{\circ}\mathrm{C}\) in 1,2- dichlorobenzene for 5 h without obvious racemization.  
+
+In order to demonstrate the potential application value of the chiral quinohelicenes, the photophysical properties of several selected quinohelicenes were investigated (Table 2). Notably, our quinohelicenes represented good quantum yields \((\Phi_{FL})\) (up to \(26.0\%\) ). Furthermore, we studied the effect of solvent on fluorescence, and found that the fluorescence gradually enhanced with increasing solvent polarity and showed a red shift (Table 2B). And the quantum yield of 5a visibly increased with the increase of solvent polarity from low- polarity dichloromethane \((\Phi_{FL} = 13.9\%)\) to high- polarity dimethyl sulfoxide \((\Phi_{FL} = 43.4\%)\) .66 As quinine containing helicenes exhibit high proton affinity, we have studied the optical properties of protonated counterpart 5a- H+. With the increase of TFA, the absorption at 240 to \(390\mathrm{nm}\) were gradually weakened, but increased at 400 to \(500\mathrm{nm}\) and the fluorescence emission peak was obviously red shifted (Table 2C).  
+
+Finally, the chiroptical properties of M- 5a, M- 5ab, M- 5ac, M- 5ae, M- 5ah, M- 5al and M- 5am, were subjected to a  
+
+preliminary evaluation (Table 2A).67 In the CD spectrums, these quinohelicenes displayed several signals at \(230\mathrm{nm}\) to \(400\mathrm{nm}\) and with fine absorption dissymmetry factors (gabs). The maximum gabs of M- 5ab, M- 5ac, M- 5ae reached 0.0072, 0.0083, and 0.0081 at 390, 393, and 388 nm respectively. The CD spectrum of enantiomer 5a appears as a mirror image, M- /P- 5al showed mirror image also (Table 2D). The CPL spectra of these enantiomers were also measured and the maximum gum of M- 5a, M- 5ab, M- 5ac, M- 5al reached 0.0038, 0.0030, 0.0051 and 0.0030 at 457, 459, 465 and 465 nm respectively. Moreover, M- /P- 5a, and M- /P- 5al also exhibited obvious mirror- image CPL emissions (Table 2E).  
+
+## CONCLUSION  
+
+In summary, we have developed a mild and multicomponent organocatalytic enantioselective Povarov/aromatization reaction of benzo[c]phenanthren- 2- amines, benzaldehydes and 3- vinyl- 1H- indoles provide a reliable tool for the preparation of highly functionalized quinohelicenes. The wide substrate scope, good yields and high enantioselectivities make this a promising method for preparing chiral helicenes, which are widely used in material science, molecular recognition, and asymmetric catalysis. Furthermore, this three- step reaction could carried out in a one- pot procedure. Moreover, the synthesized quinohelicenes showed enrich photophysical properties, including solvent and acid effects on UV and fluorescence, circular dichroic absorption (CD) and circular polarization luminescence (CPL) properties.
+
+<--- Page Split --->
+
+
+Table 2 |Photophysical properties of selected quinohelicens a   
+
+<table><tr><td>A:</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>Compound</td><td>b&amp;amp;abs. max(nm)</td><td>a&amp;amp;emi ((a&amp;amp;x)(nm))</td><td>d&amp;amp;F(%) (a&amp;amp;x(nm))</td><td>g&amp;amp;abs = Δε/ε (a&amp;amp;ab(nm))</td><td>g&amp;amp;lum (a&amp;amp;em/Δ&amp;amp;x(nm))</td></tr><tr><td>M-5a</td><td>267,325,420</td><td>440,450(325)</td><td>13.9(325)</td><td>0.0016(305), -0.0056(388)</td><td>0.0038(457/325)</td></tr><tr><td>M-5ab</td><td>269,325,420</td><td>438,457(325)</td><td>10.3(325)</td><td>0.0020(310), -0.0072(390)</td><td>0.0030(459/325)</td></tr><tr><td>M-5ac</td><td>266,330,425</td><td>440,461(330)</td><td>14.8(330)</td><td>0.0008(315), -0.0083(393)</td><td>0.0051(465/330)</td></tr><tr><td>M-5ae</td><td>266,315,420</td><td>460(315)</td><td>9.8(315)</td><td>-0.0026(363), 0.0081(388)</td><td>0.0024(468/315)</td></tr><tr><td>M-5ah</td><td>265,340,424</td><td>436,458(340)</td><td>26.0(340)</td><td>0.0024(301), -0.0065(392)</td><td>0.0015(460/340)</td></tr><tr><td>M-5al</td><td>216,320,418</td><td>458(320)</td><td>11.3(320)</td><td>-0.0017(260), 0.0045(370)</td><td>0.0030(465/320)</td></tr><tr><td>M-5am</td><td>257,313,412</td><td>436(313)</td><td>25.7(313)</td><td>-0.0036(275), 0.0048(379)</td><td>0.0013(477/313)</td></tr></table>  
+
+![PLACEHOLDER_7_0]
+
+<center>Measured at \(1\times 10^{-5}\mathrm{M}\) in DCM. Maximum UV absorption. Maximum fluorescence wavelength. Absolute fluorescence quantum efficiency under absorbance lower than 0.1. Measured with \(99\%\) ee. </center>  
+
+## Data availability  
+
+The authors declare that the data supporting the findings of this study are available within the article and its Supplementary Information files. Extra data are available from the author upon reasonable request. Crystallographic data for the structures reported in this Article have been deposited at the Cambridge Crystallographic Data Centre, under deposition numbers CCDC 2174258 (5a) and 2212769 (5al). Copies of the data can be obtained free of charge via https:  
+
+## REFERENCES  
+
+1. Shen, Y. & Chen, C. F. Helicenes: synthesis and applications. Chem. Rev. 112. 1463-1535 (2012).  
+
+2. Gingras, M. One hundred years of helicene chemistry. Part 3: applications and properties of carbochelicenes. Chem. Soc. Rev. 42. 1051-1095 (2013). 
+3. Saleh, N., Shen, C. & Crassous, J. Helicene-based transition metal complexes: synthesis, properties and applications. Chem. Sci. 5. 3680-3694 (2014). 
+4. Zhao, W. L., Li, M., Lu, H. Y. & Chen, C. F. Advances in helicene derivatives with circularly polarized luminescence. Chem. Commun. 55. 13793-13803 (2019). 
+5. Dhabibi, K., Favereau, L. & Crassous, J. Enantioenriched Helicenes and Heliconoids Containing Main-Group Elements (B, Si, N, P). Chem. Rev. 119. 8846-8953 (2019). 
+6. Mori, T. Chiroptical Properties of Symmetric Double, Triple, and Multiple Helicenes. Chem. Rev. 121. 2373-2412 (2021). 
+7. Takenaka, N., Sarangthem, R. S. & Captain, B. Helical chiral pyridine N-oxides: a new family of asymmetric catalysts. Angew. Chem. Int. Ed. 47. 9708-9710 (2008).
+
+<--- Page Split --->
+
+8. Norito Takenaka, J. C., Burjor Captain, Robindro Singh Sarangthem, and Appayee Chandrakumar. Helical Chiral 2-Aminopyridinium lons: A New Class of Hydrogen Bond Donor Catalysts. J. Am. Chem. Soc. 132. 4536-4537 (2010).9. Kos, M., et al. Enantioenriched Ruthenium-Tris-Bipyridine Complexes Bearing One Helical Bipyridine Ligand: Access to Fused Multihalecinic Systems and Chiroptical Redox Switches. Inorg. Chem. 60. 11838-11851 (2021).10. Sakamoto, D., et al. Cycloindicated Helicenes as Chiral Catalysts in the Asymmetric Transfer Hydrogenation of Imines. ACS Catal. 12. 10793-10800 (2022).11. Liang, Y., et al. Enhancement of electrocatalytic oxygen evolution by chiral molecular functionalization of hybrid 2D electrodes. Nat. Commun. 13. 3356 (2022).12. Misek, J., et al. A straightforward route to helically chiral N-heteroaromatic compounds: practical synthesis of racemic 1,14-diazal[5]helicene and optically pure 1- and 2-aza[6]helicenes. Angew. Chem. Int. Ed. 47. 3188-3191 (2008).13. Zadny, J., et al. A general approach to optically pure [5]-,[6]- and [7]heterohelicenes. Angew. Chem. Int. Ed. 51. 5857-5861 (2012).14. Otani, T., et al. Facile Two-Step Synthesis of 1,10-Phenanthroline-Derived Polyaza[7]helicenes with High Fluorescence and CPL Efficiency. Angew. Chem. Int. Ed. 56. 3906-3910 (2017).15. Li, J. K., et al. B,N-Embedded Double Hetero[7]helicenes with Strong Chiroptical Responses in the Visible Light Region. J. Am. Chem. Soc. 143. 17958-17963 (2021).16. Shen, C., et al. Oxidative cyclo-rearrangement of helicenes into chiral nanographenes. Nat. Commun. 12. 2786 (2021).17. Nakakuki, Y., et al. Doubly linked chiral phenanthrene oligomers for homogeneously \(\pi\) -extended helicenes with large effective conjugation length. Nat. Commun. 13. 1475 (2022).18. I. G. Stará, I. S., A. Kollárovic, F. Teplý, S. Vyskočil, D. Šaman., Transition Metal Catalysed Synthesis of Tetrahydro Derivatives of [5]-,[6]- and [7]Helicene. Tetrahedron Lett. 40. 1993-1996 (1999).19. Stará, I. G. & Starý, I. Helically Chiral Aromatics: The Synthesis of Helicenes by [2 + 2 + 2] Cycloisomerization of \(\pi\) -Electron Systems. Acc. Chem. Res. 53. 144-158 (2020).20. Liu, W., Qin, T., Xie, W. & Yang, X. Catalytic Enantioselective Synthesis of Helicenes. Chem. Eur. J., e202202369 (2022).21. Ken Tanaka, A. K., Takeshi Suda, Keiichi Noguchi, and Masao Hirano. Rh-Catalyzed Synthesis of Helically Chiral and Ladder-Type Molecules via [2 + 2 + 2] and Formal [2 + 1 + 2 + 1] Cycloadditions Involving C-C Triple Bond Cleavage. J. Am. Chem. Soc. 129. 12078-12079 (2007).22. Tanaka, K., Fukawa, N., Suda, T. & Noguchi, K. One-step construction of five successive rings by rhodium-catalyzed intermolecular double [2+2+2] cycloaddition: enantiomeriched [9]helicene-like molecules. Angew. Chem. Int. Ed. 48. 5470-5473 (2009).23. Fukawa, N., Osaka, T., Noguchi, K. & Tanaka, K. Asymmet ric Synthesis and Photophysical Properties of Benzopyrano- or Naphthopyrano-Fused Helical Phosphafluorenes. Org. Lett. 12. 1324-1327 (2010).24. Jancarik, A., et al. Rapid access to dibenzohelicenes and their functionalized derivatives. Angew. Chem. Int. Ed. 52. 9970-9975 (2013).25. Kimura, Y., Fukawa, N., Miyauchi, Y., Noguchi, K. & Tanaka, K. Enantioselective synthesis of [9]- and [11]helicene-like molecules: double intramolecular [2+2+2] cycloaddition. Angew. Chem. Int. Ed. 53. 8480-8483 (2014).26. Samal, M., et al. An Ultimate Stereocontrol in Asymmetric Synthesis of Optically Pure Fully Aromatic Helicenes. J. Am. Chem. Soc. 137. 8469-8474 (2015).27. Kimura, Y., Shibata, Y., Noguchi, K. & Tanaka, K. Enantioselective Synthesis and Epimerization Behavior of a Chiral S - Shaped [11]Helicene - Like Molecule Having Collision between Terminal Benzene Rings. Eur. J. Org. Chem. 2019. 1390-1396 (2019).28. Kinoshita, S., et al. Rhodium-Catalyzed Highly Diastereo- and Enantioselective Synthesis of a Configurationally Stable S- Shaped Double Helicene-Like Molecule. Angew. Chem. Int. Ed. 59. 11020-11027 (2020).  
+
+29. Nakamura, K., Furumi, S., Takeuchi, M., Shibuya, T. & Tanaka, K. Enantioselective synthesis and enhanced circularly polarized luminescence of S-shaped double azalehicenes. J. Am. Chem. Soc. 136. 5555-5558 (2014).  
+30. Tanaka, M., et al. Gold-Catalyzed Enantioselective Synthesis, Crystal Structure, and Photophysical/Chiroptical Properties of Aza[10]helicenes. Chem. Eur. J. 22. 9537-9541 (2016).  
+31. Gonzalez-Fernandez, E., Nicholls, L. D., Schaaf, L. D., Fares, C., Lehmann, C. W. & Alcarazo, M. Enantioselective Synthesis of [6]Carbochelicenes. J. Am. Chem. Soc. 139. 1428-1431 (2017).  
+32. Yamano, R., Kinoshita, S., Shibata, Y. & Tanaka, K. Substituent - Controlled and Rhodium - Catalyzed Intramolecular [2+2+2] and [2+1+2+1] Cycloadditions of Electron - Deficient Triynes. Eur. J. Org. Chem. 2018. 5916-5920 (2018).  
+33. Usui, K., et al. Internal-Edge-Substituted Coumarin-Fused [6]Helicenes: Asymmetric Synthesis, Structural Features, and Control of Self-Assembly. Chem. Eur. J. 24. 14617-14621 (2018).  
+34. Hartung, T., Machleid, R., Simon, M., Golz, C. & Alcarazo, M. Enantioselective Synthesis of 1,12-Disubstituted [4]Helicenes. Angew. Chem. Int. Ed. 59. 5660-5664 (2020).  
+35. Sako, M., et al. fficient Enantioselective Synthesis of Oxahiclenes Using Redox Acid Cooperative Catalysts. J. Am. Chem. Soc. 138. 11481-11484 (2016).  
+36. Yubuta, A., et al. Enantioselective Synthesis of Triple Helicenes by Cross-Cyclotrimerization of a Helicenyi Aryne and Alkynes via Dynamic Kinetic Resolution. J. Am. Chem. Soc. 142. 10025-10033 (2020).  
+37. Wang, Q., Zhang, W. W., Zheng, C., Gu, Q. & You, S. L. Enantioselective Synthesis of Azoniahelicenes by Rh-Catalyzed C-H Annulation with Alkynes. J. Am. Chem. Soc. 143. 114-120 (2021).  
+38. Yu, T., et al. Palladium-Catalyzed Modular Synthesis of Enantioenriched Pyridohelicenes through Double Imidoylative Cyclization. ACS Catal. 12. 13034-13041 (2022).  
+39. Kotzner, L., Webber, M. J., Martinez, A., De Fusco, C. & List, B. Asymmetric catalysis on the nanoscale: the organocatalytic approach to helicenes. Angew. Chem. Int. Ed. 53. 5202-5205 (2014).  
+40. Jia, S., Li, S., Liu, Y., Qin, W. & Yan, H. Enantioselective Control of Both Helical and Axial Stereogenic Elements though an Organocatalytic Approach. Angew. Chem. Int. Ed. 58. 18496-18501 (2019).  
+41. Li, K., Huang, S., Liu, T., Jia, S. & Yan, H. Organocatalytic Asymmetric Dearomatizing Hetero-Diels-Alder Reaction of Nonactivated Arenes. J. Am. Chem. Soc. 144. 7374-7381 (2022).  
+42. Liu, P., et al. Simultaneous Control of Central and Helical Chiralities: Expedient Helicoselective Synthesis of Dioxa[6]helicenes. J. Am. Chem. Soc. 142. 16199-16204 (2020).  
+43. Kaneko, E., Matsumoto, Y. & Kamikawa, K. Synthesis of AzaheliceneN-Oxide by Palladium-Catalyzed Direct C H Annulation of a Pendant (Z)-Bromovinyl Side Chain. Chem. Eur. J. 19. 11837-11841 (2013).  
+44. Yang, Y., da Costa, R. C., Fuchter, M. J. & Campbell, A. J. Circularly polarized light detection by a chiral organic semiconductor transistor. Nature Pha. 7. 634-638 (2013).  
+45. Yang, Y., da Costa, R. C., Smilgies, D. M., Campbell, A. J. & Fuchter, M. J. Induction of circularly polarized electroluminescence from an achiral light-emitting polymer via a chiral small-molecule dopant. Adv. Mater. 25. 2624-2628 (2013).  
+46. Abbate, S., et al. Helical Sense-Responsive and Substituent-Sensitive Features in Vibrational and Electronic Circular Dichroism, in Circularly Polarized Luminescence, and in Raman Spectra of Some Simple Optically Active Hexahelicenes. J. Phys. Chem. C 118. 1682-1695 (2014).  
+47. Klivar, J., et al. [2+2+2] Cycloisomerisation of Aromatic Cyanodiynes in the Synthesis of Pyridohelicenes and Their Analogues. Chem. Eur. J. 22. 14401-14405 (2016).  
+48. Feng, J., Wang, L., Xue, X., Chao, Z., Hong, B. & Gu, Z. Ring-Expansion Strategy for alpha-Aryl Azahelicene Construction:
+
+<--- Page Split --->
+
+Building Blocks for Optoelectronic Materials. Org. Lett. 23. 8056- 8061 (2021).49. Yen- Pon, E., et al. Heterohelicenes through 1,3- Dipolar Cycloaddition of Syndones with Arynes: Synthesis, Origins of Selectivity, and Application to pH- Triggered Chiroptical Switch with CPL Sign Reversal. JACS Au 1. 807- 818 (2021).50. Gong, X., Li, C., Cai, Z., Wan, X., Qian, H., & Yang, G. Synthesis of Nitrogen- Doped Aza- Helicenes with Chiral Optical Properties. J. Org. Chem. 87. 8406- 8412 (2022).51. Chen, J. & Takenaka, N. Helical chiral pyridine N- oxides: a new family of asymmetric catalysts. Chem. Eur. J. 15. 7268- 7276 (2009).52. Takenaka, N., Chen, J. S., Captain, B., Sarangthem, R. S., & Chandrakumar, A. Helical Chiral 2- Aminopyridinium Ions: A New Class of Hydrogen Bond Donor Catalysts. J. Am. Chem. Soc. 132. 4536- 4537 (2010).53. Huang, Q., et al. Inherently Chiral Azonia[6]helicene- Modified beta- Cyclodextrin: Synthesis, Characterization, and Chirality Sensing of Underivatized Amino Acids in Water. J. Org. Chem. 81. 3430- 3434 (2016).54. Akiyama, T., Morita, H., & Fuchibe, K. Chiral Bronsted acid- catalyzed inverse electron- demand aza Diels- Alder reaction. J. Am. Chem. Soc. 128. 13070- 13071 (2006).55. Xie, M., et al. Asymmetric three- component inverse electron- demand aza- Diels- Alder reaction: efficient synthesis of ring- fused tetrahydroquinolines. Angew. Chem. Int. Ed. 49. 3799- 3802 (2010).56. Dagousset, G., Zhu, J., & Masson, G. Chiral phosphoric acid- catalyzed enantioselective three- component Povarov reaction using encarbamates as dienophiles: highly diastereo- and enantioselective synthesis of substituted 4- aminotetrahydroquinolines. J. Am. Chem. Soc. 133. 14804- 14813 (2011).57. Chen, Z., Wang, B., Wang, Z., Zhu, G., & Sun, J. Complex bioactive alkaloid- type polycycles through efficient catalytic asymmetric multicomponent aza- Diels- Alder reaction of indoles with oxetane as directing group. Angew. Chem. Int. Ed. 52. 2027- 2031 (2013).58. Yu, J., Jiang, H. J., Zhou, Y., Luo, S. W., & Gong, L. Z. Sodium salts of anionic chiral cobalt(III) complexes as catalysts of the enantioselective Povarov reaction. Angew. Chem. Int. Ed. 54. 11209- 11213 (2015).59. Bisag, G. D., et al. Central- to- Axial Chirality Conversion Approach Designed on Organocatalytic Enantioselective Povarov Cycloadditions: First Access to Configurationally Stable Indole- Quinoline Atropisomers. Chem. Eur. J. 25. 15694- 15701 (2019).60. Wang, S. J., Wang, Z., Tang, Y., Chen, J., & Zhou, L. Asymmetric Synthesis of Quinoline- Naphthalene Atropisomers by Central- to- Axial Chirality Conversion. Org. Lett. 22. 8894- 8898 (2020).61. Clerigue, J., Ramos, M. T., & Menendez, J. C. Enantioselective catalytic Povarov reactions. Org. Biomol. Chem. 20. 1550- 1581 (2022).62. Lemos, B. C., Venturini Filho, E., Fiorot, R. G., Medici, F., Greco, S. J., & Benaglia, M. Enantioselective Povarov Reactions: An Update of a Powerful Catalytic Synthetic Methodology. Eur. J. Org. Chem. e202101171 (2022).63. Viglianisi, C., et al. Synthesis of Heterohelicenes by a Catalytic Multi- Component Povarov Reaction. Eur. J. Org. Chem. 2019. 164- 167 (2019).64. Soni, R., & Soman, S. S. Metal free synthesis of Coumarin containing hetero n helicene like molecules with TICT and AIE properties. Asian. J. Org. Chem. 11. e202100770 (2022).65. In the literature, although there have been two re- ports on the construction of helicenes through Pova- rov reaction, both of them are racemic version. More- over, Lewis acid catalysis is used, which suffered from high catalyst loading and harsh conditions.66. DFT Calculations on the Enantiomerization Process of 1- H- quinol[5]helicene and 1- (3- indol)- quinol[5]helicene. The relative Gibbs free energy (kcal/mol) was calculated at the M06- 2X/def2- TZVPP//B3LYP/def2- SVP lever. For details see Supporting Information.67. For details see Supporting Information.68. Yu, J., Zhang, P., Wu, J., & Shang, Z. Metal- free C- N bond- forming reaction: straightforward synthesis of anilines, through cleavage of aryl C- O bond and amide C- N bond. Tetrahedron Lett. 54. 3167- 3170 (2013).69. Wu, J., Yu, J., Wang, Y., & Zhang, P. Direct Animation of Phenols under Metal- Free Conditions. Synlett 24. 1448- 1454 (2013).  
+
+cleavage of aryl C- O bond and amide C- N bond. Tetrahedron Lett. 54. 3167- 3170 (2013).69. Wu, J., Yu, J., Wang, Y., & Zhang, P. Direct Animation of Phenols under Metal- Free Conditions. Synlett 24. 1448- 1454 (2013).  
+
+## Acknowledgements  
+
+We are grateful to the National Natural Science Foundation of China (Grant Nos. 22193011, 21971120 and 21933008) and National Science & Technology Fundamental Resource Investigation Program of China (No. 2018FY201200) for financial support.  
+
+## Author contributions  
+
+W. L. conceived and designed the study and performed experiments and wrote the manuscript; G. X. synthesized some substrates; Z. R. performed the theoretical calculations; H.G and M. L. performed the CPL experiments; X. L. conceived and designed the study and wrote the manuscript.  
+
+## Competing interests  
+
+The authors declare no competing financial interest  
+
+## Additional information  
+
+Supplementary information is available for this paper at  
+
+Correspondence and requests for materials should be addressed to Xin Li.
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- 5alcheckcif.pdf- SupportingInformationNat.commun.20221130.pdf
+
+<--- Page Split --->

preprint/preprint__4844999f9c48fab95f677e12ff57f18204d1027644dc64a7ffea5ccbbcee4859/preprint__4844999f9c48fab95f677e12ff57f18204d1027644dc64a7ffea5ccbbcee4859_det.mmd

@@ -0,0 +1,277 @@
+<|ref|>title<|/ref|><|det|>[[44, 106, 864, 208]]<|/det|>
+# Enantioselective Synthesis of Quinohelicenes through a Sequential Organocatalyzed Povarov Reaction/Oxidative Aromatization  
+
+<|ref|>text<|/ref|><|det|>[[44, 230, 201, 270]]<|/det|>
+Chengwen Li Nankai University  
+
+<|ref|>text<|/ref|><|det|>[[44, 277, 209, 315]]<|/det|>
+Xi Gao Nankai University  
+
+<|ref|>text<|/ref|><|det|>[[44, 323, 209, 362]]<|/det|>
+Zhiyuan Ren Nankai University  
+
+<|ref|>text<|/ref|><|det|>[[44, 369, 288, 408]]<|/det|>
+Chenhao Guo Institute of Chemistry, CAS  
+
+<|ref|>text<|/ref|><|det|>[[44, 415, 560, 502]]<|/det|>
+Meng Li chinese academy of sciences Xin Li (xin_li@nankai.edu.cn) Nankai University https://orcid.org/0000- 0001- 6020- 9170  
+
+<|ref|>text<|/ref|><|det|>[[44, 545, 102, 562]]<|/det|>
+Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 583, 136, 600]]<|/det|>
+Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 620, 319, 639]]<|/det|>
+Posted Date: January 4th, 2023  
+
+<|ref|>text<|/ref|><|det|>[[44, 658, 474, 678]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 2327586/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 696, 910, 738]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[44, 756, 530, 776]]<|/det|>
+Additional Declarations: There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[42, 812, 952, 856]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on June 8th, 2023. See the published version at https://doi.org/10.1038/s41467- 023- 39134- 9.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[88, 44, 896, 92]]<|/det|>
+# Enantioselective Synthesis of Quinohelicenes through a Sequen- tial Organocatalyzed Povarov Reaction/Oxidative Aromatization  
+
+<|ref|>text<|/ref|><|det|>[[88, 100, 700, 120]]<|/det|>
+Chengwen \(\mathsf{L}\mathsf{i}^{1}\) , Xi Gao \(^{1}\) , Zhiyuan \(\mathsf{Re n}^{1}\) , Chenhao Guo \(^{2}\) , Meng \(\mathsf{L}\mathsf{i}^{2}\) & Xin \(\mathsf{L}\mathsf{i}^{1,3}\)  
+
+<|ref|>text<|/ref|><|det|>[[388, 146, 898, 315]]<|/det|>
+Heterohelicenes are of increasing importance to the field of material science, molecular recognition, and asymmetric catalysis. However, the enantioselective construction of heterohelicenes, especially by organocatalytic method, remains rare and challenging. Herein, enantioenriched 1- (3- indol)- quino[5]helicenes have been synthesized through chiral phosphoric acid catalyzed Povarov reaction, followed by an oxidative central- to- helical chirality conversion process. The reaction has a broad scope and offers rapid access to an array of chiral quinohelicene derivatives with pretty enantioselectivities (up to \(99\%\) ee). Furthermore, optical properties of the studied quinohelicenes have been disclosed.  
+
+<|ref|>text<|/ref|><|det|>[[88, 337, 480, 533]]<|/det|>
+Helicenes, with ortho- fused \(\pi\) - conjugated rigid polycyclic aromatic structures, represents an important class of stereogenic elements. Due to their distinctive electronic properties, helicenes have been widely investigated for their potential use in materials science and molecule recognition. \(^{1,6}\) In addition, they can be used as chiral ligands and catalysts, \(^{1,7 - 10}\) and even be used to increase the oxygen evolution reaction activity of catalysts. \(^{11}\) Thereafter, their extensive applications have triggered massive investigations on the synthesis of chiral helicenes with novel structures and functional groups. Conventional syntheses of these valuable chiral molecules mainly depended on the resolution of racemic helicenes using chiral resolution reagents or through chiral HPLC separation, the chiral auxiliary- and chiral substrate- enabled asymmetric synthesis. \(^{12 - 17}\)  
+
+<|ref|>text<|/ref|><|det|>[[88, 536, 480, 823]]<|/det|>
+However, in a sharp contrast with the well- developed central chirality and axial chirality, the catalytic enantioselective synthesis of helicenes was largely under explored (Fig. 1A). In 1999, Stara and Starý realized the first example of enantioselective synthesis of helicenes, through a chiral [Ni] complex catalyzed \([2 + 2 + 2]\) cycloaddition reaction of triple alkynes. \(^{18}\) Since then, the transition- metal- catalyzed enantioselective \([2 + 2 + 2]\) cycloaddition has been applied as a universal method in the asymmetric construction of multiple chiral helicenes. \(^{19 - 28}\) Furthermore, the [Aul- catalyzed enantioselective intramolecular hydroxylation of alkynes represents another effective approach for the synthesis of chiral helicenes. \(^{29 - 34}\) In addition, transition- metal- catalyzed other approaches have been reported \(^{35 - 38}\) also, for example, a [V] catalyzed oxidative coupling of polycyclic phenol by Sasai, Takizawa and co- workers \(^{35}\) and a [Rh]- catalyzed enantioselective C- H activation/annulation by You, \(^{37}\) are also favorable supplements to the synthesis of chiral helicenes. Despite these efforts so far, specific challenges that are associated with the relatively high catalyst loading, hash reaction conditions and limited substrate scope remain unsolved.  
+
+<|ref|>text<|/ref|><|det|>[[88, 827, 480, 894], [516, 337, 907, 507]]<|/det|>
+In comparison with the transition- metal- catalyzed enantioselective synthesis of helicenes, corresponding organocatalytic strategies are very limited (Fig. 1A). In 2014, List reported an enantioselective synthesis of azalehicenes via chiral phosphoric acid catalyzed asymmetric Fischer indole reaction. \(^{39}\) Yan presented highly enantioselective synthesis of helicenes which undergo VQM intermediates enabled by asymmetric bifunctional amide catalysis. \(^{40,41}\) In 2020, Bonne and Rodriguez described an asymmetric synthesis of dioxa[6]helicenes by cinchona alkaloid- derived bifunctional catalyst through a Michael/O- alkylation heteroannulation process. \(^{42}\) It is valuable to note that the increasing demand for enantiomeric helicene compounds in various fields has stimulated the development of the highly efficient stereoselective asymmetric synthesis of structurally diverse helicenes. Therefore, the development of efficient strategy, especially orgaocatalyzed reaction, for the synthesis of helicenes is very meaningful and highly desirable.   
+
+<|ref|>text<|/ref|><|det|>[[516, 510, 907, 757]]<|/det|>
+On the other hand, quinohelicenes is a noted subclass of the heterohelicenes family, that possesses potential applications of optoelectronic materials, \(^{9,43 - 50}\) asymmetric catalysis \(^{7,9,51,52}\) and molecular recognition. \(^{53}\) (Fig. 1B). To data, only two examples of transition- metal- catalyzed methodologies were reported to synthesize chiral quinohelicenes. In 2014, Tanaka \(^{29}\) reported a [Au]- catalyzed intramolecular hydroxylation of alkynes, which suffered from high catalyst loading (30 mol%), limited substrate scope (only two examples) and moderate enantioselectivity (74% ee) (Fig. 1C). Very recently, Zhu \(^{38}\) realized the synthesis of chiral quinohelicenes through a [Pd]- catalyzed asymmetric double imidoylative cyclization (Fig. 1C). It should be noted also that the above mentioned two methods are both the quinoline heterocycles in the middle of the polycyclic aromatic structures. Accordingly, there is urgently need to develop a highly enantioselective method for the synthesis of quinohelicenes with the quinoline heterocycle on the side of polycyclic aromatic structures.  
+
+<|ref|>text<|/ref|><|det|>[[516, 761, 907, 894]]<|/det|>
+The Povarov reaction, which is a commendable means to build ring system from simple and readily available substrates, has been demonstrated to be one of the most attractive approaches in the construction of chiral molecules containing quinoline units. \(^{54 - 62}\) We speculated that chiral Bronsted acid catalyzed Povarov reaction followed with the subsequent oxidative aromatization process can synthesize chiral quinohelicenes (Fig. 1D). \(^{63 - 65}\) This process is highly desirable because it not only avoids the use of transition metal, but also has very high step- economy, which can
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[86, 44, 481, 97]]<|/det|>
+construct the chiral quinohelicene skeleton through one pot reaction of simple three component raw materials. Herein, we report a combination strategy of a chiral phosphoric acid catalyzed Povarov reaction and DDQ oxidative aromatiza-tion to prepare a wide range of chiral quinohelicenes with excellent enantioselectivities (Fig. 1D). Notably, the obtained quinohelicenes have rich optical properties.  
+
+<|ref|>image<|/ref|><|det|>[[101, 110, 910, 375]]<|/det|>  
+
+<|ref|>sub_title<|/ref|><|det|>[[97, 390, 380, 402]]<|/det|>
+### C: Enantioselective synthesis of quinohelicenes  
+
+<|ref|>text<|/ref|><|det|>[[100, 409, 330, 420]]<|/det|>
+a) Tanaka's work: hydroarylation of alkynes:  
+
+<|ref|>image<|/ref|><|det|>[[118, 424, 490, 512]]<|/det|>  
+
+<|ref|>image<|/ref|><|det|>[[510, 422, 901, 520]]<|/det|>  
+
+<|ref|>image<|/ref|><|det|>[[150, 550, 625, 680]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[146, 536, 419, 547]]<|/det|>
+<center>D: This work: Povarov reaction/aromatization </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[87, 730, 323, 746]]<|/det|>
+## RESULTS AND DISCUSSION  
+
+<|ref|>text<|/ref|><|det|>[[87, 749, 481, 840]]<|/det|>
+As is known that, the energetic barrier for the interconversion between the two enantiomers of helically chiral compounds is highly dependent on the number of ortho- fused benzene rings, in which helicenes with more than 5 rings are usually stable. To probe the feasibility of our strategy, we firstly carried out the theoretical calculations of the activation energy of racemization of the target quinohelicene.66  
+
+<|ref|>text<|/ref|><|det|>[[516, 720, 911, 833]]<|/det|>
+To our delight, the calculated energy barrier of 1- (3- indol)- quinol[5]helicene is 36.9 kcal/mol (Fig. 2), which is 11.4 kcal/mol higher than the energy barrier of corresponding hydrogen substituted quinol[5]helicene.67 This result indicated that the product our plan to synthesize has sufficient stability at room temperature. So benzo[c]phenanthren- 2- amine 1a, which can be synthesized efficiently from cheap and readily materials, has been designed and prepared.35, 68, 69
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[220, 52, 777, 241]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[100, 252, 896, 277]]<|/det|>
+<center>Fig. 2 |Process of enantiomerization for 1-(3-indol)-quino[5]helicene, the relative Gibbs free energy (kcal/mol) was calculated at the M06-2X/def2-TZVPP//B3LYP/def2-SVP lever. </center>  
+
+<|ref|>image<|/ref|><|det|>[[170, 330, 808, 760]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[100, 303, 450, 319]]<|/det|>
+<center>Table 1 |Optimization of the reaction conditionsa,d </center>  
+
+<|ref|>text<|/ref|><|det|>[[100, 789, 900, 847]]<|/det|>
+Reaction conditions: \(^a A\) : Optimization of Povarov reaction: 1a (0.05 mmol) and 2a (0.2 mmol) in toluene (1.5 mL) at \(110^{\circ}C\) for \(12\textrm{h}\) then added 3a (0.1 mmol) and CPA\\* (0.0025 mmol) at rt for \(12h\) all dr values of 4a were \(>20:1\) . \(^b\) solated yield. The ee values were determined by high- performance liquid chromatography (HPLC) analysis with a chiral stationary phase. \(^b B\) :Optimization of oxidative aromatization: 4a (0.05 mmol, with \(99\%\) ee) and 1,2- Dichloro- 4,5- Dicyanobenzoquinone (DDQ, 0.15 mmol) in solvent (2 mL) at rt for \(3\textrm{h}\) . Conversion percentage (cp) \(= \text{ee}_{4a} / \text{ee}_{5a}\times 100\%\)  
+
+<|ref|>text<|/ref|><|det|>[[87, 877, 480, 930]]<|/det|>
+The reaction conditions was then optimized (Table 1). We started by using benzo[c]phenanthren- 2- amine 1a, benzaldehyde 2a and 3- vinyl- 1H- indole 3a as the model substrates. Initially, the reaction was performed in toluene in the  
+
+<|ref|>text<|/ref|><|det|>[[515, 877, 910, 930]]<|/det|>
+presence of 5 mol% (S)- A1 as the catalyst for 12 hours. As a result, the desired Povarov reaction product tetrahydroquinoline 4a was obtained in \(36\%\) yield and \(92\%\) ee (Table 1A, entry 1). Lightly, the desired quinohilcene 5a can be
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[86, 44, 483, 150]]<|/det|>
+further synthesized in \(87\%\) yield and \(90\%\) ee by aromatization of 4a.67 Inspired by this result, then other phosphoric acids (S)- A2 to (S)- C8 were assessed, showing that the (S)- A5 was the optimal one, in which the enantioselectivity of 4a was increased to \(99\%\) ee (Table 1A, entry 5).67 However, we found that the results of enantiomeric excess of 5a were highly irregular in toluene.67 In order to obtain stable and higher enantioselective conversion percentage values, we then carried out the solvent screening for the aromatization step (Table 1B). To our delight, when dichloromethane was used as the solvent for DDQ oxidation, 5a can be obtained in \(91\%\) yield and \(99\%\) ee with \(100\%\) cp value (Table 1B). To sum up, the optimal reaction condition is locked in that the enantioselective Povarov reaction is first carried out in toluene to generate 4a, and then the oxidation of 4a by DDQ in dichloromethane to generate the final product 5a.  
+
+<|ref|>image<|/ref|><|det|>[[87, 170, 910, 680]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[99, 683, 900, 720]]<|/det|>
+<center>Fig. 3 | Scope of aromatic aldehydes and indole derivatives. Reaction conditions: 1) 1a (0.1 mmol) and 2 (0.4 mmol) in toluene (1.5 mL) at \(110^{\circ}\mathrm{C}\) for 12 h, then added 3a (0.2 mmol) and (S)-A5 (0.005 mmol) at rt for 12 h, purified by silica gel column chromatography to get 4.2) 4 and 1,2-Dichloro-4,5-Dicyanobenzoquinone (DDQ, 0.3 mmol) and DCM (2 mL) at rt for 3 h. </center>  
+
+<|ref|>text<|/ref|><|det|>[[87, 746, 483, 929]]<|/det|>
+With optimized reaction conditions in hand, we explored the substrate generality of this reaction. First, various aromatic aldehydes 2 were tested and the reaction results were summarized in Fig. 3A. The results showed that when the substituent group was present at p- position of benzaldehydes 2b- 2j, the corresponding quinohelicene products 5b- 5j can be obtained in moderate yields (40- 67%) with very good enantioselectivities (90- 98% ee), no matter whether the substituent is an electron- withdrawing (F, Cl, Br, NO2, CF3) or an electron- donating (Me, OMe, SMe, Ph) group. We also examined the reaction outcome with substituents at the m- and o- position on the benzaldehyde. As a result, the variations of either electron- donating or withdrawing character could be well- tolerated to deliver the corre  
+
+<|ref|>text<|/ref|><|det|>[[515, 746, 910, 839]]<|/det|>
+sponding products 5k- 5o in good to excellent enantioselectivities (84- 99% ee). Naphthaldehydes have also been investigated, affording the quinohelicenes 5p and 5q in 92% ee and 98% ee, respectively. Furthermore, furfural and 2- thenaldehyde were also compatible with the standard reaction conditions and delivered 5r and 5s in 95% ee and 98% ee, respectively.  
+
+<|ref|>text<|/ref|><|det|>[[515, 842, 910, 920]]<|/det|>
+To further extend the substrate scope, we evaluated the reaction of 1a and benzaldehyde 2a with various of indole substrate (Fig. 3B). To our delight, indoles with different substituents at the C- 5 position all tolerated this Povarov/oxidation strategy to give the corresponding quinohelicenes 5t- 5w in excellent 98- 99% ee values.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[95, 45, 907, 440]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[99, 444, 901, 481]]<|/det|>
+<center>Fig. 4 |Scope of various amine derivative. Reaction conditions: 1) 1 (0.1 mmol) and 2a (0.4 mmol) in toluene (1.5 mL) at \(110^{\circ}C\) for \(12h\) , then added 3a (0.2 mmol) and (S)-A5 (0.005 mmol) at rt for \(12h\) , purified by silica gel column chromatography to get 4. 2) 4 and 1,2-Dichloro-4,5-Dicyanobenzoquinone (DDQ, 0.3 mmol) and DCM (2 mL) at rt for \(3h\) . </center>  
+
+<|ref|>image<|/ref|><|det|>[[199, 512, 796, 821]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[100, 831, 900, 857]]<|/det|>
+<center>Fig. 5 |Reaction conditions: 4Large-scale reaction was conducted under optimized reaction conditions. 1a (0.1 mmol), 2a (0.4 mmol), 3a and (S)-A5 (0.005 mmol) in solvent (1.5 mL) at rt for \(24h\) then added 1,2-dichloro-4,5-dicyanobenzoquinone (DDQ 0.5 mmol) for another \(3h\) . </center>  
+
+<|ref|>text<|/ref|><|det|>[[87, 886, 480, 914]]<|/det|>
+We next investigated the aromatic amine substrates. And various substituted benzo[c]phenanthren- 2- amines 1b-  
+
+<|ref|>text<|/ref|><|det|>[[515, 886, 911, 914]]<|/det|>
+1i were designed and synthesized (Fig. 4A). Gratifyingly, all of the substituted benzo[c]phenanthren- 2- amines could
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[87, 44, 480, 123]]<|/det|>
+work well under the optimal conditions, giving the desired products in \(48 - 62\%\) yields with good to excellent enantioselectivities \((86 - 99\%)\) ee). In addition, 5- amino- 15H- benzo[c]indeno[2,1- a]phenanthren- 15- one was also apply to the current studied reaction, giving 5ai in \(45\%\) yield and \(95\%\) ee.  
+
+<|ref|>text<|/ref|><|det|>[[87, 127, 480, 271]]<|/det|>
+Further exploration of the substrate scope was focused on the skeleton of helicene. As shown in Scheme 4B, Chromen- containing quinohelicenes 5aj- 5al have been formed from chromen- containing naphthalen- 2- amines 1j- 11 in moderate to good yields \((46 - 81\%)\) and very good enantioselectivities \((86 - 97\%)\) ee). Furthermore, dinaphtho[2,1- b:1',2'- d]furan- 2- amine 1m with five rings was also attempted, which gave the quinohelicene product 5am in \(46\%\) yield and \(93\%\) ee. The absolute configuration of 5a and 5al were both assigned as \(M\) based on an X- ray diffraction analysis, and those of other products were assigned by analogy.67  
+
+<|ref|>text<|/ref|><|det|>[[517, 45, 910, 280]]<|/det|>
+To confirm the efficiency of our strategy in preparative synthesis, 1.0 mmol scale reaction was conducted with 1a, 2a and 3a (Fig. 5A). To our delight, the corresponding product 5a was obtained in \(57\%\) yield with almost maintained \(95\%\) ee. To illustrate the synthesis applicability of this protocol, some transformations of the allylation products were conducted. As shown in Fig. 5A, alkynyl- and cyanoquinol[5]helicenes 6a and 6b could been quantitative formed from 5a by simple operation without a loss in optical purity. These chiral products are prone to click chemical reactions and potential applications in biometrics. In order to further explore the practicability of this methodology, this three- step reaction was carried out in a one- pot procedure (Fig. 5B). As a result, the target chiral quinohelicene products can be obtained with satisfied enantioselectivities in the attempted three different solvents respectively. To our delight, when ethyl acetate was used as the solvent, 5a can be obtained in \(54\%\) yield and \(93\%\) ee.  
+
+<|ref|>image<|/ref|><|det|>[[246, 305, 740, 440]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[99, 451, 896, 475]]<|/det|>
+<center>Fig. 6 |Barriers to enantiomerization for quinohelicenes. aExperimental value. bThe relative Gibbs free energy (kcal/mol) was calculated at the M06-2X/def2-TZVP/P/B3LYP/def2-SVP lever. </center>  
+
+<|ref|>text<|/ref|><|det|>[[87, 504, 480, 650]]<|/det|>
+In order to investigate the stereochemical stability of the studied chiral quinohelicenes, we carried out the racemization experiment.67 As shown in Fig. 6, the racemic barrier of 5a and 5aj are \(34.9\mathrm{kcal / mol}\) and \(29.1\mathrm{kcal / mol}\) respectively, agreeing well with the calculated \(36.4\mathrm{kcal / mol}\) and \(30.6\mathrm{kcal / mol}\) .66 The reason for the observed 5.8 kcal/mol energy gap between 5a and 5aj is maybe due to the presence of C- sp in 5aj that makes the molecule more flexible. In addition, quinol[6]helicenes 5am exhibited good configurational stability even at \(170^{\circ}\mathrm{C}\) in 1,2- dichlorobenzene for 5 h without obvious racemization.  
+
+<|ref|>text<|/ref|><|det|>[[87, 654, 480, 876]]<|/det|>
+In order to demonstrate the potential application value of the chiral quinohelicenes, the photophysical properties of several selected quinohelicenes were investigated (Table 2). Notably, our quinohelicenes represented good quantum yields \((\Phi_{FL})\) (up to \(26.0\%\) ). Furthermore, we studied the effect of solvent on fluorescence, and found that the fluorescence gradually enhanced with increasing solvent polarity and showed a red shift (Table 2B). And the quantum yield of 5a visibly increased with the increase of solvent polarity from low- polarity dichloromethane \((\Phi_{FL} = 13.9\%)\) to high- polarity dimethyl sulfoxide \((\Phi_{FL} = 43.4\%)\) .66 As quinine containing helicenes exhibit high proton affinity, we have studied the optical properties of protonated counterpart 5a- H+. With the increase of TFA, the absorption at 240 to \(390\mathrm{nm}\) were gradually weakened, but increased at 400 to \(500\mathrm{nm}\) and the fluorescence emission peak was obviously red shifted (Table 2C).  
+
+<|ref|>text<|/ref|><|det|>[[86, 880, 480, 906]]<|/det|>
+Finally, the chiroptical properties of M- 5a, M- 5ab, M- 5ac, M- 5ae, M- 5ah, M- 5al and M- 5am, were subjected to a  
+
+<|ref|>text<|/ref|><|det|>[[517, 504, 910, 662]]<|/det|>
+preliminary evaluation (Table 2A).67 In the CD spectrums, these quinohelicenes displayed several signals at \(230\mathrm{nm}\) to \(400\mathrm{nm}\) and with fine absorption dissymmetry factors (gabs). The maximum gabs of M- 5ab, M- 5ac, M- 5ae reached 0.0072, 0.0083, and 0.0081 at 390, 393, and 388 nm respectively. The CD spectrum of enantiomer 5a appears as a mirror image, M- /P- 5al showed mirror image also (Table 2D). The CPL spectra of these enantiomers were also measured and the maximum gum of M- 5a, M- 5ab, M- 5ac, M- 5al reached 0.0038, 0.0030, 0.0051 and 0.0030 at 457, 459, 465 and 465 nm respectively. Moreover, M- /P- 5a, and M- /P- 5al also exhibited obvious mirror- image CPL emissions (Table 2E).  
+
+<|ref|>sub_title<|/ref|><|det|>[[517, 674, 633, 688]]<|/det|>
+## CONCLUSION  
+
+<|ref|>text<|/ref|><|det|>[[517, 693, 910, 876]]<|/det|>
+In summary, we have developed a mild and multicomponent organocatalytic enantioselective Povarov/aromatization reaction of benzo[c]phenanthren- 2- amines, benzaldehydes and 3- vinyl- 1H- indoles provide a reliable tool for the preparation of highly functionalized quinohelicenes. The wide substrate scope, good yields and high enantioselectivities make this a promising method for preparing chiral helicenes, which are widely used in material science, molecular recognition, and asymmetric catalysis. Furthermore, this three- step reaction could carried out in a one- pot procedure. Moreover, the synthesized quinohelicenes showed enrich photophysical properties, including solvent and acid effects on UV and fluorescence, circular dichroic absorption (CD) and circular polarization luminescence (CPL) properties.
+
+<--- Page Split --->
+<|ref|>table<|/ref|><|det|>[[100, 70, 833, 234]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[100, 48, 536, 64]]<|/det|>
+Table 2 |Photophysical properties of selected quinohelicens a   
+
+<table><tr><td>A:</td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>Compound</td><td>b&amp;amp;abs. max(nm)</td><td>a&amp;amp;emi ((a&amp;amp;x)(nm))</td><td>d&amp;amp;F(%) (a&amp;amp;x(nm))</td><td>g&amp;amp;abs = Δε/ε (a&amp;amp;ab(nm))</td><td>g&amp;amp;lum (a&amp;amp;em/Δ&amp;amp;x(nm))</td></tr><tr><td>M-5a</td><td>267,325,420</td><td>440,450(325)</td><td>13.9(325)</td><td>0.0016(305), -0.0056(388)</td><td>0.0038(457/325)</td></tr><tr><td>M-5ab</td><td>269,325,420</td><td>438,457(325)</td><td>10.3(325)</td><td>0.0020(310), -0.0072(390)</td><td>0.0030(459/325)</td></tr><tr><td>M-5ac</td><td>266,330,425</td><td>440,461(330)</td><td>14.8(330)</td><td>0.0008(315), -0.0083(393)</td><td>0.0051(465/330)</td></tr><tr><td>M-5ae</td><td>266,315,420</td><td>460(315)</td><td>9.8(315)</td><td>-0.0026(363), 0.0081(388)</td><td>0.0024(468/315)</td></tr><tr><td>M-5ah</td><td>265,340,424</td><td>436,458(340)</td><td>26.0(340)</td><td>0.0024(301), -0.0065(392)</td><td>0.0015(460/340)</td></tr><tr><td>M-5al</td><td>216,320,418</td><td>458(320)</td><td>11.3(320)</td><td>-0.0017(260), 0.0045(370)</td><td>0.0030(465/320)</td></tr><tr><td>M-5am</td><td>257,313,412</td><td>436(313)</td><td>25.7(313)</td><td>-0.0036(275), 0.0048(379)</td><td>0.0013(477/313)</td></tr></table>  
+
+<|ref|>image<|/ref|><|det|>[[120, 240, 850, 640]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[100, 656, 901, 682]]<|/det|>
+<center>Measured at \(1\times 10^{-5}\mathrm{M}\) in DCM. Maximum UV absorption. Maximum fluorescence wavelength. Absolute fluorescence quantum efficiency under absorbance lower than 0.1. Measured with \(99\%\) ee. </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[88, 726, 217, 741]]<|/det|>
+## Data availability  
+
+<|ref|>text<|/ref|><|det|>[[88, 744, 480, 850]]<|/det|>
+The authors declare that the data supporting the findings of this study are available within the article and its Supplementary Information files. Extra data are available from the author upon reasonable request. Crystallographic data for the structures reported in this Article have been deposited at the Cambridge Crystallographic Data Centre, under deposition numbers CCDC 2174258 (5a) and 2212769 (5al). Copies of the data can be obtained free of charge via https:  
+
+<|ref|>sub_title<|/ref|><|det|>[[88, 860, 206, 875]]<|/det|>
+## REFERENCES  
+
+<|ref|>text<|/ref|><|det|>[[88, 878, 480, 904]]<|/det|>
+1. Shen, Y. & Chen, C. F. Helicenes: synthesis and applications. Chem. Rev. 112. 1463-1535 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[515, 714, 912, 916]]<|/det|>
+2. Gingras, M. One hundred years of helicene chemistry. Part 3: applications and properties of carbochelicenes. Chem. Soc. Rev. 42. 1051-1095 (2013). 
+3. Saleh, N., Shen, C. & Crassous, J. Helicene-based transition metal complexes: synthesis, properties and applications. Chem. Sci. 5. 3680-3694 (2014). 
+4. Zhao, W. L., Li, M., Lu, H. Y. & Chen, C. F. Advances in helicene derivatives with circularly polarized luminescence. Chem. Commun. 55. 13793-13803 (2019). 
+5. Dhabibi, K., Favereau, L. & Crassous, J. Enantioenriched Helicenes and Heliconoids Containing Main-Group Elements (B, Si, N, P). Chem. Rev. 119. 8846-8953 (2019). 
+6. Mori, T. Chiroptical Properties of Symmetric Double, Triple, and Multiple Helicenes. Chem. Rev. 121. 2373-2412 (2021). 
+7. Takenaka, N., Sarangthem, R. S. & Captain, B. Helical chiral pyridine N-oxides: a new family of asymmetric catalysts. Angew. Chem. Int. Ed. 47. 9708-9710 (2008).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[85, 42, 483, 920]]<|/det|>
+8. Norito Takenaka, J. C., Burjor Captain, Robindro Singh Sarangthem, and Appayee Chandrakumar. Helical Chiral 2-Aminopyridinium lons: A New Class of Hydrogen Bond Donor Catalysts. J. Am. Chem. Soc. 132. 4536-4537 (2010).9. Kos, M., et al. Enantioenriched Ruthenium-Tris-Bipyridine Complexes Bearing One Helical Bipyridine Ligand: Access to Fused Multihalecinic Systems and Chiroptical Redox Switches. Inorg. Chem. 60. 11838-11851 (2021).10. Sakamoto, D., et al. Cycloindicated Helicenes as Chiral Catalysts in the Asymmetric Transfer Hydrogenation of Imines. ACS Catal. 12. 10793-10800 (2022).11. Liang, Y., et al. Enhancement of electrocatalytic oxygen evolution by chiral molecular functionalization of hybrid 2D electrodes. Nat. Commun. 13. 3356 (2022).12. Misek, J., et al. A straightforward route to helically chiral N-heteroaromatic compounds: practical synthesis of racemic 1,14-diazal[5]helicene and optically pure 1- and 2-aza[6]helicenes. Angew. Chem. Int. Ed. 47. 3188-3191 (2008).13. Zadny, J., et al. A general approach to optically pure [5]-,[6]- and [7]heterohelicenes. Angew. Chem. Int. Ed. 51. 5857-5861 (2012).14. Otani, T., et al. Facile Two-Step Synthesis of 1,10-Phenanthroline-Derived Polyaza[7]helicenes with High Fluorescence and CPL Efficiency. Angew. Chem. Int. Ed. 56. 3906-3910 (2017).15. Li, J. K., et al. B,N-Embedded Double Hetero[7]helicenes with Strong Chiroptical Responses in the Visible Light Region. J. Am. Chem. Soc. 143. 17958-17963 (2021).16. Shen, C., et al. Oxidative cyclo-rearrangement of helicenes into chiral nanographenes. Nat. Commun. 12. 2786 (2021).17. Nakakuki, Y., et al. Doubly linked chiral phenanthrene oligomers for homogeneously \(\pi\) -extended helicenes with large effective conjugation length. Nat. Commun. 13. 1475 (2022).18. I. G. Stará, I. S., A. Kollárovic, F. Teplý, S. Vyskočil, D. Šaman., Transition Metal Catalysed Synthesis of Tetrahydro Derivatives of [5]-,[6]- and [7]Helicene. Tetrahedron Lett. 40. 1993-1996 (1999).19. Stará, I. G. & Starý, I. Helically Chiral Aromatics: The Synthesis of Helicenes by [2 + 2 + 2] Cycloisomerization of \(\pi\) -Electron Systems. Acc. Chem. Res. 53. 144-158 (2020).20. Liu, W., Qin, T., Xie, W. & Yang, X. Catalytic Enantioselective Synthesis of Helicenes. Chem. Eur. J., e202202369 (2022).21. Ken Tanaka, A. K., Takeshi Suda, Keiichi Noguchi, and Masao Hirano. Rh-Catalyzed Synthesis of Helically Chiral and Ladder-Type Molecules via [2 + 2 + 2] and Formal [2 + 1 + 2 + 1] Cycloadditions Involving C-C Triple Bond Cleavage. J. Am. Chem. Soc. 129. 12078-12079 (2007).22. Tanaka, K., Fukawa, N., Suda, T. & Noguchi, K. One-step construction of five successive rings by rhodium-catalyzed intermolecular double [2+2+2] cycloaddition: enantiomeriched [9]helicene-like molecules. Angew. Chem. Int. Ed. 48. 5470-5473 (2009).23. Fukawa, N., Osaka, T., Noguchi, K. & Tanaka, K. Asymmet ric Synthesis and Photophysical Properties of Benzopyrano- or Naphthopyrano-Fused Helical Phosphafluorenes. Org. Lett. 12. 1324-1327 (2010).24. Jancarik, A., et al. Rapid access to dibenzohelicenes and their functionalized derivatives. Angew. Chem. Int. Ed. 52. 9970-9975 (2013).25. Kimura, Y., Fukawa, N., Miyauchi, Y., Noguchi, K. & Tanaka, K. Enantioselective synthesis of [9]- and [11]helicene-like molecules: double intramolecular [2+2+2] cycloaddition. Angew. Chem. Int. Ed. 53. 8480-8483 (2014).26. Samal, M., et al. An Ultimate Stereocontrol in Asymmetric Synthesis of Optically Pure Fully Aromatic Helicenes. J. Am. Chem. Soc. 137. 8469-8474 (2015).27. Kimura, Y., Shibata, Y., Noguchi, K. & Tanaka, K. Enantioselective Synthesis and Epimerization Behavior of a Chiral S - Shaped [11]Helicene - Like Molecule Having Collision between Terminal Benzene Rings. Eur. J. Org. Chem. 2019. 1390-1396 (2019).28. Kinoshita, S., et al. Rhodium-Catalyzed Highly Diastereo- and Enantioselective Synthesis of a Configurationally Stable S- Shaped Double Helicene-Like Molecule. Angew. Chem. Int. Ed. 59. 11020-11027 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[512, 45, 911, 920]]<|/det|>
+29. Nakamura, K., Furumi, S., Takeuchi, M., Shibuya, T. & Tanaka, K. Enantioselective synthesis and enhanced circularly polarized luminescence of S-shaped double azalehicenes. J. Am. Chem. Soc. 136. 5555-5558 (2014).  
+30. Tanaka, M., et al. Gold-Catalyzed Enantioselective Synthesis, Crystal Structure, and Photophysical/Chiroptical Properties of Aza[10]helicenes. Chem. Eur. J. 22. 9537-9541 (2016).  
+31. Gonzalez-Fernandez, E., Nicholls, L. D., Schaaf, L. D., Fares, C., Lehmann, C. W. & Alcarazo, M. Enantioselective Synthesis of [6]Carbochelicenes. J. Am. Chem. Soc. 139. 1428-1431 (2017).  
+32. Yamano, R., Kinoshita, S., Shibata, Y. & Tanaka, K. Substituent - Controlled and Rhodium - Catalyzed Intramolecular [2+2+2] and [2+1+2+1] Cycloadditions of Electron - Deficient Triynes. Eur. J. Org. Chem. 2018. 5916-5920 (2018).  
+33. Usui, K., et al. Internal-Edge-Substituted Coumarin-Fused [6]Helicenes: Asymmetric Synthesis, Structural Features, and Control of Self-Assembly. Chem. Eur. J. 24. 14617-14621 (2018).  
+34. Hartung, T., Machleid, R., Simon, M., Golz, C. & Alcarazo, M. Enantioselective Synthesis of 1,12-Disubstituted [4]Helicenes. Angew. Chem. Int. Ed. 59. 5660-5664 (2020).  
+35. Sako, M., et al. fficient Enantioselective Synthesis of Oxahiclenes Using Redox Acid Cooperative Catalysts. J. Am. Chem. Soc. 138. 11481-11484 (2016).  
+36. Yubuta, A., et al. Enantioselective Synthesis of Triple Helicenes by Cross-Cyclotrimerization of a Helicenyi Aryne and Alkynes via Dynamic Kinetic Resolution. J. Am. Chem. Soc. 142. 10025-10033 (2020).  
+37. Wang, Q., Zhang, W. W., Zheng, C., Gu, Q. & You, S. L. Enantioselective Synthesis of Azoniahelicenes by Rh-Catalyzed C-H Annulation with Alkynes. J. Am. Chem. Soc. 143. 114-120 (2021).  
+38. Yu, T., et al. Palladium-Catalyzed Modular Synthesis of Enantioenriched Pyridohelicenes through Double Imidoylative Cyclization. ACS Catal. 12. 13034-13041 (2022).  
+39. Kotzner, L., Webber, M. J., Martinez, A., De Fusco, C. & List, B. Asymmetric catalysis on the nanoscale: the organocatalytic approach to helicenes. Angew. Chem. Int. Ed. 53. 5202-5205 (2014).  
+40. Jia, S., Li, S., Liu, Y., Qin, W. & Yan, H. Enantioselective Control of Both Helical and Axial Stereogenic Elements though an Organocatalytic Approach. Angew. Chem. Int. Ed. 58. 18496-18501 (2019).  
+41. Li, K., Huang, S., Liu, T., Jia, S. & Yan, H. Organocatalytic Asymmetric Dearomatizing Hetero-Diels-Alder Reaction of Nonactivated Arenes. J. Am. Chem. Soc. 144. 7374-7381 (2022).  
+42. Liu, P., et al. Simultaneous Control of Central and Helical Chiralities: Expedient Helicoselective Synthesis of Dioxa[6]helicenes. J. Am. Chem. Soc. 142. 16199-16204 (2020).  
+43. Kaneko, E., Matsumoto, Y. & Kamikawa, K. Synthesis of AzaheliceneN-Oxide by Palladium-Catalyzed Direct C H Annulation of a Pendant (Z)-Bromovinyl Side Chain. Chem. Eur. J. 19. 11837-11841 (2013).  
+44. Yang, Y., da Costa, R. C., Fuchter, M. J. & Campbell, A. J. Circularly polarized light detection by a chiral organic semiconductor transistor. Nature Pha. 7. 634-638 (2013).  
+45. Yang, Y., da Costa, R. C., Smilgies, D. M., Campbell, A. J. & Fuchter, M. J. Induction of circularly polarized electroluminescence from an achiral light-emitting polymer via a chiral small-molecule dopant. Adv. Mater. 25. 2624-2628 (2013).  
+46. Abbate, S., et al. Helical Sense-Responsive and Substituent-Sensitive Features in Vibrational and Electronic Circular Dichroism, in Circularly Polarized Luminescence, and in Raman Spectra of Some Simple Optically Active Hexahelicenes. J. Phys. Chem. C 118. 1682-1695 (2014).  
+47. Klivar, J., et al. [2+2+2] Cycloisomerisation of Aromatic Cyanodiynes in the Synthesis of Pyridohelicenes and Their Analogues. Chem. Eur. J. 22. 14401-14405 (2016).  
+48. Feng, J., Wang, L., Xue, X., Chao, Z., Hong, B. & Gu, Z. Ring-Expansion Strategy for alpha-Aryl Azahelicene Construction:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[85, 42, 483, 940]]<|/det|>
+Building Blocks for Optoelectronic Materials. Org. Lett. 23. 8056- 8061 (2021).49. Yen- Pon, E., et al. Heterohelicenes through 1,3- Dipolar Cycloaddition of Syndones with Arynes: Synthesis, Origins of Selectivity, and Application to pH- Triggered Chiroptical Switch with CPL Sign Reversal. JACS Au 1. 807- 818 (2021).50. Gong, X., Li, C., Cai, Z., Wan, X., Qian, H., & Yang, G. Synthesis of Nitrogen- Doped Aza- Helicenes with Chiral Optical Properties. J. Org. Chem. 87. 8406- 8412 (2022).51. Chen, J. & Takenaka, N. Helical chiral pyridine N- oxides: a new family of asymmetric catalysts. Chem. Eur. J. 15. 7268- 7276 (2009).52. Takenaka, N., Chen, J. S., Captain, B., Sarangthem, R. S., & Chandrakumar, A. Helical Chiral 2- Aminopyridinium Ions: A New Class of Hydrogen Bond Donor Catalysts. J. Am. Chem. Soc. 132. 4536- 4537 (2010).53. Huang, Q., et al. Inherently Chiral Azonia[6]helicene- Modified beta- Cyclodextrin: Synthesis, Characterization, and Chirality Sensing of Underivatized Amino Acids in Water. J. Org. Chem. 81. 3430- 3434 (2016).54. Akiyama, T., Morita, H., & Fuchibe, K. Chiral Bronsted acid- catalyzed inverse electron- demand aza Diels- Alder reaction. J. Am. Chem. Soc. 128. 13070- 13071 (2006).55. Xie, M., et al. Asymmetric three- component inverse electron- demand aza- Diels- Alder reaction: efficient synthesis of ring- fused tetrahydroquinolines. Angew. Chem. Int. Ed. 49. 3799- 3802 (2010).56. Dagousset, G., Zhu, J., & Masson, G. Chiral phosphoric acid- catalyzed enantioselective three- component Povarov reaction using encarbamates as dienophiles: highly diastereo- and enantioselective synthesis of substituted 4- aminotetrahydroquinolines. J. Am. Chem. Soc. 133. 14804- 14813 (2011).57. Chen, Z., Wang, B., Wang, Z., Zhu, G., & Sun, J. Complex bioactive alkaloid- type polycycles through efficient catalytic asymmetric multicomponent aza- Diels- Alder reaction of indoles with oxetane as directing group. Angew. Chem. Int. Ed. 52. 2027- 2031 (2013).58. Yu, J., Jiang, H. J., Zhou, Y., Luo, S. W., & Gong, L. Z. Sodium salts of anionic chiral cobalt(III) complexes as catalysts of the enantioselective Povarov reaction. Angew. Chem. Int. Ed. 54. 11209- 11213 (2015).59. Bisag, G. D., et al. Central- to- Axial Chirality Conversion Approach Designed on Organocatalytic Enantioselective Povarov Cycloadditions: First Access to Configurationally Stable Indole- Quinoline Atropisomers. Chem. Eur. J. 25. 15694- 15701 (2019).60. Wang, S. J., Wang, Z., Tang, Y., Chen, J., & Zhou, L. Asymmetric Synthesis of Quinoline- Naphthalene Atropisomers by Central- to- Axial Chirality Conversion. Org. Lett. 22. 8894- 8898 (2020).61. Clerigue, J., Ramos, M. T., & Menendez, J. C. Enantioselective catalytic Povarov reactions. Org. Biomol. Chem. 20. 1550- 1581 (2022).62. Lemos, B. C., Venturini Filho, E., Fiorot, R. G., Medici, F., Greco, S. J., & Benaglia, M. Enantioselective Povarov Reactions: An Update of a Powerful Catalytic Synthetic Methodology. Eur. J. Org. Chem. e202101171 (2022).63. Viglianisi, C., et al. Synthesis of Heterohelicenes by a Catalytic Multi- Component Povarov Reaction. Eur. J. Org. Chem. 2019. 164- 167 (2019).64. Soni, R., & Soman, S. S. Metal free synthesis of Coumarin containing hetero n helicene like molecules with TICT and AIE properties. Asian. J. Org. Chem. 11. e202100770 (2022).65. In the literature, although there have been two re- ports on the construction of helicenes through Pova- rov reaction, both of them are racemic version. More- over, Lewis acid catalysis is used, which suffered from high catalyst loading and harsh conditions.66. DFT Calculations on the Enantiomerization Process of 1- H- quinol[5]helicene and 1- (3- indol)- quinol[5]helicene. The relative Gibbs free energy (kcal/mol) was calculated at the M06- 2X/def2- TZVPP//B3LYP/def2- SVP lever. For details see Supporting Information.67. For details see Supporting Information.68. Yu, J., Zhang, P., Wu, J., & Shang, Z. Metal- free C- N bond- forming reaction: straightforward synthesis of anilines, through cleavage of aryl C- O bond and amide C- N bond. Tetrahedron Lett. 54. 3167- 3170 (2013).69. Wu, J., Yu, J., Wang, Y., & Zhang, P. Direct Animation of Phenols under Metal- Free Conditions. Synlett 24. 1448- 1454 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[515, 43, 911, 101]]<|/det|>
+cleavage of aryl C- O bond and amide C- N bond. Tetrahedron Lett. 54. 3167- 3170 (2013).69. Wu, J., Yu, J., Wang, Y., & Zhang, P. Direct Animation of Phenols under Metal- Free Conditions. Synlett 24. 1448- 1454 (2013).  
+
+<|ref|>sub_title<|/ref|><|det|>[[518, 115, 680, 130]]<|/det|>
+## Acknowledgements  
+
+<|ref|>text<|/ref|><|det|>[[515, 135, 910, 210]]<|/det|>
+We are grateful to the National Natural Science Foundation of China (Grant Nos. 22193011, 21971120 and 21933008) and National Science & Technology Fundamental Resource Investigation Program of China (No. 2018FY201200) for financial support.  
+
+<|ref|>sub_title<|/ref|><|det|>[[517, 215, 689, 230]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[515, 234, 910, 320]]<|/det|>
+W. L. conceived and designed the study and performed experiments and wrote the manuscript; G. X. synthesized some substrates; Z. R. performed the theoretical calculations; H.G and M. L. performed the CPL experiments; X. L. conceived and designed the study and wrote the manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[517, 326, 684, 341]]<|/det|>
+## Competing interests  
+
+<|ref|>text<|/ref|><|det|>[[517, 346, 852, 360]]<|/det|>
+The authors declare no competing financial interest  
+
+<|ref|>sub_title<|/ref|><|det|>[[517, 364, 702, 379]]<|/det|>
+## Additional information  
+
+<|ref|>text<|/ref|><|det|>[[517, 384, 894, 398]]<|/det|>
+Supplementary information is available for this paper at  
+
+<|ref|>text<|/ref|><|det|>[[515, 422, 910, 450]]<|/det|>
+Correspondence and requests for materials should be addressed to Xin Li.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[43, 42, 312, 71]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[43, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 525, 177]]<|/det|>
+- 5alcheckcif.pdf- SupportingInformationNat.commun.20221130.pdf
+
+<--- Page Split --->

preprint/preprint__485373ce90d30c4730644ea71cd7df747a03d71497007b34dcac392a1a08f4a3/images_list.json

@@ -0,0 +1,92 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1. Association of socio-demographic and clinical characteristics with systemic immune-oncological proteins in Ghanaian (Af), AA, and EA men without prostate cancer. Association of the 82 immuno-oncological proteins with age, BMI, education, aspirin use, smoking, diabetes, and PSA was assessed in men without prostate cancer using a multivariable linear regression model. An analyte was considered significantly associated with clinical and socio-demographic covariables if the multivariable model yielded a \\(P<0.05\\) on the F-statistic. Analytes that did not have a significant association with any of the clinical/sociodemographic variables in at least one of the population groups are not presented in the heatmap. Blue represents negative association while red represents positive association. The significance level (P value-based) for each association is color-coded. TI = tumor immunity.",
+    "footnote": [],
+    "bbox": [
+      [
+        55,
+        70,
+        960,
+        789
+      ]
+    ],
+    "page_idx": 33
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2. Unsupervised hierarchical clustering associates circulating immune-oncological proteome profiles with population groups - Ghanaian (Af), AA, and EA men. Heatmap showing protein profiles for men without prostate cancer. Each row represents a protein (n=82), and each column corresponds to an individual [n=1482 (654 Af, 374 AA, and 454 EA)]. Each individual is color-coded as Af, AA, or EA in the annotation bar on top of the heatmap. Normalized z-score of proteins abundance are depicted on a low-to-high scale (blue-white-red).",
+    "footnote": [],
+    "bbox": [
+      [
+        60,
+        180,
+        940,
+        625
+      ]
+    ],
+    "page_idx": 34
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3. Immune-oncological proteins and their relationship with West-African ancestry. (A) Variance analysis for the levels of each of the 82 immune-oncological cytokines assessed as a function of genetic estimation of West African admixture among men without prostate cancer within the NCI-Maryland study. The blue plot represents the proportion of variance that can be explained by the degree of West-African admixture while the grey plot represents the residual variance that remains to be explained by other factors other than West-African ancestry. (B) The median levels of the top six West-African ancestry correlated immune-oncological proteins were compared between Af, AA, and EA. Error bars represent inter quartile range (IQR). Linearized protein abundances ( \\(2^{Npx}\\) ) were used to determine median and IQR for each of the proteins.",
+    "footnote": [],
+    "bbox": [
+      [
+        66,
+        50,
+        833,
+        789
+      ]
+    ],
+    "page_idx": 35
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4. Population differences in the abundance of proteins driving (A) apoptosis, (B) autophagy, (C) chemotaxis, (D) promotion of tumor immunity, (E) suppression of tumor immunity, and (F) vasculature. Heatmaps showing levels of process/pathway-associated proteins in relationship to population group [Ghanaian (Af), AA, EA]. Shown to the right are the mean score differences for these processes/pathways among the three population groups. Profiles for Ghanaian (n=654), AA (n=374), and EA (n=454) men without prostate cancer. The process/pathway scores are derived from the average Z-scores of all the associated proteins. These scores are shown as violin plots. TI = tumor immunity.",
+    "footnote": [],
+    "bbox": [
+      [
+        66,
+        37,
+        760,
+        856
+      ]
+    ],
+    "page_idx": 36
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig. 5",
+    "footnote": [],
+    "bbox": [
+      [
+        95,
+        12,
+        722,
+        750
+      ]
+    ],
+    "page_idx": 37
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig. 6",
+    "footnote": [],
+    "bbox": [
+      [
+        80,
+        25,
+        800,
+        800
+      ]
+    ],
+    "page_idx": 38
+  }
+]

preprint/preprint__485373ce90d30c4730644ea71cd7df747a03d71497007b34dcac392a1a08f4a3/preprint__485373ce90d30c4730644ea71cd7df747a03d71497007b34dcac392a1a08f4a3.mmd

@@ -0,0 +1,339 @@
+
+# Serum proteomics links suppression of tumor immunity to ancestry and lethal prostate cancer  
+
+Tsion Minas  Center for Cancer Research  
+
+Julian Candia  National Cancer Institute  https://orcid.org/0000- 0001- 5793- 8989  
+
+Tiffany Dorsey  National Cancer Institute, NIH, Bethesda, MD  
+
+Francine Baker  Laboratory of Translational Genomics, Division of Cancer Epidemiology and Genetics, National Cancer Institute (NCI), National Institutes of Health (NIH)  https://orcid.org/0000- 0003- 3133- 3652  
+
+Wei Tang  NCI/NIH  https://orcid.org/0000- 0002- 7089- 4391  
+
+Maeve Kiely  Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  
+
+Cheryl Smith  Center for Cancer Research  
+
+Symone Jordan  Center for Cancer Research  
+
+Obadi Obadi  Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  
+
+Anuoluwapo Ajao  Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  
+
+Yao Tettey  University of Ghana Medical School  
+
+Richard Biritwum  University of Ghana Medical School  
+
+Andrew Adjei  University of Ghana Medical School  
+
+James Mensah  University of Ghana Medical School  
+
+Robert Hoover
+
+<--- Page Split --->
+
+National Cancer Institute  
+
+Frank Jenkins  Hillman Cancer Center, University of Pittsburgh  
+
+Rick Kittles  City of Hope  https://orcid.org/0000- 0002- 5004- 4437  
+
+Ann Hsing  Stanford  
+
+Xin Wang  
+
+Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  
+
+Christopher Loffredo  Georgetown University  
+
+Clayton Yates  Tuskegee University  https://orcid.org/0000- 0001- 5420- 9852  
+
+Michael Cook  National Cancer Institute  
+
+Stefan Ambs ( ambss@mail.nih.gov )  
+
+Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  https://orcid.org/0000- 0001- 7651- 9309  
+
+## Article  
+
+Keywords: Proteomics, inflammation, prostate cancer, ancestry, survival, disparity, immune signature, African, European  
+
+Posted Date: July 13th, 2021  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 668276/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Communications on April 1st, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 29235- 2.
+
+<--- Page Split --->
+
+# Serum proteomics links suppression of tumor immunity to ancestry and lethal prostate cancer (13 words, 92 characters)  
+
+T sion Zewdu Minas \(^{1\dagger}\) , Julián Candia \(^{1\dagger}\) , Tiffany H. Dorsey \(^{1}\) , Francine Baker \(^{1}\) , Wei Tang \(^{1}\) , Maeve Kiely \(^{1}\) , Cheryl J. Smith \(^{1}\) , Symone V. Jordan \(^{1}\) , Obadi M. Obadi \(^{1}\) , Anuoluwapo Ajao \(^{1}\) , Yao Tetty \(^{2}\) , Richard B. Biritwum \(^{2}\) , Andrew A. Adjei \(^{2}\) , James E. Mensah \(^{2}\) , Robert N. Hoover \(^{3}\) , Frank J. Jenkins \(^{4}\) , Rick Kittles \(^{5}\) , Ann W. Hsing \(^{6,7}\) , Xin W. Wang \(^{1,8}\) , Christopher A. Loffredo \(^{9}\) , Clayton Yates \(^{10}\) , Michael B. Cook \(^{3}\) , and Stefan Ambs \(^{1*}\)  
+
+\(^{1}\) Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA \(^{2}\) University of Ghana Medical School, Accra, Ghana \(^{3}\) Division of Cancer Epidemiology & Genetics, National Cancer Institute, National Institutes of Health, MD, USA \(^{4}\) Hillman Cancer Center, University of Pittsburgh, Pittsburgh, PA, USA \(^{5}\) Division of Health Equities, Department of Population Sciences, City of Hope Comprehensive Cancer Center, Duarte, CA, USA \(^{6}\) Stanford Cancer Institute, Stanford School of Medicine, Palo Alto, CA, USA \(^{7}\) Stanford Prevention Research Center, Stanford School of Medicine, Palo Alto, CA, USA \(^{8}\) Liver Cancer Program, Center for Cancer Research, National Cancer Institute, Natioinal Institutes of Health, Bethesda, MD, USA \(^{9}\) Cancer Prevention and Control Program, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, Washington, DC, USA \(^{10}\) Center for Cancer Research, Tuskegee University, Tuskegee, AL, USA \(^{1*}\) These authors contributed equally  
+
+Running title: Immune- oncological markers and prostate cancer disparity  
+
+Key words: Proteomics, inflammation, prostate cancer, ancestry, survival, disparity, immune signature, African, European  
+
+Abbreviations: AA, African- American; EA, European- American; OR, odds ratio; CI, confidence interval; PSA, prostate- specific antigen.  
+
+\*Corresponding Author: Stefan Ambs, Laboratory of Human Carcinogenesis, National Cancer Institute, Bldg.37/Room 3050B, Bethesda, MD 20892- 4258, Phone: 240- 760- 6836; Email: ambss@mail.nih.gov.  
+
+One sentence summary: A serum proteome- based immune function signature is upregulated in men of African ancestry and associates with lethal prostate cancer.
+
+<--- Page Split --->
+
+## 40 Abstract (156 words):  
+
+There is evidence that tumor immunobiology and immunotherapy response may differ between African American and European American prostate cancer patients. Here, we determined if men of African descent harbor a unique systemic immune- oncological signature and measured 82 circulating proteins in almost 3000 Ghanaian, African American, and European American men. Protein signatures for suppression of tumor immunity and chemotaxis were significantly elevated in men of West African ancestry. Importantly, the suppression of tumor immunity protein signature associated with metastatic and lethal prostate cancer, pointing to clinical significance. Moreover, two markers, pleiotrophin and TNFRSF9, predicted poor disease survival specifically among African American men. These findings indicate that immune- oncology marker profiles differ between men of African and European descent. These differences may contribute to the disproportionate burden of lethal prostate cancer in men of African ancestry. The elevated peripheral suppression of tumor immunity may have important implication for guidance of cancer therapy which could particularly benefit African American patients.
+
+<--- Page Split --->
+
+Men of African origin bear the highest prostate cancer burden in the U.S. and globally<sup>1-3</sup>. They are at an increased risk of developing fatal prostate cancer in the U.S and England<sup>4</sup> and present with more aggressive disease in the Caribbean and sub- Saharan Africa<sup>2,5</sup>. The reasons for the observed global prostate cancer health disparities are unclear but may be related to an array of factors such as access to health care, lifestyle and environment, and ancestral and biological factors<sup>6-8</sup>.  
+
+Previously, we and others described that tumor immunobiology differs between African- American (AA) and European- American (EA) prostate cancer patients<sup>9-12</sup>. A tumor- specific immune- inflammation gene expression signature was more prevalent in prostate tumors of AA than EA patients<sup>11</sup>. The occurrence of this signature in prostate tumors was associated with decreased recurrence- free survival<sup>13</sup>. Furthermore, regular use of aspirin, an anti- inflammatory drug, may reduce the risk of aggressive prostate cancer, disease recurrence and the lethal disease in AA men<sup>14,15</sup>. Combined, these findings suggest that inflammation and host immunity may contribute to prostate cancer progression but with notable differences between AA and EA men.  
+
+Ancestral factors can influence immune- related pathways<sup>16</sup>. Germline genetic variant prevalence and alternative splicing in immune- inflammation- related genes can show large differences amongst population groups<sup>17-19</sup>. Hence, the immune- inflammation gene expression signature identified in the tumors of AA prostate cancer patients could be due to either tumor biology and the associated microenvironment, ancestral factors, or systemic differences in immunology marker expression. In the present study, we tested the hypothesis that a distinct systemic immune- inflammation signature exists in men of African ancestry that associates with prostate cancer. It is the novelty of our approach that we examined the serum proteome in a large cohort of
+
+<--- Page Split --->
+
+77 diverse men. Applying large- scale proteomics with Olink technology, we discovered the up- regulation of circulating immune- oncological proteins that functionally relate to chemotaxis and suppression of tumor immunity and their association with West African ancestry and lethal prostate cancer. Our findings point to the clinical importance of a serum proteomic signature in prostate cancer patients that may affect men of African ancestry more so than other men.
+
+<--- Page Split --->
+
+## RESULTS  
+
+Large- scale evaluation of immune- oncological proteins in the NCI- Maryland and NCI- Ghana prostate cancer studies. To investigate if men of African descent are differentially affected by a systemic immune inflammation, we utilized two case- control studies with large representations of men of African ancestry: the NCI- Ghana and NCI- Maryland Prostate Cancer Case- Control Studies. Characteristics of the participants in the two studies have been previously described<sup>14,20</sup>. We assayed 92 circulating immune- oncological proteins in a total of 3094 serum samples containing 1505 controls and 1432 cases along with 157 randomly selected blinded duplicates. To control for any batch effects, the serum samples were assayed in a random order along with the 5% blind duplicates for intensity normalization (see Methods). Ninety- five percent of the samples passed stringent quality control leaving 1482 controls (654 Ghanaian, 374 AA, and 454 EA) and 1308 cases (489 Ghanaian, 394 AA, and 425 EA) for our analysis (table S1). The average intra- and inter- plate CV calculated based on duplicates were very low at 1.7% and 2.6%, respectively. In addition, the proportion of variance explained by an inter- plate batch effect was rather minimal for each of the serum proteins even before intensity normalization (fig. S1). Out of the 92 serum proteins, 61 were detected in 100% of the samples tested and 78 were detected in > 50% of the samples (fig. S2). Because 10 out of the 92 serum proteins were detected in less than 20% of the samples (fig. S2), only the remaining 82 proteins were included in our analysis (table S2). Next, we assessed how the 82 serum markers correlate with one another in Ghanaian, AA, and EA men without prostate cancer using Pearson's pairwise correlation analysis (fig. S3). The top ten observed correlations for each population group is presented in table S3. Most of these relationships have not previously been described. Most notably, epidermal growth factor levels strongly correlated with CD40L [Ghanaian (r=0.71), AA (r=0.83), and EA (r=0.80) men], a marker
+
+<--- Page Split --->
+
+of activated T cells, whereas IL8 levels highly correlated with circulating caspase 8 in all three population groups [Ghanaian (r=0.69), AA (r=0.82), and EA (r=0.80)].  
+
+Clinical and socio- demographic characteristics are associated with immune- oncological proteins. Cytokine levels can be influenced by environmental exposures and disease. Therefore, we investigated the association between various socio- demographic and clinical characteristics (age, BMI, education, aspirin use, smoking, diabetes and PSA) with serum levels of immunoonological proteins using a multivariable linear regression model (Fig. 1). We restricted this analysis to the control population in the NCI- Ghana and NCI- Maryland studies to exclude the potential confounding effect of prostate cancer in the analysis. Among the exposures, aspirin use and blood PSA levels showed only few relationships with the profile of the 82 immune- oncology markers. Other exposures and several demographics showed more robust relationships.  
+
+Aging is known to impact the immune system and is a risk factor for many diseases including cancer<sup>21</sup>. In our analysis, aging was most consistently associated with the level of the analytes across the three population groups, showing a significant correlation with almost half of these circulating immune- oncological proteins. For example, PGF, CXCL9, Gal9, Gal1, CX3CL1, TNFRSF12A, CCL23, MMP7, DCN, MMP12, CXCL13, CSF1, ADGRG1, CD4, and PTN positively associated with age in all three population groups. The top- ranked biological functions that associated with these age- related proteins were cell migration and positive regulation of cell adhesion (fig. S4A). Age was also positively associated with lymphocyte activation, represented by TNFRSF9, CRTAM, PDCD1, CD27, NCR1, TNFRSF4, KLRD1, CD83, IL12, and IL12RB1, but only in the NCI- Maryland EA and AA men (fig. S4B). On the other hand, hepatocyte growth factor (HGF) and vascular endothelial growth factor- A (VEGFA), two angiogenic cytokines, were positively associated with age exclusively in men of African ancestry (Ghanaian and AA men).
+
+<--- Page Split --->
+
+Lastly, VEGFR2, a tyrosine kinase receptor for VEGF, was negatively associated with age in EA and AA men.  
+
+In contrast to the positive association of many of the immune- oncological proteins with age, BMI tended to be negatively associated with these circulating immune- oncological analytes. This finding may be surprising as obesity is generally thought to be associated with systemic inflammation. CX3CL1 was negatively associated with BMI in all three population groups. The soluble form of CX3CL1 stimulates recruitment of CX3CR1 expressing inflammatory immune cells<sup>22</sup>. CAIX and LAMP3 were inversely associated with BMI exclusively in men of African ancestry, suggesting that ancestral factors may influence the relationship between BMI and expression of these markers. CAIX is a hypoxia regulated metalloenzyme that exists as both membrane associated and soluble form<sup>23</sup> whose main cellular function is to catalyze the reversible conversion of carbon dioxide to carbonic acid<sup>24</sup>, thereby influencing local acidity, which is known to affect immune function<sup>25</sup>. LAMP3 is a member of lysosomal associated membrane glycoprotein family that have a myriad of roles including lysosomal exocytosis and cholesterol homeostasis<sup>26</sup>. On the contrary, serum GAL1, a glycan binding protein that mediates the suppressive function of \(\mathrm{T}_{\mathrm{Reg}}\) cells<sup>27</sup>, showed the opposite trend and was positively associated with BMI in all three population groups.  
+
+To explore how the social/behavioral environment may affect immune- oncological serum protein levels, we investigated their relationship with educational attainment. For Ghanaian men, 27 of the 82 immuno- oncological markers were negatively associated with their education level (Fig. 1). Yet only IL18 showed a significant inverse association with education for both Ghanaian and AA men. Among EA men, 12 of the 82 immune- oncological proteins had significant inverse
+
+<--- Page Split --->
+
+relationships with the attained level of education (Fig. 1), with some of these markers showing a similar pattern among Ghanaian and EA men.  
+
+Previous studies have shown that tobacco smoking increases inflammation<sup>28</sup>. Herein, we assessed the association between cigarette use (never, former, vs. current smoker) on the level of immune- oncological proteins in circulation. We found that current smoking was consistently associated with significantly increased level of analytes that regulate angiogenesis (ANGPT2), antigen presentation (CD83), and autophagy (LAMP3), in all three study populations (Fig. 1).  
+
+Innate immune system- driven inflammatory processes have been implicated in the pathogenesis of diabetes<sup>29</sup>. In our analysis, among the cytokines that showed an association with self- reported diabetes, serum proteins belonging to tumor necrosis factor receptor super family (CD27 and TNFRSF12A), and a matrix metalloprotease enzyme (MMP7) were positively associated with diabetes in all three population groups (Fig. 1). Others, including PGF, CX3CL1, NCR1, TNFRSF4, and TNFRSF21 were positively associated with diabetes exclusively in men with African ancestry. Functional enrichment analysis revealed that diabetes- associated CX3CL1, TNFRSF4, and TNFRSF21 are all involved in negative regulation of cytokine secretion (fig. S5). CX3CL1 is known to regulate insulin secretion<sup>30</sup>, is elevated in the serum of patients with type 2 diabetes<sup>31</sup>, and has been implicated in diabetic nephropathy<sup>32</sup>, validating the findings in our study. C- reactive protein (CRP) is a commonly measured pro- inflammatory marker in the body and has been reported to be associated with worse prostate cancer prognosis<sup>33,34</sup>. Because it was not part of our marker panel, we measured blood CRP in 156 plasma samples from population controls of the NCI- Maryland study. Smoking was the only socio- demographic variable that had a significant association with CRP (table S4), which is consistent with the literature. Furthermore, CRP showed significant positive associations with 24 of the 82 serum proteins (TNFRSF9, IL7,
+
+<--- Page Split --->
+
+PGF, IL6, Gal9, GZMH, CXCL1, TNFSF14, Gal1, PDL1, HGF, HO1, CD70, TNFRSF12A, CCL3, MMP7, ANGPT2, VEGFA, CCL20, KLRD1, CSF1, CD4, MCP3, and CXCL11).  
+
+The systemic immune- oncological cytokine profile in men of African ancestry is distinct from men of European ancestry. To investigate if ancestral population group differences may influence circulating levels of the immune- oncological markers, we performed an unsupervised clustering analysis examining how the levels of the 82 immune- oncological analytes would group men without prostate cancer from Ghana and the US. Notably, these analytes tended to cluster by population group, with levels in Ghanaian men being most distant from EA men while AA samples tended to cluster in between these two groups (Fig. 2), suggesting that the ancestral background may have a significant impact on this immune- oncological protein profile.  
+
+To further evaluate the influence of ancestry, we estimated West African ancestry in AA and EA population controls of the NCI- Maryland study and its relationship with the cytokine profile. West African ancestry was determined using 100 validated ancestry informative markers \(^{35}\) . The approach showed that, to some extent, the variance in the levels of several immune- oncological analytes is strongly influenced by the degree of West African ancestry of these individuals (Fig. 3A). The variance in 45 of the analytes were significantly \((P< 0.05)\) influenced by degree of West African ancestry (table S5). The levels of 42 analytes were significantly accounted for by West African ancestry even after adjusting for age, BMI, aspirin use, education, income, diabetes, and smoking status (table S6). CXCL5, CXCL1, MCP2, MCP1, CXCL11, CCL23, PTN, TWEAK, NCR1, IL18 and CCL17 were the top- ranked proteins (tables S5- S6). Adjusting the significance threshold by Bonferroni \((P_{\mathrm{B}} = 0.05 / 82 = 0.00061)\) , which is the most stringent criterion to adjust for multiple testing, the relationship of the top 28 proteins with West African ancestry remained significant. For instance, \(41\%\) and \(50\%\) of the variance in the serum
+
+<--- Page Split --->
+
+levels of CXCL1 and CXCL5, respectively, was accounted for by the degree of West African ancestry (Fig. 3A and tables S5- S6). When we compared the levels of these proteins across the 3 population groups, we observed a significant African ancestry- related trend (Fig. 3B), with 12 of the 82 circulating immune- oncological proteins (CXCL5, CXCL1, CXCL11, MCP2, CCL17, MCP4, CD70, MMP12, PDL2, MMP7, CCL19, and ANGPT2) being significantly elevated in both Ghanaian and AA men compared to EA men (table S7); twelve other markers (MCP1, IL12, CCL23, CD8A, NCR1, TNFRSF4, TNFSF14, TWEAK, IL7, HGF, HO1, TNFRSF21, and ANG1) were inversely related to West African ancestry (table S8).  
+
+Cytokines associated with suppression of tumor immunity and chemotaxis are upregulated in men of African ancestry. Levels of many of the 82 immune- oncology markers showed a marked association with ancestry. To better define the functional implications of these population group differences, we grouped the 82 proteins into six biological processes according to Olink guidelines (table S9): apoptosis/cell killing, autophagy/metabolism, chemotaxis/trafficking to tumor, suppression of tumor immunity (Th2 response, tolerogenic), promotion of tumor immunity (Th1 responses), or vasculature and tissue remodeling. To gain insight on how activation of these six processes/pathways may differ by population group, we compared process/pathway sum scores between Ghanaian, AA, and EA men without prostate cancer. Of these pathways, chemotaxis, promotion of tumor immunity, and suppression of tumor immunity were significantly different in their predicted activity between AA and EA men (Fig. 4). AA men had significantly higher scores for chemotaxis and suppression of tumor immunity when compared to EA men, indicating higher activity in AA men, but a lower score for promotion of tumor immunity. Ghanaian men had even higher scores for chemotaxis and suppression of tumor immunity than both AA and EA men (Fig. 4C and E), indicating a
+
+<--- Page Split --->
+
+possible association with West African ancestry. The latter was corroborated with our finding that the chemotaxis and suppression of tumor immunity scores positively correlated with the proportion of West African ancestry within the NCI- Maryland cohort (Spearman’s rho=0.23, \(P< 0.001\) , for chemotaxis score; Spearman’s rho=0.15, \(P< 0.001\) , for suppression of immunity score). Even though apoptosis and vasculature- associated cytokines were not significantly different between EA and AA men, we found both processes to be elevated in the Ghanaian men. Suppression of tumor immunity is associated with reduced survival of prostate cancer patients. Next, we examined the clinical implication of our findings and assessed the association of pathway activity with survival of prostate cancer cases or controls in the NCI- Maryland study. As of the end of 2018, out of the 819 cases, there have been 202 deaths in our case population, of whom 103 (51%) had a cancer diagnosis as the recorded primary cause of death, and 28% of all deaths ( \(n = 57\) ) were directly attributed to prostate cancer. On the other hand, 99 of the 828 population controls had died by the end of 2018. Median survival follow- up for cases and controls were 8.6 and 6.7 years, respectively. With these data, we built a multivariable Cox regression model with all biological processes/pathways and adjustment for other covariables (see Methods). Among the six defined pathways, only suppression of tumor immunity showed independent association with survival of cases (Fig. 5). Prostate cancer patients with an increased activity of this pathway had the highest risk of death from all causes (Fig. 5A, table S10). Moreover, prostate cancer patients with elevated suppression of tumor immunity at diagnosis had the highest risk of prostate cancer- specific mortality, albeit marginally significant ( \(P = 0.057\) ) (Fig. 5B, table S11). Notably, suppression of tumor immunity was not associated with all- cause mortality of population controls (table S12), suggesting that the association with all- cause mortality among cases might be prostate cancer- related. Lastly, prostate cancer
+
+<--- Page Split --->
+
+patients with increased suppression of tumor immunity were also significantly more likely to die from any cancer (prostate cancer or secondary cancer) following the prostate cancer diagnosis (Fig. 5C, table S13), indicating a more general predisposition to cancer in patients with a high suppression of tumor immunity score.  
+
+Elevated suppression of tumor immunity is associated with metastatic prostate cancer. To further corroborate the significance of suppression of tumor immunity in the development of lethal prostate cancer, we assessed its association with prostate cancer aggressiveness per NCCN guidelines (see Methods). Information on TNM stage was only obtainable for the NCI- Maryland prostate cancer patients, hence only these cases were scored according to the NCCN guidelines. Patients with a high suppression of tumor immunity score were at substantially increased odds of being diagnosed with regional or distant metastasis (HR 3.79, 95% CI 1.59- 9.04, > median vs. \(\leq\) median) (Table 1), consistent with the disease survival data. The data showed a significant trend in the association of elevated suppression of tumor immunity with disease aggressiveness ( \(P\) trend=0.004) (Table 1). Stratified analysis by race/ethnicity revealed that high suppression of tumor immunity was associated with metastatic prostate cancer more so among AA than EA men.  
+
+Blood levels of TNFRSF9 and pleiotrophin (PTN) predict lethal prostate cancer among AA men. To identify individual drivers of the relationship between immune- oncology markers and lethal prostate cancer, we applied a cross- validated, regularized Cox regression model using eNetXplorer (see Methods). Included in this model were the 82 immune- oncology markers and 6 covariates of clinical significance (age, education, BMI, smoking history, aspirin use, and diabetes). Lasso regression (alpha=1) was selected based on overall performance (fig. S6). Utilizing this method, we could not identify a robust predictive signature of lethal prostate cancer
+
+<--- Page Split --->
+
+for EA patients. However, for AA patients, a signature primarily driven by TNFRSF9 and PTN (both positively associated with the risk of lethal disease) and regular aspirin use (negatively associated with risk) were the top predictors ( \(P < 0.05\) ) based on two selection criteria: the feature frequency (Fig. 6A) and the weight of the features' contribution to the prediction (Fig. 6B). These features combined predicted prostate cancer- specific mortality with an accuracy of \(83.7\%\) (SE=3.8%). Our finding that regular aspirin use was a predictor of improved survival among AA men is consistent with previously published data from this case- control study \(^{14}\) and the Southern Community Cohort Study \(^{15}\) . The two proteins alone, TNFRSF9 and PTN, predicted prostate cancer- specific mortality with \(78.2\%\) (SE=4.2%) accuracy. AA prostate cancer patients with high levels (> median) of both TNFRSF9 and PTN in their blood at diagnosis had the worst prostate cancer- specific survival (Fig. 6C). By 10 years, 33% of cases with high levels of both TNFRSF9 and PTN died of prostate cancer compared to only 5% of cases with low levels of both or either of these proteins (Fig. 6C), highlighting the utility of these blood markers for risk stratification of AA prostate cancer patients.
+
+<--- Page Split --->
+
+## DISCUSSION  
+
+In this study, we describe consistent differences in the expression of immune and chemotaxis- related markers in men from three population groups, with two of them - AA and Ghanaian men - having an ancestral relationship due to the trans- Atlantic slave trade. Most notably, expression of immune- oncology markers related to immune suppression were up- regulated in men of West African ancestry and were associated with lethal prostate cancer. While ancestry can explain some of the observations, other and yet unknown factors may contribute to these clinically significant differences in immune function and chemotaxis.  
+
+Infections endemic to certain regions have shaped the immune response in affected populations, leaving a lasting genetic and epigenetic footprint<sup>36</sup>. As such, population differences in exposures to fatal pathogens have led to population heterogeneity in the immune. It has been estimated that as many as 360 immune- related genes have been targets of positive selection and have functional variations between populations<sup>37</sup>. Consistent with these observations, we now report population differences in circulating immune- oncological proteins among Ghanaian, AA, and EA men. We found that the serum proteome- defined immunome of Ghanaian men resembles the immunome of AA men more so than EA men. We identified CXCL5, CXCL1, MCP2, MCP1, and CXCL11 as the top immune- oncological proteins associated with West African ancestry. Four of these chemokines (CXCL5, CXCL1, MCP1, and CXCL11) are targets of Duffy Antigen Receptor for Chemokines (DARC) binding<sup>38</sup>. DARC is a non- signaling receptor that binds to both CXC and CC family of chemokines and acts as a depot for chemokines on erythrocytes and as decoy receptor on endothelial cells<sup>39</sup>. DARC expression modulates the susceptibility to clinical Plasmodium vivax malaria and loss of its expression on erythrocytes, which frequently occurs in sub- Saharan African populations due to germline
+
+<--- Page Split --->
+
+genetic variants, confers resistance against malarial infection<sup>40</sup>. Its loss may also influence cancer susceptibility<sup>41,42</sup>. Consequently, these individuals lack the ability to sequester the target chemokines, leading to elevated concentration of the chemokines in circulation<sup>43</sup>. Accordingly, we found that CXCL5, CXCL1, and CXCL11 were 2- 3- fold higher in sera of Ghanaian or AA men than EA men. Given the angiogenic properties of these chemokines<sup>44</sup>, their role in cancer progression has been proposed<sup>45</sup>.  
+
+As a key finding, we report that serum proteins driving chemotaxis and suppression of tumor immunity were elevated in men of African ancestry, suggesting persistent population differences in stimulation of leukocyte recruitment and T cell mediated immune response. Such differences may predispose men of African descent to a distinct tumor microenvironment. Although the direct impact of the peripheral immunome on the prostate tumor microenvironment requires further investigation, we and others have previously reported stark differences in the immune landscape of prostate tumors of AA men as compared to EA men<sup>9- 13</sup>. For instance, programmed cell death ligand- 1 (PD- L1), which suppresses T cell- mediated tumor immunity, was found to be overexpressed in AA prostate tumors<sup>46</sup>. Recent work by Awasthi et al. reported that AA prostate tumors tend to be enriched for immune pathways that are associated with poor clinical outcomes<sup>47</sup>. We show with our current work that elevated, peripheral suppression of tumor immunity associates with lethal prostate cancer. Hence, population differences in suppression of tumor immunity may contribute to the disproportionate burden of lethal prostate cancer among men of African ancestry. On the other hand, such differences may offer a therapeutic advantage for immunotherapeutic strategies that are tailored to target immune suppressive pathways. A recent study provided a first indication that differences in the response to cancer vaccines may lead to higher survival rates among AA men<sup>48</sup>.
+
+<--- Page Split --->
+
+Differentiating men who have lethal forms of prostate cancer from those with a more slow- growing disease remains a major challenge in clinical oncology. Risk stratification strategies are particularly needed for AA prostate cancer patients who disproportionately bear the prostate cancer burden. This study identified TNFRSF9 and PTN as candidate predictive blood markers for prostate cancer mortality among AA patients. AA patients with high levels of both TNFRSF9 and PTN in their sera had the highest risk of dying from prostate cancer. The membrane form of TNFRSF9 possesses antitumor properties and agonistic anti- TNFRSF9 antibodies are currently in clinical trials<sup>49,50</sup>. On the contrary, the soluble isoform of TNFRSF9 that we measured, generated by alternative splicing<sup>51</sup>, has been shown to antagonize antitumor immune response hence promote tumor survival most likely by acting as decoy receptor<sup>52,53</sup>. Regulatory T cells described as Tregs are thought to be a major source of secreted TNFRSF9<sup>54,55</sup>. Recently, TNFRSF9 mRNA level was shown to be a robust marker of tumor- infiltrating Tregs that suppress antitumor response<sup>56</sup>. Moreover, high numbers of TNFRSF9- expressing Tregs were associated with poor survival outcomes across multiple human cancers<sup>56</sup>, consistent with our findings that serum TNFRSF9 associates with lethal prostate cancer. Although pleiotrophin, the second protein marker associated with lethal prostate cancer in AA men, may not have the same immune function that soluble TNFRSF9 exhibits, it is a secreted cytokine with important roles in promoting angiogenesis and metastasis<sup>57</sup>. Recently, pleiotrophin was described as a serum- based biomarker of pro- metastatic prostate cancer<sup>58</sup>, consistent with our findings in this study.  
+
+To the best of our knowledge, this is the first study with a large representation of men of African descent who contributed to immune- oncological proteome profiling. With the advent of increasing immunotherapies in the drug development pipeline, such studies may inform research
+
+<--- Page Split --->
+
+on population differences in the immune landscape that need to be considered when designing therapies that exploit the immune response. Furthermore, our study may provide unique insights into variations in the manifestation and pathogenesis of different immune related diseases among different population groups.  
+
+In conclusion, it is a key finding of our study that suppression of tumor immunity was increased in Ghanaian and AA men, when compared to EA men, and associates with lethal prostate cancer. As such, these findings provide a novel insight into potential causes of the prostate cancer health disparity.
+
+<--- Page Split --->
+
+## MATERIALS and METHODS  
+
+NCI- Maryland prostate cancer case- control study. This study and the eligibility criteria have been previously described<sup>14,59</sup>. Race/ethnicity was assigned based on self- identification as either black or AA or as Caucasian or EA. The study was initiated to test the primary hypothesis that environmental exposures and ancestry- related factors contribute to the excessive prostate cancer burden among AA men. The study was approved by the NCI (protocol # 05- C- N021) and the University of Maryland (protocol #0298229) Institutional Review Boards and all participants signed an informed consent. Cases were recruited at the Baltimore Veterans Affairs Medical Center and the University of Maryland Medical Center. A total of 976 cases (489 AA and 487 EA men) were recruited into this study between 2005 and 2015. Controls were identified through the Maryland Department of Motor Vehicle Administration database and were frequency- matched to cases on age and race. A total of 1,034 population controls were recruited (486 AA and 548 EA men). At the time of enrollment, both cases and controls were administered a survey by a trained interviewer and a blood sample was collected. Serum samples were available for 846 cases (407 AA and 439 EA) and 846 controls (382 AA and 464 EA), therefore only these individuals were used for the study herein. Most of the 846 cases (85%) were recruited within a year of the disease diagnosis with a median of 5.1 months between disease diagnosis and blood collection.  
+
+NCI- Ghana prostate cancer case- control study. This case- control study has been previously described<sup>20</sup>. The study was designed to study lifestyle, environmental, and genetic risk factors for prostate cancer in African men. The study was approved by institutional review boards at the University of Ghana (protocol #001/01- 02) and at the National Cancer Institute (protocol #02CN240). Prior to study enrollment, all participants signed an informed consent. Prostate
+
+<--- Page Split --->
+
+cancer cases were recruited at Korle Bu Teaching Hospital in Accra, Ghana between 2008 and 2012. The cases were diagnosed using Digital Rectal Exam (DRE) and PSA tests, followed by biopsy confirmation. Immediately after diagnosis and before treatment, cases were consented and asked to submit blood specimen and questionnaire data. Controls were identified through probability sampling using the 2000 Ghana Population and Housing Census data to recruit approximately 1,000 men aged 50–74 years in the Greater Accra region between 2004 and 2006. These men were confirmed to not have prostate cancer by PSA testing and DRE. Serum samples were available for 586 prostate cancer cases and 659 population controls; hence, only these individuals were used for the study herein.  
+
+Serum sample processing. The participants in the two studies provided blood samples at time of recruitment. For the NCI- Maryland study, most blood samples were processed the same day, but always within 48 hours, after storage in a refrigerator. For the NCI- Ghana study, blood samples were processed within 6 hours. In this study, population controls provided fasting blood. Serum was prepared using standard procedures and aliquots were stored at \(- 80^{0}\mathrm{C}\) . Serum samples were shipped from Ghana to the NCI in dry ice boxes.  
+
+Serum protein measurement. Serum levels of 92 immuno- oncology panel proteins were measured simultaneously using a proprietary multiplex Proximal Extension Assay (PEA) by Olink Proteomics (Boston). Olink utilizes a relative quantification unit, Normalized Protein eXpression (NPX), which is in a Log2- format. Serum samples from NCI- MD study (846 cases and 846 controls) and NCI- Ghana study (586 cases and 659 controls) were completely randomized and were assayed in that order. In addition to the built- in internal controls, \(5\%\) blinded duplicates were randomly selected and were randomized along with the original set of samples. Protein levels were intensity normalized to adjust for batch effect. Because all our
+
+<--- Page Split --->
+
+samples were randomized across plates, a global adjustment was used to center the values for each assay around its median and across all plates. Ninety- five percent of the samples passed a stringent quality control (NCI- MD study: 819 cases and 828 controls; NCI- Ghana study: 489 cases and 654 controls) – with coefficients of variation (CV) among duplicates at \(< 10\%\) for every marker. Out of the 92 proteins assayed, IL33, IL35, IL21, IL2, IFNβ, IL13, IL1α, CXCL12, IFNγ, and TNF were detected in less than \(20\%\) of the samples, hence the remaining 82 proteins were used for subsequent analysis (table S2).  
+
+Functional annotation and biological processes scores. Proteins were grouped into six biological processes based on their respective biological roles following the Olink guideline: apoptosis/cell killing, autophagy/metabolism, chemotaxis/trafficking to tumor, suppression of tumor immunity (Th2 response, tolerogenic), promotion of tumor immunity (Th1 responses), or vasculature and tissue remodeling. Apoptosis, autophagy, chemotaxis, suppression of tumor immunity, promotion of tumor immunity, or vasculature scores were calculated for each study participant as the mean z- score value for the proteins belonging to the respective biological process. For survival analysis, the biological process/pathway scores were evaluated as continuous variables. To evaluate the association of suppression of tumor immunity with aggressive prostate cancer, we grouped suppression of tumor immunity scores into low (≤median) and high (>median) with cutoffs determined using the distribution of the score among population controls of the NCI- Maryland study.  
+
+Prostate Specific Antigen (PSA) measurement. For the cases in the NCI- Maryland cohort, PSA levels were obtained from medical record. For the controls of the NCI- Maryland study, total PSA was measured from stored serum aliquots using the human total PSA ELISA Kit (Abcam, ab188388). About \(7\%\) (n=56) of the controls in the NCI- Maryland cohort had PSA greater than
+
+<--- Page Split --->
+
+2.5 ng/ml, while only \(3\%\) (n=27) had blood PSA over 4 ng/ml. For the controls in the NCI- Ghana study, close to \(20\%\) (n=132) had a PSA greater than 2.5 ng/ml, while about \(11\%\) (n=73) had PSA over 4 ng/ml.  
+
+C- reactive protein (CRP) measurement. Plasma CRP was assayed using an ELISA assay (cat# ab99995, Abcam, United States) according to the manufacturer's instructions. Two microliters of plasma samples were added to \(398 \mu \mathrm{L}\) of 1x Diluent D, followed by a second 1: 200 dilution steps for each sample. One- hundred microliters of CRP standard (0 to \(600 \mathrm{pg / mL}\) ) and the diluted samples were loaded as duplicates into pre- coated 96- well plates. Samples were incubated overnight at \(4^{\circ} \mathrm{C}\) with gentle shaking, followed by incubations with the anti- human CRP antibody and the horseradish peroxidase- streptavidin solution. CRP was quantified measuring absorbance at \(450 \mathrm{nm}\) with a microplate reader.  
+
+West African ancestry estimation for participants in the NCI- Maryland case- control study. Genomic DNA was isolated from buffy coats (DNeasy Blood & Tissue Kit - Qiagen) or mouthwash samples (standard phenol- chloroform technique). Isolated DNA was genotyped for 100 ancestry informative markers using the Sequenom MassARRAY iPLEX platform, as previously described<sup>35</sup>. Single nucleotide polymorphism genotype calls were generated using Sequenom TYPER software. A genotype concordance rate of \(>99\%\) was observed for all markers. Admixture estimates for each study participant were calculated using a model- based clustering method as implemented in the program STRUCTURE v2.3. We applied STRUCTURE v2.3 with an admixture model estimating K (number of sub populations) from 2 to 5 with 100 iterations and parental population genotypes from West Africans, Europeans, and Native Americans, yielding three admixture estimations (West African, European, Native American). For a subset (83%) of the cases and controls, additional West African ancestry
+
+<--- Page Split --->
+
+estimates were provided by the Cancer Genomics Research Laboratory/NCI- Leidos from a genome- wide association study using the Infinium HumanOmni5- Quad BeadChip array. West African ancestry estimates using the two approaches were very similar (r=0.98).  
+
+Association of clinical/socio- demographic characteristics with immune- oncological  
+
+proteins. The association of age, body mass index (BMI), education, aspirin use, smoking, diabetes, and PSA levels with the relative abundance of individual analytes was assessed by means of multivariable linear regression models implemented by the function lm in the base R package stats (version 3.6.1). These variables were chosen because they have either been linked to prostate cancer risk and survival or may influence the status of inflammation and host immunity. An analyte (as response variable) was considered significantly associated with clinical and socio- demographic covariables if the multivariable model yielded \(P < 0.05\) on the F- statistic. If this condition was satisfied, the association between the target analyte and each individual covariable was characterized by the corresponding \(P\) value and coefficient.  
+
+Analysis of variance. Variance analysis for the levels of each of the 82 immune- oncological cytokines were simultaneously assessed as a function of genetic estimation of West African admixture among men without prostate cancer from the NCI- Maryland study. The analysis was implemented by the function aov in the base R package stats (version 3.6.1).  
+
+Heatmap plots. Heat map plots were generated using Broad Institute's web- based matrix visualization and analysis platform - Morpheus (https://software.broadinstitute.org/morpheus). To avoid spurious effects from outliers in heatmap plots, each protein's range of abundance values were set to saturate at the \(1^{\text{st}}\) and \(99^{\text{th}}\) percentiles. To account for widely different
+
+<--- Page Split --->
+
+abundance ranges for different proteins in the assay, each protein's measured abundances across the cohort were Z- score transformed.  
+
+Gene ontology (GO) enrichment analysis. GO terms with an enrichment in proteins of interest were identified using Over- Representation Analysis (ORA) as part of the web tool WebGestalt (WEB- based Gene SeT AnaLysis Toolkit). Enriched gene sets were further processed using affinity propagation (R package apcluster) to cluster gene sets according to functional similarity.  
+
+Survival analysis. Information on patient survival was only obtainable for the NCI- Maryland prostate cancer patients. Survival data was obtained from the National Death Index (NDI) database. We calculated survival for cases from date of diagnosis to either date of death or to the censor date of December 31, 2018. We built a multivariable Cox regression model with all biological processes scores and adjustment for other covariables to estimate adjusted hazard ratios (HR) and 95% confidence intervals (CI) for all- cause mortality, cancer- related mortality, and prostate cancer- specific mortality of cases. We adjusted for the following potential confounding factors: age at study entry (years), body- mass index (BMI, \(\mathrm{kg} / \mathrm{m}^2\) ), self- reported race (AA/EA), education (high school or less, some college, college, professional school), income (less than \(\) 10k\(,\) \ \(10 - 30K\) , \(\) 30 - 60K\(,\) \ \(60 - 90k\) , greater than \(\) 90k\(), smoking history (never, former, current), diabetes (no/yes), aspirin use (no/yes), and treatment (0=none, 1=surgery, 2=radiotherapy, 3=hormone, 4=combination). Missing values for education (n=1), smoking history (n=5), and income (n=63) were imputed using the R package missForest, which implements nonparametric missing value imputation based on random forests. In the overall survival analysis of population controls, we calculated survival from the date of interview to either date of death or to the censor date of December 31st, 2018. We applied the Cox regression model to estimate adjusted HR and 95% CI and adjusted for all the confounding factors listed
+
+<--- Page Split --->
+
+above except for treatment. Missing values for education (n=1), smoking history (n=7), and income (n=67) were imputed using the R package missForest. The reported HRs indicate the change in risk of dying when the biological process z- score value increases by 1 while holding all the other biological processes' z- scores and covariates constant.  
+
+Classification of cases using National Comprehensive Cancer Network (NCCN) Risk Score. Cases were assigned to risk groups based on the patients' TNM stage, Gleason score, Gleason pattern, and PSA level at diagnosis according to the 2019 NCCN guideline for prostate cancer<sup>60</sup>. Information on TNM stage was only obtainable for the NCI- Maryland prostate cancer patients, hence only these cases were scored. Cases were categorized as localized, regional, and metastatic prostate cancer based on their clinical parameters at the time of diagnosis. Localized prostate cancer cases were further classified into low, intermediate, high, and very high risk based on the likelihood of their disease to progress to lethal prostate cancer per the 2019 NCCN guideline<sup>60</sup>. Prostate cancer cases with lymph node involvement but no distant metastasis at diagnosis were classified as regional prostate cancer while those with distant metastasis at the time of diagnosis were classified as metastatic prostate cancer. For our analysis, we condensed these risk groups into 4 categories (low, intermediate, high/very high, and regional/metastatic). Developing a predictive proteomic signature of lethal prostate cancer. The analysis was restricted to the cases from NCI- Maryland study for whom we had survival data. We stratified by race/ethnicity into AA cases (360 censored, 34 prostate cancer deaths) and EA cases (402 censored, 23 prostate cancer deaths). To identify a multi- analyte proteomic signature that is predictive of lethal prostate cancer, 88 features were evaluated [82 immune- oncological proteins along with six demographic/clinical variables (education, age, BMI (BMI, kg/m<sup>2</sup>), smoking history, diabetes, and aspirin use)]. Missing values for education (n=1) and smoking history
+
+<--- Page Split --->
+
+(n=5) were imputed using R package missForest. R package eNetXplorer (version 1.1.2)61 was implemented to build cross- validated, regularized Cox regression models with different elastic net mixture parameters from ridge (alpha=0) to lasso (alpha=1). Alpha was selected based on overall performance assessed as a function of the 5- fold cross- validated quality function (concordance) and the empirical \(P\) value generated from comparing the model against a statistical ensemble of null models created by random permutations of the response (i.e. survival time/status randomized across subjects in the cohort). These results comprise 10,000 Cox regression elastic net realizations arising from 200 randomly generated folds, each of them compared against 50 null model permutations. Features' performance as predictors was evaluated using two different, but complementary selection criteria: feature coefficients and feature frequencies. The feature frequency measure captures the significance of how often a feature is chosen in an in- bag model. When it is chosen, the feature coefficient measure captures the significance of the feature's weight in the in- bag model. See the publication by Candia et al for more details on this method61. Using only the significant protein features from both selection criteria, a multivariate Cox regression model was run. Risk stratification was used to generate Kaplan- Meier plots and log- rank tests of significance.  
+
+Statistical analysis. Data analyses were performed using Stata/SE 16.0 and R statistical software packages. All statistical tests were two- sided, and an association was considered statistically significant with \(P< 0.05\) .
+
+<--- Page Split --->
+
+## Supplementary Materials:  
+
+Fig. S1. Variance explained by the inter- plate batch effect  Fig. S2. Detection frequencies of 92 immuno- oncological markers measured in 2937 serum samples  Fig. S3. Correlation matrix presenting Pearson pairwise correlations for each of the 82 serum protein pairs  Fig. S4. Gene ontology (GO) enrichment analysis  Fig. S5. Functional enrichment analysis of proteins positively associated with diabetes in men of African ancestry  Fig. S6. Performance of regularized cox regression models across alpha  Table S1. Characteristics of prostate cancer cases and population controls of the NCI- Maryland and NCI- Ghana Study  Table S2. List of 82 Immuno- oncological proteins detected in more than \(20\%\) of the serum samples  Table S3. Top 10 Pearson pairwise correlations in men without prostate cancer  Table S4. The association of blood CRP with clinical/sociodemographic variables estimated using multiple linear regression  Table S5. The fraction of variance in each of the serum proteins explained by degree of West African ancestry  Table S6. The fraction of variance in each of the serum proteins explained by degree of West African ancestry after adjusting for difference in age, bmi, aspirin use, education, diabetes status, smoking, and income  Table S7. Immune oncological markers that are significantly elevated in both Af and AA men compared to EA men  Table S8. Immune oncological markers that are significantly downregulated in both Af and AA men compared to EA men  Table S9. Serum proteins grouped by biological process  Table S10. Effect of biological processes scores on all- cause mortality of prostate cancer patients  Table S11. Effect of biological processes scores on prostate cancer- specific mortality  Table S12. Effect of biological processes scores on all- cause mortality of population controls  Table S13. Effect of biological processes scores on mortality from any cancer following a diagnosis with prostate cancer  
+
+Acknowledgments: We would like to thank personnel at the University of Maryland and the Baltimore Veterans Administration Hospital for their contributions with the recruitment of participants to the NCI- Maryland study. We would also like to thank Prof. Edward D. Yeboah as the original Ghana PI and Ms. Evelyn Tay as the original Study Manager for the NCI- Ghana study.  
+
+## Funding:  
+
+DoD award W81XWH1810588 (to SA, CY)  U54 CA118623- CY (NCI) and U54- MD007585- 26- CY (NIMHD) (to CY)  Intramural Research Program of the NIH, National Cancer Institute (NCI), Center for Cancer Research and Division of Cancer Epidemiology and Genetics (to SA, MBC)
+
+<--- Page Split --->
+
+584 Author contributions: 585 Conceptualization: TZM, CY, MBC, SA 586 Data curation: TZM, THD, MK, CJS, SVJ, OMO, AA, FJJ, RK 587 Formal Analysis: TZM, JC, RK 588 Funding acquisition: CY, MBC, SA 589 Investigation: TZM, JC, FJJ 590 Methodology: TZM, JC, CAL, MBC, SA 591 Project administration: THD, FB 592 Resources: WT, YT, RBB, AAA, JEM, RNH, AWH, MBC, SA 593 Supervision: WT, SA 594 Visualization: TZM, JC 595 Writing - original draft: TZM 596 Writing - review & editing: TZM, FB, WT, MK, CJS, YT, RBB, AAA, JEM, XWW, CAL, CY, 597 MBC, SA 598 Conflicts of interest: The authors declare that they have no competing interests. 600 601 Data and materials availability: All data are available in the main text or the supplementary 602 materials.
+
+<--- Page Split --->
+
+604 1 Powell, I. J. Epidemiology and pathophysiology of prostate cancer in African-American men. J Urol. 177, 444- 449 (2007). 606 2 Rebbeck, T. R. et al. Global patterns of prostate cancer incidence, aggressiveness, and mortality in men of african descent. Prostate Cancer 2013, 560857 (2013). 608 3 Culp, M. B., Soerjomataram, I., Efstathiou, J. A., Bray, F. & Jemal, A. Recent Global Patterns in Prostate Cancer Incidence and Mortality Rates. Eur Urol 77, 38- 52, doi:10.1016/j.eururo.2019.08.005 (2020). 611 4 Butler, E. N., Kelly, S. P., Coupland, V. H., Rosenberg, P. S. & Cook, M. B. Fatal prostate cancer incidence trends in the United States and England by race, stage, and treatment. Br J Cancer 123, 487- 494, doi:10.1038/s41416- 020- 0859- x (2020). 614 5 Heyns, C. F., Fisher, M., Lecuona, A. & van der Merwe, A. Prostate cancer among different racial groups in the Western Cape: presenting features and management. S Afr Med J 101, 267- 270, doi:10.7196/samj.4420 (2011). 616 6 Wallace, T. A., Martin, D. N. & Ambs, S. Interactions among genes, tumor biology and the environment in cancer health disparities: examining the evidence on a national and global scale. Carcinogenesis 32, 1107- 1121 (2011). 620 7 Rebbeck, T. R. Prostate Cancer Disparities by Race and Ethnicity: From Nucleotide to Neighborhood. Cold Spring Harb Perspect Med 8, doi:10.1101/cshperspect.a030387 (2018). 623 8 Lachance, J. et al. Genetic Hitchhiking and Population Bottlenecks Contribute to Prostate Cancer Disparities in Men of African Descent. Cancer Res 78, 2432- 2443, doi:10.1158/0008- 5472.CAN- 17- 1550 (2018). 626 9 Hardiman, G. et al. Systems analysis of the prostate transcriptome in African- American men compared with European- American men. Pharmacogenomics 17, 1129- 1143, doi:10.2217/pgs- 2016- 0025 (2016). 629 10 Powell, I. J. et al. Genes associated with prostate cancer are differentially expressed in African American and European American men. Cancer Epidemiol Biomarkers Prev 22, 891- 897, doi:10.1158/1055- 9965.EPI- 12- 1238 (2013). 632 11 Wallace, T. A. et al. Tumor immunobiological differences in prostate cancer between African- American and European- American men. Cancer Res 68, 927- 936, doi:10.1158/0008- 5472.CAN- 07- 2608 (2008). 634 12 Yuan, J. et al. Integrative comparison of the genomic and transcriptomic landscape between prostate cancer patients of predominantly African or European genetic ancestry. PLoS Genet 16, e1008641, doi:10.1371/journal.pgen.1008641 (2020). 638 13 Tang, W. et al. IFNL4- DeltaG Allele Is Associated with an Interferon Signature in Tumors and Survival of African- American Men with Prostate Cancer. Clin Cancer Res 24, 5471- 5481, doi:10.1158/1078- 0432.CCR- 18- 1060 (2018). 641 14 Smith, C. J. et al. Aspirin Use Reduces the Risk of Aggressive Prostate Cancer and Disease Recurrence in African- American Men. Cancer Epidemiol Biomarkers Prev 26, 845- 853, doi:10.1158/1055- 9965.EPI- 16- 1027 (2017). 644 15 Tang, W. et al. Aspirin Use and Prostate Cancer among African American Men in the Southern Community Cohort Study. Cancer Epidemiol Biomarkers Prev, doi:10.1158/1055- 9965.EPI- 19- 0792 (2020).
+
+<--- Page Split --->
+
+647 16 Nedelec, Y. et al. Genetic Ancestry and Natural Selection Drive Population Differences in Immune Responses to Pathogens. Cell 167, 657- 669 e621, doi:10.1016/j.cell.2016.09.025 (2016). 650 17 Ness, R. B., Haggerty, C. L., Harger, G. & Ferrell, R. Differential distribution of allelic variants in cytokine genes among African Americans and White Americans. Am J Epidemiol 160, 1033- 1038, doi:10.1093/aje/kwh325 (2004). 653 18 Rotival, M., Quach, H. & Quintana- Murci, L. Defining the genetic and evolutionary architecture of alternative splicing in response to infection. Nat Commun 10, 1671, doi:10.1038/s41467- 019- 09689- 7 (2019). 656 19 Van Dyke, A. L., Cote, M. L., Wenzlaff, A. S., Land, S. & Schwartz, A. G. Cytokine SNPs: Comparison of allele frequencies by race and implications for future studies. Cytokine 46, 236- 244, doi:10.1016/j.cyto.2009.02.003 (2009). 659 20 Cook, M. B. et al. A genome- wide association study of prostate cancer in West African men. Hum Genet 133, 509- 521, doi:10.1007/s00439- 013- 1387- z (2014). 661 21 Nikolich- Zugich, J. The twilight of immunity: emerging concepts in aging of the immune system. Nat Immunol 19, 10- 19, doi:10.1038/s41590- 017- 0006- x (2018). 662 22 Ferretti, E., Pistoia, V. & Corcione, A. Role of fractalkine/CX3CL1 and its receptor in the pathogenesis of inflammatory and malignant diseases with emphasis on B cell malignancies. Mediators Inflamm 2014, 480941, doi:10.1155/2014/480941 (2014). 663 23 Zavada, J., Zavadova, Z., Zat'ovicova, M., Hyrsl, L. & Kawaciuk, I. Soluble form of carbonic anhydrase IX (CA IX) in the serum and urine of renal carcinoma patients. Br J Cancer 89, 1067- 1071, doi:10.1038/sj.bjc.6601264 (2003). 664 24 Supuran, C. T. Carbonic anhydrases: novel therapeutic applications for inhibitors and activators. Nat Rev Drug Discov 7, 168- 181, doi:10.1038/nrd2467 (2008). 665 25 Erra Diaz, F., Dantas, E. & Geffner, J. Unravelling the Interplay between Extracellular Acidosis and Immune Cells. Mediators Inflamm 2018, 1218297, doi:10.1155/2018/1218297 (2018). 666 26 Alessandrini, F., Pezze, L. & Ciribilli, Y. LAMPs: Shedding light on cancer biology. Semin Oncol 44, 239- 253, doi:10.1053/j.seminoncol.2017.10.013 (2017). 667 27 Rabinovich, G. A. & Toscano, M. A. Turning 'sweet' on immunity: galectin- glycan interactions in immune tolerance and inflammation. Nat Rev Immunol 9, 338- 352, doi:10.1038/nri2536 (2009). 668 28 Armson, Y., Shoenfeld, Y. & Amital, H. Effects of tobacco smoke on immunity, inflammation and autoimmunity. J Autoimmun 34, J258- 265, doi:10.1016/j.jaut.2009.12.003 (2010). 669 29 Wada, J. & Makino, H. Innate immunity in diabetes and diabetic nephropathy. Nat Rev Nephrol 12, 13- 26, doi:10.1038/nr neph.2015.175 (2016). 670 30 Lee, Y. S. et al. The fractalkine/CX3CR1 system regulates beta cell function and insulin secretion. Cell 153, 413- 425, doi:10.1016/j.cell.2013.03.001 (2013). 671 31 Shah, R. et al. Fractalkine is a novel human adipocemokine associated with type 2 diabetes. Diabetes 60, 1512- 1518, doi:10.2337/db10- 0956 (2011). 672 32 Navarro- Gonzalez, J. F., Mora- Fernandez, C., Muros de Fuentes, M. & Garcia- Perez, J. Inflammatory molecules and pathways in the pathogenesis of diabetic nephropathy. Nat Rev Nephrol 7, 327- 340, doi:10.1038/nr neph.2011.51 (2011).
+
+<--- Page Split --->
+
+Liuz, Z. Q. et al. Prognostic role of C-reactive protein in prostate cancer: a systematic review and meta-analysis. Asian J Androl 16, 467- 471, doi:10.4103/1008- 682X.123686 (2014).Saito, K. & Kihara, K. C- reactive protein as a biomarker for urological cancers. Nat Rev Urol 8, 659- 666, doi:10.1038/nrurol.2011.145 (2011).Al- Alem, U. et al. Association of genetic ancestry with breast cancer in ethnically diverse women from Chicago. PLoS One 9, e112916, doi:10.1371/journal.pone.0112916 (2014).Rotimi, C. N. et al. The genomic landscape of African populations in health and disease. Hum Mol Genet 26, R225- R236, doi:10.1093/hmg/ddx253 (2017).Barreiro, L. B. & Quintana- Murci, L. From evolutionary genetics to human immunology: how selection shapes host defence genes. Nat Rev Genet 11, 17- 30, doi:10.1038/nrg2698 (2010).Gardner, L., Patterson, A. M., Ashton, B. A., Stone, M. A. & Middleton, J. The human Duffy antigen binds selected inflammatory but not homeostatic chemokines. Biochem Biophys Res Commun 321, 306- 312, doi:10.1016/j.bbrc.2004.06.146 (2004).Mantovani, A., Bonecchi, R. & Locati, M. Tuning inflammation and immunity by chemokine sequestration: decoys and more. Nat Rev Immunol 6, 907- 918, doi:10.1038/nri1964 (2006).Miller, L. H., Mason, S. J., Clyde, D. F. & McGinniss, M. H. The resistance factor to Plasmodium vivax in blacks. The Duffy- blood- group genotype, FyFy. N Engl J Med 295, 302- 304, doi:10.1056/NEJM197608052950602 (1976).Jenkins, B. D. et al. Atypical Chemokine Receptor 1 (DARC/ACKR1) in Breast Tumors Is Associated with Survival, Circulating Chemokines, Tumor- Infiltrating Immune Cells, and African Ancestry. Cancer Epidemiol Biomarkers Prev 28, 690- 700, doi:10.1158/1055- 9965.EPI- 18- 0955 (2019).Martini, R. et al. Investigation of Triple- Negative Breast Cancer Risk Alleles in An International African- Enriched Cohort. scientific reports Preprint at: https://www.researchsquare.com/article/rs- 109841/v1 (2021).Yao, S. et al. Genetic ancestry and population differences in levels of inflammatory cytokines in women: Role for evolutionary selection and environmental factors. PLoS Genet 14, e1007368, doi:10.1371/journal.pgen.1007368 (2018).Strieter, R. M. et al. The functional role of the ELR motif in CXC chemokine- mediated angiogenesis. J Biol Chem 270, 27348- 27357, doi:10.1074/jbc.270.45.27348 (1995).Bikfalvi, A. & Billottet, C. The CC and CXC chemokines: major regulators of tumor progression and the tumor microenvironment. Am J Physiol Cell Physiol 318, C542- C554, doi:10.1152/ajpcell.00378.2019 (2020).Calagua, C. et al. Expression of PD- L1 in Hormone- naive and Treated Prostate Cancer Patients Receiving Neoadjuvant Abiraterone Acetate plus Prednisone and Leuprolide. Clin Cancer Res 23, 6812- 6822, doi:10.1158/1078- 0432.CCR- 17- 0807 (2017).Awasthi, S. et al. Comparative Genomics Reveals Distinct Immune- oncologic Pathways in African American Men with Prostate Cancer. Clin Cancer Res 27, 320- 329, doi:10.1158/1078- 0432.CCR- 20- 2925 (2021).Sartor, O. et al. Survival of African- American and Caucasian men after sipuleucel- T immunotherapy: outcomes from the PROCEED registry. Prostate Cancer Prostatic Dis 23, 517- 526, doi:10.1038/s41391- 020- 0213- 7 (2020).
+
+<--- Page Split --->
+
+736 49 Segal, N. H. et al. Phase I Study of Single- Agent Utomilumab (PF- 05082566), a 4- 1BB/CD137 Agonist, in Patients with Advanced Cancer. Clin Cancer Res 24, 1816- 1823, doi:10.1158/1078- 0432.CCR- 17- 1922 (2018). 739 50 Qi, X. et al. Optimization of 4- 1BB antibody for cancer immunotherapy by balancing agonistic strength with FcgammaR affinity. Nat Commun 10, 2141, doi:10.1038/s41467- 019- 10088- 1 (2019). 741 51 Michel, J., Langstein, J., Hofstadter, F. & Schwarz, H. A soluble form of CD137 (ILA/4- 1BB), a member of the TNF receptor family, is released by activated lymphocytes and is detectable in sera of patients with rheumatoid arthritis. Eur J Immunol 28, 290- 295, doi:10.1002/(SICI)1521- 4141(199801)28:01<290::AID- IMMU290>3.0.CO;2- S (1998). 746 52 Michel, J. & Schwarz, H. Expression of soluble CD137 correlates with activation- induced cell death of lymphocytes. Cytokine 12, 742- 746, doi:10.1006/cyto.1999.0623 (2000). 748 53 Labiano, S. et al. Hypoxia- induced soluble CD137 in malignant cells blocks CD137L- costimulation as an immune escape mechanism. Oncoimmunology 5, e1062967, doi:10.1080/2162402X.2015.1062967 (2016). 752 54 Itoh, A. et al. Soluble CD137 Ameliorates Acute Type 1 Diabetes by Inducing T Cell Anergy. Front Immunol 10, 2566, doi:10.3389/fimmu.2019.02566 (2019). 754 55 Kachapati, K. et al. The B10 Idd9.3 locus mediates accumulation of functionally superior CD137(+) regulatory T cells in the nonobese diabetic type 1 diabetes model. J Immunol 189, 5001- 5015, doi:10.4049/jimmunol.1101013 (2012). 756 56 Freeman, Z. T. et al. A conserved intratumoral regulatory T cell signature identifies 4- 1BB as a pan- cancer target. J Clin Invest 130, 1405- 1416, doi:10.1172/JCI128672 (2020). 760 57 Papadimitriou, E. et al. Pleiotrophin and its receptor protein tyrosine phosphatase beta/zeta as regulators of angiogenesis and cancer. Biochim Biophys Acta 1866, 252- 265, doi:10.1016/j.bbcan.2016.09.007 (2016). 763 58 Liu, S. et al. Discovery of PTN as a serum- based biomarker of pro- metastatic prostate cancer. Br J Cancer 124, 896- 900, doi:10.1038/s41416- 020- 01200- 0 (2021). 765 59 Minas, T. Z. et al. IFNL4- DeltaG is associated with prostate cancer among men at increased risk of sexually transmitted infections. Commun Biol 1, 191, doi:10.1038/s42003- 018- 0193- 5 (2018). 767 60 Network, N. C. C. The NCCN Clinical Practice Guidelines in Oncology for Prostate Cancer, V4.2019. https://www.nccn.org/professionals/physician_gls/default.aspx#prostate (Accessed November 16, 2020). 772 61 Candia, J. & Tsang, J. S. eNetXplorer: an R package for the quantitative exploration of elastic net families for generalized linear models. BMC Bioinformatics 20, 189, doi:10.1186/s12859- 019- 2778- 5 (2019).
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig. 1. Association of socio-demographic and clinical characteristics with systemic immune-oncological proteins in Ghanaian (Af), AA, and EA men without prostate cancer. Association of the 82 immuno-oncological proteins with age, BMI, education, aspirin use, smoking, diabetes, and PSA was assessed in men without prostate cancer using a multivariable linear regression model. An analyte was considered significantly associated with clinical and socio-demographic covariables if the multivariable model yielded a \(P<0.05\) on the F-statistic. Analytes that did not have a significant association with any of the clinical/sociodemographic variables in at least one of the population groups are not presented in the heatmap. Blue represents negative association while red represents positive association. The significance level (P value-based) for each association is color-coded. TI = tumor immunity. </center>
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig. 2. Unsupervised hierarchical clustering associates circulating immune-oncological proteome profiles with population groups - Ghanaian (Af), AA, and EA men. Heatmap showing protein profiles for men without prostate cancer. Each row represents a protein (n=82), and each column corresponds to an individual [n=1482 (654 Af, 374 AA, and 454 EA)]. Each individual is color-coded as Af, AA, or EA in the annotation bar on top of the heatmap. Normalized z-score of proteins abundance are depicted on a low-to-high scale (blue-white-red). </center>
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig. 3. Immune-oncological proteins and their relationship with West-African ancestry. (A) Variance analysis for the levels of each of the 82 immune-oncological cytokines assessed as a function of genetic estimation of West African admixture among men without prostate cancer within the NCI-Maryland study. The blue plot represents the proportion of variance that can be explained by the degree of West-African admixture while the grey plot represents the residual variance that remains to be explained by other factors other than West-African ancestry. (B) The median levels of the top six West-African ancestry correlated immune-oncological proteins were compared between Af, AA, and EA. Error bars represent inter quartile range (IQR). Linearized protein abundances ( \(2^{Npx}\) ) were used to determine median and IQR for each of the proteins. </center>
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 4. Population differences in the abundance of proteins driving (A) apoptosis, (B) autophagy, (C) chemotaxis, (D) promotion of tumor immunity, (E) suppression of tumor immunity, and (F) vasculature. Heatmaps showing levels of process/pathway-associated proteins in relationship to population group [Ghanaian (Af), AA, EA]. Shown to the right are the mean score differences for these processes/pathways among the three population groups. Profiles for Ghanaian (n=654), AA (n=374), and EA (n=454) men without prostate cancer. The process/pathway scores are derived from the average Z-scores of all the associated proteins. These scores are shown as violin plots. TI = tumor immunity. </center>
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Fig. 5 </center>  
+
+Fig. 5. Suppression of the tumor immunity pathway associates with lethal prostate cancer. We assessed the association of the six pathways defined by the 82 immune-oncology markers with all-cause mortality, prostate cancer-specific mortality or mortality due to any cancer after a prostate cancer diagnosis. The pathway scores were evaluated as continuous predictor variables. Suppression of tumor immunity pathway was distinctively associated with all-cause mortality (A), prostate cancer-specific mortality (B), or a mortality due to any cancer after a prostate cancer diagnosis (C). Multivariable cox regression analyses were used to assess if the pathways were independently associated with survival of prostate cancer patients in the NCI-Maryland study. We adjusted for the following potential confounding factors: age at study entry (years), body-mass index (BMI, kg/m2), self-reported race (AA/EA), education (high school or less, some college, college, professional school), income (less than \(\) 10k\(,\) \ \(10 - 30k\) , \(\) 30 - 60k\(,\) \ \(60 - 90k\) , greater than \(\) 90k\(), smoking history (never, former, current), diabetes (no/yes), aspirin use (no/yes), and treatment (0 = none, 1 = surgery, 2 = radiotherapy, 3 = hormone, 4 = combination). The hazard ratios (HR) indicate the change in risk of dying when the biological process z-score value increases by 1 while holding all the other biological processes' z-scores and covariates constant.
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Fig. 6 </center>  
+
+Fig. 6. A signature of two serum markers is predictive of lethal prostate cancer in AA patients. Cross- validated, regularized Cox regression models with different elastic net mixture parameters from ridge (alpha=0) to lasso (alpha=1) were implemented to identify a predictive proteomic signature. (A) Heatmaps of feature frequencies across alpha. Features were ranked by \(P\) value for alpha=1. (B) Heatmaps of feature coefficients across alpha. Features were ranked by \(P\) value for alpha=1. (C) Kaplan- Meier plot comparing prostate cancer-specific mortality of AA cases with high levels (> median) of both TNFRSF9 and PTN (pleiotrophin) vs. low levels of either or both proteins. Log rank test was used to determine if there were statistically significant survival differences. Adjusted hazard ratio (HR) compares the risk of prostate cancer mortality for those with high levels of both TNFRSF9 and PTN vs. the remaining AA cases. HR estimates were adjusted for potential confounding factors: age, BMI, education, smoking history, diabetes status, aspirin use, treatment, and income. In B & C, \(P\) value significance was coded as \(< 0.001\) (***), \(< 0.01\) (**), \(< 0.05\) (*), and \(< 0.1\) (.).
+
+<--- Page Split --->
+
+
+Table 1. A high score for suppression of tumor immunity associates with National Comprehensive Cancer Network (NCCN) Risk Score for metastatic prostate cancer   
+
+<table><tr><td>NCCN Risk Score</td><td>Total<br>OR (95% CI)*</td><td>AA<br>OR (95% CI)*</td><td>EA<br>OR (95% CI)*</td></tr><tr><td>Low</td><td>Ref</td><td>Ref</td><td>Ref</td></tr><tr><td>Intermediate</td><td>1.04 (0.68-1.59)</td><td>0.89 (0.46-1.70)</td><td>1.18 (0.65, 2.13)</td></tr><tr><td>High/Very High</td><td>1.47 (0.87-2.48)</td><td>1.33 (0.59-2.98)</td><td>1.72 (0.83, 3.54)</td></tr><tr><td>Regional/Metastatic</td><td>3.79 (1.59-9.04)</td><td>5.90 (1.43-24.34)</td><td>3.16 (0.95, 10.50)</td></tr><tr><td>P value for trend</td><td>0.004</td><td>0.019</td><td>0.040</td></tr></table>
+
+Note: Bolded data indicate significant associations in the logistic regression analysis. \\*Logistic regression adjusted for age at study entry, BMI \((\mathrm{kg} / \mathrm{m}2)\) , diabetes (no/yes), aspirin (no/yes), education (high school or less, some college, college, professional school), family history of prostate cancer (first-degree relatives, yes/no), self-reported race (not included in the stratified analysis), income (less than \(\$ 10k\) , \(\$ 10 - 30k\) , \(\$ 30 - 60k\) , \(\$ 60 - 90k\) , greater than \(\$ 90k\) ), smoking history (never, former, current), treatment (0=none, 1=surgery, 2=radiation, 3=hormone, 4=combination) High suppression of tumor immunity is defined by the median score in the NCI-Maryland control population ( \(>\) median vs. \(\leq\) median)
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupplementaryMaterial.pdf CXUsersldq5835DesktopSupplementaryTable14.xlsx
+
+<--- Page Split --->

preprint/preprint__485373ce90d30c4730644ea71cd7df747a03d71497007b34dcac392a1a08f4a3/preprint__485373ce90d30c4730644ea71cd7df747a03d71497007b34dcac392a1a08f4a3_det.mmd

@@ -0,0 +1,422 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 879, 175]]<|/det|>
+# Serum proteomics links suppression of tumor immunity to ancestry and lethal prostate cancer  
+
+<|ref|>text<|/ref|><|det|>[[44, 196, 295, 236]]<|/det|>
+Tsion Minas  Center for Cancer Research  
+
+<|ref|>text<|/ref|><|det|>[[44, 242, 630, 283]]<|/det|>
+Julian Candia  National Cancer Institute  https://orcid.org/0000- 0001- 5793- 8989  
+
+<|ref|>text<|/ref|><|det|>[[44, 289, 441, 329]]<|/det|>
+Tiffany Dorsey  National Cancer Institute, NIH, Bethesda, MD  
+
+<|ref|>text<|/ref|><|det|>[[44, 335, 944, 398]]<|/det|>
+Francine Baker  Laboratory of Translational Genomics, Division of Cancer Epidemiology and Genetics, National Cancer Institute (NCI), National Institutes of Health (NIH)  https://orcid.org/0000- 0003- 3133- 3652  
+
+<|ref|>text<|/ref|><|det|>[[44, 404, 485, 444]]<|/det|>
+Wei Tang  NCI/NIH  https://orcid.org/0000- 0002- 7089- 4391  
+
+<|ref|>text<|/ref|><|det|>[[44, 450, 901, 515]]<|/det|>
+Maeve Kiely  Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  
+
+<|ref|>text<|/ref|><|det|>[[44, 520, 293, 560]]<|/det|>
+Cheryl Smith  Center for Cancer Research  
+
+<|ref|>text<|/ref|><|det|>[[44, 567, 293, 606]]<|/det|>
+Symone Jordan  Center for Cancer Research  
+
+<|ref|>text<|/ref|><|det|>[[44, 612, 901, 676]]<|/det|>
+Obadi Obadi  Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  
+
+<|ref|>text<|/ref|><|det|>[[44, 682, 901, 745]]<|/det|>
+Anuoluwapo Ajao  Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  
+
+<|ref|>text<|/ref|><|det|>[[44, 751, 368, 792]]<|/det|>
+Yao Tettey  University of Ghana Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 797, 368, 838]]<|/det|>
+Richard Biritwum  University of Ghana Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 844, 368, 885]]<|/det|>
+Andrew Adjei  University of Ghana Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 890, 368, 930]]<|/det|>
+James Mensah  University of Ghana Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 937, 170, 954]]<|/det|>
+Robert Hoover
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 46, 272, 64]]<|/det|>
+National Cancer Institute  
+
+<|ref|>text<|/ref|><|det|>[[44, 71, 463, 112]]<|/det|>
+Frank Jenkins  Hillman Cancer Center, University of Pittsburgh  
+
+<|ref|>text<|/ref|><|det|>[[44, 117, 525, 159]]<|/det|>
+Rick Kittles  City of Hope  https://orcid.org/0000- 0002- 5004- 4437  
+
+<|ref|>text<|/ref|><|det|>[[44, 164, 137, 202]]<|/det|>
+Ann Hsing  Stanford  
+
+<|ref|>text<|/ref|><|det|>[[44, 209, 130, 228]]<|/det|>
+Xin Wang  
+
+<|ref|>text<|/ref|><|det|>[[44, 230, 901, 273]]<|/det|>
+Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  
+
+<|ref|>text<|/ref|><|det|>[[44, 277, 230, 316]]<|/det|>
+Christopher Loffredo  Georgetown University  
+
+<|ref|>text<|/ref|><|det|>[[44, 323, 586, 365]]<|/det|>
+Clayton Yates  Tuskegee University  https://orcid.org/0000- 0001- 5420- 9852  
+
+<|ref|>text<|/ref|><|det|>[[44, 370, 272, 409]]<|/det|>
+Michael Cook  National Cancer Institute  
+
+<|ref|>text<|/ref|><|det|>[[44, 414, 400, 435]]<|/det|>
+Stefan Ambs ( ambss@mail.nih.gov )  
+
+<|ref|>text<|/ref|><|det|>[[44, 437, 901, 481]]<|/det|>
+Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute (NCI), National Institutes of Health (NIH)  https://orcid.org/0000- 0001- 7651- 9309  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 521, 102, 540]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 558, 914, 602]]<|/det|>
+Keywords: Proteomics, inflammation, prostate cancer, ancestry, survival, disparity, immune signature, African, European  
+
+<|ref|>text<|/ref|><|det|>[[44, 619, 293, 639]]<|/det|>
+Posted Date: July 13th, 2021  
+
+<|ref|>text<|/ref|><|det|>[[44, 657, 463, 678]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 668276/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 694, 910, 739]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 773, 949, 818]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on April 1st, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 29235- 2.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[77, 92, 880, 135]]<|/det|>
+# Serum proteomics links suppression of tumor immunity to ancestry and lethal prostate cancer (13 words, 92 characters)  
+
+<|ref|>text<|/ref|><|det|>[[57, 140, 890, 235]]<|/det|>
+T sion Zewdu Minas \(^{1\dagger}\) , Julián Candia \(^{1\dagger}\) , Tiffany H. Dorsey \(^{1}\) , Francine Baker \(^{1}\) , Wei Tang \(^{1}\) , Maeve Kiely \(^{1}\) , Cheryl J. Smith \(^{1}\) , Symone V. Jordan \(^{1}\) , Obadi M. Obadi \(^{1}\) , Anuoluwapo Ajao \(^{1}\) , Yao Tetty \(^{2}\) , Richard B. Biritwum \(^{2}\) , Andrew A. Adjei \(^{2}\) , James E. Mensah \(^{2}\) , Robert N. Hoover \(^{3}\) , Frank J. Jenkins \(^{4}\) , Rick Kittles \(^{5}\) , Ann W. Hsing \(^{6,7}\) , Xin W. Wang \(^{1,8}\) , Christopher A. Loffredo \(^{9}\) , Clayton Yates \(^{10}\) , Michael B. Cook \(^{3}\) , and Stefan Ambs \(^{1*}\)  
+
+<|ref|>text<|/ref|><|det|>[[55, 245, 886, 525]]<|/det|>
+\(^{1}\) Laboratory of Human Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA \(^{2}\) University of Ghana Medical School, Accra, Ghana \(^{3}\) Division of Cancer Epidemiology & Genetics, National Cancer Institute, National Institutes of Health, MD, USA \(^{4}\) Hillman Cancer Center, University of Pittsburgh, Pittsburgh, PA, USA \(^{5}\) Division of Health Equities, Department of Population Sciences, City of Hope Comprehensive Cancer Center, Duarte, CA, USA \(^{6}\) Stanford Cancer Institute, Stanford School of Medicine, Palo Alto, CA, USA \(^{7}\) Stanford Prevention Research Center, Stanford School of Medicine, Palo Alto, CA, USA \(^{8}\) Liver Cancer Program, Center for Cancer Research, National Cancer Institute, Natioinal Institutes of Health, Bethesda, MD, USA \(^{9}\) Cancer Prevention and Control Program, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, Washington, DC, USA \(^{10}\) Center for Cancer Research, Tuskegee University, Tuskegee, AL, USA \(^{1*}\) These authors contributed equally  
+
+<|ref|>text<|/ref|><|det|>[[57, 550, 692, 569]]<|/det|>
+Running title: Immune- oncological markers and prostate cancer disparity  
+
+<|ref|>text<|/ref|><|det|>[[57, 585, 844, 622]]<|/det|>
+Key words: Proteomics, inflammation, prostate cancer, ancestry, survival, disparity, immune signature, African, European  
+
+<|ref|>text<|/ref|><|det|>[[57, 630, 756, 667]]<|/det|>
+Abbreviations: AA, African- American; EA, European- American; OR, odds ratio; CI, confidence interval; PSA, prostate- specific antigen.  
+
+<|ref|>text<|/ref|><|det|>[[57, 700, 775, 754]]<|/det|>
+\*Corresponding Author: Stefan Ambs, Laboratory of Human Carcinogenesis, National Cancer Institute, Bldg.37/Room 3050B, Bethesda, MD 20892- 4258, Phone: 240- 760- 6836; Email: ambss@mail.nih.gov.  
+
+<|ref|>text<|/ref|><|det|>[[57, 780, 866, 818]]<|/det|>
+One sentence summary: A serum proteome- based immune function signature is upregulated in men of African ancestry and associates with lethal prostate cancer.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[57, 87, 290, 104]]<|/det|>
+## 40 Abstract (156 words):  
+
+<|ref|>text<|/ref|><|det|>[[100, 130, 890, 460]]<|/det|>
+There is evidence that tumor immunobiology and immunotherapy response may differ between African American and European American prostate cancer patients. Here, we determined if men of African descent harbor a unique systemic immune- oncological signature and measured 82 circulating proteins in almost 3000 Ghanaian, African American, and European American men. Protein signatures for suppression of tumor immunity and chemotaxis were significantly elevated in men of West African ancestry. Importantly, the suppression of tumor immunity protein signature associated with metastatic and lethal prostate cancer, pointing to clinical significance. Moreover, two markers, pleiotrophin and TNFRSF9, predicted poor disease survival specifically among African American men. These findings indicate that immune- oncology marker profiles differ between men of African and European descent. These differences may contribute to the disproportionate burden of lethal prostate cancer in men of African ancestry. The elevated peripheral suppression of tumor immunity may have important implication for guidance of cancer therapy which could particularly benefit African American patients.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[101, 129, 876, 325]]<|/det|>
+Men of African origin bear the highest prostate cancer burden in the U.S. and globally<sup>1-3</sup>. They are at an increased risk of developing fatal prostate cancer in the U.S and England<sup>4</sup> and present with more aggressive disease in the Caribbean and sub- Saharan Africa<sup>2,5</sup>. The reasons for the observed global prostate cancer health disparities are unclear but may be related to an array of factors such as access to health care, lifestyle and environment, and ancestral and biological factors<sup>6-8</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[101, 348, 890, 614]]<|/det|>
+Previously, we and others described that tumor immunobiology differs between African- American (AA) and European- American (EA) prostate cancer patients<sup>9-12</sup>. A tumor- specific immune- inflammation gene expression signature was more prevalent in prostate tumors of AA than EA patients<sup>11</sup>. The occurrence of this signature in prostate tumors was associated with decreased recurrence- free survival<sup>13</sup>. Furthermore, regular use of aspirin, an anti- inflammatory drug, may reduce the risk of aggressive prostate cancer, disease recurrence and the lethal disease in AA men<sup>14,15</sup>. Combined, these findings suggest that inflammation and host immunity may contribute to prostate cancer progression but with notable differences between AA and EA men.  
+
+<|ref|>text<|/ref|><|det|>[[101, 637, 890, 904]]<|/det|>
+Ancestral factors can influence immune- related pathways<sup>16</sup>. Germline genetic variant prevalence and alternative splicing in immune- inflammation- related genes can show large differences amongst population groups<sup>17-19</sup>. Hence, the immune- inflammation gene expression signature identified in the tumors of AA prostate cancer patients could be due to either tumor biology and the associated microenvironment, ancestral factors, or systemic differences in immunology marker expression. In the present study, we tested the hypothesis that a distinct systemic immune- inflammation signature exists in men of African ancestry that associates with prostate cancer. It is the novelty of our approach that we examined the serum proteome in a large cohort of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 84, 890, 245]]<|/det|>
+77 diverse men. Applying large- scale proteomics with Olink technology, we discovered the up- regulation of circulating immune- oncological proteins that functionally relate to chemotaxis and suppression of tumor immunity and their association with West African ancestry and lethal prostate cancer. Our findings point to the clinical importance of a serum proteomic signature in prostate cancer patients that may affect men of African ancestry more so than other men.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 90, 208, 108]]<|/det|>
+## RESULTS  
+
+<|ref|>text<|/ref|><|det|>[[111, 120, 886, 880]]<|/det|>
+Large- scale evaluation of immune- oncological proteins in the NCI- Maryland and NCI- Ghana prostate cancer studies. To investigate if men of African descent are differentially affected by a systemic immune inflammation, we utilized two case- control studies with large representations of men of African ancestry: the NCI- Ghana and NCI- Maryland Prostate Cancer Case- Control Studies. Characteristics of the participants in the two studies have been previously described<sup>14,20</sup>. We assayed 92 circulating immune- oncological proteins in a total of 3094 serum samples containing 1505 controls and 1432 cases along with 157 randomly selected blinded duplicates. To control for any batch effects, the serum samples were assayed in a random order along with the 5% blind duplicates for intensity normalization (see Methods). Ninety- five percent of the samples passed stringent quality control leaving 1482 controls (654 Ghanaian, 374 AA, and 454 EA) and 1308 cases (489 Ghanaian, 394 AA, and 425 EA) for our analysis (table S1). The average intra- and inter- plate CV calculated based on duplicates were very low at 1.7% and 2.6%, respectively. In addition, the proportion of variance explained by an inter- plate batch effect was rather minimal for each of the serum proteins even before intensity normalization (fig. S1). Out of the 92 serum proteins, 61 were detected in 100% of the samples tested and 78 were detected in > 50% of the samples (fig. S2). Because 10 out of the 92 serum proteins were detected in less than 20% of the samples (fig. S2), only the remaining 82 proteins were included in our analysis (table S2). Next, we assessed how the 82 serum markers correlate with one another in Ghanaian, AA, and EA men without prostate cancer using Pearson's pairwise correlation analysis (fig. S3). The top ten observed correlations for each population group is presented in table S3. Most of these relationships have not previously been described. Most notably, epidermal growth factor levels strongly correlated with CD40L [Ghanaian (r=0.71), AA (r=0.83), and EA (r=0.80) men], a marker
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 884, 144]]<|/det|>
+of activated T cells, whereas IL8 levels highly correlated with circulating caspase 8 in all three population groups [Ghanaian (r=0.69), AA (r=0.82), and EA (r=0.80)].  
+
+<|ref|>text<|/ref|><|det|>[[111, 158, 886, 458]]<|/det|>
+Clinical and socio- demographic characteristics are associated with immune- oncological proteins. Cytokine levels can be influenced by environmental exposures and disease. Therefore, we investigated the association between various socio- demographic and clinical characteristics (age, BMI, education, aspirin use, smoking, diabetes and PSA) with serum levels of immunoonological proteins using a multivariable linear regression model (Fig. 1). We restricted this analysis to the control population in the NCI- Ghana and NCI- Maryland studies to exclude the potential confounding effect of prostate cancer in the analysis. Among the exposures, aspirin use and blood PSA levels showed only few relationships with the profile of the 82 immune- oncology markers. Other exposures and several demographics showed more robust relationships.  
+
+<|ref|>text<|/ref|><|det|>[[111, 472, 886, 877]]<|/det|>
+Aging is known to impact the immune system and is a risk factor for many diseases including cancer<sup>21</sup>. In our analysis, aging was most consistently associated with the level of the analytes across the three population groups, showing a significant correlation with almost half of these circulating immune- oncological proteins. For example, PGF, CXCL9, Gal9, Gal1, CX3CL1, TNFRSF12A, CCL23, MMP7, DCN, MMP12, CXCL13, CSF1, ADGRG1, CD4, and PTN positively associated with age in all three population groups. The top- ranked biological functions that associated with these age- related proteins were cell migration and positive regulation of cell adhesion (fig. S4A). Age was also positively associated with lymphocyte activation, represented by TNFRSF9, CRTAM, PDCD1, CD27, NCR1, TNFRSF4, KLRD1, CD83, IL12, and IL12RB1, but only in the NCI- Maryland EA and AA men (fig. S4B). On the other hand, hepatocyte growth factor (HGF) and vascular endothelial growth factor- A (VEGFA), two angiogenic cytokines, were positively associated with age exclusively in men of African ancestry (Ghanaian and AA men).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 883, 143]]<|/det|>
+Lastly, VEGFR2, a tyrosine kinase receptor for VEGF, was negatively associated with age in EA and AA men.  
+
+<|ref|>text<|/ref|><|det|>[[111, 157, 886, 670]]<|/det|>
+In contrast to the positive association of many of the immune- oncological proteins with age, BMI tended to be negatively associated with these circulating immune- oncological analytes. This finding may be surprising as obesity is generally thought to be associated with systemic inflammation. CX3CL1 was negatively associated with BMI in all three population groups. The soluble form of CX3CL1 stimulates recruitment of CX3CR1 expressing inflammatory immune cells<sup>22</sup>. CAIX and LAMP3 were inversely associated with BMI exclusively in men of African ancestry, suggesting that ancestral factors may influence the relationship between BMI and expression of these markers. CAIX is a hypoxia regulated metalloenzyme that exists as both membrane associated and soluble form<sup>23</sup> whose main cellular function is to catalyze the reversible conversion of carbon dioxide to carbonic acid<sup>24</sup>, thereby influencing local acidity, which is known to affect immune function<sup>25</sup>. LAMP3 is a member of lysosomal associated membrane glycoprotein family that have a myriad of roles including lysosomal exocytosis and cholesterol homeostasis<sup>26</sup>. On the contrary, serum GAL1, a glycan binding protein that mediates the suppressive function of \(\mathrm{T}_{\mathrm{Reg}}\) cells<sup>27</sup>, showed the opposite trend and was positively associated with BMI in all three population groups.  
+
+<|ref|>text<|/ref|><|det|>[[112, 681, 886, 840]]<|/det|>
+To explore how the social/behavioral environment may affect immune- oncological serum protein levels, we investigated their relationship with educational attainment. For Ghanaian men, 27 of the 82 immuno- oncological markers were negatively associated with their education level (Fig. 1). Yet only IL18 showed a significant inverse association with education for both Ghanaian and AA men. Among EA men, 12 of the 82 immune- oncological proteins had significant inverse
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 884, 143]]<|/det|>
+relationships with the attained level of education (Fig. 1), with some of these markers showing a similar pattern among Ghanaian and EA men.  
+
+<|ref|>text<|/ref|><|det|>[[112, 157, 886, 320]]<|/det|>
+Previous studies have shown that tobacco smoking increases inflammation<sup>28</sup>. Herein, we assessed the association between cigarette use (never, former, vs. current smoker) on the level of immune- oncological proteins in circulation. We found that current smoking was consistently associated with significantly increased level of analytes that regulate angiogenesis (ANGPT2), antigen presentation (CD83), and autophagy (LAMP3), in all three study populations (Fig. 1).  
+
+<|ref|>text<|/ref|><|det|>[[111, 331, 886, 880]]<|/det|>
+Innate immune system- driven inflammatory processes have been implicated in the pathogenesis of diabetes<sup>29</sup>. In our analysis, among the cytokines that showed an association with self- reported diabetes, serum proteins belonging to tumor necrosis factor receptor super family (CD27 and TNFRSF12A), and a matrix metalloprotease enzyme (MMP7) were positively associated with diabetes in all three population groups (Fig. 1). Others, including PGF, CX3CL1, NCR1, TNFRSF4, and TNFRSF21 were positively associated with diabetes exclusively in men with African ancestry. Functional enrichment analysis revealed that diabetes- associated CX3CL1, TNFRSF4, and TNFRSF21 are all involved in negative regulation of cytokine secretion (fig. S5). CX3CL1 is known to regulate insulin secretion<sup>30</sup>, is elevated in the serum of patients with type 2 diabetes<sup>31</sup>, and has been implicated in diabetic nephropathy<sup>32</sup>, validating the findings in our study. C- reactive protein (CRP) is a commonly measured pro- inflammatory marker in the body and has been reported to be associated with worse prostate cancer prognosis<sup>33,34</sup>. Because it was not part of our marker panel, we measured blood CRP in 156 plasma samples from population controls of the NCI- Maryland study. Smoking was the only socio- demographic variable that had a significant association with CRP (table S4), which is consistent with the literature. Furthermore, CRP showed significant positive associations with 24 of the 82 serum proteins (TNFRSF9, IL7,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 884, 144]]<|/det|>
+PGF, IL6, Gal9, GZMH, CXCL1, TNFSF14, Gal1, PDL1, HGF, HO1, CD70, TNFRSF12A, CCL3, MMP7, ANGPT2, VEGFA, CCL20, KLRD1, CSF1, CD4, MCP3, and CXCL11).  
+
+<|ref|>text<|/ref|><|det|>[[111, 157, 884, 423]]<|/det|>
+The systemic immune- oncological cytokine profile in men of African ancestry is distinct from men of European ancestry. To investigate if ancestral population group differences may influence circulating levels of the immune- oncological markers, we performed an unsupervised clustering analysis examining how the levels of the 82 immune- oncological analytes would group men without prostate cancer from Ghana and the US. Notably, these analytes tended to cluster by population group, with levels in Ghanaian men being most distant from EA men while AA samples tended to cluster in between these two groups (Fig. 2), suggesting that the ancestral background may have a significant impact on this immune- oncological protein profile.  
+
+<|ref|>text<|/ref|><|det|>[[111, 435, 885, 877]]<|/det|>
+To further evaluate the influence of ancestry, we estimated West African ancestry in AA and EA population controls of the NCI- Maryland study and its relationship with the cytokine profile. West African ancestry was determined using 100 validated ancestry informative markers \(^{35}\) . The approach showed that, to some extent, the variance in the levels of several immune- oncological analytes is strongly influenced by the degree of West African ancestry of these individuals (Fig. 3A). The variance in 45 of the analytes were significantly \((P< 0.05)\) influenced by degree of West African ancestry (table S5). The levels of 42 analytes were significantly accounted for by West African ancestry even after adjusting for age, BMI, aspirin use, education, income, diabetes, and smoking status (table S6). CXCL5, CXCL1, MCP2, MCP1, CXCL11, CCL23, PTN, TWEAK, NCR1, IL18 and CCL17 were the top- ranked proteins (tables S5- S6). Adjusting the significance threshold by Bonferroni \((P_{\mathrm{B}} = 0.05 / 82 = 0.00061)\) , which is the most stringent criterion to adjust for multiple testing, the relationship of the top 28 proteins with West African ancestry remained significant. For instance, \(41\%\) and \(50\%\) of the variance in the serum
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 88, 886, 355]]<|/det|>
+levels of CXCL1 and CXCL5, respectively, was accounted for by the degree of West African ancestry (Fig. 3A and tables S5- S6). When we compared the levels of these proteins across the 3 population groups, we observed a significant African ancestry- related trend (Fig. 3B), with 12 of the 82 circulating immune- oncological proteins (CXCL5, CXCL1, CXCL11, MCP2, CCL17, MCP4, CD70, MMP12, PDL2, MMP7, CCL19, and ANGPT2) being significantly elevated in both Ghanaian and AA men compared to EA men (table S7); twelve other markers (MCP1, IL12, CCL23, CD8A, NCR1, TNFRSF4, TNFSF14, TWEAK, IL7, HGF, HO1, TNFRSF21, and ANG1) were inversely related to West African ancestry (table S8).  
+
+<|ref|>text<|/ref|><|det|>[[110, 367, 880, 877]]<|/det|>
+Cytokines associated with suppression of tumor immunity and chemotaxis are upregulated in men of African ancestry. Levels of many of the 82 immune- oncology markers showed a marked association with ancestry. To better define the functional implications of these population group differences, we grouped the 82 proteins into six biological processes according to Olink guidelines (table S9): apoptosis/cell killing, autophagy/metabolism, chemotaxis/trafficking to tumor, suppression of tumor immunity (Th2 response, tolerogenic), promotion of tumor immunity (Th1 responses), or vasculature and tissue remodeling. To gain insight on how activation of these six processes/pathways may differ by population group, we compared process/pathway sum scores between Ghanaian, AA, and EA men without prostate cancer. Of these pathways, chemotaxis, promotion of tumor immunity, and suppression of tumor immunity were significantly different in their predicted activity between AA and EA men (Fig. 4). AA men had significantly higher scores for chemotaxis and suppression of tumor immunity when compared to EA men, indicating higher activity in AA men, but a lower score for promotion of tumor immunity. Ghanaian men had even higher scores for chemotaxis and suppression of tumor immunity than both AA and EA men (Fig. 4C and E), indicating a
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 88, 881, 895]]<|/det|>
+possible association with West African ancestry. The latter was corroborated with our finding that the chemotaxis and suppression of tumor immunity scores positively correlated with the proportion of West African ancestry within the NCI- Maryland cohort (Spearman’s rho=0.23, \(P< 0.001\) , for chemotaxis score; Spearman’s rho=0.15, \(P< 0.001\) , for suppression of immunity score). Even though apoptosis and vasculature- associated cytokines were not significantly different between EA and AA men, we found both processes to be elevated in the Ghanaian men. Suppression of tumor immunity is associated with reduced survival of prostate cancer patients. Next, we examined the clinical implication of our findings and assessed the association of pathway activity with survival of prostate cancer cases or controls in the NCI- Maryland study. As of the end of 2018, out of the 819 cases, there have been 202 deaths in our case population, of whom 103 (51%) had a cancer diagnosis as the recorded primary cause of death, and 28% of all deaths ( \(n = 57\) ) were directly attributed to prostate cancer. On the other hand, 99 of the 828 population controls had died by the end of 2018. Median survival follow- up for cases and controls were 8.6 and 6.7 years, respectively. With these data, we built a multivariable Cox regression model with all biological processes/pathways and adjustment for other covariables (see Methods). Among the six defined pathways, only suppression of tumor immunity showed independent association with survival of cases (Fig. 5). Prostate cancer patients with an increased activity of this pathway had the highest risk of death from all causes (Fig. 5A, table S10). Moreover, prostate cancer patients with elevated suppression of tumor immunity at diagnosis had the highest risk of prostate cancer- specific mortality, albeit marginally significant ( \(P = 0.057\) ) (Fig. 5B, table S11). Notably, suppression of tumor immunity was not associated with all- cause mortality of population controls (table S12), suggesting that the association with all- cause mortality among cases might be prostate cancer- related. Lastly, prostate cancer
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 875, 214]]<|/det|>
+patients with increased suppression of tumor immunity were also significantly more likely to die from any cancer (prostate cancer or secondary cancer) following the prostate cancer diagnosis (Fig. 5C, table S13), indicating a more general predisposition to cancer in patients with a high suppression of tumor immunity score.  
+
+<|ref|>text<|/ref|><|det|>[[111, 238, 880, 640]]<|/det|>
+Elevated suppression of tumor immunity is associated with metastatic prostate cancer. To further corroborate the significance of suppression of tumor immunity in the development of lethal prostate cancer, we assessed its association with prostate cancer aggressiveness per NCCN guidelines (see Methods). Information on TNM stage was only obtainable for the NCI- Maryland prostate cancer patients, hence only these cases were scored according to the NCCN guidelines. Patients with a high suppression of tumor immunity score were at substantially increased odds of being diagnosed with regional or distant metastasis (HR 3.79, 95% CI 1.59- 9.04, > median vs. \(\leq\) median) (Table 1), consistent with the disease survival data. The data showed a significant trend in the association of elevated suppression of tumor immunity with disease aggressiveness ( \(P\) trend=0.004) (Table 1). Stratified analysis by race/ethnicity revealed that high suppression of tumor immunity was associated with metastatic prostate cancer more so among AA than EA men.  
+
+<|ref|>text<|/ref|><|det|>[[111, 666, 877, 891]]<|/det|>
+Blood levels of TNFRSF9 and pleiotrophin (PTN) predict lethal prostate cancer among AA men. To identify individual drivers of the relationship between immune- oncology markers and lethal prostate cancer, we applied a cross- validated, regularized Cox regression model using eNetXplorer (see Methods). Included in this model were the 82 immune- oncology markers and 6 covariates of clinical significance (age, education, BMI, smoking history, aspirin use, and diabetes). Lasso regression (alpha=1) was selected based on overall performance (fig. S6). Utilizing this method, we could not identify a robust predictive signature of lethal prostate cancer
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 88, 881, 565]]<|/det|>
+for EA patients. However, for AA patients, a signature primarily driven by TNFRSF9 and PTN (both positively associated with the risk of lethal disease) and regular aspirin use (negatively associated with risk) were the top predictors ( \(P < 0.05\) ) based on two selection criteria: the feature frequency (Fig. 6A) and the weight of the features' contribution to the prediction (Fig. 6B). These features combined predicted prostate cancer- specific mortality with an accuracy of \(83.7\%\) (SE=3.8%). Our finding that regular aspirin use was a predictor of improved survival among AA men is consistent with previously published data from this case- control study \(^{14}\) and the Southern Community Cohort Study \(^{15}\) . The two proteins alone, TNFRSF9 and PTN, predicted prostate cancer- specific mortality with \(78.2\%\) (SE=4.2%) accuracy. AA prostate cancer patients with high levels (> median) of both TNFRSF9 and PTN in their blood at diagnosis had the worst prostate cancer- specific survival (Fig. 6C). By 10 years, 33% of cases with high levels of both TNFRSF9 and PTN died of prostate cancer compared to only 5% of cases with low levels of both or either of these proteins (Fig. 6C), highlighting the utility of these blood markers for risk stratification of AA prostate cancer patients.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 90, 238, 108]]<|/det|>
+## DISCUSSION  
+
+<|ref|>text<|/ref|><|det|>[[112, 133, 883, 365]]<|/det|>
+In this study, we describe consistent differences in the expression of immune and chemotaxis- related markers in men from three population groups, with two of them - AA and Ghanaian men - having an ancestral relationship due to the trans- Atlantic slave trade. Most notably, expression of immune- oncology markers related to immune suppression were up- regulated in men of West African ancestry and were associated with lethal prostate cancer. While ancestry can explain some of the observations, other and yet unknown factors may contribute to these clinically significant differences in immune function and chemotaxis.  
+
+<|ref|>text<|/ref|><|det|>[[111, 377, 880, 890]]<|/det|>
+Infections endemic to certain regions have shaped the immune response in affected populations, leaving a lasting genetic and epigenetic footprint<sup>36</sup>. As such, population differences in exposures to fatal pathogens have led to population heterogeneity in the immune. It has been estimated that as many as 360 immune- related genes have been targets of positive selection and have functional variations between populations<sup>37</sup>. Consistent with these observations, we now report population differences in circulating immune- oncological proteins among Ghanaian, AA, and EA men. We found that the serum proteome- defined immunome of Ghanaian men resembles the immunome of AA men more so than EA men. We identified CXCL5, CXCL1, MCP2, MCP1, and CXCL11 as the top immune- oncological proteins associated with West African ancestry. Four of these chemokines (CXCL5, CXCL1, MCP1, and CXCL11) are targets of Duffy Antigen Receptor for Chemokines (DARC) binding<sup>38</sup>. DARC is a non- signaling receptor that binds to both CXC and CC family of chemokines and acts as a depot for chemokines on erythrocytes and as decoy receptor on endothelial cells<sup>39</sup>. DARC expression modulates the susceptibility to clinical Plasmodium vivax malaria and loss of its expression on erythrocytes, which frequently occurs in sub- Saharan African populations due to germline
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 87, 880, 283]]<|/det|>
+genetic variants, confers resistance against malarial infection<sup>40</sup>. Its loss may also influence cancer susceptibility<sup>41,42</sup>. Consequently, these individuals lack the ability to sequester the target chemokines, leading to elevated concentration of the chemokines in circulation<sup>43</sup>. Accordingly, we found that CXCL5, CXCL1, and CXCL11 were 2- 3- fold higher in sera of Ghanaian or AA men than EA men. Given the angiogenic properties of these chemokines<sup>44</sup>, their role in cancer progression has been proposed<sup>45</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[111, 297, 880, 877]]<|/det|>
+As a key finding, we report that serum proteins driving chemotaxis and suppression of tumor immunity were elevated in men of African ancestry, suggesting persistent population differences in stimulation of leukocyte recruitment and T cell mediated immune response. Such differences may predispose men of African descent to a distinct tumor microenvironment. Although the direct impact of the peripheral immunome on the prostate tumor microenvironment requires further investigation, we and others have previously reported stark differences in the immune landscape of prostate tumors of AA men as compared to EA men<sup>9- 13</sup>. For instance, programmed cell death ligand- 1 (PD- L1), which suppresses T cell- mediated tumor immunity, was found to be overexpressed in AA prostate tumors<sup>46</sup>. Recent work by Awasthi et al. reported that AA prostate tumors tend to be enriched for immune pathways that are associated with poor clinical outcomes<sup>47</sup>. We show with our current work that elevated, peripheral suppression of tumor immunity associates with lethal prostate cancer. Hence, population differences in suppression of tumor immunity may contribute to the disproportionate burden of lethal prostate cancer among men of African ancestry. On the other hand, such differences may offer a therapeutic advantage for immunotherapeutic strategies that are tailored to target immune suppressive pathways. A recent study provided a first indication that differences in the response to cancer vaccines may lead to higher survival rates among AA men<sup>48</sup>.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 881, 775]]<|/det|>
+Differentiating men who have lethal forms of prostate cancer from those with a more slow- growing disease remains a major challenge in clinical oncology. Risk stratification strategies are particularly needed for AA prostate cancer patients who disproportionately bear the prostate cancer burden. This study identified TNFRSF9 and PTN as candidate predictive blood markers for prostate cancer mortality among AA patients. AA patients with high levels of both TNFRSF9 and PTN in their sera had the highest risk of dying from prostate cancer. The membrane form of TNFRSF9 possesses antitumor properties and agonistic anti- TNFRSF9 antibodies are currently in clinical trials<sup>49,50</sup>. On the contrary, the soluble isoform of TNFRSF9 that we measured, generated by alternative splicing<sup>51</sup>, has been shown to antagonize antitumor immune response hence promote tumor survival most likely by acting as decoy receptor<sup>52,53</sup>. Regulatory T cells described as Tregs are thought to be a major source of secreted TNFRSF9<sup>54,55</sup>. Recently, TNFRSF9 mRNA level was shown to be a robust marker of tumor- infiltrating Tregs that suppress antitumor response<sup>56</sup>. Moreover, high numbers of TNFRSF9- expressing Tregs were associated with poor survival outcomes across multiple human cancers<sup>56</sup>, consistent with our findings that serum TNFRSF9 associates with lethal prostate cancer. Although pleiotrophin, the second protein marker associated with lethal prostate cancer in AA men, may not have the same immune function that soluble TNFRSF9 exhibits, it is a secreted cytokine with important roles in promoting angiogenesis and metastasis<sup>57</sup>. Recently, pleiotrophin was described as a serum- based biomarker of pro- metastatic prostate cancer<sup>58</sup>, consistent with our findings in this study.  
+
+<|ref|>text<|/ref|><|det|>[[113, 786, 877, 876]]<|/det|>
+To the best of our knowledge, this is the first study with a large representation of men of African descent who contributed to immune- oncological proteome profiling. With the advent of increasing immunotherapies in the drug development pipeline, such studies may inform research
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 880, 214]]<|/det|>
+on population differences in the immune landscape that need to be considered when designing therapies that exploit the immune response. Furthermore, our study may provide unique insights into variations in the manifestation and pathogenesis of different immune related diseases among different population groups.  
+
+<|ref|>text<|/ref|><|det|>[[111, 228, 852, 353]]<|/det|>
+In conclusion, it is a key finding of our study that suppression of tumor immunity was increased in Ghanaian and AA men, when compared to EA men, and associates with lethal prostate cancer. As such, these findings provide a novel insight into potential causes of the prostate cancer health disparity.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 90, 378, 108]]<|/det|>
+## MATERIALS and METHODS  
+
+<|ref|>text<|/ref|><|det|>[[111, 130, 872, 720]]<|/det|>
+NCI- Maryland prostate cancer case- control study. This study and the eligibility criteria have been previously described<sup>14,59</sup>. Race/ethnicity was assigned based on self- identification as either black or AA or as Caucasian or EA. The study was initiated to test the primary hypothesis that environmental exposures and ancestry- related factors contribute to the excessive prostate cancer burden among AA men. The study was approved by the NCI (protocol # 05- C- N021) and the University of Maryland (protocol #0298229) Institutional Review Boards and all participants signed an informed consent. Cases were recruited at the Baltimore Veterans Affairs Medical Center and the University of Maryland Medical Center. A total of 976 cases (489 AA and 487 EA men) were recruited into this study between 2005 and 2015. Controls were identified through the Maryland Department of Motor Vehicle Administration database and were frequency- matched to cases on age and race. A total of 1,034 population controls were recruited (486 AA and 548 EA men). At the time of enrollment, both cases and controls were administered a survey by a trained interviewer and a blood sample was collected. Serum samples were available for 846 cases (407 AA and 439 EA) and 846 controls (382 AA and 464 EA), therefore only these individuals were used for the study herein. Most of the 846 cases (85%) were recruited within a year of the disease diagnosis with a median of 5.1 months between disease diagnosis and blood collection.  
+
+<|ref|>text<|/ref|><|det|>[[112, 727, 857, 886]]<|/det|>
+NCI- Ghana prostate cancer case- control study. This case- control study has been previously described<sup>20</sup>. The study was designed to study lifestyle, environmental, and genetic risk factors for prostate cancer in African men. The study was approved by institutional review boards at the University of Ghana (protocol #001/01- 02) and at the National Cancer Institute (protocol #02CN240). Prior to study enrollment, all participants signed an informed consent. Prostate
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 864, 390]]<|/det|>
+cancer cases were recruited at Korle Bu Teaching Hospital in Accra, Ghana between 2008 and 2012. The cases were diagnosed using Digital Rectal Exam (DRE) and PSA tests, followed by biopsy confirmation. Immediately after diagnosis and before treatment, cases were consented and asked to submit blood specimen and questionnaire data. Controls were identified through probability sampling using the 2000 Ghana Population and Housing Census data to recruit approximately 1,000 men aged 50–74 years in the Greater Accra region between 2004 and 2006. These men were confirmed to not have prostate cancer by PSA testing and DRE. Serum samples were available for 586 prostate cancer cases and 659 population controls; hence, only these individuals were used for the study herein.  
+
+<|ref|>text<|/ref|><|det|>[[111, 403, 872, 596]]<|/det|>
+Serum sample processing. The participants in the two studies provided blood samples at time of recruitment. For the NCI- Maryland study, most blood samples were processed the same day, but always within 48 hours, after storage in a refrigerator. For the NCI- Ghana study, blood samples were processed within 6 hours. In this study, population controls provided fasting blood. Serum was prepared using standard procedures and aliquots were stored at \(- 80^{0}\mathrm{C}\) . Serum samples were shipped from Ghana to the NCI in dry ice boxes.  
+
+<|ref|>text<|/ref|><|det|>[[111, 610, 864, 876]]<|/det|>
+Serum protein measurement. Serum levels of 92 immuno- oncology panel proteins were measured simultaneously using a proprietary multiplex Proximal Extension Assay (PEA) by Olink Proteomics (Boston). Olink utilizes a relative quantification unit, Normalized Protein eXpression (NPX), which is in a Log2- format. Serum samples from NCI- MD study (846 cases and 846 controls) and NCI- Ghana study (586 cases and 659 controls) were completely randomized and were assayed in that order. In addition to the built- in internal controls, \(5\%\) blinded duplicates were randomly selected and were randomized along with the original set of samples. Protein levels were intensity normalized to adjust for batch effect. Because all our
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 875, 320]]<|/det|>
+samples were randomized across plates, a global adjustment was used to center the values for each assay around its median and across all plates. Ninety- five percent of the samples passed a stringent quality control (NCI- MD study: 819 cases and 828 controls; NCI- Ghana study: 489 cases and 654 controls) – with coefficients of variation (CV) among duplicates at \(< 10\%\) for every marker. Out of the 92 proteins assayed, IL33, IL35, IL21, IL2, IFNβ, IL13, IL1α, CXCL12, IFNγ, and TNF were detected in less than \(20\%\) of the samples, hence the remaining 82 proteins were used for subsequent analysis (table S2).  
+
+<|ref|>text<|/ref|><|det|>[[111, 333, 875, 737]]<|/det|>
+Functional annotation and biological processes scores. Proteins were grouped into six biological processes based on their respective biological roles following the Olink guideline: apoptosis/cell killing, autophagy/metabolism, chemotaxis/trafficking to tumor, suppression of tumor immunity (Th2 response, tolerogenic), promotion of tumor immunity (Th1 responses), or vasculature and tissue remodeling. Apoptosis, autophagy, chemotaxis, suppression of tumor immunity, promotion of tumor immunity, or vasculature scores were calculated for each study participant as the mean z- score value for the proteins belonging to the respective biological process. For survival analysis, the biological process/pathway scores were evaluated as continuous variables. To evaluate the association of suppression of tumor immunity with aggressive prostate cancer, we grouped suppression of tumor immunity scores into low (≤median) and high (>median) with cutoffs determined using the distribution of the score among population controls of the NCI- Maryland study.  
+
+<|ref|>text<|/ref|><|det|>[[111, 750, 883, 875]]<|/det|>
+Prostate Specific Antigen (PSA) measurement. For the cases in the NCI- Maryland cohort, PSA levels were obtained from medical record. For the controls of the NCI- Maryland study, total PSA was measured from stored serum aliquots using the human total PSA ELISA Kit (Abcam, ab188388). About \(7\%\) (n=56) of the controls in the NCI- Maryland cohort had PSA greater than
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 864, 179]]<|/det|>
+2.5 ng/ml, while only \(3\%\) (n=27) had blood PSA over 4 ng/ml. For the controls in the NCI- Ghana study, close to \(20\%\) (n=132) had a PSA greater than 2.5 ng/ml, while about \(11\%\) (n=73) had PSA over 4 ng/ml.  
+
+<|ref|>text<|/ref|><|det|>[[111, 203, 885, 469]]<|/det|>
+C- reactive protein (CRP) measurement. Plasma CRP was assayed using an ELISA assay (cat# ab99995, Abcam, United States) according to the manufacturer's instructions. Two microliters of plasma samples were added to \(398 \mu \mathrm{L}\) of 1x Diluent D, followed by a second 1: 200 dilution steps for each sample. One- hundred microliters of CRP standard (0 to \(600 \mathrm{pg / mL}\) ) and the diluted samples were loaded as duplicates into pre- coated 96- well plates. Samples were incubated overnight at \(4^{\circ} \mathrm{C}\) with gentle shaking, followed by incubations with the anti- human CRP antibody and the horseradish peroxidase- streptavidin solution. CRP was quantified measuring absorbance at \(450 \mathrm{nm}\) with a microplate reader.  
+
+<|ref|>text<|/ref|><|det|>[[111, 492, 860, 897]]<|/det|>
+West African ancestry estimation for participants in the NCI- Maryland case- control study. Genomic DNA was isolated from buffy coats (DNeasy Blood & Tissue Kit - Qiagen) or mouthwash samples (standard phenol- chloroform technique). Isolated DNA was genotyped for 100 ancestry informative markers using the Sequenom MassARRAY iPLEX platform, as previously described<sup>35</sup>. Single nucleotide polymorphism genotype calls were generated using Sequenom TYPER software. A genotype concordance rate of \(>99\%\) was observed for all markers. Admixture estimates for each study participant were calculated using a model- based clustering method as implemented in the program STRUCTURE v2.3. We applied STRUCTURE v2.3 with an admixture model estimating K (number of sub populations) from 2 to 5 with 100 iterations and parental population genotypes from West Africans, Europeans, and Native Americans, yielding three admixture estimations (West African, European, Native American). For a subset (83%) of the cases and controls, additional West African ancestry
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 860, 179]]<|/det|>
+estimates were provided by the Cancer Genomics Research Laboratory/NCI- Leidos from a genome- wide association study using the Infinium HumanOmni5- Quad BeadChip array. West African ancestry estimates using the two approaches were very similar (r=0.98).  
+
+<|ref|>text<|/ref|><|det|>[[111, 194, 840, 215]]<|/det|>
+Association of clinical/socio- demographic characteristics with immune- oncological  
+
+<|ref|>text<|/ref|><|det|>[[111, 230, 844, 560]]<|/det|>
+proteins. The association of age, body mass index (BMI), education, aspirin use, smoking, diabetes, and PSA levels with the relative abundance of individual analytes was assessed by means of multivariable linear regression models implemented by the function lm in the base R package stats (version 3.6.1). These variables were chosen because they have either been linked to prostate cancer risk and survival or may influence the status of inflammation and host immunity. An analyte (as response variable) was considered significantly associated with clinical and socio- demographic covariables if the multivariable model yielded \(P < 0.05\) on the F- statistic. If this condition was satisfied, the association between the target analyte and each individual covariable was characterized by the corresponding \(P\) value and coefficient.  
+
+<|ref|>text<|/ref|><|det|>[[111, 578, 844, 701]]<|/det|>
+Analysis of variance. Variance analysis for the levels of each of the 82 immune- oncological cytokines were simultaneously assessed as a function of genetic estimation of West African admixture among men without prostate cancer from the NCI- Maryland study. The analysis was implemented by the function aov in the base R package stats (version 3.6.1).  
+
+<|ref|>text<|/ref|><|det|>[[111, 718, 860, 844]]<|/det|>
+Heatmap plots. Heat map plots were generated using Broad Institute's web- based matrix visualization and analysis platform - Morpheus (https://software.broadinstitute.org/morpheus). To avoid spurious effects from outliers in heatmap plots, each protein's range of abundance values were set to saturate at the \(1^{\text{st}}\) and \(99^{\text{th}}\) percentiles. To account for widely different
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 870, 144]]<|/det|>
+abundance ranges for different proteins in the assay, each protein's measured abundances across the cohort were Z- score transformed.  
+
+<|ref|>text<|/ref|><|det|>[[111, 168, 872, 295]]<|/det|>
+Gene ontology (GO) enrichment analysis. GO terms with an enrichment in proteins of interest were identified using Over- Representation Analysis (ORA) as part of the web tool WebGestalt (WEB- based Gene SeT AnaLysis Toolkit). Enriched gene sets were further processed using affinity propagation (R package apcluster) to cluster gene sets according to functional similarity.  
+
+<|ref|>text<|/ref|><|det|>[[110, 315, 875, 900]]<|/det|>
+Survival analysis. Information on patient survival was only obtainable for the NCI- Maryland prostate cancer patients. Survival data was obtained from the National Death Index (NDI) database. We calculated survival for cases from date of diagnosis to either date of death or to the censor date of December 31, 2018. We built a multivariable Cox regression model with all biological processes scores and adjustment for other covariables to estimate adjusted hazard ratios (HR) and 95% confidence intervals (CI) for all- cause mortality, cancer- related mortality, and prostate cancer- specific mortality of cases. We adjusted for the following potential confounding factors: age at study entry (years), body- mass index (BMI, \(\mathrm{kg} / \mathrm{m}^2\) ), self- reported race (AA/EA), education (high school or less, some college, college, professional school), income (less than \(\) 10k\(,\) \ \(10 - 30K\) , \(\) 30 - 60K\(,\) \ \(60 - 90k\) , greater than \(\) 90k\(), smoking history (never, former, current), diabetes (no/yes), aspirin use (no/yes), and treatment (0=none, 1=surgery, 2=radiotherapy, 3=hormone, 4=combination). Missing values for education (n=1), smoking history (n=5), and income (n=63) were imputed using the R package missForest, which implements nonparametric missing value imputation based on random forests. In the overall survival analysis of population controls, we calculated survival from the date of interview to either date of death or to the censor date of December 31st, 2018. We applied the Cox regression model to estimate adjusted HR and 95% CI and adjusted for all the confounding factors listed
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 860, 214]]<|/det|>
+above except for treatment. Missing values for education (n=1), smoking history (n=7), and income (n=67) were imputed using the R package missForest. The reported HRs indicate the change in risk of dying when the biological process z- score value increases by 1 while holding all the other biological processes' z- scores and covariates constant.  
+
+<|ref|>text<|/ref|><|det|>[[111, 238, 875, 890]]<|/det|>
+Classification of cases using National Comprehensive Cancer Network (NCCN) Risk Score. Cases were assigned to risk groups based on the patients' TNM stage, Gleason score, Gleason pattern, and PSA level at diagnosis according to the 2019 NCCN guideline for prostate cancer<sup>60</sup>. Information on TNM stage was only obtainable for the NCI- Maryland prostate cancer patients, hence only these cases were scored. Cases were categorized as localized, regional, and metastatic prostate cancer based on their clinical parameters at the time of diagnosis. Localized prostate cancer cases were further classified into low, intermediate, high, and very high risk based on the likelihood of their disease to progress to lethal prostate cancer per the 2019 NCCN guideline<sup>60</sup>. Prostate cancer cases with lymph node involvement but no distant metastasis at diagnosis were classified as regional prostate cancer while those with distant metastasis at the time of diagnosis were classified as metastatic prostate cancer. For our analysis, we condensed these risk groups into 4 categories (low, intermediate, high/very high, and regional/metastatic). Developing a predictive proteomic signature of lethal prostate cancer. The analysis was restricted to the cases from NCI- Maryland study for whom we had survival data. We stratified by race/ethnicity into AA cases (360 censored, 34 prostate cancer deaths) and EA cases (402 censored, 23 prostate cancer deaths). To identify a multi- analyte proteomic signature that is predictive of lethal prostate cancer, 88 features were evaluated [82 immune- oncological proteins along with six demographic/clinical variables (education, age, BMI (BMI, kg/m<sup>2</sup>), smoking history, diabetes, and aspirin use)]. Missing values for education (n=1) and smoking history
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 87, 875, 633]]<|/det|>
+(n=5) were imputed using R package missForest. R package eNetXplorer (version 1.1.2)61 was implemented to build cross- validated, regularized Cox regression models with different elastic net mixture parameters from ridge (alpha=0) to lasso (alpha=1). Alpha was selected based on overall performance assessed as a function of the 5- fold cross- validated quality function (concordance) and the empirical \(P\) value generated from comparing the model against a statistical ensemble of null models created by random permutations of the response (i.e. survival time/status randomized across subjects in the cohort). These results comprise 10,000 Cox regression elastic net realizations arising from 200 randomly generated folds, each of them compared against 50 null model permutations. Features' performance as predictors was evaluated using two different, but complementary selection criteria: feature coefficients and feature frequencies. The feature frequency measure captures the significance of how often a feature is chosen in an in- bag model. When it is chosen, the feature coefficient measure captures the significance of the feature's weight in the in- bag model. See the publication by Candia et al for more details on this method61. Using only the significant protein features from both selection criteria, a multivariate Cox regression model was run. Risk stratification was used to generate Kaplan- Meier plots and log- rank tests of significance.  
+
+<|ref|>text<|/ref|><|det|>[[112, 647, 878, 736]]<|/det|>
+Statistical analysis. Data analyses were performed using Stata/SE 16.0 and R statistical software packages. All statistical tests were two- sided, and an association was considered statistically significant with \(P< 0.05\) .
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 90, 338, 108]]<|/det|>
+## Supplementary Materials:  
+
+<|ref|>text<|/ref|><|det|>[[112, 115, 875, 660]]<|/det|>
+Fig. S1. Variance explained by the inter- plate batch effect  Fig. S2. Detection frequencies of 92 immuno- oncological markers measured in 2937 serum samples  Fig. S3. Correlation matrix presenting Pearson pairwise correlations for each of the 82 serum protein pairs  Fig. S4. Gene ontology (GO) enrichment analysis  Fig. S5. Functional enrichment analysis of proteins positively associated with diabetes in men of African ancestry  Fig. S6. Performance of regularized cox regression models across alpha  Table S1. Characteristics of prostate cancer cases and population controls of the NCI- Maryland and NCI- Ghana Study  Table S2. List of 82 Immuno- oncological proteins detected in more than \(20\%\) of the serum samples  Table S3. Top 10 Pearson pairwise correlations in men without prostate cancer  Table S4. The association of blood CRP with clinical/sociodemographic variables estimated using multiple linear regression  Table S5. The fraction of variance in each of the serum proteins explained by degree of West African ancestry  Table S6. The fraction of variance in each of the serum proteins explained by degree of West African ancestry after adjusting for difference in age, bmi, aspirin use, education, diabetes status, smoking, and income  Table S7. Immune oncological markers that are significantly elevated in both Af and AA men compared to EA men  Table S8. Immune oncological markers that are significantly downregulated in both Af and AA men compared to EA men  Table S9. Serum proteins grouped by biological process  Table S10. Effect of biological processes scores on all- cause mortality of prostate cancer patients  Table S11. Effect of biological processes scores on prostate cancer- specific mortality  Table S12. Effect of biological processes scores on all- cause mortality of population controls  Table S13. Effect of biological processes scores on mortality from any cancer following a diagnosis with prostate cancer  
+
+<|ref|>text<|/ref|><|det|>[[115, 672, 880, 760]]<|/det|>
+Acknowledgments: We would like to thank personnel at the University of Maryland and the Baltimore Veterans Administration Hospital for their contributions with the recruitment of participants to the NCI- Maryland study. We would also like to thank Prof. Edward D. Yeboah as the original Ghana PI and Ms. Evelyn Tay as the original Study Manager for the NCI- Ghana study.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 779, 194, 795]]<|/det|>
+## Funding:  
+
+<|ref|>text<|/ref|><|det|>[[115, 796, 851, 864]]<|/det|>
+DoD award W81XWH1810588 (to SA, CY)  U54 CA118623- CY (NCI) and U54- MD007585- 26- CY (NIMHD) (to CY)  Intramural Research Program of the NIH, National Cancer Institute (NCI), Center for Cancer Research and Division of Cancer Epidemiology and Genetics (to SA, MBC)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 90, 880, 420]]<|/det|>
+584 Author contributions: 585 Conceptualization: TZM, CY, MBC, SA 586 Data curation: TZM, THD, MK, CJS, SVJ, OMO, AA, FJJ, RK 587 Formal Analysis: TZM, JC, RK 588 Funding acquisition: CY, MBC, SA 589 Investigation: TZM, JC, FJJ 590 Methodology: TZM, JC, CAL, MBC, SA 591 Project administration: THD, FB 592 Resources: WT, YT, RBB, AAA, JEM, RNH, AWH, MBC, SA 593 Supervision: WT, SA 594 Visualization: TZM, JC 595 Writing - original draft: TZM 596 Writing - review & editing: TZM, FB, WT, MK, CJS, YT, RBB, AAA, JEM, XWW, CAL, CY, 597 MBC, SA 598 Conflicts of interest: The authors declare that they have no competing interests. 600 601 Data and materials availability: All data are available in the main text or the supplementary 602 materials.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 100, 884, 870]]<|/det|>
+604 1 Powell, I. J. Epidemiology and pathophysiology of prostate cancer in African-American men. J Urol. 177, 444- 449 (2007). 606 2 Rebbeck, T. R. et al. Global patterns of prostate cancer incidence, aggressiveness, and mortality in men of african descent. Prostate Cancer 2013, 560857 (2013). 608 3 Culp, M. B., Soerjomataram, I., Efstathiou, J. A., Bray, F. & Jemal, A. Recent Global Patterns in Prostate Cancer Incidence and Mortality Rates. Eur Urol 77, 38- 52, doi:10.1016/j.eururo.2019.08.005 (2020). 611 4 Butler, E. N., Kelly, S. P., Coupland, V. H., Rosenberg, P. S. & Cook, M. B. Fatal prostate cancer incidence trends in the United States and England by race, stage, and treatment. Br J Cancer 123, 487- 494, doi:10.1038/s41416- 020- 0859- x (2020). 614 5 Heyns, C. F., Fisher, M., Lecuona, A. & van der Merwe, A. Prostate cancer among different racial groups in the Western Cape: presenting features and management. S Afr Med J 101, 267- 270, doi:10.7196/samj.4420 (2011). 616 6 Wallace, T. A., Martin, D. N. & Ambs, S. Interactions among genes, tumor biology and the environment in cancer health disparities: examining the evidence on a national and global scale. Carcinogenesis 32, 1107- 1121 (2011). 620 7 Rebbeck, T. R. Prostate Cancer Disparities by Race and Ethnicity: From Nucleotide to Neighborhood. Cold Spring Harb Perspect Med 8, doi:10.1101/cshperspect.a030387 (2018). 623 8 Lachance, J. et al. Genetic Hitchhiking and Population Bottlenecks Contribute to Prostate Cancer Disparities in Men of African Descent. Cancer Res 78, 2432- 2443, doi:10.1158/0008- 5472.CAN- 17- 1550 (2018). 626 9 Hardiman, G. et al. Systems analysis of the prostate transcriptome in African- American men compared with European- American men. Pharmacogenomics 17, 1129- 1143, doi:10.2217/pgs- 2016- 0025 (2016). 629 10 Powell, I. J. et al. Genes associated with prostate cancer are differentially expressed in African American and European American men. Cancer Epidemiol Biomarkers Prev 22, 891- 897, doi:10.1158/1055- 9965.EPI- 12- 1238 (2013). 632 11 Wallace, T. A. et al. Tumor immunobiological differences in prostate cancer between African- American and European- American men. Cancer Res 68, 927- 936, doi:10.1158/0008- 5472.CAN- 07- 2608 (2008). 634 12 Yuan, J. et al. Integrative comparison of the genomic and transcriptomic landscape between prostate cancer patients of predominantly African or European genetic ancestry. PLoS Genet 16, e1008641, doi:10.1371/journal.pgen.1008641 (2020). 638 13 Tang, W. et al. IFNL4- DeltaG Allele Is Associated with an Interferon Signature in Tumors and Survival of African- American Men with Prostate Cancer. Clin Cancer Res 24, 5471- 5481, doi:10.1158/1078- 0432.CCR- 18- 1060 (2018). 641 14 Smith, C. J. et al. Aspirin Use Reduces the Risk of Aggressive Prostate Cancer and Disease Recurrence in African- American Men. Cancer Epidemiol Biomarkers Prev 26, 845- 853, doi:10.1158/1055- 9965.EPI- 16- 1027 (2017). 644 15 Tang, W. et al. Aspirin Use and Prostate Cancer among African American Men in the Southern Community Cohort Study. Cancer Epidemiol Biomarkers Prev, doi:10.1158/1055- 9965.EPI- 19- 0792 (2020).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 90, 870, 860]]<|/det|>
+647 16 Nedelec, Y. et al. Genetic Ancestry and Natural Selection Drive Population Differences in Immune Responses to Pathogens. Cell 167, 657- 669 e621, doi:10.1016/j.cell.2016.09.025 (2016). 650 17 Ness, R. B., Haggerty, C. L., Harger, G. & Ferrell, R. Differential distribution of allelic variants in cytokine genes among African Americans and White Americans. Am J Epidemiol 160, 1033- 1038, doi:10.1093/aje/kwh325 (2004). 653 18 Rotival, M., Quach, H. & Quintana- Murci, L. Defining the genetic and evolutionary architecture of alternative splicing in response to infection. Nat Commun 10, 1671, doi:10.1038/s41467- 019- 09689- 7 (2019). 656 19 Van Dyke, A. L., Cote, M. L., Wenzlaff, A. S., Land, S. & Schwartz, A. G. Cytokine SNPs: Comparison of allele frequencies by race and implications for future studies. Cytokine 46, 236- 244, doi:10.1016/j.cyto.2009.02.003 (2009). 659 20 Cook, M. B. et al. A genome- wide association study of prostate cancer in West African men. Hum Genet 133, 509- 521, doi:10.1007/s00439- 013- 1387- z (2014). 661 21 Nikolich- Zugich, J. The twilight of immunity: emerging concepts in aging of the immune system. Nat Immunol 19, 10- 19, doi:10.1038/s41590- 017- 0006- x (2018). 662 22 Ferretti, E., Pistoia, V. & Corcione, A. Role of fractalkine/CX3CL1 and its receptor in the pathogenesis of inflammatory and malignant diseases with emphasis on B cell malignancies. Mediators Inflamm 2014, 480941, doi:10.1155/2014/480941 (2014). 663 23 Zavada, J., Zavadova, Z., Zat'ovicova, M., Hyrsl, L. & Kawaciuk, I. Soluble form of carbonic anhydrase IX (CA IX) in the serum and urine of renal carcinoma patients. Br J Cancer 89, 1067- 1071, doi:10.1038/sj.bjc.6601264 (2003). 664 24 Supuran, C. T. Carbonic anhydrases: novel therapeutic applications for inhibitors and activators. Nat Rev Drug Discov 7, 168- 181, doi:10.1038/nrd2467 (2008). 665 25 Erra Diaz, F., Dantas, E. & Geffner, J. Unravelling the Interplay between Extracellular Acidosis and Immune Cells. Mediators Inflamm 2018, 1218297, doi:10.1155/2018/1218297 (2018). 666 26 Alessandrini, F., Pezze, L. & Ciribilli, Y. LAMPs: Shedding light on cancer biology. Semin Oncol 44, 239- 253, doi:10.1053/j.seminoncol.2017.10.013 (2017). 667 27 Rabinovich, G. A. & Toscano, M. A. Turning 'sweet' on immunity: galectin- glycan interactions in immune tolerance and inflammation. Nat Rev Immunol 9, 338- 352, doi:10.1038/nri2536 (2009). 668 28 Armson, Y., Shoenfeld, Y. & Amital, H. Effects of tobacco smoke on immunity, inflammation and autoimmunity. J Autoimmun 34, J258- 265, doi:10.1016/j.jaut.2009.12.003 (2010). 669 29 Wada, J. & Makino, H. Innate immunity in diabetes and diabetic nephropathy. Nat Rev Nephrol 12, 13- 26, doi:10.1038/nr neph.2015.175 (2016). 670 30 Lee, Y. S. et al. The fractalkine/CX3CR1 system regulates beta cell function and insulin secretion. Cell 153, 413- 425, doi:10.1016/j.cell.2013.03.001 (2013). 671 31 Shah, R. et al. Fractalkine is a novel human adipocemokine associated with type 2 diabetes. Diabetes 60, 1512- 1518, doi:10.2337/db10- 0956 (2011). 672 32 Navarro- Gonzalez, J. F., Mora- Fernandez, C., Muros de Fuentes, M. & Garcia- Perez, J. Inflammatory molecules and pathways in the pathogenesis of diabetic nephropathy. Nat Rev Nephrol 7, 327- 340, doi:10.1038/nr neph.2011.51 (2011).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 90, 884, 875]]<|/det|>
+Liuz, Z. Q. et al. Prognostic role of C-reactive protein in prostate cancer: a systematic review and meta-analysis. Asian J Androl 16, 467- 471, doi:10.4103/1008- 682X.123686 (2014).Saito, K. & Kihara, K. C- reactive protein as a biomarker for urological cancers. Nat Rev Urol 8, 659- 666, doi:10.1038/nrurol.2011.145 (2011).Al- Alem, U. et al. Association of genetic ancestry with breast cancer in ethnically diverse women from Chicago. PLoS One 9, e112916, doi:10.1371/journal.pone.0112916 (2014).Rotimi, C. N. et al. The genomic landscape of African populations in health and disease. Hum Mol Genet 26, R225- R236, doi:10.1093/hmg/ddx253 (2017).Barreiro, L. B. & Quintana- Murci, L. From evolutionary genetics to human immunology: how selection shapes host defence genes. Nat Rev Genet 11, 17- 30, doi:10.1038/nrg2698 (2010).Gardner, L., Patterson, A. M., Ashton, B. A., Stone, M. A. & Middleton, J. The human Duffy antigen binds selected inflammatory but not homeostatic chemokines. Biochem Biophys Res Commun 321, 306- 312, doi:10.1016/j.bbrc.2004.06.146 (2004).Mantovani, A., Bonecchi, R. & Locati, M. Tuning inflammation and immunity by chemokine sequestration: decoys and more. Nat Rev Immunol 6, 907- 918, doi:10.1038/nri1964 (2006).Miller, L. H., Mason, S. J., Clyde, D. F. & McGinniss, M. H. The resistance factor to Plasmodium vivax in blacks. The Duffy- blood- group genotype, FyFy. N Engl J Med 295, 302- 304, doi:10.1056/NEJM197608052950602 (1976).Jenkins, B. D. et al. Atypical Chemokine Receptor 1 (DARC/ACKR1) in Breast Tumors Is Associated with Survival, Circulating Chemokines, Tumor- Infiltrating Immune Cells, and African Ancestry. Cancer Epidemiol Biomarkers Prev 28, 690- 700, doi:10.1158/1055- 9965.EPI- 18- 0955 (2019).Martini, R. et al. Investigation of Triple- Negative Breast Cancer Risk Alleles in An International African- Enriched Cohort. scientific reports Preprint at: https://www.researchsquare.com/article/rs- 109841/v1 (2021).Yao, S. et al. Genetic ancestry and population differences in levels of inflammatory cytokines in women: Role for evolutionary selection and environmental factors. PLoS Genet 14, e1007368, doi:10.1371/journal.pgen.1007368 (2018).Strieter, R. M. et al. The functional role of the ELR motif in CXC chemokine- mediated angiogenesis. J Biol Chem 270, 27348- 27357, doi:10.1074/jbc.270.45.27348 (1995).Bikfalvi, A. & Billottet, C. The CC and CXC chemokines: major regulators of tumor progression and the tumor microenvironment. Am J Physiol Cell Physiol 318, C542- C554, doi:10.1152/ajpcell.00378.2019 (2020).Calagua, C. et al. Expression of PD- L1 in Hormone- naive and Treated Prostate Cancer Patients Receiving Neoadjuvant Abiraterone Acetate plus Prednisone and Leuprolide. Clin Cancer Res 23, 6812- 6822, doi:10.1158/1078- 0432.CCR- 17- 0807 (2017).Awasthi, S. et al. Comparative Genomics Reveals Distinct Immune- oncologic Pathways in African American Men with Prostate Cancer. Clin Cancer Res 27, 320- 329, doi:10.1158/1078- 0432.CCR- 20- 2925 (2021).Sartor, O. et al. Survival of African- American and Caucasian men after sipuleucel- T immunotherapy: outcomes from the PROCEED registry. Prostate Cancer Prostatic Dis 23, 517- 526, doi:10.1038/s41391- 020- 0213- 7 (2020).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 90, 881, 775]]<|/det|>
+736 49 Segal, N. H. et al. Phase I Study of Single- Agent Utomilumab (PF- 05082566), a 4- 1BB/CD137 Agonist, in Patients with Advanced Cancer. Clin Cancer Res 24, 1816- 1823, doi:10.1158/1078- 0432.CCR- 17- 1922 (2018). 739 50 Qi, X. et al. Optimization of 4- 1BB antibody for cancer immunotherapy by balancing agonistic strength with FcgammaR affinity. Nat Commun 10, 2141, doi:10.1038/s41467- 019- 10088- 1 (2019). 741 51 Michel, J., Langstein, J., Hofstadter, F. & Schwarz, H. A soluble form of CD137 (ILA/4- 1BB), a member of the TNF receptor family, is released by activated lymphocytes and is detectable in sera of patients with rheumatoid arthritis. Eur J Immunol 28, 290- 295, doi:10.1002/(SICI)1521- 4141(199801)28:01<290::AID- IMMU290>3.0.CO;2- S (1998). 746 52 Michel, J. & Schwarz, H. Expression of soluble CD137 correlates with activation- induced cell death of lymphocytes. Cytokine 12, 742- 746, doi:10.1006/cyto.1999.0623 (2000). 748 53 Labiano, S. et al. Hypoxia- induced soluble CD137 in malignant cells blocks CD137L- costimulation as an immune escape mechanism. Oncoimmunology 5, e1062967, doi:10.1080/2162402X.2015.1062967 (2016). 752 54 Itoh, A. et al. Soluble CD137 Ameliorates Acute Type 1 Diabetes by Inducing T Cell Anergy. Front Immunol 10, 2566, doi:10.3389/fimmu.2019.02566 (2019). 754 55 Kachapati, K. et al. The B10 Idd9.3 locus mediates accumulation of functionally superior CD137(+) regulatory T cells in the nonobese diabetic type 1 diabetes model. J Immunol 189, 5001- 5015, doi:10.4049/jimmunol.1101013 (2012). 756 56 Freeman, Z. T. et al. A conserved intratumoral regulatory T cell signature identifies 4- 1BB as a pan- cancer target. J Clin Invest 130, 1405- 1416, doi:10.1172/JCI128672 (2020). 760 57 Papadimitriou, E. et al. Pleiotrophin and its receptor protein tyrosine phosphatase beta/zeta as regulators of angiogenesis and cancer. Biochim Biophys Acta 1866, 252- 265, doi:10.1016/j.bbcan.2016.09.007 (2016). 763 58 Liu, S. et al. Discovery of PTN as a serum- based biomarker of pro- metastatic prostate cancer. Br J Cancer 124, 896- 900, doi:10.1038/s41416- 020- 01200- 0 (2021). 765 59 Minas, T. Z. et al. IFNL4- DeltaG is associated with prostate cancer among men at increased risk of sexually transmitted infections. Commun Biol 1, 191, doi:10.1038/s42003- 018- 0193- 5 (2018). 767 60 Network, N. C. C. The NCCN Clinical Practice Guidelines in Oncology for Prostate Cancer, V4.2019. https://www.nccn.org/professionals/physician_gls/default.aspx#prostate (Accessed November 16, 2020). 772 61 Candia, J. & Tsang, J. S. eNetXplorer: an R package for the quantitative exploration of elastic net families for generalized linear models. BMC Bioinformatics 20, 189, doi:10.1186/s12859- 019- 2778- 5 (2019).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[55, 70, 960, 789]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[35, 800, 945, 963]]<|/det|>
+<center>Fig. 1. Association of socio-demographic and clinical characteristics with systemic immune-oncological proteins in Ghanaian (Af), AA, and EA men without prostate cancer. Association of the 82 immuno-oncological proteins with age, BMI, education, aspirin use, smoking, diabetes, and PSA was assessed in men without prostate cancer using a multivariable linear regression model. An analyte was considered significantly associated with clinical and socio-demographic covariables if the multivariable model yielded a \(P<0.05\) on the F-statistic. Analytes that did not have a significant association with any of the clinical/sociodemographic variables in at least one of the population groups are not presented in the heatmap. Blue represents negative association while red represents positive association. The significance level (P value-based) for each association is color-coded. TI = tumor immunity. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[60, 180, 940, 625]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[60, 633, 936, 741]]<|/det|>
+<center>Fig. 2. Unsupervised hierarchical clustering associates circulating immune-oncological proteome profiles with population groups - Ghanaian (Af), AA, and EA men. Heatmap showing protein profiles for men without prostate cancer. Each row represents a protein (n=82), and each column corresponds to an individual [n=1482 (654 Af, 374 AA, and 454 EA)]. Each individual is color-coded as Af, AA, or EA in the annotation bar on top of the heatmap. Normalized z-score of proteins abundance are depicted on a low-to-high scale (blue-white-red). </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[66, 50, 833, 789]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[66, 789, 912, 952]]<|/det|>
+<center>Fig. 3. Immune-oncological proteins and their relationship with West-African ancestry. (A) Variance analysis for the levels of each of the 82 immune-oncological cytokines assessed as a function of genetic estimation of West African admixture among men without prostate cancer within the NCI-Maryland study. The blue plot represents the proportion of variance that can be explained by the degree of West-African admixture while the grey plot represents the residual variance that remains to be explained by other factors other than West-African ancestry. (B) The median levels of the top six West-African ancestry correlated immune-oncological proteins were compared between Af, AA, and EA. Error bars represent inter quartile range (IQR). Linearized protein abundances ( \(2^{Npx}\) ) were used to determine median and IQR for each of the proteins. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[66, 37, 760, 856]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[62, 858, 945, 978]]<|/det|>
+<center>Fig. 4. Population differences in the abundance of proteins driving (A) apoptosis, (B) autophagy, (C) chemotaxis, (D) promotion of tumor immunity, (E) suppression of tumor immunity, and (F) vasculature. Heatmaps showing levels of process/pathway-associated proteins in relationship to population group [Ghanaian (Af), AA, EA]. Shown to the right are the mean score differences for these processes/pathways among the three population groups. Profiles for Ghanaian (n=654), AA (n=374), and EA (n=454) men without prostate cancer. The process/pathway scores are derived from the average Z-scores of all the associated proteins. These scores are shown as violin plots. TI = tumor immunity. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[95, 12, 722, 750]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[794, 10, 874, 35]]<|/det|>
+<center>Fig. 5 </center>  
+
+<|ref|>text<|/ref|><|det|>[[75, 747, 944, 975]]<|/det|>
+Fig. 5. Suppression of the tumor immunity pathway associates with lethal prostate cancer. We assessed the association of the six pathways defined by the 82 immune-oncology markers with all-cause mortality, prostate cancer-specific mortality or mortality due to any cancer after a prostate cancer diagnosis. The pathway scores were evaluated as continuous predictor variables. Suppression of tumor immunity pathway was distinctively associated with all-cause mortality (A), prostate cancer-specific mortality (B), or a mortality due to any cancer after a prostate cancer diagnosis (C). Multivariable cox regression analyses were used to assess if the pathways were independently associated with survival of prostate cancer patients in the NCI-Maryland study. We adjusted for the following potential confounding factors: age at study entry (years), body-mass index (BMI, kg/m2), self-reported race (AA/EA), education (high school or less, some college, college, professional school), income (less than \(\) 10k\(,\) \ \(10 - 30k\) , \(\) 30 - 60k\(,\) \ \(60 - 90k\) , greater than \(\) 90k\(), smoking history (never, former, current), diabetes (no/yes), aspirin use (no/yes), and treatment (0 = none, 1 = surgery, 2 = radiotherapy, 3 = hormone, 4 = combination). The hazard ratios (HR) indicate the change in risk of dying when the biological process z-score value increases by 1 while holding all the other biological processes' z-scores and covariates constant.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[80, 25, 800, 800]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[840, 22, 877, 45]]<|/det|>
+<center>Fig. 6 </center>  
+
+<|ref|>text<|/ref|><|det|>[[39, 796, 950, 970]]<|/det|>
+Fig. 6. A signature of two serum markers is predictive of lethal prostate cancer in AA patients. Cross- validated, regularized Cox regression models with different elastic net mixture parameters from ridge (alpha=0) to lasso (alpha=1) were implemented to identify a predictive proteomic signature. (A) Heatmaps of feature frequencies across alpha. Features were ranked by \(P\) value for alpha=1. (B) Heatmaps of feature coefficients across alpha. Features were ranked by \(P\) value for alpha=1. (C) Kaplan- Meier plot comparing prostate cancer-specific mortality of AA cases with high levels (> median) of both TNFRSF9 and PTN (pleiotrophin) vs. low levels of either or both proteins. Log rank test was used to determine if there were statistically significant survival differences. Adjusted hazard ratio (HR) compares the risk of prostate cancer mortality for those with high levels of both TNFRSF9 and PTN vs. the remaining AA cases. HR estimates were adjusted for potential confounding factors: age, BMI, education, smoking history, diabetes status, aspirin use, treatment, and income. In B & C, \(P\) value significance was coded as \(< 0.001\) (***), \(< 0.01\) (**), \(< 0.05\) (*), and \(< 0.1\) (.).
+
+<--- Page Split --->
+<|ref|>table<|/ref|><|det|>[[114, 120, 816, 277]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[120, 101, 735, 135]]<|/det|>
+Table 1. A high score for suppression of tumor immunity associates with National Comprehensive Cancer Network (NCCN) Risk Score for metastatic prostate cancer   
+
+<table><tr><td>NCCN Risk Score</td><td>Total<br>OR (95% CI)*</td><td>AA<br>OR (95% CI)*</td><td>EA<br>OR (95% CI)*</td></tr><tr><td>Low</td><td>Ref</td><td>Ref</td><td>Ref</td></tr><tr><td>Intermediate</td><td>1.04 (0.68-1.59)</td><td>0.89 (0.46-1.70)</td><td>1.18 (0.65, 2.13)</td></tr><tr><td>High/Very High</td><td>1.47 (0.87-2.48)</td><td>1.33 (0.59-2.98)</td><td>1.72 (0.83, 3.54)</td></tr><tr><td>Regional/Metastatic</td><td>3.79 (1.59-9.04)</td><td>5.90 (1.43-24.34)</td><td>3.16 (0.95, 10.50)</td></tr><tr><td>P value for trend</td><td>0.004</td><td>0.019</td><td>0.040</td></tr></table>
+
+<|ref|>table_footnote<|/ref|><|det|>[[120, 276, 796, 425]]<|/det|>
+Note: Bolded data indicate significant associations in the logistic regression analysis. \\*Logistic regression adjusted for age at study entry, BMI \((\mathrm{kg} / \mathrm{m}2)\) , diabetes (no/yes), aspirin (no/yes), education (high school or less, some college, college, professional school), family history of prostate cancer (first-degree relatives, yes/no), self-reported race (not included in the stratified analysis), income (less than \(\$ 10k\) , \(\$ 10 - 30k\) , \(\$ 30 - 60k\) , \(\$ 60 - 90k\) , greater than \(\$ 90k\) ), smoking history (never, former, current), treatment (0=none, 1=surgery, 2=radiation, 3=hormone, 4=combination) High suppression of tumor immunity is defined by the median score in the NCI-Maryland control population ( \(>\) median vs. \(\leq\) median)
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[43, 42, 311, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[43, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 541, 178]]<|/det|>
+SupplementaryMaterial.pdf CXUsersldq5835DesktopSupplementaryTable14.xlsx
+
+<--- Page Split --->

preprint/preprint__486c1efee492ebae01b70be8d249ce174803a4e1d880fe55bf307bc77f5d24e2/images_list.json

@@ -0,0 +1,47 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        77,
+        820,
+        568
+      ]
+    ],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2",
+    "footnote": [],
+    "bbox": [
+      [
+        52,
+        52,
+        949,
+        479
+      ]
+    ],
+    "page_idx": 11
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3",
+    "footnote": [],
+    "bbox": [
+      [
+        40,
+        45,
+        951,
+        456
+      ]
+    ],
+    "page_idx": 12
+  }
+]

preprint/preprint__486c1efee492ebae01b70be8d249ce174803a4e1d880fe55bf307bc77f5d24e2/preprint__486c1efee492ebae01b70be8d249ce174803a4e1d880fe55bf307bc77f5d24e2.mmd

@@ -0,0 +1,252 @@
+
+# Contributions of countries without a carbon neutrality target to limit global warming  
+
+Wei Li  
+
+wli2019@tsinghua.edu.cn  
+
+Tsinghua University https://orcid.org/0000- 0003- 2543- 2558  
+
+Jiaxin Zhou  
+
+Tsinghua University  
+
+Philippe Ciais  
+
+Laboratoire des Sciences du Climat et de l'Environnement https://orcid.org/0000- 0001- 8560- 4943  
+
+Thomas Gasser  
+
+International Institute for Applied Systems Analysis  
+
+Jingmeng Wang  
+
+Tsinghua University https://orcid.org/0000- 0001- 9191- 0744  
+
+Zhao Li  
+
+Tsinghua University https://orcid.org/0000- 0003- 1098- 6742  
+
+Lei Zhu  
+
+Tsinghua University https://orcid.org/0000- 0002- 2146- 6438  
+
+Mengjie Han  
+
+Tsinghua University  
+
+Jiaying He  
+
+Chinese Academy of Sciences  
+
+Minxuan Sun  
+
+Tsinghua University  
+
+Li Liu  
+
+Tsinghua University  
+
+Xiaomeng Huang  
+
+Tsinghua University https://orcid.org/0000- 0002- 4158- 1089  
+
+Article  
+
+Keywords:  
+
+Posted Date: January 18th, 2024
+
+<--- Page Split --->
+
+DOI: https://doi.org/10.21203/rs.3.rs- 3847798/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on January 7th, 2025. See the published version at https://doi.org/10.1038/s41467- 024- 55720- x.
+
+<--- Page Split --->
+
+## Abstract  
+
+Bioenergy with carbon capture and storage (BECCS) is a key negative emission technology in future climate mitigation. Some countries have made no commitment to carbon neutrality, but they are viewed as potential candidates for BECCS. Here we analyze the contribution of these countries with respect to BECCS and ask the question of how much would be lost for global climate change mitigation if these countries decide not to adopt BECCS. The cooling effect due to carbon- dioxide removal (CDR) through switchgrass cultivation and carbon capture in these countries is largely counterbalanced by its biophysical warming, but the net effect is still an extra cooling. These countries play a more important role in the low- warming scenario than the overshoot scenario, despite the inequality of temperature change among countries. Our study highlights the importance of efforts from all countries in global climate mitigation.  
+
+## Introduction  
+
+Bioenergy with carbon capture and storage (BECCS) has been widely used by integrated assessment models (IAMs) in future climate mitigation scenarios (Harper et al., 2018; Krause et al., 2018). It is projected to remove \(150 \sim 1200 \text{GtCO}_2\) from the atmosphere by 2100 for limiting warming to \(1.5^{\circ}\text{C}\) (Rogel et al., 2018). The net carbon- dioxide removal (CDR) capacity of BECCS is mainly determined by bioenergy crop yields (Li et al., 2020), cultivation area (Cai et al., 2011), the CCS efficiency, and land- use change (LUC) carbon emissions (Smith et al., 2013; Boysen et al., 2017; Read et al., 2008; Bui et al., 2018). In addition to the biogeochemical cooling from the reduced \(\text{CO}_2\) concentration (Wang et al., 2023), large- scale cultivation of bioenergy crops alters the land surface properties (e.g., albedo, evapotranspiration), leading to biophysical temperature changes (Wang et al., 2021). Both CDR and the biophysical effects of bioenergy cultivation show strong spatial variations (Wang et al., 2021; Wang et al., 2023). In particular, bioenergy cultivation in one region can affect the climate of others by causing changes in atmospheric circulation. However, unlike IAMs assuming a global coordinated mitigation starting this decade, currently, only 130 countries have set a target of achieving net zero or carbon neutrality (CN, hereafter, CN countries, Fig. 1), despite of varying degrees of progress (Figure S1, Methods). There are still more than 50 countries without a CN target (non- CN countries), altogether representing about \(11\%\) of global anthropogenic \(\text{CO}_2\) emissions (Friedlingstein et al. 2021). It remains unclear to what extent CDR and temperature change would be lost if non- CN countries do not implement BECCS while CN countries do.  
+
+Here, we use an Earth system model (ESM) with an explicit representation of bioenergy crops (Li et al., 2018; Wang et al., 2021) to simulate the contribution of non- CN countries to the global temperature change in future BECCS scenarios. We consider two BECCS scenarios where BECCS is the main CDR option: 1) a low- warming scenario based on Shared- Socioeconomic Pathway (SSP) 2 and Representative Concentration Pathway (RCP) 2.6 (hereafter, the low- warming scenario) and 2) an overshoot scenario based on SSP5 and RCP3.4 (hereafter, the overshoot scenario). The global cultivation maps in these two scenarios are derived from the IAM MaGPlE (Popp et al., 2014), which implements bioenergy crop
+
+<--- Page Split --->
+
+cultivation globally in both CN and non- CN countries (Fig. 1) based on cost minimization principle and suitable land use types. The cultivation area of bioenergy crops in the low- warming scenario in 2100 is about half of that in the overshoot scenario (Fig. 1), because substantial BECCS will be implemented after 2040 to offset the overshoot emissions in the latter scenario (Hurtt et al., 2020). We assumed a typical lignocellulosic bioenergy crop, switchgrass, over the BECCS regions (Fig. 1). Switchgrass is explicitly described in the land surface model with parameters calibrated from field data (Li et al., 2018). The net CDR is the sum of harvested biomass, CCS loss and LUC emissions caused by the bioenergy crop cultivation (Eq. (S1) in Methods), and it is further translated into biogeochemical air temperature change using the OSCAR ESM emulator (Methods, Gasser et al., 2017). The biophysical air temperature change (Figure S5) is simulated by the coupled ESM (Methods). The net air temperature change is thus the sum of biogeochemical and biophysical temperature change (Eq. (1) in Methods, Wang et al., 2023). We assume that the CN countries would cultivate bioenergy crops in order to realize the carbon neutrality commitment, on the top of which, non- CN countries may or may not cultivate bioenergy crops (Methods).  
+
+## Results  
+
+## Contribution of non-CN countries at the global scale  
+
+The non- CN countries account for \(14\%\) and \(20\%\) of the global total bioenergy crop cultivation area under the low- warming and overshoot scenarios (408 Mha and 803 Mha, respectively, Fig. 1 and Figure S2). Their cumulative CDR until 2100 is non- negligible, reaching 9 PgC and 20 PgC for the two scenarios. The corresponding proportions of global total CDR from BECCS in non- CN countries are \(17\%\) and \(20\%\) , higher than their proportions of cultivation area. In terms of biogeochemical temperature changes resulting from CDR, the contribution of non- CN countries is even more pronounced. The biogeochemical effects from CDR of additional cultivation in these non- CN countries will reduce global average temperature by \(0.03^{\circ}\mathrm{C}\) and \(0.05^{\circ}\mathrm{C}\) ( \(30\%\) and \(27\%\) of the total reduction) in the low- warming and overshoot scenarios (Figure S11).  
+
+Despite the biogeochemical cooling effects, the overall biophysical effect of further switchgrass cultivation in non- CN countries is warming in both scenarios. Under the low- warming scenario, the biophysical effects of cultivation in the non- CN countries contribute a temperature increase of \(0.02^{\circ}\mathrm{C}\) (from \(0.03^{\circ}\mathrm{C}\) when only cultivation in CN countries to \(0.05^{\circ}\mathrm{C}\) when cultivation in all countries). Under the overshoot scenario, by contrast, switchgrass cultivation in CN countries will cool the lands by \(0.01^{\circ}\mathrm{C}\) through biophysical feedbacks. However, the biophysical effect of cultivation in non- CN countries will increase the temperature by \(0.04^{\circ}\mathrm{C}\) , leading to an overall increase of \(0.03^{\circ}\mathrm{C}\) with cultivation in all countries.  
+
+Combining the biogeochemical effects from CDR and the biophysical effects, the net air temperature change over lands is \(- 0.03\) and \(- 0.15^{\circ}\mathrm{C}\) in the low- warming and overshoot scenarios with switchgrass cultivation only implemented in the CN countries. Cultivation in the non- CN countries will further
+
+<--- Page Split --->
+
+contribute a cooling effect of \(0.01^{\circ}C\) and \(0.02^{\circ}C\) in these two scenarios, because its biogeochemical cooling effect (- 0.03 and - \(0.05^{\circ}C\) ) is partly counterbalanced by biophysical warming effect (0.02 and \(0.04^{\circ}C\) ). The overall contribution of non- CN countries to the net temperature reduction is \(25\%\) and \(12\%\) under the low- warming and overshoot scenarios, implying that the non- CN countries play a more important role in mitigating climate in the low- warming scenario than the overshoot scenario.  
+
+## Contribution of non-CN countries in each region  
+
+At the regional scale, the air temperature changes show strong variations (Fig. 2a). Assuming that switchgrass is cultivated in the CN countries, further cultivation in the non- CN countries leads to an extra cooling (or nearly zero) effect in most regions under the two scenarios. However, it causes extra warming in western Europe and Eurasia in both scenarios, and in Eastern Asia and South Asia only in low- warming scenario, implying more challenges in controlling temperature increase in these regions. We also find that the extra air temperature change and the additional cultivation area in the non- CN countries are decoupled geographically. For instance, there is no additional cultivation area in Pacific developed region (Fig. 2b and c), but the temperature of this region would be reduced substantially if cultivation occurs in remote non- CN countries (Fig. 2a).  
+
+In the low- warming scenario, additional cultivation area in the non- CN countries is primarily located in Africa, South and central America, Western Europe, and Eurasia (Fig. 2b). However, further cultivation of switchgrass in the non- CN countries leads to significant warming effects in Western Europe and Eurasia, primarily contributed by the biophysical warming effect (Figure S12). Additionally, in the low- warming scenario, although the cultivation area in the non- CN countries in North America is marginal, it exhibits noticeable reduction in net air temperature after further cultivation in the non- CN countries, primarily attributed to the biophysical cooling effect (Figure S12).  
+
+In the overshoot scenario, the cultivation area of non- CN countries is lower in Eurasia, South Asia, and Western Europe but higher in Africa (Fig. 2c). However, after additional switchgrass cultivation in the non- CN countries, the net air temperature change in Africa remains relatively small (Fig. 2a), despite the higher CDR contributed by the non- CN countries (Figure S10). Further cultivation in the global non- CN countries induces a strong biophysical warming effect in Western Europe (Figure S10), leading to a net air temperature increase (Fig. 2a).  
+
+## Temperature changes in countries  
+
+We further analyze the net air temperature change in the non- CN countries with the largest cultivation area (e.g., Democratic Republic of the Congo, Mexico, and Paraguay in the low- warming scenario; Iran, Republic of Côte d'Ivoire, and Cameroon in the overshoot scenario, Fig. 3a, b), and the temperature changes in the CN countries (e.g., Afghanistan, Nepal, and Ukraine in the low- warming scenario; Bhutan, Bulgaria, and Hungary in the overshoot scenario) that are most affected (i.e., largest absolute temperature change) by the additional cultivation in non- CN countries (Fig. 3c, d). In the low- warming
+
+<--- Page Split --->
+
+scenario, 7 out of the top 10 non- CN countries experience an extra warming with switchgrass cultivation in the non- CN countries, and the warming magnitude in these 7 countries (e.g., Belarus) is much larger than the cooling magnitude in the remaining 3 countries with an extra cooling (orange arrows in Fig. 3a). By contrast, 7 out of the top 10 CN countries show a temperature reduction with additional cultivation in the non- CN countries (Fig. 3c), indicating further benefits of cooling in these CN countries. In the overshoot scenario, 4 and 3 out of the top 10 non- CN countries show an extra moderate cooling and warming after additional cultivation in all non- CN countries, and the temperature change in other countries is minor (Fig. 3b). However, the impacts of further cultivation in the non- CN countries on the top 10 most affected CN countries are very strong in the overshoot scenario, ranging from 0.58 to 1.13 °C (except Bhutan) driven by the biophysical effects via atmospheric teleconnection (Fig. 3b, d, Fig. 2c). It should be noted that some non- CN countries (e.g., Iran and Cameroon) and CN countries (e.g., Russia) have large cultivation area, but the CDR is low due to lower biomass yields in regions with unfavorable climate conditions (Figure S7, Figure S8b, c).  
+
+## Discussion  
+
+Our results are based on simulations from the ESM with explicit processes for bioenergy crops (Li et al., 2018; Wang et al., 2021). However, there are some uncertainties due to the simulation set- up and missing processes in the model (Text S5). For example, the amounts of CDR in different bioenergy crop cultivation scenarios were calculated using the response curves of various carbon pools derived from the offline simulations. It ignores the impact of future climate change on the bioenergy crop biomass production (Text S5.1). The CCS efficiency may also vary spatially, and thus we added a sensitivity test using different levels of CCS efficiency (Text S5.1). As expected, the CDR will increase if the CCS efficiency becomes higher (Figure S13). In addition, BECCS has other costs such as post- harvest processing such as baling and pelleting (Negri et al., 2021), transportation from the cultivation area to processing plants, pyrolysis plants and power plants (Fajardy et al., 2020; Negri et al., 2021; Sultana et al., 2011), and its conversion to available energy (Negri et al., 2021). All these additional economic constraints are not explicitly considered in our study.  
+
+Despite uncertainties in our CDR estimates arising from the idealized assumptions, our results show that additional cultivation of switchgrass in non- CN countries would induce an overall significant biogeochemical cooling effect. Although this cooling effect will be partly offset by its biophysical warming effect, the net effect is cooling at the global scale (Fig. 1a). Therefore, taking the biophysical effects into account, the contribution of additional cultivation in non- CN countries to global air temperature reduction will be weakened but still a net cooling effect, implying the non- negligible role of these countries in mitigating climate change. At the regional scale, some non- CN countries (mostly developing countries such as Mexico, Poland and Paraguay) suffer an extra warming while some CN countries gain extra cooling from cultivation in the non- CN countries, which may aggravate the inequality between the CN and non- CN countries. In addition, the relative contribution of non- CN countries to the global and regional temperature reduction is greater in the low- warming scenario than that in the overshoot scenario. Therefore, avoiding the overshooting of temperature will not only reduce cost for
+
+<--- Page Split --->
+
+climate change mitigation but also strength the effectiveness of implementing BECCS in the non- CN countries. The implementation of bioenergy crop cultivation is not likely synchronized across countries, and a delayed implementation may lead to a decrease in CDR and ultimately reduce the effectiveness of BECCS as a climate mitigation strategy (Text S5; Xu et al., 2022). Our study provides a framework for assessing the roles of non- CN countries in land- based climate mitigation options such as afforestation, and using bioenergy crop cultivation as an example, demonstrates the importance of their efforts in global climate mitigation.  
+
+## Methods  
+
+## Simulation scenario design  
+
+The status of carbon neutrality target for each country is downloaded from https://zerotracker.net/, and there were 136 countries with a carbon neutrality target but at different degrees of progress by the end of November 2021 (achieved, in law, in policy document, declaration / pledge, proposed / in discussion, Figure S1). Switchgrass is assumed to be cultivated synchronously in all CN countries or in both CN and non- CN countries. In order to separate the contribution of non- CN countries to the biophysical temperature change, we ran two sets of simulations: bioenergy crop is cultivated 1) in the CN countries only and 2) in both CN and non- CN countries. Their difference is thus the contribution of non- CN countries.  
+
+We designed four bioenergy crop cultivation scenarios based on two bioenergy crop cultivation maps and either cultivating in the CN countries only or in both CN and non- CN countries and a reference scenario without bioenergy crop cultivation (Table S1, Text S3). The contribution of non- CN countries is calculated as the difference between the scenario with switchgrass cultivated in both CN and non- CN countries and the scenario with switchgrass only cultivated in the CN countries. The cultivation maps (Figure S2) were the BECCS scenarios from the integrated assessment model of MAGPIE (Hurt et al., 2020), in which BECCS serves as the main negative emission technology to limit global warming (Text S1).  
+
+## Estimation of CDR  
+
+Following Wang et al. (2023), the offline simulations for the carbon dynamics were performed using ORCHIDEE- MICT- BIOENERGY, a dynamic vegetation model with an explicit representation of bioenergy crops (Li et al., 2018) (Text S2). In the offline simulations, ORCHIDEE- MICT- BIOENERGY simulated the changes in biomass and soil carbon pools resulting from the conversion of different vegetation types to bioenergy crops. Response curves for LUC types (from forest, grass, pasture, and cropland to switchgrass) were derived from these offline simulations, used for calculating CDR (including harvested biomass, LUC carbon emissions and CCS loss, Text S2.1) under bioenergy crop cultivation scenarios.
+
+<--- Page Split --->
+
+Besides, the CDR from bioenergy crops relies on regular harvests, impacting soil fertility (Li et al., 2021). We replenished nitrogen loss through fertilizer application, considering GHG emissions from fertilizer production and \(\mathrm{N}_2\mathrm{O}\) emissions. The study accounts for \(\mathrm{CO}_2\) reduction, fertilizer- related emissions, and \(\mathrm{N}_2\mathrm{O}\) emissions, estimating soil nitrogen loss and applied fertilizer amounts in different scenarios (Text S2.2).  
+
+## Estimation of the temperature change  
+
+The CDR were further translated into biogeochemical temperature changes using the compact ESM (OSCAR, Gasser et al., 2017; Text S2.3). OSCAR simulated temperature changes related to CDR processes and GHG emissions from fertilization, considering modeling uncertainties with a sample size of 2000. Global biogeochemical cooling effects were calculated by aggregating regional outputs.  
+
+The biophysical temperature changes were simulated by the coupled land- atmosphere model IPSL- CM (Boucher et al., 2020), in which ORCHIDEE- MICT- BIOENERGY serves as the land surface model (Wang et al., 2021), LMDz (v6) served as the atmosphere model (Hourdin et al., 2006; Contoux et al., 2012; Text S3). Ocean and sea- ice models were not activated. The simulations, spanning 50 years with 2014 atmospheric \(\mathrm{CO}_2\) levels (Sitch et al., 2015; Peng et al., 2015), employed a spatial resolution of \(1.26^{\circ}\times\) \(2.5^{\circ}\) . The study conducted five coupled simulations, including switchgrass cultivation scenarios in the CN and non- CN countries under the low- warming and overshoot scenarios, and a reference simulation without bioenergy crops (Table S1). The simulations reached a steady state between the fifth and tenth years for switchgrass, and results from the last decade (41st to 50th years) were analyzed for biophysical effects. The cultivation map in 2100 was used for the simulations of biophysical effects.  
+
+The net air temperature change \((\Delta \mathrm{T}_{\mathrm{net}})\) induced by switchgrass cultivation in this study includes 1) biogeochemical effects from the CDR of BECCS and the fertilization related greenhouse gas emissions (Text S2) and 2) biophysical effects from the changed local energy budget and the altered atmosphere circulation (Text S3):  
+
+\[\Delta \mathrm{T}_{\mathrm{net}} = \Delta \mathrm{T}_{\mathrm{bgc}} + \Delta \mathrm{T}_{\mathrm{bph}}\]  
+
+1  
+
+The subscript represents the air temperature contributed by the biogeochemical effects ("bgc") or the biophysical effects ("bph").  
+
+## Declarations  
+
+## Acknowledgments  
+
+This study was funded by the National Natural Science Foundation of China (grant number: 42175169, 72348001, to W.L.), the National Key R&D Program of China (grant number: 2019YFA0606604, to W.L.),
+
+<--- Page Split --->
+
+the Tsinghua University Initiative Scientific Research Program (grant number: 202230041, to W.L.).  
+
+## Data availability  
+
+All data are available in the main text or the supplementary information.  
+
+## Author contributions  
+
+W.L., J.W. and P.C. designed the study, J.Z., W.L. and J.W. carried out the modeling and analysis. J.Z., W.L. and J.W. wrote the first draft. P.C., T.G., Z.L., L.Z., M.H., J.H., M.S., L.L., X.H. contributed to the interpretation of the results, the draft revision and the computational tools.  
+
+## Competing interests  
+
+The authors declare no competing interests.  
+
+## References  
+
+1. Harper AB, Powell T, Cox PM et al (2018) Land-use emissions play a critical role in land-based mitigation for Paris climate targets. Nat Commun 9:2938  
+2. Krause A, Pugh TAM, Bayer AD et al (2018) Large uncertainty in carbon uptake potential of land-based climate-change mitigation efforts. Glob Change Biol 24:3025–3038  
+3. Rogelj J, Popp A, Calvin KV et al (2018) Scenarios towards limiting global mean temperature increase below 1.5°C. Nat Clim Change 8:325–332  
+4. Li W, Ciais P, Stehfest E, van Vuuren et al (2020) Mapping the yields of lignocellulosic bioenergy crops from observations at the global scale. Earth Syst Sci Data 12:789–804  
+5. Cai X et al (2011) Land availability for biofuel production. Environ Sci Technol vol 45(1):334–339  
+6. Smith LJ, Torn MS (2013) Ecological limits to terrestrial biological carbon dioxide removal. Clim Change 118:89–103  
+7. Boysen LR et al (2017) The limits to global-warming mitigation by terrestrial carbon removal. Earths Future 5(5):463–474  
+8. Read P (2008) Biosphere carbon stock management: addressing the threat of abrupt climate change in the next few decades: an editorial essay. Clim Change 87:305–320  
+9. Bui M et al (2018) Carbon capture and storage (CCS): the way forward. Energy Environ Sci 11(5):1062–1176  
+10. Wang J et al (2023) Temperature Changes Induced by Biogeochemical and Biophysical Effects of Bioenergy Crop Cultivation. Environ Sci Technol 57:2474–2483  
+11. Wang J, Li W, Ciais P et al (2021) Global cooling induced by biophysical effects of bioenergy crop cultivation. Nat Commun 12:7255  
+12. Friedlingstein P et al (2022) Global Carbon Budget 2021. Earth Syst Sci Data 14:1917–2005
+
+<--- Page Split --->
+
+13. Li W, Yue C, Ciais P et al (2018) ORCHIDEE-MICT-BIOENERGY: an attempt to represent the production of lignocellulosic crops for bioenergy in a global vegetation model. Geosci Model Dev 11:2249-2272  
+
+14. Popp A, Rose SK, Calvin K et al (2014) Land-use transition for bioenergy and climate stabilization: model comparison of drivers, impacts and interactions with other land use based mitigation options. Clim Change 123:495-509  
+
+15. Hurtt GC, Chini L, Sahajpal R et al (2020) Harmonization of global land use change and management for the period 850-2100 (LUH2) for CMIP6, Geosci. Model Dev 13:5425-5464  
+
+16. Gasser T, Ciais P, Boucher O et al (2017) The compact Earth system model OSCAR v2.2: description and first results. Geosci Model Dev 10:271-319  
+
+17. Boucher O, Servonnat J, Albright AL, Aumont O, Balkanski Y, Bastrikov V et al (2020) Presentation and evaluation of the IPSL-CM6A-LR climate model. J Adv Model Earth Syst, 12, e2019MS002010  
+
+18. Li W, Ciais P, Han M et al Bioenergy crops for low warming targets require half of the present agricultural fertilizer use. Environ Sci Technol, 55(15), 10654-1066(2021).  
+
+19. Negri V et al (2021) Life cycle optimization of BECCS supply chains in the European Union. Appl Energy 298:117252  
+
+20. Fajardy M et al (2021) The economics of bioenergy with carbon capture and storage (BECCS) deployment in a 1.5°C or 2°C world. Glob Environ Change 68:102262  
+
+21. Sultana A, Kumar A (2011) Optimal configuration and combination of multiple lignocellulosic biomass feedstocks delivery to a biorefinery. Bioresour Technol 102:9947-9956  
+
+22. Xu S, Wang R, Gasser T et al (2022) Delayed use of bioenergy crops might threaten climate and food security. Nature 609(7926):299-306  
+
+23. Hourdin F, Musat I, Bony S et al (2006) The LMDZ4 general circulation model: climate performance and sensitivity to parametrized physics with emphasis on tropical convection. Clim Dyn 27:787-813  
+
+24. Contoux C, Ramstein G, Jost A (2012) Modelling the mid-Pliocene Warm Period climate with the IPSL coupled model and its atmospheric component LMDZ5A, Geosci. Model Dev 5:903-917  
+
+25. Sitch S, Friedlingstein P, Gruber N et al (2015) Recent trends and drivers of regional sources and sinks of carbon dioxide. Biogeosciences 12, 653  
+
+26. Peng S, Ciais P, Maignan F et al (2015) Sensitivity of land use change emission estimates to historical land use and land cover mapping. Global Biogeochem Cycles 31:626-643  
+
+## Figures
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Figure 1 </center>  
+
+Bioenergy crop cultivation maps under the low- warming (a) and overshoot (b) scenarios and contributions of the CN and non- CN countries to the global total bioenergy crop cultivation area, net carbon- dioxide removal (CDR), biophysical air temperature change \((\Delta T_{\mathrm{bph}})\) and net air temperature change \((\Delta T_{\mathrm{net}})\) . Blue bars represent changes when cultivating switchgrass in the CN countries only, and yellow bars represent further changes when cultivating switchgrass in both CN and non- CN countries. Arrows represent the directions of changes. The non- CN countries with a cultivation area \(>1\) ha are marked in red. Note the scales in the bar plot are different between (a) and (b).
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Figure 2 </center>  
+
+Contributions of the CN and non- CN countries to net air temperature change at the regional scale (a), and the cultivation area in the CN and non- CN countries in each region under the low- warming (b) and overshoot (c) scenarios. In (a), blue bars represent the net air temperature change when cultivating switchgrass in the CN countries only, and orange bars represent the temperature changes after further cultivation in the non- CN countries. In (b) and (c), blue bars indicate the cultivation area in the CN countries within each region, and orange bars indicate the further cultivation area in the non- CN countries.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3 </center>  
+
+The cultivation area and net air temperature change in the top ten non- CN countries with the largest cultivation area and the top ten CN countries with the maximum net air temperature change under the low- warming (aand c) and overshoot (b and d) scenarios. Blue arrows refer to the net air temperature change when cultivating switchgrass in the CN countries only, and orange arrows indicate the temperature changes after further cultivation in the non- CN countries. The directions of arrows represent increase and decrease in air temperatures. Black dots in a- d indicate the cultivation area in each country. DRC and CIV are the Democratic Republic of the Congo and the Republic of Côte d'Ivoire.  
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- NCSIregionalBECCS240109.docx
+
+<--- Page Split --->

preprint/preprint__486c1efee492ebae01b70be8d249ce174803a4e1d880fe55bf307bc77f5d24e2/preprint__486c1efee492ebae01b70be8d249ce174803a4e1d880fe55bf307bc77f5d24e2_det.mmd

@@ -0,0 +1,345 @@
+<|ref|>title<|/ref|><|det|>[[44, 108, 802, 177]]<|/det|>
+# Contributions of countries without a carbon neutrality target to limit global warming  
+
+<|ref|>text<|/ref|><|det|>[[44, 196, 99, 213]]<|/det|>
+Wei Li  
+
+<|ref|>text<|/ref|><|det|>[[53, 222, 312, 240]]<|/det|>
+wli2019@tsinghua.edu.cn  
+
+<|ref|>text<|/ref|><|det|>[[50, 269, 585, 289]]<|/det|>
+Tsinghua University https://orcid.org/0000- 0003- 2543- 2558  
+
+<|ref|>text<|/ref|><|det|>[[44, 294, 148, 311]]<|/det|>
+Jiaxin Zhou  
+
+<|ref|>text<|/ref|><|det|>[[53, 316, 228, 334]]<|/det|>
+Tsinghua University  
+
+<|ref|>text<|/ref|><|det|>[[44, 340, 164, 358]]<|/det|>
+Philippe Ciais  
+
+<|ref|>text<|/ref|><|det|>[[50, 362, 916, 381]]<|/det|>
+Laboratoire des Sciences du Climat et de l'Environnement https://orcid.org/0000- 0001- 8560- 4943  
+
+<|ref|>text<|/ref|><|det|>[[44, 386, 184, 404]]<|/det|>
+Thomas Gasser  
+
+<|ref|>text<|/ref|><|det|>[[53, 408, 502, 427]]<|/det|>
+International Institute for Applied Systems Analysis  
+
+<|ref|>text<|/ref|><|det|>[[44, 432, 188, 451]]<|/det|>
+Jingmeng Wang  
+
+<|ref|>text<|/ref|><|det|>[[53, 455, 586, 473]]<|/det|>
+Tsinghua University https://orcid.org/0000- 0001- 9191- 0744  
+
+<|ref|>text<|/ref|><|det|>[[44, 479, 111, 496]]<|/det|>
+Zhao Li  
+
+<|ref|>text<|/ref|><|det|>[[53, 500, 586, 519]]<|/det|>
+Tsinghua University https://orcid.org/0000- 0003- 1098- 6742  
+
+<|ref|>text<|/ref|><|det|>[[44, 525, 111, 542]]<|/det|>
+Lei Zhu  
+
+<|ref|>text<|/ref|><|det|>[[53, 547, 586, 565]]<|/det|>
+Tsinghua University https://orcid.org/0000- 0002- 2146- 6438  
+
+<|ref|>text<|/ref|><|det|>[[44, 571, 164, 589]]<|/det|>
+Mengjie Han  
+
+<|ref|>text<|/ref|><|det|>[[53, 593, 228, 611]]<|/det|>
+Tsinghua University  
+
+<|ref|>text<|/ref|><|det|>[[44, 617, 138, 635]]<|/det|>
+Jiaying He  
+
+<|ref|>text<|/ref|><|det|>[[53, 640, 321, 658]]<|/det|>
+Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 664, 160, 681]]<|/det|>
+Minxuan Sun  
+
+<|ref|>text<|/ref|><|det|>[[53, 686, 228, 704]]<|/det|>
+Tsinghua University  
+
+<|ref|>text<|/ref|><|det|>[[44, 710, 110, 727]]<|/det|>
+Li Liu  
+
+<|ref|>text<|/ref|><|det|>[[53, 732, 228, 750]]<|/det|>
+Tsinghua University  
+
+<|ref|>text<|/ref|><|det|>[[44, 756, 196, 774]]<|/det|>
+Xiaomeng Huang  
+
+<|ref|>text<|/ref|><|det|>[[53, 778, 586, 797]]<|/det|>
+Tsinghua University https://orcid.org/0000- 0002- 4158- 1089  
+
+<|ref|>text<|/ref|><|det|>[[44, 840, 103, 857]]<|/det|>
+Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 877, 136, 895]]<|/det|>
+Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 914, 330, 933]]<|/det|>
+Posted Date: January 18th, 2024
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 45, 475, 64]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 3847798/v1  
+
+<|ref|>text<|/ref|><|det|>[[42, 82, 914, 125]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 143, 534, 163]]<|/det|>
+Additional Declarations: There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[42, 199, 933, 242]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on January 7th, 2025. See the published version at https://doi.org/10.1038/s41467- 024- 55720- x.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 158, 68]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[41, 82, 940, 310]]<|/det|>
+Bioenergy with carbon capture and storage (BECCS) is a key negative emission technology in future climate mitigation. Some countries have made no commitment to carbon neutrality, but they are viewed as potential candidates for BECCS. Here we analyze the contribution of these countries with respect to BECCS and ask the question of how much would be lost for global climate change mitigation if these countries decide not to adopt BECCS. The cooling effect due to carbon- dioxide removal (CDR) through switchgrass cultivation and carbon capture in these countries is largely counterbalanced by its biophysical warming, but the net effect is still an extra cooling. These countries play a more important role in the low- warming scenario than the overshoot scenario, despite the inequality of temperature change among countries. Our study highlights the importance of efforts from all countries in global climate mitigation.  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 333, 204, 358]]<|/det|>
+## Introduction  
+
+<|ref|>text<|/ref|><|det|>[[40, 370, 955, 786]]<|/det|>
+Bioenergy with carbon capture and storage (BECCS) has been widely used by integrated assessment models (IAMs) in future climate mitigation scenarios (Harper et al., 2018; Krause et al., 2018). It is projected to remove \(150 \sim 1200 \text{GtCO}_2\) from the atmosphere by 2100 for limiting warming to \(1.5^{\circ}\text{C}\) (Rogel et al., 2018). The net carbon- dioxide removal (CDR) capacity of BECCS is mainly determined by bioenergy crop yields (Li et al., 2020), cultivation area (Cai et al., 2011), the CCS efficiency, and land- use change (LUC) carbon emissions (Smith et al., 2013; Boysen et al., 2017; Read et al., 2008; Bui et al., 2018). In addition to the biogeochemical cooling from the reduced \(\text{CO}_2\) concentration (Wang et al., 2023), large- scale cultivation of bioenergy crops alters the land surface properties (e.g., albedo, evapotranspiration), leading to biophysical temperature changes (Wang et al., 2021). Both CDR and the biophysical effects of bioenergy cultivation show strong spatial variations (Wang et al., 2021; Wang et al., 2023). In particular, bioenergy cultivation in one region can affect the climate of others by causing changes in atmospheric circulation. However, unlike IAMs assuming a global coordinated mitigation starting this decade, currently, only 130 countries have set a target of achieving net zero or carbon neutrality (CN, hereafter, CN countries, Fig. 1), despite of varying degrees of progress (Figure S1, Methods). There are still more than 50 countries without a CN target (non- CN countries), altogether representing about \(11\%\) of global anthropogenic \(\text{CO}_2\) emissions (Friedlingstein et al. 2021). It remains unclear to what extent CDR and temperature change would be lost if non- CN countries do not implement BECCS while CN countries do.  
+
+<|ref|>text<|/ref|><|det|>[[41, 802, 955, 959]]<|/det|>
+Here, we use an Earth system model (ESM) with an explicit representation of bioenergy crops (Li et al., 2018; Wang et al., 2021) to simulate the contribution of non- CN countries to the global temperature change in future BECCS scenarios. We consider two BECCS scenarios where BECCS is the main CDR option: 1) a low- warming scenario based on Shared- Socioeconomic Pathway (SSP) 2 and Representative Concentration Pathway (RCP) 2.6 (hereafter, the low- warming scenario) and 2) an overshoot scenario based on SSP5 and RCP3.4 (hereafter, the overshoot scenario). The global cultivation maps in these two scenarios are derived from the IAM MaGPlE (Popp et al., 2014), which implements bioenergy crop
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[40, 44, 945, 362]]<|/det|>
+cultivation globally in both CN and non- CN countries (Fig. 1) based on cost minimization principle and suitable land use types. The cultivation area of bioenergy crops in the low- warming scenario in 2100 is about half of that in the overshoot scenario (Fig. 1), because substantial BECCS will be implemented after 2040 to offset the overshoot emissions in the latter scenario (Hurtt et al., 2020). We assumed a typical lignocellulosic bioenergy crop, switchgrass, over the BECCS regions (Fig. 1). Switchgrass is explicitly described in the land surface model with parameters calibrated from field data (Li et al., 2018). The net CDR is the sum of harvested biomass, CCS loss and LUC emissions caused by the bioenergy crop cultivation (Eq. (S1) in Methods), and it is further translated into biogeochemical air temperature change using the OSCAR ESM emulator (Methods, Gasser et al., 2017). The biophysical air temperature change (Figure S5) is simulated by the coupled ESM (Methods). The net air temperature change is thus the sum of biogeochemical and biophysical temperature change (Eq. (1) in Methods, Wang et al., 2023). We assume that the CN countries would cultivate bioenergy crops in order to realize the carbon neutrality commitment, on the top of which, non- CN countries may or may not cultivate bioenergy crops (Methods).  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 383, 144, 408]]<|/det|>
+## Results  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 421, 835, 453]]<|/det|>
+## Contribution of non-CN countries at the global scale  
+
+<|ref|>text<|/ref|><|det|>[[41, 468, 950, 671]]<|/det|>
+The non- CN countries account for \(14\%\) and \(20\%\) of the global total bioenergy crop cultivation area under the low- warming and overshoot scenarios (408 Mha and 803 Mha, respectively, Fig. 1 and Figure S2). Their cumulative CDR until 2100 is non- negligible, reaching 9 PgC and 20 PgC for the two scenarios. The corresponding proportions of global total CDR from BECCS in non- CN countries are \(17\%\) and \(20\%\) , higher than their proportions of cultivation area. In terms of biogeochemical temperature changes resulting from CDR, the contribution of non- CN countries is even more pronounced. The biogeochemical effects from CDR of additional cultivation in these non- CN countries will reduce global average temperature by \(0.03^{\circ}\mathrm{C}\) and \(0.05^{\circ}\mathrm{C}\) ( \(30\%\) and \(27\%\) of the total reduction) in the low- warming and overshoot scenarios (Figure S11).  
+
+<|ref|>text<|/ref|><|det|>[[41, 686, 950, 867]]<|/det|>
+Despite the biogeochemical cooling effects, the overall biophysical effect of further switchgrass cultivation in non- CN countries is warming in both scenarios. Under the low- warming scenario, the biophysical effects of cultivation in the non- CN countries contribute a temperature increase of \(0.02^{\circ}\mathrm{C}\) (from \(0.03^{\circ}\mathrm{C}\) when only cultivation in CN countries to \(0.05^{\circ}\mathrm{C}\) when cultivation in all countries). Under the overshoot scenario, by contrast, switchgrass cultivation in CN countries will cool the lands by \(0.01^{\circ}\mathrm{C}\) through biophysical feedbacks. However, the biophysical effect of cultivation in non- CN countries will increase the temperature by \(0.04^{\circ}\mathrm{C}\) , leading to an overall increase of \(0.03^{\circ}\mathrm{C}\) with cultivation in all countries.  
+
+<|ref|>text<|/ref|><|det|>[[42, 884, 940, 951]]<|/det|>
+Combining the biogeochemical effects from CDR and the biophysical effects, the net air temperature change over lands is \(- 0.03\) and \(- 0.15^{\circ}\mathrm{C}\) in the low- warming and overshoot scenarios with switchgrass cultivation only implemented in the CN countries. Cultivation in the non- CN countries will further
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[41, 44, 939, 157]]<|/det|>
+contribute a cooling effect of \(0.01^{\circ}C\) and \(0.02^{\circ}C\) in these two scenarios, because its biogeochemical cooling effect (- 0.03 and - \(0.05^{\circ}C\) ) is partly counterbalanced by biophysical warming effect (0.02 and \(0.04^{\circ}C\) ). The overall contribution of non- CN countries to the net temperature reduction is \(25\%\) and \(12\%\) under the low- warming and overshoot scenarios, implying that the non- CN countries play a more important role in mitigating climate in the low- warming scenario than the overshoot scenario.  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 186, 666, 214]]<|/det|>
+## Contribution of non-CN countries in each region  
+
+<|ref|>text<|/ref|><|det|>[[41, 229, 950, 432]]<|/det|>
+At the regional scale, the air temperature changes show strong variations (Fig. 2a). Assuming that switchgrass is cultivated in the CN countries, further cultivation in the non- CN countries leads to an extra cooling (or nearly zero) effect in most regions under the two scenarios. However, it causes extra warming in western Europe and Eurasia in both scenarios, and in Eastern Asia and South Asia only in low- warming scenario, implying more challenges in controlling temperature increase in these regions. We also find that the extra air temperature change and the additional cultivation area in the non- CN countries are decoupled geographically. For instance, there is no additional cultivation area in Pacific developed region (Fig. 2b and c), but the temperature of this region would be reduced substantially if cultivation occurs in remote non- CN countries (Fig. 2a).  
+
+<|ref|>text<|/ref|><|det|>[[41, 448, 949, 606]]<|/det|>
+In the low- warming scenario, additional cultivation area in the non- CN countries is primarily located in Africa, South and central America, Western Europe, and Eurasia (Fig. 2b). However, further cultivation of switchgrass in the non- CN countries leads to significant warming effects in Western Europe and Eurasia, primarily contributed by the biophysical warming effect (Figure S12). Additionally, in the low- warming scenario, although the cultivation area in the non- CN countries in North America is marginal, it exhibits noticeable reduction in net air temperature after further cultivation in the non- CN countries, primarily attributed to the biophysical cooling effect (Figure S12).  
+
+<|ref|>text<|/ref|><|det|>[[41, 622, 940, 758]]<|/det|>
+In the overshoot scenario, the cultivation area of non- CN countries is lower in Eurasia, South Asia, and Western Europe but higher in Africa (Fig. 2c). However, after additional switchgrass cultivation in the non- CN countries, the net air temperature change in Africa remains relatively small (Fig. 2a), despite the higher CDR contributed by the non- CN countries (Figure S10). Further cultivation in the global non- CN countries induces a strong biophysical warming effect in Western Europe (Figure S10), leading to a net air temperature increase (Fig. 2a).  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 773, 564, 805]]<|/det|>
+## Temperature changes in countries  
+
+<|ref|>text<|/ref|><|det|>[[41, 819, 944, 955]]<|/det|>
+We further analyze the net air temperature change in the non- CN countries with the largest cultivation area (e.g., Democratic Republic of the Congo, Mexico, and Paraguay in the low- warming scenario; Iran, Republic of Côte d'Ivoire, and Cameroon in the overshoot scenario, Fig. 3a, b), and the temperature changes in the CN countries (e.g., Afghanistan, Nepal, and Ukraine in the low- warming scenario; Bhutan, Bulgaria, and Hungary in the overshoot scenario) that are most affected (i.e., largest absolute temperature change) by the additional cultivation in non- CN countries (Fig. 3c, d). In the low- warming
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[39, 44, 950, 338]]<|/det|>
+scenario, 7 out of the top 10 non- CN countries experience an extra warming with switchgrass cultivation in the non- CN countries, and the warming magnitude in these 7 countries (e.g., Belarus) is much larger than the cooling magnitude in the remaining 3 countries with an extra cooling (orange arrows in Fig. 3a). By contrast, 7 out of the top 10 CN countries show a temperature reduction with additional cultivation in the non- CN countries (Fig. 3c), indicating further benefits of cooling in these CN countries. In the overshoot scenario, 4 and 3 out of the top 10 non- CN countries show an extra moderate cooling and warming after additional cultivation in all non- CN countries, and the temperature change in other countries is minor (Fig. 3b). However, the impacts of further cultivation in the non- CN countries on the top 10 most affected CN countries are very strong in the overshoot scenario, ranging from 0.58 to 1.13 °C (except Bhutan) driven by the biophysical effects via atmospheric teleconnection (Fig. 3b, d, Fig. 2c). It should be noted that some non- CN countries (e.g., Iran and Cameroon) and CN countries (e.g., Russia) have large cultivation area, but the CDR is low due to lower biomass yields in regions with unfavorable climate conditions (Figure S7, Figure S8b, c).  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 360, 190, 386]]<|/det|>
+## Discussion  
+
+<|ref|>text<|/ref|><|det|>[[41, 400, 944, 672]]<|/det|>
+Our results are based on simulations from the ESM with explicit processes for bioenergy crops (Li et al., 2018; Wang et al., 2021). However, there are some uncertainties due to the simulation set- up and missing processes in the model (Text S5). For example, the amounts of CDR in different bioenergy crop cultivation scenarios were calculated using the response curves of various carbon pools derived from the offline simulations. It ignores the impact of future climate change on the bioenergy crop biomass production (Text S5.1). The CCS efficiency may also vary spatially, and thus we added a sensitivity test using different levels of CCS efficiency (Text S5.1). As expected, the CDR will increase if the CCS efficiency becomes higher (Figure S13). In addition, BECCS has other costs such as post- harvest processing such as baling and pelleting (Negri et al., 2021), transportation from the cultivation area to processing plants, pyrolysis plants and power plants (Fajardy et al., 2020; Negri et al., 2021; Sultana et al., 2011), and its conversion to available energy (Negri et al., 2021). All these additional economic constraints are not explicitly considered in our study.  
+
+<|ref|>text<|/ref|><|det|>[[41, 688, 953, 960]]<|/det|>
+Despite uncertainties in our CDR estimates arising from the idealized assumptions, our results show that additional cultivation of switchgrass in non- CN countries would induce an overall significant biogeochemical cooling effect. Although this cooling effect will be partly offset by its biophysical warming effect, the net effect is cooling at the global scale (Fig. 1a). Therefore, taking the biophysical effects into account, the contribution of additional cultivation in non- CN countries to global air temperature reduction will be weakened but still a net cooling effect, implying the non- negligible role of these countries in mitigating climate change. At the regional scale, some non- CN countries (mostly developing countries such as Mexico, Poland and Paraguay) suffer an extra warming while some CN countries gain extra cooling from cultivation in the non- CN countries, which may aggravate the inequality between the CN and non- CN countries. In addition, the relative contribution of non- CN countries to the global and regional temperature reduction is greater in the low- warming scenario than that in the overshoot scenario. Therefore, avoiding the overshooting of temperature will not only reduce cost for
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[41, 44, 945, 201]]<|/det|>
+climate change mitigation but also strength the effectiveness of implementing BECCS in the non- CN countries. The implementation of bioenergy crop cultivation is not likely synchronized across countries, and a delayed implementation may lead to a decrease in CDR and ultimately reduce the effectiveness of BECCS as a climate mitigation strategy (Text S5; Xu et al., 2022). Our study provides a framework for assessing the roles of non- CN countries in land- based climate mitigation options such as afforestation, and using bioenergy crop cultivation as an example, demonstrates the importance of their efforts in global climate mitigation.  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 223, 161, 249]]<|/det|>
+## Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 263, 459, 294]]<|/det|>
+## Simulation scenario design  
+
+<|ref|>text<|/ref|><|det|>[[41, 309, 945, 489]]<|/det|>
+The status of carbon neutrality target for each country is downloaded from https://zerotracker.net/, and there were 136 countries with a carbon neutrality target but at different degrees of progress by the end of November 2021 (achieved, in law, in policy document, declaration / pledge, proposed / in discussion, Figure S1). Switchgrass is assumed to be cultivated synchronously in all CN countries or in both CN and non- CN countries. In order to separate the contribution of non- CN countries to the biophysical temperature change, we ran two sets of simulations: bioenergy crop is cultivated 1) in the CN countries only and 2) in both CN and non- CN countries. Their difference is thus the contribution of non- CN countries.  
+
+<|ref|>text<|/ref|><|det|>[[41, 506, 937, 686]]<|/det|>
+We designed four bioenergy crop cultivation scenarios based on two bioenergy crop cultivation maps and either cultivating in the CN countries only or in both CN and non- CN countries and a reference scenario without bioenergy crop cultivation (Table S1, Text S3). The contribution of non- CN countries is calculated as the difference between the scenario with switchgrass cultivated in both CN and non- CN countries and the scenario with switchgrass only cultivated in the CN countries. The cultivation maps (Figure S2) were the BECCS scenarios from the integrated assessment model of MAGPIE (Hurt et al., 2020), in which BECCS serves as the main negative emission technology to limit global warming (Text S1).  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 716, 285, 742]]<|/det|>
+## Estimation of CDR  
+
+<|ref|>text<|/ref|><|det|>[[41, 758, 945, 915]]<|/det|>
+Following Wang et al. (2023), the offline simulations for the carbon dynamics were performed using ORCHIDEE- MICT- BIOENERGY, a dynamic vegetation model with an explicit representation of bioenergy crops (Li et al., 2018) (Text S2). In the offline simulations, ORCHIDEE- MICT- BIOENERGY simulated the changes in biomass and soil carbon pools resulting from the conversion of different vegetation types to bioenergy crops. Response curves for LUC types (from forest, grass, pasture, and cropland to switchgrass) were derived from these offline simulations, used for calculating CDR (including harvested biomass, LUC carbon emissions and CCS loss, Text S2.1) under bioenergy crop cultivation scenarios.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[41, 44, 940, 161]]<|/det|>
+Besides, the CDR from bioenergy crops relies on regular harvests, impacting soil fertility (Li et al., 2021). We replenished nitrogen loss through fertilizer application, considering GHG emissions from fertilizer production and \(\mathrm{N}_2\mathrm{O}\) emissions. The study accounts for \(\mathrm{CO}_2\) reduction, fertilizer- related emissions, and \(\mathrm{N}_2\mathrm{O}\) emissions, estimating soil nitrogen loss and applied fertilizer amounts in different scenarios (Text S2.2).  
+
+<|ref|>sub_title<|/ref|><|det|>[[43, 190, 540, 217]]<|/det|>
+## Estimation of the temperature change  
+
+<|ref|>text<|/ref|><|det|>[[41, 232, 944, 322]]<|/det|>
+The CDR were further translated into biogeochemical temperature changes using the compact ESM (OSCAR, Gasser et al., 2017; Text S2.3). OSCAR simulated temperature changes related to CDR processes and GHG emissions from fertilization, considering modeling uncertainties with a sample size of 2000. Global biogeochemical cooling effects were calculated by aggregating regional outputs.  
+
+<|ref|>text<|/ref|><|det|>[[40, 338, 944, 567]]<|/det|>
+The biophysical temperature changes were simulated by the coupled land- atmosphere model IPSL- CM (Boucher et al., 2020), in which ORCHIDEE- MICT- BIOENERGY serves as the land surface model (Wang et al., 2021), LMDz (v6) served as the atmosphere model (Hourdin et al., 2006; Contoux et al., 2012; Text S3). Ocean and sea- ice models were not activated. The simulations, spanning 50 years with 2014 atmospheric \(\mathrm{CO}_2\) levels (Sitch et al., 2015; Peng et al., 2015), employed a spatial resolution of \(1.26^{\circ}\times\) \(2.5^{\circ}\) . The study conducted five coupled simulations, including switchgrass cultivation scenarios in the CN and non- CN countries under the low- warming and overshoot scenarios, and a reference simulation without bioenergy crops (Table S1). The simulations reached a steady state between the fifth and tenth years for switchgrass, and results from the last decade (41st to 50th years) were analyzed for biophysical effects. The cultivation map in 2100 was used for the simulations of biophysical effects.  
+
+<|ref|>text<|/ref|><|det|>[[41, 583, 937, 675]]<|/det|>
+The net air temperature change \((\Delta \mathrm{T}_{\mathrm{net}})\) induced by switchgrass cultivation in this study includes 1) biogeochemical effects from the CDR of BECCS and the fertilization related greenhouse gas emissions (Text S2) and 2) biophysical effects from the changed local energy budget and the altered atmosphere circulation (Text S3):  
+
+<|ref|>equation<|/ref|><|det|>[[372, 688, 625, 711]]<|/det|>
+\[\Delta \mathrm{T}_{\mathrm{net}} = \Delta \mathrm{T}_{\mathrm{bgc}} + \Delta \mathrm{T}_{\mathrm{bph}}\]  
+
+<|ref|>text<|/ref|><|det|>[[43, 730, 55, 744]]<|/det|>
+1  
+
+<|ref|>text<|/ref|><|det|>[[41, 766, 927, 810]]<|/det|>
+The subscript represents the air temperature contributed by the biogeochemical effects ("bgc") or the biophysical effects ("bph").  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 832, 210, 857]]<|/det|>
+## Declarations  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 873, 209, 892]]<|/det|>
+## Acknowledgments  
+
+<|ref|>text<|/ref|><|det|>[[42, 910, 933, 954]]<|/det|>
+This study was funded by the National Natural Science Foundation of China (grant number: 42175169, 72348001, to W.L.), the National Key R&D Program of China (grant number: 2019YFA0606604, to W.L.),
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 45, 895, 65]]<|/det|>
+the Tsinghua University Initiative Scientific Research Program (grant number: 202230041, to W.L.).  
+
+<|ref|>sub_title<|/ref|><|det|>[[43, 83, 184, 102]]<|/det|>
+## Data availability  
+
+<|ref|>text<|/ref|><|det|>[[42, 121, 664, 141]]<|/det|>
+All data are available in the main text or the supplementary information.  
+
+<|ref|>sub_title<|/ref|><|det|>[[43, 159, 225, 178]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[42, 196, 916, 263]]<|/det|>
+W.L., J.W. and P.C. designed the study, J.Z., W.L. and J.W. carried out the modeling and analysis. J.Z., W.L. and J.W. wrote the first draft. P.C., T.G., Z.L., L.Z., M.H., J.H., M.S., L.L., X.H. contributed to the interpretation of the results, the draft revision and the computational tools.  
+
+<|ref|>sub_title<|/ref|><|det|>[[43, 280, 222, 300]]<|/det|>
+## Competing interests  
+
+<|ref|>text<|/ref|><|det|>[[43, 319, 430, 338]]<|/det|>
+The authors declare no competing interests.  
+
+<|ref|>sub_title<|/ref|><|det|>[[43, 361, 193, 386]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[50, 400, 920, 944]]<|/det|>
+1. Harper AB, Powell T, Cox PM et al (2018) Land-use emissions play a critical role in land-based mitigation for Paris climate targets. Nat Commun 9:2938  
+2. Krause A, Pugh TAM, Bayer AD et al (2018) Large uncertainty in carbon uptake potential of land-based climate-change mitigation efforts. Glob Change Biol 24:3025–3038  
+3. Rogelj J, Popp A, Calvin KV et al (2018) Scenarios towards limiting global mean temperature increase below 1.5°C. Nat Clim Change 8:325–332  
+4. Li W, Ciais P, Stehfest E, van Vuuren et al (2020) Mapping the yields of lignocellulosic bioenergy crops from observations at the global scale. Earth Syst Sci Data 12:789–804  
+5. Cai X et al (2011) Land availability for biofuel production. Environ Sci Technol vol 45(1):334–339  
+6. Smith LJ, Torn MS (2013) Ecological limits to terrestrial biological carbon dioxide removal. Clim Change 118:89–103  
+7. Boysen LR et al (2017) The limits to global-warming mitigation by terrestrial carbon removal. Earths Future 5(5):463–474  
+8. Read P (2008) Biosphere carbon stock management: addressing the threat of abrupt climate change in the next few decades: an editorial essay. Clim Change 87:305–320  
+9. Bui M et al (2018) Carbon capture and storage (CCS): the way forward. Energy Environ Sci 11(5):1062–1176  
+10. Wang J et al (2023) Temperature Changes Induced by Biogeochemical and Biophysical Effects of Bioenergy Crop Cultivation. Environ Sci Technol 57:2474–2483  
+11. Wang J, Li W, Ciais P et al (2021) Global cooling induced by biophysical effects of bioenergy crop cultivation. Nat Commun 12:7255  
+12. Friedlingstein P et al (2022) Global Carbon Budget 2021. Earth Syst Sci Data 14:1917–2005
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[48, 45, 930, 112]]<|/det|>
+13. Li W, Yue C, Ciais P et al (2018) ORCHIDEE-MICT-BIOENERGY: an attempt to represent the production of lignocellulosic crops for bioenergy in a global vegetation model. Geosci Model Dev 11:2249-2272  
+
+<|ref|>text<|/ref|><|det|>[[48, 118, 930, 184]]<|/det|>
+14. Popp A, Rose SK, Calvin K et al (2014) Land-use transition for bioenergy and climate stabilization: model comparison of drivers, impacts and interactions with other land use based mitigation options. Clim Change 123:495-509  
+
+<|ref|>text<|/ref|><|det|>[[48, 190, 950, 233]]<|/det|>
+15. Hurtt GC, Chini L, Sahajpal R et al (2020) Harmonization of global land use change and management for the period 850-2100 (LUH2) for CMIP6, Geosci. Model Dev 13:5425-5464  
+
+<|ref|>text<|/ref|><|det|>[[48, 239, 945, 281]]<|/det|>
+16. Gasser T, Ciais P, Boucher O et al (2017) The compact Earth system model OSCAR v2.2: description and first results. Geosci Model Dev 10:271-319  
+
+<|ref|>text<|/ref|><|det|>[[48, 288, 936, 331]]<|/det|>
+17. Boucher O, Servonnat J, Albright AL, Aumont O, Balkanski Y, Bastrikov V et al (2020) Presentation and evaluation of the IPSL-CM6A-LR climate model. J Adv Model Earth Syst, 12, e2019MS002010  
+
+<|ref|>text<|/ref|><|det|>[[48, 337, 888, 380]]<|/det|>
+18. Li W, Ciais P, Han M et al Bioenergy crops for low warming targets require half of the present agricultural fertilizer use. Environ Sci Technol, 55(15), 10654-1066(2021).  
+
+<|ref|>text<|/ref|><|det|>[[48, 386, 912, 429]]<|/det|>
+19. Negri V et al (2021) Life cycle optimization of BECCS supply chains in the European Union. Appl Energy 298:117252  
+
+<|ref|>text<|/ref|><|det|>[[48, 435, 894, 478]]<|/det|>
+20. Fajardy M et al (2021) The economics of bioenergy with carbon capture and storage (BECCS) deployment in a 1.5°C or 2°C world. Glob Environ Change 68:102262  
+
+<|ref|>text<|/ref|><|det|>[[48, 484, 890, 527]]<|/det|>
+21. Sultana A, Kumar A (2011) Optimal configuration and combination of multiple lignocellulosic biomass feedstocks delivery to a biorefinery. Bioresour Technol 102:9947-9956  
+
+<|ref|>text<|/ref|><|det|>[[48, 533, 950, 577]]<|/det|>
+22. Xu S, Wang R, Gasser T et al (2022) Delayed use of bioenergy crops might threaten climate and food security. Nature 609(7926):299-306  
+
+<|ref|>text<|/ref|><|det|>[[48, 583, 950, 626]]<|/det|>
+23. Hourdin F, Musat I, Bony S et al (2006) The LMDZ4 general circulation model: climate performance and sensitivity to parametrized physics with emphasis on tropical convection. Clim Dyn 27:787-813  
+
+<|ref|>text<|/ref|><|det|>[[48, 632, 916, 675]]<|/det|>
+24. Contoux C, Ramstein G, Jost A (2012) Modelling the mid-Pliocene Warm Period climate with the IPSL coupled model and its atmospheric component LMDZ5A, Geosci. Model Dev 5:903-917  
+
+<|ref|>text<|/ref|><|det|>[[48, 681, 916, 724]]<|/det|>
+25. Sitch S, Friedlingstein P, Gruber N et al (2015) Recent trends and drivers of regional sources and sinks of carbon dioxide. Biogeosciences 12, 653  
+
+<|ref|>text<|/ref|><|det|>[[48, 730, 880, 773]]<|/det|>
+26. Peng S, Ciais P, Maignan F et al (2015) Sensitivity of land use change emission estimates to historical land use and land cover mapping. Global Biogeochem Cycles 31:626-643  
+
+<|ref|>sub_title<|/ref|><|det|>[[42, 797, 142, 822]]<|/det|>
+## Figures
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[130, 77, 820, 568]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[42, 628, 115, 647]]<|/det|>
+<center>Figure 1 </center>  
+
+<|ref|>text<|/ref|><|det|>[[39, 668, 941, 830]]<|/det|>
+Bioenergy crop cultivation maps under the low- warming (a) and overshoot (b) scenarios and contributions of the CN and non- CN countries to the global total bioenergy crop cultivation area, net carbon- dioxide removal (CDR), biophysical air temperature change \((\Delta T_{\mathrm{bph}})\) and net air temperature change \((\Delta T_{\mathrm{net}})\) . Blue bars represent changes when cultivating switchgrass in the CN countries only, and yellow bars represent further changes when cultivating switchgrass in both CN and non- CN countries. Arrows represent the directions of changes. The non- CN countries with a cultivation area \(>1\) ha are marked in red. Note the scales in the bar plot are different between (a) and (b).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[52, 52, 949, 479]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[42, 501, 117, 520]]<|/det|>
+<center>Figure 2 </center>  
+
+<|ref|>text<|/ref|><|det|>[[40, 542, 930, 700]]<|/det|>
+Contributions of the CN and non- CN countries to net air temperature change at the regional scale (a), and the cultivation area in the CN and non- CN countries in each region under the low- warming (b) and overshoot (c) scenarios. In (a), blue bars represent the net air temperature change when cultivating switchgrass in the CN countries only, and orange bars represent the temperature changes after further cultivation in the non- CN countries. In (b) and (c), blue bars indicate the cultivation area in the CN countries within each region, and orange bars indicate the further cultivation area in the non- CN countries.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[40, 45, 951, 456]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[42, 492, 116, 511]]<|/det|>
+<center>Figure 3 </center>  
+
+<|ref|>text<|/ref|><|det|>[[40, 534, 953, 692]]<|/det|>
+The cultivation area and net air temperature change in the top ten non- CN countries with the largest cultivation area and the top ten CN countries with the maximum net air temperature change under the low- warming (aand c) and overshoot (b and d) scenarios. Blue arrows refer to the net air temperature change when cultivating switchgrass in the CN countries only, and orange arrows indicate the temperature changes after further cultivation in the non- CN countries. The directions of arrows represent increase and decrease in air temperatures. Black dots in a- d indicate the cultivation area in each country. DRC and CIV are the Democratic Republic of the Congo and the Republic of Côte d'Ivoire.  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 715, 312, 742]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 765, 767, 785]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 803, 375, 823]]<|/det|>
+- NCSIregionalBECCS240109.docx
+
+<--- Page Split --->

preprint/preprint__487268b60d60a57ddf72c596fba8ead5ff013eeab984aee8c25ad6580afe57e1/images_list.json

@@ -0,0 +1,152 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1. Nanosensor integration with microfluidics. (a) Schematic illustration of nanosensor integration process with microfluidics using EISA. (b) Photograph of EISA process of NIM for 0 min (left) and 30 min (right). (c) Photograph of completed multi-array NIM and pristine channel. (d) Polarized Raman spectrum (G-peak) of NIM. (e) nIR images of NIM and pristine channel. (f) Magnified nIR image of NIM with single cell size resolution (20 \\(\\mu \\mathrm{m}\\) ) having \\(\\sim 720\\) nIR reporter pixel. (g) Histograms of nIR pixel intensities of top and bottom NIM surfaces NIM (inset: nIR images of top and bottom surfaces). (h) nIR fluorescence spectrum of NIM. (i) nIR images of NIM with varying composition of SWNT nanosensor integration.",
+    "footnote": [],
+    "bbox": [
+      [
+        125,
+        88,
+        875,
+        555
+      ]
+    ],
+    "page_idx": 20
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. In-vitro chemical detection performances of NIM. (a) nIR spectrum of NIM with \\(\\mathrm{H}_2\\mathrm{O}_2\\) solution flowing (1 \\(\\mu \\mathrm{M}\\) , 1 \\(\\mu \\mathrm{L} / \\mathrm{min}\\) ). (b) nIR images of NIM before and after \\(\\mathrm{H}_2\\mathrm{O}_2\\) flowing (1 M, 10 \\(\\mu \\mathrm{L} / \\mathrm{min}\\) , 10 min). (c) Schematic illustration of \\(\\mathrm{H}_2\\mathrm{O}_2\\) detection mechanism of SWNT/(GT)15 nanosensor. (d) Real-time nIR response of NIM with various concentration ( \\(10^{-6}\\) , \\(10^{-5}\\) , \\(10^{-4}\\) , \\(10^{-3}\\) , \\(10^{-2}\\) , \\(10^{-1}\\) , \\(10^{0}\\) M) of \\(\\mathrm{H}_2\\mathrm{O}_2\\) injection (10 min). (e) Maximum response amplitude and (f) response time of NIM with various concentration of \\(\\mathrm{H}_2\\mathrm{O}_2\\) . The data represent the mean value of 250 X 350 \\(\\mu \\mathrm{m}^2\\) NIM measurement. (g) nIR snapshots and intensity histogram (fire scale, ImageJ) of NIM with single cell size resolution (20 \\(\\mu \\mathrm{m}\\) ) after 10 min flowing of various concentration of \\(\\mathrm{H}_2\\mathrm{O}_2\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        117,
+        90,
+        879,
+        543
+      ]
+    ],
+    "page_idx": 21
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3. Cellular lensing effect. (a) Instrumental setup for NCC implementation: schematic illustration (left) and a photograph (right). (b) nIR images of human monocytes flowing (0.5 \\(\\mu \\mathrm{L} / \\mathrm{min}\\) ) NIM. (c) Magnified nIR image of single monocyte in NIM (inset: OM image of single monocyte). (d) FDTD numerical modeling for photonic nanojet and fitting with experimental cellular lensing profile \\((n_{\\mathrm{c}} / n_{\\mathrm{m}} = 1.04\\) , \\(\\lambda = 1 \\mu \\mathrm{m}\\) ). (e) nIR lensing profiles of a single cell with various focusing points from 5 to \\(100 \\mu \\mathrm{m}\\) along Z-stage. nIR lensing effects of (f) various live cells and (g) reference micro-particles (top-to-bottom: schematics, OM, nIR images, lensing profiles). (h) FWHM and (i) enhancement factors of various cells with numerical model. Data are mean \\(\\pm \\sigma\\) , with \\(n_{\\mathrm{cell}} = 10\\) . (j) Schematic illustrations for different lensing behavior of a high RI cell (left) and a low RI cell (right). Scale bars: \\(20 \\mu \\mathrm{m}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        132,
+        88,
+        866,
+        730
+      ]
+    ],
+    "page_idx": 22
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4. Real-time chemical efflux monitoring using the cellular lensing effect. (a) Time-series nIR images of a stationary single monocyte with different immune activation states (-PMA, +PMA, +PMA & catalase). (b) Real-time nIR intensity variations of the cells with different activation states. (c) Schematic illustrations of \\(\\mathrm{H}_2\\mathrm{O}_2\\) efflux monitoring mechanism with nIR lensing effect. (d) 3D diffusion and reaction kinetic modeling for translation of measured nIR signals to real-time local \\(\\mathrm{H}_2\\mathrm{O}_2\\) concentration. (e) Real-time \\(\\mathrm{H}_2\\mathrm{O}_2\\) efflux profiles of each single monocyte estimated by the model. 16-color scalebars represent nIR intensity from white (16833) to dark blue (0). Scale bars: \\(20\\mu \\mathrm{m}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        90,
+        881,
+        642
+      ]
+    ],
+    "page_idx": 23
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5. NCC for monitoring of multimodal immune response heterogeneities. a) Schematics and nIR images of NCC setup with distinct activation of human monocytes (-PMA and +PMA). (b) Automatic nIR image analysis using computational code for cell data extractions. (c) NCC cytometry plots of \\(\\mathrm{H}_2\\mathrm{O}_2\\) efflux rate \\(\\nu s\\) biophysical parameters ((c1) size (2D projected area), (c2) eccentricity, (c3) RI) of two monocytes populations. Data are \\(n_{\\mathrm{cell}} = 413\\) for -PMA, \\(n_{\\mathrm{cell}} = 414\\) for +PMA from \\(n = 6\\) biologically independent samples. (d) NCC distribution curves of \\(\\mathrm{H}_2\\mathrm{O}_2\\) efflux rates with data from commercial assay kit. (e) NCC cytometry plots for cell biophysical parameters ((e1) eccentricity \\(\\nu s\\) size, (e2) RI \\(\\nu s\\) eccentricity, (e3) size \\(\\nu s\\) RI). (f) NCC distribution curves of each biophysical parameters ((f1) size, (f2) eccentricity, (f3) RI). (g) Schematics illustrations for cell properties variations of human monocyte populations with immune activations. Scale bars: 20 \\(\\mu \\mathrm{m}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        123,
+        88,
+        877,
+        696
+      ]
+    ],
+    "page_idx": 24
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1",
+    "footnote": [],
+    "bbox": [
+      [
+        50,
+        95,
+        941,
+        655
+      ]
+    ],
+    "page_idx": 37
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2",
+    "footnote": [],
+    "bbox": [
+      [
+        50,
+        48,
+        944,
+        585
+      ]
+    ],
+    "page_idx": 38
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3",
+    "footnote": [],
+    "bbox": [
+      [
+        55,
+        50,
+        890,
+        788
+      ]
+    ],
+    "page_idx": 39
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4",
+    "footnote": [],
+    "bbox": [
+      [
+        45,
+        140,
+        950,
+        800
+      ]
+    ],
+    "page_idx": 40
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5",
+    "footnote": [],
+    "bbox": [
+      [
+        52,
+        115,
+        936,
+        840
+      ]
+    ],
+    "page_idx": 41
+  }
+]

preprint/preprint__487268b60d60a57ddf72c596fba8ead5ff013eeab984aee8c25ad6580afe57e1/preprint__487268b60d60a57ddf72c596fba8ead5ff013eeab984aee8c25ad6580afe57e1.mmd

@@ -0,0 +1,392 @@
+
+# Cellular Lensing and Near Infrared Fluorescent Nanosensor Arrays to Enable Chemical Efflux Cytometry  
+
+Soo- Yeon Cho Massachusetts Institute of Technology  
+
+Xun Gong Massachusetts Institute of Technology  
+
+Volodymyr Koman MIT https://orcid.org/0000- 0001- 8480- 4003  
+
+Matthias Kuehne Massachusetts Institute of Technology https://orcid.org/0000- 0002- 5096- 7522  
+
+Sun Jin Moon Massachusetts Institute of Technology  
+
+Manki Son Massachusetts Institute of Technology  
+
+Tedrick Thomas Salim Lew Massachusetts Institute of Technology (M.I.T.) https://orcid.org/0000- 0002- 4815- 9921  
+
+Pavlo Gordiichuk Massachusetts Institute of Technology  
+
+Xiaojia Jin MIT https://orcid.org/0000- 0002- 0694- 5799  
+
+Hadley Sikes Massachusetts Institute of Technology  
+
+Michael Strano ( \(\square\) strano@mit.edu) Massachusetts Institute of Technology https://orcid.org/0000- 0003- 2944- 808X  
+
+## Article  
+
+Keywords: cellular lensing, chemical efflux cytometry, nanosensor chemical cytometry  
+
+Posted Date: December 11th, 2020  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 106483/v1
+
+<--- Page Split --->
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Communications on May 25th, 2021. See the published version at https://doi.org/10.1038/s41467-021-23416-1.
+
+<--- Page Split --->
+
+Cellular Lensing and Near Infrared Fluorescent Nanosensor Arrays to Enable Chemical Efflux Cytometry  
+
+Soo- Yeon Cho, Xun Gong, Volodymyr B. Koman, Matthias Kuehne, Sun Jin Moon, Manki Son, Tedrick Thomas Salim Lew, Pavlo Gordichuk, Xiaojia Jin, Hadley D. Sikes, Michael S. Strano\* Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, United States  
+
+\*Corresponding author: strano@mit.edu
+
+<--- Page Split --->
+
+## ABSTRACT  
+
+Nanosensor have proven to be powerful tools to monitor single biological cells and organisms, achieving spatial and temporal precision even at the single molecule level. However, there has not been a way of extending this approach to statistically relevant numbers of living cells and organisms. Herein, we design and fabricate a high throughput nanosensor array in a microfluidic channel that addresses this limitation, creating a Nanosensor Chemical Cytometry (NCC). An array of nIR fluorescent single walled carbon nanotube (SWNT) nanosensors is integrated along a microfluidic channel through which a population of flowing cells is guided. We show that one can utilize the flowing cell itself as highly informative Gaussian lenses projecting nIR emission profiles and extract rich information on a per cell basis at high throughput. This unique biophotonic waveguide allows for quantified cross- correlation of the biomolecular information with physical properties such as cellular diameter, refractive index (RI), and eccentricity and creates a label- free chemical cytometer for the measurement of cellular heterogeneity with unprecedented precision. As an example, the NCC can profile the immune response heterogeneities of distinct human monocyte populations at attomolar ( \(10^{- 18}\) moles) sensitivity in a completely non- destructive and real- time manner with a rate of \(\sim 100\) cells/frame, highest range demonstrated to date for state of the art chemical cytometry. We demonstrate distinct \(\mathrm{H}_2\mathrm{O}_2\) efflux heterogeneities between 330 and 624 attomole/cell·min with cell projected areas between 271 and \(263~\mu \mathrm{m}^2\) , eccentricity values between 0.405 and 0.363 and RI values between 1.383 and 1.377 for non- activated and activated human monocytes, respectively. Hence, we show that our nanotechnology based biophotonic cytometer has significant potential and versatility to answer important questions and provide new insight in immunology, cell manufacturing and biopharmaceutical research.
+
+<--- Page Split --->
+
+Nanotechnology has produced some of the most sensitive analytical platforms for molecules in existence, with many achieving single molecule resolution, \(^{1 - 3}\) including arrays for DNA sequencing \(^{4,5}\) as well as reactive oxygen species (ROS) detection. \(^{6,7}\) There is significant interest and motivation to extend such platforms to the study of living cells \(^{8,9}\) and microbes \(^{10,11}\) where they can form the basis of non- destructive techniques to probe various biochemical mechanisms. This has obvious applications to medicine and life science research, and of particular importance to the emerging area of cell- based therapies and regenerative medicine for the treatment of cancer, leukemia, and neurodegenerative diseases. \(^{12 - 14}\) However, cellular populations are necessarily heterogeneous, and cellular therapies necessarily require characterization methods that are non- destructive and do not contaminate the cells themselves, \(^{15}\) ruling out conventional flow cytometry that requires fluorescent labels. \(^{16}\) Extending various types of nanosensor to statistically relevant numbers of living cells and organisms in non- destructive manner remains unaddressed to date with the basic problem of nanosensor including interfacing strategy, signal transducing mechanism, and mechanical robustness. \(^{17}\) In this work, we introduce a potential solution in the form of Nanosensor Chemical Cytometry (NCC), a technique that leverages cellular lensing, producing multivariate real- time single cell analysis.  
+
+Various label- free cell imaging techniques such as digital holographic microscopy (DHM) \(^{18 - 20}\) or optical diffraction tomography \(^{21 - 23}\) have been developed for high- throughput cell classification based on image analysis. For example, Ugele et al. discriminated healthy and pathological blood cells using holographic speckle images of DHM technique. \(^{18}\) Singh et al. used machine learning based hologram screening to detect tumor cells in high- throughput. \(^{19}\) However, these techniques are based on physical property measurements from cell images. Chemically quantification for heterogeneity in cell populations is still an open problem. Flow and chemical
+
+<--- Page Split --->
+
+cytometry have been widely used to quantify the molecular heterogeneities of target cell populations. While typical flow and image cytometry of living cells can sample \(10^{6} - 10^{7}\) cells in just a few minutes, \(^{24 - 26}\) the state of the art for the emerging field of chemical cytometry is between 50 to 500 cells/hr since cells need to be pre- labelled, lysed, and separated to be detected. \(^{27 - 29}\) Nevertheless, this level of throughput has elevated chemical cytometry as a valuable cell characterization tool allowing quantitative information to be gathered with high selectivity and signal- to- noise ratio. \(^{30,31}\) Nanosensors have significant potential to greatly expand the number of variables measured in chemical cytometry given the large number of new types being demonstrated in the recent literature. \(^{32 - 36}\) Organic and inorganic fluorescent nanoparticles have been used to monitor intra- and extracellular information of single cells successfully. \(^{34 - 36}\) Near infrared (nIR) fluorescent single walled carbon nanotubes (SWNT) are particularly promising components toward label- free and single molecule level cellular profiling. To date, they have been developed for the detection of single cell biochemical efflux for antibodies, neurotransmitters and ROS. \(^{9,37 - 40}\) Additionally, their rapid and direct optical readout is ideal for sensor interfacing, and carbon in particular possesses photostability, biocompatibility and tunable chemical selectivity for this purpose. \(^{17,40 - 42}\)  
+
+In this work, we develop a new class of chemical cytometry that can characterize the real- time chemical efflux of cell populations at high throughput. nIR fluorescent SWNT nanosensors were uniformly integrated within a cell- transporting microfluidic channel. Each single cell optically interacts with the underlying nanosensor array, producing an informative nIR optical lensing profile that can be modeled as a photonic nanojet. Within this biophotonic waveguide, cells can be both visualized and chemically tracked in real- time and at high- resolution, without the need for labeling or additional optical manipulation. Based on the combination of nanosensor response
+
+<--- Page Split --->
+
+and observed cellular lensing properties, the NCC platform is able to yield multivariate data that inform the heterogeneities of human monocyte populations (immune activated and non- activated) at the attomolar ( \(10^{- 18}\) moles) level of \(\mathrm{H}_2\mathrm{O}_2\) efflux. Furthermore, this type of cellular population data allows for phenotypic correlation between real- time chemical efflux and various biophysical properties of each individual cell including diameter, eccentricity, and refractive index (RI).  
+
+## Nanosensor Integration with Microfluidics  
+
+The schematic of the flow channel and nanosensor array integration for NCC are shown in Figure 1a. The array is demonstrated using a \((\mathrm{GT})_{15}\) DNA wrapped SWNT (SWNT/(GT) \(_{15}\) ), which was previously shown to exhibit nIR intensity attenuation upon selective detection of \(\mathrm{H}_2\mathrm{O}_2\) . \(^{7,42}\) \(\mathrm{H}_2\mathrm{O}_2\) efflux was targeted for the application due to its central role in cellular signaling and immune responses. \(^{6,9}\) For the first step, micro- droplet of (3- aminopropyl) triethoxysilane (APTES) was injected into a pristine channel and incubated. A commercial microfluidic channel was coated with APTES for self- assembled monolayer formation and SWNT/(GT) \(_{15}\) adhesion on both top and bottom surface of the channel. Subsequently, the channel was washed with phosphate buffer saline (PBS) and a micro- droplet of SWNT/(GT) \(_{15}\) dispersion was injected into the channel. Stable dispersions of nanosensors were confirmed via UV- vis- nIR absorption spectra of SWNT/(GT) \(_{15}\) (Figure S1). During evaporation, nanosensor particles necessarily align at the three- phase line of the micro- droplet pinned at the end of the flow channel (Figure 1b). This resulted in a uniform array on both top and bottom surfaces of the channel following the Evaporation Induced Self- Assembly (EISA). \(^{43}\) After the EISA, the channel was flushed with PBS again to remove unbounded residual nanoparticles. Completed Nanosensor Integrated Microfluidics (NIMs) were highly transparent to visible light indicating an absence of aggregation
+
+<--- Page Split --->
+
+or large array defects (Figure 1c). Polarized Raman spectroscopy of NIM showed the depolarization ratio of 0.61 from G band intensity demonstrating that the nanosensors were aligned along the flowing direction of channel during EISA (Figure 1d).<sup>44</sup> nIR imaging was used to investigate the fluorescence signal mapping of the NIM (Figure 1e). While we find NIMs to display strong nIR fluorescence, uncoated channels show no nIR signal (Figure S2). In addition, NIM without APTES treatment showed severe nanosensor aggregation during EISA process and consequently nanosensors were completely removed with PBS flowing, indicating that surface chemistry of the microfluidic channel is critically important to uniform and stable EISA process. Magnified nIR image of NIM with single cell size (20 \(\mu \mathrm{m}\) diameter) shows that nanosensors are homogeneously and continuously deposited with approximately 720 local detector pixels across a single cell (Figure 1f and S3). Atomic force microscopy (AFM) demonstrated that nanosensor bundles were densely and homogenously covered on the channel surface at the micron-scale (Figure S4). Consequently, the nanosensor array on the microfluidic channel could clearly visualize the cells flowing through the channel and maximize the signal- to- noise ratio of the signal from cell efflux for NCC.<sup>38</sup> As the concentration of nanosensor dispersion increases, uniformity of nanosensor array was enhanced with significant decrease of voids and aggregation of nanosensors and 80 mg/L coating shows highest nIR intensity with most uniform pixel distributions (Figure S5). Nanosensors were uniformly coated on the top and bottom surfaces of the channel during EISA, as shown by the comparable nIR pixel distributions along both surfaces. (Figure 1g). Peak position and relative peak intensities of nIR spectrum of NIM were almost identical with SWNT in dispersion phase, indicating that the dielectric environment surrounding the immobilized nanosensors were similar (Figure 1h).<sup>45</sup> Varying compositions of SWNT nanosensors ((GT)<sub>15</sub> DNA, (AT)<sub>15</sub> DNA, (ATT)<sub>10</sub> DNA, random DNA, chitosan) were integrated
+
+<--- Page Split --->
+
+with the microfluidic channels by our EISA based NIM fabrication process for monitoring of various chemical components of the cell (Figure 1i).<sup>46</sup>  
+
+## Chemical Detection Performances of NIM  
+
+In- vitro \(\mathrm{H}_2\mathrm{O}_2\) detection experiments were conducted to investigate the chemical sensing performance of the NIM. The fluorescence intensity from all SWNT chiralities decreased with 10 to \(20\%\) relative magnitude upon exposure to \(1\mu \mathrm{M} \mathrm{H}_2\mathrm{O}_2\) (Figure 2a). Real- time nIR images of NIM shows that the channel emission is completely quenched with \(1\mathrm{M} \mathrm{H}_2\mathrm{O}_2\) flowing (Figure 2b). This is attributed to that \(\mathrm{H}_2\mathrm{O}_2\) molecules selectively adsorbed on nanotube sidewall donate electrons directly to the conduction bands of SWNT/(GT)<sub>15</sub>, and extra electrons in the conduction bands can then quench excitons through a non- radiative recombination (Figure 2c).<sup>7,47</sup> Real- time nIR signals ((I - I<sub>0</sub>)/I<sub>0</sub>) were measured with wide range concentration of \(\mathrm{H}_2\mathrm{O}_2\) injection (Figure 2d). Here, \(I_0\) and \(I\) represent the nIR intensity of the channel at \(t = 0\) and after \(\mathrm{H}_2\mathrm{O}_2\) injection, respectively. Upon \(\mathrm{H}_2\mathrm{O}_2\) injection, the NIM showed an instantaneous and continuous decrease in nIR signal on the order of \(5 - 80\%\) depending on \(\mathrm{H}_2\mathrm{O}_2\) concentration. For the first- order reversible reaction, the relationship between the analyte and available docking sites for \(\mathrm{H}_2\mathrm{O}_2\) can be described as follows:<sup>48</sup>  
+
+\[A + \theta \rightleftharpoons A\theta \quad (1)\]  
+
+the equilibrium for this reaction can be modeled as:  
+
+\[K_{A} = \frac{[A\theta]}{[A][\theta]} \quad (2)\]  
+
+Assuming that the sensor response is proportional to the \(A\theta /\theta_{\mathrm{tot}}\) ratio, it is found that  
+
+\[\frac{I - I_0}{I_0} = \alpha \frac{[A\theta]}{[\theta_{\mathrm{tot}}]} + \beta = \alpha \frac{([A]K_A)^n}{([A]K_A)^{n + 1}} +\beta \quad (3)\]  
+
+with the total concentration of available recognition sites \([\theta ]_{\mathrm{tot}}\) and the parameter \(n\) for
+
+<--- Page Split --->
+
+cooperativity. Fitting the data in Figure 2e with equation (3) \((R^{2} = 0.9983)\) results in a proportionality factor \(\alpha = 88.74\) with \(\beta = 2.30\) , \(K_{\mathrm{D}} = 1 / K_{\mathrm{A}} = 0.00204 \mathrm{M}\) , and \(n = 0.317\) , indicating negative cooperativity in good agreement with previous papers \((n< 1)\) . \(^{42,45,48}\) The limit of detection in this mode is \(11.56 \mathrm{nM}\) ; this value was calculated by adding the NIM sensor response from the addition of only buffer (PBS) to 3- times the standard deviation \((\sigma)\) . A response time of less than 9 min was achieved based on the time it takes to reach \(90\%\) value of the minimum nIR level (Figure 2f). The NIM platform demonstrates uniform and near instantaneous nIR intensity response even when imaged at the high- resolution needed to interrogate single cells \((\sim 20 \mu \mathrm{m})\) (Figure 2g).  
+
+## Cellular Lensing Effect  
+
+For the NCC implementation, the NIM was integrated with a syringe pump and nIR microscope. 561 nm excitation laser was provided from the bottom side of the channel (right, Figure 3a). Human monocytes (U937) were cultured as chemical cytometry targets (Figure S6) since they are widely studied in biomedical fields with heterogeneous differentiation behavior into macrophages by immune activation. \(^{49,50}\) This monocyte- derived macrophage exhibits distinct ROS efflux in real- time as an immune response to various kinds of infection/inflammation. Measuring subtle molecular differences of ROS efflux can also benefit the detection and prevention of cardiovascular disease and neurodegenerative disorders. \(^{51,52}\) Therefore, a tool that would enable the precise profiling of dynamic antigenic response of single monocyte and eventually immune heterogeneities as function of different cellular physical properties could lead to mechanistic understanding and therapeutic development for these conditions. We found that the flowing cells optically interact with the underlying nanosensor emitter array and create a moving, label- free region of highest sensor signal by lensing the photoemission through the flowing cell itself (Figure
+
+<--- Page Split --->
+
+3b- 3c and Video S1). This cell visualization was directly affected by both the uniformity and intensity of underlying nanosensor array (Figure S7). A magnified nIR image of a single flowing cell shows that the contour and shape of the monocyte could be visualized as observed in an OM (inset, top- right) with highest nIR intensity from the nanosensor array corresponding to the center, and Airy rings visible around the periphery (Figure 3c). Micro- particles larger than the illumination wavelength can similarly function as focusing lens. \(^{53,54}\) When particles have a RI contrast ratio with the fluid medium less than 2:1 and a diameter \((d_{\mu})\) larger than the wavelength \((- 2\lambda < d_{\mu}< 40\lambda)\) , a highly focused propagating beam from the shadow- side of the surface is generated due to constructive interference of the light field, called a photonic nanojet. \(^{55,56}\) For our system, the nIR fluorescence \((\lambda :1 - 1.25\mu \mathrm{m})\) from the top nanosensor array passes through the membranes, cytoplasm, and nucleus of the underlying flowing cells of mean diameter \(10 - 20\mu \mathrm{m}\) . The estimated RI of the cell components are \(n_{\mathrm{n}} = 1.43\pm 0.04\) for the monocyte nucleus, \(n_{\mathrm{c}} = 1.348\pm 0.004\) for the monocyte cytoplasm (average \(n_{\mathrm{cell}} = 1.383\) ) and \(n_{\mathrm{m}} = 1.33\) for the flowing media, which are optimum optical conditions for the photonic nanojet effect \((n_{\mathrm{cell}} / n_{\mathrm{m}} = 1.039\) (< 2)). \(^{57,58}\) Consequently, nIR photoemission from the integrated nanosensor array was refracted through the flowing cell and focused at a certain focal point below it, a phenomenon called cellular lensing. Several previous papers reported such photonic nanojet based micro- lensing behavior of cells in visible spectrum. \(^{59 - 62}\) In this study, we observe the photonic nanojet phenomena through a cell in the nIR range. Based on this, we could correlate the various biophysical properties of live cells and photonic nanojet effects for the first time. Consequent nIR intensity profiles of single cell showed the highest lens intensity \((I_{0})\) with 3 - 5 \(\mu \mathrm{m}\) full width at half- maximum (FWHM) and following Airy rings corresponding to the cell diameter (red plot in Figure 3d). This nIR lensing profile was measured for multiple cells \((n = 20)\) with almost identical FWHM \((3.37\mu \mathrm{m},\sigma = 0.22)\)
+
+<--- Page Split --->
+
+and enhancement factor (9.43, \(\sigma = 1.86\) ), indicating that this lensing effect is reliable and specific to certain cell status (Figure S9).  
+
+Finite- difference time- domain (FDTD) numerical modeling can demonstrate the cellular lensing as originating from a photonic nanojet effect. \(^{55,56}\) Micro- spherical structures having similar diameters between 10 to \(20 \mu \mathrm{m}\) , eccentricity, and RI values \((n_{\mathrm{c}} / n_{\mathrm{m}} = 1.04)\) compared with cells were used as targets for FDTD modeling. The spherical target is excited by an incident plane wave of wavelength \(1 \mu \mathrm{m}\) corresponding to the fluorescence emission of the nanosensor array (modeling details in Methods and Supporting Information). The resulting optical intensity distribution map shows that light from top side of the target strongly focuses at a \(20 \mu \mathrm{m}\) distant point from the center of the cell forming a 2 to \(4 \mu \mathrm{m}\) wide light jet (left, Figure 3d). The model describes the experimental light intensity profile of the cellular nanojet at \(20 \mu \mathrm{m}\) focal distance with high fidelity in terms of \(I_{0}\) , FWHM, including Airy rings (right, Figure 3d). We note a slight deviation between the FDTD model and experiment for the Airy rings and FWHM possibly originating from the nonuniformity of nIR excitation source and ellipticity of the monocytes. When the excitation light was focused on the bottom surface of the NIM (at \(Z\) - stage \(= 100 \mu \mathrm{m}\) ), the target cell is not distinguishable above the background (red line) (Figure 3e and S10). A slight lensing peak \((I_{0})\) begins to be observed at \(80 \mu \mathrm{m}\) (orange line), and is highest in intensity at \(20\) to \(30 \mu \mathrm{m}\) distance from the top surface with an enhancement factor of 9.1 (blue line), in agreement with the focal points of the FDTD numerical model. The variation in lensing intensity as a function of focusing distance also shows excellent agreement between model and experiment (Figure S14). This agreement give confidence that cellular lensing images are indeed projected \(20 \mu \mathrm{m}\) from the cell center and therefore observable for those flowing within \(10 \mu \mathrm{m}\) of the NIM top surface.
+
+<--- Page Split --->
+
+This unique nIR lensing effect was not only observed for monocytes, but also for other type of cells including lymphocyte, macrophages, epithelial cells (e.g. human embryonic kidney cells (HEK)), and endothelial cells (e.g. human umbilical vein endothelial cells (HUVEC)) (Figure 3f). Since all the cells are composed of cytoplasm, nucleus, and membrane, \(^{58}\) which have higher RI than that of media \((n_{\mathrm{m}})\) but ratio under 2, all cell species could form photonic nanojet and nIR lensing effect following their own shape and contour. Even cells that adhere on the channel surface such as HUVEC apparently display the profile of nIR lensing albeit with weaker intensity than suspended cells due to the smaller thickness (Figure S11). In contrast, reference micro-particles similar in size with cells of interest between 15 to \(25 \mu \mathrm{m}\) such as glass spheres, polystyrene (PS) and stainless steel particles with higher RI than \(n_{\mathrm{m}}\) (1.457, 1.586, and 2.756, respectively) display no nIR lensing. Note that the nIR fluorescence is highly refracted or reflected on surfaces and overfocused within such reference particles due to the high RI values (Figure 3g). \(^{55}\) In addition, we observe significantly weaker cellular lensing with lysed monocytes. In this case, the absence of cytoplasmic content reduces the RI to close to \(n_{\mathrm{m}} \sim 1.33\) , inhibiting nIR refraction. Accordingly, the observed nIR lensing effect appears to be a phenomenon unique to live cells having optimal RI, diameter and composition for the formation of a nIR photonic nanojet.  
+
+Distinct nIR lensing profiles were observed for each cell type corresponding to unique RI ranges \((1.40 \pm 0.01, 1.384 \pm 0.015, 1.383 \pm 0.006, 1.37,\) and \(1.355 \pm 0.0007\) , for B lymphocyte, \(^{63}\) macrophage, \(^{64}\) monocyte, \(^{57}\) HEK, \(^{65}\) and HUVEC, \(^{66}\) respectively; RI values are those reported previously). The FWHM and enhancement factor of each cell can be calculated and described with a FDTD numerical model (Figure 3h and 3i, respectively) with good agreement \((R^{2} = 0.942\) and 0.950, respectively). Model predictions show that cellular lensing can be utilized to estimate a wide range of biophysical properties of the cell including diameter, eccentricity, and RI (Figure
+
+<--- Page Split --->
+
+S12- S14). For example, cells with higher RI show distinctly smaller FWHM and larger enhancement factors than cells with lower RI, in agreement with the FDTD model. High RI cells such as B lymphocyte are composed of larger nucleus volumes \((n_{n})\) than cytoplasmic components \((n_{c})\) for antibody and cytokine production. \(^{67}\) Thus, the nIR excitation wave becomes more refracted through a high RI cell and thus more tightly focused onto focal points compared with low RI HUVEC cells (Figure 3j). \(^{58}\) In this way, nIR cellular lensing in this NCC platform provides a unique opportunity to cross- correlate the chemical efflux as measured by the underlying nanosensor array with distinct biophysical properties such as cell diameter, eccentricity, and RI. Ultimately, these properties can be linked to critical attributes such as viability, membrane properties, or intracellular composition, quantitatively correlating them with biochemical signaling information.  
+
+## Real-time Chemical Efflux Detection Using Cellular Lensing Effect  
+
+We find that when human monocytes are injected into the NIM in a controlled stopped flow system, distinct nIR intensity variations can be observed for individual monocytes corresponding to different immune activation states (Figure 4a). We use phorbol 12- myristate 13- acetate (PMA) to induce immune activation of the human monocytes, since it is a known agonist of the protein kinase C (PKC) signaling cascade. PKC activates nicotinamide adenine dinucleotide phosphate (NADPH) oxidase and consequently stimulates \(\mathrm{H}_2\mathrm{O}_2\) secretion during differentiation into macrophages (Figure S16). \(^{68}\) NADPH oxidase activity generates other ROS species including superoxide anion \((\mathrm{O}_2\cdot \cdot)\) and hydroxyl radical \((\mathrm{OH}\cdot \cdot)\) of course but at significantly lower levels of \(10^{3}\) and \(10^{8}\) times less than \(\mathrm{H}_2\mathrm{O}_2\) , respectively. \(^{69,70}\) It is safe to assume that \(\mathrm{H}_2\mathrm{O}_2\) is the dominant efflux from monocyte activation. Time series nIR images show that the \(I_0\) corresponding to the
+
+<--- Page Split --->
+
+immune activated monocyte (+PMA) (middle of Figure 4a) decreases relative to non- activated monocyte (- PMA) (left of Figure 4a) with increasing time. Catalase, an enzyme that decomposes \(\mathrm{H}_2\mathrm{O}_2\) , \(^{71}\) suppresses the signal as a negative control (right of Figure 4a). To analyze quantitatively, the nIR pixels corresponding to the nanosensor array were integrated for each cell and labeled \((I_{\mathrm{cell}})\) , producing three cell populations per experiment (+PMA, - PMA, and +PMA & catalase) (Figure 4b). Activated monocytes show significant variation in their real- time nIR nanosensor response while - PMA showed slow and small variation over the 500 sec measurement window. We detect a basal \(\mathrm{H}_2\mathrm{O}_2\) level even for the non- activated monocytes without PMA activation, which is consistent with the literature. \(^{72}\) As expected, +PMA & catalase showed invariant sensor responses attributed to \(\mathrm{H}_2\mathrm{O}_2\) decomposition by the enzyme. The +PMA group \((n = 41)\) had average 4.5- and 3.4- times higher intensity variations than - PMA and +PMA & catalase groups (Figure S17). Also, the nIR image of single monocytes show distinct quenching traces after measurements consistent with a response due to \(\mathrm{H}_2\mathrm{O}_2\) efflux (Figure S18a- b).  
+
+The schematic in Figure 4c summarizes this real- time \(\mathrm{H}_2\mathrm{O}_2\) efflux detection for single cells using the cellular lensing effect. The moving cell within the flow field exhibits strong nIR lensing from the induced photonic nanojet while the \(\mathrm{H}_2\mathrm{O}_2\) efflux is minimal at the underlying nanosensor array (Figure S18c- d). During the periodic stopped flow, the \(\mathrm{H}_2\mathrm{O}_2\) efflux cloud surrounding each cell starts to register on the projected nanosensor area nearest to the cell, resulting in a quenching of the immediate spot. This quenching allows for precise quantification of the \(\mathrm{H}_2\mathrm{O}_2\) efflux. At this point, the nIR lensing power is drastically reduced with weaker fluorescence resulting from the waveguide light source. We modeled the 3D reaction and diffusion problem of the \(\mathrm{H}_2\mathrm{O}_2\) from the individual cell to translate the observed nIR quenching area above the cell into real- time local \(\mathrm{H}_2\mathrm{O}_2\) concentration (Figure 4d). An individual cell is assumed to be stationary
+
+<--- Page Split --->
+
+below the top channel surface with distance \(L\) ( \(\sim 10 \mathrm{um}\) ) and to instantaneously release \(\mathrm{H}_2\mathrm{O}_2\) molecules at \(t = 0\) sec. The effective distance between the source and nanosensor array \((L_{\mathrm{eff}})\) is then \(L_{\mathrm{eff}} = L + L_{\mathrm{cell}}\) , where \(L_{\mathrm{cell}}\) is the cell radius. The \(\mathrm{H}_2\mathrm{O}_2\) concentration \(C\) field is then  
+
+\[C(x,y,z,t) = \frac{M}{(\sqrt{4\pi D t})^3}\exp \left(-\frac{x^2 + y^2 + z^2}{4D t} -K t\right) \quad (4)\]  
+
+where \(M\) is the mass flux of \(\mathrm{H}_2\mathrm{O}_2\) release at the cell core, \(D\) is the aqueous diffusion coefficient of \(\mathrm{H}_2\mathrm{O}_2\) \((1.5\cdot 10^{- 5}\mathrm{cm}^2\cdot \mathrm{sec}^{- 1})\) , \(^{73}K\) is the first- order decay constant of \(\mathrm{H}_2\mathrm{O}_2\) (from \(K = - \ln (0.5) / t_{1 / 2} =\) \(6.93\cdot 10^{- 4}\mathrm{sec}^{- 1}\) , where \(t_{1 / 2}\) is cellular half- life of \(\mathrm{H}_2\mathrm{O}_2\) \((10^{- 3}\mathrm{sec}))^{74}\) (detail model derivations in Supporting Information). The results show that the \(\mathrm{H}_2\mathrm{O}_2\) efflux reached the nearest nanosensor array quickly at 10 milli sec with a maximum concentration \(C_{\mathrm{sensor}}\) and the ratio between \(C_{\mathrm{sensor}}\) and \(C_{\mathrm{cell}}\) was 0.193 (Figure S19). The adsorption and desorption of \(\mathrm{H}_2\mathrm{O}_2\) on nanosensor array can be described by  
+
+\[\mathrm{H}_2\mathrm{O}_2 + \mathrm{SWNT}\rightleftharpoons \mathrm{H}_2\mathrm{O}_2\mathrm{-SWNT} \quad (5)\]  
+
+Corresponding to the rate expression: \(^{75}\)  
+
+\[\frac{d[H_2O_2 - SWNT]}{dt} = k_f[H_2O_2][SWNT] - k_r[H_2O_2 - SWNT] \quad (6)\]  
+
+where \(k_{\mathrm{f}}\) , \(k_{\mathrm{r}}\) are the forward and backward rate constants, respectively, and ratio between \(k_{\mathrm{f}}\) and \(k_{\mathrm{r}}\) was calculated from the effective equilibrium dissociation constant \(K_{\mathrm{D}} = 0.00204 \mathrm{M}\) . Since the nIR intensity of the nanosensor array is proportional to the fraction of unoccupied sites for binding, [SWNT], or  
+
+\[I / I_0 = [\mathrm{SWNT}] / [\mathrm{SWNT}]_0 \quad (7)\]  
+
+And the number of binding sensor sites are conserved:  
+
+\[[\mathrm{SWNT}]_0 = [\mathrm{SWNT}] + [\mathrm{H}_2\mathrm{O}_2\mathrm{-SWNT}] \quad (8)\]  
+
+The local concentration of \(\mathrm{H}_2\mathrm{O}_2\) detected by the nanosensor array involves the measured intensity \((I)\) and its time- derivative
+
+<--- Page Split --->
+
+\[[H_{2}O_{2}] = \frac{1}{k_{f}}\frac{I_{0}}{l}\left[k_{r}\left(1 - \frac{l}{I_{0}}\right) - \frac{1}{I_{0}}\frac{dI}{dt}\right] \quad (9)\]  
+
+Integrating equation (9) yields  
+
+\[\begin{array}{l}{I(t) = \frac{I_0}{k_s} (k_r + k_f[H_2O_2]e^{-k_s t})}\\ {}\\ {k_s = k_r + k_f[H_2O_2]} \end{array} \quad (10)\]  
+
+Equation (10) can be utilized to estimate the real- time local \(\mathrm{H}_2\mathrm{O}_2\) concentration of each single cell from the measured nIR intensity (Figure 4e). Both the efflux signal ([H2O2]cell, red line) and background ([H2O2]bg, green line) for each single monocyte can be measured and differentiated. Furthermore, each monocyte (cell 1 to cell 8) demonstrates distinct \(\mathrm{H}_2\mathrm{O}_2\) efflux rates resulting in local concentrations ranging from 175 to \(750~\mu \mathrm{M}\) . This shows that our NCC platform can inform heterogeneities in the efflux rates within cell populations.  
+
+## NCC for Monitoring of Multimodal Immune Response Heterogeneities  
+
+The combination of cellular lensing and label- free nanosensor monitoring within a microfluidic channel allows for real- time chemical efflux cytometry of distinct human monocyte populations, such as those that are immune activated (+PMA) compared to non- activated (- PMA) (Figure 5a). We show that the NCC platform collects a rich, multivariate data set for each individual cell within the population, that we then easily extract and evaluate with the aid of image analysis code developed as a part of this work (Figure 5b and S20). The results allow us to plot the real- time \(\mathrm{H}_2\mathrm{O}_2\) efflux rates of two distinct groups ( \(n = 413\) for - PMA, \(n = 414\) for +PMA) versus various biophysical properties of each individual cell such as size (cell projected area), eccentricity, and RI (Figure 5c). Upon immune activation, we find that the mean size of monocytes decreases along with a narrowing of the distribution (Figure 5c1). This occurs with an increase of \(\mathrm{H}_2\mathrm{O}_2\) efflux rate. In contrast, the eccentricity (Figure 5c2) and RI (Figure 5c3) distributions show
+
+<--- Page Split --->
+
+insignificant correlation with \(\mathrm{H}_2\mathrm{O}_2\) efflux rate. To be clear, 3D cytometry and 2D Kernel density estimation show these distinct heterogeneities in detail (Figure S21 and S22, respectively). From these cytometry plots, it is clear that the average \(\mathrm{H}_2\mathrm{O}_2\) efflux rate of activated monocyte population were elevated by \(88.9\%\) with a \(44.5\%\) larger increase in the variance of the distribution and \(30\%\) larger number of high efflux cells compared to non- activated populations (Figure 5d). The nanosensor array allows us to quantify the mean \(\mathrm{H}_2\mathrm{O}_2\) efflux rates of these two populations as 330 and 624 attomole/cell·min but with \(\sigma\) of 344 and 497 attomole/cell·min for - PMA and +PMA, respectively. In comparison, we measure average values of 59 (-PMA) and 440 (+PMA) attomole/cell·min from the commercial assay Amplex UltraRed kit (Figure S23). The +PMA mean values are in good agreement for the NCC population and commercial assay. However, the mean for the - PMA as measured by NCC is larger than the commercial assay. Further analysis indicates that hyperactive outliers (>1000 attomole/cell·min) in this population are the cause of the difference. The mode in the - PMA distribution as measured by NCC of 129 attomole/cell·min is closer to the commercial assay mean, and NCC distribution curves shows the significant higher active tail in Figure 5d. The ability to detect and quantify this higher producing subpopulation is a clear advantage of NCC over the standard assay. As a consistency check, we note that both methods produce the correct order of magnitude estimate of the \(\mathrm{H}_2\mathrm{O}_2\) efflux rates.  
+
+Among the biophysical property changes, the size vs eccentricity correlation shows the most dramatic change after immune activation (Figure 5e and S24). There is a distinct change in the size distribution upon monocyte activation, with bimodal subpopulations observed for non- activated monocytes with a mean of \(271 \mu \mathrm{m}^2\) ( \(\sigma = 29\) ) but a single distribution with lower mean of \(263 \mu \mathrm{m}^2\) ( \(\sigma = 24\) ) after activation (Figure 5f1). This observation is important because one requires single cell resolution in order to quantify this type of biophysical change, underscoring an
+
+<--- Page Split --->
+
+advantage of this NCC platform. Notably, the distributions for both eccentricity (Figure 5f2) and RI (Figure 5f3) remain nearly identical comparing before and after activation but the mean values are slightly shifted from 0.405 \((\sigma = 0.14)\) to 0.363 \((\sigma = 0.13)\) for eccentricity and 1.383 \((\sigma = 0.05)\) to 1.377 \((\sigma = 0.06)\) for RI. This indicates that immune activation had a uniform effect on the cell populations with respect to these properties. The ability to detect and analyze subpopulations from a cellular population undergoing biofunctional changes has significant advantages in analytical biochemistry.  
+
+Figure 5g summarizes the variation in human monocyte characteristics before and after the immune activation process. Real- time \(\mathrm{H}_2\mathrm{O}_2\) efflux rate of monocyte populations showed \(88.9\%\) elevation. Populations showed \(- 2.92\%\) and \(- 10.31\%\) decrease in cell size and eccentricity respectively, indicating that monocytes appear to shrink and become more circular with immune activation. The RI of the populations decreased by \(- 0.3\%\) scale, which means that light refracted through activated cells produced almost identical refraction angles. This new insight may lead to additional methods of sorting human monocyte populations. As a consistency check, all of the measured values of NCC were within the ranges previously reported for monocytes, including \(\mathrm{H}_2\mathrm{O}_2\) efflux rate: \(^{72}\) 100 to 1000 attomole/cell·min, size: \(^{76}\) 78.5 to \(314 \mu \mathrm{m}^2\) , eccentricity: \(^{57}\) 0.323 to 0.473, RI: \(^{57,58}\) 1.377 to 1.389. We can safely conclude that our NCC approach is reliable in this way and allows the investigation of multiple cellular parameters of a given population in real- time and at high throughput. These cellular parameter changes of monocytes during activation are consistent with PKC translocation effects. It is known that when monocytes are activated by PMA, PKC proteins are translocated from the cytosol to the plasma membrane, activating NADPH oxidase with an increase in ROS generation. \(^{77}\) Subsequently, fluidity and permeability of the cellular membrane are both downregulated upon PKC integration. \(^{78,79}\) One expects a resulting
+
+<--- Page Split --->
+
+image, producing a profile that matches the predictions of photonic nanojet model. The result is a unique tool capable of multimodal biophysical characterization of individual cells, including their size, eccentricity, and RI, all at high throughput. With this biophotonic waveguide, the chemical efflux of single cell was label- free monitored in real- time at the attomolar level. We use this NCC tool to study the heterogeneity of the immune response of distinct human monocyte populations at highest throughput range for chemical cytometry ( \(\sim 100\) cell/frame) in a completely nondestructive manner. Mathematical analysis of the resulting rich data sets reveals new phenotypic correlations between chemical efflux and biophysical properties that can quantified, and used to understand new aspect of cellular biochemistry and mechanistic pathways. For example, we find that real- time \(\mathrm{H}_2\mathrm{O}_2\) efflux of human monocytes is unusually heterogeneous and distinctly related to biophysical parameters following immune activation. The measured \(\mathrm{H}_2\mathrm{O}_2\) efflux rates between 330 and 624 attomole/cell·min corresponded to overall cell size ranges of 271 and \(263~\mu \mathrm{m}^2\) , eccentricity values between 0.405 and 0.363 and RI values between 1.383 and 1.377 for nonactivated and activated monocytes, respectively. Thus, we highlight that NCC is able to profile immune cell heterogeneities allowing for monitoring of variances in cell therapeutics. We also demonstrate the ability to incorporate sensors for multiple molecular targets of cells. Our platform is label- free and uses the unique property of cellular lensing to extract molecular signals on a population scale. We believe that the NCC platform can be readily extended to various biochemical efflux monitoring of cell types such as neurons, cancer cells or stem cells given the appropriate choices of sensor- analyte pairs (Figure S27). We envision that our nanotechnology based biophotonic cytometry provides a unique strategy for coupling nanosensors into a form- factor that enables single- cell analysis of relevant populations for cellular manufacturing, cellular immunology, and biopharmaceutical research.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Figure 1. Nanosensor integration with microfluidics. (a) Schematic illustration of nanosensor integration process with microfluidics using EISA. (b) Photograph of EISA process of NIM for 0 min (left) and 30 min (right). (c) Photograph of completed multi-array NIM and pristine channel. (d) Polarized Raman spectrum (G-peak) of NIM. (e) nIR images of NIM and pristine channel. (f) Magnified nIR image of NIM with single cell size resolution (20 \(\mu \mathrm{m}\) ) having \(\sim 720\) nIR reporter pixel. (g) Histograms of nIR pixel intensities of top and bottom NIM surfaces NIM (inset: nIR images of top and bottom surfaces). (h) nIR fluorescence spectrum of NIM. (i) nIR images of NIM with varying composition of SWNT nanosensor integration. </center>
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Figure 2. In-vitro chemical detection performances of NIM. (a) nIR spectrum of NIM with \(\mathrm{H}_2\mathrm{O}_2\) solution flowing (1 \(\mu \mathrm{M}\) , 1 \(\mu \mathrm{L} / \mathrm{min}\) ). (b) nIR images of NIM before and after \(\mathrm{H}_2\mathrm{O}_2\) flowing (1 M, 10 \(\mu \mathrm{L} / \mathrm{min}\) , 10 min). (c) Schematic illustration of \(\mathrm{H}_2\mathrm{O}_2\) detection mechanism of SWNT/(GT)15 nanosensor. (d) Real-time nIR response of NIM with various concentration ( \(10^{-6}\) , \(10^{-5}\) , \(10^{-4}\) , \(10^{-3}\) , \(10^{-2}\) , \(10^{-1}\) , \(10^{0}\) M) of \(\mathrm{H}_2\mathrm{O}_2\) injection (10 min). (e) Maximum response amplitude and (f) response time of NIM with various concentration of \(\mathrm{H}_2\mathrm{O}_2\) . The data represent the mean value of 250 X 350 \(\mu \mathrm{m}^2\) NIM measurement. (g) nIR snapshots and intensity histogram (fire scale, ImageJ) of NIM with single cell size resolution (20 \(\mu \mathrm{m}\) ) after 10 min flowing of various concentration of \(\mathrm{H}_2\mathrm{O}_2\) . </center>
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3. Cellular lensing effect. (a) Instrumental setup for NCC implementation: schematic illustration (left) and a photograph (right). (b) nIR images of human monocytes flowing (0.5 \(\mu \mathrm{L} / \mathrm{min}\) ) NIM. (c) Magnified nIR image of single monocyte in NIM (inset: OM image of single monocyte). (d) FDTD numerical modeling for photonic nanojet and fitting with experimental cellular lensing profile \((n_{\mathrm{c}} / n_{\mathrm{m}} = 1.04\) , \(\lambda = 1 \mu \mathrm{m}\) ). (e) nIR lensing profiles of a single cell with various focusing points from 5 to \(100 \mu \mathrm{m}\) along Z-stage. nIR lensing effects of (f) various live cells and (g) reference micro-particles (top-to-bottom: schematics, OM, nIR images, lensing profiles). (h) FWHM and (i) enhancement factors of various cells with numerical model. Data are mean \(\pm \sigma\) , with \(n_{\mathrm{cell}} = 10\) . (j) Schematic illustrations for different lensing behavior of a high RI cell (left) and a low RI cell (right). Scale bars: \(20 \mu \mathrm{m}\) . </center>
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 4. Real-time chemical efflux monitoring using the cellular lensing effect. (a) Time-series nIR images of a stationary single monocyte with different immune activation states (-PMA, +PMA, +PMA & catalase). (b) Real-time nIR intensity variations of the cells with different activation states. (c) Schematic illustrations of \(\mathrm{H}_2\mathrm{O}_2\) efflux monitoring mechanism with nIR lensing effect. (d) 3D diffusion and reaction kinetic modeling for translation of measured nIR signals to real-time local \(\mathrm{H}_2\mathrm{O}_2\) concentration. (e) Real-time \(\mathrm{H}_2\mathrm{O}_2\) efflux profiles of each single monocyte estimated by the model. 16-color scalebars represent nIR intensity from white (16833) to dark blue (0). Scale bars: \(20\mu \mathrm{m}\) . </center>
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Figure 5. NCC for monitoring of multimodal immune response heterogeneities. a) Schematics and nIR images of NCC setup with distinct activation of human monocytes (-PMA and +PMA). (b) Automatic nIR image analysis using computational code for cell data extractions. (c) NCC cytometry plots of \(\mathrm{H}_2\mathrm{O}_2\) efflux rate \(\nu s\) biophysical parameters ((c1) size (2D projected area), (c2) eccentricity, (c3) RI) of two monocytes populations. Data are \(n_{\mathrm{cell}} = 413\) for -PMA, \(n_{\mathrm{cell}} = 414\) for +PMA from \(n = 6\) biologically independent samples. (d) NCC distribution curves of \(\mathrm{H}_2\mathrm{O}_2\) efflux rates with data from commercial assay kit. (e) NCC cytometry plots for cell biophysical parameters ((e1) eccentricity \(\nu s\) size, (e2) RI \(\nu s\) eccentricity, (e3) size \(\nu s\) RI). (f) NCC distribution curves of each biophysical parameters ((f1) size, (f2) eccentricity, (f3) RI). (g) Schematics illustrations for cell properties variations of human monocyte populations with immune activations. Scale bars: 20 \(\mu \mathrm{m}\) . </center>
+
+<--- Page Split --->
+
+## METHODS  
+
+Preparation and characterization of nanosensors. HiPco™ SWNTs purchased from Unidym were suspended with a 30- base (GT) sequence of ssDNA (Integrated DNA Technologies) in a 2:1 DNA:SWNT mass ratio in 0.1 M NaCl solution. (ATCAAGGCTCGAATTGTCCCTGA AATCT) sequence was used for random DNA and polystyrene sulfonate/bromostyrene was used for random copolymer in reference test. A typical DNA concentration was \(2\mathrm{mg / mL}\) . Samples were sonicated with a \(3\mathrm{mm}\) probe tip (Cole Parmer) for \(10\mathrm{min}\) at a power of \(10\mathrm{W}\) and \(40\%\) amplitude in an ice bath. Then samples were centrifuged twice for \(90\mathrm{min}\) (Eppendorf Centrifuge 5415D) at 16100 RCF (Relative Centrifugal Force). Afterwards, the supernatant was collected and the pellet was discarded. UV- Vis- nIR absorption spectra (Cary 5000, Agilent Technologies, Inc) were collected to verify successful suspension of nanosensor. Nanosensor concentration in the dispersion was estimated using an extinction coefficient of \(\ell_{632\mathrm{nm}} = 0.036\mathrm{mg / L})^{- 1}\) . Final concentration of SWNT/(GT)15 is from \(10\) to \(80\mathrm{mg / L}\) . \(80\mathrm{mg / L}\) concentration of nanosensor dispersion was used to all NIM experiments.  
+
+Nanosensors integration with microfluidic channel. Microfluidic channels (detail specification in Table S1) were purchased from ibidiR (μ- Slide VI 0.1, ibiTreat). \(2\mu \mathrm{L}\) of APTES (99%, Sigma Aldrich) in ethanol (1% APTES, 1% \(\mathrm{H}_2\mathrm{O}\) ) was injected to microfluidic channel with micropipetting and treated for \(3\mathrm{hr}\) . After APTES treatment, \(2\mu \mathrm{L}\) of nanosensor dispersions were injected. After overnight evaporation, SWNT/(GT)15 coated channel surfaces were rinsed with \(1\mathrm{mL}\) 1X PBS (pH 7.4, Life TechnologiesTM) twice to remove unbounded nanosensor. \(0.8\mathrm{mm}\) Silicone tubes were connected with NIM using Elbow Luer Connector Male (ibidiR). All experiments and characterization were done in triplicate with three different NIM fabrication. In- vitro \(\mathrm{H}_2\mathrm{O}_2\) detection experiments were conducted as below. SWNT/(GT)15 releases the nIR
+
+<--- Page Split --->
+
+fluorescence with visible range excitation laser (e.g. 516 nm) acting as an optical transducer for \(\mathrm{H}_2\mathrm{O}_2\) detection. Aqueous \(\mathrm{H}_2\mathrm{O}_2\) solution (30 wt\%, Sigma Aldrich) was diluted with distilled \(\mathrm{H}_2\mathrm{O}\) from \(1\mu \mathrm{M}\) to \(1\mathrm{M}\) to investigate chemical sensing performance of NIM. Diluted \(\mathrm{H}_2\mathrm{O}_2\) solutions were flowing through the NIM with syringe pump (0 - \(1\mu \mathrm{L} / \mathrm{min}\) , Harvard Apparatus) and averaged quenching signals from nanosensor array (250 X 350 \(\mu \mathrm{m}^2\) ) were recorded for 500 - 600 sec. Recorded nIR images were processed by ImageJ with Gray and Fire scales to clearly visualize the variations of nIR intensities.  
+
+Characterization and nIR measurements. Raman spectroscopy (Horiba Jobin Yvon LabRAM HR800) was used to investigate the nanosensor assembly direction in NIM with a 532 nm laser excitation (3 sec accumulations) and \(\sim 1\mu \mathrm{m}\) of spot size with 1800 lines/mm grating. The G band originating from tangential oscillations of the carbon atoms in the SWNT was observed in the frequency range of \(1590\mathrm{cm}^{- 1}\) . When \(\theta = 0^{\circ}\) and \(\theta = 90^{\circ}\) , the incident excitation polarization direction was parallel and perpendicular to the flowing direction of the microfluidic channel, respectively, indicating that the SWNT/(GT)15 nanosensors were aligned along the flowing direction of channel during EISA. AFM profiles of nanosensor array were scanned with Bruker Multimode 8 with Controller V. AFM images were taken in the ScanAsyst tapping mode in the air with TESPA probes having an elastic constant of \(42\mathrm{N / m}\) and tip radius of \(8\mathrm{nm}\) . The images were recorded with the scan rate of \(1\mathrm{Hz}\) and resolution of 1024 lines per image for each area respectively, recorded at three different places of single channel surface. Image analysis was done with Nanoscope Analysis software 1.4 from Bruker. nIR spectrum of NIM were collected with a fluorescence spectrometer equipped with a 785 nm photodiode laser (B&W Tek. Inc. 450 mW). Low- magnified nIR images were collected using a Zeiss AxioVision inverted microscope with appropriate optical filters. The fluorescence passed through an Acton SP2500 spectrometer
+
+<--- Page Split --->
+
+(Princeton Instruments), and measured with a liquid nitrogen cooled InGaAs 1D detector (Princeton Experiments). Inverted OM (Eclipse TS100, Nikon) was used for NIM and flowing cell imaging with visible light. NCC were implemented and recorded by nIR microscopy hyperspectral imager (IMA IR™, Photon Etc.). NCC was implemented with the help of a nIR microscope (IMA IR™, Photon Etc.) equipped with 561 nm laser excitation (MGL- FN- 561, Opto Engine LLC). The laser power was adjusted from 30 mW to 350 mW with optical density filters (laser power control in Figure S7). The laser was passed through a laser line filter, reflected by dichroic mirror, and focused onto the back focal plane of an inverted objective to illuminate the entire field of view of the NIM under study. nIR fluorescence from the NIM passed a longpass filter and was measured using a TE cooled infrared camera. All the measurements were conducted with 20X objective, 0.1 sec exposure time and medium intensity gain. In order to investigate the focal points and observed cell locations, motorized Z- stage controller was integrated with nIR microscopy. Hollow glass microspheres (0.6 g/cc & 5 - 30 μm, Cospheric LLC), PS microparticle (20 μm, Sigma Aldrich), and stainless steel metal microspheres (7.8 g/cc & 1 - 22 μm, Cospheric LLC) were used for reference particles as lensing effect observations.  
+
+FDTD numerical modeling. FDTD modeling for nIR photonic nanojet were performed using Lumerical FDTD Solution (Lumerical Inc). Micro- spherical structures having various range of size (radius: 1, 2, 3, 4, 5, 6, 7, and 8 μm), eccentricity (Z- axis distance: 2.5, 3, 3.5, 4, and 4.5 μm), and RI \((n_{\mathrm{c}} / n_{\mathrm{m}}; 1.01, 1.02, 1.03, 1.04, 1.05, 1.06, 1.07, 1.08, 1.09,\) and 1.10) were set and excited by an incident plane wave with a wavelength of \(1000\mathrm{nm}\) , corresponding to the fluorescence emission of the nanosensor array. The calculation domain was \(50\mathrm{X}50\mathrm{X}50\mathrm{μm}^3\) and uniform mesh of around \(30\mathrm{nm}\) was used. The perfectly matched layers (PML) were arranged around the boundaries. RI of media (out of cell) was set to 1.33.
+
+<--- Page Split --->
+
+Cell experiments. Monocytes (U937, ATCC CRL- 1593.2), B lymphocytes (FIB504.64, ATCC HB- 293), epithelial (HEK- 293, ATCC CRL- 1573), and endothelial (HUVEC, ATCC CRL- 1730) cells were purchased from American Type Culture Collection (ATCC) and cultivated according to the supplier's protocol. U937 and FIB504.64 were cultured in RPMI- 1640 (ATCC 30- 2001) with \(10\%\) of Fetal Bovine Serum (FBS) (A3160601, Gibco™). HEK- 293 cells were cultured in Dulbecco's Modified Essential Medium (DMEM; Lonza) with \(10\%\) FBS (ATCC 30- 2020). HUVEC were cultured in F- 12K medium supplemented with \(10\%\) FBS (ATCC 30- 2020), \(1\%\) endothelial cell growth factor (100X, Sigma), \(100 \mathrm{IU / mL}\) penicillin, and \(100 \mu \mathrm{g / mL}\) streptomycin. For the adherent HUVEC observations, microfluidic channels were initially coated with endothelial cell attachment factor (ECAF) to promote HUVEC cell adherence on channel surfaces. All the cells were cultured in \(75 \mathrm{cm}^2\) cell culture flasks (Falcon) under incubating conditions of \(5\%\) \(\mathrm{CO_2}\) at \(37^{\circ}\mathrm{C}\) (Forma™ series II 3110, ThermoFisher Scientific). Three days cultured U397 were used (cell number: \(10^{4} - 10^{5} / \mathrm{mL}\) , passage number \(= 4\) ) to implement NCC in this study. To monitor only the instantaneous \(\mathrm{H}_2\mathrm{O}_2\) efflux, cell media was changed by fresh PBS with \(10 \mathrm{min}\) 130 RCF centrifugation at \(10^{\circ}\mathrm{C}\) so that remove all the by- product, accumulated efflux and abnormal cells in media. \(10 \mu \mathrm{L}\) of \(0.5 \mathrm{mg / mL}\) PMA (Sigma Aldrich, for use in molecular biology applications, \(\geq 99\%\) ) was added in \(1 \mathrm{mL}\) of U937 cell media to activate the monocyte and induce differentiation into macrophage (final concentration of \(\mathrm{PMA} = 5 \mu \mathrm{g / mL}\) ). \(100 \mathrm{mL}\) of \(200 \mathrm{units / mL}\) Catalase (Sigma Aldrich, from bovine liver) was used for \(\mathrm{H}_2\mathrm{O}_2\) removal control experiments. For final NCC implementation, activated (+PMA) and non- activated monocytes (-PMA) were flowing through NIM using syringe pump (Harvard Apparatus) with flowing rate from \(0\) to \(10 \mu \mathrm{L / min}\) . PMA group was firstly injected through channel 1 with syringe pump and measured at stopped position for \(10 \mathrm{min}\) to accumulate the \(\mathrm{H}_2\mathrm{O}_2\) efflux on nanosensor array. Then, measured cells were flowed (10
+
+<--- Page Split --->
+
+\(\mu \mathrm{L} / \mathrm{min}\) ) and collected in empty tube for the future experiments. Lastly, \(+\mathrm{PMA}\) group was injected to channel 2 and \(\mathrm{H}_2\mathrm{O}_2\) efflux was measured. NCC were conducted for stationary cells for few min and videos were recorded to analysis efflux signals of the cells. Attomolar efflux rates were calculated from real- time local \(\mathrm{H}_2\mathrm{O}_2\) concentration multiplied with single unit volume (single monocyte volume \(= 4.18\cdot 10^{- 15} \mathrm{m}^3\) ) and divided by measurement time (10 min). Six biological replicates of U937 populations were used for NCC cytometry plot data.  
+
+Data analysis. nIR image analysis and quantitation was performed in MATLAB (Natick, MA) with the steps detailed below. Cell identification is performed by taking 1 frame of the nIR video (500 sec recorded, 0.1 sec of exposure time, 5000 frames) convolving with a Laplacian of a Gaussian filter, and then thresholded by the user for each experiment batch. For each cell, the image is then interpolated. Using the peak and Airy ring of the nIR lensing spot, the cell image is normalized, and then statistics such as cell size and eccentricity are evaluated with "Regionprops" function. The projected area (i.e. size) values are dilated appropriately to coincide with the photonic nanojet model. To avoid excess data interpolation, camera pixel intensities are used for subsequent analysis. The cellular lensing intensity \((I_0)\) is found by choosing the camera pixel closest to the centroid. To calculate background, 16 pixels outside of the secondary peak of the lensing effect is chosen. Outliers are then removed, and background traces are averaged to use as normalization for the centroid intensity traces.  
+
+## ACKNOWLEDGEMENTS  
+
+The authors are grateful for financial support from Bose Fellowship Award to M. S. S., the Juvenile Diabetes Research Foundation (JDRF), funding from the Disruptive & Sustainable Technology for Agricultural Precision (DiSTAP) and the Singapore MIT Alliance for Research and
+
+<--- Page Split --->
+
+Technology (SMART) Center, and the Walmart Foundation and the Walmart Food Safety Collaboration Center in Beijing. T. T. S. L. acknowledges a graduate fellowship by the Agency of Science, Research and Technology, Singapore. M. K. acknowledges support by the German Research Foundation (DFG) Research Fellowship KU 3952/1- 1. We give thanks to the Nanotechnology Materials Lab, and the Koch Institute for Integrative Cancer Research, MIT for AFM measurements.  
+
+## AUTHOR CONTRIBUTIONS  
+
+S.- Y. C and M. S. S. conceived the idea, designed the project and planned experiments with the assistance of X. G., V. B. K., M. K., S. J. M., M. S. and T. T. S. L.. S.- Y. C prepared the nanosensors, fabricated the NIM, implemented the NCC with cell culturing, measured and analyzed the data. X. G. coded automatic image analysis program and conducted cell data processing. V. B. K. conducted FDTD numerical modeling for photonic nanojet demonstration. M. K. assisted nIR observations of NIM. S. J. M. conducted HEK cell culturing and commercial \(\mathrm{H}_2\mathrm{O}_2\) assay. M. S. assisted with polarized Raman spectroscopy and nanosensor synthesis. T. T. S. L. commented about \(\mathrm{H}_2\mathrm{O}_2\) nanosensor and their detection mechanism. X. J. assisted with preparation of various nanosensor. P. G. assisted with AFM characterization of nanosensor array. S.- Y. C and M. S. S. wrote the manuscript with inputs from all the authors. All authors contributed to discussions informing the research.  
+
+## COMPETING INTERESTS  
+
+The Authors declare no competing financial interest
+
+<--- Page Split --->
+
+## DATA AVAILABILITY  
+
+The authors declare that all data supporting the findings of this study are available within the paper and any raw data can be obtained from the corresponding author on request.  
+
+## REFERENCES  
+
+(1) Arroyo, J. O., & Kukura, P. Non-fluorescent schemes for single-molecule detection, imaging and spectroscopy. Nat. Photonics. 10, 11-17 (2016).  
+(2) Holzmeister, P. et al. Breaking the concentration limit of optical single-molecule detection. Chem. Soc. Rev. 43, 1014-1028 (2014).  
+(3) Rigler, R. & Vogel, H. Single Molecules and Nanotechnology (Springer, 2008).  
+(4) Garoli, D. Plasmonic Nanopores for Single-molecule detection and manipulation: Toward sequencing applications. Nano Lett. 19, 7553-7562 (2019).  
+(5) Ameur, A., Kloosterman, W. P., & Hestand, M. S. Single-molecule sequencing: Towards clinical applications. Trends Biotechnol. 37, 72-85 (2019).  
+(6) Kim, J.-H. et al. Single-molecule detection of \(\mathrm{H}_2\mathrm{O}_2\) mediating angiogenic redox signaling on fluorescent single-walled carbon nanotube array. ACS Nano. 5, 7848-7857 (2010).  
+(7) Heller, D. A. et al. Multimodal optical sensing and analyte specificity using single-walled carbon nanotubes. Nat. Nanotechnol. 4, 114-120 (2009).  
+(8) Kim, J.-H. et al. The rational design of nitric oxide selectivity in single-walled carbon nanotube near-infrared fluorescence sensors for biological detection. Nat. Chem. 1, 473-481 (2009).  
+(9) Jin, H. et al. Detection of single-molecule \(\mathrm{H}_2\mathrm{O}_2\) signalling from epidermal growth factor receptor using fluorescent single-walled carbon nanotubes. Nat. Nanotechnol. 5, 302-309 (2010).  
+(10) Oh, S., Lee, M.-K., & Chi, S.-W. Single-molecule-based detection of conserved Influenza A virus RNA promoter using a protein nanopore. ACS Sens. 4, 2849-2853 (2019).  
+(11) Landry, M. P. et al. Single-molecule detection of protein efflux from microorganisms using fluorescent single-walled carbon nanotube sensor arrays. Nat. Nanotechnol. 12, 368-377 (2017).  
+(12) Chan, M. W. Y., & Viswanathan, S. Recent progress on developing exogenous monocyte/macrophage-based therapies for inflammatory and degenerative diseases. Cytotherapy 21, 393-415 (2019).
+
+<--- Page Split --->
+
+(13) Brudno, J. N., & Kochenderfer, J. N. Chimeric antigen receptor T-cell therapies for lymphoma. Nat. Rev. Clin. Oncol. 15, 31-46 (2018).  
+(14) Boese, A. C. et al. Neural stem cell therapy for subacute and chronic ischemic stroke. Stem Cell. Res. Ther. 9, 154 (2018).  
+(15) Aijaz, A. et al. Biomanufacturing for clinically advanced cell therapies. Nat. Biomed. Eng. 2, 362-376 (2018).  
+(16) Adan, A. et al. Flow cytometry: basic principles and applications. Crit. Rev. Biotechnol. 37, 163-176 (2017).  
+(17) Kozawa, D. et al. A fiber optic interface coupled to nanosensors: applications to protein aggregation and organic molecule quantification. ACS Nano 14, 10141-10152 (2020).  
+(18) Ugele, M. et al. Label-free high-throughput leukemia detection by holographic microscopy. Adv. Sci. 5, 1800761 (2018).  
+(19) Singh, D. K. et al. Label-free, high-throughput holographic screening and enumeration of tumor cells in blood. Lab Chip. 17, 2920 (2017).  
+(20) Zhang, Y. et al. Motility-based label-free detection of parasites in bodily fluids using holographic speckle analysis and deep learning. Light Sci. Appl. 7, 108 (2018).  
+(21) Kim, T. et al. White-light diffraction tomography of unlabelled live cells. Nat. Photon. 8, 256-263 (2014).  
+(22) Sung, Y. et al. Optical diffraction tomography for high resolution live cell imaging. Opt. Express. 17, 266-277 (2009).  
+(23) Yoon, J. et al. Identification of non-activated lymphocytes using three-dimensional refractive index tomography and machine learning. Sci. Rep. 7, 6654 (2017).  
+(24) Lei, C. et al. High-throughput imaging flow cytometry by optofluidic time-stretch microscopy. Nat. Protoc. 13, 1603-1631 (2018).  
+(25) Suzuki, Y. et al. Label-free chemical imaging flow cytometry by high-speed multicolor stimulated Raman scattering. Proc. Natl. Acad. Sci. U. S. A. 116, 15842-15848 (2019).  
+(26) Vembadi, A., Menachery, A., & Qasaimeh, M. A. Cell cytometry: Review and perspective on biotechnological advances. Front. Bioeng. Biotech. 7, 147 (2019).  
+(27) Dickinson, A. J., Armistead, P. M, & Allbritton, N. L. Automated capillary electrophoresis system for fast single-cell analysis. Anal. Chem. 85, 4797-4804 (2013).
+
+<--- Page Split --->
+
+(28) Abraham, D.H. et al. Design of an automated capillary electrophoresis platform for single-cell analysis. \*Method. Enzymol.\* 628, 191-221 (2019).  
+(29) Vickerman, B. M. et al. Design and application of sensors for chemical cytometry. \*ACS Chem. Biol.\* 13, 1741-1751 (2018).  
+(30) Cohen, D. et al. Chemical cytometry: Fluorescence-based single-cell analysis. \*Annu. Rev. Anal. Chem.\* 1, 165-190 (2008).  
+(31) Oomen, P. E. et al. Chemical analysis of single cells. \*Anal. Chem.\* 91, 588-621 (2019).  
+(32) Munawar, A. et al. Nanosensors for diagnosis with optical, electric and mechanical transducers. \*RSC Adv.\* 9, 6793 (2019).  
+(33) Abdel-Karim, R., Reda, Y., & Abdel-Fattah, A. Review-nanostructured materials-based nanosensors. \*J. Electrochem. Soc.\* 167, 037554 (2020).  
+(34) Liu, J. et al. A sensitive and specific nanosensor for monitoring extracellular potassium levels in the brain. \*Nat. Nanotechnol.\* 15, 321-330 (2020).  
+(35) Oh, W.-J. et al. Fluorescent polymer nanoparticle for selective sensing of intracellular hydrogen peroxide. \*ACS Nano\* 6, 8516-8524 (2012).  
+(36) Liu, J. et al. A highly sensitive and selective nanosensor for near-infrared potassium imaging. \*Sci. Adv.\* 6, eaax9757 (2020).  
+(37) Reuel, N. F. et al. Emergent properties of nanosensor arrays: Applications for monitoring IgG affinity distributions, weakly Affined hypermannosylation, and colony selection for biomanufacturing. \*ACS Nano\* 7, 7472-7482 (2013).  
+(38) Kruss, S. et al. High-resolution imaging of cellular dopamine efflux using a fluorescent nanosensor array. \*Proc. Natl. Acad. Sci. U.S.A.\* 114, 1789-1794 (2017).  
+(39) Jeong, S. et al. High-throughput evolution of near-infrared serotonin nanosensors. \*Sci. Adv.\* 5, eaay3771 (2019).  
+(40) Farrera, C. et al. Carbon nanotubes as optical sensors in biomedicine. \*ACS Nano\* 11, 10637-10643 (2017).  
+(41) Iverson, N. M. et al. In vivo biosensing via tissue-localizable near-infrared-fluorescent single-walled carbon nanotubes. \*Nat. Nanotechnol.\* 8, 873-880 (2013).  
+(42) Lew, T. T. S. et al. Real-time detection of wound-induced \(\mathrm{H}_2\mathrm{O}_2\) signaling waves in plants with optical nanosensors. \*Nat. Plants\* 6, 404-415 (2020).
+
+<--- Page Split --->
+
+(43) Engel, M. et al. Thin film nanotube transistors based on self-assembled, aligned, semiconducting carbon nanotube arrays. ACS Nano 2, 2445-2452 (2008).(44) Cho, S.-Y. et al. Continuous meter-scale synthesis of wearable tunicate cellulose/carbon nanotube fibers for high-performance wearable sensors. ACS Nano 13, 9332-9341 (2019).(45) Lee, M. A. et al. Implanted nanosensors in marine organisms for physiological biologging: Design, feasibility, and species variability. ACS Sens. 4, 32-43 (2019).(46) Zhang, J. et al. Molecular recognition using corona phase complexes made of synthetic polymers adsorbed on carbon nanotubes. Nat. Nanotechnol. 8, 959-968 (2013).(47) Dukovic, G. et al. Reversible surface oxidation and efficient luminescence quenching in semiconductor single-wall carbon nanotubes. J. Am. Chem. Soc. 126, 15269-15276 (2004).(48) Kruss, S. et al. Neurotransmitter detection using corona phase molecular recognition on fluorescent single-walled carbon nanotube sensors. J. Am. Chem. Soc. 136, 713-724 (2014).(49) Yamamoto, T. et al. Role of catalase in monocytic differentiation of U937 cells by TPA: hydrogen peroxide as a second messenger. Leukemia 23, 761-769 (2009).(50) Chateau, M. T., Rabesandratana, H. & Caravano, R. Differentiated U937 cells and human monocytes exhibit a differential production of extracellular oxygen species: \(\mathrm{O}_2\) excretion versus \(\mathrm{H}_2\mathrm{O}_2\) diffusion. FEMS Immunol. Med. Microbiol. 13, 19-28 (1996).(51) Varin, A. & Gordon, S. Alternative activation of macrophages: Immune function and cellular biology. Immunobiology 214, 630-641 (2009).(52) Satija, R. & Shalek, A. K. Heterogeneity in immune responses - From populations to single cells. Trends. Immunol. 35, 219-229 (2014).(53) Yang, H. & Gijs, M. A. M. Optical microscopy using a glass microsphere for metrology of sub-wavelength nanostructures. Microelectron. Eng. 143, 86-90 (2015).(54) Yang, H., Cornaglia, M. & Gijs, M. A. M. Photonic nanojet array for fast detection of single nanoparticles in a flow. Nano Lett. 15, 1730-1735 (2015).(55) Lee, S., Li, L. & Wang, Z. Optical resonances in microsphere photonic nanojets. J. Opt. 16, 015704 (2014).(56) Heifetz, A. et al. Photonic nanojets. J. Comput. Theor. Nanosci. 6, 1979-1992 (2009).(57) Dannhauser, D. et al. Biophysical investigation of living monocytes in flow by collaborative coherent imaging techniques. Biomed. Opt. Exp. 9, 5194-5204 (2018).
+
+<--- Page Split --->
+
+(58) Liu, P. Y. et al. Cell refractive index for cell biology and disease diagnosis: past, present and future. Lab Chip. 16, 634-644 (2016).(59) Miccio1, L. et al. Red blood cell as an adaptive optofluidic microlens. Nat. Commun. 6, 6502 (2015).(60) Li, Y. et al. Living nanospear for near-field optical probing. ACS Nano 12, 10703-10711 (2018).(61) Li, Y., Liu, X. & Li, B. Single-cell biomagnifier for optical nanoscopes and nanotweezers. Light-Sci. Appl. 8, 61 (2019).(62) Xin, H. et al. Escherichia coli-based biophotonic waveguides. Nano Lett. 13, 3408-3413 (2013).(63) Strokotov, D. I. et al. Is there a difference between T- and B-lymphocyte morphology?. J. Biomed. Opt. 146, 064036 (2009).(64) Neto, J. C. et al. Measuring optical and mechanical properties of a living cell with defocusing microscopy. Biophys. J. 91, 1108-1115 (2006).(65) Ling, T. et al. Full-field interferometric imaging of propagating action potentials. Light-Sci. Appl. 7, 107 (2018).(66) King, D. et al. Label-free multi parameter optical interrogation of endothelial activation in single cells using a lab on a disc platform. Sci. Rep. 9, 4157 (2019).(67) LeBien, T. W. & Tedder, T. F. B lymphocytes: how they develop and function. Blood 112, 1570-4580 (2008).(68) Prasad, A. et al. Simultaneous real-time monitoring of oxygen consumption and hydrogen peroxide production in cells using our newly developed chip-type biosensor device. Front. Physiol. 7, 109 (2016).(69) Giorgio, M. et al. Hydrogen peroxide: a metabolic by-product or a common mediator of ageing signals?. Nat. Rev. Mol. Cell Biol. 8, 722-728 (2007).(70) Magnani, F. & Mattevi, A. Structure and mechanisms of ROS generation by NADPH oxidases. Curr. Opin. Struc. Biol. 59, 91-97 (2019).(71) George, P. Reaction between catalase and hydrogen peroxide. Nature 160, 41-43 (1947).(72) Shen, R. et al. Sensitive detection of single-cell Secreted \(\mathrm{H}_2\mathrm{O}_2\) by integrating a microfluidic droplet sensor and Au nanoclusters. Anal. Chem. 90, 4478-4484 (2018).
+
+<--- Page Split --->
+
+(73) Tjell, A. Ø. & Almdal, K. Diffusion rate of hydrogen peroxide through water-swelled polyurethane Membranes. Sens. Biosensing. Res. 21, 35-39 (2018).  
+(74) D'Autréaux, B & Toledano, M. B. ROS as signalling molecules: Mechanisms that generate specificity in ROS Homeostasis. Nat. Rev. Mol. Cell. Biol. 8, 813-824 (2007).  
+(75) Ulissi, Z. W. et al. Spatiotemporal intracellular nitric oxide signaling captured using internalized, near-infrared fluorescent carbon nanotube nanosensors. Nano Lett. 14, 4887-4894 (2014).  
+(76) Wang, S.-Y. et al. Heterogeneity of human blood monocyte: two subpopulations with different sizes, phenotypes and functions. Immunology. 77, 298-303 (1992).  
+(77) Chakraborty, P., Ghosh, D. & Basu, M. K. Macrophage protein kinase C: its role in modulating membrane microviscosity and superoxide in leishmanial infection. J. Biochem. 127, 185-190 (2000).  
+(78) Zhang, C. et al. Regulation of basal lateral membrane mobility and permeability to divalent cations by membrane associated-protein kinase C. PLOS One 8, e80291 (2013).  
+(79) Chen, L. et al. A role for protein kinase C in the regulation of membrane fluidity and Ca²⁺(+) flux at the endoplasmic reticulum and plasma membranes of HEK293 and Jurkat cells. Cell. Signalling. 23, 497-505 (2011).  
+(80) Saha, A. K. et al. Cellular cholesterol regulates monocyte deformation. J. Biomech. 52, 83-88 (2017).  
+(81) Boss, D. et al. Measurement of absolute cell volume, osmotic membrane water permeability, and refractive index of transmembrane water and solute flux by digital holographic microscopy. J. Biomed. Opt. 18, 036007 (2013).
+
+<--- Page Split --->
+
+## Figures  
+
+![](images/Figure_1.jpg)
+
+<center>Figure 1 </center>  
+
+Nanosensor integration with microfluidics. (a) Schematic illustration of nanosensor integration process with microfluidics using EISA. (b) Photograph of EISA process of NIM for 0 min (left) and 30 min (right). (c) Photograph of completed multi- array NIM and pristine channel. (d) Polarized Raman spectrum (G- peak) of NIM. (e) nIR images of NIM and pristine channel. (f) Magnified nIR image of NIM with single cell size resolution (20 μm) having \(\sim 720\) nIR reporter pixel. (g) Histograms of nIR pixel intensities of top and bottom NIM surfaces NIM (inset: nIR images of top and bottom surfaces). (h) nIR fluorescence spectrum of NIM. (i) nIR images of NIM with varying composition of SWNT nanosensor integration.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Figure 2 </center>  
+
+In- vitro chemical detection performances of NIM. (a) nIR spectrum of NIM with H2O2 solution flowing (1 \(\mu \mathrm{M}\) , 1 \(\mu \mathrm{L} / \mathrm{min}\) ). (b) nIR images of NIM before and after H2O2 flowing (1 M, 10 \(\mu \mathrm{L} / \mathrm{min}\) , 10 min). (c) Schematic illustration of H2O2 detection mechanism of SWNT/(GT)15 nanosensor. (d) Real- time nIR response of NIM with various concentration (10- 6, 10- 5, 10- 4, 10- 3, 10- 2, 10- 1, 100 M) of H2O2 injection (10 min). (e) Maximum response amplitude and (f) response time of NIM with various concentration of H2O2. The data represent the mean value of 250 X 350 \(\mu \mathrm{m}2\) NIM measurement. (g) nIR snapshots and intensity histogram (fire scale, ImageJ) of NIM with single cell size resolution (20 \(\mu \mathrm{m}\) ) after 10 min flowing of various concentration of H2O2.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3 </center>  
+
+Cellular lensing effect. (a) Instrumental setup for NCC implementation: schematic illustration (left) and a photograph (right). (b) nIR images of human monocytes flowing (0.5 \(\mu \mathrm{L} / \mathrm{min}\) ) NIM. (c) Magnified nIR image of single monocyte in NIM (inset: OM image of single monocyte). (d) FDTD numerical modeling for photonic nanojet and fitting with experimental cellular lensing profile (nc/nm = 1.04, \(\lambda = 1 \mu \mathrm{m}\) ). (e) nIR lensing profiles of a single cell with various focusing points from 5 to 100 \(\mu \mathrm{m}\) along Z- stage. nIR lensing
+
+<--- Page Split --->
+
+effects of (f) various live cells and (g) reference micro-particles (top-to-bottom: schematics, OM, nIR images, lensing profiles). (h) FWHM and (i) enhancement factors of various cells with numerical model. Data are mean \(\pm \sigma\) , with \(\text{neel} = 10\) . (j) Schematic illustrations for different lensing behavior of a high RI cell (left) and a low RI cell (right). Scale bars: \(20 \mu \text{m}\) .  
+
+![](images/Figure_4.jpg)
+
+<center>Figure 4 </center>  
+
+Real- time chemical efflux monitoring using the cellular lensing effect. (a) Time- series nIR images of a stationary single monocyte with different immune activation states (- PMA, + PMA, + PMA & catalase). (b) Real- time nIR intensity variations of the cells with different activation states. (c) Schematic illustrations of H2O2 efflux monitoring mechanism with nIR lensing effect. (d) 3D diffusion and reaction kinetic
+
+<--- Page Split --->
+
+modeling for translation of measured nIR signals to real- time local H2O2 concentration. (e) Real- time H2O2 efflux profiles of each single monocyte estimated by the model. 16- color scalebars represent nIR intensity from white (16833) to dark blue (0). Scale bars: \(20 \mu \mathrm{m}\) .  
+
+![](images/Figure_5.jpg)
+
+<center>Figure 5 </center>  
+
+NCC for monitoring of multimodal immune response heterogeneities. a) Schematics and nIR images of NCC setup with distinct activation of human monocytes (-PMA and +PMA). (b) Automatic nIR image analysis using computational code for cell data extractions. (c) NCC cytometry plots of H2O2 efflux rate
+
+<--- Page Split --->
+
+vs biophysical parameters ((c1) size (2D projected area), (c2) eccentricity, (c3) RI) of two monocytes populations. Data are ncell = 413 for - PMA, ncell = 414 for +PMA from n = 6 biologically independent samples. (d) NCC distribution curves of H2O2 efflux rates with data from commercial assay kit. (e) NCC cytometry plots for cell biophysical parameters ((e1) eccentricity vs size, (e2) RI vs eccentricity, (e3) size vs RI). (f) NCC distribution curves of each biophysical parameters ((f1) size, (f2) eccentricity, (f3) RI). (g) Schematics illustrations for cell properties variations of human monocyte populations with immune activations. Scale bars: \(20 \mu \mathrm{m}\) .  
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- VideoS1.avi- RevisedSupportingInformationMichaelStranoNatureNanotechnology.docx
+
+<--- Page Split --->

preprint/preprint__487268b60d60a57ddf72c596fba8ead5ff013eeab984aee8c25ad6580afe57e1/preprint__487268b60d60a57ddf72c596fba8ead5ff013eeab984aee8c25ad6580afe57e1_det.mmd

@@ -0,0 +1,469 @@
+<|ref|>title<|/ref|><|det|>[[44, 107, 860, 210]]<|/det|>
+# Cellular Lensing and Near Infrared Fluorescent Nanosensor Arrays to Enable Chemical Efflux Cytometry  
+
+<|ref|>text<|/ref|><|det|>[[44, 230, 448, 273]]<|/det|>
+Soo- Yeon Cho Massachusetts Institute of Technology  
+
+<|ref|>text<|/ref|><|det|>[[44, 278, 397, 319]]<|/det|>
+Xun Gong Massachusetts Institute of Technology  
+
+<|ref|>text<|/ref|><|det|>[[44, 325, 450, 365]]<|/det|>
+Volodymyr Koman MIT https://orcid.org/0000- 0001- 8480- 4003  
+
+<|ref|>text<|/ref|><|det|>[[44, 371, 752, 412]]<|/det|>
+Matthias Kuehne Massachusetts Institute of Technology https://orcid.org/0000- 0002- 5096- 7522  
+
+<|ref|>text<|/ref|><|det|>[[44, 417, 397, 457]]<|/det|>
+Sun Jin Moon Massachusetts Institute of Technology  
+
+<|ref|>text<|/ref|><|det|>[[44, 463, 397, 503]]<|/det|>
+Manki Son Massachusetts Institute of Technology  
+
+<|ref|>text<|/ref|><|det|>[[44, 509, 820, 550]]<|/det|>
+Tedrick Thomas Salim Lew Massachusetts Institute of Technology (M.I.T.) https://orcid.org/0000- 0002- 4815- 9921  
+
+<|ref|>text<|/ref|><|det|>[[44, 555, 397, 595]]<|/det|>
+Pavlo Gordiichuk Massachusetts Institute of Technology  
+
+<|ref|>text<|/ref|><|det|>[[44, 601, 450, 641]]<|/det|>
+Xiaojia Jin MIT https://orcid.org/0000- 0002- 0694- 5799  
+
+<|ref|>text<|/ref|><|det|>[[44, 648, 397, 688]]<|/det|>
+Hadley Sikes Massachusetts Institute of Technology  
+
+<|ref|>text<|/ref|><|det|>[[44, 694, 752, 735]]<|/det|>
+Michael Strano ( \(\square\) strano@mit.edu) Massachusetts Institute of Technology https://orcid.org/0000- 0003- 2944- 808X  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 777, 102, 794]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 815, 784, 835]]<|/det|>
+Keywords: cellular lensing, chemical efflux cytometry, nanosensor chemical cytometry  
+
+<|ref|>text<|/ref|><|det|>[[44, 853, 346, 872]]<|/det|>
+Posted Date: December 11th, 2020  
+
+<|ref|>text<|/ref|><|det|>[[44, 891, 463, 910]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 106483/v1
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 44, 909, 87]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 123, 909, 167]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on May 25th, 2021. See the published version at https://doi.org/10.1038/s41467-021-23416-1.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[112, 90, 884, 255]]<|/det|>
+Cellular Lensing and Near Infrared Fluorescent Nanosensor Arrays to Enable Chemical Efflux Cytometry  
+
+<|ref|>text<|/ref|><|det|>[[112, 303, 884, 430]]<|/det|>
+Soo- Yeon Cho, Xun Gong, Volodymyr B. Koman, Matthias Kuehne, Sun Jin Moon, Manki Son, Tedrick Thomas Salim Lew, Pavlo Gordichuk, Xiaojia Jin, Hadley D. Sikes, Michael S. Strano\* Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, United States  
+
+<|ref|>text<|/ref|><|det|>[[114, 444, 435, 463]]<|/det|>
+\*Corresponding author: strano@mit.edu
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 90, 242, 111]]<|/det|>
+## ABSTRACT  
+
+<|ref|>text<|/ref|><|det|>[[112, 123, 886, 884]]<|/det|>
+Nanosensor have proven to be powerful tools to monitor single biological cells and organisms, achieving spatial and temporal precision even at the single molecule level. However, there has not been a way of extending this approach to statistically relevant numbers of living cells and organisms. Herein, we design and fabricate a high throughput nanosensor array in a microfluidic channel that addresses this limitation, creating a Nanosensor Chemical Cytometry (NCC). An array of nIR fluorescent single walled carbon nanotube (SWNT) nanosensors is integrated along a microfluidic channel through which a population of flowing cells is guided. We show that one can utilize the flowing cell itself as highly informative Gaussian lenses projecting nIR emission profiles and extract rich information on a per cell basis at high throughput. This unique biophotonic waveguide allows for quantified cross- correlation of the biomolecular information with physical properties such as cellular diameter, refractive index (RI), and eccentricity and creates a label- free chemical cytometer for the measurement of cellular heterogeneity with unprecedented precision. As an example, the NCC can profile the immune response heterogeneities of distinct human monocyte populations at attomolar ( \(10^{- 18}\) moles) sensitivity in a completely non- destructive and real- time manner with a rate of \(\sim 100\) cells/frame, highest range demonstrated to date for state of the art chemical cytometry. We demonstrate distinct \(\mathrm{H}_2\mathrm{O}_2\) efflux heterogeneities between 330 and 624 attomole/cell·min with cell projected areas between 271 and \(263~\mu \mathrm{m}^2\) , eccentricity values between 0.405 and 0.363 and RI values between 1.383 and 1.377 for non- activated and activated human monocytes, respectively. Hence, we show that our nanotechnology based biophotonic cytometer has significant potential and versatility to answer important questions and provide new insight in immunology, cell manufacturing and biopharmaceutical research.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 85, 886, 635]]<|/det|>
+Nanotechnology has produced some of the most sensitive analytical platforms for molecules in existence, with many achieving single molecule resolution, \(^{1 - 3}\) including arrays for DNA sequencing \(^{4,5}\) as well as reactive oxygen species (ROS) detection. \(^{6,7}\) There is significant interest and motivation to extend such platforms to the study of living cells \(^{8,9}\) and microbes \(^{10,11}\) where they can form the basis of non- destructive techniques to probe various biochemical mechanisms. This has obvious applications to medicine and life science research, and of particular importance to the emerging area of cell- based therapies and regenerative medicine for the treatment of cancer, leukemia, and neurodegenerative diseases. \(^{12 - 14}\) However, cellular populations are necessarily heterogeneous, and cellular therapies necessarily require characterization methods that are non- destructive and do not contaminate the cells themselves, \(^{15}\) ruling out conventional flow cytometry that requires fluorescent labels. \(^{16}\) Extending various types of nanosensor to statistically relevant numbers of living cells and organisms in non- destructive manner remains unaddressed to date with the basic problem of nanosensor including interfacing strategy, signal transducing mechanism, and mechanical robustness. \(^{17}\) In this work, we introduce a potential solution in the form of Nanosensor Chemical Cytometry (NCC), a technique that leverages cellular lensing, producing multivariate real- time single cell analysis.  
+
+<|ref|>text<|/ref|><|det|>[[113, 644, 886, 876]]<|/det|>
+Various label- free cell imaging techniques such as digital holographic microscopy (DHM) \(^{18 - 20}\) or optical diffraction tomography \(^{21 - 23}\) have been developed for high- throughput cell classification based on image analysis. For example, Ugele et al. discriminated healthy and pathological blood cells using holographic speckle images of DHM technique. \(^{18}\) Singh et al. used machine learning based hologram screening to detect tumor cells in high- throughput. \(^{19}\) However, these techniques are based on physical property measurements from cell images. Chemically quantification for heterogeneity in cell populations is still an open problem. Flow and chemical
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 80, 886, 633]]<|/det|>
+cytometry have been widely used to quantify the molecular heterogeneities of target cell populations. While typical flow and image cytometry of living cells can sample \(10^{6} - 10^{7}\) cells in just a few minutes, \(^{24 - 26}\) the state of the art for the emerging field of chemical cytometry is between 50 to 500 cells/hr since cells need to be pre- labelled, lysed, and separated to be detected. \(^{27 - 29}\) Nevertheless, this level of throughput has elevated chemical cytometry as a valuable cell characterization tool allowing quantitative information to be gathered with high selectivity and signal- to- noise ratio. \(^{30,31}\) Nanosensors have significant potential to greatly expand the number of variables measured in chemical cytometry given the large number of new types being demonstrated in the recent literature. \(^{32 - 36}\) Organic and inorganic fluorescent nanoparticles have been used to monitor intra- and extracellular information of single cells successfully. \(^{34 - 36}\) Near infrared (nIR) fluorescent single walled carbon nanotubes (SWNT) are particularly promising components toward label- free and single molecule level cellular profiling. To date, they have been developed for the detection of single cell biochemical efflux for antibodies, neurotransmitters and ROS. \(^{9,37 - 40}\) Additionally, their rapid and direct optical readout is ideal for sensor interfacing, and carbon in particular possesses photostability, biocompatibility and tunable chemical selectivity for this purpose. \(^{17,40 - 42}\)  
+
+<|ref|>text<|/ref|><|det|>[[113, 645, 886, 876]]<|/det|>
+In this work, we develop a new class of chemical cytometry that can characterize the real- time chemical efflux of cell populations at high throughput. nIR fluorescent SWNT nanosensors were uniformly integrated within a cell- transporting microfluidic channel. Each single cell optically interacts with the underlying nanosensor array, producing an informative nIR optical lensing profile that can be modeled as a photonic nanojet. Within this biophotonic waveguide, cells can be both visualized and chemically tracked in real- time and at high- resolution, without the need for labeling or additional optical manipulation. Based on the combination of nanosensor response
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 884, 250]]<|/det|>
+and observed cellular lensing properties, the NCC platform is able to yield multivariate data that inform the heterogeneities of human monocyte populations (immune activated and non- activated) at the attomolar ( \(10^{- 18}\) moles) level of \(\mathrm{H}_2\mathrm{O}_2\) efflux. Furthermore, this type of cellular population data allows for phenotypic correlation between real- time chemical efflux and various biophysical properties of each individual cell including diameter, eccentricity, and refractive index (RI).  
+
+<|ref|>sub_title<|/ref|><|det|>[[114, 299, 475, 318]]<|/det|>
+## Nanosensor Integration with Microfluidics  
+
+<|ref|>text<|/ref|><|det|>[[112, 330, 886, 880]]<|/det|>
+The schematic of the flow channel and nanosensor array integration for NCC are shown in Figure 1a. The array is demonstrated using a \((\mathrm{GT})_{15}\) DNA wrapped SWNT (SWNT/(GT) \(_{15}\) ), which was previously shown to exhibit nIR intensity attenuation upon selective detection of \(\mathrm{H}_2\mathrm{O}_2\) . \(^{7,42}\) \(\mathrm{H}_2\mathrm{O}_2\) efflux was targeted for the application due to its central role in cellular signaling and immune responses. \(^{6,9}\) For the first step, micro- droplet of (3- aminopropyl) triethoxysilane (APTES) was injected into a pristine channel and incubated. A commercial microfluidic channel was coated with APTES for self- assembled monolayer formation and SWNT/(GT) \(_{15}\) adhesion on both top and bottom surface of the channel. Subsequently, the channel was washed with phosphate buffer saline (PBS) and a micro- droplet of SWNT/(GT) \(_{15}\) dispersion was injected into the channel. Stable dispersions of nanosensors were confirmed via UV- vis- nIR absorption spectra of SWNT/(GT) \(_{15}\) (Figure S1). During evaporation, nanosensor particles necessarily align at the three- phase line of the micro- droplet pinned at the end of the flow channel (Figure 1b). This resulted in a uniform array on both top and bottom surfaces of the channel following the Evaporation Induced Self- Assembly (EISA). \(^{43}\) After the EISA, the channel was flushed with PBS again to remove unbounded residual nanoparticles. Completed Nanosensor Integrated Microfluidics (NIMs) were highly transparent to visible light indicating an absence of aggregation
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 80, 886, 880]]<|/det|>
+or large array defects (Figure 1c). Polarized Raman spectroscopy of NIM showed the depolarization ratio of 0.61 from G band intensity demonstrating that the nanosensors were aligned along the flowing direction of channel during EISA (Figure 1d).<sup>44</sup> nIR imaging was used to investigate the fluorescence signal mapping of the NIM (Figure 1e). While we find NIMs to display strong nIR fluorescence, uncoated channels show no nIR signal (Figure S2). In addition, NIM without APTES treatment showed severe nanosensor aggregation during EISA process and consequently nanosensors were completely removed with PBS flowing, indicating that surface chemistry of the microfluidic channel is critically important to uniform and stable EISA process. Magnified nIR image of NIM with single cell size (20 \(\mu \mathrm{m}\) diameter) shows that nanosensors are homogeneously and continuously deposited with approximately 720 local detector pixels across a single cell (Figure 1f and S3). Atomic force microscopy (AFM) demonstrated that nanosensor bundles were densely and homogenously covered on the channel surface at the micron-scale (Figure S4). Consequently, the nanosensor array on the microfluidic channel could clearly visualize the cells flowing through the channel and maximize the signal- to- noise ratio of the signal from cell efflux for NCC.<sup>38</sup> As the concentration of nanosensor dispersion increases, uniformity of nanosensor array was enhanced with significant decrease of voids and aggregation of nanosensors and 80 mg/L coating shows highest nIR intensity with most uniform pixel distributions (Figure S5). Nanosensors were uniformly coated on the top and bottom surfaces of the channel during EISA, as shown by the comparable nIR pixel distributions along both surfaces. (Figure 1g). Peak position and relative peak intensities of nIR spectrum of NIM were almost identical with SWNT in dispersion phase, indicating that the dielectric environment surrounding the immobilized nanosensors were similar (Figure 1h).<sup>45</sup> Varying compositions of SWNT nanosensors ((GT)<sub>15</sub> DNA, (AT)<sub>15</sub> DNA, (ATT)<sub>10</sub> DNA, random DNA, chitosan) were integrated
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 88, 884, 144]]<|/det|>
+with the microfluidic channels by our EISA based NIM fabrication process for monitoring of various chemical components of the cell (Figure 1i).<sup>46</sup>  
+
+<|ref|>sub_title<|/ref|><|det|>[[114, 193, 470, 212]]<|/det|>
+## Chemical Detection Performances of NIM  
+
+<|ref|>text<|/ref|><|det|>[[112, 227, 886, 666]]<|/det|>
+In- vitro \(\mathrm{H}_2\mathrm{O}_2\) detection experiments were conducted to investigate the chemical sensing performance of the NIM. The fluorescence intensity from all SWNT chiralities decreased with 10 to \(20\%\) relative magnitude upon exposure to \(1\mu \mathrm{M} \mathrm{H}_2\mathrm{O}_2\) (Figure 2a). Real- time nIR images of NIM shows that the channel emission is completely quenched with \(1\mathrm{M} \mathrm{H}_2\mathrm{O}_2\) flowing (Figure 2b). This is attributed to that \(\mathrm{H}_2\mathrm{O}_2\) molecules selectively adsorbed on nanotube sidewall donate electrons directly to the conduction bands of SWNT/(GT)<sub>15</sub>, and extra electrons in the conduction bands can then quench excitons through a non- radiative recombination (Figure 2c).<sup>7,47</sup> Real- time nIR signals ((I - I<sub>0</sub>)/I<sub>0</sub>) were measured with wide range concentration of \(\mathrm{H}_2\mathrm{O}_2\) injection (Figure 2d). Here, \(I_0\) and \(I\) represent the nIR intensity of the channel at \(t = 0\) and after \(\mathrm{H}_2\mathrm{O}_2\) injection, respectively. Upon \(\mathrm{H}_2\mathrm{O}_2\) injection, the NIM showed an instantaneous and continuous decrease in nIR signal on the order of \(5 - 80\%\) depending on \(\mathrm{H}_2\mathrm{O}_2\) concentration. For the first- order reversible reaction, the relationship between the analyte and available docking sites for \(\mathrm{H}_2\mathrm{O}_2\) can be described as follows:<sup>48</sup>  
+
+<|ref|>equation<|/ref|><|det|>[[113, 680, 880, 700]]<|/det|>
+\[A + \theta \rightleftharpoons A\theta \quad (1)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 716, 523, 735]]<|/det|>
+the equilibrium for this reaction can be modeled as:  
+
+<|ref|>equation<|/ref|><|det|>[[113, 748, 880, 782]]<|/det|>
+\[K_{A} = \frac{[A\theta]}{[A][\theta]} \quad (2)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 798, 780, 818]]<|/det|>
+Assuming that the sensor response is proportional to the \(A\theta /\theta_{\mathrm{tot}}\) ratio, it is found that  
+
+<|ref|>equation<|/ref|><|det|>[[112, 831, 880, 865]]<|/det|>
+\[\frac{I - I_0}{I_0} = \alpha \frac{[A\theta]}{[\theta_{\mathrm{tot}}]} + \beta = \alpha \frac{([A]K_A)^n}{([A]K_A)^{n + 1}} +\beta \quad (3)\]  
+
+<|ref|>text<|/ref|><|det|>[[112, 880, 884, 900]]<|/det|>
+with the total concentration of available recognition sites \([\theta ]_{\mathrm{tot}}\) and the parameter \(n\) for
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 87, 886, 355]]<|/det|>
+cooperativity. Fitting the data in Figure 2e with equation (3) \((R^{2} = 0.9983)\) results in a proportionality factor \(\alpha = 88.74\) with \(\beta = 2.30\) , \(K_{\mathrm{D}} = 1 / K_{\mathrm{A}} = 0.00204 \mathrm{M}\) , and \(n = 0.317\) , indicating negative cooperativity in good agreement with previous papers \((n< 1)\) . \(^{42,45,48}\) The limit of detection in this mode is \(11.56 \mathrm{nM}\) ; this value was calculated by adding the NIM sensor response from the addition of only buffer (PBS) to 3- times the standard deviation \((\sigma)\) . A response time of less than 9 min was achieved based on the time it takes to reach \(90\%\) value of the minimum nIR level (Figure 2f). The NIM platform demonstrates uniform and near instantaneous nIR intensity response even when imaged at the high- resolution needed to interrogate single cells \((\sim 20 \mu \mathrm{m})\) (Figure 2g).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 403, 313, 422]]<|/det|>
+## Cellular Lensing Effect  
+
+<|ref|>text<|/ref|><|det|>[[112, 436, 886, 879]]<|/det|>
+For the NCC implementation, the NIM was integrated with a syringe pump and nIR microscope. 561 nm excitation laser was provided from the bottom side of the channel (right, Figure 3a). Human monocytes (U937) were cultured as chemical cytometry targets (Figure S6) since they are widely studied in biomedical fields with heterogeneous differentiation behavior into macrophages by immune activation. \(^{49,50}\) This monocyte- derived macrophage exhibits distinct ROS efflux in real- time as an immune response to various kinds of infection/inflammation. Measuring subtle molecular differences of ROS efflux can also benefit the detection and prevention of cardiovascular disease and neurodegenerative disorders. \(^{51,52}\) Therefore, a tool that would enable the precise profiling of dynamic antigenic response of single monocyte and eventually immune heterogeneities as function of different cellular physical properties could lead to mechanistic understanding and therapeutic development for these conditions. We found that the flowing cells optically interact with the underlying nanosensor emitter array and create a moving, label- free region of highest sensor signal by lensing the photoemission through the flowing cell itself (Figure
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 80, 886, 880]]<|/det|>
+3b- 3c and Video S1). This cell visualization was directly affected by both the uniformity and intensity of underlying nanosensor array (Figure S7). A magnified nIR image of a single flowing cell shows that the contour and shape of the monocyte could be visualized as observed in an OM (inset, top- right) with highest nIR intensity from the nanosensor array corresponding to the center, and Airy rings visible around the periphery (Figure 3c). Micro- particles larger than the illumination wavelength can similarly function as focusing lens. \(^{53,54}\) When particles have a RI contrast ratio with the fluid medium less than 2:1 and a diameter \((d_{\mu})\) larger than the wavelength \((- 2\lambda < d_{\mu}< 40\lambda)\) , a highly focused propagating beam from the shadow- side of the surface is generated due to constructive interference of the light field, called a photonic nanojet. \(^{55,56}\) For our system, the nIR fluorescence \((\lambda :1 - 1.25\mu \mathrm{m})\) from the top nanosensor array passes through the membranes, cytoplasm, and nucleus of the underlying flowing cells of mean diameter \(10 - 20\mu \mathrm{m}\) . The estimated RI of the cell components are \(n_{\mathrm{n}} = 1.43\pm 0.04\) for the monocyte nucleus, \(n_{\mathrm{c}} = 1.348\pm 0.004\) for the monocyte cytoplasm (average \(n_{\mathrm{cell}} = 1.383\) ) and \(n_{\mathrm{m}} = 1.33\) for the flowing media, which are optimum optical conditions for the photonic nanojet effect \((n_{\mathrm{cell}} / n_{\mathrm{m}} = 1.039\) (< 2)). \(^{57,58}\) Consequently, nIR photoemission from the integrated nanosensor array was refracted through the flowing cell and focused at a certain focal point below it, a phenomenon called cellular lensing. Several previous papers reported such photonic nanojet based micro- lensing behavior of cells in visible spectrum. \(^{59 - 62}\) In this study, we observe the photonic nanojet phenomena through a cell in the nIR range. Based on this, we could correlate the various biophysical properties of live cells and photonic nanojet effects for the first time. Consequent nIR intensity profiles of single cell showed the highest lens intensity \((I_{0})\) with 3 - 5 \(\mu \mathrm{m}\) full width at half- maximum (FWHM) and following Airy rings corresponding to the cell diameter (red plot in Figure 3d). This nIR lensing profile was measured for multiple cells \((n = 20)\) with almost identical FWHM \((3.37\mu \mathrm{m},\sigma = 0.22)\)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 883, 144]]<|/det|>
+and enhancement factor (9.43, \(\sigma = 1.86\) ), indicating that this lensing effect is reliable and specific to certain cell status (Figure S9).  
+
+<|ref|>text<|/ref|><|det|>[[112, 155, 886, 844]]<|/det|>
+Finite- difference time- domain (FDTD) numerical modeling can demonstrate the cellular lensing as originating from a photonic nanojet effect. \(^{55,56}\) Micro- spherical structures having similar diameters between 10 to \(20 \mu \mathrm{m}\) , eccentricity, and RI values \((n_{\mathrm{c}} / n_{\mathrm{m}} = 1.04)\) compared with cells were used as targets for FDTD modeling. The spherical target is excited by an incident plane wave of wavelength \(1 \mu \mathrm{m}\) corresponding to the fluorescence emission of the nanosensor array (modeling details in Methods and Supporting Information). The resulting optical intensity distribution map shows that light from top side of the target strongly focuses at a \(20 \mu \mathrm{m}\) distant point from the center of the cell forming a 2 to \(4 \mu \mathrm{m}\) wide light jet (left, Figure 3d). The model describes the experimental light intensity profile of the cellular nanojet at \(20 \mu \mathrm{m}\) focal distance with high fidelity in terms of \(I_{0}\) , FWHM, including Airy rings (right, Figure 3d). We note a slight deviation between the FDTD model and experiment for the Airy rings and FWHM possibly originating from the nonuniformity of nIR excitation source and ellipticity of the monocytes. When the excitation light was focused on the bottom surface of the NIM (at \(Z\) - stage \(= 100 \mu \mathrm{m}\) ), the target cell is not distinguishable above the background (red line) (Figure 3e and S10). A slight lensing peak \((I_{0})\) begins to be observed at \(80 \mu \mathrm{m}\) (orange line), and is highest in intensity at \(20\) to \(30 \mu \mathrm{m}\) distance from the top surface with an enhancement factor of 9.1 (blue line), in agreement with the focal points of the FDTD numerical model. The variation in lensing intensity as a function of focusing distance also shows excellent agreement between model and experiment (Figure S14). This agreement give confidence that cellular lensing images are indeed projected \(20 \mu \mathrm{m}\) from the cell center and therefore observable for those flowing within \(10 \mu \mathrm{m}\) of the NIM top surface.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 87, 886, 635]]<|/det|>
+This unique nIR lensing effect was not only observed for monocytes, but also for other type of cells including lymphocyte, macrophages, epithelial cells (e.g. human embryonic kidney cells (HEK)), and endothelial cells (e.g. human umbilical vein endothelial cells (HUVEC)) (Figure 3f). Since all the cells are composed of cytoplasm, nucleus, and membrane, \(^{58}\) which have higher RI than that of media \((n_{\mathrm{m}})\) but ratio under 2, all cell species could form photonic nanojet and nIR lensing effect following their own shape and contour. Even cells that adhere on the channel surface such as HUVEC apparently display the profile of nIR lensing albeit with weaker intensity than suspended cells due to the smaller thickness (Figure S11). In contrast, reference micro-particles similar in size with cells of interest between 15 to \(25 \mu \mathrm{m}\) such as glass spheres, polystyrene (PS) and stainless steel particles with higher RI than \(n_{\mathrm{m}}\) (1.457, 1.586, and 2.756, respectively) display no nIR lensing. Note that the nIR fluorescence is highly refracted or reflected on surfaces and overfocused within such reference particles due to the high RI values (Figure 3g). \(^{55}\) In addition, we observe significantly weaker cellular lensing with lysed monocytes. In this case, the absence of cytoplasmic content reduces the RI to close to \(n_{\mathrm{m}} \sim 1.33\) , inhibiting nIR refraction. Accordingly, the observed nIR lensing effect appears to be a phenomenon unique to live cells having optimal RI, diameter and composition for the formation of a nIR photonic nanojet.  
+
+<|ref|>text<|/ref|><|det|>[[113, 645, 886, 876]]<|/det|>
+Distinct nIR lensing profiles were observed for each cell type corresponding to unique RI ranges \((1.40 \pm 0.01, 1.384 \pm 0.015, 1.383 \pm 0.006, 1.37,\) and \(1.355 \pm 0.0007\) , for B lymphocyte, \(^{63}\) macrophage, \(^{64}\) monocyte, \(^{57}\) HEK, \(^{65}\) and HUVEC, \(^{66}\) respectively; RI values are those reported previously). The FWHM and enhancement factor of each cell can be calculated and described with a FDTD numerical model (Figure 3h and 3i, respectively) with good agreement \((R^{2} = 0.942\) and 0.950, respectively). Model predictions show that cellular lensing can be utilized to estimate a wide range of biophysical properties of the cell including diameter, eccentricity, and RI (Figure
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 87, 886, 458]]<|/det|>
+S12- S14). For example, cells with higher RI show distinctly smaller FWHM and larger enhancement factors than cells with lower RI, in agreement with the FDTD model. High RI cells such as B lymphocyte are composed of larger nucleus volumes \((n_{n})\) than cytoplasmic components \((n_{c})\) for antibody and cytokine production. \(^{67}\) Thus, the nIR excitation wave becomes more refracted through a high RI cell and thus more tightly focused onto focal points compared with low RI HUVEC cells (Figure 3j). \(^{58}\) In this way, nIR cellular lensing in this NCC platform provides a unique opportunity to cross- correlate the chemical efflux as measured by the underlying nanosensor array with distinct biophysical properties such as cell diameter, eccentricity, and RI. Ultimately, these properties can be linked to critical attributes such as viability, membrane properties, or intracellular composition, quantitatively correlating them with biochemical signaling information.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 507, 678, 527]]<|/det|>
+## Real-time Chemical Efflux Detection Using Cellular Lensing Effect  
+
+<|ref|>text<|/ref|><|det|>[[112, 541, 886, 877]]<|/det|>
+We find that when human monocytes are injected into the NIM in a controlled stopped flow system, distinct nIR intensity variations can be observed for individual monocytes corresponding to different immune activation states (Figure 4a). We use phorbol 12- myristate 13- acetate (PMA) to induce immune activation of the human monocytes, since it is a known agonist of the protein kinase C (PKC) signaling cascade. PKC activates nicotinamide adenine dinucleotide phosphate (NADPH) oxidase and consequently stimulates \(\mathrm{H}_2\mathrm{O}_2\) secretion during differentiation into macrophages (Figure S16). \(^{68}\) NADPH oxidase activity generates other ROS species including superoxide anion \((\mathrm{O}_2\cdot \cdot)\) and hydroxyl radical \((\mathrm{OH}\cdot \cdot)\) of course but at significantly lower levels of \(10^{3}\) and \(10^{8}\) times less than \(\mathrm{H}_2\mathrm{O}_2\) , respectively. \(^{69,70}\) It is safe to assume that \(\mathrm{H}_2\mathrm{O}_2\) is the dominant efflux from monocyte activation. Time series nIR images show that the \(I_0\) corresponding to the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 87, 886, 530]]<|/det|>
+immune activated monocyte (+PMA) (middle of Figure 4a) decreases relative to non- activated monocyte (- PMA) (left of Figure 4a) with increasing time. Catalase, an enzyme that decomposes \(\mathrm{H}_2\mathrm{O}_2\) , \(^{71}\) suppresses the signal as a negative control (right of Figure 4a). To analyze quantitatively, the nIR pixels corresponding to the nanosensor array were integrated for each cell and labeled \((I_{\mathrm{cell}})\) , producing three cell populations per experiment (+PMA, - PMA, and +PMA & catalase) (Figure 4b). Activated monocytes show significant variation in their real- time nIR nanosensor response while - PMA showed slow and small variation over the 500 sec measurement window. We detect a basal \(\mathrm{H}_2\mathrm{O}_2\) level even for the non- activated monocytes without PMA activation, which is consistent with the literature. \(^{72}\) As expected, +PMA & catalase showed invariant sensor responses attributed to \(\mathrm{H}_2\mathrm{O}_2\) decomposition by the enzyme. The +PMA group \((n = 41)\) had average 4.5- and 3.4- times higher intensity variations than - PMA and +PMA & catalase groups (Figure S17). Also, the nIR image of single monocytes show distinct quenching traces after measurements consistent with a response due to \(\mathrm{H}_2\mathrm{O}_2\) efflux (Figure S18a- b).  
+
+<|ref|>text<|/ref|><|det|>[[113, 541, 886, 877]]<|/det|>
+The schematic in Figure 4c summarizes this real- time \(\mathrm{H}_2\mathrm{O}_2\) efflux detection for single cells using the cellular lensing effect. The moving cell within the flow field exhibits strong nIR lensing from the induced photonic nanojet while the \(\mathrm{H}_2\mathrm{O}_2\) efflux is minimal at the underlying nanosensor array (Figure S18c- d). During the periodic stopped flow, the \(\mathrm{H}_2\mathrm{O}_2\) efflux cloud surrounding each cell starts to register on the projected nanosensor area nearest to the cell, resulting in a quenching of the immediate spot. This quenching allows for precise quantification of the \(\mathrm{H}_2\mathrm{O}_2\) efflux. At this point, the nIR lensing power is drastically reduced with weaker fluorescence resulting from the waveguide light source. We modeled the 3D reaction and diffusion problem of the \(\mathrm{H}_2\mathrm{O}_2\) from the individual cell to translate the observed nIR quenching area above the cell into real- time local \(\mathrm{H}_2\mathrm{O}_2\) concentration (Figure 4d). An individual cell is assumed to be stationary
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 884, 180]]<|/det|>
+below the top channel surface with distance \(L\) ( \(\sim 10 \mathrm{um}\) ) and to instantaneously release \(\mathrm{H}_2\mathrm{O}_2\) molecules at \(t = 0\) sec. The effective distance between the source and nanosensor array \((L_{\mathrm{eff}})\) is then \(L_{\mathrm{eff}} = L + L_{\mathrm{cell}}\) , where \(L_{\mathrm{cell}}\) is the cell radius. The \(\mathrm{H}_2\mathrm{O}_2\) concentration \(C\) field is then  
+
+<|ref|>equation<|/ref|><|det|>[[113, 192, 880, 230]]<|/det|>
+\[C(x,y,z,t) = \frac{M}{(\sqrt{4\pi D t})^3}\exp \left(-\frac{x^2 + y^2 + z^2}{4D t} -K t\right) \quad (4)\]  
+
+<|ref|>text<|/ref|><|det|>[[112, 243, 885, 475]]<|/det|>
+where \(M\) is the mass flux of \(\mathrm{H}_2\mathrm{O}_2\) release at the cell core, \(D\) is the aqueous diffusion coefficient of \(\mathrm{H}_2\mathrm{O}_2\) \((1.5\cdot 10^{- 5}\mathrm{cm}^2\cdot \mathrm{sec}^{- 1})\) , \(^{73}K\) is the first- order decay constant of \(\mathrm{H}_2\mathrm{O}_2\) (from \(K = - \ln (0.5) / t_{1 / 2} =\) \(6.93\cdot 10^{- 4}\mathrm{sec}^{- 1}\) , where \(t_{1 / 2}\) is cellular half- life of \(\mathrm{H}_2\mathrm{O}_2\) \((10^{- 3}\mathrm{sec}))^{74}\) (detail model derivations in Supporting Information). The results show that the \(\mathrm{H}_2\mathrm{O}_2\) efflux reached the nearest nanosensor array quickly at 10 milli sec with a maximum concentration \(C_{\mathrm{sensor}}\) and the ratio between \(C_{\mathrm{sensor}}\) and \(C_{\mathrm{cell}}\) was 0.193 (Figure S19). The adsorption and desorption of \(\mathrm{H}_2\mathrm{O}_2\) on nanosensor array can be described by  
+
+<|ref|>equation<|/ref|><|det|>[[113, 488, 880, 509]]<|/det|>
+\[\mathrm{H}_2\mathrm{O}_2 + \mathrm{SWNT}\rightleftharpoons \mathrm{H}_2\mathrm{O}_2\mathrm{-SWNT} \quad (5)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 523, 427, 544]]<|/det|>
+Corresponding to the rate expression: \(^{75}\)  
+
+<|ref|>equation<|/ref|><|det|>[[113, 556, 880, 588]]<|/det|>
+\[\frac{d[H_2O_2 - SWNT]}{dt} = k_f[H_2O_2][SWNT] - k_r[H_2O_2 - SWNT] \quad (6)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 601, 884, 730]]<|/det|>
+where \(k_{\mathrm{f}}\) , \(k_{\mathrm{r}}\) are the forward and backward rate constants, respectively, and ratio between \(k_{\mathrm{f}}\) and \(k_{\mathrm{r}}\) was calculated from the effective equilibrium dissociation constant \(K_{\mathrm{D}} = 0.00204 \mathrm{M}\) . Since the nIR intensity of the nanosensor array is proportional to the fraction of unoccupied sites for binding, [SWNT], or  
+
+<|ref|>equation<|/ref|><|det|>[[116, 742, 880, 764]]<|/det|>
+\[I / I_0 = [\mathrm{SWNT}] / [\mathrm{SWNT}]_0 \quad (7)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 777, 545, 797]]<|/det|>
+And the number of binding sensor sites are conserved:  
+
+<|ref|>equation<|/ref|><|det|>[[113, 810, 880, 833]]<|/det|>
+\[[\mathrm{SWNT}]_0 = [\mathrm{SWNT}] + [\mathrm{H}_2\mathrm{O}_2\mathrm{-SWNT}] \quad (8)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 846, 883, 901]]<|/det|>
+The local concentration of \(\mathrm{H}_2\mathrm{O}_2\) detected by the nanosensor array involves the measured intensity \((I)\) and its time- derivative
+
+<--- Page Split --->
+<|ref|>equation<|/ref|><|det|>[[113, 85, 880, 121]]<|/det|>
+\[[H_{2}O_{2}] = \frac{1}{k_{f}}\frac{I_{0}}{l}\left[k_{r}\left(1 - \frac{l}{I_{0}}\right) - \frac{1}{I_{0}}\frac{dI}{dt}\right] \quad (9)\]  
+
+<|ref|>text<|/ref|><|det|>[[114, 137, 357, 156]]<|/det|>
+Integrating equation (9) yields  
+
+<|ref|>equation<|/ref|><|det|>[[113, 170, 880, 240]]<|/det|>
+\[\begin{array}{l}{I(t) = \frac{I_0}{k_s} (k_r + k_f[H_2O_2]e^{-k_s t})}\\ {}\\ {k_s = k_r + k_f[H_2O_2]} \end{array} \quad (10)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 255, 884, 450]]<|/det|>
+Equation (10) can be utilized to estimate the real- time local \(\mathrm{H}_2\mathrm{O}_2\) concentration of each single cell from the measured nIR intensity (Figure 4e). Both the efflux signal ([H2O2]cell, red line) and background ([H2O2]bg, green line) for each single monocyte can be measured and differentiated. Furthermore, each monocyte (cell 1 to cell 8) demonstrates distinct \(\mathrm{H}_2\mathrm{O}_2\) efflux rates resulting in local concentrations ranging from 175 to \(750~\mu \mathrm{M}\) . This shows that our NCC platform can inform heterogeneities in the efflux rates within cell populations.  
+
+<|ref|>sub_title<|/ref|><|det|>[[113, 499, 710, 519]]<|/det|>
+## NCC for Monitoring of Multimodal Immune Response Heterogeneities  
+
+<|ref|>text<|/ref|><|det|>[[112, 533, 886, 904]]<|/det|>
+The combination of cellular lensing and label- free nanosensor monitoring within a microfluidic channel allows for real- time chemical efflux cytometry of distinct human monocyte populations, such as those that are immune activated (+PMA) compared to non- activated (- PMA) (Figure 5a). We show that the NCC platform collects a rich, multivariate data set for each individual cell within the population, that we then easily extract and evaluate with the aid of image analysis code developed as a part of this work (Figure 5b and S20). The results allow us to plot the real- time \(\mathrm{H}_2\mathrm{O}_2\) efflux rates of two distinct groups ( \(n = 413\) for - PMA, \(n = 414\) for +PMA) versus various biophysical properties of each individual cell such as size (cell projected area), eccentricity, and RI (Figure 5c). Upon immune activation, we find that the mean size of monocytes decreases along with a narrowing of the distribution (Figure 5c1). This occurs with an increase of \(\mathrm{H}_2\mathrm{O}_2\) efflux rate. In contrast, the eccentricity (Figure 5c2) and RI (Figure 5c3) distributions show
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 81, 886, 670]]<|/det|>
+insignificant correlation with \(\mathrm{H}_2\mathrm{O}_2\) efflux rate. To be clear, 3D cytometry and 2D Kernel density estimation show these distinct heterogeneities in detail (Figure S21 and S22, respectively). From these cytometry plots, it is clear that the average \(\mathrm{H}_2\mathrm{O}_2\) efflux rate of activated monocyte population were elevated by \(88.9\%\) with a \(44.5\%\) larger increase in the variance of the distribution and \(30\%\) larger number of high efflux cells compared to non- activated populations (Figure 5d). The nanosensor array allows us to quantify the mean \(\mathrm{H}_2\mathrm{O}_2\) efflux rates of these two populations as 330 and 624 attomole/cell·min but with \(\sigma\) of 344 and 497 attomole/cell·min for - PMA and +PMA, respectively. In comparison, we measure average values of 59 (-PMA) and 440 (+PMA) attomole/cell·min from the commercial assay Amplex UltraRed kit (Figure S23). The +PMA mean values are in good agreement for the NCC population and commercial assay. However, the mean for the - PMA as measured by NCC is larger than the commercial assay. Further analysis indicates that hyperactive outliers (>1000 attomole/cell·min) in this population are the cause of the difference. The mode in the - PMA distribution as measured by NCC of 129 attomole/cell·min is closer to the commercial assay mean, and NCC distribution curves shows the significant higher active tail in Figure 5d. The ability to detect and quantify this higher producing subpopulation is a clear advantage of NCC over the standard assay. As a consistency check, we note that both methods produce the correct order of magnitude estimate of the \(\mathrm{H}_2\mathrm{O}_2\) efflux rates.  
+
+<|ref|>text<|/ref|><|det|>[[113, 681, 886, 876]]<|/det|>
+Among the biophysical property changes, the size vs eccentricity correlation shows the most dramatic change after immune activation (Figure 5e and S24). There is a distinct change in the size distribution upon monocyte activation, with bimodal subpopulations observed for non- activated monocytes with a mean of \(271 \mu \mathrm{m}^2\) ( \(\sigma = 29\) ) but a single distribution with lower mean of \(263 \mu \mathrm{m}^2\) ( \(\sigma = 24\) ) after activation (Figure 5f1). This observation is important because one requires single cell resolution in order to quantify this type of biophysical change, underscoring an
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 319]]<|/det|>
+advantage of this NCC platform. Notably, the distributions for both eccentricity (Figure 5f2) and RI (Figure 5f3) remain nearly identical comparing before and after activation but the mean values are slightly shifted from 0.405 \((\sigma = 0.14)\) to 0.363 \((\sigma = 0.13)\) for eccentricity and 1.383 \((\sigma = 0.05)\) to 1.377 \((\sigma = 0.06)\) for RI. This indicates that immune activation had a uniform effect on the cell populations with respect to these properties. The ability to detect and analyze subpopulations from a cellular population undergoing biofunctional changes has significant advantages in analytical biochemistry.  
+
+<|ref|>text<|/ref|><|det|>[[112, 330, 886, 880]]<|/det|>
+Figure 5g summarizes the variation in human monocyte characteristics before and after the immune activation process. Real- time \(\mathrm{H}_2\mathrm{O}_2\) efflux rate of monocyte populations showed \(88.9\%\) elevation. Populations showed \(- 2.92\%\) and \(- 10.31\%\) decrease in cell size and eccentricity respectively, indicating that monocytes appear to shrink and become more circular with immune activation. The RI of the populations decreased by \(- 0.3\%\) scale, which means that light refracted through activated cells produced almost identical refraction angles. This new insight may lead to additional methods of sorting human monocyte populations. As a consistency check, all of the measured values of NCC were within the ranges previously reported for monocytes, including \(\mathrm{H}_2\mathrm{O}_2\) efflux rate: \(^{72}\) 100 to 1000 attomole/cell·min, size: \(^{76}\) 78.5 to \(314 \mu \mathrm{m}^2\) , eccentricity: \(^{57}\) 0.323 to 0.473, RI: \(^{57,58}\) 1.377 to 1.389. We can safely conclude that our NCC approach is reliable in this way and allows the investigation of multiple cellular parameters of a given population in real- time and at high throughput. These cellular parameter changes of monocytes during activation are consistent with PKC translocation effects. It is known that when monocytes are activated by PMA, PKC proteins are translocated from the cytosol to the plasma membrane, activating NADPH oxidase with an increase in ROS generation. \(^{77}\) Subsequently, fluidity and permeability of the cellular membrane are both downregulated upon PKC integration. \(^{78,79}\) One expects a resulting
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 80, 886, 878]]<|/det|>
+image, producing a profile that matches the predictions of photonic nanojet model. The result is a unique tool capable of multimodal biophysical characterization of individual cells, including their size, eccentricity, and RI, all at high throughput. With this biophotonic waveguide, the chemical efflux of single cell was label- free monitored in real- time at the attomolar level. We use this NCC tool to study the heterogeneity of the immune response of distinct human monocyte populations at highest throughput range for chemical cytometry ( \(\sim 100\) cell/frame) in a completely nondestructive manner. Mathematical analysis of the resulting rich data sets reveals new phenotypic correlations between chemical efflux and biophysical properties that can quantified, and used to understand new aspect of cellular biochemistry and mechanistic pathways. For example, we find that real- time \(\mathrm{H}_2\mathrm{O}_2\) efflux of human monocytes is unusually heterogeneous and distinctly related to biophysical parameters following immune activation. The measured \(\mathrm{H}_2\mathrm{O}_2\) efflux rates between 330 and 624 attomole/cell·min corresponded to overall cell size ranges of 271 and \(263~\mu \mathrm{m}^2\) , eccentricity values between 0.405 and 0.363 and RI values between 1.383 and 1.377 for nonactivated and activated monocytes, respectively. Thus, we highlight that NCC is able to profile immune cell heterogeneities allowing for monitoring of variances in cell therapeutics. We also demonstrate the ability to incorporate sensors for multiple molecular targets of cells. Our platform is label- free and uses the unique property of cellular lensing to extract molecular signals on a population scale. We believe that the NCC platform can be readily extended to various biochemical efflux monitoring of cell types such as neurons, cancer cells or stem cells given the appropriate choices of sensor- analyte pairs (Figure S27). We envision that our nanotechnology based biophotonic cytometry provides a unique strategy for coupling nanosensors into a form- factor that enables single- cell analysis of relevant populations for cellular manufacturing, cellular immunology, and biopharmaceutical research.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[125, 88, 875, 555]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[114, 556, 881, 696]]<|/det|>
+<center>Figure 1. Nanosensor integration with microfluidics. (a) Schematic illustration of nanosensor integration process with microfluidics using EISA. (b) Photograph of EISA process of NIM for 0 min (left) and 30 min (right). (c) Photograph of completed multi-array NIM and pristine channel. (d) Polarized Raman spectrum (G-peak) of NIM. (e) nIR images of NIM and pristine channel. (f) Magnified nIR image of NIM with single cell size resolution (20 \(\mu \mathrm{m}\) ) having \(\sim 720\) nIR reporter pixel. (g) Histograms of nIR pixel intensities of top and bottom NIM surfaces NIM (inset: nIR images of top and bottom surfaces). (h) nIR fluorescence spectrum of NIM. (i) nIR images of NIM with varying composition of SWNT nanosensor integration. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[117, 90, 879, 543]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[114, 544, 883, 682]]<|/det|>
+<center>Figure 2. In-vitro chemical detection performances of NIM. (a) nIR spectrum of NIM with \(\mathrm{H}_2\mathrm{O}_2\) solution flowing (1 \(\mu \mathrm{M}\) , 1 \(\mu \mathrm{L} / \mathrm{min}\) ). (b) nIR images of NIM before and after \(\mathrm{H}_2\mathrm{O}_2\) flowing (1 M, 10 \(\mu \mathrm{L} / \mathrm{min}\) , 10 min). (c) Schematic illustration of \(\mathrm{H}_2\mathrm{O}_2\) detection mechanism of SWNT/(GT)15 nanosensor. (d) Real-time nIR response of NIM with various concentration ( \(10^{-6}\) , \(10^{-5}\) , \(10^{-4}\) , \(10^{-3}\) , \(10^{-2}\) , \(10^{-1}\) , \(10^{0}\) M) of \(\mathrm{H}_2\mathrm{O}_2\) injection (10 min). (e) Maximum response amplitude and (f) response time of NIM with various concentration of \(\mathrm{H}_2\mathrm{O}_2\) . The data represent the mean value of 250 X 350 \(\mu \mathrm{m}^2\) NIM measurement. (g) nIR snapshots and intensity histogram (fire scale, ImageJ) of NIM with single cell size resolution (20 \(\mu \mathrm{m}\) ) after 10 min flowing of various concentration of \(\mathrm{H}_2\mathrm{O}_2\) . </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[132, 88, 866, 730]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[114, 731, 883, 906]]<|/det|>
+<center>Figure 3. Cellular lensing effect. (a) Instrumental setup for NCC implementation: schematic illustration (left) and a photograph (right). (b) nIR images of human monocytes flowing (0.5 \(\mu \mathrm{L} / \mathrm{min}\) ) NIM. (c) Magnified nIR image of single monocyte in NIM (inset: OM image of single monocyte). (d) FDTD numerical modeling for photonic nanojet and fitting with experimental cellular lensing profile \((n_{\mathrm{c}} / n_{\mathrm{m}} = 1.04\) , \(\lambda = 1 \mu \mathrm{m}\) ). (e) nIR lensing profiles of a single cell with various focusing points from 5 to \(100 \mu \mathrm{m}\) along Z-stage. nIR lensing effects of (f) various live cells and (g) reference micro-particles (top-to-bottom: schematics, OM, nIR images, lensing profiles). (h) FWHM and (i) enhancement factors of various cells with numerical model. Data are mean \(\pm \sigma\) , with \(n_{\mathrm{cell}} = 10\) . (j) Schematic illustrations for different lensing behavior of a high RI cell (left) and a low RI cell (right). Scale bars: \(20 \mu \mathrm{m}\) . </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 90, 881, 642]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[114, 644, 883, 784]]<|/det|>
+<center>Figure 4. Real-time chemical efflux monitoring using the cellular lensing effect. (a) Time-series nIR images of a stationary single monocyte with different immune activation states (-PMA, +PMA, +PMA & catalase). (b) Real-time nIR intensity variations of the cells with different activation states. (c) Schematic illustrations of \(\mathrm{H}_2\mathrm{O}_2\) efflux monitoring mechanism with nIR lensing effect. (d) 3D diffusion and reaction kinetic modeling for translation of measured nIR signals to real-time local \(\mathrm{H}_2\mathrm{O}_2\) concentration. (e) Real-time \(\mathrm{H}_2\mathrm{O}_2\) efflux profiles of each single monocyte estimated by the model. 16-color scalebars represent nIR intensity from white (16833) to dark blue (0). Scale bars: \(20\mu \mathrm{m}\) . </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[123, 88, 877, 696]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 697, 884, 890]]<|/det|>
+<center>Figure 5. NCC for monitoring of multimodal immune response heterogeneities. a) Schematics and nIR images of NCC setup with distinct activation of human monocytes (-PMA and +PMA). (b) Automatic nIR image analysis using computational code for cell data extractions. (c) NCC cytometry plots of \(\mathrm{H}_2\mathrm{O}_2\) efflux rate \(\nu s\) biophysical parameters ((c1) size (2D projected area), (c2) eccentricity, (c3) RI) of two monocytes populations. Data are \(n_{\mathrm{cell}} = 413\) for -PMA, \(n_{\mathrm{cell}} = 414\) for +PMA from \(n = 6\) biologically independent samples. (d) NCC distribution curves of \(\mathrm{H}_2\mathrm{O}_2\) efflux rates with data from commercial assay kit. (e) NCC cytometry plots for cell biophysical parameters ((e1) eccentricity \(\nu s\) size, (e2) RI \(\nu s\) eccentricity, (e3) size \(\nu s\) RI). (f) NCC distribution curves of each biophysical parameters ((f1) size, (f2) eccentricity, (f3) RI). (g) Schematics illustrations for cell properties variations of human monocyte populations with immune activations. Scale bars: 20 \(\mu \mathrm{m}\) . </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 90, 235, 110]]<|/det|>
+## METHODS  
+
+<|ref|>text<|/ref|><|det|>[[112, 125, 886, 570]]<|/det|>
+Preparation and characterization of nanosensors. HiPco™ SWNTs purchased from Unidym were suspended with a 30- base (GT) sequence of ssDNA (Integrated DNA Technologies) in a 2:1 DNA:SWNT mass ratio in 0.1 M NaCl solution. (ATCAAGGCTCGAATTGTCCCTGA AATCT) sequence was used for random DNA and polystyrene sulfonate/bromostyrene was used for random copolymer in reference test. A typical DNA concentration was \(2\mathrm{mg / mL}\) . Samples were sonicated with a \(3\mathrm{mm}\) probe tip (Cole Parmer) for \(10\mathrm{min}\) at a power of \(10\mathrm{W}\) and \(40\%\) amplitude in an ice bath. Then samples were centrifuged twice for \(90\mathrm{min}\) (Eppendorf Centrifuge 5415D) at 16100 RCF (Relative Centrifugal Force). Afterwards, the supernatant was collected and the pellet was discarded. UV- Vis- nIR absorption spectra (Cary 5000, Agilent Technologies, Inc) were collected to verify successful suspension of nanosensor. Nanosensor concentration in the dispersion was estimated using an extinction coefficient of \(\ell_{632\mathrm{nm}} = 0.036\mathrm{mg / L})^{- 1}\) . Final concentration of SWNT/(GT)15 is from \(10\) to \(80\mathrm{mg / L}\) . \(80\mathrm{mg / L}\) concentration of nanosensor dispersion was used to all NIM experiments.  
+
+<|ref|>text<|/ref|><|det|>[[112, 581, 886, 882]]<|/det|>
+Nanosensors integration with microfluidic channel. Microfluidic channels (detail specification in Table S1) were purchased from ibidiR (μ- Slide VI 0.1, ibiTreat). \(2\mu \mathrm{L}\) of APTES (99%, Sigma Aldrich) in ethanol (1% APTES, 1% \(\mathrm{H}_2\mathrm{O}\) ) was injected to microfluidic channel with micropipetting and treated for \(3\mathrm{hr}\) . After APTES treatment, \(2\mu \mathrm{L}\) of nanosensor dispersions were injected. After overnight evaporation, SWNT/(GT)15 coated channel surfaces were rinsed with \(1\mathrm{mL}\) 1X PBS (pH 7.4, Life TechnologiesTM) twice to remove unbounded nanosensor. \(0.8\mathrm{mm}\) Silicone tubes were connected with NIM using Elbow Luer Connector Male (ibidiR). All experiments and characterization were done in triplicate with three different NIM fabrication. In- vitro \(\mathrm{H}_2\mathrm{O}_2\) detection experiments were conducted as below. SWNT/(GT)15 releases the nIR
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 319]]<|/det|>
+fluorescence with visible range excitation laser (e.g. 516 nm) acting as an optical transducer for \(\mathrm{H}_2\mathrm{O}_2\) detection. Aqueous \(\mathrm{H}_2\mathrm{O}_2\) solution (30 wt\%, Sigma Aldrich) was diluted with distilled \(\mathrm{H}_2\mathrm{O}\) from \(1\mu \mathrm{M}\) to \(1\mathrm{M}\) to investigate chemical sensing performance of NIM. Diluted \(\mathrm{H}_2\mathrm{O}_2\) solutions were flowing through the NIM with syringe pump (0 - \(1\mu \mathrm{L} / \mathrm{min}\) , Harvard Apparatus) and averaged quenching signals from nanosensor array (250 X 350 \(\mu \mathrm{m}^2\) ) were recorded for 500 - 600 sec. Recorded nIR images were processed by ImageJ with Gray and Fire scales to clearly visualize the variations of nIR intensities.  
+
+<|ref|>text<|/ref|><|det|>[[112, 330, 886, 880]]<|/det|>
+Characterization and nIR measurements. Raman spectroscopy (Horiba Jobin Yvon LabRAM HR800) was used to investigate the nanosensor assembly direction in NIM with a 532 nm laser excitation (3 sec accumulations) and \(\sim 1\mu \mathrm{m}\) of spot size with 1800 lines/mm grating. The G band originating from tangential oscillations of the carbon atoms in the SWNT was observed in the frequency range of \(1590\mathrm{cm}^{- 1}\) . When \(\theta = 0^{\circ}\) and \(\theta = 90^{\circ}\) , the incident excitation polarization direction was parallel and perpendicular to the flowing direction of the microfluidic channel, respectively, indicating that the SWNT/(GT)15 nanosensors were aligned along the flowing direction of channel during EISA. AFM profiles of nanosensor array were scanned with Bruker Multimode 8 with Controller V. AFM images were taken in the ScanAsyst tapping mode in the air with TESPA probes having an elastic constant of \(42\mathrm{N / m}\) and tip radius of \(8\mathrm{nm}\) . The images were recorded with the scan rate of \(1\mathrm{Hz}\) and resolution of 1024 lines per image for each area respectively, recorded at three different places of single channel surface. Image analysis was done with Nanoscope Analysis software 1.4 from Bruker. nIR spectrum of NIM were collected with a fluorescence spectrometer equipped with a 785 nm photodiode laser (B&W Tek. Inc. 450 mW). Low- magnified nIR images were collected using a Zeiss AxioVision inverted microscope with appropriate optical filters. The fluorescence passed through an Acton SP2500 spectrometer
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 84, 886, 600]]<|/det|>
+(Princeton Instruments), and measured with a liquid nitrogen cooled InGaAs 1D detector (Princeton Experiments). Inverted OM (Eclipse TS100, Nikon) was used for NIM and flowing cell imaging with visible light. NCC were implemented and recorded by nIR microscopy hyperspectral imager (IMA IR™, Photon Etc.). NCC was implemented with the help of a nIR microscope (IMA IR™, Photon Etc.) equipped with 561 nm laser excitation (MGL- FN- 561, Opto Engine LLC). The laser power was adjusted from 30 mW to 350 mW with optical density filters (laser power control in Figure S7). The laser was passed through a laser line filter, reflected by dichroic mirror, and focused onto the back focal plane of an inverted objective to illuminate the entire field of view of the NIM under study. nIR fluorescence from the NIM passed a longpass filter and was measured using a TE cooled infrared camera. All the measurements were conducted with 20X objective, 0.1 sec exposure time and medium intensity gain. In order to investigate the focal points and observed cell locations, motorized Z- stage controller was integrated with nIR microscopy. Hollow glass microspheres (0.6 g/cc & 5 - 30 μm, Cospheric LLC), PS microparticle (20 μm, Sigma Aldrich), and stainless steel metal microspheres (7.8 g/cc & 1 - 22 μm, Cospheric LLC) were used for reference particles as lensing effect observations.  
+
+<|ref|>text<|/ref|><|det|>[[112, 609, 886, 876]]<|/det|>
+FDTD numerical modeling. FDTD modeling for nIR photonic nanojet were performed using Lumerical FDTD Solution (Lumerical Inc). Micro- spherical structures having various range of size (radius: 1, 2, 3, 4, 5, 6, 7, and 8 μm), eccentricity (Z- axis distance: 2.5, 3, 3.5, 4, and 4.5 μm), and RI \((n_{\mathrm{c}} / n_{\mathrm{m}}; 1.01, 1.02, 1.03, 1.04, 1.05, 1.06, 1.07, 1.08, 1.09,\) and 1.10) were set and excited by an incident plane wave with a wavelength of \(1000\mathrm{nm}\) , corresponding to the fluorescence emission of the nanosensor array. The calculation domain was \(50\mathrm{X}50\mathrm{X}50\mathrm{μm}^3\) and uniform mesh of around \(30\mathrm{nm}\) was used. The perfectly matched layers (PML) were arranged around the boundaries. RI of media (out of cell) was set to 1.33.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 80, 886, 880]]<|/det|>
+Cell experiments. Monocytes (U937, ATCC CRL- 1593.2), B lymphocytes (FIB504.64, ATCC HB- 293), epithelial (HEK- 293, ATCC CRL- 1573), and endothelial (HUVEC, ATCC CRL- 1730) cells were purchased from American Type Culture Collection (ATCC) and cultivated according to the supplier's protocol. U937 and FIB504.64 were cultured in RPMI- 1640 (ATCC 30- 2001) with \(10\%\) of Fetal Bovine Serum (FBS) (A3160601, Gibco™). HEK- 293 cells were cultured in Dulbecco's Modified Essential Medium (DMEM; Lonza) with \(10\%\) FBS (ATCC 30- 2020). HUVEC were cultured in F- 12K medium supplemented with \(10\%\) FBS (ATCC 30- 2020), \(1\%\) endothelial cell growth factor (100X, Sigma), \(100 \mathrm{IU / mL}\) penicillin, and \(100 \mu \mathrm{g / mL}\) streptomycin. For the adherent HUVEC observations, microfluidic channels were initially coated with endothelial cell attachment factor (ECAF) to promote HUVEC cell adherence on channel surfaces. All the cells were cultured in \(75 \mathrm{cm}^2\) cell culture flasks (Falcon) under incubating conditions of \(5\%\) \(\mathrm{CO_2}\) at \(37^{\circ}\mathrm{C}\) (Forma™ series II 3110, ThermoFisher Scientific). Three days cultured U397 were used (cell number: \(10^{4} - 10^{5} / \mathrm{mL}\) , passage number \(= 4\) ) to implement NCC in this study. To monitor only the instantaneous \(\mathrm{H}_2\mathrm{O}_2\) efflux, cell media was changed by fresh PBS with \(10 \mathrm{min}\) 130 RCF centrifugation at \(10^{\circ}\mathrm{C}\) so that remove all the by- product, accumulated efflux and abnormal cells in media. \(10 \mu \mathrm{L}\) of \(0.5 \mathrm{mg / mL}\) PMA (Sigma Aldrich, for use in molecular biology applications, \(\geq 99\%\) ) was added in \(1 \mathrm{mL}\) of U937 cell media to activate the monocyte and induce differentiation into macrophage (final concentration of \(\mathrm{PMA} = 5 \mu \mathrm{g / mL}\) ). \(100 \mathrm{mL}\) of \(200 \mathrm{units / mL}\) Catalase (Sigma Aldrich, from bovine liver) was used for \(\mathrm{H}_2\mathrm{O}_2\) removal control experiments. For final NCC implementation, activated (+PMA) and non- activated monocytes (-PMA) were flowing through NIM using syringe pump (Harvard Apparatus) with flowing rate from \(0\) to \(10 \mu \mathrm{L / min}\) . PMA group was firstly injected through channel 1 with syringe pump and measured at stopped position for \(10 \mathrm{min}\) to accumulate the \(\mathrm{H}_2\mathrm{O}_2\) efflux on nanosensor array. Then, measured cells were flowed (10
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 283]]<|/det|>
+\(\mu \mathrm{L} / \mathrm{min}\) ) and collected in empty tube for the future experiments. Lastly, \(+\mathrm{PMA}\) group was injected to channel 2 and \(\mathrm{H}_2\mathrm{O}_2\) efflux was measured. NCC were conducted for stationary cells for few min and videos were recorded to analysis efflux signals of the cells. Attomolar efflux rates were calculated from real- time local \(\mathrm{H}_2\mathrm{O}_2\) concentration multiplied with single unit volume (single monocyte volume \(= 4.18\cdot 10^{- 15} \mathrm{m}^3\) ) and divided by measurement time (10 min). Six biological replicates of U937 populations were used for NCC cytometry plot data.  
+
+<|ref|>text<|/ref|><|det|>[[112, 298, 886, 702]]<|/det|>
+Data analysis. nIR image analysis and quantitation was performed in MATLAB (Natick, MA) with the steps detailed below. Cell identification is performed by taking 1 frame of the nIR video (500 sec recorded, 0.1 sec of exposure time, 5000 frames) convolving with a Laplacian of a Gaussian filter, and then thresholded by the user for each experiment batch. For each cell, the image is then interpolated. Using the peak and Airy ring of the nIR lensing spot, the cell image is normalized, and then statistics such as cell size and eccentricity are evaluated with "Regionprops" function. The projected area (i.e. size) values are dilated appropriately to coincide with the photonic nanojet model. To avoid excess data interpolation, camera pixel intensities are used for subsequent analysis. The cellular lensing intensity \((I_0)\) is found by choosing the camera pixel closest to the centroid. To calculate background, 16 pixels outside of the secondary peak of the lensing effect is chosen. Outliers are then removed, and background traces are averaged to use as normalization for the centroid intensity traces.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 751, 388, 772]]<|/det|>
+## ACKNOWLEDGEMENTS  
+
+<|ref|>text<|/ref|><|det|>[[114, 790, 884, 880]]<|/det|>
+The authors are grateful for financial support from Bose Fellowship Award to M. S. S., the Juvenile Diabetes Research Foundation (JDRF), funding from the Disruptive & Sustainable Technology for Agricultural Precision (DiSTAP) and the Singapore MIT Alliance for Research and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 87, 885, 284]]<|/det|>
+Technology (SMART) Center, and the Walmart Foundation and the Walmart Food Safety Collaboration Center in Beijing. T. T. S. L. acknowledges a graduate fellowship by the Agency of Science, Research and Technology, Singapore. M. K. acknowledges support by the German Research Foundation (DFG) Research Fellowship KU 3952/1- 1. We give thanks to the Nanotechnology Materials Lab, and the Koch Institute for Integrative Cancer Research, MIT for AFM measurements.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 333, 420, 355]]<|/det|>
+## AUTHOR CONTRIBUTIONS  
+
+<|ref|>text<|/ref|><|det|>[[112, 371, 886, 743]]<|/det|>
+S.- Y. C and M. S. S. conceived the idea, designed the project and planned experiments with the assistance of X. G., V. B. K., M. K., S. J. M., M. S. and T. T. S. L.. S.- Y. C prepared the nanosensors, fabricated the NIM, implemented the NCC with cell culturing, measured and analyzed the data. X. G. coded automatic image analysis program and conducted cell data processing. V. B. K. conducted FDTD numerical modeling for photonic nanojet demonstration. M. K. assisted nIR observations of NIM. S. J. M. conducted HEK cell culturing and commercial \(\mathrm{H}_2\mathrm{O}_2\) assay. M. S. assisted with polarized Raman spectroscopy and nanosensor synthesis. T. T. S. L. commented about \(\mathrm{H}_2\mathrm{O}_2\) nanosensor and their detection mechanism. X. J. assisted with preparation of various nanosensor. P. G. assisted with AFM characterization of nanosensor array. S.- Y. C and M. S. S. wrote the manuscript with inputs from all the authors. All authors contributed to discussions informing the research.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 792, 396, 813]]<|/det|>
+## COMPETING INTERESTS  
+
+<|ref|>text<|/ref|><|det|>[[115, 832, 528, 852]]<|/det|>
+The Authors declare no competing financial interest
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 90, 357, 111]]<|/det|>
+## DATA AVAILABILITY  
+
+<|ref|>text<|/ref|><|det|>[[115, 130, 883, 184]]<|/det|>
+The authors declare that all data supporting the findings of this study are available within the paper and any raw data can be obtained from the corresponding author on request.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 235, 272, 256]]<|/det|>
+## REFERENCES  
+
+<|ref|>text<|/ref|><|det|>[[108, 270, 888, 900]]<|/det|>
+(1) Arroyo, J. O., & Kukura, P. Non-fluorescent schemes for single-molecule detection, imaging and spectroscopy. Nat. Photonics. 10, 11-17 (2016).  
+(2) Holzmeister, P. et al. Breaking the concentration limit of optical single-molecule detection. Chem. Soc. Rev. 43, 1014-1028 (2014).  
+(3) Rigler, R. & Vogel, H. Single Molecules and Nanotechnology (Springer, 2008).  
+(4) Garoli, D. Plasmonic Nanopores for Single-molecule detection and manipulation: Toward sequencing applications. Nano Lett. 19, 7553-7562 (2019).  
+(5) Ameur, A., Kloosterman, W. P., & Hestand, M. S. Single-molecule sequencing: Towards clinical applications. Trends Biotechnol. 37, 72-85 (2019).  
+(6) Kim, J.-H. et al. Single-molecule detection of \(\mathrm{H}_2\mathrm{O}_2\) mediating angiogenic redox signaling on fluorescent single-walled carbon nanotube array. ACS Nano. 5, 7848-7857 (2010).  
+(7) Heller, D. A. et al. Multimodal optical sensing and analyte specificity using single-walled carbon nanotubes. Nat. Nanotechnol. 4, 114-120 (2009).  
+(8) Kim, J.-H. et al. The rational design of nitric oxide selectivity in single-walled carbon nanotube near-infrared fluorescence sensors for biological detection. Nat. Chem. 1, 473-481 (2009).  
+(9) Jin, H. et al. Detection of single-molecule \(\mathrm{H}_2\mathrm{O}_2\) signalling from epidermal growth factor receptor using fluorescent single-walled carbon nanotubes. Nat. Nanotechnol. 5, 302-309 (2010).  
+(10) Oh, S., Lee, M.-K., & Chi, S.-W. Single-molecule-based detection of conserved Influenza A virus RNA promoter using a protein nanopore. ACS Sens. 4, 2849-2853 (2019).  
+(11) Landry, M. P. et al. Single-molecule detection of protein efflux from microorganisms using fluorescent single-walled carbon nanotube sensor arrays. Nat. Nanotechnol. 12, 368-377 (2017).  
+(12) Chan, M. W. Y., & Viswanathan, S. Recent progress on developing exogenous monocyte/macrophage-based therapies for inflammatory and degenerative diseases. Cytotherapy 21, 393-415 (2019).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[108, 78, 888, 870]]<|/det|>
+(13) Brudno, J. N., & Kochenderfer, J. N. Chimeric antigen receptor T-cell therapies for lymphoma. Nat. Rev. Clin. Oncol. 15, 31-46 (2018).  
+(14) Boese, A. C. et al. Neural stem cell therapy for subacute and chronic ischemic stroke. Stem Cell. Res. Ther. 9, 154 (2018).  
+(15) Aijaz, A. et al. Biomanufacturing for clinically advanced cell therapies. Nat. Biomed. Eng. 2, 362-376 (2018).  
+(16) Adan, A. et al. Flow cytometry: basic principles and applications. Crit. Rev. Biotechnol. 37, 163-176 (2017).  
+(17) Kozawa, D. et al. A fiber optic interface coupled to nanosensors: applications to protein aggregation and organic molecule quantification. ACS Nano 14, 10141-10152 (2020).  
+(18) Ugele, M. et al. Label-free high-throughput leukemia detection by holographic microscopy. Adv. Sci. 5, 1800761 (2018).  
+(19) Singh, D. K. et al. Label-free, high-throughput holographic screening and enumeration of tumor cells in blood. Lab Chip. 17, 2920 (2017).  
+(20) Zhang, Y. et al. Motility-based label-free detection of parasites in bodily fluids using holographic speckle analysis and deep learning. Light Sci. Appl. 7, 108 (2018).  
+(21) Kim, T. et al. White-light diffraction tomography of unlabelled live cells. Nat. Photon. 8, 256-263 (2014).  
+(22) Sung, Y. et al. Optical diffraction tomography for high resolution live cell imaging. Opt. Express. 17, 266-277 (2009).  
+(23) Yoon, J. et al. Identification of non-activated lymphocytes using three-dimensional refractive index tomography and machine learning. Sci. Rep. 7, 6654 (2017).  
+(24) Lei, C. et al. High-throughput imaging flow cytometry by optofluidic time-stretch microscopy. Nat. Protoc. 13, 1603-1631 (2018).  
+(25) Suzuki, Y. et al. Label-free chemical imaging flow cytometry by high-speed multicolor stimulated Raman scattering. Proc. Natl. Acad. Sci. U. S. A. 116, 15842-15848 (2019).  
+(26) Vembadi, A., Menachery, A., & Qasaimeh, M. A. Cell cytometry: Review and perspective on biotechnological advances. Front. Bioeng. Biotech. 7, 147 (2019).  
+(27) Dickinson, A. J., Armistead, P. M, & Allbritton, N. L. Automated capillary electrophoresis system for fast single-cell analysis. Anal. Chem. 85, 4797-4804 (2013).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[108, 88, 888, 870]]<|/det|>
+(28) Abraham, D.H. et al. Design of an automated capillary electrophoresis platform for single-cell analysis. \*Method. Enzymol.\* 628, 191-221 (2019).  
+(29) Vickerman, B. M. et al. Design and application of sensors for chemical cytometry. \*ACS Chem. Biol.\* 13, 1741-1751 (2018).  
+(30) Cohen, D. et al. Chemical cytometry: Fluorescence-based single-cell analysis. \*Annu. Rev. Anal. Chem.\* 1, 165-190 (2008).  
+(31) Oomen, P. E. et al. Chemical analysis of single cells. \*Anal. Chem.\* 91, 588-621 (2019).  
+(32) Munawar, A. et al. Nanosensors for diagnosis with optical, electric and mechanical transducers. \*RSC Adv.\* 9, 6793 (2019).  
+(33) Abdel-Karim, R., Reda, Y., & Abdel-Fattah, A. Review-nanostructured materials-based nanosensors. \*J. Electrochem. Soc.\* 167, 037554 (2020).  
+(34) Liu, J. et al. A sensitive and specific nanosensor for monitoring extracellular potassium levels in the brain. \*Nat. Nanotechnol.\* 15, 321-330 (2020).  
+(35) Oh, W.-J. et al. Fluorescent polymer nanoparticle for selective sensing of intracellular hydrogen peroxide. \*ACS Nano\* 6, 8516-8524 (2012).  
+(36) Liu, J. et al. A highly sensitive and selective nanosensor for near-infrared potassium imaging. \*Sci. Adv.\* 6, eaax9757 (2020).  
+(37) Reuel, N. F. et al. Emergent properties of nanosensor arrays: Applications for monitoring IgG affinity distributions, weakly Affined hypermannosylation, and colony selection for biomanufacturing. \*ACS Nano\* 7, 7472-7482 (2013).  
+(38) Kruss, S. et al. High-resolution imaging of cellular dopamine efflux using a fluorescent nanosensor array. \*Proc. Natl. Acad. Sci. U.S.A.\* 114, 1789-1794 (2017).  
+(39) Jeong, S. et al. High-throughput evolution of near-infrared serotonin nanosensors. \*Sci. Adv.\* 5, eaay3771 (2019).  
+(40) Farrera, C. et al. Carbon nanotubes as optical sensors in biomedicine. \*ACS Nano\* 11, 10637-10643 (2017).  
+(41) Iverson, N. M. et al. In vivo biosensing via tissue-localizable near-infrared-fluorescent single-walled carbon nanotubes. \*Nat. Nanotechnol.\* 8, 873-880 (2013).  
+(42) Lew, T. T. S. et al. Real-time detection of wound-induced \(\mathrm{H}_2\mathrm{O}_2\) signaling waves in plants with optical nanosensors. \*Nat. Plants\* 6, 404-415 (2020).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[108, 78, 888, 870]]<|/det|>
+(43) Engel, M. et al. Thin film nanotube transistors based on self-assembled, aligned, semiconducting carbon nanotube arrays. ACS Nano 2, 2445-2452 (2008).(44) Cho, S.-Y. et al. Continuous meter-scale synthesis of wearable tunicate cellulose/carbon nanotube fibers for high-performance wearable sensors. ACS Nano 13, 9332-9341 (2019).(45) Lee, M. A. et al. Implanted nanosensors in marine organisms for physiological biologging: Design, feasibility, and species variability. ACS Sens. 4, 32-43 (2019).(46) Zhang, J. et al. Molecular recognition using corona phase complexes made of synthetic polymers adsorbed on carbon nanotubes. Nat. Nanotechnol. 8, 959-968 (2013).(47) Dukovic, G. et al. Reversible surface oxidation and efficient luminescence quenching in semiconductor single-wall carbon nanotubes. J. Am. Chem. Soc. 126, 15269-15276 (2004).(48) Kruss, S. et al. Neurotransmitter detection using corona phase molecular recognition on fluorescent single-walled carbon nanotube sensors. J. Am. Chem. Soc. 136, 713-724 (2014).(49) Yamamoto, T. et al. Role of catalase in monocytic differentiation of U937 cells by TPA: hydrogen peroxide as a second messenger. Leukemia 23, 761-769 (2009).(50) Chateau, M. T., Rabesandratana, H. & Caravano, R. Differentiated U937 cells and human monocytes exhibit a differential production of extracellular oxygen species: \(\mathrm{O}_2\) excretion versus \(\mathrm{H}_2\mathrm{O}_2\) diffusion. FEMS Immunol. Med. Microbiol. 13, 19-28 (1996).(51) Varin, A. & Gordon, S. Alternative activation of macrophages: Immune function and cellular biology. Immunobiology 214, 630-641 (2009).(52) Satija, R. & Shalek, A. K. Heterogeneity in immune responses - From populations to single cells. Trends. Immunol. 35, 219-229 (2014).(53) Yang, H. & Gijs, M. A. M. Optical microscopy using a glass microsphere for metrology of sub-wavelength nanostructures. Microelectron. Eng. 143, 86-90 (2015).(54) Yang, H., Cornaglia, M. & Gijs, M. A. M. Photonic nanojet array for fast detection of single nanoparticles in a flow. Nano Lett. 15, 1730-1735 (2015).(55) Lee, S., Li, L. & Wang, Z. Optical resonances in microsphere photonic nanojets. J. Opt. 16, 015704 (2014).(56) Heifetz, A. et al. Photonic nanojets. J. Comput. Theor. Nanosci. 6, 1979-1992 (2009).(57) Dannhauser, D. et al. Biophysical investigation of living monocytes in flow by collaborative coherent imaging techniques. Biomed. Opt. Exp. 9, 5194-5204 (2018).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[108, 75, 888, 870]]<|/det|>
+(58) Liu, P. Y. et al. Cell refractive index for cell biology and disease diagnosis: past, present and future. Lab Chip. 16, 634-644 (2016).(59) Miccio1, L. et al. Red blood cell as an adaptive optofluidic microlens. Nat. Commun. 6, 6502 (2015).(60) Li, Y. et al. Living nanospear for near-field optical probing. ACS Nano 12, 10703-10711 (2018).(61) Li, Y., Liu, X. & Li, B. Single-cell biomagnifier for optical nanoscopes and nanotweezers. Light-Sci. Appl. 8, 61 (2019).(62) Xin, H. et al. Escherichia coli-based biophotonic waveguides. Nano Lett. 13, 3408-3413 (2013).(63) Strokotov, D. I. et al. Is there a difference between T- and B-lymphocyte morphology?. J. Biomed. Opt. 146, 064036 (2009).(64) Neto, J. C. et al. Measuring optical and mechanical properties of a living cell with defocusing microscopy. Biophys. J. 91, 1108-1115 (2006).(65) Ling, T. et al. Full-field interferometric imaging of propagating action potentials. Light-Sci. Appl. 7, 107 (2018).(66) King, D. et al. Label-free multi parameter optical interrogation of endothelial activation in single cells using a lab on a disc platform. Sci. Rep. 9, 4157 (2019).(67) LeBien, T. W. & Tedder, T. F. B lymphocytes: how they develop and function. Blood 112, 1570-4580 (2008).(68) Prasad, A. et al. Simultaneous real-time monitoring of oxygen consumption and hydrogen peroxide production in cells using our newly developed chip-type biosensor device. Front. Physiol. 7, 109 (2016).(69) Giorgio, M. et al. Hydrogen peroxide: a metabolic by-product or a common mediator of ageing signals?. Nat. Rev. Mol. Cell Biol. 8, 722-728 (2007).(70) Magnani, F. & Mattevi, A. Structure and mechanisms of ROS generation by NADPH oxidases. Curr. Opin. Struc. Biol. 59, 91-97 (2019).(71) George, P. Reaction between catalase and hydrogen peroxide. Nature 160, 41-43 (1947).(72) Shen, R. et al. Sensitive detection of single-cell Secreted \(\mathrm{H}_2\mathrm{O}_2\) by integrating a microfluidic droplet sensor and Au nanoclusters. Anal. Chem. 90, 4478-4484 (2018).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 87, 888, 664]]<|/det|>
+(73) Tjell, A. Ø. & Almdal, K. Diffusion rate of hydrogen peroxide through water-swelled polyurethane Membranes. Sens. Biosensing. Res. 21, 35-39 (2018).  
+(74) D'Autréaux, B & Toledano, M. B. ROS as signalling molecules: Mechanisms that generate specificity in ROS Homeostasis. Nat. Rev. Mol. Cell. Biol. 8, 813-824 (2007).  
+(75) Ulissi, Z. W. et al. Spatiotemporal intracellular nitric oxide signaling captured using internalized, near-infrared fluorescent carbon nanotube nanosensors. Nano Lett. 14, 4887-4894 (2014).  
+(76) Wang, S.-Y. et al. Heterogeneity of human blood monocyte: two subpopulations with different sizes, phenotypes and functions. Immunology. 77, 298-303 (1992).  
+(77) Chakraborty, P., Ghosh, D. & Basu, M. K. Macrophage protein kinase C: its role in modulating membrane microviscosity and superoxide in leishmanial infection. J. Biochem. 127, 185-190 (2000).  
+(78) Zhang, C. et al. Regulation of basal lateral membrane mobility and permeability to divalent cations by membrane associated-protein kinase C. PLOS One 8, e80291 (2013).  
+(79) Chen, L. et al. A role for protein kinase C in the regulation of membrane fluidity and Ca²⁺(+) flux at the endoplasmic reticulum and plasma membranes of HEK293 and Jurkat cells. Cell. Signalling. 23, 497-505 (2011).  
+(80) Saha, A. K. et al. Cellular cholesterol regulates monocyte deformation. J. Biomech. 52, 83-88 (2017).  
+(81) Boss, D. et al. Measurement of absolute cell volume, osmotic membrane water permeability, and refractive index of transmembrane water and solute flux by digital holographic microscopy. J. Biomed. Opt. 18, 036007 (2013).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 44, 143, 70]]<|/det|>
+## Figures  
+
+<|ref|>image<|/ref|><|det|>[[50, 95, 941, 655]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 676, 115, 695]]<|/det|>
+<center>Figure 1 </center>  
+
+<|ref|>text<|/ref|><|det|>[[42, 716, 949, 876]]<|/det|>
+Nanosensor integration with microfluidics. (a) Schematic illustration of nanosensor integration process with microfluidics using EISA. (b) Photograph of EISA process of NIM for 0 min (left) and 30 min (right). (c) Photograph of completed multi- array NIM and pristine channel. (d) Polarized Raman spectrum (G- peak) of NIM. (e) nIR images of NIM and pristine channel. (f) Magnified nIR image of NIM with single cell size resolution (20 μm) having \(\sim 720\) nIR reporter pixel. (g) Histograms of nIR pixel intensities of top and bottom NIM surfaces NIM (inset: nIR images of top and bottom surfaces). (h) nIR fluorescence spectrum of NIM. (i) nIR images of NIM with varying composition of SWNT nanosensor integration.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[50, 48, 944, 585]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[42, 602, 118, 622]]<|/det|>
+<center>Figure 2 </center>  
+
+<|ref|>text<|/ref|><|det|>[[39, 644, 936, 825]]<|/det|>
+In- vitro chemical detection performances of NIM. (a) nIR spectrum of NIM with H2O2 solution flowing (1 \(\mu \mathrm{M}\) , 1 \(\mu \mathrm{L} / \mathrm{min}\) ). (b) nIR images of NIM before and after H2O2 flowing (1 M, 10 \(\mu \mathrm{L} / \mathrm{min}\) , 10 min). (c) Schematic illustration of H2O2 detection mechanism of SWNT/(GT)15 nanosensor. (d) Real- time nIR response of NIM with various concentration (10- 6, 10- 5, 10- 4, 10- 3, 10- 2, 10- 1, 100 M) of H2O2 injection (10 min). (e) Maximum response amplitude and (f) response time of NIM with various concentration of H2O2. The data represent the mean value of 250 X 350 \(\mu \mathrm{m}2\) NIM measurement. (g) nIR snapshots and intensity histogram (fire scale, ImageJ) of NIM with single cell size resolution (20 \(\mu \mathrm{m}\) ) after 10 min flowing of various concentration of H2O2.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[55, 50, 890, 788]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 802, 117, 821]]<|/det|>
+<center>Figure 3 </center>  
+
+<|ref|>text<|/ref|><|det|>[[42, 842, 951, 956]]<|/det|>
+Cellular lensing effect. (a) Instrumental setup for NCC implementation: schematic illustration (left) and a photograph (right). (b) nIR images of human monocytes flowing (0.5 \(\mu \mathrm{L} / \mathrm{min}\) ) NIM. (c) Magnified nIR image of single monocyte in NIM (inset: OM image of single monocyte). (d) FDTD numerical modeling for photonic nanojet and fitting with experimental cellular lensing profile (nc/nm = 1.04, \(\lambda = 1 \mu \mathrm{m}\) ). (e) nIR lensing profiles of a single cell with various focusing points from 5 to 100 \(\mu \mathrm{m}\) along Z- stage. nIR lensing
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 44, 936, 134]]<|/det|>
+effects of (f) various live cells and (g) reference micro-particles (top-to-bottom: schematics, OM, nIR images, lensing profiles). (h) FWHM and (i) enhancement factors of various cells with numerical model. Data are mean \(\pm \sigma\) , with \(\text{neel} = 10\) . (j) Schematic illustrations for different lensing behavior of a high RI cell (left) and a low RI cell (right). Scale bars: \(20 \mu \text{m}\) .  
+
+<|ref|>image<|/ref|><|det|>[[45, 140, 950, 800]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[42, 819, 118, 838]]<|/det|>
+<center>Figure 4 </center>  
+
+<|ref|>text<|/ref|><|det|>[[42, 860, 943, 949]]<|/det|>
+Real- time chemical efflux monitoring using the cellular lensing effect. (a) Time- series nIR images of a stationary single monocyte with different immune activation states (- PMA, + PMA, + PMA & catalase). (b) Real- time nIR intensity variations of the cells with different activation states. (c) Schematic illustrations of H2O2 efflux monitoring mechanism with nIR lensing effect. (d) 3D diffusion and reaction kinetic
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 45, 928, 113]]<|/det|>
+modeling for translation of measured nIR signals to real- time local H2O2 concentration. (e) Real- time H2O2 efflux profiles of each single monocyte estimated by the model. 16- color scalebars represent nIR intensity from white (16833) to dark blue (0). Scale bars: \(20 \mu \mathrm{m}\) .  
+
+<|ref|>image<|/ref|><|det|>[[52, 115, 936, 840]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[42, 850, 118, 870]]<|/det|>
+<center>Figure 5 </center>  
+
+<|ref|>text<|/ref|><|det|>[[42, 892, 941, 959]]<|/det|>
+NCC for monitoring of multimodal immune response heterogeneities. a) Schematics and nIR images of NCC setup with distinct activation of human monocytes (-PMA and +PMA). (b) Automatic nIR image analysis using computational code for cell data extractions. (c) NCC cytometry plots of H2O2 efflux rate
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[41, 44, 940, 202]]<|/det|>
+vs biophysical parameters ((c1) size (2D projected area), (c2) eccentricity, (c3) RI) of two monocytes populations. Data are ncell = 413 for - PMA, ncell = 414 for +PMA from n = 6 biologically independent samples. (d) NCC distribution curves of H2O2 efflux rates with data from commercial assay kit. (e) NCC cytometry plots for cell biophysical parameters ((e1) eccentricity vs size, (e2) RI vs eccentricity, (e3) size vs RI). (f) NCC distribution curves of each biophysical parameters ((f1) size, (f2) eccentricity, (f3) RI). (g) Schematics illustrations for cell properties variations of human monocyte populations with immune activations. Scale bars: \(20 \mu \mathrm{m}\) .  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 224, 310, 251]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 275, 765, 296]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 313, 723, 360]]<|/det|>
+- VideoS1.avi- RevisedSupportingInformationMichaelStranoNatureNanotechnology.docx
+
+<--- Page Split --->

preprint/preprint__487dd727461ac5ce1675c5210f0d2bb09de73dbbde2a7e84d0c10a7c04a6fbfd/images_list.json

@@ -0,0 +1,115 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1: Schematic of the \"peripheral kill-the-winner\" hypothesis. In the absence of phage, we expect that faster-growing antibiotic sensitive (AS) cells (cyan) will displace slower-growing antibiotic resistant (AR) cells (magenta) along the biomass periphery. In the presence of phage, however, we expect that the slower-growing AR cells will persist with the faster-growing AS cells. This is because the faster-growing AS cells will disproportionately occupy the biomass periphery, and they will therefore be more susceptible to phage lysis. This will increase the removal of AS cells from the biomass and counteract the benefits of their faster growth relative to the slower-growing AR cells, thus increasing the persistence of the slower-growing AR cells.",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        92,
+        825,
+        380
+      ]
+    ],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2: Phage lysis increases the persistence of slower-growing AR cells. a, Representative CLSM images of co-cultures of strains AS (cyan) and \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Tet}}\\) (magenta) (upper images) or of strains AS (cyan) and \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Str}}\\) (magenta) (lower images) in the absence or presence of phage T6. We imaged the biomass after ten days of incubation in the absence of antibiotic pressure. b, The proportion of the total biomass area occupied by strains \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Tet}}\\) or \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Str}}\\) when grown in coculture with strain AS in the absence or presence of phage T6. c, The biomass diameters of strains AS, \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Tet}}\\) and \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Str}}\\) when grown in monoculture in the absence of phage T6. d, The biomass diameters of co-cultures of strains AS and \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Tet}}\\) or strains AS and \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Str}}\\) in the absence or presence of phage T6. e, The biomass area (population size) of strain \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Tet}}\\) when grown in co-culture with strain AS in the absence or presence of phage T6. f, The proportion of \\(\\mathsf{AR}_{\\mathsf{C},\\mathsf{Tet}}\\) cells within co-cultures as a function of the radial distance from the centroid of the biomass. For b–f, each data point is an independent experimental replicate ( \\(n = 5\\) ), the black data points are for experiments in the absence of phage T6, and the green data points are for experiments in the presence of phage T6. For b–e, the \\(p\\) -values are for two-sample two-sided Welch tests.",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        370,
+        880,
+        616
+      ]
+    ],
+    "page_idx": 7
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3: Fitness cost and probability of phage lysis determine the persistence of slower-growing AR cells. a, Representative individual-based computational simulations of co-cultures of strains AS (cyan) and AR (magenta) in the absence or presence of phage lysis with different fitness costs of antibiotic resistance and probabilities of phage lysis. The images are the outputs at the last simulation time step. b, The proportion of AR cells as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. c, The number of AS cells removed by phage lysis as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. d, The number of AR cells removed by phage lysis as a function of the fitness cost of antibiotic resistance for different probabilities for phage lysis. For b–d, each data point is an independent simulation (n = 4), the lines connect the mean values, and the shaded regions are one standard deviation.",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        88,
+        880,
+        670
+      ]
+    ],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4: Phage lysis increases the persistence of slower-growing AR cells in the face of spontaneously generated AS cells. a, Representative CLSM images of cultures containing strain \\(\\mathsf{AR}_{\\mathsf{P},\\mathsf{CHI}}\\) (magenta) in the absence (upper image) or presence (lower image) of phage T6. Nonfluorescent regions are composed of \\(\\mathsf{AR}_{\\mathsf{P},\\mathsf{CHI}}\\) cells that lost plasmid pEF001 and became AS cells. We took the images after ten days of incubation at \\(21^{\\circ}\\mathrm{C}\\) in the absence of antibiotic pressure. b, The proportion of the total biomass area occupied by strain \\(\\mathsf{AR}_{\\mathsf{P},\\mathsf{CHI}}\\) in the absence or presence of phage T6. c, The biomass diameter in the absence or presence of phage T6. d, The proportion of \\(\\mathsf{AR}_{\\mathsf{P},\\mathsf{CHI}}\\) cells as a function of the radial distance from the centroid of the biomass. For b–d, each data point is an independent experimental replicate (n = 5), the black data points are for experiments in the absence of phage T6, and the green data points are for experiments in the presence of phage T6. For b,c, the \\(p\\) -values are for two-sample two-sided Welch tests.",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        90,
+        870,
+        528
+      ]
+    ],
+    "page_idx": 13
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Fig. 5: Properties of spontaneously emerging AS cells determine the persistence of slower-growing AR cells. a,b, Representative individual-based computational simulations of strain AR (magenta) in the (a) absence or (b) presence of phage lysis with different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance. If AR cells undergo a genetic change that causes them to lose antibiotic resistance, such as segregational loss of a plasmid, they are relieved of the fitness cost and become AS cells (grey). The images are the outputs at the last simulation time step. c, The proportion of AR cells as a function of the fitness cost of antibiotic resistance and the probability of losing antibiotic resistance. Each data point is an independent simulation (n = 4), the black data points are in the absence of phage, and the green data points are in the presence of phage.",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        88,
+        840,
+        690
+      ]
+    ],
+    "page_idx": 15
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Fig. 6: Phage lysis maintains strain diversity. Data are for individual-based computational simulations of the surface-associated growth of co-cultures consisting of between two and ten strains, where each strain has a different growth rate ranging between zero and one. The growth rate of one strain was set to a value of one and the growth rates of the other strains were sampled from a uniform distribution of growth rates. a-d, The strain diversity metrics were quantified at a fixed total biomass size and include (a) strain richness, (b) Shannon diversity, (c) Simpson diversity, and (d) evenness. For b-d, the boxplots identify the mean values, interquartile ranges, and outliers for a total of 473 independent pairs of simulations (between 35-65 replicates for each number of cell types). The black boxplots and data points are in the absence of phage and the green boxplots and data points are in the presence of phage. The \\(p\\) -values are for two-sided paired t-tests with a Bonferroni correction.",
+    "footnote": [],
+    "bbox": [
+      [
+        123,
+        118,
+        880,
+        520
+      ]
+    ],
+    "page_idx": 18
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_4.jpg",
+    "caption": "Extended Data Fig. 4: Properties of spontaneously emerging AS cells determine the persistence of slower-growing AR strains at a fixed total biomass size. All data are for individual-based computational simulations of strain AR in the absence or presence of phage lysis. The proportion of AR cells is calculated for different fitness costs of antibiotic resistance, probabilities of losing antibiotic resistance, and probabilities of phage lysis. Each data point is an independent simulation (n = 4), the black data points are in the absence of phage, and the green data points are in the presence of phage.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 35
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_5.jpg",
+    "caption": "Extended Data Fig. 5: Faster-growing emergent AS cells are disproportionately lysed by phage.",
+    "footnote": [],
+    "bbox": [
+      [
+        118,
+        90,
+        850,
+        721
+      ]
+    ],
+    "page_idx": 36
+  }
+]

preprint/preprint__487dd727461ac5ce1675c5210f0d2bb09de73dbbde2a7e84d0c10a7c04a6fbfd/preprint__487dd727461ac5ce1675c5210f0d2bb09de73dbbde2a7e84d0c10a7c04a6fbfd.mmd

@@ -0,0 +1,485 @@
+
+# Phage lysis facilitates the maintenance of costly antibiotic resistance in the absence of antibiotic pressure  
+
+David Johnson  david.johnson@eawag.ch  
+
+Swiss Federal Institute of Aquatic Science and Technology https://orcid.org/0000- 0002- 6728- 8462  Chujin Ruan  Eawag: Das Wasserforschungs- Institut des ETH- Bereichs https://orcid.org/0009- 0009- 8605- 7107  Deepthi Vinod  Swiss Federal Institute of Aquatic Science and Technology  
+
+## Article  
+
+Keywords: Antibiotic resistance, phage- bacteria interactions, plasmid dynamics, spatial organization, biofilm  
+
+Posted Date: February 7th, 2025  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 5879387/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License.  Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on July 1st, 2025. See the published version at https://doi.org/10.1038/s41467- 025- 61055- y.
+
+<--- Page Split --->
+
+## 1 Title  
+
+2 Phage lysis facilitates the maintenance of costly antibiotic resistance in the absence of antibiotic pressure  
+
+## Authors  
+
+6 Chujin Ruan \(^{1\# *}\) , Deepthi P. Vinod \(^{1,2\#}\) , David R. Johnson \(^{1,3*}\)  
+
+## Affiliations  
+
+9 1Department of Environmental Microbiology, Swiss Federal Institute of Aquatic Science and Technology (Eawag), 8600 Dübendorf, Switzerland; 2Department of Environmental Systems Science, Swiss Federal Institute of Technology (ETH), 8092 Zürich, Switzerland; 3Institute of Ecology and Evolution, University of Bern, 3012 Bern, Switzerland.  
+
+## \* Correspondence  
+
+15 David R. Johnson, david.johnson@eawag.ch; Chujin Ruan, chujin.ruan@eawag.ch  
+
+16 \*These authors contributed equally: Chujin Ruan, Deepthi P. Vinod
+
+<--- Page Split --->
+
+## Abstract  
+
+The persistence of antibiotic resistant (AR) bacteria in the absence of antibiotic pressure raises a paradox regarding the fitness costs associated with antibiotic resistance. These fitness costs should slow the growth of AR bacteria and cause them to be displaced by faster- growing antibiotic sensitive (AS) counterparts. Yet, even in the absence of antibiotic pressure, slower- growing AR bacteria can persist for prolonged periods of time. Here, we demonstrate a mechanism that can explain this apparent paradox. We hypothesize that lytic phage can modulate bacterial spatial organization to facilitate the persistence of slower- growing AR bacteria. Using surface- associated growth experiments with the bacterium Escherichia coli in conjunction with individual- based computational simulations, we show that phage disproportionately lyse the faster- growing AS counterpart cells located at the biomass periphery via a "peripheral kill- the- winner" dynamic. This enables the slower- growing AR cells to persist even when they are susceptible to the same phage. This phage- mediated selection is accompanied by enhanced bacterial diversity, further emphasizing the role of phage in shaping the assembly and evolution of bacterial systems. The mechanism is potentially relevant for any antibiotic resistance genetic determinant and has tangible implications for the management of bacterial populations via phage therapy.  
+
+## Keywords  
+
+Antibiotic resistance, phage- bacteria interactions, plasmid dynamics, spatial organization, biofilm
+
+<--- Page Split --->
+
+## Main  
+
+The rise of antibiotic resistance is an urgent threat to global public health and is causing increasing morbidity and mortality worldwide<sup>1,2</sup>. A broadly applied strategy to combat this threat is the judicious use and disposal of antibiotics. This strategy anticipates that antibiotic resistant (AR) bacteria will be displaced by antibiotic sensitive (AS) counterparts in the absence of antibiotic pressure<sup>3,4</sup>. This is because antibiotic resistance often imposes fitness costs that slow the growth of AR bacteria. This provides a growth advantage to AS counterparts that lose antibiotic resistance through genetic mutation or plasmid loss<sup>3,4</sup>. However, despite the implementation of this strategy<sup>5,6,7</sup>, AR bacteria continue to persist in the absence of antibiotic pressure in a myriad of environments<sup>8,9</sup>.  
+
+Several mechanisms have been proposed to explain how slower- growing AR bacteria can persist in the face of faster- growing AS counterparts in the absence of antibiotic pressure<sup>10,11,12,13</sup>. These mechanisms include compensatory mutations that reduce the fitness costs associated with antibiotic resistance or their associated genetic elements<sup>14,15</sup>, co- selection of linked traits<sup>16,17,18,19</sup>, and horizontal transfer of antibiotic resistance genetic determinants<sup>20,21,22</sup>. One aspect that can affect the dynamics of antibiotic resistance determinants is bacterial spatial organization<sup>23,24,25,26,27,28,29,30,31</sup>. Experiments and theoretical considerations have illustrated how spatial processes can increase the persistence of neutral and even deleterious genetic mutations at the periphery of growing biomass<sup>32,33,34</sup>. Whether these processes can explain the persistence of slower- growing AR bacteria in the face of faster- growing AS counterparts, however, remains unclear. Because surface- associated bacterial systems are important reservoirs of antibiotic resistance genetic determinants<sup>35</sup>, understanding how bacterial spatial organization and dynamics affect the persistence of antibiotic resistance could set the stage for developing more effective bacterial management and intervention strategies.  
+
+We hypothesize here that phage lysis can modulate bacterial spatial organization to increase the persistence of slower- growing AR bacteria in the face of faster- growing AS counterparts. More
+
+<--- Page Split --->
+
+precisely, we hypothesize that lytic phage can mediate a "peripheral kill- the- winner" dynamic; faster- growing AS counterpart cells are disproportionally lysed to a greater extent than slower- growing AR cells, consequently increasing the persistence of antibiotic resistance (Fig. 1). Our hypothesis is grounded in the fundamental principle that, for surface- associated bacterial systems, biomass growth is primarily driven by cells located at the biomass periphery where resources supplied from the environment are plentiful \(^{36,37,38}\) . Because of their differences in growth rates, slower- growing AR bacteria will be disproportionately located behind the biomass periphery while faster- growing AS counterpart cells will disproportionately occupy the biomass periphery (Fig. 1). Due to mass transfer limitations, phage will predominantly lyse cells located at the biomass periphery \(^{39,40}\) , which will disproportionately be AS counterpart cells (Fig 1). The faster- growing AS counterpart cells will therefore undergo more vigorous phage lysis, thereby offsetting their growth advantage and increasing the persistence of slower- growing AR cells (Fig. 1).  
+
+Our hypothesis not only provides an explanation for the persistence of slower- growing AR cells in the face of faster- growing AS counterpart cells, but also makes predictions regarding dynamic environments where spontaneous genetic changes that alter antibiotic resistance profiles can occur concurrently with phage lysis. These alterations in antibiotic resistance profiles can occur through genetic mutations or through plasmid loss, both of which can relieve cells of the fitness costs associated with antibiotic resistance \(^{41,42,43}\) . After such genetic changes occur, the faster- growing AS counterpart cells that emerge will grow towards and disproportionately occupy the biomass periphery, consequently making them more susceptible to phage lysis and increasing the persistence of slower- growing AR cells. We therefore predict that, even if AR cells are capable of reverting to AS cells via spontaneous genetic changes, the "peripheral kill- the- winner" dynamic will rapidly establish itself to disproportionately lyse the AS cells and promote the persistence of the slower- growing AR cells, thus providing a general mechanism for how antibiotic resistance genetic determinants can persist amidst ongoing evolutionary processes.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig. 1: Schematic of the "peripheral kill-the-winner" hypothesis. In the absence of phage, we expect that faster-growing antibiotic sensitive (AS) cells (cyan) will displace slower-growing antibiotic resistant (AR) cells (magenta) along the biomass periphery. In the presence of phage, however, we expect that the slower-growing AR cells will persist with the faster-growing AS cells. This is because the faster-growing AS cells will disproportionately occupy the biomass periphery, and they will therefore be more susceptible to phage lysis. This will increase the removal of AS cells from the biomass and counteract the benefits of their faster growth relative to the slower-growing AR cells, thus increasing the persistence of the slower-growing AR cells. </center>  
+
+To test our hypothesis, we performed surface- associated growth experiments with isogenic derivatives of the bacterium Escherichia coli MG1655 in the absence or presence of the lytic phage T6. The parental strain, which we refer to as strain AS for "antibiotic sensitive", is sensitive to all of the antibiotics used in this study. We then obtained antibiotic resistant variants of strain AS, which we refer to as AR strains for "antibiotic resistant". The AR strains contain antibiotic resistance determinants that are either single genetic mutations located on the chromosome (strain \(\mathrm{AR}_{\mathrm{C},\mathrm{Tet}}\) is resistant to tetracycline while strain \(\mathrm{AR}_{\mathrm{C},\mathrm{Str}}\) is resistant to streptomycin) or are genes located on the non- transmissible plasmid pEF001 (strain \(\mathrm{AR}_{\mathrm{P},\mathrm{Chl}}\) is resistant to chloramphenicol) that can be spontaneously lost during cell division. We further introduced fluorescent protein- encoding genes into the chromosomes of the strains and into plasmid pEF001 so that we can distinguish them when grown together. We then assembled
+
+<--- Page Split --->
+
+strain AS with either strain \(\mathsf{AR}_{\mathsf{C,Tet}}\) or \(\mathsf{AR}_{\mathsf{C,Str}}\) into co- cultures, propagated them across nutrient- amended agar surfaces in the absence of antibiotic pressure, and quantified the strain abundances and emergent spatial patterns using confocal laser- scanning microscopy (CLSM). We also propagated strain \(\mathsf{AR}_{\mathsf{P,Chl}}\) alone across nutrient- amended agar surfaces in the absence of antibiotic pressure, tracked the spontaneous emergence of plasmid- free AS cells, and quantified the emergent spatial patterns with CLSM. In both cases, we expected that the AR strains, all of which grow slower than the AS strain, would have increased persistence in the presence of phage T6 via our proposed "peripheral kill the winner" dynamic (Fig. 1). Finally, we complemented our experiments with individual- based computational simulations to identify general mechanisms for how phage lysis can increase the persistence of slower- growing AR strains.  
+
+## Results  
+
+## Phage lysis increases the persistence of slower-growing AR cells  
+
+We first quantified how phage lysis affects the persistence of slower- growing AR cells in the face of faster- growing AS counterparts. To accomplish this, we performed surface- associated growth experiments in the absence or presence of phage T6. We mixed strain AS with either strain \(\mathsf{AR}_{\mathsf{C,Tet}}\) or \(\mathsf{AR}_{\mathsf{C,Str}}\) , both of which contain single chromosomal mutations that bestow resistance to tetracycline or streptomycin, at a 1:1 initial cell ratio. To distinguish them, strain AS expressed a green fluorescent protein- encoding gene (falsely colored to cyan in the images) located on the chromosome while strains \(\mathsf{AR}_{\mathsf{C,Tet}}\) and \(\mathsf{AR}_{\mathsf{C,Str}}\) expressed a red fluorescent protein- encoding gene (falsely colored to magenta in the images) located on the chromosome. We then grew the cocultures across nutrient- amended agar surfaces in the absence of antibiotic pressure and quantified the patterns of spatial organization that emerged with CLSM (Fig. 2a).
+
+<--- Page Split --->
+
+In the absence of phage T6, we found that strain AS had a clear competitive advantage over both strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) (Fig. 2a). After ten days of incubation, strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) constituted less than \(0.7\%\) of the total biomass area even though they constituted approximately \(50\%\) of the initial inoculum (Fig. 2b). We attribute this effect to the fitness cost associated with the antibiotic resistance determinants for tetracycline and streptomycin. This is evident from our experiments, where the total biomass areas of strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) or \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) when grown alone in the absence of antibiotic pressure were both significantly smaller than the area of strain AS when grown alone (two- sample two- sided Welch tests; \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) , \(p = 1.2 \times 10^{- 5}\) , \(n = 5\) ; \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) , \(p = 6.1 \times 10^{- 7}\) , \(n = 5\) ) (Fig. 2c).  
+
+![](images/Figure_2.jpg)
+
+<center>Fig. 2: Phage lysis increases the persistence of slower-growing AR cells. a, Representative CLSM images of co-cultures of strains AS (cyan) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) (magenta) (upper images) or of strains AS (cyan) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) (magenta) (lower images) in the absence or presence of phage T6. We imaged the biomass after ten days of incubation in the absence of antibiotic pressure. b, The proportion of the total biomass area occupied by strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) or \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) when grown in coculture with strain AS in the absence or presence of phage T6. c, The biomass diameters of strains AS, \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) when grown in monoculture in the absence of phage T6. d, The biomass diameters of co-cultures of strains AS and \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) or strains AS and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) in the absence or presence of phage T6. e, The biomass area (population size) of strain \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) when grown in co-culture with strain AS in the absence or presence of phage T6. f, The proportion of \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) cells within co-cultures as a function of the radial distance from the centroid of the biomass. For b–f, each data point is an independent experimental replicate ( \(n = 5\) ), the black data points are for experiments in the absence of phage T6, and the green data points are for experiments in the presence of phage T6. For b–e, the \(p\) -values are for two-sample two-sided Welch tests. </center>
+
+<--- Page Split --->
+
+In the presence of phage T6, we found that the slower-growing strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) can persist in the face of the faster- growing strain AS (Fig. 2a). The proportion of the area occupied by strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) was nearly 10- fold larger when in the presence of phage T6 (two- sample two- sided Welch test; \(p = 1.1\times 10^{- 2}\) , \(n = 5\) ) (Fig. 2b). Moreover, even though the presence of phage T6 significantly reduced the overall biomass area of all the strains (two- sample two- sided Welch tests; \(p< 1.0\times\) \(10^{- 7}\) , \(n = 5\) ) (Fig. 2d), the absolute population size of the slower- growing strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) increased (two- sample two- sided Welch test; \(p = 3.5\times 10^{- 2}\) , \(n = 5\) ) (Fig. 2e). We further analyzed the spatial positionings of the \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) and AS cells. In the absence of phage T6, the faster- growing AS cells dominated the biomass periphery while the slower- growing \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) cells were positioned behind the periphery where nutrients supplied from the environment were depleted (Fig. 2f). In the presence of phage T6, in contrast, the slower- growing \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) cells persisted at the biomass periphery where nutrients supplied from the environment were plentiful (Fig. 2f).  
+
+In contrast with strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) , we found that the slower- growing strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{S}\mathsf{T}\mathsf{R}}}\) was unable to persist when co- cultured with the faster- growing strain AS in the absence of antibiotic pressure regardless of whether phage T6 was present or not (Fig. 2a,b). We attribute this to the larger growth rate difference between strains AS and \(\mathsf{AR}_{\mathsf{C}_{\mathsf{S}\mathsf{T}\mathsf{R}}}\) when compared to the growth rate difference between strains AS and \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) (i.e., for our experimental system, the fitness cost of streptomycin resistance is significantly greater than that of tetracycline resistance) (two- sample two- sided Welch test; \(p = 8.2\times 10^{- 7}\) , \(n = 5\) ) (Fig. 2c). Thus, while phage lysis can increase the persistence of slower- growing AR strains in the face of faster- growing AS strains, this is potentially only true if the fitness cost of antibiotic resistance is not excessively large.  
+
+## Fitness cost of antibiotic resistance determines the persistence of AR cells  
+
+To test the notion that the effect of phage lysis on the persistence of slower- growing AR cells depends on the fitness cost of antibiotic resistance (i.e., that excessively large fitness costs can
+
+<--- Page Split --->
+
+eliminate the positive effect of phage lysis), we adapted an individual- based computational model that allows us to simulate co- culture growth across nutrient- amended surfaces<sup>40,44,45</sup>. We performed simulations of co- cultures composed of a faster- growing AS strain (cyan) and a slower- growing AR strain (magenta) and varied the fitness cost of antibiotic resistance and the probability of phage lysis (Fig. 3a and Video 1). As with our experiments, the faster- growing AS cells do not emerge spontaneously during the simulations; rather, they are already present in the inoculum. Also consistent with our experiments, we used a 1:1 initial cell ratio of the two strains and conducted the simulations in the absence of antibiotic pressure. We simulated phage lysis by removing peripheral cells at varying probabilities (between 0.01 and 0.05), which has been found to be a reasonable approximation of the effect of phage lysis<sup>39,40</sup>.  
+
+We found that the effect of phage lysis on the persistence of slower- growing AR cells does indeed depend on the fitness cost of antibiotic resistance. When we set the probability of phage lysis to zero, the AR strain had a clear growth disadvantage due to the fitness cost associated with antibiotic resistance (Fig. 3a), which is consistent with our experimental results (Fig. 2a). Overall, the proportion of AR cells at the final simulation time- step decreased as the fitness cost of antibiotic resistance increased (Spearman rank correlation test; \(\mathrm{rho} = - 0.99\) , \(p_{BC} = 8.1 \times 10^{- 34}\) ) (Fig. 3b). When we then set the probability of phage lysis to a positive number, we observed increased persistence of the slower- growing AR cells (Pearson correlation test; \(\mathrm{rho} = 0.94\) , \(p_{BC} = 1.0 \times 10^{- 11}\) ) (Fig. 3a,b). The effect size had a unimodal relationship with the fitness cost of antibiotic resistance, where the maximum beneficial effect of phage lysis on the persistence of AR cells occurred at a fitness cost of antibiotic resistance of 0.1 (two- sample two- sided Welch tests; \(\mathrm{cost} = 0.0 \mathrm{vs. cost} = 0.1\) , \(p_{BC} = 4.8 \times 10^{- 4}\) , \(\mathrm{n} = 4\) ; \(\mathrm{cost} = 0.1 \mathrm{vs. cost} = 0.9\) , \(p_{BC} = 6.4 \times 10^{- 3}\) , \(\mathrm{n} = 4\) ) (Fig. 3b and Extended Data Fig. 1). The effect size of phage lysis then declined as the fitness cost of antibiotic resistance increased (Fig. 3b and Extended Data Fig. 1). Thus, the slower- growing AR cells persisted most effectively when the fitness cost of antibiotic resistance was relatively low (Fig. 3b and Extended Data Fig. 1), which is consistent with our experimental observations (Fig. 2a,b).
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig. 3: Fitness cost and probability of phage lysis determine the persistence of slower-growing AR cells. a, Representative individual-based computational simulations of co-cultures of strains AS (cyan) and AR (magenta) in the absence or presence of phage lysis with different fitness costs of antibiotic resistance and probabilities of phage lysis. The images are the outputs at the last simulation time step. b, The proportion of AR cells as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. c, The number of AS cells removed by phage lysis as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. d, The number of AR cells removed by phage lysis as a function of the fitness cost of antibiotic resistance for different probabilities for phage lysis. For b–d, each data point is an independent simulation (n = 4), the lines connect the mean values, and the shaded regions are one standard deviation. </center>
+
+<--- Page Split --->
+
+## Mechanism for how phage lysis increases the persistence of AR cells  
+
+To identify a plausible mechanism for how phage lysis can increase the persistence of slower- growing AR cells in the face of faster- growing AS counterparts, we again used our individual- based computational model to count the numbers of AR and AS cells that were removed from the biomass as a consequence of phage lysis. We found that more AS cells than AR cells were removed via phage lysis when there was a fitness cost of antibiotic resistance (two- sample twosided Welch tests; \(p_{BC} = 4.6 \times 10^{- 3}\) , \(n = 4\) ) (Fig. 3c,d). We refer to this outcome as “peripheral kill- the- winner”; faster- growing AS cells have a growth advantage that allows them to disproportionately occupy the biomass periphery, but cells at the biomass periphery are also more susceptible to removal via phage lysis. The disproportional removal of faster- growing AS cells diminishes the benefits of their faster growth rates, allowing slower- growing AR cells to persist. This concept is supported by our individual- based computational simulations where we varied the probability of phage lysis. As the probability of phage lysis increased, the number of faster- growing AS cells that were removed from the biomass also increased (Spearman rank correlation test; \(r_{ho} = 0.90\) , \(p_{BC} = 1.2 \times 10^{- 8}\) ) (Fig. 3c), which correspondingly increased the persistence of slower- growing AR cells (Pearson correlation test; \(r_{ho} = 0.94\) , \(p_{BC} = 1.0 \times 10^{- 11}\) ) (Fig. 3d). While we analyzed all of our simulations at a fixed simulation time for the data presented in Fig. 3, all of our outcomes remain valid when we analyzed our simulations at a fixed total biomass size (Extended Data Fig. 2). Our main outcomes are therefore robust to the simulation end- point.  
+
+## Phage lysis increases the persistence of slower-growing AR strains in the face of spontaneously emerging AS cells  
+
+We next tested whether our outcomes remain valid when antibiotic resistance is spontaneously lost during the experiment. To test this, we performed surface- associated growth experiments
+
+<--- Page Split --->
+
+with strain \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) , which contains the non- transmissible plasmid pEF001 that encodes for chloramphenicol resistance and green fluorescent protein (falsely colored magenta in our images). If plasmid pEF001 is lost from an \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cell during cell division, the cell will revert to a non- fluorescent and faster- growing version that we refer to as an AS cell (uncolored). We can therefore propagate the slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells alone on nutrient- amended agar surfaces, track the spontaneous emergence and proliferation of faster- growing AS cells with CLSM, and quantify the persistence of \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells in the face of the newly formed AS cells.  
+
+We found that phage lysis can indeed increase the persistence of slower- growing AR cells in the face of spontaneously formed and faster- growing AS cells. In the absence of phage T6, we observed substantial loss of plasmid pEF001 within the biomass (Fig. 4a). This corresponded to the proliferation of faster- growing AS cells and a decline in the proportion of slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells from \(100\%\) in the initial inoculum to only \(67\%\) of the total biomass after ten days of incubation (Fig. 4b). Thus, the plasmid- free faster- growing AS cells had a clear competitive advantage over the slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells. When we then added phage T6 to the cocultures, the proportion of slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells declined by a significantly smaller extent from \(100\%\) in the initial inoculum to \(94\%\) after ten days of incubation (two- sample two- sided Welch test; \(p = 3.6 \times 10^{- 4}\) , \(n = 5\) ) (Fig. 4a,b). Thus, the presence of phage T6 reduced the ability of the faster- growing AS cells to establish over the slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells. This was true even though the total biomass area declined (two- sample two- sided Welch test; \(p = 6.1 \times 10^{- 8}\) , \(n = 5\) ) (Fig. 4c). Finally, \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells were more abundant along the biomass periphery when phage T6 was present (two- sample two- sided Welch tests; \(p < 1.5 \times 10^{- 3}\) , \(n = 5\) ) (Fig. 4d). Thus, phage T6 allows the slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells to better occupy the biomass periphery where resources are plentiful, which is consistent with our proposed “peripheral kill- the- winner” dynamic (Fig. 1). These outcomes remain valid across various environmental conditions, including anoxic environments and different incubation temperatures (Extended Data Fig. 3).
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 4: Phage lysis increases the persistence of slower-growing AR cells in the face of spontaneously generated AS cells. a, Representative CLSM images of cultures containing strain \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) (magenta) in the absence (upper image) or presence (lower image) of phage T6. Nonfluorescent regions are composed of \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells that lost plasmid pEF001 and became AS cells. We took the images after ten days of incubation at \(21^{\circ}\mathrm{C}\) in the absence of antibiotic pressure. b, The proportion of the total biomass area occupied by strain \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) in the absence or presence of phage T6. c, The biomass diameter in the absence or presence of phage T6. d, The proportion of \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells as a function of the radial distance from the centroid of the biomass. For b–d, each data point is an independent experimental replicate (n = 5), the black data points are for experiments in the absence of phage T6, and the green data points are for experiments in the presence of phage T6. For b,c, the \(p\) -values are for two-sample two-sided Welch tests. </center>  
+
+Mechanism for how phage lysis can increase the persistence of AR cells in the face of spontaneously emerging AS cells
+
+<--- Page Split --->
+
+To identify a plausible mechanism for how phage lysis can increase the persistence of AR cells despite the spontaneous emergence of faster- growing AS cells, we incorporated the spontaneous loss of antibiotic resistance into our individual- based computational model. The loss of antibiotic resistance in our model is generic and could occur via a genetic mutation or, as in our experiments, by plasmid loss. We then counted the number of AS cells that emerge from AR cells over the course of the simulations in the absence of antibiotic pressure. We set the entire initial population to be AR cells (magenta), each of which can spontaneously transform into a faster- growing AS cell (grey) according to a defined probability. We then varied the probability of losing antibiotic resistance to be between 0.01 and 0.03 and the fitness cost of antibiotic resistance to be a reduction in the growth rate between \(5\%\) and \(30\%\) . We simulated phage lysis as described in our simulations for chromosomal antibiotic resistance genetic determinants.  
+
+When we analyzed our simulations at a fixed simulation time, we found that phage lysis does indeed increase the persistence of slower- growing AR cells in the face of faster- growing AS cells. When we set the probability of phage lysis to zero, we observed extensive emergence and proliferation of faster- growing AS cells (Fig. 5a and c and Video 2), which is consistent with our experimental data (Fig. 4). The proportion of slower- growing AR cells significantly decreased as either the probability of losing antibiotic resistance increased (Spearman rank correlation test; rho = - 0.72, \(p_{BC} = 2.6 \times 10^{- 3}\) ) or the fitness cost of antibiotic resistance increased (Spearman rank correlation test; rho = - 0.90, \(p_{BC} = 7.1 \times 10^{- 10}\) ) (Fig. 5c). When we then set the probability of phage lysis to a positive number, the proportion of slower- growing AR cells significantly increased, particularly when the probability of losing antibiotic resistance was \(> 0.02\) and the fitness cost was \(> 20\%\) (two- sample two- sided Welch test; \(p_{BC} = 4.1 \times 10^{- 2}\) , \(n = 4\) ) (Fig. 5b,c and Video 2). These outcomes remain valid when we analyzed our simulations at a fixed total biomass size (Extended Data Fig. 4), and they are therefore robust to the simulation end- point.
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Fig. 5: Properties of spontaneously emerging AS cells determine the persistence of slower-growing AR cells. a,b, Representative individual-based computational simulations of strain AR (magenta) in the (a) absence or (b) presence of phage lysis with different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance. If AR cells undergo a genetic change that causes them to lose antibiotic resistance, such as segregational loss of a plasmid, they are relieved of the fitness cost and become AS cells (grey). The images are the outputs at the last simulation time step. c, The proportion of AR cells as a function of the fitness cost of antibiotic resistance and the probability of losing antibiotic resistance. Each data point is an independent simulation (n = 4), the black data points are in the absence of phage, and the green data points are in the presence of phage. </center>
+
+<--- Page Split --->
+
+To better understand the mechanism causing the increase in AR cells in the presence of phage, we quantified the proportions of slower- growing AR and faster- growing AS cells that are lysed by phage in our simulations. In the presence of phage, a larger proportion of the faster- growing AS cells were lysed compared to the slower- growing AR cells, especially when the AR cells carried a fitness cost and had a higher probability of losing antibiotic resistance (two- sample two- sided Welch tests; \(p_{BC} = 2.4 \times 10^{- 2}\) , \(n = 4\) ) (Extended Data Fig. 5a). The increase in AR cells in the presence of phage cannot be explained by the number of AR cells that lost antibiotic resistance, as the number of such events was either not substantially affected by or was higher in the presence of phage at the end of the simulations (two- sample two- sided Welch test; \(p_{BC} = 4.7 \times 10^{- 3}\) , \(n = 4\) ) (Extended Data Fig. 6a). Thus, our proposed “peripheral kill- the- winner” dynamic remains valid even when AS cells emerge spontaneously. These outcomes remain valid when we analyzed our simulations at a fixed total biomass size (Extended Data Figs. 5b and 6b).  
+
+We further found that the properties of the spontaneously emerging AS cells are important determinants of the persistence of slower- growing AR cells. The number of AS cells that were lysed by phage significantly increased as the probability of losing antibiotic resistance increased (Spearman rank correlation test; \(\text{rho} = 0.88\) , \(p_{BC} = 1.7 \times 10^{- 6}\) ) but was relatively invariant across different fitness costs of antibiotic resistance (Extended Data Fig. 7). For a probability of losing antibiotic resistance of 0.1 and a fitness cost of 30%, over 2,000 AS cells were removed from the biomass by phage lysis, underscoring the significant impact of phage on the faster- growing AS population. The relationship between number of AS cells and the probability of losing antibiotic resistance remained valid at a fixed biomass size, while the number of AS cells lysed by the phage increased as the fitness cost of antibiotic resistance increased (Spearman rank correlation test; \(\text{rho} = 0.97\) , \(p_{BC} = 7.1 \times 10^{- 19}\) ) (Extended Data Fig. 8a). Meanwhile, the number of slower- growing AR cells lysed by phage decreased as the probability of losing antibiotic resistance increased (Spearman rank correlation test; \(\text{rho} = - 0.66\) , \(p_{BC} = 1.2 \times 10^{- 2}\) ) or the fitness cost increased (Spearman rank correlation test; \(\text{rho} = - 0.99\) , \(p_{BC} = 5.3 \times 10^{- 22}\) ) (Extended Data Fig. 7). These outcomes remain valid when we analyzed our simulations at a fixed total biomass size (Extended Data Fig. 8b).
+
+<--- Page Split --->
+
+## Phage lysis promotes biodiversity during surface-associated growth  
+
+We finally sought to test whether our main outcomes can be generalized beyond simple binary co- cultures of two microbial strains. Can our proposed "peripheral kill- the- winner" dynamic enable larger collections of microbial strains to persist together despite variance in their growth rates? To test this, we extended our individual- based computational model to simulate consortia consisting of between two and ten strains with different growth rates. We set the growth rate of one strain to a value of 1 and then randomly assigned the additional strains to have growth rates ranging between 0 and 1, where we took each growth rate from a uniform distribution. In total, we performed 473 total simulations with 35 to 65 replicates for each number of strains in the consortia. We performed simulations in the absence or presence of phage as described above and quantified strain diversity from the final spatial patterns, including strain richness, Shannon diversity, Simpson diversity, and evenness.  
+
+We found that our proposed "peripheral kill- the- winner" dynamic can indeed be generalized to systems with more than two strains. The presence of phage did not impact strain richness (Fig 6a), which is not unexpected as we did not impose a mechanism of cell death other than phage lysis. Thus, all strains remained within the inoculation area. However, we found that the Shannon diversity (Paired two- sided t- test; \(p_{\mathrm{BC}} = 1.6 \times 10^{- 7}\) , \(n > 35\) ) (Fig. 6b), Simpson diversity (Fig. 6c) (Paired two- sided t- test; \(p_{\mathrm{BC}} = 4.8 \times 10^{- 7}\) , \(n > 35\) ), and evenness (Paired two- sided t- test; \(p_{\mathrm{BC}} = 1.6 \times 10^{- 07}\) , \(n > 35\) ) (Fig. 6d) all significantly increased in the presence of phage. Thus, phage lysis not only preserved multiple strains with different growth rates (e.g. fitness costs of antibiotic resistance), but also contributed to a more even distribution of those strains (Fig. 6b- d and Extended Data Fig. 9). These outcomes remain valid when we analyzed our simulations at a fixed simulation time (Extended Data Fig. 10). Our proposed "peripheral kill- the- winner" dynamic may therefore be a general mechanism for how diversity can be maintained within surface- associated microbial systems.
+
+<--- Page Split --->
+![](images/Figure_6.jpg)
+
+<center>Fig. 6: Phage lysis maintains strain diversity. Data are for individual-based computational simulations of the surface-associated growth of co-cultures consisting of between two and ten strains, where each strain has a different growth rate ranging between zero and one. The growth rate of one strain was set to a value of one and the growth rates of the other strains were sampled from a uniform distribution of growth rates. a-d, The strain diversity metrics were quantified at a fixed total biomass size and include (a) strain richness, (b) Shannon diversity, (c) Simpson diversity, and (d) evenness. For b-d, the boxplots identify the mean values, interquartile ranges, and outliers for a total of 473 independent pairs of simulations (between 35-65 replicates for each number of cell types). The black boxplots and data points are in the absence of phage and the green boxplots and data points are in the presence of phage. The \(p\) -values are for two-sided paired t-tests with a Bonferroni correction. </center>
+
+<--- Page Split --->
+
+## Discussion  
+
+Our findings demonstrate that phage lysis can reshape the spatial organization of surface- associated microbial systems with consequences on competitive outcomes between different microorganisms. More specifically, we demonstrate a mechanism for how phage can enable the persistence of AR bacteria even in the absence of antibiotic pressure and when antibiotic resistance imposes a fitness cost. Consistent with prior studies that highlight the fitness costs of antibiotic resistance in the absence of antibiotic pressure<sup>41,42</sup>, our experiments confirm that slower- growing AR strains are outcompeted by faster- growing AS strains when phage are absent. However, the presence of phage initiates a "peripheral kill- the- winner" dynamic, where phage disproportionately lyse the faster- growing AS cells located at the biomass periphery (Figs. 2 and 4). This creates ecological niches located behind the biomass periphery that allow slower- growing AR strains to persist despite their slower growth. These findings underscore the pivotal role of spatial organization in directing microbial community dynamics and evolution and reveal phage as important drivers of these processes. Importantly, these insights highlight the dual role of phages. On the one hand, they offer therapeutic potential against disease- causing or -exacerbating microorganisms<sup>46,47,48,49,50</sup>, on the other hand, their selective pressures on microbial systems may inadvertently maintain or even promote antibiotic resistance under certain conditions<sup>40</sup>. This duality necessitates careful consideration of the ecological consequences of phage therapy, particularly in scenarios where complete eradication of target populations is challenging.  
+
+A key insight from our study is the differential impact of phage lysis on AR strains with varying fitness costs. While a tetracycline- resistant strain \((\mathsf{AR}_{\mathsf{C},\mathsf{Tet}})\) substantially benefited from phage- mediated ecological opportunities, the streptomycin- resistant strain \((\mathsf{AR}_{\mathsf{C},\mathsf{Str}})\) did not (Fig. 2). This discrepancy likely reflects the higher fitness cost of streptomycin resistance in our experimental system, which prevents the \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) strain from capitalizing on phage- mediated niche creation (Fig. 3). These findings emphasize that the persistence of AR strains under phage pressure is contingent on the specific fitness burden associated with resistance mechanisms.
+
+<--- Page Split --->
+
+Future research should explore this relationship across a broader spectrum of resistance determinants to refine predictions of resistance persistence under varying phage- bacteria dynamics.  
+
+We further extend the concept of phage- mediated persistence of antibiotic resistance to systems subjected to the spontaneous emergence of antibiotic sensitivity (i.e. spontaneous loss of antibiotic resistance and its associated fitness costs), which could occur via genetic mutation or plasmid loss. This emergence of antibiotic sensitivity is more likely to occur in rapidly dividing cells that are more prone to segregational plasmid loss or DNA replication errors, which are often positioned at the biomass periphery where nutrients are abundant. Once relieved of the fitness cost associated with antibiotic resistance, the faster growing AS cells gain further access to the biomass periphery. However, here we demonstrate that our proposed "peripheral kill- the- winner" dynamic can counteract other peripheral processes occurring in the system, ultimately preserving the antibiotic resistant population (Figs. 4 and 5). Our mechanism is thus able to explain the persistence of antibiotic resistance even in the presence of evolutionary processes that modify antibiotic resistance landscapes.  
+
+Beyond the maintenance of antibiotic resistance, our "peripheral kill- the- winner" dynamic, offering novel insights into the maintenance of metabolically costly and non- transmissible plasmids. While conventional models attribute the persistence of such plasmids to horizontal gene transfer (HGT) \(^{20,21,51,52}\) , which can facilitate their dissemination and maintenance within microbial communities even though they impose a fitness cost, our findings demonstrate that HGT is not an obligatory mechanism for maintaining these plasmids. Instead, our results reveal an alternative ecological mechanism underpinning plasmid persistence driven by phage lysis. We show that phage preferentially lyse plasmid- free cells that are relieved of their fitness burden, thereby mitigating the competitive disadvantage typically associated with plasmid- bearing strains. This selective pressure not only enables plasmid- carrying antibiotic- resistant cells to persist but also enables them to increase in frequency in the presence of phage independent of HGT. While previous studies have established the role of phage in facilitating
+
+<--- Page Split --->
+
+HGT<sup>40,53,54</sup>, our findings underscore that plasmid maintenance can occur through phage- mediated selection even in the absence of HGT. This mechanism provides a compelling explanation for the evolutionary persistence of metabolically burdensome plasmids in both natural and clinical environments, expanding our understanding of the ecological forces shaping microbial genome evolution.  
+
+Our results are not specific to simple binary systems consisting of antibiotic resistant and sensitive counterparts, but instead can be generalized to any number of strains that vary in their growth properties (Fig. 6). By selectively removing the fastest growing strains from the biomass periphery, phage will prevent any one strain from dominating the system and lead to a more balanced strain distribution. Our proposed "peripheral kill- the- winner" dynamic may therefore be useful for understanding how many natural environments, such as soils and the human microbiome, are able to sustain such incredible levels of biodiversity. These habitats contain surface- associated microbial communities and are abundant in phage<sup>55,56,57,58,59,60</sup>, and it is therefore plausible that our peripheral kill- the- winner contributes to the maintenance of diversity within these systems.  
+
+In conclusion, our study uncovers a mechanism that can explain the paradox of how antibiotic resistant bacteria can persist in the absence of antibiotic pressure. By disproportionately lysing cells positioned at the biomass periphery, phage enable the persistence of slower- growing antibiotic resistant strains, challenging traditional models of resistance loss in antibiotic- free conditions. These findings provide new insights into the ecological and evolutionary mechanisms underlying the spread of antibiotic resistance and highlight the complex interplay between phage, microbial spatial organization, and the maintenance of antibiotic resistance. Our results have implications for microbial management, phage therapy, and the global fight against antibiotic resistance, underscoring the importance of integrating spatial and ecological perspectives into resistance mitigation strategies.
+
+<--- Page Split --->
+
+## Methods  
+
+## Strains and culture conditions  
+
+We used isogenic derivatives of E. coli MG1655 for all of our experiments. We used strain TB204 as the antibiotic sensitive strain (referred to as strain AS), which expresses the gfp green fluorescent protein- encoding gene from the chromosome and is under the control of the lambda promoter61. We used strain TB205 to create all of the antibiotic resistant derivative strains used in this study, which is identical to strain TB204 except that it expresses the mcherry fluorescent protein- encoding gene from the chromosome and is under the control of the lambda promoter61. We created tetracycline and streptomycin resistant derivatives of strain TB205 (referred to as strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Test}}\) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) respectively) by growing cultures of strain TB205 in liquid lysogeny broth (LB) medium into stationary phase and then inoculating the cultures onto agar plates amended with increasingly large concentrations of tetracycline or streptomycin. We created a chloramphenicol resistant derivative of strain E. coli MG1655 (referred to as strain \(\mathsf{AR}_{\mathsf{P},\mathsf{Chl}}\) ) by introducing the non- mobile plasmid pEF001 into the strain via conjugation. This plasmid encodes for the gfp green fluorescent protein- encoding gene and chloramphenicol resistance. We routinely grew all the strains in liquid LB medium or on solid LB agar plates supplemented with \(10\mu \mathrm{g}\mathrm{mL}^{- 1}\) tetracycline, \(50\mu \mathrm{g}\mathrm{mL}^{- 1}\) streptomycin, or \(25\mu \mathrm{g}\mathrm{mL}^{- 1}\) chloramphenicol, respectively, to maintain their resistance determinants. We preserved all strains in \(15\%\) (v/v) glycerol at \(- 80^{\circ}\mathrm{C}\) . Prior to each experiment, we streaked the individual strains from their respective \(- 80^{\circ}\mathrm{C}\) glycerol stocks onto LB agar plates containing their respective antibiotic and used a single colony to initiate all of our experiments.  
+
+We used the lytic phage \(\mathsf{T}6^{62}\) for all of our experiments. We propagated the phage using strain TB204 as the host. Briefly, we grew strain TB204 in LB liquid medium at \(37^{\circ}\mathrm{C}\) for four hours with continuous shaking at 150 rpm. After incubation, we purified phage by filtering the culture through a \(0.22\mu \mathrm{m}\) membrane and storing the supernatant at \(4^{\circ}\mathrm{C}\) until further use. For long- term storage, we mixed equal volumes of strain TB204 and phage suspensions and incubated
+
+<--- Page Split --->
+
+the mixtures for ten minutes with continuous shaking at 150 rpm. We then added glycerol to the mixtures (15% v/v) and stored the stocks at - 80°C.  
+
+## Surface-associated growth experiments  
+
+We performed surface- associated growth experiments across LB agar plates<sup>40</sup>. Briefly, we prepared LB agar plates by pouring autoclaved LB medium containing 1% bacteriology- grade agar into sterile 3.5 cm diameter Petri dishes. We then allowed the medium to solidify overnight at room temperature. We next transferred the agar plates to a sterile hood and allowed them to dry with their lids open for ten minutes. Finally, we covered the plates with their lids, sealed them individually with Parafilm (Amcor, Zürich, Switzerland), and stored them at 4°C until use.  
+
+To perform the experiments, we first prepared overnight cultures of the individual strains in liquid LB medium containing the appropriate antibiotics to maintain their resistance determinants. We then diluted the overnight cultures 1:100 (v/v) into fresh LB medium in the absence of antibiotics and incubated them at 37°C with continuous shaking at 150 rpm for 4 hours to ensure that they were in the exponential growth phase. We next washed the cells to remove all traces of antibiotics by centrifugation at 3600 x g for 10 minutes at 4°C and resuspension in phosphate- buffered saline (PBS). Finally, we adjusted the optical density at 600 nm (OD<sub>600</sub>) of each culture to 1 (approximately 10<sup>8</sup> colony- forming units ml<sup>- 1</sup>).  
+
+To prepare the phage for the experiments, we mixed the refrigerated phage stock with E. coli host cells and incubated the mixtures at 37°C with continuous shaking at 150 rpm for 4 hours. After incubation, we removed the host cells by filtration through a 0.22 μm membrane, resulting in a cell- free active phage solution. We determined the phage titer using the double- layer agar plate method<sup>63</sup> and then diluted the phage solution in PBS to obtain a concentration of 10<sup>8</sup> plaque forming units ml<sup>- 1</sup>.
+
+<--- Page Split --->
+
+For the surface- associated growth experiments with chromosomal antibiotic resistance genetic determinants, we placed a \(1 \mu \text{L}\) droplet of a 1:1 mixture of strains AS and \(\text{AR}_{\text{C, Tet}}\) or strains AS and \(\text{AR}_{\text{C, Str}}\) (OD \(_{600}\) of the mixture of 1) onto the centers of individual replicated LB agar plates (n = 5 for each treatment). For surface- associated growth experiments with plasmid- encoded antibiotic resistance genetic determinants, we placed a \(1 \mu \text{L}\) droplet of strain \(\text{AR}_{\text{P, Chl}}\) onto the centers of individual replicated LB agar plates (n = 5 for each treatment). We then incubated the agar plates for 6 hours. Thereafter, we added a \(1 \mu \text{L}\) droplet of either phage solution or phage- free PBS as a control to the biomass of each agar plate. Finally, we incubated the agar plates in oxic conditions at \(21^{\circ}\text{C}\) for ten days.  
+
+## Confocal laser scanning microscopy and image analysis  
+
+At the end of the surface- associated growth experiments, we imaged the biomass using a Leica TCS SP5 II confocal laser scanning microscope (CLSM) (Leica Microsystems, Wetzlar, Germany) equipped with a \(2.5 \text{x} \text{HCX} \text{FL}\) objective, a numerical aperture of 0.12, and a frame size of \(512 \times 512\) pixels. We set the laser to 488 nm for the excitation of GFP fluorescence and 514 nm for the excitation of RFP fluorescence.  
+
+We quantified the proportions of different strains as a function of biomass radius using ImageJ (https://imagej.nih.gov/ij/) with Fiji plugins (v. 2.1.0/1.53c) (https://fiji.sc). We first loaded the image files into ImageJ and separated the channels to enable the analysis of each strain. We then selected a region of interest and applied an automated threshold using the Yen algorithm optimized for dark backgrounds. We next converted the images into binary masks to minimize background noise and enhance the signal- to- noise ratio. After isolating the target features, we applied a circular Region of Interest (ROI) of specified coordinates and radius to capture the spatial distributions of the strains within a consistent area. We executed the 'Radial Profile' function to obtain the intensity distribution across this circular ROI. This distribution served as
+
+<--- Page Split --->
+
+the basis for determining the radial intensity profile, which is essential for quantifying the community proportions along the circumference. Using the radial intensity profiles of each strain obtained from the ROI, we measured the relative abundances or proportions of each strain along the defined circumference. We repeated this process across all channels to ensure uniformity and comparability in data acquisition.  
+
+## Individual-based computational modeling  
+
+We employed the open- source modeling software Cellmodeller version 4.3 framework<sup>44,45</sup> (https://github.com/cellmodeller/CellModeller) and made modifications based on the "MicrobialEcologyToolbox" branch to carry out individual- based simulations of surface- associated microbial growth. We performed all simulations on the Euler computing cluster at ETH (https://scicomp.ethz.ch/wiki/Euler) using the Slurm workload manager for batch job submissions. We wrote sections of the code for scheduling batch jobs with the aid of ChatGPT- 4 (https://openai.com/gpt- 4). We simulated bacterial cells as three- dimensional capsules of length L that grow uniaxially. Upon reaching a predetermined target length, cells are divided into two daughter cells, each inheriting the characteristics of the parent cell. We set the biophysical parameter "gamma," which controls the ratio of drag force on cell movement to cell growth, to 20 for all our simulations.  
+
+For the simulation of \(E\) . coli cells, we defined the initial cell shape to have a radius of \(0.5 \mu \mathrm{m}\) and a length of \(2 \mu \mathrm{m}\) . Cells divided when their length reached a target length (Ldiv), where Ldiv was randomly sampled from a Gaussian distribution of 2 to \(2.5 \mu \mathrm{m}\) and daughter cells had a length of Ldiv/2. We set the growth rate of AS cells to 1, indicating that each cell would ideally grow by \(1 \mu \mathrm{m}\) per unit time step in the absence of physical constraints. To examine the impact of phage lysis, we implemented the killflag feature in CellModeller, which dynamically removed cells from the simulations and updated the cell state list to retain only living cells<sup>40</sup>. Briefly, the model determined the outermost ring of cells in the simulation space and, based on a random
+
+<--- Page Split --->
+
+probability (kill_rate), flagged cells to stop growing and be removed from the simulation. We varied the kill_rate, emulating phage lysis, between 0 (no phage lysis) and 0.5 (strong phage lysis) in our simulations, where we refer to this value as the phage lysis probability. Thus, instead of simulating the complex biophysical processes of phage replication and lysis, we modeled the phage as a signal that initiates cell removal<sup>40</sup>.  
+
+Using our model, we set the initial conditions to achieve three objectives regarding the effect of phage lysis using the parameters presented in Extended Data Table 1. First, to model the growth of \(\mathsf{AR}_{\mathsf{C}}\) cells, we initialized the simulations with 1,000 cells each of AS and AR cells (1:1 ratio) randomly distributed and rotationally oriented across a circular area of radius 50. We set the growth rate of the AR cells to be between 0.1 and 1, which corresponds to a fitness cost of antibiotic resistance between 0.9 and 0. We performed simulations until the total number of cells reached 40,000. Second, to model the growth of \(\mathsf{AR}_{\mathsf{P}}\) cells, we initialized the simulations with 2,000 \(\mathsf{AR}_{\mathsf{P}}\) cells with a fitness cost for carrying the plasmid. We simulated the spontaneous emergence of AS cells through plasmid loss during cell division by modifying the division function of cells. Briefly, with a certain probability during cell division, \(\mathsf{AR}_{\mathsf{P}}\) cells can switch to AS cells that no longer carry a fitness cost. We varied the fitness cost of antibiotic resistance to be between 0 and 0.3 and the probability that an AR cell will turn into an AS cell to between 0.01 and 0.03. We performed simulations until the total number of cells reached 40,000. Third, to model the growth of multiple strains that differ in their growth rate, we initialized the simulations with 2,000 total cells comprised of equal numbers of between two and ten strains. We next set the fitness cost of the first strain to 0 and randomly sampled from a uniform distribution between 0 and 1 to assign fitness costs to all the other strains in the simulation. Thus, the growth rate of all the other strains ranged between 1 and 0 (fitness cost between 0 and 1) relative to the first strain. We performed simulations until the total number of cells reached 40,000 and then quantified the strain richness, Shannon diversity, Simpson diversity, and evenness using standard methods<sup>64</sup>.
+
+<--- Page Split --->
+
+## Statistical analyses  
+
+We performed all statistical tests with Python (version 3.11.5) using the scipy.stats module (version 1.11.4). To test for differences between means, we employed two- sample two- sided Welch tests for comparing independent groups, which do not assume homoscedasticity in the datasets, and paired t- tests when comparing the same samples under two different treatments. For correlation measurements, we used Pearson's correlation test for linear relationships and Spearman's rank correlation test for monotonic but non- linear relationships. When performing statistical tests on the same data multiple times, we adjusted the \(p\) - values using the Bonferroni correction, which we designate as \(p_{BC}\) . We tested the normality of our datasets using the Shapiro- Wilk test and did not observe significant deviations from the assumption of normality \((p < 0.05)\) . For each statistical test, we reported the specific test used, the corresponding \(p\) - value, and the sample size in the Results section.  
+
+## Data and materials availability  
+
+All data and code generated in this study will be deposited in the publicly accessible Eawag Research Data Institutional Repository (https://opendata.eawag.ch/) at the point of revision. All data and code are currently available to the reviewers and editors at the following link: https://drive.switch.ch/index.php/s/Ts4EoWwwgpOUeFm. All bacterial strains and plasmid pEM001 are freely available from the corresponding author upon request. The sequence of pEM001 will be deposited in NCBI at the point of revision. The sequence of pEM001 is currently available to the reviewers and editors as an attached data file to this submission.
+
+<--- Page Split --->
+
+## References  
+
+1. Cassini, A. et al. Attributable deaths and disability-adjusted life-years caused by infections with antibiotic-resistant bacteria in the EU and the European Economic Area in 2015: a population-level modelling analysis. Lancet Infect. Dis. 19, 56-66 (2019).  
+2. Murray, C. J. L. et al. Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis. Lancet 399, 629-655 (2022).  
+3. Vogwill, T. & MacLean, R. C. The genetic basis of the fitness costs of antimicrobial resistance: a meta-analysis approach. Evol. Appl. 8, 284-295 (2015).  
+4. San Millan, A. & MacLean, R. C. Fitness costs of plasmids: a limit to plasmid transmission. Microbiol. Spectr. 5, 10.1128/microbiolspec.mtbp-0016-2017 (2017).  
+5. Seppälä, H. et al. The Effect of changes in the consumption of macrolide antibiotics on erythromycin resistance in Group A Streptococcus in Finland. Finnish Study Group for Antimicrobial Resistance. N. Engl. J. Med. 337, 441-446 (1997).  
+6. Enne, V. I., Livermore, D. M., Stephens, P. & Hall, L. M. Persistence of sulphonamide resistance in Escherichia coli in the UK despite national prescribing restriction. Lancet 357, 1325-1328 (2001).  
+7. Gottesman, B. S., Carmeli, Y., Shitrit, P. & Chowers, M. Impact of quinolone restriction on resistance patterns of Escherichia coli isolated from urine by culture in a community setting. Clin. Infect. Dis. 49, 869-875 (2009).  
+8. Bengtsson-Palme, J., Kristiansson, E. & Joakim Larsson, D. G. Environmental factors influencing the development and spread of antibiotic resistance. FEMS Microbiol. Rev. 42, fux053 (2018).  
+9. Joakim Larsson, D. G. & Flach, C.-F. Antibiotic resistance in the environment. Nat. Rev. Microbiol. 20, 257-269 (2022).  
+10. Hughes, D. & Andersson, D. I. Evolutionary trajectories to antibiotic resistance. Annu. Rev. Microbiol. 71, 579-596 (2017).  
+11. Wong, A. Epistasis and the evolution of antimicrobial resistance. Front. Microbiol. 8, 246 (2017).
+
+<--- Page Split --->
+
+12. Durão, P., Balbontín, R. & Gordo, I. Evolutionary mechanisms shaping the maintenance of antibiotic resistance. Trends Microbiol. 26, 677-691 (2018).  
+
+13. Darby, E. M., et al. Molecular mechanisms of antibiotic resistance revisited. Nat Rev Microbiol 21, 280-295 (2023).  
+
+14. MacLean, R. C. & San Millan, A. The evolution of antibiotic resistance. Science 365, 1082-1083 (2019).  
+
+15. Hall, J. P. J. et al. Plasmid fitness costs are caused by specific genetic conflicts enabling resolution by compensatory mutation. PLoS Biol. 19, e3001225.  
+
+16. Baker-Austin, C., Wright, M. S., Stepanauskas, R. & McArthur, J. V. Co-selection of antibiotic and metal resistance. Trends Microbiol. 14, 176-182 (2006).  
+
+17. Wales, A. & Davies, R. Co-selection of resistance to antibiotics, biocides and heavy metals, and its relevance to foodborne pathogens. Antibiotics 4, 567-604 (2015).  
+
+18. Palomino, A. et al. Metabolic genes on conjugative plasmids are highly prevalent in Escherichia coli and can protect against antibiotic treatment. ISME J. 17, 151-162 (2023).  
+
+19. Curtsinger, H. D., Martínez-Absalón, S., Liu, Y. & Lopatkin, A. J. The metabolic burden associated with plasmid acquisition: an assessment of the unrecognized benefits to host cells. BioEssays 11, e2400164 (2024).  
+
+20. Grohmann, E., Muth, G. & Espinosa, M. Conjugative plasmid transfer in Gram-positive bacteria. Microbiol. Mol. Biol. Rev. 67, 277-301 (2003).  
+
+21. Lopatkin, A. J. et al. Persistence and reversal of plasmid-mediated antibiotic resistance. Nat. Commun. 8, 1689 (2017).  
+
+22. Lopatkin, A. J. et al. Antibiotics as a selective driver for conjugation dynamics. Nat. Microbiol. 1, 16044 (2016).  
+
+23. Baym, M. et al. Spatiotemporal microbial evolution on antibiotic landscapes. Science 353, 1147-1151 (2016).  
+
+24. Ruan, C., Ramoneda, J., Chen, G., Johnson, D. R. & Wang, G. Evaporation-induced hydrodynamics promote conjugation-mediated plasmid transfer in microbial populations. ISME Commun. 1, 54 (2021).
+
+<--- Page Split --->
+
+25. Sharma, A. & Wood, K. B. Spatial segregation and cooperation in radially expanding microbial colonies under antibiotic stress. ISME J. 15, 3019-3033 (2021).  
+
+26. Metzger, G. A. et al. Biofilms preserve the transmissibility of a multi-drug resistance plasmid. NPJ Biofilms Microbiomes 8, 95 (2022).  
+
+27. Ruan, C., Ramoneda, J., Gogia, G., Wang, G. & Johnson, D. R. Fungal hyphae regulate bacterial diversity and plasmid-mediated functional novelty during range expansion. Curr. Biol. 32, 5285-5294. e4 (2022).  
+
+28. Denk-Lobnig, M. & Wood, K. B. Antibiotic resistance in bacterial communities. Curr. Opin. Microbiol. 74, 102306 (2023).  
+
+29. Ruan, C., Borer, B., Ramoneda, J., Wang, G. & Johnson, D. R. Evaporation-induced hydrodynamics control plasmid transfer during surface-associated microbial growth. NPJ Biofilms Microbiomes 9, 58 (2023).  
+
+30. Freire, T. F. A., Hu, Z., Wood, K. B. & Gjini, E. Modeling spatial evolution of multi-drug resistance under drug environmental gradients. PLoS Comput. Biol. 20, e1012098 (2024).  
+
+31. Ma, Y., Kan, A. & Johnson, D. R. Metabolic interactions control the transfer and spread of plasmid-encoded antibiotic resistance during surface-associated microbial growth. Cell Rep. 43, 114653 (2024).  
+
+32. Peischl, S., Dupanloup, I., Kirkpatrick, M. & Excoffier, L. On the accumulation of deleterious mutations during range expansions. Mol. Ecol. 22, 5972-5982 (2013).  
+
+33. Bosshard, L. et al. Accumulation of deleterious mutations during bacterial range expansions. Genetics 207, 669-684 (2017).  
+
+34. Kayser, J., Schreck, C. F., Gralka, M., Fusco, D. & Hallatschek, O. Collective motion conceals fitness differences in crowded cellular populations. Nat. Ecol. Evol. 3, 125-134 (2019).  
+
+35. Abe, K., Nomura, N. & Suzuki, S. Biofilms: hot spots of horizontal gene transfer (HGT) in aquatic environments, with a focus on a new HGT mechanism. FEMS Microbiol. Ecol. 96, fiaa031 (2020).  
+
+36. Hallatschek, O., Hersen, P., Ramanathan, S. & Nelson, D. R. Genetic drift at expanding frontiers promotes gene segregation. Proc. Natl. Acad. Sci. U. S. A. 104, 19926-19930 (2007).
+
+<--- Page Split --->
+
+37. Hallatschek, O. & Nelson, D. R. Life at the front of an expanding population. Evol. Int. J. Org. Evol. 64, 193-206 (2010). 
+
+38. Mitri, S., Clarke, E. & Foster, K. R. Resource limitation drives spatial organization in microbial groups. ISME J. 10, 1471-1482 (2016). 
+
+39. Eriksen, R. S., Svenningsen, S. L., Sneppen, K. & Mitarai, N. A growing microcolony can survive and support persistent propagation of virulent phages. Proc. Natl. Acad. Sci. U. S. A. 115, 337-342 (2018). 
+
+40. Ruan, C. et al. Phage predation accelerates the spread of plasmid-encoded antibiotic resistance. Nat. Commun. 15, 5397 (2024). 
+
+41. Andersson, D. I. & Hughes, D. Antibiotic resistance and its cost: is it possible to reverse resistance? Nat. Rev. Microbiol. 8, 260-271 (2010). 
+
+42. Carroll, A. C. & Wong, A. Plasmid persistence: costs, benefits, and the plasmid paradox. Can. J. Microbiol. 64, 293-304 (2018). 
+
+43. Yang, Q. E. et al. Compensatory mutations modulate the competitiveness and dynamics of plasmid-mediated colistin resistance in Escherichia coli clones. ISME J. 14, 861-865 (2020). 
+
+44. Rudge, T. J., Steiner, P. J., Phillips, A. & Haseloff, J. Computational modeling of synthetic microbial biofilms. ACS Synth. Biol. 1, 345-352 (2012). 
+
+45. Rudge, T. J., Federici, F., Steiner, P. J., Kan, A. & Haseloff, J. Cell polarity-driven instability generates self-organized, fractal patterning of cell layers. ACS Synth. Biol. 2, 705-714 (2013). 
+
+46. Chegini, Z. et al. Bacteriophage therapy against Pseudomonas aeruginosa biofilms: a review. Ann. Clin. Microbiol. Antimicrob. 19, 45 (2020). 
+
+47. Duan, Y., Young, R. & Schnabl, B. Bacteriophages and their potential for treatment of gastrointestinal diseases. Nat. Rev. Gastroenterol. Hepatol. 19, 135-144 (2022). 
+
+48. Uyttebroek, S. et al. Safety and efficacy of phage therapy in difficult-to-treat infections: a systematic review. Lancet Infect. Dis. 22, e208-e220 (2022). 
+
+49. Kviatcovsky, D., Valdés-Mas, R., Federici, S. & Elinav, E. Phage therapy in noncommunicable diseases. Science 382, 266-267 (2023). 
+
+50. Strathdee, S. A., Hatfull, G. F., Mutalik, V. K. & Schooley, R. T. Phage therapy: from biological mechanisms to future directions. Cell 186, 17-31 (2023).
+
+<--- Page Split --->
+
+51. Harrison, E. & Brockhurst, M. A. Plasmid-mediated horizontal gene transfer is a  
+
+coevolutionary process. Trends Microbiol. 20, 262- 267 (2012).  
+
+52. Ding, M. et al. Subinhibitory antibiotic concentrations promote the horizontal transfer of plasmid-borne resistance genes from Klebsiella pneumoniae to Escherichia coli. Front. Microbiol. 13, 1017092 (2022).  
+
+53. Schneider, C. L. Bacteriophage-Mediated Horizontal Gene Transfer: Transduction. in Bacteriophages: Biology, Technology, Therapy (eds. Harper, D. R., Abedon, S. T., Burrowes, B. H. & McConville, M. L.) 151-192 (Springer International Publishing, Cham, 2021).  
+
+54. Watson, B. N. J., Staals, R. H. J. & Fineran, P. C. CRISPR-Cas-mediated phage resistance enhances horizontal gene transfer by transduction. mBio 9, e02406-17 (2018).  
+
+55. Williamson, K. E., Fuhrmann, J. J., Wommack, K. E. & Radosevich, M. Viruses in soil ecosystems: an unknown quantity within an unexplored territory. Annu. Rev. Virol. 4, 201-219 (2017).  
+
+56. Pratama, A. A. & van Elsas, J. D. The 'neglected' soil virome - potential role and impact. Trends Microbiol. 26, 649-662 (2018).  
+
+57. Shkoporov, A. N. & Hill, C. Bacteriophages of the human gut: the "known kinknown" of the microbiome. Cell Host Microbe 24, 195-209 (2019)  
+
+58. Liang, G. & Bushman, F. D. The human virome: assembly, composition and host interactions. Nat. Rev. Microbiol. 19, 514-527 (2021).  
+
+59. Cao, Z. et al. The gut microme: a new microbiome component in health and disease. Lancet 81, 104113 (2022).  
+
+60. Jansson, J. K. & Wu, R. Soil viral diversity, ecology and climate change. Nat. Rev. Microbiol. 21, 296-311 (2023).  
+
+61. Tomasek, K., Bergmiller, T. & Guet, C. C. Lack of cations in flow cytometry buffers affect fluorescence signals by reducing membrane stability and viability of Escherichia coli strains. J. Biotechnol. 268, 40-52 (2018).  
+
+62. Huang, W. M., Wei, L. S. & Casjens, S. Relationship between bacteriophage T4 and T6 DNA topoisomerases. T6 39-protein subunit is equivalent to the combined T4 39- and 60-protein subunits. J. Biol. Chem. 260, 8973-8977 (1985).
+
+<--- Page Split --->
+
+63. Stachurska, X., Roszak, M., Jabłońska, J., Mizielińska, M. & Nawrotek, P. Double-layer agar (DLA) modifications for the first step of the phage-antibiotic synergy (PAS) identification.  
+
+Antibiotics 10, 1306 (2021).  
+
+64. Roswell, M., Dushoff, J. & Winfree, R. A conceptual guide to measuring species diversity. Oikos 130, 321-338 (2021).  
+
+## Acknowledgments  
+
+We thank Zachary Bailey for providing phage T6; Dr. Martin Ackermann and Emanuele Fara for providing all E. coli strains and plasmid pEF001; and Drs. Mireia Cordero and Josep Ramoneda for helpful discussions. We thank Mrs. Ruan (Guo Chen) for her excellent assistance with structuring the simulation videos. C.R. and D.P.V. were supported by a grant from the Swiss National Science Foundation (310030_207471) awarded to D.R.J.  
+
+## Contributions  
+
+C.R. and D.P.V. conceived and developed the research question. C.R., D.P.V., and D.R.J. designed the experiments. C.R. performed all the experiments. D.P.V. performed all the individual-based computational simulations. C.R. and D.P.V. analyzed the data. C.R. and D.P.V. wrote the first version of the manuscript with contributions from D.R.J. D.R.J. coordinated the project. All authors reviewed and approved the final version of the manuscript.  
+
+## Competing interests  
+
+The authors declare no competing interests.
+
+<--- Page Split --->
+
+850 Extended Data
+
+851
+
+852 **Extended Data Table 1: Parameters used for the individual-based computational simulations**
+
+853
+
+<table><tr><td>Parameter</td><td>Description</td><td>Value</td><td>Unit</td></tr><tr><td>\(g_{AS}\)</td><td>Specific growth rate of an AS cell</td><td>1</td><td>-</td></tr><tr><td>\(g_{AR}\)</td><td>Specific growth rate of an AR cell</td><td>0.1-0.95</td><td>-</td></tr><tr><td>\(L_{0}\)</td><td>Mean initial length of a cell</td><td>2</td><td>μm</td></tr><tr><td>\(L_{d}\)</td><td>Length of a cell at division</td><td>3.5-4</td><td>μm</td></tr><tr><td>\(W_{infection}\)</td><td>Width of the outer cell layer affected by</td><td>4</td><td>μm</td></tr><tr><td>\(d_{cell}\)</td><td>Mean diameter of a cell</td><td>0.5</td><td>μm</td></tr><tr><td>\(N\)</td><td>Initial number of cells</td><td>2000</td><td>cell</td></tr><tr><td>\(R\)</td><td>Probability of phage lysis</td><td>0.01-0.05</td><td>-</td></tr><tr><td>\(P\)</td><td>Probability of losing antibiotic resistance</td><td>0.01-0.03</td><td>-</td></tr></table>
+
+<--- Page Split --->
+![](images/Extended_Data_Figure_4.jpg)
+  
+
+Extended Data Fig. 1: Fitness cost of antibiotic resistance determines the effect of phage lysis on the persistence of antibiotic resistant (AR) cells. The effect of phage on the proportion of AR cells is the proportion of AR cells in the presence of phage minus the proportion of AR cells in the absence of phage. All quantities are those at a fixed simulation time. Each data point is an independent simulation (n = 4).
+
+<--- Page Split --->
+![](images/Extended_Data_Figure_5.jpg)
+  
+
+Extended Data Fig. 2: Fitness cost of antibiotic resistance and probability of phage lysis determine the persistence of slower- growing AR strains at a fixed total biomass size. All data are for individual- based computational simulations of co- cultures consisting of strains AS and AR with different fitness costs of antibiotic resistance and probabilities of phage lysis. a, The proportion of AR cells as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. b, The number of AS cells removed by phage as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. c, The number of AR cells removed by phage as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. For a–c, each data point is an independent simulation (n = 4), the lines connect the mean values, and the shaded regions are one standard deviation.
+
+<--- Page Split --->
+![PLACEHOLDER_37_0]
+  
+
+Extended Data Fig. 3: Phage lysis increases the persistence of slower-growing AR cells in the face of spontaneously generated AS cells across different environmental conditions. a,b Representative CLSM images of cultures of strain \(\mathrm{AR_{P,CHI}}\) in the (a) absence or (b) presence of phage T6 after 16 days of growth in oxic conditions at \(21^{\circ}\mathrm{C}\) in the absence of antibiotic pressure. c,d, Representative CLSM images of cultures of strain \(\mathrm{AR_{P,CHI}}\) in the (c) absence or (d) presence of phage T6 after ten days of growth in oxic conditions at \(30^{\circ}\mathrm{C}\) in the absence of antibiotic pressure. e,f, Representative CLSM images of cultures of strain \(\mathrm{AR_{P,CHI}}\) in the (e) absence or (f) presence of phage T6 after ten days of growth in anoxic conditions at \(21^{\circ}\mathrm{C}\) in the absence of antibiotic pressure. For all images, fluorescent (magenta) regions are \(\mathrm{AR_{P,CHI}}\) cells and non- fluorescent regions are \(\mathrm{AR_{P,CHI}}\) cells that lost plasmid pEF001 and became AS cells.
+
+<--- Page Split --->
+![PLACEHOLDER_38_0]
+
+<center>Extended Data Fig. 4: Properties of spontaneously emerging AS cells determine the persistence of slower-growing AR strains at a fixed total biomass size. All data are for individual-based computational simulations of strain AR in the absence or presence of phage lysis. The proportion of AR cells is calculated for different fitness costs of antibiotic resistance, probabilities of losing antibiotic resistance, and probabilities of phage lysis. Each data point is an independent simulation (n = 4), the black data points are in the absence of phage, and the green data points are in the presence of phage. </center>
+
+<--- Page Split --->
+![PLACEHOLDER_39_0]
+
+<center>Extended Data Fig. 5: Faster-growing emergent AS cells are disproportionately lysed by phage. </center>  
+
+All data are for individual- based computational simulations of strain AR in the absence or presence of phage lysis. The proportions of lysed AS (black) and AR (green) cells are calculated for different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance. a,b, The proportions are the number of lysed cells of one strain to the total number of cells of that strain at (a) a fixed simulation time, or (b) a fixed total biomass size. Each data point is an independent simulation (n = 4), the black data points are in the absence of phage, and the green data points are in the presence of phage.
+
+<--- Page Split --->
+![PLACEHOLDER_40_0]
+  
+
+Extended Data Fig. 6: Fitness cost of antibiotic resistance, probability of phage lysis, and probability of losing antibiotic resistance determine the total number of antibiotic resistance loss events. a,b, All data are for individual- based computational simulations of strain AR for different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance in the absence or presence of phage at (a) a fixed simulation time, or (b) a fixed total biomass size. Each data point is an independent simulation (n = 4), the black data points are for a probability of phage lysis of zero, and the green data points are for a probability of phage lysis of 0.05.
+
+<--- Page Split --->
+![PLACEHOLDER_41_0]
+  
+
+Extended Data Fig. 7: Fitness cost of antibiotic resistance, probability of phage lysis, and probability of losing antibiotic resistance determine the removal of AS and AR cells. a,b, All data are for individual- based computational simulations of strain AR in the absence or presence of phage at a fixed simulation time. a,b, The number of (a) AS cells or (b) AR cells removed by phage for different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance. For a,b, each data point is an independent simulation (n = 4), the black data points are for a probability of phage lysis of zero, and the green data points are for a probability of phage lysis of 0.05.
+
+<--- Page Split --->
+![PLACEHOLDER_42_0]
+  
+
+Extended Data Fig. 8: Fitness cost of antibiotic resistance, probability of phage lysis, and probability of losing antibiotic resistance determine the removal of AS and AR cells. All data are for individual- based computational simulations of strain AR in the absence or presence of phage at a fixed total biomass size. a,b, The number of (a) AS cells or (b) AR cells removed by phage for different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance. For a,b, each data point is an independent simulation (n = 4), the black data points are for a probability of phage lysis of zero, and the green data points are for a probability of phage lysis of 0.05.
+
+<--- Page Split --->
+![PLACEHOLDER_43_0]
+  
+
+Extended Data Fig. 9: Phage lysis maintains strain diversity. Representative individual- based computational simulations of co- cultures consisting of between two and ten distinct strains in the absence or presence of phage. The images are the outcomes at a fixed total biomass size. Each strain has a different growth rate ranging between zero and one that was randomly assigned by sampling from a uniform distribution. The color of each cell corresponds to its growth rate, represented as a gradient from magenta (growth rate = 0) to cyan (growth rate = 1).
+
+<--- Page Split --->
+![PLACEHOLDER_44_0]
+  
+
+Extended Data Fig. 10: Phage lysis maintains strain diversity when analyzing simulations at a fixed simulation time. Data are for individual- based computational simulations of the surface- associated growth of co- cultures of between two and ten strains, where each strain has a different growth rate ranging between zero and one. The growth rate of one strain was set to a value of one and the growth rates of the other strains were sampled from a uniform distribution of growth rates. a- d, The strain diversity metrics are those quantified at the last simulation time and include (a) strain richness, (b) Shannon diversity, (c) Simpson diversity, and (d) evenness. For b- d, the boxplots identify the mean values, interquartile ranges, and outliers for 473 independent pairs of simulations in total (between 35- 65 replicates for any given number of cell types). The black boxplots and data points are in the absence of phage and the green boxplots and data points are in the presence of phage. The \(p\) - values are for two- sided paired t- tests with a Bonferroni correction.
+
+<--- Page Split --->
+
+## Video Captions  
+
+Video 1: Representative individual- based computational simulations of spatial patterns formed by co- cultures of strains AS and AR for different fitness costs of antibiotic resistance and probabilities of phage lysis. We initiated the simulations with 1,000 AS cells (cyan) and 1,000 AR cells (magenta) and performed simulations until the total number of cells reached 40,000. Simulations are for three different fitness costs of antibiotic resistance and four different probabilities of phage lysis.  
+
+Video 2: Representative individual- based computational simulations of spatial patterns formed by the AR strain for different fitness costs of antibiotic resistance, probabilities of losing antibiotic resistance, and probabilities of phage lysis. The AR cells (magenta) have a fitness cost for antibiotic resistance. Each cell can stochastically transition into an AS cell (grey) with a certain probability, in which case it will be relieved of its fitness cost. We initiated the simulations with 2,000 AR cells and performed simulations until the total number of cells reached 40,000. Simulations are for three different fitness costs of antibiotic resistance and three different probabilities of losing antibiotic resistance. The upper simulations are with phage lysis and the lower simulations are without phage lysis.
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+20251119RuanVideo1. mp4 20251119RuanVideo2. mp4
+
+<--- Page Split --->

preprint/preprint__487dd727461ac5ce1675c5210f0d2bb09de73dbbde2a7e84d0c10a7c04a6fbfd/preprint__487dd727461ac5ce1675c5210f0d2bb09de73dbbde2a7e84d0c10a7c04a6fbfd_det.mmd

@@ -0,0 +1,616 @@
+<|ref|>title<|/ref|><|det|>[[44, 107, 888, 208]]<|/det|>
+# Phage lysis facilitates the maintenance of costly antibiotic resistance in the absence of antibiotic pressure  
+
+<|ref|>text<|/ref|><|det|>[[44, 229, 300, 276]]<|/det|>
+David Johnson  david.johnson@eawag.ch  
+
+<|ref|>text<|/ref|><|det|>[[44, 302, 926, 416]]<|/det|>
+Swiss Federal Institute of Aquatic Science and Technology https://orcid.org/0000- 0002- 6728- 8462  Chujin Ruan  Eawag: Das Wasserforschungs- Institut des ETH- Bereichs https://orcid.org/0009- 0009- 8605- 7107  Deepthi Vinod  Swiss Federal Institute of Aquatic Science and Technology  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 455, 103, 473]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 493, 920, 535]]<|/det|>
+Keywords: Antibiotic resistance, phage- bacteria interactions, plasmid dynamics, spatial organization, biofilm  
+
+<|ref|>text<|/ref|><|det|>[[44, 554, 325, 574]]<|/det|>
+Posted Date: February 7th, 2025  
+
+<|ref|>text<|/ref|><|det|>[[44, 592, 475, 612]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 5879387/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 630, 916, 673]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License.  Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[44, 691, 535, 711]]<|/det|>
+Additional Declarations: There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[42, 746, 950, 790]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on July 1st, 2025. See the published version at https://doi.org/10.1038/s41467- 025- 61055- y.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[75, 92, 166, 113]]<|/det|>
+## 1 Title  
+
+<|ref|>text<|/ref|><|det|>[[75, 126, 880, 175]]<|/det|>
+2 Phage lysis facilitates the maintenance of costly antibiotic resistance in the absence of antibiotic pressure  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 213, 205, 233]]<|/det|>
+## Authors  
+
+<|ref|>text<|/ref|><|det|>[[115, 245, 567, 265]]<|/det|>
+6 Chujin Ruan \(^{1\# *}\) , Deepthi P. Vinod \(^{1,2\#}\) , David R. Johnson \(^{1,3*}\)  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 304, 236, 325]]<|/det|>
+## Affiliations  
+
+<|ref|>text<|/ref|><|det|>[[112, 339, 848, 444]]<|/det|>
+9 1Department of Environmental Microbiology, Swiss Federal Institute of Aquatic Science and Technology (Eawag), 8600 Dübendorf, Switzerland; 2Department of Environmental Systems Science, Swiss Federal Institute of Technology (ETH), 8092 Zürich, Switzerland; 3Institute of Ecology and Evolution, University of Bern, 3012 Bern, Switzerland.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 479, 303, 501]]<|/det|>
+## \* Correspondence  
+
+<|ref|>text<|/ref|><|det|>[[112, 514, 764, 534]]<|/det|>
+15 David R. Johnson, david.johnson@eawag.ch; Chujin Ruan, chujin.ruan@eawag.ch  
+
+<|ref|>text<|/ref|><|det|>[[112, 543, 642, 562]]<|/det|>
+16 \*These authors contributed equally: Chujin Ruan, Deepthi P. Vinod
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[66, 92, 210, 113]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[110, 125, 876, 565]]<|/det|>
+The persistence of antibiotic resistant (AR) bacteria in the absence of antibiotic pressure raises a paradox regarding the fitness costs associated with antibiotic resistance. These fitness costs should slow the growth of AR bacteria and cause them to be displaced by faster- growing antibiotic sensitive (AS) counterparts. Yet, even in the absence of antibiotic pressure, slower- growing AR bacteria can persist for prolonged periods of time. Here, we demonstrate a mechanism that can explain this apparent paradox. We hypothesize that lytic phage can modulate bacterial spatial organization to facilitate the persistence of slower- growing AR bacteria. Using surface- associated growth experiments with the bacterium Escherichia coli in conjunction with individual- based computational simulations, we show that phage disproportionately lyse the faster- growing AS counterpart cells located at the biomass periphery via a "peripheral kill- the- winner" dynamic. This enables the slower- growing AR cells to persist even when they are susceptible to the same phage. This phage- mediated selection is accompanied by enhanced bacterial diversity, further emphasizing the role of phage in shaping the assembly and evolution of bacterial systems. The mechanism is potentially relevant for any antibiotic resistance genetic determinant and has tangible implications for the management of bacterial populations via phage therapy.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 599, 199, 615]]<|/det|>
+## Keywords  
+
+<|ref|>text<|/ref|><|det|>[[115, 625, 833, 671]]<|/det|>
+Antibiotic resistance, phage- bacteria interactions, plasmid dynamics, spatial organization, biofilm
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 93, 174, 113]]<|/det|>
+## Main  
+
+<|ref|>text<|/ref|><|det|>[[111, 155, 875, 397]]<|/det|>
+The rise of antibiotic resistance is an urgent threat to global public health and is causing increasing morbidity and mortality worldwide<sup>1,2</sup>. A broadly applied strategy to combat this threat is the judicious use and disposal of antibiotics. This strategy anticipates that antibiotic resistant (AR) bacteria will be displaced by antibiotic sensitive (AS) counterparts in the absence of antibiotic pressure<sup>3,4</sup>. This is because antibiotic resistance often imposes fitness costs that slow the growth of AR bacteria. This provides a growth advantage to AS counterparts that lose antibiotic resistance through genetic mutation or plasmid loss<sup>3,4</sup>. However, despite the implementation of this strategy<sup>5,6,7</sup>, AR bacteria continue to persist in the absence of antibiotic pressure in a myriad of environments<sup>8,9</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[111, 430, 883, 814]]<|/det|>
+Several mechanisms have been proposed to explain how slower- growing AR bacteria can persist in the face of faster- growing AS counterparts in the absence of antibiotic pressure<sup>10,11,12,13</sup>. These mechanisms include compensatory mutations that reduce the fitness costs associated with antibiotic resistance or their associated genetic elements<sup>14,15</sup>, co- selection of linked traits<sup>16,17,18,19</sup>, and horizontal transfer of antibiotic resistance genetic determinants<sup>20,21,22</sup>. One aspect that can affect the dynamics of antibiotic resistance determinants is bacterial spatial organization<sup>23,24,25,26,27,28,29,30,31</sup>. Experiments and theoretical considerations have illustrated how spatial processes can increase the persistence of neutral and even deleterious genetic mutations at the periphery of growing biomass<sup>32,33,34</sup>. Whether these processes can explain the persistence of slower- growing AR bacteria in the face of faster- growing AS counterparts, however, remains unclear. Because surface- associated bacterial systems are important reservoirs of antibiotic resistance genetic determinants<sup>35</sup>, understanding how bacterial spatial organization and dynamics affect the persistence of antibiotic resistance could set the stage for developing more effective bacterial management and intervention strategies.  
+
+<|ref|>text<|/ref|><|det|>[[112, 847, 882, 895]]<|/det|>
+We hypothesize here that phage lysis can modulate bacterial spatial organization to increase the persistence of slower- growing AR bacteria in the face of faster- growing AS counterparts. More
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 880, 443]]<|/det|>
+precisely, we hypothesize that lytic phage can mediate a "peripheral kill- the- winner" dynamic; faster- growing AS counterpart cells are disproportionally lysed to a greater extent than slower- growing AR cells, consequently increasing the persistence of antibiotic resistance (Fig. 1). Our hypothesis is grounded in the fundamental principle that, for surface- associated bacterial systems, biomass growth is primarily driven by cells located at the biomass periphery where resources supplied from the environment are plentiful \(^{36,37,38}\) . Because of their differences in growth rates, slower- growing AR bacteria will be disproportionately located behind the biomass periphery while faster- growing AS counterpart cells will disproportionately occupy the biomass periphery (Fig. 1). Due to mass transfer limitations, phage will predominantly lyse cells located at the biomass periphery \(^{39,40}\) , which will disproportionately be AS counterpart cells (Fig 1). The faster- growing AS counterpart cells will therefore undergo more vigorous phage lysis, thereby offsetting their growth advantage and increasing the persistence of slower- growing AR cells (Fig. 1).  
+
+<|ref|>text<|/ref|><|det|>[[111, 476, 880, 830]]<|/det|>
+Our hypothesis not only provides an explanation for the persistence of slower- growing AR cells in the face of faster- growing AS counterpart cells, but also makes predictions regarding dynamic environments where spontaneous genetic changes that alter antibiotic resistance profiles can occur concurrently with phage lysis. These alterations in antibiotic resistance profiles can occur through genetic mutations or through plasmid loss, both of which can relieve cells of the fitness costs associated with antibiotic resistance \(^{41,42,43}\) . After such genetic changes occur, the faster- growing AS counterpart cells that emerge will grow towards and disproportionately occupy the biomass periphery, consequently making them more susceptible to phage lysis and increasing the persistence of slower- growing AR cells. We therefore predict that, even if AR cells are capable of reverting to AS cells via spontaneous genetic changes, the "peripheral kill- the- winner" dynamic will rapidly establish itself to disproportionately lyse the AS cells and promote the persistence of the slower- growing AR cells, thus providing a general mechanism for how antibiotic resistance genetic determinants can persist amidst ongoing evolutionary processes.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 92, 825, 380]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 394, 878, 545]]<|/det|>
+<center>Fig. 1: Schematic of the "peripheral kill-the-winner" hypothesis. In the absence of phage, we expect that faster-growing antibiotic sensitive (AS) cells (cyan) will displace slower-growing antibiotic resistant (AR) cells (magenta) along the biomass periphery. In the presence of phage, however, we expect that the slower-growing AR cells will persist with the faster-growing AS cells. This is because the faster-growing AS cells will disproportionately occupy the biomass periphery, and they will therefore be more susceptible to phage lysis. This will increase the removal of AS cells from the biomass and counteract the benefits of their faster growth relative to the slower-growing AR cells, thus increasing the persistence of the slower-growing AR cells. </center>  
+
+<|ref|>text<|/ref|><|det|>[[112, 597, 864, 896]]<|/det|>
+To test our hypothesis, we performed surface- associated growth experiments with isogenic derivatives of the bacterium Escherichia coli MG1655 in the absence or presence of the lytic phage T6. The parental strain, which we refer to as strain AS for "antibiotic sensitive", is sensitive to all of the antibiotics used in this study. We then obtained antibiotic resistant variants of strain AS, which we refer to as AR strains for "antibiotic resistant". The AR strains contain antibiotic resistance determinants that are either single genetic mutations located on the chromosome (strain \(\mathrm{AR}_{\mathrm{C},\mathrm{Tet}}\) is resistant to tetracycline while strain \(\mathrm{AR}_{\mathrm{C},\mathrm{Str}}\) is resistant to streptomycin) or are genes located on the non- transmissible plasmid pEF001 (strain \(\mathrm{AR}_{\mathrm{P},\mathrm{Chl}}\) is resistant to chloramphenicol) that can be spontaneously lost during cell division. We further introduced fluorescent protein- encoding genes into the chromosomes of the strains and into plasmid pEF001 so that we can distinguish them when grown together. We then assembled
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 883, 388]]<|/det|>
+strain AS with either strain \(\mathsf{AR}_{\mathsf{C,Tet}}\) or \(\mathsf{AR}_{\mathsf{C,Str}}\) into co- cultures, propagated them across nutrient- amended agar surfaces in the absence of antibiotic pressure, and quantified the strain abundances and emergent spatial patterns using confocal laser- scanning microscopy (CLSM). We also propagated strain \(\mathsf{AR}_{\mathsf{P,Chl}}\) alone across nutrient- amended agar surfaces in the absence of antibiotic pressure, tracked the spontaneous emergence of plasmid- free AS cells, and quantified the emergent spatial patterns with CLSM. In both cases, we expected that the AR strains, all of which grow slower than the AS strain, would have increased persistence in the presence of phage T6 via our proposed "peripheral kill the winner" dynamic (Fig. 1). Finally, we complemented our experiments with individual- based computational simulations to identify general mechanisms for how phage lysis can increase the persistence of slower- growing AR strains.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 461, 196, 482]]<|/det|>
+## Results  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 524, 633, 543]]<|/det|>
+## Phage lysis increases the persistence of slower-growing AR cells  
+
+<|ref|>text<|/ref|><|det|>[[111, 577, 881, 850]]<|/det|>
+We first quantified how phage lysis affects the persistence of slower- growing AR cells in the face of faster- growing AS counterparts. To accomplish this, we performed surface- associated growth experiments in the absence or presence of phage T6. We mixed strain AS with either strain \(\mathsf{AR}_{\mathsf{C,Tet}}\) or \(\mathsf{AR}_{\mathsf{C,Str}}\) , both of which contain single chromosomal mutations that bestow resistance to tetracycline or streptomycin, at a 1:1 initial cell ratio. To distinguish them, strain AS expressed a green fluorescent protein- encoding gene (falsely colored to cyan in the images) located on the chromosome while strains \(\mathsf{AR}_{\mathsf{C,Tet}}\) and \(\mathsf{AR}_{\mathsf{C,Str}}\) expressed a red fluorescent protein- encoding gene (falsely colored to magenta in the images) located on the chromosome. We then grew the cocultures across nutrient- amended agar surfaces in the absence of antibiotic pressure and quantified the patterns of spatial organization that emerged with CLSM (Fig. 2a).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 875, 333]]<|/det|>
+In the absence of phage T6, we found that strain AS had a clear competitive advantage over both strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) (Fig. 2a). After ten days of incubation, strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) constituted less than \(0.7\%\) of the total biomass area even though they constituted approximately \(50\%\) of the initial inoculum (Fig. 2b). We attribute this effect to the fitness cost associated with the antibiotic resistance determinants for tetracycline and streptomycin. This is evident from our experiments, where the total biomass areas of strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) or \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) when grown alone in the absence of antibiotic pressure were both significantly smaller than the area of strain AS when grown alone (two- sample two- sided Welch tests; \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) , \(p = 1.2 \times 10^{- 5}\) , \(n = 5\) ; \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) , \(p = 6.1 \times 10^{- 7}\) , \(n = 5\) ) (Fig. 2c).  
+
+<|ref|>image<|/ref|><|det|>[[115, 370, 880, 616]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[112, 636, 881, 898]]<|/det|>
+<center>Fig. 2: Phage lysis increases the persistence of slower-growing AR cells. a, Representative CLSM images of co-cultures of strains AS (cyan) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) (magenta) (upper images) or of strains AS (cyan) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) (magenta) (lower images) in the absence or presence of phage T6. We imaged the biomass after ten days of incubation in the absence of antibiotic pressure. b, The proportion of the total biomass area occupied by strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) or \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) when grown in coculture with strain AS in the absence or presence of phage T6. c, The biomass diameters of strains AS, \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) when grown in monoculture in the absence of phage T6. d, The biomass diameters of co-cultures of strains AS and \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) or strains AS and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) in the absence or presence of phage T6. e, The biomass area (population size) of strain \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) when grown in co-culture with strain AS in the absence or presence of phage T6. f, The proportion of \(\mathsf{AR}_{\mathsf{C},\mathsf{Tet}}\) cells within co-cultures as a function of the radial distance from the centroid of the biomass. For b–f, each data point is an independent experimental replicate ( \(n = 5\) ), the black data points are for experiments in the absence of phage T6, and the green data points are for experiments in the presence of phage T6. For b–e, the \(p\) -values are for two-sample two-sided Welch tests. </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 115, 883, 444]]<|/det|>
+In the presence of phage T6, we found that the slower-growing strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) can persist in the face of the faster- growing strain AS (Fig. 2a). The proportion of the area occupied by strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) was nearly 10- fold larger when in the presence of phage T6 (two- sample two- sided Welch test; \(p = 1.1\times 10^{- 2}\) , \(n = 5\) ) (Fig. 2b). Moreover, even though the presence of phage T6 significantly reduced the overall biomass area of all the strains (two- sample two- sided Welch tests; \(p< 1.0\times\) \(10^{- 7}\) , \(n = 5\) ) (Fig. 2d), the absolute population size of the slower- growing strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) increased (two- sample two- sided Welch test; \(p = 3.5\times 10^{- 2}\) , \(n = 5\) ) (Fig. 2e). We further analyzed the spatial positionings of the \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) and AS cells. In the absence of phage T6, the faster- growing AS cells dominated the biomass periphery while the slower- growing \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) cells were positioned behind the periphery where nutrients supplied from the environment were depleted (Fig. 2f). In the presence of phage T6, in contrast, the slower- growing \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) cells persisted at the biomass periphery where nutrients supplied from the environment were plentiful (Fig. 2f).  
+
+<|ref|>text<|/ref|><|det|>[[111, 476, 877, 720]]<|/det|>
+In contrast with strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) , we found that the slower- growing strain \(\mathsf{AR}_{\mathsf{C}_{\mathsf{S}\mathsf{T}\mathsf{R}}}\) was unable to persist when co- cultured with the faster- growing strain AS in the absence of antibiotic pressure regardless of whether phage T6 was present or not (Fig. 2a,b). We attribute this to the larger growth rate difference between strains AS and \(\mathsf{AR}_{\mathsf{C}_{\mathsf{S}\mathsf{T}\mathsf{R}}}\) when compared to the growth rate difference between strains AS and \(\mathsf{AR}_{\mathsf{C}_{\mathsf{TET}}}\) (i.e., for our experimental system, the fitness cost of streptomycin resistance is significantly greater than that of tetracycline resistance) (two- sample two- sided Welch test; \(p = 8.2\times 10^{- 7}\) , \(n = 5\) ) (Fig. 2c). Thus, while phage lysis can increase the persistence of slower- growing AR strains in the face of faster- growing AS strains, this is potentially only true if the fitness cost of antibiotic resistance is not excessively large.  
+
+<|ref|>sub_title<|/ref|><|det|>[[113, 783, 714, 803]]<|/det|>
+## Fitness cost of antibiotic resistance determines the persistence of AR cells  
+
+<|ref|>text<|/ref|><|det|>[[112, 836, 866, 884]]<|/det|>
+To test the notion that the effect of phage lysis on the persistence of slower- growing AR cells depends on the fitness cost of antibiotic resistance (i.e., that excessively large fitness costs can
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 880, 360]]<|/det|>
+eliminate the positive effect of phage lysis), we adapted an individual- based computational model that allows us to simulate co- culture growth across nutrient- amended surfaces<sup>40,44,45</sup>. We performed simulations of co- cultures composed of a faster- growing AS strain (cyan) and a slower- growing AR strain (magenta) and varied the fitness cost of antibiotic resistance and the probability of phage lysis (Fig. 3a and Video 1). As with our experiments, the faster- growing AS cells do not emerge spontaneously during the simulations; rather, they are already present in the inoculum. Also consistent with our experiments, we used a 1:1 initial cell ratio of the two strains and conducted the simulations in the absence of antibiotic pressure. We simulated phage lysis by removing peripheral cells at varying probabilities (between 0.01 and 0.05), which has been found to be a reasonable approximation of the effect of phage lysis<sup>39,40</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[111, 393, 880, 860]]<|/det|>
+We found that the effect of phage lysis on the persistence of slower- growing AR cells does indeed depend on the fitness cost of antibiotic resistance. When we set the probability of phage lysis to zero, the AR strain had a clear growth disadvantage due to the fitness cost associated with antibiotic resistance (Fig. 3a), which is consistent with our experimental results (Fig. 2a). Overall, the proportion of AR cells at the final simulation time- step decreased as the fitness cost of antibiotic resistance increased (Spearman rank correlation test; \(\mathrm{rho} = - 0.99\) , \(p_{BC} = 8.1 \times 10^{- 34}\) ) (Fig. 3b). When we then set the probability of phage lysis to a positive number, we observed increased persistence of the slower- growing AR cells (Pearson correlation test; \(\mathrm{rho} = 0.94\) , \(p_{BC} = 1.0 \times 10^{- 11}\) ) (Fig. 3a,b). The effect size had a unimodal relationship with the fitness cost of antibiotic resistance, where the maximum beneficial effect of phage lysis on the persistence of AR cells occurred at a fitness cost of antibiotic resistance of 0.1 (two- sample two- sided Welch tests; \(\mathrm{cost} = 0.0 \mathrm{vs. cost} = 0.1\) , \(p_{BC} = 4.8 \times 10^{- 4}\) , \(\mathrm{n} = 4\) ; \(\mathrm{cost} = 0.1 \mathrm{vs. cost} = 0.9\) , \(p_{BC} = 6.4 \times 10^{- 3}\) , \(\mathrm{n} = 4\) ) (Fig. 3b and Extended Data Fig. 1). The effect size of phage lysis then declined as the fitness cost of antibiotic resistance increased (Fig. 3b and Extended Data Fig. 1). Thus, the slower- growing AR cells persisted most effectively when the fitness cost of antibiotic resistance was relatively low (Fig. 3b and Extended Data Fig. 1), which is consistent with our experimental observations (Fig. 2a,b).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 88, 880, 670]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 680, 880, 884]]<|/det|>
+<center>Fig. 3: Fitness cost and probability of phage lysis determine the persistence of slower-growing AR cells. a, Representative individual-based computational simulations of co-cultures of strains AS (cyan) and AR (magenta) in the absence or presence of phage lysis with different fitness costs of antibiotic resistance and probabilities of phage lysis. The images are the outputs at the last simulation time step. b, The proportion of AR cells as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. c, The number of AS cells removed by phage lysis as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. d, The number of AR cells removed by phage lysis as a function of the fitness cost of antibiotic resistance for different probabilities for phage lysis. For b–d, each data point is an independent simulation (n = 4), the lines connect the mean values, and the shaded regions are one standard deviation. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 118, 670, 137]]<|/det|>
+## Mechanism for how phage lysis increases the persistence of AR cells  
+
+<|ref|>text<|/ref|><|det|>[[111, 169, 876, 693]]<|/det|>
+To identify a plausible mechanism for how phage lysis can increase the persistence of slower- growing AR cells in the face of faster- growing AS counterparts, we again used our individual- based computational model to count the numbers of AR and AS cells that were removed from the biomass as a consequence of phage lysis. We found that more AS cells than AR cells were removed via phage lysis when there was a fitness cost of antibiotic resistance (two- sample twosided Welch tests; \(p_{BC} = 4.6 \times 10^{- 3}\) , \(n = 4\) ) (Fig. 3c,d). We refer to this outcome as “peripheral kill- the- winner”; faster- growing AS cells have a growth advantage that allows them to disproportionately occupy the biomass periphery, but cells at the biomass periphery are also more susceptible to removal via phage lysis. The disproportional removal of faster- growing AS cells diminishes the benefits of their faster growth rates, allowing slower- growing AR cells to persist. This concept is supported by our individual- based computational simulations where we varied the probability of phage lysis. As the probability of phage lysis increased, the number of faster- growing AS cells that were removed from the biomass also increased (Spearman rank correlation test; \(r_{ho} = 0.90\) , \(p_{BC} = 1.2 \times 10^{- 8}\) ) (Fig. 3c), which correspondingly increased the persistence of slower- growing AR cells (Pearson correlation test; \(r_{ho} = 0.94\) , \(p_{BC} = 1.0 \times 10^{- 11}\) ) (Fig. 3d). While we analyzed all of our simulations at a fixed simulation time for the data presented in Fig. 3, all of our outcomes remain valid when we analyzed our simulations at a fixed total biomass size (Extended Data Fig. 2). Our main outcomes are therefore robust to the simulation end- point.  
+
+<|ref|>sub_title<|/ref|><|det|>[[113, 755, 763, 802]]<|/det|>
+## Phage lysis increases the persistence of slower-growing AR strains in the face of spontaneously emerging AS cells  
+
+<|ref|>text<|/ref|><|det|>[[113, 839, 878, 886]]<|/det|>
+We next tested whether our outcomes remain valid when antibiotic resistance is spontaneously lost during the experiment. To test this, we performed surface- associated growth experiments
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 872, 277]]<|/det|>
+with strain \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) , which contains the non- transmissible plasmid pEF001 that encodes for chloramphenicol resistance and green fluorescent protein (falsely colored magenta in our images). If plasmid pEF001 is lost from an \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cell during cell division, the cell will revert to a non- fluorescent and faster- growing version that we refer to as an AS cell (uncolored). We can therefore propagate the slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells alone on nutrient- amended agar surfaces, track the spontaneous emergence and proliferation of faster- growing AS cells with CLSM, and quantify the persistence of \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells in the face of the newly formed AS cells.  
+
+<|ref|>text<|/ref|><|det|>[[111, 310, 880, 805]]<|/det|>
+We found that phage lysis can indeed increase the persistence of slower- growing AR cells in the face of spontaneously formed and faster- growing AS cells. In the absence of phage T6, we observed substantial loss of plasmid pEF001 within the biomass (Fig. 4a). This corresponded to the proliferation of faster- growing AS cells and a decline in the proportion of slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells from \(100\%\) in the initial inoculum to only \(67\%\) of the total biomass after ten days of incubation (Fig. 4b). Thus, the plasmid- free faster- growing AS cells had a clear competitive advantage over the slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells. When we then added phage T6 to the cocultures, the proportion of slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells declined by a significantly smaller extent from \(100\%\) in the initial inoculum to \(94\%\) after ten days of incubation (two- sample two- sided Welch test; \(p = 3.6 \times 10^{- 4}\) , \(n = 5\) ) (Fig. 4a,b). Thus, the presence of phage T6 reduced the ability of the faster- growing AS cells to establish over the slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells. This was true even though the total biomass area declined (two- sample two- sided Welch test; \(p = 6.1 \times 10^{- 8}\) , \(n = 5\) ) (Fig. 4c). Finally, \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells were more abundant along the biomass periphery when phage T6 was present (two- sample two- sided Welch tests; \(p < 1.5 \times 10^{- 3}\) , \(n = 5\) ) (Fig. 4d). Thus, phage T6 allows the slower- growing \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells to better occupy the biomass periphery where resources are plentiful, which is consistent with our proposed “peripheral kill- the- winner” dynamic (Fig. 1). These outcomes remain valid across various environmental conditions, including anoxic environments and different incubation temperatures (Extended Data Fig. 3).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 90, 870, 528]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 536, 876, 740]]<|/det|>
+<center>Fig. 4: Phage lysis increases the persistence of slower-growing AR cells in the face of spontaneously generated AS cells. a, Representative CLSM images of cultures containing strain \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) (magenta) in the absence (upper image) or presence (lower image) of phage T6. Nonfluorescent regions are composed of \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells that lost plasmid pEF001 and became AS cells. We took the images after ten days of incubation at \(21^{\circ}\mathrm{C}\) in the absence of antibiotic pressure. b, The proportion of the total biomass area occupied by strain \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) in the absence or presence of phage T6. c, The biomass diameter in the absence or presence of phage T6. d, The proportion of \(\mathsf{AR}_{\mathsf{P},\mathsf{CHI}}\) cells as a function of the radial distance from the centroid of the biomass. For b–d, each data point is an independent experimental replicate (n = 5), the black data points are for experiments in the absence of phage T6, and the green data points are for experiments in the presence of phage T6. For b,c, the \(p\) -values are for two-sample two-sided Welch tests. </center>  
+
+<|ref|>text<|/ref|><|det|>[[113, 794, 806, 841]]<|/det|>
+Mechanism for how phage lysis can increase the persistence of AR cells in the face of spontaneously emerging AS cells
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 879, 415]]<|/det|>
+To identify a plausible mechanism for how phage lysis can increase the persistence of AR cells despite the spontaneous emergence of faster- growing AS cells, we incorporated the spontaneous loss of antibiotic resistance into our individual- based computational model. The loss of antibiotic resistance in our model is generic and could occur via a genetic mutation or, as in our experiments, by plasmid loss. We then counted the number of AS cells that emerge from AR cells over the course of the simulations in the absence of antibiotic pressure. We set the entire initial population to be AR cells (magenta), each of which can spontaneously transform into a faster- growing AS cell (grey) according to a defined probability. We then varied the probability of losing antibiotic resistance to be between 0.01 and 0.03 and the fitness cost of antibiotic resistance to be a reduction in the growth rate between \(5\%\) and \(30\%\) . We simulated phage lysis as described in our simulations for chromosomal antibiotic resistance genetic determinants.  
+
+<|ref|>text<|/ref|><|det|>[[111, 450, 880, 806]]<|/det|>
+When we analyzed our simulations at a fixed simulation time, we found that phage lysis does indeed increase the persistence of slower- growing AR cells in the face of faster- growing AS cells. When we set the probability of phage lysis to zero, we observed extensive emergence and proliferation of faster- growing AS cells (Fig. 5a and c and Video 2), which is consistent with our experimental data (Fig. 4). The proportion of slower- growing AR cells significantly decreased as either the probability of losing antibiotic resistance increased (Spearman rank correlation test; rho = - 0.72, \(p_{BC} = 2.6 \times 10^{- 3}\) ) or the fitness cost of antibiotic resistance increased (Spearman rank correlation test; rho = - 0.90, \(p_{BC} = 7.1 \times 10^{- 10}\) ) (Fig. 5c). When we then set the probability of phage lysis to a positive number, the proportion of slower- growing AR cells significantly increased, particularly when the probability of losing antibiotic resistance was \(> 0.02\) and the fitness cost was \(> 20\%\) (two- sample two- sided Welch test; \(p_{BC} = 4.1 \times 10^{- 2}\) , \(n = 4\) ) (Fig. 5b,c and Video 2). These outcomes remain valid when we analyzed our simulations at a fixed total biomass size (Extended Data Fig. 4), and they are therefore robust to the simulation end- point.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 88, 840, 690]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 699, 872, 884]]<|/det|>
+<center>Fig. 5: Properties of spontaneously emerging AS cells determine the persistence of slower-growing AR cells. a,b, Representative individual-based computational simulations of strain AR (magenta) in the (a) absence or (b) presence of phage lysis with different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance. If AR cells undergo a genetic change that causes them to lose antibiotic resistance, such as segregational loss of a plasmid, they are relieved of the fitness cost and become AS cells (grey). The images are the outputs at the last simulation time step. c, The proportion of AR cells as a function of the fitness cost of antibiotic resistance and the probability of losing antibiotic resistance. Each data point is an independent simulation (n = 4), the black data points are in the absence of phage, and the green data points are in the presence of phage. </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 878, 420]]<|/det|>
+To better understand the mechanism causing the increase in AR cells in the presence of phage, we quantified the proportions of slower- growing AR and faster- growing AS cells that are lysed by phage in our simulations. In the presence of phage, a larger proportion of the faster- growing AS cells were lysed compared to the slower- growing AR cells, especially when the AR cells carried a fitness cost and had a higher probability of losing antibiotic resistance (two- sample two- sided Welch tests; \(p_{BC} = 2.4 \times 10^{- 2}\) , \(n = 4\) ) (Extended Data Fig. 5a). The increase in AR cells in the presence of phage cannot be explained by the number of AR cells that lost antibiotic resistance, as the number of such events was either not substantially affected by or was higher in the presence of phage at the end of the simulations (two- sample two- sided Welch test; \(p_{BC} = 4.7 \times 10^{- 3}\) , \(n = 4\) ) (Extended Data Fig. 6a). Thus, our proposed “peripheral kill- the- winner” dynamic remains valid even when AS cells emerge spontaneously. These outcomes remain valid when we analyzed our simulations at a fixed total biomass size (Extended Data Figs. 5b and 6b).  
+
+<|ref|>text<|/ref|><|det|>[[111, 447, 883, 888]]<|/det|>
+We further found that the properties of the spontaneously emerging AS cells are important determinants of the persistence of slower- growing AR cells. The number of AS cells that were lysed by phage significantly increased as the probability of losing antibiotic resistance increased (Spearman rank correlation test; \(\text{rho} = 0.88\) , \(p_{BC} = 1.7 \times 10^{- 6}\) ) but was relatively invariant across different fitness costs of antibiotic resistance (Extended Data Fig. 7). For a probability of losing antibiotic resistance of 0.1 and a fitness cost of 30%, over 2,000 AS cells were removed from the biomass by phage lysis, underscoring the significant impact of phage on the faster- growing AS population. The relationship between number of AS cells and the probability of losing antibiotic resistance remained valid at a fixed biomass size, while the number of AS cells lysed by the phage increased as the fitness cost of antibiotic resistance increased (Spearman rank correlation test; \(\text{rho} = 0.97\) , \(p_{BC} = 7.1 \times 10^{- 19}\) ) (Extended Data Fig. 8a). Meanwhile, the number of slower- growing AR cells lysed by phage decreased as the probability of losing antibiotic resistance increased (Spearman rank correlation test; \(\text{rho} = - 0.66\) , \(p_{BC} = 1.2 \times 10^{- 2}\) ) or the fitness cost increased (Spearman rank correlation test; \(\text{rho} = - 0.99\) , \(p_{BC} = 5.3 \times 10^{- 22}\) ) (Extended Data Fig. 7). These outcomes remain valid when we analyzed our simulations at a fixed total biomass size (Extended Data Fig. 8b).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 145, 664, 165]]<|/det|>
+## Phage lysis promotes biodiversity during surface-associated growth  
+
+<|ref|>text<|/ref|><|det|>[[112, 200, 880, 499]]<|/det|>
+We finally sought to test whether our main outcomes can be generalized beyond simple binary co- cultures of two microbial strains. Can our proposed "peripheral kill- the- winner" dynamic enable larger collections of microbial strains to persist together despite variance in their growth rates? To test this, we extended our individual- based computational model to simulate consortia consisting of between two and ten strains with different growth rates. We set the growth rate of one strain to a value of 1 and then randomly assigned the additional strains to have growth rates ranging between 0 and 1, where we took each growth rate from a uniform distribution. In total, we performed 473 total simulations with 35 to 65 replicates for each number of strains in the consortia. We performed simulations in the absence or presence of phage as described above and quantified strain diversity from the final spatial patterns, including strain richness, Shannon diversity, Simpson diversity, and evenness.  
+
+<|ref|>text<|/ref|><|det|>[[111, 532, 883, 886]]<|/det|>
+We found that our proposed "peripheral kill- the- winner" dynamic can indeed be generalized to systems with more than two strains. The presence of phage did not impact strain richness (Fig 6a), which is not unexpected as we did not impose a mechanism of cell death other than phage lysis. Thus, all strains remained within the inoculation area. However, we found that the Shannon diversity (Paired two- sided t- test; \(p_{\mathrm{BC}} = 1.6 \times 10^{- 7}\) , \(n > 35\) ) (Fig. 6b), Simpson diversity (Fig. 6c) (Paired two- sided t- test; \(p_{\mathrm{BC}} = 4.8 \times 10^{- 7}\) , \(n > 35\) ), and evenness (Paired two- sided t- test; \(p_{\mathrm{BC}} = 1.6 \times 10^{- 07}\) , \(n > 35\) ) (Fig. 6d) all significantly increased in the presence of phage. Thus, phage lysis not only preserved multiple strains with different growth rates (e.g. fitness costs of antibiotic resistance), but also contributed to a more even distribution of those strains (Fig. 6b- d and Extended Data Fig. 9). These outcomes remain valid when we analyzed our simulations at a fixed simulation time (Extended Data Fig. 10). Our proposed "peripheral kill- the- winner" dynamic may therefore be a general mechanism for how diversity can be maintained within surface- associated microbial systems.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[123, 118, 880, 520]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 528, 875, 729]]<|/det|>
+<center>Fig. 6: Phage lysis maintains strain diversity. Data are for individual-based computational simulations of the surface-associated growth of co-cultures consisting of between two and ten strains, where each strain has a different growth rate ranging between zero and one. The growth rate of one strain was set to a value of one and the growth rates of the other strains were sampled from a uniform distribution of growth rates. a-d, The strain diversity metrics were quantified at a fixed total biomass size and include (a) strain richness, (b) Shannon diversity, (c) Simpson diversity, and (d) evenness. For b-d, the boxplots identify the mean values, interquartile ranges, and outliers for a total of 473 independent pairs of simulations (between 35-65 replicates for each number of cell types). The black boxplots and data points are in the absence of phage and the green boxplots and data points are in the presence of phage. The \(p\) -values are for two-sided paired t-tests with a Bonferroni correction. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 231, 113]]<|/det|>
+## Discussion  
+
+<|ref|>text<|/ref|><|det|>[[111, 150, 884, 677]]<|/det|>
+Our findings demonstrate that phage lysis can reshape the spatial organization of surface- associated microbial systems with consequences on competitive outcomes between different microorganisms. More specifically, we demonstrate a mechanism for how phage can enable the persistence of AR bacteria even in the absence of antibiotic pressure and when antibiotic resistance imposes a fitness cost. Consistent with prior studies that highlight the fitness costs of antibiotic resistance in the absence of antibiotic pressure<sup>41,42</sup>, our experiments confirm that slower- growing AR strains are outcompeted by faster- growing AS strains when phage are absent. However, the presence of phage initiates a "peripheral kill- the- winner" dynamic, where phage disproportionately lyse the faster- growing AS cells located at the biomass periphery (Figs. 2 and 4). This creates ecological niches located behind the biomass periphery that allow slower- growing AR strains to persist despite their slower growth. These findings underscore the pivotal role of spatial organization in directing microbial community dynamics and evolution and reveal phage as important drivers of these processes. Importantly, these insights highlight the dual role of phages. On the one hand, they offer therapeutic potential against disease- causing or -exacerbating microorganisms<sup>46,47,48,49,50</sup>, on the other hand, their selective pressures on microbial systems may inadvertently maintain or even promote antibiotic resistance under certain conditions<sup>40</sup>. This duality necessitates careful consideration of the ecological consequences of phage therapy, particularly in scenarios where complete eradication of target populations is challenging.  
+
+<|ref|>text<|/ref|><|det|>[[112, 708, 876, 895]]<|/det|>
+A key insight from our study is the differential impact of phage lysis on AR strains with varying fitness costs. While a tetracycline- resistant strain \((\mathsf{AR}_{\mathsf{C},\mathsf{Tet}})\) substantially benefited from phage- mediated ecological opportunities, the streptomycin- resistant strain \((\mathsf{AR}_{\mathsf{C},\mathsf{Str}})\) did not (Fig. 2). This discrepancy likely reflects the higher fitness cost of streptomycin resistance in our experimental system, which prevents the \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) strain from capitalizing on phage- mediated niche creation (Fig. 3). These findings emphasize that the persistence of AR strains under phage pressure is contingent on the specific fitness burden associated with resistance mechanisms.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 90, 835, 164]]<|/det|>
+Future research should explore this relationship across a broader spectrum of resistance determinants to refine predictions of resistance persistence under varying phage- bacteria dynamics.  
+
+<|ref|>text<|/ref|><|det|>[[111, 201, 881, 526]]<|/det|>
+We further extend the concept of phage- mediated persistence of antibiotic resistance to systems subjected to the spontaneous emergence of antibiotic sensitivity (i.e. spontaneous loss of antibiotic resistance and its associated fitness costs), which could occur via genetic mutation or plasmid loss. This emergence of antibiotic sensitivity is more likely to occur in rapidly dividing cells that are more prone to segregational plasmid loss or DNA replication errors, which are often positioned at the biomass periphery where nutrients are abundant. Once relieved of the fitness cost associated with antibiotic resistance, the faster growing AS cells gain further access to the biomass periphery. However, here we demonstrate that our proposed "peripheral kill- the- winner" dynamic can counteract other peripheral processes occurring in the system, ultimately preserving the antibiotic resistant population (Figs. 4 and 5). Our mechanism is thus able to explain the persistence of antibiotic resistance even in the presence of evolutionary processes that modify antibiotic resistance landscapes.  
+
+<|ref|>text<|/ref|><|det|>[[111, 561, 872, 886]]<|/det|>
+Beyond the maintenance of antibiotic resistance, our "peripheral kill- the- winner" dynamic, offering novel insights into the maintenance of metabolically costly and non- transmissible plasmids. While conventional models attribute the persistence of such plasmids to horizontal gene transfer (HGT) \(^{20,21,51,52}\) , which can facilitate their dissemination and maintenance within microbial communities even though they impose a fitness cost, our findings demonstrate that HGT is not an obligatory mechanism for maintaining these plasmids. Instead, our results reveal an alternative ecological mechanism underpinning plasmid persistence driven by phage lysis. We show that phage preferentially lyse plasmid- free cells that are relieved of their fitness burden, thereby mitigating the competitive disadvantage typically associated with plasmid- bearing strains. This selective pressure not only enables plasmid- carrying antibiotic- resistant cells to persist but also enables them to increase in frequency in the presence of phage independent of HGT. While previous studies have established the role of phage in facilitating
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 884, 220]]<|/det|>
+HGT<sup>40,53,54</sup>, our findings underscore that plasmid maintenance can occur through phage- mediated selection even in the absence of HGT. This mechanism provides a compelling explanation for the evolutionary persistence of metabolically burdensome plasmids in both natural and clinical environments, expanding our understanding of the ecological forces shaping microbial genome evolution.  
+
+<|ref|>text<|/ref|><|det|>[[111, 255, 880, 526]]<|/det|>
+Our results are not specific to simple binary systems consisting of antibiotic resistant and sensitive counterparts, but instead can be generalized to any number of strains that vary in their growth properties (Fig. 6). By selectively removing the fastest growing strains from the biomass periphery, phage will prevent any one strain from dominating the system and lead to a more balanced strain distribution. Our proposed "peripheral kill- the- winner" dynamic may therefore be useful for understanding how many natural environments, such as soils and the human microbiome, are able to sustain such incredible levels of biodiversity. These habitats contain surface- associated microbial communities and are abundant in phage<sup>55,56,57,58,59,60</sup>, and it is therefore plausible that our peripheral kill- the- winner contributes to the maintenance of diversity within these systems.  
+
+<|ref|>text<|/ref|><|det|>[[111, 560, 875, 830]]<|/det|>
+In conclusion, our study uncovers a mechanism that can explain the paradox of how antibiotic resistant bacteria can persist in the absence of antibiotic pressure. By disproportionately lysing cells positioned at the biomass periphery, phage enable the persistence of slower- growing antibiotic resistant strains, challenging traditional models of resistance loss in antibiotic- free conditions. These findings provide new insights into the ecological and evolutionary mechanisms underlying the spread of antibiotic resistance and highlight the complex interplay between phage, microbial spatial organization, and the maintenance of antibiotic resistance. Our results have implications for microbial management, phage therapy, and the global fight against antibiotic resistance, underscoring the importance of integrating spatial and ecological perspectives into resistance mitigation strategies.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 92, 216, 113]]<|/det|>
+## Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 156, 361, 174]]<|/det|>
+## Strains and culture conditions  
+
+<|ref|>text<|/ref|><|det|>[[110, 202, 883, 736]]<|/det|>
+We used isogenic derivatives of E. coli MG1655 for all of our experiments. We used strain TB204 as the antibiotic sensitive strain (referred to as strain AS), which expresses the gfp green fluorescent protein- encoding gene from the chromosome and is under the control of the lambda promoter61. We used strain TB205 to create all of the antibiotic resistant derivative strains used in this study, which is identical to strain TB204 except that it expresses the mcherry fluorescent protein- encoding gene from the chromosome and is under the control of the lambda promoter61. We created tetracycline and streptomycin resistant derivatives of strain TB205 (referred to as strains \(\mathsf{AR}_{\mathsf{C},\mathsf{Test}}\) and \(\mathsf{AR}_{\mathsf{C},\mathsf{Str}}\) respectively) by growing cultures of strain TB205 in liquid lysogeny broth (LB) medium into stationary phase and then inoculating the cultures onto agar plates amended with increasingly large concentrations of tetracycline or streptomycin. We created a chloramphenicol resistant derivative of strain E. coli MG1655 (referred to as strain \(\mathsf{AR}_{\mathsf{P},\mathsf{Chl}}\) ) by introducing the non- mobile plasmid pEF001 into the strain via conjugation. This plasmid encodes for the gfp green fluorescent protein- encoding gene and chloramphenicol resistance. We routinely grew all the strains in liquid LB medium or on solid LB agar plates supplemented with \(10\mu \mathrm{g}\mathrm{mL}^{- 1}\) tetracycline, \(50\mu \mathrm{g}\mathrm{mL}^{- 1}\) streptomycin, or \(25\mu \mathrm{g}\mathrm{mL}^{- 1}\) chloramphenicol, respectively, to maintain their resistance determinants. We preserved all strains in \(15\%\) (v/v) glycerol at \(- 80^{\circ}\mathrm{C}\) . Prior to each experiment, we streaked the individual strains from their respective \(- 80^{\circ}\mathrm{C}\) glycerol stocks onto LB agar plates containing their respective antibiotic and used a single colony to initiate all of our experiments.  
+
+<|ref|>text<|/ref|><|det|>[[112, 763, 882, 895]]<|/det|>
+We used the lytic phage \(\mathsf{T}6^{62}\) for all of our experiments. We propagated the phage using strain TB204 as the host. Briefly, we grew strain TB204 in LB liquid medium at \(37^{\circ}\mathrm{C}\) for four hours with continuous shaking at 150 rpm. After incubation, we purified phage by filtering the culture through a \(0.22\mu \mathrm{m}\) membrane and storing the supernatant at \(4^{\circ}\mathrm{C}\) until further use. For long- term storage, we mixed equal volumes of strain TB204 and phage suspensions and incubated
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 857, 137]]<|/det|>
+the mixtures for ten minutes with continuous shaking at 150 rpm. We then added glycerol to the mixtures (15% v/v) and stored the stocks at - 80°C.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 202, 440, 220]]<|/det|>
+## Surface-associated growth experiments  
+
+<|ref|>text<|/ref|><|det|>[[112, 255, 872, 441]]<|/det|>
+We performed surface- associated growth experiments across LB agar plates<sup>40</sup>. Briefly, we prepared LB agar plates by pouring autoclaved LB medium containing 1% bacteriology- grade agar into sterile 3.5 cm diameter Petri dishes. We then allowed the medium to solidify overnight at room temperature. We next transferred the agar plates to a sterile hood and allowed them to dry with their lids open for ten minutes. Finally, we covered the plates with their lids, sealed them individually with Parafilm (Amcor, Zürich, Switzerland), and stored them at 4°C until use.  
+
+<|ref|>text<|/ref|><|det|>[[112, 478, 875, 691]]<|/det|>
+To perform the experiments, we first prepared overnight cultures of the individual strains in liquid LB medium containing the appropriate antibiotics to maintain their resistance determinants. We then diluted the overnight cultures 1:100 (v/v) into fresh LB medium in the absence of antibiotics and incubated them at 37°C with continuous shaking at 150 rpm for 4 hours to ensure that they were in the exponential growth phase. We next washed the cells to remove all traces of antibiotics by centrifugation at 3600 x g for 10 minutes at 4°C and resuspension in phosphate- buffered saline (PBS). Finally, we adjusted the optical density at 600 nm (OD<sub>600</sub>) of each culture to 1 (approximately 10<sup>8</sup> colony- forming units ml<sup>- 1</sup>).  
+
+<|ref|>text<|/ref|><|det|>[[112, 727, 872, 886]]<|/det|>
+To prepare the phage for the experiments, we mixed the refrigerated phage stock with E. coli host cells and incubated the mixtures at 37°C with continuous shaking at 150 rpm for 4 hours. After incubation, we removed the host cells by filtration through a 0.22 μm membrane, resulting in a cell- free active phage solution. We determined the phage titer using the double- layer agar plate method<sup>63</sup> and then diluted the phage solution in PBS to obtain a concentration of 10<sup>8</sup> plaque forming units ml<sup>- 1</sup>.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 116, 875, 360]]<|/det|>
+For the surface- associated growth experiments with chromosomal antibiotic resistance genetic determinants, we placed a \(1 \mu \text{L}\) droplet of a 1:1 mixture of strains AS and \(\text{AR}_{\text{C, Tet}}\) or strains AS and \(\text{AR}_{\text{C, Str}}\) (OD \(_{600}\) of the mixture of 1) onto the centers of individual replicated LB agar plates (n = 5 for each treatment). For surface- associated growth experiments with plasmid- encoded antibiotic resistance genetic determinants, we placed a \(1 \mu \text{L}\) droplet of strain \(\text{AR}_{\text{P, Chl}}\) onto the centers of individual replicated LB agar plates (n = 5 for each treatment). We then incubated the agar plates for 6 hours. Thereafter, we added a \(1 \mu \text{L}\) droplet of either phage solution or phage- free PBS as a control to the biomass of each agar plate. Finally, we incubated the agar plates in oxic conditions at \(21^{\circ}\text{C}\) for ten days.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 422, 564, 443]]<|/det|>
+## Confocal laser scanning microscopy and image analysis  
+
+<|ref|>text<|/ref|><|det|>[[113, 476, 875, 608]]<|/det|>
+At the end of the surface- associated growth experiments, we imaged the biomass using a Leica TCS SP5 II confocal laser scanning microscope (CLSM) (Leica Microsystems, Wetzlar, Germany) equipped with a \(2.5 \text{x} \text{HCX} \text{FL}\) objective, a numerical aperture of 0.12, and a frame size of \(512 \times 512\) pixels. We set the laser to 488 nm for the excitation of GFP fluorescence and 514 nm for the excitation of RFP fluorescence.  
+
+<|ref|>text<|/ref|><|det|>[[112, 643, 874, 887]]<|/det|>
+We quantified the proportions of different strains as a function of biomass radius using ImageJ (https://imagej.nih.gov/ij/) with Fiji plugins (v. 2.1.0/1.53c) (https://fiji.sc). We first loaded the image files into ImageJ and separated the channels to enable the analysis of each strain. We then selected a region of interest and applied an automated threshold using the Yen algorithm optimized for dark backgrounds. We next converted the images into binary masks to minimize background noise and enhance the signal- to- noise ratio. After isolating the target features, we applied a circular Region of Interest (ROI) of specified coordinates and radius to capture the spatial distributions of the strains within a consistent area. We executed the 'Radial Profile' function to obtain the intensity distribution across this circular ROI. This distribution served as
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 875, 221]]<|/det|>
+the basis for determining the radial intensity profile, which is essential for quantifying the community proportions along the circumference. Using the radial intensity profiles of each strain obtained from the ROI, we measured the relative abundances or proportions of each strain along the defined circumference. We repeated this process across all channels to ensure uniformity and comparability in data acquisition.  
+
+<|ref|>sub_title<|/ref|><|det|>[[114, 285, 458, 303]]<|/det|>
+## Individual-based computational modeling  
+
+<|ref|>text<|/ref|><|det|>[[111, 338, 880, 636]]<|/det|>
+We employed the open- source modeling software Cellmodeller version 4.3 framework<sup>44,45</sup> (https://github.com/cellmodeller/CellModeller) and made modifications based on the "MicrobialEcologyToolbox" branch to carry out individual- based simulations of surface- associated microbial growth. We performed all simulations on the Euler computing cluster at ETH (https://scicomp.ethz.ch/wiki/Euler) using the Slurm workload manager for batch job submissions. We wrote sections of the code for scheduling batch jobs with the aid of ChatGPT- 4 (https://openai.com/gpt- 4). We simulated bacterial cells as three- dimensional capsules of length L that grow uniaxially. Upon reaching a predetermined target length, cells are divided into two daughter cells, each inheriting the characteristics of the parent cell. We set the biophysical parameter "gamma," which controls the ratio of drag force on cell movement to cell growth, to 20 for all our simulations.  
+
+<|ref|>text<|/ref|><|det|>[[111, 671, 884, 886]]<|/det|>
+For the simulation of \(E\) . coli cells, we defined the initial cell shape to have a radius of \(0.5 \mu \mathrm{m}\) and a length of \(2 \mu \mathrm{m}\) . Cells divided when their length reached a target length (Ldiv), where Ldiv was randomly sampled from a Gaussian distribution of 2 to \(2.5 \mu \mathrm{m}\) and daughter cells had a length of Ldiv/2. We set the growth rate of AS cells to 1, indicating that each cell would ideally grow by \(1 \mu \mathrm{m}\) per unit time step in the absence of physical constraints. To examine the impact of phage lysis, we implemented the killflag feature in CellModeller, which dynamically removed cells from the simulations and updated the cell state list to retain only living cells<sup>40</sup>. Briefly, the model determined the outermost ring of cells in the simulation space and, based on a random
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 857, 221]]<|/det|>
+probability (kill_rate), flagged cells to stop growing and be removed from the simulation. We varied the kill_rate, emulating phage lysis, between 0 (no phage lysis) and 0.5 (strong phage lysis) in our simulations, where we refer to this value as the phage lysis probability. Thus, instead of simulating the complex biophysical processes of phage replication and lysis, we modeled the phage as a signal that initiates cell removal<sup>40</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[111, 255, 880, 830]]<|/det|>
+Using our model, we set the initial conditions to achieve three objectives regarding the effect of phage lysis using the parameters presented in Extended Data Table 1. First, to model the growth of \(\mathsf{AR}_{\mathsf{C}}\) cells, we initialized the simulations with 1,000 cells each of AS and AR cells (1:1 ratio) randomly distributed and rotationally oriented across a circular area of radius 50. We set the growth rate of the AR cells to be between 0.1 and 1, which corresponds to a fitness cost of antibiotic resistance between 0.9 and 0. We performed simulations until the total number of cells reached 40,000. Second, to model the growth of \(\mathsf{AR}_{\mathsf{P}}\) cells, we initialized the simulations with 2,000 \(\mathsf{AR}_{\mathsf{P}}\) cells with a fitness cost for carrying the plasmid. We simulated the spontaneous emergence of AS cells through plasmid loss during cell division by modifying the division function of cells. Briefly, with a certain probability during cell division, \(\mathsf{AR}_{\mathsf{P}}\) cells can switch to AS cells that no longer carry a fitness cost. We varied the fitness cost of antibiotic resistance to be between 0 and 0.3 and the probability that an AR cell will turn into an AS cell to between 0.01 and 0.03. We performed simulations until the total number of cells reached 40,000. Third, to model the growth of multiple strains that differ in their growth rate, we initialized the simulations with 2,000 total cells comprised of equal numbers of between two and ten strains. We next set the fitness cost of the first strain to 0 and randomly sampled from a uniform distribution between 0 and 1 to assign fitness costs to all the other strains in the simulation. Thus, the growth rate of all the other strains ranged between 1 and 0 (fitness cost between 0 and 1) relative to the first strain. We performed simulations until the total number of cells reached 40,000 and then quantified the strain richness, Shannon diversity, Simpson diversity, and evenness using standard methods<sup>64</sup>.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[113, 91, 270, 108]]<|/det|>
+## Statistical analyses  
+
+<|ref|>text<|/ref|><|det|>[[111, 145, 880, 444]]<|/det|>
+We performed all statistical tests with Python (version 3.11.5) using the scipy.stats module (version 1.11.4). To test for differences between means, we employed two- sample two- sided Welch tests for comparing independent groups, which do not assume homoscedasticity in the datasets, and paired t- tests when comparing the same samples under two different treatments. For correlation measurements, we used Pearson's correlation test for linear relationships and Spearman's rank correlation test for monotonic but non- linear relationships. When performing statistical tests on the same data multiple times, we adjusted the \(p\) - values using the Bonferroni correction, which we designate as \(p_{BC}\) . We tested the normality of our datasets using the Shapiro- Wilk test and did not observe significant deviations from the assumption of normality \((p < 0.05)\) . For each statistical test, we reported the specific test used, the corresponding \(p\) - value, and the sample size in the Results section.  
+
+<|ref|>sub_title<|/ref|><|det|>[[113, 516, 448, 539]]<|/det|>
+## Data and materials availability  
+
+<|ref|>text<|/ref|><|det|>[[111, 577, 880, 764]]<|/det|>
+All data and code generated in this study will be deposited in the publicly accessible Eawag Research Data Institutional Repository (https://opendata.eawag.ch/) at the point of revision. All data and code are currently available to the reviewers and editors at the following link: https://drive.switch.ch/index.php/s/Ts4EoWwwgpOUeFm. All bacterial strains and plasmid pEM001 are freely available from the corresponding author upon request. The sequence of pEM001 will be deposited in NCBI at the point of revision. The sequence of pEM001 is currently available to the reviewers and editors as an attached data file to this submission.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 239, 112]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[111, 123, 884, 870]]<|/det|>
+1. Cassini, A. et al. Attributable deaths and disability-adjusted life-years caused by infections with antibiotic-resistant bacteria in the EU and the European Economic Area in 2015: a population-level modelling analysis. Lancet Infect. Dis. 19, 56-66 (2019).  
+2. Murray, C. J. L. et al. Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis. Lancet 399, 629-655 (2022).  
+3. Vogwill, T. & MacLean, R. C. The genetic basis of the fitness costs of antimicrobial resistance: a meta-analysis approach. Evol. Appl. 8, 284-295 (2015).  
+4. San Millan, A. & MacLean, R. C. Fitness costs of plasmids: a limit to plasmid transmission. Microbiol. Spectr. 5, 10.1128/microbiolspec.mtbp-0016-2017 (2017).  
+5. Seppälä, H. et al. The Effect of changes in the consumption of macrolide antibiotics on erythromycin resistance in Group A Streptococcus in Finland. Finnish Study Group for Antimicrobial Resistance. N. Engl. J. Med. 337, 441-446 (1997).  
+6. Enne, V. I., Livermore, D. M., Stephens, P. & Hall, L. M. Persistence of sulphonamide resistance in Escherichia coli in the UK despite national prescribing restriction. Lancet 357, 1325-1328 (2001).  
+7. Gottesman, B. S., Carmeli, Y., Shitrit, P. & Chowers, M. Impact of quinolone restriction on resistance patterns of Escherichia coli isolated from urine by culture in a community setting. Clin. Infect. Dis. 49, 869-875 (2009).  
+8. Bengtsson-Palme, J., Kristiansson, E. & Joakim Larsson, D. G. Environmental factors influencing the development and spread of antibiotic resistance. FEMS Microbiol. Rev. 42, fux053 (2018).  
+9. Joakim Larsson, D. G. & Flach, C.-F. Antibiotic resistance in the environment. Nat. Rev. Microbiol. 20, 257-269 (2022).  
+10. Hughes, D. & Andersson, D. I. Evolutionary trajectories to antibiotic resistance. Annu. Rev. Microbiol. 71, 579-596 (2017).  
+11. Wong, A. Epistasis and the evolution of antimicrobial resistance. Front. Microbiol. 8, 246 (2017).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 90, 857, 137]]<|/det|>
+12. Durão, P., Balbontín, R. & Gordo, I. Evolutionary mechanisms shaping the maintenance of antibiotic resistance. Trends Microbiol. 26, 677-691 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[110, 145, 812, 191]]<|/det|>
+13. Darby, E. M., et al. Molecular mechanisms of antibiotic resistance revisited. Nat Rev Microbiol 21, 280-295 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[110, 200, 852, 247]]<|/det|>
+14. MacLean, R. C. & San Millan, A. The evolution of antibiotic resistance. Science 365, 1082-1083 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[110, 255, 828, 303]]<|/det|>
+15. Hall, J. P. J. et al. Plasmid fitness costs are caused by specific genetic conflicts enabling resolution by compensatory mutation. PLoS Biol. 19, e3001225.  
+
+<|ref|>text<|/ref|><|det|>[[110, 310, 868, 358]]<|/det|>
+16. Baker-Austin, C., Wright, M. S., Stepanauskas, R. & McArthur, J. V. Co-selection of antibiotic and metal resistance. Trends Microbiol. 14, 176-182 (2006).  
+
+<|ref|>text<|/ref|><|det|>[[110, 366, 861, 414]]<|/det|>
+17. Wales, A. & Davies, R. Co-selection of resistance to antibiotics, biocides and heavy metals, and its relevance to foodborne pathogens. Antibiotics 4, 567-604 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[110, 421, 850, 469]]<|/det|>
+18. Palomino, A. et al. Metabolic genes on conjugative plasmids are highly prevalent in Escherichia coli and can protect against antibiotic treatment. ISME J. 17, 151-162 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[110, 477, 850, 552]]<|/det|>
+19. Curtsinger, H. D., Martínez-Absalón, S., Liu, Y. & Lopatkin, A. J. The metabolic burden associated with plasmid acquisition: an assessment of the unrecognized benefits to host cells. BioEssays 11, e2400164 (2024).  
+
+<|ref|>text<|/ref|><|det|>[[110, 560, 828, 607]]<|/det|>
+20. Grohmann, E., Muth, G. & Espinosa, M. Conjugative plasmid transfer in Gram-positive bacteria. Microbiol. Mol. Biol. Rev. 67, 277-301 (2003).  
+
+<|ref|>text<|/ref|><|det|>[[110, 615, 872, 662]]<|/det|>
+21. Lopatkin, A. J. et al. Persistence and reversal of plasmid-mediated antibiotic resistance. Nat. Commun. 8, 1689 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[110, 670, 797, 717]]<|/det|>
+22. Lopatkin, A. J. et al. Antibiotics as a selective driver for conjugation dynamics. Nat. Microbiol. 1, 16044 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[110, 726, 856, 773]]<|/det|>
+23. Baym, M. et al. Spatiotemporal microbial evolution on antibiotic landscapes. Science 353, 1147-1151 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[110, 782, 857, 857]]<|/det|>
+24. Ruan, C., Ramoneda, J., Chen, G., Johnson, D. R. & Wang, G. Evaporation-induced hydrodynamics promote conjugation-mediated plasmid transfer in microbial populations. ISME Commun. 1, 54 (2021).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 88, 825, 137]]<|/det|>
+25. Sharma, A. & Wood, K. B. Spatial segregation and cooperation in radially expanding microbial colonies under antibiotic stress. ISME J. 15, 3019-3033 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[110, 145, 880, 193]]<|/det|>
+26. Metzger, G. A. et al. Biofilms preserve the transmissibility of a multi-drug resistance plasmid. NPJ Biofilms Microbiomes 8, 95 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[110, 200, 858, 276]]<|/det|>
+27. Ruan, C., Ramoneda, J., Gogia, G., Wang, G. & Johnson, D. R. Fungal hyphae regulate bacterial diversity and plasmid-mediated functional novelty during range expansion. Curr. Biol. 32, 5285-5294. e4 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[110, 283, 852, 331]]<|/det|>
+28. Denk-Lobnig, M. & Wood, K. B. Antibiotic resistance in bacterial communities. Curr. Opin. Microbiol. 74, 102306 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[110, 339, 850, 414]]<|/det|>
+29. Ruan, C., Borer, B., Ramoneda, J., Wang, G. & Johnson, D. R. Evaporation-induced hydrodynamics control plasmid transfer during surface-associated microbial growth. NPJ Biofilms Microbiomes 9, 58 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[110, 421, 852, 469]]<|/det|>
+30. Freire, T. F. A., Hu, Z., Wood, K. B. & Gjini, E. Modeling spatial evolution of multi-drug resistance under drug environmental gradients. PLoS Comput. Biol. 20, e1012098 (2024).  
+
+<|ref|>text<|/ref|><|det|>[[110, 476, 872, 552]]<|/det|>
+31. Ma, Y., Kan, A. & Johnson, D. R. Metabolic interactions control the transfer and spread of plasmid-encoded antibiotic resistance during surface-associated microbial growth. Cell Rep. 43, 114653 (2024).  
+
+<|ref|>text<|/ref|><|det|>[[110, 560, 870, 608]]<|/det|>
+32. Peischl, S., Dupanloup, I., Kirkpatrick, M. & Excoffier, L. On the accumulation of deleterious mutations during range expansions. Mol. Ecol. 22, 5972-5982 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[110, 615, 876, 663]]<|/det|>
+33. Bosshard, L. et al. Accumulation of deleterious mutations during bacterial range expansions. Genetics 207, 669-684 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[110, 670, 860, 718]]<|/det|>
+34. Kayser, J., Schreck, C. F., Gralka, M., Fusco, D. & Hallatschek, O. Collective motion conceals fitness differences in crowded cellular populations. Nat. Ecol. Evol. 3, 125-134 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[110, 725, 848, 802]]<|/det|>
+35. Abe, K., Nomura, N. & Suzuki, S. Biofilms: hot spots of horizontal gene transfer (HGT) in aquatic environments, with a focus on a new HGT mechanism. FEMS Microbiol. Ecol. 96, fiaa031 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[110, 810, 879, 858]]<|/det|>
+36. Hallatschek, O., Hersen, P., Ramanathan, S. & Nelson, D. R. Genetic drift at expanding frontiers promotes gene segregation. Proc. Natl. Acad. Sci. U. S. A. 104, 19926-19930 (2007).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 90, 872, 137]]<|/det|>
+37. Hallatschek, O. & Nelson, D. R. Life at the front of an expanding population. Evol. Int. J. Org. Evol. 64, 193-206 (2010). 
+
+<|ref|>text<|/ref|><|det|>[[110, 146, 875, 192]]<|/det|>
+38. Mitri, S., Clarke, E. & Foster, K. R. Resource limitation drives spatial organization in microbial groups. ISME J. 10, 1471-1482 (2016). 
+
+<|ref|>text<|/ref|><|det|>[[110, 202, 872, 275]]<|/det|>
+39. Eriksen, R. S., Svenningsen, S. L., Sneppen, K. & Mitarai, N. A growing microcolony can survive and support persistent propagation of virulent phages. Proc. Natl. Acad. Sci. U. S. A. 115, 337-342 (2018). 
+
+<|ref|>text<|/ref|><|det|>[[110, 285, 820, 330]]<|/det|>
+40. Ruan, C. et al. Phage predation accelerates the spread of plasmid-encoded antibiotic resistance. Nat. Commun. 15, 5397 (2024). 
+
+<|ref|>text<|/ref|><|det|>[[110, 340, 833, 386]]<|/det|>
+41. Andersson, D. I. & Hughes, D. Antibiotic resistance and its cost: is it possible to reverse resistance? Nat. Rev. Microbiol. 8, 260-271 (2010). 
+
+<|ref|>text<|/ref|><|det|>[[110, 396, 875, 442]]<|/det|>
+42. Carroll, A. C. & Wong, A. Plasmid persistence: costs, benefits, and the plasmid paradox. Can. J. Microbiol. 64, 293-304 (2018). 
+
+<|ref|>text<|/ref|><|det|>[[110, 451, 864, 497]]<|/det|>
+43. Yang, Q. E. et al. Compensatory mutations modulate the competitiveness and dynamics of plasmid-mediated colistin resistance in Escherichia coli clones. ISME J. 14, 861-865 (2020). 
+
+<|ref|>text<|/ref|><|det|>[[110, 506, 835, 552]]<|/det|>
+44. Rudge, T. J., Steiner, P. J., Phillips, A. & Haseloff, J. Computational modeling of synthetic microbial biofilms. ACS Synth. Biol. 1, 345-352 (2012). 
+
+<|ref|>text<|/ref|><|det|>[[110, 562, 874, 608]]<|/det|>
+45. Rudge, T. J., Federici, F., Steiner, P. J., Kan, A. & Haseloff, J. Cell polarity-driven instability generates self-organized, fractal patterning of cell layers. ACS Synth. Biol. 2, 705-714 (2013). 
+
+<|ref|>text<|/ref|><|det|>[[110, 618, 875, 663]]<|/det|>
+46. Chegini, Z. et al. Bacteriophage therapy against Pseudomonas aeruginosa biofilms: a review. Ann. Clin. Microbiol. Antimicrob. 19, 45 (2020). 
+
+<|ref|>text<|/ref|><|det|>[[110, 673, 819, 719]]<|/det|>
+47. Duan, Y., Young, R. & Schnabl, B. Bacteriophages and their potential for treatment of gastrointestinal diseases. Nat. Rev. Gastroenterol. Hepatol. 19, 135-144 (2022). 
+
+<|ref|>text<|/ref|><|det|>[[110, 728, 850, 774]]<|/det|>
+48. Uyttebroek, S. et al. Safety and efficacy of phage therapy in difficult-to-treat infections: a systematic review. Lancet Infect. Dis. 22, e208-e220 (2022). 
+
+<|ref|>text<|/ref|><|det|>[[110, 784, 867, 830]]<|/det|>
+49. Kviatcovsky, D., Valdés-Mas, R., Federici, S. & Elinav, E. Phage therapy in noncommunicable diseases. Science 382, 266-267 (2023). 
+
+<|ref|>text<|/ref|><|det|>[[110, 840, 870, 886]]<|/det|>
+50. Strathdee, S. A., Hatfull, G. F., Mutalik, V. K. & Schooley, R. T. Phage therapy: from biological mechanisms to future directions. Cell 186, 17-31 (2023).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 770, 109]]<|/det|>
+51. Harrison, E. & Brockhurst, M. A. Plasmid-mediated horizontal gene transfer is a  
+
+<|ref|>text<|/ref|><|det|>[[141, 118, 639, 137]]<|/det|>
+coevolutionary process. Trends Microbiol. 20, 262- 267 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[111, 145, 852, 220]]<|/det|>
+52. Ding, M. et al. Subinhibitory antibiotic concentrations promote the horizontal transfer of plasmid-borne resistance genes from Klebsiella pneumoniae to Escherichia coli. Front. Microbiol. 13, 1017092 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[111, 228, 877, 303]]<|/det|>
+53. Schneider, C. L. Bacteriophage-Mediated Horizontal Gene Transfer: Transduction. in Bacteriophages: Biology, Technology, Therapy (eds. Harper, D. R., Abedon, S. T., Burrowes, B. H. & McConville, M. L.) 151-192 (Springer International Publishing, Cham, 2021).  
+
+<|ref|>text<|/ref|><|det|>[[111, 310, 830, 359]]<|/det|>
+54. Watson, B. N. J., Staals, R. H. J. & Fineran, P. C. CRISPR-Cas-mediated phage resistance enhances horizontal gene transfer by transduction. mBio 9, e02406-17 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[111, 366, 860, 442]]<|/det|>
+55. Williamson, K. E., Fuhrmann, J. J., Wommack, K. E. & Radosevich, M. Viruses in soil ecosystems: an unknown quantity within an unexplored territory. Annu. Rev. Virol. 4, 201-219 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[111, 450, 840, 498]]<|/det|>
+56. Pratama, A. A. & van Elsas, J. D. The 'neglected' soil virome - potential role and impact. Trends Microbiol. 26, 649-662 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[111, 505, 866, 554]]<|/det|>
+57. Shkoporov, A. N. & Hill, C. Bacteriophages of the human gut: the "known kinknown" of the microbiome. Cell Host Microbe 24, 195-209 (2019)  
+
+<|ref|>text<|/ref|><|det|>[[111, 561, 875, 609]]<|/det|>
+58. Liang, G. & Bushman, F. D. The human virome: assembly, composition and host interactions. Nat. Rev. Microbiol. 19, 514-527 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[111, 616, 872, 664]]<|/det|>
+59. Cao, Z. et al. The gut microme: a new microbiome component in health and disease. Lancet 81, 104113 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[111, 672, 860, 719]]<|/det|>
+60. Jansson, J. K. & Wu, R. Soil viral diversity, ecology and climate change. Nat. Rev. Microbiol. 21, 296-311 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[111, 727, 870, 802]]<|/det|>
+61. Tomasek, K., Bergmiller, T. & Guet, C. C. Lack of cations in flow cytometry buffers affect fluorescence signals by reducing membrane stability and viability of Escherichia coli strains. J. Biotechnol. 268, 40-52 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[111, 810, 872, 886]]<|/det|>
+62. Huang, W. M., Wei, L. S. & Casjens, S. Relationship between bacteriophage T4 and T6 DNA topoisomerases. T6 39-protein subunit is equivalent to the combined T4 39- and 60-protein subunits. J. Biol. Chem. 260, 8973-8977 (1985).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 856, 140]]<|/det|>
+63. Stachurska, X., Roszak, M., Jabłońska, J., Mizielińska, M. & Nawrotek, P. Double-layer agar (DLA) modifications for the first step of the phage-antibiotic synergy (PAS) identification.  
+
+<|ref|>text<|/ref|><|det|>[[142, 146, 365, 165]]<|/det|>
+Antibiotics 10, 1306 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[111, 173, 847, 222]]<|/det|>
+64. Roswell, M., Dushoff, J. & Winfree, R. A conceptual guide to measuring species diversity. Oikos 130, 321-338 (2021).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 285, 320, 308]]<|/det|>
+## Acknowledgments  
+
+<|ref|>text<|/ref|><|det|>[[112, 348, 872, 480]]<|/det|>
+We thank Zachary Bailey for providing phage T6; Dr. Martin Ackermann and Emanuele Fara for providing all E. coli strains and plasmid pEF001; and Drs. Mireia Cordero and Josep Ramoneda for helpful discussions. We thank Mrs. Ruan (Guo Chen) for her excellent assistance with structuring the simulation videos. C.R. and D.P.V. were supported by a grant from the Swiss National Science Foundation (310030_207471) awarded to D.R.J.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 545, 285, 568]]<|/det|>
+## Contributions  
+
+<|ref|>text<|/ref|><|det|>[[112, 611, 868, 742]]<|/det|>
+C.R. and D.P.V. conceived and developed the research question. C.R., D.P.V., and D.R.J. designed the experiments. C.R. performed all the experiments. D.P.V. performed all the individual-based computational simulations. C.R. and D.P.V. analyzed the data. C.R. and D.P.V. wrote the first version of the manuscript with contributions from D.R.J. D.R.J. coordinated the project. All authors reviewed and approved the final version of the manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 808, 365, 832]]<|/det|>
+## Competing interests  
+
+<|ref|>text<|/ref|><|det|>[[115, 876, 469, 895]]<|/det|>
+The authors declare no competing interests.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[63, 95, 297, 113]]<|/det|>
+850 Extended Data
+
+<|ref|>text<|/ref|><|det|>[[60, 137, 90, 146]]<|/det|>
+851
+
+<|ref|>text<|/ref|><|det|>[[60, 155, 870, 168]]<|/det|>
+852 **Extended Data Table 1: Parameters used for the individual-based computational simulations**
+
+<|ref|>text<|/ref|><|det|>[[60, 174, 90, 183]]<|/det|>
+853
+
+<|ref|>table<|/ref|><|det|>[[115, 188, 880, 450]]<|/det|>
+
+<table><tr><td>Parameter</td><td>Description</td><td>Value</td><td>Unit</td></tr><tr><td>\(g_{AS}\)</td><td>Specific growth rate of an AS cell</td><td>1</td><td>-</td></tr><tr><td>\(g_{AR}\)</td><td>Specific growth rate of an AR cell</td><td>0.1-0.95</td><td>-</td></tr><tr><td>\(L_{0}\)</td><td>Mean initial length of a cell</td><td>2</td><td>μm</td></tr><tr><td>\(L_{d}\)</td><td>Length of a cell at division</td><td>3.5-4</td><td>μm</td></tr><tr><td>\(W_{infection}\)</td><td>Width of the outer cell layer affected by</td><td>4</td><td>μm</td></tr><tr><td>\(d_{cell}\)</td><td>Mean diameter of a cell</td><td>0.5</td><td>μm</td></tr><tr><td>\(N\)</td><td>Initial number of cells</td><td>2000</td><td>cell</td></tr><tr><td>\(R\)</td><td>Probability of phage lysis</td><td>0.01-0.05</td><td>-</td></tr><tr><td>\(P\)</td><td>Probability of losing antibiotic resistance</td><td>0.01-0.03</td><td>-</td></tr></table>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[144, 100, 852, 415]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[113, 444, 881, 536]]<|/det|>
+Extended Data Fig. 1: Fitness cost of antibiotic resistance determines the effect of phage lysis on the persistence of antibiotic resistant (AR) cells. The effect of phage on the proportion of AR cells is the proportion of AR cells in the presence of phage minus the proportion of AR cells in the absence of phage. All quantities are those at a fixed simulation time. Each data point is an independent simulation (n = 4).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 92, 870, 230]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[112, 270, 883, 456]]<|/det|>
+Extended Data Fig. 2: Fitness cost of antibiotic resistance and probability of phage lysis determine the persistence of slower- growing AR strains at a fixed total biomass size. All data are for individual- based computational simulations of co- cultures consisting of strains AS and AR with different fitness costs of antibiotic resistance and probabilities of phage lysis. a, The proportion of AR cells as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. b, The number of AS cells removed by phage as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. c, The number of AR cells removed by phage as a function of the fitness cost of antibiotic resistance for different probabilities of phage lysis. For a–c, each data point is an independent simulation (n = 4), the lines connect the mean values, and the shaded regions are one standard deviation.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[262, 90, 732, 625]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[112, 661, 884, 848]]<|/det|>
+Extended Data Fig. 3: Phage lysis increases the persistence of slower-growing AR cells in the face of spontaneously generated AS cells across different environmental conditions. a,b Representative CLSM images of cultures of strain \(\mathrm{AR_{P,CHI}}\) in the (a) absence or (b) presence of phage T6 after 16 days of growth in oxic conditions at \(21^{\circ}\mathrm{C}\) in the absence of antibiotic pressure. c,d, Representative CLSM images of cultures of strain \(\mathrm{AR_{P,CHI}}\) in the (c) absence or (d) presence of phage T6 after ten days of growth in oxic conditions at \(30^{\circ}\mathrm{C}\) in the absence of antibiotic pressure. e,f, Representative CLSM images of cultures of strain \(\mathrm{AR_{P,CHI}}\) in the (e) absence or (f) presence of phage T6 after ten days of growth in anoxic conditions at \(21^{\circ}\mathrm{C}\) in the absence of antibiotic pressure. For all images, fluorescent (magenta) regions are \(\mathrm{AR_{P,CHI}}\) cells and non- fluorescent regions are \(\mathrm{AR_{P,CHI}}\) cells that lost plasmid pEF001 and became AS cells.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[130, 90, 861, 401]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 417, 880, 546]]<|/det|>
+<center>Extended Data Fig. 4: Properties of spontaneously emerging AS cells determine the persistence of slower-growing AR strains at a fixed total biomass size. All data are for individual-based computational simulations of strain AR in the absence or presence of phage lysis. The proportion of AR cells is calculated for different fitness costs of antibiotic resistance, probabilities of losing antibiotic resistance, and probabilities of phage lysis. Each data point is an independent simulation (n = 4), the black data points are in the absence of phage, and the green data points are in the presence of phage. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[118, 90, 850, 721]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 733, 876, 750]]<|/det|>
+<center>Extended Data Fig. 5: Faster-growing emergent AS cells are disproportionately lysed by phage. </center>  
+
+<|ref|>text<|/ref|><|det|>[[113, 752, 876, 881]]<|/det|>
+All data are for individual- based computational simulations of strain AR in the absence or presence of phage lysis. The proportions of lysed AS (black) and AR (green) cells are calculated for different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance. a,b, The proportions are the number of lysed cells of one strain to the total number of cells of that strain at (a) a fixed simulation time, or (b) a fixed total biomass size. Each data point is an independent simulation (n = 4), the black data points are in the absence of phage, and the green data points are in the presence of phage.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[118, 90, 850, 720]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[113, 733, 875, 863]]<|/det|>
+Extended Data Fig. 6: Fitness cost of antibiotic resistance, probability of phage lysis, and probability of losing antibiotic resistance determine the total number of antibiotic resistance loss events. a,b, All data are for individual- based computational simulations of strain AR for different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance in the absence or presence of phage at (a) a fixed simulation time, or (b) a fixed total biomass size. Each data point is an independent simulation (n = 4), the black data points are for a probability of phage lysis of zero, and the green data points are for a probability of phage lysis of 0.05.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[117, 88, 850, 720]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[112, 733, 877, 880]]<|/det|>
+Extended Data Fig. 7: Fitness cost of antibiotic resistance, probability of phage lysis, and probability of losing antibiotic resistance determine the removal of AS and AR cells. a,b, All data are for individual- based computational simulations of strain AR in the absence or presence of phage at a fixed simulation time. a,b, The number of (a) AS cells or (b) AR cells removed by phage for different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance. For a,b, each data point is an independent simulation (n = 4), the black data points are for a probability of phage lysis of zero, and the green data points are for a probability of phage lysis of 0.05.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[118, 90, 848, 720]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[112, 732, 863, 880]]<|/det|>
+Extended Data Fig. 8: Fitness cost of antibiotic resistance, probability of phage lysis, and probability of losing antibiotic resistance determine the removal of AS and AR cells. All data are for individual- based computational simulations of strain AR in the absence or presence of phage at a fixed total biomass size. a,b, The number of (a) AS cells or (b) AR cells removed by phage for different fitness costs of antibiotic resistance and probabilities of losing antibiotic resistance. For a,b, each data point is an independent simulation (n = 4), the black data points are for a probability of phage lysis of zero, and the green data points are for a probability of phage lysis of 0.05.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[133, 90, 840, 500]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[113, 517, 861, 644]]<|/det|>
+Extended Data Fig. 9: Phage lysis maintains strain diversity. Representative individual- based computational simulations of co- cultures consisting of between two and ten distinct strains in the absence or presence of phage. The images are the outcomes at a fixed total biomass size. Each strain has a different growth rate ranging between zero and one that was randomly assigned by sampling from a uniform distribution. The color of each cell corresponds to its growth rate, represented as a gradient from magenta (growth rate = 0) to cyan (growth rate = 1).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[122, 91, 850, 470]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[112, 483, 880, 707]]<|/det|>
+Extended Data Fig. 10: Phage lysis maintains strain diversity when analyzing simulations at a fixed simulation time. Data are for individual- based computational simulations of the surface- associated growth of co- cultures of between two and ten strains, where each strain has a different growth rate ranging between zero and one. The growth rate of one strain was set to a value of one and the growth rates of the other strains were sampled from a uniform distribution of growth rates. a- d, The strain diversity metrics are those quantified at the last simulation time and include (a) strain richness, (b) Shannon diversity, (c) Simpson diversity, and (d) evenness. For b- d, the boxplots identify the mean values, interquartile ranges, and outliers for 473 independent pairs of simulations in total (between 35- 65 replicates for any given number of cell types). The black boxplots and data points are in the absence of phage and the green boxplots and data points are in the presence of phage. The \(p\) - values are for two- sided paired t- tests with a Bonferroni correction.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 93, 301, 117]]<|/det|>
+## Video Captions  
+
+<|ref|>text<|/ref|><|det|>[[115, 157, 860, 268]]<|/det|>
+Video 1: Representative individual- based computational simulations of spatial patterns formed by co- cultures of strains AS and AR for different fitness costs of antibiotic resistance and probabilities of phage lysis. We initiated the simulations with 1,000 AS cells (cyan) and 1,000 AR cells (magenta) and performed simulations until the total number of cells reached 40,000. Simulations are for three different fitness costs of antibiotic resistance and four different probabilities of phage lysis.  
+
+<|ref|>text<|/ref|><|det|>[[115, 306, 881, 473]]<|/det|>
+Video 2: Representative individual- based computational simulations of spatial patterns formed by the AR strain for different fitness costs of antibiotic resistance, probabilities of losing antibiotic resistance, and probabilities of phage lysis. The AR cells (magenta) have a fitness cost for antibiotic resistance. Each cell can stochastically transition into an AS cell (grey) with a certain probability, in which case it will be relieved of its fitness cost. We initiated the simulations with 2,000 AR cells and performed simulations until the total number of cells reached 40,000. Simulations are for three different fitness costs of antibiotic resistance and three different probabilities of losing antibiotic resistance. The upper simulations are with phage lysis and the lower simulations are without phage lysis.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[42, 42, 312, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[42, 92, 768, 112]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 323, 176]]<|/det|>
+20251119RuanVideo1. mp4 20251119RuanVideo2. mp4
+
+<--- Page Split --->

preprint/preprint__488dd5e7f45f267645d4ecafeb0063d839323925821422c8cf4a42428a5d1df4/images_list.json

@@ -0,0 +1,70 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig. 1: Cell morphology of Eleftheros. A–j, Living cells of E. xomoi (strain Cur-11),",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        115,
+        593,
+        714
+      ]
+    ],
+    "page_idx": 19
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig 2. Phylogenomic reconstruction of the Marine Alveolates. a, Maximum likelihood",
+    "footnote": [],
+    "bbox": [
+      [
+        130,
+        90,
+        860,
+        545
+      ]
+    ],
+    "page_idx": 21
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig 3. Contextualising new MALVs using SSU data. Maximum likelihood analysis (GTR + F+ R10) showing MALV I lineages cluster by host, supporting the need for a distinct epithet for dinoflagellate-infecting MALV I species and the creation of Deorella g. n. as a separate genus. The eleftherids again form a sister clade to the remaining Syndinales but here cluster with uncultured environmental sequenced previously classified within MALV III. Black dots at nodes represent full statistical support (UltraFast bootstrap \\(= 100\\%\\) ), values are shown for support below \\(100\\%\\) . Coloured circles accompanying MALV I lineages reflects host identity.",
+    "footnote": [],
+    "bbox": [
+      [
+        170,
+        92,
+        858,
+        610
+      ]
+    ],
+    "page_idx": 23
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig 4. Character and plastid metabolic pathway evolution in dinoflagellates, apicomplexans and related lineages. Schematic tree indicating abundance of and character evolution in dinoflagellates, apicomplexans and related lineages. Black triangles/lines denote an estimation of the relative diversity of each lineage, while crossed-out circles indicate plastid loss (i.e. in the ancestor of the MALVs and within the apicomplexans). Abbreviations on branches indicate the inferred origin of: L,Ldiaminopimelate aminotransferase, DapL; SL, spliced leader; DVNPs, Dinoflagellate Viral NucleoProteins; HLPs, histone-like proteins. Next to each branch, Coulson plots depict selected plastidial metabolic pathways in the corresponding lineage. Grey segments indicate a cytosolic form of a given protein, while an offset, yellow segment denotes the mitochondrial enzyme 5-aminolevulinate synthase (ALAS) that synthesizes 5-aminolevulinate in all lineages but the core dinoflagellates, where the same intermediate is synthesized by the plastidial enzymes glutamyl-tRNA reductase (HemA) and glutamate-1-semialdehyde 2,1-aminomutase (HemL). Red nucleotides in SL sequence indicate deviation from canonical SL. Blue ovals surrounding cell depictions depict host cell, indicating parasitism. Coulson plot headers refer to metabolic pathways: isoprenoid,",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        98,
+        856,
+        400
+      ]
+    ],
+    "page_idx": 24
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_6.jpg",
+    "caption": "Extended Data Fig. 6: Plastid metabolic pathways in MALV-related lineages. Related to Fig. 3.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 42
+  }
+]

preprint/preprint__488dd5e7f45f267645d4ecafeb0063d839323925821422c8cf4a42428a5d1df4/preprint__488dd5e7f45f267645d4ecafeb0063d839323925821422c8cf4a42428a5d1df4.mmd

@@ -0,0 +1,566 @@
+
+# Multiple parallel origins of parasitic Marine Alveolates  
+
+Corey C. Holt ( corey.holt@ubc.ca )  University of British Columbia https://orcid.org/0000- 0003- 4222- 6086  Elisabeth Hehenberger ( elisabeth.hehenberger@paru.cas.cz )  Institute of Parasitology, Biology Centre Czech Academy of Sciences https://orcid.org/0000- 0001- 7810- 1336  
+
+Denis V. Tikhonenkov  Papanin Institute for Biology of Inland Waters, Russian Academy of Sciences https://orcid.org/0000- 0002- 4882- 2148  
+
+Victoria K. L. Jacko- Reynolds  University of British Columbia  
+
+Noriko Okamoto  Hakai Institute  
+
+Elizabeth C. Cooney  University of British Columbia  
+
+Nicholas A. T. Irwin  Merton College, University of Oxford https://orcid.org/0000- 0002- 2904- 8214  Patrick J. Keeling ( pkeeling@mail.ubc.ca )  University of British Columbia https://orcid.org/0000- 0002- 7644- 0745  
+
+Research Article  
+
+Keywords:  
+
+Posted Date: May 1st, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 1472581/v2  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on November 3rd, 2023. See the published version at https://doi.org/10.1038/s41467- 023- 42807- 0.
+
+<--- Page Split --->
+
+# 1. 1. 1. 1. 1. 1. 1. 2. 2. 2. 2. 2. 2. 2.
+
+<--- Page Split --->
+
+# Multiple parallel origins of parasitic Marine Alveolates  
+
+2 3 Corey C. Holt<sup>1,2</sup>†\*, Elisabeth Hehenberger<sup>1,3</sup>†\*, Denis V. Tikhonenkov<sup>1,4,5</sup>, Victoria K. L. 4 Jacko- Reynolds<sup>1</sup>, Noriko Okamoto<sup>1,2</sup>, Elizabeth C. Cooney<sup>1,2</sup>, Nicholas A. T. Irwin<sup>1,6</sup>, Patrick 5 J. Keeling<sup>1</sup>\* 6 7 <sup>1</sup>Department of Botany, University of British Columbia, Vancouver, British Columbia, 8 Canada 9 <sup>2</sup>Hakai Institute, Heriot Bay, British Columbia, Canada 10 <sup>3</sup>Institute of Parasitology, Biology Centre Czech Academy of Sciences, České Budějovice, 11 Czech Republic 12 <sup>4</sup>Papanin Institute for Biology of Inland Waters, Russian Academy of Sciences, Borok, 13 Russia 14 <sup>5</sup>AquaBioSafe Laboratory, University of Tyumen, Tyumen, Russia 15 <sup>6</sup>Present address: Merton College, University of Oxford, Oxford, UK 16 †Equal contribution 17 18 Summary 19 Microbial eukaryotes are important components of marine ecosystems, and the Marine 20 Alveolates (MALVs) are consistently one of the most abundant and diverse groups of 21 eukaryotes in global environmental sequencing surveys. Relatives of the dinoflagellates, 22 MALVs are thought to be parasites of animals and other protists but the absence of data 23 beyond ribosomal RNA gene sequences from all but two species means much of their 24 biology and evolution remain unknown. Here we show that MALVs evolved independently 25 from two distinct, free- living ancestors and that parasitism evolved twice, in parallel, prior to
+
+<--- Page Split --->
+
+the divergence of the core dinoflagellates. Phylogenomics shows one subgroup (MALV II and IV, or Syndinales) to be related to a novel lineage of free-living, eukaryovorous predators, the eleftherids, while the other (MALV I, now the Ichthyodinales) is related to Oxyrrhis marina. Reconstructing the evolution of photosynthesis, plastids, and parasitism in early- diverging dinoflagellates show a number of parallels with the evolution of their sister group, the parasitic apicomplexans. In both groups, similar forms of parasitism evolved multiple times, photosynthesis was lost many times, and the plastid organelle was lost infrequently, leaving no trace in the genome that they ever existed.  
+
+## Main  
+
+The marine alveolates (MALVs) are an elusive group of microbial eukaryotes first discovered through amplicon surveys of environmental rRNA \(^{1,2}\) , and since found to consistently dominate eukaryotic metabarcoding surveys in global oceans \(^{3,4,5,6,7}\) . Virtually all MALVs are known only from these rRNA SSU gene fragments, which have been phylogenetically linked to a handful of described parasites related to dinoflagellates \(^{1,9,10,11}\) . MALV sequences in marine environments are accordingly interpreted to represent a large and diverse population of infectious dinospores that reproduce inside an animal or protist host. Their abundance, diversity, and phylogenetic position all emphasize the evolutionary and ecological significance of MALVs but, since their discovery, new insights beyond their SSU rRNA gene diversity have been limited by a lack of genomic data and the inability to culture all but a few species.  
+
+MALVs are subdivided into five groups based on rRNA (MALV I- V), and are generally considered to be a monophyletic clade called Syndinales \(^{7}\) . All phylogenetic analyses support the branching of the Syndinales at the base of the dinoflagellates, but concatenating SSU and LSU rRNA genes suggested the Syndinales may be paraphyletic,
+
+<--- Page Split --->
+
+with MALV II and IV forming one group and MALV I diverging earlier<sup>12</sup>. Genomic data are only available for two MALVs, Amoebophyra and Hematodinium, which are in the MALV II and IV group, respectively<sup>13,14,15</sup>. Both of which have no evidence of a plastid, despite the fact that plastids are found in all other closely-related dinoflagellate lineages<sup>14,15</sup>. Losing photosynthesis is relatively common, but losing the plastid entirely is not<sup>14,16</sup>, and this absence of a plastid has created some controversy about the evolution of the organelle in close relatives<sup>17</sup>. However, the lack of data beyond rRNA gene sequencing for most MALVs, and in particular the major MALV I group, together with the lack of data from free-living relatives of these apparently obligate parasites, makes it difficult to untangle the evolution of parasitism and plastids.  
+
+## MALV II and IV are closely related to the eleftherids, a new group of free-living heterotrophs.  
+
+We isolated and cultured several strains collectively representing two new species of colourless, eukaryovorous flagellates, which we formally name Eleftheros xomoi (strains Colp37 and Cur- 11, isolated from the surface of a coral in the Caribbean Sea, Curaçao) and Eleftheros karadeniz (strain Colp- 25, isolated from marine near-shore sediments in the Black Sea, Crimea) (see Supplementary Information for taxonomic diagnosis). Sequences sufficiently similar to eleftherid SSU rRNA genes were not found in any of the largest global planktonic environmental surveys<sup>18,19</sup>, but we did recover small numbers of amplicons from marine sediment data, with a higher prevalence (albeit with often low read counts) found in deep sea samples<sup>20</sup> (Extended Data Fig. 1; Supplementary Table 1), suggesting that eleftherids are rare and benthic. Eleftheros are very small ( \(\sim 4 \mu \mathrm{m}\) ), bean-shaped or roundish, fast-swimming cells with two heterodynamic flagella, externally resembling the biflagellate dinospores of MALVs (Fig. 1, Extended Data Fig. 2 and Supplementary Video 1). Many of
+
+<--- Page Split --->
+
+their morphological features are reminiscent of dinoflagellates, including longitudinal and transverse grooves similar to cingulum and sulcus (Fig. 1a, c, f, j-m, o, q). Eleftherids possess large alveoli beneath the plasma membrane (Fig. 1n-r), flagellar basal bodies orientated predominantly at an acute angle to each other, a flagellar transition zone including an axosome with extended central microtubules and a transverse plate just below the cell surface (Fig. 1p and Extended Data Fig. 2a), enlarged perinuclear space containing tubular mastigomenes (Fig. 1n, q and Extended Data Fig. 2b), bowling pin-shaped trichocysts with square cross-sections scattered throughout the cytoplasm that are cross-striated after discharging (Fig. 1n, o, s and Extended Data Fig. 2c-e), large convoluted mitochondrion with tubular cristae (Fig. 1n, o, q and Extended Data Fig. 2a, f), and storage compounds in the form of roundish granules (Fig. 1n and Extended Data Fig. 2a, g). Interestingly we also observed large and distinctive vesicular compartments resembling the rhoptries of apicomplexan and perkinsid parasites, which are also present in the parasitic MALVs (Fig. 1o, r).  
+
+To determine their phylogenetic position, we generated transcriptomic data from both whole cultures and manually isolated eleftherid cells and built a concatenated 77- taxon/239 protein dataset (64,690 sites). Maximum- likelihood and Bayesian analyses consistently recovered eleftherids as sister to Amoebophyra and Hematodinium (MALV II and IV) with maximum statistical support (Fig. 2a and Extended Data Fig. 3). To examine the possibility that long- branch attraction affected the position of eleftherids, we performed fast- evolving site removal analysis until 50% of sites were removed, which showed that the support for the eleftherids as sisters to MALV II and IV remained 100% (Fig. 2a).  
+
+MALV I is closely related to Oxyrrhis, another free- living heterotroph.
+
+<--- Page Split --->
+
+Currently only rRNA gene fragments are available from the other major MALV group, MALV I, and none of this group are in culture. Accordingly, we sought infected hosts from marine samples and obtained one infected cell from a deep- water plankton tow near Quadra Island, British Columbia. This cell was identified from its morphology as the dinoflagellate Polykrikos but was also observed to contain a large and distinctive "fried egg- like" inclusion with centripetal grooves on a cuticular episcopal disc (Fig. 2c); a common morphological characteristic of a mature trophonts in MALV I parasites, including others infecting dinoflagellates<sup>21</sup>.  
+
+The Polykrikos cell was isolated, and a single- cell transcriptome sequenced, which revealed two phylogenetically- distinct signals: one being Polykrikos and the other an uncharacterized MALV I (Fig 2a). Phylogenetic analysis of full- length SSU rRNA gene sequences extracted from the transcriptomic data showed the MALV I parasite clustering with full- support with a clade of unidentified environmental sequences in the MALV I clade (Fig 3). This fell within a larger group of parasites isolated from dinoflagellates, which were ascribed to Eudubosquella based on conserved morphological traits<sup>21</sup>. Eudubosquella has been isolated from both ciliates and dinoflagellates<sup>21,22,23,24</sup>, however the SSU rRNA gene tree (Fig. 3; Extended Data Fig. 4) shows MALV sequences isolated from these two host types form two phylogenetically discrete groups. Indeed, MALV- I diversity seems to fall into a small number of highly host- specific clades, infecting ciliates, fish eggs, dinoflagellates, and radiolarians. The type host of Eudubosquella is a ciliate (Favella panamensis), so we propose Deorella gen. n. for the dinoflagellate- infecting lineage and the Polykrikos- infecting specimen as Deorella krika (see Supplementary Information for taxonomic diagnosis).  
+
+Phylogenomic analyses including D. krika rejects the monophyly of MALV- I and MALV- II/IV, instead showing complete support for MALV- I branching specifically with another lineage of free- living heterotrophs, the common marine flagellate Oxyrrhis (Fig 2a).
+
+<--- Page Split --->
+
+All topologies in which MALVs formed a monophyletic group were rejected by approximately unbiased (AU) tests (Extended Data Fig. 5). Indeed, Syndinales, as currently defined, are neither monophyletic nor paraphyletic, since we can demonstrate complete support for separate lineages being sister groups to different free- living heterotrophs. Parasitic transitions in MALV- I and MALV- II therefore trace back to different free- living ancestors and must have evolved their obligate parasitic life independently. We propose to retain the name Syndinales for MALV- II/IV and to create a new order for MALV- I, Ichthyodinales, after the earliest defined taxon within the group. Whether MALV III (and the little- studied MALV V) can all be included in Ichthyodinales seems likely based on SSU rRNA gene phylogenies (Ref 6 and Fig 3, Extended Data Fig. 4), but confirmation will await genomic data.  
+
+## Plastids and parasitism.  
+
+The complete absence of evidence for a plastid in Amoebophyra and Hematodinium<sup>14,15</sup> led to new hypotheses to account for the apparently complex distribution of plastids in early- diverging dinoflagellates and their sisters, the apicomplexan parasites<sup>17,25,26</sup>. But it is difficult to evaluate any hypothesis about the origin of plastids in parasites without data from close free- living relatives. Now that such relatives are identified, we examined plastid function across the deeply- branching dinoflagellate groups, including another relevant lineage, Psammosa. Psammosa is a free- living heterotroph related to dinoflagellates, but its exact position and whether it has a plastid are unclear in the absence of genomic data<sup>27</sup>. We generated transcriptomes from two strains of Psammosa, and phylogenomic analyses strongly support them forming an independent lineage at an important juncture in the tree, basal to Oxyrrhis, Ichthyodinales, Syndinales, and core dinoflagellates (Fig. 2a), or perhaps within the Oxyrrhis/Ichthyodinales clade itself (Fig. 2d, Extended Data Fig. 5).
+
+<--- Page Split --->
+
+The Polykrikos/Deorella data had too few parasite genes to determine whether Deorella has a plastid or to assess its function. In contrast, however, eleftherids and (to a lesser extent) psammosids encoded substantial numbers of nucleus- encoded genes for plastid- targeted proteins (as does Oxyrrhis), and these relate to metabolic pathways typical for non- photosynthetic plastids<sup>28</sup>. Specifically, we identified genes for the methylerythritol phosphate (MEP) pathway for isoprenoid biosynthesis and a plastidial SUF- type FeS cluster biosynthesis pathway including a ferredoxin redox system (Fig. 4 and Extended Data Fig. 6; phylogenetic trees available at https://figshare.com/s/fd3170a4b4a05027d9d2). These are both widespread in non- photosynthetic dinoflagellates and apicomplexans but absent in MALVs<sup>14</sup>. The complete transcripts encode N- terminal extensions with one or more predicted transmembrane domains that are consistent with plastid- targeting (IspF, SufC, SufD, SufE, PetF, PetH – in eleftherids), but do differ from canonical plastid- targeting sequences in core dinoflagellates<sup>29</sup> (alignments available at https://figshare.com/s/fd3170a4b4a05027d9d2). No evidence for the plastidial type II fatty acid biosynthesis pathway was found, which is also absent in MALVs and all other dinoflagellates with non- photosynthetic plastids<sup>27,30</sup> (Fig. 4 and Extended Data Fig. 6). Fatty acid biosynthesis likely occurs via a cytosolic type I fatty acid synthase (Supplementary Table 2).  
+
+acid biosynthesis pathway was found, which is also absent in MALVs and all other dinoflagellates with non- photosynthetic plastids<sup>27,30</sup> (Fig. 4 and Extended Data Fig. 6). Fatty acid biosynthesis likely occurs via a cytosolic type I fatty acid synthase (Supplementary Table 2). Heme is also synthesized in the plastids of some apicomplexans and dinoflagellates, but in eleftherids all these enzymes were mitochondrial (5- aminolevulinate synthase ALAS), or cytosolic (porphobilinogen synthase HemB, hydroxymethylbilane synthase HemC, uroporphyrinogen decarboxylase HemE), which is also the case in MALVs and Perkinsus. Oxyrrhis, by contrast, retains a mixed pathway with some enzymes in the plastid, like dinoflagellates and apicomplexans (Extended Data Fig. 6; phylogenetic trees available at https://figshare.com/s/fd3170a4b4a05027d9d2). The only enzyme proposed to have originated in the plastid of MALV II genomes is uroporphyrinogen-III synthase, or HemD<sup>14</sup>.
+
+<--- Page Split --->
+
+We did not find a HemD orthologue in our transcriptome data (Fig. 4 and Extended Data Fig. 6), and it is also absent from many dinoflagellates, suggesting a low level of expression.  
+
+However, in our phylogenetic reconstruction of HemD (Extended Data Fig. 7), photosynthetic eukaryotes and MALV II do not group with cyanobacteria, suggesting a bacterial origin for HemD independent from the plastid. Unlike MALVs and the eleftherids, Oxyrrhis has retained plastidial versions of enzymes involved in heme biosynthesis (Extended Data Fig. 6). This indicates a functional redundancy in heme biosynthesis persisted long into the evolution of dinoflagellates, which may have facilitated the repeated loss of plastid heme biosynthesis, and perhaps even the loss of the plastid itself in MALV II.  
+
+The genome of Hematodinium (MALV IV) contains another potential plastid pathway, the lysine biosynthesis enzyme L,Ldiaminopimelate aminotransferase (DapL), which is only present in bacteria and a subset of plastids. Its phylogenetic origin in Hematodinium could not be resolved and support for its plastid origin was concluded to be ambiguous<sup>14</sup>. We identified DapL in eleftherids and psammosids (Fig. 4; protein absent in apicomplexans and most dinoflagellates) and phylogenetic reconstruction shows it to group with plastid homologues in cryptophytes, whereas the Hematodinium homologue appears to have originated independently from bacteria (Extended Data Fig. 8). While most eukaryotic DapL genes encode plastid-targeting sequences, eleftherid and cryptophyte genes do not, indicating a likely cytosolic function (alignment available at https://figshare.com/s/fd3170a4b4a05027d9d2).  
+
+We found no proteins related to photosynthesis or chlorophyll biosynthesis in eleftherids, psammosids, or Oxyrrhis (as expected). We also identified no plastid import proteins; however, the import machinery is understudied in dinoflagellates and only a single component has been reported<sup>31</sup>. There is no evidence for a plastid genome, which is also expected since the plastid genome of photosynthetic dinoflagellates encodes only 12 proteins
+
+<--- Page Split --->
+
+that are exclusively involved in photosynthesis<sup>32</sup>. We also investigated the 129 proteins predicted to localize to the apicoplast in the apicomplexan Toxoplasma by spatial proteomics<sup>33</sup> and found no clear candidates for plastid- targeted homologs, beyond the plastid pathways noted above. Taken together, the plastid in these non- photosynthetic lineages has been profoundly reduced.  
+
+## Early character evolution in dinoflagellate-related lineages.  
+
+Dinoflagellates evolved in many strange ways, beyond parasitism and plastids<sup>28,34,35</sup>, and we now have a more accurate glimpse into these events as well. The nucleus, or "dinokaryon" is a good example: unlike other eukaryotes, dinoflagellates lack bulk nucleosomal chromatin and have huge genomes with permanently- condensed chromosomes structured by some combination of genes derived from horizontal gene transfer from viruses (Dinoflagellate Viral NucleoProtein, or DVNP)<sup>36,37</sup> and bacteria (two Histone- Like Proteins, or HLPs). DVNP appears in the tree after the divergence of the psammosids (Fig. 4 and Supplementary Table 2), whereas HLPs are present only in core dinoflagellates<sup>28</sup>. The absence of DVNP (which is highly expressed and easily detected in dinoflagellate transcriptomes) in psammosids strongly supports the phylogenomic topology in Fig. 2a, since the dinokaryon/DVNP is arguably the clearest synapomorphy of dinoflagellates and integral to dinoflagellate chromatin. If the psammosids were sister to Oxyrrhis and the Ichthyodiniales, it would require the loss of DVNP, which has not been described in any dinoflagellate lineage thus far.  
+
+Dinoflagellate nuclei also perform trans- splicing of a highly- conserved spliced leader (SL) to the 5' end of many or most mRNAs<sup>38</sup>. The same is also true of the perkinsids however the SL sequence is more variable<sup>39</sup>. We identified complete SL sequences in a subset of transcripts from eleftherids, MALV I, and psammosids, and found variations in
+
+<--- Page Split --->
+
+sequence from the canonical SL sequence identified in core dinoflagellates and MALVs \(^{15,38}\) (Fig. 4).  
+
+The ultrastructure of eleftherids also contains links to MALVs, which shed light on their diversification. Electron microscopy revealed the presence of rhoptry- like structures (Fig. 1o, r), similar to the electron- dense bodies found in the MALV II, Amoebophyra \(^{40}\) . Rhoptries are components of the infection apparatus of apicomplexan parasites, their free- living relatives, as well as Perkinsus and Psammosa \(^{41}\) . In eleftherids the presence of rhoptry- like structures may indicate components of an apical complex were also retained with a role in feeding, as is the case in Psammosa and a handful of free- living lineages related to apicomplexans \(^{42}\) . Altogether the distribution of rhoptry- like structures and plastids support a mixotrophic ancestor of dinoflagellates and apicomplexans (Fig. 4).  
+
+## Parallelism in the evolution of parasitism  
+
+While we have focused mostly on the parasitic dinoflagellates, of course the sisters to dinoflagellates are the much more famously parasitic apicomplexans (e.g., the malaria parasite Plasmodium). Due to their medical significance, apicomplexans are far better studied, and since it became clear they too evolved from algae and retained a cryptic plastid, their evolution has been thoroughly investigated \(^{43,44}\) . On the apicomplexan side of the tree, it is now clear that many of the major transitions seemingly associated with the origin of parasitism actually evolved multiple times in parallel. This includes the loss of photosynthesis, the loss of the plastid organelle, and even the origin of parasitism itself \(^{45}\) . Marine environmental surveys show MALVs dominate eukaryotic communities in terms of both diversity and abundance. Accordingly, they are likely to play an important role in marine ecology, yet we know shockingly little about their biology or evolution. It is now at least becoming clear that some of the same patterns emerging from apicomplexan research
+
+<--- Page Split --->
+
+are also true of parasitic dinoflagellates, in particular parallel evolution in seemingly similar parasites.  
+
+Syndinales (MALV II and IV) and Ichthyodinales (MALV I and potentially MALV III and V) were previously thought to have evolved form a single parasitic ancestor \(^{1,2,7}\) , but phylogenomics now shows they evolved independently from different free- living ancestors, or in other words that parasitism evolved twice in the Marine Alveolates. Similarly, given that plastids are present in nearly every apicomplexan and dinoflagellate subgroup, and are demonstrably related \(^{45}\) , the ancestor of all these groups can be inferred to have had a plastid and be capable of photosynthesis. But photosynthesis was lost several times independently in both the basal- branching dinoflagellates and apicomplexans. The prevalence of non- photosynthetic organisms branching at the base of the dinoflagellate tree has led to complex theories about the origin of their plastids \(^{17}\) . However, based on current data it now appears that a parallel loss of photosynthesis in early dinoflagellate evolution may not be that remarkable, and indeed would be very similar to what is observed in apicomplexans: roughly the same number of independent loses of photosynthesis (six or seven) are required to explain both. It has been suggested that the loss of photosynthesis may be linked to the origin of parasitism in apicomplexans through a “photoparasitic” intermediate \(^{46}\) . However, the demonstration that both syndineans and ichthyodinians are specifically related to free- living, non- photosynthetic lineages suggest that photosynthesis was likely lost prior to the origin of their parasitic lifestyles, which might also extend to the origin of apicomplexans.  
+
+The outright loss of the plastid organelle is very rare compared to the loss of photosynthesis, but this also happened in parallel in both dinoflagellate and apicomplexan lineages. Plastids were lost at least twice in basal- branching apicomplexans (cryptosporidia and one lineage of gregarines), and at least once in dinoflagellates (Syndinales): more data are required to clarify whether the Ichthyodinales retain any plastid- related genes or the
+
+<--- Page Split --->
+
+organelle. In all these cases, it is interesting to note how little evidence of the existence of a plastid is retained once the organelle has been lost. The idea that the endosymbiotic origin of plastids necessarily had an extensive and lasting genetic impact on their host genome has penetrated deeply into our thinking about organelle evolution, but the relatively recent presence of a plastid in the ancestor of eleftherids and Syndinales has left no impression on the genome of Syndinales \(^{15}\) , and indeed the only reason we know a plastid ever existed is due to their phylogenetic position, not because of any genetic "footprint" of the organelle. This is also true of apicomplexan lineages that have lost their plastid entirely \(^{44}\) and is further reinforced by the recent demonstration that the Picozoa are plastid- lacking members of the archaeoplastids \(^{47}\) . All intuition aside, this seems to be the rule, rather than the exception. We argue that our baseline expectation should be that when an organelle is lost, it leaves no genetic trace whatsoever, and it will therefore be very difficult, without clear phylogenetic evidence, to distinguish a cell where an organelle was lost from one that never had the organelle in the first place.
+
+<--- Page Split --->
+
+## References  
+
+1. Moon-van der Staay, S. Y., De Wachter, R. & Vaulot, D. Oceanic 18S rDNA sequences from picoplankton reveal unsuspected eukaryotic diversity. Nature 409, 607-610 (2001).  
+
+2. López-García, P., Rodríguez-Valera, F., Pedrós-Alió, C. & Moreira, D. Unexpected diversity of small eukaryotes in deep-sea Antarctic plankton. Nature 409, 603-607 (2001).  
+
+3. de Vargas, C. et al. Eukaryotic plankton diversity in the sunlit ocean. Science 348, 1261605 (2015).  
+
+4. Massana, R. et al. Marine protist diversity in European coastal waters and sediments as revealed by high-throughput sequencing: Protist diversity in European coastal areas. Environ Microbiol 17, 4035-4049 (2015).  
+
+5. Clarke, L. J., Bestley, S., Bissett, A. & Deagle, B. E. A globally distributed Syndiniales parasite dominates the Southern Ocean micro-eukaryote community near the sea-ice edge. ISME J 13, 734-737 (2019).  
+
+6. Holt, C. C. et al. Microscopic marine invertebrates are reservoirs for cryptic and diverse protists and fungi. Microbiome 10, 161 (2022).  
+
+7. Guillou, L. et al. Widespread occurrence and genetic diversity of marine parasitoids belonging to Syndiniales (Alveolata). Environmental Microbiology 10, 3349-3365 (2008).  
+
+8. Rizos, I. et al. Beyond the limits of the unassigned protist microbiome: inferring large-scale spatio-temporal patterns of Syndiniales marine parasites. ISME Commun. 3, 16 (2023).  
+
+9. Skovgaard, A., Massana, R., Balagué, V. & Saiz, E. Phylogenetic Position of the Copepod-Infesting Parasite Syndinium turbo (Dinoflagellata, Syndinea). Protist 156, 413-423 (2005).
+
+<--- Page Split --->
+
+10. Gestal, C., Novoa, B., Posada, D., Figueras, A. & Azevedo, C. Perkinsoide chabelardi n. gen., a protozoan parasite with an intermediate evolutionary position: possible cause of the decrease of sardine fisheries? Environ Microbiol 8, 1105-1114 (2006). 
+
+11. Harada, A., Ohtsuka, S. & Horiguchi, T. Species of the Parasitic Genus Duboscquella are Members of the Enigmatic Marine Alveolate Group I. Protist 158, 337-347 (2007). 
+
+12. Strassert, J. F. H. et al. Single cell genomics of uncultured marine alveolates shows paraphyly of basal dinoflagellates. ISME J 12, 304-308 (2018). 
+
+13. Jackson, C. J., Gornik, S. G. & Waller, R. F. The Mitochondrial genome and transcriptome of the basal dinoflagellate Hematodinium sp.: character evolution within the highly derived mitochondrial genomes of dinoflagellates. Genome Biology and Evolution 4, 59-72 (2012). 
+
+14. Gornik, S. G. et al. Endosymbiosis undone by stepwise elimination of the plastid in a parasitic dinoflagellate. Proc. Natl. Acad. Sci. U.S.A. 112, 5767-5772 (2015). 
+
+15. Farhat, S. et al. Rapid protein evolution, organellar reductions, and invasive intronic elements in the marine aerobic parasite dinoflagellate Amoebophyra spp. BMC Biol 19, 1 (2021). 
+
+16. Xu, P. et al. The genome of Cryptosporidium hominis. Nature 431, 1107-1112 (2004). 
+
+17. Waller, R. F. & Kořený, L. Plastid complexity in dinoflagellates: a picture of gains, losses, replacements and revisions. In Advances in Botanical Research vol. 84 105-143 (Elsevier, 2017). 
+
+18. Villar, E. et al. The Ocean Gene Atlas: exploring the biogeography of plankton genes online. Nucleic Acids Research 46, W289-W295 (2018). 
+
+19. Duarte, C. M. Seafaring in the 21st Century: The Malaspina 2010 Circumnavigation Expedition. Limnology and Oceanography Bulletin 24, 11-14 (2015).
+
+<--- Page Split --->
+
+20. Cordier, T. et al. Patterns of eukaryotic diversity from the surface to the deep-ocean sediment. Sci. Adv. 8, eabj9309 (2022).
+21. Yoo, J., Coats, D. W. & Kim, S. Syndinean dinoflagellates of the genus Euduboscquella are paraphyletic. J Eukaryotic Microbiology 70, (2023).
+22. Coats, D. W. & Bachvaroff, T. R. Parasites of Tintinnids. In The Biology and Ecology of Tintinnid Ciliates (eds. Dolan, J. R., Montagnes, D. J. S., Agatha, S., Coats, D. W. & Stoecker, D. K.) 145–170 (Wiley, 2012).
+23. Jung, J.-H., Choi, J. M., Coats, D. W. & Kim, Y.-O. Euduboscquella costata n. sp. (Dinoflagellata, Syndinea), an intracellular parasite of the ciliate Schmidingerella arcuata: morphology, molecular phylogeny, life cycle, prevalence, and infection intensity. J. Eukaryot. Microbiol. 63, 3–15 (2016).
+24. Choi, J. M., Jung, J. H., Kim, K. H., Coats, D. W. & Kim, Y. O. A novel parasitic, syndinean dinoflagellate Euduboscquella triangula infecting the tintinnid Helicostomella longa. Front. Mar. Sci. 8, 720424 (2021).
+25. Bodyl, A., Stiller, J. W. & Mackiewicz, P. Chromalveolate plastids: direct descent or multiple endosymbioses? Trends in Ecology & Evolution 24, 119–121 (2009).
+26. Ševčíkov��, T. et al. Updating algal evolutionary relationships through plastid genome sequencing: did alveolate plastids emerge through endosymbiosis of an ochrophyte? Sci Rep 5, 10134 (2015).
+27. Okamoto, N., Horák, A. & Keeling, P. J. Description of two species of early branching dinoflagellates, Psammosa pacifica n. g., n. sp. and P. atlantica n. sp. PloS ONE 7, e34900 (2012).
+28. Janouškovec, J. et al. Major transitions in dinoflagellate evolution unveiled by phylotranscriptomics. Proc. Natl. Acad. Sci. U.S.A. 114, (2017).
+
+<--- Page Split --->
+
+29. Patron, N. J., Waller, R. F., Archibald, J. M. & Keeling, P. J. Complex protein targeting to dinoflagellate plastids. Journal of Molecular Biology 348, 1015–1024 (2005).  
+
+30. Hehenberger, E., Gast, R. J. & Keeling, P. J. A kleptoplastic dinoflagellate and the tipping point between transient and fully integrated plastid endosymbiosis. Proc. Natl. Acad. Sci. U.S.A. 116, 17934–17942 (2019).  
+
+31. Hehenberger, E., Imanian, B., Burki, F. & Keeling, P. J. Evidence for the retention of two evolutionary distinct plastids in dinoflagellates with diatom endosymbionts. Genome Biology and Evolution 6, 2321–2334 (2014).  
+
+32. Dorrell, R. G. et al. Progressive and biased divergent evolution underpins the origin and diversification of peridinin dinoflagellate plastids. Mol Biol Evol msw235 (2016)  
+
+33. Barylyuk, K. et al. A comprehensive subcellular atlas of the Toxoplasma proteome via hyperLOPIT provides spatial context for protein functions. Cell Host & Microbe 28, 752-766.e9 (2020).  
+
+34. Gavelis, G. S. et al. Microbial arms race: Ballistic “nematocysts” in dinoflagellates represent a new extreme in organelle complexity. Sci. Adv. 3, e1602552 (2017).  
+
+35. Hackett, J. D., Anderson, D. M., Erdner, D. L. & Bhattacharya, D. Dinoflagellates: a remarkable evolutionary experiment. Am. J. Bot. 91, 1523–1534 (2004).  
+
+36. Gornik, S. G. et al. Loss of nucleosomal DNA condensation coincides with appearance of a novel nuclear protein in dinoflagellates. Current Biology 22, 2303–2312 (2012).  
+
+37. Irwin, N. A. T. et al. Viral proteins as a potential driver of histone depletion in dinoflagellates. Nat Commun 9, 1535 (2018).  
+
+38. Zhang, H. et al. Spliced leader RNA trans-splicing in dinoflagellates. Proc. Natl. Acad. Sci. U.S.A. 104, 4618–4623 (2007).  
+
+39. Alacid, E. et al. A diversified and segregated mRNA spliced-leader system in the parasitic Perkinsozoa. Open Biol. 12, 220126 (2022).
+
+<--- Page Split --->
+
+40. Miller, J. J., Delwiche, C. F. & Coats, D. W. Ultrastructure of Amoebophrya sp. and its changes during the course of infection. Protist 163, 720–745 (2012). 
+
+41. Okamoto, N. & Keeling, P. J. The 3D structure of the apical complex and association with the flagellar apparatus revealed by serial TEM tomography in *Psammosa pacifica*, a distant relative of the Apicomplexa. *PloS ONE* 9, e84653 (2014). 
+
+42. Leander, B. S. & Keeling, P. J. Morphostasis in alveolate evolution. *Trends in Ecology & Evolution* 18, 395–402 (2003). 
+
+43. Janouškovec, J. et al. Apicomplexan-like parasites are polyphyletic and widely but selectively dependent on cryptic plastid organelles. *eLife* 8, e49662 (2019). 
+
+44. Mathur, V. et al. Multiple independent origins of apicomplexan-like parasites. *Current Biology* 29, 2936-2941.e5 (2019). 
+
+45. Janouškovec, J., Horák, A., Oborník, M., Lukeš, J. & Keeling, P. J. A common red algal origin of the apicomplexan, dinoflagellate, and heterokont plastids. *Proc. Natl. Acad. Sci. U.S.A.* 107, 10949–10954 (2010). 
+
+46. Oborník, M. Photoparasitism as an intermediate state in the evolution of apicomplexan parasites. *Trends in Parasitology* 36, 727–734 (2020). 
+
+47. Schön, M. E. et al. Single cell genomics reveals plastid-lacking Picozoa are close relatives of red algae. *Nat Commun* 12, 6651 (2021).
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig. 1: Cell morphology of Eleftheros. A–j, Living cells of E. xomoi (strain Cur-11), </center>  
+
+visualized by light microscopy. K- m, Living cells of E. karadeniz, visualized by light microscopy. N- s, Cells of E. karadeniz, visualized by transmission electron microscopy. N, Longitudinal section. O, Oblique section showing the transverse groove. P, Arrangement of basal bodies. Q, transverse section showing the longitudinal groove. R, Structure of the distal part with trophy- like structures and alveoli. S, Longitudinal and cross section (inset) of
+
+<--- Page Split --->
+
+419 trichocysts. A, alveoli; ax, axosome; bb, basal bodies; cm, central microtubules; f, flagella; 420 fv, food vacuole; lf, longitudinal flagellum; lg, longitudinal groove; m, mitochondrion; n, 421 nucleus; ps, perinuclear space; r, trophy-like structures; sc, storage compounds; t, trichocyst; 422 tf, transverse flagellum; tg, transverse groove; tp, transverse plate. Scale bars, \(4\mu \mathrm{m}\) (a-m), 1 423 \(\mu \mathrm{m}\) (n, o, q) and \(0,5\mu \mathrm{m}\) (p, r, s).
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig 2. Phylogenomic reconstruction of the Marine Alveolates. a, Maximum likelihood </center>  
+
+Fig 2. Phylogenomic reconstruction of the Marine Alveolates. a, Maximum likelihood analysis of a 239 multi-protein alignment (LG+C60+F+G4) separate MALV I from the remaining Syndinales (MALV II and IV) to which the eleftherids are the free-living sister group. Black dots represent full statistical support (non- parametric bootstrap = 100%), values are shown for support below 100%. The percentage of sites for each taxon are shown as bars to the right. b, UltraFast bootstrap support for given topologies, indicated by highlighted node labels in a, following removal of the fastest-evolving sites as inferred using the LG + C 40 + F + G4 model. * indicates nodes that remained at 100% c, Living cell of Polykrikos sp. infected with a MALV 1- like parasite showing characteristic centripetal grooves. Scale bar, 12 μm. d, All supported topologies from approximately unbiased (AU) tests, showing
+
+<--- Page Split --->
+
+alternative position of psammosids. Rejected topologies (including those showing the Syndinales as a monophyletic clade) shown in Extended Data Figure. 5. Support for core dinoflagellate monophyly serves as a control for the presence of sufficient information for phylogenomic inference.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig 3. Contextualising new MALVs using SSU data. Maximum likelihood analysis (GTR + F+ R10) showing MALV I lineages cluster by host, supporting the need for a distinct epithet for dinoflagellate-infecting MALV I species and the creation of Deorella g. n. as a separate genus. The eleftherids again form a sister clade to the remaining Syndinales but here cluster with uncultured environmental sequenced previously classified within MALV III. Black dots at nodes represent full statistical support (UltraFast bootstrap \(= 100\%\) ), values are shown for support below \(100\%\) . Coloured circles accompanying MALV I lineages reflects host identity. </center>
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig 4. Character and plastid metabolic pathway evolution in dinoflagellates, apicomplexans and related lineages. Schematic tree indicating abundance of and character evolution in dinoflagellates, apicomplexans and related lineages. Black triangles/lines denote an estimation of the relative diversity of each lineage, while crossed-out circles indicate plastid loss (i.e. in the ancestor of the MALVs and within the apicomplexans). Abbreviations on branches indicate the inferred origin of: L,Ldiaminopimelate aminotransferase, DapL; SL, spliced leader; DVNPs, Dinoflagellate Viral NucleoProteins; HLPs, histone-like proteins. Next to each branch, Coulson plots depict selected plastidial metabolic pathways in the corresponding lineage. Grey segments indicate a cytosolic form of a given protein, while an offset, yellow segment denotes the mitochondrial enzyme 5-aminolevulinate synthase (ALAS) that synthesizes 5-aminolevulinate in all lineages but the core dinoflagellates, where the same intermediate is synthesized by the plastidial enzymes glutamyl-tRNA reductase (HemA) and glutamate-1-semialdehyde 2,1-aminomutase (HemL). Red nucleotides in SL sequence indicate deviation from canonical SL. Blue ovals surrounding cell depictions depict host cell, indicating parasitism. Coulson plot headers refer to metabolic pathways: isoprenoid, </center>
+
+<--- Page Split --->
+
+466 methylerythritol phosphate (MEP) pathway for isoprenoid biosynthesis; FASII, type 2 fatty 467 acid biosynthesis pathway; FeS cluster, plastidial Fe-S cluster biosynthesis pathway including 468 a ferrodoxin system; heme, heme biosynthesis. Complete protein names are listed in 469 Supplementary Table 2.
+
+<--- Page Split --->
+
+## Methods  
+
+Data reporting. No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.  
+
+Cell isolation and culture establishment. Strains Colp- 37 and Cur- 11 (Eleftheros xomoi gen. et sp. nov.) were obtained from the surface of brain coral Colpophylia natans Houttuyn, 1772 in coastal waters of Curacao (Caribbean Sea) in April 2016 and 2018, respectively. Strain Colp- 25 (Eleftheros karadeniz gen. et sp. nov.) was isolated from the near shore bottom sediments in the Black Sea near T.I. Vyazemsky Karadag Scientific Station, Crimea, May 2015. The samples were enriched with a suspension of Aeromonas sobria bacteria and examined on the third, sixth and ninth day of incubation ( \(25^{\circ}\mathrm{C}\) , darkness) in accordance with methods described previously \(^{48}\) . To obtain clonal cultures, the individual cells were transferred using a drawn- out glass micropipette into Petri dishes containing a clonal culture of eukaryotic prey Procryptobia sorokini (strain B- 69), which were grown in marine Schmalz- Pratt's medium at a final salinity of \(20\%\) using the bacterium A. sobria as food \(^{49}\) . Strains perished after several months of cultivation.  
+
+The Deorella- infected Polykrikos cell was collected with a \(20\mu \mathrm{m}\) mesh net towed at Heriot Bay near the Hakai Quadra Island Ecological Observatory in British Columbia, October 2022. The infected cell was isolated with a drawn- out glass micropipette and washed with \((0.2\mu \mathrm{m})\) filtered water before imaging and lysis.  
+
+Psammosa pacifica Psp strain was isolated from Boundary Bay, British Columbia Canada (45.25580°N; - 64.34907°W) and maintained according to Okamoto et al. 2012 \(^{27}\) .  
+
+Psammosa sp. (strain Colp- 34) was isolated from the near shore bottom sediments in the Kapsel Bay, Black Sea, Crimea, May 2016 and maintained following the abovementioned protocol.
+
+<--- Page Split --->
+
+Light and electron microscopy. To observe living cells, an AxioScope A1 light microscope (Carl Zeiss, Jena, Germany) with DIC water immersion objective \(63 \times\) and an inverted microscope Leica DM IL LED with DIC objectives \(40 \times\) and \(63 \times\) were used. Images were captured with a Sony \(\alpha 7R\) camera.  
+
+For transmission electron microscopy (TEM), cells were centrifuged, fixed at \(1^{\circ}\mathrm{C}\) for 60 min in a cocktail of \(0.6\%\) glutaraldehyde and \(2\% \mathrm{OsO_4}\) (final concentration) prepared using a \(0.1\mathrm{M}\) cacodylate buffer (pH 7.2). Fixed cells were dehydrated in alcohol and acetone series (30, 50, 70, 96, and \(100\%\) , 20 min in each step). Afterward, the cells were embedded in a mixture of Araldite and Epon (Fluka, 45345). Ultrathin sections (60 nm) were prepared with a Leica EM UC6 ultramicrotome (Leica Microsystems, Germany) and observed by using a JEM 1011 transmission electron microscope (JEOL, Japan).  
+
+## Preparation of libraries and sequencing.  
+
+RNA isolation and cDNA preparation. Cells from clonal culture were collected by centrifugation (1000 x g, room temperature) onto the \(0.8 \mu \mathrm{m}\) membrane of a Vivaclear mini column (Sartorium Stedim Biotech Gmmg, VK01P042). Total RNA was then extracted using a RNAqueous- Micro Kit (Invitrogen, AM1931) and reverse transcribed into cDNA using the Smart- Seq2 protocol<sup>50</sup>, which uses poly- A selection to enrich mRNA. Additionally, cDNA of Cur11 and Colp- 37 were obtained from 20 single cells using the Smart- Seq2 protocol (cells were manually picked from culture using a drawn- out glass micropipette and transferred to a \(0.2 \mathrm{mL}\) thin- walled PCR tube containing \(2 \mu \mathrm{L}\) of cell lysis buffer - \(0.2\%\) Triton X- 100 and RNase inhibitor (Invitrogen)). Likewise, cDNA from the infected Polykrikos cell and Psammosa spp. was prepared following the same protocol.  
+
+Sequencing dataset assembly and decontamination. Deorella libraries were sequenced with an Illumina NextSeq 500 system using a Mid Output Flow Cell (2 x 150bp reads). Eleftheros and Psammosa libraries were sequenced on an Illumina MiSeq platform with read lengths of
+
+<--- Page Split --->
+
+\(2 \times 300 \mathrm{bp}\) (strains Colp- 25, Colp- 37 and Colp- 34) and on an Illumina HiSeq 2500 machine, \(2 \times 125 \mathrm{bp}\) reads (strain Cur- 11). Sequence quality and adapter contamination of reads from transcriptomic datasets were assessed with FastQC v.0.10.15.  
+
+Reads of clone Colp- 37 (culture and 20- cell preparations) and clone Colp- 25 (culture) were merged with PEAR v.0.9.6 \(^{52}\) and resulting assembled as well as unassembled reads were separately trimmed with Trimmomatic \(^{53}\) as implemented in Trinity v.2.0.6 \(^{54}\) , removing Illumina adapters with ILLUMINACLIP, with a maximum of two mismatches, a palindrome clip threshold of 30 and a simple clip threshold of 10. Low- quality sequences were discarded, using a sliding window of 4 bp and a minimum trimmed length of 25 bp. Trimmed assembled and unassembled reads were combined into a single file and transcriptomes were assembled with Trinity, using the --single flag. Contaminating non- eukaryotic (bacterial, archaeal, and viral) and prey contigs were identified using BLASTn \(^{55}\) queries of the NCBI nt database. Sequences that aligned with \(\geq 100\) nucleotides, had a query coverage of \(\geq 80\%\) and were \(\geq 90\%\) identical to noneukaryotic entries were removed, while all sequences resulting in a kinetoplastid best hit were removed.  
+
+Reads of the later sampled clone Cur- 11 (culture and 20- cell preparations) were trimmed and assembled without prior read merging in a single step, using the Trimmomatic plugin embedded in Trinity. Trimming parameters remained unchanged except for a simple clip threshold of 9. Due to results from initial analyses of clone Colp- 37, the transcriptome was not subjected to any cleaning steps to retain putative bacterial horizontal gene transfers as well as sequences with bacterial best hits in nt that were identified as likely being eukaryotic in phylogenetic reconstructions in clone Colp- 37. Instead, all genes of interest were subjected to phylogenetic analysis, as described below, to clarify the taxonomic identity of the gene.
+
+<--- Page Split --->
+
+Reads from the Deorella/Polykrikos and psammosid libraries were trimmed using Cutadapt v3.2<sup>56</sup> before assembly with rnaSPAdes v3.15.1<sup>57</sup>. Contaminating sequences were removed using BLASTx<sup>58</sup> and BLASTn<sup>55</sup> searches against the NCBI nt and UniProt databases (E- value cut- off = 1x10<sup>-25</sup>). Removal of prey contigs, Spumella elongata for Psammosa pacifica and Procryptobia sorokini for Psammosa sp. were characterized using BLASTn<sup>55</sup> against prey transcriptome data and removed.  
+
+Prediction of protein coding regions in all assemblies was performed by TransDecoder v.5.0.2, including BLASTp<sup>58</sup> queries of the Swiss- Prot database (E- value cut- off = 1x10<sup>-5</sup>).  
+
+Phylogenomic dataset preparation and analysis. In addition to the taxa described above we added to or updated the following available transcriptome or genome data in an existing phylogenomic framework<sup>59</sup>. We added the re- assembled transcriptomes of several core dinoflagellates generated by the MMETSP project<sup>60</sup>, several transcriptomes from the EukProt database<sup>61</sup>, the transcriptomes of the dinoflagellates Lepidodinium chlorophorum (https://www.ncbi.nlm.nih.gov/bioproject/481676), TGD and  
+
+MGD<sup>62</sup>, Abedinium<sup>63</sup>, Amyloodinium ocellatum<sup>64</sup> and additional taxa from Cooney et al. 2021 including Spatulodinium, Kofoidinium and Fabadinium<sup>65</sup>. We updated the data for the MALVs Amoebophyra ceratii<sup>66</sup> and  
+
+Hematodinium sp.<sup>14</sup>, for Oxyrrhis marina<sup>60</sup>, for the perkinsids Perkinsus marinus (https://protists.ensembl.org/Perkinsus_marinus_atcc_50983_gca_000006405/Info/Index), Maranthos nigrum and Parvilucifera sinerae<sup>39</sup>, the apicomplexa Ancora sagittata<sup>43</sup>, and for the chromerids Vitrella brassicaformis (https://cryptodb.org/cryptodb/app/record/dataset/NCBITAXXON_1169540) and Symbiont X<sup>43</sup>. Transcriptomic datasets were subjected to TransDecoder coding region prediction to extract peptides for downstream analyses.
+
+<--- Page Split --->
+
+All listed datasets were searched for 263 proteins to generate single- protein trees as described in Burki et al. 2016 \(^{59}\) . In brief: BLASTp was used to identify homologues to the 263 genes in the new datasets. After a parsing step (E- value \(\leq 1 \times 10^{- 20}\) ), a maximum of four non- redundant hits was added to the initial 263- protein set. The expanded gene sets were aligned with MAFFT L- ins- i v.7.222 \(^{67}\) and trimmed automatically with trimAl v1. \(^{26}\) , with a gap threshold of 80%. Single- protein maximum- likelihood phylogenies were reconstructed under the LG + G4 model using RAxML v.8.1. \(^{69}\) in combination with 100 rapid bootstraps and resulting trees were manually screened to flag paralogues and sequences derived from prey or other contamination. Cleaned protein sets were aligned and trimmed as above and taxa were selected upon concatenation with SCaFOs v.1.2. \(^{50}\) to select proteins sequences present in \(\geq 60\%\) of all taxa. To improve data presence, the concatenated sequences of clone Colp- 37 derived from culture and 20- cell preparations were merged, as were the sequences for the two clone Cur- 11 datasets. The final concatenated alignment included 77- taxa, 239 proteins and 64,690 amino acid sites. Clone Cur- 11 was represented with 91% of proteins and 85% of sites, clone Colp- 37 with 81% of proteins and 71% of sites and clone Colp- 25 with 49% of proteins and 37% of sites. The host Polykrikos sp. was represented with 75% of proteins and 66% of sites and Deorella with 26% and 15% respectively. Psammosa pacifica Psp was represented with 75% and 62% of sites, Psammosa sp. was represented with 74% of proteins and 64% of sites.  
+
+Maximum- likelihood phylogenomic tree reconstruction was performed using IQ- TREE v. 1.6. \(^{57}\) using the C60 empirical mixture model in combination with the LG matrix, amino acid frequencies computed from the data and four gamma categories for handling the rate heterogeneity across sites (LG + C60 + F + G4 model with 1000 UFBoot replicates \(^{72}\) ). Bayesian analyses were carried out by running four independent Markov chain Monte Carlo chains with PhyloBayes MPI v.1.8c \(^{73}\) , using the GTR matrix in combination with an infinite
+
+<--- Page Split --->
+
+mixture model and four discrete gamma categories \((\mathrm{CAT} + \mathrm{GTR} + \mathrm{G4})\) . Chains were run for more than 10,000 generations saving every second tree, and the first 200 generations (or \(20\%\) ) were discarded as burn- in. As is frequently seen in large- scale phylogenomic analyses, the chains failed to converge (maxdiff \(= 1\) and meandiff \(= 0.0171456\) ), with the topologies differing within the core dinoflagellates and the apicomplexan plus chromerid clade.  
+
+Bayesian posterior probabilities are reported as a measure of statistical support for bipartitions. Phylogenetic trees were visualized in \(\mathrm{R}^{74}\) with the ggtree \(^{75}\) and treelio \(^{76}\) packages.  
+
+Fast- evolving sites were estimated with IQ- TREE using the - wsr option, using the LG \(+ \mathrm{C20} + \mathrm{F} + \mathrm{G}\) model. Sites were removed in increments of \(5\%\) of the original alignment length, up to \(50\%\) , and for each subsample with a reduced number of sites trees were reconstructed with the \(\mathrm{LG} + \mathrm{C20} + \mathrm{F} + \mathrm{G}\) model. The support for the sister relationship of eleftherids and MALVs was assessed at each removal increment. The monophyly of core dinoflagellates was tested as a control.  
+
+Small subunit ribosomal RNA (SSU rRNA) phylogeny. The most complete sequences for SSU rRNA genes were identified from each eleftherid transcriptome with BLASTn, using a known dinoflagellate SSU query, whereas barnap \(^{77}\) was used to extract Deorella SSU rRNA sequences prior to identification with BLAST. An alignment of all near full- length MALV sequences deposited in GenBank was curated to include all five previously described MALV groups in Guillou et al. \(^{7}\) . The resulting SSU collection was aligned using MAFFT with the E- INS- I algorithm and inspected for misaligned and chimeric sequences. Shorter sequences obtained from surveys of environmental data (see below) were aligned using the addfragments flag. The final alignment was trimmed with trimAl (- gt 0.1, - st 0.001) before generating a maximum likelihood tree in IQTREE with 1000 ultrafast bootstrap replicates \(^{71}\) , using the \(\mathrm{GTR} + \mathrm{F} + \mathrm{R10}\) model (and \(\mathrm{GTR} + \mathrm{R} + \mathrm{R60}\) for Extended Data Figure 4), selected
+
+<--- Page Split --->
+
+with ModelFinder<sup>78</sup>. Oxyrrhis marina and Ellobiopsis chattonii were omitted to minimize long branch attraction.  
+
+Identification of putative plastid- targeted proteins. Putative plastid- targeted proteins were identified combining a BLASTp- based similarity search and a hidden Markov models (HMMs) based screen. Known dinoflagellate<sup>30</sup> or MALV<sup>14</sup> proteins involved in plastid metabolic pathways were used as queries in a BLASTp search against a comprehensive custom database containing representatives from most major eukaryotic groups (excluding the long- branching excavates and the data- poor group of Rhizaria) and RefSeq data from all bacterial phyla at NCBI (last accessed December 2017). The database was subjected to CD- HIT<sup>79</sup> clustering with a similarity threshold of 85% to reduce redundant sequences and paralogues, except for the data sets created in this study (clustered at 98%). The search results of the BLASTp step were parsed for hits with an E- value threshold \(\leq 1 \times 10^{- 25}\) and a query coverage of \(\geq 30\%\) to reduce the possibility of paralogs and extremely short sequences and at the same time recover possibly fragmented eleftherid homologues. The number of bacterial hits was restrained to 20 hits per phylum (for FCB group, most classes of Proteobacteria, PVC group, Spirochaetes, Actinobacteria, Cyanobacteria (unranked) and Firmicutes) or 10 per phylum (remaining bacterial phyla) as defined by NCBI taxonomy. In some cases (HemD and DapL) these numbers were expanded to 20 and 40, respectively, for a more representative bacterial sampling. In addition, the Colp- 37 and Cur- 11 protein data were used to search the Pfam- A database release 33.0 with hmmscan (HMMER3.1; hmmer.org), employing the manually curated Pfam gathering threshold. The results were queried, using a keyword search, for Pfam domains present in plastid- associated proteins, including proteins involved in metabolic pathways in Hehenberger et al. 2019<sup>30</sup>, photosynthesis, plastid import (TIC/TOC components) and Calvin cycle (RuBisCO). Candidates with domains of interest were used as BLASTp queries as described above and parsed hits (query coverage of \(\geq 50\%\) ).
+
+<--- Page Split --->
+
+were combined with the recovered hits from the known proteins. After a deduplication step, sequences were aligned with MAFFT using the --auto option, trimmed using trimAl (-gt 0.8) and Maximum-likelihood tree reconstructions were performed with FastTree v.2.1.7<sup>80</sup> using the default options in a preliminary analysis. The resulting phylogenies and underlying alignments were inspected manually to remove contaminant, divergent, and/or low-quality sequences. The cleaned, unaligned sequences were then subjected to filtering with PREQUAL<sup>81</sup> using the default options, followed by alignment with MAFFT G- INS-i using the VSM option (--unalignlevel 0.6). The alignments were subjected to Divvier<sup>82</sup> using the - mincol 4 and the -divvygap option before trimming with trimAl (-gt 0.01). Final trees were calculated with IQ- TREE, using the -mset option to restrict model selection to LG for ModelFinder<sup>78</sup>, while branch support was assessed with 1000 ultrafast bootstrap replicates<sup>72</sup>. We also searched for homologues to the 129 proteins predicted to localize to the apicoplast in Toxoplasma<sup>33</sup>. The predicted Toxoplasma proteins were used as queries in a BLASTp search against our custom database, using an initial E-value threshold of \(\leq 1 \times 10^{- 25}\) and a query coverage of \(\geq 30\%\) for parsing. All proteins recovering putative eleftherid homologues were used in a preliminary tree reconstruction analysis as described above. After manual inspection of the phylogenies, potential plastid-targeted candidates were further investigated by using the recovered eleftherid homologues as additional BLASTp queries and combining the resulting hits (query coverage of \(\geq 50\%\) ) with the initial BLASTp output as described above. The BLASTp search using the Toxoplasma proteins was repeated with relaxed parameters (E-value threshold \(\leq 1 \times 10^{- 5}\) ) to recover additional eleftherid candidates. All candidates investigated were also submitted to a web BLASTp search against nr to exclude the possibility of a contamination not present in our database. N-terminal extensions of putative plastid-targeted sequences were investigated in the respective protein alignments, visualized with AliView<sup>83</sup>. For easier recognition of such
+
+<--- Page Split --->
+
+extensions, only the dinoflagellate sequences of the protein of interest plus the prokaryotic sequence with the highest sequence similarity were viewed. Eleftherid sequences with N- terminal extensions relative to prokaryotic sequences were submitted to SignalP- 3.0<sup>84</sup> and TMHMM 2.0<sup>85</sup> to predict putative signal peptides and transmembrane domains, respectively.  
+
+For proteins with known bacterial orthologs, bacterial naming conventions were applied, with the exception of the multifunctional enzyme Acetyl- CoA carboxylase (ACC) not present in bacteria and the enzyme 5- aminolevulinate synthase (ALAS), where eukaryotic conventions were applied.  
+
+Identification of molecular characteristics of dinoflagellates. Dinoflagellate Viral NucleoProteins (DVNPs) in eleftherids were identified using a BLASTp search against our custom database using all described Hematodinium DVNP proteins<sup>36</sup> as well as by performing an hmmsearch (HMMER3.1; hmmer.org) against our database using the DVNP profile HMM downloaded from pfam.xfam.org, employing default thresholds. The two approaches recovered the same set of eleftherid DVNP candidates.  
+
+We also searched for histone- like proteins (HLPs), using known representatives of the two known types of HLPs in dinoflagellates<sup>28</sup> as query, Crypthecodinium cohnii AAM97522.1 (HLPI) and Noctiluca scintillans ABV22345.1 (HLPII) in a BLASTp search against the predicted eleftherid peptides (E- value threshold \(\leq 1 \times 10^{- 25}\) ). Additionally, we performed an hmmsearch using a profile HMM constructed from a curated alignment of dinoflagellate and prokaryotic HLPs.  
+
+Dinoflagellate spliced leader (SL) sequences were identified using BLASTn searches (using the option - task blastn to allow for short input queries and to identify short matches) against the all transcriptomes, using the canonical 21- nucleotide dinoflagellate SL sequence<sup>38</sup> as query. The recovered 21- nucleotide sequence in eleftherids, differing in 2 nucleotides (5A>T and 17T>A) from the canonical sequence, was used as a search string to identify all
+
+<--- Page Split --->
+
+eleftherids transcripts containing a full- length SL. To avoid SL sequences in host Polykrikos sequences, Deorella peptides identified in single- gene trees were aligned against Polykrikos/Deorella transcripts with tBLASTn to obtain Deorella only transcripts.  
+
+Environmental distribution and abundance. To uncover the global distribution and abundance of eleftherids, we searched for sequences similar to the eleftherid SSU rRNA gene among published environmental SSU rRNA gene amplicon studies.  
+
+Full- length eleftherid SSU rRNA genes and their V4 and V9 hypervariable regions were used as queries in BLASTn searches (E- value threshold \(\leq 1 \times 10^{- 10}\) ) against the complete Tara Ocean database available on Ocean Gene Atlas<sup>18</sup>. We recovered no sequences with \(\geq 95\%\) identity and a query cover of \(\geq 90\%\) , neither when using full- length nor hypervariable regions of eleftherid SSU rRNA gene queries.  
+
+Additionally, we searched a custom environmental sequence database containing data from 26 environmental amplicon studies with a focus on marine sediment, also including the large sequencing projects BioMarKs<sup>4</sup> and Malaspina<sup>19</sup>. NCBI SRA files of amplicon data were downloaded using fastq- dump from stratolkit 2.10.8. Raw sequence data were processed with MICrobial Community Analysis, using micca merge or mergepairs (- l 100 - d 30) for single- end or paired- end data respectively<sup>86</sup>, and then concatenated and converted them into a BLAST database composed of 113,203,549 sequences with a total of 25,261,490,053 letters. Alternatively, paired- end data were processed in Qiime<sup>28</sup> with the DADA2<sup>88</sup> denoising algorithm before creation of the BLAST database. This database was searched using full- length eleftherid SSU rRNA gene sequences as well as their V4 and V9 hypervariable regions in BLASTn searches (E- value threshold \(\leq 1 \times 10^{- 25}\) ). In addition to the environmental database, sequences similar to the eleftherid 18S SSU gene were searched using BLASTn against a V9 amplicon dataset collected from sandy beaches off the central coast of British Columbia (the European Nucleotide Archive project PRJEB14727). A total
+
+<--- Page Split --->
+
+of 73 unique sequences with \(\geq 97\%\) identity to and \(\geq 90\%\) coverage of the eleftherid V9 and V4 hypervariable regions resulted from both search approaches. Full- length and hypervariable regions V9 and V4 of \(P\) pacifica were used as queries against the environmental database which resulted in a total of 4 unique sequences with with \(\geq 97\%\) identity to and \(\geq 90\%\) coverage to psammosids. These sequences were then clustered at \(99\%\) identity using CD- HIT prior to being added to the SSU rRNA gene phylogeny to confirm their phylogenetic position as described above. Sequences less than 150bp were removed from the final phylogeny. Sample coordinates corresponding to BLAST hits in the custom database were extracted and plotted in \(\mathrm{R}^{74}\) with the nraturearth \(^{89}\) package.  
+
+## Data availability  
+
+Raw transcriptome reads will be deposited in the GenBank Sequence Read Archive (SRA). SSU rRNA gene sequences retrieved from the transcriptomes will also be deposited in GenBank. Assembled transcriptomes, along with individual gene alignments, concatenated and trimmed alignments, and maximum- likelihood and Bayesian tree files for the phylogenomic dataset will be available at Figshare. The untrimmed and trimmed alignments, alignments depicting N- terminal extensions and tree files in nexus and pdf format for plastid- associated and other proteins of interest will be available at Figshare. The genus Eleftheros and species Eleftheros xomoi and Eleftheros karadeniz will be registered with the Zoobank database (http://zoobank.org/). As will the family Deorellidae, genus Deorella, and species Deorella krika.  
+
+## Code availability  
+
+All unpublished code is available upon reasonable request from the corresponding authors.
+
+<--- Page Split --->
+
+## Methods References  
+
+48. Tikhonenkov, D. V., Mazei, Y. A. & Embulaeva, E. A. Degradation succession of heterotrophic flagellate communities in microcosms. Zh. Obs. Biol. 69, 57-64 (2008).  
+
+49. Tikhonenkov, D. V. et al. Description of Colponema vietnamica sp.n. and Acavomonas peruviana n. gen. n. sp., two new alveolate phyla (Colponemidia nom. nov. and Acavomonidia nom.nov.) and their contributions to reconstructing the ancestral state of alveolates and eukaryotes. PLoS One 9, (2014).  
+
+50. Picelli, S. et al. Full-length RNA-seq from single cells using Smart-seq2. Nat Protoc 9, 171-181 (2014).  
+
+51. Andrews, S. FastQC: A quality control tool for high throughput sequence data. version 0.10.1. https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ (2010).  
+
+52. Zhang, J., Kobert, K., Flouri, T. & Stamatakis, A. PEAR: A fast and accurate Illumina Paired-End reAd mergeR. Bioinformatics 30, 614-620 (2014).  
+
+53. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114-2120 (2014).  
+
+54. Grabherr, M. G. et al. Trinity: reconstructing a full-length transcriptome without a genome from RNA-Seq data. Nat. Biotechnol. 29, 644-652 (2013).  
+
+55. Zhang, Z., Schwartz, S., Wagner, L. & Miller, W. A greedy algorithm for aligning DNA sequences. Journal of Computational Biology vol. 7 203-214 (2000).  
+
+56. Martin., M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal. 17, 10-12 (2011).  
+
+57. Bankevich, A. et al. SPAdes: A new genome assembly algorithm and its applications to single-cell sequencing. J. Comput. Biol. 19, 455-477 (2012).  
+
+58. Camacho, C. et al. BLAST+: Architecture and applications. BMC Bioinformatics 10, 1-9 (2009).
+
+<--- Page Split --->
+
+59. Burki, F. et al. Untangling the early diversification of eukaryotes: A phylogenomic study of the evolutionary origins of centrohelida, haptophyta and cryptista. Proc. R. Soc. B Biol. Sci. 283, (2016).  
+
+60. Johnson, L. K., Alexander, H. & Brown, C. T. Re-assembly, quality evaluation, and annotation of 678 microbial eukaryotic reference transcriptomes. Gigascience 8, 1–12 (2019).  
+
+61. Richter, D. J. et al. EukProt: A database of genome-scale predicted proteins across the diversity of eukaryotes. Peer Community Journal 2, e56 (2022).  
+
+62. Sarai, C. et al. Dinoflagellates with relic endosymbiont nuclei as models for elucidating organellogenesis. Proc. Natl. Acad. Sci. U. S. A. 117, 5364–5375 (2020).  
+
+63. Cooney, E. C. et al. Single-cell transcriptomics of Abedinium reveals a new earlybranching dinoflagellate lineage. Genome Biol. Evol. 12, 2417–2428 (2020).  
+
+64. Byadgi, O. et al. Transcriptome Analysis of Amyloodinium ocellatum Tomonts Revealed Basic Information on the Major Potential Virulence Factors. Genes 11, 1252 (2020).  
+
+65. Cooney, E. C., Leander, B. S. & Keeling, P. J. Phylogenomics shows unique traits in Noctilucales are derived rather than ancestral. PNAS Nexus 1, pgac202 (2022).  
+
+66. John, U. et al. An aerobic eukaryotic parasite with functional mitochondria that likely lacks a mitochondrial genome. Sci. Adv. 5, 1–12 (2019).  
+
+67. Katoh, K. & Standley, D. M. MAFFT multiple sequence alignment software version 7: Improvements in performance and usability. Mol. Biol. Evol. 30, 772–780 (2013).  
+
+68. Capella-Gutierrez, S., Silla-Martinez, J. M. & Gabaldón, T. trimAl: A tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25, 1972–1973 (2009).  
+
+69. Stamatakis, A. RAxML version 8: A tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 30, 1312–1313 (2014).
+
+<--- Page Split --->
+
+70. Roure, B., Rodriguez-Ezpeleta, N. & Philippe, H. SCaFoS: A tool for selection, concatenation and fusion of sequences for phylogenomics. BMC Evol. Biol. 7, 1–12 (2007).  
+
+71. Nguyen, L. T., Schmidt, H. A., Von Haeseler, A. & Minh, B. Q. IQ-TREE: A fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol. Biol. Evol. 32, 268–274 (2015).  
+
+72. Hoang, D. T., Chernomor, O., Von Haeseler, A., Minh, B. Q. & Vinh, L. S. UFBoot2: Improving the ultrafast bootstrap approximation. Mol. Biol. Evol. 35, 518–522 (2018).  
+
+73. Lartillot, N., Lepage, T. & Blanquart, S. PhyloBayes 3: A Bayesian software package for phylogenetic reconstruction and molecular dating. Bioinformatics 25, 2286–2288 (2009).  
+
+74. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (2021).  
+
+75. Yu, G., Smith, D. K., Zhu, H., Guan, Y. & Lam, T. T. GGtree: an R package for visualization and annotation of phylogenetic trees with their covariates and other associated data. Methods Ecol Evol 8, 28–36 (2017).  
+
+76. Wang, L.-G. et al. Trecio: An R Package for Phylogenetic Tree Input and Output with Richly Annotated and Associated Data. Molecular Biology and Evolution 37, 599–603 (2020).  
+
+77. Seemann T. barrnap 0.9: rapid ribosomal RNA prediction. https://github.com/tseemann/barrnap  
+
+78. Kalyaanamoothry, S., Minh, B. Q., Wong, T. K. F., Von Haeseler, A. & Jermiin, L. S. ModelFinder: fast model selection for accurate phylogenetic estimates. Nat. Methods 14, 587–589 (2017).  
+
+79. Fu, L., Niu, B., Zhu, Z., Wu, S. & Li, W. CD-HIT: Accelerated for clustering the nextgeneration sequencing data. Bioinformatics 28, 3150–3152 (2012).
+
+<--- Page Split --->
+
+80. Price, M. N., Dehal, P. S. & Arkin, A. P. FastTree 2 - Approximately maximum-likelihood trees for large alignments. PLoS One 5, (2010). 
+
+81. Whelan, S., Irisarri, I. & Burki, F. PREQUAL: Detecting non-homologous characters in sets of unaligned homologous sequences. Bioinformatics 34, 3929–3930 (2018). 
+
+82. Ali, R. H., Bogusz, M., Whelan, S. & Tamura, K. Identifying Clusters of High Confidence Homologies in Multiple Sequence Alignments. Mol. Biol. Evol. 36, 2340–2351 (2019). 
+
+83. Larsson, A. AliView: a fast and lightweight alignment viewer and editor for large datasets. Bioinformatics 30, 3276–3278 (2014). 
+
+84. Dyrløv Bendtsen, J., Nielsen, H., Von Heijne, G. & Brunak, S. Improved prediction of signal peptides: SignalP 3.0. Journal of Molecular Biology 340, 783–795 (2004). 
+
+85. Krogh, A., Larsson, B., Von Heijne, G. & Sonnhammer, E. L. L. Predicting transmembrane protein topology with a hidden markov model: application to complete genomes11Edited by F. Cohen. Journal of Molecular Biology 305, 567–580 (2001). 
+
+86. Albanese, D., Fontana, P., De Filippo, C., Cavalieri, D. & Donati, C. MICCA: A complete and accurate software for taxonomic profiling of metagenomic data. Sci. Rep. 5, 1–7 (2015). 
+
+87. Bolyen, E. et al. Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nat Biotechnol 37, 852–857 (2019). 
+
+88. Callahan, B. J. et al. DADA2: High-resolution sample inference from Illumina amplicon data. Nat Methods 13, 581–583 (2016). 
+
+89. Massicotte P, South A. rnaturalearth: World Map Data from Natural Earth. R package version0.3.2.9000, https://docs.ropensci.org/rnaturalearth/https://github.com/ropensci/rnaturalearth (2023).
+
+<--- Page Split --->
+
+Acknowledgements We thank A.P. Mylnikov and T.G. Simdyanov for help with sample collection, fixation, and interpretation of transmission electron microscopy images. This research was supported by grants from the Gordon and Betty Moore Foundation (to P.J.K., https://doi.org/10.37807/GBMF9201), the Natural Sciences and Engineering Research Council of Canada (to P.J.K., Grant Number 2019- 03994), the Czech Academy of Sciences (to E.H., Grant Number LQ200962204) the Russian Foundation for Basic Research (to D.V.T., Grant Number 20- 04- 00583), Tyumen Oblast Government, as part of the West- Siberian Interregional Science and Education Center's project No. 89- DON (2) and carried out within the framework of State Assignment no. 121051100102- 2.  
+
+Author contributions C.C.H., E.H., D.V.T. and P.J.K. designed the study. D.V.T. isolated and cultured eleftherid cells. C.C.H. isolated Polykrikos/Deorella. V.K.L.J.- R and N.O. isolated and cultured psammosid cells. C.C.H., D.V.T. and E.H. generated material for sequencing. D.V.T. performed microscopy experiments. C.C.H., E.H., E.C.C. and N.A.T.I. performed phylogenomic analyses. C.C.H. and E.C.C. performed phylogenetic analysis of the SSU rRNA genes. V.K.L.J.- R. performed the environmental distribution and abundance analysis. E.H. and V.K.L.J.- R. performed transcriptomic analyses and phylogenetic analysis of plastid and other proteins. C.C.H., E.H. and P.J.K. wrote the manuscript with input from all authors.  
+
+Competing interests The authors declare no competing interests.  
+
+Additional information  
+
+Supplementary information is available for this paper.  
+
+Correspondence and requests for materials should be addressed to C.C.H., E.H. or P.J.K.
+
+<--- Page Split --->
+![](images/Extended_Data_Figure_6.jpg)
+  
+
+870  
+
+Extended Data Fig. 1: Global distribution and abundance of eleftherids. The global distribution of eleftherids within sequenced transcriptomes (squares) and environmental SSU rRNA amplicon studies (circles). All eleftherid reads were obtained from marine sediment with the exception of two SSU rRNA reads from the Coral Sea study PRJNA369575 which were discovered in the water column at 798m depth.
+
+<--- Page Split --->
+![PLACEHOLDER_43_0]
+  
+
+877  
+
+Extended Data Fig. 2 Cell structure of Eleftheros karadeniz, visualized by transmission electron microscopy. Related to Fig. 1. a, Longitudinal section showing arrangement of basal bodies and transitional zone of flagella with transverse plates. b, Perinuclear space with mastigonemes, tubular in cross-sections (arrows). c, Longitudinal sections and arrangement of trichocysts. d, square cross-sections of trichocysts. e, cross-striated filaments of discharged trichocysts. f, mitochondrion with tubular cristae (arrows). g, storage compounds. ax, axosome; bb, basal bodies; f, flagella; fv, food vacuole; m, mitochondrion; n, nucleus; ps, perinuclear space; sc, storage compounds; t, trichocyst; tp, transverse plate. Scale bars, 1 μm (a, c), 0,2 μm (b, d, e) and 0,5 μm (f, g).
+
+<--- Page Split --->
+![PLACEHOLDER_44_0]
+
+ 
+
+887 
+
+Extended Data Fig. 3 Bayesian phylogenomic analysis. Related to Fig. 2. Consensus 
+
+Bayesian tree based on four independent chains (CAT + GTR + G4), places the eleftherids as 
+
+sister to the Syndinales (now MALV II and IV). MALV I (now the Ichthyodinales) groups 
+
+sister to Oxyrrhis marina, while the psammosids occupy a basal position relative to MALVs 
+
+and core dinoflagellates. Black dots denote full statistical support (Bayesian posterior 
+
+probability = 1.0), values are shown for support below 1.0.
+
+<--- Page Split --->
+![PLACEHOLDER_45_0]
+  
+
+Extended Data Fig. 4 SSU rRNA gene phylogeny including short read sequences.  
+
+Maximum likelihood analysis (GTR+F+R6) of MALV related sequences. Black dots at nodes represent full statistical support (UltraFast bootstrap \(= 100\%\) ). Grey dots represent support above \(90\%\) . Support values are shown for support below \(90\%\) . Clades of interest coloured
+
+<--- Page Split --->
+
+899 with proposed taxonomic groups: purple, Ichthyodinales (MALV I); green, Syndinales; red, 900 eleftherids; and blue, psammosids. 901
+
+<--- Page Split --->
+![PLACEHOLDER_47_0]
+  
+
+Extended Data Fig. 5 All topologies tested with approximately unbiased (AU) test. Maximum likelihood analysis of nine alternative topologies. Associated \(p\) values show two supported topologies (1 and 3) reflecting Fig 2a and psammosids branching sister to Oxyrrhis and Deorella. All topologies showing monophyly of MALVs (7,8,9) were rejected ( \(p\) value \(< 0.05\) ). Red tip labels \(=\) eleftherids, green \(=\) Syndinales (MALV II and IV), purple \(=\) Ichthyodinales (MALV I) and blue \(=\) psammosids.
+
+<--- Page Split --->
+![PLACEHOLDER_48_0]
+
+<center>Extended Data Fig. 6: Plastid metabolic pathways in MALV-related lineages. Related to Fig. 3. </center>  
+
+Presence/absence table of plastid metabolic pathways in core dinoflagellates, eleftherids, MALVs, Oxyrrhis and Perkinsus including presence /absence of N- terminal extensions or incomplete N- termini. The pathways presented, from top to bottom, are: isoprenoid, methylerythritol phosphate (MEP) pathway for isoprenoid biosynthesis; FASII, type 2 fatty acid biosynthesis pathway; FeS cluster, plastidial Fe- S cluster biosynthesis pathway including a ferrodoxin system; heme, heme biosynthesis. \*, eleftherid sequences cluster with bacteria; m, mitochondrial; p/c, plastidial/cytosolic clade.
+
+<--- Page Split --->
+![PLACEHOLDER_49_0]
+  
+
+Extended Data Fig. 7: Maximum likelihood phylogeny of HemD. The scale bar and the number beneath it indicate the estimated number of substitutions per site, above the scale bar the model for tree reconstruction is indicated. Node numbers represent ultrafast bootstrap support values of \(>70\%\) , black dots indicate support values of \(> = 95\%\) . Eukaryotic groups are
+
+<--- Page Split --->
+
+925 indicated by colored taxon names: green, Viridiplantae; dark red, rhodophytes; grey, 926 glaucophytes; turquoise, cryptophytes; pink, stramenopiles; dark blue, haptophytes; yellow, 927 Dinozoa; light green, eleftherids. Black taxa/clades outlined in black are prokaryotic. 928 Annotated orthologs in model species are indicated by red taxon name and protein identifier. 929 For species represented by more than one strain or taxa identified on genus level only, the 930 strain information is provided where available.
+
+<--- Page Split --->
+![PLACEHOLDER_51_0]
+  
+
+Extended Data Fig. 8: Maximum likelihood phylogeny of DapL. The scale bar and the number beneath it indicate the estimated number of substitutions per site, above the scale bar the model for tree reconstruction is indicated. Node numbers represent ultrafast bootstrap support values of \(>70\%\) , black dots indicate support values of \(> = 95\%\) . Eukaryotic groups are
+
+<--- Page Split --->
+
+937 indicated by colored taxon names: green, Viridiplantae; dark red, rhodophytes; grey, glaucophytes; turquoise, cryptophytes; pink, stramenopiles; dark blue, haptophytes; yellow, Dinozoa; light green, eleftherids. Black taxa/clades outlined in black are prokaryotic. Annotated orthologs in model species are indicated by red taxon name and protein identifier. For species represented by more than one strain or taxa identified on genus level only, the strain information is provided where available.
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupplementaryInformation.pdf  SIGuide.docx  SupplementaryTable1.xlsx  SupplementaryTable2.xlsx  SupplementaryVideo1.mp4
+
+<--- Page Split --->

preprint/preprint__488dd5e7f45f267645d4ecafeb0063d839323925821422c8cf4a42428a5d1df4/preprint__488dd5e7f45f267645d4ecafeb0063d839323925821422c8cf4a42428a5d1df4_det.mmd

@@ -0,0 +1,727 @@
+<|ref|>title<|/ref|><|det|>[[44, 107, 790, 175]]<|/det|>
+# Multiple parallel origins of parasitic Marine Alveolates  
+
+<|ref|>text<|/ref|><|det|>[[44, 195, 916, 280]]<|/det|>
+Corey C. Holt ( corey.holt@ubc.ca )  University of British Columbia https://orcid.org/0000- 0003- 4222- 6086  Elisabeth Hehenberger ( elisabeth.hehenberger@paru.cas.cz )  Institute of Parasitology, Biology Centre Czech Academy of Sciences https://orcid.org/0000- 0001- 7810- 1336  
+
+<|ref|>text<|/ref|><|det|>[[44, 285, 940, 375]]<|/det|>
+Denis V. Tikhonenkov  Papanin Institute for Biology of Inland Waters, Russian Academy of Sciences https://orcid.org/0000- 0002- 4882- 2148  
+
+<|ref|>text<|/ref|><|det|>[[44, 380, 315, 420]]<|/det|>
+Victoria K. L. Jacko- Reynolds  University of British Columbia  
+
+<|ref|>text<|/ref|><|det|>[[44, 426, 220, 465]]<|/det|>
+Noriko Okamoto  Hakai Institute  
+
+<|ref|>text<|/ref|><|det|>[[44, 472, 315, 512]]<|/det|>
+Elizabeth C. Cooney  University of British Columbia  
+
+<|ref|>text<|/ref|><|det|>[[44, 518, 728, 607]]<|/det|>
+Nicholas A. T. Irwin  Merton College, University of Oxford https://orcid.org/0000- 0002- 2904- 8214  Patrick J. Keeling ( pkeeling@mail.ubc.ca )  University of British Columbia https://orcid.org/0000- 0002- 7644- 0745  
+
+<|ref|>text<|/ref|><|det|>[[44, 648, 186, 666]]<|/det|>
+Research Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 686, 136, 704]]<|/det|>
+Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 723, 285, 743]]<|/det|>
+Posted Date: May 1st, 2023  
+
+<|ref|>text<|/ref|><|det|>[[44, 761, 477, 781]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 1472581/v2  
+
+<|ref|>text<|/ref|><|det|>[[44, 799, 910, 842]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[44, 860, 531, 880]]<|/det|>
+Additional Declarations: There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[42, 916, 945, 959]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on November 3rd, 2023. See the published version at https://doi.org/10.1038/s41467- 023- 42807- 0.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 997]]<|/det|>
+# 1. 1. 1. 1. 1. 1. 1. 2. 2. 2. 2. 2. 2. 2.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[72, 84, 596, 103]]<|/det|>
+# Multiple parallel origins of parasitic Marine Alveolates  
+
+<|ref|>text<|/ref|><|det|>[[70, 117, 884, 920]]<|/det|>
+2 3 Corey C. Holt<sup>1,2</sup>†\*, Elisabeth Hehenberger<sup>1,3</sup>†\*, Denis V. Tikhonenkov<sup>1,4,5</sup>, Victoria K. L. 4 Jacko- Reynolds<sup>1</sup>, Noriko Okamoto<sup>1,2</sup>, Elizabeth C. Cooney<sup>1,2</sup>, Nicholas A. T. Irwin<sup>1,6</sup>, Patrick 5 J. Keeling<sup>1</sup>\* 6 7 <sup>1</sup>Department of Botany, University of British Columbia, Vancouver, British Columbia, 8 Canada 9 <sup>2</sup>Hakai Institute, Heriot Bay, British Columbia, Canada 10 <sup>3</sup>Institute of Parasitology, Biology Centre Czech Academy of Sciences, České Budějovice, 11 Czech Republic 12 <sup>4</sup>Papanin Institute for Biology of Inland Waters, Russian Academy of Sciences, Borok, 13 Russia 14 <sup>5</sup>AquaBioSafe Laboratory, University of Tyumen, Tyumen, Russia 15 <sup>6</sup>Present address: Merton College, University of Oxford, Oxford, UK 16 †Equal contribution 17 18 Summary 19 Microbial eukaryotes are important components of marine ecosystems, and the Marine 20 Alveolates (MALVs) are consistently one of the most abundant and diverse groups of 21 eukaryotes in global environmental sequencing surveys. Relatives of the dinoflagellates, 22 MALVs are thought to be parasites of animals and other protists but the absence of data 23 beyond ribosomal RNA gene sequences from all but two species means much of their 24 biology and evolution remain unknown. Here we show that MALVs evolved independently 25 from two distinct, free- living ancestors and that parasitism evolved twice, in parallel, prior to
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 868, 333]]<|/det|>
+the divergence of the core dinoflagellates. Phylogenomics shows one subgroup (MALV II and IV, or Syndinales) to be related to a novel lineage of free-living, eukaryovorous predators, the eleftherids, while the other (MALV I, now the Ichthyodinales) is related to Oxyrrhis marina. Reconstructing the evolution of photosynthesis, plastids, and parasitism in early- diverging dinoflagellates show a number of parallels with the evolution of their sister group, the parasitic apicomplexans. In both groups, similar forms of parasitism evolved multiple times, photosynthesis was lost many times, and the plastid organelle was lost infrequently, leaving no trace in the genome that they ever existed.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 381, 168, 397]]<|/det|>
+## Main  
+
+<|ref|>text<|/ref|><|det|>[[115, 411, 872, 759]]<|/det|>
+The marine alveolates (MALVs) are an elusive group of microbial eukaryotes first discovered through amplicon surveys of environmental rRNA \(^{1,2}\) , and since found to consistently dominate eukaryotic metabarcoding surveys in global oceans \(^{3,4,5,6,7}\) . Virtually all MALVs are known only from these rRNA SSU gene fragments, which have been phylogenetically linked to a handful of described parasites related to dinoflagellates \(^{1,9,10,11}\) . MALV sequences in marine environments are accordingly interpreted to represent a large and diverse population of infectious dinospores that reproduce inside an animal or protist host. Their abundance, diversity, and phylogenetic position all emphasize the evolutionary and ecological significance of MALVs but, since their discovery, new insights beyond their SSU rRNA gene diversity have been limited by a lack of genomic data and the inability to culture all but a few species.  
+
+<|ref|>text<|/ref|><|det|>[[117, 772, 845, 889]]<|/det|>
+MALVs are subdivided into five groups based on rRNA (MALV I- V), and are generally considered to be a monophyletic clade called Syndinales \(^{7}\) . All phylogenetic analyses support the branching of the Syndinales at the base of the dinoflagellates, but concatenating SSU and LSU rRNA genes suggested the Syndinales may be paraphyletic,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 880, 400]]<|/det|>
+with MALV II and IV forming one group and MALV I diverging earlier<sup>12</sup>. Genomic data are only available for two MALVs, Amoebophyra and Hematodinium, which are in the MALV II and IV group, respectively<sup>13,14,15</sup>. Both of which have no evidence of a plastid, despite the fact that plastids are found in all other closely-related dinoflagellate lineages<sup>14,15</sup>. Losing photosynthesis is relatively common, but losing the plastid entirely is not<sup>14,16</sup>, and this absence of a plastid has created some controversy about the evolution of the organelle in close relatives<sup>17</sup>. However, the lack of data beyond rRNA gene sequencing for most MALVs, and in particular the major MALV I group, together with the lack of data from free-living relatives of these apparently obligate parasites, makes it difficult to untangle the evolution of parasitism and plastids.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 445, 812, 494]]<|/det|>
+## MALV II and IV are closely related to the eleftherids, a new group of free-living heterotrophs.  
+
+<|ref|>text<|/ref|><|det|>[[115, 510, 875, 895]]<|/det|>
+We isolated and cultured several strains collectively representing two new species of colourless, eukaryovorous flagellates, which we formally name Eleftheros xomoi (strains Colp37 and Cur- 11, isolated from the surface of a coral in the Caribbean Sea, Curaçao) and Eleftheros karadeniz (strain Colp- 25, isolated from marine near-shore sediments in the Black Sea, Crimea) (see Supplementary Information for taxonomic diagnosis). Sequences sufficiently similar to eleftherid SSU rRNA genes were not found in any of the largest global planktonic environmental surveys<sup>18,19</sup>, but we did recover small numbers of amplicons from marine sediment data, with a higher prevalence (albeit with often low read counts) found in deep sea samples<sup>20</sup> (Extended Data Fig. 1; Supplementary Table 1), suggesting that eleftherids are rare and benthic. Eleftheros are very small ( \(\sim 4 \mu \mathrm{m}\) ), bean-shaped or roundish, fast-swimming cells with two heterodynamic flagella, externally resembling the biflagellate dinospores of MALVs (Fig. 1, Extended Data Fig. 2 and Supplementary Video 1). Many of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 82, 880, 536]]<|/det|>
+their morphological features are reminiscent of dinoflagellates, including longitudinal and transverse grooves similar to cingulum and sulcus (Fig. 1a, c, f, j-m, o, q). Eleftherids possess large alveoli beneath the plasma membrane (Fig. 1n-r), flagellar basal bodies orientated predominantly at an acute angle to each other, a flagellar transition zone including an axosome with extended central microtubules and a transverse plate just below the cell surface (Fig. 1p and Extended Data Fig. 2a), enlarged perinuclear space containing tubular mastigomenes (Fig. 1n, q and Extended Data Fig. 2b), bowling pin-shaped trichocysts with square cross-sections scattered throughout the cytoplasm that are cross-striated after discharging (Fig. 1n, o, s and Extended Data Fig. 2c-e), large convoluted mitochondrion with tubular cristae (Fig. 1n, o, q and Extended Data Fig. 2a, f), and storage compounds in the form of roundish granules (Fig. 1n and Extended Data Fig. 2a, g). Interestingly we also observed large and distinctive vesicular compartments resembling the rhoptries of apicomplexan and perkinsid parasites, which are also present in the parasitic MALVs (Fig. 1o, r).  
+
+<|ref|>text<|/ref|><|det|>[[115, 547, 875, 796]]<|/det|>
+To determine their phylogenetic position, we generated transcriptomic data from both whole cultures and manually isolated eleftherid cells and built a concatenated 77- taxon/239 protein dataset (64,690 sites). Maximum- likelihood and Bayesian analyses consistently recovered eleftherids as sister to Amoebophyra and Hematodinium (MALV II and IV) with maximum statistical support (Fig. 2a and Extended Data Fig. 3). To examine the possibility that long- branch attraction affected the position of eleftherids, we performed fast- evolving site removal analysis until 50% of sites were removed, which showed that the support for the eleftherids as sisters to MALV II and IV remained 100% (Fig. 2a).  
+
+<|ref|>text<|/ref|><|det|>[[115, 842, 725, 861]]<|/det|>
+MALV I is closely related to Oxyrrhis, another free- living heterotroph.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 875, 333]]<|/det|>
+Currently only rRNA gene fragments are available from the other major MALV group, MALV I, and none of this group are in culture. Accordingly, we sought infected hosts from marine samples and obtained one infected cell from a deep- water plankton tow near Quadra Island, British Columbia. This cell was identified from its morphology as the dinoflagellate Polykrikos but was also observed to contain a large and distinctive "fried egg- like" inclusion with centripetal grooves on a cuticular episcopal disc (Fig. 2c); a common morphological characteristic of a mature trophonts in MALV I parasites, including others infecting dinoflagellates<sup>21</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[115, 346, 875, 796]]<|/det|>
+The Polykrikos cell was isolated, and a single- cell transcriptome sequenced, which revealed two phylogenetically- distinct signals: one being Polykrikos and the other an uncharacterized MALV I (Fig 2a). Phylogenetic analysis of full- length SSU rRNA gene sequences extracted from the transcriptomic data showed the MALV I parasite clustering with full- support with a clade of unidentified environmental sequences in the MALV I clade (Fig 3). This fell within a larger group of parasites isolated from dinoflagellates, which were ascribed to Eudubosquella based on conserved morphological traits<sup>21</sup>. Eudubosquella has been isolated from both ciliates and dinoflagellates<sup>21,22,23,24</sup>, however the SSU rRNA gene tree (Fig. 3; Extended Data Fig. 4) shows MALV sequences isolated from these two host types form two phylogenetically discrete groups. Indeed, MALV- I diversity seems to fall into a small number of highly host- specific clades, infecting ciliates, fish eggs, dinoflagellates, and radiolarians. The type host of Eudubosquella is a ciliate (Favella panamensis), so we propose Deorella gen. n. for the dinoflagellate- infecting lineage and the Polykrikos- infecting specimen as Deorella krika (see Supplementary Information for taxonomic diagnosis).  
+
+<|ref|>text<|/ref|><|det|>[[115, 806, 866, 891]]<|/det|>
+Phylogenomic analyses including D. krika rejects the monophyly of MALV- I and MALV- II/IV, instead showing complete support for MALV- I branching specifically with another lineage of free- living heterotrophs, the common marine flagellate Oxyrrhis (Fig 2a).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 875, 430]]<|/det|>
+All topologies in which MALVs formed a monophyletic group were rejected by approximately unbiased (AU) tests (Extended Data Fig. 5). Indeed, Syndinales, as currently defined, are neither monophyletic nor paraphyletic, since we can demonstrate complete support for separate lineages being sister groups to different free- living heterotrophs. Parasitic transitions in MALV- I and MALV- II therefore trace back to different free- living ancestors and must have evolved their obligate parasitic life independently. We propose to retain the name Syndinales for MALV- II/IV and to create a new order for MALV- I, Ichthyodinales, after the earliest defined taxon within the group. Whether MALV III (and the little- studied MALV V) can all be included in Ichthyodinales seems likely based on SSU rRNA gene phylogenies (Ref 6 and Fig 3, Extended Data Fig. 4), but confirmation will await genomic data.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 478, 325, 495]]<|/det|>
+## Plastids and parasitism.  
+
+<|ref|>text<|/ref|><|det|>[[115, 509, 876, 895]]<|/det|>
+The complete absence of evidence for a plastid in Amoebophyra and Hematodinium<sup>14,15</sup> led to new hypotheses to account for the apparently complex distribution of plastids in early- diverging dinoflagellates and their sisters, the apicomplexan parasites<sup>17,25,26</sup>. But it is difficult to evaluate any hypothesis about the origin of plastids in parasites without data from close free- living relatives. Now that such relatives are identified, we examined plastid function across the deeply- branching dinoflagellate groups, including another relevant lineage, Psammosa. Psammosa is a free- living heterotroph related to dinoflagellates, but its exact position and whether it has a plastid are unclear in the absence of genomic data<sup>27</sup>. We generated transcriptomes from two strains of Psammosa, and phylogenomic analyses strongly support them forming an independent lineage at an important juncture in the tree, basal to Oxyrrhis, Ichthyodinales, Syndinales, and core dinoflagellates (Fig. 2a), or perhaps within the Oxyrrhis/Ichthyodinales clade itself (Fig. 2d, Extended Data Fig. 5).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 85, 870, 504]]<|/det|>
+The Polykrikos/Deorella data had too few parasite genes to determine whether Deorella has a plastid or to assess its function. In contrast, however, eleftherids and (to a lesser extent) psammosids encoded substantial numbers of nucleus- encoded genes for plastid- targeted proteins (as does Oxyrrhis), and these relate to metabolic pathways typical for non- photosynthetic plastids<sup>28</sup>. Specifically, we identified genes for the methylerythritol phosphate (MEP) pathway for isoprenoid biosynthesis and a plastidial SUF- type FeS cluster biosynthesis pathway including a ferredoxin redox system (Fig. 4 and Extended Data Fig. 6; phylogenetic trees available at https://figshare.com/s/fd3170a4b4a05027d9d2). These are both widespread in non- photosynthetic dinoflagellates and apicomplexans but absent in MALVs<sup>14</sup>. The complete transcripts encode N- terminal extensions with one or more predicted transmembrane domains that are consistent with plastid- targeting (IspF, SufC, SufD, SufE, PetF, PetH – in eleftherids), but do differ from canonical plastid- targeting sequences in core dinoflagellates<sup>29</sup> (alignments available at https://figshare.com/s/fd3170a4b4a05027d9d2). No evidence for the plastidial type II fatty acid biosynthesis pathway was found, which is also absent in MALVs and all other dinoflagellates with non- photosynthetic plastids<sup>27,30</sup> (Fig. 4 and Extended Data Fig. 6). Fatty acid biosynthesis likely occurs via a cytosolic type I fatty acid synthase (Supplementary Table 2).  
+
+<|ref|>text<|/ref|><|det|>[[112, 515, 880, 899]]<|/det|>
+acid biosynthesis pathway was found, which is also absent in MALVs and all other dinoflagellates with non- photosynthetic plastids<sup>27,30</sup> (Fig. 4 and Extended Data Fig. 6). Fatty acid biosynthesis likely occurs via a cytosolic type I fatty acid synthase (Supplementary Table 2). Heme is also synthesized in the plastids of some apicomplexans and dinoflagellates, but in eleftherids all these enzymes were mitochondrial (5- aminolevulinate synthase ALAS), or cytosolic (porphobilinogen synthase HemB, hydroxymethylbilane synthase HemC, uroporphyrinogen decarboxylase HemE), which is also the case in MALVs and Perkinsus. Oxyrrhis, by contrast, retains a mixed pathway with some enzymes in the plastid, like dinoflagellates and apicomplexans (Extended Data Fig. 6; phylogenetic trees available at https://figshare.com/s/fd3170a4b4a05027d9d2). The only enzyme proposed to have originated in the plastid of MALV II genomes is uroporphyrinogen-III synthase, or HemD<sup>14</sup>.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 872, 150]]<|/det|>
+We did not find a HemD orthologue in our transcriptome data (Fig. 4 and Extended Data Fig. 6), and it is also absent from many dinoflagellates, suggesting a low level of expression.  
+
+<|ref|>text<|/ref|><|det|>[[115, 150, 868, 375]]<|/det|>
+However, in our phylogenetic reconstruction of HemD (Extended Data Fig. 7), photosynthetic eukaryotes and MALV II do not group with cyanobacteria, suggesting a bacterial origin for HemD independent from the plastid. Unlike MALVs and the eleftherids, Oxyrrhis has retained plastidial versions of enzymes involved in heme biosynthesis (Extended Data Fig. 6). This indicates a functional redundancy in heme biosynthesis persisted long into the evolution of dinoflagellates, which may have facilitated the repeated loss of plastid heme biosynthesis, and perhaps even the loss of the plastid itself in MALV II.  
+
+<|ref|>text<|/ref|><|det|>[[115, 383, 864, 728]]<|/det|>
+The genome of Hematodinium (MALV IV) contains another potential plastid pathway, the lysine biosynthesis enzyme L,Ldiaminopimelate aminotransferase (DapL), which is only present in bacteria and a subset of plastids. Its phylogenetic origin in Hematodinium could not be resolved and support for its plastid origin was concluded to be ambiguous<sup>14</sup>. We identified DapL in eleftherids and psammosids (Fig. 4; protein absent in apicomplexans and most dinoflagellates) and phylogenetic reconstruction shows it to group with plastid homologues in cryptophytes, whereas the Hematodinium homologue appears to have originated independently from bacteria (Extended Data Fig. 8). While most eukaryotic DapL genes encode plastid-targeting sequences, eleftherid and cryptophyte genes do not, indicating a likely cytosolic function (alignment available at https://figshare.com/s/fd3170a4b4a05027d9d2).  
+
+<|ref|>text<|/ref|><|det|>[[115, 745, 872, 899]]<|/det|>
+We found no proteins related to photosynthesis or chlorophyll biosynthesis in eleftherids, psammosids, or Oxyrrhis (as expected). We also identified no plastid import proteins; however, the import machinery is understudied in dinoflagellates and only a single component has been reported<sup>31</sup>. There is no evidence for a plastid genome, which is also expected since the plastid genome of photosynthetic dinoflagellates encodes only 12 proteins
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 872, 235]]<|/det|>
+that are exclusively involved in photosynthesis<sup>32</sup>. We also investigated the 129 proteins predicted to localize to the apicoplast in the apicomplexan Toxoplasma by spatial proteomics<sup>33</sup> and found no clear candidates for plastid- targeted homologs, beyond the plastid pathways noted above. Taken together, the plastid in these non- photosynthetic lineages has been profoundly reduced.  
+
+<|ref|>sub_title<|/ref|><|det|>[[120, 283, 630, 302]]<|/det|>
+## Early character evolution in dinoflagellate-related lineages.  
+
+<|ref|>text<|/ref|><|det|>[[115, 315, 872, 765]]<|/det|>
+Dinoflagellates evolved in many strange ways, beyond parasitism and plastids<sup>28,34,35</sup>, and we now have a more accurate glimpse into these events as well. The nucleus, or "dinokaryon" is a good example: unlike other eukaryotes, dinoflagellates lack bulk nucleosomal chromatin and have huge genomes with permanently- condensed chromosomes structured by some combination of genes derived from horizontal gene transfer from viruses (Dinoflagellate Viral NucleoProtein, or DVNP)<sup>36,37</sup> and bacteria (two Histone- Like Proteins, or HLPs). DVNP appears in the tree after the divergence of the psammosids (Fig. 4 and Supplementary Table 2), whereas HLPs are present only in core dinoflagellates<sup>28</sup>. The absence of DVNP (which is highly expressed and easily detected in dinoflagellate transcriptomes) in psammosids strongly supports the phylogenomic topology in Fig. 2a, since the dinokaryon/DVNP is arguably the clearest synapomorphy of dinoflagellates and integral to dinoflagellate chromatin. If the psammosids were sister to Oxyrrhis and the Ichthyodiniales, it would require the loss of DVNP, which has not been described in any dinoflagellate lineage thus far.  
+
+<|ref|>text<|/ref|><|det|>[[115, 779, 872, 903]]<|/det|>
+Dinoflagellate nuclei also perform trans- splicing of a highly- conserved spliced leader (SL) to the 5' end of many or most mRNAs<sup>38</sup>. The same is also true of the perkinsids however the SL sequence is more variable<sup>39</sup>. We identified complete SL sequences in a subset of transcripts from eleftherids, MALV I, and psammosids, and found variations in
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 864, 135]]<|/det|>
+sequence from the canonical SL sequence identified in core dinoflagellates and MALVs \(^{15,38}\) (Fig. 4).  
+
+<|ref|>text<|/ref|><|det|>[[115, 149, 870, 433]]<|/det|>
+The ultrastructure of eleftherids also contains links to MALVs, which shed light on their diversification. Electron microscopy revealed the presence of rhoptry- like structures (Fig. 1o, r), similar to the electron- dense bodies found in the MALV II, Amoebophyra \(^{40}\) . Rhoptries are components of the infection apparatus of apicomplexan parasites, their free- living relatives, as well as Perkinsus and Psammosa \(^{41}\) . In eleftherids the presence of rhoptry- like structures may indicate components of an apical complex were also retained with a role in feeding, as is the case in Psammosa and a handful of free- living lineages related to apicomplexans \(^{42}\) . Altogether the distribution of rhoptry- like structures and plastids support a mixotrophic ancestor of dinoflagellates and apicomplexans (Fig. 4).  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 479, 475, 497]]<|/det|>
+## Parallelism in the evolution of parasitism  
+
+<|ref|>text<|/ref|><|det|>[[115, 510, 872, 893]]<|/det|>
+While we have focused mostly on the parasitic dinoflagellates, of course the sisters to dinoflagellates are the much more famously parasitic apicomplexans (e.g., the malaria parasite Plasmodium). Due to their medical significance, apicomplexans are far better studied, and since it became clear they too evolved from algae and retained a cryptic plastid, their evolution has been thoroughly investigated \(^{43,44}\) . On the apicomplexan side of the tree, it is now clear that many of the major transitions seemingly associated with the origin of parasitism actually evolved multiple times in parallel. This includes the loss of photosynthesis, the loss of the plastid organelle, and even the origin of parasitism itself \(^{45}\) . Marine environmental surveys show MALVs dominate eukaryotic communities in terms of both diversity and abundance. Accordingly, they are likely to play an important role in marine ecology, yet we know shockingly little about their biology or evolution. It is now at least becoming clear that some of the same patterns emerging from apicomplexan research
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 866, 135]]<|/det|>
+are also true of parasitic dinoflagellates, in particular parallel evolution in seemingly similar parasites.  
+
+<|ref|>text<|/ref|><|det|>[[115, 149, 876, 732]]<|/det|>
+Syndinales (MALV II and IV) and Ichthyodinales (MALV I and potentially MALV III and V) were previously thought to have evolved form a single parasitic ancestor \(^{1,2,7}\) , but phylogenomics now shows they evolved independently from different free- living ancestors, or in other words that parasitism evolved twice in the Marine Alveolates. Similarly, given that plastids are present in nearly every apicomplexan and dinoflagellate subgroup, and are demonstrably related \(^{45}\) , the ancestor of all these groups can be inferred to have had a plastid and be capable of photosynthesis. But photosynthesis was lost several times independently in both the basal- branching dinoflagellates and apicomplexans. The prevalence of non- photosynthetic organisms branching at the base of the dinoflagellate tree has led to complex theories about the origin of their plastids \(^{17}\) . However, based on current data it now appears that a parallel loss of photosynthesis in early dinoflagellate evolution may not be that remarkable, and indeed would be very similar to what is observed in apicomplexans: roughly the same number of independent loses of photosynthesis (six or seven) are required to explain both. It has been suggested that the loss of photosynthesis may be linked to the origin of parasitism in apicomplexans through a “photoparasitic” intermediate \(^{46}\) . However, the demonstration that both syndineans and ichthyodinians are specifically related to free- living, non- photosynthetic lineages suggest that photosynthesis was likely lost prior to the origin of their parasitic lifestyles, which might also extend to the origin of apicomplexans.  
+
+<|ref|>text<|/ref|><|det|>[[115, 740, 860, 891]]<|/det|>
+The outright loss of the plastid organelle is very rare compared to the loss of photosynthesis, but this also happened in parallel in both dinoflagellate and apicomplexan lineages. Plastids were lost at least twice in basal- branching apicomplexans (cryptosporidia and one lineage of gregarines), and at least once in dinoflagellates (Syndinales): more data are required to clarify whether the Ichthyodinales retain any plastid- related genes or the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 82, 868, 540]]<|/det|>
+organelle. In all these cases, it is interesting to note how little evidence of the existence of a plastid is retained once the organelle has been lost. The idea that the endosymbiotic origin of plastids necessarily had an extensive and lasting genetic impact on their host genome has penetrated deeply into our thinking about organelle evolution, but the relatively recent presence of a plastid in the ancestor of eleftherids and Syndinales has left no impression on the genome of Syndinales \(^{15}\) , and indeed the only reason we know a plastid ever existed is due to their phylogenetic position, not because of any genetic "footprint" of the organelle. This is also true of apicomplexan lineages that have lost their plastid entirely \(^{44}\) and is further reinforced by the recent demonstration that the Picozoa are plastid- lacking members of the archaeoplastids \(^{47}\) . All intuition aside, this seems to be the rule, rather than the exception. We argue that our baseline expectation should be that when an organelle is lost, it leaves no genetic trace whatsoever, and it will therefore be very difficult, without clear phylogenetic evidence, to distinguish a cell where an organelle was lost from one that never had the organelle in the first place.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[147, 85, 247, 102]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[115, 115, 870, 168]]<|/det|>
+1. Moon-van der Staay, S. Y., De Wachter, R. & Vaulot, D. Oceanic 18S rDNA sequences from picoplankton reveal unsuspected eukaryotic diversity. Nature 409, 607-610 (2001).  
+
+<|ref|>text<|/ref|><|det|>[[115, 180, 830, 268]]<|/det|>
+2. López-García, P., Rodríguez-Valera, F., Pedrós-Alió, C. & Moreira, D. Unexpected diversity of small eukaryotes in deep-sea Antarctic plankton. Nature 409, 603-607 (2001).  
+
+<|ref|>text<|/ref|><|det|>[[115, 280, 816, 333]]<|/det|>
+3. de Vargas, C. et al. Eukaryotic plankton diversity in the sunlit ocean. Science 348, 1261605 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[115, 345, 858, 430]]<|/det|>
+4. Massana, R. et al. Marine protist diversity in European coastal waters and sediments as revealed by high-throughput sequencing: Protist diversity in European coastal areas. Environ Microbiol 17, 4035-4049 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[115, 442, 860, 529]]<|/det|>
+5. Clarke, L. J., Bestley, S., Bissett, A. & Deagle, B. E. A globally distributed Syndiniales parasite dominates the Southern Ocean micro-eukaryote community near the sea-ice edge. ISME J 13, 734-737 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[115, 541, 864, 594]]<|/det|>
+6. Holt, C. C. et al. Microscopic marine invertebrates are reservoirs for cryptic and diverse protists and fungi. Microbiome 10, 161 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[115, 606, 833, 691]]<|/det|>
+7. Guillou, L. et al. Widespread occurrence and genetic diversity of marine parasitoids belonging to Syndiniales (Alveolata). Environmental Microbiology 10, 3349-3365 (2008).  
+
+<|ref|>text<|/ref|><|det|>[[115, 704, 852, 789]]<|/det|>
+8. Rizos, I. et al. Beyond the limits of the unassigned protist microbiome: inferring large-scale spatio-temporal patterns of Syndiniales marine parasites. ISME Commun. 3, 16 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[115, 802, 839, 888]]<|/det|>
+9. Skovgaard, A., Massana, R., Balagué, V. & Saiz, E. Phylogenetic Position of the Copepod-Infesting Parasite Syndinium turbo (Dinoflagellata, Syndinea). Protist 156, 413-423 (2005).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 84, 870, 164]]<|/det|>
+10. Gestal, C., Novoa, B., Posada, D., Figueras, A. & Azevedo, C. Perkinsoide chabelardi n. gen., a protozoan parasite with an intermediate evolutionary position: possible cause of the decrease of sardine fisheries? Environ Microbiol 8, 1105-1114 (2006). 
+
+<|ref|>text<|/ref|><|det|>[[111, 181, 875, 233]]<|/det|>
+11. Harada, A., Ohtsuka, S. & Horiguchi, T. Species of the Parasitic Genus Duboscquella are Members of the Enigmatic Marine Alveolate Group I. Protist 158, 337-347 (2007). 
+
+<|ref|>text<|/ref|><|det|>[[111, 249, 831, 300]]<|/det|>
+12. Strassert, J. F. H. et al. Single cell genomics of uncultured marine alveolates shows paraphyly of basal dinoflagellates. ISME J 12, 304-308 (2018). 
+
+<|ref|>text<|/ref|><|det|>[[111, 315, 855, 430]]<|/det|>
+13. Jackson, C. J., Gornik, S. G. & Waller, R. F. The Mitochondrial genome and transcriptome of the basal dinoflagellate Hematodinium sp.: character evolution within the highly derived mitochondrial genomes of dinoflagellates. Genome Biology and Evolution 4, 59-72 (2012). 
+
+<|ref|>text<|/ref|><|det|>[[111, 445, 850, 496]]<|/det|>
+14. Gornik, S. G. et al. Endosymbiosis undone by stepwise elimination of the plastid in a parasitic dinoflagellate. Proc. Natl. Acad. Sci. U.S.A. 112, 5767-5772 (2015). 
+
+<|ref|>text<|/ref|><|det|>[[111, 511, 869, 595]]<|/det|>
+15. Farhat, S. et al. Rapid protein evolution, organellar reductions, and invasive intronic elements in the marine aerobic parasite dinoflagellate Amoebophyra spp. BMC Biol 19, 1 (2021). 
+
+<|ref|>text<|/ref|><|det|>[[111, 610, 855, 630]]<|/det|>
+16. Xu, P. et al. The genome of Cryptosporidium hominis. Nature 431, 1107-1112 (2004). 
+
+<|ref|>text<|/ref|><|det|>[[111, 644, 860, 725]]<|/det|>
+17. Waller, R. F. & Kořený, L. Plastid complexity in dinoflagellates: a picture of gains, losses, replacements and revisions. In Advances in Botanical Research vol. 84 105-143 (Elsevier, 2017). 
+
+<|ref|>text<|/ref|><|det|>[[111, 741, 841, 792]]<|/det|>
+18. Villar, E. et al. The Ocean Gene Atlas: exploring the biogeography of plankton genes online. Nucleic Acids Research 46, W289-W295 (2018). 
+
+<|ref|>text<|/ref|><|det|>[[111, 807, 832, 858]]<|/det|>
+19. Duarte, C. M. Seafaring in the 21st Century: The Malaspina 2010 Circumnavigation Expedition. Limnology and Oceanography Bulletin 24, 11-14 (2015).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[58, 83, 875, 863]]<|/det|>
+20. Cordier, T. et al. Patterns of eukaryotic diversity from the surface to the deep-ocean sediment. Sci. Adv. 8, eabj9309 (2022).
+21. Yoo, J., Coats, D. W. & Kim, S. Syndinean dinoflagellates of the genus Euduboscquella are paraphyletic. J Eukaryotic Microbiology 70, (2023).
+22. Coats, D. W. & Bachvaroff, T. R. Parasites of Tintinnids. In The Biology and Ecology of Tintinnid Ciliates (eds. Dolan, J. R., Montagnes, D. J. S., Agatha, S., Coats, D. W. & Stoecker, D. K.) 145–170 (Wiley, 2012).
+23. Jung, J.-H., Choi, J. M., Coats, D. W. & Kim, Y.-O. Euduboscquella costata n. sp. (Dinoflagellata, Syndinea), an intracellular parasite of the ciliate Schmidingerella arcuata: morphology, molecular phylogeny, life cycle, prevalence, and infection intensity. J. Eukaryot. Microbiol. 63, 3–15 (2016).
+24. Choi, J. M., Jung, J. H., Kim, K. H., Coats, D. W. & Kim, Y. O. A novel parasitic, syndinean dinoflagellate Euduboscquella triangula infecting the tintinnid Helicostomella longa. Front. Mar. Sci. 8, 720424 (2021).
+25. Bodyl, A., Stiller, J. W. & Mackiewicz, P. Chromalveolate plastids: direct descent or multiple endosymbioses? Trends in Ecology & Evolution 24, 119–121 (2009).
+26. Ševčíková, T. et al. Updating algal evolutionary relationships through plastid genome sequencing: did alveolate plastids emerge through endosymbiosis of an ochrophyte? Sci Rep 5, 10134 (2015).
+27. Okamoto, N., Horák, A. & Keeling, P. J. Description of two species of early branching dinoflagellates, Psammosa pacifica n. g., n. sp. and P. atlantica n. sp. PloS ONE 7, e34900 (2012).
+28. Janouškovec, J. et al. Major transitions in dinoflagellate evolution unveiled by phylotranscriptomics. Proc. Natl. Acad. Sci. U.S.A. 114, (2017).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 83, 870, 164]]<|/det|>
+29. Patron, N. J., Waller, R. F., Archibald, J. M. & Keeling, P. J. Complex protein targeting to dinoflagellate plastids. Journal of Molecular Biology 348, 1015–1024 (2005).  
+
+<|ref|>text<|/ref|><|det|>[[110, 150, 852, 232]]<|/det|>
+30. Hehenberger, E., Gast, R. J. & Keeling, P. J. A kleptoplastic dinoflagellate and the tipping point between transient and fully integrated plastid endosymbiosis. Proc. Natl. Acad. Sci. U.S.A. 116, 17934–17942 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[110, 246, 878, 333]]<|/det|>
+31. Hehenberger, E., Imanian, B., Burki, F. & Keeling, P. J. Evidence for the retention of two evolutionary distinct plastids in dinoflagellates with diatom endosymbionts. Genome Biology and Evolution 6, 2321–2334 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[110, 345, 875, 399]]<|/det|>
+32. Dorrell, R. G. et al. Progressive and biased divergent evolution underpins the origin and diversification of peridinin dinoflagellate plastids. Mol Biol Evol msw235 (2016)  
+
+<|ref|>text<|/ref|><|det|>[[110, 411, 875, 499]]<|/det|>
+33. Barylyuk, K. et al. A comprehensive subcellular atlas of the Toxoplasma proteome via hyperLOPIT provides spatial context for protein functions. Cell Host & Microbe 28, 752-766.e9 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[110, 510, 825, 565]]<|/det|>
+34. Gavelis, G. S. et al. Microbial arms race: Ballistic “nematocysts” in dinoflagellates represent a new extreme in organelle complexity. Sci. Adv. 3, e1602552 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[110, 576, 837, 630]]<|/det|>
+35. Hackett, J. D., Anderson, D. M., Erdner, D. L. & Bhattacharya, D. Dinoflagellates: a remarkable evolutionary experiment. Am. J. Bot. 91, 1523–1534 (2004).  
+
+<|ref|>text<|/ref|><|det|>[[110, 641, 880, 695]]<|/det|>
+36. Gornik, S. G. et al. Loss of nucleosomal DNA condensation coincides with appearance of a novel nuclear protein in dinoflagellates. Current Biology 22, 2303–2312 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[110, 706, 787, 760]]<|/det|>
+37. Irwin, N. A. T. et al. Viral proteins as a potential driver of histone depletion in dinoflagellates. Nat Commun 9, 1535 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[110, 772, 856, 826]]<|/det|>
+38. Zhang, H. et al. Spliced leader RNA trans-splicing in dinoflagellates. Proc. Natl. Acad. Sci. U.S.A. 104, 4618–4623 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[110, 838, 809, 892]]<|/det|>
+39. Alacid, E. et al. A diversified and segregated mRNA spliced-leader system in the parasitic Perkinsozoa. Open Biol. 12, 220126 (2022).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 84, 868, 135]]<|/det|>
+40. Miller, J. J., Delwiche, C. F. & Coats, D. W. Ultrastructure of Amoebophrya sp. and its changes during the course of infection. Protist 163, 720–745 (2012). 
+
+<|ref|>text<|/ref|><|det|>[[115, 150, 874, 234]]<|/det|>
+41. Okamoto, N. & Keeling, P. J. The 3D structure of the apical complex and association with the flagellar apparatus revealed by serial TEM tomography in *Psammosa pacifica*, a distant relative of the Apicomplexa. *PloS ONE* 9, e84653 (2014). 
+
+<|ref|>text<|/ref|><|det|>[[115, 249, 875, 299]]<|/det|>
+42. Leander, B. S. & Keeling, P. J. Morphostasis in alveolate evolution. *Trends in Ecology & Evolution* 18, 395–402 (2003). 
+
+<|ref|>text<|/ref|><|det|>[[115, 314, 825, 365]]<|/det|>
+43. Janouškovec, J. et al. Apicomplexan-like parasites are polyphyletic and widely but selectively dependent on cryptic plastid organelles. *eLife* 8, e49662 (2019). 
+
+<|ref|>text<|/ref|><|det|>[[115, 380, 858, 430]]<|/det|>
+44. Mathur, V. et al. Multiple independent origins of apicomplexan-like parasites. *Current Biology* 29, 2936-2941.e5 (2019). 
+
+<|ref|>text<|/ref|><|det|>[[115, 445, 872, 527]]<|/det|>
+45. Janouškovec, J., Horák, A., Oborník, M., Lukeš, J. & Keeling, P. J. A common red algal origin of the apicomplexan, dinoflagellate, and heterokont plastids. *Proc. Natl. Acad. Sci. U.S.A.* 107, 10949–10954 (2010). 
+
+<|ref|>text<|/ref|><|det|>[[115, 542, 853, 593]]<|/det|>
+46. Oborník, M. Photoparasitism as an intermediate state in the evolution of apicomplexan parasites. *Trends in Parasitology* 36, 727–734 (2020). 
+
+<|ref|>text<|/ref|><|det|>[[115, 608, 810, 659]]<|/det|>
+47. Schön, M. E. et al. Single cell genomics reveals plastid-lacking Picozoa are close relatives of red algae. *Nat Commun* 12, 6651 (2021).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 115, 593, 714]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 728, 816, 749]]<|/det|>
+<center>Fig. 1: Cell morphology of Eleftheros. A–j, Living cells of E. xomoi (strain Cur-11), </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 761, 870, 912]]<|/det|>
+visualized by light microscopy. K- m, Living cells of E. karadeniz, visualized by light microscopy. N- s, Cells of E. karadeniz, visualized by transmission electron microscopy. N, Longitudinal section. O, Oblique section showing the transverse groove. P, Arrangement of basal bodies. Q, transverse section showing the longitudinal groove. R, Structure of the distal part with trophy- like structures and alveoli. S, Longitudinal and cross section (inset) of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 82, 875, 234]]<|/det|>
+419 trichocysts. A, alveoli; ax, axosome; bb, basal bodies; cm, central microtubules; f, flagella; 420 fv, food vacuole; lf, longitudinal flagellum; lg, longitudinal groove; m, mitochondrion; n, 421 nucleus; ps, perinuclear space; r, trophy-like structures; sc, storage compounds; t, trichocyst; 422 tf, transverse flagellum; tg, transverse groove; tp, transverse plate. Scale bars, \(4\mu \mathrm{m}\) (a-m), 1 423 \(\mu \mathrm{m}\) (n, o, q) and \(0,5\mu \mathrm{m}\) (p, r, s).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[130, 90, 860, 545]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 570, 858, 590]]<|/det|>
+<center>Fig 2. Phylogenomic reconstruction of the Marine Alveolates. a, Maximum likelihood </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 600, 872, 884]]<|/det|>
+Fig 2. Phylogenomic reconstruction of the Marine Alveolates. a, Maximum likelihood analysis of a 239 multi-protein alignment (LG+C60+F+G4) separate MALV I from the remaining Syndinales (MALV II and IV) to which the eleftherids are the free-living sister group. Black dots represent full statistical support (non- parametric bootstrap = 100%), values are shown for support below 100%. The percentage of sites for each taxon are shown as bars to the right. b, UltraFast bootstrap support for given topologies, indicated by highlighted node labels in a, following removal of the fastest-evolving sites as inferred using the LG + C 40 + F + G4 model. * indicates nodes that remained at 100% c, Living cell of Polykrikos sp. infected with a MALV 1- like parasite showing characteristic centripetal grooves. Scale bar, 12 μm. d, All supported topologies from approximately unbiased (AU) tests, showing
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 851, 202]]<|/det|>
+alternative position of psammosids. Rejected topologies (including those showing the Syndinales as a monophyletic clade) shown in Extended Data Figure. 5. Support for core dinoflagellate monophyly serves as a control for the presence of sufficient information for phylogenomic inference.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[170, 92, 858, 610]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[57, 640, 880, 860]]<|/det|>
+<center>Fig 3. Contextualising new MALVs using SSU data. Maximum likelihood analysis (GTR + F+ R10) showing MALV I lineages cluster by host, supporting the need for a distinct epithet for dinoflagellate-infecting MALV I species and the creation of Deorella g. n. as a separate genus. The eleftherids again form a sister clade to the remaining Syndinales but here cluster with uncultured environmental sequenced previously classified within MALV III. Black dots at nodes represent full statistical support (UltraFast bootstrap \(= 100\%\) ), values are shown for support below \(100\%\) . Coloured circles accompanying MALV I lineages reflects host identity. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 98, 856, 400]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 432, 875, 911]]<|/det|>
+<center>Fig 4. Character and plastid metabolic pathway evolution in dinoflagellates, apicomplexans and related lineages. Schematic tree indicating abundance of and character evolution in dinoflagellates, apicomplexans and related lineages. Black triangles/lines denote an estimation of the relative diversity of each lineage, while crossed-out circles indicate plastid loss (i.e. in the ancestor of the MALVs and within the apicomplexans). Abbreviations on branches indicate the inferred origin of: L,Ldiaminopimelate aminotransferase, DapL; SL, spliced leader; DVNPs, Dinoflagellate Viral NucleoProteins; HLPs, histone-like proteins. Next to each branch, Coulson plots depict selected plastidial metabolic pathways in the corresponding lineage. Grey segments indicate a cytosolic form of a given protein, while an offset, yellow segment denotes the mitochondrial enzyme 5-aminolevulinate synthase (ALAS) that synthesizes 5-aminolevulinate in all lineages but the core dinoflagellates, where the same intermediate is synthesized by the plastidial enzymes glutamyl-tRNA reductase (HemA) and glutamate-1-semialdehyde 2,1-aminomutase (HemL). Red nucleotides in SL sequence indicate deviation from canonical SL. Blue ovals surrounding cell depictions depict host cell, indicating parasitism. Coulson plot headers refer to metabolic pathways: isoprenoid, </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 82, 881, 202]]<|/det|>
+466 methylerythritol phosphate (MEP) pathway for isoprenoid biosynthesis; FASII, type 2 fatty 467 acid biosynthesis pathway; FeS cluster, plastidial Fe-S cluster biosynthesis pathway including 468 a ferrodoxin system; heme, heme biosynthesis. Complete protein names are listed in 469 Supplementary Table 2.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[117, 85, 197, 101]]<|/det|>
+## Methods  
+
+<|ref|>text<|/ref|><|det|>[[115, 116, 874, 200]]<|/det|>
+Data reporting. No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.  
+
+<|ref|>text<|/ref|><|det|>[[115, 214, 872, 594]]<|/det|>
+Cell isolation and culture establishment. Strains Colp- 37 and Cur- 11 (Eleftheros xomoi gen. et sp. nov.) were obtained from the surface of brain coral Colpophylia natans Houttuyn, 1772 in coastal waters of Curacao (Caribbean Sea) in April 2016 and 2018, respectively. Strain Colp- 25 (Eleftheros karadeniz gen. et sp. nov.) was isolated from the near shore bottom sediments in the Black Sea near T.I. Vyazemsky Karadag Scientific Station, Crimea, May 2015. The samples were enriched with a suspension of Aeromonas sobria bacteria and examined on the third, sixth and ninth day of incubation ( \(25^{\circ}\mathrm{C}\) , darkness) in accordance with methods described previously \(^{48}\) . To obtain clonal cultures, the individual cells were transferred using a drawn- out glass micropipette into Petri dishes containing a clonal culture of eukaryotic prey Procryptobia sorokini (strain B- 69), which were grown in marine Schmalz- Pratt's medium at a final salinity of \(20\%\) using the bacterium A. sobria as food \(^{49}\) . Strains perished after several months of cultivation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 607, 877, 727]]<|/det|>
+The Deorella- infected Polykrikos cell was collected with a \(20\mu \mathrm{m}\) mesh net towed at Heriot Bay near the Hakai Quadra Island Ecological Observatory in British Columbia, October 2022. The infected cell was isolated with a drawn- out glass micropipette and washed with \((0.2\mu \mathrm{m})\) filtered water before imaging and lysis.  
+
+<|ref|>text<|/ref|><|det|>[[115, 739, 845, 795]]<|/det|>
+Psammosa pacifica Psp strain was isolated from Boundary Bay, British Columbia Canada (45.25580°N; - 64.34907°W) and maintained according to Okamoto et al. 2012 \(^{27}\) .  
+
+<|ref|>text<|/ref|><|det|>[[115, 820, 850, 905]]<|/det|>
+Psammosa sp. (strain Colp- 34) was isolated from the near shore bottom sediments in the Kapsel Bay, Black Sea, Crimea, May 2016 and maintained following the abovementioned protocol.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 875, 202]]<|/det|>
+Light and electron microscopy. To observe living cells, an AxioScope A1 light microscope (Carl Zeiss, Jena, Germany) with DIC water immersion objective \(63 \times\) and an inverted microscope Leica DM IL LED with DIC objectives \(40 \times\) and \(63 \times\) were used. Images were captured with a Sony \(\alpha 7R\) camera.  
+
+<|ref|>text<|/ref|><|det|>[[115, 216, 881, 435]]<|/det|>
+For transmission electron microscopy (TEM), cells were centrifuged, fixed at \(1^{\circ}\mathrm{C}\) for 60 min in a cocktail of \(0.6\%\) glutaraldehyde and \(2\% \mathrm{OsO_4}\) (final concentration) prepared using a \(0.1\mathrm{M}\) cacodylate buffer (pH 7.2). Fixed cells were dehydrated in alcohol and acetone series (30, 50, 70, 96, and \(100\%\) , 20 min in each step). Afterward, the cells were embedded in a mixture of Araldite and Epon (Fluka, 45345). Ultrathin sections (60 nm) were prepared with a Leica EM UC6 ultramicrotome (Leica Microsystems, Germany) and observed by using a JEM 1011 transmission electron microscope (JEOL, Japan).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 448, 468, 466]]<|/det|>
+## Preparation of libraries and sequencing.  
+
+<|ref|>text<|/ref|><|det|>[[115, 479, 880, 794]]<|/det|>
+RNA isolation and cDNA preparation. Cells from clonal culture were collected by centrifugation (1000 x g, room temperature) onto the \(0.8 \mu \mathrm{m}\) membrane of a Vivaclear mini column (Sartorium Stedim Biotech Gmmg, VK01P042). Total RNA was then extracted using a RNAqueous- Micro Kit (Invitrogen, AM1931) and reverse transcribed into cDNA using the Smart- Seq2 protocol<sup>50</sup>, which uses poly- A selection to enrich mRNA. Additionally, cDNA of Cur11 and Colp- 37 were obtained from 20 single cells using the Smart- Seq2 protocol (cells were manually picked from culture using a drawn- out glass micropipette and transferred to a \(0.2 \mathrm{mL}\) thin- walled PCR tube containing \(2 \mu \mathrm{L}\) of cell lysis buffer - \(0.2\%\) Triton X- 100 and RNase inhibitor (Invitrogen)). Likewise, cDNA from the infected Polykrikos cell and Psammosa spp. was prepared following the same protocol.  
+
+<|ref|>text<|/ref|><|det|>[[115, 807, 878, 891]]<|/det|>
+Sequencing dataset assembly and decontamination. Deorella libraries were sequenced with an Illumina NextSeq 500 system using a Mid Output Flow Cell (2 x 150bp reads). Eleftheros and Psammosa libraries were sequenced on an Illumina MiSeq platform with read lengths of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 870, 169]]<|/det|>
+\(2 \times 300 \mathrm{bp}\) (strains Colp- 25, Colp- 37 and Colp- 34) and on an Illumina HiSeq 2500 machine, \(2 \times 125 \mathrm{bp}\) reads (strain Cur- 11). Sequence quality and adapter contamination of reads from transcriptomic datasets were assessed with FastQC v.0.10.15.  
+
+<|ref|>text<|/ref|><|det|>[[115, 180, 881, 562]]<|/det|>
+Reads of clone Colp- 37 (culture and 20- cell preparations) and clone Colp- 25 (culture) were merged with PEAR v.0.9.6 \(^{52}\) and resulting assembled as well as unassembled reads were separately trimmed with Trimmomatic \(^{53}\) as implemented in Trinity v.2.0.6 \(^{54}\) , removing Illumina adapters with ILLUMINACLIP, with a maximum of two mismatches, a palindrome clip threshold of 30 and a simple clip threshold of 10. Low- quality sequences were discarded, using a sliding window of 4 bp and a minimum trimmed length of 25 bp. Trimmed assembled and unassembled reads were combined into a single file and transcriptomes were assembled with Trinity, using the --single flag. Contaminating non- eukaryotic (bacterial, archaeal, and viral) and prey contigs were identified using BLASTn \(^{55}\) queries of the NCBI nt database. Sequences that aligned with \(\geq 100\) nucleotides, had a query coverage of \(\geq 80\%\) and were \(\geq 90\%\) identical to noneukaryotic entries were removed, while all sequences resulting in a kinetoplastid best hit were removed.  
+
+<|ref|>text<|/ref|><|det|>[[115, 575, 866, 857]]<|/det|>
+Reads of the later sampled clone Cur- 11 (culture and 20- cell preparations) were trimmed and assembled without prior read merging in a single step, using the Trimmomatic plugin embedded in Trinity. Trimming parameters remained unchanged except for a simple clip threshold of 9. Due to results from initial analyses of clone Colp- 37, the transcriptome was not subjected to any cleaning steps to retain putative bacterial horizontal gene transfers as well as sequences with bacterial best hits in nt that were identified as likely being eukaryotic in phylogenetic reconstructions in clone Colp- 37. Instead, all genes of interest were subjected to phylogenetic analysis, as described below, to clarify the taxonomic identity of the gene.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 863, 268]]<|/det|>
+Reads from the Deorella/Polykrikos and psammosid libraries were trimmed using Cutadapt v3.2<sup>56</sup> before assembly with rnaSPAdes v3.15.1<sup>57</sup>. Contaminating sequences were removed using BLASTx<sup>58</sup> and BLASTn<sup>55</sup> searches against the NCBI nt and UniProt databases (E- value cut- off = 1x10<sup>-25</sup>). Removal of prey contigs, Spumella elongata for Psammosa pacifica and Procryptobia sorokini for Psammosa sp. were characterized using BLASTn<sup>55</sup> against prey transcriptome data and removed.  
+
+<|ref|>text<|/ref|><|det|>[[115, 280, 870, 365]]<|/det|>
+Prediction of protein coding regions in all assemblies was performed by TransDecoder v.5.0.2, including BLASTp<sup>58</sup> queries of the Swiss- Prot database (E- value cut- off = 1x10<sup>-5</sup>).  
+
+<|ref|>text<|/ref|><|det|>[[115, 378, 880, 560]]<|/det|>
+Phylogenomic dataset preparation and analysis. In addition to the taxa described above we added to or updated the following available transcriptome or genome data in an existing phylogenomic framework<sup>59</sup>. We added the re- assembled transcriptomes of several core dinoflagellates generated by the MMETSP project<sup>60</sup>, several transcriptomes from the EukProt database<sup>61</sup>, the transcriptomes of the dinoflagellates Lepidodinium chlorophorum (https://www.ncbi.nlm.nih.gov/bioproject/481676), TGD and  
+
+<|ref|>text<|/ref|><|det|>[[115, 573, 864, 680]]<|/det|>
+MGD<sup>62</sup>, Abedinium<sup>63</sup>, Amyloodinium ocellatum<sup>64</sup> and additional taxa from Cooney et al. 2021 including Spatulodinium, Kofoidinium and Fabadinium<sup>65</sup>. We updated the data for the MALVs Amoebophyra ceratii<sup>66</sup> and  
+
+<|ref|>text<|/ref|><|det|>[[115, 671, 861, 891]]<|/det|>
+Hematodinium sp.<sup>14</sup>, for Oxyrrhis marina<sup>60</sup>, for the perkinsids Perkinsus marinus (https://protists.ensembl.org/Perkinsus_marinus_atcc_50983_gca_000006405/Info/Index), Maranthos nigrum and Parvilucifera sinerae<sup>39</sup>, the apicomplexa Ancora sagittata<sup>43</sup>, and for the chromerids Vitrella brassicaformis (https://cryptodb.org/cryptodb/app/record/dataset/NCBITAXXON_1169540) and Symbiont X<sup>43</sup>. Transcriptomic datasets were subjected to TransDecoder coding region prediction to extract peptides for downstream analyses.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 80, 875, 696]]<|/det|>
+All listed datasets were searched for 263 proteins to generate single- protein trees as described in Burki et al. 2016 \(^{59}\) . In brief: BLASTp was used to identify homologues to the 263 genes in the new datasets. After a parsing step (E- value \(\leq 1 \times 10^{- 20}\) ), a maximum of four non- redundant hits was added to the initial 263- protein set. The expanded gene sets were aligned with MAFFT L- ins- i v.7.222 \(^{67}\) and trimmed automatically with trimAl v1. \(^{26}\) , with a gap threshold of 80%. Single- protein maximum- likelihood phylogenies were reconstructed under the LG + G4 model using RAxML v.8.1. \(^{69}\) in combination with 100 rapid bootstraps and resulting trees were manually screened to flag paralogues and sequences derived from prey or other contamination. Cleaned protein sets were aligned and trimmed as above and taxa were selected upon concatenation with SCaFOs v.1.2. \(^{50}\) to select proteins sequences present in \(\geq 60\%\) of all taxa. To improve data presence, the concatenated sequences of clone Colp- 37 derived from culture and 20- cell preparations were merged, as were the sequences for the two clone Cur- 11 datasets. The final concatenated alignment included 77- taxa, 239 proteins and 64,690 amino acid sites. Clone Cur- 11 was represented with 91% of proteins and 85% of sites, clone Colp- 37 with 81% of proteins and 71% of sites and clone Colp- 25 with 49% of proteins and 37% of sites. The host Polykrikos sp. was represented with 75% of proteins and 66% of sites and Deorella with 26% and 15% respectively. Psammosa pacifica Psp was represented with 75% and 62% of sites, Psammosa sp. was represented with 74% of proteins and 64% of sites.  
+
+<|ref|>text<|/ref|><|det|>[[115, 705, 870, 890]]<|/det|>
+Maximum- likelihood phylogenomic tree reconstruction was performed using IQ- TREE v. 1.6. \(^{57}\) using the C60 empirical mixture model in combination with the LG matrix, amino acid frequencies computed from the data and four gamma categories for handling the rate heterogeneity across sites (LG + C60 + F + G4 model with 1000 UFBoot replicates \(^{72}\) ). Bayesian analyses were carried out by running four independent Markov chain Monte Carlo chains with PhyloBayes MPI v.1.8c \(^{73}\) , using the GTR matrix in combination with an infinite
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 83, 875, 250]]<|/det|>
+mixture model and four discrete gamma categories \((\mathrm{CAT} + \mathrm{GTR} + \mathrm{G4})\) . Chains were run for more than 10,000 generations saving every second tree, and the first 200 generations (or \(20\%\) ) were discarded as burn- in. As is frequently seen in large- scale phylogenomic analyses, the chains failed to converge (maxdiff \(= 1\) and meandiff \(= 0.0171456\) ), with the topologies differing within the core dinoflagellates and the apicomplexan plus chromerid clade.  
+
+<|ref|>text<|/ref|><|det|>[[115, 247, 878, 301]]<|/det|>
+Bayesian posterior probabilities are reported as a measure of statistical support for bipartitions. Phylogenetic trees were visualized in \(\mathrm{R}^{74}\) with the ggtree \(^{75}\) and treelio \(^{76}\) packages.  
+
+<|ref|>text<|/ref|><|det|>[[115, 312, 877, 496]]<|/det|>
+Fast- evolving sites were estimated with IQ- TREE using the - wsr option, using the LG \(+ \mathrm{C20} + \mathrm{F} + \mathrm{G}\) model. Sites were removed in increments of \(5\%\) of the original alignment length, up to \(50\%\) , and for each subsample with a reduced number of sites trees were reconstructed with the \(\mathrm{LG} + \mathrm{C20} + \mathrm{F} + \mathrm{G}\) model. The support for the sister relationship of eleftherids and MALVs was assessed at each removal increment. The monophyly of core dinoflagellates was tested as a control.  
+
+<|ref|>text<|/ref|><|det|>[[115, 508, 875, 858]]<|/det|>
+Small subunit ribosomal RNA (SSU rRNA) phylogeny. The most complete sequences for SSU rRNA genes were identified from each eleftherid transcriptome with BLASTn, using a known dinoflagellate SSU query, whereas barnap \(^{77}\) was used to extract Deorella SSU rRNA sequences prior to identification with BLAST. An alignment of all near full- length MALV sequences deposited in GenBank was curated to include all five previously described MALV groups in Guillou et al. \(^{7}\) . The resulting SSU collection was aligned using MAFFT with the E- INS- I algorithm and inspected for misaligned and chimeric sequences. Shorter sequences obtained from surveys of environmental data (see below) were aligned using the addfragments flag. The final alignment was trimmed with trimAl (- gt 0.1, - st 0.001) before generating a maximum likelihood tree in IQTREE with 1000 ultrafast bootstrap replicates \(^{71}\) , using the \(\mathrm{GTR} + \mathrm{F} + \mathrm{R10}\) model (and \(\mathrm{GTR} + \mathrm{R} + \mathrm{R60}\) for Extended Data Figure 4), selected
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 850, 135]]<|/det|>
+with ModelFinder<sup>78</sup>. Oxyrrhis marina and Ellobiopsis chattonii were omitted to minimize long branch attraction.  
+
+<|ref|>text<|/ref|><|det|>[[112, 144, 880, 895]]<|/det|>
+Identification of putative plastid- targeted proteins. Putative plastid- targeted proteins were identified combining a BLASTp- based similarity search and a hidden Markov models (HMMs) based screen. Known dinoflagellate<sup>30</sup> or MALV<sup>14</sup> proteins involved in plastid metabolic pathways were used as queries in a BLASTp search against a comprehensive custom database containing representatives from most major eukaryotic groups (excluding the long- branching excavates and the data- poor group of Rhizaria) and RefSeq data from all bacterial phyla at NCBI (last accessed December 2017). The database was subjected to CD- HIT<sup>79</sup> clustering with a similarity threshold of 85% to reduce redundant sequences and paralogues, except for the data sets created in this study (clustered at 98%). The search results of the BLASTp step were parsed for hits with an E- value threshold \(\leq 1 \times 10^{- 25}\) and a query coverage of \(\geq 30\%\) to reduce the possibility of paralogs and extremely short sequences and at the same time recover possibly fragmented eleftherid homologues. The number of bacterial hits was restrained to 20 hits per phylum (for FCB group, most classes of Proteobacteria, PVC group, Spirochaetes, Actinobacteria, Cyanobacteria (unranked) and Firmicutes) or 10 per phylum (remaining bacterial phyla) as defined by NCBI taxonomy. In some cases (HemD and DapL) these numbers were expanded to 20 and 40, respectively, for a more representative bacterial sampling. In addition, the Colp- 37 and Cur- 11 protein data were used to search the Pfam- A database release 33.0 with hmmscan (HMMER3.1; hmmer.org), employing the manually curated Pfam gathering threshold. The results were queried, using a keyword search, for Pfam domains present in plastid- associated proteins, including proteins involved in metabolic pathways in Hehenberger et al. 2019<sup>30</sup>, photosynthesis, plastid import (TIC/TOC components) and Calvin cycle (RuBisCO). Candidates with domains of interest were used as BLASTp queries as described above and parsed hits (query coverage of \(\geq 50\%\) ).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 82, 875, 900]]<|/det|>
+were combined with the recovered hits from the known proteins. After a deduplication step, sequences were aligned with MAFFT using the --auto option, trimmed using trimAl (-gt 0.8) and Maximum-likelihood tree reconstructions were performed with FastTree v.2.1.7<sup>80</sup> using the default options in a preliminary analysis. The resulting phylogenies and underlying alignments were inspected manually to remove contaminant, divergent, and/or low-quality sequences. The cleaned, unaligned sequences were then subjected to filtering with PREQUAL<sup>81</sup> using the default options, followed by alignment with MAFFT G- INS-i using the VSM option (--unalignlevel 0.6). The alignments were subjected to Divvier<sup>82</sup> using the - mincol 4 and the -divvygap option before trimming with trimAl (-gt 0.01). Final trees were calculated with IQ- TREE, using the -mset option to restrict model selection to LG for ModelFinder<sup>78</sup>, while branch support was assessed with 1000 ultrafast bootstrap replicates<sup>72</sup>. We also searched for homologues to the 129 proteins predicted to localize to the apicoplast in Toxoplasma<sup>33</sup>. The predicted Toxoplasma proteins were used as queries in a BLASTp search against our custom database, using an initial E-value threshold of \(\leq 1 \times 10^{- 25}\) and a query coverage of \(\geq 30\%\) for parsing. All proteins recovering putative eleftherid homologues were used in a preliminary tree reconstruction analysis as described above. After manual inspection of the phylogenies, potential plastid-targeted candidates were further investigated by using the recovered eleftherid homologues as additional BLASTp queries and combining the resulting hits (query coverage of \(\geq 50\%\) ) with the initial BLASTp output as described above. The BLASTp search using the Toxoplasma proteins was repeated with relaxed parameters (E-value threshold \(\leq 1 \times 10^{- 5}\) ) to recover additional eleftherid candidates. All candidates investigated were also submitted to a web BLASTp search against nr to exclude the possibility of a contamination not present in our database. N-terminal extensions of putative plastid-targeted sequences were investigated in the respective protein alignments, visualized with AliView<sup>83</sup>. For easier recognition of such
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 863, 199]]<|/det|>
+extensions, only the dinoflagellate sequences of the protein of interest plus the prokaryotic sequence with the highest sequence similarity were viewed. Eleftherid sequences with N- terminal extensions relative to prokaryotic sequences were submitted to SignalP- 3.0<sup>84</sup> and TMHMM 2.0<sup>85</sup> to predict putative signal peptides and transmembrane domains, respectively.  
+
+<|ref|>text<|/ref|><|det|>[[115, 213, 880, 333]]<|/det|>
+For proteins with known bacterial orthologs, bacterial naming conventions were applied, with the exception of the multifunctional enzyme Acetyl- CoA carboxylase (ACC) not present in bacteria and the enzyme 5- aminolevulinate synthase (ALAS), where eukaryotic conventions were applied.  
+
+<|ref|>text<|/ref|><|det|>[[115, 346, 864, 530]]<|/det|>
+Identification of molecular characteristics of dinoflagellates. Dinoflagellate Viral NucleoProteins (DVNPs) in eleftherids were identified using a BLASTp search against our custom database using all described Hematodinium DVNP proteins<sup>36</sup> as well as by performing an hmmsearch (HMMER3.1; hmmer.org) against our database using the DVNP profile HMM downloaded from pfam.xfam.org, employing default thresholds. The two approaches recovered the same set of eleftherid DVNP candidates.  
+
+<|ref|>text<|/ref|><|det|>[[115, 542, 877, 725]]<|/det|>
+We also searched for histone- like proteins (HLPs), using known representatives of the two known types of HLPs in dinoflagellates<sup>28</sup> as query, Crypthecodinium cohnii AAM97522.1 (HLPI) and Noctiluca scintillans ABV22345.1 (HLPII) in a BLASTp search against the predicted eleftherid peptides (E- value threshold \(\leq 1 \times 10^{- 25}\) ). Additionally, we performed an hmmsearch using a profile HMM constructed from a curated alignment of dinoflagellate and prokaryotic HLPs.  
+
+<|ref|>text<|/ref|><|det|>[[115, 738, 876, 890]]<|/det|>
+Dinoflagellate spliced leader (SL) sequences were identified using BLASTn searches (using the option - task blastn to allow for short input queries and to identify short matches) against the all transcriptomes, using the canonical 21- nucleotide dinoflagellate SL sequence<sup>38</sup> as query. The recovered 21- nucleotide sequence in eleftherids, differing in 2 nucleotides (5A>T and 17T>A) from the canonical sequence, was used as a search string to identify all
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 870, 168]]<|/det|>
+eleftherids transcripts containing a full- length SL. To avoid SL sequences in host Polykrikos sequences, Deorella peptides identified in single- gene trees were aligned against Polykrikos/Deorella transcripts with tBLASTn to obtain Deorella only transcripts.  
+
+<|ref|>text<|/ref|><|det|>[[115, 180, 878, 268]]<|/det|>
+Environmental distribution and abundance. To uncover the global distribution and abundance of eleftherids, we searched for sequences similar to the eleftherid SSU rRNA gene among published environmental SSU rRNA gene amplicon studies.  
+
+<|ref|>text<|/ref|><|det|>[[115, 279, 873, 431]]<|/det|>
+Full- length eleftherid SSU rRNA genes and their V4 and V9 hypervariable regions were used as queries in BLASTn searches (E- value threshold \(\leq 1 \times 10^{- 10}\) ) against the complete Tara Ocean database available on Ocean Gene Atlas<sup>18</sup>. We recovered no sequences with \(\geq 95\%\) identity and a query cover of \(\geq 90\%\) , neither when using full- length nor hypervariable regions of eleftherid SSU rRNA gene queries.  
+
+<|ref|>text<|/ref|><|det|>[[114, 443, 872, 892]]<|/det|>
+Additionally, we searched a custom environmental sequence database containing data from 26 environmental amplicon studies with a focus on marine sediment, also including the large sequencing projects BioMarKs<sup>4</sup> and Malaspina<sup>19</sup>. NCBI SRA files of amplicon data were downloaded using fastq- dump from stratolkit 2.10.8. Raw sequence data were processed with MICrobial Community Analysis, using micca merge or mergepairs (- l 100 - d 30) for single- end or paired- end data respectively<sup>86</sup>, and then concatenated and converted them into a BLAST database composed of 113,203,549 sequences with a total of 25,261,490,053 letters. Alternatively, paired- end data were processed in Qiime<sup>28</sup> with the DADA2<sup>88</sup> denoising algorithm before creation of the BLAST database. This database was searched using full- length eleftherid SSU rRNA gene sequences as well as their V4 and V9 hypervariable regions in BLASTn searches (E- value threshold \(\leq 1 \times 10^{- 25}\) ). In addition to the environmental database, sequences similar to the eleftherid 18S SSU gene were searched using BLASTn against a V9 amplicon dataset collected from sandy beaches off the central coast of British Columbia (the European Nucleotide Archive project PRJEB14727). A total
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 864, 366]]<|/det|>
+of 73 unique sequences with \(\geq 97\%\) identity to and \(\geq 90\%\) coverage of the eleftherid V9 and V4 hypervariable regions resulted from both search approaches. Full- length and hypervariable regions V9 and V4 of \(P\) pacifica were used as queries against the environmental database which resulted in a total of 4 unique sequences with with \(\geq 97\%\) identity to and \(\geq 90\%\) coverage to psammosids. These sequences were then clustered at \(99\%\) identity using CD- HIT prior to being added to the SSU rRNA gene phylogeny to confirm their phylogenetic position as described above. Sequences less than 150bp were removed from the final phylogeny. Sample coordinates corresponding to BLAST hits in the custom database were extracted and plotted in \(\mathrm{R}^{74}\) with the nraturearth \(^{89}\) package.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 412, 263, 430]]<|/det|>
+## Data availability  
+
+<|ref|>text<|/ref|><|det|>[[115, 444, 875, 758]]<|/det|>
+Raw transcriptome reads will be deposited in the GenBank Sequence Read Archive (SRA). SSU rRNA gene sequences retrieved from the transcriptomes will also be deposited in GenBank. Assembled transcriptomes, along with individual gene alignments, concatenated and trimmed alignments, and maximum- likelihood and Bayesian tree files for the phylogenomic dataset will be available at Figshare. The untrimmed and trimmed alignments, alignments depicting N- terminal extensions and tree files in nexus and pdf format for plastid- associated and other proteins of interest will be available at Figshare. The genus Eleftheros and species Eleftheros xomoi and Eleftheros karadeniz will be registered with the Zoobank database (http://zoobank.org/). As will the family Deorellidae, genus Deorella, and species Deorella krika.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 806, 266, 824]]<|/det|>
+## Code availability  
+
+<|ref|>text<|/ref|><|det|>[[115, 838, 853, 858]]<|/det|>
+All unpublished code is available upon reasonable request from the corresponding authors.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[118, 84, 297, 101]]<|/det|>
+## Methods References  
+
+<|ref|>text<|/ref|><|det|>[[115, 115, 850, 170]]<|/det|>
+48. Tikhonenkov, D. V., Mazei, Y. A. & Embulaeva, E. A. Degradation succession of heterotrophic flagellate communities in microcosms. Zh. Obs. Biol. 69, 57-64 (2008).  
+
+<|ref|>text<|/ref|><|det|>[[115, 183, 860, 299]]<|/det|>
+49. Tikhonenkov, D. V. et al. Description of Colponema vietnamica sp.n. and Acavomonas peruviana n. gen. n. sp., two new alveolate phyla (Colponemidia nom. nov. and Acavomonidia nom.nov.) and their contributions to reconstructing the ancestral state of alveolates and eukaryotes. PLoS One 9, (2014).  
+
+<|ref|>text<|/ref|><|det|>[[115, 313, 857, 367]]<|/det|>
+50. Picelli, S. et al. Full-length RNA-seq from single cells using Smart-seq2. Nat Protoc 9, 171-181 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[115, 381, 855, 435]]<|/det|>
+51. Andrews, S. FastQC: A quality control tool for high throughput sequence data. version 0.10.1. https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ (2010).  
+
+<|ref|>text<|/ref|><|det|>[[115, 448, 850, 502]]<|/det|>
+52. Zhang, J., Kobert, K., Flouri, T. & Stamatakis, A. PEAR: A fast and accurate Illumina Paired-End reAd mergeR. Bioinformatics 30, 614-620 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[115, 515, 850, 569]]<|/det|>
+53. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114-2120 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[115, 582, 816, 635]]<|/det|>
+54. Grabherr, M. G. et al. Trinity: reconstructing a full-length transcriptome without a genome from RNA-Seq data. Nat. Biotechnol. 29, 644-652 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[115, 648, 866, 702]]<|/det|>
+55. Zhang, Z., Schwartz, S., Wagner, L. & Miller, W. A greedy algorithm for aligning DNA sequences. Journal of Computational Biology vol. 7 203-214 (2000).  
+
+<|ref|>text<|/ref|><|det|>[[115, 714, 875, 767]]<|/det|>
+56. Martin., M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal. 17, 10-12 (2011).  
+
+<|ref|>text<|/ref|><|det|>[[115, 780, 864, 834]]<|/det|>
+57. Bankevich, A. et al. SPAdes: A new genome assembly algorithm and its applications to single-cell sequencing. J. Comput. Biol. 19, 455-477 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[115, 847, 874, 900]]<|/det|>
+58. Camacho, C. et al. BLAST+: Architecture and applications. BMC Bioinformatics 10, 1-9 (2009).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 84, 875, 168]]<|/det|>
+59. Burki, F. et al. Untangling the early diversification of eukaryotes: A phylogenomic study of the evolutionary origins of centrohelida, haptophyta and cryptista. Proc. R. Soc. B Biol. Sci. 283, (2016).  
+
+<|ref|>text<|/ref|><|det|>[[110, 181, 848, 268]]<|/det|>
+60. Johnson, L. K., Alexander, H. & Brown, C. T. Re-assembly, quality evaluation, and annotation of 678 microbial eukaryotic reference transcriptomes. Gigascience 8, 1–12 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[110, 281, 852, 335]]<|/det|>
+61. Richter, D. J. et al. EukProt: A database of genome-scale predicted proteins across the diversity of eukaryotes. Peer Community Journal 2, e56 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[110, 348, 864, 402]]<|/det|>
+62. Sarai, C. et al. Dinoflagellates with relic endosymbiont nuclei as models for elucidating organellogenesis. Proc. Natl. Acad. Sci. U. S. A. 117, 5364–5375 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[110, 414, 816, 468]]<|/det|>
+63. Cooney, E. C. et al. Single-cell transcriptomics of Abedinium reveals a new earlybranching dinoflagellate lineage. Genome Biol. Evol. 12, 2417–2428 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[110, 480, 875, 534]]<|/det|>
+64. Byadgi, O. et al. Transcriptome Analysis of Amyloodinium ocellatum Tomonts Revealed Basic Information on the Major Potential Virulence Factors. Genes 11, 1252 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[110, 547, 840, 600]]<|/det|>
+65. Cooney, E. C., Leander, B. S. & Keeling, P. J. Phylogenomics shows unique traits in Noctilucales are derived rather than ancestral. PNAS Nexus 1, pgac202 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[110, 613, 844, 666]]<|/det|>
+66. John, U. et al. An aerobic eukaryotic parasite with functional mitochondria that likely lacks a mitochondrial genome. Sci. Adv. 5, 1–12 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[110, 679, 857, 732]]<|/det|>
+67. Katoh, K. & Standley, D. M. MAFFT multiple sequence alignment software version 7: Improvements in performance and usability. Mol. Biol. Evol. 30, 772–780 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[110, 745, 870, 829]]<|/det|>
+68. Capella-Gutierrez, S., Silla-Martinez, J. M. & Gabaldón, T. trimAl: A tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25, 1972–1973 (2009).  
+
+<|ref|>text<|/ref|><|det|>[[110, 842, 863, 896]]<|/det|>
+69. Stamatakis, A. RAxML version 8: A tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 30, 1312–1313 (2014).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 84, 833, 168]]<|/det|>
+70. Roure, B., Rodriguez-Ezpeleta, N. & Philippe, H. SCaFoS: A tool for selection, concatenation and fusion of sequences for phylogenomics. BMC Evol. Biol. 7, 1–12 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[111, 181, 875, 270]]<|/det|>
+71. Nguyen, L. T., Schmidt, H. A., Von Haeseler, A. & Minh, B. Q. IQ-TREE: A fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol. Biol. Evol. 32, 268–274 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[111, 281, 850, 335]]<|/det|>
+72. Hoang, D. T., Chernomor, O., Von Haeseler, A., Minh, B. Q. & Vinh, L. S. UFBoot2: Improving the ultrafast bootstrap approximation. Mol. Biol. Evol. 35, 518–522 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[111, 347, 877, 401]]<|/det|>
+73. Lartillot, N., Lepage, T. & Blanquart, S. PhyloBayes 3: A Bayesian software package for phylogenetic reconstruction and molecular dating. Bioinformatics 25, 2286–2288 (2009).  
+
+<|ref|>text<|/ref|><|det|>[[111, 412, 880, 467]]<|/det|>
+74. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (2021).  
+
+<|ref|>text<|/ref|><|det|>[[111, 479, 806, 567]]<|/det|>
+75. Yu, G., Smith, D. K., Zhu, H., Guan, Y. & Lam, T. T. GGtree: an R package for visualization and annotation of phylogenetic trees with their covariates and other associated data. Methods Ecol Evol 8, 28–36 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[111, 579, 852, 666]]<|/det|>
+76. Wang, L.-G. et al. Trecio: An R Package for Phylogenetic Tree Input and Output with Richly Annotated and Associated Data. Molecular Biology and Evolution 37, 599–603 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[111, 679, 627, 732]]<|/det|>
+77. Seemann T. barrnap 0.9: rapid ribosomal RNA prediction. https://github.com/tseemann/barrnap  
+
+<|ref|>text<|/ref|><|det|>[[111, 744, 868, 833]]<|/det|>
+78. Kalyaanamoothry, S., Minh, B. Q., Wong, T. K. F., Von Haeseler, A. & Jermiin, L. S. ModelFinder: fast model selection for accurate phylogenetic estimates. Nat. Methods 14, 587–589 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[111, 845, 808, 900]]<|/det|>
+79. Fu, L., Niu, B., Zhu, Z., Wu, S. & Li, W. CD-HIT: Accelerated for clustering the nextgeneration sequencing data. Bioinformatics 28, 3150–3152 (2012).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[110, 85, 808, 135]]<|/det|>
+80. Price, M. N., Dehal, P. S. & Arkin, A. P. FastTree 2 - Approximately maximum-likelihood trees for large alignments. PLoS One 5, (2010). 
+
+<|ref|>text<|/ref|><|det|>[[110, 152, 864, 202]]<|/det|>
+81. Whelan, S., Irisarri, I. & Burki, F. PREQUAL: Detecting non-homologous characters in sets of unaligned homologous sequences. Bioinformatics 34, 3929–3930 (2018). 
+
+<|ref|>text<|/ref|><|det|>[[110, 219, 850, 299]]<|/det|>
+82. Ali, R. H., Bogusz, M., Whelan, S. & Tamura, K. Identifying Clusters of High Confidence Homologies in Multiple Sequence Alignments. Mol. Biol. Evol. 36, 2340–2351 (2019). 
+
+<|ref|>text<|/ref|><|det|>[[110, 317, 811, 367]]<|/det|>
+83. Larsson, A. AliView: a fast and lightweight alignment viewer and editor for large datasets. Bioinformatics 30, 3276–3278 (2014). 
+
+<|ref|>text<|/ref|><|det|>[[110, 384, 853, 434]]<|/det|>
+84. Dyrløv Bendtsen, J., Nielsen, H., Von Heijne, G. & Brunak, S. Improved prediction of signal peptides: SignalP 3.0. Journal of Molecular Biology 340, 783–795 (2004). 
+
+<|ref|>text<|/ref|><|det|>[[110, 451, 854, 530]]<|/det|>
+85. Krogh, A., Larsson, B., Von Heijne, G. & Sonnhammer, E. L. L. Predicting transmembrane protein topology with a hidden markov model: application to complete genomes11Edited by F. Cohen. Journal of Molecular Biology 305, 567–580 (2001). 
+
+<|ref|>text<|/ref|><|det|>[[110, 548, 875, 631]]<|/det|>
+86. Albanese, D., Fontana, P., De Filippo, C., Cavalieri, D. & Donati, C. MICCA: A complete and accurate software for taxonomic profiling of metagenomic data. Sci. Rep. 5, 1–7 (2015). 
+
+<|ref|>text<|/ref|><|det|>[[110, 648, 831, 698]]<|/det|>
+87. Bolyen, E. et al. Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nat Biotechnol 37, 852–857 (2019). 
+
+<|ref|>text<|/ref|><|det|>[[110, 715, 870, 764]]<|/det|>
+88. Callahan, B. J. et al. DADA2: High-resolution sample inference from Illumina amplicon data. Nat Methods 13, 581–583 (2016). 
+
+<|ref|>text<|/ref|><|det|>[[110, 782, 876, 864]]<|/det|>
+89. Massicotte P, South A. rnaturalearth: World Map Data from Natural Earth. R package version0.3.2.9000, https://docs.ropensci.org/rnaturalearth/https://github.com/ropensci/rnaturalearth (2023).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 860, 365]]<|/det|>
+Acknowledgements We thank A.P. Mylnikov and T.G. Simdyanov for help with sample collection, fixation, and interpretation of transmission electron microscopy images. This research was supported by grants from the Gordon and Betty Moore Foundation (to P.J.K., https://doi.org/10.37807/GBMF9201), the Natural Sciences and Engineering Research Council of Canada (to P.J.K., Grant Number 2019- 03994), the Czech Academy of Sciences (to E.H., Grant Number LQ200962204) the Russian Foundation for Basic Research (to D.V.T., Grant Number 20- 04- 00583), Tyumen Oblast Government, as part of the West- Siberian Interregional Science and Education Center's project No. 89- DON (2) and carried out within the framework of State Assignment no. 121051100102- 2.  
+
+<|ref|>text<|/ref|><|det|>[[115, 411, 880, 692]]<|/det|>
+Author contributions C.C.H., E.H., D.V.T. and P.J.K. designed the study. D.V.T. isolated and cultured eleftherid cells. C.C.H. isolated Polykrikos/Deorella. V.K.L.J.- R and N.O. isolated and cultured psammosid cells. C.C.H., D.V.T. and E.H. generated material for sequencing. D.V.T. performed microscopy experiments. C.C.H., E.H., E.C.C. and N.A.T.I. performed phylogenomic analyses. C.C.H. and E.C.C. performed phylogenetic analysis of the SSU rRNA genes. V.K.L.J.- R. performed the environmental distribution and abundance analysis. E.H. and V.K.L.J.- R. performed transcriptomic analyses and phylogenetic analysis of plastid and other proteins. C.C.H., E.H. and P.J.K. wrote the manuscript with input from all authors.  
+
+<|ref|>text<|/ref|><|det|>[[117, 740, 650, 759]]<|/det|>
+Competing interests The authors declare no competing interests.  
+
+<|ref|>text<|/ref|><|det|>[[117, 805, 320, 822]]<|/det|>
+Additional information  
+
+<|ref|>text<|/ref|><|det|>[[117, 838, 575, 856]]<|/det|>
+Supplementary information is available for this paper.  
+
+<|ref|>text<|/ref|><|det|>[[117, 871, 865, 889]]<|/det|>
+Correspondence and requests for materials should be addressed to C.C.H., E.H. or P.J.K.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[142, 135, 853, 377]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[57, 384, 90, 399]]<|/det|>
+870  
+
+<|ref|>text<|/ref|><|det|>[[58, 412, 873, 562]]<|/det|>
+Extended Data Fig. 1: Global distribution and abundance of eleftherids. The global distribution of eleftherids within sequenced transcriptomes (squares) and environmental SSU rRNA amplicon studies (circles). All eleftherid reads were obtained from marine sediment with the exception of two SSU rRNA reads from the Coral Sea study PRJNA369575 which were discovered in the water column at 798m depth.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[144, 100, 625, 585]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[63, 595, 93, 610]]<|/det|>
+877  
+
+<|ref|>text<|/ref|><|det|>[[65, 624, 876, 909]]<|/det|>
+Extended Data Fig. 2 Cell structure of Eleftheros karadeniz, visualized by transmission electron microscopy. Related to Fig. 1. a, Longitudinal section showing arrangement of basal bodies and transitional zone of flagella with transverse plates. b, Perinuclear space with mastigonemes, tubular in cross-sections (arrows). c, Longitudinal sections and arrangement of trichocysts. d, square cross-sections of trichocysts. e, cross-striated filaments of discharged trichocysts. f, mitochondrion with tubular cristae (arrows). g, storage compounds. ax, axosome; bb, basal bodies; f, flagella; fv, food vacuole; m, mitochondrion; n, nucleus; ps, perinuclear space; sc, storage compounds; t, trichocyst; tp, transverse plate. Scale bars, 1 μm (a, c), 0,2 μm (b, d, e) and 0,5 μm (f, g).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[155, 90, 825, 714]]<|/det|>
+ 
+
+<|ref|>text<|/ref|><|det|>[[57, 720, 88, 736]]<|/det|>
+887 
+
+<|ref|>text<|/ref|><|det|>[[57, 747, 830, 767]]<|/det|>
+Extended Data Fig. 3 Bayesian phylogenomic analysis. Related to Fig. 2. Consensus 
+
+<|ref|>text<|/ref|><|det|>[[57, 774, 875, 791]]<|/det|>
+Bayesian tree based on four independent chains (CAT + GTR + G4), places the eleftherids as 
+
+<|ref|>text<|/ref|><|det|>[[57, 803, 866, 822]]<|/det|>
+sister to the Syndinales (now MALV II and IV). MALV I (now the Ichthyodinales) groups 
+
+<|ref|>text<|/ref|><|det|>[[57, 833, 866, 852]]<|/det|>
+sister to Oxyrrhis marina, while the psammosids occupy a basal position relative to MALVs 
+
+<|ref|>text<|/ref|><|det|>[[57, 864, 812, 883]]<|/det|>
+and core dinoflagellates. Black dots denote full statistical support (Bayesian posterior 
+
+<|ref|>text<|/ref|><|det|>[[57, 895, 598, 914]]<|/det|>
+probability = 1.0), values are shown for support below 1.0.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 87, 415, 775]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[115, 797, 816, 816]]<|/det|>
+Extended Data Fig. 4 SSU rRNA gene phylogeny including short read sequences.  
+
+<|ref|>text<|/ref|><|det|>[[115, 828, 880, 912]]<|/det|>
+Maximum likelihood analysis (GTR+F+R6) of MALV related sequences. Black dots at nodes represent full statistical support (UltraFast bootstrap \(= 100\%\) ). Grey dots represent support above \(90\%\) . Support values are shown for support below \(90\%\) . Clades of interest coloured
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[57, 83, 875, 166]]<|/det|>
+899 with proposed taxonomic groups: purple, Ichthyodinales (MALV I); green, Syndinales; red, 900 eleftherids; and blue, psammosids. 901
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 95, 861, 412]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[45, 425, 878, 655]]<|/det|>
+Extended Data Fig. 5 All topologies tested with approximately unbiased (AU) test. Maximum likelihood analysis of nine alternative topologies. Associated \(p\) values show two supported topologies (1 and 3) reflecting Fig 2a and psammosids branching sister to Oxyrrhis and Deorella. All topologies showing monophyly of MALVs (7,8,9) were rejected ( \(p\) value \(< 0.05\) ). Red tip labels \(=\) eleftherids, green \(=\) Syndinales (MALV II and IV), purple \(=\) Ichthyodinales (MALV I) and blue \(=\) psammosids.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[128, 88, 530, 604]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[118, 625, 872, 675]]<|/det|>
+<center>Extended Data Fig. 6: Plastid metabolic pathways in MALV-related lineages. Related to Fig. 3. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 690, 880, 907]]<|/det|>
+Presence/absence table of plastid metabolic pathways in core dinoflagellates, eleftherids, MALVs, Oxyrrhis and Perkinsus including presence /absence of N- terminal extensions or incomplete N- termini. The pathways presented, from top to bottom, are: isoprenoid, methylerythritol phosphate (MEP) pathway for isoprenoid biosynthesis; FASII, type 2 fatty acid biosynthesis pathway; FeS cluster, plastidial Fe- S cluster biosynthesis pathway including a ferrodoxin system; heme, heme biosynthesis. \*, eleftherid sequences cluster with bacteria; m, mitochondrial; p/c, plastidial/cytosolic clade.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[117, 85, 500, 760]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[60, 772, 875, 891]]<|/det|>
+Extended Data Fig. 7: Maximum likelihood phylogeny of HemD. The scale bar and the number beneath it indicate the estimated number of substitutions per site, above the scale bar the model for tree reconstruction is indicated. Node numbers represent ultrafast bootstrap support values of \(>70\%\) , black dots indicate support values of \(> = 95\%\) . Eukaryotic groups are
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 82, 870, 270]]<|/det|>
+925 indicated by colored taxon names: green, Viridiplantae; dark red, rhodophytes; grey, 926 glaucophytes; turquoise, cryptophytes; pink, stramenopiles; dark blue, haptophytes; yellow, 927 Dinozoa; light green, eleftherids. Black taxa/clades outlined in black are prokaryotic. 928 Annotated orthologs in model species are indicated by red taxon name and protein identifier. 929 For species represented by more than one strain or taxa identified on genus level only, the 930 strain information is provided where available.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[155, 90, 560, 760]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[57, 780, 880, 900]]<|/det|>
+Extended Data Fig. 8: Maximum likelihood phylogeny of DapL. The scale bar and the number beneath it indicate the estimated number of substitutions per site, above the scale bar the model for tree reconstruction is indicated. Node numbers represent ultrafast bootstrap support values of \(>70\%\) , black dots indicate support values of \(> = 95\%\) . Eukaryotic groups are
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[60, 83, 870, 270]]<|/det|>
+937 indicated by colored taxon names: green, Viridiplantae; dark red, rhodophytes; grey, glaucophytes; turquoise, cryptophytes; pink, stramenopiles; dark blue, haptophytes; yellow, Dinozoa; light green, eleftherids. Black taxa/clades outlined in black are prokaryotic. Annotated orthologs in model species are indicated by red taxon name and protein identifier. For species represented by more than one strain or taxa identified on genus level only, the strain information is provided where available.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 312, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 368, 258]]<|/det|>
+SupplementaryInformation.pdf  SIGuide.docx  SupplementaryTable1.xlsx  SupplementaryTable2.xlsx  SupplementaryVideo1.mp4
+
+<--- Page Split --->

preprint/preprint__48961379f14baa7d8ee3a9307154d8bf615c946407f087937d7589936c43fad7/images_list.json

@@ -0,0 +1,62 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Fig 1. Chip-based parallel data communication architecture for short-reach inter-DCls. a Schematics of the data transmission system. At the transmitter, a multi-wavelength laser source with an equal channel spacing is generated by a \\(\\mathrm{Si}_3\\mathrm{N}_4\\) Kerr MRR, and then individually intensity-modulated by integrated MRR modulator arrays. After that, the modulated multi-wavelength optical signal transmits through an SMF. At the receiver side, the parallel WDM signals are first processed by a cascaded-MRRs-based optical signal processor for dispersion compensation, and then demultiplexed by MRR filter arrays and received by photodetectors separately. b Microscope image of the \\(\\mathrm{Si}_3\\mathrm{N}_4\\) micro-resonator for single-soliton generation. c Microscope image of the silicon-MRRs-based dispersion compensator. The numbers label the self-defined sequence of each ring, and the input and output ports are labeled. d Enlarged view of a silicon MRR in the dispersion compensator. e Schematics of the silicon MRR.",
+    "footnote": [],
+    "bbox": [
+      [
+        153,
+        88,
+        840,
+        355
+      ]
+    ],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2. Characterizations of the dispersion compensator and data transmission using a CW-laser. a, b Measured (a) transmission and (b) group delay responses with and without thermal crosstalk compensation for the dispersion compensation of 40-km-long SMF. TCC: thermal crosstalk compensation. c Measured transmission spectra for dispersion compensation of 10-km-, 20-km-, 30-km-, and 40-km-long SMFs in the wavelength range of 1540 nm to 1580 nm. d Enlarged view of the transmission spectra at the wavelength around 1558.7 nm. The operation bandwidth is calibrated to 32 GHz. e Measured group delay responses in the 40-nm wavelength span. f Enlarged view of the group delay responses at the wavelength of about 1558.7 nm. g Extracted insertion loss of each channel for dispersion compensation of different lengths of SMFs. The insertion loss is obtained at the center of the operation bandwidth. h Averaged dispersion of each channel for the 4 kinds",
+    "footnote": [],
+    "bbox": [
+      [
+        153,
+        87,
+        844,
+        736
+      ]
+    ],
+    "page_idx": 7
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Fig. 3. 80-Gbit/s PAM4-based WDM data transmission with microcombs. a Schematic experimental setup of a single-soliton microcomb driven parallel signal transmission system. PC: polarization controller; EDFA: erbium-doped fiber amplifier; WS: waveshaper; EA: electrical amplifier; AWG: arbitrary waveform generator; RTO: real-time oscilloscope; RRC: root-raised cosine; FFE: feed-forward equalization. b Measured spectrum of the generated single-soliton microcomb, which shows a sech²-like spectral shape. The utilized 15 comb lines are indicated by a dashed box. c Measured spectrum of the chosen-out 15 comb lines after the waveshaper. d Measured spectrum of the 15 comb lines after the 40-Gbaud PAM4 modulation. e Measured eye diagrams of the \\(4^{\\text{th}}\\) , \\(6^{\\text{th}}\\) , \\(8^{\\text{th}}\\) , \\(10^{\\text{th}}\\) , and \\(12^{\\text{th}}\\) channels after the dispersion compensation. f Measured BERs for the 15 WDM channels under the BTB transmission and the 20-km-long SMF transmission with dispersion compensation. All BERs are below the \\(7\\%\\) HD-FEC threshold after dispersion compensation. DC: dispersion compensation. g BERs versus received optical power for the 20-km-long SMF transmissions of the \\(3^{\\text{rd}}\\) , \\(6^{\\text{th}}\\) , and \\(9^{\\text{th}}\\) channels with dispersion compensation using the CW-laser and the microcomb as the light source, respectively.",
+    "footnote": [],
+    "bbox": [
+      [
+        163,
+        88,
+        840,
+        568
+      ]
+    ],
+    "page_idx": 9
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Fig. 4. 112-Gbit/s DMT-based WDM transmission with microcombs. a Measured signal-to-noise-ratio (SNR) response of the transmission system after the DMT training under the BTB transmission. b Bit allocations for DMT transmission within 32-GHz bandwidth. c Measured spectrum of the 15 comb lines after 112-Gbit/s DMT modulation. d Measured total BERs for the \\(8^{\\text{th}}\\) channel with and without the dispersion compensation. Inset: The enlarged view of BER for transmission with dispersion compensation. DC: dispersion compensation. e BERs for WDM transmissions with dispersion compensation. All BERs are within the \\(20\\%\\) SD-FEC threshold after dispersion compensation. f Measured 16-QAM constellations for the \\(4^{\\text{th}}\\) , \\(6^{\\text{th}}\\) , \\(8^{\\text{th}}\\) , \\(10^{\\text{th}}\\) , and \\(12^{\\text{th}}\\) comb line channels at the \\(7^{\\text{th}}\\) subcarrier.",
+    "footnote": [],
+    "bbox": [
+      [
+        153,
+        90,
+        844,
+        547
+      ]
+    ],
+    "page_idx": 10
+  }
+]

preprint/preprint__48961379f14baa7d8ee3a9307154d8bf615c946407f087937d7589936c43fad7/preprint__48961379f14baa7d8ee3a9307154d8bf615c946407f087937d7589936c43fad7.mmd

@@ -0,0 +1,227 @@
+
+Parallel Wavelength-Division-Multiplexed Signal Transmission and Dispersion Compensation Enabled by Soliton Microcombs and Microrings  
+
+Liangjun Lu luliangjun@sjtu.edu.cn  
+
+Shanghai Jiao Tong University Yuanbin Liu  
+
+Shanghai Jiao Tong University  
+
+Hongyi Zhang Shanghai Jiao Tong University  
+
+Jiacheng Liu Shanghai Jiao Tong University  
+
+Jiangbing Du Shanghai Jiao Tong University https://orcid.org/0000- 0002- 6333- 824X  
+
+Yu Li Shanghai Jiao Tong University  
+
+Zuyuan He Shanghai Jiao Tong University  
+
+Jianping Chen Shanghai Jiaotong University  
+
+Linjie Zhou Shanghai Jiao Tong University https://orcid.org/0000- 0002- 2792- 2959  
+
+Andrew Poon Hong Kong University of Science and Technology https://orcid.org/0000- 0002- 5222- 8184  
+
+## Article  
+
+# Keywords:  
+
+Posted Date: September 18th, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 3282850/v1
+
+<--- Page Split --->
+
+License: © ① This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on April 29th, 2024. See the published version at https://doi.org/10.1038/s41467-024-47904-2.
+
+<--- Page Split --->
+
+# Parallel Wavelength-Division-Multiplexed Signal Transmission and Dispersion Compensation Enabled by Soliton Microcombs and Microrings  
+
+Yuanbin Liu \(^{1,4}\) , Hongyi Zhang \(^{1,4}\) , Jiacheng Liu \(^{1,4}\) , Liangjun Lu \(^{1,2,*}\) , Jiangbing Du \(^{1,*}\) , Yu Li \(^{1,2}\) , Zuyuan He \(^{1}\) , Jianping Chen \(^{1,2}\) , Linjie Zhou \(^{1,2}\) , and Andrew W. Poon \(^{3}\) \(^{1}\) State Key Laboratory of Advanced Optical Communication Systems and Networks, Shanghai Key Lab of Navigation and Location Services, Department of Electronic Engineering, Shanghai Jiao Tong University, Shanghai 200240, China \(^{2}\) SJTU-Pinghu Institute of Intelligent Optoelectronics, Pinghu 314200, China \(^{3}\) Photonic Device Laboratory, Department of Electronic and Computer Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong \(^{4}\) Corresponding authors' e-mail address: luliangjun@sjtuedu.cn, dujiangbing@sjtuedu.cn \(^{5}\) These authors contributed equally to this work.  
+
+## Abstract  
+
+The proliferation of computation- intensive technologies has led to a significant rise in the number of datacenters, posing challenges for high- speed and power- efficient datacenter interconnects (DCIs). Although inter- DCIs based on intensity modulation and direct detection (IM- DD) along with wavelength- division multiplexing (WDM) technologies exhibit cost- effective, power- efficient, and large- capacity properties, the requirement of multiple laser sources leads to high costs and limited scalability. Moreover, careful considerations must be given to the chromatic dispersion in the C- band as it restricts the transmission length of the optical signals. Electronic and optical approaches based on digital signal processing algorithms or dispersion- compensating fibers suffer from either a high power consumption or a lack of full reconfigurability. In this study, we present an original scalable on- chip parallel IM- DD data transmission system enabled by a single- soliton Kerr microcomb and a reconfigurable microring resonator (MRR)- based dispersion compensator. The highly compact MRR- based Kerr microcomb and dispersion compensator are intrinsically compatible with the parallel processing nature of the WDM link. Besides, the reconfigurability of the dispersion compensator shows the validation for various data- rate transmissions over multiple single- mode fiber lengths of up to 40 km, with a power consumption of below 160 mW, regardless of the modulation format or of the number of transmission channels utilized. Through our experimental validation, we demonstrate an aggregate bit rate of 1.68 Tbit/s over a 20- km- long single- mode fiber using 15 independent wavelength channels spaced at 100 GHz. Our approach holds significant promise for achieving data communications at a scale exceeding 10 terabits, making it highly valuable for future hyper- scale DCIs.  
+
+## Introduction  
+
+According to the International Data Corporation (IDC), the exponential growth of digital data
+
+<--- Page Split --->
+
+generation is projected to reach 175 zettabytes by \(2025^{1}\) due to the surgen growth of computation- intensive technologies such as artificial intelligence, Internet of Things (IoT), and autonomous vehicles. The total data capacity grows 80 times from 640 GHz in 2010 to 51.2 THz in 2022 for data switching in datacenter interconnects (DCls) while the total power consumption grows 22 times \(^{2}\) . As a result, there is a critical need for energy- efficient and cost- effective DCls that can accommodate a substantial capacity. For short- reach DCls spanning distances of less than \(40 \text{km}\) , the utilization of intensity modulation and direct detection (IM- DD) has emerged as a practical and preferred solution \(^{3}\) . This choice is motivated by the inherent advantages IM- DD offers, including cost- effectiveness, a low power consumption, and a compact physical footprint. Due to the lower data rate of the IM- DD system for a single wavelength channel than the coherent scheme, wavelength- division multiplexing (WDM) technology is commonly employed to economically enhance the data capacity. However, multiple laser sources are required to provide multiple operation wavelengths in the WDM systems, which leads to a higher cost. Moreover, in IM- DD systems operating within the C- band, the transmission distance is primarily limited by chromatic dispersion (CD), necessitating effective mitigation strategies \(^{6 - 8}\) . The CD- induced frequency- selective fading (FSF) significantly impacts transmission performance \(^{9, 10}\) . To address the dispersion effect, various electronic and optical approaches have been employed, such as digital signal processing (DSP) \(^{11 - 13}\) and dispersion- compensating fibers (DCFs). However, these methods are burdened with limitations, particularly in terms of a high energy consumption and associated costs \(^{14}\) .  
+
+Chip- scaled soliton microcombs, known for their wide spectral range, low noise, and high repetition rate, have emerged as a promising light source for WDM applications \(^{15 - 24}\) . Recent advancements have demonstrated massively parallel data transmission using various types of microcomb sources, including bright single- solitons \(^{25}\) , perfect soliton crystals \(^{26}\) , and dark solitons \(^{27}\) . For example, the utilization of two interleaved single- soliton microcombs in a silicon nitride (Si \(_{3}\) N \(_{4}\) ) microresonator enabled coherent data transmission with 179 carriers in the C- and L- bands, achieving a total bit rate of 55 Tb/s \(^{25}\) . Similarly, the soliton crystal with defects, generated in a doped silica glass microresonator, facilitated a line rate of 40.1 Tb/s and a spectral efficiency of 10.2 b/s/Hz \(^{26}\) . However, it is critical to note that the majority of WDM transmission systems utilizing soliton microcombs rely on expensive coherent schemes, necessitating costly transmitters and receivers \(^{25 - 30}\) . Recently, massively parallel optical interconnects based on a Si \(_{3}\) N \(_{4}\) Kerr dark soliton comb source and a silicon microdisk modulator array with a data rate of 512 Gb/s for IM- DD transmission have been demonstrated \(^{31}\) . In addition, a parallel optical data link driven by an AlGaAs Kerr microcomb has achieved a data rate of 2 Tbit/s in the IM- DD system \(^{32}\) . However, we note that these transmission techniques are limited in term of transmission lengths and are primarily suitable for intra- DCI applications.  
+
+In recent years, the adoption of silicon photonics facilitated the integration of various silicon- based photonic devices, including chirped Bragg gratings \(^{33 - 37}\) , Mach- Zehnder interferometers (MZIs) \(^{38, 39}\) , and microring resonators (MRRs) \(^{40 - 46}\) , to effectively mitigate the impact of CD in single- mode fibers (SMFs). These integrated devices offer several advantages, including a compact form factor, cost efficiency, a low power consumption, and a high reconfigurability. While imbalanced MZIs and MRRs demonstrate desirable characteristics for parallel signal processing due to their periodic spectral properties, their potential in WDM systems is currently limited. As a result, existing studies have reported only a modest maximum data transmission rate of 224 Gbit/s \(^{44}\) .  
+
+In this paper, we present an original approach for short- reach inter- DCls based on an IM- DD
+
+<--- Page Split --->
+
+optical communications architecture. Our proposed system leverages a single- soliton microcomb source generated by a \(\mathrm{Si}_3\mathrm{N}_4\) MRR as an WDM light source. Furthermore, we introduce a fully reconfigurable optical dispersion compensator based on cascaded silicon MRRs for parallel dispersion compensation. This WDM- assisted IM- DD scheme offers numerous advantages, including a high scalability, compactness, energy efficiency, and cost- effectiveness. To evaluate the performance of our proposed system, we conducted experiments demonstrating parallel signal transmission using up to 15 wavelength channels within the C- band. By employing pulse- amplitude four- level (PAM4) modulation signals at a rate of 80 Gbit/s, we achieved an aggregate data rate of 1.2 Tbit/s. Additionally, by employing discrete multi- tone (DMT) modulation signals at a rate of 112 Gbit/s, we achieved an aggregate data rate of 1.68 Tbit/s. These transmission rates were achieved over a 20- km- length SMF. Importantly, we assessed the power consumption of the dispersion compensator to be consistently below 160 mW, regardless of the modulation format or the number of transmission channels utilized. By aligning the free spectral range (FSR) of the Kerr MRR and the dispersion compensator and further increasing the number of wavelength channels, we expect the transmission capacity to exceed 10 Tbit/s. This capability holds significant promise for meeting the demands of future hyperscale DCIs.  
+
+## Results  
+
+## Parallel IM-DD data transmission and dispersion compensation architecture  
+
+Figure 1a shows the schematics of the integrated parallel data communications system for IM- DD applications. At the transmitter, a bright single- soliton microcomb with a smooth spectrum, which is generated by a \(\mathrm{Si}_3\mathrm{N}_4\) microresonator, is utilized as the multi- wavelength laser source with evenly distributed and low- noise frequency tones. The comb source is then demultiplexed, independently encoded with electrical data, and multiplexed by silicon microring modulator (MRM) arrays working at several individual wavelength channels. The modulated data are transmitted through an SMF before being sent to the receiver. The MRMs have the advantages of a compact size and of a high modulation bandwidth, which have been demonstrated with a modulation bandwidth of 110 GHz<sup>47</sup>. To increase the channels and to reduce the crosstalk, de-/multiplexers based on MRRs<sup>48- 50</sup>, asymmetric MZIs<sup>51, 52</sup>, and MRR- coupled MZIs<sup>53, 54</sup> can be inserted before and after the silicon MRR modulator arrays. At the receiver end, we employ a silicon- based dispersion compensator utilizing cascaded MRRs to simultaneously compensate for the dispersion induced by SMF transmission across all the wavelength channels. The FSR of the cascaded MRRs is designed to match the wavelength spacing of the frequency combs. Moreover, the resonant wavelengths and the coupling coefficients of each MRR can be independently adjusted, offering a fully reconfigurable configuration that caters to various transmission ranges and operation bandwidths. Following the compensation process, the signal is then separated into distinct wavelength channels with MRR- based demultiplexers for photodetection. This comprehensive system enables parallel data transmission and dispersion compensation through the integration of photonic devices, featuring a simple arrangement and remarkable scalability. Consequently, our approach holds great promise for the development of fully integrated photonic circuits, especially for high- capacity inter- DCIs.  
+
+In this proof- of- concept study, we used a \(\mathrm{Si}_3\mathrm{N}_4\) microresonator with a bending radius of \(232\mu \mathrm{m}\) and an average quality (Q) factor of \(7\times 10^{5}\) for single- soliton generation. Figure 1b shows the microscope image of the fabricated \(\mathrm{Si}_3\mathrm{N}_4\) microresonator. The waveguide cross- section is \(1800\times 800 \mathrm{nm}^2\) and the measured second- order group- velocity dispersion \((\mathrm{D}_2 / 2\pi)\) is \(601.2 \mathrm{kHz}^{55}\) .
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Fig 1. Chip-based parallel data communication architecture for short-reach inter-DCls. a Schematics of the data transmission system. At the transmitter, a multi-wavelength laser source with an equal channel spacing is generated by a \(\mathrm{Si}_3\mathrm{N}_4\) Kerr MRR, and then individually intensity-modulated by integrated MRR modulator arrays. After that, the modulated multi-wavelength optical signal transmits through an SMF. At the receiver side, the parallel WDM signals are first processed by a cascaded-MRRs-based optical signal processor for dispersion compensation, and then demultiplexed by MRR filter arrays and received by photodetectors separately. b Microscope image of the \(\mathrm{Si}_3\mathrm{N}_4\) micro-resonator for single-soliton generation. c Microscope image of the silicon-MRRs-based dispersion compensator. The numbers label the self-defined sequence of each ring, and the input and output ports are labeled. d Enlarged view of a silicon MRR in the dispersion compensator. e Schematics of the silicon MRR. </center>  
+
+The FSR is about 97.7 GHz. We utilized a program- controlled scheme based on thermal auxiliary compensation to generate and to stabilize the microcomb in the whole experiments<sup>56</sup>. See Supplementary Note 1 for the detailed program- controlled scheme for the single- soliton generation and stabilization. At the receiver side, the fully reconfigurable silicon dispersion compensator comprises 8 identical MRRs. Figure 1c shows the microscope image of the device. Figures 1d and 1e demonstrate the enlarged microscope image and the schematics of one MRR, respectively. The FSR of these MRRs is around 99.5 GHz for dense WDM (DWDM) scenarios in the C- band. The coupling region of each MRR is replaced by an MZI- based tunable coupler. Both the ring and the coupling region are integrated with a titanium micro- heater for independent thermal adjustment of the resonant wavelength and of the coupling coefficient. A pair of grating couplers are utilized to couple the optical signal in and out of this chip. The footprint of the dispersion compensator is \(2 \mathrm{~mm} \times 1 \mathrm{~mm}\). See Methods for the detailed design, fabrication, and packaging of the \(\mathrm{Si}_3\mathrm{N}_4\) Kerr comb and of the dispersion compensator.  
+
+## Characterizations of the dispersion compensator and data transmission using a CW laser  
+
+To achieve reconfiguration of the dispersion compensator for varying lengths of SMF, we devised an optimization algorithm that combines the transfer matrix method with the sequential quadratic programming (SQP) algorithm. This algorithm allowed us to determine the target parameters for the MRR- based dispersion compensator. We successfully demonstrated dispersion compensation for
+
+<--- Page Split --->
+
+SMF transmission up to \(40km\) within a bandwidth of \(32GHz\) . See Supplementary Note 2 for a detailed description of the optimization methodology and of the obtained results. Then, to facilitate an efficient dispersion compensation for diverse dispersion values, the optimized resonant wavelengths and coupling coefficients of all the MRRs for different dispersion conditions were documented in a look- up table. Accounting for the inherent fabrication deviations in silicon photonic devices, we deem it necessary the calibration of the random states for each MRR. Furthermore, we investigated the impact of thermal crosstalk on the tuning of the chip to their optimal configurations, considering the intra- and inter- thermal crosstalk within and between MRRs, respectively. By characterizing the tuning efficiencies influenced by the intra- and inter- thermal crosstalk and adjusting the applied voltages of MRRs correspondingly according to these efficiencies, we successfully compensated for the thermal crosstalk- induced resonant wavelength shifts.  
+
+Figures 2a and 2b show the measured transmission and group delay responses of dispersion compensation for a \(40 - km\) - long SMF with and without thermal crosstalk compensation (TCC). The implementation of TCC significantly alleviated these shifts, leading to a smoother and more linearly varied response within the operation bandwidth. For detailed information on the calibration process, see Supplementary Note 3. We note that the proposed TCC method is applicable to other photonic devices comprising multiple thermal phase shifters, which can significantly simplify the adjustment of the working states.  
+
+Figures 2c and 2e show the measured transmission spectra and the corresponding group delay responses covering a wavelength span from \(1540nm\) to \(1580nm\) after the calibration. The MRRs were calibrated at the wavelength of \(1558.7nm\) with a \(32 - GHz\) operation bandwidth for \(10 - km\) to \(40 - km\) SMFs. The enlarged views at the wavelength around \(1558.7nm\) are shown in Figs. 2d and 2f. The transmission spectra include the coupling loss of the grating couplers. The grating coupler has a maximum loss non- uniformity of \(\sim 3.5dB\) over the \(40nm\) wavelength, which can be improved by using an edge coupler. The increased on- resonance transmission loss at the shorter wavelength side is due to the wavelength- dependent loss and the coupling coefficient of the MRRs. Figures 2g and 2h show the extracted insertion loss and the averaged dispersion of the dispersion compensator. The insertion loss is obtained at the center of the operation bandwidth of each channel, and the averaged dispersion is calculated as \(D_{avg} = (\tau_1 - \tau_2) / (\lambda_1 - \lambda_2)\) , where \(\lambda_1\) and \(\lambda_2\) are the shorter and longer wavelength ends across the operation bandwidth of each channel, \(\tau_1\) and \(\tau_2\) are the corresponding group delays, respectively. The insertion loss increases from \(\sim 8.4dB\) to \(\sim 14.8dB\) at the wavelength around \(1558.7nm\) for dispersion compensation of \(10 - km\) to \(40 - km\) SMFs, respectively. We can reduce the insertion loss by lowering the transmission loss of the silicon MRR waveguides. The averaged dispersion is \(- 169.5ps/nm\) , \(- 325.4ps/nm\) , \(- 506.6ps/nm\) , and \(- 682.9ps/nm\) for dispersion compensation of \(10 - km\) , \(20 - km\) , \(30 - km\) , and \(40 - km\) - long SMFs at the same wavelength. The group delay increases slightly with the wavelength, which we attribute to the coupling coefficient variations of MRRs across wavelengths. The group delay degradation around \(1540nm\) is limited by the dynamic range of our measurement equipment, the Photonic Dispersion and Loss Analyzer (PDLA, Agilent 86038B).  
+
+Next, we conducted a 64 Gbit/s PAM4 signal transmission experiment using a continuous- wave (CW) laser as the light source to demonstrate the reconfigurability and the wide- band signal processing capability of our dispersion compensator in a wide wavelength span. We utilized an offline digital signal processing (DSP) algorithm with a time- domain feed- forward equalization (FFE) to recover the received signal. We note that no dispersion compensation algorithm was used in the
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Fig. 2. Characterizations of the dispersion compensator and data transmission using a CW-laser. a, b Measured (a) transmission and (b) group delay responses with and without thermal crosstalk compensation for the dispersion compensation of 40-km-long SMF. TCC: thermal crosstalk compensation. c Measured transmission spectra for dispersion compensation of 10-km-, 20-km-, 30-km-, and 40-km-long SMFs in the wavelength range of 1540 nm to 1580 nm. d Enlarged view of the transmission spectra at the wavelength around 1558.7 nm. The operation bandwidth is calibrated to 32 GHz. e Measured group delay responses in the 40-nm wavelength span. f Enlarged view of the group delay responses at the wavelength of about 1558.7 nm. g Extracted insertion loss of each channel for dispersion compensation of different lengths of SMFs. The insertion loss is obtained at the center of the operation bandwidth. h Averaged dispersion of each channel for the 4 kinds </center>
+
+<--- Page Split --->
+
+of dispersion compensation. i Measured PAM4 eye diagrams for the data transmission using a CW- laser as the light source at the wavelength of around 1558.96 nm for the four lengths of SMFs with and without dispersion compensation. 64- Gbit/s PAM4 signal is transmitted through SMFs. DC: dispersion compensation. j Measured BERs for data transmission under BtB transmission and SMFs transmissions with and without dispersion compensation at 5 different wavelengths in the 30- nm wavelength span ranging from 1540 nm to 1570 nm. BERs are much higher than the 20% SD- FEC threshold \((2\times 10^{- 2})\) without the dispersion compensation, while they are all within the \(7\%\) HD- FEC threshold \((3.8\times 10^{- 3})\) after the dispersion compensation.  
+
+signal receiving end. See Supplementary Note 4 for the detailed experimental setup. We measured 5 different wavelengths distributed from 1540 nm to 1570 nm under the 10-, 20-, 30-, and 40- km- long SMFs with and without dispersion compensation. The wavelength span was limited by our tunable bandpass filter. Figure 2i shows the measured eye diagrams for the transmissions with and without dispersion compensation at the wavelength of 1558.96 nm under the four lengths of SMFs. In the absence of dispersion compensation, the eye diagrams exhibit significant blurring. In contrast, with our dispersion compensator, the eye diagrams are successfully restored. The rest of the measured eye diagrams at the other wavelengths are included in Supplementary Note 4. The bit- error ratios (BERs) were also measured and shown in Fig. 2j. The BERs without dispersion compensator are all far beyond the 20% soft- decision forward- error correction (SD- FEC) threshold while all of them are below the 7% hard- decision forward- error correction (HD- FEC) threshold after dispersion compensation, demonstrating that our dispersion compensator exhibits a fairly good continuous dispersion compensation tunability and is capable of parallelly processing the data of at least 37 channels over the 30- nm wavelength range. The slightly higher BERs at wavelengths around 1540 nm and 1570 nm are due to the grating coupler- induced loss non- uniformity. The successful data transmission at the wavelength around 1540 nm indicates the degradation of the averaged dispersion of the 40- km- long SMF shown in Fig. 2h is limited by our instrument.  
+
+## Parallel data transmission using a comb light source  
+
+We demonstrated parallel data transmission and dispersion compensation using the \(\mathrm{Si}_3\mathrm{N}_4\) soliton comb lines as the light source. In this demonstration, we utilized a 20- km- long SMF in the transmission, and we calibrated our dispersion compensator for a 50- GHz operation bandwidth. See Supplementary Note 3 for the measured transmission spectra and group delay responses. The 80- Gbit/s PAM4 and 112- Gbit/s discrete multi- tone (DMT) signals were transmitted through the SMF, respectively.  
+
+Figure 3a shows the experimental setup of the PAM4 signal transmission. The spectrum of the generated single- soliton microcombs is depicted in Fig. 3b. The 10- dB spectral bandwidth of the single soliton is 62.6 nm, with 80 comb lines ranging from 1527 nm to 1590 nm included. Here the 10- dB spectral bandwidth describes the optical wavelength span of the spectrum with the optical power of the comb lines decreased by 10 dB from the comb line with the maximum power. As there is a slight FSR discrepancy between the comb lines ( \(\sim 97.7\) GHz) and the dispersion compensators ( \(\sim 99.5\) GHz), we only selected 15 comb lines from 1555 nm to 1563 nm by a programmable optical filter for the following data transmission. In principle, by carefully designing the FSRs of both devices with a smaller or even no discrepancy, we can select more wavelength channels for parallel data transmission. We utilized a single modulator to simultaneously modulate all the wavelength channels for simplicity. We detail the experimental settings and the DSP flows in Methods. The optical spectra of the reshaped and of the modulated comb lines are depicted in Figs. 3c and 3d, respectively.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Fig. 3. 80-Gbit/s PAM4-based WDM data transmission with microcombs. a Schematic experimental setup of a single-soliton microcomb driven parallel signal transmission system. PC: polarization controller; EDFA: erbium-doped fiber amplifier; WS: waveshaper; EA: electrical amplifier; AWG: arbitrary waveform generator; RTO: real-time oscilloscope; RRC: root-raised cosine; FFE: feed-forward equalization. b Measured spectrum of the generated single-soliton microcomb, which shows a sech²-like spectral shape. The utilized 15 comb lines are indicated by a dashed box. c Measured spectrum of the chosen-out 15 comb lines after the waveshaper. d Measured spectrum of the 15 comb lines after the 40-Gbaud PAM4 modulation. e Measured eye diagrams of the \(4^{\text{th}}\) , \(6^{\text{th}}\) , \(8^{\text{th}}\) , \(10^{\text{th}}\) , and \(12^{\text{th}}\) channels after the dispersion compensation. f Measured BERs for the 15 WDM channels under the BTB transmission and the 20-km-long SMF transmission with dispersion compensation. All BERs are below the \(7\%\) HD-FEC threshold after dispersion compensation. DC: dispersion compensation. g BERs versus received optical power for the 20-km-long SMF transmissions of the \(3^{\text{rd}}\) , \(6^{\text{th}}\) , and \(9^{\text{th}}\) channels with dispersion compensation using the CW-laser and the microcomb as the light source, respectively. </center>  
+
+Figure 3e shows the measured eye diagrams of the \(4^{\text{th}}\) , \(6^{\text{th}}\) , \(8^{\text{th}}\) , \(10^{\text{th}}\) , and \(12^{\text{th}}\) channels after transmission and dispersion compensation. See Supplementary Note 5 for the eye diagrams of all the 15 comb lines. All the 15 wavelength channels have good and comparable eye diagrams for the back- to- back (BtB) and for the 20- km- long SMF transmission. Figure 3f shows the measured BERs after dispersion compensation. Although a minor FSR difference between the comb lines and the channels of the dispersion compensator leads to a slightly higher BER level around the left and right
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Fig. 4. 112-Gbit/s DMT-based WDM transmission with microcombs. a Measured signal-to-noise-ratio (SNR) response of the transmission system after the DMT training under the BTB transmission. b Bit allocations for DMT transmission within 32-GHz bandwidth. c Measured spectrum of the 15 comb lines after 112-Gbit/s DMT modulation. d Measured total BERs for the \(8^{\text{th}}\) channel with and without the dispersion compensation. Inset: The enlarged view of BER for transmission with dispersion compensation. DC: dispersion compensation. e BERs for WDM transmissions with dispersion compensation. All BERs are within the \(20\%\) SD-FEC threshold after dispersion compensation. f Measured 16-QAM constellations for the \(4^{\text{th}}\) , \(6^{\text{th}}\) , \(8^{\text{th}}\) , \(10^{\text{th}}\) , and \(12^{\text{th}}\) comb line channels at the \(7^{\text{th}}\) subcarrier. </center>  
+
+sides of the wavelength span, all the BERs are below the \(7\%\) HD- FEC threshold, enabling a transmitted bit rate of 1.2 Tbit/s in total (1.12 Tbit/s net rate after FEC overhead subtraction). The BERs under different received optical powers for the CW- laser and the microcombs after dispersion compensation are illustrated in Fig. 3g. The \(3^{\text{rd}}\) , \(6^{\text{th}}\) , and \(9^{\text{th}}\) channels were chosen in the experiments, and the wavelengths of the CW- laser were kept the same as the ones of the selected comb lines. A trivial deterioration is observed between the BERs of the microcomb and the CW- laser transmission, demonstrating a comparable performance between these two light sources. The power penalty is \(\sim 1\) dB at the \(7\%\) HD- FEC threshold.  
+
+To facilitate the next- generation data transmission rates, Common Electrical I/O (CEI)- 112G standard based on 112 Gbit/s data rate transmission has been established by Optical Internetworking Forum (OIF)57. To achieve a 112 Gbit/s data rate per lane, we implemented the DMT
+
+<--- Page Split --->
+
+modulation format to increase the total data capacity. Compared with the PAM4 modulation format, DMT signals use a higher- order modulation format and can achieve a higher spectral efficiency for the microcomb- based WDM transmission system. The experimental setup of the DMT transmission is the same as that for the PAM4 transmission except that the DSP procedure is different, and the sampling rate of the arbitrary waveform generator (AWG) is set to be 64 GSa/s because of the limitation of the maximum length of the data sequence. The bandwidth of the dispersion compensator is still 50 GHz, and the DMT signal was mapped to a 32- GHz bandwidth with 160 subcarriers. The details of the experimental setup are discussed in Methods. Before the data transmission, training of the DMT is essential to obtain the bit allocation of each tone. The training was carried out at the \(9^{\text{th}}\) channel under the BtB condition. The signal- to- noise- ratio (SNR) response and the bit allocation were obtained, which are shown in Figs. 4a and 4b. The maximum bit allocation is 4, corresponding to the modulation format of 16 quadratic amplitude modulation (16- QAM). Each sub- carrier was modulated with its own allocated modulation format. Then, we applied the identical bit allocation to all the 15 comb lines to ensure a uniform DMT setting.  
+
+Figure 4c illustrates the spectrum of the 112- Gbit/s DMT- modulated comb lines. Figure 4f depicts the measured 16- QAM constellations of the \(7^{\text{th}}\) sub- carrier for the \(4^{\text{th}}\) , \(6^{\text{th}}\) , \(8^{\text{th}}\) , \(10^{\text{th}}\) , and \(12^{\text{th}}\) channels. See Supplementary Note 6 for the constellations of all the 15 comb lines. Figure 4d shows the measured BERs for the \(8^{\text{th}}\) channel with and without dispersion compensation. The inset illustrates the enlarged view of the BER for transmission with dispersion compensation. At the frequency around 13 GHz and 23 GHz, the BERs reach as high as 0.49 and 0.47 before the dispersion compensation, respectively. We attribute this to the FSF of the 20- km SMF. See Supplementary Note 6 for the measured \(S_{21}\) response of the 20- km SMF. The BERs can be suppressed below 0.01 after dispersion compensation. The BERs of all the 15 channels after dispersion compensation are all below the \(20\%\) SD- FEC threshold, as shown in Fig. 4e. No FFE was introduced in the DMT transmission. In this case, an aggregate data rate of 1.68 Tbit/s (1.46 Tbit/s net rate) was achieved. The maximum power consumption of our dispersion compensator is as low as 160 mW. See Supplementary Note 6 for the detailed calculation of the bit rate and Supplementary Note 3 for the power consumption.  
+
+## Discussion  
+
+To our knowledge, we present the first demonstration of on- chip parallel signal transmission and dispersion compensation using a single- soliton microcomb and a dispersion compensator based on MRRs. Our optical dispersion compensator outperforms electronic dispersion compensation approaches based on DSP algorithms, which necessitate individual processing for each wavelength channel. In contrast, our optical dispersion compensator supports parallel dispersion compensation capabilities, irrespective of the modulation format and of the number of transmission channels. Compared with the dispersion compensators enabled by Bragg gratings and cascaded MZIs, the MRR- based dispersion compensator features a compact footprint and a flexible tuning capability. Notably, we achieved this parallel dispersion compensation with a low power consumption below 160 mW, underscoring its energy- efficient nature.  
+
+Furthermore, under the assumption of a matched FSR between the microcomb and the dispersion compensator, we can in principle utilize up to 51 channels from the microcomb for data transmission, which encompasses a wavelength span ranging from 1540 nm to 1580 nm. To achieve a total bit rate exceeding 10 Tbit/s, the spacing between adjacent comb lines can be further reduced to 50 GHz. It is worth noting that we can integrate various components such as a \(\mathrm{Si}_3\mathrm{N}_4\) microcomb,
+
+<--- Page Split --->
+
+a silicon MRR- based dispersion compensator, MRR modulators, photodetectors, and optical demultiplexers onto a single chip. By accomplishing this, we can realize a fully integrated optical communications system tailored specifically for short- reach DCIs.  
+
+In the experiment, the single- soliton microcomb in the microresonator with anomalous dispersion demonstrates a conversion energy efficiency of approximately \(1\%\) . The power of most of the comb lines is below - 20 dBm (considering the coupling and link loss). Investigations highlighting the effectiveness of utilizing a double- microring structure have been proposed with a high conversion energy efficiency of \(55\%\) for single soliton generation \(^{58}\) , resulting in increased power levels for the majority of the comb lines to - 10 dBm, which reduces the gain required by the erbium- doped fiber amplifier (EDFA). Consequently, this leads to improved noise characteristics and facilitates higher bit rates for each carrier. To further enhance the overall performance of the system, it is plausible to minimize the fiber- to- fiber loss associated with the dispersion compensator by introducing low- loss and broadband edge couplers \(^{59 - 61}\) . Additionally, MRRs made of widened waveguides can be incorporated to mitigate on- chip losses. By implementing these strategies, we believe a higher number of microcomb channels can be effectively utilized for data transmission.  
+
+## Conclusion  
+
+We have presented parallel data transmission and dispersion compensation utilizing a \(\mathrm{Si}_3\mathrm{N}_4\) single- soliton microcomb and a silicon MRR- based dispersion compensator. With the full reconfigurability of the MRRs, our dispersion compensator achieves a continuously tunable dispersion compensation of up to \(40\mathrm{km}\) SMFs within a 32- GHz bandwidth, and a dispersion compensation of \(20\mathrm{- km}\) - long SMF within a 50- GHz bandwidth. We proposed a method for calibration of the MRR- based dispersion compensator, which can eliminate the thermal crosstalk and is applicable to other integrated photonic circuits. Our experimentation with the \(\mathrm{Si}_3\mathrm{N}_4\) single- soliton microcomb revealed the utilization of 15 comb lines for parallel data transmission, resulting in an aggregate total bit rate of 1.2 Tbit/s and of 1.68 Tbit/s using 80- Gbit/s PAM4 and 112- Gbit/s DMT modulation signals for 20- km SMF transmission, respectively. These demonstrations in utilizing the microcomb and the silicon photonic circuits enable a higher data transmission capacity, potentially lower costs, and power- efficient solutions for parallel data processing, especially in WDM- assisted IM- DD systems. Our findings represent a significant advancement toward the pragmatic applications of cost- effective and practical short- reach DCIs.  
+
+## Methods  
+
+## Design, fabrication, and packaging of the silicon photonic dispersion compensator chip  
+
+The coupling region of each microring comprises an MZI- based tunable coupler, which comprises two 50:50 multimode interferometers (MMIs), two \(165 - \mu \mathrm{m}\) - long straight arms, and waveguide connections between them. The feedback region of each microring comprises a \(335 - \mu \mathrm{m}\) - long straight waveguide and two \(10 - \mu \mathrm{m}\) - radius arc bends. The dimensions of the waveguide cross- section are \(500\mathrm{nm}\) (width) \(\times 220\mathrm{nm}\) (height). Two \(160 - \mu \mathrm{m}\) - long and \(2 - \mu \mathrm{m}\) - wide titanium microheaters, with a resistance of about \(1800\Omega\) , are integrated \(2 - \mu \mathrm{m}\) above the center of the \(335 - \mu \mathrm{m}\) - long straight waveguide and of the lower arm of the MZI coupler. Deep trenches are positioned beside each micro- heater to suppress the thermal crosstalk within the MRR.  
+
+The device was fabricated on a silicon- on- insulator (SOI) platform with a 220- nm- thick silicon top layer and a \(2 - \mu \mathrm{m}\) buried oxide (BOX) using electron- beam lithography. All the devices except for
+
+<--- Page Split --->
+
+the grating couplers were fully etched to a depth of \(220 \text{nm}\) . A 70- nm partial etch was implemented on the grating coupler regions. The gold metal connections were deposited above the titanium microheaters for electrical contacts. The metal pads were wire- bonded to an external printed- circuit board (PCB). The chip and the PCB were placed on a metal shell, and a thermo- electric cooler (TEC) was placed beneath the chip to control the on- chip temperature.  
+
+## Design, fabrication, and packaging of the \(\text{Si}_3\text{N}_4\) microcomb chip  
+
+The \(\text{Si}_3\text{N}_4\) microresonator has waveguide cross- sectional dimensions of \(1800 \text{nm}\) (width) \(\times 800 \text{nm}\) (height), with the fundamental TE mode exhibiting anomalous dispersion. The radius of the microresonator is \(232 \mu \text{m}\) , resulting in an FSR of \(97.7 \text{GHz}\) . The ultra- low- loss waveguide was fabricated on an \(800 \text{nm}\) - thick \(\text{Si}_3\text{N}_4\) layer, deposited by low- pressure chemical vapor deposition (LPCVD) by Ligentec. The device was patterned using \(193 \text{nm}\) photolithography and the quality (Q) factor of the microring resonator is \(7 \times 10^5\) to support the generation of the soliton microcomb. The \(\text{Si}_3\text{N}_4\) chip was packaged with a pair of fiber arrays and the coupling loss is \(\sim 3 \text{dB/facet after being fixed by an ultra- violet (UV) curable adhesive. We utilized a TEC to control the on- chip temperature for the long- term single- soliton generation.  
+
+## Experimental settings and DSP flows of PAM4 parallel data transmission  
+
+For the 80- Gbit/s PAM4- based parallel data transmission, the generated comb lines were first amplified to a total power of \(20 \text{dBm}\) , and then they were sent into a programmable optical filter (Finisar Waveshaper 1000S) to select 15 comb lines and to ensure that the power of each comb line was almost identical to each other. Then, another EDFA was utilized to amplify the 15 comb lines to a total power of \(17.5 \text{dBm}\) . Next, the optical comb lines were simultaneously modulated by a commercial 40- GHz intensity modulator (iXblue MXAN- LN- 40) driven by an PAM4 signal generated from an AWG (Keysight 8199A). The modulated comb lines were sent into the \(20 \text{km SMF}\) , and the dispersion was then compensated for by our dispersion compensator. Then, one channel of the dispersion- compensated parallel signals was filtered out by a tunable narrow- bandwidth optical filter and was directly detected by a commercial 50- GHz photodetector (Finisar XPDV2320R). The received electrical signal was finally recorded by a real- time oscilloscope (RTO, Keysight Z592A), and a DSP algorithm, which is identical to the one utilized in the CW- laser transmission, was applied to process offline the received data. At the transmitter side, a high- speed PAM4 signal was transmitted using the root- raised cosine (RRC) filter with a roll- off factor of 0.01 to compress the signal bandwidth. The matched filter and the time- domain FFE were applied at the receiver side to obtain lower BERs. The instruments and their settings were consistent with those in the CW- laser transmission. The optical power before the photodetector was set to the same power to ensure consistency.  
+
+## Experimental setup and DSP flows of DMT parallel data transmission  
+
+Extended Data Figure 1 shows the schematic experimental setup of the microcomb- based data transmission using the DMT modulation format. A high- speed pseudo- random binary sequence (PRBS) data stream was divided into several parallel low- speed data streams and was mapped to a QAM constellation map. The complex values corresponding to each point in the constellation map were converted to real numbers by an inverse fast Fourier transform (I- FFT). A total number of 160 sub- carriers was mapped to the 32- GHz frequency range. Finally, the signals were converted to a serial data stream after adding the cyclic prefix (CP) to compensate for the multipath delay
+
+<--- Page Split --->
+
+broadening. At the receiver side, the photocurrent signal was sent to the RTO for data collection. The CP was first removed from the sampled data sequence. Next, the serial data was converted to parallel data and mapped to the QAM constellation map through a fast Fourier transform (FFT). The BERs of the full link were then calculated by comparing the measured data sequence with the transmitter data sequence.  
+
+![PLACEHOLDER_14_0]
+  
+
+Extended Data Fig. 1 Experimental setup of the DMT parallel data transmission. P/S Conversion: parallel to serial conversion; S/P Conversion: serial to parallel conversion; FFT: fast Fourier transform; I- FFT: inverse fast Fourier transform; QAM: quadratic amplitude modulation; CP: cyclic prefix.  
+
+## Acknowledgements  
+
+This work was supported in part by the National Key Research and Development Program of China (2018YFB2201702), and the National Natural Science Foundation of China (62090052, 62135010, 62075128).  
+
+## Author contributions  
+
+Y. Liu designed, simulated, and characterized the dispersion compensator. Y. Liu performed the experiments of the CW-laser-based 64-Gbit/s PAM4 data transmission. H. Zhang designed, simulated, and characterized the Kerr microcomb. Y. Liu and H. Zhang conceived the link architecture and performed the high-speed data-transmission experiments of the microcomb-based parallel signal processing. J. Liu conducted the offline DSP. All authors helped analyze the data. L. Lu, J. Du, Y. Liu, H. Zhang, and J. Liu prepared the manuscript.  
+
+## Conflict of interest  
+
+The authors declare that they have no conflict of interest.  
+
+## References  
+
+1. Forbes. 175 Zettabytes by 2025 https://www.forbescom/sites/tomcoughlin/2018/11/27/175-zettabytes-by-2025/ (2018).  
+
+2. Chopra R. Looking beyond 400G - A system vendor perspective. Cisco Systems Inc https://www.ethernetalliance.org/wp-content/uploads/2021/02/TEF21Day1_KeynoteRChopra.pdf, (2020).  
+
+3. Pang XD, et al. 200 Gbps/Lane IM/DD Technologies for Short Reach Optical Interconnects. Journal of Lightwave Technology 38, 492-503 (2020).
+
+<--- Page Split --->
+
+447 4. Alam MS, et al. Net 220 Gbps/A IM/DD Transmission in O-Band and C-Band With Silicon Photonic Traveling-Wave MZM. Journal of Lightwave Technology 39, 4270-4278 (2021). 449 5. Zhang J, et al. Receive-diversity-aided power-fading compensation for C-band IM/DD OFDM systems. Opt Lett 48, 2237-2240 (2023). 451 6. Zou D, Li F, Wang W, Li Z, Li Z. Amplifier-less transmission of beyond 100-Gbit/s/A signal for 40-km DCI-Edge with 10G-class O-band DML. Journal of Lightwave Technology 38, 5649-5655 (2020). 453 7. Wu X, Karar AS, Zhong K, Lau APT, Lu C. Experimental demonstration of pre-electronic dispersion compensation in IM/DD systems using an iterative algorithm. Opt Express 29, 24735-24749 (2021). 455 8. Karar AS. Iterative Algorithm for Electronic Dispersion Compensation in IM/DD Systems. Journal of Lightwave Technology 38, 698-704 (2020). 457 9. Zou DD, Li F, Wang W, Yin MZ, Sui Q, Li ZH. Modified Gerchberg-Saxton Algorithm Based Electrical Dispersion Pre-Compensation for Intensity-modulation and Direct-detection Systems. Journal of Lightwave Technology 40, 2840-2849 (2022). 460 10. Zou DD, et al. 100G PAM-6 and PAM-8 Signal Transmission Enabled by Pre-Chirping for 10-km Intra-DCI Utilizing MZM in C-band. Journal of Lightwave Technology 38, 3445-3453 (2020). 462 11. Yin M, Zou D, Wang W, Li F, Li Z. Transmission of a 56 Gbit/s PAM4 signal with low-resolution DAC and pre- equalization only over 80 km fiber in C-band IM/DD systems for optical interconnects. Opt Lett 46, 5615-5618 (2021). 465 12. Zhong KP, Zhou X, Huo JH, Yu CY, Lu C, Lau APT. Digital Signal Processing for Short-Reach Optical Communications: A Review of Current Technologies and Future Trends. Journal of Lightwave Technology 36, 377-400 (2018). 468 13. Wu X, Zhang J, Lau APT, Lu C. C-band 100-GBand PS-PAM-4 transmission over 50-km SSMF enabled by FIR-filter-based pre-electronic dispersion compensation. Opt Express 31, 17759-17768 (2023). 470 14. Kilper DC, Rastegarfar H. Energy challenges in optical access and aggregation networks. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 374, 20140435 (2016). 472 15. Yang KY, et al. Bridging ultrahigh-Q devices and photonic circuits. Nature Photonics 12, 297-303 (2018). 473 16. Brasch V, et al. Photonic chip-based optical frequency comb using soliton Cherenkov radiation. Science 351, 357-360 (2016). 475 17. Guo H, et al. Universal dynamics and deterministic switching of dissipative Kerr solitons in optical microresonators. Nature Physics 13, 94-102 (2016). 477 18. Herr T, et al. Temporal solitons in optical microresonators. Nature Photonics 8, 145-152 (2013). 478 19. Liu JQ, et al. Ultralow-power chip-based soliton microcombs for photonic integration. Optica 5, 1347-1353 (2018). 479 20. Shen B, et al. Integrated turnkey soliton microcombs. Nature 582, 365-369 (2020). 480 21. Stern B, Ji X, Okawachi Y, Gaeta AL, Lipson M. Battery-operated integrated frequency comb generator. Nature 562, 401-405 (2018). 482 22. Yi X, Yang QF, Yang KY, Suh MG, Vahala K. Soliton frequency comb at microwave rates in a high-Q silica microresonator. Optica 2, 1078-1085 (2015). 484 23. Zhang SY, et al. Sub-milliwatt-level microresonator solitons with extended access range using an auxiliary laser. Optica 6, 206-212 (2019). 486 24. Zhou H, et al. Soliton bursts and deterministic dissipative Kerr soliton generation in auxiliary-assisted microcavities. Light: Science & Application 8, 50 (2019). 488 25. Marin-Palomo P, et al. Microresonator-based solitons for massively parallel coherent optical communications. Nature 546, 274-279 (2017). 490 26. Corcoran B, et al. Ultra-dense optical data transmission over standard fibre with a single chip source. Nature
+
+<--- Page Split --->
+
+491 Communications 11, 2568 (2020).492 27. Jørgensen AA, et al. Petabit-per-second data transmission using a chip-scale microcomb ring resonator source. 493 Nature Photonics 16, 798- 802 (2022).494 28. Geng Y, et al. Coherent optical communications using coherence-cloned Kerr soliton microcombs. Nature 495 Communications 13, 1070 (2022).496 29. Mazur M, et al. High Spectral Efficiency Coherent Superchannel Transmission With Soliton Microcombs. Journal 497 of Lightwave Technology 39, 4367- 4373 (2021).498 30. Fulop A, et al. High- order coherent communications using mode- locked dark- pulse Kerr combs from 499 microresonators. Nature Communications 9, 1598 (2018).500 31. Rizzo A, et al. Massively scalable Kerr comb- driven silicon photonic link. Nature Photonics, (2023).501 32. Shu H, et al. Microcomb- driven silicon photonic systems. Nature 605, 457- 463 (2022).502 33. Liu SJ, Zhang CP, Cao HY, Hong SH, Yu ZJ, Dai DX. On- Chip Circulator- Free Chirped Spiral Multimode Waveguide 503 Grating for Dispersion Management. ACS Photonics 10, 1654- 1661 (2023).504 34. Kaushal S, Roberge A, Kashyap R, Azana J. Ultra- compact silicon photonics highly dispersive elements for low- 505 latency signal processing. Opt Express 31, 3467- 3478 (2023).506 35. Chen GF, Ong KYK, Chia XX, Cao Y, Tan DTH. Dispersion Compensation of 30GBaud/s NRZ and PAM4 data using 507 integrated Silicon Nitride Gratings. In: Optical Fiber Communications Conference and Exhibition (OFC), paper 508 Th1D.4 (2022).509 36. Liu SJ, Liu DJ, Yu ZJ, Liu L, Shi YC, Dai DX. Digitally tunable dispersion controller using chirped multimode 510 waveguide gratings. Optica 10, 316- 323 (2023).511 37. Giuntoni I, et al. Integrated Drop- Filter for Dispersion Compensation Based on SOI Rib Waveguides. Conference 512 on Optical Fiber Communication and Collocated National Fiber Optic Engineers Conference (OFC- NFOEC), paper 513 OTjh5, (2010).514 38. Moreira R, Gundavarapu S, Blumenthal DJ. Programmable eye- opener lattice filter for multi- channel dispersion 515 compensation using an integrated compact low- loss silicon nitride platform. Opt Express 24, 16732- 16742 (2016).516 39. Brodnik GM, Pinho C, Chang F, Blumenthal DJ. Extended Reach 40km Transmission of C- Band Real- Time 53.125 517 Gbps PAM- 4 Enabled with a Photonic Integrated Tunable Lattice Filter Dispersion Compensator. In: Optical Fiber 518 Communications Conference and Exposition (OFC), paper W2A.30 (2018).519 40. Madsen CK, et al. An All- Pass Filter Dispersion Compensator Using Planar Waveguide Ring Resonators. In: Optical 520 Fiber Communication Conference and the International Conference on Integrated Optics and Optical Fiber 521 Communication, paper FE6 (1999).522 41. Madsen CK, Lenz G, Bruce AJ, Cappuzzo MA, Gomez LT, Scotti RE. Integrated all- pass filters for tunable 523 dispersion and dispersion slope compensation. IEEE Photonics Technology Letters 11, 1623- 1625 (1999).524 42. Notaros J, et al. Programmable dispersion on a photonic integrated circuit for classical and quantum applications. 525 Opt Express 25, 21275- 21285 (2017).526 43. Sorianello V, et al. 100Gb/s PolMux- NRZ Transmission at 1550nm over 30km Single Mode Fiber Enabled by a 527 Silicon Photonics Optical Dispersion Compensator. In: Optical Fiber Communication Conference (OFC), paper 528 W2A.31 (2018).529 44. Alam MS, et al. 224 Gb/s Transmission over 10 km of SMF at 1550 nm Enabled by a SiN Optical Dispersion 530 Compensator and Stokes Vector Direct Detect Receiver. In: OSA Advanced Photonics Congress (APC), paper 531 SpM4I.5 (2020).532 45. Liu Y, Lu L, Ni Z, Chen J, Chen J, Zhou L. Silicon Integrated Continuously Tunable Dispersion Compensator Based 533 on Cascaded Micro- Ring Resonators. In: 2022 Asia Communications and Photonics Conference (ACP) (2022).534 46. Ong KYK, Chen GFR, Xing P, Gao H, Tan DTH. Dispersion compensation of high- speed data using an integrated
+
+<--- Page Split --->
+
+535 silicon nitride ring resonator. Opt Express 30, 13959- 13967 (2022).536 47. Zhang Y, et al. 240 Gb/s optical transmission based on an ultrafast silicon microring modulator. Photon Res 10, 1127- 1133 (2022).537 48. Dong P. Silicon Photonic Integrated Circuits for Wavelength- Division Multiplexing Applications. IEEE Journal of Selected Topics in Quantum Electronics 22, 370- 378 (2016).538 49. Liu D, Zhang L, Tan Y, Dai D. High- Order Adiabatic Elliptical- Microring Filter with an Ultra- Large Free- Spectral- Range. Journal of Lightwave Technology 39, 5910- 5916 (2021).539 50. Liu DJ, He JH, Xiang YL, Xu Y, Dai DX. High- performance silicon photonic filters based on all- passive tenth- order adiabatic elliptical- microrings. Apl Photonics 7, 051303 (2022).540 51. Horst F, Green WMJ, Assefa S, Shank SM, Vlasov YA, Offrein BJ. Cascaded Mach- Zehnder wavelength filters in silicon photonics for low loss and flat pass- band WDM (de- )multiplexing. Opt Express 21, 11652- 11658 (2013).541 52. Jeong SH, Shimura D, Simoyama T, Horikawa T, Tanaka Y, Morito K. Si- nanowire- based multistage delayed Mach- Zehnder interferometer optical MUX/DeMUX fabricated by an ArF- immersion lithography process on a 300 mm SOI wafer. Opt Lett 39, 3702- 3705 (2014).542 53. Sun Q, Zhou L, Lu L, Zhou G, Chen J. Reconfigurable High- Resolution Microwave Photonic Filter Based on Dual- Ring- Assisted MZIs on the Si3N4 Platform. IEEE Photonics Journal 10, 1- 12 (2018).543 54. Kohli N, Sang BL, Nabki F, Menard M. Tunable Bandpass Filter With Serially Coupled Ring Resonators Assisted MZI. IEEE Photonics Journal 13, 1- 8 (2021).544 55. Fujii S, Tanabe T. Dispersion engineering and measurement of whispering gallery mode microresonator for Kerr frequency comb generation. Nanophotonics 9, 1087- 1104 (2020).545 56. Zhang HY, Lu LJ, Chen JP, Zhou LJ. Program- Controlled Single Soliton Generation Driven by the Thermal- Compensated Avoided Mode Crossing. Journal of Lightwave Technology 41, 1801- 1810 (2023).546 57. OIF. Common Electrical I/O (CEI) - Electrical and Jitter Interoperability agreements for 6G+ bps, 11G+ bps, 25G+ bps, 56G+ bps and 112G+ bps I/O. (2022).547 58. Helgason oB, Girardi M, Ye Z, Lei F, Schroder J, Torres- Company V. Power- efficient soliton microcombs. arXiv:202209410, (2022).548 59. He A, Guo X, Wang K, Zhang Y, Su Y. Low Loss, Large Bandwidth Fiber- Chip Edge Couplers Based on Silicon- on- Insulator Platform. Journal of Lightwave Technology 38, 4780- 4786 (2020).549 60. Liu W, Zhang J, Liu L, Dai D, Shi Y. High Efficiency Silicon Edge Coupler Based On Uniform Arrayed Waveguides With Un- Patterned Cladding. IEEE Photonics Technology Letters 32, 1077- 1080 (2020).550 61. Marchetti R, Lacava C, Carroll L, Gradkowski K, Minzioni P. Coupling strategies for silicon photonics integrated chips [Invited]. Photon Res 7, 201- 239 (2019).
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupplementaryInformation.pdf
+
+<--- Page Split --->

preprint/preprint__48961379f14baa7d8ee3a9307154d8bf615c946407f087937d7589936c43fad7/preprint__48961379f14baa7d8ee3a9307154d8bf615c946407f087937d7589936c43fad7_det.mmd

@@ -0,0 +1,298 @@
+<|ref|>title<|/ref|><|det|>[[44, 106, 872, 210]]<|/det|>
+Parallel Wavelength-Division-Multiplexed Signal Transmission and Dispersion Compensation Enabled by Soliton Microcombs and Microrings  
+
+<|ref|>text<|/ref|><|det|>[[44, 230, 300, 277]]<|/det|>
+Liangjun Lu luliangjun@sjtu.edu.cn  
+
+<|ref|>text<|/ref|><|det|>[[44, 303, 325, 350]]<|/det|>
+Shanghai Jiao Tong University Yuanbin Liu  
+
+<|ref|>text<|/ref|><|det|>[[44, 353, 325, 370]]<|/det|>
+Shanghai Jiao Tong University  
+
+<|ref|>text<|/ref|><|det|>[[44, 375, 325, 415]]<|/det|>
+Hongyi Zhang Shanghai Jiao Tong University  
+
+<|ref|>text<|/ref|><|det|>[[44, 420, 325, 460]]<|/det|>
+Jiacheng Liu Shanghai Jiao Tong University  
+
+<|ref|>text<|/ref|><|det|>[[44, 465, 682, 505]]<|/det|>
+Jiangbing Du Shanghai Jiao Tong University https://orcid.org/0000- 0002- 6333- 824X  
+
+<|ref|>text<|/ref|><|det|>[[44, 510, 325, 551]]<|/det|>
+Yu Li Shanghai Jiao Tong University  
+
+<|ref|>text<|/ref|><|det|>[[44, 556, 325, 596]]<|/det|>
+Zuyuan He Shanghai Jiao Tong University  
+
+<|ref|>text<|/ref|><|det|>[[44, 601, 325, 641]]<|/det|>
+Jianping Chen Shanghai Jiaotong University  
+
+<|ref|>text<|/ref|><|det|>[[44, 646, 675, 686]]<|/det|>
+Linjie Zhou Shanghai Jiao Tong University https://orcid.org/0000- 0002- 2792- 2959  
+
+<|ref|>text<|/ref|><|det|>[[44, 692, 845, 736]]<|/det|>
+Andrew Poon Hong Kong University of Science and Technology https://orcid.org/0000- 0002- 5222- 8184  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 781, 102, 798]]<|/det|>
+## Article  
+
+<|ref|>title<|/ref|><|det|>[[44, 818, 135, 836]]<|/det|>
+# Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 856, 352, 875]]<|/det|>
+Posted Date: September 18th, 2023  
+
+<|ref|>text<|/ref|><|det|>[[44, 894, 473, 913]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 3282850/v1
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 44, 908, 87]]<|/det|>
+License: © ① This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 105, 530, 125]]<|/det|>
+Additional Declarations: There is NO Competing Interest.  
+
+<|ref|>text<|/ref|><|det|>[[42, 160, 909, 204]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on April 29th, 2024. See the published version at https://doi.org/10.1038/s41467-024-47904-2.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[110, 120, 831, 217]]<|/det|>
+# Parallel Wavelength-Division-Multiplexed Signal Transmission and Dispersion Compensation Enabled by Soliton Microcombs and Microrings  
+
+<|ref|>text<|/ref|><|det|>[[100, 238, 852, 430]]<|/det|>
+Yuanbin Liu \(^{1,4}\) , Hongyi Zhang \(^{1,4}\) , Jiacheng Liu \(^{1,4}\) , Liangjun Lu \(^{1,2,*}\) , Jiangbing Du \(^{1,*}\) , Yu Li \(^{1,2}\) , Zuyuan He \(^{1}\) , Jianping Chen \(^{1,2}\) , Linjie Zhou \(^{1,2}\) , and Andrew W. Poon \(^{3}\) \(^{1}\) State Key Laboratory of Advanced Optical Communication Systems and Networks, Shanghai Key Lab of Navigation and Location Services, Department of Electronic Engineering, Shanghai Jiao Tong University, Shanghai 200240, China \(^{2}\) SJTU-Pinghu Institute of Intelligent Optoelectronics, Pinghu 314200, China \(^{3}\) Photonic Device Laboratory, Department of Electronic and Computer Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong \(^{4}\) Corresponding authors' e-mail address: luliangjun@sjtuedu.cn, dujiangbing@sjtuedu.cn \(^{5}\) These authors contributed equally to this work.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 459, 216, 472]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[144, 483, 852, 851]]<|/det|>
+The proliferation of computation- intensive technologies has led to a significant rise in the number of datacenters, posing challenges for high- speed and power- efficient datacenter interconnects (DCIs). Although inter- DCIs based on intensity modulation and direct detection (IM- DD) along with wavelength- division multiplexing (WDM) technologies exhibit cost- effective, power- efficient, and large- capacity properties, the requirement of multiple laser sources leads to high costs and limited scalability. Moreover, careful considerations must be given to the chromatic dispersion in the C- band as it restricts the transmission length of the optical signals. Electronic and optical approaches based on digital signal processing algorithms or dispersion- compensating fibers suffer from either a high power consumption or a lack of full reconfigurability. In this study, we present an original scalable on- chip parallel IM- DD data transmission system enabled by a single- soliton Kerr microcomb and a reconfigurable microring resonator (MRR)- based dispersion compensator. The highly compact MRR- based Kerr microcomb and dispersion compensator are intrinsically compatible with the parallel processing nature of the WDM link. Besides, the reconfigurability of the dispersion compensator shows the validation for various data- rate transmissions over multiple single- mode fiber lengths of up to 40 km, with a power consumption of below 160 mW, regardless of the modulation format or of the number of transmission channels utilized. Through our experimental validation, we demonstrate an aggregate bit rate of 1.68 Tbit/s over a 20- km- long single- mode fiber using 15 independent wavelength channels spaced at 100 GHz. Our approach holds significant promise for achieving data communications at a scale exceeding 10 terabits, making it highly valuable for future hyper- scale DCIs.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 862, 244, 875]]<|/det|>
+## Introduction  
+
+<|ref|>text<|/ref|><|det|>[[144, 886, 848, 903]]<|/det|>
+According to the International Data Corporation (IDC), the exponential growth of digital data
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[144, 85, 852, 435]]<|/det|>
+generation is projected to reach 175 zettabytes by \(2025^{1}\) due to the surgen growth of computation- intensive technologies such as artificial intelligence, Internet of Things (IoT), and autonomous vehicles. The total data capacity grows 80 times from 640 GHz in 2010 to 51.2 THz in 2022 for data switching in datacenter interconnects (DCls) while the total power consumption grows 22 times \(^{2}\) . As a result, there is a critical need for energy- efficient and cost- effective DCls that can accommodate a substantial capacity. For short- reach DCls spanning distances of less than \(40 \text{km}\) , the utilization of intensity modulation and direct detection (IM- DD) has emerged as a practical and preferred solution \(^{3}\) . This choice is motivated by the inherent advantages IM- DD offers, including cost- effectiveness, a low power consumption, and a compact physical footprint. Due to the lower data rate of the IM- DD system for a single wavelength channel than the coherent scheme, wavelength- division multiplexing (WDM) technology is commonly employed to economically enhance the data capacity. However, multiple laser sources are required to provide multiple operation wavelengths in the WDM systems, which leads to a higher cost. Moreover, in IM- DD systems operating within the C- band, the transmission distance is primarily limited by chromatic dispersion (CD), necessitating effective mitigation strategies \(^{6 - 8}\) . The CD- induced frequency- selective fading (FSF) significantly impacts transmission performance \(^{9, 10}\) . To address the dispersion effect, various electronic and optical approaches have been employed, such as digital signal processing (DSP) \(^{11 - 13}\) and dispersion- compensating fibers (DCFs). However, these methods are burdened with limitations, particularly in terms of a high energy consumption and associated costs \(^{14}\) .  
+
+<|ref|>text<|/ref|><|det|>[[145, 437, 852, 731]]<|/det|>
+Chip- scaled soliton microcombs, known for their wide spectral range, low noise, and high repetition rate, have emerged as a promising light source for WDM applications \(^{15 - 24}\) . Recent advancements have demonstrated massively parallel data transmission using various types of microcomb sources, including bright single- solitons \(^{25}\) , perfect soliton crystals \(^{26}\) , and dark solitons \(^{27}\) . For example, the utilization of two interleaved single- soliton microcombs in a silicon nitride (Si \(_{3}\) N \(_{4}\) ) microresonator enabled coherent data transmission with 179 carriers in the C- and L- bands, achieving a total bit rate of 55 Tb/s \(^{25}\) . Similarly, the soliton crystal with defects, generated in a doped silica glass microresonator, facilitated a line rate of 40.1 Tb/s and a spectral efficiency of 10.2 b/s/Hz \(^{26}\) . However, it is critical to note that the majority of WDM transmission systems utilizing soliton microcombs rely on expensive coherent schemes, necessitating costly transmitters and receivers \(^{25 - 30}\) . Recently, massively parallel optical interconnects based on a Si \(_{3}\) N \(_{4}\) Kerr dark soliton comb source and a silicon microdisk modulator array with a data rate of 512 Gb/s for IM- DD transmission have been demonstrated \(^{31}\) . In addition, a parallel optical data link driven by an AlGaAs Kerr microcomb has achieved a data rate of 2 Tbit/s in the IM- DD system \(^{32}\) . However, we note that these transmission techniques are limited in term of transmission lengths and are primarily suitable for intra- DCI applications.  
+
+<|ref|>text<|/ref|><|det|>[[145, 734, 851, 878]]<|/det|>
+In recent years, the adoption of silicon photonics facilitated the integration of various silicon- based photonic devices, including chirped Bragg gratings \(^{33 - 37}\) , Mach- Zehnder interferometers (MZIs) \(^{38, 39}\) , and microring resonators (MRRs) \(^{40 - 46}\) , to effectively mitigate the impact of CD in single- mode fibers (SMFs). These integrated devices offer several advantages, including a compact form factor, cost efficiency, a low power consumption, and a high reconfigurability. While imbalanced MZIs and MRRs demonstrate desirable characteristics for parallel signal processing due to their periodic spectral properties, their potential in WDM systems is currently limited. As a result, existing studies have reported only a modest maximum data transmission rate of 224 Gbit/s \(^{44}\) .  
+
+<|ref|>text<|/ref|><|det|>[[168, 882, 849, 897]]<|/det|>
+In this paper, we present an original approach for short- reach inter- DCls based on an IM- DD
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 85, 852, 380]]<|/det|>
+optical communications architecture. Our proposed system leverages a single- soliton microcomb source generated by a \(\mathrm{Si}_3\mathrm{N}_4\) MRR as an WDM light source. Furthermore, we introduce a fully reconfigurable optical dispersion compensator based on cascaded silicon MRRs for parallel dispersion compensation. This WDM- assisted IM- DD scheme offers numerous advantages, including a high scalability, compactness, energy efficiency, and cost- effectiveness. To evaluate the performance of our proposed system, we conducted experiments demonstrating parallel signal transmission using up to 15 wavelength channels within the C- band. By employing pulse- amplitude four- level (PAM4) modulation signals at a rate of 80 Gbit/s, we achieved an aggregate data rate of 1.2 Tbit/s. Additionally, by employing discrete multi- tone (DMT) modulation signals at a rate of 112 Gbit/s, we achieved an aggregate data rate of 1.68 Tbit/s. These transmission rates were achieved over a 20- km- length SMF. Importantly, we assessed the power consumption of the dispersion compensator to be consistently below 160 mW, regardless of the modulation format or the number of transmission channels utilized. By aligning the free spectral range (FSR) of the Kerr MRR and the dispersion compensator and further increasing the number of wavelength channels, we expect the transmission capacity to exceed 10 Tbit/s. This capability holds significant promise for meeting the demands of future hyperscale DCIs.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 390, 208, 404]]<|/det|>
+## Results  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 415, 725, 431]]<|/det|>
+## Parallel IM-DD data transmission and dispersion compensation architecture  
+
+<|ref|>text<|/ref|><|det|>[[145, 440, 852, 828]]<|/det|>
+Figure 1a shows the schematics of the integrated parallel data communications system for IM- DD applications. At the transmitter, a bright single- soliton microcomb with a smooth spectrum, which is generated by a \(\mathrm{Si}_3\mathrm{N}_4\) microresonator, is utilized as the multi- wavelength laser source with evenly distributed and low- noise frequency tones. The comb source is then demultiplexed, independently encoded with electrical data, and multiplexed by silicon microring modulator (MRM) arrays working at several individual wavelength channels. The modulated data are transmitted through an SMF before being sent to the receiver. The MRMs have the advantages of a compact size and of a high modulation bandwidth, which have been demonstrated with a modulation bandwidth of 110 GHz<sup>47</sup>. To increase the channels and to reduce the crosstalk, de-/multiplexers based on MRRs<sup>48- 50</sup>, asymmetric MZIs<sup>51, 52</sup>, and MRR- coupled MZIs<sup>53, 54</sup> can be inserted before and after the silicon MRR modulator arrays. At the receiver end, we employ a silicon- based dispersion compensator utilizing cascaded MRRs to simultaneously compensate for the dispersion induced by SMF transmission across all the wavelength channels. The FSR of the cascaded MRRs is designed to match the wavelength spacing of the frequency combs. Moreover, the resonant wavelengths and the coupling coefficients of each MRR can be independently adjusted, offering a fully reconfigurable configuration that caters to various transmission ranges and operation bandwidths. Following the compensation process, the signal is then separated into distinct wavelength channels with MRR- based demultiplexers for photodetection. This comprehensive system enables parallel data transmission and dispersion compensation through the integration of photonic devices, featuring a simple arrangement and remarkable scalability. Consequently, our approach holds great promise for the development of fully integrated photonic circuits, especially for high- capacity inter- DCIs.  
+
+<|ref|>text<|/ref|><|det|>[[147, 830, 850, 901]]<|/det|>
+In this proof- of- concept study, we used a \(\mathrm{Si}_3\mathrm{N}_4\) microresonator with a bending radius of \(232\mu \mathrm{m}\) and an average quality (Q) factor of \(7\times 10^{5}\) for single- soliton generation. Figure 1b shows the microscope image of the fabricated \(\mathrm{Si}_3\mathrm{N}_4\) microresonator. The waveguide cross- section is \(1800\times 800 \mathrm{nm}^2\) and the measured second- order group- velocity dispersion \((\mathrm{D}_2 / 2\pi)\) is \(601.2 \mathrm{kHz}^{55}\) .
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[153, 88, 840, 355]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 364, 852, 546]]<|/det|>
+<center>Fig 1. Chip-based parallel data communication architecture for short-reach inter-DCls. a Schematics of the data transmission system. At the transmitter, a multi-wavelength laser source with an equal channel spacing is generated by a \(\mathrm{Si}_3\mathrm{N}_4\) Kerr MRR, and then individually intensity-modulated by integrated MRR modulator arrays. After that, the modulated multi-wavelength optical signal transmits through an SMF. At the receiver side, the parallel WDM signals are first processed by a cascaded-MRRs-based optical signal processor for dispersion compensation, and then demultiplexed by MRR filter arrays and received by photodetectors separately. b Microscope image of the \(\mathrm{Si}_3\mathrm{N}_4\) micro-resonator for single-soliton generation. c Microscope image of the silicon-MRRs-based dispersion compensator. The numbers label the self-defined sequence of each ring, and the input and output ports are labeled. d Enlarged view of a silicon MRR in the dispersion compensator. e Schematics of the silicon MRR. </center>  
+
+<|ref|>text<|/ref|><|det|>[[147, 556, 852, 794]]<|/det|>
+The FSR is about 97.7 GHz. We utilized a program- controlled scheme based on thermal auxiliary compensation to generate and to stabilize the microcomb in the whole experiments<sup>56</sup>. See Supplementary Note 1 for the detailed program- controlled scheme for the single- soliton generation and stabilization. At the receiver side, the fully reconfigurable silicon dispersion compensator comprises 8 identical MRRs. Figure 1c shows the microscope image of the device. Figures 1d and 1e demonstrate the enlarged microscope image and the schematics of one MRR, respectively. The FSR of these MRRs is around 99.5 GHz for dense WDM (DWDM) scenarios in the C- band. The coupling region of each MRR is replaced by an MZI- based tunable coupler. Both the ring and the coupling region are integrated with a titanium micro- heater for independent thermal adjustment of the resonant wavelength and of the coupling coefficient. A pair of grating couplers are utilized to couple the optical signal in and out of this chip. The footprint of the dispersion compensator is \(2 \mathrm{~mm} \times 1 \mathrm{~mm}\). See Methods for the detailed design, fabrication, and packaging of the \(\mathrm{Si}_3\mathrm{N}_4\) Kerr comb and of the dispersion compensator.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 804, 830, 820]]<|/det|>
+## Characterizations of the dispersion compensator and data transmission using a CW laser  
+
+<|ref|>text<|/ref|><|det|>[[147, 829, 852, 901]]<|/det|>
+To achieve reconfiguration of the dispersion compensator for varying lengths of SMF, we devised an optimization algorithm that combines the transfer matrix method with the sequential quadratic programming (SQP) algorithm. This algorithm allowed us to determine the target parameters for the MRR- based dispersion compensator. We successfully demonstrated dispersion compensation for
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 86, 851, 286]]<|/det|>
+SMF transmission up to \(40km\) within a bandwidth of \(32GHz\) . See Supplementary Note 2 for a detailed description of the optimization methodology and of the obtained results. Then, to facilitate an efficient dispersion compensation for diverse dispersion values, the optimized resonant wavelengths and coupling coefficients of all the MRRs for different dispersion conditions were documented in a look- up table. Accounting for the inherent fabrication deviations in silicon photonic devices, we deem it necessary the calibration of the random states for each MRR. Furthermore, we investigated the impact of thermal crosstalk on the tuning of the chip to their optimal configurations, considering the intra- and inter- thermal crosstalk within and between MRRs, respectively. By characterizing the tuning efficiencies influenced by the intra- and inter- thermal crosstalk and adjusting the applied voltages of MRRs correspondingly according to these efficiencies, we successfully compensated for the thermal crosstalk- induced resonant wavelength shifts.  
+
+<|ref|>text<|/ref|><|det|>[[147, 289, 851, 415]]<|/det|>
+Figures 2a and 2b show the measured transmission and group delay responses of dispersion compensation for a \(40 - km\) - long SMF with and without thermal crosstalk compensation (TCC). The implementation of TCC significantly alleviated these shifts, leading to a smoother and more linearly varied response within the operation bandwidth. For detailed information on the calibration process, see Supplementary Note 3. We note that the proposed TCC method is applicable to other photonic devices comprising multiple thermal phase shifters, which can significantly simplify the adjustment of the working states.  
+
+<|ref|>text<|/ref|><|det|>[[147, 419, 851, 805]]<|/det|>
+Figures 2c and 2e show the measured transmission spectra and the corresponding group delay responses covering a wavelength span from \(1540nm\) to \(1580nm\) after the calibration. The MRRs were calibrated at the wavelength of \(1558.7nm\) with a \(32 - GHz\) operation bandwidth for \(10 - km\) to \(40 - km\) SMFs. The enlarged views at the wavelength around \(1558.7nm\) are shown in Figs. 2d and 2f. The transmission spectra include the coupling loss of the grating couplers. The grating coupler has a maximum loss non- uniformity of \(\sim 3.5dB\) over the \(40nm\) wavelength, which can be improved by using an edge coupler. The increased on- resonance transmission loss at the shorter wavelength side is due to the wavelength- dependent loss and the coupling coefficient of the MRRs. Figures 2g and 2h show the extracted insertion loss and the averaged dispersion of the dispersion compensator. The insertion loss is obtained at the center of the operation bandwidth of each channel, and the averaged dispersion is calculated as \(D_{avg} = (\tau_1 - \tau_2) / (\lambda_1 - \lambda_2)\) , where \(\lambda_1\) and \(\lambda_2\) are the shorter and longer wavelength ends across the operation bandwidth of each channel, \(\tau_1\) and \(\tau_2\) are the corresponding group delays, respectively. The insertion loss increases from \(\sim 8.4dB\) to \(\sim 14.8dB\) at the wavelength around \(1558.7nm\) for dispersion compensation of \(10 - km\) to \(40 - km\) SMFs, respectively. We can reduce the insertion loss by lowering the transmission loss of the silicon MRR waveguides. The averaged dispersion is \(- 169.5ps/nm\) , \(- 325.4ps/nm\) , \(- 506.6ps/nm\) , and \(- 682.9ps/nm\) for dispersion compensation of \(10 - km\) , \(20 - km\) , \(30 - km\) , and \(40 - km\) - long SMFs at the same wavelength. The group delay increases slightly with the wavelength, which we attribute to the coupling coefficient variations of MRRs across wavelengths. The group delay degradation around \(1540nm\) is limited by the dynamic range of our measurement equipment, the Photonic Dispersion and Loss Analyzer (PDLA, Agilent 86038B).  
+
+<|ref|>text<|/ref|><|det|>[[147, 808, 851, 898]]<|/det|>
+Next, we conducted a 64 Gbit/s PAM4 signal transmission experiment using a continuous- wave (CW) laser as the light source to demonstrate the reconfigurability and the wide- band signal processing capability of our dispersion compensator in a wide wavelength span. We utilized an offline digital signal processing (DSP) algorithm with a time- domain feed- forward equalization (FFE) to recover the received signal. We note that no dispersion compensation algorithm was used in the
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[153, 87, 844, 736]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 743, 851, 906]]<|/det|>
+<center>Fig. 2. Characterizations of the dispersion compensator and data transmission using a CW-laser. a, b Measured (a) transmission and (b) group delay responses with and without thermal crosstalk compensation for the dispersion compensation of 40-km-long SMF. TCC: thermal crosstalk compensation. c Measured transmission spectra for dispersion compensation of 10-km-, 20-km-, 30-km-, and 40-km-long SMFs in the wavelength range of 1540 nm to 1580 nm. d Enlarged view of the transmission spectra at the wavelength around 1558.7 nm. The operation bandwidth is calibrated to 32 GHz. e Measured group delay responses in the 40-nm wavelength span. f Enlarged view of the group delay responses at the wavelength of about 1558.7 nm. g Extracted insertion loss of each channel for dispersion compensation of different lengths of SMFs. The insertion loss is obtained at the center of the operation bandwidth. h Averaged dispersion of each channel for the 4 kinds </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 87, 850, 212]]<|/det|>
+of dispersion compensation. i Measured PAM4 eye diagrams for the data transmission using a CW- laser as the light source at the wavelength of around 1558.96 nm for the four lengths of SMFs with and without dispersion compensation. 64- Gbit/s PAM4 signal is transmitted through SMFs. DC: dispersion compensation. j Measured BERs for data transmission under BtB transmission and SMFs transmissions with and without dispersion compensation at 5 different wavelengths in the 30- nm wavelength span ranging from 1540 nm to 1570 nm. BERs are much higher than the 20% SD- FEC threshold \((2\times 10^{- 2})\) without the dispersion compensation, while they are all within the \(7\%\) HD- FEC threshold \((3.8\times 10^{- 3})\) after the dispersion compensation.  
+
+<|ref|>text<|/ref|><|det|>[[145, 222, 851, 534]]<|/det|>
+signal receiving end. See Supplementary Note 4 for the detailed experimental setup. We measured 5 different wavelengths distributed from 1540 nm to 1570 nm under the 10-, 20-, 30-, and 40- km- long SMFs with and without dispersion compensation. The wavelength span was limited by our tunable bandpass filter. Figure 2i shows the measured eye diagrams for the transmissions with and without dispersion compensation at the wavelength of 1558.96 nm under the four lengths of SMFs. In the absence of dispersion compensation, the eye diagrams exhibit significant blurring. In contrast, with our dispersion compensator, the eye diagrams are successfully restored. The rest of the measured eye diagrams at the other wavelengths are included in Supplementary Note 4. The bit- error ratios (BERs) were also measured and shown in Fig. 2j. The BERs without dispersion compensator are all far beyond the 20% soft- decision forward- error correction (SD- FEC) threshold while all of them are below the 7% hard- decision forward- error correction (HD- FEC) threshold after dispersion compensation, demonstrating that our dispersion compensator exhibits a fairly good continuous dispersion compensation tunability and is capable of parallelly processing the data of at least 37 channels over the 30- nm wavelength range. The slightly higher BERs at wavelengths around 1540 nm and 1570 nm are due to the grating coupler- induced loss non- uniformity. The successful data transmission at the wavelength around 1540 nm indicates the degradation of the averaged dispersion of the 40- km- long SMF shown in Fig. 2h is limited by our instrument.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 545, 550, 560]]<|/det|>
+## Parallel data transmission using a comb light source  
+
+<|ref|>text<|/ref|><|det|>[[147, 571, 850, 679]]<|/det|>
+We demonstrated parallel data transmission and dispersion compensation using the \(\mathrm{Si}_3\mathrm{N}_4\) soliton comb lines as the light source. In this demonstration, we utilized a 20- km- long SMF in the transmission, and we calibrated our dispersion compensator for a 50- GHz operation bandwidth. See Supplementary Note 3 for the measured transmission spectra and group delay responses. The 80- Gbit/s PAM4 and 112- Gbit/s discrete multi- tone (DMT) signals were transmitted through the SMF, respectively.  
+
+<|ref|>text<|/ref|><|det|>[[147, 682, 851, 901]]<|/det|>
+Figure 3a shows the experimental setup of the PAM4 signal transmission. The spectrum of the generated single- soliton microcombs is depicted in Fig. 3b. The 10- dB spectral bandwidth of the single soliton is 62.6 nm, with 80 comb lines ranging from 1527 nm to 1590 nm included. Here the 10- dB spectral bandwidth describes the optical wavelength span of the spectrum with the optical power of the comb lines decreased by 10 dB from the comb line with the maximum power. As there is a slight FSR discrepancy between the comb lines ( \(\sim 97.7\) GHz) and the dispersion compensators ( \(\sim 99.5\) GHz), we only selected 15 comb lines from 1555 nm to 1563 nm by a programmable optical filter for the following data transmission. In principle, by carefully designing the FSRs of both devices with a smaller or even no discrepancy, we can select more wavelength channels for parallel data transmission. We utilized a single modulator to simultaneously modulate all the wavelength channels for simplicity. We detail the experimental settings and the DSP flows in Methods. The optical spectra of the reshaped and of the modulated comb lines are depicted in Figs. 3c and 3d, respectively.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[163, 88, 840, 568]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[145, 576, 852, 793]]<|/det|>
+<center>Fig. 3. 80-Gbit/s PAM4-based WDM data transmission with microcombs. a Schematic experimental setup of a single-soliton microcomb driven parallel signal transmission system. PC: polarization controller; EDFA: erbium-doped fiber amplifier; WS: waveshaper; EA: electrical amplifier; AWG: arbitrary waveform generator; RTO: real-time oscilloscope; RRC: root-raised cosine; FFE: feed-forward equalization. b Measured spectrum of the generated single-soliton microcomb, which shows a sech²-like spectral shape. The utilized 15 comb lines are indicated by a dashed box. c Measured spectrum of the chosen-out 15 comb lines after the waveshaper. d Measured spectrum of the 15 comb lines after the 40-Gbaud PAM4 modulation. e Measured eye diagrams of the \(4^{\text{th}}\) , \(6^{\text{th}}\) , \(8^{\text{th}}\) , \(10^{\text{th}}\) , and \(12^{\text{th}}\) channels after the dispersion compensation. f Measured BERs for the 15 WDM channels under the BTB transmission and the 20-km-long SMF transmission with dispersion compensation. All BERs are below the \(7\%\) HD-FEC threshold after dispersion compensation. DC: dispersion compensation. g BERs versus received optical power for the 20-km-long SMF transmissions of the \(3^{\text{rd}}\) , \(6^{\text{th}}\) , and \(9^{\text{th}}\) channels with dispersion compensation using the CW-laser and the microcomb as the light source, respectively. </center>  
+
+<|ref|>text<|/ref|><|det|>[[147, 803, 852, 912]]<|/det|>
+Figure 3e shows the measured eye diagrams of the \(4^{\text{th}}\) , \(6^{\text{th}}\) , \(8^{\text{th}}\) , \(10^{\text{th}}\) , and \(12^{\text{th}}\) channels after transmission and dispersion compensation. See Supplementary Note 5 for the eye diagrams of all the 15 comb lines. All the 15 wavelength channels have good and comparable eye diagrams for the back- to- back (BtB) and for the 20- km- long SMF transmission. Figure 3f shows the measured BERs after dispersion compensation. Although a minor FSR difference between the comb lines and the channels of the dispersion compensator leads to a slightly higher BER level around the left and right
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[153, 90, 844, 547]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 550, 851, 693]]<|/det|>
+<center>Fig. 4. 112-Gbit/s DMT-based WDM transmission with microcombs. a Measured signal-to-noise-ratio (SNR) response of the transmission system after the DMT training under the BTB transmission. b Bit allocations for DMT transmission within 32-GHz bandwidth. c Measured spectrum of the 15 comb lines after 112-Gbit/s DMT modulation. d Measured total BERs for the \(8^{\text{th}}\) channel with and without the dispersion compensation. Inset: The enlarged view of BER for transmission with dispersion compensation. DC: dispersion compensation. e BERs for WDM transmissions with dispersion compensation. All BERs are within the \(20\%\) SD-FEC threshold after dispersion compensation. f Measured 16-QAM constellations for the \(4^{\text{th}}\) , \(6^{\text{th}}\) , \(8^{\text{th}}\) , \(10^{\text{th}}\) , and \(12^{\text{th}}\) comb line channels at the \(7^{\text{th}}\) subcarrier. </center>  
+
+<|ref|>text<|/ref|><|det|>[[147, 704, 851, 849]]<|/det|>
+sides of the wavelength span, all the BERs are below the \(7\%\) HD- FEC threshold, enabling a transmitted bit rate of 1.2 Tbit/s in total (1.12 Tbit/s net rate after FEC overhead subtraction). The BERs under different received optical powers for the CW- laser and the microcombs after dispersion compensation are illustrated in Fig. 3g. The \(3^{\text{rd}}\) , \(6^{\text{th}}\) , and \(9^{\text{th}}\) channels were chosen in the experiments, and the wavelengths of the CW- laser were kept the same as the ones of the selected comb lines. A trivial deterioration is observed between the BERs of the microcomb and the CW- laser transmission, demonstrating a comparable performance between these two light sources. The power penalty is \(\sim 1\) dB at the \(7\%\) HD- FEC threshold.  
+
+<|ref|>text<|/ref|><|det|>[[148, 852, 850, 905]]<|/det|>
+To facilitate the next- generation data transmission rates, Common Electrical I/O (CEI)- 112G standard based on 112 Gbit/s data rate transmission has been established by Optical Internetworking Forum (OIF)57. To achieve a 112 Gbit/s data rate per lane, we implemented the DMT
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 85, 851, 343]]<|/det|>
+modulation format to increase the total data capacity. Compared with the PAM4 modulation format, DMT signals use a higher- order modulation format and can achieve a higher spectral efficiency for the microcomb- based WDM transmission system. The experimental setup of the DMT transmission is the same as that for the PAM4 transmission except that the DSP procedure is different, and the sampling rate of the arbitrary waveform generator (AWG) is set to be 64 GSa/s because of the limitation of the maximum length of the data sequence. The bandwidth of the dispersion compensator is still 50 GHz, and the DMT signal was mapped to a 32- GHz bandwidth with 160 subcarriers. The details of the experimental setup are discussed in Methods. Before the data transmission, training of the DMT is essential to obtain the bit allocation of each tone. The training was carried out at the \(9^{\text{th}}\) channel under the BtB condition. The signal- to- noise- ratio (SNR) response and the bit allocation were obtained, which are shown in Figs. 4a and 4b. The maximum bit allocation is 4, corresponding to the modulation format of 16 quadratic amplitude modulation (16- QAM). Each sub- carrier was modulated with its own allocated modulation format. Then, we applied the identical bit allocation to all the 15 comb lines to ensure a uniform DMT setting.  
+
+<|ref|>text<|/ref|><|det|>[[145, 345, 851, 601]]<|/det|>
+Figure 4c illustrates the spectrum of the 112- Gbit/s DMT- modulated comb lines. Figure 4f depicts the measured 16- QAM constellations of the \(7^{\text{th}}\) sub- carrier for the \(4^{\text{th}}\) , \(6^{\text{th}}\) , \(8^{\text{th}}\) , \(10^{\text{th}}\) , and \(12^{\text{th}}\) channels. See Supplementary Note 6 for the constellations of all the 15 comb lines. Figure 4d shows the measured BERs for the \(8^{\text{th}}\) channel with and without dispersion compensation. The inset illustrates the enlarged view of the BER for transmission with dispersion compensation. At the frequency around 13 GHz and 23 GHz, the BERs reach as high as 0.49 and 0.47 before the dispersion compensation, respectively. We attribute this to the FSF of the 20- km SMF. See Supplementary Note 6 for the measured \(S_{21}\) response of the 20- km SMF. The BERs can be suppressed below 0.01 after dispersion compensation. The BERs of all the 15 channels after dispersion compensation are all below the \(20\%\) SD- FEC threshold, as shown in Fig. 4e. No FFE was introduced in the DMT transmission. In this case, an aggregate data rate of 1.68 Tbit/s (1.46 Tbit/s net rate) was achieved. The maximum power consumption of our dispersion compensator is as low as 160 mW. See Supplementary Note 6 for the detailed calculation of the bit rate and Supplementary Note 3 for the power consumption.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 611, 236, 625]]<|/det|>
+## Discussion  
+
+<|ref|>text<|/ref|><|det|>[[147, 636, 851, 820]]<|/det|>
+To our knowledge, we present the first demonstration of on- chip parallel signal transmission and dispersion compensation using a single- soliton microcomb and a dispersion compensator based on MRRs. Our optical dispersion compensator outperforms electronic dispersion compensation approaches based on DSP algorithms, which necessitate individual processing for each wavelength channel. In contrast, our optical dispersion compensator supports parallel dispersion compensation capabilities, irrespective of the modulation format and of the number of transmission channels. Compared with the dispersion compensators enabled by Bragg gratings and cascaded MZIs, the MRR- based dispersion compensator features a compact footprint and a flexible tuning capability. Notably, we achieved this parallel dispersion compensation with a low power consumption below 160 mW, underscoring its energy- efficient nature.  
+
+<|ref|>text<|/ref|><|det|>[[147, 823, 851, 912]]<|/det|>
+Furthermore, under the assumption of a matched FSR between the microcomb and the dispersion compensator, we can in principle utilize up to 51 channels from the microcomb for data transmission, which encompasses a wavelength span ranging from 1540 nm to 1580 nm. To achieve a total bit rate exceeding 10 Tbit/s, the spacing between adjacent comb lines can be further reduced to 50 GHz. It is worth noting that we can integrate various components such as a \(\mathrm{Si}_3\mathrm{N}_4\) microcomb,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 86, 850, 138]]<|/det|>
+a silicon MRR- based dispersion compensator, MRR modulators, photodetectors, and optical demultiplexers onto a single chip. By accomplishing this, we can realize a fully integrated optical communications system tailored specifically for short- reach DCIs.  
+
+<|ref|>text<|/ref|><|det|>[[147, 141, 851, 360]]<|/det|>
+In the experiment, the single- soliton microcomb in the microresonator with anomalous dispersion demonstrates a conversion energy efficiency of approximately \(1\%\) . The power of most of the comb lines is below - 20 dBm (considering the coupling and link loss). Investigations highlighting the effectiveness of utilizing a double- microring structure have been proposed with a high conversion energy efficiency of \(55\%\) for single soliton generation \(^{58}\) , resulting in increased power levels for the majority of the comb lines to - 10 dBm, which reduces the gain required by the erbium- doped fiber amplifier (EDFA). Consequently, this leads to improved noise characteristics and facilitates higher bit rates for each carrier. To further enhance the overall performance of the system, it is plausible to minimize the fiber- to- fiber loss associated with the dispersion compensator by introducing low- loss and broadband edge couplers \(^{59 - 61}\) . Additionally, MRRs made of widened waveguides can be incorporated to mitigate on- chip losses. By implementing these strategies, we believe a higher number of microcomb channels can be effectively utilized for data transmission.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 371, 238, 385]]<|/det|>
+## Conclusion  
+
+<|ref|>text<|/ref|><|det|>[[147, 396, 851, 653]]<|/det|>
+We have presented parallel data transmission and dispersion compensation utilizing a \(\mathrm{Si}_3\mathrm{N}_4\) single- soliton microcomb and a silicon MRR- based dispersion compensator. With the full reconfigurability of the MRRs, our dispersion compensator achieves a continuously tunable dispersion compensation of up to \(40\mathrm{km}\) SMFs within a 32- GHz bandwidth, and a dispersion compensation of \(20\mathrm{- km}\) - long SMF within a 50- GHz bandwidth. We proposed a method for calibration of the MRR- based dispersion compensator, which can eliminate the thermal crosstalk and is applicable to other integrated photonic circuits. Our experimentation with the \(\mathrm{Si}_3\mathrm{N}_4\) single- soliton microcomb revealed the utilization of 15 comb lines for parallel data transmission, resulting in an aggregate total bit rate of 1.2 Tbit/s and of 1.68 Tbit/s using 80- Gbit/s PAM4 and 112- Gbit/s DMT modulation signals for 20- km SMF transmission, respectively. These demonstrations in utilizing the microcomb and the silicon photonic circuits enable a higher data transmission capacity, potentially lower costs, and power- efficient solutions for parallel data processing, especially in WDM- assisted IM- DD systems. Our findings represent a significant advancement toward the pragmatic applications of cost- effective and practical short- reach DCIs.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 664, 217, 678]]<|/det|>
+## Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 689, 816, 705]]<|/det|>
+## Design, fabrication, and packaging of the silicon photonic dispersion compensator chip  
+
+<|ref|>text<|/ref|><|det|>[[147, 714, 851, 858]]<|/det|>
+The coupling region of each microring comprises an MZI- based tunable coupler, which comprises two 50:50 multimode interferometers (MMIs), two \(165 - \mu \mathrm{m}\) - long straight arms, and waveguide connections between them. The feedback region of each microring comprises a \(335 - \mu \mathrm{m}\) - long straight waveguide and two \(10 - \mu \mathrm{m}\) - radius arc bends. The dimensions of the waveguide cross- section are \(500\mathrm{nm}\) (width) \(\times 220\mathrm{nm}\) (height). Two \(160 - \mu \mathrm{m}\) - long and \(2 - \mu \mathrm{m}\) - wide titanium microheaters, with a resistance of about \(1800\Omega\) , are integrated \(2 - \mu \mathrm{m}\) above the center of the \(335 - \mu \mathrm{m}\) - long straight waveguide and of the lower arm of the MZI coupler. Deep trenches are positioned beside each micro- heater to suppress the thermal crosstalk within the MRR.  
+
+<|ref|>text<|/ref|><|det|>[[147, 862, 850, 896]]<|/det|>
+The device was fabricated on a silicon- on- insulator (SOI) platform with a 220- nm- thick silicon top layer and a \(2 - \mu \mathrm{m}\) buried oxide (BOX) using electron- beam lithography. All the devices except for
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 85, 850, 176]]<|/det|>
+the grating couplers were fully etched to a depth of \(220 \text{nm}\) . A 70- nm partial etch was implemented on the grating coupler regions. The gold metal connections were deposited above the titanium microheaters for electrical contacts. The metal pads were wire- bonded to an external printed- circuit board (PCB). The chip and the PCB were placed on a metal shell, and a thermo- electric cooler (TEC) was placed beneath the chip to control the on- chip temperature.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 186, 633, 202]]<|/det|>
+## Design, fabrication, and packaging of the \(\text{Si}_3\text{N}_4\) microcomb chip  
+
+<|ref|>text<|/ref|><|det|>[[147, 211, 851, 376]]<|/det|>
+The \(\text{Si}_3\text{N}_4\) microresonator has waveguide cross- sectional dimensions of \(1800 \text{nm}\) (width) \(\times 800 \text{nm}\) (height), with the fundamental TE mode exhibiting anomalous dispersion. The radius of the microresonator is \(232 \mu \text{m}\) , resulting in an FSR of \(97.7 \text{GHz}\) . The ultra- low- loss waveguide was fabricated on an \(800 \text{nm}\) - thick \(\text{Si}_3\text{N}_4\) layer, deposited by low- pressure chemical vapor deposition (LPCVD) by Ligentec. The device was patterned using \(193 \text{nm}\) photolithography and the quality (Q) factor of the microring resonator is \(7 \times 10^5\) to support the generation of the soliton microcomb. The \(\text{Si}_3\text{N}_4\) chip was packaged with a pair of fiber arrays and the coupling loss is \(\sim 3 \text{dB/facet after being fixed by an ultra- violet (UV) curable adhesive. We utilized a TEC to control the on- chip temperature for the long- term single- soliton generation.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 386, 700, 402]]<|/det|>
+## Experimental settings and DSP flows of PAM4 parallel data transmission  
+
+<|ref|>text<|/ref|><|det|>[[147, 410, 851, 742]]<|/det|>
+For the 80- Gbit/s PAM4- based parallel data transmission, the generated comb lines were first amplified to a total power of \(20 \text{dBm}\) , and then they were sent into a programmable optical filter (Finisar Waveshaper 1000S) to select 15 comb lines and to ensure that the power of each comb line was almost identical to each other. Then, another EDFA was utilized to amplify the 15 comb lines to a total power of \(17.5 \text{dBm}\) . Next, the optical comb lines were simultaneously modulated by a commercial 40- GHz intensity modulator (iXblue MXAN- LN- 40) driven by an PAM4 signal generated from an AWG (Keysight 8199A). The modulated comb lines were sent into the \(20 \text{km SMF}\) , and the dispersion was then compensated for by our dispersion compensator. Then, one channel of the dispersion- compensated parallel signals was filtered out by a tunable narrow- bandwidth optical filter and was directly detected by a commercial 50- GHz photodetector (Finisar XPDV2320R). The received electrical signal was finally recorded by a real- time oscilloscope (RTO, Keysight Z592A), and a DSP algorithm, which is identical to the one utilized in the CW- laser transmission, was applied to process offline the received data. At the transmitter side, a high- speed PAM4 signal was transmitted using the root- raised cosine (RRC) filter with a roll- off factor of 0.01 to compress the signal bandwidth. The matched filter and the time- domain FFE were applied at the receiver side to obtain lower BERs. The instruments and their settings were consistent with those in the CW- laser transmission. The optical power before the photodetector was set to the same power to ensure consistency.  
+
+<|ref|>sub_title<|/ref|><|det|>[[147, 751, 672, 767]]<|/det|>
+## Experimental setup and DSP flows of DMT parallel data transmission  
+
+<|ref|>text<|/ref|><|det|>[[147, 777, 851, 905]]<|/det|>
+Extended Data Figure 1 shows the schematic experimental setup of the microcomb- based data transmission using the DMT modulation format. A high- speed pseudo- random binary sequence (PRBS) data stream was divided into several parallel low- speed data streams and was mapped to a QAM constellation map. The complex values corresponding to each point in the constellation map were converted to real numbers by an inverse fast Fourier transform (I- FFT). A total number of 160 sub- carriers was mapped to the 32- GHz frequency range. Finally, the signals were converted to a serial data stream after adding the cyclic prefix (CP) to compensate for the multipath delay
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 85, 851, 175]]<|/det|>
+broadening. At the receiver side, the photocurrent signal was sent to the RTO for data collection. The CP was first removed from the sampled data sequence. Next, the serial data was converted to parallel data and mapped to the QAM constellation map through a fast Fourier transform (FFT). The BERs of the full link were then calculated by comparing the measured data sequence with the transmitter data sequence.  
+
+<|ref|>image<|/ref|><|det|>[[150, 182, 852, 415]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[147, 419, 851, 471]]<|/det|>
+Extended Data Fig. 1 Experimental setup of the DMT parallel data transmission. P/S Conversion: parallel to serial conversion; S/P Conversion: serial to parallel conversion; FFT: fast Fourier transform; I- FFT: inverse fast Fourier transform; QAM: quadratic amplitude modulation; CP: cyclic prefix.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 483, 301, 497]]<|/det|>
+## Acknowledgements  
+
+<|ref|>text<|/ref|><|det|>[[147, 507, 850, 560]]<|/det|>
+This work was supported in part by the National Key Research and Development Program of China (2018YFB2201702), and the National Natural Science Foundation of China (62090052, 62135010, 62075128).  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 572, 310, 586]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[147, 596, 851, 705]]<|/det|>
+Y. Liu designed, simulated, and characterized the dispersion compensator. Y. Liu performed the experiments of the CW-laser-based 64-Gbit/s PAM4 data transmission. H. Zhang designed, simulated, and characterized the Kerr microcomb. Y. Liu and H. Zhang conceived the link architecture and performed the high-speed data-transmission experiments of the microcomb-based parallel signal processing. J. Liu conducted the offline DSP. All authors helped analyze the data. L. Lu, J. Du, Y. Liu, H. Zhang, and J. Liu prepared the manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 715, 292, 729]]<|/det|>
+## Conflict of interest  
+
+<|ref|>text<|/ref|><|det|>[[149, 740, 552, 755]]<|/det|>
+The authors declare that they have no conflict of interest.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 766, 237, 780]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[147, 792, 850, 825]]<|/det|>
+1. Forbes. 175 Zettabytes by 2025 https://www.forbescom/sites/tomcoughlin/2018/11/27/175-zettabytes-by-2025/ (2018).  
+
+<|ref|>text<|/ref|><|det|>[[147, 829, 850, 862]]<|/det|>
+2. Chopra R. Looking beyond 400G - A system vendor perspective. Cisco Systems Inc https://www.ethernetalliance.org/wp-content/uploads/2021/02/TEF21Day1_KeynoteRChopra.pdf, (2020).  
+
+<|ref|>text<|/ref|><|det|>[[147, 866, 850, 899]]<|/det|>
+3. Pang XD, et al. 200 Gbps/Lane IM/DD Technologies for Short Reach Optical Interconnects. Journal of Lightwave Technology 38, 492-503 (2020).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[90, 80, 854, 900]]<|/det|>
+447 4. Alam MS, et al. Net 220 Gbps/A IM/DD Transmission in O-Band and C-Band With Silicon Photonic Traveling-Wave MZM. Journal of Lightwave Technology 39, 4270-4278 (2021). 449 5. Zhang J, et al. Receive-diversity-aided power-fading compensation for C-band IM/DD OFDM systems. Opt Lett 48, 2237-2240 (2023). 451 6. Zou D, Li F, Wang W, Li Z, Li Z. Amplifier-less transmission of beyond 100-Gbit/s/A signal for 40-km DCI-Edge with 10G-class O-band DML. Journal of Lightwave Technology 38, 5649-5655 (2020). 453 7. Wu X, Karar AS, Zhong K, Lau APT, Lu C. Experimental demonstration of pre-electronic dispersion compensation in IM/DD systems using an iterative algorithm. Opt Express 29, 24735-24749 (2021). 455 8. Karar AS. Iterative Algorithm for Electronic Dispersion Compensation in IM/DD Systems. Journal of Lightwave Technology 38, 698-704 (2020). 457 9. Zou DD, Li F, Wang W, Yin MZ, Sui Q, Li ZH. Modified Gerchberg-Saxton Algorithm Based Electrical Dispersion Pre-Compensation for Intensity-modulation and Direct-detection Systems. Journal of Lightwave Technology 40, 2840-2849 (2022). 460 10. Zou DD, et al. 100G PAM-6 and PAM-8 Signal Transmission Enabled by Pre-Chirping for 10-km Intra-DCI Utilizing MZM in C-band. Journal of Lightwave Technology 38, 3445-3453 (2020). 462 11. Yin M, Zou D, Wang W, Li F, Li Z. Transmission of a 56 Gbit/s PAM4 signal with low-resolution DAC and pre- equalization only over 80 km fiber in C-band IM/DD systems for optical interconnects. Opt Lett 46, 5615-5618 (2021). 465 12. Zhong KP, Zhou X, Huo JH, Yu CY, Lu C, Lau APT. Digital Signal Processing for Short-Reach Optical Communications: A Review of Current Technologies and Future Trends. Journal of Lightwave Technology 36, 377-400 (2018). 468 13. Wu X, Zhang J, Lau APT, Lu C. C-band 100-GBand PS-PAM-4 transmission over 50-km SSMF enabled by FIR-filter-based pre-electronic dispersion compensation. Opt Express 31, 17759-17768 (2023). 470 14. Kilper DC, Rastegarfar H. Energy challenges in optical access and aggregation networks. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 374, 20140435 (2016). 472 15. Yang KY, et al. Bridging ultrahigh-Q devices and photonic circuits. Nature Photonics 12, 297-303 (2018). 473 16. Brasch V, et al. Photonic chip-based optical frequency comb using soliton Cherenkov radiation. Science 351, 357-360 (2016). 475 17. Guo H, et al. Universal dynamics and deterministic switching of dissipative Kerr solitons in optical microresonators. Nature Physics 13, 94-102 (2016). 477 18. Herr T, et al. Temporal solitons in optical microresonators. Nature Photonics 8, 145-152 (2013). 478 19. Liu JQ, et al. Ultralow-power chip-based soliton microcombs for photonic integration. Optica 5, 1347-1353 (2018). 479 20. Shen B, et al. Integrated turnkey soliton microcombs. Nature 582, 365-369 (2020). 480 21. Stern B, Ji X, Okawachi Y, Gaeta AL, Lipson M. Battery-operated integrated frequency comb generator. Nature 562, 401-405 (2018). 482 22. Yi X, Yang QF, Yang KY, Suh MG, Vahala K. Soliton frequency comb at microwave rates in a high-Q silica microresonator. Optica 2, 1078-1085 (2015). 484 23. Zhang SY, et al. Sub-milliwatt-level microresonator solitons with extended access range using an auxiliary laser. Optica 6, 206-212 (2019). 486 24. Zhou H, et al. Soliton bursts and deterministic dissipative Kerr soliton generation in auxiliary-assisted microcavities. Light: Science & Application 8, 50 (2019). 488 25. Marin-Palomo P, et al. Microresonator-based solitons for massively parallel coherent optical communications. Nature 546, 274-279 (2017). 490 26. Corcoran B, et al. Ultra-dense optical data transmission over standard fibre with a single chip source. Nature
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[90, 88, 854, 900]]<|/det|>
+491 Communications 11, 2568 (2020).492 27. Jørgensen AA, et al. Petabit-per-second data transmission using a chip-scale microcomb ring resonator source. 493 Nature Photonics 16, 798- 802 (2022).494 28. Geng Y, et al. Coherent optical communications using coherence-cloned Kerr soliton microcombs. Nature 495 Communications 13, 1070 (2022).496 29. Mazur M, et al. High Spectral Efficiency Coherent Superchannel Transmission With Soliton Microcombs. Journal 497 of Lightwave Technology 39, 4367- 4373 (2021).498 30. Fulop A, et al. High- order coherent communications using mode- locked dark- pulse Kerr combs from 499 microresonators. Nature Communications 9, 1598 (2018).500 31. Rizzo A, et al. Massively scalable Kerr comb- driven silicon photonic link. Nature Photonics, (2023).501 32. Shu H, et al. Microcomb- driven silicon photonic systems. Nature 605, 457- 463 (2022).502 33. Liu SJ, Zhang CP, Cao HY, Hong SH, Yu ZJ, Dai DX. On- Chip Circulator- Free Chirped Spiral Multimode Waveguide 503 Grating for Dispersion Management. ACS Photonics 10, 1654- 1661 (2023).504 34. Kaushal S, Roberge A, Kashyap R, Azana J. Ultra- compact silicon photonics highly dispersive elements for low- 505 latency signal processing. Opt Express 31, 3467- 3478 (2023).506 35. Chen GF, Ong KYK, Chia XX, Cao Y, Tan DTH. Dispersion Compensation of 30GBaud/s NRZ and PAM4 data using 507 integrated Silicon Nitride Gratings. In: Optical Fiber Communications Conference and Exhibition (OFC), paper 508 Th1D.4 (2022).509 36. Liu SJ, Liu DJ, Yu ZJ, Liu L, Shi YC, Dai DX. Digitally tunable dispersion controller using chirped multimode 510 waveguide gratings. Optica 10, 316- 323 (2023).511 37. Giuntoni I, et al. Integrated Drop- Filter for Dispersion Compensation Based on SOI Rib Waveguides. Conference 512 on Optical Fiber Communication and Collocated National Fiber Optic Engineers Conference (OFC- NFOEC), paper 513 OTjh5, (2010).514 38. Moreira R, Gundavarapu S, Blumenthal DJ. Programmable eye- opener lattice filter for multi- channel dispersion 515 compensation using an integrated compact low- loss silicon nitride platform. Opt Express 24, 16732- 16742 (2016).516 39. Brodnik GM, Pinho C, Chang F, Blumenthal DJ. Extended Reach 40km Transmission of C- Band Real- Time 53.125 517 Gbps PAM- 4 Enabled with a Photonic Integrated Tunable Lattice Filter Dispersion Compensator. In: Optical Fiber 518 Communications Conference and Exposition (OFC), paper W2A.30 (2018).519 40. Madsen CK, et al. An All- Pass Filter Dispersion Compensator Using Planar Waveguide Ring Resonators. In: Optical 520 Fiber Communication Conference and the International Conference on Integrated Optics and Optical Fiber 521 Communication, paper FE6 (1999).522 41. Madsen CK, Lenz G, Bruce AJ, Cappuzzo MA, Gomez LT, Scotti RE. Integrated all- pass filters for tunable 523 dispersion and dispersion slope compensation. IEEE Photonics Technology Letters 11, 1623- 1625 (1999).524 42. Notaros J, et al. Programmable dispersion on a photonic integrated circuit for classical and quantum applications. 525 Opt Express 25, 21275- 21285 (2017).526 43. Sorianello V, et al. 100Gb/s PolMux- NRZ Transmission at 1550nm over 30km Single Mode Fiber Enabled by a 527 Silicon Photonics Optical Dispersion Compensator. In: Optical Fiber Communication Conference (OFC), paper 528 W2A.31 (2018).529 44. Alam MS, et al. 224 Gb/s Transmission over 10 km of SMF at 1550 nm Enabled by a SiN Optical Dispersion 530 Compensator and Stokes Vector Direct Detect Receiver. In: OSA Advanced Photonics Congress (APC), paper 531 SpM4I.5 (2020).532 45. Liu Y, Lu L, Ni Z, Chen J, Chen J, Zhou L. Silicon Integrated Continuously Tunable Dispersion Compensator Based 533 on Cascaded Micro- Ring Resonators. In: 2022 Asia Communications and Photonics Conference (ACP) (2022).534 46. Ong KYK, Chen GFR, Xing P, Gao H, Tan DTH. Dispersion compensation of high- speed data using an integrated
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[90, 87, 855, 678]]<|/det|>
+535 silicon nitride ring resonator. Opt Express 30, 13959- 13967 (2022).536 47. Zhang Y, et al. 240 Gb/s optical transmission based on an ultrafast silicon microring modulator. Photon Res 10, 1127- 1133 (2022).537 48. Dong P. Silicon Photonic Integrated Circuits for Wavelength- Division Multiplexing Applications. IEEE Journal of Selected Topics in Quantum Electronics 22, 370- 378 (2016).538 49. Liu D, Zhang L, Tan Y, Dai D. High- Order Adiabatic Elliptical- Microring Filter with an Ultra- Large Free- Spectral- Range. Journal of Lightwave Technology 39, 5910- 5916 (2021).539 50. Liu DJ, He JH, Xiang YL, Xu Y, Dai DX. High- performance silicon photonic filters based on all- passive tenth- order adiabatic elliptical- microrings. Apl Photonics 7, 051303 (2022).540 51. Horst F, Green WMJ, Assefa S, Shank SM, Vlasov YA, Offrein BJ. Cascaded Mach- Zehnder wavelength filters in silicon photonics for low loss and flat pass- band WDM (de- )multiplexing. Opt Express 21, 11652- 11658 (2013).541 52. Jeong SH, Shimura D, Simoyama T, Horikawa T, Tanaka Y, Morito K. Si- nanowire- based multistage delayed Mach- Zehnder interferometer optical MUX/DeMUX fabricated by an ArF- immersion lithography process on a 300 mm SOI wafer. Opt Lett 39, 3702- 3705 (2014).542 53. Sun Q, Zhou L, Lu L, Zhou G, Chen J. Reconfigurable High- Resolution Microwave Photonic Filter Based on Dual- Ring- Assisted MZIs on the Si3N4 Platform. IEEE Photonics Journal 10, 1- 12 (2018).543 54. Kohli N, Sang BL, Nabki F, Menard M. Tunable Bandpass Filter With Serially Coupled Ring Resonators Assisted MZI. IEEE Photonics Journal 13, 1- 8 (2021).544 55. Fujii S, Tanabe T. Dispersion engineering and measurement of whispering gallery mode microresonator for Kerr frequency comb generation. Nanophotonics 9, 1087- 1104 (2020).545 56. Zhang HY, Lu LJ, Chen JP, Zhou LJ. Program- Controlled Single Soliton Generation Driven by the Thermal- Compensated Avoided Mode Crossing. Journal of Lightwave Technology 41, 1801- 1810 (2023).546 57. OIF. Common Electrical I/O (CEI) - Electrical and Jitter Interoperability agreements for 6G+ bps, 11G+ bps, 25G+ bps, 56G+ bps and 112G+ bps I/O. (2022).547 58. Helgason oB, Girardi M, Ye Z, Lei F, Schroder J, Torres- Company V. Power- efficient soliton microcombs. arXiv:202209410, (2022).548 59. He A, Guo X, Wang K, Zhang Y, Su Y. Low Loss, Large Bandwidth Fiber- Chip Edge Couplers Based on Silicon- on- Insulator Platform. Journal of Lightwave Technology 38, 4780- 4786 (2020).549 60. Liu W, Zhang J, Liu L, Dai D, Shi Y. High Efficiency Silicon Edge Coupler Based On Uniform Arrayed Waveguides With Un- Patterned Cladding. IEEE Photonics Technology Letters 32, 1077- 1080 (2020).550 61. Marchetti R, Lacava C, Carroll L, Gradkowski K, Minzioni P. Coupling strategies for silicon photonics integrated chips [Invited]. Photon Res 7, 201- 239 (2019).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 311, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 353, 150]]<|/det|>
+SupplementaryInformation.pdf
+
+<--- Page Split --->

preprint/preprint__48b464b07d1a9fb2491f3443e735e9f38e6b8abe90e64efd27de8474ea209aaf/images_list.json

@@ -0,0 +1,245 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 21
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 21
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 23
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_0.jpg",
+    "caption": "Figure S1",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 23
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_1.jpg",
+    "caption": "D. Against WT Leukemia",
+    "footnote": [],
+    "bbox": [
+      [
+        62,
+        300,
+        940,
+        400
+      ]
+    ],
+    "page_idx": 25
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_2.jpg",
+    "caption": "H.",
+    "footnote": [],
+    "bbox": [
+      [
+        88,
+        500,
+        910,
+        600
+      ]
+    ],
+    "page_idx": 27
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_3.jpg",
+    "caption": "Antigen Density",
+    "footnote": [],
+    "bbox": [
+      [
+        88,
+        612,
+        910,
+        712
+      ]
+    ],
+    "page_idx": 27
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_4.jpg",
+    "caption": "Antigen Density",
+    "footnote": [],
+    "bbox": [
+      [
+        88,
+        725,
+        910,
+        825
+      ]
+    ],
+    "page_idx": 27
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_5.jpg",
+    "caption": "Figure S3: CAR T cells and leukemia counts per tibia for in vivo data (Related to Figures 2 & 3)",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 29
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_6.jpg",
+    "caption": "A.",
+    "footnote": [],
+    "bbox": [
+      [
+        97,
+        99,
+        700,
+        256
+      ]
+    ],
+    "page_idx": 31
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_7.jpg",
+    "caption": "B.",
+    "footnote": [],
+    "bbox": [
+      [
+        97,
+        303,
+        912,
+        410
+      ]
+    ],
+    "page_idx": 33
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_8.jpg",
+    "caption": "Day 11",
+    "footnote": [],
+    "bbox": [
+      [
+        97,
+        425,
+        912,
+        531
+      ]
+    ],
+    "page_idx": 33
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_9.jpg",
+    "caption": "D.",
+    "footnote": [],
+    "bbox": [
+      [
+        97,
+        560,
+        912,
+        666
+      ]
+    ],
+    "page_idx": 33
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_10.jpg",
+    "caption": "Day 11",
+    "footnote": [],
+    "bbox": [
+      [
+        97,
+        692,
+        912,
+        800
+      ]
+    ],
+    "page_idx": 33
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_11.jpg",
+    "caption": "Figure S6",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 33
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_12.jpg",
+    "caption": "Comparisons of Naive T cell derived populations",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        312,
+        530,
+        420
+      ]
+    ],
+    "page_idx": 35
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_13.jpg",
+    "caption": "Comparisons of Memory T cell derived populations",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        449,
+        530,
+        556
+      ]
+    ],
+    "page_idx": 37
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_14.jpg",
+    "caption": "Comparisons between Naive and Memory T cell derived populations",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        585,
+        530,
+        693
+      ]
+    ],
+    "page_idx": 37
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_15.jpg",
+    "caption": "D.",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        725,
+        590,
+        940
+      ]
+    ],
+    "page_idx": 38
+  }
+]

preprint/preprint__48b464b07d1a9fb2491f3443e735e9f38e6b8abe90e64efd27de8474ea209aaf/preprint__48b464b07d1a9fb2491f3443e735e9f38e6b8abe90e64efd27de8474ea209aaf.mmd

@@ -0,0 +1,577 @@
+
+# Antigen experience history directs distinct functional states of CD8+ CAR T cells during the anti-leukemia response  
+
+Terry Fry  TERRY.FRY@CUANSCHUTZ.EDU  
+
+University of Colorado Anschutz Medical Campus  Kole DeGolier  
+
+University of Colorado Anschutz Medical Campus https://orcid.org/0000- 0002- 4144- 7277  
+
+Etienne Danis  University of Colorado Anschutz Medical Campus  
+
+Marc D'Antonio  University of Colorado School  
+
+Jennifer Cimons  University of Colorado Anschutz Medical Campus https://orcid.org/0000- 0002- 8111- 8697  
+
+Michael Yarnell  University of Colorado Anschutz Medical Campus https://orcid.org/0000- 0002- 4557- 7255  
+
+Ross Keld  University of Colorado Anschutz Medical Campus https://orcid.org/0000- 0002- 9065- 7895  
+
+Mark kohler  University of Colorado School of Medicine  
+
+James Scott- Browne  National Jewish Health  
+
+## Article  
+
+Keywords:  
+
+Posted Date: December 21st, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 3712137/v1  
+
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License.  Read Full License
+
+<--- Page Split --->
+
+Additional Declarations: Yes there is potential Competing Interest. Patent applicant (whether author or institution): University of Colorado Anschutz Medical Campus Name of inventor(s): Kole R DeGolier, James P Scott- Browne, Terry J Fry Application number: U.S. Provisional Application No. 63/595,612 Status of application: Pending Aspect covered: Methods of enhancing potency of engineered immune cells via Runx2 modulation  
+
+Version of Record: A version of this preprint was published at Nature Immunology on January 2nd, 2025. See the published version at https://doi.org/10.1038/s41590-024-02034-1.
+
+<--- Page Split --->
+
+# Antigen experience history directs distinct functional states of CD8+ CAR T cells during the anti-leukemia response  
+
+Kole R. DeGolier \(^{1,2}\) , Etienne Danis \(^{3}\) , Marc D'Antonio \(^{1}\) , Jennifer Cimons \(^{1,2}\) , Michael Yarnell \(^{2,4}\) , Ross M. Kedl \(^{1}\) , M. Eric Kohler \(^{1,2,4}\) , James P. Scott- Browne \(^{1,5}\) , Terry J. Fry \(^{1,2,4}\)  
+
+\(^{1}\) Department of Immunology, University of Colorado Anschutz Medical Campus; Aurora, CO, USA \(^{2}\) Department of Pediatrics, University of Colorado Anschutz Medical Campus; Aurora, CO, USA \(^{3}\) Biostatistics and Bioinformatics Shared Resource, University of Colorado Cancer Center, University of Colorado Anschutz Medical Campus; Aurora, CO, USA \(^{4}\) Center for Cancer and Blood Disorders, Children's Hospital Colorado; Aurora, CO, USA \(^{5}\) Department of Immunology and Genomic Medicine, National Jewish Health, Denver, CO, USA  
+
+Corresponding author:  
+
+Terry J Fry, MD  University of Colorado Anschutz Medical Campus and Children's Hospital Colorado  13123 E 16 \(^{\text{th}}\) Avenue  Aurora, Colorado 80045  Phone: 303.724.7293  Email: terry.fry@cuanschutz.edu  
+
+Running title: Antigen experience history directs distinct functional states of CD8+ CAR T cells during the anti- leukemia response  
+
+Nature Immunology  
+
+Abstract word count: 200 (200)  Manuscript word count: 4147 (4300)  Tables and Figure count: 6 figures, no tables (5- 6 modest length (1/4 page))  Extended Data: 9 figures, no tables (10)  References: 39 (50)  Methods References: 12
+
+<--- Page Split --->
+
+## ABSTRACT  
+
+Chimeric antigen receptor T cells are an effective therapy for B- lineage malignancies. However, many patients relapse and this therapeutic has yet to show strong efficacy in other hematologic or solid tumors. One opportunity for improvement lies in the ability to generate T cells with desirable functional characteristics. Here, we dissect the biology of CD8+ CAR T cells (CAR8) by controlling whether the T cell has encountered cognate TCR antigen prior to CAR generation. We find that prior antigen experience influences multiple aspects of in vitro and in vivo CAR8 functionality, resulting in superior effector function and leukemia clearance in the setting of limiting target antigen density compared to antigen- inexperienced T cells. However, this comes at the expense of inferior proliferative capacity, susceptibility to phenotypic exhaustion and dysfunction, and inability to clear wildtype leukemia in the setting of limiting CAR+ cell dose. Epigenomic and transcriptomic comparisons of these cell populations identified overexpression of the Runx2 transcription factor as a novel strategy to enhance CAR8 function, with a differential impact depending on prior cell state. Collectively, our data demonstrate that prior antigen experience determines functional attributes of a CAR T cell, as well as amenability to functional enhancement by transcription factor modulation.
+
+<--- Page Split --->
+
+relapsed and treatment- refractory B- lineage hematologic malignancies. However, many patients do not achieve complete remission, or relapse. Poor response or lack of remission durability results from cancer cell resistance or suboptimal CAR T cell function'. Thus, further studies into the immunobiology of these engineered cells are warranted to enhance remissions and expand therapeutic potential to other hematologic and solid tumors. CAR T cells are commonly generated from a heterogeneous population of peripheral blood T cells that varies between patients, likely impacting the quality of a CAR T cell product?. Although it has been difficult to track cell fate through the manufacturing process and into patients, previous reports have shown differential function of CAR T cell products generated from memory versus naive T cells sorted by surface marker phenotypes, which are not always an accurate representation of cellular differentiation state2,3,4. Emerging studies have demonstrated that phenotypic, transcriptomic and epigenomic attributes of the CAR product can influence patient outcomes5. During acute infections, naive CD8+ T cells become activated through the T cell antigen receptor (TCR) by antigen presenting cells displaying cognate antigen and co- stimulatory ligands, and subsequently enter a highly regulated differentiation trajectory. A phase of rapid expansion and differentiation into effector cells is followed by contraction and formation of long- lived memory cells that rapidly respond to future exposures. However, if the pathogen is not cleared, antigen- specific T cell populations will receive recurring antigen stimulation. In this setting, rather than forming functional memory, T cells differentiate down a trajectory characterized by progressive dysfunction, preventing immune- mediated pathology, but simultaneously failing to clear the challenge. A growing body of work demonstrates that these differentiation trajectories (and resulting functional characteristics imbued on T cells) are controlled epigenetically in traditional T cell responses to viral infections and tumors. These programs are defined by progressive changes to the epigenome, associated with DNA methylation and histone modifications which are driven by a variety of transcription factors (TFs) and modulated by antigen receptor signaling6. These molecular modifications alter chromatin accessibility and transcriptional profiles which characterize cellular differentiation state and functional capacity. Epigenetic modulation of T cells via stimulation through the physiologic TCR has a well- established role in impacting the differentiation program and functional capacity of a pool of antigen- experienced T cells7. Emerging data also highlight the importance of epigenetic remodeling in CAR T cell responses to tumors5.  
+
+Here, we carefully examine and compare the biology of CAR- transduced CD8+ T cells that differ as to whether cognate antigen has been encountered through the TCR prior to transduction with a CAR. We hypothesize that 1) T cells exhibit functional characteristics after CAR transduction that are dictated by prior antigen experience via the TCR 2)the functional characteristics of CAR8 derived from naive or memory cells are the result of epigenetic attributes maintained through CAR transduction and reinfusion, and that 3) TF modulation as a modality to enhance CAR8 function may be dependent on the epigenetic and transcriptomic contexts determined by prior antigen experience status. Prior work has shown dose- dependent effects in the anti- tumor responses of adoptively- transferred T cells2 and CAR T cells have been shown to elicit poor responses to tumors with low antigen density1,8,9,10. Using limiting target antigen density or limiting T cell dose as stressors, we show that prior T cell antigen experience influences in vitro and in vivo functional characteristics of T cells stimulated through a CAR. Comparison of the epigenetic and transcriptomic states of CAR8 stratified by prior antigen- experience
+
+<--- Page Split --->
+
+status revealed differential chromatin accessibility and transcriptional programming. We pinpoint divergent RUNX2 activity within the two populations as a potential driver of differential function and show that ectopic expression of RUNX2 enhances the anti- leukemia response and mediates exhaustion resistance in CAR T cells in a manner dependent on prior T cell antigen experience status.  
+
+## RESULTS  
+
+T cell antigen experience prior to transduction with a CAR directs in vitro proliferative and effector capacities of CD8+ CAR T cells.  
+
+Memory T cells demonstrate superior antigen sensitivity compared to naive T cells in some contexts<sup>11,12</sup>. Thus, we hypothesized that CAR T cells derived from a memory T cell population would exhibit enhanced responsiveness to low antigen density leukemias compared to naive- derived CAR T cells. T cells expressing a CAR containing an anti- mouse CD19 scFv incorporating a FLAG sequence and a CD28 costimulatory domain fused to mouse CD3zeta, followed by a 2A sequence and a truncated EGFR<sup>13</sup> (Figure S1A) were used to target a murine leukemia driven by the E2A- PBX1 fusion protein (E2A- PBX)<sup>14, 15, 16</sup>. FLAG- specific antibody detection of the CAR correlated strongly with EGFR expression, allowing for use of EGFR as a marker for long term tracking of CAR+ cells in vivo (Figure S1B). We expanded this model by generating a set of clones of E2A- PBX which express differing CD19 densities (Figure 1A, S1C). Memory OT- I T cells generated using a well- characterized ovalbumin vaccination model<sup>17, 18, 19</sup> (Figure 1B). were used to produce memory- derived CD8+ CAR T cells (CAR<sub>MD</sub>) for comparison to naive- derived CD8+ OT- I CAR T cells (CAR<sub>ND</sub>). As no difference was seen in leukemia control by memory or naive- derived control T cells (Figure S1D), we used naive- derived (EGFR8) in all subsequent experiments. A functional duality began to emerge upon in vitro testing. As predicted, a greater proportion of CAR<sub>MD</sub> cells had a polyfunctional effector profile, producing both TNFa and IFNg, or degranulating (as measured by CD107a), most pronounced in response to low target antigen (Figure 1C- H; S1E- G). Interestingly, while the proportion of IFNg+ cells was greater in CAR<sub>MD</sub>, the proportion of TNFa+ cells was slightly increased in CAR<sub>ND</sub>, suggesting a predisposition toward either IFNg or TNFa (Figure 1C & F). However, CAR<sub>ND</sub> outperformed CAR<sub>MD</sub> in cell cycle entry (Ki67 expression; Figure 1I, S1H) and extended proliferative capacity (Figure 1J, S1I) across antigen densities. To compare polyclonal antigen- experienced and naive T cells more analogous to human CAR T cells, we generated pathogen- elicited polyclonal T cells by infecting WT C57BL/6 mice with the common acute viral infection model LCMV- Armstrong. Memory (CD8+/CD44+/CD49d<sup>hi</sup>) and naive (CD8+/CD44-/CD49d<sup>lo</sup>/CD62L+) T cell populations were FACS- sorted from the same mice 28 days after LCMV infection and used for CAR T cell manufacturing (Figure S2A). Polyclonal pathogen- elicited T cells behaved similarly in vitro to memory and naive OT- I cells: CAR<sub>MD</sub> demonstrated superior effector function (increased proportions of cells producing IFNg) and CAR<sub>ND</sub> demonstrated superior proliferative capacity (Figure S2B- E). Thus, CD8+ T cell antigen experience prior to transduction with a CAR promotes effector functions at the expense of proliferative capacity.  
+
+Treatment of leukemia- bearing mice with a high CAR+ cell dose reveals enhanced cytotoxic profile and clearance of antigen- low leukemia by memory- derived CAR8.
+
+<--- Page Split --->
+
+Given the opposing functional profiles of naive and memory- derived CAR8, we next compared the ability of these two populations to mediate tumor clearance in vivo. Mice were engrafted with WT (35,000 antigens per cell), CD19<sup>lo</sup> (10,000 antigens per cell) or CD19<sup>Neg</sup> leukemia followed 3 days later by a dose of 1e6 CAR T cells. The CD19<sup>lo</sup> clone antigen density was chosen based on differential in vitro responses and, although higher than antigen density reported for CAR relapses post- CD22 CAR treatment<sup>9</sup>, is consistent with the drop- off in CAR sensitivity against other antigens<sup>8, 10</sup>. Rag1- deficient hosts enabled CAR T cell expansion without irradiation and limited CAR T cell antigen exposure to CD19 densities expressed on leukemia rather than endogenous B cells. While we did not observe differences in proportions of CAR T cells in the marrow at peak expansion on day 4 (Figure 2A), post- contraction (day 11) CAR8<sub>ND</sub> had increased proportions and total counts of CAR T cells in mice bearing WT and CD19<sup>lo</sup> leukemia (Figure 2B- C, Figure S3A- B). Both CAR groups mediated robust clearance of WT leukemia by day 11. Although there was no significant difference in clearance of CD19<sup>lo</sup> leukemia, 4/10 mice treated by CAR8<sub>ND</sub> had detectable leukemia at \(>15\%\) of live bone marrow cells while all 10 mice treated with CAR8<sub>MD</sub> had minimal leukemic burdens (<5%) (Figure 2D). We next tested whether the enhanced clearance of CD19<sup>lo</sup> leukemia was associated with maintenance of the superior cytotoxic capacity of CAR8<sub>MD</sub> observed in vitro. Upon ex vivo restitution of CAR8 in the bone marrow, we found that, while IFNg production was highly variable, GZMB production was markedly greater in CAR8<sub>MD</sub> (Figure 2E- F). CAR8<sub>MD</sub> had significantly higher proportions of cells falling into short- lived effector cell (SLEC, IL7Ra-/KLRG1+) and effector memory precursor (EMP, CD27+/CD62L-) phenotypes, fewer cells in the central memory precursor phenotype (CMP, CD27+/CD62L+), and no change in memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) populations (Figure S4B- E). Additionally, early expression of effector- associated TFs IRF4, T- bet and EOMES was greater in CAR8<sub>MD</sub> (Figure 2G- I). Finally, while mice bearing WT high- antigen leukemia showed no survival difference after treatment with CAR8<sub>MD</sub> versus CAR8<sub>ND</sub>, mice bearing CD19<sup>lo</sup> leukemia treated with CAR8<sub>MD</sub> showed a significant survival benefit, with 20% of mice surviving to the 80 day experimental endpoint (Figure 2J). Together, these data show that CAR8<sub>MD</sub> mediate superior clearance of CD19<sup>lo</sup> leukemia relative to CAR8<sub>ND</sub>, associated with maintenance of effector function and expression of effector- associated markers.  
+
+## Treatment of leukemia-bearing mice with a low CAR+ cell dose reveals enhanced proliferative capacity and clearance of WT leukemia by naive-derived CAR8.  
+
+We next hypothesized that the benefit of enhanced proliferative capacity of naive- derived CAR8 would emerge at a lower CAR+ cell dose (3e5). As anticipated, CAR8<sub>ND</sub> expanded to significantly higher numbers in the bone marrow by day 4 regardless of leukemia antigen density, mirroring in vitro proliferative assays (Figure 3A- B, S3C, 11- J). While CAR8<sub>ND</sub> mediated enhanced clearance and survival in mice bearing WT leukemia, there was no improvement in leukemia clearance or survival of mice bearing CD19<sup>lo</sup> leukemia (Figure 3C, 3I, S3D), potentially due to reduced potency. Indeed, ex vivo IFNg production was greater in CAR8<sub>MD</sub>, although there was no difference in GZMB production or expression of IRF4, T- bet or EOMES (Figure 3D- H). CAR8<sub>MD</sub> consistently demonstrated significantly higher proportions of SLECs at the early timepoint consistent with high CAR doses, but these differences disappeared by day 11 and no differences were seen in the MPEC population (Figure S5A,B). While EMP and CMP patterns mimicked high dose experiments, the differences were much less
+
+<--- Page Split --->
+
+pronounced, indicating that naive- derived cells largely became more "effector- like" with greater proliferative drive (Figure S5C,D), consistent with effector- polarization in the setting of low numbers of antigen- specific precursor populations \(^{20,21}\) . However, these changes, combined with the strong expansion, did not mediate survival benefit against CD19 \(^{1,0}\) leukemia (Figure 3l). Finally, we predicted that at this lower cell dose, T cell dysfunction could emerge. Indeed, CAR \(_{8\mathrm{MD}}\) expressed higher levels of exhaustion- associated markers against WT leukemia with failure of CAR \(_{8\mathrm{MD}}\) to control leukemia (Figure S5E- F, I- J). Interestingly, we found that CD19 \(^{1,0}\) leukemia drove similar proportions of exhaustion phenotypes in both CAR8 populations, demonstrating that chronic, uncleared antigen exposure, even at low antigen density, can drive dysfunction (Figure S5G- H, K- L). These findings highlight the importance of proliferative capacity and resistance to dysfunction afforded by CAR \(_{8\mathrm{ND}}\) at limiting cell dose.  
+
+## Epigenetic profiling of naive and memory-derived CAR8 shows differential chromatin accessibility at binding sites for bZIP, Tcf, Runx and other TF families.  
+
+We predicted that functional traits were a product of distinct epigenetic states, given that functional distinctions of naive and memory- derived CAR8 were dictated by status prior to CAR transduction. To test this, we performed bulk ATAC- seq on naive and memory- derived cells at three timepoints: ex vivo prior to CAR transduction (Day - 5, "PreCAR"), in vitro after CAR transduction (Day 0, "PostCAR"), and after reinfusion into mice bearing CD19 \(^{1,0}\) leukemia (Day 4, "Tumor") (Figure 4A). Comparison of experimental replicates showed tight concordance of chromatin accessibility at each condition and timepoint (Figure S6A). Broadly, the data showed several thousand differentially accessible regions between either cell type compared to itself across timepoints, and between naive and memory- derived cells at each timepoint (Figure S6B). We found predictable patterns of ATAC- seq signal at genetic loci involved in T cell activation or effector function, including higher accessibility in CAR \(_{8\mathrm{MD}}\) at Gzmb, Gzmc, and the Pdcd1 loci encoding for the PD1 protein. Concurrently, we found greater accessibility in CAR \(_{8\mathrm{ND}}\) at the Tcf7 loci encoding TCF1, a TF important for maintaining self- renewal capacity (Figure 4B). We used ChromVAR \(^{22}\) , to associate these changes in chromatin accessibility to previously defined datasets and potential TF activities. Based on relative chromatin accessibility at regions that were differentially accessible in a published comparison of effector and memory CD8+ T cells after acute viral infection with LCMV- Armstrong \(^{23}\) , memory- derived CAR8 acquired effector- associated changes in chromatin accessibility during CAR generation in culture that were maintained after transfer into tumor bearing mice. CAR8 generated from memory T cells also had reduced chromatin accessibility at features associated with memory T cells. By comparison, naive- derived CAR8 maintained chromatin accessibility patterns at regions associated with memory T cells and showed minimal skewing toward an effector- like profile \(^{23}\) (Figure 4C). To associate these changes with specific TF activities, we used ChromVAR to compare chromatin accessibility at regions containing DNA sequence motifs bound by different TFs (Figure 4D). Classifying this data using a kmeans clustering strategy, we found that there were distinct patterns of motif- associated chromatin accessibility between conditions and across each of the timepoints (Figure 4E). While motifs for bZIP and Irf family TFs broadly looked similar at the PreCAR timepoint, and became progressively enriched in memory cells, Tcf family motifs started similar and became enriched in naive cells at the latter timepoints, while E2A family motifs started highly enriched in naive and progressively
+
+<--- Page Split --->
+
+converged. Uniquely, motifs for Runx family members were always more accessible in memory- derived cells and did not converge or diverge (Figure 4D- E, S6C). Overall, these data show epigenetic features imprinted in the starting CD8+ T cell population are maintained through CAR engineering.  
+
+Prior antigen experience directs distinct transcriptomic patterns of naïve and memory- derived CAR8. To test whether the epigenetic states of naïve and memory- derived CAR8 resulted in concurrent transcriptomic changes, we performed bulk RNA- seq at the same timepoints as for ATAC- seq (Figure 4A). We found predictable differential gene expression at each timepoint, with genes associated with self- renewal and proliferative capacity (Lef1, Sell, Id3, Tcf7, Slamf6, Il7r) upregulated in the naïve- derived cells and genes associated with effector capacity and activation (Prf1, Ifng, Klrg1, Gzmb, Prdm1, Id2, Pdcd1, Tbx21) upregulated in the memory- derived cells (Figure 5A). Gene set enrichment analysis (GSEA) showed progressive bias by normalized enrichment score (NES) toward effector- like in memory- derived CAR8, and toward memory- like in naïve- derived CAR8<sup>24, 25, 26</sup> (Figure 5B- C). Analysis with gene sets comparing memory and naïve T cells showed progressive decrease in the normalized enrichment score of memory or naïve- derived CAR8 toward the derivative cell population of each, suggesting the effector/memory gene set enrichment axis as the more accurate indicator of cell fate over time<sup>24, 25</sup> (Figure S7A). Looking at the top differentially- expressed TFs between the populations at the PreCAR timepoint, we found many expected hits, including Bhlhe40, Klf4, Tbx21, Id2 and many bZIP family members (Jun, JunB, Fos, Cebpb) represented in the memory- derived group, while Zeb1, Myb and Lef1, encoding TFs associated with self- renewal, were upregulated in the naïve- derived cells<sup>23, 27</sup> (Figure S7D). Notably, among the Runx family, which showed uniquely stable differential motif accessibility between naïve and memory cells (Figure 4D), Runx2 was among the most differentially expressed TF genes with marked overexpression in memory derived cells (Figure S7D). Ingenuity Pathway Analysis of global transcriptional profile implicated similar TF drivers<sup>28</sup> (Figure S7B) with numerous distinct patterns of differential TF expression between memory and naïve- derived T cells. However, a very common pattern among ChromVAR- implicated TFs was high initial expression in memory cells at the PreCAR timepoint, followed by a convergence in expression between memory and naïve- derived CAR T cells at the PostCAR and Tumor timepoints, as seen with bZIP family members Jun, Fos and Atf3, along with the gene Tbx21, encoding canonical effector TF T- bet (Figure 5D). Among the Runx family, Runx1 and Runx3 gene expression tracked relatively closely between memory and naïve- derived cells at each timepoint, while Runx2 followed the “high in memory, then converging” pattern which was commonly found among other TF families (Figure 5E). In summary, naïve and memory- derived T cells show differential gene expression and gene set association with self- renewal or memory- associated genes and activation or effector- associated genes, respectively. Many relevant TF genes show a pattern of high initial expression in memory cells at the PreCAR timepoint which converges between the cell derivations upon transduction with a CAR and reinfusion into tumor- bearing hosts.  
+
+RUNX2 overexpression boosts leukemia clearance, CAR T cell potency and CAR proportions in bone marrow.
+
+<--- Page Split --->
+
+To validate the epigenetic and transcriptomic data, we overexpressed two TFs from the ChromVAR- implicated bZIP family, BATF and c- Jun, both of which have been previously reported to impact CAR T cell function (Figure 6A- B) \(^{29,30,31}\) . Although neither TF increased cytokine production or proliferation in vitro (Figure S8C- E), overexpression of either TF enhanced leukemia clearance by memory and naive- derived CAR T cells (Figure 6C- D). There was no difference between BATF- CAR8 or JUN- CAR8 and control CAR8 in the PD1+ proportion (Figure S9C,E), co- expression of PD1 with markers of exhaustion (PD1+/CD39+ and PD1+/TOX+), or in the terminally exhausted Tcf1-/Tim3+ population (Figure S9D,F- H).  
+
+Due to the memory- like state of CAR8 \(_{\mathrm{ND}}\) , we anticipated that comparison of factors enriched in memory cells over naive cells could reveal important drivers of memory cell function that were not fully induced in naive cells during the synthetic engineering process. Given the unique profile of chromatin accessibility for Runx- family binding motifs coupled with the pattern of Runx2 transcript expression which was high in PreCAR memory CD8+ T cells and then lost upon CAR transduction, we hypothesized that establishing RUNX2 expression in CAR8 \(_{\mathrm{ND}}\) could enhance the existing memory- like profile of these T cells and boost T cell potency and anti- leukemia response. Murine RUNX2 was introduced into the pMSCV- IRES- eGFP (pMIG) backbone, containing a GFP reporter gene for long- term tracking of RUNX2- transduced T cell populations (RUNX2). Co- transduction of naive CD8+ T cells with CAR- EGFR reporter and RUNX2- GFP reporter resulted in a large proportion of cells expressing both EGFR and GFP (Figure 6A). Upon intracellular staining for the RUNX2 protein, we found that the EGFR+ population in the RUNX2- transduced group showed approximately a 10- fold increase in RUNX2 expression relative to empty pMIG- transduced cells (Figure 6B). Co- culture of RUNX2- CAR8 and leukemia with a range of antigen densities revealed similar cytokine production and proliferation relative to pMIG- CAR8 (Figure 6C- D). To stress the ability of RUNX2- CAR8 to clear WT leukemia, we used an ultra- low CAR+ dose (1e5), against which both CAR8 \(_{\mathrm{ND}}\) and CAR8 \(_{\mathrm{MD}}\) exhibit markers of exhaustion and fail to control leukemia (Figure S7A- C). RUNX2 overexpression in CAR8 \(_{\mathrm{ND}}\) strongly enhanced leukemia clearance and increased CAR proportions and absolute numbers in the marrow at 11 days post- CAR infusion (Figure 6E- F). While there was no difference in the PD1+ proportion, consistent with similar activation, mice treated with RUNX2- CAR8 \(_{\mathrm{ND}}\) exhibited dramatically reduced proportion of PD1+/TOX+ cells, a lower proportion of PD1+/CD39+ cells and reduced proportions of TCF1-/TIM3+ cells, suggesting that RUNX2 overexpression counteracts the differentiation trajectory toward terminal exhaustion (Figure 6L, S9M- N,P) \(^{27,32}\) . CAR8 \(_{\mathrm{MD}}\) showed less of an increase in RUNX2 following transduction with RUNX2- eGFP (Figure S7F) potentially due to higher RUNX2 at baseline (Figure 5E). Nonetheless, RUNX2- overexpression resulted in a significant reduction in the PD1+/CD39+ exhaustion phenotype of RUNX2- CAR8 \(_{\mathrm{MD}}\) responding to WT leukemia and reduction in leukemia counts in marrow (Figure S9I) but no difference in other exhaustion phenotypes, CAR proportions or CAR counts (Figure 6K, S9K,L,O). We demonstrate that Runx2 overexpression in naive- derived T cells enhances maintenance of CAR T cells in the marrow, boosts leukemia clearance and mediates a favorable exhaustion profile at a highly sub- curative CAR T cell dose with less impact in memory- derived CAR T cells, demonstrating that TF overexpression has a differential impact depending on starting T cell state.  
+
+## DISCUSSION
+
+<--- Page Split --->
+
+Factors underlying tumor relapse after CAR T cell therapy are a central focus of study in the field of cell therapies for leukemia. Advances have been made in understanding and engineering solutions to prevent tumor cell escape via antigen modulation, T cell dysfunction, and poor T cell trafficking/persistence'. However, defining in vitro and in vivo functional strengths and cellular profiles associated with different starting T cell populations may be an opportunity to specifically identify approaches to arm CAR T cells to overcome different tumor escape modalities. Importantly, refining qualities of the starting cell population will likely be a large contributor to efficacy of cellular therapeutics derived from healthy allogeneic donors or induced pluripotent stem cells, or in the case of in vivo transduction platforms targeting genetic payloads to specific cell populations. Recent work has sought to use targeted modulation of TFs to enhance CAR T cell function or prevent dysfunction, with several publications focusing on the bZIP TF family, including forced expression of BATF and c- Jun, or genetic deletion of the Nr4a family of nuclear receptors29, 30, 31, 33, 34. However, the impact of modulation of the bZIP family has been variable. Therefore, we set out to characterize functional attributes programmed by prior T cell antigen experience, with the prediction that these would be tied to epigenetic traits. We anticipated that downstream modulation of TFs implicated by this framework might have divergent functional outcomes depending on starting cell population.  
+
+In this study, we use a syngeneic murine model with anti- mouse CD19 CAR T cells targeting murine pre- B cell leukemia enabling more natural T cell differentiation trajectories without xenogeneic TCR stimulation. We also used a well- defined vaccine model for precise control of the antigen experience history of CAR T cells with a clonotypic TCR, with confirmation in a polyclonal memory response. With limiting T cell dose or low target antigen density as "stressors," we report that antigen experience dictates multiple functional outputs of CAR T cells. Memory- derived CAR T cells exhibited stronger cytotoxic function across target antigen densities, while naive- derived CAR T cells show greater proliferative capacity and more rapid cell cycle entry. This was associated with enhanced activity against low- antigen density leukemia by memory derived CAR T cells and enhanced activity of naive- derived cells at limiting cell dose, a setting that drove phenotypic exhaustion and dysfunction of memory- derived cells.  
+
+T cell differentiation is a product of epigenetic and transcriptomic state23, 27 and while CAR T cells have been extensively profiled post- manufacturing, little work has been done to characterize effects of prior T cell state on post- transduction CAR T cell profiles. We demonstrate that features of these states are maintained through CAR manufacturing and associate with differences in functional profiles. Specifically, we find significant differences in bZIP family transcription factors, which have been previously implicated in CAR T cell function29, 30, 31. BATF or JUN mediated enhanced leukemia clearance in our model independent of starting cell state, indicating that these TFs may derive most of their early in vivo activity via binding to NFAT- AP1 composite motifs, which show high accessibility in both cell types. Surprisingly, there was no difference in phenotypic exhaustion in BATF or JUN- overexpressing CAR T cells relative to control, indicating preservation of function in an exhausted state rather than prevention of exhaustion.  
+
+As a novel finding, we use epigenomic and transcriptomic assays and implicate modulation of Runx- family TFs, particularly Runx2, as having a likelihood for higher impact in naive- derived cells compared to memory. Ectopic RUNX2 expression in naive- derived CAR T cells resulted in superior clearance of leukemia, higher proportions
+
+<--- Page Split --->
+
+of cells in the marrow, and reduced proportions of cells displaying terminally exhausted phenotypes relative to control. Our data suggest that RUNX2 overexpression, in contrast to overexpression of bZIP family members, can enhance functional potency of naive- derived CD8+ CAR T cells while preventing entry into the exhaustion differentiation trajectory. In addition to their activity as transcriptional activators, Runx family members have been shown to recruit chromatin remodeling factors to Runx binding sites to open these sites and allow for transcriptional activation. In other model systems, RUNX2 has been shown to interact with SWI/SNF complexes, histone acetyltransferases (MOZ, p300), histone deacetylases (HDAC3, HDAC4, HDAC6) and histone methyltransferases (SUV39H1), along with all three TET family enzymes, indicating a plausibility for the ability for RUNX2 to recruit enzymes which participate in chromatin remodeling at RUNX binding motifs<sup>35, 36, 37, 38</sup>. These features could help explain the contribution of RUNX2 overexpression to the enhanced functionality and exhaustion resistance of CAR<sub>8ND</sub> seen in our experiments. Additional studies will be necessary to fully elucidate the effects of RUNX2 in CAR T cells, and to confirm our findings in human CAR T cells. Nonetheless, using a model in which antigen history can be precisely controlled, we show that RUNX2 overexpression enhances in vivo CAR T cell function dependent on the starting T cell. Finally, we have generated a framework for the role of antigen experience on function of a CAR T cell in stress situations of limiting T cell dose or target antigen density and highlight the importance of considering this framework when assessing the impact of approaches to apply synthetic immunology to manipulate therapeutic immune effector cell functions. METHODS See Supplemental Material. AUTHOR CONTRIBUTIONS K.R.D. Conceptualized the studies, performed experiments and data analysis, and wrote the manuscript. E.D. Performed data analysis and provided expertise related to ATAC/RNA sequencing. M.D. Performed experiments. J.C. Conceptualized the studies and provided expertise related to the murine CAR and leukemia models. M.Y. Designed and generated DNA constructs. R.M.K. Provided expertise related to the vaccine model. M.E.K. Conceptualized the studies and provided expertise related to the murine CAR and leukemia models. J.P.S-B. Performed data analysis and provided expertise related to ATAC/RNA sequencing. T.J.F. Conceptualized, supervised and provided funding for the studies, and wrote the manuscript. All authors contributed to the article and approved the submitted version. ACKNOWLEDGEMENTS We thank Lillie Leach for laboratory management, Amanda Novak for animal colony management, Garrett Hedlund and Henry Chu at the CU Anschutz Clinical Immunology Flow Core for their assistance in cell sorting, the CU Cancer Center Genomics Shared Resource (RRID: SCR_021984) for their help with sequencing/genomics, and the CU Anschutz OLAR and the animal facility for their support. This work was funded in part by Department of Defense W81XWH-19-1-0196 and partly supported by the National Institutes of
+
+<--- Page Split --->
+
+Health P30CA046934 Bioinformatics and Biostatistics Shared Resource Core (RRID: SCR_021983). Some figures were created with BioRender.com.  
+
+## SUPPLEMENTAL METHODS  
+
+## Mouse Strains  
+
+Mouse StrainsB6.129S6- Rag2tm1Fwa Tg(TcraTcrb)1100Mjb ("OT- I," Model #: 2334- F) mice were obtained from Taconic Biosciences. B6. SJL- Ptprca Pepcb/BoyJ ("PepBoy," Strain #: 002014), B6.129S7- Rag1tm1Mom/J ("Rag1- ", Strain #: 002216), C57BL/6J mice ("B6," Strain #: 000664) were obtained from The Jackson Laboratory. Female mice were used for all experiments with B6 background mice. All mice were bred and/or maintained in the animal facility at University of Colorado Anschutz Medical Campus. All experiments were performed in compliance with the study protocol approved by University of Colorado Anschutz Medical Campus Institutional Animal Care and Use Committee (IACUC).  
+
+## Mouse CAR Constructs  
+
+Mouse CAR ConstructsThe basic construction of the murine 1928z CAR was previously described<sup>39</sup>. The murine anti- CD19 scFv was Flag- tagged to enable CAR detection, and all ITAMs in the CD3zeta domain were kept intact. A truncated human EGFR reporter protein was incorporated following a 2A skip sequence to provide an additional method for detection of CAR- transduced cells<sup>13</sup>. The DNA was codon optimized, ordered from ThermoFisher GeneArt, and cloned into the MSCV- IRES- GFP backbone, a gift from Tannishtha Reya (Addgene plasmid # 20672; http://n2t.net/addgene:20672; RRID:Addgene_20672), using Xhol and Clal enzyme sites. A control plasmid with just the truncated EGFR reporter in the MSCV backbone was generated using similar methods.  
+
+## Cell lines and media  
+
+Cell lines and mediaE2A- PBX pre- B cell acute lymphoblastic leukemia was developed in the laboratory as previously described<sup>14, 15, 16</sup>. Murine T cells and leukemia were cultured in Complete Mouse Media (CMM), consisting of RPMI 1640 medium (Gibco) with 10% heat- inactivated fetal calf serum (Omega Bio), 1% nonessential amino acids (Gibco), 1% sodium pyruvate (Gibco), 1% penicillin/streptomycin (Gibco), 1% L- glutamine (Gibco), 1% HEPES buffer (Gibco) and 50uM 2- mercaptoethanol (Sigma- Aldrich).  
+
+## Mouse CAR Transduction  
+
+Mouse CAR TransductionCAR transduction was performed as previously described<sup>14, 15, 16</sup>. Briefly, spleens from 6- 10 week old donor mice were harvested and CD8+ T cells were isolated using EasySep Mouse CD8+ T cell Isolation Kit from STEMCell Technologies or bulk T cells were isolated using the Mouse CD3+ T Cell Enrichment Column Kit (R&D Biosciences, Cat No. MTCC- 25). On day 1, T cells were activated on anti- CD3/anti- CD28 Mouse T cell Activator DynaBeads (Invitrogen) at a 1:1 cell:bead ratio and cultured at 1e6/mL in CMM in the presence of rhIL- 2 (40IU/mL) and rhIL- 7(10ng/mL) from R&D Systems. On days 2 and 3, retroviral supernatant was added to Retro nectin- coated (Takara Biosciences) 6 well plates and spun at 2000xg and 32°C for 2- 3 hours. Supernatant
+
+<--- Page Split --->
+
+was then removed and activated T cells were added to the wells at 1.67mL/well. On day 4, beads were removed and T cells were resuspended at 1e6/mL in fresh media with cytokines. CAR transduction was determined post- debeading by analyzing T cells by flow cytometry for a FLAG/EGFR double- positive population (or EGFR single- positive for control T cells), and T cells were used in assays or infused into mice on day 5 or 6.  
+
+## Vaccine Model  
+
+The ovalbumin vaccine consists of 100ug whole ovalbumin protein (InvivoGen, Cat. code: vac- pova- 100), 40ug anti- mouse CD40 (BioXCell, Catalog #BE0016- 2) and 40ug Polyinosinic:polycytidylic acid [Poly (l:C)] (InvivoGen, Cat. code: tlrl- pic- 5) per mouse, resuspended to 200uL total volume in PBS<sup>17, 18, 19</sup>. CD8+ T cells were isolated from naive 6 to 8 week old OT- I mouse splenocytes using the Mouse CD3+ T Cell Enrichment Column Kit (R&D Biosciences, Cat No. MTCC- 25). PepBoy mice were given 5e3 OT- I T cells retro- orbitally and concurrently vaccinated intravenously. 3- 4 weeks later, spleens from 5- 20 vaccinated mice were pooled and CD45.2+ OT- I memory T cells were isolated using the EasySep Mouse CD8+ T cell Isolation Kit, followed by column isolation using biotinylated anti- mouse CD45.2 (BioLegend, Cat # 109804), LS Columns (Miltenyi Biotec, Order No. 130- 042- 401), and anti- Biotin MicroBeads (Miltenyi Biotec, Order No. 130- 090- 485). Naive T cells from 1- 5 naive OT- I donors were isolated in parallel. T cells were then activated and transduced as described for downstream experiments.  
+
+## Generation of CD19<sup>lo</sup> E2A-PBX leukemia cell lines  
+
+The E2A- PBX murine leukemia was generated in our lab as previously described<sup>14</sup>. CD19 knockout leukemia was produced using CRISPR/Cas9. A previously- validated murine CD19- targeting sgRNA<sup>15</sup> from Integrated DNA Technologies was incubated with recombinant Cas9 from TakaraBio (Cat# 632641) to create an RNP complex. RNP was then electroporated into E2A- PBX using the Lonza 4D- Nucleofector X with nucleofector solution SG and pulse program CM- 147. Electroporated cells were allowed to recover for 48 hours and then FACS- sorted twice to obtain a pure CD19 knockout cell line. This cell population was additionally single cell cloned to create a CD19 knockout single cell clone prior to transduction with murine CD19. A truncated/non- signaling murine CD19 was cloned into the pLV.SP146.gp91.GP91.cHS4 plasmid, a gift from Didier Trono (Addgene plasmid # 30480; http://n2t.net/addgene:30480; RRID:Addgene_30480). Backbones were generated with the hEF1a promoter (pLV.hEF1a.cHS4) or the hUbC promoter (pLV.hUbC.cHS4) from the pLenti6/UbC/mSlc7a1 plasmid, a gift from Shinya Yamanaka (Addgene plasmid # 17224; http://n2t.net/addgene:17224; RRID:Addgene_17224). VSV- G pseudotyped lentivirus was generated as described and E2A- PBX CD19KO underwent a single round of transduction using standard protocols, followed by single cell cloning to obtain clonally- derived lines expressing defined levels of CD19 target antigen.  
+
+## Flow Cytometry  
+
+Flow cytometry analysis was performed using an LSR- Fortessa X- 20 flow cytometer (BD Biosciences) and analyzed using FlowJo (BD Biosciences). Monoclonal antibodies used in staining are listed in the supplemental methods. Intracellular flow cytometry staining was performed using the TrueNuclear Transcription Factor Buffer
+
+<--- Page Split --->
+
+Set (BioLegend) for ex vivo staining of transcription factors, Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences) for intracellular cytokine staining, and Mouse Foxp3 Buffer Set (BD Biosciences) for intracellular staining of Ki67 and Runx2.  
+
+CD107a Degranulation, Intracellular Cytokine Staining (ICCS), Ki67 and CellTrace Dilution In Vitro Assays In vitro assays were performed using a 1:1 effector to target cell ratio with 1e5 of each cell type in a 96- well round- bottom plate followed by analysis by flow cytometry at the indicated timepoints. Degranulation assays were performed by incubation for 4 hours in the presence of 2uM monensin and 1uL of CD107a antibody. ICCS was performed by incubation for 6 hours, with 1uM monensin and 2.5uM Brefeldin A added at 1 hour in. Ki67 was performed by incubation for 18 hours, followed by intracellular staining for Ki67. CellTrace dilution assays were performed by staining T cells with CellTrace Violet (Thermo Fisher Scientific) per manufacturer protocols followed by incubation with target cells for 72 hours.  
+
+## LCMV infection and T cell isolation  
+
+6 week old female C57BL/6 mice were injected retro- orbitally with 2e5 PFU of LCMV- Armstrong. 4 weeks later, CD8+ T cells were isolated from 5 pooled spleens using the EasySep Mouse CD8+ T cell Isolation Kit from STEMCell Technologies and then FACS- sorted to obtain Memory (CD8+/CD44+/CD49dhi) and Naive (CD8+/CD44-/CD49d0/CD62L+) populations from the same mice. T cells were then transduced using the standard transduction protocol as described.  
+
+## In vivo experiments in Rag1- hosts  
+
+Experiments were carried out using a timeline previously optimized in the lab<sup>14</sup>. Briefly, Rag1- hosts were inoculated with 1e6 E2A- PBX by tail vein I.V. injection on day - 3 followed by CAR T cells via retroorbital injection at either 1e5, 3e5 or 1e6 CAR+ cell dose on day 0. Bone marrow was harvested and analyzed by flow cytometry on day 4 or 11 post- CAR infusion, or mice were euthanized at humane endpoints for survival experiments. Ex vivo stimulation for cytokine production was performed using 1e6 E2A- PBX WT to stimulate approximately 1.5e6 whole bone marrow cells from each individual mouse, with pooled bone marrow from each n=5 experimental group stimulated by E2A- PBX CD19<sup>Neg</sup> as a negative control. Cells were co- cultured for 6 hours, with 1uM monensin and 2.5uM Brefeldin A added at 1 hour in and then analyzed by flow for cytokine production.  
+
+## Bulk ATAC and RNA sequencing experimental setup and workflows  
+
+OT- I CD8+ T cells were isolated from vaccinated or naive donors and CARs were transduced into T cells as described above. CAR8 Rag1- hosts were inoculated with 1e6 E2A- PBX CD19<sup>10,000</sup> followed by 1e6 CAR8<sub>MD</sub> or CAR8<sub>ND</sub> on the timeline described above. At day 4 post- CAR infusion, bone marrow from 10 mice per CAR group was harvested and pooled. At each of 3 timepoints, CD8+ cells were isolated using the EasySep Mouse CD8+ T cell Isolation Kit from STEMCell Technologies and then FACS- sorted to obtain 50,000 cells per condition. ATAC- seq and RNA- seq were performed in triplicate on separate sorted aliquots of 50,000 cells at "Pre- CAR/Day -5" (ex vivo, directly after isolation of memory or naive CD8+ T cells from donor mice), "Post- CAR/Day 0" (in
+
+<--- Page Split --->
+
+vitro, after CAR manufacturing) and "Tumor/Day 4" (ex vivo, after reinfusion into leukemia bearing mice). Experimental analyses were performed on the first technical replicate from 2 separate experimental replicates. For RNA-seq, cells were homogenized in QIAzol Lysis Reagent (Qiagen, Cat. No. 79306) and then frozen at - 80C for processing within 2 weeks. Samples were thawed and processed using the miRNeasy Micro Kit (Qiagen, Mat. No. 1071023), with on- column DNase treatment (RNase- Free DNase Set, Qiagen, Cat. No. 79254), both according to manufacturer protocols. RNA purity, quantity and integrity was determined with NanoDrop (ThermoFisher Scientific) and TapeStation 4200 (Agilent) analysis prior to RNA-seq library preparation. The Universal Plus mRNA-Seq library preparation kit with NuQuant was used (Tecan) with an input of 200ng of total RNA to generate RNA-seq libraries. Paired- end sequencing reads of 150bp were generated on NovaSeq 6000 (Illumina) sequencer at a target depth of 40 million clusters/80 million paired- end reads per sample. Raw sequencing reads were de- multiplexed using bcl2fastq. For ATAC- seq, cells were immediately processed using the Omni- ATAC protocol as previously described<sup>40</sup>. Briefly, sorted cells were washed once in 1X PBS, lysed, washed once in Wash Buffer and then the transposition reaction was carried out at 32°C for 30 minutes on a thermomixer set to 1000 rpm. Transposed chromatin was then purified using the Zymo Clean and Concentrator 5 Kit (Zymo Research, Cat # D4013) using manufacturer protocols. DNA was then ran on PCR for 12 total cycles with matched barcoding primers<sup>41</sup>. PCR reactions were then size- selected using AMPure XP beads (Beckman Coulter Life Sciences, Product No: A63880) and checked for quality and size distribution using TapeStation 4200 with D5000 reagents (Agilent). Libraries were pooled at equimolar ratios for sequencing and paired- end sequencing reads of 150bp for the first replicate and 50bp for the second replicate were generated on NovaSeq 6000 (Illumina) sequencer at a target depth of 40 million clusters/80 million paired- end reads per sample. Raw sequencing reads for replicate 1 were shortened to match the read lengths for replicate 2 using trimmomatic function CROP. Raw sequencing reads were de- multiplexed using bcl2fastq.  
+
+## RNA-seq Data Analysis  
+
+Quality of fastq files was accessed using FastQC (v.0.11.8) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc), FastQ Screen (v.0.13.0)<sup>42</sup> and MultiQC (v.1.8)<sup>43</sup>. Illumina adapters and low- quality reads were filtered out using BBDuk (v. 38.87) (http://jgi.doe.gov/data- and- tools/bb- tools). Trimmed fastqc files were aligned to the mm10 murine reference genome and aligned counts per gene were quantified using STAR (v.2.7.9a)<sup>44</sup>. Differential gene expression analysis was performed using the DESeq2 package<sup>45</sup>. Pathway enrichment analysis was performed using GSEA (UC San Diego/Broad Institute)<sup>26</sup>,<sup>46</sup>, Metascape<sup>47</sup> for gene mapping and IPA (Qiagen)<sup>28, 48</sup>. Differential gene expression was plotted using GraphPad Prism or ggplot2 (R package). RNA- seq differential gene expression statistics were run using the DESeq2 R package, with filtering threshold at 10 with greater than 2- fold change and adjusted p value < 0.05.  
+
+## ATAC-seq Data Analysis  
+
+Fastq files were used to map to the mm10 genome using the ENCODE ATAC- seq pipeline (https://www.encodeproject.org/atac- seq/), with default parameters, except bam files used for peak calling were randomly downsampled to a maximum of 50 million mapped reads. Peaks with a
+
+<--- Page Split --->
+
+MACS2(https://pypi.org/project/MACS2/) computed q value of less than 1e- 6 and a signalValue of more than 4 in at least one replicate were merged with bedtools<sup>49</sup> function intersect and processed to uniform peaks with the functions getPeaks and resize from R package ChromVAR<sup>22</sup>. Reads overlapping peaks were enumerated with getCounts function from ChromVAR and normalized and log2- transformed with loom from R package limma<sup>50</sup>. Peaks with 3 or more normalized counts per million mapped reads at least one replicate were included to define a global peak set of 82,410 peaks. Pairwise Euclidean distances were computed between all samples using log2- transformed counts per million mapped reads among the global peak set. Differentially accessible peaks were identified in pairwise comparisons based on fdr adjusted p values of less than 0.01, fold change of at least 4 and with an average of 3 normalized counts per million mapped reads using R package limma. Motif associated variability in ATAC- seq signal was computed with R package ChromVAR. Genome- wide visualization of ATAC- seq coverage was computed with deeptools<sup>51</sup> function coveragebam, using manually computed scale factors based on the number of reads within the total peak set.  
+
+## Statistics  
+
+Statistical tests for all experiments except sequencing analyses were performed using GraphPad Prism v9.0 for Macintosh (GraphPad Software). Comparisons between three groups were made with ordinary one- way ANOVA with Holm- Sidak's multiple comparisons test, Brown- Forsythe and Welch one- way ANOVA with Dunnett's T3 multiple comparisons test, or Kruskal- Wallis non- parametric test with Dunn's multiple comparisons test were used depending on variance in standard deviations. Two- way ANOVA or mixed effects analysis with Tukey's multiple comparisons test was used for in vitro experimental comparisons with multiple antigen densities and in vivo CAR expansion data. Two- tailed ordinary t test, Welch's t test or Mann- Whitney test were performed for comparisons with two groups depending on normality of distributions. For multiple comparisons of two groups, multiple unpaired t tests or multiple Welch's t tests, both with Holm- Sidak's multiple comparisons test, were performed when appropriate depending on variance in standard deviations. Log- rank (Mantel- Cox) test was used for survival curve comparisons. All data represented as mean +/- standard deviation. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. Technical and experimental replicates in each dataset are indicated in figure legends.  
+
+## Data and Materials Availability  
+
+All data is readily available from authors upon request or accessible at Gene Expression Omnibus (GEO Accession Number will be provided before paper acceptance). All materials are either commercially available as described or available from authors upon request.
+
+<--- Page Split --->
+
+1. Labanieh, L. & Mackall, C.L. CAR immune cells: design principles, resistance and the next generation. Nature 614, 635-648 (2023).  
+2. Sommermeyer, D. et al. Chimeric antigen receptor-modified T cells derived from defined CD8+ and CD4+ subsets confer superior antitumor reactivity in vivo. Leukemia 30, 492-500 (2016).  
+3. Qin, H. et al. CAR T cells targeting BAFF-R can overcome CD19 antigen loss in B cell malignancies. Sci Transl Med 11 (2019).  
+4. Aldoss, I. et al. Favorable Activity and Safety Profile of Memory-Enriched CD19-Targeted Chimeric Antigen Receptor T-Cell Therapy in Adults with High-Risk Relapsed/Refractory ALL. Clin Cancer Res 29, 742-753 (2023).  
+5. Alvanou, M. et al. Empowering the Potential of CAR-T Cell Immunotherapies by Epigenetic Reprogramming. Cancers (Basel) 15 (2023).  
+6. Frias, A.B., Boi, S.K., Lan, X. & Youngblood, B. Epigenetic regulation of T cell adaptive immunity. Immunol Rev 300, 9-21 (2021).  
+7. Wherry, E.J. & Kurachi, M. Molecular and cellular insights into T cell exhaustion. Nat Rev Immunol 15, 486-499 (2015).  
+8. Heitzeneder, S. et al. GPC2-CAR T cells tuned for low antigen density mediate potent activity against neuroblastoma without toxicity. Cancer Cell 40, 53-69.e59 (2022).  
+9. Fry, T.J. et al. CD22-targeted CAR T cells induce remission in B-ALL that is naive or resistant to CD19-targeted CAR immunotherapy. Nat Med 24, 20-28 (2018).  
+10. Majzner, R.G. et al. Tuning the Antigen Density Requirement for CAR T-cell Activity. Cancer Discov 10, 702-723 (2020).  
+11. Kumar, R. et al. Increased sensitivity of antigen-experienced T cells through the enrichment of oligomeric T cell receptor complexes. Immunity 35, 375-387 (2011).  
+12. Mehlhop-Williams, E.R. & Bevan, M.J. Memory CD8+ T cells exhibit increased antigen threshold requirements for recall proliferation. J Exp Med 211, 345-356 (2014).  
+13. Wang, X. et al. A transgene-encoded cell surface polypeptide for selection, in vivo tracking, and ablation of engineered cells. Blood 118, 1255-1263 (2011).  
+14. Qin, H. et al. Murine pre-B-cell ALL induces T-cell dysfunction not fully reversed by introduction of a chimeric antigen receptor. Blood 132, 1899-1910 (2018).  
+15. Jacoby, E. et al. CD19 CAR immune pressure induces B-precursor acute lymphoblastic leukaemia lineage switch exposing inherent leukaemic plasticity. Nat Commun 7, 12320 (2016).  
+16. Yang, Y. et al. TCR engagement negatively affects CD8 but not CD4 CAR T cell expansion and leukemic clearance. Sci Transl Med 9 (2017).
+
+<--- Page Split --->
+
+17. Ivanova, D.L. et al. Vaccine adjuvant-elicited CD8(+) T cell immunity is co-dependent on T-bet and FOXO1. Cell Rep 42, 112911 (2023). 
+
+18. Klarquist, J. et al. B cells promote CD8 T cell primary and memory responses to subunit vaccines. Cell Rep 36, 109591 (2021). 
+
+19. Klarquist, J. et al. Clonal expansion of vaccine-elicited T cells is independent of aerobic glycolysis. Sci Immunol 3 (2018). 
+
+20. Badovinac, V.P., Haring, J.S. & Harty, J.T. Initial T cell receptor transgenic cell precursor frequency dictates critical aspects of the CD8(+) T cell response to infection. Immunity 26, 827-841 (2007). 
+
+21. Marzo, A.L. et al. Initial T cell frequency dictates memory CD8+ T cell lineage commitment. Nat Immunol 6, 793-799 (2005). 
+
+22. Schep, A.N., Wu, B., Buenrostro, J.D. & Greenleaf, W.J. chromVAR: inferring transcription-factor-associated accessibility from single-cell epigenomic data. Nat Methods 14, 975-978 (2017). 
+
+23. Scott-Browne, J.P. et al. Dynamic Changes in Chromatin Accessibility Occur in CD8(+) T Cells Responding to Viral Infection. Immunity 45, 1327-1340 (2016). 
+
+24. Wherry, E.J. et al. Molecular signature of CD8+ T cell exhaustion during chronic viral infection. Immunity 27, 670-684 (2007). 
+
+25. Luckey, C.J. et al. Memory T and memory B cells share a transcriptional program of self-renewal with long-term hematopoietic stem cells. Proc Natl Acad Sci U S A 103, 3304-3309 (2006). 
+
+26. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 102, 15545-15550 (2005). 
+
+27. Giles, J.R. et al. Human epigenetic and transcriptional T cell differentiation atlas for identifying functional T cell-specific enhancers. Immunity 55, 557-574.e557 (2022). 
+
+28. Krämer, A., Green, J., Pollard, J., Jr. & Tugendreich, S. Causal analysis approaches in Ingenuity Pathway Analysis. Bioinformatics 30, 523-530 (2014). 
+
+29. Zhang, X. et al. Depletion of BATF in CAR-T cells enhances antitumor activity by inducing resistance against exhaustion and formation of central memory cells. Cancer Cell 40, 1407-1422.e1407 (2022). 
+
+30. Seo, H. et al. BATF and IRF4 cooperate to counter exhaustion in tumor-infiltrating CAR T cells. Nat Immunol 22, 983-995 (2021). 
+
+31. Lynn, R.C. et al. c-Jun overexpression in CAR T cells induces exhaustion resistance. Nature 576, 293-300 (2019). 
+
+32. van der Heide, V., Humblin, E., Vaidya, A. & Kamphorst, A.O. Advancing beyond the twists and turns of T cell exhaustion in cancer. Sci Transl Med 14, eabo4997 (2022).
+
+<--- Page Split --->
+
+33. Chen, J. et al. NR4A transcription factors limit CAR T cell function in solid tumours. Nature 567, 530-534 (2019).  
+
+34. Jung, I.Y. et al. BLIMP1 and NR4A3 transcription factors reciprocally regulate antitumor CAR T cell stemness and exhaustion. Sci Transl Med 14, eabn7336 (2022).  
+
+35. Tandon, M. et al. Runx2 mediates epigenetic silencing of the bone morphogenetic protein-3B (BMP-3B/GDF10) in lung cancer cells. Mol Cancer 11, 27 (2012).  
+
+36. Hojo, H. et al. Runx2 regulates chromatin accessibility to direct the osteoblast program at neonatal stages. Cell Rep 40, 111315 (2022).  
+
+37. Wang, L. et al. TET enzymes regulate skeletal development through increasing chromatin accessibility of RUNX2 target genes. Nat Commun 13, 4709 (2022).  
+
+38. Korinfskaya, S., Parameswaran, S., Weirauch, M.T. & Barski, A. Runx Transcription Factors in T Cells-What Is Beyond Thymic Development? Front Immunol 12, 701924 (2021).  
+
+39. Kochenderfer, J.N., Yu, Z., Frasheri, D., Restifo, N.P. & Rosenberg, S.A. Adoptive transfer of syngeneic T cells transduced with a chimeric antigen receptor that recognizes murine CD19 can eradicate lymphoma and normal B cells. Blood 116, 3875-3886 (2010).  
+
+## METHODS REFERENCES  
+
+40. Corces, M.R. et al. An improved ATAC-seq protocol reduces background and enables interrogation of frozen tissues. Nat Methods 14, 959-962 (2017).  
+
+41. Buenrostro, J.D. et al. Single-cell chromatin accessibility reveals principles of regulatory variation. Nature 523, 486-490 (2015).  
+
+42. Wingett, S.W. & Andrews, S. FastQ Screen: A tool for multi-genome mapping and quality control. F1000Res 7, 1338 (2018).  
+
+43. Ewels, P., Magnusson, M., Lundin, S. & Käller, M. MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics 32, 3047-3048 (2016).  
+
+44. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15-21 (2013).  
+
+45. Love, M.I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550 (2014).  
+
+46. Zhu, A., Srivastava, A., Ibrahim, J.G., Patro, R. & Love, M.I. Nonparametric expression analysis using inferential replicate counts. Nucleic Acids Res 47, e105 (2019).  
+
+47. Zhou, Y. et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun 10, 1523 (2019).  
+
+48. Mootha, V.K. et al. PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet 34, 267-273 (2003).
+
+<--- Page Split --->
+
+49. Quinlan, A.R. & Hall, I.M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841-842 (2010).  
+50. Ritchie, M.E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43, e47 (2015).  
+51. Ramírez, F., Dündar, F., Diehl, S., Grüning, B.A. & Manke, T. deepTools: a flexible platform for exploring deep-sequencing data. Nucleic Acids Res 42, W187-191 (2014).
+
+<--- Page Split --->
+
+# FIGURES 
+
+## Figure 1 
+
+### A. E2A-PBX Leukemia Molecules/cell 
+
+![](images/Figure_2.jpg)
+
+ 
+
+C. 
+
+Increasing Target Antigen Expression 
+
+![](images/Figure_4.jpg)
+
+
+<--- Page Split --->
+
+1 Figure 1: Antigen experience history directs multiple aspects of in vitro functional capacity of murine CD8+ CAR T cells. 2 1A: E2A- PBX murine leukemia was engineered to knockout CD19, followed by reintroduction of CD19 at different levels to generate a range of antigen density clones. 1B: Schematic: Vaccine model for generating memory CD8+ OT- I T cells. 5e3 OT- I T cells were transferred into congenically distinct hosts which were concurrently vaccinated with antigen and adjuvants. 3- 5 weeks later, CAR T cells were manufactured from memory OT- I's (CAR8MD, memory- derived) or naive OT- I's (CAR8ND, naive- derived) 1C: Intracellular cytokine staining of IFNg and TNFa after 6 hour co- culture assay. 1D: Degranulation as measured by CD107a expression after 4 hour coculture assay. 1E- G: Quantification of cytokine data, % positive cells for indicated cytokine. 1H: Quantification of CD107a data, % positive cells. 1I: Cell- cycle entry as measured by Ki- 67 staining after 18 hour co- culture assay. 1J: Proliferation as measured by dilution of CellTrace Violet dye after 72 hour co- culture assay. All in vitro assays were performed with \(n = 3\) technical replicates, and are representative of 2 independent experiments. Data represent mean \(+ / -\) SD. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) .
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+ 
+
+<center>Figure 2</center> 
+
+![](images/Figure_unknown_0.jpg)
+
+
+<--- Page Split --->
+
+1 Figure 2: CAR8MD exhibit enhanced cytotoxicity and clearance of CD19<sup>lo</sup> leukemia in vivo (high CAR  
+
+2 dose).  
+
+3 2A: Schematic: Timeline for in vivo experiments. Rag1<sup>- t</sup> mice were injected with 1e6 E2A- PBX1 leukemia on  
+
+4 day - 3, followed by 1e6 OT- I CD8+/EGFR+ T cells from indicated T cell condition on day 0. Bone marrow was  
+
+5 analyzed by flow cytometry on day +4 or day +11. T cell populations were isolated memory- derived CAR T cells  
+
+6 (CAR8MD), isolated naive- derived CAR T cells (CAR8ND) or EGFR control T cells (EGFR8). Leukemia populations  
+
+7 were CD19<sup>Neg</sup>, CD19<sup>lo</sup>(10,000 antigens/cell), or WT (35,000 antigens/cell). 2B- C: Early T cell expansion (day  
+
+8 +4) or persistence (day +11) after infusion of transduced T cells against WT leukemia (B) and CD19<sup>lo</sup> leukemia  
+
+9 (C). Transduced T cell populations measured by coexpression of CD8a+/TCRbeta+/EGFR+. 2D: Clearance of  
+
+10 WT and CD19<sup>lo</sup> leukemia at day +11 after CAR infusion. E2A- PBX measured by coexpression of B220+/CD22+.  
+
+11 2E- F: Intracellular cytokine staining of interferon gamma (E) or granzyme B (F) in CAR T cells from whole bone marrow restimulated ex vivo with leukemia. Data represent mean +/- SD. 2G- I: Intranuclear transcription factor staining of IRF4 (G), EOMES (H), or T- bet (I) on CAR+ T cells from mice bearing the indicated leukemia at day +4 after CAR infusion. Violin plot data represent median with quartiles. Data are from 2 pooled, independent experiments with n=10 mice per condition. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. 2J: Survival of mice after treatment with 1e6 EGFR+ CAR or control T cells. Survival statistics were performed using log- rank (Mantel- Cox) test \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. Data is from 2 independent pooled experiments, total n=10 mice per group.  
+
+12 marrow restimulated ex vivo with leukemia. Data represent mean +/- SD. 2G- I: Intranuclear transcription factor  
+
+13 staining of IRF4 (G), EOMES (H), or T- bet (I) on CAR+ T cells from mice bearing the indicated leukemia at day  
+
+14 +4 after CAR infusion. Violin plot data represent median with quartiles. Data are from 2 pooled, independent  
+
+15 experiments with n=10 mice per condition. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. 2J: Survival of mice  
+
+16 after treatment with 1e6 EGFR+ CAR or control T cells. Survival statistics were performed using log- rank (Mantel-  
+
+17 Cox) test \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. Data is from 2 independent pooled experiments, total  
+
+18 n=10 mice per group.  
+
+19  
+
+20  
+
+21  
+
+22  
+
+23  
+
+24  
+
+25  
+
+26
+
+<--- Page Split --->
+![](images/Figure_unknown_1.jpg)
+
+
+<--- Page Split --->
+
+Figure 3: CAR8ND exhibit enhanced expansion capacity and clearance of WT leukemia in vivo (low CAR dose).3A-B: Early T cell expansion (day +4) or persistence (day +11) after infusion of transduced T cells against WT leukemia (A) and CD19<sup>lo</sup> leukemia (B). Transduced T cell populations measured by coexpression of CD8a+/TCRbeta+/EGFR+. 3C: Clearance of WT and CD19<sup>lo</sup> leukemia at day +11 after CAR infusion. E2A-PBX measured by coexpression of B220+/CD22+. 3D-E: Intracellular cytokine staining of interferon gamma (D) or granzyme B (E) in CAR T cells from whole bone marrow restimulated ex vivo with leukemia. Data represent mean +/- SD. 2F-H: Intranuclear transcription factor staining of IRF4 (F), EOMES (G), or T-bet (H) on CAR+ T cells from mice bearing the indicated leukemia at day +4 after CAR infusion. Violin plot data represent median with quartiles. Data are from 2 pooled, independent experiments with \(n = 10\) mice per condition. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) . 2I: Survival of mice after treatment with 1e6 EGFR+ CAR or control T cells. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) . Data are from 2 independent pooled experiments, total \(n = 10\) mice per group.
+
+<--- Page Split --->
+![](images/Figure_unknown_2.jpg)
+
+<center>Figure 4 </center>  
+
+![](images/Figure_unknown_3.jpg)
+  
+
+![](images/Figure_unknown_4.jpg)
+
+
+<--- Page Split --->
+
+Figure 4: Prior antigen experience imprints chromatin accessibility states which follow unique patterns during CAR transduction and reinfusion. 4A: Schematic: Layout for paired ATAC-seq/RNA-seq experiments. Memory-derived or naïve derived OT-I CD8+ T cells were sorted at three sequential timepoints: Ex vivo from donor mice before CAR transduction ("PreCAR"), in vitro after CAR transduction ("PostCAR"), and ex vivo after reinfusion into CD19<sup>Lo</sup> leukemia-bearing Rag1<sup>-/-</sup> mice ("Tumor"). 4B: Chromatin accessibility at Gzmb, Gzmc, Ifng, Tcf7 and Pdcd1 gene loci for naïve and memory-derived T cells at each timepoint. 4C: ChromVAR deviation z-scores between indicated populations at differentially accessible regions between Effector and Memory T cells after LCMV-Armstrong infection<sup>23</sup>. Data are mean +/- range of two biological replicates. 4D: Motif-associated ChromVAR deviation z-scores between indicated populations. Data are mean +/- range of two biological replicates 4E: K-means clustering of relative ATAC-seq signal at differentially accessible regions (top, data from two biological replicates are shown) and motif enrichment in each cluster vs all regions (bottom).
+
+<--- Page Split --->
+![](images/Figure_unknown_5.jpg)
+
+<center>Figure 5 </center>
+
+<--- Page Split --->
+
+Figure 5: Prior antigen experience drives differential CAR8 transcriptomic states which follow unique patterns during CAR transduction and reinfusion. RNA- seq analysis was run on the timepoints/conditions indicated in the previous figure. 5A: Volcano plots of significant differentially expressed genes between naïve and memory- derived cells at each of the three timepoints. 5B: Normalized enrichment scores from gene set enrichment analysis (GSEA) of differentially enriched genesets between indicated CD8+ T cell subsets after LCMV- Armstrong acute viral infection24 5C: GSEA plots at each timepoint. 5D: Top differentially expressed transcription factors at the "PreCAR" timepoint, generated using Ingenuity Pathway Analysis (IPA). 5E: DESeq2- normalized counts of indicated transcription factors at each timepoint for naïve and memory- derived cells. 5F: DESeq2- normalized counts of Runx family transcription factors at each timepoint for naïve and memory- derived cells. All statistics performed using DESeq2 with filtering threshold at 10, log2foldchange >2 and padj < 0.05.
+
+<--- Page Split --->
+![](images/Figure_unknown_6.jpg)
+
+
+<--- Page Split --->
+
+Figure 6: Runx2 overexpression as a novel strategy for enhancement of naïve-derived CD8+ CAR T cell potency and resistance to dysfunction. 6A-B: Cotransduction of memory (A) or naïve (B) CD8+ T cells with CAR and pMIG-Empty, pMIG- BATF, or pMIG- JUN. For 6C-F & I-L, Rag1- mice were given leukemia on day -3, followed by 1e5 pMIG- Runx2 or pMIG- Empty co- transduced CAR8 on day 0. Bone marrow was analyzed by flow cytometry on day 11 post- CAR. 6C & D: CAR T cell and leukemia proportions for naïve (C) and memory- derived (D) CAR T cells cotransduced with BATF, JUN or pMIG control. 6E & F: Proportion of CAR T cells displaying PD1+/TOX+ phenotype. 6G-H: Cotransduction of memory (G) or naïve (H) CD8+ T cells with CAR and pMIG-Empty or pMIG- Runx2 and intracellular staining for Runx2. 6I & J: CAR T cell and leukemia proportions for naïve (C) and memory- derived (D) CAR T cells cotransduced with RUNX2 or pMIG control. 6K & L: Proportion of CAR T cells displaying PD1+/TOX+ phenotype. Data in 6A,B,G & H are representative of 3-4 independent experiments. Data in 6C-F are from 1 experiment with n=5 mice per condition. Data in 6I-L are from 2 pooled, independent experiments with n=9 mice per condition. Data represent mean +/- SD. * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001.
+
+<--- Page Split --->
+![](images/Figure_unknown_7.jpg)
+
+<center>Figure S1 </center>  
+
+![](images/Figure_unknown_8.jpg)
+
+<center>D. Against WT Leukemia </center>  
+
+![](images/Figure_unknown_9.jpg)
+
+<center>H. </center>  
+
+![](images/Figure_unknown_10.jpg)
+
+<center>Antigen Density </center>  
+
+![](images/Figure_unknown_11.jpg)
+
+<center>Antigen Density </center>
+
+<--- Page Split --->
+
+Figure S1: E2A-PBX/mCD19 antigen density model and murine anti-CD19 CAR T cells, and additional statistical comparisons of in vitro data (Related to Figure 1) S1A: Schematic of the anti-mouse CD19 CAR contained in pMSCV backbone. S1B: Coexpression of CAR and EGFR on murine CAR T cells. S1C: Engineering of murine leukemia with lentiviral vectors containing hUbC or hEF1a promoters driving the CD19 transgene. S1D: Survival of mice after treatment with 1e6 EGFR+ (EGFR8, non- CAR expressing) naïve or memory- derived CD8+ T cells. Data is from 1 experiment, total n=5 mice per group. S1E: Mean fluorescence intensity of IFNg+ cell population. S1F: Mean fluorescence intensity of TNFa+ cell population. S1G: Mean fluorescence intensity of CD107a+ population. S1H: Statistical comparisons of Ki67Ne9 (% Ki67Neg of EGFR+), Ki67L0 (% Ki67L0 of EGFR+, MFI Ki67L0 of EGFR+) and Ki67H1 (% Ki67H1 of EGFR+, MFI Ki67H1 of EGFR+) populations. S1I: Statistical comparisons of CellTraceL0 (% CellTraceL0 of EGFR+, MFI CellTraceL0 of EGFR+) and total EGFR+ (GFMI CellTrace, GFMI CellTrace with zoomed axis) populations. Data represent mean +/- SD. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001.
+
+<--- Page Split --->
+![](images/Figure_unknown_12.jpg)
+
+
+<--- Page Split --->
+
+Figure S2: Polyclonal pathogen- elicited CAR8MD function similarly to vaccine- elicited CAR8MD (Related to Figure 1). S3A: Schematic: LCMV model for generating memory CD8+ T cells. C57BL/6 hosts were infected with LCMV- Armstrong. 4 weeks later, naïve and memory CD8+ T cells were sorted from the same hosts using the indicated FACS markers and used to manufacture CAR8MD, memory- derived or CAR8ND, naïve- derived or EGFR8 control cells. S3B: Intracellular cytokine staining of IFNg and TNFA after 6 hour co- culture assay. S3C: Quantifications of proportions of IFNg+ and TNFA+ cells of EGFR+ population. S3D: Proliferation as measured by dilution of CellTrace Violet dye after 72 hour co- culture assay. S3E: Quantification of CellTrace assay, proportions of CellTrace<sup>Lo</sup> cells. All assays were performed with \(n = 3\) technical replicates, and are representative of 2 independent experiments. Data represent mean \(+ / -\) SD. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) .
+
+<--- Page Split --->
+
+# Figure S3  
+
+# 1e6 CAR+ Cell Dose  
+
+![](images/Figure_unknown_13.jpg)
+  
+
+# 3e5 CAR+ Cell Dose  
+
+![](images/Figure_unknown_14.jpg)
+
+<center>Figure S3: CAR T cells and leukemia counts per tibia for in vivo data (Related to Figures 2 & 3) </center>  
+
+All analyses in this figure are done on the same experiments described in Figures 2 and 3. Counts data was generated by flushing a single tibia and using total tibia counts and cytometer proportions data to calculate CAR and leukemia cell counts per tibia. S8A: CAR counts for 1e6 CAR dose experiments. S8B: Leukemia counts for 1e6 CAR dose experiments. S8C: CAR counts for 3e5 CAR dose experiments. S8D: Leukemia counts for 3e5 CAR dose experiments. Data are from 2 pooled, independent experiments with \(n = 10\) mice per condition, apart from the 1e6 CAR dose day 11 timepoint, which contains data from one experiment with \(n = 5\) mice per condition. Data represent mean \(+ / -\) SD. \(^*\) \(p< 0.05\) \(^{**}\) \(p< 0.01\) \(^{***}\) \(p< 0.0001\)
+
+<--- Page Split --->
+![](images/Figure_unknown_15.jpg)
+
+ 
+
+<center>A.</center> 
+
+![PLACEHOLDER_38_1]
+
+ 
+
+<center>B.</center> 
+
+![PLACEHOLDER_38_2]
+
+ 
+
+<center>Day 11</center> 
+
+![PLACEHOLDER_38_3]
+
+ 
+
+<center>D.</center> 
+
+![PLACEHOLDER_38_4]
+
+ 
+
+<center>Day 11</center>
+
+<--- Page Split --->
+
+Figure S4: Basic characterization of in vivo model and additional in vivo effector/memory phenotyping at high CAR dose (Related to Figure 2).S4A: Basic flow cytometry gating strategy for in vivo experiments. Total events were gated by Singlets, Live Cells and then Lymphocytes, followed by CD8a+/TCRbeta+/EGFR+ cells for CAR8/EGFR8 or B220+/CD22+ cells for E2A-PBX. S4C-F are from experiments with the 1e6 EGFR+ cell dose. S4B: Proportions of CAR8 with the short-lived effector cell (SLEC, IL7Ra-/KLRG1+) or memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) phenotypes at the indicated timepoint against WT leukemia. S4C: Proportions of CAR8 with the short-lived effector cell (SLEC, IL7Ra-/KLRG1+) or memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) phenotypes at the indicated timepoint against CD19<sup>lo</sup> leukemia. S4D: Proportions of CAR8 with the effector memory precursor (EMP, CD27+/CD62L-) or central memory precursor (CMP, CD27+/CD62L+) phenotypes at the indicated timepoint against WT leukemia. S4E: Proportions of CAR8 with the effector memory precursor (EMP, CD27+/CD62L-) or central memory precursor (CMP, CD27+/CD62L+) phenotypes at the indicated timepoint against CD19<sup>lo</sup> leukemia. Data in S4C-F are from 2 pooled, independent experiments with n=10 mice per condition. Data represent mean +/- SD. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001.
+
+<--- Page Split --->
+![PLACEHOLDER_40_0]
+
+
+<--- Page Split --->
+
+Figure S5: Additional in vivo effector/memory and exhaustion phenotyping at low CAR dose (Related to Figure 3). All data in this figure are from experiments with the 3e5 EGFR+ cell dose. S5A: Proportions of CAR8 with the short-lived effector cell (SLEC, IL7Ra-/KLRG1+) or memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) phenotypes at the indicated timepoint against WT leukemia. S5B: Proportions of CAR8 with the short-lived effector cell (SLEC, IL7Ra-/KLRG1+) or memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) phenotypes at the indicated timepoint against CD19<sup>lo</sup> leukemia. S5C: Proportions of CAR8 with the effector memory precursor (EMP, CD27+/CD62L-) or central memory precursor (CMP, CD27+/CD62L+) phenotypes at the indicated timepoint against WT leukemia. S5D: Proportions of CAR8 with the effector memory precursor (EMP, CD27+/CD62L-) or central memory precursor (CMP, CD27+/CD62L+) phenotypes at the indicated timepoint against CD19<sup>lo</sup> leukemia. Figures S4E-L display proportions of CAR8 with the indicated phenotype at 11 days post-CAR injection against either WT (left, E,F,I,J) or CD19<sup>lo</sup> (right, G,H,K,L) leukemia. S5E & G: PD1+/TOX+ S5F & H: PD1+/CD39+ S5I & K: TCF1+/TIM3- S5J & L: TCF1-/TIM3+. Data are from 2 pooled, independent experiments with \(n = 10\) mice per condition. Data represent mean \(+ / -\) SD. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) .
+
+<--- Page Split --->
+![PLACEHOLDER_42_0]
+
+<center>Figure S6 </center>  
+
+![PLACEHOLDER_42_1]
+
+<center>Comparisons of Naive T cell derived populations </center>  
+
+![PLACEHOLDER_42_2]
+
+<center>Comparisons of Memory T cell derived populations </center>  
+
+![PLACEHOLDER_42_3]
+
+<center>Comparisons between Naive and Memory T cell derived populations </center>  
+
+![PLACEHOLDER_42_4]
+
+<center>D. </center>  
+
+![PLACEHOLDER_42_5]
+
+
+<--- Page Split --->
+
+1 Figure S6: Additional analyses of ATAC-seq data (Related to Figure 4).2 All analyses in this figure are from the same timeline/experimental layout described in Figure 4A. S6A: Inter-3 replicate Euclidian distance of boom-normalized ATAC-seq counts per peak between biological replicates. S6B: Pairwise comparisons of differentially accessible chromatin regions within conditions between different 5 timepoints of the same condition, or between different conditions at each timepoint. Data points are mean of 6 boom-normalized ATAC-seq counts per peak between biological replicates of each group. S6C: Heatmap of 7 motif-associated ChromVAR deviation z-scores patterns of motif-associated ATAC-seq signal for indicated 8 transcription factors. List comprises all significant differentially accessible comparisons. S6D: Representative 9 gating for sorting of cells in sequencing experiments.
+
+<--- Page Split --->
+![PLACEHOLDER_44_0]
+
+
+<--- Page Split --->
+
+## Figure S7: Additional analyses of RNA-seq data (Related to Figure 5).  
+
+All analyses in this figure are from the same timeline/experimental layout described in Figure 4A. S7A: Normalized enrichment scores from GSEA of differentially enriched genesets between indicated CD8+ T cell subsets after LCMV- armstrong acute viral infection<sup>24, 25</sup>. S7B: Top transcriptional activators predicted to be activated and driving differential transcriptional state between naive versus memory- derived cells at the indicated timepoint, as predicted by Qiagen Ingenuity Pathway Analysis<sup>28</sup> (IPA). S7C: IPA activation map for the Cebpb transcription factor, the top predicted driver of transcriptional state in memory- derived cells at the PostCAR and Tumor timepoints. S7D-F: Top differentially expressed transcription factors, at the indicated timepoint. All statistics performed using DESeq2 with filtering threshold at 10, log2foldchange >2 and padj > 0.05.
+
+<--- Page Split --->
+![PLACEHOLDER_46_0]
+
+
+<--- Page Split --->
+
+1 Figure S8 (Related to Figure 6): Characterization of 1e5 CAR T cell dose in vivo experiments and in vitro comparisons of BATF, JUN or RUNX2 overexpressing cells to pMIG. 2 All analyses in this figure are the same timeline/experimental layout described in Figure 3A except with 1e5 CAR+ cell dose, at 11 days post- CAR timepoint. S8A- B are characterization of the 1e5 cell dose with standard T cell groups (no ectopic transcription factor expression). S8A: Leukemia burden. S8B: Proportions of CAR8 with the PD1+/TOX+ phenotype. Data in S7A- B are from 1 experiment with n=5 mice per condition. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. S8C- D: Quantification of intracellular cytokine staining of IFNg and TNFa after 6 hour co- culture assay, % positive of EGFR+, for memory (C) or naive- derived (D) cells cotransduced with BATF, JUN or pMIG. Data in S8C- D are from 3 independent experiments. S8E: Proliferation as measured by dilution of CellTrace Violet dye dilution of EGFR+ cells after 72 hour co- culture assay, for memory or naive derived cells cotransduced with BATF, JUN or pMIG. Data representative of 3 independent experiments. S8F- G: Quantification of intracellular cytokine staining of IFNg and TNFa after 6 hour co- culture assay, % positive of EGFR+, for memory (C) or naive- derived (D) cells cotransduced with RUNX2 or pMIG. Data in S8C- D are from 3- 4 independent experiments. S8H: Proliferation as measured by dilution of CellTrace Violet dye dilution of EGFR+ cells after 72 hour co- culture assay, for memory or naive derived cells cotransduced with RUNX2 or pMIG. Data representative of 3 independent experiments. No statistically significant differences were found between BATF, JUN or RUNX2 engineered CAR T cells and pMIG control T cells for in vitro data. Data represent mean +/- SD.
+
+<--- Page Split --->
+![PLACEHOLDER_48_0]
+
+
+<--- Page Split --->
+
+1 Figure S9 (Related to Figure 6): Counts and additional exhaustion phenotyping data for BATF, JUN or RUNX2 overexpression in vivo experiments. 2 All analyses in this figure are done on the same experiments described in Figure 6. Counts data was generated by flushing a single tibia and using total tibia counts and cytometer proportions data to calculate CAR and leukemia cell counts per tibia. S9A- B: CAR and leukemia counts for BATF or JUN overexpressing memory (A) or naive- derived (B) CAR T cells compared to pMIG control. S9C- H: Proportions of EGFR+ cells from BATF, JUN or pMIG CAR8 with the indicated phenotype. S9C,D,G are memory- derived cells, S9E,F,H are naive- derived cells. S9C,E: PD1+ S9D,F: PD1+/CD39+ S9G- H: Indicated TCF1/TIM3 phenotype. Data in S9A- H are from one experiment with n=5 mice per condition. S9I- J: CAR and leukemia counts for RUNX2 overexpressing memory (A) or naive- derived (B) CAR T cells compared to pMIG control. S9K- N: Proportions of EGFR+ cells from RUNX2 or pMIG CAR8 with the indicated phenotype. S9C,D,G are memory- derived cells, S9E,F,H are naive- derived cells. S9K,M: PD1+ S9L,N: PD1+/CD39+ S9O,P: Indicated TCF1/TIM3 phenotype. Data in S9I- P are from 2 pooled, independent experiments with n=9 mice per condition. Data represent mean +/- SD. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001.
+
+<--- Page Split --->

preprint/preprint__48b464b07d1a9fb2491f3443e735e9f38e6b8abe90e64efd27de8474ea209aaf/preprint__48b464b07d1a9fb2491f3443e735e9f38e6b8abe90e64efd27de8474ea209aaf_det.mmd

@@ -0,0 +1,697 @@
+<|ref|>title<|/ref|><|det|>[[42, 108, 882, 207]]<|/det|>
+# Antigen experience history directs distinct functional states of CD8+ CAR T cells during the anti-leukemia response  
+
+<|ref|>text<|/ref|><|det|>[[44, 230, 323, 275]]<|/det|>
+Terry Fry  TERRY.FRY@CUANSCHUTZ.EDU  
+
+<|ref|>text<|/ref|><|det|>[[45, 303, 490, 345]]<|/det|>
+University of Colorado Anschutz Medical Campus  Kole DeGolier  
+
+<|ref|>text<|/ref|><|det|>[[45, 348, 848, 369]]<|/det|>
+University of Colorado Anschutz Medical Campus https://orcid.org/0000- 0002- 4144- 7277  
+
+<|ref|>text<|/ref|><|det|>[[44, 374, 488, 416]]<|/det|>
+Etienne Danis  University of Colorado Anschutz Medical Campus  
+
+<|ref|>text<|/ref|><|det|>[[44, 421, 316, 462]]<|/det|>
+Marc D'Antonio  University of Colorado School  
+
+<|ref|>text<|/ref|><|det|>[[44, 468, 848, 509]]<|/det|>
+Jennifer Cimons  University of Colorado Anschutz Medical Campus https://orcid.org/0000- 0002- 8111- 8697  
+
+<|ref|>text<|/ref|><|det|>[[44, 513, 848, 555]]<|/det|>
+Michael Yarnell  University of Colorado Anschutz Medical Campus https://orcid.org/0000- 0002- 4557- 7255  
+
+<|ref|>text<|/ref|><|det|>[[44, 560, 848, 602]]<|/det|>
+Ross Keld  University of Colorado Anschutz Medical Campus https://orcid.org/0000- 0002- 9065- 7895  
+
+<|ref|>text<|/ref|><|det|>[[44, 606, 422, 647]]<|/det|>
+Mark kohler  University of Colorado School of Medicine  
+
+<|ref|>text<|/ref|><|det|>[[44, 653, 260, 694]]<|/det|>
+James Scott- Browne  National Jewish Health  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 735, 103, 752]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 772, 137, 790]]<|/det|>
+Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 810, 348, 829]]<|/det|>
+Posted Date: December 21st, 2023  
+
+<|ref|>text<|/ref|><|det|>[[44, 848, 475, 867]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 3712137/v1  
+
+<|ref|>text<|/ref|><|det|>[[42, 886, 912, 928]]<|/det|>
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License.  Read Full License
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[41, 44, 931, 156]]<|/det|>
+Additional Declarations: Yes there is potential Competing Interest. Patent applicant (whether author or institution): University of Colorado Anschutz Medical Campus Name of inventor(s): Kole R DeGolier, James P Scott- Browne, Terry J Fry Application number: U.S. Provisional Application No. 63/595,612 Status of application: Pending Aspect covered: Methods of enhancing potency of engineered immune cells via Runx2 modulation  
+
+<|ref|>text<|/ref|><|det|>[[41, 191, 950, 234]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Immunology on January 2nd, 2025. See the published version at https://doi.org/10.1038/s41590-024-02034-1.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[55, 42, 940, 87]]<|/det|>
+# Antigen experience history directs distinct functional states of CD8+ CAR T cells during the anti-leukemia response  
+
+<|ref|>text<|/ref|><|det|>[[55, 111, 940, 159]]<|/det|>
+Kole R. DeGolier \(^{1,2}\) , Etienne Danis \(^{3}\) , Marc D'Antonio \(^{1}\) , Jennifer Cimons \(^{1,2}\) , Michael Yarnell \(^{2,4}\) , Ross M. Kedl \(^{1}\) , M. Eric Kohler \(^{1,2,4}\) , James P. Scott- Browne \(^{1,5}\) , Terry J. Fry \(^{1,2,4}\)  
+
+<|ref|>text<|/ref|><|det|>[[55, 185, 943, 325]]<|/det|>
+\(^{1}\) Department of Immunology, University of Colorado Anschutz Medical Campus; Aurora, CO, USA \(^{2}\) Department of Pediatrics, University of Colorado Anschutz Medical Campus; Aurora, CO, USA \(^{3}\) Biostatistics and Bioinformatics Shared Resource, University of Colorado Cancer Center, University of Colorado Anschutz Medical Campus; Aurora, CO, USA \(^{4}\) Center for Cancer and Blood Disorders, Children's Hospital Colorado; Aurora, CO, USA \(^{5}\) Department of Immunology and Genomic Medicine, National Jewish Health, Denver, CO, USA  
+
+<|ref|>text<|/ref|><|det|>[[55, 379, 238, 398]]<|/det|>
+Corresponding author:  
+
+<|ref|>text<|/ref|><|det|>[[55, 424, 722, 567]]<|/det|>
+Terry J Fry, MD  University of Colorado Anschutz Medical Campus and Children's Hospital Colorado  13123 E 16 \(^{\text{th}}\) Avenue  Aurora, Colorado 80045  Phone: 303.724.7293  Email: terry.fry@cuanschutz.edu  
+
+<|ref|>text<|/ref|><|det|>[[55, 592, 940, 637]]<|/det|>
+Running title: Antigen experience history directs distinct functional states of CD8+ CAR T cells during the anti- leukemia response  
+
+<|ref|>text<|/ref|><|det|>[[55, 664, 217, 683]]<|/det|>
+Nature Immunology  
+
+<|ref|>text<|/ref|><|det|>[[55, 710, 662, 856]]<|/det|>
+Abstract word count: 200 (200)  Manuscript word count: 4147 (4300)  Tables and Figure count: 6 figures, no tables (5- 6 modest length (1/4 page))  Extended Data: 9 figures, no tables (10)  References: 39 (50)  Methods References: 12
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[55, 43, 158, 60]]<|/det|>
+## ABSTRACT  
+
+<|ref|>text<|/ref|><|det|>[[52, 66, 944, 375]]<|/det|>
+Chimeric antigen receptor T cells are an effective therapy for B- lineage malignancies. However, many patients relapse and this therapeutic has yet to show strong efficacy in other hematologic or solid tumors. One opportunity for improvement lies in the ability to generate T cells with desirable functional characteristics. Here, we dissect the biology of CD8+ CAR T cells (CAR8) by controlling whether the T cell has encountered cognate TCR antigen prior to CAR generation. We find that prior antigen experience influences multiple aspects of in vitro and in vivo CAR8 functionality, resulting in superior effector function and leukemia clearance in the setting of limiting target antigen density compared to antigen- inexperienced T cells. However, this comes at the expense of inferior proliferative capacity, susceptibility to phenotypic exhaustion and dysfunction, and inability to clear wildtype leukemia in the setting of limiting CAR+ cell dose. Epigenomic and transcriptomic comparisons of these cell populations identified overexpression of the Runx2 transcription factor as a novel strategy to enhance CAR8 function, with a differential impact depending on prior cell state. Collectively, our data demonstrate that prior antigen experience determines functional attributes of a CAR T cell, as well as amenability to functional enhancement by transcription factor modulation.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[48, 68, 944, 688]]<|/det|>
+relapsed and treatment- refractory B- lineage hematologic malignancies. However, many patients do not achieve complete remission, or relapse. Poor response or lack of remission durability results from cancer cell resistance or suboptimal CAR T cell function'. Thus, further studies into the immunobiology of these engineered cells are warranted to enhance remissions and expand therapeutic potential to other hematologic and solid tumors. CAR T cells are commonly generated from a heterogeneous population of peripheral blood T cells that varies between patients, likely impacting the quality of a CAR T cell product?. Although it has been difficult to track cell fate through the manufacturing process and into patients, previous reports have shown differential function of CAR T cell products generated from memory versus naive T cells sorted by surface marker phenotypes, which are not always an accurate representation of cellular differentiation state2,3,4. Emerging studies have demonstrated that phenotypic, transcriptomic and epigenomic attributes of the CAR product can influence patient outcomes5. During acute infections, naive CD8+ T cells become activated through the T cell antigen receptor (TCR) by antigen presenting cells displaying cognate antigen and co- stimulatory ligands, and subsequently enter a highly regulated differentiation trajectory. A phase of rapid expansion and differentiation into effector cells is followed by contraction and formation of long- lived memory cells that rapidly respond to future exposures. However, if the pathogen is not cleared, antigen- specific T cell populations will receive recurring antigen stimulation. In this setting, rather than forming functional memory, T cells differentiate down a trajectory characterized by progressive dysfunction, preventing immune- mediated pathology, but simultaneously failing to clear the challenge. A growing body of work demonstrates that these differentiation trajectories (and resulting functional characteristics imbued on T cells) are controlled epigenetically in traditional T cell responses to viral infections and tumors. These programs are defined by progressive changes to the epigenome, associated with DNA methylation and histone modifications which are driven by a variety of transcription factors (TFs) and modulated by antigen receptor signaling6. These molecular modifications alter chromatin accessibility and transcriptional profiles which characterize cellular differentiation state and functional capacity. Epigenetic modulation of T cells via stimulation through the physiologic TCR has a well- established role in impacting the differentiation program and functional capacity of a pool of antigen- experienced T cells7. Emerging data also highlight the importance of epigenetic remodeling in CAR T cell responses to tumors5.  
+
+<|ref|>text<|/ref|><|det|>[[52, 688, 945, 949]]<|/det|>
+Here, we carefully examine and compare the biology of CAR- transduced CD8+ T cells that differ as to whether cognate antigen has been encountered through the TCR prior to transduction with a CAR. We hypothesize that 1) T cells exhibit functional characteristics after CAR transduction that are dictated by prior antigen experience via the TCR 2)the functional characteristics of CAR8 derived from naive or memory cells are the result of epigenetic attributes maintained through CAR transduction and reinfusion, and that 3) TF modulation as a modality to enhance CAR8 function may be dependent on the epigenetic and transcriptomic contexts determined by prior antigen experience status. Prior work has shown dose- dependent effects in the anti- tumor responses of adoptively- transferred T cells2 and CAR T cells have been shown to elicit poor responses to tumors with low antigen density1,8,9,10. Using limiting target antigen density or limiting T cell dose as stressors, we show that prior T cell antigen experience influences in vitro and in vivo functional characteristics of T cells stimulated through a CAR. Comparison of the epigenetic and transcriptomic states of CAR8 stratified by prior antigen- experience
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 42, 943, 135]]<|/det|>
+status revealed differential chromatin accessibility and transcriptional programming. We pinpoint divergent RUNX2 activity within the two populations as a potential driver of differential function and show that ectopic expression of RUNX2 enhances the anti- leukemia response and mediates exhaustion resistance in CAR T cells in a manner dependent on prior T cell antigen experience status.  
+
+<|ref|>sub_title<|/ref|><|det|>[[56, 163, 142, 180]]<|/det|>
+## RESULTS  
+
+<|ref|>text<|/ref|><|det|>[[55, 185, 943, 228]]<|/det|>
+T cell antigen experience prior to transduction with a CAR directs in vitro proliferative and effector capacities of CD8+ CAR T cells.  
+
+<|ref|>text<|/ref|><|det|>[[52, 225, 944, 880]]<|/det|>
+Memory T cells demonstrate superior antigen sensitivity compared to naive T cells in some contexts<sup>11,12</sup>. Thus, we hypothesized that CAR T cells derived from a memory T cell population would exhibit enhanced responsiveness to low antigen density leukemias compared to naive- derived CAR T cells. T cells expressing a CAR containing an anti- mouse CD19 scFv incorporating a FLAG sequence and a CD28 costimulatory domain fused to mouse CD3zeta, followed by a 2A sequence and a truncated EGFR<sup>13</sup> (Figure S1A) were used to target a murine leukemia driven by the E2A- PBX1 fusion protein (E2A- PBX)<sup>14, 15, 16</sup>. FLAG- specific antibody detection of the CAR correlated strongly with EGFR expression, allowing for use of EGFR as a marker for long term tracking of CAR+ cells in vivo (Figure S1B). We expanded this model by generating a set of clones of E2A- PBX which express differing CD19 densities (Figure 1A, S1C). Memory OT- I T cells generated using a well- characterized ovalbumin vaccination model<sup>17, 18, 19</sup> (Figure 1B). were used to produce memory- derived CD8+ CAR T cells (CAR<sub>MD</sub>) for comparison to naive- derived CD8+ OT- I CAR T cells (CAR<sub>ND</sub>). As no difference was seen in leukemia control by memory or naive- derived control T cells (Figure S1D), we used naive- derived (EGFR8) in all subsequent experiments. A functional duality began to emerge upon in vitro testing. As predicted, a greater proportion of CAR<sub>MD</sub> cells had a polyfunctional effector profile, producing both TNFa and IFNg, or degranulating (as measured by CD107a), most pronounced in response to low target antigen (Figure 1C- H; S1E- G). Interestingly, while the proportion of IFNg+ cells was greater in CAR<sub>MD</sub>, the proportion of TNFa+ cells was slightly increased in CAR<sub>ND</sub>, suggesting a predisposition toward either IFNg or TNFa (Figure 1C & F). However, CAR<sub>ND</sub> outperformed CAR<sub>MD</sub> in cell cycle entry (Ki67 expression; Figure 1I, S1H) and extended proliferative capacity (Figure 1J, S1I) across antigen densities. To compare polyclonal antigen- experienced and naive T cells more analogous to human CAR T cells, we generated pathogen- elicited polyclonal T cells by infecting WT C57BL/6 mice with the common acute viral infection model LCMV- Armstrong. Memory (CD8+/CD44+/CD49d<sup>hi</sup>) and naive (CD8+/CD44-/CD49d<sup>lo</sup>/CD62L+) T cell populations were FACS- sorted from the same mice 28 days after LCMV infection and used for CAR T cell manufacturing (Figure S2A). Polyclonal pathogen- elicited T cells behaved similarly in vitro to memory and naive OT- I cells: CAR<sub>MD</sub> demonstrated superior effector function (increased proportions of cells producing IFNg) and CAR<sub>ND</sub> demonstrated superior proliferative capacity (Figure S2B- E). Thus, CD8+ T cell antigen experience prior to transduction with a CAR promotes effector functions at the expense of proliferative capacity.  
+
+<|ref|>text<|/ref|><|det|>[[55, 904, 942, 949]]<|/det|>
+Treatment of leukemia- bearing mice with a high CAR+ cell dose reveals enhanced cytotoxic profile and clearance of antigen- low leukemia by memory- derived CAR8.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[50, 42, 944, 640]]<|/det|>
+Given the opposing functional profiles of naive and memory- derived CAR8, we next compared the ability of these two populations to mediate tumor clearance in vivo. Mice were engrafted with WT (35,000 antigens per cell), CD19<sup>lo</sup> (10,000 antigens per cell) or CD19<sup>Neg</sup> leukemia followed 3 days later by a dose of 1e6 CAR T cells. The CD19<sup>lo</sup> clone antigen density was chosen based on differential in vitro responses and, although higher than antigen density reported for CAR relapses post- CD22 CAR treatment<sup>9</sup>, is consistent with the drop- off in CAR sensitivity against other antigens<sup>8, 10</sup>. Rag1- deficient hosts enabled CAR T cell expansion without irradiation and limited CAR T cell antigen exposure to CD19 densities expressed on leukemia rather than endogenous B cells. While we did not observe differences in proportions of CAR T cells in the marrow at peak expansion on day 4 (Figure 2A), post- contraction (day 11) CAR8<sub>ND</sub> had increased proportions and total counts of CAR T cells in mice bearing WT and CD19<sup>lo</sup> leukemia (Figure 2B- C, Figure S3A- B). Both CAR groups mediated robust clearance of WT leukemia by day 11. Although there was no significant difference in clearance of CD19<sup>lo</sup> leukemia, 4/10 mice treated by CAR8<sub>ND</sub> had detectable leukemia at \(>15\%\) of live bone marrow cells while all 10 mice treated with CAR8<sub>MD</sub> had minimal leukemic burdens (<5%) (Figure 2D). We next tested whether the enhanced clearance of CD19<sup>lo</sup> leukemia was associated with maintenance of the superior cytotoxic capacity of CAR8<sub>MD</sub> observed in vitro. Upon ex vivo restitution of CAR8 in the bone marrow, we found that, while IFNg production was highly variable, GZMB production was markedly greater in CAR8<sub>MD</sub> (Figure 2E- F). CAR8<sub>MD</sub> had significantly higher proportions of cells falling into short- lived effector cell (SLEC, IL7Ra-/KLRG1+) and effector memory precursor (EMP, CD27+/CD62L-) phenotypes, fewer cells in the central memory precursor phenotype (CMP, CD27+/CD62L+), and no change in memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) populations (Figure S4B- E). Additionally, early expression of effector- associated TFs IRF4, T- bet and EOMES was greater in CAR8<sub>MD</sub> (Figure 2G- I). Finally, while mice bearing WT high- antigen leukemia showed no survival difference after treatment with CAR8<sub>MD</sub> versus CAR8<sub>ND</sub>, mice bearing CD19<sup>lo</sup> leukemia treated with CAR8<sub>MD</sub> showed a significant survival benefit, with 20% of mice surviving to the 80 day experimental endpoint (Figure 2J). Together, these data show that CAR8<sub>MD</sub> mediate superior clearance of CD19<sup>lo</sup> leukemia relative to CAR8<sub>ND</sub>, associated with maintenance of effector function and expression of effector- associated markers.  
+
+<|ref|>sub_title<|/ref|><|det|>[[55, 664, 940, 707]]<|/det|>
+## Treatment of leukemia-bearing mice with a low CAR+ cell dose reveals enhanced proliferative capacity and clearance of WT leukemia by naive-derived CAR8.  
+
+<|ref|>text<|/ref|><|det|>[[50, 712, 944, 949]]<|/det|>
+We next hypothesized that the benefit of enhanced proliferative capacity of naive- derived CAR8 would emerge at a lower CAR+ cell dose (3e5). As anticipated, CAR8<sub>ND</sub> expanded to significantly higher numbers in the bone marrow by day 4 regardless of leukemia antigen density, mirroring in vitro proliferative assays (Figure 3A- B, S3C, 11- J). While CAR8<sub>ND</sub> mediated enhanced clearance and survival in mice bearing WT leukemia, there was no improvement in leukemia clearance or survival of mice bearing CD19<sup>lo</sup> leukemia (Figure 3C, 3I, S3D), potentially due to reduced potency. Indeed, ex vivo IFNg production was greater in CAR8<sub>MD</sub>, although there was no difference in GZMB production or expression of IRF4, T- bet or EOMES (Figure 3D- H). CAR8<sub>MD</sub> consistently demonstrated significantly higher proportions of SLECs at the early timepoint consistent with high CAR doses, but these differences disappeared by day 11 and no differences were seen in the MPEC population (Figure S5A,B). While EMP and CMP patterns mimicked high dose experiments, the differences were much less
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[52, 42, 944, 277]]<|/det|>
+pronounced, indicating that naive- derived cells largely became more "effector- like" with greater proliferative drive (Figure S5C,D), consistent with effector- polarization in the setting of low numbers of antigen- specific precursor populations \(^{20,21}\) . However, these changes, combined with the strong expansion, did not mediate survival benefit against CD19 \(^{1,0}\) leukemia (Figure 3l). Finally, we predicted that at this lower cell dose, T cell dysfunction could emerge. Indeed, CAR \(_{8\mathrm{MD}}\) expressed higher levels of exhaustion- associated markers against WT leukemia with failure of CAR \(_{8\mathrm{MD}}\) to control leukemia (Figure S5E- F, I- J). Interestingly, we found that CD19 \(^{1,0}\) leukemia drove similar proportions of exhaustion phenotypes in both CAR8 populations, demonstrating that chronic, uncleared antigen exposure, even at low antigen density, can drive dysfunction (Figure S5G- H, K- L). These findings highlight the importance of proliferative capacity and resistance to dysfunction afforded by CAR \(_{8\mathrm{ND}}\) at limiting cell dose.  
+
+<|ref|>sub_title<|/ref|><|det|>[[55, 304, 941, 348]]<|/det|>
+## Epigenetic profiling of naive and memory-derived CAR8 shows differential chromatin accessibility at binding sites for bZIP, Tcf, Runx and other TF families.  
+
+<|ref|>text<|/ref|><|det|>[[48, 350, 944, 952]]<|/det|>
+We predicted that functional traits were a product of distinct epigenetic states, given that functional distinctions of naive and memory- derived CAR8 were dictated by status prior to CAR transduction. To test this, we performed bulk ATAC- seq on naive and memory- derived cells at three timepoints: ex vivo prior to CAR transduction (Day - 5, "PreCAR"), in vitro after CAR transduction (Day 0, "PostCAR"), and after reinfusion into mice bearing CD19 \(^{1,0}\) leukemia (Day 4, "Tumor") (Figure 4A). Comparison of experimental replicates showed tight concordance of chromatin accessibility at each condition and timepoint (Figure S6A). Broadly, the data showed several thousand differentially accessible regions between either cell type compared to itself across timepoints, and between naive and memory- derived cells at each timepoint (Figure S6B). We found predictable patterns of ATAC- seq signal at genetic loci involved in T cell activation or effector function, including higher accessibility in CAR \(_{8\mathrm{MD}}\) at Gzmb, Gzmc, and the Pdcd1 loci encoding for the PD1 protein. Concurrently, we found greater accessibility in CAR \(_{8\mathrm{ND}}\) at the Tcf7 loci encoding TCF1, a TF important for maintaining self- renewal capacity (Figure 4B). We used ChromVAR \(^{22}\) , to associate these changes in chromatin accessibility to previously defined datasets and potential TF activities. Based on relative chromatin accessibility at regions that were differentially accessible in a published comparison of effector and memory CD8+ T cells after acute viral infection with LCMV- Armstrong \(^{23}\) , memory- derived CAR8 acquired effector- associated changes in chromatin accessibility during CAR generation in culture that were maintained after transfer into tumor bearing mice. CAR8 generated from memory T cells also had reduced chromatin accessibility at features associated with memory T cells. By comparison, naive- derived CAR8 maintained chromatin accessibility patterns at regions associated with memory T cells and showed minimal skewing toward an effector- like profile \(^{23}\) (Figure 4C). To associate these changes with specific TF activities, we used ChromVAR to compare chromatin accessibility at regions containing DNA sequence motifs bound by different TFs (Figure 4D). Classifying this data using a kmeans clustering strategy, we found that there were distinct patterns of motif- associated chromatin accessibility between conditions and across each of the timepoints (Figure 4E). While motifs for bZIP and Irf family TFs broadly looked similar at the PreCAR timepoint, and became progressively enriched in memory cells, Tcf family motifs started similar and became enriched in naive cells at the latter timepoints, while E2A family motifs started highly enriched in naive and progressively
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 44, 943, 110]]<|/det|>
+converged. Uniquely, motifs for Runx family members were always more accessible in memory- derived cells and did not converge or diverge (Figure 4D- E, S6C). Overall, these data show epigenetic features imprinted in the starting CD8+ T cell population are maintained through CAR engineering.  
+
+<|ref|>text<|/ref|><|det|>[[55, 123, 940, 860]]<|/det|>
+Prior antigen experience directs distinct transcriptomic patterns of naïve and memory- derived CAR8. To test whether the epigenetic states of naïve and memory- derived CAR8 resulted in concurrent transcriptomic changes, we performed bulk RNA- seq at the same timepoints as for ATAC- seq (Figure 4A). We found predictable differential gene expression at each timepoint, with genes associated with self- renewal and proliferative capacity (Lef1, Sell, Id3, Tcf7, Slamf6, Il7r) upregulated in the naïve- derived cells and genes associated with effector capacity and activation (Prf1, Ifng, Klrg1, Gzmb, Prdm1, Id2, Pdcd1, Tbx21) upregulated in the memory- derived cells (Figure 5A). Gene set enrichment analysis (GSEA) showed progressive bias by normalized enrichment score (NES) toward effector- like in memory- derived CAR8, and toward memory- like in naïve- derived CAR8<sup>24, 25, 26</sup> (Figure 5B- C). Analysis with gene sets comparing memory and naïve T cells showed progressive decrease in the normalized enrichment score of memory or naïve- derived CAR8 toward the derivative cell population of each, suggesting the effector/memory gene set enrichment axis as the more accurate indicator of cell fate over time<sup>24, 25</sup> (Figure S7A). Looking at the top differentially- expressed TFs between the populations at the PreCAR timepoint, we found many expected hits, including Bhlhe40, Klf4, Tbx21, Id2 and many bZIP family members (Jun, JunB, Fos, Cebpb) represented in the memory- derived group, while Zeb1, Myb and Lef1, encoding TFs associated with self- renewal, were upregulated in the naïve- derived cells<sup>23, 27</sup> (Figure S7D). Notably, among the Runx family, which showed uniquely stable differential motif accessibility between naïve and memory cells (Figure 4D), Runx2 was among the most differentially expressed TF genes with marked overexpression in memory derived cells (Figure S7D). Ingenuity Pathway Analysis of global transcriptional profile implicated similar TF drivers<sup>28</sup> (Figure S7B) with numerous distinct patterns of differential TF expression between memory and naïve- derived T cells. However, a very common pattern among ChromVAR- implicated TFs was high initial expression in memory cells at the PreCAR timepoint, followed by a convergence in expression between memory and naïve- derived CAR T cells at the PostCAR and Tumor timepoints, as seen with bZIP family members Jun, Fos and Atf3, along with the gene Tbx21, encoding canonical effector TF T- bet (Figure 5D). Among the Runx family, Runx1 and Runx3 gene expression tracked relatively closely between memory and naïve- derived cells at each timepoint, while Runx2 followed the “high in memory, then converging” pattern which was commonly found among other TF families (Figure 5E). In summary, naïve and memory- derived T cells show differential gene expression and gene set association with self- renewal or memory- associated genes and activation or effector- associated genes, respectively. Many relevant TF genes show a pattern of high initial expression in memory cells at the PreCAR timepoint which converges between the cell derivations upon transduction with a CAR and reinfusion into tumor- bearing hosts.  
+
+<|ref|>text<|/ref|><|det|>[[55, 881, 941, 924]]<|/det|>
+RUNX2 overexpression boosts leukemia clearance, CAR T cell potency and CAR proportions in bone marrow.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[50, 42, 944, 202]]<|/det|>
+To validate the epigenetic and transcriptomic data, we overexpressed two TFs from the ChromVAR- implicated bZIP family, BATF and c- Jun, both of which have been previously reported to impact CAR T cell function (Figure 6A- B) \(^{29,30,31}\) . Although neither TF increased cytokine production or proliferation in vitro (Figure S8C- E), overexpression of either TF enhanced leukemia clearance by memory and naive- derived CAR T cells (Figure 6C- D). There was no difference between BATF- CAR8 or JUN- CAR8 and control CAR8 in the PD1+ proportion (Figure S9C,E), co- expression of PD1 with markers of exhaustion (PD1+/CD39+ and PD1+/TOX+), or in the terminally exhausted Tcf1-/Tim3+ population (Figure S9D,F- H).  
+
+<|ref|>text<|/ref|><|det|>[[50, 201, 944, 904]]<|/det|>
+Due to the memory- like state of CAR8 \(_{\mathrm{ND}}\) , we anticipated that comparison of factors enriched in memory cells over naive cells could reveal important drivers of memory cell function that were not fully induced in naive cells during the synthetic engineering process. Given the unique profile of chromatin accessibility for Runx- family binding motifs coupled with the pattern of Runx2 transcript expression which was high in PreCAR memory CD8+ T cells and then lost upon CAR transduction, we hypothesized that establishing RUNX2 expression in CAR8 \(_{\mathrm{ND}}\) could enhance the existing memory- like profile of these T cells and boost T cell potency and anti- leukemia response. Murine RUNX2 was introduced into the pMSCV- IRES- eGFP (pMIG) backbone, containing a GFP reporter gene for long- term tracking of RUNX2- transduced T cell populations (RUNX2). Co- transduction of naive CD8+ T cells with CAR- EGFR reporter and RUNX2- GFP reporter resulted in a large proportion of cells expressing both EGFR and GFP (Figure 6A). Upon intracellular staining for the RUNX2 protein, we found that the EGFR+ population in the RUNX2- transduced group showed approximately a 10- fold increase in RUNX2 expression relative to empty pMIG- transduced cells (Figure 6B). Co- culture of RUNX2- CAR8 and leukemia with a range of antigen densities revealed similar cytokine production and proliferation relative to pMIG- CAR8 (Figure 6C- D). To stress the ability of RUNX2- CAR8 to clear WT leukemia, we used an ultra- low CAR+ dose (1e5), against which both CAR8 \(_{\mathrm{ND}}\) and CAR8 \(_{\mathrm{MD}}\) exhibit markers of exhaustion and fail to control leukemia (Figure S7A- C). RUNX2 overexpression in CAR8 \(_{\mathrm{ND}}\) strongly enhanced leukemia clearance and increased CAR proportions and absolute numbers in the marrow at 11 days post- CAR infusion (Figure 6E- F). While there was no difference in the PD1+ proportion, consistent with similar activation, mice treated with RUNX2- CAR8 \(_{\mathrm{ND}}\) exhibited dramatically reduced proportion of PD1+/TOX+ cells, a lower proportion of PD1+/CD39+ cells and reduced proportions of TCF1-/TIM3+ cells, suggesting that RUNX2 overexpression counteracts the differentiation trajectory toward terminal exhaustion (Figure 6L, S9M- N,P) \(^{27,32}\) . CAR8 \(_{\mathrm{MD}}\) showed less of an increase in RUNX2 following transduction with RUNX2- eGFP (Figure S7F) potentially due to higher RUNX2 at baseline (Figure 5E). Nonetheless, RUNX2- overexpression resulted in a significant reduction in the PD1+/CD39+ exhaustion phenotype of RUNX2- CAR8 \(_{\mathrm{MD}}\) responding to WT leukemia and reduction in leukemia counts in marrow (Figure S9I) but no difference in other exhaustion phenotypes, CAR proportions or CAR counts (Figure 6K, S9K,L,O). We demonstrate that Runx2 overexpression in naive- derived T cells enhances maintenance of CAR T cells in the marrow, boosts leukemia clearance and mediates a favorable exhaustion profile at a highly sub- curative CAR T cell dose with less impact in memory- derived CAR T cells, demonstrating that TF overexpression has a differential impact depending on starting T cell state.  
+
+<|ref|>sub_title<|/ref|><|det|>[[55, 930, 170, 947]]<|/det|>
+## DISCUSSION
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[52, 42, 944, 396]]<|/det|>
+Factors underlying tumor relapse after CAR T cell therapy are a central focus of study in the field of cell therapies for leukemia. Advances have been made in understanding and engineering solutions to prevent tumor cell escape via antigen modulation, T cell dysfunction, and poor T cell trafficking/persistence'. However, defining in vitro and in vivo functional strengths and cellular profiles associated with different starting T cell populations may be an opportunity to specifically identify approaches to arm CAR T cells to overcome different tumor escape modalities. Importantly, refining qualities of the starting cell population will likely be a large contributor to efficacy of cellular therapeutics derived from healthy allogeneic donors or induced pluripotent stem cells, or in the case of in vivo transduction platforms targeting genetic payloads to specific cell populations. Recent work has sought to use targeted modulation of TFs to enhance CAR T cell function or prevent dysfunction, with several publications focusing on the bZIP TF family, including forced expression of BATF and c- Jun, or genetic deletion of the Nr4a family of nuclear receptors29, 30, 31, 33, 34. However, the impact of modulation of the bZIP family has been variable. Therefore, we set out to characterize functional attributes programmed by prior T cell antigen experience, with the prediction that these would be tied to epigenetic traits. We anticipated that downstream modulation of TFs implicated by this framework might have divergent functional outcomes depending on starting cell population.  
+
+<|ref|>text<|/ref|><|det|>[[52, 400, 944, 635]]<|/det|>
+In this study, we use a syngeneic murine model with anti- mouse CD19 CAR T cells targeting murine pre- B cell leukemia enabling more natural T cell differentiation trajectories without xenogeneic TCR stimulation. We also used a well- defined vaccine model for precise control of the antigen experience history of CAR T cells with a clonotypic TCR, with confirmation in a polyclonal memory response. With limiting T cell dose or low target antigen density as "stressors," we report that antigen experience dictates multiple functional outputs of CAR T cells. Memory- derived CAR T cells exhibited stronger cytotoxic function across target antigen densities, while naive- derived CAR T cells show greater proliferative capacity and more rapid cell cycle entry. This was associated with enhanced activity against low- antigen density leukemia by memory derived CAR T cells and enhanced activity of naive- derived cells at limiting cell dose, a setting that drove phenotypic exhaustion and dysfunction of memory- derived cells.  
+
+<|ref|>text<|/ref|><|det|>[[52, 639, 944, 876]]<|/det|>
+T cell differentiation is a product of epigenetic and transcriptomic state23, 27 and while CAR T cells have been extensively profiled post- manufacturing, little work has been done to characterize effects of prior T cell state on post- transduction CAR T cell profiles. We demonstrate that features of these states are maintained through CAR manufacturing and associate with differences in functional profiles. Specifically, we find significant differences in bZIP family transcription factors, which have been previously implicated in CAR T cell function29, 30, 31. BATF or JUN mediated enhanced leukemia clearance in our model independent of starting cell state, indicating that these TFs may derive most of their early in vivo activity via binding to NFAT- AP1 composite motifs, which show high accessibility in both cell types. Surprisingly, there was no difference in phenotypic exhaustion in BATF or JUN- overexpressing CAR T cells relative to control, indicating preservation of function in an exhausted state rather than prevention of exhaustion.  
+
+<|ref|>text<|/ref|><|det|>[[52, 880, 943, 949]]<|/det|>
+As a novel finding, we use epigenomic and transcriptomic assays and implicate modulation of Runx- family TFs, particularly Runx2, as having a likelihood for higher impact in naive- derived cells compared to memory. Ectopic RUNX2 expression in naive- derived CAR T cells resulted in superior clearance of leukemia, higher proportions
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 0, 945, 970]]<|/det|>
+of cells in the marrow, and reduced proportions of cells displaying terminally exhausted phenotypes relative to control. Our data suggest that RUNX2 overexpression, in contrast to overexpression of bZIP family members, can enhance functional potency of naive- derived CD8+ CAR T cells while preventing entry into the exhaustion differentiation trajectory. In addition to their activity as transcriptional activators, Runx family members have been shown to recruit chromatin remodeling factors to Runx binding sites to open these sites and allow for transcriptional activation. In other model systems, RUNX2 has been shown to interact with SWI/SNF complexes, histone acetyltransferases (MOZ, p300), histone deacetylases (HDAC3, HDAC4, HDAC6) and histone methyltransferases (SUV39H1), along with all three TET family enzymes, indicating a plausibility for the ability for RUNX2 to recruit enzymes which participate in chromatin remodeling at RUNX binding motifs<sup>35, 36, 37, 38</sup>. These features could help explain the contribution of RUNX2 overexpression to the enhanced functionality and exhaustion resistance of CAR<sub>8ND</sub> seen in our experiments. Additional studies will be necessary to fully elucidate the effects of RUNX2 in CAR T cells, and to confirm our findings in human CAR T cells. Nonetheless, using a model in which antigen history can be precisely controlled, we show that RUNX2 overexpression enhances in vivo CAR T cell function dependent on the starting T cell. Finally, we have generated a framework for the role of antigen experience on function of a CAR T cell in stress situations of limiting T cell dose or target antigen density and highlight the importance of considering this framework when assessing the impact of approaches to apply synthetic immunology to manipulate therapeutic immune effector cell functions. METHODS See Supplemental Material. AUTHOR CONTRIBUTIONS K.R.D. Conceptualized the studies, performed experiments and data analysis, and wrote the manuscript. E.D. Performed data analysis and provided expertise related to ATAC/RNA sequencing. M.D. Performed experiments. J.C. Conceptualized the studies and provided expertise related to the murine CAR and leukemia models. M.Y. Designed and generated DNA constructs. R.M.K. Provided expertise related to the vaccine model. M.E.K. Conceptualized the studies and provided expertise related to the murine CAR and leukemia models. J.P.S-B. Performed data analysis and provided expertise related to ATAC/RNA sequencing. T.J.F. Conceptualized, supervised and provided funding for the studies, and wrote the manuscript. All authors contributed to the article and approved the submitted version. ACKNOWLEDGEMENTS We thank Lillie Leach for laboratory management, Amanda Novak for animal colony management, Garrett Hedlund and Henry Chu at the CU Anschutz Clinical Immunology Flow Core for their assistance in cell sorting, the CU Cancer Center Genomics Shared Resource (RRID: SCR_021984) for their help with sequencing/genomics, and the CU Anschutz OLAR and the animal facility for their support. This work was funded in part by Department of Defense W81XWH-19-1-0196 and partly supported by the National Institutes of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 42, 941, 87]]<|/det|>
+Health P30CA046934 Bioinformatics and Biostatistics Shared Resource Core (RRID: SCR_021983). Some figures were created with BioRender.com.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 115, 300, 133]]<|/det|>
+## SUPPLEMENTAL METHODS  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 163, 182, 180]]<|/det|>
+## Mouse Strains  
+
+<|ref|>text<|/ref|><|det|>[[55, 185, 944, 350]]<|/det|>
+Mouse StrainsB6.129S6- Rag2tm1Fwa Tg(TcraTcrb)1100Mjb ("OT- I," Model #: 2334- F) mice were obtained from Taconic Biosciences. B6. SJL- Ptprca Pepcb/BoyJ ("PepBoy," Strain #: 002014), B6.129S7- Rag1tm1Mom/J ("Rag1- ", Strain #: 002216), C57BL/6J mice ("B6," Strain #: 000664) were obtained from The Jackson Laboratory. Female mice were used for all experiments with B6 background mice. All mice were bred and/or maintained in the animal facility at University of Colorado Anschutz Medical Campus. All experiments were performed in compliance with the study protocol approved by University of Colorado Anschutz Medical Campus Institutional Animal Care and Use Committee (IACUC).  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 379, 260, 396]]<|/det|>
+## Mouse CAR Constructs  
+
+<|ref|>text<|/ref|><|det|>[[54, 401, 944, 565]]<|/det|>
+Mouse CAR ConstructsThe basic construction of the murine 1928z CAR was previously described<sup>39</sup>. The murine anti- CD19 scFv was Flag- tagged to enable CAR detection, and all ITAMs in the CD3zeta domain were kept intact. A truncated human EGFR reporter protein was incorporated following a 2A skip sequence to provide an additional method for detection of CAR- transduced cells<sup>13</sup>. The DNA was codon optimized, ordered from ThermoFisher GeneArt, and cloned into the MSCV- IRES- GFP backbone, a gift from Tannishtha Reya (Addgene plasmid # 20672; http://n2t.net/addgene:20672; RRID:Addgene_20672), using Xhol and Clal enzyme sites. A control plasmid with just the truncated EGFR reporter in the MSCV backbone was generated using similar methods.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 594, 232, 611]]<|/det|>
+## Cell lines and media  
+
+<|ref|>text<|/ref|><|det|>[[55, 616, 943, 732]]<|/det|>
+Cell lines and mediaE2A- PBX pre- B cell acute lymphoblastic leukemia was developed in the laboratory as previously described<sup>14, 15, 16</sup>. Murine T cells and leukemia were cultured in Complete Mouse Media (CMM), consisting of RPMI 1640 medium (Gibco) with 10% heat- inactivated fetal calf serum (Omega Bio), 1% nonessential amino acids (Gibco), 1% sodium pyruvate (Gibco), 1% penicillin/streptomycin (Gibco), 1% L- glutamine (Gibco), 1% HEPES buffer (Gibco) and 50uM 2- mercaptoethanol (Sigma- Aldrich).  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 761, 278, 778]]<|/det|>
+## Mouse CAR Transduction  
+
+<|ref|>text<|/ref|><|det|>[[54, 783, 944, 949]]<|/det|>
+Mouse CAR TransductionCAR transduction was performed as previously described<sup>14, 15, 16</sup>. Briefly, spleens from 6- 10 week old donor mice were harvested and CD8+ T cells were isolated using EasySep Mouse CD8+ T cell Isolation Kit from STEMCell Technologies or bulk T cells were isolated using the Mouse CD3+ T Cell Enrichment Column Kit (R&D Biosciences, Cat No. MTCC- 25). On day 1, T cells were activated on anti- CD3/anti- CD28 Mouse T cell Activator DynaBeads (Invitrogen) at a 1:1 cell:bead ratio and cultured at 1e6/mL in CMM in the presence of rhIL- 2 (40IU/mL) and rhIL- 7(10ng/mL) from R&D Systems. On days 2 and 3, retroviral supernatant was added to Retro nectin- coated (Takara Biosciences) 6 well plates and spun at 2000xg and 32°C for 2- 3 hours. Supernatant
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[54, 42, 944, 136]]<|/det|>
+was then removed and activated T cells were added to the wells at 1.67mL/well. On day 4, beads were removed and T cells were resuspended at 1e6/mL in fresh media with cytokines. CAR transduction was determined post- debeading by analyzing T cells by flow cytometry for a FLAG/EGFR double- positive population (or EGFR single- positive for control T cells), and T cells were used in assays or infused into mice on day 5 or 6.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 163, 184, 180]]<|/det|>
+## Vaccine Model  
+
+<|ref|>text<|/ref|><|det|>[[52, 185, 944, 445]]<|/det|>
+The ovalbumin vaccine consists of 100ug whole ovalbumin protein (InvivoGen, Cat. code: vac- pova- 100), 40ug anti- mouse CD40 (BioXCell, Catalog #BE0016- 2) and 40ug Polyinosinic:polycytidylic acid [Poly (l:C)] (InvivoGen, Cat. code: tlrl- pic- 5) per mouse, resuspended to 200uL total volume in PBS<sup>17, 18, 19</sup>. CD8+ T cells were isolated from naive 6 to 8 week old OT- I mouse splenocytes using the Mouse CD3+ T Cell Enrichment Column Kit (R&D Biosciences, Cat No. MTCC- 25). PepBoy mice were given 5e3 OT- I T cells retro- orbitally and concurrently vaccinated intravenously. 3- 4 weeks later, spleens from 5- 20 vaccinated mice were pooled and CD45.2+ OT- I memory T cells were isolated using the EasySep Mouse CD8+ T cell Isolation Kit, followed by column isolation using biotinylated anti- mouse CD45.2 (BioLegend, Cat # 109804), LS Columns (Miltenyi Biotec, Order No. 130- 042- 401), and anti- Biotin MicroBeads (Miltenyi Biotec, Order No. 130- 090- 485). Naive T cells from 1- 5 naive OT- I donors were isolated in parallel. T cells were then activated and transduced as described for downstream experiments.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 473, 470, 492]]<|/det|>
+## Generation of CD19<sup>lo</sup> E2A-PBX leukemia cell lines  
+
+<|ref|>text<|/ref|><|det|>[[51, 497, 944, 830]]<|/det|>
+The E2A- PBX murine leukemia was generated in our lab as previously described<sup>14</sup>. CD19 knockout leukemia was produced using CRISPR/Cas9. A previously- validated murine CD19- targeting sgRNA<sup>15</sup> from Integrated DNA Technologies was incubated with recombinant Cas9 from TakaraBio (Cat# 632641) to create an RNP complex. RNP was then electroporated into E2A- PBX using the Lonza 4D- Nucleofector X with nucleofector solution SG and pulse program CM- 147. Electroporated cells were allowed to recover for 48 hours and then FACS- sorted twice to obtain a pure CD19 knockout cell line. This cell population was additionally single cell cloned to create a CD19 knockout single cell clone prior to transduction with murine CD19. A truncated/non- signaling murine CD19 was cloned into the pLV.SP146.gp91.GP91.cHS4 plasmid, a gift from Didier Trono (Addgene plasmid # 30480; http://n2t.net/addgene:30480; RRID:Addgene_30480). Backbones were generated with the hEF1a promoter (pLV.hEF1a.cHS4) or the hUbC promoter (pLV.hUbC.cHS4) from the pLenti6/UbC/mSlc7a1 plasmid, a gift from Shinya Yamanaka (Addgene plasmid # 17224; http://n2t.net/addgene:17224; RRID:Addgene_17224). VSV- G pseudotyped lentivirus was generated as described and E2A- PBX CD19KO underwent a single round of transduction using standard protocols, followed by single cell cloning to obtain clonally- derived lines expressing defined levels of CD19 target antigen.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 858, 194, 875]]<|/det|>
+## Flow Cytometry  
+
+<|ref|>text<|/ref|><|det|>[[52, 881, 943, 949]]<|/det|>
+Flow cytometry analysis was performed using an LSR- Fortessa X- 20 flow cytometer (BD Biosciences) and analyzed using FlowJo (BD Biosciences). Monoclonal antibodies used in staining are listed in the supplemental methods. Intracellular flow cytometry staining was performed using the TrueNuclear Transcription Factor Buffer
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 42, 943, 110]]<|/det|>
+Set (BioLegend) for ex vivo staining of transcription factors, Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences) for intracellular cytokine staining, and Mouse Foxp3 Buffer Set (BD Biosciences) for intracellular staining of Ki67 and Runx2.  
+
+<|ref|>text<|/ref|><|det|>[[55, 138, 943, 326]]<|/det|>
+CD107a Degranulation, Intracellular Cytokine Staining (ICCS), Ki67 and CellTrace Dilution In Vitro Assays In vitro assays were performed using a 1:1 effector to target cell ratio with 1e5 of each cell type in a 96- well round- bottom plate followed by analysis by flow cytometry at the indicated timepoints. Degranulation assays were performed by incubation for 4 hours in the presence of 2uM monensin and 1uL of CD107a antibody. ICCS was performed by incubation for 6 hours, with 1uM monensin and 2.5uM Brefeldin A added at 1 hour in. Ki67 was performed by incubation for 18 hours, followed by intracellular staining for Ki67. CellTrace dilution assays were performed by staining T cells with CellTrace Violet (Thermo Fisher Scientific) per manufacturer protocols followed by incubation with target cells for 72 hours.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 355, 357, 372]]<|/det|>
+## LCMV infection and T cell isolation  
+
+<|ref|>text<|/ref|><|det|>[[55, 377, 943, 493]]<|/det|>
+6 week old female C57BL/6 mice were injected retro- orbitally with 2e5 PFU of LCMV- Armstrong. 4 weeks later, CD8+ T cells were isolated from 5 pooled spleens using the EasySep Mouse CD8+ T cell Isolation Kit from STEMCell Technologies and then FACS- sorted to obtain Memory (CD8+/CD44+/CD49dhi) and Naive (CD8+/CD44-/CD49d0/CD62L+) populations from the same mice. T cells were then transduced using the standard transduction protocol as described.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 522, 360, 540]]<|/det|>
+## In vivo experiments in Rag1- hosts  
+
+<|ref|>text<|/ref|><|det|>[[54, 545, 944, 732]]<|/det|>
+Experiments were carried out using a timeline previously optimized in the lab<sup>14</sup>. Briefly, Rag1- hosts were inoculated with 1e6 E2A- PBX by tail vein I.V. injection on day - 3 followed by CAR T cells via retroorbital injection at either 1e5, 3e5 or 1e6 CAR+ cell dose on day 0. Bone marrow was harvested and analyzed by flow cytometry on day 4 or 11 post- CAR infusion, or mice were euthanized at humane endpoints for survival experiments. Ex vivo stimulation for cytokine production was performed using 1e6 E2A- PBX WT to stimulate approximately 1.5e6 whole bone marrow cells from each individual mouse, with pooled bone marrow from each n=5 experimental group stimulated by E2A- PBX CD19<sup>Neg</sup> as a negative control. Cells were co- cultured for 6 hours, with 1uM monensin and 2.5uM Brefeldin A added at 1 hour in and then analyzed by flow for cytokine production.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 761, 633, 779]]<|/det|>
+## Bulk ATAC and RNA sequencing experimental setup and workflows  
+
+<|ref|>text<|/ref|><|det|>[[54, 784, 944, 950]]<|/det|>
+OT- I CD8+ T cells were isolated from vaccinated or naive donors and CARs were transduced into T cells as described above. CAR8 Rag1- hosts were inoculated with 1e6 E2A- PBX CD19<sup>10,000</sup> followed by 1e6 CAR8<sub>MD</sub> or CAR8<sub>ND</sub> on the timeline described above. At day 4 post- CAR infusion, bone marrow from 10 mice per CAR group was harvested and pooled. At each of 3 timepoints, CD8+ cells were isolated using the EasySep Mouse CD8+ T cell Isolation Kit from STEMCell Technologies and then FACS- sorted to obtain 50,000 cells per condition. ATAC- seq and RNA- seq were performed in triplicate on separate sorted aliquots of 50,000 cells at "Pre- CAR/Day -5" (ex vivo, directly after isolation of memory or naive CD8+ T cells from donor mice), "Post- CAR/Day 0" (in
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 42, 944, 570]]<|/det|>
+vitro, after CAR manufacturing) and "Tumor/Day 4" (ex vivo, after reinfusion into leukemia bearing mice). Experimental analyses were performed on the first technical replicate from 2 separate experimental replicates. For RNA-seq, cells were homogenized in QIAzol Lysis Reagent (Qiagen, Cat. No. 79306) and then frozen at - 80C for processing within 2 weeks. Samples were thawed and processed using the miRNeasy Micro Kit (Qiagen, Mat. No. 1071023), with on- column DNase treatment (RNase- Free DNase Set, Qiagen, Cat. No. 79254), both according to manufacturer protocols. RNA purity, quantity and integrity was determined with NanoDrop (ThermoFisher Scientific) and TapeStation 4200 (Agilent) analysis prior to RNA-seq library preparation. The Universal Plus mRNA-Seq library preparation kit with NuQuant was used (Tecan) with an input of 200ng of total RNA to generate RNA-seq libraries. Paired- end sequencing reads of 150bp were generated on NovaSeq 6000 (Illumina) sequencer at a target depth of 40 million clusters/80 million paired- end reads per sample. Raw sequencing reads were de- multiplexed using bcl2fastq. For ATAC- seq, cells were immediately processed using the Omni- ATAC protocol as previously described<sup>40</sup>. Briefly, sorted cells were washed once in 1X PBS, lysed, washed once in Wash Buffer and then the transposition reaction was carried out at 32°C for 30 minutes on a thermomixer set to 1000 rpm. Transposed chromatin was then purified using the Zymo Clean and Concentrator 5 Kit (Zymo Research, Cat # D4013) using manufacturer protocols. DNA was then ran on PCR for 12 total cycles with matched barcoding primers<sup>41</sup>. PCR reactions were then size- selected using AMPure XP beads (Beckman Coulter Life Sciences, Product No: A63880) and checked for quality and size distribution using TapeStation 4200 with D5000 reagents (Agilent). Libraries were pooled at equimolar ratios for sequencing and paired- end sequencing reads of 150bp for the first replicate and 50bp for the second replicate were generated on NovaSeq 6000 (Illumina) sequencer at a target depth of 40 million clusters/80 million paired- end reads per sample. Raw sequencing reads for replicate 1 were shortened to match the read lengths for replicate 2 using trimmomatic function CROP. Raw sequencing reads were de- multiplexed using bcl2fastq.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 595, 258, 612]]<|/det|>
+## RNA-seq Data Analysis  
+
+<|ref|>text<|/ref|><|det|>[[55, 617, 944, 830]]<|/det|>
+Quality of fastq files was accessed using FastQC (v.0.11.8) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc), FastQ Screen (v.0.13.0)<sup>42</sup> and MultiQC (v.1.8)<sup>43</sup>. Illumina adapters and low- quality reads were filtered out using BBDuk (v. 38.87) (http://jgi.doe.gov/data- and- tools/bb- tools). Trimmed fastqc files were aligned to the mm10 murine reference genome and aligned counts per gene were quantified using STAR (v.2.7.9a)<sup>44</sup>. Differential gene expression analysis was performed using the DESeq2 package<sup>45</sup>. Pathway enrichment analysis was performed using GSEA (UC San Diego/Broad Institute)<sup>26</sup>,<sup>46</sup>, Metascape<sup>47</sup> for gene mapping and IPA (Qiagen)<sup>28, 48</sup>. Differential gene expression was plotted using GraphPad Prism or ggplot2 (R package). RNA- seq differential gene expression statistics were run using the DESeq2 R package, with filtering threshold at 10 with greater than 2- fold change and adjusted p value < 0.05.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 858, 265, 875]]<|/det|>
+## ATAC-seq Data Analysis  
+
+<|ref|>text<|/ref|><|det|>[[55, 881, 943, 949]]<|/det|>
+Fastq files were used to map to the mm10 genome using the ENCODE ATAC- seq pipeline (https://www.encodeproject.org/atac- seq/), with default parameters, except bam files used for peak calling were randomly downsampled to a maximum of 50 million mapped reads. Peaks with a
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[52, 42, 944, 326]]<|/det|>
+MACS2(https://pypi.org/project/MACS2/) computed q value of less than 1e- 6 and a signalValue of more than 4 in at least one replicate were merged with bedtools<sup>49</sup> function intersect and processed to uniform peaks with the functions getPeaks and resize from R package ChromVAR<sup>22</sup>. Reads overlapping peaks were enumerated with getCounts function from ChromVAR and normalized and log2- transformed with loom from R package limma<sup>50</sup>. Peaks with 3 or more normalized counts per million mapped reads at least one replicate were included to define a global peak set of 82,410 peaks. Pairwise Euclidean distances were computed between all samples using log2- transformed counts per million mapped reads among the global peak set. Differentially accessible peaks were identified in pairwise comparisons based on fdr adjusted p values of less than 0.01, fold change of at least 4 and with an average of 3 normalized counts per million mapped reads using R package limma. Motif associated variability in ATAC- seq signal was computed with R package ChromVAR. Genome- wide visualization of ATAC- seq coverage was computed with deeptools<sup>51</sup> function coveragebam, using manually computed scale factors based on the number of reads within the total peak set.  
+
+<|ref|>sub_title<|/ref|><|det|>[[56, 355, 140, 371]]<|/det|>
+## Statistics  
+
+<|ref|>text<|/ref|><|det|>[[52, 378, 945, 660]]<|/det|>
+Statistical tests for all experiments except sequencing analyses were performed using GraphPad Prism v9.0 for Macintosh (GraphPad Software). Comparisons between three groups were made with ordinary one- way ANOVA with Holm- Sidak's multiple comparisons test, Brown- Forsythe and Welch one- way ANOVA with Dunnett's T3 multiple comparisons test, or Kruskal- Wallis non- parametric test with Dunn's multiple comparisons test were used depending on variance in standard deviations. Two- way ANOVA or mixed effects analysis with Tukey's multiple comparisons test was used for in vitro experimental comparisons with multiple antigen densities and in vivo CAR expansion data. Two- tailed ordinary t test, Welch's t test or Mann- Whitney test were performed for comparisons with two groups depending on normality of distributions. For multiple comparisons of two groups, multiple unpaired t tests or multiple Welch's t tests, both with Holm- Sidak's multiple comparisons test, were performed when appropriate depending on variance in standard deviations. Log- rank (Mantel- Cox) test was used for survival curve comparisons. All data represented as mean +/- standard deviation. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. Technical and experimental replicates in each dataset are indicated in figure legends.  
+
+<|ref|>sub_title<|/ref|><|det|>[[57, 688, 316, 706]]<|/det|>
+## Data and Materials Availability  
+
+<|ref|>text<|/ref|><|det|>[[56, 712, 943, 780]]<|/det|>
+All data is readily available from authors upon request or accessible at Gene Expression Omnibus (GEO Accession Number will be provided before paper acceptance). All materials are either commercially available as described or available from authors upon request.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 997]]<|/det|>
+1. Labanieh, L. & Mackall, C.L. CAR immune cells: design principles, resistance and the next generation. Nature 614, 635-648 (2023).  
+2. Sommermeyer, D. et al. Chimeric antigen receptor-modified T cells derived from defined CD8+ and CD4+ subsets confer superior antitumor reactivity in vivo. Leukemia 30, 492-500 (2016).  
+3. Qin, H. et al. CAR T cells targeting BAFF-R can overcome CD19 antigen loss in B cell malignancies. Sci Transl Med 11 (2019).  
+4. Aldoss, I. et al. Favorable Activity and Safety Profile of Memory-Enriched CD19-Targeted Chimeric Antigen Receptor T-Cell Therapy in Adults with High-Risk Relapsed/Refractory ALL. Clin Cancer Res 29, 742-753 (2023).  
+5. Alvanou, M. et al. Empowering the Potential of CAR-T Cell Immunotherapies by Epigenetic Reprogramming. Cancers (Basel) 15 (2023).  
+6. Frias, A.B., Boi, S.K., Lan, X. & Youngblood, B. Epigenetic regulation of T cell adaptive immunity. Immunol Rev 300, 9-21 (2021).  
+7. Wherry, E.J. & Kurachi, M. Molecular and cellular insights into T cell exhaustion. Nat Rev Immunol 15, 486-499 (2015).  
+8. Heitzeneder, S. et al. GPC2-CAR T cells tuned for low antigen density mediate potent activity against neuroblastoma without toxicity. Cancer Cell 40, 53-69.e59 (2022).  
+9. Fry, T.J. et al. CD22-targeted CAR T cells induce remission in B-ALL that is naive or resistant to CD19-targeted CAR immunotherapy. Nat Med 24, 20-28 (2018).  
+10. Majzner, R.G. et al. Tuning the Antigen Density Requirement for CAR T-cell Activity. Cancer Discov 10, 702-723 (2020).  
+11. Kumar, R. et al. Increased sensitivity of antigen-experienced T cells through the enrichment of oligomeric T cell receptor complexes. Immunity 35, 375-387 (2011).  
+12. Mehlhop-Williams, E.R. & Bevan, M.J. Memory CD8+ T cells exhibit increased antigen threshold requirements for recall proliferation. J Exp Med 211, 345-356 (2014).  
+13. Wang, X. et al. A transgene-encoded cell surface polypeptide for selection, in vivo tracking, and ablation of engineered cells. Blood 118, 1255-1263 (2011).  
+14. Qin, H. et al. Murine pre-B-cell ALL induces T-cell dysfunction not fully reversed by introduction of a chimeric antigen receptor. Blood 132, 1899-1910 (2018).  
+15. Jacoby, E. et al. CD19 CAR immune pressure induces B-precursor acute lymphoblastic leukaemia lineage switch exposing inherent leukaemic plasticity. Nat Commun 7, 12320 (2016).  
+16. Yang, Y. et al. TCR engagement negatively affects CD8 but not CD4 CAR T cell expansion and leukemic clearance. Sci Transl Med 9 (2017).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 44, 930, 81]]<|/det|>
+17. Ivanova, D.L. et al. Vaccine adjuvant-elicited CD8(+) T cell immunity is co-dependent on T-bet and FOXO1. Cell Rep 42, 112911 (2023). 
+
+<|ref|>text<|/ref|><|det|>[[55, 97, 880, 133]]<|/det|>
+18. Klarquist, J. et al. B cells promote CD8 T cell primary and memory responses to subunit vaccines. Cell Rep 36, 109591 (2021). 
+
+<|ref|>text<|/ref|><|det|>[[55, 149, 875, 185]]<|/det|>
+19. Klarquist, J. et al. Clonal expansion of vaccine-elicited T cells is independent of aerobic glycolysis. Sci Immunol 3 (2018). 
+
+<|ref|>text<|/ref|><|det|>[[55, 201, 940, 254]]<|/det|>
+20. Badovinac, V.P., Haring, J.S. & Harty, J.T. Initial T cell receptor transgenic cell precursor frequency dictates critical aspects of the CD8(+) T cell response to infection. Immunity 26, 827-841 (2007). 
+
+<|ref|>text<|/ref|><|det|>[[55, 270, 940, 306]]<|/det|>
+21. Marzo, A.L. et al. Initial T cell frequency dictates memory CD8+ T cell lineage commitment. Nat Immunol 6, 793-799 (2005). 
+
+<|ref|>text<|/ref|><|det|>[[55, 322, 904, 375]]<|/det|>
+22. Schep, A.N., Wu, B., Buenrostro, J.D. & Greenleaf, W.J. chromVAR: inferring transcription-factor-associated accessibility from single-cell epigenomic data. Nat Methods 14, 975-978 (2017). 
+
+<|ref|>text<|/ref|><|det|>[[55, 392, 932, 427]]<|/det|>
+23. Scott-Browne, J.P. et al. Dynamic Changes in Chromatin Accessibility Occur in CD8(+) T Cells Responding to Viral Infection. Immunity 45, 1327-1340 (2016). 
+
+<|ref|>text<|/ref|><|det|>[[55, 444, 930, 479]]<|/det|>
+24. Wherry, E.J. et al. Molecular signature of CD8+ T cell exhaustion during chronic viral infection. Immunity 27, 670-684 (2007). 
+
+<|ref|>text<|/ref|><|det|>[[55, 496, 910, 548]]<|/det|>
+25. Luckey, C.J. et al. Memory T and memory B cells share a transcriptional program of self-renewal with long-term hematopoietic stem cells. Proc Natl Acad Sci U S A 103, 3304-3309 (2006). 
+
+<|ref|>text<|/ref|><|det|>[[55, 565, 904, 618]]<|/det|>
+26. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 102, 15545-15550 (2005). 
+
+<|ref|>text<|/ref|><|det|>[[55, 634, 928, 669]]<|/det|>
+27. Giles, J.R. et al. Human epigenetic and transcriptional T cell differentiation atlas for identifying functional T cell-specific enhancers. Immunity 55, 557-574.e557 (2022). 
+
+<|ref|>text<|/ref|><|det|>[[55, 686, 866, 721]]<|/det|>
+28. Krämer, A., Green, J., Pollard, J., Jr. & Tugendreich, S. Causal analysis approaches in Ingenuity Pathway Analysis. Bioinformatics 30, 523-530 (2014). 
+
+<|ref|>text<|/ref|><|det|>[[55, 738, 911, 790]]<|/det|>
+29. Zhang, X. et al. Depletion of BATF in CAR-T cells enhances antitumor activity by inducing resistance against exhaustion and formation of central memory cells. Cancer Cell 40, 1407-1422.e1407 (2022). 
+
+<|ref|>text<|/ref|><|det|>[[55, 808, 934, 842]]<|/det|>
+30. Seo, H. et al. BATF and IRF4 cooperate to counter exhaustion in tumor-infiltrating CAR T cells. Nat Immunol 22, 983-995 (2021). 
+
+<|ref|>text<|/ref|><|det|>[[55, 860, 916, 895]]<|/det|>
+31. Lynn, R.C. et al. c-Jun overexpression in CAR T cells induces exhaustion resistance. Nature 576, 293-300 (2019). 
+
+<|ref|>text<|/ref|><|det|>[[55, 912, 936, 947]]<|/det|>
+32. van der Heide, V., Humblin, E., Vaidya, A. & Kamphorst, A.O. Advancing beyond the twists and turns of T cell exhaustion in cancer. Sci Transl Med 14, eabo4997 (2022).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[50, 60, 940, 98]]<|/det|>
+33. Chen, J. et al. NR4A transcription factors limit CAR T cell function in solid tumours. Nature 567, 530-534 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[50, 112, 936, 150]]<|/det|>
+34. Jung, I.Y. et al. BLIMP1 and NR4A3 transcription factors reciprocally regulate antitumor CAR T cell stemness and exhaustion. Sci Transl Med 14, eabn7336 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[50, 164, 927, 202]]<|/det|>
+35. Tandon, M. et al. Runx2 mediates epigenetic silencing of the bone morphogenetic protein-3B (BMP-3B/GDF10) in lung cancer cells. Mol Cancer 11, 27 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[50, 216, 896, 254]]<|/det|>
+36. Hojo, H. et al. Runx2 regulates chromatin accessibility to direct the osteoblast program at neonatal stages. Cell Rep 40, 111315 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[50, 268, 897, 306]]<|/det|>
+37. Wang, L. et al. TET enzymes regulate skeletal development through increasing chromatin accessibility of RUNX2 target genes. Nat Commun 13, 4709 (2022).  
+
+<|ref|>text<|/ref|><|det|>[[50, 320, 930, 358]]<|/det|>
+38. Korinfskaya, S., Parameswaran, S., Weirauch, M.T. & Barski, A. Runx Transcription Factors in T Cells-What Is Beyond Thymic Development? Front Immunol 12, 701924 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[50, 372, 920, 428]]<|/det|>
+39. Kochenderfer, J.N., Yu, Z., Frasheri, D., Restifo, N.P. & Rosenberg, S.A. Adoptive transfer of syngeneic T cells transduced with a chimeric antigen receptor that recognizes murine CD19 can eradicate lymphoma and normal B cells. Blood 116, 3875-3886 (2010).  
+
+<|ref|>sub_title<|/ref|><|det|>[[55, 444, 297, 463]]<|/det|>
+## METHODS REFERENCES  
+
+<|ref|>text<|/ref|><|det|>[[50, 477, 860, 515]]<|/det|>
+40. Corces, M.R. et al. An improved ATAC-seq protocol reduces background and enables interrogation of frozen tissues. Nat Methods 14, 959-962 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[50, 529, 880, 567]]<|/det|>
+41. Buenrostro, J.D. et al. Single-cell chromatin accessibility reveals principles of regulatory variation. Nature 523, 486-490 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[50, 581, 886, 619]]<|/det|>
+42. Wingett, S.W. & Andrews, S. FastQ Screen: A tool for multi-genome mapping and quality control. F1000Res 7, 1338 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[50, 633, 900, 671]]<|/det|>
+43. Ewels, P., Magnusson, M., Lundin, S. & Käller, M. MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics 32, 3047-3048 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[50, 685, 904, 705]]<|/det|>
+44. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15-21 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[50, 719, 900, 757]]<|/det|>
+45. Love, M.I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[50, 771, 870, 809]]<|/det|>
+46. Zhu, A., Srivastava, A., Ibrahim, J.G., Patro, R. & Love, M.I. Nonparametric expression analysis using inferential replicate counts. Nucleic Acids Res 47, e105 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[50, 823, 911, 861]]<|/det|>
+47. Zhou, Y. et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat Commun 10, 1523 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[50, 875, 911, 914]]<|/det|>
+48. Mootha, V.K. et al. PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat Genet 34, 267-273 (2003).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 0, 936, 997]]<|/det|>
+49. Quinlan, A.R. & Hall, I.M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841-842 (2010).  
+50. Ritchie, M.E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43, e47 (2015).  
+51. Ramírez, F., Dündar, F., Diehl, S., Grüning, B.A. & Manke, T. deepTools: a flexible platform for exploring deep-sequencing data. Nucleic Acids Res 42, W187-191 (2014).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[58, 42, 146, 58]]<|/det|>
+# FIGURES 
+
+<|ref|>sub_title<|/ref|><|det|>[[72, 65, 148, 83]]<|/det|>
+## Figure 1 
+
+<|ref|>sub_title<|/ref|><|det|>[[85, 87, 370, 103]]<|/det|>
+### A. E2A-PBX Leukemia Molecules/cell 
+
+<|ref|>image<|/ref|><|det|>[[110, 102, 325, 253]]<|/det|>
+ 
+
+<|ref|>text<|/ref|><|det|>[[85, 268, 105, 283]]<|/det|>
+C. 
+
+<|ref|>sub_title<|/ref|><|det|>[[179, 260, 410, 273]]<|/det|>
+Increasing Target Antigen Expression 
+
+<|ref|>image<|/ref|><|det|>[[80, 275, 940, 960]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 42, 944, 350]]<|/det|>
+1 Figure 1: Antigen experience history directs multiple aspects of in vitro functional capacity of murine CD8+ CAR T cells. 2 1A: E2A- PBX murine leukemia was engineered to knockout CD19, followed by reintroduction of CD19 at different levels to generate a range of antigen density clones. 1B: Schematic: Vaccine model for generating memory CD8+ OT- I T cells. 5e3 OT- I T cells were transferred into congenically distinct hosts which were concurrently vaccinated with antigen and adjuvants. 3- 5 weeks later, CAR T cells were manufactured from memory OT- I's (CAR8MD, memory- derived) or naive OT- I's (CAR8ND, naive- derived) 1C: Intracellular cytokine staining of IFNg and TNFa after 6 hour co- culture assay. 1D: Degranulation as measured by CD107a expression after 4 hour coculture assay. 1E- G: Quantification of cytokine data, % positive cells for indicated cytokine. 1H: Quantification of CD107a data, % positive cells. 1I: Cell- cycle entry as measured by Ki- 67 staining after 18 hour co- culture assay. 1J: Proliferation as measured by dilution of CellTrace Violet dye after 72 hour co- culture assay. All in vitro assays were performed with \(n = 3\) technical replicates, and are representative of 2 independent experiments. Data represent mean \(+ / -\) SD. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) .
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[60, 78, 740, 181]]<|/det|>
+ 
+
+<|ref|>image_caption<|/ref|><|det|>[[66, 52, 150, 70]]<|/det|>
+<center>Figure 2</center> 
+
+<|ref|>image<|/ref|><|det|>[[99, 220, 860, 900]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+1 Figure 2: CAR8MD exhibit enhanced cytotoxicity and clearance of CD19<sup>lo</sup> leukemia in vivo (high CAR  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+2 dose).  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+3 2A: Schematic: Timeline for in vivo experiments. Rag1<sup>- t</sup> mice were injected with 1e6 E2A- PBX1 leukemia on  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+4 day - 3, followed by 1e6 OT- I CD8+/EGFR+ T cells from indicated T cell condition on day 0. Bone marrow was  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+5 analyzed by flow cytometry on day +4 or day +11. T cell populations were isolated memory- derived CAR T cells  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+6 (CAR8MD), isolated naive- derived CAR T cells (CAR8ND) or EGFR control T cells (EGFR8). Leukemia populations  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+7 were CD19<sup>Neg</sup>, CD19<sup>lo</sup>(10,000 antigens/cell), or WT (35,000 antigens/cell). 2B- C: Early T cell expansion (day  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+8 +4) or persistence (day +11) after infusion of transduced T cells against WT leukemia (B) and CD19<sup>lo</sup> leukemia  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+9 (C). Transduced T cell populations measured by coexpression of CD8a+/TCRbeta+/EGFR+. 2D: Clearance of  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+10 WT and CD19<sup>lo</sup> leukemia at day +11 after CAR infusion. E2A- PBX measured by coexpression of B220+/CD22+.  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+11 2E- F: Intracellular cytokine staining of interferon gamma (E) or granzyme B (F) in CAR T cells from whole bone marrow restimulated ex vivo with leukemia. Data represent mean +/- SD. 2G- I: Intranuclear transcription factor staining of IRF4 (G), EOMES (H), or T- bet (I) on CAR+ T cells from mice bearing the indicated leukemia at day +4 after CAR infusion. Violin plot data represent median with quartiles. Data are from 2 pooled, independent experiments with n=10 mice per condition. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. 2J: Survival of mice after treatment with 1e6 EGFR+ CAR or control T cells. Survival statistics were performed using log- rank (Mantel- Cox) test \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. Data is from 2 independent pooled experiments, total n=10 mice per group.  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+12 marrow restimulated ex vivo with leukemia. Data represent mean +/- SD. 2G- I: Intranuclear transcription factor  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+13 staining of IRF4 (G), EOMES (H), or T- bet (I) on CAR+ T cells from mice bearing the indicated leukemia at day  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+14 +4 after CAR infusion. Violin plot data represent median with quartiles. Data are from 2 pooled, independent  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+15 experiments with n=10 mice per condition. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. 2J: Survival of mice  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+16 after treatment with 1e6 EGFR+ CAR or control T cells. Survival statistics were performed using log- rank (Mantel-  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+17 Cox) test \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. Data is from 2 independent pooled experiments, total  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+18 n=10 mice per group.  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+19  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+20  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+21  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+22  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+23  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+24  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+25  
+
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 0]]<|/det|>
+26
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[70, 75, 888, 800]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 40, 945, 348]]<|/det|>
+Figure 3: CAR8ND exhibit enhanced expansion capacity and clearance of WT leukemia in vivo (low CAR dose).3A-B: Early T cell expansion (day +4) or persistence (day +11) after infusion of transduced T cells against WT leukemia (A) and CD19<sup>lo</sup> leukemia (B). Transduced T cell populations measured by coexpression of CD8a+/TCRbeta+/EGFR+. 3C: Clearance of WT and CD19<sup>lo</sup> leukemia at day +11 after CAR infusion. E2A-PBX measured by coexpression of B220+/CD22+. 3D-E: Intracellular cytokine staining of interferon gamma (D) or granzyme B (E) in CAR T cells from whole bone marrow restimulated ex vivo with leukemia. Data represent mean +/- SD. 2F-H: Intranuclear transcription factor staining of IRF4 (F), EOMES (G), or T-bet (H) on CAR+ T cells from mice bearing the indicated leukemia at day +4 after CAR infusion. Violin plot data represent median with quartiles. Data are from 2 pooled, independent experiments with \(n = 10\) mice per condition. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) . 2I: Survival of mice after treatment with 1e6 EGFR+ CAR or control T cells. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) . Data are from 2 independent pooled experiments, total \(n = 10\) mice per group.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[75, 92, 561, 290]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[62, 50, 144, 68]]<|/det|>
+<center>Figure 4 </center>  
+
+<|ref|>image<|/ref|><|det|>[[530, 90, 912, 290]]<|/det|>  
+
+<|ref|>image<|/ref|><|det|>[[70, 310, 920, 883]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 42, 944, 328]]<|/det|>
+Figure 4: Prior antigen experience imprints chromatin accessibility states which follow unique patterns during CAR transduction and reinfusion. 4A: Schematic: Layout for paired ATAC-seq/RNA-seq experiments. Memory-derived or naïve derived OT-I CD8+ T cells were sorted at three sequential timepoints: Ex vivo from donor mice before CAR transduction ("PreCAR"), in vitro after CAR transduction ("PostCAR"), and ex vivo after reinfusion into CD19<sup>Lo</sup> leukemia-bearing Rag1<sup>-/-</sup> mice ("Tumor"). 4B: Chromatin accessibility at Gzmb, Gzmc, Ifng, Tcf7 and Pdcd1 gene loci for naïve and memory-derived T cells at each timepoint. 4C: ChromVAR deviation z-scores between indicated populations at differentially accessible regions between Effector and Memory T cells after LCMV-Armstrong infection<sup>23</sup>. Data are mean +/- range of two biological replicates. 4D: Motif-associated ChromVAR deviation z-scores between indicated populations. Data are mean +/- range of two biological replicates 4E: K-means clustering of relative ATAC-seq signal at differentially accessible regions (top, data from two biological replicates are shown) and motif enrichment in each cluster vs all regions (bottom).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[128, 88, 911, 940]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[92, 52, 176, 70]]<|/det|>
+<center>Figure 5 </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 44, 944, 315]]<|/det|>
+Figure 5: Prior antigen experience drives differential CAR8 transcriptomic states which follow unique patterns during CAR transduction and reinfusion. RNA- seq analysis was run on the timepoints/conditions indicated in the previous figure. 5A: Volcano plots of significant differentially expressed genes between naïve and memory- derived cells at each of the three timepoints. 5B: Normalized enrichment scores from gene set enrichment analysis (GSEA) of differentially enriched genesets between indicated CD8+ T cell subsets after LCMV- Armstrong acute viral infection24 5C: GSEA plots at each timepoint. 5D: Top differentially expressed transcription factors at the "PreCAR" timepoint, generated using Ingenuity Pathway Analysis (IPA). 5E: DESeq2- normalized counts of indicated transcription factors at each timepoint for naïve and memory- derived cells. 5F: DESeq2- normalized counts of Runx family transcription factors at each timepoint for naïve and memory- derived cells. All statistics performed using DESeq2 with filtering threshold at 10, log2foldchange >2 and padj < 0.05.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[66, 70, 880, 940]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 42, 944, 351]]<|/det|>
+Figure 6: Runx2 overexpression as a novel strategy for enhancement of naïve-derived CD8+ CAR T cell potency and resistance to dysfunction. 6A-B: Cotransduction of memory (A) or naïve (B) CD8+ T cells with CAR and pMIG-Empty, pMIG- BATF, or pMIG- JUN. For 6C-F & I-L, Rag1- mice were given leukemia on day -3, followed by 1e5 pMIG- Runx2 or pMIG- Empty co- transduced CAR8 on day 0. Bone marrow was analyzed by flow cytometry on day 11 post- CAR. 6C & D: CAR T cell and leukemia proportions for naïve (C) and memory- derived (D) CAR T cells cotransduced with BATF, JUN or pMIG control. 6E & F: Proportion of CAR T cells displaying PD1+/TOX+ phenotype. 6G-H: Cotransduction of memory (G) or naïve (H) CD8+ T cells with CAR and pMIG-Empty or pMIG- Runx2 and intracellular staining for Runx2. 6I & J: CAR T cell and leukemia proportions for naïve (C) and memory- derived (D) CAR T cells cotransduced with RUNX2 or pMIG control. 6K & L: Proportion of CAR T cells displaying PD1+/TOX+ phenotype. Data in 6A,B,G & H are representative of 3-4 independent experiments. Data in 6C-F are from 1 experiment with n=5 mice per condition. Data in 6I-L are from 2 pooled, independent experiments with n=9 mice per condition. Data represent mean +/- SD. * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[85, 72, 560, 285]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[86, 50, 168, 66]]<|/det|>
+<center>Figure S1 </center>  
+
+<|ref|>image<|/ref|><|det|>[[62, 300, 940, 400]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[87, 300, 245, 312]]<|/det|>
+<center>D. Against WT Leukemia </center>  
+
+<|ref|>image<|/ref|><|det|>[[88, 500, 910, 600]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[62, 465, 83, 480]]<|/det|>
+<center>H. </center>  
+
+<|ref|>image<|/ref|><|det|>[[88, 612, 910, 712]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[88, 611, 201, 620]]<|/det|>
+<center>Antigen Density </center>  
+
+<|ref|>image<|/ref|><|det|>[[88, 725, 910, 825]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[88, 725, 201, 734]]<|/det|>
+<center>Antigen Density </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 42, 945, 330]]<|/det|>
+Figure S1: E2A-PBX/mCD19 antigen density model and murine anti-CD19 CAR T cells, and additional statistical comparisons of in vitro data (Related to Figure 1) S1A: Schematic of the anti-mouse CD19 CAR contained in pMSCV backbone. S1B: Coexpression of CAR and EGFR on murine CAR T cells. S1C: Engineering of murine leukemia with lentiviral vectors containing hUbC or hEF1a promoters driving the CD19 transgene. S1D: Survival of mice after treatment with 1e6 EGFR+ (EGFR8, non- CAR expressing) naïve or memory- derived CD8+ T cells. Data is from 1 experiment, total n=5 mice per group. S1E: Mean fluorescence intensity of IFNg+ cell population. S1F: Mean fluorescence intensity of TNFa+ cell population. S1G: Mean fluorescence intensity of CD107a+ population. S1H: Statistical comparisons of Ki67Ne9 (% Ki67Neg of EGFR+), Ki67L0 (% Ki67L0 of EGFR+, MFI Ki67L0 of EGFR+) and Ki67H1 (% Ki67H1 of EGFR+, MFI Ki67H1 of EGFR+) populations. S1I: Statistical comparisons of CellTraceL0 (% CellTraceL0 of EGFR+, MFI CellTraceL0 of EGFR+) and total EGFR+ (GFMI CellTrace, GFMI CellTrace with zoomed axis) populations. Data represent mean +/- SD. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[66, 55, 880, 940]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 42, 945, 295]]<|/det|>
+Figure S2: Polyclonal pathogen- elicited CAR8MD function similarly to vaccine- elicited CAR8MD (Related to Figure 1). S3A: Schematic: LCMV model for generating memory CD8+ T cells. C57BL/6 hosts were infected with LCMV- Armstrong. 4 weeks later, naïve and memory CD8+ T cells were sorted from the same hosts using the indicated FACS markers and used to manufacture CAR8MD, memory- derived or CAR8ND, naïve- derived or EGFR8 control cells. S3B: Intracellular cytokine staining of IFNg and TNFA after 6 hour co- culture assay. S3C: Quantifications of proportions of IFNg+ and TNFA+ cells of EGFR+ population. S3D: Proliferation as measured by dilution of CellTrace Violet dye after 72 hour co- culture assay. S3E: Quantification of CellTrace assay, proportions of CellTrace<sup>Lo</sup> cells. All assays were performed with \(n = 3\) technical replicates, and are representative of 2 independent experiments. Data represent mean \(+ / -\) SD. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) .
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[65, 49, 168, 69]]<|/det|>
+# Figure S3  
+
+<|ref|>title<|/ref|><|det|>[[392, 68, 610, 87]]<|/det|>
+# 1e6 CAR+ Cell Dose  
+
+<|ref|>image<|/ref|><|det|>[[66, 95, 933, 300]]<|/det|>  
+
+<|ref|>title<|/ref|><|det|>[[390, 360, 610, 380]]<|/det|>
+# 3e5 CAR+ Cell Dose  
+
+<|ref|>image<|/ref|><|det|>[[66, 386, 933, 650]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[55, 680, 866, 700]]<|/det|>
+<center>Figure S3: CAR T cells and leukemia counts per tibia for in vivo data (Related to Figures 2 & 3) </center>  
+
+<|ref|>text<|/ref|><|det|>[[52, 704, 945, 867]]<|/det|>
+All analyses in this figure are done on the same experiments described in Figures 2 and 3. Counts data was generated by flushing a single tibia and using total tibia counts and cytometer proportions data to calculate CAR and leukemia cell counts per tibia. S8A: CAR counts for 1e6 CAR dose experiments. S8B: Leukemia counts for 1e6 CAR dose experiments. S8C: CAR counts for 3e5 CAR dose experiments. S8D: Leukemia counts for 3e5 CAR dose experiments. Data are from 2 pooled, independent experiments with \(n = 10\) mice per condition, apart from the 1e6 CAR dose day 11 timepoint, which contains data from one experiment with \(n = 5\) mice per condition. Data represent mean \(+ / -\) SD. \(^*\) \(p< 0.05\) \(^{**}\) \(p< 0.01\) \(^{***}\) \(p< 0.0001\)
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[97, 99, 700, 256]]<|/det|>
+ 
+
+<|ref|>image_caption<|/ref|><|det|>[[88, 78, 110, 92]]<|/det|>
+<center>A.</center> 
+
+<|ref|>image<|/ref|><|det|>[[97, 303, 912, 410]]<|/det|>
+ 
+
+<|ref|>image_caption<|/ref|><|det|>[[88, 291, 110, 305]]<|/det|>
+<center>B.</center> 
+
+<|ref|>image<|/ref|><|det|>[[97, 425, 912, 531]]<|/det|>
+ 
+
+<|ref|>image_caption<|/ref|><|det|>[[88, 460, 110, 474]]<|/det|>
+<center>Day 11</center> 
+
+<|ref|>image<|/ref|><|det|>[[97, 560, 912, 666]]<|/det|>
+ 
+
+<|ref|>image_caption<|/ref|><|det|>[[88, 547, 110, 560]]<|/det|>
+<center>D.</center> 
+
+<|ref|>image<|/ref|><|det|>[[97, 692, 912, 800]]<|/det|>
+ 
+
+<|ref|>image_caption<|/ref|><|det|>[[88, 744, 110, 758]]<|/det|>
+<center>Day 11</center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 42, 944, 377]]<|/det|>
+Figure S4: Basic characterization of in vivo model and additional in vivo effector/memory phenotyping at high CAR dose (Related to Figure 2).S4A: Basic flow cytometry gating strategy for in vivo experiments. Total events were gated by Singlets, Live Cells and then Lymphocytes, followed by CD8a+/TCRbeta+/EGFR+ cells for CAR8/EGFR8 or B220+/CD22+ cells for E2A-PBX. S4C-F are from experiments with the 1e6 EGFR+ cell dose. S4B: Proportions of CAR8 with the short-lived effector cell (SLEC, IL7Ra-/KLRG1+) or memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) phenotypes at the indicated timepoint against WT leukemia. S4C: Proportions of CAR8 with the short-lived effector cell (SLEC, IL7Ra-/KLRG1+) or memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) phenotypes at the indicated timepoint against CD19<sup>lo</sup> leukemia. S4D: Proportions of CAR8 with the effector memory precursor (EMP, CD27+/CD62L-) or central memory precursor (CMP, CD27+/CD62L+) phenotypes at the indicated timepoint against WT leukemia. S4E: Proportions of CAR8 with the effector memory precursor (EMP, CD27+/CD62L-) or central memory precursor (CMP, CD27+/CD62L+) phenotypes at the indicated timepoint against CD19<sup>lo</sup> leukemia. Data in S4C-F are from 2 pooled, independent experiments with n=10 mice per condition. Data represent mean +/- SD. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[85, 70, 925, 896]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 42, 944, 395]]<|/det|>
+Figure S5: Additional in vivo effector/memory and exhaustion phenotyping at low CAR dose (Related to Figure 3). All data in this figure are from experiments with the 3e5 EGFR+ cell dose. S5A: Proportions of CAR8 with the short-lived effector cell (SLEC, IL7Ra-/KLRG1+) or memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) phenotypes at the indicated timepoint against WT leukemia. S5B: Proportions of CAR8 with the short-lived effector cell (SLEC, IL7Ra-/KLRG1+) or memory precursor effector cell (MPEC, IL7Ra+/KLRG1-) phenotypes at the indicated timepoint against CD19<sup>lo</sup> leukemia. S5C: Proportions of CAR8 with the effector memory precursor (EMP, CD27+/CD62L-) or central memory precursor (CMP, CD27+/CD62L+) phenotypes at the indicated timepoint against WT leukemia. S5D: Proportions of CAR8 with the effector memory precursor (EMP, CD27+/CD62L-) or central memory precursor (CMP, CD27+/CD62L+) phenotypes at the indicated timepoint against CD19<sup>lo</sup> leukemia. Figures S4E-L display proportions of CAR8 with the indicated phenotype at 11 days post-CAR injection against either WT (left, E,F,I,J) or CD19<sup>lo</sup> (right, G,H,K,L) leukemia. S5E & G: PD1+/TOX+ S5F & H: PD1+/CD39+ S5I & K: TCF1+/TIM3- S5J & L: TCF1-/TIM3+. Data are from 2 pooled, independent experiments with \(n = 10\) mice per condition. Data represent mean \(+ / -\) SD. \* \(p< 0.05\) , \*\* \(p< 0.01\) , \*\*\* \(p< 0.001\) , \*\*\*\* \(p< 0.0001\) .
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 120, 530, 264]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[63, 52, 160, 69]]<|/det|>
+<center>Figure S6 </center>  
+
+<|ref|>image<|/ref|><|det|>[[120, 312, 530, 420]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[190, 299, 512, 312]]<|/det|>
+<center>Comparisons of Naive T cell derived populations </center>  
+
+<|ref|>image<|/ref|><|det|>[[120, 449, 530, 556]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[131, 430, 519, 444]]<|/det|>
+<center>Comparisons of Memory T cell derived populations </center>  
+
+<|ref|>image<|/ref|><|det|>[[120, 585, 530, 693]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[123, 565, 570, 579]]<|/det|>
+<center>Comparisons between Naive and Memory T cell derived populations </center>  
+
+<|ref|>image<|/ref|><|det|>[[120, 725, 590, 940]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[90, 725, 110, 738]]<|/det|>
+<center>D. </center>  
+
+<|ref|>image<|/ref|><|det|>[[639, 120, 930, 940]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[9, 44, 944, 255]]<|/det|>
+1 Figure S6: Additional analyses of ATAC-seq data (Related to Figure 4).2 All analyses in this figure are from the same timeline/experimental layout described in Figure 4A. S6A: Inter-3 replicate Euclidian distance of boom-normalized ATAC-seq counts per peak between biological replicates. S6B: Pairwise comparisons of differentially accessible chromatin regions within conditions between different 5 timepoints of the same condition, or between different conditions at each timepoint. Data points are mean of 6 boom-normalized ATAC-seq counts per peak between biological replicates of each group. S6C: Heatmap of 7 motif-associated ChromVAR deviation z-scores patterns of motif-associated ATAC-seq signal for indicated 8 transcription factors. List comprises all significant differentially accessible comparisons. S6D: Representative 9 gating for sorting of cells in sequencing experiments.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[66, 66, 840, 960]]<|/det|>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[55, 42, 648, 61]]<|/det|>
+## Figure S7: Additional analyses of RNA-seq data (Related to Figure 5).  
+
+<|ref|>text<|/ref|><|det|>[[52, 66, 944, 255]]<|/det|>
+All analyses in this figure are from the same timeline/experimental layout described in Figure 4A. S7A: Normalized enrichment scores from GSEA of differentially enriched genesets between indicated CD8+ T cell subsets after LCMV- armstrong acute viral infection<sup>24, 25</sup>. S7B: Top transcriptional activators predicted to be activated and driving differential transcriptional state between naive versus memory- derived cells at the indicated timepoint, as predicted by Qiagen Ingenuity Pathway Analysis<sup>28</sup> (IPA). S7C: IPA activation map for the Cebpb transcription factor, the top predicted driver of transcriptional state in memory- derived cells at the PostCAR and Tumor timepoints. S7D-F: Top differentially expressed transcription factors, at the indicated timepoint. All statistics performed using DESeq2 with filtering threshold at 10, log2foldchange >2 and padj > 0.05.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[65, 90, 930, 930]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 44, 945, 472]]<|/det|>
+1 Figure S8 (Related to Figure 6): Characterization of 1e5 CAR T cell dose in vivo experiments and in vitro comparisons of BATF, JUN or RUNX2 overexpressing cells to pMIG. 2 All analyses in this figure are the same timeline/experimental layout described in Figure 3A except with 1e5 CAR+ cell dose, at 11 days post- CAR timepoint. S8A- B are characterization of the 1e5 cell dose with standard T cell groups (no ectopic transcription factor expression). S8A: Leukemia burden. S8B: Proportions of CAR8 with the PD1+/TOX+ phenotype. Data in S7A- B are from 1 experiment with n=5 mice per condition. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001. S8C- D: Quantification of intracellular cytokine staining of IFNg and TNFa after 6 hour co- culture assay, % positive of EGFR+, for memory (C) or naive- derived (D) cells cotransduced with BATF, JUN or pMIG. Data in S8C- D are from 3 independent experiments. S8E: Proliferation as measured by dilution of CellTrace Violet dye dilution of EGFR+ cells after 72 hour co- culture assay, for memory or naive derived cells cotransduced with BATF, JUN or pMIG. Data representative of 3 independent experiments. S8F- G: Quantification of intracellular cytokine staining of IFNg and TNFa after 6 hour co- culture assay, % positive of EGFR+, for memory (C) or naive- derived (D) cells cotransduced with RUNX2 or pMIG. Data in S8C- D are from 3- 4 independent experiments. S8H: Proliferation as measured by dilution of CellTrace Violet dye dilution of EGFR+ cells after 72 hour co- culture assay, for memory or naive derived cells cotransduced with RUNX2 or pMIG. Data representative of 3 independent experiments. No statistically significant differences were found between BATF, JUN or RUNX2 engineered CAR T cells and pMIG control T cells for in vitro data. Data represent mean +/- SD.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[90, 80, 912, 925]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 44, 944, 375]]<|/det|>
+1 Figure S9 (Related to Figure 6): Counts and additional exhaustion phenotyping data for BATF, JUN or RUNX2 overexpression in vivo experiments. 2 All analyses in this figure are done on the same experiments described in Figure 6. Counts data was generated by flushing a single tibia and using total tibia counts and cytometer proportions data to calculate CAR and leukemia cell counts per tibia. S9A- B: CAR and leukemia counts for BATF or JUN overexpressing memory (A) or naive- derived (B) CAR T cells compared to pMIG control. S9C- H: Proportions of EGFR+ cells from BATF, JUN or pMIG CAR8 with the indicated phenotype. S9C,D,G are memory- derived cells, S9E,F,H are naive- derived cells. S9C,E: PD1+ S9D,F: PD1+/CD39+ S9G- H: Indicated TCF1/TIM3 phenotype. Data in S9A- H are from one experiment with n=5 mice per condition. S9I- J: CAR and leukemia counts for RUNX2 overexpressing memory (A) or naive- derived (B) CAR T cells compared to pMIG control. S9K- N: Proportions of EGFR+ cells from RUNX2 or pMIG CAR8 with the indicated phenotype. S9C,D,G are memory- derived cells, S9E,F,H are naive- derived cells. S9K,M: PD1+ S9L,N: PD1+/CD39+ S9O,P: Indicated TCF1/TIM3 phenotype. Data in S9I- P are from 2 pooled, independent experiments with n=9 mice per condition. Data represent mean +/- SD. \* p<0.05, \*\* p<0.01, \*\*\* p<0.001, \*\*\*\* p<0.0001.
+
+<--- Page Split --->

preprint/preprint__48de5430c259c6879fe0e3cc8610c93e4199c05191d114610b87fb8ee3f4f045/images_list.json

@@ -0,0 +1,115 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1. Prediction errors of the SpatioTemporal eXtreme Gradient Boosting (STXGB) models in one-week prediction horizon against a temporal autoregressive model without spatial lags (TXGB). Error comparison of STXGB models with different feature sets in predicting weekly new cases (top row) and new cases per 10k population (bottom row) for one-week ahead horizon. STXGB-FB, which incorporates Facebook-derived features, including spatial lags based on Social Connectedness Index, outperforms other models. Left column: prediction RMSE. Right column: prediction MAE.",
+    "footnote": [],
+    "bbox": [
+      [
+        137,
+        90,
+        901,
+        620
+      ]
+    ],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. Long-term prediction error comparison. Prediction errors of the STXGB-FB model (dashed line) and the Ensemble baseline (solid line) over four prediction horizons on five forecasting dates. a) Prediction RMSE and b) Prediction MAE.",
+    "footnote": [],
+    "bbox": [
+      [
+        168,
+        95,
+        933,
+        358
+      ]
+    ],
+    "page_idx": 12
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3. Map of COVID-19 cases per 10k population and errors in predicting them. (a) number of confirmed new cases per 10k population over the week ahead of forecasting date Oct. 31, 2020 (b) prediction errors for the same forecasting date(c) number of new cases over the week ahead of Nov. 7, 2020 forecasting date (d) prediction errors for the same forecasting date. The pattern of errors in Georgia, and Texas, and Kentucky flip from Oct. 31 to Nov. 7, indicating potential lags in testing and reporting.",
+    "footnote": [],
+    "bbox": [
+      [
+        120,
+        95,
+        965,
+        544
+      ]
+    ],
+    "page_idx": 16
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4. Prediction errors in urban vs. rural counties. Prediction errors of the number of new cases (left column) and new cases per 10k population (right column) in rural and urban counties on the Nov. 7 forecast date across four prediction horizons. The higher and lower 3% of counties are trimmed from the plot view.",
+    "footnote": [],
+    "bbox": [
+      [
+        163,
+        88,
+        833,
+        860
+      ]
+    ],
+    "page_idx": 19
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1",
+    "footnote": [],
+    "bbox": [
+      [
+        68,
+        106,
+        480,
+        707
+      ]
+    ],
+    "page_idx": 57
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 57
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3",
+    "footnote": [],
+    "bbox": [
+      [
+        99,
+        65,
+        940,
+        518
+      ]
+    ],
+    "page_idx": 58
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4",
+    "footnote": [],
+    "bbox": [
+      [
+        70,
+        50,
+        700,
+        787
+      ]
+    ],
+    "page_idx": 59
+  }
+]

preprint/preprint__48de5430c259c6879fe0e3cc8610c93e4199c05191d114610b87fb8ee3f4f045/preprint__48de5430c259c6879fe0e3cc8610c93e4199c05191d114610b87fb8ee3f4f045.mmd

@@ -0,0 +1,570 @@
+
+# Predicting County-Level COVID-19 Cases using Spatiotemporal Machine Learning: Modeling Human Interactions using Social Media and Cell-Phone Data  
+
+Behzad Vahedi ( \(\boxed{ \begin{array}{r l} \end{array} }\) behzad@colorado.edu ) University of Colorado Boulder https://orcid.org/0000- 0001- 5782- 3831 Morteza Karimzadeh University of Colorado Boulder Hamidreza Zoraghein Social and Behavioral Science Research, Population Council  
+
+## Article  
+
+Keywords: spatiotemporal machine learning, COVID- 19  
+
+Posted Date: February 10th, 2021  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 203188/v1  
+
+License: \(\circledcirc\) This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Communications on November 8th, 2021. See the published version at https://doi.org/10.1038/s41467- 021- 26742- 6.
+
+<--- Page Split --->
+
+# Predicting County-Level COVID-19 Cases using Spatiotemporal Machine Learning: Modeling Human Interactions using Social Media and Cell-Phone Data  
+
+Behzad Vahedi<sup>1\*</sup>, Morteza Karimzadeh<sup>1\*</sup>, Hamidreza Zoraghein<sup>2</sup>  <sup>1</sup> Department of Geography, University of Colorado Boulder; Behzad@colorado.edu; karimzadeh@colorado.edu  <sup>2</sup> Social and Behavioral Science Research, Population Council, New York, USA; hzoraghein@popcouncil.org  
+
+## Abstract  
+
+Measurements of human interaction through proxies such as social connectedness or movement patterns have proved useful for predictive modeling of COVID- 19. In this study, we first compare the power of Facebook's social connectedness with cell phone- derived human mobility for predicting county- level new cases of COVID- 19. Our experiments show that social connectedness is a better proxy for measuring human interactions leading to new infections. Next, we develop a SpatioTemporal autoregressive eXtreme Gradient Boosting (STXGB) model to predict county- level new cases of COVID- 19 in the coterminous US. We
+
+<--- Page Split --->
+
+evaluate the model on five weekly forecast dates between October 24 and November 28, 2020 over one- to four- week prediction horizons. Comparing our predictions with a baseline Ensemble of 32- models currently used by the CDC indicates an average \(58\%\) improvement in prediction RMSEs over two- to four- week prediction horizons, pointing to the strong predictive power of our model.  
+
+## 1. Introduction  
+
+Human interaction in close physical proximity is the primary cause of the transmission of highly contagious diseases such as COVID- 19<sup>1</sup>. Measuring human interaction is therefore an important step in understanding and predicting the spread of COVID- 19<sup>2,3</sup>. However, tracking human interactions requires rigorous contact tracing at national and regional scales which has not been implemented in the United States due to the economic, legal, and sociocultural concerns, as well as inadequate testing supplies, and insufficient national coordination<sup>4</sup>.  
+
+As a result, researchers have adopted different proxies to track human interaction levels. One such proxy is the "Social Connectedness Index" (SCI), generated from Facebook's friendship data. SCI represents the probability that two users in a pair of regions (e.g., U.S. counties) are friends (i.e., connected) on Facebook<sup>5</sup>. Kuchler et al.<sup>6</sup> reported on the strong correlation between early
+
+<--- Page Split --->
+
+hotspots of COVID- 19 outbreak and their level of social connectedness. The underlying assumption in leveraging SCl as a proxy for physical human interactions is that individuals who are socially connected on Facebook have a higher probability for physical interaction, thereby, potentially contributing to the spread of communicable diseases.  
+
+Human mobility flow, as measured by anonymized cell phone data, serves as another proxy for quantifying human interactions<sup>7,8</sup>. Widely used to study the spread of COVID- 19, most studies incorporating cell- phone data have focused on the change in mobility within a spatial unit<sup>9,10</sup>, while a few others have also incorporated the flow between different spatial units<sup>11</sup> to predict transmissions across units, albeit mostly in the early stages of the pandemic with limited evaluation data. The underlying assumption in this approach is that more movements between spatial units leads to higher interactions, and consequently, an elevated risk of disease spread.  
+
+It is unclear, however, which of these approaches—using social media connectedness versus cell phone- derived human mobility flow—is a better indicator of physical interaction within and between different regions. Furthermore, the underlying assumption in each approach may not necessarily be valid in the case of COVID- 19: considering the sporadic and regional stay- at- home
+
+<--- Page Split --->
+
+orders across the United States, social connectedness may not lead to physical interaction, at least not to the same level as pre- pandemic. Similarly, given the recommended preventative measures such as mask- wearing and physical distancing<sup>12</sup>, human flow from one location to another may not necessarily lead to physical interactions that could communicate the disease, especially in public places, where preventative measures are enforced more strictly.  
+
+In this paper, we compare the predictive power of Facebook's social connectedness index, as an example of social media proxy, with cell phone- derived human mobility data, as an example of human flow proxy, in forecasting county- level new cases of COVID- 19 in the conterminous US over multiple prediction horizons. County- level prediction is more challenging than state- level prediction<sup>13–15</sup>, yet it serves as the highest spatial resolution for national models in the U.S., since cases are aggregated and reported at the county level. Longer term County- level predictions are also essential for policy making and resource allocation.  
+
+The unique characteristics of COVID- 19, including its presymptomatic and asymptomatic contagiousness, rapid spread, along with variations in regional response policies, such as inconsistent and sporadic testing and contact tracing, make forecasting the spatial patterns of this disease challenging. Researchers
+
+<--- Page Split --->
+
+have used a variety of methods including time- series autoregressive models<sup>16–18</sup>, machine learning techniques<sup>19–21</sup>, epidemiologic models such as SIR model and its variants<sup>22,23</sup>, and combinations of these methods<sup>24</sup> for forecasting COVID- 19. We experiment with five different machine learning- based spatiotemporal autoregressive algorithms to perform county- level predictions, and use the best algorithm, i.e. the one with the lowest average prediction RMSE and MAE, to compare between Facebook- and cell phone- derived features.  
+
+We compare our best model predictions against one of the most prominent collective efforts in forecasting COVID- 19 in the U.S., namely, the Ensemble model developed by the "COVID- 19 Forecast Hub" team<sup>25</sup> which is used by the Centers for Disease Control and Prevention (CDC) to report predictions of new cases and deaths in U.S. counties in one- to four- weeks ahead horizons<sup>25,26</sup>.  
+
+Our specific contributions are as follows: (a) designing inter- county and intra- county features for spatiotemporal autoregressive machine learning of county- level new cases, (b) comparing the performance of social media connectedness (derived from Facebook) and human flow connectedness (derived from SafeGraph's cell phone data) by incorporating inter- county spatial lags for predicting county- level new COVID- 19 cases, and (c) improving the long- term
+
+<--- Page Split --->
+
+prediction of county- level new cases of COVID- 19 in the coterminous U.S. in comparison to a baseline Ensemble model, using an end- to- end model.  
+
+## 2. Results  
+
+## Algorithm Selection  
+
+Five different machine learning algorithms were trained and tuned using each set of features named in the next section (details in Section 4), and tested over the last 5 weeks of our dataset (same dates as Tables 2 & 5), by holding out one week at a time for testing. Table 1 reports the average performance for each algorithm. Extreme Gradient Boosting (XGB) performed better on unseen data compared with other tree- based ensemble algorithms and the neural networks, including Feed Forward Neural Network (FFNN) and Long Short- Term Memory (LSTM) network (Table 1). Therefore, we used XGB for developing short- term and long- term prediction models. The RMSE and MAE values reported in Table 1 are for the natural log values of [new cases per 10k population + 1], which we used as a transformed target variable in the models, given the skewed distribution of new cases (or new cases per 10k) in counties.
+
+<--- Page Split --->
+
+
+Table 1. Performance comparison of machine learning regressors. The best performance in each category is bolded.   
+
+<table><tr><td>Model</td><td>RMSE Train</td><td>MAE Train</td><td>RMSE Test</td><td>MAE Test</td></tr><tr><td>Random Forest</td><td>0.486</td><td>0.359</td><td>0.511</td><td>0.38</td></tr><tr><td>Stochastic Gradient Boosting</td><td>0.438</td><td>0.313</td><td>0.494</td><td>0.348</td></tr><tr><td>Extreme Gradient Boosting</td><td>0.441</td><td>0.316</td><td>0.47</td><td>0.330</td></tr><tr><td>Feed Forward Neural Network</td><td>0.524</td><td>0.391</td><td>0.566</td><td>0.438</td></tr><tr><td>Long Short-Term Memory</td><td>1.179</td><td>0.964</td><td>1.209</td><td>1.007</td></tr></table>  
+
+## Comparing social media- and cell phone-derived features  
+
+To compare the relative strength of Facebook- derived movement and connectedness against SafeGraph- derived human mobility flows, as proxies for physical human interactions, we designed a set of intra- county and inter- county interaction features using each proxy, and incorporated each set of features separately to develop spatiotemporally lagged autoregressive prediction models of new cases of COVID- 19 (i.e. target variable). We then compared the predictions of these models against each other as well as a base model (not to be confused with the baseline model for final evaluations), all of which were trained using the XGB algorithm.  
+
+Our base model incorporates a series of socioeconomic, demographic and temperature variables, as well as temporal lags of the target variable in the same
+
+<--- Page Split --->
+
+county only, thus, we call it Temporal XGB (TGXB), whereas the SpatioTemporal XGB (STXGB) models, in addition to temporal lags, also incorporate intra- county movement features and spatiotemporal lags of the target variable weighted by the inter- county connectedness strength. Specifically, the spatial lags in STXGB are calculated by multiplying the target variable (natural log of weekly new cases per 10k population + 1) in "connected counties" by either (a) inter- county Facebook Social Media Connectedness Index, (in the STXGB- FB model), or (b) inter- county Flow Connectedness Index derived from SafeGraph's cell- phone movement data, forming STXGB- SG and STXGB- SGR models (described in detail in Section 4).  
+
+Table 2 and Fig. 1 present the error values of predicted new cases and new cases per 10k population in the one- week prediction horizon using the TXGB and STXGB models. The incorporation of spatiotemporal lags using county connectedness indices (in STXGB) was advantageous across the board, compared to the temporal lags only (TXGB). All variants of STXGB (- FB, - SG, and - SGR) achieved lower errors compared to TXGB. Furthermore, STXGB- FB, which uses the Facebook- derived features, outperformed all other models in average RMSEs and MAEs as well as on all forecast dates, except the fifth date when the STXGB- SG model generated slightly lower errors.
+
+<--- Page Split --->
+
+Table 2. RMSE and MAE of county-level predicted weekly new cases and new cases per 10k population. Lowest values of each error metric are highlighted. Average values across forecasting dates for each model is bold faced. 
+
+<table><tr><td></td><td>Model</td><td>Forecast Date</td><td>RMSE New Case Prediction</td><td>MAE New Case Prediction</td><td>RMSE New Case/10k Prediction</td><td>MAE New Case/10k Prediction</td></tr><tr><td rowspan="5">including temporal lags</td><td rowspan="5">Base Model (TXGB)</td><td>10/24/2020</td><td>136.255</td><td>30.894</td><td>16.084</td><td>6.134</td></tr><tr><td>10/31/2020</td><td>192.993</td><td>50.91</td><td>22.319</td><td>11.176</td></tr><tr><td>11/07/2020</td><td>203.678</td><td>70.689</td><td>22.899</td><td>12.907</td></tr><tr><td>11/14/2020</td><td>237.113</td><td>80.1</td><td>26.45</td><td>15.104</td></tr><tr><td>11/21/2020</td><td>166.855</td><td>50.684</td><td>16.384</td><td>9.611</td></tr><tr><td></td><td></td><td>Average</td><td>187.379</td><td>56.655</td><td>20.827</td><td>10.986</td></tr><tr><td rowspan="5">including spatial-temporal lags</td><td rowspan="5">STXGB with Facebook-derived features (STXGB-FB)</td><td>10/24/2020</td><td>116.312</td><td>25.909</td><td>15.083</td><td>5.708</td></tr><tr><td>10/31/2020</td><td>172.582</td><td>46.398</td><td>21.938</td><td>10.817</td></tr><tr><td>11/07/2020</td><td>169.602</td><td>54.613</td><td>20.925</td><td>11.072</td></tr><tr><td>11/14/2020</td><td>185.391</td><td>62.243</td><td>23.263</td><td>12.477</td></tr><tr><td>11/21/2020</td><td>142.312</td><td>48.625</td><td>16.297</td><td>9.373</td></tr><tr><td></td><td></td><td>Average</td><td>157.24</td><td>47.557</td><td>19.501</td><td>9.89</td></tr><tr><td rowspan="5">including spatial-temporal lags</td><td rowspan="5">STXGB with SafeGraph-derived features (STXGB-SG)</td><td>10/24/2020</td><td>120.049</td><td>27.312</td><td>15.101</td><td>5.785</td></tr><tr><td>10/31/2020</td><td>195.03</td><td>50.487</td><td>22.267</td><td>11.171</td></tr><tr><td>11/07/2020</td><td>193.263</td><td>62.506</td><td>21.156</td><td>11.421</td></tr><tr><td>11/14/2020</td><td>203.675</td><td>68.962</td><td>24.423</td><td>13.371</td></tr><tr><td>11/21/2020</td><td>140.748</td><td>48.383</td><td>16.683</td><td>9.596</td></tr><tr><td></td><td></td><td>Average</td><td>170.553</td><td>51.53</td><td>19.926</td><td>10.269</td></tr><tr><td rowspan="5">including spatial-temporal lags</td><td rowspan="5">STXGB with SafeGraph-derived features-rich (STXGB-SGR)</td><td>10/24/2020</td><td>122.482</td><td>27.952</td><td>15.362</td><td>5.739</td></tr><tr><td>10/31/2020</td><td>207.411</td><td>54.531</td><td>22.626</td><td>11.541</td></tr><tr><td>11/07/2020</td><td>178.934</td><td>60.061</td><td>20.998</td><td>11.324</td></tr><tr><td>11/14/2020</td><td>186.997</td><td>64.745</td><td>24.079</td><td>13.032</td></tr><tr><td>11/21/2020</td><td>141.356</td><td>47.491</td><td>16.758</td><td>9.539</td></tr><tr><td></td><td></td><td>Average</td><td>167.436</td><td>50.956</td><td>19.965</td><td>10.235</td></tr></table>
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Figure 1. Prediction errors of the SpatioTemporal eXtreme Gradient Boosting (STXGB) models in one-week prediction horizon against a temporal autoregressive model without spatial lags (TXGB). Error comparison of STXGB models with different feature sets in predicting weekly new cases (top row) and new cases per 10k population (bottom row) for one-week ahead horizon. STXGB-FB, which incorporates Facebook-derived features, including spatial lags based on Social Connectedness Index, outperforms other models. Left column: prediction RMSE. Right column: prediction MAE. </center>
+
+<--- Page Split --->
+
+## Long-term predictions and evaluation against the COVID-19 Forecast Hub  
+
+## Ensemble  
+
+We compared the predictions of our best model, STXGB- FB, against the predictions of the COVID- 19 Forecast Hub Ensemble of 32 models (used by the CDC in reporting forecasts of new cases<sup>26</sup>) over one-, two-, three-, and four- week horizons. We trained and tuned STXGB- FB for each prediction horizon separately. We then used the reported new cases by Johns Hopkins University Center for Systems Science and Engineering (JHU CSSE)<sup>27</sup> as ground- truth to calculate RMSE and MAE of each prediction set, shown in Fig. 2 and Table 3, over varying prediction horizons across the five forecast dates. Our model considerably improves RMSEs and MAEs compared with the Ensemble model in the two- week, three- week, and four- week ahead prediction horizons, with an average 58% reduction in RMSEs and 61% reduction in MAEs.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Figure 2. Long-term prediction error comparison. Prediction errors of the STXGB-FB model (dashed line) and the Ensemble baseline (solid line) over four prediction horizons on five forecasting dates. a) Prediction RMSE and b) Prediction MAE. </center>  
+
+In the one- week prediction horizons, the Ensemble model outperformed our STXGB- FB model on the first three forecasting dates (Oct 24, Oct. 31, Nov. 7), but performed similarly on the Nov. 14 forecasting date. Nevertheless, the differences in the one- week horizon predictions are relatively small. STXGB- FB outperformed the Ensemble prediction on Nov. 21 across all prediction horizons. This is noteworthy since the prediction horizons (one- to four- week) on Nov. 21 overlap the post- Thanksgiving holidays in the US, which caused a surge in the number of cases<sup>28</sup>. In summary, out of the 20 predictions performed, the STXGB- FB model outperformed the Ensemble model in 17 of them, including in all longer than one- week prediction horizons (Table 3).
+
+<--- Page Split --->
+
+To investigate the potential of SafeGraph cell phone-derived features in long-term predictions, we generated one-, two-, three-, and four-week forecasts using the STXGB-SG model as well. This model does not perform as well as STXGB-FB, pointing to the superiority of Facebook-derived features in our models consistent with the one-week predictions (Supplementary Table 4, Supplementary Information). However, while STXGB-SG generates larger errors compared to STXGB-FB, it still outperforms the Ensemble model in long-term prediction horizons.  
+
+Table 3. Comparison of the prediction errors generated by the COvID-19 Forecast Hub Ensemble model and our STXGB-FB model in 1-to 4-week prediction horizons.   
+
+<table><tr><td rowspan="2">Forecast Date</td><td rowspan="2">Model</td><td colspan="4">Prediction RMSE in Prediction Horizon</td></tr><tr><td>1 wk ahead</td><td>2 wk ahead</td><td>3 wk ahead</td><td>4 wk ahead</td></tr><tr><td rowspan="3">10/24/2020</td><td>Ensemble</td><td>126.05</td><td>894.41</td><td>1915.31</td><td>3087.32</td></tr><tr><td>STXGB-FB</td><td>164.80</td><td>489.81</td><td>832.36</td><td>1157.87</td></tr><tr><td>Pct. Improvement</td><td>-30.74%</td><td>45.24%</td><td>56.54%</td><td>62.50%</td></tr><tr><td rowspan="3">10/31/2020</td><td>Ensemble</td><td>205.12</td><td>1172.21</td><td>2342.40</td><td>3461.53</td></tr><tr><td>STXGB-FB</td><td>247.67</td><td>554.25</td><td>892.22</td><td>1376.14</td></tr><tr><td>Pct. Improvement</td><td>-20.74%</td><td>52.72%</td><td>61.91%</td><td>60.24%</td></tr><tr><td rowspan="3">11/07/2020</td><td>Ensemble</td><td>223.96</td><td>1311.84</td><td>2424.17</td><td>3864.08</td></tr><tr><td>STXGB-FB</td><td>268.78</td><td>649.08</td><td>906.63</td><td>1380.22</td></tr><tr><td>Pct. Improvement</td><td>-20.01%</td><td>50.52%</td><td>62.60%</td><td>64.28%</td></tr><tr><td rowspan="3">11/14/2020</td><td>Ensemble</td><td>275.24</td><td>1148.63</td><td>2558.68</td><td>4372.37</td></tr><tr><td>STXGB-FB</td><td>274.35</td><td>468.53</td><td>848.68</td><td>1450.59</td></tr><tr><td>Pct. Improvement</td><td>0.32%</td><td>59.21%</td><td>66.83%</td><td>66.82%</td></tr><tr><td rowspan="3">11/21/2020</td><td>Ensemble</td><td>301.87</td><td>1292.31</td><td>3086.82</td><td>5390.16</td></tr><tr><td>STXGB-FB</td><td>210.64</td><td>592.83</td><td>1457.29</td><td>2646.68</td></tr><tr><td>Pct. Improvement</td><td>30.22%</td><td>54.13%</td><td>52.79%</td><td>50.90%</td></tr></table>
+
+<--- Page Split --->
+
+## Spatial distribution of errors  
+
+Our STXGB- FB performed worse than the baseline Ensemble model on the Oct.24, Oct. 31 and Nov. 7 forecasting dates over the one- week horizon, but considerably outperformed the Ensemble over longer- term horizons on the same forecasting dates. Further, STXGB- FB outperformed the Ensemble baseline on Nov. 14 and Nov. 21 across all prediction horizons. To find potential explanations for this inconsistency, we inspected the spatial patterns of errors. Figure 3 illustrates maps of confirmed new cases per 10k population along with prediction errors per 10k population generated by the STXGB- FB model for two forecasting dates of Oct. 31 and Nov. 7. The purple- shaded counties in the error maps are those with model underestimation of new cases, and the brown shades indicate overestimations of observed values. As can be seen in this figure, the majority of counties with high prediction errors (per 10K) are located in the rural Midwest with relatively high numbers of cases per 10k population during the November surge, albeit these are counties with fewer total cases compared to more populated, urban ones. It is worth noting that we use normalized (by 10K population) maps in Fig. 3, since choropleth maps would be biased by patterns of population distribution otherwise.
+
+<--- Page Split --->
+
+Figure 3 also demonstrates clusters of apparent underestimations in Georgia and Texas on the Oct. 31 forecasting date, followed by apparent overestimations in the same areas for the week after. The opposite pattern is the case for Kentucky. In the case of Georgia, the high- error clusters can almost perfectly delineate the boundary of the state. This discrepancy could be a result of lags or different policies in testing and reporting COVID- 19 cases. These potential short- term lags in reporting by some states may explain why our model performs considerably better in the longer- term prediction horizons but underperforms in the one- week horizon on Oct. 31 and Nov. 7.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3. Map of COVID-19 cases per 10k population and errors in predicting them. (a) number of confirmed new cases per 10k population over the week ahead of forecasting date Oct. 31, 2020 (b) prediction errors for the same forecasting date(c) number of new cases over the week ahead of Nov. 7, 2020 forecasting date (d) prediction errors for the same forecasting date. The pattern of errors in Georgia, and Texas, and Kentucky flip from Oct. 31 to Nov. 7, indicating potential lags in testing and reporting. </center>  
+
+The majority of counties in the U.S are rural, which are also the ones with fewer medical resources, and where social media data or cell- phone mobility data, which underlie our models, might be less representative \(^{29 - 31}\) . To investigate our models' performance in rural- majority counties compared to the Ensemble baseline, we categorized the counties into urban- and rural- majority by
+
+<--- Page Split --->
+
+calculating an urbanization index for each county (Supplementary Information, Section B). 2391 counties (\~77%) were identified as rural and 712 as urban.  
+
+We then calculated the prediction errors of the number of cases and the number of cases per 10k population for the Ensemble and STXGB- FB models in each category across four prediction horizons for the Nov. 7 forecasting date (Fig. 4). Both models generate considerably lower median errors and narrower interquartile error ranges in rural counties when predicting the total number of new cases (not normalized by population), which could be attributed to the overall higher prevalence and higher variance of COVID- 19 cases in urban counties in our prediction horizons. However, the opposite is the case when predicting the number of weekly new cases per 10k population; both models have wider interquartile ranges in rural counties across all prediction horizons. This could be due to the overall higher prevalence of COVID- 19 per population in rural counties during the selected prediction horizons.  
+
+As evident in Fig. 4, STXGB- FB has lower prediction errors with a narrower IQR in both urban and rural counties compared to the Ensemble model, across all prediction horizons except for the shortest one (one- week), which may be attributed to temporal fluctuations and policy variations in testing and case reporting as discussed above. The overall superior performance of STXGB- FB is
+
+<--- Page Split --->
+
+observed when predicting both the total number of new cases and new cases per 10k population, affirming the higher robustness of our model. Furthermore, the difference between the median prediction errors of STXGB- FB in urban and rural counties, when predicting the number of cases per 10k population, is smaller compared to the Ensemble model. This points to the more consistent performance of STXGB- FB in majority- rural counties, even though Facebook might not be as representative in these areas<sup>29</sup>.
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 4. Prediction errors in urban vs. rural counties. Prediction errors of the number of new cases (left column) and new cases per 10k population (right column) in rural and urban counties on the Nov. 7 forecast date across four prediction horizons. The higher and lower 3% of counties are trimmed from the plot view. </center>
+
+<--- Page Split --->
+
+## 3. Discussion  
+
+We demonstrated that incorporating (1) spatiotemporal lags using inter- county indices of connectedness and (2) intra- county measurements of movement can improve the performance of high resolution COVID- 19 predictive models, especially over long- term horizons. Short- term and long- term predictions of COVID- 19 cases help the federal and local governments make informed decisions such as imposing or relaxing business restrictions or planning resource allocation in response to the forecasted trends of COVID- 19<sup>32,33</sup>.  
+
+Our results showed that using Facebook- derived features implemented in our SpatioTemporal Autoregressive eXtreme Gradient Boosted (STXGB) model generate lower prediction errors across all prediction horizons compared to SafeGraph cell phone- derived features within the same architecture (Supplementary Table 5). Notably, however, to maintain compatibility, we used a formulation similar to Facebook's Social Connectedness Index when creating a corresponding index from SafeGraph data (which we call Flow Connectedness Index, refer to Section 4). This might have had adverse effects on the predictive power of the cell- phone- derived features. Nevertheless, the resulting model performs considerably better than the Ensemble baseline in long- term predictions (Supplementary Table 4). We will investigate alternative designs of inter- county
+
+<--- Page Split --->
+
+connectedness metrics from SafeGraph mobility data in the future to ensure the utilization of the full potential of this dataset.  
+
+The superior performance of Facebook- derived features means that stronger county- level social (media) connections could lead to a higher chance of (unsafe) human interaction (e.g., house parties) and thus, COVID spread, compared to human flow from one location to another. The findings of this paper also suggest that Facebook's Social Connectedness Index can be used for successful predictive modeling of COVID- 19 in data- poor countries without cell- phone movement datasets, assuming that the Facebook usage in those countries is of a comparable size and representativeness to the U.S<sup>34</sup>. With more than 2.5 billion active users globally, Facebook provides social connectedness for many countries. Conversely, human mobility data, to the best of our knowledge, is available in far fewer numbers of countries.  
+
+The STXGB- FB and STXGB- SG models significantly outperformed the Ensemble model currently used by the CDC in predicting county- level new cases of COVID- 19 in two-, three-, and four- weeks prediction horizons, with inconsistent comparisons in the one- week horizon. Our error maps suggest that this inconsistency might be partly due to inconsistent and delayed testing and reporting by some states.
+
+<--- Page Split --->
+
+The superiority of STXGB over purely temporal models (such as our TXGB) points to the importance of incorporating both within- unit (e.g., intra- county) and between- unit (e.g., inter- county) interactions when predicting a highly contagious disease such as COVID- 19. Predictive models that focus on within state boundaries are likely to underperform, because after all, the disease does spread across geographic unit borderlines.  
+
+As evident in table 1, the XGB and SGB methods performed better than other machine learning- based algorithms. A potential reason for the lower performance of FFNN and LSTM is the relatively small size of the training data (34 weeks of observation at most). Neural networks' main advantage is in their ability to learn features from data<sup>35</sup>; however, they require higher amounts of training data compared to tree- based models for optimizing the model's parameters. Our results are a testament to the advantages of high- performance tree- based ensemble algorithms such as XGB with more limited training data, especially if features are well- engineered. Our spatiotemporal lag features provide a template for such features to improve machine learning- based predictive modeling of infectious diseases.  
+
+It is also worth noting that between XGB and SGB, SGB generated lower average training errors when using the base and SafeGraph- derived features, but
+
+<--- Page Split --->
+
+XGB outperformed SGB in testing RMSE and MAE across all 4 models (Supplementary Table 2). This is somewhat expected, as XGB uses the second- order derivatives of the loss function for optimization, and more importantly, a regularized model formalization to control over- fitting, which is otherwise a disadvantage with regression trees<sup>36</sup>. This regularized model results in better performance on unseen data. To ensure consistency, we ran all regression methods 10 times; and XGB had lower testing errors compared to all other regression methods in all 10 runs.  
+
+## 4. Methods  
+
+This section outlines the details of feature engineering, algorithm selection, and implementation of spatiotemporal autoregressive machine learning models for predicting new cases of COVID- 19 in the conterminous United States. We describe our experimental setup for comparing the predictive power of Facebook- derived features and SafeGraph's cell phone- derived mobility features (as proxies for human physical interaction) for this purpose, and the evaluation of our models against the COVID- 19 Forecast Hub Ensemble baseline.
+
+<--- Page Split --->
+
+## Base Features  
+
+We engineered features for machine learning that can be categorized into five groups: (A) a set of county- level demographic and socioeconomic features, (B) minimum and maximum temperatures of inhabited areas in counties, (C) temporally- lagged (i) weekly average and (ii) weekly change of incident rates (COVID- 19 cases per 10k population) in each county, (D) Facebook- derived features of (i) intra- county movement measurements and (ii) exposure to COVID- 19 through inter- county connectedness, and (E) SafeGraph- derived features of (i) intra- county movement measurements and (ii) exposure to COVID- 19 through inter- county connectedness.  
+
+Socioeconomic, demographic, and climatic variables have been shown to be correlated with the spread of COVID- 19<sup>21,37- 40</sup>, therefore we include category A and B features in all of our models to control for these factors. Section A of the supplementary information outlines the detailed methodology for generating features in these categories.  
+
+Features in category C indirectly capture population compartments of the susceptible- infected- recovered (SIR) epidemiological models<sup>41,42</sup>. We defined the COVID- 19 incidence rate of a county as its number of cases per 10,000 population
+
+<--- Page Split --->
+
+and included a four- week lagged (t- 4) weekly average of total incidence rates in each county as a feature (category C.i) to capture the Susceptible and Recovered compartments. If the feature value is small, many individuals in the unit have not yet contracted the disease; and therefore, are still susceptible. If the feature value is sufficiently large (level of sufficiency is learned by the model), the compartment is approaching higher levels of immunity as a whole. We use machine learning algorithms that are capable of learning such non- linear relationships. It is worth noting that vaccination rates can similarly be incorporated as a feature. However, inoculations in the U.S. started after our study period, and thus, not included here.  
+
+To account for the latency associated with the effects of temperature, new and historical incidences, and human interaction on the spread of COVID- 19, we generated four temporal weekly lags of features in categories B- E (we assume that the features in category A are static during our study period). Notably, for category C, we also included (natural log- transformed values of) change in incidence rate( \(\ln (\Delta \text{incidence rate} + 1)\) ), i.e., the subtraction of observed start- of- week incident rate from end- of- week incidence rate during the four weekly temporal- lags (t- 1,...,t- 4), as another set of features in all of our models (category C.ii) (more details in the supplementary information). These features conceptually capture
+
+<--- Page Split --->
+
+the Infected compartment in the SIR model, i.e., the currently infected populations in the spatial unit. Our autoregressive modeling with multiple temporal lags allows the models to learn the rate of spread in a unit, as well as the varying incubation periods of the disease in relation to the change in temperature and demographic features<sup>43,44</sup>.  
+
+Features in categories D and E also model the SIR population compartments, but in connected counties (through either social connectedness or flow connectedness). Features in category D represent the social media proxy (of physical human interactions), whereas features in category E represent the cell phone- derived human mobility flow proxy.  
+
+## Conceptualization of models with social media and cell phone features  
+
+We evaluate the predictive power of the Facebook- derived features (category D) and the SafeGraph- derived features (category E) against a base model by developing four different model setups (not to be confused with the final evaluations against the baseline Ensemble model). Here, we provide an outline of these models, with more details on the specific algorithms and features mentioned in Table 4 in the following sections.
+
+<--- Page Split --->
+
+week-ahead prediction horizon on forecast date \(d\) is the number of new cases per 10k from the forecast date through the end of prediction horizon \(t\) in each county, i.e., \(\Delta\) (incidence rate) \(t,d\) . The target for the two week-horizon prediction is \(\Delta\) (incidence rate) \(t+1,d\) , three week-horizon is \(\Delta\) (incidence rate) \(t+2,d\) , and four week horizon is \(\Delta\) (incidence rate) \(t+3,d\) . \(\ln (t)\) in the table indicates natural logarithm, Mean () indicates weekly average, \(\Delta\) indicates weekly change, i.e., difference (calculated by subtracting the value of the feature at the beginning of the week from its value at the end of the week) and Slope indicates the slope of a fitted linear regression model to the standardized daily measures of metric value as the dependent variable and standardized day of week as the independent variable.  
+
+<table><tr><td>Category</td><td>Model(s)</td><td>Variables</td><td>Temporal Lag</td></tr><tr><td>A- socioeconomic and demographic</td><td>All of the models</td><td>population density<br>percentage of male population<br>percentage of African American population<br>percentage of Hispanic population<br>percentage of Native American population<br>percentage of the rural population<br>percentage of the population with a college degree<br>median household income<br>percentage of the population who voted republican in 2016 presidential election</td><td>None (constant)</td></tr><tr><td>B- Temperature</td><td>All of the models</td><td>mean(daily minimum temperature) <br>mean(daily maximum temperature) <br>Ln (Δ incidence rate +1) <br>Ln (mean (incidence rate) +1)</td><td>t-1, t-2, t-3, t-4 <br>t-1, t-2, t-3, t-4 <br>t4</td></tr><tr><td>C- COVID-19 incidence rate</td><td>All of the models</td><td>Δ SPC <br>mean (SPC) <br>mean (Stay Put) <br>and slope (Stay Put) <br>t</td><td>t-1, t-2, t-3, t-4 <br>t-4 <br>t-1, t-2, t-3, t-4</td></tr></table>
+
+<--- Page Split --->
+
+
+<table><tr><td></td><td></td><td>mean (Change in Movement)t and slope (Change in Movement)t</td><td>t-1, t-2, t-3, t-4</td></tr><tr><td rowspan="4">E- SafeGraph</td><td rowspan="2">-SG and -SGR models</td><td>Δ FPC t<br>mean (FPC)t<br>mean (% completely_home_device_count)t and slope(% completely_home_device_count)t</td><td>t-1, t-2, t-3, t-4</td></tr><tr><td>mean(baselined distance_traveled_from_home)t<br>slope(baselined distance_traveled_from_home)t</td><td>t-1, t-2, t-3, t-4</td></tr><tr><td rowspan="2">-SGR model</td><td>mean (baselined median_home_dwell_time)t and slope(baselined median_home_dwell_time)t<br>mean (baselined % full_time_work_behavior_devices)t and slope(baselined % full_time_work_behavior_devices)t</td><td>t-1, t-2, t-3, t-4</td></tr><tr><td></td><td></td></tr></table>
+
+The first model (base model) only includes base features: socioeconomic features (category A), four temporally-lagged weekly temperature features (category B), and four temporally-lagged weekly change in incidence rates in each county, as well as weekly average of incidence rates during the fourth lagged week (category C). Therefore, the base model only incorporates temporal lags of the features and the target variable in predicting new cases of COVID-19.
+
+The second model, which we identify by the “-FB” suffix (to note the inclusion of Facebook features), includes the base features as well as category D features,i.e. Facebook-derived intra-county movement features and inter-county spatiotemporal lags of the target variable, i.e. exposure to COVID-19 through
+
+<--- Page Split --->
+
+social connectedness (Social Proximity to Cases), across four temporal lags. The third model, which we identify by the "- SG" suffix, is conceptually similar to the - FB model, but with features derived from SafeGraph cell phone mobility data instead of Facebook data. Specifically, the - SG models include the base features in addition to inter- county spatiotemporal lags of the target variable, i.e. exposure to COVID- 19 through human flow connectedness (which we call Flow Proximity to Cases), and a subset of category E SafeGraph- derived intra- county movement features, across four temporal lags.  
+
+To explore the full potential of the movement features provided by the SafeGraph Social Distancing Metrics (SDM) dataset, we developed a fourth model, in which two additional mobility- related measurements provided in the SDM dataset (that are least correlated with other features in Category E) are added to the - SG model. This model thus includes categories A- C and all features in category E, and is identified by the "- SGR" suffix.
+
+<--- Page Split --->
+
+## Features derived from Facebook  
+
+### i. Intra-county movement features  
+
+Facebook publishes the Movement Range dataset for 14 countries<sup>45</sup> and it includes two metrics called "Change in Movement" and "Stay Put", each providing a different perspective on movement trends as measured by mobile devices carrying the Facebook app. The Change in Movement metric for each county is a measure of relative change in aggregated movement compared to the baseline of February \(2^{\text{nd}}\) to February \(29^{\text{th}}\) 2020 (excluding the February \(17^{\text{th}}\) 2020 President Day holiday in the US)<sup>45</sup>. The Stay Put metric measures "the fraction of the population that have stayed within a small area during an entire day"<sup>45</sup>. We used four temporal lags of weekly averages and slopes of each metric as a feature in our - FB model. We calculated the slopes by fitting a linear regression model to the metric value as the dependent variable and day of week as the independent variable, both transformed to standard scale N(0,1). The slope feature characterizes the overall trend in a week, as compared to the baseline period.
+
+<--- Page Split --->
+
+## ii. Inter-county features and spatial lag modeling  
+
+The intra- county features capture the intrinsic movement- related characteristics of a county and ignore its interactions (i.e. spatial lags) with the counties to which it is connected. Therefore, we calculated inter- county metrics of connectivity as a basis for incorporating spatiotemporal lags in our models. Notably, the connectedness in this context transcends spatial connectedness in the form of mere physical adjacency.  
+
+Social connectedness Index (SCI), another dataset published by Facebook, is a measure of the intensity of connectedness between administrative units, calculated from Facebook friendship data. Social connectedness between two counties \(i\) and \(j\) is defined as<sup>46</sup>:
+
+<--- Page Split --->
+
+\[Social connectedness (SC)_{i,j} = \frac{FB Connections_{i,j}}{FB Users_{i} * FB Users_{j}} \quad (1)\]  
+
+where FB Connections \(_{i,j}\) is the number of friendships between Facebook users who live in county \(i\) and those who live in county \(j;\) while FB Users \(i\) and FB Users \(j\) are the total number of active Facebook users in counties \(i j\) and, respectively. Social Connectedness is scaled to a range between 1 and 1,000,000,000 and rounded to the nearest integer to generate SCI, as published by Facebook47. Therefore, if the SCI value between a pair of counties is twice as large as another pair, it means the users in the first county- pair are almost twice as likely to be friends on Facebook than the second county- pair46. We used the latest version of the SCI dataset (at the time of our analyses), which was released in August 202047.  
+
+While SCI provides a measure of connectivity, our goal is to capture the spatiotemporal lags of COVID- 19 cases in county \(i\) , i.e., the number of recent COVID- 19 cases in other counties connected to county \(i\) . Using SCI, Kuchler et al. \(^{6}\) created a new metric, called Social Proximity to Cases (SPC) for each county, which is a measure of the level of exposure to COVID- 19 cases in connected counties through social connectedness. We use a slight variation of SPC, defined as follows for county \(i\) at time \(t\) :
+
+<--- Page Split --->
+
+\[\begin{array}{l}{{S o c i a l P r o x i m i t y t o C a s e s(S P C)_{i,t}}}\\ {{=\sum_{j}C a s e s P e r10k_{j,t}\times\frac{S o c i a l C o n n e c t e d n e s s_{i,j}}{\sum_{h}S o c i a l C o n n e c t e d n e s s_{h}}}}\end{array} \quad (2)\]  
+
+where Cases Per \(10k_{j,t}\) is the number of COVID- 19 cases per 10k population (i.e., incidence rate) in county \(j\) as of time \(t\) . For county \(i\) , the sums \(j\) and \(h\) are over all counties. In other words, SPC for a county \(i\) , in time \(t\) , is the average of COVID- 19 incidence rates in connected counties weighted by their social connectedness to county \(i\) , i.e., the spatial lag of incidence rates. To the best of our knowledge, SPC data has not been published, but we were able to generate this feature using the original method \(^{6}\) , modified for our weekly temporal lagged features and calculated using incidence rates (cases per 10k population) rather than total number of cases. In the - FB models (Table 4), we incorporated features of weekly change \((\Delta)\) in SPC at four temporally lagged weeks (difference between the end and start of the lag week) to model the Infected SIR compartment in connected counties, as well as weekly average of SPC in the fourth lagged week (t- 4), to capture the Susceptible and Recovered SIR compartments in connected counties, similar to the rational for features in category C, as explained earlier.
+
+<--- Page Split --->
+
+## Features derived from SafeGraph  
+
+### i. Intra-county movement features  
+
+To generate movement features from cell phone data, we used SafeGraph's SDM dataset that is "generated using a panel of GPS pings from anonymous mobile devices"48. The SDM dataset contains multiple mobility metrics published at the Census Block Group (CBG) level. Among these metrics, distance_traveled_from_home (median distance traveled by the observed devices in meters) and completely_home_device_count (the number of devices that did not leave their home location during a day)48 are conceptually closest to the metrics included in the Facebook's Movement Range Dataset. We used these two features in our - SG model, which is the conceptual equivalent of the - FB model, but with cell phone- derived features instead of the Facebook- derived features (Table 4).  
+
+We included the SafeGraph's median_home_dwell_time (median dwell time at home in minutes for all observed devices during the time period), and full_time_work_behavior_devices (the number of devices that spent more than 6 hours at a location other than their home during the day)48 in addition to the
+
+<--- Page Split --->
+
+previous two features in the - SGR model to take fuller advantage of the metrics available in the SDM dataset.  
+
+We derived baselined features from the SDM metrics as such: To address the potential effect of fewer cell phone observations in some CBGs, we used a Bayesian hierarchical model<sup>49,50</sup> with two levels (states and counties), and then smoothed the daily measurements using a seven- day rolling average to reduce the effect of outliers in the data. We then aggregated CBG- level completely_home_device_count and full_time_work_behavior_devices values up to the county level, divided by the total device_count in the county on the same day. For full_time_work_behavior_devices, we subtracted the final proportion from the February baseline of the same metric. For the median_home_dwell_time and distance_traveled_from_home variables, we calculated the weighted mean (by CBG population) of values per county, and then calculated percent of change compared to February baseline.  
+
+We used weekly averages and slopes (calculated by fitting a linear regression model to the values as the response variable and day of week as the independent variable) of these four metrics as features in our models (Table 4).
+
+<--- Page Split --->
+
+## ii. Inter-county features and spatial lag modeling  
+
+Building on the conceptual structure of SCI, we derived a novel and daily inter- county connectivity index from SafeGraph's SDM dataset to quantify connectedness between counties based on the level of human flow from one county to the other (measured through cell- phone pings). We call this index "Flow Connectedness Index" (FCI). Using FCI, we then calculated a spatial lag metric that we call "Flow Proximity to Cases" (FPC) for each county. FPC captures the average of COVID- 19 incidence rates in connected (by human movement) counties weighted by the FCI. Again, it is worth noting that connectedness in this sense goes beyond the physical connectivity of counties, and considers daily human movement between them as the basis for determining connectivity. The similar formulations of FCI and SCI, as well as FPC and SPC, allow for direct comparison of the two networks (i.e. FB's friendship network and SafeGraph's human flow network) in their capability to capture inter- county physical human interactions, and subsequently, to predict new COVID- 19 cases.  
+
+The SafeGraph's SDM contains the number of visits between different CBGs. We aggregate these values to the county level to measure the daily number of devices that move (flow) between each county pair. Leveraging these flow measurements, we define Flow Connectedness Index (FCI) as:
+
+<--- Page Split --->
+
+\[Flow connectedness index(FCI)_{i,j} = \frac{Device flow_{i,j} + Device flow_{j,i}}{Device count_{i} * Device count_{j}} \quad (3)\]  
+
+where for counties \(i\) and \(j\) , Device flow \(i,j\) is the sum of visits with origin \(i\) and destination \(j\) . Device count \(i\) is the number of devices whose home location is in county \(i\) . We then scale FCI to a range between 1 to 1,000,000,000.  
+
+We defined FPC as:  
+
+\[\begin{array}{l}{{Flow P r o x i m i t y t o C a s e s(F P C)_{i,t}}}\\ {{=\sum_{j}C a s e s P e r10k_{j,t}\times\frac{F l o w C o n n e c t e d n e s s_{i,j}}{\sum_{h}F l o w C o n n e c t e d n e s s_{i,h}}}}\end{array} \quad (4)\]  
+
+Where Cases Per \(10k_{j,t}\) is the number of confirmed COVID- 19 cases per 10k population in county \(j\) at time \(t\) , and Flow Connectedness \(i,j\) is the value of FCI between county \(i\) and \(j\) .  
+
+Facebook's social network and friendship connections do not change significantly over time, and therefore, SCI is a static index in a one- year period. Conversely, inter- county human flow from SafeGraph is dynamic and can change dramatically, even within a week. We generated daily FCI (and FPC) for each county- pair in the US to utilize the full temporal resolution of the SDM dataset. We used weekly change \((\Delta)\) of FPC for the four temporally lagged weeks, and its average only in the fourth week as features - SG and - SGR models, with the same
+
+<--- Page Split --->
+
+rationale as features in Category C and D to capture SIR compartments in connected counties (Table 4).  
+
+## Model Implementation  
+
+The ultimate target variable in all of our autoregressive models is the number of new cases of COVID- 19. For training and tuning the models, however, we used a transformed target variable, namely the natural log- transformed values of new cases per 10k population plus one (to avoid zero values). For reporting the model predictions, we computed the number of new cases by applying an inverse transformation, i.e. an exponential transformation minus one (Formula 5a- c). The rationale for using the log- transformed target variable, as opposed to directly predicting the weekly new cases, was to minimize skewness, and more importantly, minimize the sensitivity of the models to the population of counties. Our exploratory work did confirm that using this logged of incidence rates produced better results.  
+
+\[\begin{array}{r l} & {y_{p r e d i c t e d(i,t)} = l n(\Delta i n c i d e n c e r a t e_{(i,t)} + 1)}\\ & {\Delta i n c i d e n c e r a t e_{p r e d i c t e d(i,t)} = \mathrm{e}^{y_{p r e d i c t e d(i,t)}} - 1}\\ & {\Delta C a s e_{p r e d i c t e d(i,t)} = \left(\Delta i n c i d e n c e r a t e_{p r e d i c t e d(i,t)}\right)*P o p u l a t i o n_{i} / 10,000} \end{array} \quad (5.c)\]
+
+<--- Page Split --->
+
+\(\Delta C a s e_{p r e d i c t e d_{(i,t)}}\) in 5.c denotes the number of new cases in a county in the prediction horizon.  
+
+Our training dataset includes up to 34 training samples per county (number of total samples \(n = 3103 \times 34\) ), with each sample holding various features in one- to four- weekly temporal lags (Table 4). The weekly calculation of features is based on weeks starting on Sundays and ending on Saturdays, with predictions also made for horizons spanning Sunday- Saturday periods as in common practice<sup>25</sup>. Our features, models, evaluations and comparisons are limited to the counties in the coterminous US. Table 4 summarizes the features that we used and the number of temporal lags (if any) used for each feature. All features were standardized for use in machine learning algorithms.  
+
+Our general approach to training, validation and testing of our models for different prediction horizons is similar, only, with target variables calculated separately for the specific prediction horizon. We first outline our approach for one- week ahead prediction horizons, which is used as the basis for algorithm selection and comparison of Facebook- and SafeGraph- derived features (- FB and - SG models); It is worth noting that we compare the two feature sets in longer
+
+<--- Page Split --->
+
+term predictions too (Supplementary Table 5). We then provide an overview of the implementation of the models for longer term prediction horizons.  
+
+We trained and tuned the models using randomized search and 5- fold cross validation, and tested the best tuned model for predicting new cases in the week following the forecast date (for the one- week prediction horizons, as listed in Table 5). For instance, for the forecast date of October \(24^{\text{th}}\) , we used features which were generated using data collected before October \(24^{\text{th}}\) for tuning and training. The tuned model was then used for predicting new cases in each county during the October \(24^{\text{th}}\) to October \(31^{\text{st}}\) period. We used the reported cases by the JHU CSSE. The temporally lagged features for this forecast date were generated for t- 1, t- 2, t- 3 and t- 4 weekly lags, namely, the weeks ending on Oct. 24, Oct. 17, Oct. 10, and Oct. 3 respectively.  
+
+For the next forecast date, October 31, the training size increased by one week (per county), and the target week was also shifted by one week. Table 5 summarizes the forecast dates, one- week ahead prediction horizons, and training data size. The data used in generating these features spans a period from March 29 to November 28, 2020 to cover the temporal lags, and the target variable is collected through December 12, 2020 for the evaluation of four- week ahead predictions on the last forecast date. More details on cross validation,
+
+<--- Page Split --->
+
+hyperparameters, and evaluation are presented in the supplementary information (Section C).  
+
+Table 5. Summary of training and testing data used for machine learning algorithm selection, as well as comparison of -FB and -SG models in short-term predictions   
+
+<table><tr><td>Forecast Date</td><td>One-week prediction Horizon</td><td>Training data start date</td><td>Training data end date</td><td># training samples per county</td></tr><tr><td>2020-10-24</td><td>2020-10-25 to 2020-10-31</td><td>2020-03-29</td><td>2020-10-24</td><td>30</td></tr><tr><td>2020-10-31</td><td>2020-11-01 to 2020-11-07</td><td>2020-03-29</td><td>2020-10-31</td><td>31</td></tr><tr><td>2020-11-07</td><td>2020-11-08 to 2020-11-14</td><td>2020-03-29</td><td>2020-11-07</td><td>32</td></tr><tr><td>2020-11-14</td><td>2020-11-15 to 2020-11-21</td><td>2020-03-29</td><td>2020-11-14</td><td>33</td></tr><tr><td>2020-11-21</td><td>2020-11-22 to 2020-11-28</td><td>2020-03-29</td><td>2020-11-21</td><td>34</td></tr></table>  
+
+We experimented with five different supervised machine learning regression algorithms, namely Random Forest \(^{51}\) (RF), Stochastic Gradient Boosting \(^{52}\) (SGB), eXtreme Gradient Boosting \(^{36,53}\) (XGB), Feed Forward Neural Network \(^{54}\) (FFNN), and Long Short- Term Memory \(^{55}\) (LSTM) network to build the autoregressive machine learning models with features described in Table 4. We evaluated the models using the dates listed in Table 5. Results are presented in Table 1. The
+
+<--- Page Split --->
+
+details of hyperparameter candidates and specific architectures are presented in the supplementary information (Section C).  
+
+## Comparing Facebook-derived features with SafeGraph-derived features  
+
+Since the XGB algorithm performed best (Table 1), we chose it as the selected machine learning algorithm, and trained the base, - FB, - SG, and - SGR models using the XGB algorithm to predict new cases of COVID- 19 in short- term (one week) and long- term (two to four weeks) prediction horizons. To name a specific model in this article, we use a prefix that denotes the type of lag included in the model features (i.e. T for temporal or ST for spatiotemporal), followed by the name of the algorithm (XGB), followed by a suffix denoting the features included in the model, namely, - FB, - SG, and - SGR. Thus, TXGB (Temporal eXtreme Gradient Boosting) denotes the model that is built using the XGB algorithm and includes the base, Temporally lagged features; and STXGB- FB (SpatioTemporal eXtreme Gradient Boosting) denotes the model that includes Facebook- derived features (and thus, spatiotemporal lags) and is built using XGB.  
+
+We evaluated the performance of TXGB, STXGB- FB, STXGB- SG, and STXGB- SGR by comparing the RMSE and MAE scores of the predictions against the observed numbers of new cases in the corresponding prediction horizon (results in Table 2).
+
+<--- Page Split --->
+
+We also tested the - FB and - SG models for all prediction horizons across all forecast dates (results in Supplementary Table 5).  
+
+Re- tuning and re- training all different variations of the STXGB model for each forecast date and prediction horizon resulted in considerably improved predictions compared to the baseline model (Table 4 and Supplementary Tables 4 and 5). Each STXGB model was tuned and trained on a regular desktop machine (with a 6 core Ryzen 5 3600X CPU and 64GB of RAM) in approximately 12- 13 minutes for a single prediction horizon, and thus, in almost one hour for all of the four prediction horizons.  
+
+## Evaluation against the COVID-19 Forecast Hub Ensemble baseline  
+
+In addition to the one- week short- term predictions, we performed long- term predictions of new COVID- 19 cases in two-, three-, and four- week ahead prediction horizons. We only used the STXGB algorithm to develop long- term prediction models since it outperformed other algorithms in short- term predictions (see Section 2). We used the same set of features for long- term predictions, with modifications on the target variable to reflect different prediction horizons. For instance, the two-, three-, and four- week ahead horizons
+
+<--- Page Split --->
+
+of the Forecast date Oct. 24, were Oct 24 to Nov. 7, Oct. 24 to Nov. 14, and Oct 24 to Nov. 21, respectively.  
+
+The model for each horizon was trained and validated separately using the same training data and approach described in the previous Section (Table 5), and was tested on two, three and four weeks of unseen data, respectively, for each horizon. We evaluated the models' predictions by comparing them against the predictions generated by the COVID- 19 Forecast Hub's Ensemble model as well as the ground- truth values of new cases derived from JHU CSSE COVID- 19 reports. Additionally, we compared the long- term predictions of STXGB- FB and STXGB- SG (Supplementary Table 5).  
+
+## Data Availability  
+
+All of the datasets used in this study are publicly available (at the time of writing this manuscript). We created socioeconomic features from the 5- year survey data- - between 2014- 2018- - provided by the American Community Survey (ACS) and available at IPUMS National Historical GIS portal (https://www.nhgis.org/). Daily maximum and minimum temperature surfaces of the U.S. published by the NOAA are available at https://ftp.cpc.ncep.noaa.gov/GIS/GRADS_GIS/GeoTIFF/TEMP/. We used the
+
+<--- Page Split --->
+
+cumulative confirmed COVID- 19 cases published by the Johns Hopkins University Center for Systems Science and Engineering (JHU CSSE) to generate COVID- related features. Facebook's Social Connectedness Index (SCI) database is available at https://dataforgood.fb.com/tools/social-connectedness- index/ and the movement range dataset can be found at https://data.humdata.org/dataset/movement- range- maps. Finally, the instructions for accessing SafeGraph's Social Distancing Metrics dataset is available at https://docs.safegraph.com/docs/social- distancing- metrics.  
+
+## Code Availability  
+
+All code necessary for the replication of our results is available for reviewers upon request. The code will be published publicly on GitHub under MIT license upon acceptance of this article.  
+
+## Summary of Contributions  
+
+Morteza Karimzadeh conceptualized the project, designed the features and contributed \(30\%\) of data processing and implementation, and contributed equally to writing. Behzad Vahedi conducted the majority of data processing, implementation, literature review, and contributed equally to writing. Hamidreza
+
+<--- Page Split --->
+
+Zoraghein contributed to the study design and \(10\%\) of implementation and writing.  
+
+## Competing interests  
+
+The authors do not report any competing interests.
+
+<--- Page Split --->
+
+## References  
+
+1. Chu, D. K. et al. Physical distancing, face masks, and eye protection to prevent person-to-person transmission of SARS-CoV-2 and COVID-19: a systematic review and meta-analysis. The Lancet 395, 1973-1987 (2020).  
+
+2. Chiu, W. A., Fischer, R. & Ndeffo-Mbah, M. L. State-level needs for social distancing and contact tracing to contain COVID-19 in the United States. Nat. Hum. Behav. 4, 1080-1090 (2020).  
+
+3. Eames, K. T. D. & Keeling, M. J. Contact tracing and disease control. Proc. R. Soc. Lond. B Biol. Sci. 270, 2565-2571 (2003).  
+
+4. Clark, E., Chiao, E. Y. & Amirian, E. S. Why Contact Tracing Efforts Have Failed to Curb Coronavirus Disease 2019 (COVID-19) Transmission in Much of the United States. Clin. Infect. Dis. (2020) doi:10.1093/cid/ciaa1155.  
+
+5. Bailey, M., Cao, R., Kuchler, T. & Stroebel, J. The Economic Effects of Social Networks: Evidence from the Housing Market. J. Polit. Econ. 126, 2224-2276 (2018).  
+
+6. Kuchler, T., Russel, D. & Stroebel, J. The Geographic Spread of COVID-19 Correlates with the Structure of Social Networks as Measured by Facebook. J. Urban Econ., 103314 (2020).
+
+<--- Page Split --->
+
+7. Büchel, K. & Ehrlich, M. v. Cities and the structure of social interactions: Evidence from mobile phone data. J. Urban Econ. 119, 103276 (2020).  
+
+8. Kang, Y. et al. Multiscale dynamic human mobility flow dataset in the U.S. during the COVID-19 epidemic. Sci. Data 7, 390 (2020).  
+
+9. Bullinger, L. R., Carr, J. B. & Packham, A. COVID-19 and Crime: Effects of Stay-at-Home Orders on Domestic Violence. https://www.nber.org/papers/w27667 (2020) doi:10.3386/w27667.  
+
+10. Killeen, B. D. et al. A County-level Dataset for Informing the United States' Response to COVID-19. Preprint at https://arxiv.org/abs/2004.00756 (2020).  
+
+11. Gao, S., Rao, J., Kang, Y., Liang, Y. & Kruse, J. Mapping county-level mobility pattern changes in the United States in response to COVID-19. (2020).  
+
+12. Sen-Crowe, B., McKenney, M. & Elkbuli, A. Social distancing during the COVID-19 pandemic: Staying home save lives. Am. J. Emerg. Med. 38, 1519–1520 (2020).  
+
+13. Unwin, H. J. T. et al. State-level tracking of COVID-19 in the United States. Nat. Commun. 11, 6189 (2020).  
+
+14. Bertozzi, A. L., Franco, E., Mohler, G., Short, M. B. & Sledge, D. The challenges of modeling and forecasting the spread of COVID-19. Proc. Natl. Acad. Sci. 117, 16732–16738 (2020).
+
+<--- Page Split --->
+
+15. Zhou, Y. et al. A Spatiotemporal Epidemiological Prediction Model to Inform County-Level COVID-19 Risk in the United States. *Harv. Data Sci. Rev.* (2020) doi:10.1162/99608f92.79e1f45e.  
+
+16. Singh, R. K. et al. Prediction of the COVID-19 Pandemic for the Top 15 Affected Countries: Advanced Autoregressive Integrated Moving Average (ARIMA) Model. *JMIR Public Health Surveill.* 6, e19115 (2020).  
+
+17. Dansana, D. et al. Global Forecasting Confirmed and Fatal Cases of COVID-19 Outbreak Using Autoregressive Integrated Moving Average Model. *Front. Public Health* 8, (2020).  
+
+18. Xiang, J. et al. Impacts of the COVID-19 responses on traffic-related air pollution in a Northwestern US city. *Sci. Total Environ.* 747, 141325 (2020).  
+
+19. Zeroual, A., Harrou, F., Dairi, A. & Sun, Y. Deep learning methods for forecasting COVID-19 time-Series data: A Comparative study. *Chaos Solitons Fractals* 140, 110121 (2020).  
+
+20. Jo, H., Kim, J., Huang, T.-C. & Ni, Y.-L. condLSTM-Q: A novel deep learning model for predicting Covid-19 mortality in fine geographical Scale. Preprint at https://arxiv.org/abs/2011.11507 (2020).
+
+<--- Page Split --->
+
+21. Mollalo, A., Rivera, K. M. & Vahedi, B. Artificial Neural Network Modeling of Novel Coronavirus (COVID-19) Incidence Rates across the Continental United States. Int. J. Environ. Res. Public. Health 17, 4204 (2020).  
+
+22. Buchwald, A. G., Adams, J., Bortz, D. M. & Carlton, E. J. Infectious Disease Transmission Models to Predict, Evaluate, and Improve Understanding of COVID-19 Trajectory and Interventions. Ann. Am. Thorac. Soc. 17, 1204–1206 (2020).  
+
+23. Reiner, R. C. et al. Modeling COVID-19 scenarios for the United States. Nat. Med. 27, 94–105 (2021).  
+
+24. Santosh, K. C. COVID-19 Prediction Models and Unexploited Data. J. Med. Syst. 44, 170 (2020).  
+
+25. Ray, E. L. et al. Ensemble Forecasts of Coronavirus Disease 2019 (COVID-19) in the U.S. Preprint at https://www.medrxiv.org/content/10.1101/2020.08.19.20177493v1 (2020).  
+
+26. CDC. Cases, Data, and Surveillance. Centers for Disease Control and Prevention https://www.cdc.gov/coronavirus/2019-ncov/cases-updates/forecasts-cases.html (Retrieved on Jan 30, 2020).  
+
+27. Dong, E., Du, H. & Gardner, L. An interactive web-based dashboard to track COVID-19 in real time. Lancet Infect. Dis. 20, 533–534 (2020).
+
+<--- Page Split --->
+
+28. Epidemiologists Urge A Cautious Christmas, After Thanksgiving Surge in Some States. NPR.org https://www.npr.org/sections/healthshots/2020/12/21/948809129/epidemiologists-urge-a-cautious-christmas-after-thanksgiving-surge-in-some-state (Retrieved on Jan 30, 2020).  
+
+29. Rama, D., Mejova, Y., Tizzoni, M., Kalimeri, K. & Weber, I. Facebook Ads as a Demographic Tool to Measure the Urban-Rural Divide. in Proceedings of The Web Conference 2020 327–338 (Association for Computing Machinery, 2020). doi:10.1145/3366423.3380118.  
+
+30. Wang, Q., Phillips, N. E., Small, M. L. & Sampson, R. J. Urban mobility and neighborhood isolation in America’s 50 largest cities. Proc. Natl. Acad. Sci. 115, 7735–7740 (2018).  
+
+31. Coston, A. et al. Leveraging Administrative Data for Bias Audits: Assessing Disparate Coverage with Mobility Data for COVID-19 Policy. Preprint at https://arxiv.org/abs/2011.07194 (2020).  
+
+32. Nikolopoulos, K., Punia, S., Schäfers, A., Tsinopoulos, C. & Vasilakis, C. Forecasting and planning during a pandemic: COVID-19 growth rates, supply chain disruptions, and governmental decisions. Eur. J. Oper. Res. 290, 99–115 (2021).
+
+<--- Page Split --->
+
+33. CDC. Coronavirus Disease 2019 (COVID-19). Centers for Disease Control and Prevention https://www.cdc.gov/coronavirus/2019-ncov/more/masking- science-sars-cov2.html (Retrieved on Jan 30, 2020).  
+
+34. Wyche, S. P., Schoenebeck, S. Y. & Forte, A. 'Facebook is a luxury': an exploratory study of social media use in rural Kenya. in Proceedings of the 2013 conference on Computer supported cooperative work 33–44 (Association for Computing Machinery, 2013).  
+
+35. LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444 (2015).  
+
+36. Chen, T. & Guestrin, C. XGBoost: A Scalable Tree Boosting System. in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining 785–794 (Association for Computing Machinery, 2016). doi:10.1145/2939672.2939785.  
+
+37. Bashir, M. F. et al. Correlation between climate indicators and COVID-19 pandemic in New York, USA. Sci. Total Environ. 728, 138835 (2020).  
+
+38. Mollalo, A., Vahedi, B. & Rivera, K. M. GIS-based spatial modeling of COVID-19 incidence rate in the continental United States. Sci. Total Environ. 728, 138884 (2020).
+
+<--- Page Split --->
+
+39. Khalatbari-Soltani, S., Cumming, R. C., Delpierre, C. & Kelly-Irving, M. Importance of collecting data on socioeconomic determinants from the early stage of the COVID-19 outbreak onwards. J Epidemiol Community Health 74, 620–623 (2020).  
+
+40. Li, H. et al. Air pollution and temperature are associated with increased COVID-19 incidence: A time series study. Int. J. Infect. Dis. 97, 278–282 (2020).  
+
+41. Toda, A. A. Susceptible-Infected-Recovered (SIR) Dynamics of COVID-19 and Economic Impact. Preprint at https://arxiv.org/abs/2003.11221(2020).  
+
+42. Volz, E. & Meyers, L. A. Susceptible–infected–recovered epidemics in dynamic contact networks. Proc. R. Soc. B Biol. Sci. 274, 2925–2934 (2007).  
+
+43. Lauer, S. A. et al. The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application. Ann. Intern. Med. 172, 577–582 (2020).  
+
+44. Chen, N. et al. Epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study. The Lancet 395, 507–513 (2020).  
+
+45. Protecting privacy in Facebook mobility data during the COVID-19 response. Facebook Research https://research.fb.com/blog/2020/06/protecting-privacy-
+
+<--- Page Split --->
+
+in- facebook- mobility- data- during- the- covid- 19- response/ ( Retrieved on Jan 30, 2020).  
+
+46. Bailey, M., Cao, R., Kuchler, T., Stroebel, J. & Wong, A. Social Connectedness: Measurement, Determinants, and Effects. J. Econ. Perspect. 32, 259–280 (2018).  
+
+47. Social Connectedness Index Methodology. Facebook Data for Good https://dataforgood.fb.com/docs/social-connectedness-index-methodology/ (Retrieved on Jan 30, 2020).  
+
+48. Social Distancing Metrics. SafeGraph https://docs.safegraph.com/docs/social-distancing-metrics ( Retrieved on Jan 30, 2020).  
+
+49. Gelman, A. et al. Bayesian Data Analysis, Third Edition. (CRC Press, 2013).  
+
+50. Devine, O. J., Louis, T. A. & Halloran, M. E. Empirical Bayes Methods for Stabilizing Incidence Rates before Mapping. Epidemiology 5, 622–630 (1994).  
+
+51. Breiman, L. Random forests. Mach. Learn. 45, 5–32 (2001).  
+
+52. Friedman, J. H. Stochastic gradient boosting. Comput. Stat. Data Anal. 38, 367–378 (2002).  
+
+53. Friedman, J. H. Greedy Function Approximation: A Gradient Boosting Machine. Ann. Stat. 29, 1189–1232 (2001).
+
+<--- Page Split --->
+
+54. Fine, T. L. \*Feedforward Neural Network Methodology\*. (Springer Science & Business Media, 2006).55. Hochreiter, S. & Schmidhuber, J. Long Short-Term Memory. \*Neural Comput.\* 9, 1735–1780 (1997).
+
+<--- Page Split --->
+
+55
+
+<--- Page Split --->
+
+## Figures  
+
+![](images/Figure_1.jpg)
+
+<center>Figure 1 </center>  
+
+![](images/Figure_2.jpg)
+  
+
+Prediction errors of the SpatioTemporal eXtreme Gradient Boosting (STXGB) models in one- week prediction horizon against a temporal autoregressive model without spatial lags (TXGB). Error comparison of STXGB models with different feature sets in predicting weekly new cases (top row) and new cases per 10k population (bottom row) for one- week ahead horizon. STXGB- FB, which incorporates Facebook- derived features, including spatial lags based on Social Connectedness Index, outperforms other models. Left column: prediction RMSE. Right column: prediction MAE.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 2 </center>  
+
+Long- term prediction error comparison. Prediction errors of the STXGB- FB model (dashed line) and the Ensemble baseline (solid line) over four prediction horizons on five forecasting dates. a) Prediction RMSE and b) Prediction MAE.
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 3 </center>  
+
+Map of COVID- 19 cases per 10k population and errors in predicting them. (a) number of confirmed new cases per 10k population over the week ahead of forecasting date Oct. 31, 2020 (b) prediction errors for the same forecasting date(c) number of new cases over the week ahead of Nov. 7, 2020 forecasting date (d) prediction errors for the same forecasting date. The pattern of errors in Georgia, and Texas, and Kentucky flip from Oct. 31 to Nov. 7, indicating potential lags in testing and reporting.
+
+<--- Page Split --->
+![PLACEHOLDER_60_0]
+
+<center>Figure 4 </center>  
+
+Prediction errors in urban vs. rural counties. Prediction errors of the number of new cases (left column) and new cases per 10k population (right column) in rural and urban counties on the Nov. 7 forecast date across four prediction horizons. The higher and lower \(3\%\) of counties are trimmed from the plot view.  
+
+## Supplementary Files
+
+<--- Page Split --->
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- SPCandFPCSupplementaryinformationsubmissionready.docx
+
+<--- Page Split --->

preprint/preprint__48de5430c259c6879fe0e3cc8610c93e4199c05191d114610b87fb8ee3f4f045/preprint__48de5430c259c6879fe0e3cc8610c93e4199c05191d114610b87fb8ee3f4f045_det.mmd

@@ -0,0 +1,722 @@
+<|ref|>title<|/ref|><|det|>[[44, 106, 952, 241]]<|/det|>
+# Predicting County-Level COVID-19 Cases using Spatiotemporal Machine Learning: Modeling Human Interactions using Social Media and Cell-Phone Data  
+
+<|ref|>text<|/ref|><|det|>[[44, 262, 675, 402]]<|/det|>
+Behzad Vahedi ( \(\boxed{ \begin{array}{r l} \end{array} }\) behzad@colorado.edu ) University of Colorado Boulder https://orcid.org/0000- 0001- 5782- 3831 Morteza Karimzadeh University of Colorado Boulder Hamidreza Zoraghein Social and Behavioral Science Research, Population Council  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 440, 102, 457]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 477, 519, 497]]<|/det|>
+Keywords: spatiotemporal machine learning, COVID- 19  
+
+<|ref|>text<|/ref|><|det|>[[44, 515, 333, 535]]<|/det|>
+Posted Date: February 10th, 2021  
+
+<|ref|>text<|/ref|><|det|>[[44, 553, 463, 572]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 203188/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 590, 910, 633]]<|/det|>
+License: \(\circledcirc\) This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 668, 945, 711]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on November 8th, 2021. See the published version at https://doi.org/10.1038/s41467- 021- 26742- 6.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[115, 115, 881, 181]]<|/det|>
+# Predicting County-Level COVID-19 Cases using Spatiotemporal Machine Learning: Modeling Human Interactions using Social Media and Cell-Phone Data  
+
+<|ref|>text<|/ref|><|det|>[[130, 323, 880, 450]]<|/det|>
+Behzad Vahedi<sup>1\*</sup>, Morteza Karimzadeh<sup>1\*</sup>, Hamidreza Zoraghein<sup>2</sup>  <sup>1</sup> Department of Geography, University of Colorado Boulder; Behzad@colorado.edu; karimzadeh@colorado.edu  <sup>2</sup> Social and Behavioral Science Research, Population Council, New York, USA; hzoraghein@popcouncil.org  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 504, 198, 522]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[112, 552, 880, 877]]<|/det|>
+Measurements of human interaction through proxies such as social connectedness or movement patterns have proved useful for predictive modeling of COVID- 19. In this study, we first compare the power of Facebook's social connectedness with cell phone- derived human mobility for predicting county- level new cases of COVID- 19. Our experiments show that social connectedness is a better proxy for measuring human interactions leading to new infections. Next, we develop a SpatioTemporal autoregressive eXtreme Gradient Boosting (STXGB) model to predict county- level new cases of COVID- 19 in the coterminous US. We
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 860, 286]]<|/det|>
+evaluate the model on five weekly forecast dates between October 24 and November 28, 2020 over one- to four- week prediction horizons. Comparing our predictions with a baseline Ensemble of 32- models currently used by the CDC indicates an average \(58\%\) improvement in prediction RMSEs over two- to four- week prediction horizons, pointing to the strong predictive power of our model.  
+
+<|ref|>sub_title<|/ref|><|det|>[[144, 331, 295, 352]]<|/det|>
+## 1. Introduction  
+
+<|ref|>text<|/ref|><|det|>[[112, 380, 884, 664]]<|/det|>
+Human interaction in close physical proximity is the primary cause of the transmission of highly contagious diseases such as COVID- 19<sup>1</sup>. Measuring human interaction is therefore an important step in understanding and predicting the spread of COVID- 19<sup>2,3</sup>. However, tracking human interactions requires rigorous contact tracing at national and regional scales which has not been implemented in the United States due to the economic, legal, and sociocultural concerns, as well as inadequate testing supplies, and insufficient national coordination<sup>4</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[112, 690, 861, 887]]<|/det|>
+As a result, researchers have adopted different proxies to track human interaction levels. One such proxy is the "Social Connectedness Index" (SCI), generated from Facebook's friendship data. SCI represents the probability that two users in a pair of regions (e.g., U.S. counties) are friends (i.e., connected) on Facebook<sup>5</sup>. Kuchler et al.<sup>6</sup> reported on the strong correlation between early
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 879, 285]]<|/det|>
+hotspots of COVID- 19 outbreak and their level of social connectedness. The underlying assumption in leveraging SCl as a proxy for physical human interactions is that individuals who are socially connected on Facebook have a higher probability for physical interaction, thereby, potentially contributing to the spread of communicable diseases.  
+
+<|ref|>text<|/ref|><|det|>[[112, 313, 872, 682]]<|/det|>
+Human mobility flow, as measured by anonymized cell phone data, serves as another proxy for quantifying human interactions<sup>7,8</sup>. Widely used to study the spread of COVID- 19, most studies incorporating cell- phone data have focused on the change in mobility within a spatial unit<sup>9,10</sup>, while a few others have also incorporated the flow between different spatial units<sup>11</sup> to predict transmissions across units, albeit mostly in the early stages of the pandemic with limited evaluation data. The underlying assumption in this approach is that more movements between spatial units leads to higher interactions, and consequently, an elevated risk of disease spread.  
+
+<|ref|>text<|/ref|><|det|>[[112, 709, 880, 904]]<|/det|>
+It is unclear, however, which of these approaches—using social media connectedness versus cell phone- derived human mobility flow—is a better indicator of physical interaction within and between different regions. Furthermore, the underlying assumption in each approach may not necessarily be valid in the case of COVID- 19: considering the sporadic and regional stay- at- home
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 861, 328]]<|/det|>
+orders across the United States, social connectedness may not lead to physical interaction, at least not to the same level as pre- pandemic. Similarly, given the recommended preventative measures such as mask- wearing and physical distancing<sup>12</sup>, human flow from one location to another may not necessarily lead to physical interactions that could communicate the disease, especially in public places, where preventative measures are enforced more strictly.  
+
+<|ref|>text<|/ref|><|det|>[[112, 355, 884, 724]]<|/det|>
+In this paper, we compare the predictive power of Facebook's social connectedness index, as an example of social media proxy, with cell phone- derived human mobility data, as an example of human flow proxy, in forecasting county- level new cases of COVID- 19 in the conterminous US over multiple prediction horizons. County- level prediction is more challenging than state- level prediction<sup>13–15</sup>, yet it serves as the highest spatial resolution for national models in the U.S., since cases are aggregated and reported at the county level. Longer term County- level predictions are also essential for policy making and resource allocation.  
+
+<|ref|>text<|/ref|><|det|>[[113, 751, 860, 903]]<|/det|>
+The unique characteristics of COVID- 19, including its presymptomatic and asymptomatic contagiousness, rapid spread, along with variations in regional response policies, such as inconsistent and sporadic testing and contact tracing, make forecasting the spatial patterns of this disease challenging. Researchers
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 877, 372]]<|/det|>
+have used a variety of methods including time- series autoregressive models<sup>16–18</sup>, machine learning techniques<sup>19–21</sup>, epidemiologic models such as SIR model and its variants<sup>22,23</sup>, and combinations of these methods<sup>24</sup> for forecasting COVID- 19. We experiment with five different machine learning- based spatiotemporal autoregressive algorithms to perform county- level predictions, and use the best algorithm, i.e. the one with the lowest average prediction RMSE and MAE, to compare between Facebook- and cell phone- derived features.  
+
+<|ref|>text<|/ref|><|det|>[[113, 399, 883, 595]]<|/det|>
+We compare our best model predictions against one of the most prominent collective efforts in forecasting COVID- 19 in the U.S., namely, the Ensemble model developed by the "COVID- 19 Forecast Hub" team<sup>25</sup> which is used by the Centers for Disease Control and Prevention (CDC) to report predictions of new cases and deaths in U.S. counties in one- to four- weeks ahead horizons<sup>25,26</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[113, 622, 861, 862]]<|/det|>
+Our specific contributions are as follows: (a) designing inter- county and intra- county features for spatiotemporal autoregressive machine learning of county- level new cases, (b) comparing the performance of social media connectedness (derived from Facebook) and human flow connectedness (derived from SafeGraph's cell phone data) by incorporating inter- county spatial lags for predicting county- level new COVID- 19 cases, and (c) improving the long- term
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 824, 155]]<|/det|>
+prediction of county- level new cases of COVID- 19 in the coterminous U.S. in comparison to a baseline Ensemble model, using an end- to- end model.  
+
+<|ref|>sub_title<|/ref|><|det|>[[144, 203, 245, 222]]<|/det|>
+## 2. Results  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 271, 305, 291]]<|/det|>
+## Algorithm Selection  
+
+<|ref|>text<|/ref|><|det|>[[112, 317, 876, 820]]<|/det|>
+Five different machine learning algorithms were trained and tuned using each set of features named in the next section (details in Section 4), and tested over the last 5 weeks of our dataset (same dates as Tables 2 & 5), by holding out one week at a time for testing. Table 1 reports the average performance for each algorithm. Extreme Gradient Boosting (XGB) performed better on unseen data compared with other tree- based ensemble algorithms and the neural networks, including Feed Forward Neural Network (FFNN) and Long Short- Term Memory (LSTM) network (Table 1). Therefore, we used XGB for developing short- term and long- term prediction models. The RMSE and MAE values reported in Table 1 are for the natural log values of [new cases per 10k population + 1], which we used as a transformed target variable in the models, given the skewed distribution of new cases (or new cases per 10k) in counties.
+
+<--- Page Split --->
+<|ref|>table<|/ref|><|det|>[[117, 133, 879, 370]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[115, 90, 840, 125]]<|/det|>
+Table 1. Performance comparison of machine learning regressors. The best performance in each category is bolded.   
+
+<table><tr><td>Model</td><td>RMSE Train</td><td>MAE Train</td><td>RMSE Test</td><td>MAE Test</td></tr><tr><td>Random Forest</td><td>0.486</td><td>0.359</td><td>0.511</td><td>0.38</td></tr><tr><td>Stochastic Gradient Boosting</td><td>0.438</td><td>0.313</td><td>0.494</td><td>0.348</td></tr><tr><td>Extreme Gradient Boosting</td><td>0.441</td><td>0.316</td><td>0.47</td><td>0.330</td></tr><tr><td>Feed Forward Neural Network</td><td>0.524</td><td>0.391</td><td>0.566</td><td>0.438</td></tr><tr><td>Long Short-Term Memory</td><td>1.179</td><td>0.964</td><td>1.209</td><td>1.007</td></tr></table>  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 395, 660, 418]]<|/det|>
+## Comparing social media- and cell phone-derived features  
+
+<|ref|>text<|/ref|><|det|>[[112, 445, 879, 814]]<|/det|>
+To compare the relative strength of Facebook- derived movement and connectedness against SafeGraph- derived human mobility flows, as proxies for physical human interactions, we designed a set of intra- county and inter- county interaction features using each proxy, and incorporated each set of features separately to develop spatiotemporally lagged autoregressive prediction models of new cases of COVID- 19 (i.e. target variable). We then compared the predictions of these models against each other as well as a base model (not to be confused with the baseline model for final evaluations), all of which were trained using the XGB algorithm.  
+
+<|ref|>text<|/ref|><|det|>[[115, 842, 876, 907]]<|/det|>
+Our base model incorporates a series of socioeconomic, demographic and temperature variables, as well as temporal lags of the target variable in the same
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 876, 503]]<|/det|>
+county only, thus, we call it Temporal XGB (TGXB), whereas the SpatioTemporal XGB (STXGB) models, in addition to temporal lags, also incorporate intra- county movement features and spatiotemporal lags of the target variable weighted by the inter- county connectedness strength. Specifically, the spatial lags in STXGB are calculated by multiplying the target variable (natural log of weekly new cases per 10k population + 1) in "connected counties" by either (a) inter- county Facebook Social Media Connectedness Index, (in the STXGB- FB model), or (b) inter- county Flow Connectedness Index derived from SafeGraph's cell- phone movement data, forming STXGB- SG and STXGB- SGR models (described in detail in Section 4).  
+
+<|ref|>text<|/ref|><|det|>[[112, 529, 880, 896]]<|/det|>
+Table 2 and Fig. 1 present the error values of predicted new cases and new cases per 10k population in the one- week prediction horizon using the TXGB and STXGB models. The incorporation of spatiotemporal lags using county connectedness indices (in STXGB) was advantageous across the board, compared to the temporal lags only (TXGB). All variants of STXGB (- FB, - SG, and - SGR) achieved lower errors compared to TXGB. Furthermore, STXGB- FB, which uses the Facebook- derived features, outperformed all other models in average RMSEs and MAEs as well as on all forecast dates, except the fifth date when the STXGB- SG model generated slightly lower errors.
+
+<--- Page Split --->
+<|ref|>table_caption<|/ref|><|det|>[[113, 141, 870, 191]]<|/det|>
+Table 2. RMSE and MAE of county-level predicted weekly new cases and new cases per 10k population. Lowest values of each error metric are highlighted. Average values across forecasting dates for each model is bold faced. 
+
+<|ref|>table<|/ref|><|det|>[[113, 202, 832, 864]]<|/det|>
+<table><tr><td></td><td>Model</td><td>Forecast Date</td><td>RMSE New Case Prediction</td><td>MAE New Case Prediction</td><td>RMSE New Case/10k Prediction</td><td>MAE New Case/10k Prediction</td></tr><tr><td rowspan="5">including temporal lags</td><td rowspan="5">Base Model (TXGB)</td><td>10/24/2020</td><td>136.255</td><td>30.894</td><td>16.084</td><td>6.134</td></tr><tr><td>10/31/2020</td><td>192.993</td><td>50.91</td><td>22.319</td><td>11.176</td></tr><tr><td>11/07/2020</td><td>203.678</td><td>70.689</td><td>22.899</td><td>12.907</td></tr><tr><td>11/14/2020</td><td>237.113</td><td>80.1</td><td>26.45</td><td>15.104</td></tr><tr><td>11/21/2020</td><td>166.855</td><td>50.684</td><td>16.384</td><td>9.611</td></tr><tr><td></td><td></td><td>Average</td><td>187.379</td><td>56.655</td><td>20.827</td><td>10.986</td></tr><tr><td rowspan="5">including spatial-temporal lags</td><td rowspan="5">STXGB with Facebook-derived features (STXGB-FB)</td><td>10/24/2020</td><td>116.312</td><td>25.909</td><td>15.083</td><td>5.708</td></tr><tr><td>10/31/2020</td><td>172.582</td><td>46.398</td><td>21.938</td><td>10.817</td></tr><tr><td>11/07/2020</td><td>169.602</td><td>54.613</td><td>20.925</td><td>11.072</td></tr><tr><td>11/14/2020</td><td>185.391</td><td>62.243</td><td>23.263</td><td>12.477</td></tr><tr><td>11/21/2020</td><td>142.312</td><td>48.625</td><td>16.297</td><td>9.373</td></tr><tr><td></td><td></td><td>Average</td><td>157.24</td><td>47.557</td><td>19.501</td><td>9.89</td></tr><tr><td rowspan="5">including spatial-temporal lags</td><td rowspan="5">STXGB with SafeGraph-derived features (STXGB-SG)</td><td>10/24/2020</td><td>120.049</td><td>27.312</td><td>15.101</td><td>5.785</td></tr><tr><td>10/31/2020</td><td>195.03</td><td>50.487</td><td>22.267</td><td>11.171</td></tr><tr><td>11/07/2020</td><td>193.263</td><td>62.506</td><td>21.156</td><td>11.421</td></tr><tr><td>11/14/2020</td><td>203.675</td><td>68.962</td><td>24.423</td><td>13.371</td></tr><tr><td>11/21/2020</td><td>140.748</td><td>48.383</td><td>16.683</td><td>9.596</td></tr><tr><td></td><td></td><td>Average</td><td>170.553</td><td>51.53</td><td>19.926</td><td>10.269</td></tr><tr><td rowspan="5">including spatial-temporal lags</td><td rowspan="5">STXGB with SafeGraph-derived features-rich (STXGB-SGR)</td><td>10/24/2020</td><td>122.482</td><td>27.952</td><td>15.362</td><td>5.739</td></tr><tr><td>10/31/2020</td><td>207.411</td><td>54.531</td><td>22.626</td><td>11.541</td></tr><tr><td>11/07/2020</td><td>178.934</td><td>60.061</td><td>20.998</td><td>11.324</td></tr><tr><td>11/14/2020</td><td>186.997</td><td>64.745</td><td>24.079</td><td>13.032</td></tr><tr><td>11/21/2020</td><td>141.356</td><td>47.491</td><td>16.758</td><td>9.539</td></tr><tr><td></td><td></td><td>Average</td><td>167.436</td><td>50.956</td><td>19.965</td><td>10.235</td></tr></table>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[137, 90, 901, 620]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 628, 875, 756]]<|/det|>
+<center>Figure 1. Prediction errors of the SpatioTemporal eXtreme Gradient Boosting (STXGB) models in one-week prediction horizon against a temporal autoregressive model without spatial lags (TXGB). Error comparison of STXGB models with different feature sets in predicting weekly new cases (top row) and new cases per 10k population (bottom row) for one-week ahead horizon. STXGB-FB, which incorporates Facebook-derived features, including spatial lags based on Social Connectedness Index, outperforms other models. Left column: prediction RMSE. Right column: prediction MAE. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[114, 91, 812, 112]]<|/det|>
+## Long-term predictions and evaluation against the COVID-19 Forecast Hub  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 135, 210, 153]]<|/det|>
+## Ensemble  
+
+<|ref|>text<|/ref|><|det|>[[112, 182, 876, 640]]<|/det|>
+We compared the predictions of our best model, STXGB- FB, against the predictions of the COVID- 19 Forecast Hub Ensemble of 32 models (used by the CDC in reporting forecasts of new cases<sup>26</sup>) over one-, two-, three-, and four- week horizons. We trained and tuned STXGB- FB for each prediction horizon separately. We then used the reported new cases by Johns Hopkins University Center for Systems Science and Engineering (JHU CSSE)<sup>27</sup> as ground- truth to calculate RMSE and MAE of each prediction set, shown in Fig. 2 and Table 3, over varying prediction horizons across the five forecast dates. Our model considerably improves RMSEs and MAEs compared with the Ensemble model in the two- week, three- week, and four- week ahead prediction horizons, with an average 58% reduction in RMSEs and 61% reduction in MAEs.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[168, 95, 933, 358]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[133, 366, 875, 418]]<|/det|>
+<center>Figure 2. Long-term prediction error comparison. Prediction errors of the STXGB-FB model (dashed line) and the Ensemble baseline (solid line) over four prediction horizons on five forecasting dates. a) Prediction RMSE and b) Prediction MAE. </center>  
+
+<|ref|>text<|/ref|><|det|>[[112, 430, 881, 844]]<|/det|>
+In the one- week prediction horizons, the Ensemble model outperformed our STXGB- FB model on the first three forecasting dates (Oct 24, Oct. 31, Nov. 7), but performed similarly on the Nov. 14 forecasting date. Nevertheless, the differences in the one- week horizon predictions are relatively small. STXGB- FB outperformed the Ensemble prediction on Nov. 21 across all prediction horizons. This is noteworthy since the prediction horizons (one- to four- week) on Nov. 21 overlap the post- Thanksgiving holidays in the US, which caused a surge in the number of cases<sup>28</sup>. In summary, out of the 20 predictions performed, the STXGB- FB model outperformed the Ensemble model in 17 of them, including in all longer than one- week prediction horizons (Table 3).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 91, 876, 412]]<|/det|>
+To investigate the potential of SafeGraph cell phone-derived features in long-term predictions, we generated one-, two-, three-, and four-week forecasts using the STXGB-SG model as well. This model does not perform as well as STXGB-FB, pointing to the superiority of Facebook-derived features in our models consistent with the one-week predictions (Supplementary Table 4, Supplementary Information). However, while STXGB-SG generates larger errors compared to STXGB-FB, it still outperforms the Ensemble model in long-term prediction horizons.  
+
+<|ref|>table<|/ref|><|det|>[[113, 480, 912, 825]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[115, 435, 839, 470]]<|/det|>
+Table 3. Comparison of the prediction errors generated by the COvID-19 Forecast Hub Ensemble model and our STXGB-FB model in 1-to 4-week prediction horizons.   
+
+<table><tr><td rowspan="2">Forecast Date</td><td rowspan="2">Model</td><td colspan="4">Prediction RMSE in Prediction Horizon</td></tr><tr><td>1 wk ahead</td><td>2 wk ahead</td><td>3 wk ahead</td><td>4 wk ahead</td></tr><tr><td rowspan="3">10/24/2020</td><td>Ensemble</td><td>126.05</td><td>894.41</td><td>1915.31</td><td>3087.32</td></tr><tr><td>STXGB-FB</td><td>164.80</td><td>489.81</td><td>832.36</td><td>1157.87</td></tr><tr><td>Pct. Improvement</td><td>-30.74%</td><td>45.24%</td><td>56.54%</td><td>62.50%</td></tr><tr><td rowspan="3">10/31/2020</td><td>Ensemble</td><td>205.12</td><td>1172.21</td><td>2342.40</td><td>3461.53</td></tr><tr><td>STXGB-FB</td><td>247.67</td><td>554.25</td><td>892.22</td><td>1376.14</td></tr><tr><td>Pct. Improvement</td><td>-20.74%</td><td>52.72%</td><td>61.91%</td><td>60.24%</td></tr><tr><td rowspan="3">11/07/2020</td><td>Ensemble</td><td>223.96</td><td>1311.84</td><td>2424.17</td><td>3864.08</td></tr><tr><td>STXGB-FB</td><td>268.78</td><td>649.08</td><td>906.63</td><td>1380.22</td></tr><tr><td>Pct. Improvement</td><td>-20.01%</td><td>50.52%</td><td>62.60%</td><td>64.28%</td></tr><tr><td rowspan="3">11/14/2020</td><td>Ensemble</td><td>275.24</td><td>1148.63</td><td>2558.68</td><td>4372.37</td></tr><tr><td>STXGB-FB</td><td>274.35</td><td>468.53</td><td>848.68</td><td>1450.59</td></tr><tr><td>Pct. Improvement</td><td>0.32%</td><td>59.21%</td><td>66.83%</td><td>66.82%</td></tr><tr><td rowspan="3">11/21/2020</td><td>Ensemble</td><td>301.87</td><td>1292.31</td><td>3086.82</td><td>5390.16</td></tr><tr><td>STXGB-FB</td><td>210.64</td><td>592.83</td><td>1457.29</td><td>2646.68</td></tr><tr><td>Pct. Improvement</td><td>30.22%</td><td>54.13%</td><td>52.79%</td><td>50.90%</td></tr></table>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 387, 112]]<|/det|>
+## Spatial distribution of errors  
+
+<|ref|>text<|/ref|><|det|>[[112, 140, 876, 858]]<|/det|>
+Our STXGB- FB performed worse than the baseline Ensemble model on the Oct.24, Oct. 31 and Nov. 7 forecasting dates over the one- week horizon, but considerably outperformed the Ensemble over longer- term horizons on the same forecasting dates. Further, STXGB- FB outperformed the Ensemble baseline on Nov. 14 and Nov. 21 across all prediction horizons. To find potential explanations for this inconsistency, we inspected the spatial patterns of errors. Figure 3 illustrates maps of confirmed new cases per 10k population along with prediction errors per 10k population generated by the STXGB- FB model for two forecasting dates of Oct. 31 and Nov. 7. The purple- shaded counties in the error maps are those with model underestimation of new cases, and the brown shades indicate overestimations of observed values. As can be seen in this figure, the majority of counties with high prediction errors (per 10K) are located in the rural Midwest with relatively high numbers of cases per 10k population during the November surge, albeit these are counties with fewer total cases compared to more populated, urban ones. It is worth noting that we use normalized (by 10K population) maps in Fig. 3, since choropleth maps would be biased by patterns of population distribution otherwise.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 90, 870, 459]]<|/det|>
+Figure 3 also demonstrates clusters of apparent underestimations in Georgia and Texas on the Oct. 31 forecasting date, followed by apparent overestimations in the same areas for the week after. The opposite pattern is the case for Kentucky. In the case of Georgia, the high- error clusters can almost perfectly delineate the boundary of the state. This discrepancy could be a result of lags or different policies in testing and reporting COVID- 19 cases. These potential short- term lags in reporting by some states may explain why our model performs considerably better in the longer- term prediction horizons but underperforms in the one- week horizon on Oct. 31 and Nov. 7.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 95, 965, 544]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 550, 864, 655]]<|/det|>
+<center>Figure 3. Map of COVID-19 cases per 10k population and errors in predicting them. (a) number of confirmed new cases per 10k population over the week ahead of forecasting date Oct. 31, 2020 (b) prediction errors for the same forecasting date(c) number of new cases over the week ahead of Nov. 7, 2020 forecasting date (d) prediction errors for the same forecasting date. The pattern of errors in Georgia, and Texas, and Kentucky flip from Oct. 31 to Nov. 7, indicating potential lags in testing and reporting. </center>  
+
+<|ref|>text<|/ref|><|det|>[[112, 666, 872, 860]]<|/det|>
+The majority of counties in the U.S are rural, which are also the ones with fewer medical resources, and where social media data or cell- phone mobility data, which underlie our models, might be less representative \(^{29 - 31}\) . To investigate our models' performance in rural- majority counties compared to the Ensemble baseline, we categorized the counties into urban- and rural- majority by
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 852, 156]]<|/det|>
+calculating an urbanization index for each county (Supplementary Information, Section B). 2391 counties (\~77%) were identified as rural and 712 as urban.  
+
+<|ref|>text<|/ref|><|det|>[[112, 180, 880, 685]]<|/det|>
+We then calculated the prediction errors of the number of cases and the number of cases per 10k population for the Ensemble and STXGB- FB models in each category across four prediction horizons for the Nov. 7 forecasting date (Fig. 4). Both models generate considerably lower median errors and narrower interquartile error ranges in rural counties when predicting the total number of new cases (not normalized by population), which could be attributed to the overall higher prevalence and higher variance of COVID- 19 cases in urban counties in our prediction horizons. However, the opposite is the case when predicting the number of weekly new cases per 10k population; both models have wider interquartile ranges in rural counties across all prediction horizons. This could be due to the overall higher prevalence of COVID- 19 per population in rural counties during the selected prediction horizons.  
+
+<|ref|>text<|/ref|><|det|>[[113, 710, 868, 904]]<|/det|>
+As evident in Fig. 4, STXGB- FB has lower prediction errors with a narrower IQR in both urban and rural counties compared to the Ensemble model, across all prediction horizons except for the shortest one (one- week), which may be attributed to temporal fluctuations and policy variations in testing and case reporting as discussed above. The overall superior performance of STXGB- FB is
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 881, 373]]<|/det|>
+observed when predicting both the total number of new cases and new cases per 10k population, affirming the higher robustness of our model. Furthermore, the difference between the median prediction errors of STXGB- FB in urban and rural counties, when predicting the number of cases per 10k population, is smaller compared to the Ensemble model. This points to the more consistent performance of STXGB- FB in majority- rural counties, even though Facebook might not be as representative in these areas<sup>29</sup>.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[163, 88, 833, 860]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[120, 864, 875, 907]]<|/det|>
+<center>Figure 4. Prediction errors in urban vs. rural counties. Prediction errors of the number of new cases (left column) and new cases per 10k population (right column) in rural and urban counties on the Nov. 7 forecast date across four prediction horizons. The higher and lower 3% of counties are trimmed from the plot view. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[144, 91, 275, 111]]<|/det|>
+## 3. Discussion  
+
+<|ref|>text<|/ref|><|det|>[[112, 140, 877, 423]]<|/det|>
+We demonstrated that incorporating (1) spatiotemporal lags using inter- county indices of connectedness and (2) intra- county measurements of movement can improve the performance of high resolution COVID- 19 predictive models, especially over long- term horizons. Short- term and long- term predictions of COVID- 19 cases help the federal and local governments make informed decisions such as imposing or relaxing business restrictions or planning resource allocation in response to the forecasted trends of COVID- 19<sup>32,33</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[111, 450, 876, 905]]<|/det|>
+Our results showed that using Facebook- derived features implemented in our SpatioTemporal Autoregressive eXtreme Gradient Boosted (STXGB) model generate lower prediction errors across all prediction horizons compared to SafeGraph cell phone- derived features within the same architecture (Supplementary Table 5). Notably, however, to maintain compatibility, we used a formulation similar to Facebook's Social Connectedness Index when creating a corresponding index from SafeGraph data (which we call Flow Connectedness Index, refer to Section 4). This might have had adverse effects on the predictive power of the cell- phone- derived features. Nevertheless, the resulting model performs considerably better than the Ensemble baseline in long- term predictions (Supplementary Table 4). We will investigate alternative designs of inter- county
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 865, 155]]<|/det|>
+connectedness metrics from SafeGraph mobility data in the future to ensure the utilization of the full potential of this dataset.  
+
+<|ref|>text<|/ref|><|det|>[[112, 181, 877, 640]]<|/det|>
+The superior performance of Facebook- derived features means that stronger county- level social (media) connections could lead to a higher chance of (unsafe) human interaction (e.g., house parties) and thus, COVID spread, compared to human flow from one location to another. The findings of this paper also suggest that Facebook's Social Connectedness Index can be used for successful predictive modeling of COVID- 19 in data- poor countries without cell- phone movement datasets, assuming that the Facebook usage in those countries is of a comparable size and representativeness to the U.S<sup>34</sup>. With more than 2.5 billion active users globally, Facebook provides social connectedness for many countries. Conversely, human mobility data, to the best of our knowledge, is available in far fewer numbers of countries.  
+
+<|ref|>text<|/ref|><|det|>[[113, 666, 868, 904]]<|/det|>
+The STXGB- FB and STXGB- SG models significantly outperformed the Ensemble model currently used by the CDC in predicting county- level new cases of COVID- 19 in two-, three-, and four- weeks prediction horizons, with inconsistent comparisons in the one- week horizon. Our error maps suggest that this inconsistency might be partly due to inconsistent and delayed testing and reporting by some states.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 882, 330]]<|/det|>
+The superiority of STXGB over purely temporal models (such as our TXGB) points to the importance of incorporating both within- unit (e.g., intra- county) and between- unit (e.g., inter- county) interactions when predicting a highly contagious disease such as COVID- 19. Predictive models that focus on within state boundaries are likely to underperform, because after all, the disease does spread across geographic unit borderlines.  
+
+<|ref|>text<|/ref|><|det|>[[112, 355, 880, 812]]<|/det|>
+As evident in table 1, the XGB and SGB methods performed better than other machine learning- based algorithms. A potential reason for the lower performance of FFNN and LSTM is the relatively small size of the training data (34 weeks of observation at most). Neural networks' main advantage is in their ability to learn features from data<sup>35</sup>; however, they require higher amounts of training data compared to tree- based models for optimizing the model's parameters. Our results are a testament to the advantages of high- performance tree- based ensemble algorithms such as XGB with more limited training data, especially if features are well- engineered. Our spatiotemporal lag features provide a template for such features to improve machine learning- based predictive modeling of infectious diseases.  
+
+<|ref|>text<|/ref|><|det|>[[114, 838, 875, 904]]<|/det|>
+It is also worth noting that between XGB and SGB, SGB generated lower average training errors when using the base and SafeGraph- derived features, but
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 853, 415]]<|/det|>
+XGB outperformed SGB in testing RMSE and MAE across all 4 models (Supplementary Table 2). This is somewhat expected, as XGB uses the second- order derivatives of the loss function for optimization, and more importantly, a regularized model formalization to control over- fitting, which is otherwise a disadvantage with regression trees<sup>36</sup>. This regularized model results in better performance on unseen data. To ensure consistency, we ran all regression methods 10 times; and XGB had lower testing errors compared to all other regression methods in all 10 runs.  
+
+<|ref|>sub_title<|/ref|><|det|>[[143, 461, 262, 481]]<|/det|>
+## 4. Methods  
+
+<|ref|>text<|/ref|><|det|>[[112, 510, 886, 792]]<|/det|>
+This section outlines the details of feature engineering, algorithm selection, and implementation of spatiotemporal autoregressive machine learning models for predicting new cases of COVID- 19 in the conterminous United States. We describe our experimental setup for comparing the predictive power of Facebook- derived features and SafeGraph's cell phone- derived mobility features (as proxies for human physical interaction) for this purpose, and the evaluation of our models against the COVID- 19 Forecast Hub Ensemble baseline.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 92, 250, 111]]<|/det|>
+## Base Features  
+
+<|ref|>text<|/ref|><|det|>[[112, 140, 876, 510]]<|/det|>
+We engineered features for machine learning that can be categorized into five groups: (A) a set of county- level demographic and socioeconomic features, (B) minimum and maximum temperatures of inhabited areas in counties, (C) temporally- lagged (i) weekly average and (ii) weekly change of incident rates (COVID- 19 cases per 10k population) in each county, (D) Facebook- derived features of (i) intra- county movement measurements and (ii) exposure to COVID- 19 through inter- county connectedness, and (E) SafeGraph- derived features of (i) intra- county movement measurements and (ii) exposure to COVID- 19 through inter- county connectedness.  
+
+<|ref|>text<|/ref|><|det|>[[113, 536, 853, 732]]<|/det|>
+Socioeconomic, demographic, and climatic variables have been shown to be correlated with the spread of COVID- 19<sup>21,37- 40</sup>, therefore we include category A and B features in all of our models to control for these factors. Section A of the supplementary information outlines the detailed methodology for generating features in these categories.  
+
+<|ref|>text<|/ref|><|det|>[[113, 759, 880, 867]]<|/det|>
+Features in category C indirectly capture population compartments of the susceptible- infected- recovered (SIR) epidemiological models<sup>41,42</sup>. We defined the COVID- 19 incidence rate of a county as its number of cases per 10,000 population
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 882, 502]]<|/det|>
+and included a four- week lagged (t- 4) weekly average of total incidence rates in each county as a feature (category C.i) to capture the Susceptible and Recovered compartments. If the feature value is small, many individuals in the unit have not yet contracted the disease; and therefore, are still susceptible. If the feature value is sufficiently large (level of sufficiency is learned by the model), the compartment is approaching higher levels of immunity as a whole. We use machine learning algorithms that are capable of learning such non- linear relationships. It is worth noting that vaccination rates can similarly be incorporated as a feature. However, inoculations in the U.S. started after our study period, and thus, not included here.  
+
+<|ref|>text<|/ref|><|det|>[[112, 529, 880, 896]]<|/det|>
+To account for the latency associated with the effects of temperature, new and historical incidences, and human interaction on the spread of COVID- 19, we generated four temporal weekly lags of features in categories B- E (we assume that the features in category A are static during our study period). Notably, for category C, we also included (natural log- transformed values of) change in incidence rate( \(\ln (\Delta \text{incidence rate} + 1)\) ), i.e., the subtraction of observed start- of- week incident rate from end- of- week incidence rate during the four weekly temporal- lags (t- 1,...,t- 4), as another set of features in all of our models (category C.ii) (more details in the supplementary information). These features conceptually capture
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 852, 285]]<|/det|>
+the Infected compartment in the SIR model, i.e., the currently infected populations in the spatial unit. Our autoregressive modeling with multiple temporal lags allows the models to learn the rate of spread in a unit, as well as the varying incubation periods of the disease in relation to the change in temperature and demographic features<sup>43,44</sup>.  
+
+<|ref|>text<|/ref|><|det|>[[113, 313, 857, 509]]<|/det|>
+Features in categories D and E also model the SIR population compartments, but in connected counties (through either social connectedness or flow connectedness). Features in category D represent the social media proxy (of physical human interactions), whereas features in category E represent the cell phone- derived human mobility flow proxy.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 554, 788, 576]]<|/det|>
+## Conceptualization of models with social media and cell phone features  
+
+<|ref|>text<|/ref|><|det|>[[113, 604, 876, 843]]<|/det|>
+We evaluate the predictive power of the Facebook- derived features (category D) and the SafeGraph- derived features (category E) against a base model by developing four different model setups (not to be confused with the final evaluations against the baseline Ensemble model). Here, we provide an outline of these models, with more details on the specific algorithms and features mentioned in Table 4 in the following sections.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 91, 882, 225]]<|/det|>
+week-ahead prediction horizon on forecast date \(d\) is the number of new cases per 10k from the forecast date through the end of prediction horizon \(t\) in each county, i.e., \(\Delta\) (incidence rate) \(t,d\) . The target for the two week-horizon prediction is \(\Delta\) (incidence rate) \(t+1,d\) , three week-horizon is \(\Delta\) (incidence rate) \(t+2,d\) , and four week horizon is \(\Delta\) (incidence rate) \(t+3,d\) . \(\ln (t)\) in the table indicates natural logarithm, Mean () indicates weekly average, \(\Delta\) indicates weekly change, i.e., difference (calculated by subtracting the value of the feature at the beginning of the week from its value at the end of the week) and Slope indicates the slope of a fitted linear regression model to the standardized daily measures of metric value as the dependent variable and standardized day of week as the independent variable.  
+
+<|ref|>table<|/ref|><|det|>[[113, 231, 941, 911]]<|/det|>
+
+<table><tr><td>Category</td><td>Model(s)</td><td>Variables</td><td>Temporal Lag</td></tr><tr><td>A- socioeconomic and demographic</td><td>All of the models</td><td>population density<br>percentage of male population<br>percentage of African American population<br>percentage of Hispanic population<br>percentage of Native American population<br>percentage of the rural population<br>percentage of the population with a college degree<br>median household income<br>percentage of the population who voted republican in 2016 presidential election</td><td>None (constant)</td></tr><tr><td>B- Temperature</td><td>All of the models</td><td>mean(daily minimum temperature) <br>mean(daily maximum temperature) <br>Ln (Δ incidence rate +1) <br>Ln (mean (incidence rate) +1)</td><td>t-1, t-2, t-3, t-4 <br>t-1, t-2, t-3, t-4 <br>t4</td></tr><tr><td>C- COVID-19 incidence rate</td><td>All of the models</td><td>Δ SPC <br>mean (SPC) <br>mean (Stay Put) <br>and slope (Stay Put) <br>t</td><td>t-1, t-2, t-3, t-4 <br>t-4 <br>t-1, t-2, t-3, t-4</td></tr></table>
+
+<--- Page Split --->
+<|ref|>table<|/ref|><|det|>[[114, 87, 944, 400]]<|/det|>
+
+<table><tr><td></td><td></td><td>mean (Change in Movement)t and slope (Change in Movement)t</td><td>t-1, t-2, t-3, t-4</td></tr><tr><td rowspan="4">E- SafeGraph</td><td rowspan="2">-SG and -SGR models</td><td>Δ FPC t<br>mean (FPC)t<br>mean (% completely_home_device_count)t and slope(% completely_home_device_count)t</td><td>t-1, t-2, t-3, t-4</td></tr><tr><td>mean(baselined distance_traveled_from_home)t<br>slope(baselined distance_traveled_from_home)t</td><td>t-1, t-2, t-3, t-4</td></tr><tr><td rowspan="2">-SGR model</td><td>mean (baselined median_home_dwell_time)t and slope(baselined median_home_dwell_time)t<br>mean (baselined % full_time_work_behavior_devices)t and slope(baselined % full_time_work_behavior_devices)t</td><td>t-1, t-2, t-3, t-4</td></tr><tr><td></td><td></td></tr></table>
+
+<|ref|>text<|/ref|><|det|>[[114, 452, 880, 686]]<|/det|>
+The first model (base model) only includes base features: socioeconomic features (category A), four temporally-lagged weekly temperature features (category B), and four temporally-lagged weekly change in incidence rates in each county, as well as weekly average of incidence rates during the fourth lagged week (category C). Therefore, the base model only incorporates temporal lags of the features and the target variable in predicting new cases of COVID-19.
+
+<|ref|>text<|/ref|><|det|>[[114, 719, 868, 865]]<|/det|>
+The second model, which we identify by the “-FB” suffix (to note the inclusion of Facebook features), includes the base features as well as category D features,i.e. Facebook-derived intra-county movement features and inter-county spatiotemporal lags of the target variable, i.e. exposure to COVID-19 through
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 881, 416]]<|/det|>
+social connectedness (Social Proximity to Cases), across four temporal lags. The third model, which we identify by the "- SG" suffix, is conceptually similar to the - FB model, but with features derived from SafeGraph cell phone mobility data instead of Facebook data. Specifically, the - SG models include the base features in addition to inter- county spatiotemporal lags of the target variable, i.e. exposure to COVID- 19 through human flow connectedness (which we call Flow Proximity to Cases), and a subset of category E SafeGraph- derived intra- county movement features, across four temporal lags.  
+
+<|ref|>text<|/ref|><|det|>[[112, 442, 880, 681]]<|/det|>
+To explore the full potential of the movement features provided by the SafeGraph Social Distancing Metrics (SDM) dataset, we developed a fourth model, in which two additional mobility- related measurements provided in the SDM dataset (that are least correlated with other features in Category E) are added to the - SG model. This model thus includes categories A- C and all features in category E, and is identified by the "- SGR" suffix.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 425, 111]]<|/det|>
+## Features derived from Facebook  
+
+<|ref|>sub_title<|/ref|><|det|>[[137, 159, 468, 180]]<|/det|>
+### i. Intra-county movement features  
+
+<|ref|>text<|/ref|><|det|>[[112, 202, 883, 753]]<|/det|>
+Facebook publishes the Movement Range dataset for 14 countries<sup>45</sup> and it includes two metrics called "Change in Movement" and "Stay Put", each providing a different perspective on movement trends as measured by mobile devices carrying the Facebook app. The Change in Movement metric for each county is a measure of relative change in aggregated movement compared to the baseline of February \(2^{\text{nd}}\) to February \(29^{\text{th}}\) 2020 (excluding the February \(17^{\text{th}}\) 2020 President Day holiday in the US)<sup>45</sup>. The Stay Put metric measures "the fraction of the population that have stayed within a small area during an entire day"<sup>45</sup>. We used four temporal lags of weekly averages and slopes of each metric as a feature in our - FB model. We calculated the slopes by fitting a linear regression model to the metric value as the dependent variable and day of week as the independent variable, both transformed to standard scale N(0,1). The slope feature characterizes the overall trend in a week, as compared to the baseline period.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[137, 91, 603, 113]]<|/det|>
+## ii. Inter-county features and spatial lag modeling  
+
+<|ref|>text<|/ref|><|det|>[[112, 140, 884, 379]]<|/det|>
+The intra- county features capture the intrinsic movement- related characteristics of a county and ignore its interactions (i.e. spatial lags) with the counties to which it is connected. Therefore, we calculated inter- county metrics of connectivity as a basis for incorporating spatiotemporal lags in our models. Notably, the connectedness in this context transcends spatial connectedness in the form of mere physical adjacency.  
+
+<|ref|>text<|/ref|><|det|>[[113, 405, 861, 560]]<|/det|>
+Social connectedness Index (SCI), another dataset published by Facebook, is a measure of the intensity of connectedness between administrative units, calculated from Facebook friendship data. Social connectedness between two counties \(i\) and \(j\) is defined as<sup>46</sup>:
+
+<--- Page Split --->
+<|ref|>equation<|/ref|><|det|>[[165, 94, 855, 142]]<|/det|>
+\[Social connectedness (SC)_{i,j} = \frac{FB Connections_{i,j}}{FB Users_{i} * FB Users_{j}} \quad (1)\]  
+
+<|ref|>text<|/ref|><|det|>[[112, 161, 867, 576]]<|/det|>
+where FB Connections \(_{i,j}\) is the number of friendships between Facebook users who live in county \(i\) and those who live in county \(j;\) while FB Users \(i\) and FB Users \(j\) are the total number of active Facebook users in counties \(i j\) and, respectively. Social Connectedness is scaled to a range between 1 and 1,000,000,000 and rounded to the nearest integer to generate SCI, as published by Facebook47. Therefore, if the SCI value between a pair of counties is twice as large as another pair, it means the users in the first county- pair are almost twice as likely to be friends on Facebook than the second county- pair46. We used the latest version of the SCI dataset (at the time of our analyses), which was released in August 202047.  
+
+<|ref|>text<|/ref|><|det|>[[112, 604, 872, 886]]<|/det|>
+While SCI provides a measure of connectivity, our goal is to capture the spatiotemporal lags of COVID- 19 cases in county \(i\) , i.e., the number of recent COVID- 19 cases in other counties connected to county \(i\) . Using SCI, Kuchler et al. \(^{6}\) created a new metric, called Social Proximity to Cases (SPC) for each county, which is a measure of the level of exposure to COVID- 19 cases in connected counties through social connectedness. We use a slight variation of SPC, defined as follows for county \(i\) at time \(t\) :
+
+<--- Page Split --->
+<|ref|>equation<|/ref|><|det|>[[179, 94, 874, 171]]<|/det|>
+\[\begin{array}{l}{{S o c i a l P r o x i m i t y t o C a s e s(S P C)_{i,t}}}\\ {{=\sum_{j}C a s e s P e r10k_{j,t}\times\frac{S o c i a l C o n n e c t e d n e s s_{i,j}}{\sum_{h}S o c i a l C o n n e c t e d n e s s_{h}}}}\end{array} \quad (2)\]  
+
+<|ref|>text<|/ref|><|det|>[[112, 181, 882, 771]]<|/det|>
+where Cases Per \(10k_{j,t}\) is the number of COVID- 19 cases per 10k population (i.e., incidence rate) in county \(j\) as of time \(t\) . For county \(i\) , the sums \(j\) and \(h\) are over all counties. In other words, SPC for a county \(i\) , in time \(t\) , is the average of COVID- 19 incidence rates in connected counties weighted by their social connectedness to county \(i\) , i.e., the spatial lag of incidence rates. To the best of our knowledge, SPC data has not been published, but we were able to generate this feature using the original method \(^{6}\) , modified for our weekly temporal lagged features and calculated using incidence rates (cases per 10k population) rather than total number of cases. In the - FB models (Table 4), we incorporated features of weekly change \((\Delta)\) in SPC at four temporally lagged weeks (difference between the end and start of the lag week) to model the Infected SIR compartment in connected counties, as well as weekly average of SPC in the fourth lagged week (t- 4), to capture the Susceptible and Recovered SIR compartments in connected counties, similar to the rational for features in category C, as explained earlier.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 91, 433, 112]]<|/det|>
+## Features derived from SafeGraph  
+
+<|ref|>sub_title<|/ref|><|det|>[[137, 159, 469, 180]]<|/det|>
+### i. Intra-county movement features  
+
+<|ref|>text<|/ref|><|det|>[[111, 207, 881, 666]]<|/det|>
+To generate movement features from cell phone data, we used SafeGraph's SDM dataset that is "generated using a panel of GPS pings from anonymous mobile devices"48. The SDM dataset contains multiple mobility metrics published at the Census Block Group (CBG) level. Among these metrics, distance_traveled_from_home (median distance traveled by the observed devices in meters) and completely_home_device_count (the number of devices that did not leave their home location during a day)48 are conceptually closest to the metrics included in the Facebook's Movement Range Dataset. We used these two features in our - SG model, which is the conceptual equivalent of the - FB model, but with cell phone- derived features instead of the Facebook- derived features (Table 4).  
+
+<|ref|>text<|/ref|><|det|>[[112, 691, 880, 844]]<|/det|>
+We included the SafeGraph's median_home_dwell_time (median dwell time at home in minutes for all observed devices during the time period), and full_time_work_behavior_devices (the number of devices that spent more than 6 hours at a location other than their home during the day)48 in addition to the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 89, 857, 156]]<|/det|>
+previous two features in the - SGR model to take fuller advantage of the metrics available in the SDM dataset.  
+
+<|ref|>text<|/ref|><|det|>[[111, 180, 884, 686]]<|/det|>
+We derived baselined features from the SDM metrics as such: To address the potential effect of fewer cell phone observations in some CBGs, we used a Bayesian hierarchical model<sup>49,50</sup> with two levels (states and counties), and then smoothed the daily measurements using a seven- day rolling average to reduce the effect of outliers in the data. We then aggregated CBG- level completely_home_device_count and full_time_work_behavior_devices values up to the county level, divided by the total device_count in the county on the same day. For full_time_work_behavior_devices, we subtracted the final proportion from the February baseline of the same metric. For the median_home_dwell_time and distance_traveled_from_home variables, we calculated the weighted mean (by CBG population) of values per county, and then calculated percent of change compared to February baseline.  
+
+<|ref|>text<|/ref|><|det|>[[113, 709, 879, 819]]<|/det|>
+We used weekly averages and slopes (calculated by fitting a linear regression model to the values as the response variable and day of week as the independent variable) of these four metrics as features in our models (Table 4).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[125, 91, 592, 113]]<|/det|>
+## ii. Inter-county features and spatial lag modeling  
+
+<|ref|>text<|/ref|><|det|>[[112, 139, 884, 730]]<|/det|>
+Building on the conceptual structure of SCI, we derived a novel and daily inter- county connectivity index from SafeGraph's SDM dataset to quantify connectedness between counties based on the level of human flow from one county to the other (measured through cell- phone pings). We call this index "Flow Connectedness Index" (FCI). Using FCI, we then calculated a spatial lag metric that we call "Flow Proximity to Cases" (FPC) for each county. FPC captures the average of COVID- 19 incidence rates in connected (by human movement) counties weighted by the FCI. Again, it is worth noting that connectedness in this sense goes beyond the physical connectivity of counties, and considers daily human movement between them as the basis for determining connectivity. The similar formulations of FCI and SCI, as well as FPC and SPC, allow for direct comparison of the two networks (i.e. FB's friendship network and SafeGraph's human flow network) in their capability to capture inter- county physical human interactions, and subsequently, to predict new COVID- 19 cases.  
+
+<|ref|>text<|/ref|><|det|>[[113, 752, 852, 904]]<|/det|>
+The SafeGraph's SDM contains the number of visits between different CBGs. We aggregate these values to the county level to measure the daily number of devices that move (flow) between each county pair. Leveraging these flow measurements, we define Flow Connectedness Index (FCI) as:
+
+<--- Page Split --->
+<|ref|>equation<|/ref|><|det|>[[130, 93, 872, 140]]<|/det|>
+\[Flow connectedness index(FCI)_{i,j} = \frac{Device flow_{i,j} + Device flow_{j,i}}{Device count_{i} * Device count_{j}} \quad (3)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 152, 861, 263]]<|/det|>
+where for counties \(i\) and \(j\) , Device flow \(i,j\) is the sum of visits with origin \(i\) and destination \(j\) . Device count \(i\) is the number of devices whose home location is in county \(i\) . We then scale FCI to a range between 1 to 1,000,000,000.  
+
+<|ref|>text<|/ref|><|det|>[[137, 298, 320, 319]]<|/det|>
+We defined FPC as:  
+
+<|ref|>equation<|/ref|><|det|>[[163, 351, 872, 436]]<|/det|>
+\[\begin{array}{l}{{Flow P r o x i m i t y t o C a s e s(F P C)_{i,t}}}\\ {{=\sum_{j}C a s e s P e r10k_{j,t}\times\frac{F l o w C o n n e c t e d n e s s_{i,j}}{\sum_{h}F l o w C o n n e c t e d n e s s_{i,h}}}}\end{array} \quad (4)\]  
+
+<|ref|>text<|/ref|><|det|>[[113, 448, 842, 561]]<|/det|>
+Where Cases Per \(10k_{j,t}\) is the number of confirmed COVID- 19 cases per 10k population in county \(j\) at time \(t\) , and Flow Connectedness \(i,j\) is the value of FCI between county \(i\) and \(j\) .  
+
+<|ref|>text<|/ref|><|det|>[[112, 586, 870, 870]]<|/det|>
+Facebook's social network and friendship connections do not change significantly over time, and therefore, SCI is a static index in a one- year period. Conversely, inter- county human flow from SafeGraph is dynamic and can change dramatically, even within a week. We generated daily FCI (and FPC) for each county- pair in the US to utilize the full temporal resolution of the SDM dataset. We used weekly change \((\Delta)\) of FPC for the four temporally lagged weeks, and its average only in the fourth week as features - SG and - SGR models, with the same
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 797, 155]]<|/det|>
+rationale as features in Category C and D to capture SIR compartments in connected counties (Table 4).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 202, 339, 222]]<|/det|>
+## Model Implementation  
+
+<|ref|>text<|/ref|><|det|>[[112, 250, 876, 708]]<|/det|>
+The ultimate target variable in all of our autoregressive models is the number of new cases of COVID- 19. For training and tuning the models, however, we used a transformed target variable, namely the natural log- transformed values of new cases per 10k population plus one (to avoid zero values). For reporting the model predictions, we computed the number of new cases by applying an inverse transformation, i.e. an exponential transformation minus one (Formula 5a- c). The rationale for using the log- transformed target variable, as opposed to directly predicting the weekly new cases, was to minimize skewness, and more importantly, minimize the sensitivity of the models to the population of counties. Our exploratory work did confirm that using this logged of incidence rates produced better results.  
+
+<|ref|>equation<|/ref|><|det|>[[203, 737, 870, 821]]<|/det|>
+\[\begin{array}{r l} & {y_{p r e d i c t e d(i,t)} = l n(\Delta i n c i d e n c e r a t e_{(i,t)} + 1)}\\ & {\Delta i n c i d e n c e r a t e_{p r e d i c t e d(i,t)} = \mathrm{e}^{y_{p r e d i c t e d(i,t)}} - 1}\\ & {\Delta C a s e_{p r e d i c t e d(i,t)} = \left(\Delta i n c i d e n c e r a t e_{p r e d i c t e d(i,t)}\right)*P o p u l a t i o n_{i} / 10,000} \end{array} \quad (5.c)\]
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 89, 857, 160]]<|/det|>
+\(\Delta C a s e_{p r e d i c t e d_{(i,t)}}\) in 5.c denotes the number of new cases in a county in the prediction horizon.  
+
+<|ref|>text<|/ref|><|det|>[[112, 187, 870, 558]]<|/det|>
+Our training dataset includes up to 34 training samples per county (number of total samples \(n = 3103 \times 34\) ), with each sample holding various features in one- to four- weekly temporal lags (Table 4). The weekly calculation of features is based on weeks starting on Sundays and ending on Saturdays, with predictions also made for horizons spanning Sunday- Saturday periods as in common practice<sup>25</sup>. Our features, models, evaluations and comparisons are limited to the counties in the coterminous US. Table 4 summarizes the features that we used and the number of temporal lags (if any) used for each feature. All features were standardized for use in machine learning algorithms.  
+
+<|ref|>text<|/ref|><|det|>[[113, 662, 866, 902]]<|/det|>
+Our general approach to training, validation and testing of our models for different prediction horizons is similar, only, with target variables calculated separately for the specific prediction horizon. We first outline our approach for one- week ahead prediction horizons, which is used as the basis for algorithm selection and comparison of Facebook- and SafeGraph- derived features (- FB and - SG models); It is worth noting that we compare the two feature sets in longer
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 89, 857, 156]]<|/det|>
+term predictions too (Supplementary Table 5). We then provide an overview of the implementation of the models for longer term prediction horizons.  
+
+<|ref|>text<|/ref|><|det|>[[112, 182, 881, 596]]<|/det|>
+We trained and tuned the models using randomized search and 5- fold cross validation, and tested the best tuned model for predicting new cases in the week following the forecast date (for the one- week prediction horizons, as listed in Table 5). For instance, for the forecast date of October \(24^{\text{th}}\) , we used features which were generated using data collected before October \(24^{\text{th}}\) for tuning and training. The tuned model was then used for predicting new cases in each county during the October \(24^{\text{th}}\) to October \(31^{\text{st}}\) period. We used the reported cases by the JHU CSSE. The temporally lagged features for this forecast date were generated for t- 1, t- 2, t- 3 and t- 4 weekly lags, namely, the weeks ending on Oct. 24, Oct. 17, Oct. 10, and Oct. 3 respectively.  
+
+<|ref|>text<|/ref|><|det|>[[113, 621, 879, 904]]<|/det|>
+For the next forecast date, October 31, the training size increased by one week (per county), and the target week was also shifted by one week. Table 5 summarizes the forecast dates, one- week ahead prediction horizons, and training data size. The data used in generating these features spans a period from March 29 to November 28, 2020 to cover the temporal lags, and the target variable is collected through December 12, 2020 for the evaluation of four- week ahead predictions on the last forecast date. More details on cross validation,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 91, 770, 156]]<|/det|>
+hyperparameters, and evaluation are presented in the supplementary information (Section C).  
+
+<|ref|>table<|/ref|><|det|>[[177, 255, 820, 575]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[120, 211, 875, 246]]<|/det|>
+Table 5. Summary of training and testing data used for machine learning algorithm selection, as well as comparison of -FB and -SG models in short-term predictions   
+
+<table><tr><td>Forecast Date</td><td>One-week prediction Horizon</td><td>Training data start date</td><td>Training data end date</td><td># training samples per county</td></tr><tr><td>2020-10-24</td><td>2020-10-25 to 2020-10-31</td><td>2020-03-29</td><td>2020-10-24</td><td>30</td></tr><tr><td>2020-10-31</td><td>2020-11-01 to 2020-11-07</td><td>2020-03-29</td><td>2020-10-31</td><td>31</td></tr><tr><td>2020-11-07</td><td>2020-11-08 to 2020-11-14</td><td>2020-03-29</td><td>2020-11-07</td><td>32</td></tr><tr><td>2020-11-14</td><td>2020-11-15 to 2020-11-21</td><td>2020-03-29</td><td>2020-11-14</td><td>33</td></tr><tr><td>2020-11-21</td><td>2020-11-22 to 2020-11-28</td><td>2020-03-29</td><td>2020-11-21</td><td>34</td></tr></table>  
+
+<|ref|>text<|/ref|><|det|>[[113, 600, 855, 840]]<|/det|>
+We experimented with five different supervised machine learning regression algorithms, namely Random Forest \(^{51}\) (RF), Stochastic Gradient Boosting \(^{52}\) (SGB), eXtreme Gradient Boosting \(^{36,53}\) (XGB), Feed Forward Neural Network \(^{54}\) (FFNN), and Long Short- Term Memory \(^{55}\) (LSTM) network to build the autoregressive machine learning models with features described in Table 4. We evaluated the models using the dates listed in Table 5. Results are presented in Table 1. The
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 868, 156]]<|/det|>
+details of hyperparameter candidates and specific architectures are presented in the supplementary information (Section C).  
+
+<|ref|>sub_title<|/ref|><|det|>[[114, 201, 797, 223]]<|/det|>
+## Comparing Facebook-derived features with SafeGraph-derived features  
+
+<|ref|>text<|/ref|><|det|>[[112, 249, 871, 754]]<|/det|>
+Since the XGB algorithm performed best (Table 1), we chose it as the selected machine learning algorithm, and trained the base, - FB, - SG, and - SGR models using the XGB algorithm to predict new cases of COVID- 19 in short- term (one week) and long- term (two to four weeks) prediction horizons. To name a specific model in this article, we use a prefix that denotes the type of lag included in the model features (i.e. T for temporal or ST for spatiotemporal), followed by the name of the algorithm (XGB), followed by a suffix denoting the features included in the model, namely, - FB, - SG, and - SGR. Thus, TXGB (Temporal eXtreme Gradient Boosting) denotes the model that is built using the XGB algorithm and includes the base, Temporally lagged features; and STXGB- FB (SpatioTemporal eXtreme Gradient Boosting) denotes the model that includes Facebook- derived features (and thus, spatiotemporal lags) and is built using XGB.  
+
+<|ref|>text<|/ref|><|det|>[[113, 777, 880, 888]]<|/det|>
+We evaluated the performance of TXGB, STXGB- FB, STXGB- SG, and STXGB- SGR by comparing the RMSE and MAE scores of the predictions against the observed numbers of new cases in the corresponding prediction horizon (results in Table 2).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 814, 156]]<|/det|>
+We also tested the - FB and - SG models for all prediction horizons across all forecast dates (results in Supplementary Table 5).  
+
+<|ref|>text<|/ref|><|det|>[[112, 183, 880, 466]]<|/det|>
+Re- tuning and re- training all different variations of the STXGB model for each forecast date and prediction horizon resulted in considerably improved predictions compared to the baseline model (Table 4 and Supplementary Tables 4 and 5). Each STXGB model was tuned and trained on a regular desktop machine (with a 6 core Ryzen 5 3600X CPU and 64GB of RAM) in approximately 12- 13 minutes for a single prediction horizon, and thus, in almost one hour for all of the four prediction horizons.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 540, 737, 562]]<|/det|>
+## Evaluation against the COVID-19 Forecast Hub Ensemble baseline  
+
+<|ref|>text<|/ref|><|det|>[[112, 590, 872, 873]]<|/det|>
+In addition to the one- week short- term predictions, we performed long- term predictions of new COVID- 19 cases in two-, three-, and four- week ahead prediction horizons. We only used the STXGB algorithm to develop long- term prediction models since it outperformed other algorithms in short- term predictions (see Section 2). We used the same set of features for long- term predictions, with modifications on the target variable to reflect different prediction horizons. For instance, the two-, three-, and four- week ahead horizons
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 89, 880, 156]]<|/det|>
+of the Forecast date Oct. 24, were Oct 24 to Nov. 7, Oct. 24 to Nov. 14, and Oct 24 to Nov. 21, respectively.  
+
+<|ref|>text<|/ref|><|det|>[[112, 183, 879, 510]]<|/det|>
+The model for each horizon was trained and validated separately using the same training data and approach described in the previous Section (Table 5), and was tested on two, three and four weeks of unseen data, respectively, for each horizon. We evaluated the models' predictions by comparing them against the predictions generated by the COVID- 19 Forecast Hub's Ensemble model as well as the ground- truth values of new cases derived from JHU CSSE COVID- 19 reports. Additionally, we compared the long- term predictions of STXGB- FB and STXGB- SG (Supplementary Table 5).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 554, 275, 575]]<|/det|>
+## Data Availability  
+
+<|ref|>text<|/ref|><|det|>[[112, 604, 872, 880]]<|/det|>
+All of the datasets used in this study are publicly available (at the time of writing this manuscript). We created socioeconomic features from the 5- year survey data- - between 2014- 2018- - provided by the American Community Survey (ACS) and available at IPUMS National Historical GIS portal (https://www.nhgis.org/). Daily maximum and minimum temperature surfaces of the U.S. published by the NOAA are available at https://ftp.cpc.ncep.noaa.gov/GIS/GRADS_GIS/GeoTIFF/TEMP/. We used the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 884, 417]]<|/det|>
+cumulative confirmed COVID- 19 cases published by the Johns Hopkins University Center for Systems Science and Engineering (JHU CSSE) to generate COVID- related features. Facebook's Social Connectedness Index (SCI) database is available at https://dataforgood.fb.com/tools/social-connectedness- index/ and the movement range dataset can be found at https://data.humdata.org/dataset/movement- range- maps. Finally, the instructions for accessing SafeGraph's Social Distancing Metrics dataset is available at https://docs.safegraph.com/docs/social- distancing- metrics.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 460, 278, 481]]<|/det|>
+## Code Availability  
+
+<|ref|>text<|/ref|><|det|>[[114, 510, 852, 619]]<|/det|>
+All code necessary for the replication of our results is available for reviewers upon request. The code will be published publicly on GitHub under MIT license upon acceptance of this article.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 666, 369, 687]]<|/det|>
+## Summary of Contributions  
+
+<|ref|>text<|/ref|><|det|>[[113, 715, 879, 867]]<|/det|>
+Morteza Karimzadeh conceptualized the project, designed the features and contributed \(30\%\) of data processing and implementation, and contributed equally to writing. Behzad Vahedi conducted the majority of data processing, implementation, literature review, and contributed equally to writing. Hamidreza
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 812, 156]]<|/det|>
+Zoraghein contributed to the study design and \(10\%\) of implementation and writing.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 204, 310, 223]]<|/det|>
+## Competing interests  
+
+<|ref|>text<|/ref|><|det|>[[137, 253, 617, 273]]<|/det|>
+The authors do not report any competing interests.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 92, 223, 111]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[115, 140, 864, 250]]<|/det|>
+1. Chu, D. K. et al. Physical distancing, face masks, and eye protection to prevent person-to-person transmission of SARS-CoV-2 and COVID-19: a systematic review and meta-analysis. The Lancet 395, 1973-1987 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 269, 856, 378]]<|/det|>
+2. Chiu, W. A., Fischer, R. & Ndeffo-Mbah, M. L. State-level needs for social distancing and contact tracing to contain COVID-19 in the United States. Nat. Hum. Behav. 4, 1080-1090 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 398, 880, 464]]<|/det|>
+3. Eames, K. T. D. & Keeling, M. J. Contact tracing and disease control. Proc. R. Soc. Lond. B Biol. Sci. 270, 2565-2571 (2003).  
+
+<|ref|>text<|/ref|><|det|>[[115, 485, 880, 595]]<|/det|>
+4. Clark, E., Chiao, E. Y. & Amirian, E. S. Why Contact Tracing Efforts Have Failed to Curb Coronavirus Disease 2019 (COVID-19) Transmission in Much of the United States. Clin. Infect. Dis. (2020) doi:10.1093/cid/ciaa1155.  
+
+<|ref|>text<|/ref|><|det|>[[115, 614, 852, 724]]<|/det|>
+5. Bailey, M., Cao, R., Kuchler, T. & Stroebel, J. The Economic Effects of Social Networks: Evidence from the Housing Market. J. Polit. Econ. 126, 2224-2276 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[115, 744, 852, 852]]<|/det|>
+6. Kuchler, T., Russel, D. & Stroebel, J. The Geographic Spread of COVID-19 Correlates with the Structure of Social Networks as Measured by Facebook. J. Urban Econ., 103314 (2020).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 802, 156]]<|/det|>
+7. Büchel, K. & Ehrlich, M. v. Cities and the structure of social interactions: Evidence from mobile phone data. J. Urban Econ. 119, 103276 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 177, 822, 243]]<|/det|>
+8. Kang, Y. et al. Multiscale dynamic human mobility flow dataset in the U.S. during the COVID-19 epidemic. Sci. Data 7, 390 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 262, 880, 373]]<|/det|>
+9. Bullinger, L. R., Carr, J. B. & Packham, A. COVID-19 and Crime: Effects of Stay-at-Home Orders on Domestic Violence. https://www.nber.org/papers/w27667 (2020) doi:10.3386/w27667.  
+
+<|ref|>text<|/ref|><|det|>[[113, 392, 850, 459]]<|/det|>
+10. Killeen, B. D. et al. A County-level Dataset for Informing the United States' Response to COVID-19. Preprint at https://arxiv.org/abs/2004.00756 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 478, 852, 544]]<|/det|>
+11. Gao, S., Rao, J., Kang, Y., Liang, Y. & Kruse, J. Mapping county-level mobility pattern changes in the United States in response to COVID-19. (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 564, 877, 672]]<|/det|>
+12. Sen-Crowe, B., McKenney, M. & Elkbuli, A. Social distancing during the COVID-19 pandemic: Staying home save lives. Am. J. Emerg. Med. 38, 1519–1520 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 693, 877, 760]]<|/det|>
+13. Unwin, H. J. T. et al. State-level tracking of COVID-19 in the United States. Nat. Commun. 11, 6189 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 780, 872, 890]]<|/det|>
+14. Bertozzi, A. L., Franco, E., Mohler, G., Short, M. B. & Sledge, D. The challenges of modeling and forecasting the spread of COVID-19. Proc. Natl. Acad. Sci. 117, 16732–16738 (2020).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 88, 881, 199]]<|/det|>
+15. Zhou, Y. et al. A Spatiotemporal Epidemiological Prediction Model to Inform County-Level COVID-19 Risk in the United States. *Harv. Data Sci. Rev.* (2020) doi:10.1162/99608f92.79e1f45e.  
+
+<|ref|>text<|/ref|><|det|>[[115, 218, 825, 328]]<|/det|>
+16. Singh, R. K. et al. Prediction of the COVID-19 Pandemic for the Top 15 Affected Countries: Advanced Autoregressive Integrated Moving Average (ARIMA) Model. *JMIR Public Health Surveill.* 6, e19115 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 348, 866, 457]]<|/det|>
+17. Dansana, D. et al. Global Forecasting Confirmed and Fatal Cases of COVID-19 Outbreak Using Autoregressive Integrated Moving Average Model. *Front. Public Health* 8, (2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 477, 835, 544]]<|/det|>
+18. Xiang, J. et al. Impacts of the COVID-19 responses on traffic-related air pollution in a Northwestern US city. *Sci. Total Environ.* 747, 141325 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 563, 848, 672]]<|/det|>
+19. Zeroual, A., Harrou, F., Dairi, A. & Sun, Y. Deep learning methods for forecasting COVID-19 time-Series data: A Comparative study. *Chaos Solitons Fractals* 140, 110121 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 692, 861, 803]]<|/det|>
+20. Jo, H., Kim, J., Huang, T.-C. & Ni, Y.-L. condLSTM-Q: A novel deep learning model for predicting Covid-19 mortality in fine geographical Scale. Preprint at https://arxiv.org/abs/2011.11507 (2020).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 880, 200]]<|/det|>
+21. Mollalo, A., Rivera, K. M. & Vahedi, B. Artificial Neural Network Modeling of Novel Coronavirus (COVID-19) Incidence Rates across the Continental United States. Int. J. Environ. Res. Public. Health 17, 4204 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 218, 860, 373]]<|/det|>
+22. Buchwald, A. G., Adams, J., Bortz, D. M. & Carlton, E. J. Infectious Disease Transmission Models to Predict, Evaluate, and Improve Understanding of COVID-19 Trajectory and Interventions. Ann. Am. Thorac. Soc. 17, 1204–1206 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 391, 872, 459]]<|/det|>
+23. Reiner, R. C. et al. Modeling COVID-19 scenarios for the United States. Nat. Med. 27, 94–105 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[113, 477, 860, 545]]<|/det|>
+24. Santosh, K. C. COVID-19 Prediction Models and Unexploited Data. J. Med. Syst. 44, 170 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 563, 880, 672]]<|/det|>
+25. Ray, E. L. et al. Ensemble Forecasts of Coronavirus Disease 2019 (COVID-19) in the U.S. Preprint at https://www.medrxiv.org/content/10.1101/2020.08.19.20177493v1 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 692, 801, 804]]<|/det|>
+26. CDC. Cases, Data, and Surveillance. Centers for Disease Control and Prevention https://www.cdc.gov/coronavirus/2019-ncov/cases-updates/forecasts-cases.html (Retrieved on Jan 30, 2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 822, 875, 889]]<|/det|>
+27. Dong, E., Du, H. & Gardner, L. An interactive web-based dashboard to track COVID-19 in real time. Lancet Infect. Dis. 20, 533–534 (2020).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 840, 243]]<|/det|>
+28. Epidemiologists Urge A Cautious Christmas, After Thanksgiving Surge in Some States. NPR.org https://www.npr.org/sections/healthshots/2020/12/21/948809129/epidemiologists-urge-a-cautious-christmas-after-thanksgiving-surge-in-some-state (Retrieved on Jan 30, 2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 261, 881, 413]]<|/det|>
+29. Rama, D., Mejova, Y., Tizzoni, M., Kalimeri, K. & Weber, I. Facebook Ads as a Demographic Tool to Measure the Urban-Rural Divide. in Proceedings of The Web Conference 2020 327–338 (Association for Computing Machinery, 2020). doi:10.1145/3366423.3380118.  
+
+<|ref|>text<|/ref|><|det|>[[113, 433, 872, 544]]<|/det|>
+30. Wang, Q., Phillips, N. E., Small, M. L. & Sampson, R. J. Urban mobility and neighborhood isolation in America’s 50 largest cities. Proc. Natl. Acad. Sci. 115, 7735–7740 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[113, 563, 861, 674]]<|/det|>
+31. Coston, A. et al. Leveraging Administrative Data for Bias Audits: Assessing Disparate Coverage with Mobility Data for COVID-19 Policy. Preprint at https://arxiv.org/abs/2011.07194 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 693, 857, 845]]<|/det|>
+32. Nikolopoulos, K., Punia, S., Schäfers, A., Tsinopoulos, C. & Vasilakis, C. Forecasting and planning during a pandemic: COVID-19 growth rates, supply chain disruptions, and governmental decisions. Eur. J. Oper. Res. 290, 99–115 (2021).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 880, 200]]<|/det|>
+33. CDC. Coronavirus Disease 2019 (COVID-19). Centers for Disease Control and Prevention https://www.cdc.gov/coronavirus/2019-ncov/more/masking- science-sars-cov2.html (Retrieved on Jan 30, 2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 218, 879, 373]]<|/det|>
+34. Wyche, S. P., Schoenebeck, S. Y. & Forte, A. 'Facebook is a luxury': an exploratory study of social media use in rural Kenya. in Proceedings of the 2013 conference on Computer supported cooperative work 33–44 (Association for Computing Machinery, 2013).  
+
+<|ref|>text<|/ref|><|det|>[[113, 391, 830, 459]]<|/det|>
+35. LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[113, 478, 861, 629]]<|/det|>
+36. Chen, T. & Guestrin, C. XGBoost: A Scalable Tree Boosting System. in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining 785–794 (Association for Computing Machinery, 2016). doi:10.1145/2939672.2939785.  
+
+<|ref|>text<|/ref|><|det|>[[113, 649, 848, 717]]<|/det|>
+37. Bashir, M. F. et al. Correlation between climate indicators and COVID-19 pandemic in New York, USA. Sci. Total Environ. 728, 138835 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 735, 879, 847]]<|/det|>
+38. Mollalo, A., Vahedi, B. & Rivera, K. M. GIS-based spatial modeling of COVID-19 incidence rate in the continental United States. Sci. Total Environ. 728, 138884 (2020).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 857, 242]]<|/det|>
+39. Khalatbari-Soltani, S., Cumming, R. C., Delpierre, C. & Kelly-Irving, M. Importance of collecting data on socioeconomic determinants from the early stage of the COVID-19 outbreak onwards. J Epidemiol Community Health 74, 620–623 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 262, 866, 328]]<|/det|>
+40. Li, H. et al. Air pollution and temperature are associated with increased COVID-19 incidence: A time series study. Int. J. Infect. Dis. 97, 278–282 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 348, 881, 413]]<|/det|>
+41. Toda, A. A. Susceptible-Infected-Recovered (SIR) Dynamics of COVID-19 and Economic Impact. Preprint at https://arxiv.org/abs/2003.11221(2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 433, 836, 500]]<|/det|>
+42. Volz, E. & Meyers, L. A. Susceptible–infected–recovered epidemics in dynamic contact networks. Proc. R. Soc. B Biol. Sci. 274, 2925–2934 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[113, 520, 816, 630]]<|/det|>
+43. Lauer, S. A. et al. The Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Application. Ann. Intern. Med. 172, 577–582 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 650, 860, 760]]<|/det|>
+44. Chen, N. et al. Epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study. The Lancet 395, 507–513 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[113, 780, 880, 846]]<|/det|>
+45. Protecting privacy in Facebook mobility data during the COVID-19 response. Facebook Research https://research.fb.com/blog/2020/06/protecting-privacy-
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[135, 90, 880, 155]]<|/det|>
+in- facebook- mobility- data- during- the- covid- 19- response/ ( Retrieved on Jan 30, 2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 176, 880, 286]]<|/det|>
+46. Bailey, M., Cao, R., Kuchler, T., Stroebel, J. & Wong, A. Social Connectedness: Measurement, Determinants, and Effects. J. Econ. Perspect. 32, 259–280 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[115, 304, 860, 415]]<|/det|>
+47. Social Connectedness Index Methodology. Facebook Data for Good https://dataforgood.fb.com/docs/social-connectedness-index-methodology/ (Retrieved on Jan 30, 2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 435, 863, 544]]<|/det|>
+48. Social Distancing Metrics. SafeGraph https://docs.safegraph.com/docs/social-distancing-metrics ( Retrieved on Jan 30, 2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 564, 857, 589]]<|/det|>
+49. Gelman, A. et al. Bayesian Data Analysis, Third Edition. (CRC Press, 2013).  
+
+<|ref|>text<|/ref|><|det|>[[115, 608, 856, 674]]<|/det|>
+50. Devine, O. J., Louis, T. A. & Halloran, M. E. Empirical Bayes Methods for Stabilizing Incidence Rates before Mapping. Epidemiology 5, 622–630 (1994).  
+
+<|ref|>text<|/ref|><|det|>[[115, 694, 722, 717]]<|/det|>
+51. Breiman, L. Random forests. Mach. Learn. 45, 5–32 (2001).  
+
+<|ref|>text<|/ref|><|det|>[[115, 737, 860, 804]]<|/det|>
+52. Friedman, J. H. Stochastic gradient boosting. Comput. Stat. Data Anal. 38, 367–378 (2002).  
+
+<|ref|>text<|/ref|><|det|>[[115, 824, 816, 889]]<|/det|>
+53. Friedman, J. H. Greedy Function Approximation: A Gradient Boosting Machine. Ann. Stat. 29, 1189–1232 (2001).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 89, 864, 234]]<|/det|>
+54. Fine, T. L. \*Feedforward Neural Network Methodology\*. (Springer Science & Business Media, 2006).55. Hochreiter, S. & Schmidhuber, J. Long Short-Term Memory. \*Neural Comput.\* 9, 1735–1780 (1997).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[496, 911, 520, 925]]<|/det|>
+55
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[43, 44, 143, 69]]<|/det|>
+## Figures  
+
+<|ref|>image<|/ref|><|det|>[[68, 106, 480, 707]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[43, 720, 115, 740]]<|/det|>
+<center>Figure 1 </center>  
+
+<|ref|>image<|/ref|><|det|>[[496, 115, 892, 707]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[41, 760, 940, 896]]<|/det|>
+Prediction errors of the SpatioTemporal eXtreme Gradient Boosting (STXGB) models in one- week prediction horizon against a temporal autoregressive model without spatial lags (TXGB). Error comparison of STXGB models with different feature sets in predicting weekly new cases (top row) and new cases per 10k population (bottom row) for one- week ahead horizon. STXGB- FB, which incorporates Facebook- derived features, including spatial lags based on Social Connectedness Index, outperforms other models. Left column: prediction RMSE. Right column: prediction MAE.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[78, 68, 930, 352]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 383, 117, 402]]<|/det|>
+<center>Figure 2 </center>  
+
+<|ref|>text<|/ref|><|det|>[[42, 426, 951, 490]]<|/det|>
+Long- term prediction error comparison. Prediction errors of the STXGB- FB model (dashed line) and the Ensemble baseline (solid line) over four prediction horizons on five forecasting dates. a) Prediction RMSE and b) Prediction MAE.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[99, 65, 940, 518]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 538, 116, 558]]<|/det|>
+<center>Figure 3 </center>  
+
+<|ref|>text<|/ref|><|det|>[[41, 579, 949, 693]]<|/det|>
+Map of COVID- 19 cases per 10k population and errors in predicting them. (a) number of confirmed new cases per 10k population over the week ahead of forecasting date Oct. 31, 2020 (b) prediction errors for the same forecasting date(c) number of new cases over the week ahead of Nov. 7, 2020 forecasting date (d) prediction errors for the same forecasting date. The pattern of errors in Georgia, and Texas, and Kentucky flip from Oct. 31 to Nov. 7, indicating potential lags in testing and reporting.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[70, 50, 700, 787]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[44, 800, 117, 819]]<|/det|>
+<center>Figure 4 </center>  
+
+<|ref|>text<|/ref|><|det|>[[42, 842, 945, 907]]<|/det|>
+Prediction errors in urban vs. rural counties. Prediction errors of the number of new cases (left column) and new cases per 10k population (right column) in rural and urban counties on the Nov. 7 forecast date across four prediction horizons. The higher and lower \(3\%\) of counties are trimmed from the plot view.  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 931, 310, 957]]<|/det|>
+## Supplementary Files
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[44, 45, 765, 65]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 82, 618, 102]]<|/det|>
+- SPCandFPCSupplementaryinformationsubmissionready.docx
+
+<--- Page Split --->

preprint/preprint__48fdc4dc3cce1523cd0346c874c274f187408b5762031fb3d036d6b4eed72790/images_list.json

@@ -0,0 +1,33 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Fig. 2 | Heterologous vaccination induces neutralizing antibodies as well as CD4 and CD8 T cell responses. a, Heterologous ChAd/BNT/BNT or ChAd/ChAd/BNT vaccination induces neutralizing antibodies against Wuhan, B.1.1.7 (Alpha), B.1.351 (Beta), P.1 (B.1.1.28.1; Gamma), B.1.617.2 (Delta) and the B.1.1.529 (Omicron) SARS-CoV-2-S variants measured using the sVNT. Data are from \\(n =\\) biologically independent samples as indicated. For better visualization of identical titer values, data were randomly and proportionally adjusted closely around the precise titer results. Boost vaccination increased total percentage of cytokine-secreting CD4+ (b) and CD8+ (c) T cells. We calculated the",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 13
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_1.jpg",
+    "caption": "Extended Data Fig. 1 | Antibody panel a and gating strategy b for SARS-CoV-2-S (Spike)-specific B cell populations in blood. Pseudocolor plots show representative data from a female donor<br>283 days after priming with ChAd; 213 days after a second dose with BNT and 14 days after a third dose with BNT.",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        300,
+        846,
+        572
+      ]
+    ],
+    "page_idx": 15
+  },
+  {
+    "type": "image",
+    "img_path": "images/Extended_Data_Figure_5.jpg",
+    "caption": "Extended Data Fig. 5 | Frequency of cytokine-producing CD4<sup>+</sup> T cells and CD8<sup>+</sup> T cells after ex vivo re-stimulation with DMSO or the pool of Spike-specific peptides for 12–16 hr.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 18
+  }
+]

preprint/preprint__48fdc4dc3cce1523cd0346c874c274f187408b5762031fb3d036d6b4eed72790/preprint__48fdc4dc3cce1523cd0346c874c274f187408b5762031fb3d036d6b4eed72790.mmd

@@ -0,0 +1,355 @@
+
+# BNT162b2 boosted immune responses six months after heterologous or homologous ChAdOx1nCoV-19/BNT162b2 vaccination against COVID-19  
+
+Georg Behrens ( \(\boxed{ \begin{array}{r l} \end{array} }\) Behrens.georg@mh- hannover.de) Medizinische Hochschule Hannover (MHH) https://orcid.org/0000- 0003- 3111- 621X  
+
+Joana Barros- Martins Medizinische Hochschule Hannover https://orcid.org/0000- 0002- 3370- 4990  
+
+Anne Cossmann Hannover Medical School  
+
+Gema Morillas Ramos Hannover Medical School https://orcid.org/0000- 0002- 4835- 5172  
+
+Metodi Stankov Hannover Medical School  
+
+Ivan Odak Hannover Medical School  
+
+Alexandra Dopfer- Jablonka Hannover Medical School https://orcid.org/0000- 0001- 7129- 100X  
+
+Laura Hetzel Hannover Medical School  
+
+Miriam Köhler Hannover Medical School  
+
+Gwendolyn Patzer Hannover Medical School  
+
+Christoph Binz Hannover Medical School  
+
+Christiane Ritter Hannover Medical School  
+
+Michaela Friedrichsen Hannover Medical School  
+
+Christian Schultze- Florey Hannover Medical School https://orcid.org/0000- 0002- 3307- 2639  
+
+Inga Ravens Hannover Medical School  
+
+Stefanie Willenzon
+
+<--- Page Split --->
+
+Hannover Medical School  
+
+Anja Bubke Hannover Medical School  
+
+Jasmin Ristenpart Hannover Medical School  
+
+Anika Janssen Hannover Medical School  
+
+George Ssebyatika University of Lübeck  
+
+Günter Bernhardt Hannover Medical School https://orcid.org/0000- 0002- 0510- 2853  
+
+Markus Hoffmann German Primate Center https://orcid.org/0000- 0003- 4603- 7696  
+
+Stefan Pöhlmann German Primate Center - Leibniz Institute for Primate Research https://orcid.org/0000- 0001- 6086- 9136  
+
+Thomas Krey University of Lübeck https://orcid.org/0000- 0002- 4548- 7241  
+
+Berislav Bosnjak Hannover Medical School https://orcid.org/0000- 0003- 3374- 7488  
+
+Swantje Hammerschmidt Hannover Medical School https://orcid.org/0000- 0003- 0718- 0422  
+
+Reinhold Forster Institute of Immunology, Hannover Medical School, Hannover https://orcid.org/0000- 0001- 6190- 7923  
+
+## Article  
+
+## Keywords:  
+
+Posted Date: December 29th, 2021  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 1200506/v1  
+
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Communications on August 18th, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 32527- 2.
+
+<--- Page Split --->
+
+# BNT162b2 boosted immune responses six months after heterologous or homologous ChAdOx1nCoV-19/BNT162b2 vaccination against COVID-19  
+
+Georg M. N. Behrens<sup>1,2,3,\*</sup>, Joana Barros- Martins<sup>4\*</sup>, Anne Cossmann<sup>1\*</sup>, Gema Morillas Ramos<sup>1\*</sup>, Metodi V. Stankov<sup>1</sup>, Ivan Odak<sup>2</sup>, Alexandra Dopfer- Jablonka<sup>1,2</sup>, Laura Hetzel<sup>1</sup>, Miriam Köhler<sup>4</sup>, Gwendolyn Patzer<sup>4</sup>, Christoph Binz<sup>4</sup>, Christiane Ritter<sup>4</sup>, Michaela Friedrichsen<sup>4</sup>, Christian Schultze- Florey<sup>4,5</sup>, Inga Ravens<sup>4</sup>, Stefanie Willenzon<sup>4</sup>, Anja Bubke<sup>4</sup>, Jasmin Ristenpart<sup>4</sup>, Anika Janssen<sup>4</sup>, George Ssebyatika<sup>6</sup>, Günter Bernhardt<sup>4</sup>, Markus Hoffmann<sup>7,8</sup>, Stefan Pöhlmann<sup>7,8</sup>, Thomas Krey<sup>5,9</sup>, Berislav Bošnjak<sup>4</sup>, Swantje I. Hammerschmidt<sup>4\*</sup>, Reinhold Förster<sup>2,4,10,\*</sup>  
+
+Affiliations:  
+
+<sup>1</sup> Department for Rheumatology and Immunology, Hannover Medical School, 30625 Hannover, Germany  
+
+<sup>2</sup> German Center for Infection Research (DZIF), Partner Site Hannover- Braunschweig, 30625 Hannover, Germany  
+
+<sup>3</sup> CiiM, Centre for Individualized Infection Medicine, Hannover, Germany  
+
+<sup>4</sup> Institute of Immunology, Hannover Medical School, 30625 Hannover, Germany  
+
+<sup>5</sup> Department of Hematology, Hemostasis, Oncology and Stem- Cell Transplantation, Hannover Medical School, 30625 Hannover, Germany  
+
+<sup>6</sup> Institute of Biochemistry, University of Lübeck, 23562 Lübeck, Germany  
+
+<sup>7</sup> Infection Biology Unit, German Primate Center, 37077 Göttingen, Germany  
+
+<sup>8</sup> Faculty of Biology and Psychology, Georg- August- University Göttingen, 37073 Göttingen, Germany  
+
+<sup>9</sup> German Center for Infection Research (DZIF), Partner Site Hannover- Braunschweig and Partner Site Hamburg- Lübeck- Borstel- Riems, Hamburg, Germany  
+
+<sup>10</sup> Cluster of Excellence RESIST (EXC 2155), Hannover Medical School, 30625 Hannover, Germany\* These authors contributed equally  
+
+Corresponding authors:  
+
+Georg M.N. Behrens, Department of Rheumatology and Immunology, Hannover Medical School, Carl- Neuberg- Straße 1, D - 30625 Hannover, Germany, Tel: +49 511 532 5337, Fax: +49 511 532 5324, Email: behrens.georg@mh- hannover.de  
+
+https://orcid.org/0000- 0003- 3111- 621X  
+
+Reinhold Förster, Institute for Immunology, Hannover Medical School, Carl- Neuberg- Straße 1, D - 30625 Hannover, Germany, Tel: +49 511 532 9721, Fax: +49 511 532 9722, Email: foerster.reinhold@mh- hannover.de  
+
+https://orcid.org/0000- 0001- 6190- 7923
+
+<--- Page Split --->
+
+## Abstract  
+
+AbstractReports suggest that COVID- 19 vaccine effectiveness is decreasing, either due to waning immune protection, emergence of new variants of concern, or both. Heterologous prime/boost vaccination with a vector- based approach (ChAdOx- 1nCov- 19, ChAd) followed by an mRNA vaccine (e.g. BNT162b2, BNT) appeared to be superior in inducing protective immunity, and large scale second booster vaccination is ongoing. However, data comparing declining immunity after homologous and heterologous vaccination as well as effects of a third vaccine application after heterologous ChAd/BNT vaccination are lacking. We longitudinally monitored immunity in ChAd/ChAd (n=41) and ChAd/BNT (n=88) vaccinated individuals and assessed the impact of a second booster with BNT in both groups. The second booster greatly augmented waning anti- spike IgG but only moderately increased spike- specific \(\mathrm{CD4^{+}}\) and \(\mathrm{CD8^{+}}\) T cells in both groups to cell frequencies already present after the boost. More importantly, the second booster efficiently restored neutralizing antibody responses against Alpha, Beta, Gamma, and Delta, but neutralizing activity against B.1.1.529 (Omicron) stayed severely impaired. Our data suggest that inferior SARS- CoV- 2 specific immune responses after homologous ChAd/ChAd vaccination can be cured by a heterologous BNT vaccination. However, prior heterologous ChAd/BNT vaccination provides no additional benefit for spike- specific T cell immunity or neutralizing Omicron after the second boost.  
+
+## Main text  
+
+While the COVID- 19 vaccines currently approved by the European Medicine Agency (EMA) and U.S. Food and Drug Administration (FDA) provide high levels of protection against severe illness, the emergence of the Delta variant resulted in increasing numbers of breakthrough infections in fully vaccinated individuals (1). This coincided with evidence of waning immunity in vaccinated individuals (2, 3). Thus, booster vaccinations were proposed to reconstitute immunity and to potentially expand the breadth of immunity against SARS- CoV- 2 variants of concern (Voc). Also policy makers have begun to promote booster vaccination not only for vulnerable patients but also to mitigate healthcare and economic impact. Real world data confirm that booster vaccination is effective in preventing COVID- 19 (4- 6).  
+
+Concomitant to booster vaccination campaigns, the novel Omicron variant was recently identified in South Africa and its emergence was associated with a steep increase in cases and hospitalizations. Omicron is spreading rapidly in European, African and Asian countries as well as the USA, with case numbers doubling every two to three days (7). The S protein of the Omicron variant harbors an unusually high number of mutations, which increases immune evasion and potentially transmissibility (8- 11). Thus, the Omicron variant constitutes a rapidly emerging threat to public health and might undermine global efforts to control the COVID- 19 pandemic.  
+
+Heterologous prime boost strategies appear to offer immunological advantages to strengthen protection against COVID- 19 achieved with currently available vaccines. Administration of mRNA vaccines like BNT162b2 (Comirnaty; BNT) after the initial ChAdOx1- nCov- 19 (Vaxzevria, ChAd) dose as the second dose of a two- dose regimen was safe and had enhanced immunogenicity compared to homologous ChAdOx vaccination (12- 18). We have previously reported on the results after homologous and heterologous vaccination after ChAd priming (16). Here we present findings from a subsequent analysis assessing the efficacy of a second booster vaccination after heterologous and homologous prime- boost vaccination on neutralization of Voc including Omicron.  
+
+In addition to our previously reported findings, we longitudinally monitored immunity after prime- boost COVID- 19 vaccine treatment schedules and determined thereafter the impact of BNT booster
+
+<--- Page Split --->
+
+(Methods). Health care professional vaccines without previous SARS- CoV- 2 infection, who had received ChAd/ChAd or ChAd/BNT, donated further blood four and six months after first booster vaccination and about two weeks after the second booster. The vaccination and blood collection schedule is depicted in Fig. 1A with additional demographic information (age and sex) in Extended Data Table 1. A third group of BNT/BNT vaccines served as an independent control group for serologic analysis only and was monitored for up to nine months (Extended Data Table 1). As described (16), anti- SARS- CoV- 2 spike IgG (anti- S IgG) were significantly higher in the ChAd/BNT group short after prime- boost vaccination when compared to the ChAd/ChAd group but declined significantly over time in both groups, with lower anti- S IgG after homologous vaccination prior to the second boost (Fig. 1B).  
+
+Following second booster immunization, we found greatly increased anti- S IgG responses in both groups. Boosting of the heterologous ChAd/BNT immunized group led to a significant 47.9 - fold increase for anti- S IgG \((p< 0.0001)\) and 8.0- fold increase in individuals after homologous ChAd vaccination \((p< 0.0001)\) (Fig. 1B). In both groups, anti- S IgG were considerably higher as compared to the first boost time point. More importantly, the second booster diminished previous differences between the heterologous and homologous prime- boost vaccination groups, since anti- S IgG were comparable in both groups after the second BNT boost and were within the range of triple BNT vaccinated individuals (Fig. 1C).  
+
+Next, we measured the frequency and phenotype of memory B cells carrying membrane- bound immunoglobulins specific for the Spike protein (Methods, Extended Data Fig. 1) over time. Interestingly, numbers of spike- specific memory B cells generated after prime- boost vaccination gradually increased during the following months with no significant difference between the ChAd/ChAd and the ChAd/BNT group (Fig. 1D). Again, the second booster with BNT led to a further and significant expansion of spike- specific memory B cells in both groups (Fig. 1D) in line with increased amounts of spike- specific antibodies, highlighting the impact of the second booster vaccination for better protection from SARS- CoV- 2 infection.  
+
+For testing neutralizing activity of antibodies induced by vaccination, we employed our ELISA- based surrogate virus neutralization test (sVNT). We modified the sVNT to include Spike proteins of the Omicron variant (Methods) and for validation applied sera from vaccines that had been recently tested for their neutralizing capacity applying vesicular stomatitis virus (VSV)- based pseudotyped virus neutralization assays (pVNT) (9, 20, 21). As for other VoC (16), we obtained a high degree of correlation between both assays with a R square value of 0.7044 (Extended Data Fig. 2). Thus, the sVNT is a reliable tool to quantitatively assess the neutralization capacity of vaccination- induced antibodies not only against the Wuhan but also against the Alpha, Beta, Gamma, Delta and Omicron variants of SARS- CoV- 2.  
+
+Using the sVNT assays and consistent with declining anti- S IgG, we confirmed waning neutralizing activity against the Wuhan variant and particularly against VoC tested. Whilst the majority of participants had neutralizing antibodies against the Wuhan strain in pre- second boost plasma, neutralizing antibodies against the Alpha, Beta, Gamma and Delta variants were particularly in the ChAd/ChAd group less frequent or virtually absent (Fig. 2A). At 2 weeks after the second booster immunization, frequencies and titers of neutralizing antibodies against the Wuhan strain increased profoundly in the ChAd/BNT and ChAd/ChAd group with titers reaching values above those after the initial two injections in the latter group (Fig. 2A).  
+
+Differences between the pre- booster vaccination regimens became even more evident when analyzing the neutralization capacity of antibodies induced against the VoC. In the ChAd/ChAd group, second booster immunization profoundly increased neutralization of the Alpha, Beta, Gamma and
+
+<--- Page Split --->
+
+Delta variants (Fig. 2A), which was low for Alpha and Delta after prime/boost and virtually absent for Beta and Gamma. Whilst initial ChAd/BNT immunization had induced neutralizing antibodies at high levels against all analyzed VoC, the following decline was more than restored by the second booster (Fig. 2A). In fact, nearly all BNT boosted ChAd/BNT vaccines had efficient neutralizing activity against Alpha, Beta, Gamma and Delta and amounts mostly above those from after the first booster. Importantly, the neutralization capacity against the Omicron variant was virtually absent before the second booster and remained low thereafter in comparison to the other VoC (Fig. 2A), with 9/29 (31%) and 27/58 (47%) of vaccines in the ChAd/ChAd and ChAD/BNT group, respectively having no detectable neutralization activity after boost. We obtained very similar results after BNT booster in BNT/BNT vaccinated individuals (Extended Data Fig. 3). Altogether, these data indicate that the second booster immunization led to an increase of neutralizing antibodies in both vaccination groups against all tested VoC including the Omicron variant to only some extent.  
+
+Finally, we also analyzed frequencies and phenotypes of spike- specific T cells (Methods, Extended Data Fig. 4 and 5). We quantified numbers of spike- specific T cells as the sum of all cells producing IFN- \(\gamma\) or TNF- \(\alpha\) as described previously (16). The frequencies of spike- specific \(\mathrm{CD4^{+}}\) and \(\mathrm{CD8^{+}}\) T cells in blood samples collected after the first booster vaccination were significantly higher in the ChAd/BNT group (Fig. 2B and C) and both cell populations declined over time, while they remained more or less stable after homologous vaccination. Whilst spike- specific \(\mathrm{CD4^{+}}\) T cells declined to frequencies similar to individuals after homologous ChAd/ChAd vaccination (Fig. 2B), spike- specific \(\mathrm{CD8^{+}}\) T cells remained above the frequencies of the heterologous vaccinated group (Fig. 2C). More interestingly, heterologous BNT booster in the ChAd/ChAd vaccination group significantly raised numbers of spike- specific \(\mathrm{CD4^{+}}\) T cells above amounts observed after the first boost. In contrast, individuals with heterologous ChAd/BNT vaccination only regained spike- specific \(\mathrm{CD4^{+}}\) T cell levels corresponding to levels after the first boost (Fig. 2B). Similarly, reboot with BNT did not result in an expansion of spike- specific \(\mathrm{CD8^{+}}\) T cells above levels after first boost in ChAd/BNT vaccines, but did so in ChAd/ChAd vaccinated individuals (Fig. 2C). Like for spike- specific \(\mathrm{CD4^{+}}\) T cells, raised numbers in spike- specific IFN- \(\gamma\) - producing T cells in the ChAd/ChAd as well as the ChAd/BNT after BNT boost group was confirmed by cytokine measurement in supernatants after SARS- CoV- 2 spike peptide stimulation (Fig. 2D). Again, BNT boost did not further increase spike- specific IFN- \(\gamma\) - producing T cells in ChAd/BNT vaccinated subjects above levels obtained already after first boost vaccination.  
+
+BNT booster potently increased anti- S IgG in all heterologous and homologous vaccinated individuals tested and this rise was accompanied by further strengthened neutralizing capacity against the Wuhan variant and Alpha, Beta, Gamma, and Delta. These data corroborate findings after homologous vaccination (22) and support current recommendations by the European Medicines Agency (EMA) and the European Centre for Disease Prevention and Control (ECDC). However, neutralization of the Omicron variant was absent before second booster and remained clearly inferior thereafter, irrespective of the previous vaccination scheme. Considering the kinetics of waning neutralizing antibodies against the other VoC after the first booster, we expect remaining neutralization against Omicron to vanish rapidly in the majority of vaccines despite persisting high anti- S IgG concentrations. This suggests that variant- specific vaccines are required and need to be tested soon to better combat COVID- 19 caused by Omicron and other emerging variants. In contrast, the second BNT booster only made up for absent rise in spike- specific T cell responses after homologous ChAd/ChAd vaccination and merely restored spike- specific \(\mathrm{CD4^{+}}\) T cell responses after heterologous vaccination. These data confirm and extent reports that ChAd does not boost cellular responses after ChAd/ChAd vaccination (13, 23), but we are yet unable to make conclusion about the long- term protection and immunological memory. Although the relative role for T cell immunity remains unclear, spike- specific T memory cells are probably of great importance for protection against severe COVID- 19, hospitalization and death. Our data reveal that additional BNT booster
+
+<--- Page Split --->
+
+poorly support spike- specific \(\mathrm{CD8^{+}}\) expansion and suggest that novel vaccines and vaccine schedules should be explored for further strengthening of adaptive cellular immunity against SARS- CoV- 2 and its variants (24). Such vaccines should also aim to target other structural viral proteins including nucleocapsid and membrane proteins, which are less likely to be able to escape from capable immune recognition.  
+
+## Acknowledgements  
+
+This work was supported by the German Center for Infection Research TTU 01.938 (grant no 80018019238 to G.M.N.B and R.F.), and TTU 04.820 to G.M.N.B., by Deutsche Forschungsgemeinschaft, (DFG, German Research Foundation) Excellence Strategy EXC 2155 "RESIST" (Project ID39087428 to R.F.), by funds of the State of Lower Saxony (14- 76103- 184 CORONA- 11/20 to R.F.), by funds of the BMBF (NaFoUniMedCovid19 FKZ: 01KX2021; Projects B- FAST to R.F.) and Deutsche Forschungsgemeinschaft, SFB 900/3 (Projects B1, 158989968 to R.F.), and the European Regional Development Fund (Defeat Corona, ZW7- 8515131 and ZW7- 85151373 to G.M.N.B.). We thank the CoCo Study participants for their support and the entire CoCo study team for help. We would like to thank Luis Manthey, Annika Heidemann, Till Redeker, Madeleine Rommel, Christian Sturm, Marie Mikuteit, Jacqueline Niewolik, Ruth Sikora, Janine Topal, Kerstin Sträche, Birgit Heinisch, Michael Stephan, Mariel Nöhre, Simone Müller, Olivera Dragicevic, Kim Do Thi Hoang, Amy Kempf, and Inga Nehlmeier for technical and logistical support.  
+
+## Author Contributions Statement  
+
+Study design: G.M.N.B and R.F.  
+
+Data collection: J.B.- M., S.I.H., A.C., I.O., M.V.S., G.M.R., A.D.- J., L.H., M.K, G.P., C.B., C.R., M.F., C. S.- F., I.R., S.W., A.B., J.M., J.R., A.J., G.S., G.B., J.M., M.H., S.P., T.K.  
+
+Data analysis: G.M.N.B, J.B.- M., S.I.H., A.C., I.O., G.M.R., M.H., B.B, M.V.S.  
+
+Data interpretation: R.F., G.M.N.B.  
+
+Writing: G.M.N.B., R.F. with comments from all authors.  
+
+## Competing Interests Statement  
+
+The authors declare no competing interests.
+
+<--- Page Split --->
+
+## Methods  
+
+## Participants  
+
+ParticipantsParticipants for this analysis were from the COVID- 19 Contact (CoCo) Study (German Clinical Trial Registry, DRKS00021152), an ongoing, prospective observational study monitoring anti- SARS- CoV- 2 IgG immunoglobulin and immune responses in health care professionals (HCP) at Hannover Medical School and individuals with potential contact to SARS- CoV- 2 (19, 25). An amendment from Dec 2020 allowed us to study the immune responses after COVID- 19 vaccination. We followed the study cohort described previously (16) after heterologous ChAdOx/BNT or homologous ChAdOx/ ChAdOx and BNT/BNT vaccination. Scheduling appointments for a third booster vaccination with BNT was coordinated by an independent vaccination team according to vaccine availability.  
+
+One individual with previous SARS- CoV- 2 infection as determined by positive anti- SARS- CoV- 2 NCP IgG before vaccinations was excluded from this analysis. Two additional individuals of the ChAdOx/ ChAdOx group developed anti- SARS- CoV- 2 NCP IgG after prime/boost vaccination and were excluded from follow up analysis. Demographics (sex and age) are depicted in Extended Data Table 1. After blood collection, we separated plasma from EDTA or lithium heparin blood (S- Monovette, Sarstedt) and stored it at \(- 80^{\circ}C\) until use. We used full blood or isolated PBMCs from whole blood samples by Ficoll gradient centrifugation and for stimulation with SARS- CoV- 2 peptide pools.  
+
+## Pseudotyped virus neutralization assay (pVNT)  
+
+pVNTs were performed at the Infection Biology Unit of the German Primate Center in Gottingen as described recently (21). Briefly, the rhabdoviral pseudotyped particles were produced in 293T cells transfected to express the desired SARS- CoV- 2- S variant inoculated with VSV\*DG- Fluc, a replication- deficient VSV vector that encodes for enhanced green fluorescent protein and firefly luciferase (Fluc) instead of VSV- G protein (kindly provided by Gert Zimmer, Institute of Virology and Immunology, Mittelhäusern, Switzerland). Produced pseudoparticles were collected, cleared from cellular debris by centrifugation and stored at \(- 80^{\circ}C\) until used. For neutralization experiments, equal volumes of pseudotyped particles and heat- inactivated (56 \(^\circ C\) , 30 min) plasma samples serially diluted in culture medium were mixed and incubated for 30 min at \(37^{\circ}C\) . Afterwards, the samples together with non- plasma- exposed pseudotyped particles were used for transduction experiments. The assay was performed in 96- well plates in which Vero cells were inoculated with the respective pseudotyped particles/plasma mixtures. The transduction efficacy was analyzed at 16- 18 hr post inoculation by measuring Fluc activity in lysed cells (Cell culture lysis reagent, Promega) using a commercial substrate (Beetle- Juice, PJK) and a plate luminometer (Hidex Sense Plate Reader, Hidex) with the Hidex Sense Microplate Reader Software (version 0.5.41.0).  
+
+## Serology  
+
+We measured SARS- CoV- 2 IgG by quantitative ELISA (anti- SARS- CoV- 2 S1 Spike protein domain/receptor binding domain IgG SARS- CoV- 2- QuantivVac, Euroimmun, Lubeck, Germany) according to the manufacturer's instructions (dilution up to 1:4000). We provide anti- S1 concentrations expressed as RU/mL as assessed from a calibration curve with values above 11 RU/mL defined as positive. These values can be converted in binding antibody units (BAU/mL) by multiplying RU/mL by 3.2. We performed anti SARS- CoV- 2 nucleocapsid (NCP) IgG measurements according to the manufacturer's instructions (Euroimmun, Lubeck, Germany). We used an AESKU.READER (AESKU.GROUP, Wendelsheim, Germany) and the Gen5 2.01 Software for analysis.
+
+<--- Page Split --->
+
+## Surrogate virus neutralization assay (sVNT) for SARS-CoV-2 variants  
+
+To determine neutralizing antibodies against the Wuhan- Spike, the B.1.1.7- Spike (Alpha), the B.1.351- Spike (Beta), the P.1- Spike (B.1.1.28.1; Gamma), the B.1.617.2 (Delta), and the B.1.1.529 (Omicron) variants of SARS- CoV- 2- S in plasma, we modified our recently established surrogate virus neutralization test (sVNT) (20, 26). In this assay, the soluble receptor for SARS- CoV- 2, ACE2, is bound to 96- well- plates to which different purified tagged receptor binding domains (RBDs) of the Spike- protein of SARS- CoV- 2 can bind once added to the assay. Binding is further revealed by an anti- tag peroxidase- labelled antibody and colorimetric quantification. Pre- incubation of the Spike- protein with serum or plasma of convalescent patients or vaccines prevents subsequent binding to ACE2 to various degrees, depending on the amount of neutralizing antibodies present. In detail, MaxiSorp 96F plates (Nunc) were coated with recombinant soluble hACE2- Fc(IgG1) protein at 300 ng per well in \(50\mu \mathrm{L}\) coating buffer (30 mM Na2CO3, 70 mM NaHCO3, pH 9.6) at \(4^{\circ}C\) overnight. After blocking with hACE2- Fc(IgG1), plates were washed with phosphate- buffered saline, \(0.05\%\) Tween- 20 (PBST) and blocked with BD OptEIA Assay Diluent for \(1.5\mathrm{h}\) at \(37^{\circ}C\) . In the meantime, plasma samples were serially diluted threefold starting at 1:20 and then pre- incubated for \(1\mathrm{h}\) at \(37^{\circ}C\) with \(1.5\mathrm{ng}\) recombinant SARS- CoV- 2 Spike RBD of either the Wuhan strain (Trenzyme), the B.1.1.7 variant (N501Y; Alpha), the B.1.351 variant (K417N, E484K, N501Y; Beta) or the P.1 variant (K417T, E484K, N501Y; Gamma) (the latter three from SinoBiological), all with a C- terminal His- Tag. BD OptEIA Assay Diluent was used for preparing plasma sample as well as RBD dilutions. After pre- incubation with SARS- CoV- 2 Spike RBDs, plasma samples were given onto the hACE2- coated MaxiSorp ELISA plates for \(1\mathrm{h}\) at \(37^{\circ}C\) . SARS- CoV- 2 Spike RBDs pre- incubated with buffer only served as negative controls for inhibition. Plates were washed three times with PBST and incubated with an HRP- conjugated anti- His- tag antibody (clone HIS 3D5, provided by Helmholtz Zentrum München) for \(1\mathrm{h}\) at \(37^{\circ}C\) . Unbound antibody was removed by six washes with PBST. A colorimetric signal was developed on the enzymatic reaction of HRP with the chromogenic substrate \(3,3',5,5'\) - tetramethylbenzidine (BD OptEIA TMB Substrate Reagent Set). An equal volume of \(0.2\mathrm{MH}_2\mathrm{SO}_4\) was added to stop the reaction, and the absorbance readings at \(450\mathrm{nm}\) and \(570\mathrm{nm}\) were acquired using a SpectraMax iD3 microplate reader (Molecular Devices) using SoftMAX Pro v7.03 software. For each well, the percent inhibition was calculated from optical density (OD) values after subtraction of background values as: Inhibition \((\%) = (1 - \text{Sample OD value / Average SARS - CoV - 2 S RBD OD value})\times 100\) . Neutralizing sVNT titers were determined as the dilution with binding reduction \(> \mathrm{mean} + 2\mathrm{SD}\) of values from a plasma pool consisting of three pre- pandemic plasma samples.  
+
+## SARS-CoV-2 protein peptide pools  
+
+We ordered 15 amino acid (aa) long and 10 aa overlapping peptide pools spanning the whole length of SARS- CoV- 2- Spike (- S) (total 253 peptides), - Membrane (- M) (43 peptides), - Nucleocapsid (- N) (82 peptides) or - Envelope (- E) (12 peptides; peptide no 4 could not be synthesized) proteins from GenScript. All lyophilized peptides were synthesized at \(>95\%\) purity and reconstituted at a stock concentration of \(50\mathrm{mg / mL}\) in DMSO (Sigma- Aldrich), except for 9 SARS- CoV- 2- S overlapping peptides (number 24, 190, 191, 225, 226, 234, 244, 245 and 246), 2 for SARS- CoV- 2- M (number 15 and 16), 1 for SARS- CoV- 2- N (number 61) and all 12 SARS- CoV- 2- E peptides that were dissolved at \(25\mathrm{mg / mL}\) due to solubility issues. All peptides in DMSO stocks were stored at \(- 80^{\circ}\mathrm{C}\) until used.  
+
+## T cell re-stimulation assay  
+
+PBMCs, isolated using a Ficoll gradient, were re- suspended at concentration of \(20\times 10^{6}\) cells/mL in complete RPMI medium [RPMI 1640 (Gibco) supplemented with \(10\%\) FBS (GE Healthcare Life Sciences, Logan, UT), \(1\mathrm{mM}\) sodium pyruvate, \(50\mu \mathrm{M}\beta\) - mercaptoethanol, \(1\%\) streptomycin/penicillin (all Gibco)]. For stimulation, cells were diluted with equal volume of peptide pools containing S
+
+<--- Page Split --->
+
+protein or mixture of M-, N- and E- proteins. Peptide pools were prepared in complete RPMI containing brefeldin A (Sigma- Aldrich) at final concentration of \(10\mu \mathrm{g / mL}\) . In the final mixture each peptide had concentration of \(2\mu \mathrm{g}\) ( \(\sim 1.2\) nmol)/mL, except for SARS- CoV2- S peptides number 24, 190, 191, 225, 226, 234, 244, 245 and 246, SARS- CoV2- M peptides 15 and 16, and SARS- CoV2- N peptide 61, which were used at final concentration of \(1\mu \mathrm{g / mL}\) due to solubility issues. As a negative control, we stimulated the cells with DMSO, used in maximal volume corresponding to DMSO amount in peptide pools (equaling to \(5\%\) DMSO in final medium volume) Extended Fig. 5. In each experiment, we used cells stimulated with Phorbol- 12- myristate- 13- acetate (PMA; Calbiochem) and ionomycin (Invitrogen) at final concentration of \(50\mathrm{ng / mL}\) and \(1500\mathrm{ng / mL}\) , respectively, as an internal positive control. Cells were then incubated for 12- 16 hr at \(37^{\circ}C\) , \(5\%\) CO2. After washing, cells were resuspended in MACS buffer (PBS supplemented with \(3\%\) FBS and \(2\mathrm{mM}\) EDTA). Non- specific antibody binding was blocked by incubating samples with \(10\%\) mouse serum at \(4^{\circ}C\) for 15 min. Next, without washing, an antibody mix of anti- CD3- AF532 (UCHT1; #58- 0038- 42; Lot # 2288218; Invitrogen; 1:50), anti- CD4- BUV563 (RPA- T4; #741353; Lot # 9333607; BD Biosciences; 1:200), anti- CD8- SparkBlue 550 (SK1; #344760; Lot #B326454; Biolegend; 1:200), anti- CD45RA (HI100, #740298, Lot # 0295003; BD Biosciences; 1:200), anti- CCR7 (G043H7; #353230; Lot # B335328; Biolegend; 1:50), anti- CD38 PerCP- eF710 (HB7; #46- 0388- 42; Lot # 2044748; Invitrogen; 1:100) and Zombie NIR™ Fixable Viability Kit (#423106; Lot # B323372; BioLegend) was added. After staining for 20 min at RT, cells were washed before they were fixed and permeabilized (#554714; BD Biosciences) according to the manufacturers' protocol. Next, intracellular cytokines were stained using anti- IFN- PE- Cy7 (B27; #506518; Lot # B326674; Biolegend; 1:100), anti- TNF- AF700 (Mab11; #502928; Lot # B326186; Biolegend; 1:50) and anti- IL- 17A- BV421 (BL168; #512322; Lot # B317903; Biolegend; 1:50) for 45 min on RT. Excess antibodies were washed away and cells were then acquired on Cytek Aurora spectral flow cytometer (Cytek) equipped with five lasers operating on \(355\mathrm{nm}\) , \(405\mathrm{nm}\) , \(488\mathrm{nm}\) , \(561\mathrm{nm}\) and \(640\mathrm{nm}\) (for gating strategy, see Extended Data Fig. 4). All flow cytometry data was acquired using SpectroFlo v2.2.0 (Cytek) and analyzed FCS Express V7 (Denovo).  
+
+## Flow cytometric analysis of Spike-specific B cells  
+
+Total leukocytes were isolated from whole blood using erythrolysis in \(0.83\%\) ammonium chloride solution. Isolated cells were then washed, counted and resuspended in PBS and stained for 20 min on RT with an antibody mix containing antibodies listed in Extended Data Figure 1A together with Spike- mNEONGreen protein ( \(5\mu \mathrm{g}\) per reaction; production will be described elsewhere). After one wash, samples were acquired on spectral flow cytometer and the data was analyzed as described above (for gating strategy, see Extended Data Fig. 1B).  
+
+## Quantification of IFN- \(\gamma\) release  
+
+0.5 mL full blood were stimulated with manufacturer's selected parts of the SARS- CoV- 2 S1 domain of the Spike Protein for a period of 20- 24 h. We carried out negative and positive controls according to the manufacturer's instruction and measured IFN- \(\gamma\) using an ELISA (SARS- CoV- 2 Interferon Gamma Release Assay, IGRA (Euroimmun, Lübeck, Germany). For analysis, we used an AESKU.READER (AESKU.GROUP, Wendelsheim, Germany) and the Gen5 2.01 Software.  
+
+## Statistics  
+
+Statistical analysis was done using GraphPad Prism 8.4 (GraphPad Software, USA) and SPSS 20.0.0 (IBM SPSS Statistics, USA). For comparison of levels of Spike- specific IgG levels, as well as for comparison of percentages of cytokine- secreting T cells, for comparison of frequencies of Spike- specific B cells, or cytokine concentrations in the IGRA assay and sVNT values we used mixed effect analysis with Sidiak's multiple comparison paired t- test (within groups) or unpaired T test with Welch
+
+<--- Page Split --->
+
+correction 2- way ANOVA followed by Sidak's multiple comparison test (between groups). Percentages of cytokine secreting T cells were log transformed prior comparison. For comparison of sVNT titers we used Chi- square test for trend. Differences were considered significant if \(p < 0.05\) . Correlation between sVNT and pVNT values was calculated using single linear regression analysis.  
+
+Ethics committee approval.  
+
+The CoCo Study and the analysis conducted for this article were approved by the Internal Review Board of Hannover Medical School (institutional review board no. 8973_BO- K_2020, amendment Dec 2020).  
+
+## Data availability statement  
+
+All requests for raw and analyzed data that underlie the results reported in this article will be reviewed by the CoCo Study Team, Hannover Medical School (cocostudie@mh- hannover.de) to determine whether the request is subject to confidentiality and data protection obligations. Data that can be shared will be released via a material transfer agreement.  
+
+## References  
+
+1. Powell AA, Power L, Westrop S, McOwat K, Campbell H, Simmons R, et al. Real-world data shows increased reactogenicity in adults after heterologous compared to homologous prime-boost COVID-19 vaccination, March-June 2021, England. Euro Surveill. 2021;26(28).  
+2. Evans JP, Zeng C, Carlin C, Lozanski G, Saif LJ, Oltz EM, et al. Loss of Neutralizing Antibody Response to mRNA Vaccination against SARS-CoV-2 Variants: Differing Kinetics and Strong Boosting by Breakthrough Infection. bioRxiv. 2021.  
+3. Tober-Lau P, Schwarz T, Vanshylla K, Hillus D, Gruell H, Group ECS, et al. Long-term immunogenicity of BNT162b2 vaccination in older people and younger health-care workers. Lancet Respir Med. 2021;9(11):e104-e5.  
+4. Arbel R, Hammerman A, Sergienko R, Friger M, Peretz A, Netzer D, et al. BNT162b2 Vaccine Booster and Mortality Due to Covid-19. N Engl J Med. 2021.  
+5. Hansen CH, Schelde AB, Moustsen-Helms IR, Emborg H-D, Krause TG, Moelbak K, et al. Vaccine effectiveness against SARS-CoV-2 infection with the Omicron or Delta variants following a two-dose or booster BNT162b2 or mRNA-1273 vaccination series: A Danish cohort study. medRxiv. 2021:2021.12.20.21267966.  
+6. Bar-On YM, Goldberg Y, Mandel M, Bodenheimer O, Freedman L, Alroy-Preis S, et al. Protection against Covid-19 by BNT162b2 Booster across Age Groups. N Engl J Med. 2021;385(26):2421-30.  
+7. Torjesen I. Covid restrictions tighten as omicron cases double every two to three days. BMJ. 2021;375:n3051.  
+8. Cele S, Jackson L, Khan K, Khoury D, Moyo-Gwete T, Tegally H, et al. SARS-CoV-2 Omicron has extensive but incomplete escape of Pfizer BNT162b2 elicited neutralization and requires ACE2 for infection. medRxiv. 2021:2021.12.08.21267417.  
+9. Hoffmann M, Kruger N, Schulz S, Cossmann A, Rocha C, Kempf A, et al. The Omicron variant is highly resistant against antibody-mediated neutralization – implications for control of the COVID-19 pandemic. bioRxiv. 2021:2021.12.12.472286.  
+10. Schmidt F, Muecksch F, Weisblum Y, Silva JD, Bednarski E, Cho A, et al. Plasma neutralization properties of the SARS-CoV-2 Omicron variant. medRxiv. 2021:2021.12.12.21267646.
+
+<--- Page Split --->
+
+11. Tenbusch M, Schumacher S, Vogel E, Priller A, Held J, Steininger P, et al. Heterologous prime-boost vaccination with ChAdOx1 nCoV-19 and BNT162b2. Lancet Infect Dis. 2021;21(9):1212-3. 
+
+12. Hillus D, Schwarz T, Tober-Lau P, Vanshylla K, Hastor H, Thibeault C, et al. Safety, reactogenicity, and immunogenicity of homologous and heterologous prime-boost immunisation with ChAdOx1 nCoV-19 and BNT162b2: a prospective cohort study. Lancet Respir Med. 2021;9(11):1255-65. 
+
+13. Flaxman A, Marchevsky NG, Jenkin D, Aboagye J, Aley PK, Angus B, et al. Reactogenicity and immunogenicity after a late second dose or a third dose of ChAdOx1 nCoV-19 in the UK: a substudy of two randomised controlled trials (COV001 and COV002). Lancet. 2021;398(10304):981-90. 
+
+14. Ramasamy MN, Minassian AM, Ewer KJ, Flaxman AL, Folegatti PM, Owens DR, et al. Safety and immunogenicity of ChAdOx1 nCoV-19 vaccine administered in a prime-boost regimen in young and old adults (COV002): a single-blind, randomised, controlled, phase 2/3 trial. Lancet. 2021;396(10267):1979-93. 
+
+15. Borobia AM, Carcas AJ, Perez-Olmeda M, Castano L, Bertran MJ, Garcia-Perez J, et al. Immunogenicity and reactogenicity of BNT162b2 booster in ChAdOx1-S-primed participants (CombiVacS): a multicentre, open-label, randomised, controlled, phase 2 trial. Lancet. 2021;398(10295):121-30. 
+
+16. Barros-Martins J, Hammerschmidt SI, Cossmann A, Odak I, Stankov MV, Morillas Ramos G, et al. Immune responses against SARS-CoV-2 variants after heterologous and homologous ChAdOx1 nCoV-19/BNT162b2 vaccination. Nat Med. 2021;27(9):1525-9. 
+
+17. Schmidt T, Klemis V, Schub D, Mihm J, Hielscher F, Marx S, et al. Immunogenicity and reactogenicity of heterologous ChAdOx1 nCoV-19/mRNA vaccination. Nat Med. 2021;27(9):1530-5. 
+
+18. Pozzetto B, Legros V, Djebali S, Barateau V, Guibert N, Villard M, et al. Immunogenicity and efficacy of heterologous ChAdOx1-BNT162b2 vaccination. Nature. 2021. 
+
+19. Behrens GMN, Cossmann A, Stankov MV, Witte T, Ernst D, Happle C, et al. Perceived versus proven SARS-CoV-2-specific immune responses in health-care professionals. Infection. 2020;48(4):631-4. 
+
+20. Bosnjak B, Stein SC, Willenzon S, Cordes AK, Puppe W, Bernhardt G, et al. Low serum neutralizing anti-SARS-CoV-2 antibody levels in mildly affected COVID-19 convalescent patients revealed by two different detection methods. Cell Mol Immunol. 2021;18(4):936-44. 
+
+21. Hoffmann M, Arora P, Gross R, Seidel A, Hornich BF, Hahn AS, et al. SARS-CoV-2 variants B.1.351 and P.1 escape from neutralizing antibodies. Cell. 2021;184(9):2384-93 e12. 
+
+22. Munro APS, Janani L, Cornelius V, Aley PK, Babbage G, Baxter D, et al. Safety and immunogenicity of seven COVID-19 vaccines as a third dose (booster) following two doses of ChAdOx1 nCoV-19 or BNT162b2 in the UK (COV-BOOST): a blinded, multicentre, randomised, controlled, phase 2 trial. Lancet. 2021;398(10318):2258-76. 
+
+23. Liu X, Shaw RH, Stuart ASV, Greenland M, Aley PK, Andrews NJ, et al. Safety and immunogenicity of heterologous versus homologous prime-boost schedules with an adenoviral vectored and mRNA COVID-19 vaccine (Com-COV): a single-blind, randomised, non-inferiority trial. Lancet. 2021;398(10303):856-69. 
+
+24. Madelon N, Heikkilä N, Sabater Royo I, Fontannaz P, Breville G, Lauper K, et al. Omicron-specific cytotoxic T-cell responses are boosted following a third dose of mRNA COVID-19 vaccine in anti-CD20-treated multiple sclerosis patients. medRxiv. 2021:2021.12.20.21268128. 
+
+25. Jablonka A, Happle C, Cossmann A, Stankov MV, Scharff AZ, Ernst D, et al. Protocol for longitudinal assessment of SARS-CoV-2-specific immune responses in healthcare professionals in Hannover, Germany: the prospective, longitudinal, observational COVID-19 Contact (CoCo) study. medRxiv. 2020:2020.12.02.20242479.
+
+<--- Page Split --->
+
+26. Hammerschmidt SI, Bosnjak B, Bernhardt G, Friedrichsen M, Ravens I, Dopfer-Jablonka A, et al. Neutralization of the SARS-CoV-2 Delta variant after heterologous and homologous BNT162b2 or ChAdOx1 nCoV-19 vaccination. Cell Mol Immunol. 2021;18(10):2455-6.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+
+<--- Page Split --->
+
+Fig. 1 | Participant recruitment schemes and humoral immune response. a, Participant recruitment and vaccination and blood sampling scheme. C, ChAd; B, BNT. b, A third heterologous or c, a third homologous immunization with BNT induce strong increases in anti S1 IgG5. Data are from \(n\) biologically independent samples as shown in the figure. d, A third heterologous immunization with BNT leads to increased frequencies of S-specific memory B cells. Data are from \(n\) biologically independent samples as shown in the figure. Statistics: b,c,d, Mixed effect analysis followed by Sidak's multiple comparison test (within groups) and unpaired t test with Welch's correction (between groups). The majority of the symbols depicted in grey had been published before (16, 26).
+
+<--- Page Split --->
+![](images/Extended_Data_Figure_1.jpg)
+
+<center>Fig. 2 | Heterologous vaccination induces neutralizing antibodies as well as CD4 and CD8 T cell responses. a, Heterologous ChAd/BNT/BNT or ChAd/ChAd/BNT vaccination induces neutralizing antibodies against Wuhan, B.1.1.7 (Alpha), B.1.351 (Beta), P.1 (B.1.1.28.1; Gamma), B.1.617.2 (Delta) and the B.1.1.529 (Omicron) SARS-CoV-2-S variants measured using the sVNT. Data are from \(n =\) biologically independent samples as indicated. For better visualization of identical titer values, data were randomly and proportionally adjusted closely around the precise titer results. Boost vaccination increased total percentage of cytokine-secreting CD4+ (b) and CD8+ (c) T cells. We calculated the </center>
+
+<--- Page Split --->
+
+total number of cytokine secreting cells as the sum of IFN- \(\gamma +\) TNF- \(\alpha -\) , IFN- \(\gamma +\) TNF- \(\alpha +\) and IFN- \(\gamma -\) TNF- \(\alpha +\) cells in the gates indicated in Extended Data Fig. 4. Data are from \(n\) biologically independent samples as indicated.  
+
+d, IFN- \(\gamma\) concentration in full blood supernatants after stimulation with SARS- CoV- 2 S1 domain for 20- 24 h measured in duplicate by IGRA (Euroimmun). Data are from \(n\) biologically independent samples as indicated.  
+
+a, b, c and d. Mixed effect analysis followed by Sidak's multiple comparison test (within groups) and unpaired t test with Welch's correction (between groups). The majority of the symbols depicted in grey had been published before (16, 26).
+
+<--- Page Split --->
+
+# Extended Data
+
+<table><tr><td rowspan="2"></td><td rowspan="2">Mean age,<br>years<br>(range)</td><td rowspan="2">Sex, m/f<br>(%)</td><td colspan="6">Median (IQR) days past last vaccination</td></tr><tr><td>1st vaccination</td><td>2 mo</td><td>14 d</td><td>4 mo</td><td>6 mo</td><td>14 d</td></tr><tr><td>ChAd<br>ChAd<br>BNT<br>n=41</td><td>40<br>(21-64)</td><td>14/27<br>(34/66)</td><td rowspan="3">1st vaccination</td><td>68<br>(12.75)</td><td rowspan="3">14<br>(4)</td><td>15<br>(13)</td><td>119<br>(5)</td><td rowspan="3">196<br>(6.5)</td></tr><tr><td>ChAd<br>BNT<br>n=82</td><td>37<br>(19-61)</td><td>17/65<br>(21/79)</td><td>70<br>(8)</td><td>17<br>(5)</td><td>117<br>(13)</td><td>195<br>(8)</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr></table>
+
+<table><tr><td rowspan="2"></td><td rowspan="2">Mean age,<br>years (range)</td><td rowspan="2">Sex, m/f<br>(%)</td><td colspan="6">Median (IQR) days past last vaccination</td></tr><tr><td></td><td>21 d</td><td>1 mo</td><td>7 mo</td><td>9 mo</td><td>21 d</td></tr><tr><td>BNT<br>BNT<br>BNT<br>n=57</td><td>42<br>(23-63)</td><td>21/36<br>(38/62)</td><td rowspan="2">1st vaccination</td><td>20<br>(1.25)</td><td rowspan="2">29<br>(8.25)</td><td>211<br>(9)</td><td>267<br>(22.5)</td><td rowspan="2">23<br>(10.2<br>5)</td></tr><tr><td></td><td></td><td></td><td>2nd vaccination</td><td></td><td></td></tr></table>
+
+**Extended Data Table. 1 |** Demographic data and median time in days since last vaccination for the five blood collection time points (months = mo and days = d) of the three vaccination groups.
+
+<--- Page Split --->
+
+a
+
+<table><tr><td>Antigen</td><td>Conjugate</td><td>Clone</td><td>Orde no.</td><td>Company</td><td>Dilution</td></tr><tr><td>CD14</td><td>BB700</td><td>MP9</td><td>566465</td><td>BD</td><td>100</td></tr><tr><td>CD16</td><td>BUV496</td><td>3G8</td><td>612944</td><td>BD</td><td>100</td></tr><tr><td>CD19</td><td>PECy7</td><td>HIB19</td><td>982410</td><td>BioLegend</td><td>200</td></tr><tr><td>CD20</td><td>BV421</td><td>2H7</td><td>302330</td><td>BioLegend</td><td>100</td></tr><tr><td>CD27</td><td>BUV805</td><td>L128</td><td>748704</td><td>BD</td><td>100</td></tr><tr><td>CD38</td><td>PerCP-eF710</td><td>HB7</td><td>46-0388-42</td><td>Invitrogen</td><td>100</td></tr><tr><td>IgD</td><td>BV480</td><td>IA6-2</td><td>566138</td><td>BD</td><td>200</td></tr><tr><td>IgM</td><td>AF647</td><td>MHM-88</td><td>314436</td><td>BioLegend</td><td>100</td></tr><tr><td>Viability</td><td>Zombie NIRTM</td><td>-</td><td>423106</td><td>BioLegend</td><td>400</td></tr><tr><td>Anti-S BCR</td><td>mNeonGreen</td><td colspan="3">Produced by T.Krey</td><td>\(5\mu L/sample\)</td></tr></table>
+
+![](images/Extended_Data_Figure_5.jpg)
+
+<center>Extended Data Fig. 1 | Antibody panel a and gating strategy b for SARS-CoV-2-S (Spike)-specific B cell populations in blood. Pseudocolor plots show representative data from a female donor<br>283 days after priming with ChAd; 213 days after a second dose with BNT and 14 days after a third dose with BNT.</center>
+
+<--- Page Split --->
+![PLACEHOLDER_19_0]
+  
+
+Extended Data Fig. 2 | Antibody neutralization measurements against the Omicron SARS- CoV- 2 variant is positively correlated between the virus neutralization tests (sVNT) and pseudotyped virus neutralization tests (pVNT). Correlation (solid line) and 95% confidence intervals (dotted lines) between sVNT1:20 and antibody titers resulting in 50% reduction of luciferase activity in pVNT, indicated as pVNT50. Open circles, values from individual donors, outliers are marked with X and were defined as values with absolute residual value \(> 2\) SD of all residual values in each group of samples. Correlation was calculated using single linear regression.
+
+<--- Page Split --->
+![PLACEHOLDER_20_0]
+  
+
+Extended Data Fig. 3 | Humoral immune response against all SARS- CoV- 2 variants following homologous BNT162b2 (BNT) / BNT /BNT vaccination. Reciprocal titers of neutralizing antibodies against Wuhan, B.1.1.7 (Alpha), B.1.351 (Beta), P.1 (B.1.1.28.1; Gamma), B.1.617.2 (Delta) and the B.1.1.529 (Omicron) SARS- CoV- 2- S variants measured using the sVNT. Data are from \(n =\) biologically independent samples as indicated. Mixed effect analysis followed by Sidak's multiple comparison test (within groups). For better visualization of identical titer values, data were randomly and proportionally adjusted closely around the precise titer results. The majority of the symbols depicted in grey had been published before (16, 26).
+
+<--- Page Split --->
+![PLACEHOLDER_21_0]
+  
+
+Extended Data Fig. 4 | Gating strategy used for detection of cytokine producing CD4+ and CD8+ T cells after ex vivo re- stimulation with DMSO or the pool of Spike- specific peptides for 12–16 hr.
+
+<--- Page Split --->
+![PLACEHOLDER_22_0]
+
+<center>Extended Data Fig. 5 | Frequency of cytokine-producing CD4<sup>+</sup> T cells and CD8<sup>+</sup> T cells after ex vivo re-stimulation with DMSO or the pool of Spike-specific peptides for 12–16 hr.</center>
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- flatBehrensrs2.pdf
+
+<--- Page Split --->

preprint/preprint__48fdc4dc3cce1523cd0346c874c274f187408b5762031fb3d036d6b4eed72790/preprint__48fdc4dc3cce1523cd0346c874c274f187408b5762031fb3d036d6b4eed72790_det.mmd

@@ -0,0 +1,477 @@
+<|ref|>title<|/ref|><|det|>[[44, 106, 936, 208]]<|/det|>
+# BNT162b2 boosted immune responses six months after heterologous or homologous ChAdOx1nCoV-19/BNT162b2 vaccination against COVID-19  
+
+<|ref|>text<|/ref|><|det|>[[44, 228, 789, 270]]<|/det|>
+Georg Behrens ( \(\boxed{ \begin{array}{r l} \end{array} }\) Behrens.georg@mh- hannover.de) Medizinische Hochschule Hannover (MHH) https://orcid.org/0000- 0003- 3111- 621X  
+
+<|ref|>text<|/ref|><|det|>[[44, 275, 728, 316]]<|/det|>
+Joana Barros- Martins Medizinische Hochschule Hannover https://orcid.org/0000- 0002- 3370- 4990  
+
+<|ref|>text<|/ref|><|det|>[[44, 322, 281, 363]]<|/det|>
+Anne Cossmann Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 369, 639, 410]]<|/det|>
+Gema Morillas Ramos Hannover Medical School https://orcid.org/0000- 0002- 4835- 5172  
+
+<|ref|>text<|/ref|><|det|>[[44, 415, 281, 456]]<|/det|>
+Metodi Stankov Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 462, 281, 503]]<|/det|>
+Ivan Odak Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 508, 281, 549]]<|/det|>
+Alexandra Dopfer- Jablonka Hannover Medical School https://orcid.org/0000- 0001- 7129- 100X  
+
+<|ref|>text<|/ref|><|det|>[[44, 555, 281, 595]]<|/det|>
+Laura Hetzel Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 601, 281, 641]]<|/det|>
+Miriam Köhler Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 648, 281, 688]]<|/det|>
+Gwendolyn Patzer Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 694, 281, 734]]<|/det|>
+Christoph Binz Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 740, 281, 780]]<|/det|>
+Christiane Ritter Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 787, 281, 827]]<|/det|>
+Michaela Friedrichsen Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 833, 281, 873]]<|/det|>
+Christian Schultze- Florey Hannover Medical School https://orcid.org/0000- 0002- 3307- 2639  
+
+<|ref|>text<|/ref|><|det|>[[44, 879, 281, 920]]<|/det|>
+Inga Ravens Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 925, 209, 944]]<|/det|>
+Stefanie Willenzon
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[52, 46, 280, 64]]<|/det|>
+Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 71, 280, 110]]<|/det|>
+Anja Bubke Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 116, 280, 156]]<|/det|>
+Jasmin Ristenpart Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 163, 280, 202]]<|/det|>
+Anika Janssen Hannover Medical School  
+
+<|ref|>text<|/ref|><|det|>[[44, 209, 235, 248]]<|/det|>
+George Ssebyatika University of Lübeck  
+
+<|ref|>text<|/ref|><|det|>[[44, 255, 636, 295]]<|/det|>
+Günter Bernhardt Hannover Medical School https://orcid.org/0000- 0002- 0510- 2853  
+
+<|ref|>text<|/ref|><|det|>[[44, 301, 617, 342]]<|/det|>
+Markus Hoffmann German Primate Center https://orcid.org/0000- 0003- 4603- 7696  
+
+<|ref|>text<|/ref|><|det|>[[44, 348, 951, 389]]<|/det|>
+Stefan Pöhlmann German Primate Center - Leibniz Institute for Primate Research https://orcid.org/0000- 0001- 6086- 9136  
+
+<|ref|>text<|/ref|><|det|>[[44, 395, 590, 435]]<|/det|>
+Thomas Krey University of Lübeck https://orcid.org/0000- 0002- 4548- 7241  
+
+<|ref|>text<|/ref|><|det|>[[44, 441, 636, 481]]<|/det|>
+Berislav Bosnjak Hannover Medical School https://orcid.org/0000- 0003- 3374- 7488  
+
+<|ref|>text<|/ref|><|det|>[[44, 487, 636, 527]]<|/det|>
+Swantje Hammerschmidt Hannover Medical School https://orcid.org/0000- 0003- 0718- 0422  
+
+<|ref|>text<|/ref|><|det|>[[44, 533, 941, 574]]<|/det|>
+Reinhold Forster Institute of Immunology, Hannover Medical School, Hannover https://orcid.org/0000- 0001- 6190- 7923  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 616, 101, 634]]<|/det|>
+## Article  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 654, 135, 672]]<|/det|>
+## Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 692, 344, 711]]<|/det|>
+Posted Date: December 29th, 2021  
+
+<|ref|>text<|/ref|><|det|>[[44, 730, 473, 749]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 1200506/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 767, 910, 810]]<|/det|>
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 846, 930, 889]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on August 18th, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 32527- 2.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[115, 84, 857, 164]]<|/det|>
+# BNT162b2 boosted immune responses six months after heterologous or homologous ChAdOx1nCoV-19/BNT162b2 vaccination against COVID-19  
+
+<|ref|>text<|/ref|><|det|>[[115, 214, 880, 316]]<|/det|>
+Georg M. N. Behrens<sup>1,2,3,\*</sup>, Joana Barros- Martins<sup>4\*</sup>, Anne Cossmann<sup>1\*</sup>, Gema Morillas Ramos<sup>1\*</sup>, Metodi V. Stankov<sup>1</sup>, Ivan Odak<sup>2</sup>, Alexandra Dopfer- Jablonka<sup>1,2</sup>, Laura Hetzel<sup>1</sup>, Miriam Köhler<sup>4</sup>, Gwendolyn Patzer<sup>4</sup>, Christoph Binz<sup>4</sup>, Christiane Ritter<sup>4</sup>, Michaela Friedrichsen<sup>4</sup>, Christian Schultze- Florey<sup>4,5</sup>, Inga Ravens<sup>4</sup>, Stefanie Willenzon<sup>4</sup>, Anja Bubke<sup>4</sup>, Jasmin Ristenpart<sup>4</sup>, Anika Janssen<sup>4</sup>, George Ssebyatika<sup>6</sup>, Günter Bernhardt<sup>4</sup>, Markus Hoffmann<sup>7,8</sup>, Stefan Pöhlmann<sup>7,8</sup>, Thomas Krey<sup>5,9</sup>, Berislav Bošnjak<sup>4</sup>, Swantje I. Hammerschmidt<sup>4\*</sup>, Reinhold Förster<sup>2,4,10,\*</sup>  
+
+<|ref|>text<|/ref|><|det|>[[115, 345, 199, 359]]<|/det|>
+Affiliations:  
+
+<|ref|>text<|/ref|><|det|>[[115, 361, 827, 400]]<|/det|>
+<sup>1</sup> Department for Rheumatology and Immunology, Hannover Medical School, 30625 Hannover, Germany  
+
+<|ref|>text<|/ref|><|det|>[[115, 393, 805, 425]]<|/det|>
+<sup>2</sup> German Center for Infection Research (DZIF), Partner Site Hannover- Braunschweig, 30625 Hannover, Germany  
+
+<|ref|>text<|/ref|><|det|>[[115, 425, 662, 440]]<|/det|>
+<sup>3</sup> CiiM, Centre for Individualized Infection Medicine, Hannover, Germany  
+
+<|ref|>text<|/ref|><|det|>[[115, 442, 722, 457]]<|/det|>
+<sup>4</sup> Institute of Immunology, Hannover Medical School, 30625 Hannover, Germany  
+
+<|ref|>text<|/ref|><|det|>[[115, 458, 822, 488]]<|/det|>
+<sup>5</sup> Department of Hematology, Hemostasis, Oncology and Stem- Cell Transplantation, Hannover Medical School, 30625 Hannover, Germany  
+
+<|ref|>text<|/ref|><|det|>[[115, 489, 666, 504]]<|/det|>
+<sup>6</sup> Institute of Biochemistry, University of Lübeck, 23562 Lübeck, Germany  
+
+<|ref|>text<|/ref|><|det|>[[115, 505, 690, 520]]<|/det|>
+<sup>7</sup> Infection Biology Unit, German Primate Center, 37077 Göttingen, Germany  
+
+<|ref|>text<|/ref|><|det|>[[115, 521, 870, 536]]<|/det|>
+<sup>8</sup> Faculty of Biology and Psychology, Georg- August- University Göttingen, 37073 Göttingen, Germany  
+
+<|ref|>text<|/ref|><|det|>[[115, 537, 870, 567]]<|/det|>
+<sup>9</sup> German Center for Infection Research (DZIF), Partner Site Hannover- Braunschweig and Partner Site Hamburg- Lübeck- Borstel- Riems, Hamburg, Germany  
+
+<|ref|>text<|/ref|><|det|>[[115, 568, 840, 599]]<|/det|>
+<sup>10</sup> Cluster of Excellence RESIST (EXC 2155), Hannover Medical School, 30625 Hannover, Germany\* These authors contributed equally  
+
+<|ref|>text<|/ref|><|det|>[[115, 641, 295, 656]]<|/det|>
+Corresponding authors:  
+
+<|ref|>text<|/ref|><|det|>[[115, 668, 880, 715]]<|/det|>
+Georg M.N. Behrens, Department of Rheumatology and Immunology, Hannover Medical School, Carl- Neuberg- Straße 1, D - 30625 Hannover, Germany, Tel: +49 511 532 5337, Fax: +49 511 532 5324, Email: behrens.georg@mh- hannover.de  
+
+<|ref|>text<|/ref|><|det|>[[115, 727, 420, 743]]<|/det|>
+https://orcid.org/0000- 0003- 3111- 621X  
+
+<|ref|>text<|/ref|><|det|>[[115, 755, 848, 803]]<|/det|>
+Reinhold Förster, Institute for Immunology, Hannover Medical School, Carl- Neuberg- Straße 1, D - 30625 Hannover, Germany, Tel: +49 511 532 9721, Fax: +49 511 532 9722, Email: foerster.reinhold@mh- hannover.de  
+
+<|ref|>text<|/ref|><|det|>[[115, 815, 420, 830]]<|/det|>
+https://orcid.org/0000- 0001- 6190- 7923
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[117, 84, 184, 98]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[115, 109, 880, 383]]<|/det|>
+AbstractReports suggest that COVID- 19 vaccine effectiveness is decreasing, either due to waning immune protection, emergence of new variants of concern, or both. Heterologous prime/boost vaccination with a vector- based approach (ChAdOx- 1nCov- 19, ChAd) followed by an mRNA vaccine (e.g. BNT162b2, BNT) appeared to be superior in inducing protective immunity, and large scale second booster vaccination is ongoing. However, data comparing declining immunity after homologous and heterologous vaccination as well as effects of a third vaccine application after heterologous ChAd/BNT vaccination are lacking. We longitudinally monitored immunity in ChAd/ChAd (n=41) and ChAd/BNT (n=88) vaccinated individuals and assessed the impact of a second booster with BNT in both groups. The second booster greatly augmented waning anti- spike IgG but only moderately increased spike- specific \(\mathrm{CD4^{+}}\) and \(\mathrm{CD8^{+}}\) T cells in both groups to cell frequencies already present after the boost. More importantly, the second booster efficiently restored neutralizing antibody responses against Alpha, Beta, Gamma, and Delta, but neutralizing activity against B.1.1.529 (Omicron) stayed severely impaired. Our data suggest that inferior SARS- CoV- 2 specific immune responses after homologous ChAd/ChAd vaccination can be cured by a heterologous BNT vaccination. However, prior heterologous ChAd/BNT vaccination provides no additional benefit for spike- specific T cell immunity or neutralizing Omicron after the second boost.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 421, 193, 435]]<|/det|>
+## Main text  
+
+<|ref|>text<|/ref|><|det|>[[115, 446, 877, 602]]<|/det|>
+While the COVID- 19 vaccines currently approved by the European Medicine Agency (EMA) and U.S. Food and Drug Administration (FDA) provide high levels of protection against severe illness, the emergence of the Delta variant resulted in increasing numbers of breakthrough infections in fully vaccinated individuals (1). This coincided with evidence of waning immunity in vaccinated individuals (2, 3). Thus, booster vaccinations were proposed to reconstitute immunity and to potentially expand the breadth of immunity against SARS- CoV- 2 variants of concern (Voc). Also policy makers have begun to promote booster vaccination not only for vulnerable patients but also to mitigate healthcare and economic impact. Real world data confirm that booster vaccination is effective in preventing COVID- 19 (4- 6).  
+
+<|ref|>text<|/ref|><|det|>[[115, 611, 872, 732]]<|/det|>
+Concomitant to booster vaccination campaigns, the novel Omicron variant was recently identified in South Africa and its emergence was associated with a steep increase in cases and hospitalizations. Omicron is spreading rapidly in European, African and Asian countries as well as the USA, with case numbers doubling every two to three days (7). The S protein of the Omicron variant harbors an unusually high number of mutations, which increases immune evasion and potentially transmissibility (8- 11). Thus, the Omicron variant constitutes a rapidly emerging threat to public health and might undermine global efforts to control the COVID- 19 pandemic.  
+
+<|ref|>text<|/ref|><|det|>[[115, 742, 875, 878]]<|/det|>
+Heterologous prime boost strategies appear to offer immunological advantages to strengthen protection against COVID- 19 achieved with currently available vaccines. Administration of mRNA vaccines like BNT162b2 (Comirnaty; BNT) after the initial ChAdOx1- nCov- 19 (Vaxzevria, ChAd) dose as the second dose of a two- dose regimen was safe and had enhanced immunogenicity compared to homologous ChAdOx vaccination (12- 18). We have previously reported on the results after homologous and heterologous vaccination after ChAd priming (16). Here we present findings from a subsequent analysis assessing the efficacy of a second booster vaccination after heterologous and homologous prime- boost vaccination on neutralization of Voc including Omicron.  
+
+<|ref|>text<|/ref|><|det|>[[115, 889, 866, 922]]<|/det|>
+In addition to our previously reported findings, we longitudinally monitored immunity after prime- boost COVID- 19 vaccine treatment schedules and determined thereafter the impact of BNT booster
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 857, 255]]<|/det|>
+(Methods). Health care professional vaccines without previous SARS- CoV- 2 infection, who had received ChAd/ChAd or ChAd/BNT, donated further blood four and six months after first booster vaccination and about two weeks after the second booster. The vaccination and blood collection schedule is depicted in Fig. 1A with additional demographic information (age and sex) in Extended Data Table 1. A third group of BNT/BNT vaccines served as an independent control group for serologic analysis only and was monitored for up to nine months (Extended Data Table 1). As described (16), anti- SARS- CoV- 2 spike IgG (anti- S IgG) were significantly higher in the ChAd/BNT group short after prime- boost vaccination when compared to the ChAd/ChAd group but declined significantly over time in both groups, with lower anti- S IgG after homologous vaccination prior to the second boost (Fig. 1B).  
+
+<|ref|>text<|/ref|><|det|>[[115, 264, 870, 400]]<|/det|>
+Following second booster immunization, we found greatly increased anti- S IgG responses in both groups. Boosting of the heterologous ChAd/BNT immunized group led to a significant 47.9 - fold increase for anti- S IgG \((p< 0.0001)\) and 8.0- fold increase in individuals after homologous ChAd vaccination \((p< 0.0001)\) (Fig. 1B). In both groups, anti- S IgG were considerably higher as compared to the first boost time point. More importantly, the second booster diminished previous differences between the heterologous and homologous prime- boost vaccination groups, since anti- S IgG were comparable in both groups after the second BNT boost and were within the range of triple BNT vaccinated individuals (Fig. 1C).  
+
+<|ref|>text<|/ref|><|det|>[[115, 410, 863, 548]]<|/det|>
+Next, we measured the frequency and phenotype of memory B cells carrying membrane- bound immunoglobulins specific for the Spike protein (Methods, Extended Data Fig. 1) over time. Interestingly, numbers of spike- specific memory B cells generated after prime- boost vaccination gradually increased during the following months with no significant difference between the ChAd/ChAd and the ChAd/BNT group (Fig. 1D). Again, the second booster with BNT led to a further and significant expansion of spike- specific memory B cells in both groups (Fig. 1D) in line with increased amounts of spike- specific antibodies, highlighting the impact of the second booster vaccination for better protection from SARS- CoV- 2 infection.  
+
+<|ref|>text<|/ref|><|det|>[[115, 558, 866, 713]]<|/det|>
+For testing neutralizing activity of antibodies induced by vaccination, we employed our ELISA- based surrogate virus neutralization test (sVNT). We modified the sVNT to include Spike proteins of the Omicron variant (Methods) and for validation applied sera from vaccines that had been recently tested for their neutralizing capacity applying vesicular stomatitis virus (VSV)- based pseudotyped virus neutralization assays (pVNT) (9, 20, 21). As for other VoC (16), we obtained a high degree of correlation between both assays with a R square value of 0.7044 (Extended Data Fig. 2). Thus, the sVNT is a reliable tool to quantitatively assess the neutralization capacity of vaccination- induced antibodies not only against the Wuhan but also against the Alpha, Beta, Gamma, Delta and Omicron variants of SARS- CoV- 2.  
+
+<|ref|>text<|/ref|><|det|>[[115, 723, 872, 860]]<|/det|>
+Using the sVNT assays and consistent with declining anti- S IgG, we confirmed waning neutralizing activity against the Wuhan variant and particularly against VoC tested. Whilst the majority of participants had neutralizing antibodies against the Wuhan strain in pre- second boost plasma, neutralizing antibodies against the Alpha, Beta, Gamma and Delta variants were particularly in the ChAd/ChAd group less frequent or virtually absent (Fig. 2A). At 2 weeks after the second booster immunization, frequencies and titers of neutralizing antibodies against the Wuhan strain increased profoundly in the ChAd/BNT and ChAd/ChAd group with titers reaching values above those after the initial two injections in the latter group (Fig. 2A).  
+
+<|ref|>text<|/ref|><|det|>[[117, 870, 877, 920]]<|/det|>
+Differences between the pre- booster vaccination regimens became even more evident when analyzing the neutralization capacity of antibodies induced against the VoC. In the ChAd/ChAd group, second booster immunization profoundly increased neutralization of the Alpha, Beta, Gamma and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 881, 288]]<|/det|>
+Delta variants (Fig. 2A), which was low for Alpha and Delta after prime/boost and virtually absent for Beta and Gamma. Whilst initial ChAd/BNT immunization had induced neutralizing antibodies at high levels against all analyzed VoC, the following decline was more than restored by the second booster (Fig. 2A). In fact, nearly all BNT boosted ChAd/BNT vaccines had efficient neutralizing activity against Alpha, Beta, Gamma and Delta and amounts mostly above those from after the first booster. Importantly, the neutralization capacity against the Omicron variant was virtually absent before the second booster and remained low thereafter in comparison to the other VoC (Fig. 2A), with 9/29 (31%) and 27/58 (47%) of vaccines in the ChAd/ChAd and ChAD/BNT group, respectively having no detectable neutralization activity after boost. We obtained very similar results after BNT booster in BNT/BNT vaccinated individuals (Extended Data Fig. 3). Altogether, these data indicate that the second booster immunization led to an increase of neutralizing antibodies in both vaccination groups against all tested VoC including the Omicron variant to only some extent.  
+
+<|ref|>text<|/ref|><|det|>[[115, 298, 881, 609]]<|/det|>
+Finally, we also analyzed frequencies and phenotypes of spike- specific T cells (Methods, Extended Data Fig. 4 and 5). We quantified numbers of spike- specific T cells as the sum of all cells producing IFN- \(\gamma\) or TNF- \(\alpha\) as described previously (16). The frequencies of spike- specific \(\mathrm{CD4^{+}}\) and \(\mathrm{CD8^{+}}\) T cells in blood samples collected after the first booster vaccination were significantly higher in the ChAd/BNT group (Fig. 2B and C) and both cell populations declined over time, while they remained more or less stable after homologous vaccination. Whilst spike- specific \(\mathrm{CD4^{+}}\) T cells declined to frequencies similar to individuals after homologous ChAd/ChAd vaccination (Fig. 2B), spike- specific \(\mathrm{CD8^{+}}\) T cells remained above the frequencies of the heterologous vaccinated group (Fig. 2C). More interestingly, heterologous BNT booster in the ChAd/ChAd vaccination group significantly raised numbers of spike- specific \(\mathrm{CD4^{+}}\) T cells above amounts observed after the first boost. In contrast, individuals with heterologous ChAd/BNT vaccination only regained spike- specific \(\mathrm{CD4^{+}}\) T cell levels corresponding to levels after the first boost (Fig. 2B). Similarly, reboot with BNT did not result in an expansion of spike- specific \(\mathrm{CD8^{+}}\) T cells above levels after first boost in ChAd/BNT vaccines, but did so in ChAd/ChAd vaccinated individuals (Fig. 2C). Like for spike- specific \(\mathrm{CD4^{+}}\) T cells, raised numbers in spike- specific IFN- \(\gamma\) - producing T cells in the ChAd/ChAd as well as the ChAd/BNT after BNT boost group was confirmed by cytokine measurement in supernatants after SARS- CoV- 2 spike peptide stimulation (Fig. 2D). Again, BNT boost did not further increase spike- specific IFN- \(\gamma\) - producing T cells in ChAd/BNT vaccinated subjects above levels obtained already after first boost vaccination.  
+
+<|ref|>text<|/ref|><|det|>[[115, 619, 880, 929]]<|/det|>
+BNT booster potently increased anti- S IgG in all heterologous and homologous vaccinated individuals tested and this rise was accompanied by further strengthened neutralizing capacity against the Wuhan variant and Alpha, Beta, Gamma, and Delta. These data corroborate findings after homologous vaccination (22) and support current recommendations by the European Medicines Agency (EMA) and the European Centre for Disease Prevention and Control (ECDC). However, neutralization of the Omicron variant was absent before second booster and remained clearly inferior thereafter, irrespective of the previous vaccination scheme. Considering the kinetics of waning neutralizing antibodies against the other VoC after the first booster, we expect remaining neutralization against Omicron to vanish rapidly in the majority of vaccines despite persisting high anti- S IgG concentrations. This suggests that variant- specific vaccines are required and need to be tested soon to better combat COVID- 19 caused by Omicron and other emerging variants. In contrast, the second BNT booster only made up for absent rise in spike- specific T cell responses after homologous ChAd/ChAd vaccination and merely restored spike- specific \(\mathrm{CD4^{+}}\) T cell responses after heterologous vaccination. These data confirm and extent reports that ChAd does not boost cellular responses after ChAd/ChAd vaccination (13, 23), but we are yet unable to make conclusion about the long- term protection and immunological memory. Although the relative role for T cell immunity remains unclear, spike- specific T memory cells are probably of great importance for protection against severe COVID- 19, hospitalization and death. Our data reveal that additional BNT booster
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[116, 82, 870, 168]]<|/det|>
+poorly support spike- specific \(\mathrm{CD8^{+}}\) expansion and suggest that novel vaccines and vaccine schedules should be explored for further strengthening of adaptive cellular immunity against SARS- CoV- 2 and its variants (24). Such vaccines should also aim to target other structural viral proteins including nucleocapsid and membrane proteins, which are less likely to be able to escape from capable immune recognition.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 205, 269, 219]]<|/det|>
+## Acknowledgements  
+
+<|ref|>text<|/ref|><|det|>[[115, 229, 883, 448]]<|/det|>
+This work was supported by the German Center for Infection Research TTU 01.938 (grant no 80018019238 to G.M.N.B and R.F.), and TTU 04.820 to G.M.N.B., by Deutsche Forschungsgemeinschaft, (DFG, German Research Foundation) Excellence Strategy EXC 2155 "RESIST" (Project ID39087428 to R.F.), by funds of the State of Lower Saxony (14- 76103- 184 CORONA- 11/20 to R.F.), by funds of the BMBF (NaFoUniMedCovid19 FKZ: 01KX2021; Projects B- FAST to R.F.) and Deutsche Forschungsgemeinschaft, SFB 900/3 (Projects B1, 158989968 to R.F.), and the European Regional Development Fund (Defeat Corona, ZW7- 8515131 and ZW7- 85151373 to G.M.N.B.). We thank the CoCo Study participants for their support and the entire CoCo study team for help. We would like to thank Luis Manthey, Annika Heidemann, Till Redeker, Madeleine Rommel, Christian Sturm, Marie Mikuteit, Jacqueline Niewolik, Ruth Sikora, Janine Topal, Kerstin Sträche, Birgit Heinisch, Michael Stephan, Mariel Nöhre, Simone Müller, Olivera Dragicevic, Kim Do Thi Hoang, Amy Kempf, and Inga Nehlmeier for technical and logistical support.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 482, 366, 497]]<|/det|>
+## Author Contributions Statement  
+
+<|ref|>text<|/ref|><|det|>[[117, 507, 350, 522]]<|/det|>
+Study design: G.M.N.B and R.F.  
+
+<|ref|>text<|/ref|><|det|>[[115, 531, 880, 566]]<|/det|>
+Data collection: J.B.- M., S.I.H., A.C., I.O., M.V.S., G.M.R., A.D.- J., L.H., M.K, G.P., C.B., C.R., M.F., C. S.- F., I.R., S.W., A.B., J.M., J.R., A.J., G.S., G.B., J.M., M.H., S.P., T.K.  
+
+<|ref|>text<|/ref|><|det|>[[115, 576, 666, 592]]<|/det|>
+Data analysis: G.M.N.B, J.B.- M., S.I.H., A.C., I.O., G.M.R., M.H., B.B, M.V.S.  
+
+<|ref|>text<|/ref|><|det|>[[117, 602, 377, 617]]<|/det|>
+Data interpretation: R.F., G.M.N.B.  
+
+<|ref|>text<|/ref|><|det|>[[115, 627, 538, 642]]<|/det|>
+Writing: G.M.N.B., R.F. with comments from all authors.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 678, 359, 694]]<|/det|>
+## Competing Interests Statement  
+
+<|ref|>text<|/ref|><|det|>[[117, 704, 450, 719]]<|/det|>
+The authors declare no competing interests.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[117, 83, 187, 97]]<|/det|>
+## Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 110, 208, 125]]<|/det|>
+## Participants  
+
+<|ref|>text<|/ref|><|det|>[[116, 136, 882, 273]]<|/det|>
+ParticipantsParticipants for this analysis were from the COVID- 19 Contact (CoCo) Study (German Clinical Trial Registry, DRKS00021152), an ongoing, prospective observational study monitoring anti- SARS- CoV- 2 IgG immunoglobulin and immune responses in health care professionals (HCP) at Hannover Medical School and individuals with potential contact to SARS- CoV- 2 (19, 25). An amendment from Dec 2020 allowed us to study the immune responses after COVID- 19 vaccination. We followed the study cohort described previously (16) after heterologous ChAdOx/BNT or homologous ChAdOx/ ChAdOx and BNT/BNT vaccination. Scheduling appointments for a third booster vaccination with BNT was coordinated by an independent vaccination team according to vaccine availability.  
+
+<|ref|>text<|/ref|><|det|>[[116, 283, 878, 402]]<|/det|>
+One individual with previous SARS- CoV- 2 infection as determined by positive anti- SARS- CoV- 2 NCP IgG before vaccinations was excluded from this analysis. Two additional individuals of the ChAdOx/ ChAdOx group developed anti- SARS- CoV- 2 NCP IgG after prime/boost vaccination and were excluded from follow up analysis. Demographics (sex and age) are depicted in Extended Data Table 1. After blood collection, we separated plasma from EDTA or lithium heparin blood (S- Monovette, Sarstedt) and stored it at \(- 80^{\circ}C\) until use. We used full blood or isolated PBMCs from whole blood samples by Ficoll gradient centrifugation and for stimulation with SARS- CoV- 2 peptide pools.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 413, 465, 428]]<|/det|>
+## Pseudotyped virus neutralization assay (pVNT)  
+
+<|ref|>text<|/ref|><|det|>[[115, 439, 877, 698]]<|/det|>
+pVNTs were performed at the Infection Biology Unit of the German Primate Center in Gottingen as described recently (21). Briefly, the rhabdoviral pseudotyped particles were produced in 293T cells transfected to express the desired SARS- CoV- 2- S variant inoculated with VSV\*DG- Fluc, a replication- deficient VSV vector that encodes for enhanced green fluorescent protein and firefly luciferase (Fluc) instead of VSV- G protein (kindly provided by Gert Zimmer, Institute of Virology and Immunology, Mittelhäusern, Switzerland). Produced pseudoparticles were collected, cleared from cellular debris by centrifugation and stored at \(- 80^{\circ}C\) until used. For neutralization experiments, equal volumes of pseudotyped particles and heat- inactivated (56 \(^\circ C\) , 30 min) plasma samples serially diluted in culture medium were mixed and incubated for 30 min at \(37^{\circ}C\) . Afterwards, the samples together with non- plasma- exposed pseudotyped particles were used for transduction experiments. The assay was performed in 96- well plates in which Vero cells were inoculated with the respective pseudotyped particles/plasma mixtures. The transduction efficacy was analyzed at 16- 18 hr post inoculation by measuring Fluc activity in lysed cells (Cell culture lysis reagent, Promega) using a commercial substrate (Beetle- Juice, PJK) and a plate luminometer (Hidex Sense Plate Reader, Hidex) with the Hidex Sense Microplate Reader Software (version 0.5.41.0).  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 709, 184, 723]]<|/det|>
+## Serology  
+
+<|ref|>text<|/ref|><|det|>[[116, 734, 881, 871]]<|/det|>
+We measured SARS- CoV- 2 IgG by quantitative ELISA (anti- SARS- CoV- 2 S1 Spike protein domain/receptor binding domain IgG SARS- CoV- 2- QuantivVac, Euroimmun, Lubeck, Germany) according to the manufacturer's instructions (dilution up to 1:4000). We provide anti- S1 concentrations expressed as RU/mL as assessed from a calibration curve with values above 11 RU/mL defined as positive. These values can be converted in binding antibody units (BAU/mL) by multiplying RU/mL by 3.2. We performed anti SARS- CoV- 2 nucleocapsid (NCP) IgG measurements according to the manufacturer's instructions (Euroimmun, Lubeck, Germany). We used an AESKU.READER (AESKU.GROUP, Wendelsheim, Germany) and the Gen5 2.01 Software for analysis.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[117, 83, 622, 98]]<|/det|>
+## Surrogate virus neutralization assay (sVNT) for SARS-CoV-2 variants  
+
+<|ref|>text<|/ref|><|det|>[[115, 108, 881, 644]]<|/det|>
+To determine neutralizing antibodies against the Wuhan- Spike, the B.1.1.7- Spike (Alpha), the B.1.351- Spike (Beta), the P.1- Spike (B.1.1.28.1; Gamma), the B.1.617.2 (Delta), and the B.1.1.529 (Omicron) variants of SARS- CoV- 2- S in plasma, we modified our recently established surrogate virus neutralization test (sVNT) (20, 26). In this assay, the soluble receptor for SARS- CoV- 2, ACE2, is bound to 96- well- plates to which different purified tagged receptor binding domains (RBDs) of the Spike- protein of SARS- CoV- 2 can bind once added to the assay. Binding is further revealed by an anti- tag peroxidase- labelled antibody and colorimetric quantification. Pre- incubation of the Spike- protein with serum or plasma of convalescent patients or vaccines prevents subsequent binding to ACE2 to various degrees, depending on the amount of neutralizing antibodies present. In detail, MaxiSorp 96F plates (Nunc) were coated with recombinant soluble hACE2- Fc(IgG1) protein at 300 ng per well in \(50\mu \mathrm{L}\) coating buffer (30 mM Na2CO3, 70 mM NaHCO3, pH 9.6) at \(4^{\circ}C\) overnight. After blocking with hACE2- Fc(IgG1), plates were washed with phosphate- buffered saline, \(0.05\%\) Tween- 20 (PBST) and blocked with BD OptEIA Assay Diluent for \(1.5\mathrm{h}\) at \(37^{\circ}C\) . In the meantime, plasma samples were serially diluted threefold starting at 1:20 and then pre- incubated for \(1\mathrm{h}\) at \(37^{\circ}C\) with \(1.5\mathrm{ng}\) recombinant SARS- CoV- 2 Spike RBD of either the Wuhan strain (Trenzyme), the B.1.1.7 variant (N501Y; Alpha), the B.1.351 variant (K417N, E484K, N501Y; Beta) or the P.1 variant (K417T, E484K, N501Y; Gamma) (the latter three from SinoBiological), all with a C- terminal His- Tag. BD OptEIA Assay Diluent was used for preparing plasma sample as well as RBD dilutions. After pre- incubation with SARS- CoV- 2 Spike RBDs, plasma samples were given onto the hACE2- coated MaxiSorp ELISA plates for \(1\mathrm{h}\) at \(37^{\circ}C\) . SARS- CoV- 2 Spike RBDs pre- incubated with buffer only served as negative controls for inhibition. Plates were washed three times with PBST and incubated with an HRP- conjugated anti- His- tag antibody (clone HIS 3D5, provided by Helmholtz Zentrum München) for \(1\mathrm{h}\) at \(37^{\circ}C\) . Unbound antibody was removed by six washes with PBST. A colorimetric signal was developed on the enzymatic reaction of HRP with the chromogenic substrate \(3,3',5,5'\) - tetramethylbenzidine (BD OptEIA TMB Substrate Reagent Set). An equal volume of \(0.2\mathrm{MH}_2\mathrm{SO}_4\) was added to stop the reaction, and the absorbance readings at \(450\mathrm{nm}\) and \(570\mathrm{nm}\) were acquired using a SpectraMax iD3 microplate reader (Molecular Devices) using SoftMAX Pro v7.03 software. For each well, the percent inhibition was calculated from optical density (OD) values after subtraction of background values as: Inhibition \((\%) = (1 - \text{Sample OD value / Average SARS - CoV - 2 S RBD OD value})\times 100\) . Neutralizing sVNT titers were determined as the dilution with binding reduction \(> \mathrm{mean} + 2\mathrm{SD}\) of values from a plasma pool consisting of three pre- pandemic plasma samples.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 653, 370, 668]]<|/det|>
+## SARS-CoV-2 protein peptide pools  
+
+<|ref|>text<|/ref|><|det|>[[115, 679, 877, 816]]<|/det|>
+We ordered 15 amino acid (aa) long and 10 aa overlapping peptide pools spanning the whole length of SARS- CoV- 2- Spike (- S) (total 253 peptides), - Membrane (- M) (43 peptides), - Nucleocapsid (- N) (82 peptides) or - Envelope (- E) (12 peptides; peptide no 4 could not be synthesized) proteins from GenScript. All lyophilized peptides were synthesized at \(>95\%\) purity and reconstituted at a stock concentration of \(50\mathrm{mg / mL}\) in DMSO (Sigma- Aldrich), except for 9 SARS- CoV- 2- S overlapping peptides (number 24, 190, 191, 225, 226, 234, 244, 245 and 246), 2 for SARS- CoV- 2- M (number 15 and 16), 1 for SARS- CoV- 2- N (number 61) and all 12 SARS- CoV- 2- E peptides that were dissolved at \(25\mathrm{mg / mL}\) due to solubility issues. All peptides in DMSO stocks were stored at \(- 80^{\circ}\mathrm{C}\) until used.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 827, 312, 842]]<|/det|>
+## T cell re-stimulation assay  
+
+<|ref|>text<|/ref|><|det|>[[117, 853, 870, 921]]<|/det|>
+PBMCs, isolated using a Ficoll gradient, were re- suspended at concentration of \(20\times 10^{6}\) cells/mL in complete RPMI medium [RPMI 1640 (Gibco) supplemented with \(10\%\) FBS (GE Healthcare Life Sciences, Logan, UT), \(1\mathrm{mM}\) sodium pyruvate, \(50\mu \mathrm{M}\beta\) - mercaptoethanol, \(1\%\) streptomycin/penicillin (all Gibco)]. For stimulation, cells were diluted with equal volume of peptide pools containing S
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 81, 880, 530]]<|/det|>
+protein or mixture of M-, N- and E- proteins. Peptide pools were prepared in complete RPMI containing brefeldin A (Sigma- Aldrich) at final concentration of \(10\mu \mathrm{g / mL}\) . In the final mixture each peptide had concentration of \(2\mu \mathrm{g}\) ( \(\sim 1.2\) nmol)/mL, except for SARS- CoV2- S peptides number 24, 190, 191, 225, 226, 234, 244, 245 and 246, SARS- CoV2- M peptides 15 and 16, and SARS- CoV2- N peptide 61, which were used at final concentration of \(1\mu \mathrm{g / mL}\) due to solubility issues. As a negative control, we stimulated the cells with DMSO, used in maximal volume corresponding to DMSO amount in peptide pools (equaling to \(5\%\) DMSO in final medium volume) Extended Fig. 5. In each experiment, we used cells stimulated with Phorbol- 12- myristate- 13- acetate (PMA; Calbiochem) and ionomycin (Invitrogen) at final concentration of \(50\mathrm{ng / mL}\) and \(1500\mathrm{ng / mL}\) , respectively, as an internal positive control. Cells were then incubated for 12- 16 hr at \(37^{\circ}C\) , \(5\%\) CO2. After washing, cells were resuspended in MACS buffer (PBS supplemented with \(3\%\) FBS and \(2\mathrm{mM}\) EDTA). Non- specific antibody binding was blocked by incubating samples with \(10\%\) mouse serum at \(4^{\circ}C\) for 15 min. Next, without washing, an antibody mix of anti- CD3- AF532 (UCHT1; #58- 0038- 42; Lot # 2288218; Invitrogen; 1:50), anti- CD4- BUV563 (RPA- T4; #741353; Lot # 9333607; BD Biosciences; 1:200), anti- CD8- SparkBlue 550 (SK1; #344760; Lot #B326454; Biolegend; 1:200), anti- CD45RA (HI100, #740298, Lot # 0295003; BD Biosciences; 1:200), anti- CCR7 (G043H7; #353230; Lot # B335328; Biolegend; 1:50), anti- CD38 PerCP- eF710 (HB7; #46- 0388- 42; Lot # 2044748; Invitrogen; 1:100) and Zombie NIR™ Fixable Viability Kit (#423106; Lot # B323372; BioLegend) was added. After staining for 20 min at RT, cells were washed before they were fixed and permeabilized (#554714; BD Biosciences) according to the manufacturers' protocol. Next, intracellular cytokines were stained using anti- IFN- PE- Cy7 (B27; #506518; Lot # B326674; Biolegend; 1:100), anti- TNF- AF700 (Mab11; #502928; Lot # B326186; Biolegend; 1:50) and anti- IL- 17A- BV421 (BL168; #512322; Lot # B317903; Biolegend; 1:50) for 45 min on RT. Excess antibodies were washed away and cells were then acquired on Cytek Aurora spectral flow cytometer (Cytek) equipped with five lasers operating on \(355\mathrm{nm}\) , \(405\mathrm{nm}\) , \(488\mathrm{nm}\) , \(561\mathrm{nm}\) and \(640\mathrm{nm}\) (for gating strategy, see Extended Data Fig. 4). All flow cytometry data was acquired using SpectroFlo v2.2.0 (Cytek) and analyzed FCS Express V7 (Denovo).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 539, 475, 555]]<|/det|>
+## Flow cytometric analysis of Spike-specific B cells  
+
+<|ref|>text<|/ref|><|det|>[[117, 565, 881, 669]]<|/det|>
+Total leukocytes were isolated from whole blood using erythrolysis in \(0.83\%\) ammonium chloride solution. Isolated cells were then washed, counted and resuspended in PBS and stained for 20 min on RT with an antibody mix containing antibodies listed in Extended Data Figure 1A together with Spike- mNEONGreen protein ( \(5\mu \mathrm{g}\) per reaction; production will be described elsewhere). After one wash, samples were acquired on spectral flow cytometer and the data was analyzed as described above (for gating strategy, see Extended Data Fig. 1B).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 680, 345, 696]]<|/det|>
+## Quantification of IFN- \(\gamma\) release  
+
+<|ref|>text<|/ref|><|det|>[[117, 706, 878, 791]]<|/det|>
+0.5 mL full blood were stimulated with manufacturer's selected parts of the SARS- CoV- 2 S1 domain of the Spike Protein for a period of 20- 24 h. We carried out negative and positive controls according to the manufacturer's instruction and measured IFN- \(\gamma\) using an ELISA (SARS- CoV- 2 Interferon Gamma Release Assay, IGRA (Euroimmun, Lübeck, Germany). For analysis, we used an AESKU.READER (AESKU.GROUP, Wendelsheim, Germany) and the Gen5 2.01 Software.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 803, 186, 817]]<|/det|>
+## Statistics  
+
+<|ref|>text<|/ref|><|det|>[[117, 828, 877, 914]]<|/det|>
+Statistical analysis was done using GraphPad Prism 8.4 (GraphPad Software, USA) and SPSS 20.0.0 (IBM SPSS Statistics, USA). For comparison of levels of Spike- specific IgG levels, as well as for comparison of percentages of cytokine- secreting T cells, for comparison of frequencies of Spike- specific B cells, or cytokine concentrations in the IGRA assay and sVNT values we used mixed effect analysis with Sidiak's multiple comparison paired t- test (within groups) or unpaired T test with Welch
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 82, 870, 150]]<|/det|>
+correction 2- way ANOVA followed by Sidak's multiple comparison test (between groups). Percentages of cytokine secreting T cells were log transformed prior comparison. For comparison of sVNT titers we used Chi- square test for trend. Differences were considered significant if \(p < 0.05\) . Correlation between sVNT and pVNT values was calculated using single linear regression analysis.  
+
+<|ref|>text<|/ref|><|det|>[[118, 161, 321, 176]]<|/det|>
+Ethics committee approval.  
+
+<|ref|>text<|/ref|><|det|>[[117, 188, 880, 237]]<|/det|>
+The CoCo Study and the analysis conducted for this article were approved by the Internal Review Board of Hannover Medical School (institutional review board no. 8973_BO- K_2020, amendment Dec 2020).  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 276, 325, 291]]<|/det|>
+## Data availability statement  
+
+<|ref|>text<|/ref|><|det|>[[117, 302, 850, 368]]<|/det|>
+All requests for raw and analyzed data that underlie the results reported in this article will be reviewed by the CoCo Study Team, Hannover Medical School (cocostudie@mh- hannover.de) to determine whether the request is subject to confidentiality and data protection obligations. Data that can be shared will be released via a material transfer agreement.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 407, 203, 421]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[110, 456, 884, 907]]<|/det|>
+1. Powell AA, Power L, Westrop S, McOwat K, Campbell H, Simmons R, et al. Real-world data shows increased reactogenicity in adults after heterologous compared to homologous prime-boost COVID-19 vaccination, March-June 2021, England. Euro Surveill. 2021;26(28).  
+2. Evans JP, Zeng C, Carlin C, Lozanski G, Saif LJ, Oltz EM, et al. Loss of Neutralizing Antibody Response to mRNA Vaccination against SARS-CoV-2 Variants: Differing Kinetics and Strong Boosting by Breakthrough Infection. bioRxiv. 2021.  
+3. Tober-Lau P, Schwarz T, Vanshylla K, Hillus D, Gruell H, Group ECS, et al. Long-term immunogenicity of BNT162b2 vaccination in older people and younger health-care workers. Lancet Respir Med. 2021;9(11):e104-e5.  
+4. Arbel R, Hammerman A, Sergienko R, Friger M, Peretz A, Netzer D, et al. BNT162b2 Vaccine Booster and Mortality Due to Covid-19. N Engl J Med. 2021.  
+5. Hansen CH, Schelde AB, Moustsen-Helms IR, Emborg H-D, Krause TG, Moelbak K, et al. Vaccine effectiveness against SARS-CoV-2 infection with the Omicron or Delta variants following a two-dose or booster BNT162b2 or mRNA-1273 vaccination series: A Danish cohort study. medRxiv. 2021:2021.12.20.21267966.  
+6. Bar-On YM, Goldberg Y, Mandel M, Bodenheimer O, Freedman L, Alroy-Preis S, et al. Protection against Covid-19 by BNT162b2 Booster across Age Groups. N Engl J Med. 2021;385(26):2421-30.  
+7. Torjesen I. Covid restrictions tighten as omicron cases double every two to three days. BMJ. 2021;375:n3051.  
+8. Cele S, Jackson L, Khan K, Khoury D, Moyo-Gwete T, Tegally H, et al. SARS-CoV-2 Omicron has extensive but incomplete escape of Pfizer BNT162b2 elicited neutralization and requires ACE2 for infection. medRxiv. 2021:2021.12.08.21267417.  
+9. Hoffmann M, Kruger N, Schulz S, Cossmann A, Rocha C, Kempf A, et al. The Omicron variant is highly resistant against antibody-mediated neutralization – implications for control of the COVID-19 pandemic. bioRxiv. 2021:2021.12.12.472286.  
+10. Schmidt F, Muecksch F, Weisblum Y, Silva JD, Bednarski E, Cho A, et al. Plasma neutralization properties of the SARS-CoV-2 Omicron variant. medRxiv. 2021:2021.12.12.21267646.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 82, 878, 127]]<|/det|>
+11. Tenbusch M, Schumacher S, Vogel E, Priller A, Held J, Steininger P, et al. Heterologous prime-boost vaccination with ChAdOx1 nCoV-19 and BNT162b2. Lancet Infect Dis. 2021;21(9):1212-3. 
+
+<|ref|>text<|/ref|><|det|>[[115, 130, 848, 194]]<|/det|>
+12. Hillus D, Schwarz T, Tober-Lau P, Vanshylla K, Hastor H, Thibeault C, et al. Safety, reactogenicity, and immunogenicity of homologous and heterologous prime-boost immunisation with ChAdOx1 nCoV-19 and BNT162b2: a prospective cohort study. Lancet Respir Med. 2021;9(11):1255-65. 
+
+<|ref|>text<|/ref|><|det|>[[115, 196, 870, 259]]<|/det|>
+13. Flaxman A, Marchevsky NG, Jenkin D, Aboagye J, Aley PK, Angus B, et al. Reactogenicity and immunogenicity after a late second dose or a third dose of ChAdOx1 nCoV-19 in the UK: a substudy of two randomised controlled trials (COV001 and COV002). Lancet. 2021;398(10304):981-90. 
+
+<|ref|>text<|/ref|><|det|>[[115, 261, 875, 323]]<|/det|>
+14. Ramasamy MN, Minassian AM, Ewer KJ, Flaxman AL, Folegatti PM, Owens DR, et al. Safety and immunogenicity of ChAdOx1 nCoV-19 vaccine administered in a prime-boost regimen in young and old adults (COV002): a single-blind, randomised, controlled, phase 2/3 trial. Lancet. 2021;396(10267):1979-93. 
+
+<|ref|>text<|/ref|><|det|>[[115, 325, 872, 386]]<|/det|>
+15. Borobia AM, Carcas AJ, Perez-Olmeda M, Castano L, Bertran MJ, Garcia-Perez J, et al. Immunogenicity and reactogenicity of BNT162b2 booster in ChAdOx1-S-primed participants (CombiVacS): a multicentre, open-label, randomised, controlled, phase 2 trial. Lancet. 2021;398(10295):121-30. 
+
+<|ref|>text<|/ref|><|det|>[[115, 388, 875, 435]]<|/det|>
+16. Barros-Martins J, Hammerschmidt SI, Cossmann A, Odak I, Stankov MV, Morillas Ramos G, et al. Immune responses against SARS-CoV-2 variants after heterologous and homologous ChAdOx1 nCoV-19/BNT162b2 vaccination. Nat Med. 2021;27(9):1525-9. 
+
+<|ref|>text<|/ref|><|det|>[[115, 437, 820, 483]]<|/det|>
+17. Schmidt T, Klemis V, Schub D, Mihm J, Hielscher F, Marx S, et al. Immunogenicity and reactogenicity of heterologous ChAdOx1 nCoV-19/mRNA vaccination. Nat Med. 2021;27(9):1530-5. 
+
+<|ref|>text<|/ref|><|det|>[[115, 485, 863, 515]]<|/det|>
+18. Pozzetto B, Legros V, Djebali S, Barateau V, Guibert N, Villard M, et al. Immunogenicity and efficacy of heterologous ChAdOx1-BNT162b2 vaccination. Nature. 2021. 
+
+<|ref|>text<|/ref|><|det|>[[115, 516, 866, 562]]<|/det|>
+19. Behrens GMN, Cossmann A, Stankov MV, Witte T, Ernst D, Happle C, et al. Perceived versus proven SARS-CoV-2-specific immune responses in health-care professionals. Infection. 2020;48(4):631-4. 
+
+<|ref|>text<|/ref|><|det|>[[115, 564, 875, 610]]<|/det|>
+20. Bosnjak B, Stein SC, Willenzon S, Cordes AK, Puppe W, Bernhardt G, et al. Low serum neutralizing anti-SARS-CoV-2 antibody levels in mildly affected COVID-19 convalescent patients revealed by two different detection methods. Cell Mol Immunol. 2021;18(4):936-44. 
+
+<|ref|>text<|/ref|><|det|>[[115, 612, 875, 643]]<|/det|>
+21. Hoffmann M, Arora P, Gross R, Seidel A, Hornich BF, Hahn AS, et al. SARS-CoV-2 variants B.1.351 and P.1 escape from neutralizing antibodies. Cell. 2021;184(9):2384-93 e12. 
+
+<|ref|>text<|/ref|><|det|>[[115, 645, 875, 707]]<|/det|>
+22. Munro APS, Janani L, Cornelius V, Aley PK, Babbage G, Baxter D, et al. Safety and immunogenicity of seven COVID-19 vaccines as a third dose (booster) following two doses of ChAdOx1 nCoV-19 or BNT162b2 in the UK (COV-BOOST): a blinded, multicentre, randomised, controlled, phase 2 trial. Lancet. 2021;398(10318):2258-76. 
+
+<|ref|>text<|/ref|><|det|>[[115, 709, 850, 770]]<|/det|>
+23. Liu X, Shaw RH, Stuart ASV, Greenland M, Aley PK, Andrews NJ, et al. Safety and immunogenicity of heterologous versus homologous prime-boost schedules with an adenoviral vectored and mRNA COVID-19 vaccine (Com-COV): a single-blind, randomised, non-inferiority trial. Lancet. 2021;398(10303):856-69. 
+
+<|ref|>text<|/ref|><|det|>[[115, 772, 843, 833]]<|/det|>
+24. Madelon N, Heikkilä N, Sabater Royo I, Fontannaz P, Breville G, Lauper K, et al. Omicron-specific cytotoxic T-cell responses are boosted following a third dose of mRNA COVID-19 vaccine in anti-CD20-treated multiple sclerosis patients. medRxiv. 2021:2021.12.20.21268128. 
+
+<|ref|>text<|/ref|><|det|>[[115, 835, 866, 897]]<|/det|>
+25. Jablonka A, Happle C, Cossmann A, Stankov MV, Scharff AZ, Ernst D, et al. Protocol for longitudinal assessment of SARS-CoV-2-specific immune responses in healthcare professionals in Hannover, Germany: the prospective, longitudinal, observational COVID-19 Contact (CoCo) study. medRxiv. 2020:2020.12.02.20242479.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 866, 130]]<|/det|>
+26. Hammerschmidt SI, Bosnjak B, Bernhardt G, Friedrichsen M, Ravens I, Dopfer-Jablonka A, et al. Neutralization of the SARS-CoV-2 Delta variant after heterologous and homologous BNT162b2 or ChAdOx1 nCoV-19 vaccination. Cell Mol Immunol. 2021;18(10):2455-6.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[117, 85, 802, 936]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 873, 213]]<|/det|>
+Fig. 1 | Participant recruitment schemes and humoral immune response. a, Participant recruitment and vaccination and blood sampling scheme. C, ChAd; B, BNT. b, A third heterologous or c, a third homologous immunization with BNT induce strong increases in anti S1 IgG5. Data are from \(n\) biologically independent samples as shown in the figure. d, A third heterologous immunization with BNT leads to increased frequencies of S-specific memory B cells. Data are from \(n\) biologically independent samples as shown in the figure. Statistics: b,c,d, Mixed effect analysis followed by Sidak's multiple comparison test (within groups) and unpaired t test with Welch's correction (between groups). The majority of the symbols depicted in grey had been published before (16, 26).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 80, 884, 803]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 814, 880, 925]]<|/det|>
+<center>Fig. 2 | Heterologous vaccination induces neutralizing antibodies as well as CD4 and CD8 T cell responses. a, Heterologous ChAd/BNT/BNT or ChAd/ChAd/BNT vaccination induces neutralizing antibodies against Wuhan, B.1.1.7 (Alpha), B.1.351 (Beta), P.1 (B.1.1.28.1; Gamma), B.1.617.2 (Delta) and the B.1.1.529 (Omicron) SARS-CoV-2-S variants measured using the sVNT. Data are from \(n =\) biologically independent samples as indicated. For better visualization of identical titer values, data were randomly and proportionally adjusted closely around the precise titer results. Boost vaccination increased total percentage of cytokine-secreting CD4+ (b) and CD8+ (c) T cells. We calculated the </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 878, 130]]<|/det|>
+total number of cytokine secreting cells as the sum of IFN- \(\gamma +\) TNF- \(\alpha -\) , IFN- \(\gamma +\) TNF- \(\alpha +\) and IFN- \(\gamma -\) TNF- \(\alpha +\) cells in the gates indicated in Extended Data Fig. 4. Data are from \(n\) biologically independent samples as indicated.  
+
+<|ref|>text<|/ref|><|det|>[[115, 131, 857, 178]]<|/det|>
+d, IFN- \(\gamma\) concentration in full blood supernatants after stimulation with SARS- CoV- 2 S1 domain for 20- 24 h measured in duplicate by IGRA (Euroimmun). Data are from \(n\) biologically independent samples as indicated.  
+
+<|ref|>text<|/ref|><|det|>[[115, 180, 869, 227]]<|/det|>
+a, b, c and d. Mixed effect analysis followed by Sidak's multiple comparison test (within groups) and unpaired t test with Welch's correction (between groups). The majority of the symbols depicted in grey had been published before (16, 26).
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[115, 85, 232, 97]]<|/det|>
+# Extended Data
+
+<|ref|>table<|/ref|><|det|>[[115, 161, 881, 323]]<|/det|>
+
+<table><tr><td rowspan="2"></td><td rowspan="2">Mean age,<br>years<br>(range)</td><td rowspan="2">Sex, m/f<br>(%)</td><td colspan="6">Median (IQR) days past last vaccination</td></tr><tr><td>1st vaccination</td><td>2 mo</td><td>14 d</td><td>4 mo</td><td>6 mo</td><td>14 d</td></tr><tr><td>ChAd<br>ChAd<br>BNT<br>n=41</td><td>40<br>(21-64)</td><td>14/27<br>(34/66)</td><td rowspan="3">1st vaccination</td><td>68<br>(12.75)</td><td rowspan="3">14<br>(4)</td><td>15<br>(13)</td><td>119<br>(5)</td><td rowspan="3">196<br>(6.5)</td></tr><tr><td>ChAd<br>BNT<br>n=82</td><td>37<br>(19-61)</td><td>17/65<br>(21/79)</td><td>70<br>(8)</td><td>17<br>(5)</td><td>117<br>(13)</td><td>195<br>(8)</td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr></table>
+
+<|ref|>table<|/ref|><|det|>[[115, 353, 881, 496]]<|/det|>
+
+<table><tr><td rowspan="2"></td><td rowspan="2">Mean age,<br>years (range)</td><td rowspan="2">Sex, m/f<br>(%)</td><td colspan="6">Median (IQR) days past last vaccination</td></tr><tr><td></td><td>21 d</td><td>1 mo</td><td>7 mo</td><td>9 mo</td><td>21 d</td></tr><tr><td>BNT<br>BNT<br>BNT<br>n=57</td><td>42<br>(23-63)</td><td>21/36<br>(38/62)</td><td rowspan="2">1st vaccination</td><td>20<br>(1.25)</td><td rowspan="2">29<br>(8.25)</td><td>211<br>(9)</td><td>267<br>(22.5)</td><td rowspan="2">23<br>(10.2<br>5)</td></tr><tr><td></td><td></td><td></td><td>2nd vaccination</td><td></td><td></td></tr></table>
+
+<|ref|>text<|/ref|><|det|>[[115, 530, 859, 559]]<|/det|>
+**Extended Data Table. 1 |** Demographic data and median time in days since last vaccination for the five blood collection time points (months = mo and days = d) of the three vaccination groups.
+
+<--- Page Split --->
+<|ref|>table_caption<|/ref|><|det|>[[115, 85, 130, 97]]<|/det|>
+a
+
+<|ref|>table<|/ref|><|det|>[[150, 100, 820, 275]]<|/det|>
+
+<table><tr><td>Antigen</td><td>Conjugate</td><td>Clone</td><td>Orde no.</td><td>Company</td><td>Dilution</td></tr><tr><td>CD14</td><td>BB700</td><td>MP9</td><td>566465</td><td>BD</td><td>100</td></tr><tr><td>CD16</td><td>BUV496</td><td>3G8</td><td>612944</td><td>BD</td><td>100</td></tr><tr><td>CD19</td><td>PECy7</td><td>HIB19</td><td>982410</td><td>BioLegend</td><td>200</td></tr><tr><td>CD20</td><td>BV421</td><td>2H7</td><td>302330</td><td>BioLegend</td><td>100</td></tr><tr><td>CD27</td><td>BUV805</td><td>L128</td><td>748704</td><td>BD</td><td>100</td></tr><tr><td>CD38</td><td>PerCP-eF710</td><td>HB7</td><td>46-0388-42</td><td>Invitrogen</td><td>100</td></tr><tr><td>IgD</td><td>BV480</td><td>IA6-2</td><td>566138</td><td>BD</td><td>200</td></tr><tr><td>IgM</td><td>AF647</td><td>MHM-88</td><td>314436</td><td>BioLegend</td><td>100</td></tr><tr><td>Viability</td><td>Zombie NIRTM</td><td>-</td><td>423106</td><td>BioLegend</td><td>400</td></tr><tr><td>Anti-S BCR</td><td>mNeonGreen</td><td colspan="3">Produced by T.Krey</td><td>\(5\mu L/sample\)</td></tr></table>
+
+<|ref|>image<|/ref|><|det|>[[115, 300, 846, 572]]<|/det|>
+
+<|ref|>image_caption<|/ref|><|det|>[[115, 592, 877, 653]]<|/det|>
+<center>Extended Data Fig. 1 | Antibody panel a and gating strategy b for SARS-CoV-2-S (Spike)-specific B cell populations in blood. Pseudocolor plots show representative data from a female donor<br>283 days after priming with ChAd; 213 days after a second dose with BNT and 14 days after a third dose with BNT.</center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 80, 420, 291]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[115, 318, 864, 430]]<|/det|>
+Extended Data Fig. 2 | Antibody neutralization measurements against the Omicron SARS- CoV- 2 variant is positively correlated between the virus neutralization tests (sVNT) and pseudotyped virus neutralization tests (pVNT). Correlation (solid line) and 95% confidence intervals (dotted lines) between sVNT1:20 and antibody titers resulting in 50% reduction of luciferase activity in pVNT, indicated as pVNT50. Open circles, values from individual donors, outliers are marked with X and were defined as values with absolute residual value \(> 2\) SD of all residual values in each group of samples. Correlation was calculated using single linear regression.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 78, 884, 375]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[115, 393, 883, 522]]<|/det|>
+Extended Data Fig. 3 | Humoral immune response against all SARS- CoV- 2 variants following homologous BNT162b2 (BNT) / BNT /BNT vaccination. Reciprocal titers of neutralizing antibodies against Wuhan, B.1.1.7 (Alpha), B.1.351 (Beta), P.1 (B.1.1.28.1; Gamma), B.1.617.2 (Delta) and the B.1.1.529 (Omicron) SARS- CoV- 2- S variants measured using the sVNT. Data are from \(n =\) biologically independent samples as indicated. Mixed effect analysis followed by Sidak's multiple comparison test (within groups). For better visualization of identical titer values, data were randomly and proportionally adjusted closely around the precise titer results. The majority of the symbols depicted in grey had been published before (16, 26).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 80, 870, 355]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[115, 400, 852, 433]]<|/det|>
+Extended Data Fig. 4 | Gating strategy used for detection of cytokine producing CD4+ and CD8+ T cells after ex vivo re- stimulation with DMSO or the pool of Spike- specific peptides for 12–16 hr.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[130, 80, 884, 525]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 548, 870, 580]]<|/det|>
+<center>Extended Data Fig. 5 | Frequency of cytokine-producing CD4<sup>+</sup> T cells and CD8<sup>+</sup> T cells after ex vivo re-stimulation with DMSO or the pool of Spike-specific peptides for 12–16 hr.</center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 312, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[61, 130, 245, 149]]<|/det|>
+- flatBehrensrs2.pdf
+
+<--- Page Split --->

preprint/preprint__4917d2ec82456765310e4fd0e81bd9c6564f10de32d29557ebc886beb9b7789e/images_list.json

@@ -0,0 +1,77 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1. Schematic diagram illustrating that SMB and \\(\\delta^{15}\\mathrm{N}_{\\mathrm{NO3arc}}\\) values covary due to photolytic \\(\\mathrm{NO}_3^-\\) mass loss. After \\(\\mathrm{NO}_3^-\\) containing either \\(^{14}\\mathrm{N}\\) (blue) and \\(^{15}\\mathrm{N}\\) (red) is deposited on the Antarctic snowpack surface (1), sunlight in the photic zone can trigger photolysis of \\(\\mathrm{NO}_3^-\\) that favors \\(\\mathrm{NO}_3^-\\) with a \\(^{14}\\mathrm{N}\\) atom, which leaves the residual \\(\\mathrm{NO}_3^-\\) enriched in \\(^{15}\\mathrm{N}\\) (2). Because sites with lower SMBs will accumulate less snow over a given period of time than high SMB sites (3), the \\(\\mathrm{NO}_3^-\\) at lower SMB sites will remain in the photic zone longer, experience more photolytic mass loss before burial in the archived zone, and have higher \\(\\delta^{15}\\mathrm{N}_{\\mathrm{NO3arc}}\\) values (4).",
+    "footnote": [],
+    "bbox": [
+      [
+        117,
+        81,
+        702,
+        368
+      ]
+    ],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. A) Map of East Antarctic sites sampled for \\(\\delta^{15}\\mathrm{N}_{\\mathrm{NO3arc}}\\) along different scientific and logistic transect routes<sup>36</sup>. The SMBs were modeled by \\(\\mathrm{MAR}^{12}\\) and adjusted for dry site bias (see Methods). Regions with SMBs less than or greater than \\(40–200\\mathrm{kg}\\mathrm{m}^{-2}\\mathrm{a}^{-1}\\) (i.e., the SMB range targeted by the proxy described here) are illustrated with hatching and crosshatching, respectively. Preservation of \\(\\mathrm{NO}_3^-\\) is not expected in blue ice zones (gray solid) due to very low or negative SMBs and wind scouring<sup>37</sup>. B) Scatter plot and linear regression of Eq. (1) using all sites in the field dataset. Note that axis labels have been converted to show simpler SMB and \\(\\delta^{15}\\mathrm{N}_{\\mathrm{NO3arc}}\\) values. The linear regression (gray solid line) is shown with shaded \\(95\\%\\) confidence intervals, and regression parameters are displayed at upper right. The point colors correspond to the transect of origin shown in (a), and the point shapes correspond to the sampling method (i.e., snow core, snow pit, or 1-m depth layer).",
+    "footnote": [],
+    "bbox": [
+      [
+        124,
+        84,
+        873,
+        277
+      ]
+    ],
+    "page_idx": 7
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3. (a) Surface mass balance reconstruction for Aurora Basin North based on \\(\\delta^{15}\\mathrm{N}_{\\mathrm{NO3}}\\) arc data from the ABN1314-103 ice core. Reconstructed \\(\\mathrm{SMB}_{\\delta 15\\mathrm{N}}\\) values are shown by the red stepped lines with the 50-yr running mean \\(\\pm 1\\sigma\\) overlaid as a darker thick line and shaded zone. (b) Comparison of SMBs reconstructed from \\(\\delta^{15}\\mathrm{N}_{\\mathrm{NO3}}\\) (red) with those from ice density (gray) and upstream GPR isochron depth \\(^{39}\\) . The \\(\\mathrm{SMB}_{\\delta 15\\mathrm{N}}\\) and \\(\\mathrm{SMB}_{\\mathrm{GPR}}\\) values were aggregated to match the 1-m resolution of the \\(\\mathrm{SMB}_{\\mathrm{density}}\\) data. For \\(\\mathrm{SMB}_{\\delta 15\\mathrm{N}}\\) and \\(\\mathrm{SMB}_{\\mathrm{density}}\\) , smoothed LOESS curves are overlaid to more clearly show long-term patterns. (c) \\(\\mathrm{SMB}_{\\delta 15\\mathrm{N}}\\) values after the upstream topographic impact on SMBs has been removed, with 50-yr running mean \\(\\pm 1\\sigma\\) values overlaid. The resulting residuals may better illustrate SMB variability due to climate change.",
+    "footnote": [],
+    "bbox": [
+      [
+        122,
+        78,
+        825,
+        645
+      ]
+    ],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4. Linear regressions of $\\mathrm {SMB}_{\\mathrm {ground}}$ versus $\\mathrm {SMB}_{\\mathrm {MAR}}$ for the period 1979-2017 at the 51 sites with $\\mathrm {SMB}_{\\mathrm {ground}}$ observations, with 95% confidence intervals of the regressions shaded. Sites are subset for two overlapping regressions that intersect at (138, 130). These linear regressions were applied to the $\\mathrm {SMB}_{\\mathrm {MAR}}$ values for all sampling sites to produce the $\\mathrm {SMB}_{\\mathrm {adjMA}}$ used in analyses. The dashed line represents a slope of 1 (i.e., if the $\\mathrm {SMB}_{\\mathrm {MAR}}$ perfectly matched the $\\mathrm {SMB}_{\\mathrm {ground}}$ ).",
+    "footnote": [],
+    "bbox": [
+      [
+        132,
+        268,
+        810,
+        732
+      ]
+    ],
+    "page_idx": 24
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5. Local accumulation rate variability with depth along the upstream ABN transect determined from GPR identification of isochronic internal reflective horizons. Accumulation rates have an original depth resolution of \\(0.5 \\mathrm{m}\\) which is smoothed through a moving age-depth average with a cosine weighting window to reduce isochron artifacts<sup>41</sup>.",
+    "footnote": [],
+    "bbox": [
+      [
+        122,
+        90,
+        854,
+        496
+      ]
+    ],
+    "page_idx": 27
+  }
+]

preprint/preprint__4917d2ec82456765310e4fd0e81bd9c6564f10de32d29557ebc886beb9b7789e/preprint__4917d2ec82456765310e4fd0e81bd9c6564f10de32d29557ebc886beb9b7789e.mmd

@@ -0,0 +1,542 @@
+
+# Sunlight-driven nitrate loss records Antarctic surface mass balance  
+
+Pete D. Akers ( pete.d.akers@gmail.com )  
+
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France https://orcid.org/0000- 0002- 2266- 5551  
+
+Joël Savarino  
+
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+Nicolas Caillon  
+
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+Aymeric P. M. Servettaz  
+
+Japan Agency for Marine- Earth Science and Technology, Yokosuka, Japan  
+
+Emmanuel Le Meur  
+
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+Olivier Magand  
+
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+Jean Martins  
+
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+Cécile Agosta  
+
+Laboratoire des Sciences du Climat et de l'Environnement, LSCE- IPSL, CEA- CNRS- UVSQ, Université Paris- Saclay, Gif- sur- Yvette, France https://orcid.org/0000- 0003- 4091- 1653  
+
+Peter Crockford  
+
+Department of Earth and Planetary Sciences, Weizmann Institute of Science, Rehovot, Israel; Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA, USA  
+
+Kanon Kobayashi  
+
+Department of Chemical Science and Engineering, Tokyo Institute of Technology, Yokohama, Japan  
+
+Shohei Hattori  
+
+Department of Chemical Science and Engineering, Tokyo Institute of Technology, Yokohama, Japan; International Center for Isotope Effects Research, Nanjing University, Nanjing, China  
+
+Mark Curran  
+
+Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia; Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia  
+
+Tas van Ommen  
+
+Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia; Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University
+
+<--- Page Split --->
+
+of Tasmania, Hobart, Tasmania, Australia https://orcid.org/0000- 0002- 2463- 1718  
+
+## Lenneke Jong  
+
+Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia; Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia  
+
+## Jason L. Roberts  
+
+Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia; Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia  
+
+## Article  
+
+## Keywords:  
+
+Posted Date: February 8th, 2022  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 1307003/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Communications on July 25th, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 31855- 7.
+
+<--- Page Split --->
+
+# Sunlight-driven nitrate loss records Antarctic surface mass balance  
+
+Authors: Pete D. Akers \(^{1*}\) , Joël Savarino \(^{1*}\) , Nicolas Caillon \(^{1}\) , Aymeric P. M. Servettaz \(^{2}\) , Emmanuel Le Meur \(^{1}\) , Olivier Magand \(^{1}\) , Jean Martins \(^{1}\) , Cécile Agosta \(^{3}\) , Peter Crockford \(^{4,5}\) , Kanon Kobayashi \(^{6}\) , Shohei Hattori \(^{6,7}\) , Mark Curran \(^{8,9}\) , Tas van Ommen \(^{8,9}\) , Lenneke Jong \(^{8,9}\) , Jason L. Roberts \(^{8,9}\)  
+
+## Affiliations:  
+
+\(^{1}\) Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France. \(^{2}\) Japan Agency for Marine- Earth Science and Technology, Yokosuka, Japan. \(^{3}\) Laboratoire des Sciences du Climat et de l'Environnement, LSCE- IPSL, CEA- CNRS- UVSQ, Université Paris- Saclay, Gif- sur- Yvette, France. \(^{4}\) Department of Earth and Planetary Sciences, Weizmann Institute of Science, Rehovot, Israel. \(^{5}\) Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA, USA. \(^{6}\) Department of Chemical Science and Engineering, Tokyo Institute of Technology, Yokohama, Japan. \(^{7}\) International Center for Isotope Effects Research, Nanjing University, Nanjing, China. \(^{8}\) Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia. \(^{9}\) Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia.  
+
+\* Correspondence to: pete- akers@univ- grenoble- alpes.fr; joel.savarino@cnrs.fr  
+
+## Abstract:  
+
+Standard proxies for reconstructing surface mass balance (SMB) in Antarctic ice cores are often inaccurate or coarsely resolved when applied to more complicated environments away from dome summits. Here, we propose an alternative SMB proxy based on photolytic fractionation of nitrogen isotopes in nitrate observed at 114 sites throughout East Antarctica. Applying this proxy approach to nitrate in a shallow core drilled at a moderate SMB site (Aurora Basin North), we reconstruct 700 years of SMB changes that agree well with changes estimated from ice core density and upstream surface topography. For the under- sampled transition zones between dome summits and the coast, this proxy can considerably
+
+<--- Page Split --->
+
+expand our SMB records by providing high- resolution SMBs that better reflect the local environment and are easier to sample than existing techniques.  
+
+One Sentence Summary: Nitrate isotopes offer a new way to track past and present changes in Antarctic snowfall and ice sheet mass balance.  
+
+Main Text: Antarctica holds a critical role in the Earth's hydrosphere, providing long- term storage of 27 million \(\mathrm{km}^3\) of ice<sup>1</sup> and impacting global ocean and atmosphere circulation through its albedo, topography, export of calved glacial ice, and function as an atmospheric heat sink<sup>2- 5</sup>. Since even small shifts in the surface mass balance (SMB) across Antarctic ice sheets can redistribute huge masses of water between the cryosphere, ocean, and atmosphere, a clear understanding of how its SMB has responded to past climate change is crucial for calibrating forecast models of the global environment and properly interpreting ice cores<sup>6- 10</sup>. Despite this pressing importance, a comprehensive understanding of past SMB changes in Antarctica is limited by insufficient long- term records for sites between of the wet coastal periphery and dry dome summits.  
+
+Reconstructing SMBs in this moderate SMB transition zone can be challenging with existing SMB proxies. Ice density- based reconstructions become less effective and more uncertain with depth due to thinning and deformation of ice layers<sup>11</sup>, while the frequent minor damage and breakage of cores during the drilling process can make accurate physical measurements of mass and volume challenging. Water isotopes ( \(\delta^2\mathrm{H}\) or \(\delta^{18}\mathrm{O}\) ) can be used as a proxy temperature to derive snow accumulation through water vapor saturation<sup>10</sup>, but this approach does not account for wind- driven transport and sublimation of surface snow at warmer and lower elevation sites<sup>12- 14</sup>. Additionally, water isotopes reflect many environmental factors other than temperature, such as atmospheric circulation changes, which can lead to large uncertainty and/or bias in reconstructed SMBs<sup>15,16</sup>. There is thus a strong need for alternative
+
+<--- Page Split --->
+
+proxies that better record local conditions to provide SMB estimates for paleoclimate reconstructions and ice sheet models.  
+
+Here, we present one such independent SMB proxy based on photolysis- induced changes in the \(^{15}\mathrm{N / ^{14}N}\) ratio ( \(\delta^{15}\mathrm{N}\) , defined as \(\delta = \frac{^{15}\mathrm{N / ^{14}N}_{\mathrm{sample}}}{^{15}\mathrm{N / ^{14}N}_{\mathrm{standard}}} - 1\) , relative to the \(\mathrm{N}_2\) - air standard) of nitrate \(\mathrm{(NO_3^- )}\) (Figure 1). Naturally deposited on the Antarctic ice sheet surface as the end product of the atmospheric oxidation of reactive nitrogen \(^{17 - 20}\) , \(\mathrm{NO_3^- }\) within the Antarctic snowpack can be photolytically converted to gaseous nitrogen oxides \(\mathrm{(NO_x = NO + NO_2)}\) when exposed to ultraviolet light ( \(\lambda = 290 - 350 \mathrm{nm}\) ). Because \(^{14}\mathrm{NO_3^- }\) is more readily photolyzed than \(^{15}\mathrm{NO_3}\) , the \(\delta^{15}\mathrm{NNO_3}\) of \(\mathrm{NO_3^- }\) remaining in the snow will increase from its initial depositional value of \(\approx - 20\) to \(+20\%\) to values as high as \(+400\%\) \(^{19 - 26}\) as the isotopically lighter photolytic \(\mathrm{NO_x}\) is ventilated and lost to the atmosphere. Although \(\mathrm{NO_3^- }\) can also be lost through \(\mathrm{HNO_3}\) volatilization, we interpret \(\delta^{15}\mathrm{NNO_3}\) are solely through photolysis as volatilization does not strongly fractionate \(\mathrm{NO_3^- }\) and is a very minor component of \(\mathrm{NO_3^- }\) loss outside of the warmest coastal zones \(^{22,27,28}\) . Additionally, while the oxygen in \(\mathrm{NO_3^- }\) also undergoes isotopic fractionation through photolysis, its interpretation is complicated by isotopic interactions with snow and water vapor \(^{22,23,29}\) and is not further discussed here.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Figure 1. Schematic diagram illustrating that SMB and \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values covary due to photolytic \(\mathrm{NO}_3^-\) mass loss. After \(\mathrm{NO}_3^-\) containing either \(^{14}\mathrm{N}\) (blue) and \(^{15}\mathrm{N}\) (red) is deposited on the Antarctic snowpack surface (1), sunlight in the photic zone can trigger photolysis of \(\mathrm{NO}_3^-\) that favors \(\mathrm{NO}_3^-\) with a \(^{14}\mathrm{N}\) atom, which leaves the residual \(\mathrm{NO}_3^-\) enriched in \(^{15}\mathrm{N}\) (2). Because sites with lower SMBs will accumulate less snow over a given period of time than high SMB sites (3), the \(\mathrm{NO}_3^-\) at lower SMB sites will remain in the photic zone longer, experience more photolytic mass loss before burial in the archived zone, and have higher \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values (4). </center>  
+
+Photolysis is limited to the depth where light penetrates and initiates photochemical reactions, and so the snowpack can be divided into an uppermost photic zone (generally 10- 100 cm in East Antarctica) and a deeper archived zone \(^{29 - 33}\) . Photolysis and the resulting isotopic fractionation of \(\mathrm{NO}_3^-\) cease once snowfall buries \(\mathrm{NO}_3^-\) beneath the photic zone, and the \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) value of the buried \(\mathrm{NO}_3^-\) ( \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) ) is assumed to be preserved indefinitely in glacial ice \(^{22,23,29,30}\) . The final \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value reflects the total sum of photolysis inducing radiation experienced by \(\mathrm{NO}_3^-\) during the burial process, which, assuming stable insolation and photic zone depth, is itself determined by the rate at which the \(\mathrm{NO}_3^-\) is buried and thus inversely related to SMB \(^{17,23,26,34}\) . Modeling (Supplementary Text 1) and field observations support SMB as the primary driver of spatial variability in \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values. Based on a new
+
+<--- Page Split --->
+
+simplified theoretical framework (Methods, Supplementary Text 1), this relationship can be expressed as:  
+
+\[\ln (\delta^{15}N_{NO3arc} + 1) = \frac{A}{SMB} +B \quad (1)\]  
+
+where the regression coefficients \(A\) and \(B\) are parameters that subsume constants and linearly co- varying variables associated with photolytic and fractionation processes. The inverse function of Eq. (1) can then be used as a transfer function to reconstruct SMBs from \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values (SMB \(\delta_{15}\mathrm{N}\) ).  
+
+## Results: SMB \(\delta_{15}\mathrm{N}\) relationship and spatial applicability  
+
+To obtain parameter estimates for Eq. (1), we sampled \(\mathrm{NO}_3^-\) in snow and firn from 92 East Antarctic shallow pits and cores that are newly reported here. Combined with 43 previously published \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) samples \(^{21 - 23,26,29,35}\) , this constitutes a database of 135 total \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values representing 114 distinct sites across East Antarctica (Figure 2a). These \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) data were spatially paired with local SMBs either observed directly onsite (SMB \(\mathrm{ground}\) ) or as an output from the Modèle Atmosphérique Régional (MAR) using ERA- interim reanalysis data \(^{12}\) and adjusted for a dry- site bias (SMB \(\mathrm{adjMAR}\) ) (Methods, Supplementary Text 2). The sites in our database cover a comprehensive range of East Antarctic SMBs, from 20–30 kg m \(^{- 2}\) a \(^{- 1}\) at dome summits on the high plateau to \(>300 \mathrm{kg m}^{- 2} \mathrm{a}^{- 1}\) for sites on the coastal periphery.
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Figure 2. A) Map of East Antarctic sites sampled for \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) along different scientific and logistic transect routes<sup>36</sup>. The SMBs were modeled by \(\mathrm{MAR}^{12}\) and adjusted for dry site bias (see Methods). Regions with SMBs less than or greater than \(40–200\mathrm{kg}\mathrm{m}^{-2}\mathrm{a}^{-1}\) (i.e., the SMB range targeted by the proxy described here) are illustrated with hatching and crosshatching, respectively. Preservation of \(\mathrm{NO}_3^-\) is not expected in blue ice zones (gray solid) due to very low or negative SMBs and wind scouring<sup>37</sup>. B) Scatter plot and linear regression of Eq. (1) using all sites in the field dataset. Note that axis labels have been converted to show simpler SMB and \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values. The linear regression (gray solid line) is shown with shaded \(95\%\) confidence intervals, and regression parameters are displayed at upper right. The point colors correspond to the transect of origin shown in (a), and the point shapes correspond to the sampling method (i.e., snow core, snow pit, or 1-m depth layer). </center>  
+
+The SMB and \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) in our field dataset are correlated with a high degree of confidence, producing a linear regression where \(\ln (\delta^{15}\mathrm{N}_{\mathrm{NO3arc}} + 1) = 6.98\pm 0.19\mathrm{SMB}^{- 1} - 0.02\pm 0.01\) (Figure 2b, \(r^2 = 0.91\) , \(p\ll 0.001\) , \(n = 135\) ). This relationship is within modeled expectations (Figure S3, Supplementary Text 1) and reproduces the spatial variability of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) observed throughout East Antarctica (Table S6). Although millennial- scale changes in global nitrogen dynamics and atmospheric oxidative capacity are not currently well known, the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy should be broadly applicable to Holocene- age ice as the factors parameterized in Eq. (1) have likely been relatively stable during this time (Supplementary Text 1). For pre- Holocene ice, \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) can still offer important insight into relative
+
+<--- Page Split --->
+
+changes in SMB and into how nitrate dynamics varied during the dramatically different Antarctic and global environments of the Pleistocene.  
+
+While our field dataset covers sites with SMBs from 22 to \(548\mathrm{kgm^{- 2}a^{- 1}}\) , the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy appears best suited for sites with SMBs between 40 and \(200\mathrm{kgm^{- 2}a^{- 1}}\) . Shallow cores from very dry Dome A and Dome C have lower \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values at 2- 6 m below the surface than at the \(\sim 1\mathrm{m}\) base of the photic zone, possibly because photolytic \(\mathrm{NO_x}\) can be transported downward through firn air convection and re- oxidized into \(\mathrm{NO_3}\) with low \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) values (Supplementary Text 3, Figure S5). Although this phenomenon violates the foundational assumption of “locked- in” \(\mathrm{NO_3}\) - beneath the photic zone, we observe only it at the ultra- dry interior sites where \(\mathrm{SMB} > 40\mathrm{kgm^{- 2}a^{- 1}}\) . For sites with \(\mathrm{SMB} > 200\mathrm{kgm^{- 2}a^{- 1}}\) , the expected \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value falls within the general range of atmospheric \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) ( \(< +20\%\) ) because \(\mathrm{NO_3}\) - is buried below the photic zone in less than a year. Since more than \(80\%\) of \(\mathrm{NO_3}\) - is deposited during months with sunshine outside of winter polar night \(^{22,38}\) , samples that integrate multiple years of accumulation at high SMB sites might still resolve differences in SMB, but the greater \(\mathrm{HNO_3}\) volatilization at these warmer and wetter sites also warrant caution due to possible interference. Additionally, the asymptotic nature of \(\mathrm{SMB^{- 1}}\) means that \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values are increasingly less sensitive to SMB changes with higher SMB values. Despite these restrictions, over \(59\%\) of Antarctica has a SMB between 40 and \(200\mathrm{kgm^{- 2}a^{- 1}}\) \(^{12}\) (Figure 2), and additional study of \(\mathrm{NO_3}\) - dynamics in wet and dry extremes may reveal regional adjustments that allow our proxy’s SMB range to be expanded.  
+
+## Results: Aurora Basin North SMB reconstruction  
+
+As a proof of concept, we applied the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) transfer function to \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) data from the \(103\mathrm{m}\) deep ABN1314- 103 ice core. This core was one of three drilled in the Australian
+
+<--- Page Split --->
+
+Antarctic Program's 2013–2014 summer campaign at Aurora Basin North (ABN; 71.17 °S 111.37 °E, 2679 m above sea level), a site with moderate modern SMBs (≈120 kg m⁻² a⁻¹) located midway between coastal Casey Station and the Dome C summit (Figure 2a). The SMBδ15N history reconstructed from ABN1314- 103 covers the period from –47 to 649 years before present (BP, where present = 1950 CE) and has values ranging from 49 to 208 kg m⁻² a⁻¹ (Figure 3a). Each SMBδ15N value integrates an average of 2.4 years of accumulation (total range: 0.7–4.5 years), and thus any impacts from individual precipitation events or seasonal extremes are largely moderated. Overall, the SMBs have fairly high variability (coefficient of variation = 0.21). The mean SMBδ15N in the 20th century (126±26.5 kg m⁻² a⁻¹) is 34% greater than the mean SMBδ15N before 1900 CE (94±18 kg m⁻² a⁻¹) and nearly 52% greater than the driest century that spans the 1600s CE (83±20 kg m⁻² a⁻¹) (Figure 3a).
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3. (a) Surface mass balance reconstruction for Aurora Basin North based on \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) arc data from the ABN1314-103 ice core. Reconstructed \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) values are shown by the red stepped lines with the 50-yr running mean \(\pm 1\sigma\) overlaid as a darker thick line and shaded zone. (b) Comparison of SMBs reconstructed from \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) (red) with those from ice density (gray) and upstream GPR isochron depth \(^{39}\) . The \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) and \(\mathrm{SMB}_{\mathrm{GPR}}\) values were aggregated to match the 1-m resolution of the \(\mathrm{SMB}_{\mathrm{density}}\) data. For \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) and \(\mathrm{SMB}_{\mathrm{density}}\) , smoothed LOESS curves are overlaid to more clearly show long-term patterns. (c) \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) values after the upstream topographic impact on SMBs has been removed, with 50-yr running mean \(\pm 1\sigma\) values overlaid. The resulting residuals may better illustrate SMB variability due to climate change. </center>
+
+<--- Page Split --->
+
+## Discussion: Validating the SMB \(\delta 15\mathrm{N}\) proxy reconstruction  
+
+We verified our new proxy's accuracy by comparing the SMB \(\delta 15\mathrm{N}\) values with SMBs derived from the physical ice density (SMBdensity) measurements of the same core. For each 1- m core segment of ABN1314- 103, we calculated a SMBdensity value by dividing the segment's mass (kg) by both its volume \((\mathrm{m}^3)\) and the age difference between the top and bottom of the segment (a \(\mathrm{m}^{- 1}\) ). The SMB \(\delta 15\mathrm{N}\) (aggregated to match the 1- m resolution) and SMB density share very similar mean values (100.8 vs. 98.0 kg \(\mathrm{m}^{- 2}\mathrm{a}^{- 1}\) , respectively) and total SMB ranges (62.0–157.3 vs. 61.7–153.4 kg \(\mathrm{m}^{- 2}\mathrm{a}^{- 1}\) , respectively), and the two SMB reconstructions have a similar pattern of variation with a moderate linear correlation \((\mathrm{r} = +0.46\) , \(\mathrm{p}< 0.001\) , \(\mathrm{n} = 90\) ) (Figure 3b). This agreement in mean value, range, and variability strongly validates our SMB \(\delta 15\mathrm{N}\) approach and the potential of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) as an accurate proxy for paleoenvironmental change.  
+
+Interpreting the ABN1314- 103 SMB profile is more complicated than for ice cores drilled at dome summits because the ice sheet at the ABN drilling site is flowing at a rate of \(16.2\mathrm{m}\mathrm{a}^{- 1}\) 40. This means that the ice in ABN1314- 103 actually fell as snow along a continuous \(11.5\mathrm{km}\) transect upstream of the current ABN drilling site, with the oldest and deepest ice originating from the most distant upstream position. As a result, the \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) and core density have recorded any spatial SMB variability that existed along the upstream transect in addition to any SMB changes due to wetting or drying of the regional climate. Although overall elevation gain is small along the transect ( \(< 15\mathrm{m}\) ), the region has abundant \(0.5–1\mathrm{m}\) undulations in surface topography extending over horizontal extents of \(3–10\mathrm{km}^{36}\) . While the MAR's horizontal grid size (35 km) cannot resolve the SMB impact from these features, ground penetrating radar (GPR) performed along the upstream transect revealed that these
+
+<--- Page Split --->
+
+surface slope changes correlate with SMB variations of up to \(40\mathrm{kgm^{- 2}a^{- 1}}\) as determined by isochronic internal reflection horizons \(^{39,41}\) (Figure 5). Although the long- term stability of such features is not well understood, the current surface features are still largely identifiable as buried horizons to depths below the deepest segment of ABN1314- 103 with only steady horizontal offset due to ice flow.  
+
+By relating the upstream topographic- driven changes in SMB revealed by GPR to core depth through the horizontal ice flow rate and the core age- depth model \(^{39}\) , we can determine the expected SMB signal due only to upstream surface topography (SMB \(_{\mathrm{GPR}}\) ). We find that the general pattern of variability in SMB \(_{\mathrm{GPR}}\) correlates very well with the patterns recorded in the SMB \(_{\delta 15\mathrm{N}}\) ( \(r = +0.74\) ) and SMB \(_{\mathrm{density}}\) ( \(r = +0.63\) ) records (Figure 3b). Thus, it appears that the primary SMB pattern preserved in ABN1314- 103 is driven by upstream changes in surface slope, which is important for properly interpreting other environmental proxies contained in the ice and for understanding the local ice flow history.  
+
+## Discussion: Extracting a climate-driven SMB record  
+
+To examine whether a secondary signal related to climate change was also preserved, we removed the spatial impact of upstream topography by subtracting the SMB \(_{\mathrm{GPR}}\) data from the SMB \(_{\delta 15\mathrm{N}}\) record. After this “upstream effect detrending” and accounting for the small offset in mean SMB values ( \(3.7\mathrm{kgm^{- 2}a^{- 1}}\) ), we find that the multi- decadal SMB values have been generally stable over the past 700 years (Figure 3c), with 50- yr running averages of the SMB never greater or less than \(15\mathrm{kgm^{- 2}a^{- 1}}\) from the detrended mean. These running averages suggest that drier conditions existed at ABN between 1600 and 1890 CE (partially corresponding to the Little Ice Age) and that precipitation has increased in the most recent 100–150 years. This is generally consistent with what has been observed at other East
+
+<--- Page Split --->
+
+Antarctic sites \(^{42 - 44}\) and for Antarctica as a whole \(^{11}\) , but we recognize that this pattern is similar to the upstream topographic effect and that it might also arise if the SMB \(_{\text{GRR}}\) record is excessively smoothed relative to true topographic- driven SMB variability (perhaps by the GPR data processing).  
+
+On shorter timescales, SMBs frequently change by \(\approx 50 \mathrm{kg m^{- 2} a^{- 1}}\) around a common mean within 10- 20 year periods. This pattern likely reflects the high interannual snowfall variability expected at sites like ABN \(^{13}\) . Located at the transition between the coast and the interior East Antarctic Plateau, annual snow accumulation at ABN is sensitive to chance intrusions of extreme precipitation events and atmospheric rivers \(^{45,46}\) , and the observed sub- decadal SMB \(_{\beta 15N}\) variability may represent the frequency of their stochastic occurrence at the site. Additionally, small scale surface roughness features like sastrugi may affect hyperlocal SMB (i.e., the SMB at scales of \(< 1 \mathrm{m}\) ) through periods of enhanced accumulation and erosion as they migrate and evolve on the snow surface \(^{47 - 49}\) . While the temporal evolution and possible life cycle cyclicity of surface roughness features are as yet poorly known, hyperlocal changes in SMB could also explain some of the short- term SMB variability observed in the ABN record if the sampling interval is shorter than the average duration of a surface feature at a given location.  
+
+## Discussion: Applied use and potential of the SMB \(\alpha \beta \mathrm{SN}\) proxy  
+
+With over 8 million \(\mathrm{km^2}\) of Antarctica having a SMB between 40 and \(200 \mathrm{kg m^{- 2} a^{- 1}}\) \(^{12}\) and over \(70\%\) of the ice sheet area modeled to have \(\delta^{15}\mathrm{NNO_3}\) values markedly elevated by photolysis (Figure S6, Supplemental Text 4), the SMB \(\alpha \beta \mathrm{SN}\) proxy holds great potential for vastly expanding our knowledge of Antarctic SMB variability over time and space. Currently, regions with moderate SMBs have only a handful of sites with SMB records older
+
+<--- Page Split --->
+
+than 200 years, with the East Antarctic Plateau particularly poorly represented<sup>11</sup>. For ice coring projects in these regions, the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy will excel at capturing the local effects of strong winds, irregular surface topography, and high interannual snowfall variability better than water isotopic techniques while avoiding problems with layer thinning and density modeling that affect \(\mathrm{SMB}_{\mathrm{density}}\) methods. As regional climate models still struggle to accurately simulate drifting snow and sublimation fluxes in the coast- to- plateau transition<sup>12</sup>, \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) can provide critical ground- based data for models predicting future contributions to sea level rise. The \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy also holds particular value for helping constrain and validate models of upstream flow effects in research targeting ice streams and broad- scale glacial flow patterns.  
+
+Additionally, sampling for the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy can save valuable time and cost compared to existing alternatives in order to expand current records of modern SMBs. Obtaining new ground- based SMBs for sites without annually resolved layers requires either coring several meters to the increasingly buried Pinatubo volcanic horizon or repeated visits to newly installed stake transects. However, limited time and resources for research expeditions to remote areas precludes intensive SMB surveys with these methods. With the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy, a mean site SMB could be determined with only a series of shallow snow or firm samples extending deep enough into the archived zone to cover only a few seasonal cycles (much shallower than the Pinatubo horizon). After proper mixing, only \(\sim 0.3 - 1.0 \mathrm{kg}\) would need to be kept, transported, and analyzed for each sample, which logistically allows for the rapid collection of robust SMB site means in many locations. On- site melting and \(\mathrm{NO}_3^-\) concentration could further reduce logistical requirements.  
+
+The \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy promises to grow and adapt as studies on Antarctic \(\mathrm{NO}_3^-\) dynamics continue. Because the resolution of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) sampling is limited only by the minimum
+
+<--- Page Split --->
+
+amount of \(\mathrm{NO_3^- }\) needed for analysis, very finely- resolved \(\delta^{15}\mathrm{N_{NO3arc}}\) records can be obtained by increasing the mass of ice collected per depth unit (e.g., by specifically drilling whole cores or replicate cores for \(\mathrm{NO_3^- }\) isotopes) and with advances in \(\mathrm{NO_3^- }\) isotopic analysis expected in the near future<sup>50</sup>. This may allow for more precise multi- annual aggregations for \(\mathrm{SMB_{\delta 15N}}\) reconstructions and permit a deeper examination of subannual \(\mathrm{NO_3^- }\) dynamics. Finally, SMBs from parts of the West Antarctic ice sheet and the highest elevations of the northern Greenland ice sheet fall within the appropriate range for the \(\mathrm{SMB_{\delta 15N}}\) proxy, and additional field sampling at those locations may allow us to reconstruct SMBs by verifying or adapting the relationship defined here for regional use outside of East Antarctica. Given the great potential of the \(\mathrm{SMB_{\delta 15N}}\) proxy to advance our understanding of the Antarctic environment and its sensitivity to climate change, we strongly recommend that potential ice coring projects incorporate \(\mathrm{NO_3^- }\) analyses into their planning and urge continued studies on Antarctic \(\mathrm{NO_3^- }\) dynamics.  
+
+## Acknowledgements:  
+
+We express thanks to the following individuals for project assistance and data support: Sarah Albertin, Selin Bagci, Albane Barbero, Mathieu Casado, Armelle Crouzet, Vincent Favier, Elsa Gautier, Gaspard Jannot, Alexis Lamothe, Anaïs Orsi, Fred Parrenin, Holly Winton, and the overwintering crews at Concordia Station. We acknowledge the logistical support of IPEV for the French missions in Antarctica, the IPEV and PNRA colleagues and overwintering crews at Concordia Station, and the JARE54 traverse team for fieldwork assistance and access to the S80 site data. We thank the co- investigators for the ABN drilling project (Jérome Chappellaz, Dorthe Dahl- Jensen, David Etheridge, Joe McConnell, Andrew Moy, Steven Phipps, Andrew Smith, Tessa Vance, Meredith Nation) and the French National
+
+<--- Page Split --->
+
+Center for Coring and Drilling (C2FN, funded by INSU) for critical drilling, logistic, and analytical support at ABN and other sites. Finally, we acknowledge the Glacioclism- SAMBA, ITASE, and IPICS 2kyr Array programs for SMB data, the Air- O- Sol facility at IGE for microbial culturing, and additional support from the MITACS Globalink program and JSPS- CNRS joint research program.  
+
+## Funding:  
+
+SCADI: European Horizon Marie Sklodowska- Curie individual fellowship 889508  LabEx OSUG@2020 Investissements d'avenir: ANR10 LABX56  EAIIST: ANR- 16- CE01- 0011- 01  BNP- Paribas Climate Initiative programs 1115 (CHICTABA), 1117 (CAPOXI 35- 75), and 1169 (EAIIST)  CLIMCOR Equipex: ANR- 11- EQPX- 0009  MEXT/JSPS KAKENHI: 20H0496  
+
+## Author contributions:  
+
+Conceptualization: PDA, JS, NC, MC  Investigation: All authors  Formal analysis: PDA, APMS, ELM  Visualization: PDA, ELM  Funding acquisition: PDA, JS, MC  Writing - original draft: PDA, APMS, PC  Writing - review & editing: All authors  
+
+The authors declare that they have no competing interests  
+
+Data and materials availability: Data and code are available for reviewers at doi:10.5281/zenodo.5793694. Data have been submitted to PANGAEA and are in review for publishing.  
+
+## Methods  
+
+## Mathematical framework for \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) and SMB relationships  
+
+A linear relationship between \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) and the reciprocal of surface mass balance (SMB \(^{- 1}\) ) has been previously observed and reported in Antarctica \(^{17,34,26}\) . Here, we mathematically illustrate how this relationship between \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) and SMB arises through photolysis of
+
+<--- Page Split --->
+
+\(\mathrm{NO}_3^-\) . We focus solely on the characteristics of \(\mathrm{NO}_3^-\) contained within a given horizontal plane of snow that is located at the snowpack surface at \(t = 0\) . We assume simplified sites with a stable surface mass balance \((SMB)\) , clear sky conditions, no surface roughness, and no significant compaction with burial in the photic zone. Any \(\mathrm{NO}_3^-\) that is photolyzed is immediately and permanently removed from the plane of snow, and \(\mathrm{NO}_3^-\) recycling \(^{29,34}\) is assumed not to affect \(\mathrm{NO}_3^-\) in the plane of snow during the burial process modeled here (i.e., after \(t = 0\) ).  
+
+Defining the relationship between \(\delta^{5}N_{\mathrm{NO3arc}}\) and SMB The time that it takes for a given horizontal plane of snow to be buried from the surface to a particular depth \(z\) is determined by the SMB \((\mathrm{kg m}^{- 2} \mathrm{a}^{- 1}\) , converted to \(\mathrm{cm s}^{- 1}\) ):  
+
+\[t_{(z)} = \frac{z}{SMB} \quad (2)\]  
+
+The concentration of \(\mathrm{NO}_3^-\) within a plane of snow decays through time according to:  
+
+\[\frac{d[N03]}{dt} = -J_{(z)}[NO3]_{(t)} \quad (3)\]  
+
+where \(J_{(z)}\) is the photolytic rate constant at a given depth defined as:  
+
+\[J_{(z)} = \sigma \phi I_{(z)} \quad (4)\]  
+
+where \(\sigma\) is the absorption cross section for \(\mathrm{NO}_3^-\) photolysis \((\mathrm{cm}^2)\) , \(\phi\) is the quantum yield for \(\mathrm{NO}_3^-\) photolysis (molec photon \(^{- 1}\) ), and \(I_{(z)}\) is the actinic flux of ultraviolet irradiance (photon \(\mathrm{cm}^{- 2} \mathrm{s}^{- 1} \mathrm{nm}^{- 1}\) ) integrated over wavelengths that can induce photolysis of \(\mathrm{NO}_3^-\) . However, this photolytic rate "constant" changes with depth because actinic flux exponentially decays with depth as:  
+
+\[I_{(z)} = I_0 e^{\frac{-z}{z_e}} \quad (5)\]  
+
+where \(I_0\) is the initial actinic flux that strikes the snow surface and \(z_e\) is the \(e\) - folding depth (cm) of the snowpack. Note that non- exponential decay of \(I\) in the top \(\sim 2\) cm of snowpack \(^{30}\) is simplified here by assuming the decay to be exponential from the snow surface. Equation (3) can then be expressed as:  
+
+\[\frac{d[N03]}{dt} = -\sigma \phi I_0 e^{\frac{-z}{z_e}} [NO3]_{(t)} \quad (6)\]  
+
+Through Eq. (2), we can rewrite Eq. (6) as:
+
+<--- Page Split --->
+
+\[\frac{d[N O3]}{d t} = -\sigma \phi I_{o} e^{\frac{-S M B t}{z_{e}}} [N O3]_{(t)} \quad (7)\]  
+
+In order to determine the \(\mathrm{NO}_3^-\) concentration at a given depth (i.e., \(SMB \cdot t\) ), we derive:  
+
+\[\frac{d[N O3]}{[N O3]_{(t)}} = -\sigma \phi I_{o} e^{\frac{-S M B t}{z_{e}}} d t \quad (8)\]  
+
+And integrate to produce:  
+
+\[\ln [N O3]_{(t)} = \frac{\sigma\phi I_{o}z_{e}e^{\frac{-S M B t}{z_{e}}}}{S M B} +C \quad (9)\]  
+
+Which simplifies to:  
+
+\[[N O3]_{(t)} = e^{c}\frac{\sigma\phi I_{o}z_{e}e^{\frac{-S M B t}{z_{e}}}}{S M B} \quad (10)\]  
+
+At \(t = 0\) , \([N O_3^- ]_{(t)} = [N O_3^- ]_0\) and therefore:  
+
+\[e^{c} = [N O3]_{0}e^{\frac{-\sigma\phi I_{o}z_{e}}{S M B}} \quad (11)\]  
+
+And thus combining Eq. (10) and Eq. (11):  
+
+\[[N O3]_{(t)} = [N O3]_{0}e^{\frac{-\sigma\phi I_{o}z_{e}}{S M B}}e^{\frac{\sigma\phi I_{o}z_{e}e^{\frac{-S M B t}{z_{e}}}}{S M B}} = [N O3]_{0}e^{\frac{\sigma\phi I_{o}z_{e}e^{\frac{-S M B t}{z_{e}}}}{S M B}} \quad (12)\]  
+
+According to Eq. (12), as time (i.e., burial depth) increases, the \(\mathrm{NO}_3^-\) concentration will decrease. However, the rate of decrease will lessen over time as the value of \(SMB \cdot t\) approaches \(3z_{e}\) , and below the photic zone (i.e., \(z > 3z_{e}\) ) the \(\mathrm{NO}_3^-\) concentration is largely stable and equal to \(e^{c}\) .  
+
+Therefore, we can calculate the fraction of \(\mathrm{NO}_3^-\) archived below the photic zone \((f_{NO3arc})\) as:  
+
+\[f_{NO3arc} = \frac{e^{c}}{[NO3]_{0}} = \frac{[\sigma\phi I_{o}z_{e}e^{\frac{-\sigma\phi I_{o}z_{e}}{S M B}}]}{[NO3]_{0}} = \frac{-\sigma\phi I_{o}z_{e}e^{\frac{-\sigma\phi I_{o}z_{e}}{S M B}}}{[NO3]_{0}} \quad (13)\]  
+
+To determine the \(\delta^{15}\mathrm{N}_{NO3arc}\) of this \(\mathrm{NO}_3^-\) , Rayleigh fractionation states that \(\delta^{15}\mathrm{N}_{NO3}\) can be calculated with the fractionation factor \(a\) by:  
+
+\[\ln \left(\delta^{15}N_{NO3arc} + 1\right) = (\alpha -1)\ln \left(f_{NO3arc}\right) + \ln \left(\delta^{15}N_{NO3} + 1\right) \quad (14)\]  
+
+Through our prior calculation of \(f_{NO3arc}\) in Eq. (13), we thus produce:  
+
+\[\ln \left(\delta^{15}N_{NO3arc} + 1\right) = (\alpha -1)\frac{-\sigma\phi I_{o}z_{e}}{S M B} +\ln \left(\delta^{15}N_{NO3} + 1\right) \quad (15)\]  
+
+Because \((\alpha - 1)\) is negative for nitrogen during photolysis of \(\mathrm{NO}_3^{- 21,22,31,51 - 53}\) and the other parameters are positive, this means that \(\delta^{15}\mathrm{N}_{NO3arc}\) will vary linearly and positively with \(SMB^{- 1}\) when other parameters are held constant or scale linearly with \(SMB^{- 1}\) . We examine the
+
+<--- Page Split --->
+
+potential impacts of variability in these other parameters more thoroughly in Supplementary Text 1.  
+
+Based on modeling and field observations, SMB is the primary driver of change in \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values. Thus, the non- SMB variables can be subsumed into two parameters \(A\) and \(B\) to function as linear regression coefficients, producing Eq. (1) of the main text:  
+
+\[\ln \left(\delta^{15}N_{NO3arc} + 1\right) = \frac{A}{SMB} +B \quad (Eq. (1))\]  
+
+The inverse function of Eq. (1) can be used as a transfer function to calculate an SMB based on a \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value:  
+
+\[\frac{1}{SMB} = \frac{\ln(\delta^{15}N_{NO3arc} + 1) - B}{A} \quad (Eq. (16))\]  
+
+Finally, since \(\ln (x + 1) \approx x\) when \(x \approx 0\) , a simpler relationship of Eq. (15) can be approximated, in a form similar to that previously reported from field observations \(^{23,26,34}\) :  
+
+\[\delta^{15}N_{NO3arc} = (\alpha -1)\frac{-\sigma\phi l_oze}{SMB} +\delta^{15}N_{NO3o} \quad (Eq. (17))\]  
+
+## Snow sampling techniques  
+
+The \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values in our database are taken from a mix of previously reported values from Antarctic research traverses and values newly reported here (Figure 2). For all values, snow and ice containing \(\mathrm{NO}_3^-\) was sampled in the field in one of three techniques: 1) 1- 2 m deep snow pit with continuous sampling at regular intervals from top to bottom, 2) single sample taken of a well- mixed 5- 10 cm layer around the 1- m depth layer, and 3) drilled core later cut at desired intervals. Since current \(\mathrm{NO}_3^-\) isotopic analysis requires 50- 150 nmol of \(\mathrm{NO}_3^-\) , 0.25- 1.50 kg of snow or ice per sample \(^{19,21}\) are gathered to ensure a sufficient amount of \(\mathrm{NO}_3^-\) . Generally, the multiple samples produced by the snow pit technique offers the best and most flexible results, but the 1- m depth layer technique is valuable for quick sampling during limited stops, and cores are necessary to collect samples deeper than \(\approx 5 \mathrm{~m}\) .  
+
+## Laboratory analyses  
+
+For \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) results included in our database that have been previously reported, readers are directed to the original papers for specific analytical and sampling techniques. For the
+
+<--- Page Split --->
+
+\(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) data newly reported here, snow and ice samples were collected into clean sealed plastic bags or tubs and stored frozen until melted at room temperature for analysis. The \(\mathrm{NO_3^- }\) mass fraction \((\omega (\mathrm{NO}_3^- ))\) was determined on aliquots by either a colorimetric method or ion chromatography with detection limits \(< 0.5\mathrm{ng}\mathrm{g}^{- 1}\) and precision of \(< 3\%^{21,22}\) . The remaining melted samples were passed through an anionic exchange resin (Bio- Rad™ AG 1- X8, chloride form), and the resulting trapped \(\mathrm{NO_3^- }\) was eluted with \(10\mathrm{ml}\) of NaCl 1 M solution. Isotopic analysis occurred at IGE- CNRS, Grenoble, France, where \(\mathrm{NO_3^- }\) in these samples was converted to \(\mathrm{N}_2\mathrm{O}\) with the denitrifying bacteria Pseudomonas aureofaciens (lacking nitrous oxide reductase), thermally decomposed into \(\mathrm{O_2}\) and \(\mathrm{N}_2\) on a \(900^{\circ}\mathrm{C}\) gold surface, and separated by gas chromatography with a GasBench IITM. Oxygen and nitrogen isotopic ratios were then measured on a Thermo Finnigan™ MAT 253 mass spectrometer \(^{54 - 57}\) . Isotopic effects from this analysis were corrected as described by Morin et al. (2009) and Frey at al. (2009), using the international reference materials USGS 32, USGS 34, and USGS 35 with ultrapure Dome C water used for standards and samples throughout the analyses to account for potential oxygen isotopic exchanges. Results are reported relative to Vienna Standard Mean Ocean Water (V- SMOW) for oxygen isotopes \(^{58}\) and \(\mathrm{N}_2\) - Air for nitrogen isotopes \(^{59}\) .  
+
+For snow pits with multiple sequential \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values, a single \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value was calculated as the aggregate of samples \(30+\) cm deep, weighted by the relative mass of \(\mathrm{NO_3^- }\) per sample. Although the photic zone boundary can extend lower than \(30\mathrm{cm}\) at some sites \(^{29,30}\) , this cutoff was deemed an acceptable compromise to include more data from pits that stopped at \(50\mathrm{cm}\) depth as the great majority of photolysis will have occurred within the top \(30\mathrm{cm}\) due to exponential decay of actinic flux and \(\omega (\mathrm{NO}_3^- )\) with depth. Exceptions to this were made for three coastal pits from Cap Prud'homme (weighted- means of \(3+\) cm samples), where high accumulation greatly reduces photolytic impact, higher snow impurities reduce the photic zone depth, and a broader aggregation is necessary to smooth seasonal cycles.
+
+<--- Page Split --->
+
+Additionally, two pits from Dronning Maud Land were aggregated with \(15+\) cm samples based on shallow \(3z_{e}\) values (2–5 cm) calculated on site during snow pit sampling<sup>29</sup>. For cores included in our database, a single \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value was calculated as the isotopic mean of samples extending from present back to no earlier than 1800 CE.  
+
+Noro et al. (2018) reported \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) values for 16 pits along the JARE54 and JARE57 transects, but the sampling methodology for these pits took a single well- mixed sample of the entire pit depth which included the entire photic zone. In order to estimate the \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values of these sites (i.e., the value as if the photic zone snow had been excluded), we applied a correction factor calculated using data from other pits in our database that were taken on two similar transects spanning from the coast to other interior domes (Dome A and Dome C) of East Antarctica<sup>22,23</sup>. Because each of the pits on the Dome A and Dome C transects were continuously sampled at discrete intervals from the surface to a point below the photic zone, we calculated different weighted- mean \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) values for selected depth spans that matched the three extents of the JARE pits: 0–30 cm, 0–50 cm, and 0–80 cm. Corrective factors were calculated through the linear regression of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) vs. \(\delta^{15}\mathrm{N}_{\mathrm{NO3.X}}\) from Dome A/Dome C transect pits (where \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) is our database's \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) value from the archived zone and \(\delta^{15}\mathrm{N}_{\mathrm{NO3.X}}\) is the weighted- mean value of samples from the surface to depth \(x\) : 30, 50, or 80 cm) and applied to the JARE pit data through the appropriate depth correction (Table 1, 2). Corrections were not made for JARE samples where \(\delta^{15}\mathrm{N}_{\mathrm{NO3}} < 0\%\) , as these low \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) values strongly suggest that photolysis was not a significant factor at these coastal sites, and photic zone corrections were thus not warranted.  
+
+Table 1. Linear regressions of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) vs. \(\delta^{15}\mathrm{N}_{\mathrm{NO3.X}}\) (where X is 30, 50, or 80 cm) calculated from nonJARE pit data in the \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) database.   
+
+<table><tr><td>Depth correction</td><td>Slope (% / %)</td><td>Intercept (%)</td><td>r²</td></tr><tr><td>0–30 cm</td><td>1.9±0.1</td><td>-2.4±11.3</td><td>0.89</td></tr><tr><td>0–50 cm</td><td>1.6±0.1</td><td>-1.7±8.2</td><td>0.94</td></tr><tr><td>0–80 cm</td><td>1.5±0.1</td><td>-0.9±7.8</td><td>0.94</td></tr></table>
+
+<--- Page Split --->
+
+
+Table 2. The \(\delta^{15}\mathrm{N_{NO3}}\) values for JARE sites included in our database as originally reported by Noro et al. (2018) and the \(\delta^{15}\mathrm{N_{NO3arc}}\) values corrected here to account for photic zone snow included in the original samples. Samples with original \(\delta^{15}\mathrm{N_{NO3}}\) values \(< 0\%\) (italicized) were not corrected.   
+
+<table><tr><td>JARE site</td><td>Depth (cm)</td><td>Original δ15NNO3 (%)</td><td>Corrected δ15NNO3arc (%)</td></tr><tr><td>Z2</td><td>0–80</td><td>20.6</td><td>30.8</td></tr><tr><td>IM0</td><td>0–50</td><td>25.7</td><td>40.3</td></tr><tr><td>NMD304</td><td>0–50</td><td>41.1</td><td>65.4</td></tr><tr><td>MD590</td><td>0–50</td><td>83.5</td><td>134.7</td></tr><tr><td>DF1</td><td>0–30</td><td>127.3</td><td>236.5</td></tr><tr><td>NDF</td><td>0–30</td><td>111.7</td><td>207.2</td></tr><tr><td>Plateau S</td><td>0–30</td><td>165.5</td><td>308.1</td></tr><tr><td>S80</td><td>0–30</td><td>90.7</td><td>167.8</td></tr><tr><td>Fuji Pass</td><td>0–30</td><td>74.3</td><td>137.0</td></tr><tr><td>DF2</td><td>0–30</td><td>118.6</td><td>220.1</td></tr><tr><td>S30</td><td>0–50</td><td>-19.0</td><td>-19.0</td></tr><tr><td>H42</td><td>0–50</td><td>-6.6</td><td>-6.6</td></tr><tr><td>H68</td><td>0–50</td><td>-14.5</td><td>-14.5</td></tr><tr><td>H88</td><td>0–50</td><td>-19.4</td><td>-19.4</td></tr><tr><td>H108</td><td>0–50</td><td>-6.4</td><td>-6.4</td></tr><tr><td>H128</td><td>0–50</td><td>14.1</td><td>21.3</td></tr></table>  
+
+## SMB data  
+
+In our database, 74 \(\delta^{15}\mathrm{N_{NO3arc}}\) samples are represented by 51 unique direct ground measurements of SMB (SMBground) values observed at or near the \(\mathrm{NO_3}\) sampling site, with the numerical discrepancy due to some sites having replicate \(\delta^{15}\mathrm{N_{NO3arc}}\) samples. These previously reported SMBground values were determined by measuring the change in surface height on established stakes or poles, by measuring the mass between known volcanic or radioactivity horizons in an ice core, or by ground penetrating radar (GPR) identification of dated horizons \(^{10,22,23,60 - 67}\) .  
+
+Regional climate models can be used to estimate modern SMB rates for sites lacking ground observations \(^{7,12}\) , and we used the Modèle Atmosphérique Régional (MAR) version 3.6.4 with European Centre for Medium- Range Weather Forecasts “Interim” re- analysis data (ERA- interim) data as applied by Agosta et al. (2019) to model mean annual SMBs at all database sites for the period 1979–2017 \(^{12}\) . Because the MAR overestimates SMBs at higher and more interior sites of the East Antarctic plateau \(^{68}\) , we calculated a correction factor through linear
+
+<--- Page Split --->
+
+regressions of \(\mathrm{SMB}_{\mathrm{ground}}\) vs. MAR-estimated SMBs ( \(\mathrm{SMB}_{\mathrm{MAR}}\) ) for our 51 sites that have both values (Table 3, Figure 4). This correction was applied to all original MAR estimates to produce "adjusted-MAR" SMBs ( \(\mathrm{SMB}_{\mathrm{adjMAR}}\) ) that match more closely with ground observations. 
+
+Table 3. A list of all sampling sites that have a \(\mathrm{SMB}_{\mathrm{ground}}\) observation with corresponding values of original \(\mathrm{SMB}_{\mathrm{MAR}}\) and \(\mathrm{SMB}_{\mathrm{adjMAR}}\) (Figure 4). The difference between the \(\mathrm{SMB}_{\mathrm{adjMAR}}\) and \(\mathrm{SMB}_{\mathrm{MAR}}\) values is given in the final column. 
+
+<table><tr><td>Site</td><td>\(\mathrm{SMB}_{\mathrm{ground}}\) (kg m⁻² a⁻¹)</td><td>\(\mathrm{SMB}_{\mathrm{MAR}}\) (kg m⁻² a⁻¹)</td><td>\(\mathrm{SMB}_{\mathrm{adjMAR}}\) (kg m⁻² a⁻¹)</td><td>\(\mathrm{SMB}_{\mathrm{AdjMAR}} - \mathrm{SMB}_{\mathrm{MAR}}\) (kg m⁻² a⁻¹)</td><td>\(\mathrm{SMB}_{\mathrm{adjMAR}} / \mathrm{SMB}_{\mathrm{MAR}}\)</td></tr><tr><td>Vostok</td><td>22.6</td><td>30.4</td><td>24.1</td><td>-6.3</td><td>0.79</td></tr><tr><td>DomeA</td><td>22.9</td><td>40.9</td><td>34.4</td><td>-6.5</td><td>0.84</td></tr><tr><td>ZtoA-P6</td><td>25.4</td><td>62.1</td><td>55.2</td><td>-6.9</td><td>0.89</td></tr><tr><td>DomeC</td><td>28.4</td><td>41.0</td><td>34.5</td><td>-6.5</td><td>0.84</td></tr><tr><td>DomeF</td><td>29.2</td><td>35.6</td><td>29.2</td><td>-6.4</td><td>0.82</td></tr><tr><td>NDF</td><td>30.9</td><td>33.2</td><td>26.8</td><td>-6.4</td><td>0.81</td></tr><tr><td>Plateau S</td><td>32.4</td><td>31.2</td><td>24.8</td><td>-6.4</td><td>0.79</td></tr><tr><td>ZtoA-P5</td><td>33.3</td><td>61.5</td><td>54.6</td><td>-6.9</td><td>0.89</td></tr><tr><td>preeaiist.18</td><td>34.0</td><td>47.8</td><td>41.1</td><td>-6.7</td><td>0.86</td></tr><tr><td>S80Jare</td><td>37.5</td><td>31.3</td><td>24.9</td><td>-6.4</td><td>0.80</td></tr><tr><td>MD590</td><td>37.9</td><td>42.6</td><td>36.0</td><td>-6.6</td><td>0.85</td></tr><tr><td>Fuji Pass</td><td>40.7</td><td>35.0</td><td>28.6</td><td>-6.4</td><td>0.82</td></tr><tr><td>ZtoA-P4</td><td>54.8</td><td>55.4</td><td>48.6</td><td>-6.8</td><td>0.88</td></tr><tr><td>NMD304</td><td>65.8</td><td>75.6</td><td>68.4</td><td>-7.2</td><td>0.90</td></tr><tr><td>IM0</td><td>68.5</td><td>99.2</td><td>91.6</td><td>-7.6</td><td>0.92</td></tr><tr><td>Kohnen</td><td>75.0</td><td>97.6</td><td>90.0</td><td>-7.6</td><td>0.92</td></tr><tr><td>posteaiist.asuma05</td><td>76.0</td><td>288.9</td><td>266.9</td><td>-22.0</td><td>0.92</td></tr><tr><td>preeaiist.15</td><td>80.0</td><td>70.0</td><td>62.9</td><td>-7.1</td><td>0.90</td></tr><tr><td>preeaiist.13</td><td>86.0</td><td>95.1</td><td>87.5</td><td>-7.6</td><td>0.92</td></tr><tr><td>ZtoA-P3</td><td>90.7</td><td>76.3</td><td>69.1</td><td>-7.2</td><td>0.91</td></tr><tr><td>CPH.D5</td><td>97.6</td><td>139.7</td><td>124.8</td><td>-8.3</td><td>0.89</td></tr><tr><td>ZtoA-P2</td><td>99.4</td><td>95.6</td><td>88.0</td><td>-7.6</td><td>0.92</td></tr><tr><td>Z2</td><td>113.5</td><td>116.9</td><td>108.9</td><td>-8.0</td><td>0.93</td></tr><tr><td>posteaiist.stop36</td><td>118.0</td><td>147.5</td><td>138.3</td><td>-9.2</td><td>0.94</td></tr><tr><td>CPH.D24</td><td>120.0</td><td>247.2</td><td>228.9</td><td>-18.3</td><td>0.93</td></tr><tr><td>preeaiist.12</td><td>130.0</td><td>113.3</td><td>105.4</td><td>-7.9</td><td>0.93</td></tr><tr><td>ABN</td><td>130.0</td><td>122.0</td><td>113.9</td><td>-8.1</td><td>0.93</td></tr><tr><td>posteaiist.asuma06</td><td>131.0</td><td>280.4</td><td>259.1</td><td>-21.3</td><td>0.92</td></tr><tr><td>H128</td><td>158.8</td><td>210.2</td><td>195.3</td><td>-14.9</td><td>0.93</td></tr><tr><td>preeaiist.06</td><td>160.0</td><td>286.2</td><td>264.4</td><td>-21.8</td><td>0.92</td></tr><tr><td>preeaiist.07</td><td>171.0</td><td>269.5</td><td>249.2</td><td>-20.3</td><td>0.92</td></tr><tr><td>ZtoA-P1</td><td>172.0</td><td>153.4</td><td>143.7</td><td>-9.7</td><td>0.94</td></tr><tr><td>H108</td><td>185.0</td><td>250.7</td><td>232.1</td><td>-18.6</td><td>0.93</td></tr><tr><td>preeaiist.09</td><td>198.0</td><td>201.0</td><td>186.9</td><td>-14.1</td><td>0.93</td></tr><tr><td>H88</td><td>207.3</td><td>245.5</td><td>227.4</td><td>-18.1</td><td>0.93</td></tr><tr><td>posteaiist.asuma04</td><td>215.0</td><td>304.2</td><td>280.8</td><td>-23.4</td><td>0.92</td></tr><tr><td>posteaiist.asuma09</td><td>219.0</td><td>256.6</td><td>237.5</td><td>-19.1</td><td>0.93</td></tr><tr><td>H68</td><td>225.4</td><td>255.7</td><td>236.7</td><td>-19.0</td><td>0.93</td></tr><tr><td>H42</td><td>234.9</td><td>272.5</td><td>251.9</td><td>-20.6</td><td>0.92</td></tr><tr><td>posteaiist.asuma02</td><td>239.0</td><td>343.0</td><td>316.0</td><td>-27.0</td><td>0.92</td></tr><tr><td>posteaiist.asuma10</td><td>240.0</td><td>232.0</td><td>215.1</td><td>-16.9</td><td>0.93</td></tr></table>
+
+<--- Page Split --->
+
+
+<table><tr><td>posteaiist.asuma11</td><td>243.0</td><td>203.6</td><td>189.3</td><td>-14.3</td><td>0.93</td></tr><tr><td>S30-JARE</td><td>271.9</td><td>288.6</td><td>266.6</td><td>-22.0</td><td>0.92</td></tr><tr><td>posteaiist.asuma07</td><td>273.0</td><td>271.1</td><td>250.7</td><td>-20.4</td><td>0.92</td></tr><tr><td>preeaiist.04</td><td>280.0</td><td>330.1</td><td>304.3</td><td>-25.8</td><td>0.92</td></tr><tr><td>posteaiist.asuma01</td><td>321.0</td><td>337.4</td><td>310.9</td><td>-26.5</td><td>0.92</td></tr><tr><td>preeaiist.03</td><td>337.7</td><td>366.0</td><td>337.0</td><td>-29.0</td><td>0.92</td></tr><tr><td>cph.d17</td><td>446.0</td><td>178.1</td><td>166.1</td><td>-12.0</td><td>0.93</td></tr><tr><td>preeaiist.02</td><td>487.8</td><td>439.0</td><td>403.3</td><td>-35.7</td><td>0.92</td></tr><tr><td>asuma.2016.2</td><td>488.0</td><td>312.5</td><td>288.3</td><td>-24.2</td><td>0.92</td></tr><tr><td>asuma.2016.1</td><td>548.0</td><td>366.5</td><td>337.4</td><td>-29.1</td><td>0.92</td></tr></table>
+
+![](images/Figure_4.jpg)
+
+<center>Figure 4. Linear regressions of $\mathrm {SMB}_{\mathrm {ground}}$  versus $\mathrm {SMB}_{\mathrm {MAR}}$  for the period 1979-2017 at the 51 sites with $\mathrm {SMB}_{\mathrm {ground}}$  observations, with 95% confidence intervals of the regressions shaded. Sites are subset for two overlapping regressions that intersect at (138, 130). These linear regressions were applied to the $\mathrm {SMB}_{\mathrm {MAR}}$  values for all sampling sites to produce the $\mathrm {SMB}_{\mathrm {adjMA}}$  used in analyses. The dashed line represents a slope of 1 (i.e., if the $\mathrm {SMB}_{\mathrm {MAR}}$  perfectly matched the $\mathrm {SMB}_{\mathrm {ground}}$ ).</center>
+
+A linear regression was calculated for two overlapping subsets of sites: one for the set of
+
+well-grouped sites where the $\mathrm {SMB}_{\mathrm {MAR}}$  is $<175kgm^{-2}a^{-1}$  and another for all sites where
+
+$\mathrm {SMB}_{\mathrm {MAR}}$  is $>110kgm^{-2}a^{-1}$ . This first regression is tightly constrained $(SMB_{\mathrm {ground}}=1.0\pm$
+
+<--- Page Split --->
+
+\(0.1 \times \mathrm{SMB}_{\mathrm{MAR}} - 5.8 \pm 7.1\) , \(\mathrm{r}^2 = 0.84\) , and it performs well to better align the \(\mathrm{SMB}_{\mathrm{MAR}}\) estimates with the \(\mathrm{SMB}_{\mathrm{ground}}\) values at low SMB sites. The second regression covers samples with where some differences between \(\mathrm{SMB}_{\mathrm{MAR}}\) and \(\mathrm{SMB}_{\mathrm{ground}}\) are very large, particularly at lower elevation sites where intense aeolian erosion and deposition can produce highly variable local SMB rates that are difficult to accurately model \(^{12,13}\) . As a result, this regression is weaker \(\mathrm{(SMB_{ground} = 0.9 \pm 0.2 \times SMB_{MAR} + 4.2 \pm 57.9, r^2 = 0.35)}\) than the first regression, but we apply it while acknowledging the possibility of wide deviations. The two regressions intersect at \(\mathrm{(SMB_{MAR} = 138 kg m^{-2} a^{-1}}\) , \(\mathrm{SMB}_{\mathrm{ground}} = 130 \mathrm{kg m^{-2} a^{-1}}\) ), and thus \(\mathrm{SMB}_{\mathrm{adjMAR}}\) values were calculated by applying the first regression to all sites where \(\mathrm{SMB}_{\mathrm{MAR}} \leq 138 \mathrm{kg m^{-2} a^{-1}}\) and applying the second regression to all sites where \(\mathrm{SMB}_{\mathrm{MAR}} > 138 \mathrm{kg m^{-2} a^{-1}}\) . We constructed our final primary SMB dataset for the analysis of \(\delta^{5}\mathrm{N}_{\mathrm{NO3arc}}\) samples by using the best quality SMB data for each site: \(\mathrm{SMB}_{\mathrm{ground}}\) if available and \(\mathrm{SMB}_{\mathrm{adjMAR}}\) otherwise.  
+
+## Transfer function and SMB reconstruction  
+
+We modeled linear relationships between \(\ln (\delta^{5}\mathrm{N}_{\mathrm{NO3}} + 1)\) and \(\mathrm{SMB}^{- 1}\) based on Eq. (15) using previously reported parameter values to compare our theoretical framework to field results and to better understand the sensitivity of the relationships to photolytic and fractionation factors (Supplementary Text 1). To determine the coefficients in Eq. (1) from our field data, we performed linear regressions using all database samples and the primary SMB dataset. Additional regressions (Supplementary Text 2) were performed for subsets of the database based on SMB type ( \(\mathrm{SMB}_{\mathrm{ground}}\) vs. \(\mathrm{SMB}_{\mathrm{adjMAR}}\) ). With regression coefficients determined for Eq. (1), we modeled the spatial distribution of \(\delta^{5}\mathrm{N}_{\mathrm{NO3arc}}\) values across Antarctica using gridded mean SMBs (MAR- ERA- interim, 1979–2015) at a \(35 \mathrm{km}\) resolution \(^{12}\) that were converted to \(\mathrm{SMB}_{\mathrm{adjMAR}}\) as previously described.
+
+<--- Page Split --->
+
+For reconstructing the ABN \(\mathrm{SMB}_{\mathrm{015N}}\) history, the ABN1314- 103 ice core was cut into \(0.33\mathrm{m}\) samples from 5 to \(103\mathrm{m}\) , and these were processed for \(\mathrm{NO}_3^-\) isotopes in 2016 as previously described. We applied an annually- resolved age model (ALC01112018) based on seasonal ion and water isotope cycles and constrained by volcanic horizons that was originally developed for a longer core also taken at ABN. Each \(1\mathrm{m}\) ice core segment was individually weighed prior to cutting, and the mass and volume were used to calculate a SMB profile based on dated ice density changes ( \(\mathrm{SMB}_{\mathrm{density}}\) ).  
+
+To determine past topographical effects on SMBs, a MALA GPR device towing a RTA antenna on the surface (50 MHz out, 100 MHz in) was operated for a \(65\mathrm{km}\) transect upstream of the coring site as part of the 2013–2014 campaign. Radar was triggered every 2 seconds (i.e., every 6–7 m along the transect) with a recording time window of 3000 nanoseconds that captured returns down to \(300\mathrm{m}\) depth. After postprocessing<sup>41</sup>, isochronic internal reflecting horizons were identified to \(220\mathrm{m}\) depth, digitized with ReflexW software, and dated by connecting to the ALC01112018 age-depth model. Using a density profile taken from a longer ice core simultaneously drilled at ABN, 2D fields (depth by transect distance) were calculated for age, mean accumulation rate, and local accumulation rate. The mean accumulation rate to the most shallow reflecting horizon was taken as the upstream topographical effect on SMBs (i.e., \(\mathrm{SMB}_{\mathrm{GPR}}\) ).  
+
+Statistical analyses, regressions, SMB reconstructions, visualizations, and other statistical analyses were perform using the R programming language with packages ggplot2, RColorBrewer, gridExtra, cowplot, and tidyverse.
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Figure 5. Local accumulation rate variability with depth along the upstream ABN transect determined from GPR identification of isochronic internal reflective horizons. Accumulation rates have an original depth resolution of \(0.5 \mathrm{m}\) which is smoothed through a moving age-depth average with a cosine weighting window to reduce isochron artifacts<sup>41</sup>. </center>
+
+<--- Page Split --->
+
+1. Fretwell, P. et al. Bedmap2: improved ice bed, surface and thickness datasets for Antarctica. The Cryosphere 7, 375-393 (2013).  
+
+2. Juckes, M. N., James, I. N. & Blackburn, M. The influence of Antarctica on the momentum budget of the southern extratropics. Quarterly Journal of the Royal Meteorological Society 120, 1017-1044 (1994).  
+
+3. M.R. van den Broeke. On the role of Antarctica as heat sink for the global atmosphere. J. Phys. IV France 121, 115-124 (2004).  
+
+4. Bronselaer, B. et al. Change in future climate due to Antarctic meltwater. Nature 564, 53-58 (2018).  
+
+5. Starr, A. et al. Antarctic icebergs reorganize ocean circulation during Pleistocene glacials. Nature 589, 236-241 (2021).  
+
+6. IPCC. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. 151 (2014).  
+
+7. Shepherd, A. et al. Mass balance of the Antarctic Ice Sheet from 1992 to 2017. Nature 558, 219-222 (2018).  
+
+8. Martin-Español, A. et al. Spatial and temporal Antarctic Ice Sheet mass trends, glacio-isostatic adjustment, and surface processes from a joint inversion of satellite altimeter, gravity, and GPS data. Journal of Geophysical Research: Earth Surface 121, 182-200 (2016).  
+
+9. Stauffer, B., Flückiger, J., Wolff, E. & Barnes, P. The EPICA deep ice cores: first results and perspectives. Annals of Glaciology 39, 93-100 (2004).  
+
+10. Parrenin, F. et al. 1-D-ice flow modelling at EPICA Dome C and Dome Fuji, East Antarctica. Clim. Past 3, 243-259 (2007).  
+
+11. Thomas, E. R. et al. Regional Antarctic snow accumulation over the past 1000 years. Clim. Past 13, 1491-1513 (2017).  
+
+12. Agosta, C. et al. Estimation of the Antarctic surface mass balance using the regional climate model MAR (1979-2015) and identification of dominant processes. The Cryosphere 13, 281-296 (2019).  
+
+13. Agosta, C. et al. A 40-year accumulation dataset for Adelie Land, Antarctica and its application for model validation. Climate Dynamics 38, 75-86 (2012).  
+
+14. Gallee, H. et al. Transport of Snow by the Wind: A Comparison Between Observations in Adélie Land, Antarctica, and Simulations Made with the Regional Climate Model MAR. Boundary-Layer Meteorology 146, 133-147 (2013).
+
+<--- Page Split --->
+
+15. Vimeux, F., Cuffey, K. M. & Jouzel, J. New insights into Southern Hemisphere temperature changes from Vostok ice cores using deuterium excess correction. Earth and Planetary Science Letters 203, 829–843 (2002).  
+
+16. Cauquoin, A. et al. Comparing past accumulation rate reconstructions in East Antarctic ice cores using 10Be, water isotopes and CMIP5-PMIP3 models. Clim. Past 11, 355–367 (2015).  
+
+17. Freyer, H. D., Kobel, K., Delmas, R. J., Kley, D. & Legrand, M. R. First results of 15N/14N ratios in nitrate from alpine and polar ice cores. Tellus B 48, 93–105 (1996).  
+
+18. Legrand, M., Wolff, E., Wagenbach, D. & Jacka, T. Antarctic aerosol and snowfall chemistry: implications for deep Antarctic ice-core chemistry. Annals of Glaciology, Vol 29, 1999 29, 66–72 (1999).  
+
+19. Wolff, E. Nitrate in Polar Ice. in Ice Core Studies of Global Biogeochemical Cycles (Springer-Verlag, 1995).  
+
+20. Röhlisberger, R. et al. Nitrate in Greenland and Antarctic ice cores: a detailed description of post-depositional processes. Annals of Glaciology 35, 209–216 (2002).  
+
+21. Frey, M., Savarino, J., Morin, S., Erbland, J. & Martins, J. Photolysis imprint in the nitrate stable isotope signal in snow and atmosphere of East Antarctica and implications for reactive nitrogen cycling. Atmospheric Chemistry and Physics 9, 8681–8696 (2009).  
+
+22. Erbland, J. et al. Air-snow transfer of nitrate on the East Antarctic Plateau - Part 1: Isotopic evidence for a photolytically driven dynamic equilibrium in summer. Atmospheric Chemistry and Physics 13, 6403–6419 (2013).  
+
+23. Shi, G. et al. Investigation of post-depositional processing of nitrate in East Antarctic snow: isotopic constraints on photolytic loss, re-oxidation, and source inputs. Atmospheric Chemistry and Physics 15, 9435–9453 (2015).  
+
+24. Grannas, A. et al. An overview of snow photochemistry: evidence, mechanisms and impacts. Atmospheric Chemistry and Physics 7, 4329–4373 (2007).  
+
+25. Berhanu, T. et al. Laboratory study of nitrate photolysis in Antarctic snow. II. Isotopic effects and wavelength dependence. Journal of Chemical Physics 140, (2014).  
+
+26. Noro, K. et al. Spatial variation of isotopic compositions of snowpack nitrate related to post-depositional processes in eastern Dronning Maud Land, East Antarctica. Geochemical Journal 52, e7–e14 (2018).  
+
+27. Shi, G. et al. Isotope fractionation of nitrate during volatilization in snow: A field investigation in Antarctica. Geophysical Research Letters 46, 3287–3297 (2019).
+
+<--- Page Split --->
+
+28. Noro, K. & Takenaka, N. Post-depositional loss of nitrate and chloride in Antarctic snow by photolysis and sublimation: a field investigation. POLAR 39, (2020).  
+
+29. Winton, V. H. L. et al. Deposition, recycling, and archival of nitrate stable isotopes between the air-snow interface: comparison between Dronning Maud Land and Dome C, Antarctica. Atmos. Chem. Phys. 20, 5861-5885 (2020).  
+
+30. Zatko, M. C. et al. The influence of snow grain size and impurities on the vertical profiles of actinic flux and associated NOx emissions on the Antarctic and Greenland ice sheets. Atmos. Chem. Phys. 13, 3547-3567 (2013).  
+
+31. Zatko, M., Geng, L., Alexander, B., Sofen, E. & Klein, K. The impact of snow nitrate photolysis on boundary layer chemistry and the recycling and redistribution of reactive nitrogen across Antarctica and Greenland in a global chemical transport model. Atmos. Chem. Phys. 16, 2819-2842 (2016).  
+
+32. France, J. L. et al. Snow optical properties at Dome C (Concordia), Antarctica; implications for snow emissions and snow chemistry of reactive nitrogen. Atmos. Chem. Phys. 11, 9787-9801 (2011).  
+
+33. Wolff, E., Jones, A., Martin, T. & Grenfell, T. Modelling photochemical NOX production and nitrate loss in the upper snowpack of Antarctica. Geophysical Research Letters 29, (2002).  
+
+34. Erbland, J. et al. Air-snow transfer of nitrate on the East Antarctic Plateau - Part 2: An isotopic model for the interpretation of deep ice-core records. Atmospheric Chemistry and Physics 15, 12079-12113 (2015).  
+
+35. Jiang, S. et al. Nitrate preservation in snow at Dome A, East Antarctica from ice core concentration and isotope records. Atmospheric Environment 213, 405-412 (2019).  
+
+36. Howat, I. M., Porter, C., Smith, B. E., Noh, M.-J. & Morin, P. The Reference Elevation Model of Antarctica. The Cryosphere 13, 665-674 (2019).  
+
+37. Hui, F. et al. Mapping blue-ice areas in Antarctica using ETM+ and MODIS data. Annals of Glaciology 55, 129-137 (2014).  
+
+38. Savarino, J., Kaiser, J., Morin, S., Sigman, D. & Thiemens, M. Nitrogen and oxygen isotopic constraints on the origin of atmospheric nitrate in coastal Antarctica. Atmospheric Chemistry and Physics 7, 1925-1945 (2007).  
+
+39. Servettaz, A., Landais, A. & Orsi, A. Two thousand years of temperature variability on the lower East Antarctic Plateau inferred from the analysis of stable isotopes of water and inert gases in the Aurora Basin North ice core. (Université Paris-Saclay, 2021).
+
+<--- Page Split --->
+
+40. Mouginot, J., Rignot, E. & Scheuchl, B. Continent-wide, interferometric SAR phase, mapping of Antarctic ice velocity. Geophysical Research Letters 46, 9710-9718 (2019).  
+
+41. Le Meur, E. et al. Spatial and temporal distributions of surface mass balance between Concordia and Vostok stations, Antarctica, from combined radar and ice core data: first results and detailed error analysis. The Cryosphere 12, 1831-1850 (2018).  
+
+42. Stenni, B. et al. Eight centuries of volcanic signal and climate change at Talos Dome (East Antarctica). J. Geophys. Res.-Atmos. 107, 13 (2002).  
+
+43. Frezzotti, M., Scarchilli, C., Becagli, S., Proposito, M. & Urbini, S. A synthesis of the Antarctic surface mass balance during the last 800 yr. The Cryosphere 7, 303-319 (2013).  
+
+44. Urbini, S. et al. Historical behaviour of Dome C and Talos Dome (East Antarctica) as investigated by snow accumulation and ice velocity measurements. Global and Planetary Change 60, 576-588 (2008).  
+
+45. Wille, J. D. et al. Antarctic atmospheric river climatology and precipitation impacts. Journal of Geophysical Research: Atmospheres 126, (2021).  
+
+46. Turner, J. et al. The Dominant Role of Extreme Precipitation Events in Antarctic Snowfall Variability. Geophysical Research Letters 46, 3502-3511 (2019).  
+
+47. Frezzotti, M., Gandolfi, S., Marca, F. L. & Urbini, S. Snow dunes and glazed surfaces in Antarctica: new field and remote-sensing data. Annals of Glaciology 34, 81-88 (2002).  
+
+48. Libois, Q., Picard, G., Arnaud, L., Morin, S. & Brun, E. Modeling the impact of snow drift on the decameter-scale variability of snow properties on the Antarctic Plateau. Journal of Geophysical Research: Atmospheres 119, 11662-11681 (2014).  
+
+49. Picard, G., Arnaud, L., Caneill, R., Lefebvre, E. & Lamare, M. Observation of the process of snow accumulation on the Antarctic Plateau by time lapse laser scanning. The Cryosphere 13, 1983-1999 (2019).  
+
+50. Neubauer, C. et al. Stable isotope analysis of intact oxyanions using electrospray quadrupole-orbitrap mass spectrometry. Anal. Chem. 92, 3077-3085 (2020).  
+
+51. Meusinger, C., Berhanu, T. A., Erbland, J., Savarino, J. & Johnson, M. S. Laboratory study of nitrate photolysis in Antarctic snow. I. Observed quantum yield, domain of photolysis, and secondary chemistry. The Journal of Chemical Physics 140, 244305 (2014).  
+
+52. Chu, L. & Anastasio, C. Quantum yields of hydroxyl radical and nitrogen dioxide from the photolysis of nitrate on ice. J. Phys. Chem. A 107, 9594-9602 (2003).
+
+<--- Page Split --->
+
+53. Benedict, K. B., McFall, A. S. & Anastasio, C. Quantum yield of nitrite from the photolysis of aqueous nitrate above 300 nm. Environ. Sci. Technol. 51, 4387-4395 (2017).  
+
+54. Kaiser, J., Hastings, M. G., Houlton, B. Z., Röckmann, T. & Sigman, D. M. Triple oxygen isotope analysis of nitrate Using the denitrifier method and thermal decomposition of N2O. Analytical Chemistry 79, 599-607 (2007).  
+
+55. Morin, S. et al. Comprehensive isotopic composition of atmospheric nitrate in the Atlantic Ocean boundary layer from 65 degrees S to 79 degrees N. Journal of Geophysical Research-Atmospheres 114, (2009).  
+
+56. Sigman, D. M. et al. A bacterial method for the nitrogen isotopic analysis of nitrate in seawater and freshwater. Anal. Chem. 73, 4145-4153 (2001).  
+
+57. Casciotti, K. L., Sigman, D. M., Hastings, M. G., Bohlke, J. K. & Hilkert, A. Measurement of the oxygen isotopic composition of nitrate in seawater and freshwater using the denitrifier method. Anal. Chem. 74, 4905-4912 (2002).  
+
+58. Baertschi, P. Absolute 180 content of standard mean ocean water. Earth and Planetary Science Letters 31, 341-344 (1976).  
+
+59. Mariotti, A. Atmospheric nitrogen is a reliable standard for natural 15N abundance measurements. Nature 303, 685-687 (1983).  
+
+60. Pourchet, M. et al. Distribution and fall-out of 137Cs and other radionuclides over Antarctica. Journal of Glaciology 43, 435-445 (1997).  
+
+61. Ding, M. et al. Spatial variability of surface mass balance along a traverse route from Zhongshan station to Dome A, Antarctica. Journal of Glaciology 57, 658-666 (2011).  
+
+62. Verfaillie, D. et al. Snow accumulation variability derived from radar and firn core data along a 600 km transect in Adelie Land, East Antarctic plateau. The Cryosphere 6, 1345-1358 (2012).  
+
+63. Favier, V. et al. An updated and quality controlled surface mass balance dataset for Antarctica. The Cryosphere 7, 583-597 (2013).  
+
+64. Hoshina, Y., Fujita, K., Iizuka, Y. & Motoyama, H. Inconsistent relationships between major ions and water stable isotopes in Antarctic snow under different accumulation environments. Polar Science 10, 1-10 (2016).  
+
+65. Ding, M. et al. Re-assessment of recent (2008 - 2013) surface mass balance over Dome Argus, Antarctica. POLAR 35, (2016).
+
+<--- Page Split --->
+
+66. Ekaykin, A. A. et al. Underestimation of Snow Accumulation Rate in Central Antarctica (Vostok Station) Derived from Stake Measurements. Russian Meteorology and Hydrology 45, 132–140 (2020).  
+
+67. Sommer, S., Wagenbach, D., Mulvaney, R. & Fischer, H. Glacio-chemical study spanning the past 2 kyr on three ice cores from Dronning Maud Land, Antarctica: 2. Seasonally resolved chemical records. Journal of Geophysical Research: Atmospheres 105, 29423–29433 (2000).  
+
+68. Richter, A. et al. Surface mass balance models vs. stake observations: a comparison in the Lake Vostok region, central East Antarctica. Frontiers in Earth Science 9, 388 (2021).
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- Akersd15NManuscript20220128NComSupplement.docx
+
+<--- Page Split --->

preprint/preprint__4917d2ec82456765310e4fd0e81bd9c6564f10de32d29557ebc886beb9b7789e/preprint__4917d2ec82456765310e4fd0e81bd9c6564f10de32d29557ebc886beb9b7789e_det.mmd

@@ -0,0 +1,738 @@
+<|ref|>title<|/ref|><|det|>[[44, 107, 825, 174]]<|/det|>
+# Sunlight-driven nitrate loss records Antarctic surface mass balance  
+
+<|ref|>text<|/ref|><|det|>[[44, 195, 432, 215]]<|/det|>
+Pete D. Akers ( pete.d.akers@gmail.com )  
+
+<|ref|>text<|/ref|><|det|>[[44, 217, 949, 260]]<|/det|>
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France https://orcid.org/0000- 0002- 2266- 5551  
+
+<|ref|>text<|/ref|><|det|>[[44, 265, 168, 283]]<|/det|>
+Joël Savarino  
+
+<|ref|>text<|/ref|><|det|>[[50, 287, 686, 306]]<|/det|>
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+<|ref|>text<|/ref|><|det|>[[44, 311, 178, 329]]<|/det|>
+Nicolas Caillon  
+
+<|ref|>text<|/ref|><|det|>[[50, 333, 686, 352]]<|/det|>
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+<|ref|>text<|/ref|><|det|>[[44, 357, 249, 375]]<|/det|>
+Aymeric P. M. Servettaz  
+
+<|ref|>text<|/ref|><|det|>[[50, 379, 697, 399]]<|/det|>
+Japan Agency for Marine- Earth Science and Technology, Yokosuka, Japan  
+
+<|ref|>text<|/ref|><|det|>[[44, 404, 216, 422]]<|/det|>
+Emmanuel Le Meur  
+
+<|ref|>text<|/ref|><|det|>[[50, 426, 686, 445]]<|/det|>
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+<|ref|>text<|/ref|><|det|>[[44, 450, 178, 468]]<|/det|>
+Olivier Magand  
+
+<|ref|>text<|/ref|><|det|>[[50, 472, 686, 491]]<|/det|>
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+<|ref|>text<|/ref|><|det|>[[44, 496, 160, 514]]<|/det|>
+Jean Martins  
+
+<|ref|>text<|/ref|><|det|>[[50, 518, 686, 537]]<|/det|>
+Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France  
+
+<|ref|>text<|/ref|><|det|>[[44, 543, 165, 561]]<|/det|>
+Cécile Agosta  
+
+<|ref|>text<|/ref|><|det|>[[44, 565, 955, 607]]<|/det|>
+Laboratoire des Sciences du Climat et de l'Environnement, LSCE- IPSL, CEA- CNRS- UVSQ, Université Paris- Saclay, Gif- sur- Yvette, France https://orcid.org/0000- 0003- 4091- 1653  
+
+<|ref|>text<|/ref|><|det|>[[44, 612, 180, 630]]<|/det|>
+Peter Crockford  
+
+<|ref|>text<|/ref|><|det|>[[44, 633, 855, 676]]<|/det|>
+Department of Earth and Planetary Sciences, Weizmann Institute of Science, Rehovot, Israel; Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA, USA  
+
+<|ref|>text<|/ref|><|det|>[[44, 681, 200, 700]]<|/det|>
+Kanon Kobayashi  
+
+<|ref|>text<|/ref|><|det|>[[44, 703, 919, 723]]<|/det|>
+Department of Chemical Science and Engineering, Tokyo Institute of Technology, Yokohama, Japan  
+
+<|ref|>text<|/ref|><|det|>[[44, 728, 170, 746]]<|/det|>
+Shohei Hattori  
+
+<|ref|>text<|/ref|><|det|>[[44, 749, 923, 792]]<|/det|>
+Department of Chemical Science and Engineering, Tokyo Institute of Technology, Yokohama, Japan; International Center for Isotope Effects Research, Nanjing University, Nanjing, China  
+
+<|ref|>text<|/ref|><|det|>[[44, 797, 152, 815]]<|/det|>
+Mark Curran  
+
+<|ref|>text<|/ref|><|det|>[[44, 818, 936, 882]]<|/det|>
+Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia; Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia  
+
+<|ref|>text<|/ref|><|det|>[[44, 888, 190, 906]]<|/det|>
+Tas van Ommen  
+
+<|ref|>text<|/ref|><|det|>[[44, 910, 936, 953]]<|/det|>
+Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia; Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[42, 45, 763, 65]]<|/det|>
+of Tasmania, Hobart, Tasmania, Australia https://orcid.org/0000- 0002- 2463- 1718  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 70, 169, 88]]<|/det|>
+## Lenneke Jong  
+
+<|ref|>text<|/ref|><|det|>[[42, 91, 937, 157]]<|/det|>
+Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia; Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 163, 193, 181]]<|/det|>
+## Jason L. Roberts  
+
+<|ref|>text<|/ref|><|det|>[[42, 185, 937, 250]]<|/det|>
+Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia; Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 290, 101, 308]]<|/det|>
+## Article  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 327, 135, 346]]<|/det|>
+## Keywords:  
+
+<|ref|>text<|/ref|><|det|>[[44, 365, 325, 384]]<|/det|>
+Posted Date: February 8th, 2022  
+
+<|ref|>text<|/ref|><|det|>[[44, 403, 474, 422]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 1307003/v1  
+
+<|ref|>text<|/ref|><|det|>[[42, 440, 910, 484]]<|/det|>
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 519, 908, 562]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Communications on July 25th, 2022. See the published version at https://doi.org/10.1038/s41467- 022- 31855- 7.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[157, 91, 836, 111]]<|/det|>
+# Sunlight-driven nitrate loss records Antarctic surface mass balance  
+
+<|ref|>text<|/ref|><|det|>[[133, 115, 867, 184]]<|/det|>
+Authors: Pete D. Akers \(^{1*}\) , Joël Savarino \(^{1*}\) , Nicolas Caillon \(^{1}\) , Aymeric P. M. Servettaz \(^{2}\) , Emmanuel Le Meur \(^{1}\) , Olivier Magand \(^{1}\) , Jean Martins \(^{1}\) , Cécile Agosta \(^{3}\) , Peter Crockford \(^{4,5}\) , Kanon Kobayashi \(^{6}\) , Shohei Hattori \(^{6,7}\) , Mark Curran \(^{8,9}\) , Tas van Ommen \(^{8,9}\) , Lenneke Jong \(^{8,9}\) , Jason L. Roberts \(^{8,9}\)  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 205, 222, 221]]<|/det|>
+## Affiliations:  
+
+<|ref|>text<|/ref|><|det|>[[115, 227, 850, 520]]<|/det|>
+\(^{1}\) Université Grenoble Alpes, CNRS, IRD, Grenoble INP, IGE, Grenoble, France. \(^{2}\) Japan Agency for Marine- Earth Science and Technology, Yokosuka, Japan. \(^{3}\) Laboratoire des Sciences du Climat et de l'Environnement, LSCE- IPSL, CEA- CNRS- UVSQ, Université Paris- Saclay, Gif- sur- Yvette, France. \(^{4}\) Department of Earth and Planetary Sciences, Weizmann Institute of Science, Rehovot, Israel. \(^{5}\) Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA, USA. \(^{6}\) Department of Chemical Science and Engineering, Tokyo Institute of Technology, Yokohama, Japan. \(^{7}\) International Center for Isotope Effects Research, Nanjing University, Nanjing, China. \(^{8}\) Australian Antarctic Division, Department of Agriculture, Water and Environment, Kingston, Tasmania, Australia. \(^{9}\) Australian Antarctic Program Partnership, Institute of Marine and Antarctic Studies, University of Tasmania, Hobart, Tasmania, Australia.  
+
+<|ref|>text<|/ref|><|det|>[[118, 545, 760, 563]]<|/det|>
+\* Correspondence to: pete- akers@univ- grenoble- alpes.fr; joel.savarino@cnrs.fr  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 593, 201, 609]]<|/det|>
+## Abstract:  
+
+<|ref|>text<|/ref|><|det|>[[115, 631, 867, 881]]<|/det|>
+Standard proxies for reconstructing surface mass balance (SMB) in Antarctic ice cores are often inaccurate or coarsely resolved when applied to more complicated environments away from dome summits. Here, we propose an alternative SMB proxy based on photolytic fractionation of nitrogen isotopes in nitrate observed at 114 sites throughout East Antarctica. Applying this proxy approach to nitrate in a shallow core drilled at a moderate SMB site (Aurora Basin North), we reconstruct 700 years of SMB changes that agree well with changes estimated from ice core density and upstream surface topography. For the under- sampled transition zones between dome summits and the coast, this proxy can considerably
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 835, 135]]<|/det|>
+expand our SMB records by providing high- resolution SMBs that better reflect the local environment and are easier to sample than existing techniques.  
+
+<|ref|>text<|/ref|><|det|>[[117, 156, 875, 208]]<|/det|>
+One Sentence Summary: Nitrate isotopes offer a new way to track past and present changes in Antarctic snowfall and ice sheet mass balance.  
+
+<|ref|>text<|/ref|><|det|>[[115, 228, 875, 544]]<|/det|>
+Main Text: Antarctica holds a critical role in the Earth's hydrosphere, providing long- term storage of 27 million \(\mathrm{km}^3\) of ice<sup>1</sup> and impacting global ocean and atmosphere circulation through its albedo, topography, export of calved glacial ice, and function as an atmospheric heat sink<sup>2- 5</sup>. Since even small shifts in the surface mass balance (SMB) across Antarctic ice sheets can redistribute huge masses of water between the cryosphere, ocean, and atmosphere, a clear understanding of how its SMB has responded to past climate change is crucial for calibrating forecast models of the global environment and properly interpreting ice cores<sup>6- 10</sup>. Despite this pressing importance, a comprehensive understanding of past SMB changes in Antarctica is limited by insufficient long- term records for sites between of the wet coastal periphery and dry dome summits.  
+
+<|ref|>text<|/ref|><|det|>[[115, 563, 875, 882]]<|/det|>
+Reconstructing SMBs in this moderate SMB transition zone can be challenging with existing SMB proxies. Ice density- based reconstructions become less effective and more uncertain with depth due to thinning and deformation of ice layers<sup>11</sup>, while the frequent minor damage and breakage of cores during the drilling process can make accurate physical measurements of mass and volume challenging. Water isotopes ( \(\delta^2\mathrm{H}\) or \(\delta^{18}\mathrm{O}\) ) can be used as a proxy temperature to derive snow accumulation through water vapor saturation<sup>10</sup>, but this approach does not account for wind- driven transport and sublimation of surface snow at warmer and lower elevation sites<sup>12- 14</sup>. Additionally, water isotopes reflect many environmental factors other than temperature, such as atmospheric circulation changes, which can lead to large uncertainty and/or bias in reconstructed SMBs<sup>15,16</sup>. There is thus a strong need for alternative
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 810, 135]]<|/det|>
+proxies that better record local conditions to provide SMB estimates for paleoclimate reconstructions and ice sheet models.  
+
+<|ref|>text<|/ref|><|det|>[[115, 155, 875, 655]]<|/det|>
+Here, we present one such independent SMB proxy based on photolysis- induced changes in the \(^{15}\mathrm{N / ^{14}N}\) ratio ( \(\delta^{15}\mathrm{N}\) , defined as \(\delta = \frac{^{15}\mathrm{N / ^{14}N}_{\mathrm{sample}}}{^{15}\mathrm{N / ^{14}N}_{\mathrm{standard}}} - 1\) , relative to the \(\mathrm{N}_2\) - air standard) of nitrate \(\mathrm{(NO_3^- )}\) (Figure 1). Naturally deposited on the Antarctic ice sheet surface as the end product of the atmospheric oxidation of reactive nitrogen \(^{17 - 20}\) , \(\mathrm{NO_3^- }\) within the Antarctic snowpack can be photolytically converted to gaseous nitrogen oxides \(\mathrm{(NO_x = NO + NO_2)}\) when exposed to ultraviolet light ( \(\lambda = 290 - 350 \mathrm{nm}\) ). Because \(^{14}\mathrm{NO_3^- }\) is more readily photolyzed than \(^{15}\mathrm{NO_3}\) , the \(\delta^{15}\mathrm{NNO_3}\) of \(\mathrm{NO_3^- }\) remaining in the snow will increase from its initial depositional value of \(\approx - 20\) to \(+20\%\) to values as high as \(+400\%\) \(^{19 - 26}\) as the isotopically lighter photolytic \(\mathrm{NO_x}\) is ventilated and lost to the atmosphere. Although \(\mathrm{NO_3^- }\) can also be lost through \(\mathrm{HNO_3}\) volatilization, we interpret \(\delta^{15}\mathrm{NNO_3}\) are solely through photolysis as volatilization does not strongly fractionate \(\mathrm{NO_3^- }\) and is a very minor component of \(\mathrm{NO_3^- }\) loss outside of the warmest coastal zones \(^{22,27,28}\) . Additionally, while the oxygen in \(\mathrm{NO_3^- }\) also undergoes isotopic fractionation through photolysis, its interpretation is complicated by isotopic interactions with snow and water vapor \(^{22,23,29}\) and is not further discussed here.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[117, 81, 702, 368]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 380, 880, 538]]<|/det|>
+<center>Figure 1. Schematic diagram illustrating that SMB and \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values covary due to photolytic \(\mathrm{NO}_3^-\) mass loss. After \(\mathrm{NO}_3^-\) containing either \(^{14}\mathrm{N}\) (blue) and \(^{15}\mathrm{N}\) (red) is deposited on the Antarctic snowpack surface (1), sunlight in the photic zone can trigger photolysis of \(\mathrm{NO}_3^-\) that favors \(\mathrm{NO}_3^-\) with a \(^{14}\mathrm{N}\) atom, which leaves the residual \(\mathrm{NO}_3^-\) enriched in \(^{15}\mathrm{N}\) (2). Because sites with lower SMBs will accumulate less snow over a given period of time than high SMB sites (3), the \(\mathrm{NO}_3^-\) at lower SMB sites will remain in the photic zone longer, experience more photolytic mass loss before burial in the archived zone, and have higher \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values (4). </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 559, 880, 881]]<|/det|>
+Photolysis is limited to the depth where light penetrates and initiates photochemical reactions, and so the snowpack can be divided into an uppermost photic zone (generally 10- 100 cm in East Antarctica) and a deeper archived zone \(^{29 - 33}\) . Photolysis and the resulting isotopic fractionation of \(\mathrm{NO}_3^-\) cease once snowfall buries \(\mathrm{NO}_3^-\) beneath the photic zone, and the \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) value of the buried \(\mathrm{NO}_3^-\) ( \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) ) is assumed to be preserved indefinitely in glacial ice \(^{22,23,29,30}\) . The final \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value reflects the total sum of photolysis inducing radiation experienced by \(\mathrm{NO}_3^-\) during the burial process, which, assuming stable insolation and photic zone depth, is itself determined by the rate at which the \(\mathrm{NO}_3^-\) is buried and thus inversely related to SMB \(^{17,23,26,34}\) . Modeling (Supplementary Text 1) and field observations support SMB as the primary driver of spatial variability in \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values. Based on a new
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 864, 135]]<|/det|>
+simplified theoretical framework (Methods, Supplementary Text 1), this relationship can be expressed as:  
+
+<|ref|>equation<|/ref|><|det|>[[117, 155, 812, 185]]<|/det|>
+\[\ln (\delta^{15}N_{NO3arc} + 1) = \frac{A}{SMB} +B \quad (1)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 204, 872, 325]]<|/det|>
+where the regression coefficients \(A\) and \(B\) are parameters that subsume constants and linearly co- varying variables associated with photolytic and fractionation processes. The inverse function of Eq. (1) can then be used as a transfer function to reconstruct SMBs from \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values (SMB \(\delta_{15}\mathrm{N}\) ).  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 386, 590, 406]]<|/det|>
+## Results: SMB \(\delta_{15}\mathrm{N}\) relationship and spatial applicability  
+
+<|ref|>text<|/ref|><|det|>[[115, 426, 875, 715]]<|/det|>
+To obtain parameter estimates for Eq. (1), we sampled \(\mathrm{NO}_3^-\) in snow and firn from 92 East Antarctic shallow pits and cores that are newly reported here. Combined with 43 previously published \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) samples \(^{21 - 23,26,29,35}\) , this constitutes a database of 135 total \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values representing 114 distinct sites across East Antarctica (Figure 2a). These \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) data were spatially paired with local SMBs either observed directly onsite (SMB \(\mathrm{ground}\) ) or as an output from the Modèle Atmosphérique Régional (MAR) using ERA- interim reanalysis data \(^{12}\) and adjusted for a dry- site bias (SMB \(\mathrm{adjMAR}\) ) (Methods, Supplementary Text 2). The sites in our database cover a comprehensive range of East Antarctic SMBs, from 20–30 kg m \(^{- 2}\) a \(^{- 1}\) at dome summits on the high plateau to \(>300 \mathrm{kg m}^{- 2} \mathrm{a}^{- 1}\) for sites on the coastal periphery.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[124, 84, 873, 277]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 291, 872, 533]]<|/det|>
+<center>Figure 2. A) Map of East Antarctic sites sampled for \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) along different scientific and logistic transect routes<sup>36</sup>. The SMBs were modeled by \(\mathrm{MAR}^{12}\) and adjusted for dry site bias (see Methods). Regions with SMBs less than or greater than \(40–200\mathrm{kg}\mathrm{m}^{-2}\mathrm{a}^{-1}\) (i.e., the SMB range targeted by the proxy described here) are illustrated with hatching and crosshatching, respectively. Preservation of \(\mathrm{NO}_3^-\) is not expected in blue ice zones (gray solid) due to very low or negative SMBs and wind scouring<sup>37</sup>. B) Scatter plot and linear regression of Eq. (1) using all sites in the field dataset. Note that axis labels have been converted to show simpler SMB and \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values. The linear regression (gray solid line) is shown with shaded \(95\%\) confidence intervals, and regression parameters are displayed at upper right. The point colors correspond to the transect of origin shown in (a), and the point shapes correspond to the sampling method (i.e., snow core, snow pit, or 1-m depth layer). </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 590, 872, 885]]<|/det|>
+The SMB and \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) in our field dataset are correlated with a high degree of confidence, producing a linear regression where \(\ln (\delta^{15}\mathrm{N}_{\mathrm{NO3arc}} + 1) = 6.98\pm 0.19\mathrm{SMB}^{- 1} - 0.02\pm 0.01\) (Figure 2b, \(r^2 = 0.91\) , \(p\ll 0.001\) , \(n = 135\) ). This relationship is within modeled expectations (Figure S3, Supplementary Text 1) and reproduces the spatial variability of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) observed throughout East Antarctica (Table S6). Although millennial- scale changes in global nitrogen dynamics and atmospheric oxidative capacity are not currently well known, the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy should be broadly applicable to Holocene- age ice as the factors parameterized in Eq. (1) have likely been relatively stable during this time (Supplementary Text 1). For pre- Holocene ice, \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) can still offer important insight into relative
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 830, 135]]<|/det|>
+changes in SMB and into how nitrate dynamics varied during the dramatically different Antarctic and global environments of the Pleistocene.  
+
+<|ref|>text<|/ref|><|det|>[[115, 155, 870, 746]]<|/det|>
+While our field dataset covers sites with SMBs from 22 to \(548\mathrm{kgm^{- 2}a^{- 1}}\) , the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy appears best suited for sites with SMBs between 40 and \(200\mathrm{kgm^{- 2}a^{- 1}}\) . Shallow cores from very dry Dome A and Dome C have lower \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values at 2- 6 m below the surface than at the \(\sim 1\mathrm{m}\) base of the photic zone, possibly because photolytic \(\mathrm{NO_x}\) can be transported downward through firn air convection and re- oxidized into \(\mathrm{NO_3}\) with low \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) values (Supplementary Text 3, Figure S5). Although this phenomenon violates the foundational assumption of “locked- in” \(\mathrm{NO_3}\) - beneath the photic zone, we observe only it at the ultra- dry interior sites where \(\mathrm{SMB} > 40\mathrm{kgm^{- 2}a^{- 1}}\) . For sites with \(\mathrm{SMB} > 200\mathrm{kgm^{- 2}a^{- 1}}\) , the expected \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value falls within the general range of atmospheric \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) ( \(< +20\%\) ) because \(\mathrm{NO_3}\) - is buried below the photic zone in less than a year. Since more than \(80\%\) of \(\mathrm{NO_3}\) - is deposited during months with sunshine outside of winter polar night \(^{22,38}\) , samples that integrate multiple years of accumulation at high SMB sites might still resolve differences in SMB, but the greater \(\mathrm{HNO_3}\) volatilization at these warmer and wetter sites also warrant caution due to possible interference. Additionally, the asymptotic nature of \(\mathrm{SMB^{- 1}}\) means that \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values are increasingly less sensitive to SMB changes with higher SMB values. Despite these restrictions, over \(59\%\) of Antarctica has a SMB between 40 and \(200\mathrm{kgm^{- 2}a^{- 1}}\) \(^{12}\) (Figure 2), and additional study of \(\mathrm{NO_3}\) - dynamics in wet and dry extremes may reveal regional adjustments that allow our proxy’s SMB range to be expanded.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 805, 549, 824]]<|/det|>
+## Results: Aurora Basin North SMB reconstruction  
+
+<|ref|>text<|/ref|><|det|>[[118, 844, 861, 898]]<|/det|>
+As a proof of concept, we applied the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) transfer function to \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) data from the \(103\mathrm{m}\) deep ABN1314- 103 ice core. This core was one of three drilled in the Australian
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 876, 442]]<|/det|>
+Antarctic Program's 2013–2014 summer campaign at Aurora Basin North (ABN; 71.17 °S 111.37 °E, 2679 m above sea level), a site with moderate modern SMBs (≈120 kg m⁻² a⁻¹) located midway between coastal Casey Station and the Dome C summit (Figure 2a). The SMBδ15N history reconstructed from ABN1314- 103 covers the period from –47 to 649 years before present (BP, where present = 1950 CE) and has values ranging from 49 to 208 kg m⁻² a⁻¹ (Figure 3a). Each SMBδ15N value integrates an average of 2.4 years of accumulation (total range: 0.7–4.5 years), and thus any impacts from individual precipitation events or seasonal extremes are largely moderated. Overall, the SMBs have fairly high variability (coefficient of variation = 0.21). The mean SMBδ15N in the 20th century (126±26.5 kg m⁻² a⁻¹) is 34% greater than the mean SMBδ15N before 1900 CE (94±18 kg m⁻² a⁻¹) and nearly 52% greater than the driest century that spans the 1600s CE (83±20 kg m⁻² a⁻¹) (Figure 3a).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[122, 78, 825, 645]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 660, 876, 879]]<|/det|>
+<center>Figure 3. (a) Surface mass balance reconstruction for Aurora Basin North based on \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) arc data from the ABN1314-103 ice core. Reconstructed \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) values are shown by the red stepped lines with the 50-yr running mean \(\pm 1\sigma\) overlaid as a darker thick line and shaded zone. (b) Comparison of SMBs reconstructed from \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) (red) with those from ice density (gray) and upstream GPR isochron depth \(^{39}\) . The \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) and \(\mathrm{SMB}_{\mathrm{GPR}}\) values were aggregated to match the 1-m resolution of the \(\mathrm{SMB}_{\mathrm{density}}\) data. For \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) and \(\mathrm{SMB}_{\mathrm{density}}\) , smoothed LOESS curves are overlaid to more clearly show long-term patterns. (c) \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) values after the upstream topographic impact on SMBs has been removed, with 50-yr running mean \(\pm 1\sigma\) values overlaid. The resulting residuals may better illustrate SMB variability due to climate change. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[118, 125, 610, 144]]<|/det|>
+## Discussion: Validating the SMB \(\delta 15\mathrm{N}\) proxy reconstruction  
+
+<|ref|>text<|/ref|><|det|>[[115, 165, 880, 520]]<|/det|>
+We verified our new proxy's accuracy by comparing the SMB \(\delta 15\mathrm{N}\) values with SMBs derived from the physical ice density (SMBdensity) measurements of the same core. For each 1- m core segment of ABN1314- 103, we calculated a SMBdensity value by dividing the segment's mass (kg) by both its volume \((\mathrm{m}^3)\) and the age difference between the top and bottom of the segment (a \(\mathrm{m}^{- 1}\) ). The SMB \(\delta 15\mathrm{N}\) (aggregated to match the 1- m resolution) and SMB density share very similar mean values (100.8 vs. 98.0 kg \(\mathrm{m}^{- 2}\mathrm{a}^{- 1}\) , respectively) and total SMB ranges (62.0–157.3 vs. 61.7–153.4 kg \(\mathrm{m}^{- 2}\mathrm{a}^{- 1}\) , respectively), and the two SMB reconstructions have a similar pattern of variation with a moderate linear correlation \((\mathrm{r} = +0.46\) , \(\mathrm{p}< 0.001\) , \(\mathrm{n} = 90\) ) (Figure 3b). This agreement in mean value, range, and variability strongly validates our SMB \(\delta 15\mathrm{N}\) approach and the potential of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) as an accurate proxy for paleoenvironmental change.  
+
+<|ref|>text<|/ref|><|det|>[[115, 537, 880, 891]]<|/det|>
+Interpreting the ABN1314- 103 SMB profile is more complicated than for ice cores drilled at dome summits because the ice sheet at the ABN drilling site is flowing at a rate of \(16.2\mathrm{m}\mathrm{a}^{- 1}\) 40. This means that the ice in ABN1314- 103 actually fell as snow along a continuous \(11.5\mathrm{km}\) transect upstream of the current ABN drilling site, with the oldest and deepest ice originating from the most distant upstream position. As a result, the \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) and core density have recorded any spatial SMB variability that existed along the upstream transect in addition to any SMB changes due to wetting or drying of the regional climate. Although overall elevation gain is small along the transect ( \(< 15\mathrm{m}\) ), the region has abundant \(0.5–1\mathrm{m}\) undulations in surface topography extending over horizontal extents of \(3–10\mathrm{km}^{36}\) . While the MAR's horizontal grid size (35 km) cannot resolve the SMB impact from these features, ground penetrating radar (GPR) performed along the upstream transect revealed that these
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 867, 234]]<|/det|>
+surface slope changes correlate with SMB variations of up to \(40\mathrm{kgm^{- 2}a^{- 1}}\) as determined by isochronic internal reflection horizons \(^{39,41}\) (Figure 5). Although the long- term stability of such features is not well understood, the current surface features are still largely identifiable as buried horizons to depths below the deepest segment of ABN1314- 103 with only steady horizontal offset due to ice flow.  
+
+<|ref|>text<|/ref|><|det|>[[115, 254, 878, 506]]<|/det|>
+By relating the upstream topographic- driven changes in SMB revealed by GPR to core depth through the horizontal ice flow rate and the core age- depth model \(^{39}\) , we can determine the expected SMB signal due only to upstream surface topography (SMB \(_{\mathrm{GPR}}\) ). We find that the general pattern of variability in SMB \(_{\mathrm{GPR}}\) correlates very well with the patterns recorded in the SMB \(_{\delta 15\mathrm{N}}\) ( \(r = +0.74\) ) and SMB \(_{\mathrm{density}}\) ( \(r = +0.63\) ) records (Figure 3b). Thus, it appears that the primary SMB pattern preserved in ABN1314- 103 is driven by upstream changes in surface slope, which is important for properly interpreting other environmental proxies contained in the ice and for understanding the local ice flow history.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 567, 572, 586]]<|/det|>
+## Discussion: Extracting a climate-driven SMB record  
+
+<|ref|>text<|/ref|><|det|>[[115, 606, 875, 892]]<|/det|>
+To examine whether a secondary signal related to climate change was also preserved, we removed the spatial impact of upstream topography by subtracting the SMB \(_{\mathrm{GPR}}\) data from the SMB \(_{\delta 15\mathrm{N}}\) record. After this “upstream effect detrending” and accounting for the small offset in mean SMB values ( \(3.7\mathrm{kgm^{- 2}a^{- 1}}\) ), we find that the multi- decadal SMB values have been generally stable over the past 700 years (Figure 3c), with 50- yr running averages of the SMB never greater or less than \(15\mathrm{kgm^{- 2}a^{- 1}}\) from the detrended mean. These running averages suggest that drier conditions existed at ABN between 1600 and 1890 CE (partially corresponding to the Little Ice Age) and that precipitation has increased in the most recent 100–150 years. This is generally consistent with what has been observed at other East
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 875, 202]]<|/det|>
+Antarctic sites \(^{42 - 44}\) and for Antarctica as a whole \(^{11}\) , but we recognize that this pattern is similar to the upstream topographic effect and that it might also arise if the SMB \(_{\text{GRR}}\) record is excessively smoothed relative to true topographic- driven SMB variability (perhaps by the GPR data processing).  
+
+<|ref|>text<|/ref|><|det|>[[115, 220, 875, 638]]<|/det|>
+On shorter timescales, SMBs frequently change by \(\approx 50 \mathrm{kg m^{- 2} a^{- 1}}\) around a common mean within 10- 20 year periods. This pattern likely reflects the high interannual snowfall variability expected at sites like ABN \(^{13}\) . Located at the transition between the coast and the interior East Antarctic Plateau, annual snow accumulation at ABN is sensitive to chance intrusions of extreme precipitation events and atmospheric rivers \(^{45,46}\) , and the observed sub- decadal SMB \(_{\beta 15N}\) variability may represent the frequency of their stochastic occurrence at the site. Additionally, small scale surface roughness features like sastrugi may affect hyperlocal SMB (i.e., the SMB at scales of \(< 1 \mathrm{m}\) ) through periods of enhanced accumulation and erosion as they migrate and evolve on the snow surface \(^{47 - 49}\) . While the temporal evolution and possible life cycle cyclicity of surface roughness features are as yet poorly known, hyperlocal changes in SMB could also explain some of the short- term SMB variability observed in the ABN record if the sampling interval is shorter than the average duration of a surface feature at a given location.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 699, 631, 718]]<|/det|>
+## Discussion: Applied use and potential of the SMB \(\alpha \beta \mathrm{SN}\) proxy  
+
+<|ref|>text<|/ref|><|det|>[[115, 738, 860, 890]]<|/det|>
+With over 8 million \(\mathrm{km^2}\) of Antarctica having a SMB between 40 and \(200 \mathrm{kg m^{- 2} a^{- 1}}\) \(^{12}\) and over \(70\%\) of the ice sheet area modeled to have \(\delta^{15}\mathrm{NNO_3}\) values markedly elevated by photolysis (Figure S6, Supplemental Text 4), the SMB \(\alpha \beta \mathrm{SN}\) proxy holds great potential for vastly expanding our knowledge of Antarctic SMB variability over time and space. Currently, regions with moderate SMBs have only a handful of sites with SMB records older
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 81, 880, 405]]<|/det|>
+than 200 years, with the East Antarctic Plateau particularly poorly represented<sup>11</sup>. For ice coring projects in these regions, the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy will excel at capturing the local effects of strong winds, irregular surface topography, and high interannual snowfall variability better than water isotopic techniques while avoiding problems with layer thinning and density modeling that affect \(\mathrm{SMB}_{\mathrm{density}}\) methods. As regional climate models still struggle to accurately simulate drifting snow and sublimation fluxes in the coast- to- plateau transition<sup>12</sup>, \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) can provide critical ground- based data for models predicting future contributions to sea level rise. The \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy also holds particular value for helping constrain and validate models of upstream flow effects in research targeting ice streams and broad- scale glacial flow patterns.  
+
+<|ref|>text<|/ref|><|det|>[[115, 425, 872, 811]]<|/det|>
+Additionally, sampling for the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy can save valuable time and cost compared to existing alternatives in order to expand current records of modern SMBs. Obtaining new ground- based SMBs for sites without annually resolved layers requires either coring several meters to the increasingly buried Pinatubo volcanic horizon or repeated visits to newly installed stake transects. However, limited time and resources for research expeditions to remote areas precludes intensive SMB surveys with these methods. With the \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy, a mean site SMB could be determined with only a series of shallow snow or firm samples extending deep enough into the archived zone to cover only a few seasonal cycles (much shallower than the Pinatubo horizon). After proper mixing, only \(\sim 0.3 - 1.0 \mathrm{kg}\) would need to be kept, transported, and analyzed for each sample, which logistically allows for the rapid collection of robust SMB site means in many locations. On- site melting and \(\mathrm{NO}_3^-\) concentration could further reduce logistical requirements.  
+
+<|ref|>text<|/ref|><|det|>[[118, 831, 830, 886]]<|/det|>
+The \(\mathrm{SMB}_{\delta 15\mathrm{N}}\) proxy promises to grow and adapt as studies on Antarctic \(\mathrm{NO}_3^-\) dynamics continue. Because the resolution of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) sampling is limited only by the minimum
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 82, 877, 506]]<|/det|>
+amount of \(\mathrm{NO_3^- }\) needed for analysis, very finely- resolved \(\delta^{15}\mathrm{N_{NO3arc}}\) records can be obtained by increasing the mass of ice collected per depth unit (e.g., by specifically drilling whole cores or replicate cores for \(\mathrm{NO_3^- }\) isotopes) and with advances in \(\mathrm{NO_3^- }\) isotopic analysis expected in the near future<sup>50</sup>. This may allow for more precise multi- annual aggregations for \(\mathrm{SMB_{\delta 15N}}\) reconstructions and permit a deeper examination of subannual \(\mathrm{NO_3^- }\) dynamics. Finally, SMBs from parts of the West Antarctic ice sheet and the highest elevations of the northern Greenland ice sheet fall within the appropriate range for the \(\mathrm{SMB_{\delta 15N}}\) proxy, and additional field sampling at those locations may allow us to reconstruct SMBs by verifying or adapting the relationship defined here for regional use outside of East Antarctica. Given the great potential of the \(\mathrm{SMB_{\delta 15N}}\) proxy to advance our understanding of the Antarctic environment and its sensitivity to climate change, we strongly recommend that potential ice coring projects incorporate \(\mathrm{NO_3^- }\) analyses into their planning and urge continued studies on Antarctic \(\mathrm{NO_3^- }\) dynamics.  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 567, 293, 584]]<|/det|>
+## Acknowledgements:  
+
+<|ref|>text<|/ref|><|det|>[[115, 604, 879, 888]]<|/det|>
+We express thanks to the following individuals for project assistance and data support: Sarah Albertin, Selin Bagci, Albane Barbero, Mathieu Casado, Armelle Crouzet, Vincent Favier, Elsa Gautier, Gaspard Jannot, Alexis Lamothe, Anaïs Orsi, Fred Parrenin, Holly Winton, and the overwintering crews at Concordia Station. We acknowledge the logistical support of IPEV for the French missions in Antarctica, the IPEV and PNRA colleagues and overwintering crews at Concordia Station, and the JARE54 traverse team for fieldwork assistance and access to the S80 site data. We thank the co- investigators for the ABN drilling project (Jérome Chappellaz, Dorthe Dahl- Jensen, David Etheridge, Joe McConnell, Andrew Moy, Steven Phipps, Andrew Smith, Tessa Vance, Meredith Nation) and the French National
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 875, 234]]<|/det|>
+Center for Coring and Drilling (C2FN, funded by INSU) for critical drilling, logistic, and analytical support at ABN and other sites. Finally, we acknowledge the Glacioclism- SAMBA, ITASE, and IPICS 2kyr Array programs for SMB data, the Air- O- Sol facility at IGE for microbial culturing, and additional support from the MITACS Globalink program and JSPS- CNRS joint research program.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 256, 200, 273]]<|/det|>
+## Funding:  
+
+<|ref|>text<|/ref|><|det|>[[115, 279, 860, 425]]<|/det|>
+SCADI: European Horizon Marie Sklodowska- Curie individual fellowship 889508  LabEx OSUG@2020 Investissements d'avenir: ANR10 LABX56  EAIIST: ANR- 16- CE01- 0011- 01  BNP- Paribas Climate Initiative programs 1115 (CHICTABA), 1117 (CAPOXI 35- 75), and 1169 (EAIIST)  CLIMCOR Equipex: ANR- 11- EQPX- 0009  MEXT/JSPS KAKENHI: 20H0496  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 439, 310, 454]]<|/det|>
+## Author contributions:  
+
+<|ref|>text<|/ref|><|det|>[[115, 460, 585, 630]]<|/det|>
+Conceptualization: PDA, JS, NC, MC  Investigation: All authors  Formal analysis: PDA, APMS, ELM  Visualization: PDA, ELM  Funding acquisition: PDA, JS, MC  Writing - original draft: PDA, APMS, PC  Writing - review & editing: All authors  
+
+<|ref|>text<|/ref|><|det|>[[118, 625, 585, 642]]<|/det|>
+The authors declare that they have no competing interests  
+
+<|ref|>text<|/ref|><|det|>[[118, 649, 873, 700]]<|/det|>
+Data and materials availability: Data and code are available for reviewers at doi:10.5281/zenodo.5793694. Data have been submitted to PANGAEA and are in review for publishing.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 738, 196, 754]]<|/det|>
+## Methods  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 771, 645, 789]]<|/det|>
+## Mathematical framework for \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) and SMB relationships  
+
+<|ref|>text<|/ref|><|det|>[[117, 804, 860, 892]]<|/det|>
+A linear relationship between \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) and the reciprocal of surface mass balance (SMB \(^{- 1}\) ) has been previously observed and reported in Antarctica \(^{17,34,26}\) . Here, we mathematically illustrate how this relationship between \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) and SMB arises through photolysis of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 874, 300]]<|/det|>
+\(\mathrm{NO}_3^-\) . We focus solely on the characteristics of \(\mathrm{NO}_3^-\) contained within a given horizontal plane of snow that is located at the snowpack surface at \(t = 0\) . We assume simplified sites with a stable surface mass balance \((SMB)\) , clear sky conditions, no surface roughness, and no significant compaction with burial in the photic zone. Any \(\mathrm{NO}_3^-\) that is photolyzed is immediately and permanently removed from the plane of snow, and \(\mathrm{NO}_3^-\) recycling \(^{29,34}\) is assumed not to affect \(\mathrm{NO}_3^-\) in the plane of snow during the burial process modeled here (i.e., after \(t = 0\) ).  
+
+<|ref|>text<|/ref|><|det|>[[117, 345, 864, 399]]<|/det|>
+Defining the relationship between \(\delta^{5}N_{\mathrm{NO3arc}}\) and SMB The time that it takes for a given horizontal plane of snow to be buried from the surface to a particular depth \(z\) is determined by the SMB \((\mathrm{kg m}^{- 2} \mathrm{a}^{- 1}\) , converted to \(\mathrm{cm s}^{- 1}\) ):  
+
+<|ref|>equation<|/ref|><|det|>[[455, 408, 812, 440]]<|/det|>
+\[t_{(z)} = \frac{z}{SMB} \quad (2)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 451, 803, 470]]<|/det|>
+The concentration of \(\mathrm{NO}_3^-\) within a plane of snow decays through time according to:  
+
+<|ref|>equation<|/ref|><|det|>[[395, 480, 812, 511]]<|/det|>
+\[\frac{d[N03]}{dt} = -J_{(z)}[NO3]_{(t)} \quad (3)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 525, 667, 544]]<|/det|>
+where \(J_{(z)}\) is the photolytic rate constant at a given depth defined as:  
+
+<|ref|>equation<|/ref|><|det|>[[433, 556, 812, 579]]<|/det|>
+\[J_{(z)} = \sigma \phi I_{(z)} \quad (4)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 592, 870, 678]]<|/det|>
+where \(\sigma\) is the absorption cross section for \(\mathrm{NO}_3^-\) photolysis \((\mathrm{cm}^2)\) , \(\phi\) is the quantum yield for \(\mathrm{NO}_3^-\) photolysis (molec photon \(^{- 1}\) ), and \(I_{(z)}\) is the actinic flux of ultraviolet irradiance (photon \(\mathrm{cm}^{- 2} \mathrm{s}^{- 1} \mathrm{nm}^{- 1}\) ) integrated over wavelengths that can induce photolysis of \(\mathrm{NO}_3^-\) . However, this photolytic rate "constant" changes with depth because actinic flux exponentially decays with depth as:  
+
+<|ref|>equation<|/ref|><|det|>[[444, 688, 812, 720]]<|/det|>
+\[I_{(z)} = I_0 e^{\frac{-z}{z_e}} \quad (5)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 732, 863, 802]]<|/det|>
+where \(I_0\) is the initial actinic flux that strikes the snow surface and \(z_e\) is the \(e\) - folding depth (cm) of the snowpack. Note that non- exponential decay of \(I\) in the top \(\sim 2\) cm of snowpack \(^{30}\) is simplified here by assuming the decay to be exponential from the snow surface. Equation (3) can then be expressed as:  
+
+<|ref|>equation<|/ref|><|det|>[[367, 814, 812, 847]]<|/det|>
+\[\frac{d[N03]}{dt} = -\sigma \phi I_0 e^{\frac{-z}{z_e}} [NO3]_{(t)} \quad (6)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 860, 469, 879]]<|/det|>
+Through Eq. (2), we can rewrite Eq. (6) as:
+
+<--- Page Split --->
+<|ref|>equation<|/ref|><|det|>[[353, 83, 812, 115]]<|/det|>
+\[\frac{d[N O3]}{d t} = -\sigma \phi I_{o} e^{\frac{-S M B t}{z_{e}}} [N O3]_{(t)} \quad (7)\]  
+
+<|ref|>text<|/ref|><|det|>[[115, 129, 822, 147]]<|/det|>
+In order to determine the \(\mathrm{NO}_3^-\) concentration at a given depth (i.e., \(SMB \cdot t\) ), we derive:  
+
+<|ref|>equation<|/ref|><|det|>[[375, 159, 812, 196]]<|/det|>
+\[\frac{d[N O3]}{[N O3]_{(t)}} = -\sigma \phi I_{o} e^{\frac{-S M B t}{z_{e}}} d t \quad (8)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 213, 328, 230]]<|/det|>
+And integrate to produce:  
+
+<|ref|>equation<|/ref|><|det|>[[359, 228, 812, 268]]<|/det|>
+\[\ln [N O3]_{(t)} = \frac{\sigma\phi I_{o}z_{e}e^{\frac{-S M B t}{z_{e}}}}{S M B} +C \quad (9)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 280, 285, 297]]<|/det|>
+Which simplifies to:  
+
+<|ref|>equation<|/ref|><|det|>[[374, 293, 812, 333]]<|/det|>
+\[[N O3]_{(t)} = e^{c}\frac{\sigma\phi I_{o}z_{e}e^{\frac{-S M B t}{z_{e}}}}{S M B} \quad (10)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 350, 459, 368]]<|/det|>
+At \(t = 0\) , \([N O_3^- ]_{(t)} = [N O_3^- ]_0\) and therefore:  
+
+<|ref|>equation<|/ref|><|det|>[[400, 380, 812, 410]]<|/det|>
+\[e^{c} = [N O3]_{0}e^{\frac{-\sigma\phi I_{o}z_{e}}{S M B}} \quad (11)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 423, 468, 441]]<|/det|>
+And thus combining Eq. (10) and Eq. (11):  
+
+<|ref|>equation<|/ref|><|det|>[[115, 453, 812, 495]]<|/det|>
+\[[N O3]_{(t)} = [N O3]_{0}e^{\frac{-\sigma\phi I_{o}z_{e}}{S M B}}e^{\frac{\sigma\phi I_{o}z_{e}e^{\frac{-S M B t}{z_{e}}}}{S M B}} = [N O3]_{0}e^{\frac{\sigma\phi I_{o}z_{e}e^{\frac{-S M B t}{z_{e}}}}{S M B}} \quad (12)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 509, 840, 577]]<|/det|>
+According to Eq. (12), as time (i.e., burial depth) increases, the \(\mathrm{NO}_3^-\) concentration will decrease. However, the rate of decrease will lessen over time as the value of \(SMB \cdot t\) approaches \(3z_{e}\) , and below the photic zone (i.e., \(z > 3z_{e}\) ) the \(\mathrm{NO}_3^-\) concentration is largely stable and equal to \(e^{c}\) .  
+
+<|ref|>text<|/ref|><|det|>[[117, 591, 860, 610]]<|/det|>
+Therefore, we can calculate the fraction of \(\mathrm{NO}_3^-\) archived below the photic zone \((f_{NO3arc})\) as:  
+
+<|ref|>equation<|/ref|><|det|>[[305, 622, 812, 662]]<|/det|>
+\[f_{NO3arc} = \frac{e^{c}}{[NO3]_{0}} = \frac{[\sigma\phi I_{o}z_{e}e^{\frac{-\sigma\phi I_{o}z_{e}}{S M B}}]}{[NO3]_{0}} = \frac{-\sigma\phi I_{o}z_{e}e^{\frac{-\sigma\phi I_{o}z_{e}}{S M B}}}{[NO3]_{0}} \quad (13)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 676, 853, 712]]<|/det|>
+To determine the \(\delta^{15}\mathrm{N}_{NO3arc}\) of this \(\mathrm{NO}_3^-\) , Rayleigh fractionation states that \(\delta^{15}\mathrm{N}_{NO3}\) can be calculated with the fractionation factor \(a\) by:  
+
+<|ref|>equation<|/ref|><|det|>[[189, 723, 812, 750]]<|/det|>
+\[\ln \left(\delta^{15}N_{NO3arc} + 1\right) = (\alpha -1)\ln \left(f_{NO3arc}\right) + \ln \left(\delta^{15}N_{NO3} + 1\right) \quad (14)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 763, 675, 782]]<|/det|>
+Through our prior calculation of \(f_{NO3arc}\) in Eq. (13), we thus produce:  
+
+<|ref|>equation<|/ref|><|det|>[[201, 793, 812, 824]]<|/det|>
+\[\ln \left(\delta^{15}N_{NO3arc} + 1\right) = (\alpha -1)\frac{-\sigma\phi I_{o}z_{e}}{S M B} +\ln \left(\delta^{15}N_{NO3} + 1\right) \quad (15)\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 836, 876, 890]]<|/det|>
+Because \((\alpha - 1)\) is negative for nitrogen during photolysis of \(\mathrm{NO}_3^{- 21,22,31,51 - 53}\) and the other parameters are positive, this means that \(\delta^{15}\mathrm{N}_{NO3arc}\) will vary linearly and positively with \(SMB^{- 1}\) when other parameters are held constant or scale linearly with \(SMB^{- 1}\) . We examine the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 84, 866, 118]]<|/det|>
+potential impacts of variability in these other parameters more thoroughly in Supplementary Text 1.  
+
+<|ref|>text<|/ref|><|det|>[[117, 133, 874, 187]]<|/det|>
+Based on modeling and field observations, SMB is the primary driver of change in \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values. Thus, the non- SMB variables can be subsumed into two parameters \(A\) and \(B\) to function as linear regression coefficients, producing Eq. (1) of the main text:  
+
+<|ref|>equation<|/ref|><|det|>[[320, 199, 812, 229]]<|/det|>
+\[\ln \left(\delta^{15}N_{NO3arc} + 1\right) = \frac{A}{SMB} +B \quad (Eq. (1))\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 241, 866, 277]]<|/det|>
+The inverse function of Eq. (1) can be used as a transfer function to calculate an SMB based on a \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value:  
+
+<|ref|>equation<|/ref|><|det|>[[345, 288, 812, 321]]<|/det|>
+\[\frac{1}{SMB} = \frac{\ln(\delta^{15}N_{NO3arc} + 1) - B}{A} \quad (Eq. (16))\]  
+
+<|ref|>text<|/ref|><|det|>[[117, 333, 842, 369]]<|/det|>
+Finally, since \(\ln (x + 1) \approx x\) when \(x \approx 0\) , a simpler relationship of Eq. (15) can be approximated, in a form similar to that previously reported from field observations \(^{23,26,34}\) :  
+
+<|ref|>equation<|/ref|><|det|>[[272, 381, 812, 412]]<|/det|>
+\[\delta^{15}N_{NO3arc} = (\alpha -1)\frac{-\sigma\phi l_oze}{SMB} +\delta^{15}N_{NO3o} \quad (Eq. (17))\]  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 424, 344, 443]]<|/det|>
+## Snow sampling techniques  
+
+<|ref|>text<|/ref|><|det|>[[115, 457, 866, 775]]<|/det|>
+The \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values in our database are taken from a mix of previously reported values from Antarctic research traverses and values newly reported here (Figure 2). For all values, snow and ice containing \(\mathrm{NO}_3^-\) was sampled in the field in one of three techniques: 1) 1- 2 m deep snow pit with continuous sampling at regular intervals from top to bottom, 2) single sample taken of a well- mixed 5- 10 cm layer around the 1- m depth layer, and 3) drilled core later cut at desired intervals. Since current \(\mathrm{NO}_3^-\) isotopic analysis requires 50- 150 nmol of \(\mathrm{NO}_3^-\) , 0.25- 1.50 kg of snow or ice per sample \(^{19,21}\) are gathered to ensure a sufficient amount of \(\mathrm{NO}_3^-\) . Generally, the multiple samples produced by the snow pit technique offers the best and most flexible results, but the 1- m depth layer technique is valuable for quick sampling during limited stops, and cores are necessary to collect samples deeper than \(\approx 5 \mathrm{~m}\) .  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 821, 290, 839]]<|/det|>
+## Laboratory analyses  
+
+<|ref|>text<|/ref|><|det|>[[117, 859, 875, 913]]<|/det|>
+For \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) results included in our database that have been previously reported, readers are directed to the original papers for specific analytical and sampling techniques. For the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 80, 880, 610]]<|/det|>
+\(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) data newly reported here, snow and ice samples were collected into clean sealed plastic bags or tubs and stored frozen until melted at room temperature for analysis. The \(\mathrm{NO_3^- }\) mass fraction \((\omega (\mathrm{NO}_3^- ))\) was determined on aliquots by either a colorimetric method or ion chromatography with detection limits \(< 0.5\mathrm{ng}\mathrm{g}^{- 1}\) and precision of \(< 3\%^{21,22}\) . The remaining melted samples were passed through an anionic exchange resin (Bio- Rad™ AG 1- X8, chloride form), and the resulting trapped \(\mathrm{NO_3^- }\) was eluted with \(10\mathrm{ml}\) of NaCl 1 M solution. Isotopic analysis occurred at IGE- CNRS, Grenoble, France, where \(\mathrm{NO_3^- }\) in these samples was converted to \(\mathrm{N}_2\mathrm{O}\) with the denitrifying bacteria Pseudomonas aureofaciens (lacking nitrous oxide reductase), thermally decomposed into \(\mathrm{O_2}\) and \(\mathrm{N}_2\) on a \(900^{\circ}\mathrm{C}\) gold surface, and separated by gas chromatography with a GasBench IITM. Oxygen and nitrogen isotopic ratios were then measured on a Thermo Finnigan™ MAT 253 mass spectrometer \(^{54 - 57}\) . Isotopic effects from this analysis were corrected as described by Morin et al. (2009) and Frey at al. (2009), using the international reference materials USGS 32, USGS 34, and USGS 35 with ultrapure Dome C water used for standards and samples throughout the analyses to account for potential oxygen isotopic exchanges. Results are reported relative to Vienna Standard Mean Ocean Water (V- SMOW) for oxygen isotopes \(^{58}\) and \(\mathrm{N}_2\) - Air for nitrogen isotopes \(^{59}\) .  
+
+<|ref|>text<|/ref|><|det|>[[115, 624, 875, 910]]<|/det|>
+For snow pits with multiple sequential \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values, a single \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value was calculated as the aggregate of samples \(30+\) cm deep, weighted by the relative mass of \(\mathrm{NO_3^- }\) per sample. Although the photic zone boundary can extend lower than \(30\mathrm{cm}\) at some sites \(^{29,30}\) , this cutoff was deemed an acceptable compromise to include more data from pits that stopped at \(50\mathrm{cm}\) depth as the great majority of photolysis will have occurred within the top \(30\mathrm{cm}\) due to exponential decay of actinic flux and \(\omega (\mathrm{NO}_3^- )\) with depth. Exceptions to this were made for three coastal pits from Cap Prud'homme (weighted- means of \(3+\) cm samples), where high accumulation greatly reduces photolytic impact, higher snow impurities reduce the photic zone depth, and a broader aggregation is necessary to smooth seasonal cycles.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 880, 201]]<|/det|>
+Additionally, two pits from Dronning Maud Land were aggregated with \(15+\) cm samples based on shallow \(3z_{e}\) values (2–5 cm) calculated on site during snow pit sampling<sup>29</sup>. For cores included in our database, a single \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) value was calculated as the isotopic mean of samples extending from present back to no earlier than 1800 CE.  
+
+<|ref|>text<|/ref|><|det|>[[115, 220, 880, 778]]<|/det|>
+Noro et al. (2018) reported \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) values for 16 pits along the JARE54 and JARE57 transects, but the sampling methodology for these pits took a single well- mixed sample of the entire pit depth which included the entire photic zone. In order to estimate the \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) values of these sites (i.e., the value as if the photic zone snow had been excluded), we applied a correction factor calculated using data from other pits in our database that were taken on two similar transects spanning from the coast to other interior domes (Dome A and Dome C) of East Antarctica<sup>22,23</sup>. Because each of the pits on the Dome A and Dome C transects were continuously sampled at discrete intervals from the surface to a point below the photic zone, we calculated different weighted- mean \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) values for selected depth spans that matched the three extents of the JARE pits: 0–30 cm, 0–50 cm, and 0–80 cm. Corrective factors were calculated through the linear regression of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) vs. \(\delta^{15}\mathrm{N}_{\mathrm{NO3.X}}\) from Dome A/Dome C transect pits (where \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) is our database's \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) value from the archived zone and \(\delta^{15}\mathrm{N}_{\mathrm{NO3.X}}\) is the weighted- mean value of samples from the surface to depth \(x\) : 30, 50, or 80 cm) and applied to the JARE pit data through the appropriate depth correction (Table 1, 2). Corrections were not made for JARE samples where \(\delta^{15}\mathrm{N}_{\mathrm{NO3}} < 0\%\) , as these low \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) values strongly suggest that photolysis was not a significant factor at these coastal sites, and photic zone corrections were thus not warranted.  
+
+<|ref|>table<|/ref|><|det|>[[117, 821, 510, 898]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[115, 792, 866, 823]]<|/det|>
+Table 1. Linear regressions of \(\delta^{15}\mathrm{N}_{\mathrm{NO3arc}}\) vs. \(\delta^{15}\mathrm{N}_{\mathrm{NO3.X}}\) (where X is 30, 50, or 80 cm) calculated from nonJARE pit data in the \(\delta^{15}\mathrm{N}_{\mathrm{NO3}}\) database.   
+
+<table><tr><td>Depth correction</td><td>Slope (% / %)</td><td>Intercept (%)</td><td>r²</td></tr><tr><td>0–30 cm</td><td>1.9±0.1</td><td>-2.4±11.3</td><td>0.89</td></tr><tr><td>0–50 cm</td><td>1.6±0.1</td><td>-1.7±8.2</td><td>0.94</td></tr><tr><td>0–80 cm</td><td>1.5±0.1</td><td>-0.9±7.8</td><td>0.94</td></tr></table>
+
+<--- Page Split --->
+<|ref|>table<|/ref|><|det|>[[117, 128, 585, 403]]<|/det|>
+<|ref|>table_caption<|/ref|><|det|>[[115, 85, 860, 131]]<|/det|>
+Table 2. The \(\delta^{15}\mathrm{N_{NO3}}\) values for JARE sites included in our database as originally reported by Noro et al. (2018) and the \(\delta^{15}\mathrm{N_{NO3arc}}\) values corrected here to account for photic zone snow included in the original samples. Samples with original \(\delta^{15}\mathrm{N_{NO3}}\) values \(< 0\%\) (italicized) were not corrected.   
+
+<table><tr><td>JARE site</td><td>Depth (cm)</td><td>Original δ15NNO3 (%)</td><td>Corrected δ15NNO3arc (%)</td></tr><tr><td>Z2</td><td>0–80</td><td>20.6</td><td>30.8</td></tr><tr><td>IM0</td><td>0–50</td><td>25.7</td><td>40.3</td></tr><tr><td>NMD304</td><td>0–50</td><td>41.1</td><td>65.4</td></tr><tr><td>MD590</td><td>0–50</td><td>83.5</td><td>134.7</td></tr><tr><td>DF1</td><td>0–30</td><td>127.3</td><td>236.5</td></tr><tr><td>NDF</td><td>0–30</td><td>111.7</td><td>207.2</td></tr><tr><td>Plateau S</td><td>0–30</td><td>165.5</td><td>308.1</td></tr><tr><td>S80</td><td>0–30</td><td>90.7</td><td>167.8</td></tr><tr><td>Fuji Pass</td><td>0–30</td><td>74.3</td><td>137.0</td></tr><tr><td>DF2</td><td>0–30</td><td>118.6</td><td>220.1</td></tr><tr><td>S30</td><td>0–50</td><td>-19.0</td><td>-19.0</td></tr><tr><td>H42</td><td>0–50</td><td>-6.6</td><td>-6.6</td></tr><tr><td>H68</td><td>0–50</td><td>-14.5</td><td>-14.5</td></tr><tr><td>H88</td><td>0–50</td><td>-19.4</td><td>-19.4</td></tr><tr><td>H108</td><td>0–50</td><td>-6.4</td><td>-6.4</td></tr><tr><td>H128</td><td>0–50</td><td>14.1</td><td>21.3</td></tr></table>  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 424, 205, 440]]<|/det|>
+## SMB data  
+
+<|ref|>text<|/ref|><|det|>[[115, 461, 860, 680]]<|/det|>
+In our database, 74 \(\delta^{15}\mathrm{N_{NO3arc}}\) samples are represented by 51 unique direct ground measurements of SMB (SMBground) values observed at or near the \(\mathrm{NO_3}\) sampling site, with the numerical discrepancy due to some sites having replicate \(\delta^{15}\mathrm{N_{NO3arc}}\) samples. These previously reported SMBground values were determined by measuring the change in surface height on established stakes or poles, by measuring the mass between known volcanic or radioactivity horizons in an ice core, or by ground penetrating radar (GPR) identification of dated horizons \(^{10,22,23,60 - 67}\) .  
+
+<|ref|>text<|/ref|><|det|>[[115, 702, 875, 887]]<|/det|>
+Regional climate models can be used to estimate modern SMB rates for sites lacking ground observations \(^{7,12}\) , and we used the Modèle Atmosphérique Régional (MAR) version 3.6.4 with European Centre for Medium- Range Weather Forecasts “Interim” re- analysis data (ERA- interim) data as applied by Agosta et al. (2019) to model mean annual SMBs at all database sites for the period 1979–2017 \(^{12}\) . Because the MAR overestimates SMBs at higher and more interior sites of the East Antarctic plateau \(^{68}\) , we calculated a correction factor through linear
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 880, 199]]<|/det|>
+regressions of \(\mathrm{SMB}_{\mathrm{ground}}\) vs. MAR-estimated SMBs ( \(\mathrm{SMB}_{\mathrm{MAR}}\) ) for our 51 sites that have both values (Table 3, Figure 4). This correction was applied to all original MAR estimates to produce "adjusted-MAR" SMBs ( \(\mathrm{SMB}_{\mathrm{adjMAR}}\) ) that match more closely with ground observations. 
+
+<|ref|>table_caption<|/ref|><|det|>[[115, 215, 880, 260]]<|/det|>
+Table 3. A list of all sampling sites that have a \(\mathrm{SMB}_{\mathrm{ground}}\) observation with corresponding values of original \(\mathrm{SMB}_{\mathrm{MAR}}\) and \(\mathrm{SMB}_{\mathrm{adjMAR}}\) (Figure 4). The difference between the \(\mathrm{SMB}_{\mathrm{adjMAR}}\) and \(\mathrm{SMB}_{\mathrm{MAR}}\) values is given in the final column. 
+
+<|ref|>table<|/ref|><|det|>[[115, 258, 896, 904]]<|/det|>
+<table><tr><td>Site</td><td>\(\mathrm{SMB}_{\mathrm{ground}}\) (kg m⁻² a⁻¹)</td><td>\(\mathrm{SMB}_{\mathrm{MAR}}\) (kg m⁻² a⁻¹)</td><td>\(\mathrm{SMB}_{\mathrm{adjMAR}}\) (kg m⁻² a⁻¹)</td><td>\(\mathrm{SMB}_{\mathrm{AdjMAR}} - \mathrm{SMB}_{\mathrm{MAR}}\) (kg m⁻² a⁻¹)</td><td>\(\mathrm{SMB}_{\mathrm{adjMAR}} / \mathrm{SMB}_{\mathrm{MAR}}\)</td></tr><tr><td>Vostok</td><td>22.6</td><td>30.4</td><td>24.1</td><td>-6.3</td><td>0.79</td></tr><tr><td>DomeA</td><td>22.9</td><td>40.9</td><td>34.4</td><td>-6.5</td><td>0.84</td></tr><tr><td>ZtoA-P6</td><td>25.4</td><td>62.1</td><td>55.2</td><td>-6.9</td><td>0.89</td></tr><tr><td>DomeC</td><td>28.4</td><td>41.0</td><td>34.5</td><td>-6.5</td><td>0.84</td></tr><tr><td>DomeF</td><td>29.2</td><td>35.6</td><td>29.2</td><td>-6.4</td><td>0.82</td></tr><tr><td>NDF</td><td>30.9</td><td>33.2</td><td>26.8</td><td>-6.4</td><td>0.81</td></tr><tr><td>Plateau S</td><td>32.4</td><td>31.2</td><td>24.8</td><td>-6.4</td><td>0.79</td></tr><tr><td>ZtoA-P5</td><td>33.3</td><td>61.5</td><td>54.6</td><td>-6.9</td><td>0.89</td></tr><tr><td>preeaiist.18</td><td>34.0</td><td>47.8</td><td>41.1</td><td>-6.7</td><td>0.86</td></tr><tr><td>S80Jare</td><td>37.5</td><td>31.3</td><td>24.9</td><td>-6.4</td><td>0.80</td></tr><tr><td>MD590</td><td>37.9</td><td>42.6</td><td>36.0</td><td>-6.6</td><td>0.85</td></tr><tr><td>Fuji Pass</td><td>40.7</td><td>35.0</td><td>28.6</td><td>-6.4</td><td>0.82</td></tr><tr><td>ZtoA-P4</td><td>54.8</td><td>55.4</td><td>48.6</td><td>-6.8</td><td>0.88</td></tr><tr><td>NMD304</td><td>65.8</td><td>75.6</td><td>68.4</td><td>-7.2</td><td>0.90</td></tr><tr><td>IM0</td><td>68.5</td><td>99.2</td><td>91.6</td><td>-7.6</td><td>0.92</td></tr><tr><td>Kohnen</td><td>75.0</td><td>97.6</td><td>90.0</td><td>-7.6</td><td>0.92</td></tr><tr><td>posteaiist.asuma05</td><td>76.0</td><td>288.9</td><td>266.9</td><td>-22.0</td><td>0.92</td></tr><tr><td>preeaiist.15</td><td>80.0</td><td>70.0</td><td>62.9</td><td>-7.1</td><td>0.90</td></tr><tr><td>preeaiist.13</td><td>86.0</td><td>95.1</td><td>87.5</td><td>-7.6</td><td>0.92</td></tr><tr><td>ZtoA-P3</td><td>90.7</td><td>76.3</td><td>69.1</td><td>-7.2</td><td>0.91</td></tr><tr><td>CPH.D5</td><td>97.6</td><td>139.7</td><td>124.8</td><td>-8.3</td><td>0.89</td></tr><tr><td>ZtoA-P2</td><td>99.4</td><td>95.6</td><td>88.0</td><td>-7.6</td><td>0.92</td></tr><tr><td>Z2</td><td>113.5</td><td>116.9</td><td>108.9</td><td>-8.0</td><td>0.93</td></tr><tr><td>posteaiist.stop36</td><td>118.0</td><td>147.5</td><td>138.3</td><td>-9.2</td><td>0.94</td></tr><tr><td>CPH.D24</td><td>120.0</td><td>247.2</td><td>228.9</td><td>-18.3</td><td>0.93</td></tr><tr><td>preeaiist.12</td><td>130.0</td><td>113.3</td><td>105.4</td><td>-7.9</td><td>0.93</td></tr><tr><td>ABN</td><td>130.0</td><td>122.0</td><td>113.9</td><td>-8.1</td><td>0.93</td></tr><tr><td>posteaiist.asuma06</td><td>131.0</td><td>280.4</td><td>259.1</td><td>-21.3</td><td>0.92</td></tr><tr><td>H128</td><td>158.8</td><td>210.2</td><td>195.3</td><td>-14.9</td><td>0.93</td></tr><tr><td>preeaiist.06</td><td>160.0</td><td>286.2</td><td>264.4</td><td>-21.8</td><td>0.92</td></tr><tr><td>preeaiist.07</td><td>171.0</td><td>269.5</td><td>249.2</td><td>-20.3</td><td>0.92</td></tr><tr><td>ZtoA-P1</td><td>172.0</td><td>153.4</td><td>143.7</td><td>-9.7</td><td>0.94</td></tr><tr><td>H108</td><td>185.0</td><td>250.7</td><td>232.1</td><td>-18.6</td><td>0.93</td></tr><tr><td>preeaiist.09</td><td>198.0</td><td>201.0</td><td>186.9</td><td>-14.1</td><td>0.93</td></tr><tr><td>H88</td><td>207.3</td><td>245.5</td><td>227.4</td><td>-18.1</td><td>0.93</td></tr><tr><td>posteaiist.asuma04</td><td>215.0</td><td>304.2</td><td>280.8</td><td>-23.4</td><td>0.92</td></tr><tr><td>posteaiist.asuma09</td><td>219.0</td><td>256.6</td><td>237.5</td><td>-19.1</td><td>0.93</td></tr><tr><td>H68</td><td>225.4</td><td>255.7</td><td>236.7</td><td>-19.0</td><td>0.93</td></tr><tr><td>H42</td><td>234.9</td><td>272.5</td><td>251.9</td><td>-20.6</td><td>0.92</td></tr><tr><td>posteaiist.asuma02</td><td>239.0</td><td>343.0</td><td>316.0</td><td>-27.0</td><td>0.92</td></tr><tr><td>posteaiist.asuma10</td><td>240.0</td><td>232.0</td><td>215.1</td><td>-16.9</td><td>0.93</td></tr></table>
+
+<--- Page Split --->
+<|ref|>table<|/ref|><|det|>[[118, 82, 895, 237]]<|/det|>
+
+<table><tr><td>posteaiist.asuma11</td><td>243.0</td><td>203.6</td><td>189.3</td><td>-14.3</td><td>0.93</td></tr><tr><td>S30-JARE</td><td>271.9</td><td>288.6</td><td>266.6</td><td>-22.0</td><td>0.92</td></tr><tr><td>posteaiist.asuma07</td><td>273.0</td><td>271.1</td><td>250.7</td><td>-20.4</td><td>0.92</td></tr><tr><td>preeaiist.04</td><td>280.0</td><td>330.1</td><td>304.3</td><td>-25.8</td><td>0.92</td></tr><tr><td>posteaiist.asuma01</td><td>321.0</td><td>337.4</td><td>310.9</td><td>-26.5</td><td>0.92</td></tr><tr><td>preeaiist.03</td><td>337.7</td><td>366.0</td><td>337.0</td><td>-29.0</td><td>0.92</td></tr><tr><td>cph.d17</td><td>446.0</td><td>178.1</td><td>166.1</td><td>-12.0</td><td>0.93</td></tr><tr><td>preeaiist.02</td><td>487.8</td><td>439.0</td><td>403.3</td><td>-35.7</td><td>0.92</td></tr><tr><td>asuma.2016.2</td><td>488.0</td><td>312.5</td><td>288.3</td><td>-24.2</td><td>0.92</td></tr><tr><td>asuma.2016.1</td><td>548.0</td><td>366.5</td><td>337.4</td><td>-29.1</td><td>0.92</td></tr></table>
+
+<|ref|>image<|/ref|><|det|>[[132, 268, 810, 732]]<|/det|>
+
+<|ref|>image_caption<|/ref|><|det|>[[117, 748, 877, 820]]<|/det|>
+<center>Figure 4. Linear regressions of $\mathrm {SMB}_{\mathrm {ground}}$  versus $\mathrm {SMB}_{\mathrm {MAR}}$  for the period 1979-2017 at the 51 sites with $\mathrm {SMB}_{\mathrm {ground}}$  observations, with 95% confidence intervals of the regressions shaded. Sites are subset for two overlapping regressions that intersect at (138, 130). These linear regressions were applied to the $\mathrm {SMB}_{\mathrm {MAR}}$  values for all sampling sites to produce the $\mathrm {SMB}_{\mathrm {adjMA}}$  used in analyses. The dashed line represents a slope of 1 (i.e., if the $\mathrm {SMB}_{\mathrm {MAR}}$  perfectly matched the $\mathrm {SMB}_{\mathrm {ground}}$ ).</center>
+
+<|ref|>text<|/ref|><|det|>[[117, 832, 839, 846]]<|/det|>
+A linear regression was calculated for two overlapping subsets of sites: one for the set of
+
+<|ref|>text<|/ref|><|det|>[[117, 864, 835, 880]]<|/det|>
+well-grouped sites where the $\mathrm {SMB}_{\mathrm {MAR}}$  is $<175kgm^{-2}a^{-1}$  and another for all sites where
+
+<|ref|>text<|/ref|><|det|>[[117, 897, 855, 914]]<|/det|>
+$\mathrm {SMB}_{\mathrm {MAR}}$  is $>110kgm^{-2}a^{-1}$ . This first regression is tightly constrained $(SMB_{\mathrm {ground}}=1.0\pm$
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 876, 470]]<|/det|>
+\(0.1 \times \mathrm{SMB}_{\mathrm{MAR}} - 5.8 \pm 7.1\) , \(\mathrm{r}^2 = 0.84\) , and it performs well to better align the \(\mathrm{SMB}_{\mathrm{MAR}}\) estimates with the \(\mathrm{SMB}_{\mathrm{ground}}\) values at low SMB sites. The second regression covers samples with where some differences between \(\mathrm{SMB}_{\mathrm{MAR}}\) and \(\mathrm{SMB}_{\mathrm{ground}}\) are very large, particularly at lower elevation sites where intense aeolian erosion and deposition can produce highly variable local SMB rates that are difficult to accurately model \(^{12,13}\) . As a result, this regression is weaker \(\mathrm{(SMB_{ground} = 0.9 \pm 0.2 \times SMB_{MAR} + 4.2 \pm 57.9, r^2 = 0.35)}\) than the first regression, but we apply it while acknowledging the possibility of wide deviations. The two regressions intersect at \(\mathrm{(SMB_{MAR} = 138 kg m^{-2} a^{-1}}\) , \(\mathrm{SMB}_{\mathrm{ground}} = 130 \mathrm{kg m^{-2} a^{-1}}\) ), and thus \(\mathrm{SMB}_{\mathrm{adjMAR}}\) values were calculated by applying the first regression to all sites where \(\mathrm{SMB}_{\mathrm{MAR}} \leq 138 \mathrm{kg m^{-2} a^{-1}}\) and applying the second regression to all sites where \(\mathrm{SMB}_{\mathrm{MAR}} > 138 \mathrm{kg m^{-2} a^{-1}}\) . We constructed our final primary SMB dataset for the analysis of \(\delta^{5}\mathrm{N}_{\mathrm{NO3arc}}\) samples by using the best quality SMB data for each site: \(\mathrm{SMB}_{\mathrm{ground}}\) if available and \(\mathrm{SMB}_{\mathrm{adjMAR}}\) otherwise.  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 521, 480, 539]]<|/det|>
+## Transfer function and SMB reconstruction  
+
+<|ref|>text<|/ref|><|det|>[[115, 560, 870, 880]]<|/det|>
+We modeled linear relationships between \(\ln (\delta^{5}\mathrm{N}_{\mathrm{NO3}} + 1)\) and \(\mathrm{SMB}^{- 1}\) based on Eq. (15) using previously reported parameter values to compare our theoretical framework to field results and to better understand the sensitivity of the relationships to photolytic and fractionation factors (Supplementary Text 1). To determine the coefficients in Eq. (1) from our field data, we performed linear regressions using all database samples and the primary SMB dataset. Additional regressions (Supplementary Text 2) were performed for subsets of the database based on SMB type ( \(\mathrm{SMB}_{\mathrm{ground}}\) vs. \(\mathrm{SMB}_{\mathrm{adjMAR}}\) ). With regression coefficients determined for Eq. (1), we modeled the spatial distribution of \(\delta^{5}\mathrm{N}_{\mathrm{NO3arc}}\) values across Antarctica using gridded mean SMBs (MAR- ERA- interim, 1979–2015) at a \(35 \mathrm{km}\) resolution \(^{12}\) that were converted to \(\mathrm{SMB}_{\mathrm{adjMAR}}\) as previously described.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 875, 304]]<|/det|>
+For reconstructing the ABN \(\mathrm{SMB}_{\mathrm{015N}}\) history, the ABN1314- 103 ice core was cut into \(0.33\mathrm{m}\) samples from 5 to \(103\mathrm{m}\) , and these were processed for \(\mathrm{NO}_3^-\) isotopes in 2016 as previously described. We applied an annually- resolved age model (ALC01112018) based on seasonal ion and water isotope cycles and constrained by volcanic horizons that was originally developed for a longer core also taken at ABN. Each \(1\mathrm{m}\) ice core segment was individually weighed prior to cutting, and the mass and volume were used to calculate a SMB profile based on dated ice density changes ( \(\mathrm{SMB}_{\mathrm{density}}\) ).  
+
+<|ref|>text<|/ref|><|det|>[[115, 323, 875, 670]]<|/det|>
+To determine past topographical effects on SMBs, a MALA GPR device towing a RTA antenna on the surface (50 MHz out, 100 MHz in) was operated for a \(65\mathrm{km}\) transect upstream of the coring site as part of the 2013–2014 campaign. Radar was triggered every 2 seconds (i.e., every 6–7 m along the transect) with a recording time window of 3000 nanoseconds that captured returns down to \(300\mathrm{m}\) depth. After postprocessing<sup>41</sup>, isochronic internal reflecting horizons were identified to \(220\mathrm{m}\) depth, digitized with ReflexW software, and dated by connecting to the ALC01112018 age-depth model. Using a density profile taken from a longer ice core simultaneously drilled at ABN, 2D fields (depth by transect distance) were calculated for age, mean accumulation rate, and local accumulation rate. The mean accumulation rate to the most shallow reflecting horizon was taken as the upstream topographical effect on SMBs (i.e., \(\mathrm{SMB}_{\mathrm{GPR}}\) ).  
+
+<|ref|>text<|/ref|><|det|>[[115, 690, 846, 776]]<|/det|>
+Statistical analyses, regressions, SMB reconstructions, visualizations, and other statistical analyses were perform using the R programming language with packages ggplot2, RColorBrewer, gridExtra, cowplot, and tidyverse.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[122, 90, 854, 496]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[117, 510, 876, 569]]<|/det|>
+<center>Figure 5. Local accumulation rate variability with depth along the upstream ABN transect determined from GPR identification of isochronic internal reflective horizons. Accumulation rates have an original depth resolution of \(0.5 \mathrm{m}\) which is smoothed through a moving age-depth average with a cosine weighting window to reduce isochron artifacts<sup>41</sup>. </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 100, 808, 144]]<|/det|>
+1. Fretwell, P. et al. Bedmap2: improved ice bed, surface and thickness datasets for Antarctica. The Cryosphere 7, 375-393 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[115, 155, 866, 199]]<|/det|>
+2. Juckes, M. N., James, I. N. & Blackburn, M. The influence of Antarctica on the momentum budget of the southern extratropics. Quarterly Journal of the Royal Meteorological Society 120, 1017-1044 (1994).  
+
+<|ref|>text<|/ref|><|det|>[[115, 209, 868, 253]]<|/det|>
+3. M.R. van den Broeke. On the role of Antarctica as heat sink for the global atmosphere. J. Phys. IV France 121, 115-124 (2004).  
+
+<|ref|>text<|/ref|><|det|>[[115, 264, 831, 280]]<|/det|>
+4. Bronselaer, B. et al. Change in future climate due to Antarctic meltwater. Nature 564, 53-58 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[115, 291, 844, 334]]<|/det|>
+5. Starr, A. et al. Antarctic icebergs reorganize ocean circulation during Pleistocene glacials. Nature 589, 236-241 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[115, 345, 860, 418]]<|/det|>
+6. IPCC. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. 151 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[115, 428, 834, 472]]<|/det|>
+7. Shepherd, A. et al. Mass balance of the Antarctic Ice Sheet from 1992 to 2017. Nature 558, 219-222 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[115, 483, 874, 555]]<|/det|>
+8. Martin-Español, A. et al. Spatial and temporal Antarctic Ice Sheet mass trends, glacio-isostatic adjustment, and surface processes from a joint inversion of satellite altimeter, gravity, and GPS data. Journal of Geophysical Research: Earth Surface 121, 182-200 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[115, 565, 872, 609]]<|/det|>
+9. Stauffer, B., Flückiger, J., Wolff, E. & Barnes, P. The EPICA deep ice cores: first results and perspectives. Annals of Glaciology 39, 93-100 (2004).  
+
+<|ref|>text<|/ref|><|det|>[[115, 619, 875, 662]]<|/det|>
+10. Parrenin, F. et al. 1-D-ice flow modelling at EPICA Dome C and Dome Fuji, East Antarctica. Clim. Past 3, 243-259 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[115, 673, 872, 716]]<|/det|>
+11. Thomas, E. R. et al. Regional Antarctic snow accumulation over the past 1000 years. Clim. Past 13, 1491-1513 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[115, 728, 865, 771]]<|/det|>
+12. Agosta, C. et al. Estimation of the Antarctic surface mass balance using the regional climate model MAR (1979-2015) and identification of dominant processes. The Cryosphere 13, 281-296 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[115, 782, 868, 826]]<|/det|>
+13. Agosta, C. et al. A 40-year accumulation dataset for Adelie Land, Antarctica and its application for model validation. Climate Dynamics 38, 75-86 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[115, 837, 866, 909]]<|/det|>
+14. Gallee, H. et al. Transport of Snow by the Wind: A Comparison Between Observations in Adélie Land, Antarctica, and Simulations Made with the Regional Climate Model MAR. Boundary-Layer Meteorology 146, 133-147 (2013).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 872, 155]]<|/det|>
+15. Vimeux, F., Cuffey, K. M. & Jouzel, J. New insights into Southern Hemisphere temperature changes from Vostok ice cores using deuterium excess correction. Earth and Planetary Science Letters 203, 829–843 (2002).  
+
+<|ref|>text<|/ref|><|det|>[[117, 165, 848, 210]]<|/det|>
+16. Cauquoin, A. et al. Comparing past accumulation rate reconstructions in East Antarctic ice cores using 10Be, water isotopes and CMIP5-PMIP3 models. Clim. Past 11, 355–367 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[117, 220, 835, 264]]<|/det|>
+17. Freyer, H. D., Kobel, K., Delmas, R. J., Kley, D. & Legrand, M. R. First results of 15N/14N ratios in nitrate from alpine and polar ice cores. Tellus B 48, 93–105 (1996).  
+
+<|ref|>text<|/ref|><|det|>[[117, 274, 876, 319]]<|/det|>
+18. Legrand, M., Wolff, E., Wagenbach, D. & Jacka, T. Antarctic aerosol and snowfall chemistry: implications for deep Antarctic ice-core chemistry. Annals of Glaciology, Vol 29, 1999 29, 66–72 (1999).  
+
+<|ref|>text<|/ref|><|det|>[[117, 329, 840, 373]]<|/det|>
+19. Wolff, E. Nitrate in Polar Ice. in Ice Core Studies of Global Biogeochemical Cycles (Springer-Verlag, 1995).  
+
+<|ref|>text<|/ref|><|det|>[[117, 384, 818, 429]]<|/det|>
+20. Röhlisberger, R. et al. Nitrate in Greenland and Antarctic ice cores: a detailed description of post-depositional processes. Annals of Glaciology 35, 209–216 (2002).  
+
+<|ref|>text<|/ref|><|det|>[[117, 438, 860, 511]]<|/det|>
+21. Frey, M., Savarino, J., Morin, S., Erbland, J. & Martins, J. Photolysis imprint in the nitrate stable isotope signal in snow and atmosphere of East Antarctica and implications for reactive nitrogen cycling. Atmospheric Chemistry and Physics 9, 8681–8696 (2009).  
+
+<|ref|>text<|/ref|><|det|>[[117, 520, 872, 592]]<|/det|>
+22. Erbland, J. et al. Air-snow transfer of nitrate on the East Antarctic Plateau - Part 1: Isotopic evidence for a photolytically driven dynamic equilibrium in summer. Atmospheric Chemistry and Physics 13, 6403–6419 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[117, 602, 844, 674]]<|/det|>
+23. Shi, G. et al. Investigation of post-depositional processing of nitrate in East Antarctic snow: isotopic constraints on photolytic loss, re-oxidation, and source inputs. Atmospheric Chemistry and Physics 15, 9435–9453 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[117, 685, 868, 729]]<|/det|>
+24. Grannas, A. et al. An overview of snow photochemistry: evidence, mechanisms and impacts. Atmospheric Chemistry and Physics 7, 4329–4373 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[117, 739, 816, 783]]<|/det|>
+25. Berhanu, T. et al. Laboratory study of nitrate photolysis in Antarctic snow. II. Isotopic effects and wavelength dependence. Journal of Chemical Physics 140, (2014).  
+
+<|ref|>text<|/ref|><|det|>[[117, 794, 860, 838]]<|/det|>
+26. Noro, K. et al. Spatial variation of isotopic compositions of snowpack nitrate related to post-depositional processes in eastern Dronning Maud Land, East Antarctica. Geochemical Journal 52, e7–e14 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[117, 848, 811, 892]]<|/det|>
+27. Shi, G. et al. Isotope fractionation of nitrate during volatilization in snow: A field investigation in Antarctica. Geophysical Research Letters 46, 3287–3297 (2019).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 875, 128]]<|/det|>
+28. Noro, K. & Takenaka, N. Post-depositional loss of nitrate and chloride in Antarctic snow by photolysis and sublimation: a field investigation. POLAR 39, (2020).  
+
+<|ref|>text<|/ref|><|det|>[[117, 138, 866, 209]]<|/det|>
+29. Winton, V. H. L. et al. Deposition, recycling, and archival of nitrate stable isotopes between the air-snow interface: comparison between Dronning Maud Land and Dome C, Antarctica. Atmos. Chem. Phys. 20, 5861-5885 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[117, 220, 861, 291]]<|/det|>
+30. Zatko, M. C. et al. The influence of snow grain size and impurities on the vertical profiles of actinic flux and associated NOx emissions on the Antarctic and Greenland ice sheets. Atmos. Chem. Phys. 13, 3547-3567 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[117, 302, 857, 375]]<|/det|>
+31. Zatko, M., Geng, L., Alexander, B., Sofen, E. & Klein, K. The impact of snow nitrate photolysis on boundary layer chemistry and the recycling and redistribution of reactive nitrogen across Antarctica and Greenland in a global chemical transport model. Atmos. Chem. Phys. 16, 2819-2842 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[117, 385, 844, 429]]<|/det|>
+32. France, J. L. et al. Snow optical properties at Dome C (Concordia), Antarctica; implications for snow emissions and snow chemistry of reactive nitrogen. Atmos. Chem. Phys. 11, 9787-9801 (2011).  
+
+<|ref|>text<|/ref|><|det|>[[117, 439, 866, 483]]<|/det|>
+33. Wolff, E., Jones, A., Martin, T. & Grenfell, T. Modelling photochemical NOX production and nitrate loss in the upper snowpack of Antarctica. Geophysical Research Letters 29, (2002).  
+
+<|ref|>text<|/ref|><|det|>[[117, 494, 865, 538]]<|/det|>
+34. Erbland, J. et al. Air-snow transfer of nitrate on the East Antarctic Plateau - Part 2: An isotopic model for the interpretation of deep ice-core records. Atmospheric Chemistry and Physics 15, 12079-12113 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[117, 549, 856, 593]]<|/det|>
+35. Jiang, S. et al. Nitrate preservation in snow at Dome A, East Antarctica from ice core concentration and isotope records. Atmospheric Environment 213, 405-412 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[117, 603, 811, 647]]<|/det|>
+36. Howat, I. M., Porter, C., Smith, B. E., Noh, M.-J. & Morin, P. The Reference Elevation Model of Antarctica. The Cryosphere 13, 665-674 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[117, 657, 875, 701]]<|/det|>
+37. Hui, F. et al. Mapping blue-ice areas in Antarctica using ETM+ and MODIS data. Annals of Glaciology 55, 129-137 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[117, 712, 877, 784]]<|/det|>
+38. Savarino, J., Kaiser, J., Morin, S., Sigman, D. & Thiemens, M. Nitrogen and oxygen isotopic constraints on the origin of atmospheric nitrate in coastal Antarctica. Atmospheric Chemistry and Physics 7, 1925-1945 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[117, 795, 871, 866]]<|/det|>
+39. Servettaz, A., Landais, A. & Orsi, A. Two thousand years of temperature variability on the lower East Antarctic Plateau inferred from the analysis of stable isotopes of water and inert gases in the Aurora Basin North ice core. (Université Paris-Saclay, 2021).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 84, 874, 128]]<|/det|>
+40. Mouginot, J., Rignot, E. & Scheuchl, B. Continent-wide, interferometric SAR phase, mapping of Antarctic ice velocity. Geophysical Research Letters 46, 9710-9718 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[115, 139, 875, 209]]<|/det|>
+41. Le Meur, E. et al. Spatial and temporal distributions of surface mass balance between Concordia and Vostok stations, Antarctica, from combined radar and ice core data: first results and detailed error analysis. The Cryosphere 12, 1831-1850 (2018).  
+
+<|ref|>text<|/ref|><|det|>[[115, 220, 864, 264]]<|/det|>
+42. Stenni, B. et al. Eight centuries of volcanic signal and climate change at Talos Dome (East Antarctica). J. Geophys. Res.-Atmos. 107, 13 (2002).  
+
+<|ref|>text<|/ref|><|det|>[[115, 275, 855, 319]]<|/det|>
+43. Frezzotti, M., Scarchilli, C., Becagli, S., Proposito, M. & Urbini, S. A synthesis of the Antarctic surface mass balance during the last 800 yr. The Cryosphere 7, 303-319 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[115, 330, 876, 374]]<|/det|>
+44. Urbini, S. et al. Historical behaviour of Dome C and Talos Dome (East Antarctica) as investigated by snow accumulation and ice velocity measurements. Global and Planetary Change 60, 576-588 (2008).  
+
+<|ref|>text<|/ref|><|det|>[[115, 385, 798, 429]]<|/det|>
+45. Wille, J. D. et al. Antarctic atmospheric river climatology and precipitation impacts. Journal of Geophysical Research: Atmospheres 126, (2021).  
+
+<|ref|>text<|/ref|><|det|>[[115, 439, 843, 483]]<|/det|>
+46. Turner, J. et al. The Dominant Role of Extreme Precipitation Events in Antarctic Snowfall Variability. Geophysical Research Letters 46, 3502-3511 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[115, 494, 861, 538]]<|/det|>
+47. Frezzotti, M., Gandolfi, S., Marca, F. L. & Urbini, S. Snow dunes and glazed surfaces in Antarctica: new field and remote-sensing data. Annals of Glaciology 34, 81-88 (2002).  
+
+<|ref|>text<|/ref|><|det|>[[115, 549, 870, 619]]<|/det|>
+48. Libois, Q., Picard, G., Arnaud, L., Morin, S. & Brun, E. Modeling the impact of snow drift on the decameter-scale variability of snow properties on the Antarctic Plateau. Journal of Geophysical Research: Atmospheres 119, 11662-11681 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[115, 630, 878, 675]]<|/det|>
+49. Picard, G., Arnaud, L., Caneill, R., Lefebvre, E. & Lamare, M. Observation of the process of snow accumulation on the Antarctic Plateau by time lapse laser scanning. The Cryosphere 13, 1983-1999 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[115, 685, 875, 729]]<|/det|>
+50. Neubauer, C. et al. Stable isotope analysis of intact oxyanions using electrospray quadrupole-orbitrap mass spectrometry. Anal. Chem. 92, 3077-3085 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[115, 740, 866, 812]]<|/det|>
+51. Meusinger, C., Berhanu, T. A., Erbland, J., Savarino, J. & Johnson, M. S. Laboratory study of nitrate photolysis in Antarctic snow. I. Observed quantum yield, domain of photolysis, and secondary chemistry. The Journal of Chemical Physics 140, 244305 (2014).  
+
+<|ref|>text<|/ref|><|det|>[[115, 823, 860, 866]]<|/det|>
+52. Chu, L. & Anastasio, C. Quantum yields of hydroxyl radical and nitrogen dioxide from the photolysis of nitrate on ice. J. Phys. Chem. A 107, 9594-9602 (2003).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 82, 856, 128]]<|/det|>
+53. Benedict, K. B., McFall, A. S. & Anastasio, C. Quantum yield of nitrite from the photolysis of aqueous nitrate above 300 nm. Environ. Sci. Technol. 51, 4387-4395 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[115, 138, 875, 210]]<|/det|>
+54. Kaiser, J., Hastings, M. G., Houlton, B. Z., Röckmann, T. & Sigman, D. M. Triple oxygen isotope analysis of nitrate Using the denitrifier method and thermal decomposition of N2O. Analytical Chemistry 79, 599-607 (2007).  
+
+<|ref|>text<|/ref|><|det|>[[115, 220, 877, 266]]<|/det|>
+55. Morin, S. et al. Comprehensive isotopic composition of atmospheric nitrate in the Atlantic Ocean boundary layer from 65 degrees S to 79 degrees N. Journal of Geophysical Research-Atmospheres 114, (2009).  
+
+<|ref|>text<|/ref|><|det|>[[115, 276, 825, 320]]<|/det|>
+56. Sigman, D. M. et al. A bacterial method for the nitrogen isotopic analysis of nitrate in seawater and freshwater. Anal. Chem. 73, 4145-4153 (2001).  
+
+<|ref|>text<|/ref|><|det|>[[115, 330, 866, 402]]<|/det|>
+57. Casciotti, K. L., Sigman, D. M., Hastings, M. G., Bohlke, J. K. & Hilkert, A. Measurement of the oxygen isotopic composition of nitrate in seawater and freshwater using the denitrifier method. Anal. Chem. 74, 4905-4912 (2002).  
+
+<|ref|>text<|/ref|><|det|>[[115, 412, 872, 457]]<|/det|>
+58. Baertschi, P. Absolute 180 content of standard mean ocean water. Earth and Planetary Science Letters 31, 341-344 (1976).  
+
+<|ref|>text<|/ref|><|det|>[[115, 467, 868, 512]]<|/det|>
+59. Mariotti, A. Atmospheric nitrogen is a reliable standard for natural 15N abundance measurements. Nature 303, 685-687 (1983).  
+
+<|ref|>text<|/ref|><|det|>[[115, 521, 866, 567]]<|/det|>
+60. Pourchet, M. et al. Distribution and fall-out of 137Cs and other radionuclides over Antarctica. Journal of Glaciology 43, 435-445 (1997).  
+
+<|ref|>text<|/ref|><|det|>[[115, 576, 872, 621]]<|/det|>
+61. Ding, M. et al. Spatial variability of surface mass balance along a traverse route from Zhongshan station to Dome A, Antarctica. Journal of Glaciology 57, 658-666 (2011).  
+
+<|ref|>text<|/ref|><|det|>[[115, 631, 855, 676]]<|/det|>
+62. Verfaillie, D. et al. Snow accumulation variability derived from radar and firn core data along a 600 km transect in Adelie Land, East Antarctic plateau. The Cryosphere 6, 1345-1358 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[115, 686, 824, 731]]<|/det|>
+63. Favier, V. et al. An updated and quality controlled surface mass balance dataset for Antarctica. The Cryosphere 7, 583-597 (2013).  
+
+<|ref|>text<|/ref|><|det|>[[115, 741, 878, 812]]<|/det|>
+64. Hoshina, Y., Fujita, K., Iizuka, Y. & Motoyama, H. Inconsistent relationships between major ions and water stable isotopes in Antarctic snow under different accumulation environments. Polar Science 10, 1-10 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[115, 822, 868, 867]]<|/det|>
+65. Ding, M. et al. Re-assessment of recent (2008 - 2013) surface mass balance over Dome Argus, Antarctica. POLAR 35, (2016).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 868, 127]]<|/det|>
+66. Ekaykin, A. A. et al. Underestimation of Snow Accumulation Rate in Central Antarctica (Vostok Station) Derived from Stake Measurements. Russian Meteorology and Hydrology 45, 132–140 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[117, 137, 880, 210]]<|/det|>
+67. Sommer, S., Wagenbach, D., Mulvaney, R. & Fischer, H. Glacio-chemical study spanning the past 2 kyr on three ice cores from Dronning Maud Land, Antarctica: 2. Seasonally resolved chemical records. Journal of Geophysical Research: Atmospheres 105, 29423–29433 (2000).  
+
+<|ref|>text<|/ref|><|det|>[[117, 220, 852, 265]]<|/det|>
+68. Richter, A. et al. Surface mass balance models vs. stake observations: a comparison in the Lake Vostok region, central East Antarctica. Frontiers in Earth Science 9, 388 (2021).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 312, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[60, 130, 572, 150]]<|/det|>
+- Akersd15NManuscript20220128NComSupplement.docx
+
+<--- Page Split --->

preprint/preprint__491c12817894379960e50a02a9c4da0c41dd179e729ff4fd51850bf7dd1a6580/images_list.json

@@ -0,0 +1,62 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1 Temperature dependence of in-plane resistance for Bi₂Sr₂CaCu₂O₈+δ",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        196,
+        841,
+        377
+      ]
+    ],
+    "page_idx": 14
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. In-plane resistance \\((R)\\) and \\(ac\\) susceptibility \\((\\Delta \\chi^{\\prime})\\) as a function of",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        90,
+        850,
+        375
+      ]
+    ],
+    "page_idx": 15
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3 Pressure- \\(T_{c}\\) phase diagrams for \\(\\mathrm{Bi_2Sr_2CaCu_2O_{8 + \\delta}}\\) superconductors. The Right panels are the phase diagrams established by the experimental results from the under-doped (UD), optimally-doped (OP) and over-doped (OD) samples, together with the mapping information of temperature and pressure dependent \\(R\\) (shown in color scale). The left panel is a normalized phase diagram that is built on the basis of the experimental phase diagrams (the right panels). \\(P_{Tc - max}\\) and \\(P_{c}\\) stand for the critical pressures where \\(T_{c}\\) reaches the maximum and the zero, respectively. In the normalizing analysis, we define the pressure as \\(P_{Tc - max}\\) when \\((P - P_{Tc - max}) / (P_{c} - P_{Tc - max}) = 0\\) , and the pressure as \\(P_{c}\\) when \\((P - P_{Tc - max}) / (P_{c} - P_{Tc - max}) = 1\\) . The results of the normalizing analysis for \\(T_{c} / T_{c - max}\\) versus \\((P - P_{Tc - max}) / (P_{c} - P_{Tc - max})\\) and \\(T_{ins} / T_{c - max}\\) versus \\((P - P_{Tc - max}) / (P_{c} - P_{Tc - max})\\) show that the three kinds of samples display a universal quantum phase transition from the superconducting state to an insulating-like state. SC, M and I stand for superconducting state, metallic state and insulating-like state, respectively. The region of the M phase is determined by the critical value of \\(R / R_{290K}\\) where the quantum phase",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        90,
+        848,
+        393
+      ]
+    ],
+    "page_idx": 16
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4 Magnetoresistance \\((MR)\\) as a function of magnetic field \\((B)\\) for the UD, OP and OD \\(\\mathrm{Bi_2Sr_2CaCu_2O_{8 + \\delta}}\\) superconductors when they enter into an insulating-like state, and Hall coefficient information of the OP sample under pressure. (a)- (b) Plots of \\(MR\\) versus \\(B\\) for the UD, OP and OD samples measured at \\(4\\mathrm{K}\\) at \\(36.2\\mathrm{GPa}\\) , 41.2 GPa and 49 GPa, respectively. It is seen that all of them exhibit a positive magnetic effect. The red and beige arrows represent the directions of increasing and decreasing magnetic field. (d) Pressure dependence of Hall coefficient \\((R_H)\\) for the OD superconductor measured at \\(100\\mathrm{K}\\) .",
+    "footnote": [],
+    "bbox": [
+      [
+        150,
+        92,
+        845,
+        500
+      ]
+    ],
+    "page_idx": 18
+  }
+]

preprint/preprint__491c12817894379960e50a02a9c4da0c41dd179e729ff4fd51850bf7dd1a6580/preprint__491c12817894379960e50a02a9c4da0c41dd179e729ff4fd51850bf7dd1a6580.mmd

@@ -0,0 +1,244 @@
+
+# Universal quantum phase transition from superconducting to insulating-like states in pressurized Bi2Sr2CaCu2O8+δ superconductors  
+
+Liling Sun (llsun@iphy.ac.cn) Institute of Physics, Chinese Academy of Sciences https://orcid.org/0000- 0002- 6653- 7826  
+
+Yazhou Zhou Institute of Physics, Chinese Academy of Sciences  
+
+Jing Guo Institute of Physics, Chinese Academy of Sciences https://orcid.org/0000- 0002- 2164- 5065  
+
+Shu Cai Institute of Physics, Chinese Academy of Sciences  
+
+Jinyu Zhao Institute of Physics, Chinese Academy of Sciences  
+
+Genda Gu Brookhaven National Laboratory  
+
+Chengtian Lin Max- Planck- Inst of Solid State Res  
+
+Hongtao Yan Institute of Physics, Chinese Academy of Sciences  
+
+Cheng Huang Institute of Physics, Chinese Academy of Sciences  
+
+Chongli Yang Institute of Physics, Chinese Academy of Sciences  
+
+Sijin Long Institute of Physics, Chinese Academy of Sciences  
+
+Yu Gong Institute of Physics, Chinese Academy of Sciences  
+
+Yanchun Li Institute of High Energy Physics https://orcid.org/0000- 0003- 4540- 4546  
+
+Xiaodong Li Institute of High Energy Physics https://orcid.org/0000- 0002- 2290- 1198  
+
+Qi Wu Institute of Physics, Chinese Academy of Sciences  
+
+Jiangping Hu
+
+<--- Page Split --->
+
+Institute of Physics, Chinese Academy of Sciences  
+
+Xingjiang Zhou National Lab for Superconductivity, Institute of Physics, Chinese Academy of Sciences https://orcid.org/0000- 0002- 5261- 1386  
+
+Tao Xiang Institute of Physics, Chinese Academy of Sciences, Beijing National Laboratory for Condensed Matter Physics  
+
+## Article  
+
+Keywords: copper oxide, superconductions, Bi2Sr2CaCu2O8+δ, high- Tc superconductivity  
+
+Posted Date: June 9th, 2021  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 551119/v1  
+
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Version of Record: A version of this preprint was published at Nature Physics on February 17th, 2022. See the published version at https://doi.org/10.1038/s41567- 022- 01513- 2.
+
+<--- Page Split --->
+
+# Universal quantum phase transition from superconducting to insulating-like states in pressurized \(\mathrm{Bi_2Sr_2CaCu_2O_8 + \delta}\)  
+
+## superconductors  
+
+Yazhou Zhou \(^{1*}\) , Jing Guo \(^{1,6*}\) , Shu Cai \(^{1*}\) , Jinyu Zhao \(^{1,4}\) , Genda Gu \(^{2}\) , Chengtian Lin \(^{3}\) , Hongtao Yan \(^{1,4}\) , Cheng Huang \(^{1,4}\) , Chongli Yang \(^{1}\) , Sijin Long \(^{1,4}\) , Yu Gong \(^{5}\) , Yanchun Li \(^{5}\) , Xiaodong Li \(^{5}\) , Qi Wu \(^{1}\) , Jiangping Hu \(^{1,4}\) , Xingjiang Zhou \(^{1,4,6}\) , Tao Xiang \(^{1,4}\) and Liling Sun \(^{1,4,6}\)  
+
+\(^{1}\) Institute of Physics, National Laboratory for Condensed Matter Physics, Chinese Academy of Sciences, Beijing, 100190, China \(^{2}\) Condensed Matter Physics & Materials Science Department, Brookhaven National Laboratory, NY, 11973- 5000 USA \(^{3}\) Max- Planck- Institut für Festko'perforschung, D- 70569 Stuttgart, Germany \(^{4}\) University of Chinese Academy of Sciences, Department of Physics, Beijing 100190, China \(^{5}\) Institute of High Energy Physics, Chinese Academy of Science, Beijing 100049, China \(^{6}\) Songshan Lake Materials Laboratory, Dongguan, Guangdong 523808, China  
+
+Copper oxide superconductors have been continually fascinating the communities of condensed matter physics and material sciences because they host the highest ambient- pressure superconducting transition temperature \((T_{c})\) and mysterious physics \(^{1 - 3}\) . Searching for the universal correlation between the superconducting state and its normal state or neighboring ground state is believed to be an effective way for finding clues to elucidate the underlying mechanism of the superconductivity. One of the common pictures for the copper oxide superconductors is that a well- behaved metallic phase will present after the superconductivity is entirely suppressed by chemical doping \(^{4 - 8}\) or application of the magnetic field \(^{9}\) . Here, we report the first observation of universal quantum phase transition from superconducting state to insulating- like state under pressure in the under- , optimally- and over- doped \(\mathrm{Bi_2Sr_2CaCu_2O_8 + \delta}\) (Bi2212) superconductors with two \(\mathrm{CuO_2}\) planes in a unit cell. The same phenomenon has also been found in the \(\mathrm{Bi_2Sr_{1.63}La_{0.37}CuO_6 + \delta}\) (Bi2201) superconductor with one \(\mathrm{CuO_2}\) plane and the \(\mathrm{Bi_2.1Sr_{1.9}Ca_2Cu_3O_{10 + \delta}}\) (Bi2223) superconductor with three \(\mathrm{CuO_2}\) planes in a unit cell. These results not only provide fresh information on the cuprate superconductors but also pose a new challenge for achieving unified
+
+<--- Page Split --->
+
+## understandings on the mechanism of the high- \(T_{c}\) superconductivity.  
+
+Although a huge body of experimental investigations has been made for the copper oxide (cuprate) superconductors since they were discovered for more than thirty years \(^{10,11}\) , the correlation between the superconducting state and its normal state or the neighboring ground state is widely debated \(^{2,6,12 - 14}\) . By changing the chemical makeup of interleaved charge- reservoir layers, electrons can be added to or removed from the \(\mathrm{CuO_2}\) planes, resulting in the suppression of the antiferromagnetic insulating state of the parent compound \(^{2}\) . As the doping level reaches a critical one, superconductivity presents and its transition temperature \((T_{c})\) grows to a maximum upon doping to an optimal one, then declines for higher doping, and finally vanishes at a maximum doping level \(^{2,5,7,9,15}\) . \(T_{c - \mathrm{max}}\) are referred to as optimal- doped ones. It is important to recognize that once the superconducting state is completely suppressed by the chemical doping, the material undergoes a quantum phase transition from a superconducting state to a metallic state \(^{16 - 18}\) . However, the detailed experimental studies on the breakdown of the quantum state in cuprates are still lacking, which may be crucial for understanding how the superconducting state melts into or emerges from its neighboring ground states.  
+
+Pressure is an alternative method of tuning superconductivity beyond the chemical doping or external magnetic field, and it can provide significant information on the evolution among superconductivity, electronic state, and crystal structure without changing the chemical composition. On the other hand, it can also provide valuable assistance in the search for the superconductors with higher values of \(T_{c}\) at ambient pressure by the substitution of the smaller ions \(^{19}\) . A notably successful application of
+
+<--- Page Split --->
+
+this strategy leads to the discoveries of the important cuprate and Fe- based superconductors \(^{11,20,21}\) . Therefore, high- pressure studies on superconductivity can benefit not only for searching new superconductors but also for deeper understandings on the correlation between the superconducting and its neighbor normal or ground states \(^{22 - 26}\) . To reveal how the superconducting state or non- superconducting state develops, a central issue for understanding the high- \(T_{c}\) superconductivity in cuprates, we performed a series of high- pressure investigations by employing our newly developed state- of- the- art technique, a combined in- situ high- pressure measurements of the resistance and alternating current (ac) susceptibility for the same sample at the same pressure. The studied samples that have been investigated broadly by variety of methods \(^{26 - 31}\) are the under- doped (UD), optimally- doped (OP) and over- doped (OD) \(\mathrm{Bi}_{2}\mathrm{Sr}_{2}\mathrm{CaCu}_{2}\mathrm{O}_{8 + \delta}\) (Bi2212) superconductors with two \(\mathrm{CuO}_{2}\) planes in a unit cell.  
+
+Figure 1 shows the results of temperature versus in- plane resistance for the UD sample with \(T_{c} = 74 \mathrm{~K}\) (Fig.1a), the OP sample with \(T_{c} = 91 \mathrm{~K}\) (Fig.1b) and the OD sample with \(T_{c} = 82 \mathrm{~K}\) (Fig.1c) at different pressures. It is found that the onset \(T_{c}\) of these samples exhibits the same high- pressure behavior: a slight increase initially and then a monotonous decrease upon elevating pressure until not detectable. Subsequently, an unexpected insulating- like state presents at a pressure \((P_{i})\) of \(34.3 \mathrm{GPa}\) for the UD- doped sample, \(39.9 \mathrm{GPa}\) for the OP sample and \(42.2 \mathrm{GPa}\) for the OD sample, respectively. And the insulating- like behavior becomes pronounced when the pressure is higher than the \(P_{i}\) (Fig.1a- 1c). It is a grand surprise because naively one expects that by applying pressure the bandwidth should increase and thereby the system should
+
+<--- Page Split --->
+
+become more metallic, but instead it becomes insulating- like. We repeated the measurements on new samples and found the results are reproducible [see Supplementary Information].  
+
+To investigate whether the insulating behavior observed is due to pressure- induced cracks in the material, we performed the experiments for the over- doped Bi2212 sample, and found a reversible transition between the superconducting state and the insulating- like state (see Supplementary Information). These results rule out the possibility that the transition from a superconducting state to an insulating- like state is caused by cracks.  
+
+The combined high- pressure measurements of \(ac\) susceptibility and in- plane resistance were performed for the above three kinds of samples. As shown in Fig.2, the superconducting transitions of the samples detected by the \(ac\) susceptibility can be clearly identified by the onset signal of the deviation from the almost constant background on the high- temperature side (see the blue plots) and the plunge of the resistance to zero (see the red plots). Upon compression to 34.3 GPa for the UD sample, 39.9 GPa for the OP sample and 42.2 for the OD sample, the samples show an insulating- like behavior (see the red plots in Fig.2d, 2h and 2l) and no diamagnetic signal is captured by the \(ac\) susceptibility measurements (see the blue plots in Fig.2d, 2h and 2l). These results manifest that the pressure induces a quantum phase transition from a superconducting state to an insulating- like state in all these superconductors.  
+
+We summarize the experimental results in the normalized pressure- \(T_{c}\) phase diagram in the left panel of Fig.3, which is established on the basis of the pressure- \(T_{c}\) phase diagrams of the UD, OP and OD Bi2212 samples (the right panels of Fig.3). The
+
+<--- Page Split --->
+
+phase diagram for the three kinds of samples shows two distinct regions: the superconducting state (SC) and the insulating- like state (I), and demonstrates a universal quantum phase transition from the superconducting to the insulating- like states. It is seen that \(T_{c}\) displays a slight increase initially within a small pressure range, and then a continuous decrease with elevating pressure until fully suppressed at a critical pressure \((P_{c}\) , the determination of \(P_{c}\) can be found in the caption of Figure 3), above which an insulating- like state emerges, as shown in the left panel of Fig.3 (the detail of the normalizing analysis can be found in the Supplementary information).  
+
+In order to know whether the quantum phase transition discovered in this study is a common phenomenon beyond the Bi2212 superconductors investigated, we conducted the same measurements on the \(\mathrm{Bi}_{2}\mathrm{Sr}_{1.63}\mathrm{La}_{0.37}\mathrm{CuO}_{6 + \delta}\) (Bi2201) superconductor with one \(\mathrm{CuO}_{2}\) plane and the \(\mathrm{Bi}_{2}\mathrm{Sr}_{2}\mathrm{Ca}_{2}\mathrm{Cu}_{3}\mathrm{O}_{10 + \delta}\) (Bi2223) superconductor with three \(\mathrm{CuO}_{2}\) planes in a unit cell. The same phenomenon is also found in these superconductors (see Supplementary Information), indicating that the observed quantum phase transition is universal in these bismuth- bearing cuprate superconductors, regardless of the doping level and the number of \(\mathrm{CuO}_{2}\) planes in a unit cell.  
+
+These results impact our knowledge about the cuprate superconductors that, after the superconducting state is destroyed, the sample should show a well- behaved metallic state because pressure generally increases the bandwidth. To clarify the possible origin that leads to the destruction of the superconducting state and the emergence of the insulating- like state under pressure, we carried out more experiments.
+
+<--- Page Split --->
+
+First, we conducted the high- pressure synchrotron X- ray diffraction measurements at \(50\mathrm{K}\) for the OD sample on beamline 4W2 at the Beijing Synchrotron Radiation Facility. Our results indicated that there is no structural phase transition in the range of pressure up to \(43.1\mathrm{GPa}\) , except that the volume of the lattice is apparently compressed (see Supplementary Information). These results ruled out the possibility that the quantum phase transition from superconducting to insulating- like states connects with a pressure- induced structural phase transition.  
+
+Second, we measured the magnetoresistance \((MR)\) at \(4\mathrm{K}\) for the compressed UD, OP and OD samples that host the insulating- like state. The magnetic field was applied perpendicular to the \(ab\) - plane of these samples. As shown in Fig.4a- c, the \(MR\) of all the samples exhibits a positive effect, the in- plane resistance increases upon elevating magnetic field. Considering that the \(MR\) is very weak \((\sim 1\%)\) and the appearance of the insulating- like state is close to the superconducting- insulating transition, we presume that the origin of the positive \(MR\) may be related to the superconducting fluctuation.  
+
+Third, we performed the high- pressure Hall coefficient \((R_H)\) measurements for the OD sample (Fig.4d) and find that \(R_H(P)\) decreases remarkably with increasing pressure up to \(\sim 18\) GPa. Because the Hall resistance versus magnetic field displays a linear behavior in the pressure range investigated (Supplementary Information), a typical feature of the single band, the decrease of \(R_H(P)\) below \(18\) GPa ought to be associated with the enhancement of carrier density. However, \(R_H\) remains almost unchanged for pressures ranging from \(\sim 18\) GPa to \(\sim 35\) GPa and then shows a slow increase from \(\sim 35\) GPa to \(48.3\) GPa. No apparent change in \(R_H(P)\) at \(P_c = 39.5\) GPa implies that the total
+
+<--- Page Split --->
+
+density of charge carriers seems to remain in a steady state across the quantum criticality. The reproducible result is also obtained in the Bi2201 superconductor (see Supplemental Information).  
+
+It is noted that, unlike the usual insulator, the low- temperature resistance in the insulating- like state rises way too slowly to be exponential. We attempted to fit the low temperature resistance with exponential dependence and power law, but they fail [see Supplementary Information]. Slow rises of the kind have been found in the low temperature orthorhombic \(\mathrm{La}_{2 - x}\mathrm{Sr}_x\mathrm{CuO}_4\) , \(\mathrm{Y}\mathrm{Bi}\mathrm{Ba}_2\mathrm{Cu}_3\mathrm{O}_{7 - \delta}\) cuprates and \(\mathrm{La}_{1 - x}\mathrm{M}_x\mathrm{OBiS}_2\) , which are perceived as quite mysterious \(^{32 - 35}\) .  
+
+There is in fact no precedence anywhere else for such a transition from a superconducting state to an insulating- like state without a coincident structural phase transition. Therefore, some questions are raised naturally: why do the itinerant superconducting electrons become localized after the quantum phase transition, and what is the exotic pathway that results in the quantum phase transition? If considering the buckling of the \(\mathrm{CuO}_2\) planes due to the asymmetric lattice structure around the planes and nonuniform deformation from the doped atoms, when pressure shrinks the lattice constant of the planes, the plane buckling should be enhanced, which may help to develop an exotic band structure with a reduction of the bandwidth \(^{36}\) and eventually result in the emergence of the insulating- like state. All the above are the attractive issues in searching for the new physics behind the pressure- induced quantum phase transition from a superconducting state to insulating- like state instead of to a metallic state, which deserves further investigation with other advanced experimental probes and
+
+<--- Page Split --->
+
+sophisticated theoretical studies.  
+
+## Acknowledgements  
+
+We thank Prof. Jan Zaanen and Prof. Yi Zhou for helpful discussions on this work. The work in China was supported by the National Key Research and Development Program of China (Grant No. 2017YFA0302900, 2016YFA0300300 and 2017YFA0303103), the NSF of China (Grants No. U2032214, 11888101 and 12004419) and the Strategic Priority Research Program (B) of the Chinese Academy of Sciences (Grant No. XDB25000000). We thank the support from the Users with Excellence Program of Hefei Science Center CAS (2020HSC- UE015). Part of the work is supported by the Synergic Extreme Condition User System. J. G. is grateful for support from the Youth Innovation Promotion Association of the CAS (2019008). The work in BNL was supported by the US Department of Energy, office of Basic Energy Sciences (contract No. desc0012704).  
+
+## Author contributions  
+
+L.S., T.X. and Q.W. designed the study and supervised the project. G. G. grew the \(\mathrm{Bi}_2\mathrm{Sr}_2\mathrm{CaCu}_2\mathrm{O}_{8 + \delta}\) single crystals. H.T.Y and X.J.Z. grew the \(\mathrm{Bi}_2\mathrm{Sr}_{1.63}\mathrm{La}_{0.37}\mathrm{CuO}_{6 + \delta}\) single crystals. C.T. L. grew the \(\mathrm{Bi}_2\mathrm{Sr}_2\mathrm{Ca}_2\mathrm{Cu}_3\mathrm{O}_{10 + \delta}\) single crystals. Y.Z., J.G., S.C., and L.S. performed the high pressure resistance, \(ac\) susceptibility and magnetoresistance measurements. J.G., Y.Z., Y.G., Y.C.L., X.D.L. and C.L.Y. performed the high pressure X- ray diffraction measurements. J.G, C.H. and S.J.L. performed the high
+
+<--- Page Split --->
+
+pressure Hall measurements. L.S., T.X., Q.W., J.P.H. and Y.Z. wrote the manuscript in consultation with all authors.  
+
+## Author information  
+
+The authors declare no competing financial interest. Correspondence and requests for materials should be addressed to L.S. (llsun@iphv.ac.cn). These authors with star (\*) contributed equally to this work.  
+
+## References  
+
+1 Chu, C. W., Deng, L. Z. & Lv, B. Hole-doped cuprate high temperature superconductors. Physica C Supercond 514, 290- 313 (2015).  2 Keimer, B., Kivelson, S. A., Norman, M. R., Uchida, S. & Zaanen, J. From quantum matter to high- temperature superconductivity in copper oxides. Nature 518, 179- 186 (2015).  3 Zaanen, J. Superconducting electrons go missing. Nature 536, 282- 283 (2016).  4 Norman, M. R. & Pepin, C. The electronic nature of high temperature cuprate superconductors. Rep. Prog. Phys. 66, 1547- 1610 (2003).  5 Barišić, N. et al. Universal sheet resistance and revised phase diagram of the cuprate high- temperature superconductors. Proc. Natl. Acad. Sci. U.S.A. 110, 12235- 12240 (2013).  6 Božović, I., He, X., Wu, J. & Bollinger, A. T. Dependence of the critical temperature in overdoped copper oxides on superfluid density. Nature 536, 309-
+
+<--- Page Split --->
+
+311 (2016).  
+
+Greene, R. L., Mandal, P. R., Poniatowski, N. R. & Sarkar, T. The Strange Metal State of the Electron- Doped Cuprates. Annu. Rev. Condens. Matter Phys. 11, 213- 229 (2020).  
+
+Sacépé, B., Feigel'man, M. & Klapwijk, T. M. Quantum breakdown of superconductivity in low- dimensional materials. Nat. Phys. 16, 734- 746 (2020).  
+
+Proust, C. & Taillefer, L. The Remarkable Underlying Ground States of Cuprate Superconductors. Annu. Rev. Condens. Matter Phys. 10, 409- 429 (2019).  
+
+Bednorz, J. G. & Müller, K. A. Possible high \(T_{c}\) superconductivity in the Ba- La- Cu- O system. Z Phys B Con Mat 64, 189- 193 (1986).  
+
+Wu, M. K. et al. Superconductivity at 93 K in a new mixed- phase Y- Ba- Cu- O compound system at ambient pressure. Phys. Rev. Lett. 58, 908- 910 (1987).  
+
+Zaanen, J. et al. Towards a complete theory of high \(T_{c}\) . Nat. Phys. 2, 138- 143 (2006).  
+
+Lee, P. A., Nagaosa, N. & Wen, X.- G. Doping a Mott insulator: Physics of high- temperature superconductivity. Rev. Mod. Phys. 78, 17- 85 (2006).  
+
+Zaanen, J. Planckian dissipation, minimal viscosity and the transport in cuprate strange metals. Scipost Phys 6, 061 (2019).  
+
+Nakamae, S. et al. Electronic ground state of heavily overdoped nonsuperconducting La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub>. Phys. Rev. B 68, 100502 (2003).  
+
+Hussey, N. E., Abdel- Jawad, M., Carrington, A., Mackenzie, A. P. & Balicas, L. A coherent three- dimensional Fermi surface in a high- transition- temperature
+
+<--- Page Split --->
+
+superconductor. Nature 425, 814- 817 (2003).  
+
+17 Plate, M. et al. Fermi Surface and Quasiparticle Excitations of Overdoped Tl₂Ba₂CuO₆+δ. Phys. Rev. Lett. 95, 077001 (2005).  
+
+18 Vignolle, B. et al. Quantum oscillations in an overdoped high- Tc superconductor. Nature 455, 952- 955 (2008).  
+
+19 Chu, C. W. et al. Evidence for superconductivity above 40 K in the La- Ba- Cu- O compound system. Phys. Rev. Lett. 58, 405- 407 (1987).  
+
+20 Takahashi, H. et al. Superconductivity at 43 K in an iron- based layered compound LaO₁- xFxFeAs. Nature 453, 376- 378 (2008).  
+
+21 Zhi- An, R. et al. Superconductivity at 55 K in Iron- Based F- Doped Layered Quaternary Compound Sm[O₁- xFx]FeAs. Chinese Phys. Lett. 25, 2215- 2216 (2008).  
+
+22 S. Schilling, J. in Frontiers of High Pressure Research II: Application of High Pressure to Low- Dimensional Novel Electronic Materials (eds Hans D. Hochheimer, Bogdan Kuchta, Peter K. Dorhout, & Jeffery L. Yarger) 345- 360 (Springer Netherlands, 2001).  
+
+23 Sun, L. et al. Re- emerging superconductivity at 48 kelvin in iron chalcogenides. Nature 483, 67 (2012).  
+
+24 Deng, L. et al. Higher superconducting transition temperature by breaking the universal pressure relation. Proc. Natl. Acad. Sci. U.S.A. 116, 2004- 2008 (2019).  
+
+25 Guo, J. et al. Crossover from two- dimensional to three- dimensional superconducting states in bismuth- based cuprate superconductor. Nat. Phys. 16,
+
+<--- Page Split --->
+
+295- 300 (2020).  
+
+26 Parker, C. V. et al. Fluctuating stripes at the onset of the pseudogap in the high- \(T_{c}\) superconductor \(\mathrm{Bi}_{2}\mathrm{Sr}_{2}\mathrm{CaCu}_{2}\mathrm{O}_{8 + x}\) . Nature 468, 677- 680 (2010).  
+
+27 Kondo, T. et al. Disentangling Cooper- pair formation above the transition temperature from the pseudogap state in the cuprates. Nat. Phys. 7, 21- 25 (2011).  
+
+28 Zhao, H. et al. Charge- stripe crystal phase in an insulating cuprate. Nat. Mater. 18, 103- 107 (2019).  
+
+29 Du, Z. et al. Imaging the energy gap modulations of the cuprate pair- density- wave state. Nature 580, 65- 70 (2020).  
+
+30 Gao, Q. et al. Electronic Evolution from the Parent Mott Insulator to a Superconductor in Lightly Hole- Doped \(\mathrm{Bi}_{2}\mathrm{Sr}_{2}\mathrm{CaCu}_{2}\mathrm{O}_{8 + \delta}\) . Chinese Phys. Lett. 37, 087402 (2020).  
+
+31 Adachi, T. et al. Magnetic- field effects on the charge- spin stripe order in La- 214 high- \(T_{C}\) cuprates. J. Phys. Conf. Ser. 51, 259- 262 (2006).  
+
+32 Muramatsu, T., Pham, D. & Chu, C. W. A possible pressure- induced superconducting- semiconducting transition in nearly optimally doped single crystalline \(\mathrm{YBa}_{2}\mathrm{Cu}_{3}\mathrm{O}_{7 - \delta}\) . Appl. Phys. Lett. 99, 052508 (2011).  
+
+33 Fang, Y., Yazici, D., Jeon, I. & Maple, M. B. High- pressure effects on nonfluorinated \(\mathrm{BiS}_{2}\) - based superconductors \(\mathrm{La}_{1 - x}\mathrm{M}_{x}\mathrm{OBiS}_{2}\) ( \(M = \mathrm{Ti}\) and Th). Phys. Rev. B 96, 214505 (2017).  
+
+34 Moritomo, Y., Kuwahara, H. & Tokura, Y. Bandwidth- and Doping- Dependent Pressure Effects on the Ferromagnetic Transition in Perovskite Manganites. J.
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Figure 1 Temperature dependence of in-plane resistance for Bi₂Sr₂CaCu₂O₈+δ </center>  
+
+(Bi2212) at different pressures: (a) for the under- doped (UD) superconductor with superconducting transition temperature \((T_{c})\) about \(74\mathrm{K}\) in the pressure range of \(1.5\mathrm{GPa}\) - \(36.2\mathrm{GPa}\) ; (b) for the optimally- doped (OP) sample with \(T_{c}\) about \(91\mathrm{K}\) in the pressure range of \(0.7\mathrm{GPa} - 41.2\mathrm{GPa}\) ; (c) for the over- doped (OD) sample with \(T_{c}\) about \(82\mathrm{K}\) in the pressure range of \(1.0\mathrm{GPa} - 49\mathrm{GPa}\) . The three kinds of samples display the same behavior of an insulating- like state above the pressure \((P_{i})\) .
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Figure 2. In-plane resistance \((R)\) and \(ac\) susceptibility \((\Delta \chi^{\prime})\) as a function of </center>  
+
+temperature \((T)\) for the \(\mathrm{Bi_2Sr_2CaCu_2O_{8 + \delta}}\) superconductors at different pressures: (a)-(d) for the under- doped (UD) superconductor; (e)-(h) for the optimally- doped (OP) superconductor; (i)-(l) for the over- doped (OD) superconductor. The blue lines in the figures are the data of \(\Delta \chi^{\prime}(T)\) , while the red lines are the data of \(R(T)\) . The red and blue arrows indicate the temperatures of the onset superconducting transition detected by resistance and \(ac\) susceptibility measurements, respectively.
+
+<--- Page Split --->
+![](images/Figure_3.jpg)
+
+<center>Figure 3 Pressure- \(T_{c}\) phase diagrams for \(\mathrm{Bi_2Sr_2CaCu_2O_{8 + \delta}}\) superconductors. The Right panels are the phase diagrams established by the experimental results from the under-doped (UD), optimally-doped (OP) and over-doped (OD) samples, together with the mapping information of temperature and pressure dependent \(R\) (shown in color scale). The left panel is a normalized phase diagram that is built on the basis of the experimental phase diagrams (the right panels). \(P_{Tc - max}\) and \(P_{c}\) stand for the critical pressures where \(T_{c}\) reaches the maximum and the zero, respectively. In the normalizing analysis, we define the pressure as \(P_{Tc - max}\) when \((P - P_{Tc - max}) / (P_{c} - P_{Tc - max}) = 0\) , and the pressure as \(P_{c}\) when \((P - P_{Tc - max}) / (P_{c} - P_{Tc - max}) = 1\) . The results of the normalizing analysis for \(T_{c} / T_{c - max}\) versus \((P - P_{Tc - max}) / (P_{c} - P_{Tc - max})\) and \(T_{ins} / T_{c - max}\) versus \((P - P_{Tc - max}) / (P_{c} - P_{Tc - max})\) show that the three kinds of samples display a universal quantum phase transition from the superconducting state to an insulating-like state. SC, M and I stand for superconducting state, metallic state and insulating-like state, respectively. The region of the M phase is determined by the critical value of \(R / R_{290K}\) where the quantum phase </center>
+
+<--- Page Split --->
+
+transition occurs. For example, when \(R / R_{290K}\) is greater than 3, the over- doped sample is in the insulating- like state while, \(R / R_{290K}\) is less than 3, the sample is in the metallic state (see right bottom panel). \(T_{c - R - onset}\) and \(T_{c - \Delta \chi ' - onset}\) denote the onset temperatures of the superconducting transition detected by the resistance and \(ac\) susceptibility measurements, respectively. \(T_{c - max}\) and \(T_{ins}\) are the maximum value of \(T_{c}\) and the characteristic temperature of the insulating- like transition (the method of determining the \(T_{ins}\) can be found in the Supplementary information), respectively. The \(P_{c}\) value is determined by the average pressure of two experimental runs \([P_{c} = (P_{c - run1} + P_{c - run2}) / 2]\) , in which \(P_{c}\) of each experimental run is determined by the highest experimental pressure where the superconducting transition can still be observed and the lowest experimental pressure where the insulating- like state appears. The error bar of \(P_{c}\) is the difference between \(P_{c - run1}\) and \(P_{c - run2}\) .
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 4 Magnetoresistance \((MR)\) as a function of magnetic field \((B)\) for the UD, OP and OD \(\mathrm{Bi_2Sr_2CaCu_2O_{8 + \delta}}\) superconductors when they enter into an insulating-like state, and Hall coefficient information of the OP sample under pressure. (a)- (b) Plots of \(MR\) versus \(B\) for the UD, OP and OD samples measured at \(4\mathrm{K}\) at \(36.2\mathrm{GPa}\) , 41.2 GPa and 49 GPa, respectively. It is seen that all of them exhibit a positive magnetic effect. The red and beige arrows represent the directions of increasing and decreasing magnetic field. (d) Pressure dependence of Hall coefficient \((R_H)\) for the OD superconductor measured at \(100\mathrm{K}\) . </center>
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+- Slf.docx
+
+<--- Page Split --->

preprint/preprint__491c12817894379960e50a02a9c4da0c41dd179e729ff4fd51850bf7dd1a6580/preprint__491c12817894379960e50a02a9c4da0c41dd179e729ff4fd51850bf7dd1a6580_det.mmd

@@ -0,0 +1,324 @@
+<|ref|>title<|/ref|><|det|>[[44, 107, 895, 208]]<|/det|>
+# Universal quantum phase transition from superconducting to insulating-like states in pressurized Bi2Sr2CaCu2O8+δ superconductors  
+
+<|ref|>text<|/ref|><|det|>[[44, 228, 855, 270]]<|/det|>
+Liling Sun (llsun@iphy.ac.cn) Institute of Physics, Chinese Academy of Sciences https://orcid.org/0000- 0002- 6653- 7826  
+
+<|ref|>text<|/ref|><|det|>[[44, 275, 495, 317]]<|/det|>
+Yazhou Zhou Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 324, 855, 365]]<|/det|>
+Jing Guo Institute of Physics, Chinese Academy of Sciences https://orcid.org/0000- 0002- 2164- 5065  
+
+<|ref|>text<|/ref|><|det|>[[44, 370, 495, 412]]<|/det|>
+Shu Cai Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 418, 495, 459]]<|/det|>
+Jinyu Zhao Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 465, 340, 506]]<|/det|>
+Genda Gu Brookhaven National Laboratory  
+
+<|ref|>text<|/ref|><|det|>[[44, 512, 357, 552]]<|/det|>
+Chengtian Lin Max- Planck- Inst of Solid State Res  
+
+<|ref|>text<|/ref|><|det|>[[44, 558, 495, 599]]<|/det|>
+Hongtao Yan Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 604, 495, 645]]<|/det|>
+Cheng Huang Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 651, 495, 692]]<|/det|>
+Chongli Yang Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 697, 495, 738]]<|/det|>
+Sijin Long Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 744, 495, 785]]<|/det|>
+Yu Gong Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 790, 691, 831]]<|/det|>
+Yanchun Li Institute of High Energy Physics https://orcid.org/0000- 0003- 4540- 4546  
+
+<|ref|>text<|/ref|><|det|>[[44, 836, 691, 877]]<|/det|>
+Xiaodong Li Institute of High Energy Physics https://orcid.org/0000- 0002- 2290- 1198  
+
+<|ref|>text<|/ref|><|det|>[[44, 882, 495, 923]]<|/det|>
+Qi Wu Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 928, 163, 947]]<|/det|>
+Jiangping Hu
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[52, 46, 494, 66]]<|/det|>
+Institute of Physics, Chinese Academy of Sciences  
+
+<|ref|>text<|/ref|><|det|>[[44, 70, 802, 135]]<|/det|>
+Xingjiang Zhou National Lab for Superconductivity, Institute of Physics, Chinese Academy of Sciences https://orcid.org/0000- 0002- 5261- 1386  
+
+<|ref|>text<|/ref|><|det|>[[44, 140, 933, 204]]<|/det|>
+Tao Xiang Institute of Physics, Chinese Academy of Sciences, Beijing National Laboratory for Condensed Matter Physics  
+
+<|ref|>sub_title<|/ref|><|det|>[[44, 245, 102, 262]]<|/det|>
+## Article  
+
+<|ref|>text<|/ref|><|det|>[[44, 281, 821, 303]]<|/det|>
+Keywords: copper oxide, superconductions, Bi2Sr2CaCu2O8+δ, high- Tc superconductivity  
+
+<|ref|>text<|/ref|><|det|>[[44, 320, 290, 340]]<|/det|>
+Posted Date: June 9th, 2021  
+
+<|ref|>text<|/ref|><|det|>[[44, 358, 463, 378]]<|/det|>
+DOI: https://doi.org/10.21203/rs.3.rs- 551119/v1  
+
+<|ref|>text<|/ref|><|det|>[[44, 395, 910, 439]]<|/det|>
+License: © © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+<|ref|>text<|/ref|><|det|>[[42, 472, 955, 517]]<|/det|>
+Version of Record: A version of this preprint was published at Nature Physics on February 17th, 2022. See the published version at https://doi.org/10.1038/s41567- 022- 01513- 2.
+
+<--- Page Split --->
+<|ref|>title<|/ref|><|det|>[[189, 93, 808, 150]]<|/det|>
+# Universal quantum phase transition from superconducting to insulating-like states in pressurized \(\mathrm{Bi_2Sr_2CaCu_2O_8 + \delta}\)  
+
+<|ref|>sub_title<|/ref|><|det|>[[413, 168, 583, 187]]<|/det|>
+## superconductors  
+
+<|ref|>text<|/ref|><|det|>[[166, 215, 833, 277]]<|/det|>
+Yazhou Zhou \(^{1*}\) , Jing Guo \(^{1,6*}\) , Shu Cai \(^{1*}\) , Jinyu Zhao \(^{1,4}\) , Genda Gu \(^{2}\) , Chengtian Lin \(^{3}\) , Hongtao Yan \(^{1,4}\) , Cheng Huang \(^{1,4}\) , Chongli Yang \(^{1}\) , Sijin Long \(^{1,4}\) , Yu Gong \(^{5}\) , Yanchun Li \(^{5}\) , Xiaodong Li \(^{5}\) , Qi Wu \(^{1}\) , Jiangping Hu \(^{1,4}\) , Xingjiang Zhou \(^{1,4,6}\) , Tao Xiang \(^{1,4}\) and Liling Sun \(^{1,4,6}\)  
+
+<|ref|>text<|/ref|><|det|>[[150, 298, 849, 444]]<|/det|>
+\(^{1}\) Institute of Physics, National Laboratory for Condensed Matter Physics, Chinese Academy of Sciences, Beijing, 100190, China \(^{2}\) Condensed Matter Physics & Materials Science Department, Brookhaven National Laboratory, NY, 11973- 5000 USA \(^{3}\) Max- Planck- Institut für Festko'perforschung, D- 70569 Stuttgart, Germany \(^{4}\) University of Chinese Academy of Sciences, Department of Physics, Beijing 100190, China \(^{5}\) Institute of High Energy Physics, Chinese Academy of Science, Beijing 100049, China \(^{6}\) Songshan Lake Materials Laboratory, Dongguan, Guangdong 523808, China  
+
+<|ref|>text<|/ref|><|det|>[[147, 470, 852, 909]]<|/det|>
+Copper oxide superconductors have been continually fascinating the communities of condensed matter physics and material sciences because they host the highest ambient- pressure superconducting transition temperature \((T_{c})\) and mysterious physics \(^{1 - 3}\) . Searching for the universal correlation between the superconducting state and its normal state or neighboring ground state is believed to be an effective way for finding clues to elucidate the underlying mechanism of the superconductivity. One of the common pictures for the copper oxide superconductors is that a well- behaved metallic phase will present after the superconductivity is entirely suppressed by chemical doping \(^{4 - 8}\) or application of the magnetic field \(^{9}\) . Here, we report the first observation of universal quantum phase transition from superconducting state to insulating- like state under pressure in the under- , optimally- and over- doped \(\mathrm{Bi_2Sr_2CaCu_2O_8 + \delta}\) (Bi2212) superconductors with two \(\mathrm{CuO_2}\) planes in a unit cell. The same phenomenon has also been found in the \(\mathrm{Bi_2Sr_{1.63}La_{0.37}CuO_6 + \delta}\) (Bi2201) superconductor with one \(\mathrm{CuO_2}\) plane and the \(\mathrm{Bi_2.1Sr_{1.9}Ca_2Cu_3O_{10 + \delta}}\) (Bi2223) superconductor with three \(\mathrm{CuO_2}\) planes in a unit cell. These results not only provide fresh information on the cuprate superconductors but also pose a new challenge for achieving unified
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[148, 91, 730, 110]]<|/det|>
+## understandings on the mechanism of the high- \(T_{c}\) superconductivity.  
+
+<|ref|>text<|/ref|><|det|>[[146, 120, 853, 660]]<|/det|>
+Although a huge body of experimental investigations has been made for the copper oxide (cuprate) superconductors since they were discovered for more than thirty years \(^{10,11}\) , the correlation between the superconducting state and its normal state or the neighboring ground state is widely debated \(^{2,6,12 - 14}\) . By changing the chemical makeup of interleaved charge- reservoir layers, electrons can be added to or removed from the \(\mathrm{CuO_2}\) planes, resulting in the suppression of the antiferromagnetic insulating state of the parent compound \(^{2}\) . As the doping level reaches a critical one, superconductivity presents and its transition temperature \((T_{c})\) grows to a maximum upon doping to an optimal one, then declines for higher doping, and finally vanishes at a maximum doping level \(^{2,5,7,9,15}\) . \(T_{c - \mathrm{max}}\) are referred to as optimal- doped ones. It is important to recognize that once the superconducting state is completely suppressed by the chemical doping, the material undergoes a quantum phase transition from a superconducting state to a metallic state \(^{16 - 18}\) . However, the detailed experimental studies on the breakdown of the quantum state in cuprates are still lacking, which may be crucial for understanding how the superconducting state melts into or emerges from its neighboring ground states.  
+
+<|ref|>text<|/ref|><|det|>[[147, 675, 853, 880]]<|/det|>
+Pressure is an alternative method of tuning superconductivity beyond the chemical doping or external magnetic field, and it can provide significant information on the evolution among superconductivity, electronic state, and crystal structure without changing the chemical composition. On the other hand, it can also provide valuable assistance in the search for the superconductors with higher values of \(T_{c}\) at ambient pressure by the substitution of the smaller ions \(^{19}\) . A notably successful application of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 94, 853, 523]]<|/det|>
+this strategy leads to the discoveries of the important cuprate and Fe- based superconductors \(^{11,20,21}\) . Therefore, high- pressure studies on superconductivity can benefit not only for searching new superconductors but also for deeper understandings on the correlation between the superconducting and its neighbor normal or ground states \(^{22 - 26}\) . To reveal how the superconducting state or non- superconducting state develops, a central issue for understanding the high- \(T_{c}\) superconductivity in cuprates, we performed a series of high- pressure investigations by employing our newly developed state- of- the- art technique, a combined in- situ high- pressure measurements of the resistance and alternating current (ac) susceptibility for the same sample at the same pressure. The studied samples that have been investigated broadly by variety of methods \(^{26 - 31}\) are the under- doped (UD), optimally- doped (OP) and over- doped (OD) \(\mathrm{Bi}_{2}\mathrm{Sr}_{2}\mathrm{CaCu}_{2}\mathrm{O}_{8 + \delta}\) (Bi2212) superconductors with two \(\mathrm{CuO}_{2}\) planes in a unit cell.  
+
+<|ref|>text<|/ref|><|det|>[[147, 539, 853, 891]]<|/det|>
+Figure 1 shows the results of temperature versus in- plane resistance for the UD sample with \(T_{c} = 74 \mathrm{~K}\) (Fig.1a), the OP sample with \(T_{c} = 91 \mathrm{~K}\) (Fig.1b) and the OD sample with \(T_{c} = 82 \mathrm{~K}\) (Fig.1c) at different pressures. It is found that the onset \(T_{c}\) of these samples exhibits the same high- pressure behavior: a slight increase initially and then a monotonous decrease upon elevating pressure until not detectable. Subsequently, an unexpected insulating- like state presents at a pressure \((P_{i})\) of \(34.3 \mathrm{GPa}\) for the UD- doped sample, \(39.9 \mathrm{GPa}\) for the OP sample and \(42.2 \mathrm{GPa}\) for the OD sample, respectively. And the insulating- like behavior becomes pronounced when the pressure is higher than the \(P_{i}\) (Fig.1a- 1c). It is a grand surprise because naively one expects that by applying pressure the bandwidth should increase and thereby the system should
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 94, 851, 187]]<|/det|>
+become more metallic, but instead it becomes insulating- like. We repeated the measurements on new samples and found the results are reproducible [see Supplementary Information].  
+
+<|ref|>text<|/ref|><|det|>[[147, 205, 855, 374]]<|/det|>
+To investigate whether the insulating behavior observed is due to pressure- induced cracks in the material, we performed the experiments for the over- doped Bi2212 sample, and found a reversible transition between the superconducting state and the insulating- like state (see Supplementary Information). These results rule out the possibility that the transition from a superconducting state to an insulating- like state is caused by cracks.  
+
+<|ref|>text<|/ref|><|det|>[[147, 391, 853, 781]]<|/det|>
+The combined high- pressure measurements of \(ac\) susceptibility and in- plane resistance were performed for the above three kinds of samples. As shown in Fig.2, the superconducting transitions of the samples detected by the \(ac\) susceptibility can be clearly identified by the onset signal of the deviation from the almost constant background on the high- temperature side (see the blue plots) and the plunge of the resistance to zero (see the red plots). Upon compression to 34.3 GPa for the UD sample, 39.9 GPa for the OP sample and 42.2 for the OD sample, the samples show an insulating- like behavior (see the red plots in Fig.2d, 2h and 2l) and no diamagnetic signal is captured by the \(ac\) susceptibility measurements (see the blue plots in Fig.2d, 2h and 2l). These results manifest that the pressure induces a quantum phase transition from a superconducting state to an insulating- like state in all these superconductors.  
+
+<|ref|>text<|/ref|><|det|>[[147, 798, 851, 891]]<|/det|>
+We summarize the experimental results in the normalized pressure- \(T_{c}\) phase diagram in the left panel of Fig.3, which is established on the basis of the pressure- \(T_{c}\) phase diagrams of the UD, OP and OD Bi2212 samples (the right panels of Fig.3). The
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 93, 852, 373]]<|/det|>
+phase diagram for the three kinds of samples shows two distinct regions: the superconducting state (SC) and the insulating- like state (I), and demonstrates a universal quantum phase transition from the superconducting to the insulating- like states. It is seen that \(T_{c}\) displays a slight increase initially within a small pressure range, and then a continuous decrease with elevating pressure until fully suppressed at a critical pressure \((P_{c}\) , the determination of \(P_{c}\) can be found in the caption of Figure 3), above which an insulating- like state emerges, as shown in the left panel of Fig.3 (the detail of the normalizing analysis can be found in the Supplementary information).  
+
+<|ref|>text<|/ref|><|det|>[[147, 389, 852, 706]]<|/det|>
+In order to know whether the quantum phase transition discovered in this study is a common phenomenon beyond the Bi2212 superconductors investigated, we conducted the same measurements on the \(\mathrm{Bi}_{2}\mathrm{Sr}_{1.63}\mathrm{La}_{0.37}\mathrm{CuO}_{6 + \delta}\) (Bi2201) superconductor with one \(\mathrm{CuO}_{2}\) plane and the \(\mathrm{Bi}_{2}\mathrm{Sr}_{2}\mathrm{Ca}_{2}\mathrm{Cu}_{3}\mathrm{O}_{10 + \delta}\) (Bi2223) superconductor with three \(\mathrm{CuO}_{2}\) planes in a unit cell. The same phenomenon is also found in these superconductors (see Supplementary Information), indicating that the observed quantum phase transition is universal in these bismuth- bearing cuprate superconductors, regardless of the doping level and the number of \(\mathrm{CuO}_{2}\) planes in a unit cell.  
+
+<|ref|>text<|/ref|><|det|>[[147, 724, 852, 890]]<|/det|>
+These results impact our knowledge about the cuprate superconductors that, after the superconducting state is destroyed, the sample should show a well- behaved metallic state because pressure generally increases the bandwidth. To clarify the possible origin that leads to the destruction of the superconducting state and the emergence of the insulating- like state under pressure, we carried out more experiments.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 93, 853, 336]]<|/det|>
+First, we conducted the high- pressure synchrotron X- ray diffraction measurements at \(50\mathrm{K}\) for the OD sample on beamline 4W2 at the Beijing Synchrotron Radiation Facility. Our results indicated that there is no structural phase transition in the range of pressure up to \(43.1\mathrm{GPa}\) , except that the volume of the lattice is apparently compressed (see Supplementary Information). These results ruled out the possibility that the quantum phase transition from superconducting to insulating- like states connects with a pressure- induced structural phase transition.  
+
+<|ref|>text<|/ref|><|det|>[[147, 353, 852, 595]]<|/det|>
+Second, we measured the magnetoresistance \((MR)\) at \(4\mathrm{K}\) for the compressed UD, OP and OD samples that host the insulating- like state. The magnetic field was applied perpendicular to the \(ab\) - plane of these samples. As shown in Fig.4a- c, the \(MR\) of all the samples exhibits a positive effect, the in- plane resistance increases upon elevating magnetic field. Considering that the \(MR\) is very weak \((\sim 1\%)\) and the appearance of the insulating- like state is close to the superconducting- insulating transition, we presume that the origin of the positive \(MR\) may be related to the superconducting fluctuation.  
+
+<|ref|>text<|/ref|><|det|>[[147, 612, 852, 891]]<|/det|>
+Third, we performed the high- pressure Hall coefficient \((R_H)\) measurements for the OD sample (Fig.4d) and find that \(R_H(P)\) decreases remarkably with increasing pressure up to \(\sim 18\) GPa. Because the Hall resistance versus magnetic field displays a linear behavior in the pressure range investigated (Supplementary Information), a typical feature of the single band, the decrease of \(R_H(P)\) below \(18\) GPa ought to be associated with the enhancement of carrier density. However, \(R_H\) remains almost unchanged for pressures ranging from \(\sim 18\) GPa to \(\sim 35\) GPa and then shows a slow increase from \(\sim 35\) GPa to \(48.3\) GPa. No apparent change in \(R_H(P)\) at \(P_c = 39.5\) GPa implies that the total
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 94, 852, 187]]<|/det|>
+density of charge carriers seems to remain in a steady state across the quantum criticality. The reproducible result is also obtained in the Bi2201 superconductor (see Supplemental Information).  
+
+<|ref|>text<|/ref|><|det|>[[147, 205, 852, 410]]<|/det|>
+It is noted that, unlike the usual insulator, the low- temperature resistance in the insulating- like state rises way too slowly to be exponential. We attempted to fit the low temperature resistance with exponential dependence and power law, but they fail [see Supplementary Information]. Slow rises of the kind have been found in the low temperature orthorhombic \(\mathrm{La}_{2 - x}\mathrm{Sr}_x\mathrm{CuO}_4\) , \(\mathrm{Y}\mathrm{Bi}\mathrm{Ba}_2\mathrm{Cu}_3\mathrm{O}_{7 - \delta}\) cuprates and \(\mathrm{La}_{1 - x}\mathrm{M}_x\mathrm{OBiS}_2\) , which are perceived as quite mysterious \(^{32 - 35}\) .  
+
+<|ref|>text<|/ref|><|det|>[[147, 428, 852, 892]]<|/det|>
+There is in fact no precedence anywhere else for such a transition from a superconducting state to an insulating- like state without a coincident structural phase transition. Therefore, some questions are raised naturally: why do the itinerant superconducting electrons become localized after the quantum phase transition, and what is the exotic pathway that results in the quantum phase transition? If considering the buckling of the \(\mathrm{CuO}_2\) planes due to the asymmetric lattice structure around the planes and nonuniform deformation from the doped atoms, when pressure shrinks the lattice constant of the planes, the plane buckling should be enhanced, which may help to develop an exotic band structure with a reduction of the bandwidth \(^{36}\) and eventually result in the emergence of the insulating- like state. All the above are the attractive issues in searching for the new physics behind the pressure- induced quantum phase transition from a superconducting state to insulating- like state instead of to a metallic state, which deserves further investigation with other advanced experimental probes and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 95, 412, 111]]<|/det|>
+sophisticated theoretical studies.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 169, 318, 186]]<|/det|>
+## Acknowledgements  
+
+<|ref|>text<|/ref|><|det|>[[146, 205, 853, 594]]<|/det|>
+We thank Prof. Jan Zaanen and Prof. Yi Zhou for helpful discussions on this work. The work in China was supported by the National Key Research and Development Program of China (Grant No. 2017YFA0302900, 2016YFA0300300 and 2017YFA0303103), the NSF of China (Grants No. U2032214, 11888101 and 12004419) and the Strategic Priority Research Program (B) of the Chinese Academy of Sciences (Grant No. XDB25000000). We thank the support from the Users with Excellence Program of Hefei Science Center CAS (2020HSC- UE015). Part of the work is supported by the Synergic Extreme Condition User System. J. G. is grateful for support from the Youth Innovation Promotion Association of the CAS (2019008). The work in BNL was supported by the US Department of Energy, office of Basic Energy Sciences (contract No. desc0012704).  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 650, 334, 667]]<|/det|>
+## Author contributions  
+
+<|ref|>text<|/ref|><|det|>[[146, 685, 853, 890]]<|/det|>
+L.S., T.X. and Q.W. designed the study and supervised the project. G. G. grew the \(\mathrm{Bi}_2\mathrm{Sr}_2\mathrm{CaCu}_2\mathrm{O}_{8 + \delta}\) single crystals. H.T.Y and X.J.Z. grew the \(\mathrm{Bi}_2\mathrm{Sr}_{1.63}\mathrm{La}_{0.37}\mathrm{CuO}_{6 + \delta}\) single crystals. C.T. L. grew the \(\mathrm{Bi}_2\mathrm{Sr}_2\mathrm{Ca}_2\mathrm{Cu}_3\mathrm{O}_{10 + \delta}\) single crystals. Y.Z., J.G., S.C., and L.S. performed the high pressure resistance, \(ac\) susceptibility and magnetoresistance measurements. J.G., Y.Z., Y.G., Y.C.L., X.D.L. and C.L.Y. performed the high pressure X- ray diffraction measurements. J.G, C.H. and S.J.L. performed the high
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 93, 850, 148]]<|/det|>
+pressure Hall measurements. L.S., T.X., Q.W., J.P.H. and Y.Z. wrote the manuscript in consultation with all authors.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 205, 322, 222]]<|/det|>
+## Author information  
+
+<|ref|>text<|/ref|><|det|>[[147, 241, 850, 335]]<|/det|>
+The authors declare no competing financial interest. Correspondence and requests for materials should be addressed to L.S. (llsun@iphv.ac.cn). These authors with star (\*) contributed equally to this work.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 390, 245, 407]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[145, 426, 852, 895]]<|/det|>
+1 Chu, C. W., Deng, L. Z. & Lv, B. Hole-doped cuprate high temperature superconductors. Physica C Supercond 514, 290- 313 (2015).  2 Keimer, B., Kivelson, S. A., Norman, M. R., Uchida, S. & Zaanen, J. From quantum matter to high- temperature superconductivity in copper oxides. Nature 518, 179- 186 (2015).  3 Zaanen, J. Superconducting electrons go missing. Nature 536, 282- 283 (2016).  4 Norman, M. R. & Pepin, C. The electronic nature of high temperature cuprate superconductors. Rep. Prog. Phys. 66, 1547- 1610 (2003).  5 Barišić, N. et al. Universal sheet resistance and revised phase diagram of the cuprate high- temperature superconductors. Proc. Natl. Acad. Sci. U.S.A. 110, 12235- 12240 (2013).  6 Božović, I., He, X., Wu, J. & Bollinger, A. T. Dependence of the critical temperature in overdoped copper oxides on superfluid density. Nature 536, 309-
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[205, 95, 304, 111]]<|/det|>
+311 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[145, 130, 852, 225]]<|/det|>
+Greene, R. L., Mandal, P. R., Poniatowski, N. R. & Sarkar, T. The Strange Metal State of the Electron- Doped Cuprates. Annu. Rev. Condens. Matter Phys. 11, 213- 229 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[145, 240, 852, 300]]<|/det|>
+Sacépé, B., Feigel'man, M. & Klapwijk, T. M. Quantum breakdown of superconductivity in low- dimensional materials. Nat. Phys. 16, 734- 746 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[145, 316, 851, 373]]<|/det|>
+Proust, C. & Taillefer, L. The Remarkable Underlying Ground States of Cuprate Superconductors. Annu. Rev. Condens. Matter Phys. 10, 409- 429 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[145, 389, 850, 447]]<|/det|>
+Bednorz, J. G. & Müller, K. A. Possible high \(T_{c}\) superconductivity in the Ba- La- Cu- O system. Z Phys B Con Mat 64, 189- 193 (1986).  
+
+<|ref|>text<|/ref|><|det|>[[145, 463, 851, 521]]<|/det|>
+Wu, M. K. et al. Superconductivity at 93 K in a new mixed- phase Y- Ba- Cu- O compound system at ambient pressure. Phys. Rev. Lett. 58, 908- 910 (1987).  
+
+<|ref|>text<|/ref|><|det|>[[145, 537, 850, 594]]<|/det|>
+Zaanen, J. et al. Towards a complete theory of high \(T_{c}\) . Nat. Phys. 2, 138- 143 (2006).  
+
+<|ref|>text<|/ref|><|det|>[[145, 611, 850, 669]]<|/det|>
+Lee, P. A., Nagaosa, N. & Wen, X.- G. Doping a Mott insulator: Physics of high- temperature superconductivity. Rev. Mod. Phys. 78, 17- 85 (2006).  
+
+<|ref|>text<|/ref|><|det|>[[145, 686, 850, 743]]<|/det|>
+Zaanen, J. Planckian dissipation, minimal viscosity and the transport in cuprate strange metals. Scipost Phys 6, 061 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[145, 760, 850, 817]]<|/det|>
+Nakamae, S. et al. Electronic ground state of heavily overdoped nonsuperconducting La<sub>2-x</sub>Sr<sub>x</sub>CuO<sub>4</sub>. Phys. Rev. B 68, 100502 (2003).  
+
+<|ref|>text<|/ref|><|det|>[[145, 834, 850, 891]]<|/det|>
+Hussey, N. E., Abdel- Jawad, M., Carrington, A., Mackenzie, A. P. & Balicas, L. A coherent three- dimensional Fermi surface in a high- transition- temperature
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[207, 94, 576, 112]]<|/det|>
+superconductor. Nature 425, 814- 817 (2003).  
+
+<|ref|>text<|/ref|><|det|>[[147, 130, 850, 188]]<|/det|>
+17 Plate, M. et al. Fermi Surface and Quasiparticle Excitations of Overdoped Tl₂Ba₂CuO₆+δ. Phys. Rev. Lett. 95, 077001 (2005).  
+
+<|ref|>text<|/ref|><|det|>[[147, 205, 850, 261]]<|/det|>
+18 Vignolle, B. et al. Quantum oscillations in an overdoped high- Tc superconductor. Nature 455, 952- 955 (2008).  
+
+<|ref|>text<|/ref|><|det|>[[147, 279, 850, 336]]<|/det|>
+19 Chu, C. W. et al. Evidence for superconductivity above 40 K in the La- Ba- Cu- O compound system. Phys. Rev. Lett. 58, 405- 407 (1987).  
+
+<|ref|>text<|/ref|><|det|>[[147, 353, 850, 410]]<|/det|>
+20 Takahashi, H. et al. Superconductivity at 43 K in an iron- based layered compound LaO₁- xFxFeAs. Nature 453, 376- 378 (2008).  
+
+<|ref|>text<|/ref|><|det|>[[147, 427, 850, 521]]<|/det|>
+21 Zhi- An, R. et al. Superconductivity at 55 K in Iron- Based F- Doped Layered Quaternary Compound Sm[O₁- xFx]FeAs. Chinese Phys. Lett. 25, 2215- 2216 (2008).  
+
+<|ref|>text<|/ref|><|det|>[[147, 539, 850, 670]]<|/det|>
+22 S. Schilling, J. in Frontiers of High Pressure Research II: Application of High Pressure to Low- Dimensional Novel Electronic Materials (eds Hans D. Hochheimer, Bogdan Kuchta, Peter K. Dorhout, & Jeffery L. Yarger) 345- 360 (Springer Netherlands, 2001).  
+
+<|ref|>text<|/ref|><|det|>[[147, 687, 848, 743]]<|/det|>
+23 Sun, L. et al. Re- emerging superconductivity at 48 kelvin in iron chalcogenides. Nature 483, 67 (2012).  
+
+<|ref|>text<|/ref|><|det|>[[147, 760, 850, 817]]<|/det|>
+24 Deng, L. et al. Higher superconducting transition temperature by breaking the universal pressure relation. Proc. Natl. Acad. Sci. U.S.A. 116, 2004- 2008 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[147, 834, 850, 891]]<|/det|>
+25 Guo, J. et al. Crossover from two- dimensional to three- dimensional superconducting states in bismuth- based cuprate superconductor. Nat. Phys. 16,
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[208, 95, 343, 111]]<|/det|>
+295- 300 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[147, 131, 850, 188]]<|/det|>
+26 Parker, C. V. et al. Fluctuating stripes at the onset of the pseudogap in the high- \(T_{c}\) superconductor \(\mathrm{Bi}_{2}\mathrm{Sr}_{2}\mathrm{CaCu}_{2}\mathrm{O}_{8 + x}\) . Nature 468, 677- 680 (2010).  
+
+<|ref|>text<|/ref|><|det|>[[147, 205, 853, 262]]<|/det|>
+27 Kondo, T. et al. Disentangling Cooper- pair formation above the transition temperature from the pseudogap state in the cuprates. Nat. Phys. 7, 21- 25 (2011).  
+
+<|ref|>text<|/ref|><|det|>[[147, 279, 850, 336]]<|/det|>
+28 Zhao, H. et al. Charge- stripe crystal phase in an insulating cuprate. Nat. Mater. 18, 103- 107 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[147, 354, 850, 410]]<|/det|>
+29 Du, Z. et al. Imaging the energy gap modulations of the cuprate pair- density- wave state. Nature 580, 65- 70 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[147, 428, 850, 521]]<|/det|>
+30 Gao, Q. et al. Electronic Evolution from the Parent Mott Insulator to a Superconductor in Lightly Hole- Doped \(\mathrm{Bi}_{2}\mathrm{Sr}_{2}\mathrm{CaCu}_{2}\mathrm{O}_{8 + \delta}\) . Chinese Phys. Lett. 37, 087402 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[147, 539, 850, 595]]<|/det|>
+31 Adachi, T. et al. Magnetic- field effects on the charge- spin stripe order in La- 214 high- \(T_{C}\) cuprates. J. Phys. Conf. Ser. 51, 259- 262 (2006).  
+
+<|ref|>text<|/ref|><|det|>[[147, 613, 851, 706]]<|/det|>
+32 Muramatsu, T., Pham, D. & Chu, C. W. A possible pressure- induced superconducting- semiconducting transition in nearly optimally doped single crystalline \(\mathrm{YBa}_{2}\mathrm{Cu}_{3}\mathrm{O}_{7 - \delta}\) . Appl. Phys. Lett. 99, 052508 (2011).  
+
+<|ref|>text<|/ref|><|det|>[[147, 724, 850, 817]]<|/det|>
+33 Fang, Y., Yazici, D., Jeon, I. & Maple, M. B. High- pressure effects on nonfluorinated \(\mathrm{BiS}_{2}\) - based superconductors \(\mathrm{La}_{1 - x}\mathrm{M}_{x}\mathrm{OBiS}_{2}\) ( \(M = \mathrm{Ti}\) and Th). Phys. Rev. B 96, 214505 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[147, 835, 850, 891]]<|/det|>
+34 Moritomo, Y., Kuwahara, H. & Tokura, Y. Bandwidth- and Doping- Dependent Pressure Effects on the Ferromagnetic Transition in Perovskite Manganites. J.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 196, 841, 377]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 392, 850, 411]]<|/det|>
+<center>Figure 1 Temperature dependence of in-plane resistance for Bi₂Sr₂CaCu₂O₈+δ </center>  
+
+<|ref|>text<|/ref|><|det|>[[147, 428, 852, 632]]<|/det|>
+(Bi2212) at different pressures: (a) for the under- doped (UD) superconductor with superconducting transition temperature \((T_{c})\) about \(74\mathrm{K}\) in the pressure range of \(1.5\mathrm{GPa}\) - \(36.2\mathrm{GPa}\) ; (b) for the optimally- doped (OP) sample with \(T_{c}\) about \(91\mathrm{K}\) in the pressure range of \(0.7\mathrm{GPa} - 41.2\mathrm{GPa}\) ; (c) for the over- doped (OD) sample with \(T_{c}\) about \(82\mathrm{K}\) in the pressure range of \(1.0\mathrm{GPa} - 49\mathrm{GPa}\) . The three kinds of samples display the same behavior of an insulating- like state above the pressure \((P_{i})\) .
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 90, 850, 375]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 392, 850, 415]]<|/det|>
+<center>Figure 2. In-plane resistance \((R)\) and \(ac\) susceptibility \((\Delta \chi^{\prime})\) as a function of </center>  
+
+<|ref|>text<|/ref|><|det|>[[147, 430, 850, 632]]<|/det|>
+temperature \((T)\) for the \(\mathrm{Bi_2Sr_2CaCu_2O_{8 + \delta}}\) superconductors at different pressures: (a)-(d) for the under- doped (UD) superconductor; (e)-(h) for the optimally- doped (OP) superconductor; (i)-(l) for the over- doped (OD) superconductor. The blue lines in the figures are the data of \(\Delta \chi^{\prime}(T)\) , while the red lines are the data of \(R(T)\) . The red and blue arrows indicate the temperatures of the onset superconducting transition detected by resistance and \(ac\) susceptibility measurements, respectively.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 90, 848, 393]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 409, 852, 911]]<|/det|>
+<center>Figure 3 Pressure- \(T_{c}\) phase diagrams for \(\mathrm{Bi_2Sr_2CaCu_2O_{8 + \delta}}\) superconductors. The Right panels are the phase diagrams established by the experimental results from the under-doped (UD), optimally-doped (OP) and over-doped (OD) samples, together with the mapping information of temperature and pressure dependent \(R\) (shown in color scale). The left panel is a normalized phase diagram that is built on the basis of the experimental phase diagrams (the right panels). \(P_{Tc - max}\) and \(P_{c}\) stand for the critical pressures where \(T_{c}\) reaches the maximum and the zero, respectively. In the normalizing analysis, we define the pressure as \(P_{Tc - max}\) when \((P - P_{Tc - max}) / (P_{c} - P_{Tc - max}) = 0\) , and the pressure as \(P_{c}\) when \((P - P_{Tc - max}) / (P_{c} - P_{Tc - max}) = 1\) . The results of the normalizing analysis for \(T_{c} / T_{c - max}\) versus \((P - P_{Tc - max}) / (P_{c} - P_{Tc - max})\) and \(T_{ins} / T_{c - max}\) versus \((P - P_{Tc - max}) / (P_{c} - P_{Tc - max})\) show that the three kinds of samples display a universal quantum phase transition from the superconducting state to an insulating-like state. SC, M and I stand for superconducting state, metallic state and insulating-like state, respectively. The region of the M phase is determined by the critical value of \(R / R_{290K}\) where the quantum phase </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[146, 92, 852, 523]]<|/det|>
+transition occurs. For example, when \(R / R_{290K}\) is greater than 3, the over- doped sample is in the insulating- like state while, \(R / R_{290K}\) is less than 3, the sample is in the metallic state (see right bottom panel). \(T_{c - R - onset}\) and \(T_{c - \Delta \chi ' - onset}\) denote the onset temperatures of the superconducting transition detected by the resistance and \(ac\) susceptibility measurements, respectively. \(T_{c - max}\) and \(T_{ins}\) are the maximum value of \(T_{c}\) and the characteristic temperature of the insulating- like transition (the method of determining the \(T_{ins}\) can be found in the Supplementary information), respectively. The \(P_{c}\) value is determined by the average pressure of two experimental runs \([P_{c} = (P_{c - run1} + P_{c - run2}) / 2]\) , in which \(P_{c}\) of each experimental run is determined by the highest experimental pressure where the superconducting transition can still be observed and the lowest experimental pressure where the insulating- like state appears. The error bar of \(P_{c}\) is the difference between \(P_{c - run1}\) and \(P_{c - run2}\) .
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 92, 845, 500]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 520, 850, 802]]<|/det|>
+<center>Figure 4 Magnetoresistance \((MR)\) as a function of magnetic field \((B)\) for the UD, OP and OD \(\mathrm{Bi_2Sr_2CaCu_2O_{8 + \delta}}\) superconductors when they enter into an insulating-like state, and Hall coefficient information of the OP sample under pressure. (a)- (b) Plots of \(MR\) versus \(B\) for the UD, OP and OD samples measured at \(4\mathrm{K}\) at \(36.2\mathrm{GPa}\) , 41.2 GPa and 49 GPa, respectively. It is seen that all of them exhibit a positive magnetic effect. The red and beige arrows represent the directions of increasing and decreasing magnetic field. (d) Pressure dependence of Hall coefficient \((R_H)\) for the OD superconductor measured at \(100\mathrm{K}\) . </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[44, 42, 311, 70]]<|/det|>
+## Supplementary Files  
+
+<|ref|>text<|/ref|><|det|>[[44, 93, 765, 113]]<|/det|>
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+<|ref|>text<|/ref|><|det|>[[61, 131, 156, 149]]<|/det|>
+- Slf.docx
+
+<--- Page Split --->

preprint/preprint__491dd82e431bd9db89911af3357d8436b6d13fcb4ae9cd9d6e41df61934cee6d/images_list.json

@@ -0,0 +1,33 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_0.jpg",
+    "caption": "b",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 41
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_1.jpg",
+    "caption": "d",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 42
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_unknown_2.jpg",
+    "caption": "e",
+    "footnote": [],
+    "bbox": [
+      [
+        420,
+        420,
+        947,
+        500
+      ]
+    ],
+    "page_idx": 43
+  }
+]

preprint/preprint__491dd82e431bd9db89911af3357d8436b6d13fcb4ae9cd9d6e41df61934cee6d/preprint__491dd82e431bd9db89911af3357d8436b6d13fcb4ae9cd9d6e41df61934cee6d.mmd

@@ -0,0 +1,514 @@
+
+# The Greatwall-Endosulfine-PP2A/B55 pathway controls entry into quiescence by promoting translation of Elongator-tuneable transcripts  
+
+Sergio Moreno smo@usal.es  
+
+CSIC / University of Salamanca https://orcid.org/0000- 0002- 8039- 1413  
+
+Javier Encinar del Dedo CSIC / University of Salamanca  
+
+Rafael López- San Segundo CSIC / University of Salamanca  
+
+Alicia Vázquez- Bolado CSIC / University of Salamanca  
+
+Jingjing Sun  
+
+Antimicrobial Resistance Interdisciplinary Research Group, Singapore- MIT Alliance for Research and Technology  
+
+Natalia García- Blanco CSIC / University of Salamanca  
+
+M. Belén Suárez CSIC / University of Salamanca  
+
+Patricia García CSIC / University of Salamanca https://orcid.org/0000- 0001- 7513- 1847  
+
+Pauline Tricquet URPHYM- GEMO. The University of Namur  
+
+Jun- Song Chen Vanderbilt University School of Medicine  
+
+Peter Dedon Massachusetts Institute of Technology https://orcid.org/0000- 0003- 0011- 3067  
+
+Kathleen Gould Vanderbilt University  
+
+Elena Hidalgo Universitat Pompeu Fabra https://orcid.org/0000- 0002- 3768- 6785  
+
+Damien Hermand
+
+<--- Page Split --->
+
+## Article  
+
+Keywords: Quiescence, Nitrogen starvation, TORC1, TORC2, Greatwall, Endosulfine, PP2A/B55, tRNA modifications, Elongator, translation  
+
+Posted Date: December 5th, 2023  
+
+DOI: https://doi.org/10.21203/rs.3.rs- 3616701/v1  
+
+License: © This work is licensed under a Creative Commons Attribution 4.0 International License. Read Full License  
+
+Additional Declarations: There is NO Competing Interest.  
+
+Version of Record: A version of this preprint was published at Nature Communications on December 5th, 2024. See the published version at https://doi.org/10.1038/s41467- 024- 55004- 4.
+
+<--- Page Split --->
+
+1 The Greatwall-Endosulfine-PP2A/B55 pathway controls entry 2 into quiescence by promoting translation of Elongator-tuneable transcripts  
+
+4 5 6 Javier Encinar del Dedo \(^{1,*}\) , Rafael López- San Segundo \(^{1}\) , Alicia Vázquez- Bolado \(^{1}\) , Jingjing Sun \(^{2}\) , Natalia García- Blanco \(^{1}\) , M. Belén Suárez \(^{3,4}\) , Patricia García \(^{3,4}\) , Pauline Tricquet \(^{5}\) , Jun- Song Chen \(^{6}\) , Peter C. Dedon \(^{2,7}\) , Kathleen L. Gould \(^{6}\) , Elena Hidalgo \(^{8}\) , Damien Hermand \(^{5}\) and Sergio Moreno \(^{1,9,*}\)  
+
+1 Instituto de Biología Funcional y Genómica, CSIC, University of Salamanca, 37007 Salamanca, Spain. 2 Antimicrobial Resistance Interdisciplinary Research Group, Singapore-MIT Alliance for Research and Technology, Singapore, Singapore. 3 Instituto de Biología Funcional y Genómica, University of Salamanca, CSIC, 37007 Salamanca, Spain. 4 Departamento de Microbiología y Genética, University of Salamanca, 37007 Salamanca, Spain. 5 URPHYM-GEMO, University of Namur, rue de Bruxelles, 61, Namur 5000, Belgium. 6 Department of Cell and Developmental Biology, Vanderbilt University School of Medicine, Nashville, United States. 7 Department of Biological Engineering and Center for Environmental Health Science, Massachusetts Institute of Technology, Cambridge, MA, United States. 8 Oxidative Stress and Cell Cycle Group, Universitat Pompeu Fabra, 08003 Barcelona, Spain. 9 Lead contact.  
+
+Keywords: Quiescence, Nitrogen starvation, TORC1, TORC2, Greatwall, Endosulfine, PP2A/B55, tRNA modifications, Elongator, translation  
+
+Running title: PP2A/B55 activity controls translation during quiescence
+
+<--- Page Split --->
+
+## 41 Summary  
+
+42 Quiescent cells require a continuous supply of proteins to maintain protein 43 homeostasis. In fission yeast, entry into quiescence is triggered by nitrogen 44 stress, leading to the inactivation of TORC1 and the activation of TORC2. Here, 45 we report that the Greatwall- Endosulfine- PPA/B55 pathway connects the 46 downregulation of TORC1 with the upregulation of TORC2, resulting in the 47 activation of Elongator- dependent tRNA modifications essential for sustaining the 48 translation programme during entry into quiescence. This process promotes 49 \(\mathsf{U}_{34}\) and \(\mathsf{A}_{37}\) tRNA modifications at the anticodon stem loop, enhancing translation 50 efficiency and fidelity of mRNAs enriched for AAA versus AAG lysine codons. 51 Notably, some of these mRNAs encode inhibitors of TORC1, activators of 52 TORC2, tRNA modifiers, and proteins necessary for telomeric and subtelomeric 53 functions. Therefore, we propose a novel mechanism by which cells respond to 54 nitrogen stress at the level of translation, involving a coordinated interplay 55 between the tRNA epitranscriptome and biased codon usage.
+
+<--- Page Split --->
+
+## 59 Introduction  
+
+Most cells in living organisms rest in a non- dividing state called the \(\mathtt{G_0}\) phase, also known as quiescence. Quiescent cells can re- enter the cell cycle with full viability when provided with the appropriate signals. Examples of quiescent cells include stem cells, neuronal progenitor cells, memory T cells, eggs, and spores. Despite their importance, the molecular mechanisms governing quiescence entry, its maintenance, and exit are not yet fully understood.  
+
+In nature, unicellular organisms continuously enter and exit quiescence depending on nutrient availability. In fission yeast, entry into quiescence is induced by nitrogen starvation, resulting in the inactivation of TOR complex 1 (TORC1) and the activation of TOR complex 2 (TORC2). TORC1 promotes cell growth in response to nutrients, growth factors, or cellular energy \(^{1,2}\) , while TORC2 is required for the nutrient stress response, cell survival during quiescence, and cell differentiation \(^{3 - 5}\) . The activation of the Greatwall- Endosulfine switch upon TORC1 inactivation leads to inhibition of the PP2A/B55 protein phosphatase, which is necessary for switching on TORC2 activity by increasing Gad8 phosphorylation \(^{4 - 7}\) .  
+
+TORC2 activation also regulates protein translation by controlling tRNA modifications through the Elongator complex \(^{8}\) . Elongator is a multiprotein complex that modifies the anticodon stem loop of tRNA \(^{Lys}Uuu\) , tRNA \(^{Glu}Uuc\) and tRNA \(^{Gln}Uug\) by introducing an acetyl group at position 5 of U \(_{34}\) (cm \(^{5}U_{34}\) ), which is further modified by Trm112- Trm9, a methyltransferase complex involved in the formation of mcm \(^{5}U_{34}\) , and by Ctu1- Ctu2 complex, which catalyses the thiolation at carbon 2 of U \(_{34}\) (mcm \(^{5}s^{2}U_{34}\) ). These U \(_{34}\) modifications counteract codon misreading resulting from low effective stacking interactions between A- U bases \(^{9 - 11}\) . They also play a crucial role in maintaining translational fidelity under stress conditions \(^{12 - 15}\) . Thus, Elongator is necessary for the efficient translation of mRNAs with a high AAA codon usage.  
+
+Previous studies have reported a feedback loop between the TORC1- TORC2 signalling cascade and the Elongator complex. In this loop, Elongator plays an essential role in the translation of key components of TORC2 and repressors of TORC1. Additionally, the TORC2 pathway functions as an activator of Elongator
+
+<--- Page Split --->
+
+by down- regulating Gsk3, a glycogen synthase kinase that inhibits Elongator by phosphorylating the Elp4 subunit at Serine114 8,16,17. In this study, we report that elevated PP2A/B55 phosphatase activity, resulting from the deletion of Endosulfine (igo1Δ), impairs the translation efficiency of mRNAs enriched in AAA codons during entry into quiescence. Additionally, we demonstrate a physical and functional interaction between PP2A/B55 and Gad8, Trm112, Ctu1, and the Elongator complex. Furthermore, hyperactivation of PP2A/B55 protein phosphatase reduces the function of the Elongator complex and the amount of Trm112, Ctu1 and Cgi121 proteins, which are essential for U34 and A37 tRNA modifications at the anticodon stem loop. This reduction in translational efficiency leads to decreased protein levels from transcripts containing high AAA codon usage, such as rap1, sgo2, clr2, or clr3, all of which are crucial for telomeric and subtelomeric organisation. This induces telomeric detachment, upregulation of subtelomeric gene expression, and eventually, cell death. Our work suggests that the Greatwall-Endosulfine-PP2A/B55 pathway governs the translational programme during entry into quiescence through the control of U34 and A37 tRNA modifications. We propose that the implementation of an alternative gene expression programme in response to nitrogen starvation is based on translation of mRNAs enriched in sub- optimal AAA codons by activation of tRNA- modifying complexes.
+
+<--- Page Split --->
+
+## Results  
+
+The Greatwall-Endosulfine switch regulates telomere silencing and telomere attachment to the nuclear envelope  
+
+To understand the function of the Greatwall- Endosulfine- PP2A/B55 pathway during entry into quiescence we compared the transcriptome of the wild- type (WT) and the Endosulfine mutant (igo1) by RNAseq after shifting cells from nitrogen- rich (EMM2) to nitrogen- free (EMM2- N) media at times 0 and 4 hours. In nitrogen- rich medium (EMM2, time 0) the transcriptome was almost identical between the two strains. However, after 4 hours of nitrogen starvation, we found significant changes in subtelomeric gene expression, as igo1 cells showed a high expression level (more than 10- fold) of a group of subtelomeric genes in chromosomes I and II compared to WT cells (Fig. 1a; Supplementary Table 1). Similar results were obtained when we analysed the transcriptome of the Greatwall (ppk18 \(\Delta\) cek1) mutant (Supplementary Fig. 1a; Supplementary Table 2), consistent with the fact that the Greatwall- Endosulfine- PP2A/B55 is a linear pathway 4. These results suggest that downregulation of PP2A/B55 plays a key role in transcriptional silencing of subtelomeric genes during quiescence entry.  
+
+The ends of chromosomes I and II are composed of telomeric repeats and the subtelomeric regions. While the telomeric repeats extend approximately 300 bp, the subtelomeric regions consist of about 100 kilobases between the telomeric repeats and the euchromatin (Fig. 1b). The heterochromatin present in the subtelomeric regions can be divided into SH chromatin, characterised by highly methylated histone H3K9, and ST chromatin, in which histone modifications are kept at low levels, but exhibit highly condensed chromosome structures called knobs 18- 22. Several protein complexes essential for maintaining the telomeric and subtelomeric structure have been identified. For example, Rap1 (a component of the shelterin complex) and Bqt4 (a component of the bouquet complex) create a molecular link between telomeres and the nuclear envelope 23- 26. Proteins such as Swi6, the SHREC complex or the CLRC complex play a role in H3K9 methylation, control nucleosome maintenance and genome stability 27,28. Finally, shugoshin 2, Sgo2, is an essential protein for condensation of ST chromatin and knob stability 21,22 (Fig. 1b).
+
+<--- Page Split --->
+
+To study the role of the Greatwall- Endosulfine- PP2A/B55 pathway in telomeric organisation during quiescence, we analysed nuclear- telomeric attachment in the wild- type (WT) and in the Endosulfine (igo1Δ) mutant in nitrogen- rich media (EMM2) and after 8 hours of nitrogen starvation (EMM2- N) using Super- Resolution Radial Fluctuations (SRRF) microscopy. Wild- type and igo1Δ cells tagged with Cut11- mCherry (a nuclear envelope - NE- marker), Sad1- CFP (a spindle pole body - SPB- marker) and Taz1- YFP (a telomeric marker), showed no significant differences in nitrogen- rich media. In contrast, after 8 hours of nitrogen starvation, the igo1Δ mutant showed telomeric detachment from the NE (Fig. 1c). To analyse the defect of the igo1Δ mutant in more detail, we combined SRRF microscopy with Radial Profile Analysis (see details in Supplementary Fig. 1b and Methods). The wild- type and igo1Δ mutant showed a perfect overlap between the NE signal (red line) and the SPB signal (blue line). However, we detected differences in the telomeric signals (yellow line) between strains. While in the wild- type strain the three signals overlapped more with time (4 and 8 hours of nitrogen starvation), in the igo1Δ mutant, the telomeric signal separated from the NE and the SPB signals (Fig. 1d). A similar result was obtained when we analysed the overlap between the mean NE signal and the mean SPB signal or the mean NE signal and the mean telomeric signal at different time points (Fig. 1e). Pearson's correlation coefficients allowed us to identify significant differences between Cut11/Sad1 and Cut11/Taz1 signals in the wild- type and the igo1Δ mutant (Fig. 1f). These results indicate that the interaction between the NE and telomeres is lost in the igo1Δ mutant.  
+
+## Telomeric detachment is mediated by reduced levels of the Rap1 protein, a component of the shelterin complex  
+
+A high level of PP2A/B55 activity, caused by deleting igo1, delays entry into mitosis during vegetative growth when fission yeast cells are shifted from a nitrogen- rich to a nitrogen- poor medium, or during entry into quiescence 4. Therefore, it seemed possible that elevated PP2A/B55 activity caused telomeric detachment during entry into quiescence. In S. pombe, two different complexes have been described as essential for maintaining telomeric- NE attachment, the bouquet complex and the shelterin complex. Interestingly, two subunits of the
+
+<--- Page Split --->
+
+shelterin complex, Rap1 and Ccq1, have been described as heavily phosphorylated proteins \(^{25,26,29,30}\) . Thus, we considered that the phosphorylation state of these proteins might be affected by the high PP2A/B55 phosphatase activity in the \(igo1\Delta\) mutant, triggering telomeric detachment. However, we did not detect changes in the phosphorylation state of either Rap1 or Ccq1. Surprisingly, we detected a dramatic reduction in the amount of Rap1 protein levels in the \(igo1\Delta\) mutant, while the Ccq1 levels remained constant during the experiment (Fig. 2a). To confirm this data and improve our temporal resolution, we repeated the experiment taking samples every 30 minutes during the first 2 hours and then after 4 hours of nitrogen starvation. Once more, we detected a very significant decrease in Rap1 protein levels after 2- 4 hours of nitrogen deprivation in the \(igo1\Delta\) mutant (Fig. 2b; Fig. 2c, left panel).  
+
+Previous results from our lab have shown that \(igo1\Delta\) phenotypes could be restored by decreasing PP2A/B55 activity \(^{31}\) , including the reduction in viability during quiescence (Supplementary Fig. 2a). This prompted us to investigate whether a reduction of PP2A/B55 activity could restore Rap1 protein levels. To modulate PP2A/B55 activity, we placed the \(pab1\) open reading frame, encoding the PP2A B55 regulatory subunit, under the control of the thiamine- repressible \(nmt41\) promoter at its chromosomal locus. We found that repressing \(Pab1\) production and therefore PP2A/B55 activity reinstated Rap1 protein levels (Fig. 2c). Data quantification further confirmed the restoration of Rap1 protein levels when PP2A/B55 activity was reduced (Fig. 2d).  
+
+To investigate whether reducing PP2A/B55 activity could also prevent telomeric detachment in vivo, we examined the localization of Taz1- YFP in strains exhibiting low PP2A/B55 activity (wild- type and \(igo1\Delta\) \(nmt41:GST:pab1 +\) Thiamine) in comparison to strains with elevated PP2A/B55 activity ( \(igo1\Delta\) and \(igo1\Delta\) \(nmt41:GST:pab1 -\) Thiamine) during nitrogen starvation. Our analysis showed that low PP2A/B55 activity restored the telomeric detachment phenotype, whereas high PP2A/B55 activity maintained the telomeric attachment defect (Fig. 2e; Supplementary Fig. 2b). Statistical analysis of the data confirmed that reduced PP2A/B55 activity during nitrogen starvation was necessary for preserving telomeric organisation during quiescence (Fig. 2f). In summary, these findings suggest that the downregulation of PP2A/B55 activity during entry into
+
+<--- Page Split --->
+
+quiescence is crucial for maintaining Rap1 protein levels and for anchoring telomeres to the NE.  
+
+## Sgo2, Clr2 and Clr3 protein levels are reduced in quiescent igo1Δ mutant cells  
+
+Different protein complexes coordinately maintain chromatin silencing in subtelomeric regions during quiescence in fission yeast. One of the critical factors for this regulation is shugoshin 2, Sgo2. Sgo2 is essential for the formation of knobs, highly condensed chromatin structures organised close to the ends of chromosomes I and II \(^{21}\) . Lack of Sgo2 (sgo2Δ) induces transcription of genes located at subtelomeric regions on chromosomes I and II \(^{22}\) , similar to what we observed in cells lacking Endosulfine (igo1Δ) or Greatwall (ppk18Δ cek1Δ) after nitrogen starvation (Fig. 1a; Supplementary Fig. 1a; Supplementary Tables 1, 2). This correlation prompted us to investigate whether Sgo2 levels might be altered in the igo1Δ mutant. Western- blot analysis revealed that igo1- deleted cells show a severe reduction of Sgo2 levels during entry into quiescence (Fig. 3a). As in the case of Rap1, reducing Pab1 levels in the igo1Δ mutant restored Sgo2 levels (Fig. 3b). To confirm the role of Igo1 in maintenance of Sgo2 levels and knob formation, we studied the localisation of Sgo2 during quiescence entry. Sgo2 protein was tagged with GFP and its localisation during nitrogen starvation was examined. As previously described, in the wild- type strain, Sgo2- GFP localised as nuclear dots in most cells, ranging from 1 to 3 dots per cell (Fig. 3c). In the igo1Δ mutant, we detected no significant differences with the wild- type in nitrogen- rich media (t=0 hours), only a slight decrease in dot size and brightness. However, when we shifted the cells to nitrogen- free media we observed a clear decrease in the number of dots per cell in the igo1Δ mutant (Fig. 3c). These data indicate that Igo1 is required for maintaining Sgo2 protein levels and for the formation of knobs, a structure essential for maintaining the transcriptional repression of subtelomeric genes.  
+
+Another key protein complex required for silencing subtelomeric regions is the heterochromatic repressor complex SHREC (Snf2- like/HDAC- containing repressor complex) \(^{32,33}\) , composed of Mit1, Clr1, Clr2, and Clr3. The SHREC complex plays regulatory roles in histone acetylation, as a chromatin remodeller
+
+<--- Page Split --->
+
+and in the stability of subtelomeric nucleosomes \(^{27,32,33}\) . To determine if SHREC was also affected by PP2A activity, we examined Clr2 and Clr3 levels. In both cases, we detected a modest but reproducible decrease in protein levels during nitrogen starvation in the \(igo1\Delta\) mutant (Fig. 3d,e). ChIP analysis showed that lack of Igo1 caused an increase in histone H3- K14 acetylation in the overexpressed subtelomeric genes SPCC977.15 and SPAC186.06 after 4 hours of nitrogen starvation consistent with loss of SHREC function (Fig. 3f). These results suggest that the Greatwall- Endosulfine- PP2A/B55 pathway modulates SH chromatin organisation and subtelomeric gene silencing.  
+
+# PP2A/B55 regulates translation through its physical and functional interaction with protein complexes involved in tRNA modification  
+
+Our data indicate that the \(igo1\Delta\) mutant exhibits reduced levels of proteins essential for maintaining telomeric and subtelomeric organisation. However, what is the molecular mechanism underlying these phenotypes? We explored three possibilities: reduced protein stability, reduced transcription, or reduced translation.  
+
+To examine protein stability, we treated cells with cycloheximide and monitored Rap1 protein levels over time. After treatment with cycloheximide, Rap1 was degraded with similar kinetics in wild- type and \(igo1\Delta\) cells (Supplementary Fig. 3a). Similarly, rap1 mRNA levels were not reduced in \(igo1\Delta\) cells; on the contrary, the rap1 gene exhibited higher transcript levels in the \(igo1\Delta\) mutant compared to wild- type cells (Supplementary Fig. 3b).  
+
+Evidence of a translation defect in the \(igo1\Delta\) mutant was obtained by mass- spectrometry analysis of proteins co- purifying with the PP2A/Pab1 protein phosphatase. Paa1, the structural subunit of the PP2A complex, was tagged with YFP and expressed from its endogenous promoter at its chromosomal locus. After one hour of nitrogen starvation, Paa1- YFP was pulled down and co- purifying proteins were analysed by mass- spectrometry. Our results revealed the presence of all components of PP2A protein complexes including Paa1, the catalytic subunits Ppa1, Ppa2, and Ppa3, and the regulatory subunits Pab1, Par1 and Par2 (Supplementary Table 3). Additionally, several PP2A regulators (Igo1, Zds1,
+
+<--- Page Split --->
+
+Dis2, Ppe1 and Ekc1) and components of the PP2A SIP/STRIPAK complex 34 were also detected, confirming that the pull- down approach was successful.  
+
+The mass- spectrometry analysis also showed an over- representation of proteins related to ribosome structure, translation initiation, aminoacylation, and tRNA modification in the Paa1 interactome (Fig. 4a). We focused on Trm112, a widely conserved protein with a crucial role in translation. Specifically, Trm112 regulates methyltransferase enzymes (Trm9, Trm11, Mtq2 and Bud23) during ribosome biogenesis, tRNA modification and stop codon recognition 35,36. We confirmed an interaction between Trm112 and the PP2A- Pab1 complex by repeating the mass- spectrometry analysis using Pab1, the B55 regulatory subunit of the PP2A complex, as bait. Trm112 was pulled down as an interacting partner of PP2A/Pab1 (Supplementary Fig. 4a,b; Supplementary Table 4). The Trm112- Paa1 interaction was further validated by co- immunoprecipitation, showing a stronger association in nitrogen- depleted than in nitrogen- rich media (Fig. 4b). Interestingly, several subunits of the Elongator complex (Elp1, Elp2 and Elp3) were also pulled down as interacting partners of PP2A/Pab1 when Pab1 was slightly overexpressed from the nmt41 promoter (Supplementary Fig. 4b,c; Supplementary Table 5).  
+
+The Elongator complex, along with Trm112/Trm9 and the Ctu1/Ctu2 complexes, plays a critical role in the formation of the 5- methoxycarbonylmethyl (mcm5) and 5- methoxycarbonylmethyl- 2- thiouridine (mcm5s2) side chains on uridine 34 (U34) at the tRNA wobble position during vegetative growth and under stress conditions 8,12,13,37- 40. We conducted an analysis of Trm112 protein levels during nitrogen starvation in both wild- type and igo1 deleted cells. In the wild- type, Trm112 levels remained constant during the first two hours and then exhibited a slight decrease after four hours. In contrast, in the igo1Δ mutant, we observed a more pronounced reduction in Trm112 proteins levels during entry into quiescence (Fig. 4c, left panel). Interestingly, as shown previously for other proteins, the reduction of PP2A/B55 activity restored Trm112 protein levels (Fig. 4c, right panel). We also found that the levels of Ctu1 protein, which cooperates with Ctu2 in tRNA U34 thiolation 41,42, were also diminished in the igo1Δ background (Fig. 4d).
+
+<--- Page Split --->
+
+Collectively, these data suggest potential impairment of tRNA modifications in the \(igo1\Delta\) mutant. To assess this possibility, we tested the sensitivity of the \(igo1\Delta\) mutant to drugs that affect translation, such as paromomycin, puromycin or cycloheximide, in nitrogen- rich (EMM2) and nitrogen- poor (MMPhe) media. Among all the drugs tested, only paromomycin, which induces codon misreading \(^{43}\) , exhibited an effect on the \(igo1\Delta\) mutant (Fig. 4e), particularly in MMPhe.  
+
+In summary, our data indicate a potential role for the Greatwall- Endosulfine- PP2A/B55 pathway in translation during the onset of quiescence. This role likely involves tRNA modifications that enhance codon- anticodon recognition.  
+
+## The \(igo1\Delta\) mutant is defective in \(\mathbf{U}_{34}\) and \(\mathbf{A}_{37}\) tRNA modifications  
+
+In all organisms, modifications of uridine 34 at the wobble position ( \(\mathbf{U}_{34}\) ) of certain tRNAs are necessary to enhance codon- anticodon recognition \(^{37}\) . These modifications are mediated by the Elongator complex, which introduces an acetyl group at position 5 of \(\mathbf{U}_{34}\) (cm \(^{5}\mathbf{U}_{34}\) ), the Trm112/Trm9 methyltransferase complex involved in the formation of mcm \(^{5}\mathbf{U}_{34}\) , and the Ctu1- Ctu2 complex, which catalyses the thiolation at carbon 2 of \(\mathbf{U}_{34}\) (mcm \(^{5}\mathbf{s}^{2}\mathbf{U}_{34}\) ) (Supplementary Fig. 5a).  
+
+Previous studies in S. pombe have reported that differences in codon usage and tRNA modifications play a crucial role in regulating translation efficiency during the cell cycle and under oxidative stress. The mRNAs of the cell cycle regulator Cdr2 \(^{39}\) and of the stress- responsive transcription factors Atf1 and Pcr1 \(^{12,13}\) exhibit a high usage of lysine AAA codons compared to AAG codons, and their translational rate is particularly sensitive to deficiencies in tRNA modifications mediated by the Elongator, Trm112/Trm9 and Ctu1/Ctu2 complexes \(^{8,12,13,39}\) . Therefore, we hypothesised that differences in \(\mathrm{AAA}_{\mathrm{lys}}\) codon usage might be responsible for the translation phenotype observed in the \(igo1\Delta\) mutant. To test this hypothesis, we examined the use of \(\mathrm{AAA}_{\mathrm{lys}}\) versus AAG \(\mathrm{lys}\) codons for some of the proteins analysed in our study and found that all the proteins defective in the \(igo1\Delta\) mutant (Rap1, Sgo2, Clr3, Clr2, and Ctu1) primarily utilise the \(\mathrm{AAA}_{\mathrm{lys}}\) codon (Supplementary Fig. 5b). Proteins with reduced \(\mathrm{AAA}_{\mathrm{lys}}\) codon usage, such as Ccq1, Pgk1, Krs1, or Swi6, did not exhibit translation deficiencies during nitrogen starvation in the \(igo1\Delta\) mutant (Fig. 2a; Supplementary Fig. 5b,c).
+
+<--- Page Split --->
+
+To confirm the potential defect in \(\mathsf{U}_{34}\) tRNA modification in the \(igo1\Delta\) mutant, we employed quantitative liquid chromatography coupled to mass spectrometry (LC- MS) \(^{8}\) to analyse tRNAs extracted from wild- type and \(igo1\Delta\) cells. Samples were collected during exponential growth and at 2 and 4 hours of nitrogen starvation. This analysis revealed an increase in mcm \(^{5s}2U_{34}\) levels in wild- type cells upon entry into quiescence. However, in the \(igo1\Delta\) mutant, the levels of mcm \(^{5s}2U_{34}\) diminished after 4 hours of nitrogen starvation (Fig. 5a,b). Furthermore, this analysis also indicated a reduction in \(\mathsf{A}_{37}\) N6- threonylcarbamoyladenosine (t \(^{6}\) ) modification in the \(igo1\Delta\) mutant (Fig. 5a; Supplementary Fig. 6a). The t \(^{6}\mathsf{A}_{37}\) tRNA modification, present in Archaea and Eukarya, mediated by the protein Sua5 and the KEOPS/EKC complex, is essential for cell growth and accurate translation \(^{44 - }\) . Once again, we hypothesised that differences in \(\mathsf{AAA}_{\mathsf{lys}}\) codon usage might be responsible for the reduction in t6 \(\mathsf{A}_{37}\) modification in the \(igo1\Delta\) mutant. When we examined the use of \(\mathsf{AAA}_{\mathsf{lys}}\) versus \(\mathsf{AA}\mathsf{G}_{\mathsf{lys}}\) codons for Sua5 and the KEOPS/EKC subunits, we found that Pcc1 and Cgi121 (two components of KEOPS/EKC complex) primarily use the \(\mathsf{AAA}_{\mathsf{lys}}\) codons (Supplementary Fig. 6b). Taking Cgi121 as an example, we detected a significant reduction in Cgi121 level in the \(igo1\) mutant during nitrogen starvation (Supplementary Fig. 6c). These data suggest that a defect in the translation of Cgi121 protein could be responsible of the reduction in t \(^{6}\mathsf{A}_{37}\) tRNA modification in the \(igo1\Delta\) mutant. Both mcm \(^{5s}2U_{34}\) and t \(^{6}\mathsf{A}_{37}\) modifications are involved in decoding codons that start with adenosine, promoting codon- anticodon pairing and enhancing translation fidelity \(^{47,48}\) . These findings provide a molecular explanation for the paromomycin hypersensitivity and translation defect observed in the \(igo1\Delta\) mutant under nitrogen- stress conditions (Fig. 4e).  
+
+As cells enter quiescence, a notable reduction in tRNA \(^{Lys}\) UUU levels was observed in both wild- type and \(igo1\Delta\) cells (Supplementary Fig. 5d), suggesting that this tRNA becomes limiting in quiescent cells. Previous studies in yeast and worms have shown that over- expression of tRNAs can effectively restore translation rates and protein homeostasis in mutants defective in tRNA modification \(^{8,12,13,39,47}\) . Using Rap1 as an example, we assessed whether over- expression of tRNA \(^{Lys}\) UUU would lead to recovery of Rap1 protein levels in the \(igo1\Delta\) mutant, with tRNA \(^{Lys}\) CUU over- expression serving as a control. As anticipated, tRNA \(^{Lys}\) CUU
+
+<--- Page Split --->
+
+overexpression had no impact on Rap1 levels, while overexpression of tRNA \(^{lys}_{uuu}\) partially restored Rap1 protein levels (Fig. 5c, d). To confirm that the reduced levels of Rap1 protein in \(igo1\Delta\) cells resulted from defective translation of its mRNA, we engineered a mutant version of the rap1 gene in which the 40 AAA codons were substituted with AAG, making all lysine codons independent of tRNA modification. Consistent with previous findings for other proteins, the translation deficiency of Rap1 in the \(igo1\Delta\) mutant was completely rescued by expressing the rap1- allAAG allele (Fig. 5e).  
+
+In summary, considering all the data, we conclude that during the transition into quiescence, the Endosulfine Igo1 is necessary for facilitating U34 and A37 tRNA modifications, which are critical for enhancing the translation efficiency and fidelity of proteins encoded by mRNAs with high AAA \(_{lys}\) codon usage.  
+
+## Gad8 phosphorylation is required to enhance translation during quiescence entry  
+
+We then investigated the underlying molecular mechanism that triggers the translational defect due to a failure in tRNA modification. We identified Gad8 as an interactor of PP2A/Pab1 (Supplementary Fig. 4c). Previous studies demonstrated that PP2A/Pab1 counteracts the phosphorylation of Gad8 by TORC2 at serine 546, suggesting that Gad8 is a direct target of PP2A/Pab1 \(^{6,7}\) . Additionally, active Gad8 phosphorylated at S546 activates Elongator by inhibiting Gsk3, a glycogen synthase kinase that inhibits Elongator by phosphorylating the Elp4 subunit at serine 114 \(^{8}\) .  
+
+These findings prompted us to investigate the potential role of Igo1 in Gad8 phosphorylation during quiescence entry. Notably, it has previously been reported that deletion of \(elp3\) , which encodes the tRNA acetyltransferase subunit of the Elongator complex, leads to a reduction in Gad8 protein levels \(^{8}\) . A sequence analysis of the gad8 mRNA revealed a high usage of AAA \(_{lys}\) codons, like rap1 (Supplementary Fig. 5b, gad8 z- scoreAAA/AAG = 0.73/- 0.73 vs. rap1 z- scoreAAA/AAG = 0.72/- 0.71). This observation led us to investigate whether high PP2A/B55 activity, in the absence of Igo1, could be relevant for maintaining Gad8 protein levels during nitrogen starvation. Western blot analysis clearly showed a
+
+<--- Page Split --->
+
+decrease in Gad8 protein levels in the \(igo1\Delta\) mutant (Fig. 6a). Furthermore, the phosphorylation status of Gad8 at S546 was also reduced in the absence of Igo1, compared to wild- type cells (Fig. 6b). These results suggest that during nitrogen starvation, inhibition of PP2A/Pab1 leads to the accumulation of phosphorylated Gad8 at S546 by TORC2 and consequent activation of the Elongator complex. This mechanism generates a positive feedback loop that enhances translation of Gad8 and promotes more \(U_{34}\) tRNA modifications (see Fig. 7).  
+
+If our model is correct, and PP2A/B55 indeed regulates Elongator activity through Gad8 protein homeostasis, the deletion of the \(gad8\) gene should lead to a substantial decrease in proteins with high \(\mathrm{AAA_{lys}}\) codon usage. To test this hypothesis, we examined the levels of Sgo2, a protein with a very high \(\mathrm{AAA_{lys}}\) codon usage (z- scoreAAA/AAAG = 1.28/- 1.28), in a \(gad8\Delta\) mutant background. Western blot analysis showed a reduction in Sgo2 protein levels in the \(gad8\Delta\) mutant (Fig. 6c). As a positive control, we used the \(elp3\Delta\) mutant, where the reduction in Sgo2 levels was even greater (Fig. 6d). Moreover, both \(gad8\Delta\) and \(elp3\Delta\) mutants displayed sensitivity to paromomycin in both nitrogen- rich and nitrogen- poor media (Supplementary Fig. 6d), although the sensitivity to paromomycin in nitrogen- poor medium was more pronounced in the \(igo1\Delta\) mutant (Figs. 6e; Supplementary Fig. 6d). Thus, our data strongly suggests that defects in the Elongator activation pathway led to decreased translation efficiency of mRNAs with high \(\mathrm{AAA_{lys}}\) codon usage.  
+
+Finally, to confirm the connection between Igo1 and the Elongator activation pathway, we generated a double mutant, \(igo1\Delta\) gsk3Δ. Gsk3 is an Elongator inhibitor in S. pombe 8. As mentioned earlier, the \(igo1\Delta\) mutant exhibited sensitivity to paromomycin, particularly in nitrogen- poor media containing phenylalanine (Fig. 4e). If this phenotype is indeed related to reduced Elongator activity, the \(igo1\Delta\) gsk3Δ double deletion should rescue this defect. Wild- type, \(igo1\Delta\) and gsk3Δ single mutants, along with the \(igo1\Delta\) gsk3Δ double mutant, were cultivated in nitrogen- poor (MMPhe) medium with or without paromomycin, and their growth phenotype was assessed. The \(igo1\Delta\) mutant displayed hypersensitivity to paromomycin, while the gsk3Δ mutant exhibited mild resistance compared to the wild- type strain. A partial improvement in cell growth in the presence of paromomycin was detected in the \(igo1\Delta\) gsk3Δ double mutant
+
+<--- Page Split --->
+
+439 clones compared to the \(igo1\Delta\) mutant (Fig. 6e). These findings strongly suggest that proper Igo1- mediated activation of Elongator is crucial for maintaining the rate of translation during quiescence entry.
+
+<--- Page Split --->
+
+## Discussion  
+
+Entry into quiescence in yeasts is regulated by diverse signalling cascades that converge at the Greatwall- Endosulfine- PP2A/B55 pathway \(^{3,5,49 - 51}\) . In S. pombe, the primary signal regulating entry into quiescence is nitrogen starvation, which reduces the activity of TORC1 and increases the activity of TORC2 \(^{4,6,7,52 - 58}\) . Inactivation of TORC1 results in reduced protein synthesis and the activation of protein degradation through autophagy. However, quiescent cells must maintain a continuous supply of specific proteins to remain viable.  
+
+In this study, we demonstrate that the Greatwall- Endosulfine- PP2A/B55 pathway links the inactivation of TORC1 with the activation of TORC2 signalling to promote the activation of the Elongator complex and other tRNA modification complexes essential for sustaining the translation programme during quiescence. This is achieved by facilitating \(\mathrm{U}_{34}\) and \(\mathrm{A}_{37}\) tRNA modifications, which increase translation efficiency and fidelity of critical proteins, including those necessary for telomeric and subtelomeric functions.  
+
+The reduction of PP2A/B55 activity, achieved through the activation of Greatwall and Endosulfine, is required to accumulate phosphorylated Gad8 at S546 \(^{7}\) . This phosphorylation event increases the activity of the Elongator complex by inhibiting glycogen synthase kinase, Gsk3 \(^{8}\) . The increased Elongator activity promotes the efficient translation of mRNAs containing high \(\mathrm{AAA}_{\mathrm{lys}}\) codon usage, such as tsc2 (an inhibitor of TORC1) \(^{8}\) , gad8 (a positive effector of TORC2), trm112, ctu1 and cgi121 (involved in \(\mathrm{U}_{34}\) and \(\mathrm{A}_{37}\) tRNA modifications). All of these facilitate the switch from high TORC1 to high TORC2 activity as cells enter quiescence. Furthermore, the synthesis of key proteins with roles in telomeric and subtelomeric organisation, such as Rap1, Sgo2, Clr2 or Clr3, which also exhibit a high \(\mathrm{AAA}_{\mathrm{lys}}\) codon usage, is dependent on the correct activation of Elongator (Fig. 7).  
+
+Previous studies have demonstrated that the deletion of rap1 (encoding a component of the shelterin complex) or bqt4 (encoding a component of the bouquet complex) causes telomeric detachment from the NE \(^{23,25,26}\) . Our findings suggest that the telomeric detachment defect in the igo1Δ mutant is probably caused by a reduction in Rap1 protein levels (Fig. 2a- d). Interestingly, Bqt4, the other protein that creates a molecular link between telomeres and the NE, has a
+
+<--- Page Split --->
+
+low AAAIys codon usage (z- scoreAAA/AAG = - 0.94/0.95), making it unlikely to be responsible for the telomeric detachment phenotype in the \(igo1\Delta\) mutant. However, we cannot rule out the possibility that other proteins may be involved in the telomeric attachment to the NE, such as Lem2, a member of the Lap2/Emerin/Man1 (LEM) family of lamin- associated proteins, which is known to be involved in telomere anchoring and heterochromatic gene silencing \(^{59 - 61}\) . Lem2 mRNA has a high AAAIys codon usage (z- scoreAAA/AAG = 0.88/- 0.88), suggesting that it may also be subject to translation defects in the \(igo1\Delta\) mutant, which could potentially affect telomere attachment to the NE.  
+
+In addition to the telomeric anchoring defect, the \(igo1\Delta\) mutant exhibited upregulation of genes located in the subtelomeric regions of chromosomes I and II (Fig. 1a; Supplementary Fig. 1a). Several proteins related to subtelomeric organisation were defective in the \(igo1\Delta\) background, particularly Sgo2, whose protein levels were significantly reduced (Fig. 3a). Sgo2 plays a role in \(knobs\) assembly, and its deletion leads to derepression of genes located in subtelomeric regions \(^{21,22}\) . Our data demonstrate that Igo1 is involved in \(knobs\) assembly, and its deletion results in defects in the regulation of subtelomeric genes. The overlapping roles and phenotypes between Sgo2 and Igo1 suggest that the reduction in Sgo2 protein levels is responsible for the derepression of subtelomeric genes in the \(igo1\Delta\) mutant. However, other proteins, such as Clr2 or Clr3, which are involved in silencing could also contribute to this phenotype.  
+
+All the proteins tested in our study had their levels restored after reducing PP2A/B55 activity, indicating that low PP2A/B55 activity was necessary to maintain protein homeostasis during quiescence entry. But what is the link between PP2A/B55 activity, translation and the telomeric/subtelomeric organisation?  
+
+Previous research has shown that deleting specific components of TORC2 signalling, such as tor1 or gad8, leads to a significant derepression of genes located in subtelomeric regions \(^{62}\) . However, the molecular mechanism linking these processes remained unclear. Our data reveal that not only components of the TORC2 complex but also elements acting downstream of the TORC1 complex, such as Greatwall (Ppk18 and Cek1) or Endosulfine (Igo1), play a crucial role in regulating subtelomeric gene expression. These findings establish
+
+<--- Page Split --->
+
+a connection between the Greatwall-Endosulfine- PP2A/B55 pathway and telomeric/subtelomeric organisation.  
+
+A link between the TORC2 signalling pathway, the Elongator complex, and tRNA modifications has been demonstrated 8. Our data further demonstrate that this connection is particularly relevant during entry into quiescence by elucidating the molecular details linking nitrogen starvation, TORC1 inactivation, activation of Greatwall- Endosulfine, inactivation of PP2A/Pab1, activation of Gad8, and upregulation of tRNA- modifying complexes (Fig. 7). Indeed, these molecular events affect not only the Elongator complex but also other tRNA modifiers like Trm112 or Ctu1. These interactions create positive feedback loops that are responsible for the transition from high TORC1 to high TORC2 activity, subsequently leading to further tRNA modifications. Mutations in Greatwall (Ppk18 and Cek1) or Endosulfine (igo1) result in a defect in this transition, causing a failure in tRNA modifications that affect the translation of mRNAs with a high AAAlys codon usage encoding for proteins such as Rap1, Sgo2, Clr2 and Clr3. This, in turn, triggers telomeric detachment and the derepression of subtelomeric genes.  
+
+In addition to the mcm \(^5\) s \(^2\) U \(_{34}\) defect, we also observed a reduction in t \(^6\) A \(_{37}\) in the Endosulfine (igo1 \(\Delta\) ) mutant (Fig. 5a). These two closely located tRNA modifications play a crucial role in ensuring accurate codon- anticodon interactions by stabilising codon- anticodon pairings \(^{63 - 66}\) . Interestingly, tRNA \(^{Lys}UUU\) carries both the mcm \(^5\) s \(^2\) U \(_{34}\) and t \(^6\) A \(_{37}\) modifications, both of which are defective in the Endosulfine (igo1 \(\Delta\) ) mutant. This finding provides an explanation for the strong sensitivity of igo1 \(\Delta\) cells to paromomycin compared to mutants in Elongator (elp3 \(\Delta\) ) or Gad8 (gad8 \(\Delta\) ). The defect in the t \(^6\) A \(_{37}\) modification could be explained by a reduction in Cgi121 protein levels (Supplementary Fig. 6c), one of the subunits of the KEOPS complex responsible for this modification \(^{48,64}\) , which is encoded by an mRNA with a high AAAlys codon usage.  
+
+Finally, an interesting concept has emerged over the last few years on tRNA modifications and stress, known as tRNA modification tuneable transcripts (MoTTs). These transcripts are characterised by a specific use of degenerate codons and codon biases to encode essential stress response proteins. Translation of these transcripts is affected by modifications at the wobble position
+
+<--- Page Split --->
+
+of the tRNAs \(^{14,15}\) . Our work supports the idea that mRNAs encoding proteins involved in nutrient starvation, stress response or even the translation of viral RNA genomes present MoTTs, which allow for an increase in translation efficiency under stress conditions \(^{8,12,13,67 - 70}\) . One of these transcripts, the Hif1α mRNA, is translated in drug-resistant melanomas through a mechanism involving the activation of Elp1 by the PI3 kinase signalling pathway. The activation of Hif1α promotes drug resistance by inducing anaerobic glycolysis \(^{71}\) . Therefore, the mcm \(^{5}\) s \(^{2}\) U \(_{34}\) modification in the tRNA anticodon promotes the translation of mRNAs enriched in AAA codons, including Hif1α mRNA. This discovery opens new avenues for identifying inhibitors of Elongator and other tRNA modifiers to treat drug-resistant tumours and combat viral infections.
+
+<--- Page Split --->
+
+## Methods  
+
+## Strains and Growth Conditions  
+
+Yeast strains are listed in Supplementary Table 6. Fission yeast cells were cultured and genetically manipulated according to standard protocols 72. Genetic crosses were performed on malt extract agar plates. Cells were typically cultured overnight at the appropriate temperatures in yeast extract supplemented with adenine, leucine, histidine, lysine, and uracil (YES), or in Edinburgh minimal medium containing \(93.5 \text{mM}\) ammonium chloride (EMM2) as a nitrogen source. For nitrogen starvation experiments, exponentially growing cells were shifted from Edinburgh minimal medium (EMM2) at \(28^{\circ}\text{C}\) to minimal medium without nitrogen (EMM2- N) at \(25^{\circ}\text{C}\) .  
+
+In overexpression experiments using the \(nmt1^{+}\) promoter, cells were grown to the logarithmic phase in EMM2 containing \(15 \mu \text{M}\) thiamine. Then, the cells were harvested and inoculated in fresh EMM2 medium without thiamine.  
+
+## DNA Techniques and Plasmid Construction  
+
+DNA Manipulations were performed as described in Sambrook, J., Fritsch, E.F., and Maniatis, T. (1989). Enzymes for molecular biology were obtained from Fermentas and Thermo Fisher. PCRs were performed with using Velocity DNA polymerase (Bioline). Oligonucleotides employed for strain and plasmid construction are listed in Supplementary Table 7. Information regarding construction strategies is available upon request. Plasmids used in this study carry the ampicillin resistance gene for selection in \(E\) coli and are listed in Supplementary Table 7.  
+
+## RNA Isolation, RNAseq and RT-qPCR  
+
+Wild type (2666), \(igo1\Delta\) (2727) and \(ppk18\Delta\) cek1\(\Delta\) (2883) cells were grown to mid- exponential phase in EMM2, centrifuged and washed three times in EMM2- N, and cultured in EMM2- N at \(25^{\circ}\text{C}\) . For RNAseq, \(2 \times 10^{8}\) cells were harvested at times 0 and 4 hours, washed with cold DEPC- \(H_{2}O\) , and snap frozen. RNA extraction was carried out by disrupting the cells with glass beads using RNAeasy Mini kit (Qiagen) and following the manufacturer's instructions. RNA quality was evaluated using the Bioanalyzer 2100 (Agilent). Library preparation, using the
+
+<--- Page Split --->
+
+Illumina Ribo Zero and TruSeq Stranded kits, and subsequent NGS sequencing were performed by Macrogen. Sequencing quality was assessed with FastQC (v 0.11.8, Babraham Bioinformatics). If necessary, adaptors were trimmed using Trimmomatic (v 0.38) \(^{73}\) . Alignment was performed with HISAT2 v 2.1.0 (CCB in John Hopkins University) \(^{74}\) using S. pombe reference genome from Pombase (downloaded on 30/11/2018). Samtools (v 1.9) and deepTools (v 3.3.0) were used to obtain bigWig files to visualize in IGV (v 2.4.16) and JBrowse (v 1.15.4) browsers. Read counts were obtained with featureCounts (Subread package v 1.6.3, Walter+Eliza Hall Bioinformatics) \(^{75}\) . DESeq2 (v1.22.2) \(^{76}\) was used for the differential expression analysis. Plots representing upregulated genes in Endosulfine (igo1Δ) and Greatwall (cek1Δ ppk18Δ) mutants shown in Fig. 1a and Supplementary Fig. 1a were generated with karyoploteR (v 1.12.4) \(^{77}\) .  
+
+For RT- qPCR, total RNA was isolated from 2x10 \(^{8}\) S. pombe cells in exponential phase by disrupting the cells with glass beads in TRIzol® Reagent (Invitrogen) and following the manufacturer's instructions. The integrity of the RNA was verified through 1% agarose gel electrophoresis, and its quality and quantity were determined using a microspectrophotometer. RNA was treated with RNase- free DNase I (Invitrogen) at 25°C for 30 minutes, following the manufacturer's instructions. Each RNA sample (1.2- 1.5 μg) was then reverse transcribed with the SuperScript™ First- Strand Synthesis System (Invitrogen) using the oligo(dT) primer supplied with the kit or the tRNA \(^{Lys}\) UUU specific reverse primer in combination with the act1 gene reverse primer (Supplementary Table 7) at 50°C for 30 minutes in a 20- µl total volume. Quantitative PCR amplification of cDNA (1 µl) was carried out using TB Green Premix Ex Taq™ (TaKaRa) and the primer pairs indicated in the Supplementary Table 7, in a 20- µl total volume with the following cycling parameters: 95°C for 45 seconds, 40 cycles of 95°C for 5 seconds and 60°C for 31 seconds, followed by a dissociation step at 95°C for 15 seconds, 60°C for 1 minute and 95°C for 15 seconds. The reactions were run in duplicate or triplicate in an Applied Biosystems 7300 Real- Time PCR System. Negative controls without reverse transcriptase, without RT- primer and without cDNA were included to control for DNA contaminations. Fold changes in the expression levels relative to the wild- type strain grown in EMM2 were calculated according to the mathematic model described by \(^{78}\) , with normalization to act1
+
+<--- Page Split --->
+
+expression levels. The experiments were performed at least twice with cDNA from different biological repeats.  
+
+## SRFF microscopy  
+
+Samples were observed using a Confocal Andor Dragonfly 200 microscope, equipped with a 100x/1.45 Oil Plan Apo objective, an Andor sCMOS Sona 4.2B- 11 camera and controlled by Fusion (SRRF- STREAM) software. Image J software was used for general image and movie manipulation. Radial Profile Analysis and the calculations of Pearson's Correlation Coefficients were performed using ImageJ. More than 100 nuclei were measured for each strain.  
+
+## S. pombe protein extracts and western blot  
+
+TCA extraction was performed as previously described 79. For immunoblotting, PVDF membranes were probed with anti- HA (12CA5, Roche), anti- GFP (3H9, Chromotek), anti- GST (RPN1236V, Cytiva) or anti- P- Gad8 (kindly provided by José Cansado, University of Murcia, Spain). Standard procedures were employed for protein transfer, blotting and chemiluminescence detection. Protein detection was performed using the ECL kit (BioRad).  
+
+## Chromatin immunoprecipitation (ChIP)  
+
+Chromatin isolation and immunoprecipitation was performed as previously described 80. S. pombe cell cultures were grown in EMM2 or after 4 hours of nitrogen starvation in EMM2- N to \(\mathrm{OD}_{600}\) of 0.5- 0.6 and crosslinked with 1% formaldehyde for 10 min at room temperature. To terminate crosslinking, 2.5M glycine was added to a final concentration of 125 mM for 5 min. Cells were pelleted by centrifugation, washed twice with 10 ml of cold PBS, frozen on dry ice and stored at \(- 80^{\circ}\mathrm{C}\) . Cell pellets from 50 ml cultures were resuspended in 0.25 ml of Breaking buffer (0.1M Tris- HCl pH 8.0, 20% glycerol, 1 mM PMSF) and lysed in a Fast- prep (2 cycles of 45 s) in the presence of glass beads (50 micron; Sigma) at \(4^{\circ}\mathrm{C}\) . Lysates were centrifuged at 14,000 g for 1 min at \(4^{\circ}\mathrm{C}\) . Pellets were washed with 1 ml of Lysis buffer (50 mM HEPES pH 7.6, 140 mM NaCl, 1 mM EDTA, 1% Triton X- 100, 0.1% sodium deoxycholate, 0.1% SDS, 1 mM PMSF). Pellets containing chromatin were resuspended in 0.25 ml of Lysis buffer. Lysates were sonicated for 6 min at \(4^{\circ}\mathrm{C}\) (30 seconds on, 30 seconds off), using
+
+<--- Page Split --->
+
+a water bath sonicator (Diagenode Bioruptor Plus), transferred to new 1.5- ml Eppendorf tubes, added 0.75 ml of Lysis Buffer and centrifuged at 14,000 g for 30 min at \(4^{\circ}C\) . 50 \(\mu\) l of supernatant was kept as 'input' and the remainder ( \(\sim 950\) \(\mu\) l) was subjected to immunoprecipitation with antibodies against K14- acetylated histone H3 (07- 353, Upstate Biotechnology) and 20 \(\mu\) l of protein G agarose beads (100.04D, Dynabeads Protein G, Thermo Fisher). After an overnight incubation at \(4^{\circ}C\) with mixing, beads were washed sequentially with 1 ml of Lysis buffer once, Lysis + 500 mM NaCl twice, Wash buffer (10 mM Tris pH 8.0, 1 mM EDTA, 250 mM LiCl, 0.5% sodium deoxycholate, 0.5% NP- 40, 1 mM PMSF) twice, and TE buffer (10 mM Tris pH 7.5, 1 mM EDTA) once. Each wash was for 5 min with mixing at room temperature. Immune complexes were eluted in 100 \(\mu\) l elution buffer (50 mM Tris pH 8.0, 10 mM EDTA, 1% SDS) at \(65^{\circ}C\) for 20 min. Beads were washed with 150 \(\mu\) l TE + 0.67% SDS, which was combined with the eluate. 150 \(\mu\) l TE + 0.67% SDS was also added to the input samples, and both IP and input samples were incubated at \(65^{\circ}C\) overnight to reverse protein- DNA crosslinks. DNA was purified by phenol/chloroform extraction. Analysis by qPCR was carried out using a Bio- Rad CFX96 instrument, Takara TB Green premix Ex- Taq, and primers listed in the Key resources table. ChIP signals were calculated as IP/input and normalized to WT Oh with an assigned value of 1.  
+
+## Immunoprecipitations and mass-spectrometry analysis  
+
+Immunoprecipitation was performed following previously established protocols \(^{81}\) . For the immunoprecipitation of Paa1- L- YFP (strain JED62) and Pab1- L- YFP (strain JED56), 1 l of cells was grown to mid- log phase in EMM2, then shifted to EMM2- N for 1 hour and crosslinked with 1% formaldehyde for 10 min at \(25^{\circ}C\) . The reaction was quenched by adding glycine to 250 mM and incubating for 5 min on ice. Cells were collected by centrifugation, washed with PBS 1x, frozen in liquid nitrogen and broken with a Freezer/Mill in lysis buffer (25 mM Tris HCl pH 7.5, 150 mM NaCl, 0.5% SDS, 1% NP40, 1 mM PMSF, 1 \(\mu\) g/ml aprotinin, 1 \(\mu\) g/ml leupeptin, 1 \(\mu\) g/ml pepstatin). Cell lysates were then slowly diluted to 0.1% SDS final concentration for immunoprecipitation in lysis buffer without SDS at \(4^{\circ}C\) for 30 min. Clarified extracts were immunoprecipitated by adding 40 \(\mu\) l of GFP- Trap beads (gta- 20, Chromotek) for 1 hour at \(4^{\circ}C\) . The beads were washed six times
+
+<--- Page Split --->
+
+with lysis buffer containing 500 mM NaCl. Finally, the beads were sent to the Proteomics Facility of the Salamanca Cancer Research Center for Massspectrometry analysis. Analysis and interpretation of the results were carried out using the String Database.  
+
+For the immunoprecipitation of GFP- Pab1 (strain 2895), 16 l of cells were grown to mid- log phase in EMM2 at \(25^{\circ}C\) . After one day, half of the culture was shifted to EMM2- N for 1 hour. Subsequently, cells were harvested by centrifugation and frozen at \(- 80^{\circ}C\) . Cells were lysed using glass beads and a bead beater in \(100~\mathrm{ml}\) of NP- 40 buffer (6 mM \(\mathrm{Na_2HPO_4}\) , 4 mM \(\mathrm{NaH_2PO_4}\) , \(1\%\) NP- 40, \(150~\mathrm{mM~NaCl}\) , 2 mM EDTA, \(50~\mathrm{mM~NaF}\) and \(0.1\mathrm{mMNa_3VO_4}\) ) supplemented with complete EDTA- free protease inhibitor cocktail (Roche), \(1.3\mathrm{mM}\) benzamidine (Sigma) and \(1\mathrm{mM}\) PMSF (Sigma). Lysates were cleared by centrifugation, and supernatants were mixed with \(60~\mu \mathrm{l}\) of \(50\%\) slurry GFP- TRAP magnetic agarose beads (GFP- Trap® magnetic agarose, ChromoTek) equilibrated with NP- 40 buffer. After 90 minutes of incubation at \(4^{\circ}C\) , beads were magnetically separated from lysates and washed twice with \(5\mathrm{ml}\) of NP- 40 buffer. Samples were washed with \(5\mathrm{ml}\) of low- NP- 40 buffer ( \(0.02\%\) NP- 40) to reduce total detergent in purified proteins and subsequently resuspended in \(1\mathrm{ml}\) of low- NP- 40 buffer. Proteins were eluted twice with \(150~\mu \mathrm{l}\) of elution buffer ( \(200~\mathrm{mM}\) glycine- HCl pH 2.5) and precipitated for 30 minutes on ice using \(100~\mu \mathrm{l}\) of \(100\%\) TCA. Samples were then spun down for 30 minutes at \(13000~\mathrm{rpm}\) and \(4^{\circ}C\) , washed with \(1\mathrm{ml}\) of cold acetone containing \(0.05\mathrm{NHCl}\) and \(1\mathrm{ml}\) of cold acetone. Finally, pellets were dried at room temperature and stored at \(4^{\circ}C\) for mass spectrometry analysis. A small amount of each sample was used to confirmed proper purification of GFP- tagged proteins. For that purpose, the Plus One Silver Staining protein kit (GE Healthcare) was employed following manufacturer instructions. TCA- precipitated proteins were digested with trypsin and analyzed by two- dimensional liquid chromatography tandem MS (2D- LC- MS/MS) as previously described \(^{82}\) . MS2 and MS3 spectra were extracted separately from RAW files, and converted to DTA files using Scansifter software \(^{83}\) (v2.1.25). Spectra with less than 20 peaks were excluded and the remaining spectra were searched using the SEQUEST algorithm (Thermo Fisher Scientific, San Jose, CA, USA; version 27, rev. 12). Sequest was set up to search the S. pombe protein database
+
+<--- Page Split --->
+
+(pombe_contams_20151012_rev database, created in October 2015 from pombase.org). Common contaminants were added, and all sequences were reversed to estimate the false discovery rate (FDR), yielding 10390 total entries. Variable modifications (C+57, M+16, [STY]+80 for all spectra and [STY]- 18 for MS3), strict trypsin cleavage, \(< 10\) missed cleavages, fragment mass tolerance: 0.00 Da (because of rounding in SEQUEST, this results in 0.5 Da tolerance), and parent mass tolerance: 2.5 Da were allowed. Peptide identifications were assembled and filtered in Scaffold (v4.8.4, Proteome Software, Portland, OR) using the following criteria: minimum of \(99.0\%\) protein identification probability; minimum of two unique peptides; minimum of \(95\%\) peptide identification probability. FDRs were estimated in Scaffold based on the percentage of decoy sequences identified after using the above filtering criteria; the protein level FDR was \(0.7\%\) and the peptide level FDR was \(0.3\%\) . Proteins containing the same or similar peptides that could not be differentiated based on MS/MS alone were grouped to satisfy the principles of parsimony. Mass spectrometry identified proteins were exported from Scaffold to Excel for further analysis. Further analysis and interpretation of the results was carried out using the String Database (https://string-db.org/).  
+
+## tRNA purification  
+
+tRNA purification assay was performed following established procedures 8.  
+
+## Quantification of tRNA modifications  
+
+The purified tRNAs (500 ng per sample) were subjected to hydrolysis in a \(40~\mu \mathrm{L}\) digestion cocktail containing 10 U benzonase, 4 U calf intestinal alkaline phosphatase, 0.12 U phosphodiesterase I, 0.1 mM deferoxamine, 0.1 mM butylated hydroxytoluene, 4 ng pentosatin, \(2.5\mathrm{mM}\mathrm{MgCl}_2\) and \(5\mathrm{mM}\) tris buffer (pH 8.0). The digestion mixture was incubated at \(37^{\circ}C\) for \(6\mathrm{h}\) . For the verification of HPLC retention times of RNA modifications, synthetic standards were employed. Analytical separation was facilitated by a Thermo Hypersil Gold aQ C18 column ( \(100\times 2.1\mathrm{mm}\) , \(1.9\mu \mathrm{m}\) ), which was interfaced with an Agilent 1290 HPLC system and an Agilent 6495 triple quadrupole mass spectrometer. The employed LC system operated at \(35^{\circ}C\) , maintaining a flow rate of \(0.35\mathrm{mL / min}\) . The gradient starts with \(100\%\) solution A ( \(0.1\%\) formic acid in water) for 4 min,
+
+<--- Page Split --->
+
+followed by a 4- 15 min phase involving a transition from \(0\%\) to \(20\%\) solution B \((0.1\%)\) formic acid in acetonitrile). The HPLC column was coupled with an Agilent 6495 triple quadrupole mass spectrometer, utilizing an electrospray ionization source in positive ion mode. Operational parameters were set as follows: gas temperature at \(120^{\circ}C\) ; gas flow rate at \(11L / min\) ; nebulizer pressure at \(20psi\) ; sheath gas temperature at \(400^{\circ}C\) ; sheath gas flow rate at \(12L / min\) ; capillary voltage at \(1500V\) ; and nozzle voltage maintained at \(0V\) . The dynamic multiple reaction monitoring mode was used for detection of product ions derived from their respective precursor ions for all the RNA modifications. The collision energy was optimized to ensure maximal detection sensitivity for each modification. To ensure the same sample input, the MS signal intensity for each ribonucleoside was normalized with the UV signal intensity of canonical ribonucleosides. The fold change of the modified ribonucleosides in experiment group was calculated relative to the control group.  
+
+## Drug sensitivity assays  
+
+For survival on agar plates, S. pombe strains were cultured in YES, diluted and the cells were spotted onto plates with minimal medium containing \(93.5mM\) \(\mathrm{NH_4Cl}\) (EMM2) or \(20mM\) phenylalanine (MMPhe) without or with paromomycin \((0.5mg / ml)\) , puromycin \((0.5mg / ml)\) or cycloheximide \((2.5\mu g / ml)\) . The plates were then incubated at the indicated temperatures for 4- 8 days.  
+
+## Statistical methods  
+
+Average, standard deviation, and P values for the two- sided Student's t test of statistically significant differences were calculated with Microsoft Excel. Data distribution was assumed to be normal, but this was not formally tested.  
+
+## Data availability  
+
+Further information and requests for reagents may be directed to Sergio Moreno (smo@usal.es) or Javier Encinar del Dedo (jodel_dedo@usal.es).
+
+<--- Page Split --->
+
+787 1 Gonzalez, A. & Hall, M. N. Nutrient sensing and TOR signaling in yeast and mammals. EMBO J 36, 397- 408, doi:10.15252/embj.201696010 (2017). 791 2 Gonzalez, A., Hall, M. N., Lin, S. C. & Hardie, D. G. AMPK and TOR: The Yin and Yang of Cellular Nutrient Sensing and Growth Control. Cell Metab 31, 472- 492, doi:10.1016/j.cmet.2020.01.015 (2020). 794 3 Bontron, S. et al. Yeast endosulfines control entry into quiescence and chronological life span by inhibiting protein phosphatase 2A. Cell Rep 3, 16- 22, doi:10.1016/j.celrep.2012.11.025 (2013). 797 4 Chica, N. et al. Nutritional Control of Cell Size by the Greatwall- Endosulfine- PP2A.B55 Pathway. Curr Biol 26, 319- 330, doi:10.1016/j.cub.2015.12.035 (2016). 800 5 Aono, S., Haruna, Y., Watanabe, Y. H., Mochida, S. & Takeda, K. The fission yeast Greatwall- Endosulfine pathway is required for proper quiescence/G(0) phase entry and maintenance. Genes Cells 24, 172- 186, doi:10.1111/gtc.12665 (2019). 804 6 Laboucarie, T. et al. TORC1 and TORC2 converge to regulate the SAGA co- activator in response to nutrient availability. EMBO Rep 18, 2197- 2218, doi:10.15252/embr.201744942 (2017). 807 7 Martin, R. et al. A PP2A- B55- Mediated Crosstalk between TORC1 and TORC2 Regulates the Differentiation Response in Fission Yeast. Curr Biol 27, 175- 188, doi:10.1016/j.cub.2016.11.037 (2017). 810 8 Candiracci, J. et al. Reciprocal regulation of TORC signaling and tRNA modifications by Elongator enforces nutrient- dependent cell fate. Sci Adv 5, eaav0184, doi:10.1126/sciadv.aav0184 (2019). 813 9 Yarian, C. et al. Modified nucleoside dependent Watson- Crick and wobble codon binding by tRNALysUUU species. Biochemistry 39, 13390- 13395, doi:10.1021/bi001302g (2000). 816 10 Murphy, F. V. t., Ramakrishnan, V., Malkiewicz, A. & Agris, P. F. The role of modifications in codon discrimination by tRNA(Lys)UUU. Nat Struct Mol Biol 11, 1186- 1191, doi:10.1038/nsmb861 (2004). 819 11 Kruger, M. K., Pedersen, S., Hagervall, T. G. & Sorensen, M. A. The modification of the wobble base of tRNAGlul modulates the translation rate of glutamic acid codons in vivo. J Mol Biol 284, 621- 631, doi:10.1006/jmbi.1998.2196 (1998). 822 12 Garcia, P., Encinar Del Dedo, J., Ayte, J. & Hidalgo, E. Genome- wide Screening of Regulators of Catalase Expression: role of a transcription complex and histone and tRNA modification complexes on adaptation to stress. J Biol Chem 291, 790- 799, doi:10.1074/jbc.M115.696658 (2016). 827 13 Fernandez- Vazquez, J. et al. Modification of tRNA(Lys) UUU by elongator is essential for efficient translation of stress mRNAs. PLoS Genet 9, e1003647, doi:10.1371/journal.pgen.1003647 (2013). 830 14 Endres, L., Dedon, P. C. & Begley, T. J. Codon- biased translation can be regulated by wobble- base tRNA modification systems during cellular stress responses. RNA Biol 12, 603- 614, doi:10.1080/15476286.2015.1031947 (2015).
+
+<--- Page Split --->
+
+834 15 Gu, C., Begley, T. J. & Dedon, P. C. tRNA modifications regulate 835 translation during cellular stress. FEBS Lett 588, 4287- 4296, 836 doi:10.1016/j.febslet.2014.09.038 (2014). 837 16 Hermand, D. Anticodon Wobble Uridine Modification by Elongator at the 838 Crossroad of Cell Signaling, Differentiation, and Diseases. Epigenomes 4, 839 doi:10.3390/epigenomes4020007 (2020). 840 17 Doi, A. et al. Chemical genomics approach to identify genes associated 841 with sensitivity to rapamycin in the fission yeast Schizosaccharomyces 842 pombe. Genes Cells 20, 292- 309, doi:10.1111/gtc.12223 (2015). 843 18 Yadav, R. K., Matsuda, A., Lowe, B. R., Hiraoka, Y. & Partridge, J. F. 844 Subtelomeric Chromatin in the Fission Yeast S. pombe. Microorganisms 845 9, doi:10.3390/microorganisms9091977 (2021). 846 19 Kanoh, J. Unexpected roles of a shugoshin protein at subtelomeres. 847 Genes Genet Syst 92, 127- 133, doi:10.1266/ggs.17- 00016 (2018). 848 20 Hirano, Y., Asakawa, H., Sakuno, T., Haraguchi, T. & Hiraoka, Y. Nuclear 849 Envelope Proteins Modulating the Heterochromatin Formation and 850 Functions in Fission Yeast. Cells 9, doi:10.3390/cells9081908 (2020). 851 21 Matsuda, A. et al. Highly condensed chromatins are formed adjacent to 852 subtelomeric and decondensed silent chromatin in fission yeast. Nat 853 Commun 6, 7753, doi:10.1038/ncomms8753 (2015). 854 22 Tashiro, S. et al. Shugoshin forms a specialized chromatin domain at 855 subtelomeres that regulates transcription and replication timing. Nat 856 Commun 7, 10393, doi:10.1038/ncomms10393 (2016). 857 23 Maestroni, L. et al. Nuclear envelope attachment of telomeres limits 858 TERRA and telomeric rearrangements in quiescent fission yeast cells. 859 Nucleic Acids Res 48, 3029- 3041, doi:10.1093/nar/gkaa043 (2020). 860 24 Chikashige, Y. et al. Membrane proteins Bqt3 and - 4 anchor telomeres to 861 the nuclear envelope to ensure chromosomal bouquet formation. J Cell 862 Biol 187, 413- 427, doi:10.1083/jcb.200902122 (2009). 863 25 Inoue, H., Horiguchi, M., Ono, K. & Kanoh, J. Casein kinase 2 regulates 864 telomere protein complex formation through Rap1 phosphorylation. 865 Nucleic Acids Res 47, 6871- 6884, doi:10.1093/nar/gkz458 (2019). 866 26 Fujita, I. et al. Telomere- nuclear envelope dissociation promoted by Rap1 867 phosphorylation ensures faithful chromosome segregation. Curr Biol 22, 868 1932- 1937, doi:10.1016/j.cub.2012.08.019 (2012). 869 27 van Emden, T. S. et al. Shelterin and subtelomeric DNA sequences control 870 nucleosome maintenance and genome stability. EMBO Rep 20, 871 doi:10.15252/embr.201847181 (2019). 872 28 Kanoh, J., Sadaie, M., Urano, T. & Ishikawa, F. Telomere binding protein 873 Taz1 establishes Swi6 heterochromatin independently of RNAi at 874 telomeres. Curr Biol 15, 1808- 1819, doi:10.1016/j.cub.2005.09.041 875 (2005). 876 29 Harland, J. L., Chang, Y. T., Moser, B. A. & Nakamura, T. M. Tpz1- Ccq1 877 and Tpz1- Poz1 interactions within fission yeast shelterin modulate Ccq1 878 Thr93 phosphorylation and telomerase recruitment. PLoS Genet 10, 879 e1004708, doi:10.1371/journal.pgen.1004708 (2014). 880 30 Hu, X., Liu, J., Jun, H. I., Kim, J. K. & Qiao, F. Multi- step coordination of 881 telomerase recruitment in fission yeast through two coupled telomere- 882 telomerase interfaces. Elife 5, doi:10.7554/eLife.15470 (2016).
+
+<--- Page Split --->
+
+883 31 Vazquez-Bolado, A. et al. The Greatwall-Endosulfine Switch Accelerates 884 Autophagic Flux during the Cell Divisions Leading to G1 Arrest and Entry 885 into Quiescence in Fission Yeast. Int J Mol Sci 24, 886 doi:10.3390/ijms24010148 (2022). 887 32 Sugiyama, T. et al. SHREC, an effector complex for heterochromatic 888 transcriptional silencing. Cell 128, 491-504, doi:10.1016/j.cell.2006.12.035 889 (2007). 890 33 Job, G. et al. SHREC Silences Heterochromatin via Distinct Remodeling 891 and Deacetylation Modules. Mol Cell 62, 207-221, 892 doi:10.1016/j.molcel.2016.03.016 (2016). 893 34 Singh, N. S. et al. SIN-inhibitory phosphatase complex promotes Cdc11p 894 dephosphorylation and propagates SIN asymmetry in fission yeast. Curr 895 Biol 21, 1968-1978, doi:10.1016/j.cub.2011.10.051 (2011). 896 35 Bourgeois, G., Letoquart, J., van Tran, N. & Graille, M. Trm112, a Protein 897 Activator of Methyltransferases Modifying Actors of the Eukaryotic 898 Translational Apparatus. Biomolecules 7, doi:10.3390/biom7010007 899 (2017). 900 36 Liger, D. et al. Mechanism of activation of methyltransferases involved in 901 translation by the Trm112 'hub' protein. Nucleic Acids Res 39, 6249-6259, 902 doi:10.1093/nar/gkr176 (2011). 903 37 Agris, P. F., Vendeix, F. A. & Graham, W. D. tRNA's wobble decoding of 904 the genome: 40 years of modification. J Mol Biol 366, 1-13, 905 doi:10.1016/j.jmb.2006.11.046 (2007). 906 38 Karlsborn, T. et al. Elongator, a conserved complex required for wobble 907 uridine modifications in eukaryotes. RNA Biol 11, 1519-1528, 908 doi:10.4161/15476286.2014.992276 (2014). 909 39 Bauer, F. et al. Translational control of cell division by Elongator. Cell Rep 910 1, 424-433, doi:10.1016/j.celrep.2012.04.001 (2012). 911 40 Huang, B., Johansson, M. J. & Bystrom, A. S. An early step in wobble 912 uridine tRNA modification requires the Elongator complex. RNA 11, 424- 913 436, doi:10.1261/rna.7247705 (2005). 914 41 Bjork, G. R., Huang, B., Persson, O. P. & Bystrom, A. S. A conserved 915 modified wobble nucleoside (mcm5s2U) in lysyl-tRNA is required for 916 viability in yeast. RNA 13, 1245-1255, doi:10.1261/rna.558707 (2007). 917 42 Dewez, M. et al. The conserved Wobble uridine tRNA thiolase Ctu1-Ctu2 918 is required to maintain genome integrity. Proc Natl Acad Sci U S A 105, 919 5459-5464, doi:10.1073/pnas.0709404105 (2008). 920 43 Palmer, E., Wilhelm, J. M. & Sherman, F. Variation of phenotypic 921 suppression due to the psi+ and psi- extrachromosomal determinants in 922 yeast. J Mol Biol 128, 107-110, doi:10.1016/0022-2836(79)90311-5 923 (1979). 924 44 Deutsch, C., El Yacoubi, B., de Crecy-Lagard, V. & Iwata-Reuyl, D. 925 Biosynthesis of threonylcarbamoyl adenosine (t6A), a universal tRNA 926 nucleoside. J Biol Chem 287, 13666-13673, 927 doi:10.1074/jbc.M112.344028 (2012). 928 45 El Yacoubi, B. et al. A role for the universal Kae1/Qri7/YgjD (COG0533) 929 family in tRNA modification. EMBO J 30, 882-893, 930 doi:10.1038/emboj.2010.363 (2011).
+
+<--- Page Split --->
+
+931 46 El Yacoubi, B. et al. The universal YrdC/Sua5 family is required for the 932 formation of threonylcarbamoyladenosine in tRNA. Nucleic Acids Res 37, 933 2894- 2909, doi:10.1093/nar/gkp152 (2009). 934 47 Nedialkova, D. D. & Leidel, S. A. Optimization of Codon Translation Rates 935 via tRNA Modifications Maintains Proteome Integrity. Cell 161, 1606- 1618, 936 doi:10.1016/j.cell.2015.05.022 (2015). 937 48 Su, C., Jin, M. & Zhang, W. Conservation and Diversification of tRNA 938 t(6)A- Modifying Enzymes across the Three Domains of Life. Int J Mol Sci 939 23, doi:10.3390/ijms232113600 (2022). 940 49 Sarkar, S., Dalgaard, J. Z., Millar, J. B. & Arumugam, P. The Rim15- 941 endosulfine- PP2ACdc55 signalling module regulates entry into 942 gametogenesis and quiescence via distinct mechanisms in budding yeast. 943 PLoS Genet 10, e1004456, doi:10.1371/journal.pgen.1004456 (2014). 944 50 Luo, X., Talarek, N. & De Virgilio, C. Initiation of the yeast G0 program 945 requires Igo1 and Igo2, which antagonize activation of decapping of 946 specific nutrient- regulated mRNAs. RNA Biol 8, 14- 17, 947 doi:10.4161/rna.8.1.13483 (2011). 948 51 Talarek, N. et al. Initiation of the TORC1- regulated G0 program requires 949 Igo1/2, which license specific mRNAs to evade degradation via the 5'- 3' 950 mRNA decay pathway. Mol Cell 38, 345- 355, 951 doi:10.1016/j.molcel.2010.02.039 (2010). 952 52 Alvarez, B. & Moreno, S. Fission yeast Tor2 promotes cell growth and 953 represses cell differentiation. J Cell Sci 119, 4475- 4485, 954 doi:10.1242/jcs.03241 (2006). 955 53 Hayashi, T. et al. Rapamycin sensitivity of the Schizosaccharomyces 956 pombe tor2 mutant and organization of two highly phosphorylated TOR 957 complexes by specific and common subunits. Genes Cells 12, 1357- 1370, 958 doi:10.1111/j.1365- 2443.2007.01141. x (2007). 959 54 Ikai, N., Nakazawa, N., Hayashi, T. & Yanagida, M. The reverse, but 960 coordinated, roles of Tor2 (TORC1) and Tor1 (TORC2) kinases for growth, 961 cell cycle and separate- mediated mitosis in Schizosaccharomyces 962 pombe. Open Biol 1, 110007, doi:10.1098/rsob.110007 (2011). 963 55 Matsuo, T., Otsubo, Y., Urano, J., Tamanoi, F. & Yamamoto, M. Loss of 964 the TOR kinase Tor2 mimics nitrogen starvation and activates the sexual 965 development pathway in fission yeast. Mol Cell Biol 27, 3154- 3164, 966 doi:10.1128/MCB.01039- 06 (2007). 967 56 Uritani, M. et al. Fission yeast Tor2 links nitrogen signals to cell 968 proliferation and acts downstream of the Rheb GTPase. Genes Cells 11, 969 1367- 1379, doi:10.1111/j.1365- 2443.2006.01025. x (2006). 970 57 Weisman, R. & Choder, M. The fission yeast TOR homolog, tor1+, is 971 required for the response to starvation and other stresses via a conserved 972 serine. J Biol Chem 276, 7027- 7032, doi:10.1074/jbc.M010446200 (2001). 973 58 Weisman, R., Roitburg, I., Schonbrun, M., Harari, R. & Kupiec, M. 974 Opposite effects of tor1 and tor2 on nitrogen starvation responses in 975 fission yeast. Genetics 175, 1153- 1162, doi:10.1534/genetics.106.064170 976 (2007). 977 59 Ebrahimi, H., Masuda, H., Jain, D. & Cooper, J. P. Distinct 'safe zones' at 978 the nuclear envelope ensure robust replication of heterochromatic 979 chromosome regions. Elife 7, doi:10.7554/eLife.32911 (2018).
+
+<--- Page Split --->
+
+980 60 Banday, S., Farooq, Z., Rashid, R., Abdullah, E. & Altaf, M. Role of Inner 981 Nuclear Membrane Protein Complex Lem2- Nur1 in Heterochromatic Gene 982 Silencing. J Biol Chem 291, 20021- 20029, doi:10.1074/jbc.M116.743211 983 (2016). 984 61 Barrales, R. R., Forn, M., Georgescu, P. R., Sarkadi, Z. & Braun, S. 985 Control of heterochromatin localization and silencing by the nuclear 986 membrane protein Lem2. Genes Dev 30, 133- 148, 987 doi:10.1101/gad.271288.115 (2016). 988 62 Cohen, A. et al. TOR complex 2 in fission yeast is required for chromatin- 989 mediated gene silencing and assembly of heterochromatic domains at 990 subtelomeres. J Biol Chem 293, 8138- 8150, 991 doi:10.1074/jbc.RA118.002270 (2018). 992 63 Atkins, J. F. & Bjork, G. R. A gripping tale of ribosomal frameshifting: 993 extragenic suppressors of frameshift mutations spotlight P- site 994 realignment. Microbiol Mol Biol Rev 73, 178- 210, 995 doi:10.1128/MMBR.00010- 08 (2009). 996 64 El Yacoubi, B., Bailly, M. & de Crecy- Lagard, V. Biosynthesis and function 997 of posttranscriptional modifications of transfer RNAs. Annu Rev Genet 46, 998 99- 95, doi:10.1146/annurev- genet- 110711- 155641 (2012). 999 65 Gustilo, E. M., Vendeix, F. A. & Agris, P. F. tRNA's modifications bring 1000 order to gene expression. Curr Opin Microbiol 11, 134- 140, 1001 doi:10.1016/j.mib.2008.02.003 (2008). 1002 66 Jenner, L. B., Demeshkina, N., Yusupova, G. & Yusupov, M. Structural 1003 aspects of messenger RNA reading frame maintenance by the ribosome. 1004 Nat Struct Mol Biol 17, 555- 560, doi:10.1038/nsmb.1790 (2010). 1005 67 Chan, C. T. et al. Reprogramming of tRNA modifications controls the 1006 oxidative stress response by codon- biased translation of proteins. Nat 1007 Commun 3, 937, doi:10.1038/ncomms1938 (2012). 1008 68 Dedon, P. C. & Begley, T. J. A system of RNA modifications and biased 1009 codon use controls cellular stress response at the level of translation. 1010 Chem Res Toxicol 27, 330- 337, doi:10.1021/tx400438d (2014). 1011 69 Jungfleisch, J. et al. CHIKV infection reprograms codon optimality to favor 1012 viral RNA translation by altering the tRNA epitranscriptome. Nat Commun 1013 13, 4725, doi:10.1038/s41467- 022- 31835- x (2022). 1014 70 Patil, A. et al. Translational infidelity- induced protein stress results from a 1015 deficiency in Trm9- catalyzed tRNA modifications. RNA Biol 9, 990- 1001, 1016 doi:10.4161/rna.20531 (2012). 1017 71 Rapino, F. et al. Codon- specific translation reprogramming promotes 1018 resistance to targeted therapy. Nature 558, 605- 609, doi:10.1038/s41586- 1019 018- 0243- 7 (2018). 1020 72 Moreno, S., Klar, A. & Nurse, P. Molecular genetic analysis of fission yeast 1021 Schizosaccharomyces pombe. Methods Enzymol 194, 795- 823, 1022 doi:10.1016/0076- 6879(91)94059- I (1991). 1023 73 Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for 1024 Illumina sequence data. Bioinformatics 30, 2114- 2120, 1025 doi:10.1093/bioinformatics/btu170 (2014). 1026 74 Kim, D., Langmead, B. & Salzberg, S. L. HISAT: a fast spliced aligner with 1027 low memory requirements. Nat Methods 12, 357- 360, 1028 doi:10.1038/nmeth.3317 (2015).
+
+<--- Page Split --->
+
+1029 75 Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general 1030 purpose program for assigning sequence reads to genomic features. 1031 Bioinformatics 30, 923- 930, doi:10.1093/bioinformatics/btt656 (2014). 1032 76 Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change 1033 and dispersion for RNA- seq data with DESeq2. Genome Biol 15, 550, doi:10.1186/s13059- 014- 0550- 8 (2014). 1035 77 Gel, B. & Serra, E. karyplotReR: an R/Bioconductor package to plot 1036 customizable genomes displaying arbitrary data. Bioinformatics 33, 3088- 1037 3090, doi:10.1093/bioinformatics/btx346 (2017). 1038 78 Pfaffl, M. W. A new mathematical model for relative quantification in real- 1039 time RT- PCR. Nucleic Acids Res 29, e45, doi:10.1093/nar/29.9.e45 1040 (2001). 1041 79 Sanso, M., Gogol, M., Ayte, J., Seidel, C. & Hidalgo, E. Transcription 1042 factors Pcr1 and Atf1 have distinct roles in stress- and Sty1- dependent 1043 gene regulation. Eukaryot Cell 7, 826- 835, doi:10.1128/EC.00465- 07 1044 (2008). 1045 80 Sanso, M. et al. Cdk9 and H2Bub1 signal to Clr6- Cll/Rpd3S to suppress 1046 aberrant antisense transcription. Nucleic Acids Res 48, 7154- 7168, 1047 doi:10.1093/nar/gkaa474 (2020). 1048 81 Garcia, P. et al. Eng2, a new player involved in feedback loop regulation 1049 of Cdc42 activity in fission yeast. Sci Rep 11, 17872, doi:10.1038/s41598- 1050 021- 97311- 6 (2021). 1051 82 Roberts- Galbraith, R. H., Chen, J. S., Wang, J. & Gould, K. L. The SH3 domains of two PCH family members cooperate in assembly of the 1052 Schizosaccharomyces pombe contractile ring. J Cell Biol 184, 113- 127, 1053 doi:10.1083/jcb.200806044 (2009). 1054 83 Ma, Z. Q. et al. Supporting tool suite for production proteomics. 1056 Bioinformatics 27, 3214- 3215, doi:10.1093/bioinformatics/bt544 (2011). 1057 84 Kanoh, J. Telomeres and subtelomeres: new insights into the chromatin 1058 structures and functions of chromosome ends. Genes Genet Syst 92, 105, 1059 doi:10.1266/ggs.17- 10001 (2017). 1060
+
+<--- Page Split --->
+
+## Acknowledgements  
+
+We thank Rosa Degano and Nieves Ibarrola from the Salamanca Cancer Research Institute Proteomic Facility for their technical assistance in mass spectrometry analyses, Jorge Fernández- Vázquez, José Cansado, Yolanda Sánchez, Carlos R. Vázquez and Mercedes Tamame for sharing plasmids, antibodies and reagents, and Pilar Pérez and members of the Moreno lab for their valuable discussions and comments on the manuscript. This work was funded by the Spanish Ministry of Science and Innovation- MCIN (grants BFU2017- 88335- R and PID2020- 115929RB- I00) and from the Castile and Leon government (grants CSI259P20, CSI010P23 and IBFG Unit of Excellence programmes CLU- 2017- 03 and CL- EI- 2021- 08 co- funded by the P.O. Feder of Castile and Leon 14- 20 and European Union ERDF "Europe drives our growth"), and by the National Research Foundation of Singapore under its Singapore- MIT Alliance for Research and Technology Antimicrobial Resistance Interdisciplinary Research Group (J.S., P.D.). The work in K.L.G.'s lab was supported by the National Institutes of Health R35GM131799, and in D.H.'s lab by PDR T.0012.14, CDR J.0066.16 and PDR T.0112.21 grants. D.H. is a FNRS Director of Research. N.G.- B. and A.V.- B. were funded by XFPU15/03654, BES- 2015- 073171, predoctoral training contracts. R.L.- S.S. was funded by a predoctoral fellowship from the Castile and Leon government.  
+
+## Authors contributions  
+
+J.E.- D. led the project and conducted the experiments, excluding the RNAseq analysis performed by A.V.- B., the proteomic analyses conducted by R.L.- S.S. and J.- S.C. under the supervision of J.E.- D. and S.M., and K.L.G., respectively. The protein purifications used for MS analysis and cell viability assay performed by N.G.- B. RNA quantification was carried out by B.S., ChIP assay was performed by P.G., tRNA purification was carried out by P.T. under the supervision of D.H. and tRNA modification analysis was conducted by J.J. under P.C.D.'s supervision. J.E.D., E.H., D.M. and S.M. discussed and interpreted the results. J.E.- D. and S.M. wrote the original draft. All authors contributed to editing the manuscript. S.M. supervised the work.
+
+<--- Page Split --->
+
+1093  
+
+1094 Competing interests  
+
+1095 The authors declare no competing interests.  
+
+1096  
+
+1097 Data Availability Statement: The RNAseq data in this study has been deposited in GEO database with the following accession number GSE217398.  
+
+1099  
+
+1100 Additional information  
+
+1101 Supplementary information can be found online at:  
+
+1102 Supplementary Figs. 1- 6.  
+
+1103 Supplementary Table 1: List of the top 50 genes overexpressed in \(igo1\Delta\) at 4 hours in EMM2- N.  
+
+1105 Supplementary Table 2: List of the top 50 genes overexpressed in \(ppk18\Delta\) cek1 \(\Delta\) at 4 hours in EMM2- N.  
+
+1107 Supplementary Table 3: List of the proteins interacting with Paa1:L:GFP.  
+
+1108 Supplementary Table 4: List of the proteins interacting with Pab1:L:GFP.  
+
+1109 Supplementary Table 5: List of the proteins interacting with GFP:Pab1  
+
+1110 Supplementary Table 6: List of fission yeast strains used in this work.  
+
+1111 Supplementary Table 7: List of oligonucleotides and plasmids used in this work.  
+
+1112  
+
+1113  
+
+1114  
+
+1115  
+
+1116  
+
+1117
+
+<--- Page Split --->
+
+## Figure legends  
+
+Fig. 1. The Greatwall-Endosulfine switch regulates subtelomeric gene silencing and telomeric anchoring to the nuclear envelope. a Schematic representation of transcriptionally upregulated genes in the Endosulfine (igo1Δ) mutant. Genes overexpressed more than 10-fold in igo1Δ cells compared to the wild-type after 4 hours in nitrogen-free EMM2 medium. Subtelomeric genes are highlighted in red. b Schematic illustration of S. pombe subtelomeric chromatin structure (modified from 84). c Representative Super-Resolution Radial Fluctuations (SRRF) micrographs of wild-type (WT) and igo1Δ cells expressing Cut11:mCherry, Sad1:CFP and Taz1:YFP in nitrogen-rich EMM2 media (0 hours) and after 8 hours of nitrogen starvation in EMM2-N. The merged image and a detail view are shown. Bar: 2 μm. d Radial Profile Analysis for WT and igo1Δ cells after 0, 4 or 8 hours of nitrogen deprivation (see details in Supplementary Fig. 1b). The average projection signals for the NE (in red), the SPB (in cyan) and the telomeres (in yellow) are shown. The graphs represent the normalized integrated intensity as a function of distance in microns. The red lines correspond to the NE signal, the cyan lines correspond to the SPB signal and the yellow lines correspond to the telomeric signal. Over 100 nuclei were analysed at each time point. e Overlay between the average projection signals for Cut11/Sad1 or Cut11/Taz1 in WT and igo1Δ cells during entry into quiescence. The images were generated by projecting at least 100 nuclei. f Co-localization between Cut11/Sad1 and Cut11/Taz1 signals was quantified as Pearson correlation coefficients using ImageJ software. Student's t-test p-values are indicated, significant differences are in orange or red.  
+
+Fig. 2. Telomeric detachment from the nuclear envelope in igo1Δ cells is mediated by reduced Rap1 protein levels. a Extracts from rap1:L:HA and ccq1:L:HA cells in a WT and igo1Δ background were collected at 0, 1, 2, 4 and 8 hours of nitrogen starvation. These extracts were analysed by SDS- PAGE and western blotting using anti- HA antibodies. Ponceau staining was used as the loading control. b Extracts from rap1:L:HA cells in a WT and igo1Δ background, collected every 30 minutes during the first 2 hours and then at 4 hours of nitrogen starvation. These extracts were analysed by SDS- PAGE and western blotting
+
+<--- Page Split --->
+
+using anti- HA antibodies. Ponceau staining was used as the loading control. c Extracts from rap1:L:HA and rap1:L:HA \(P_{nmt}41x:GST:pab1\) cells in a WT and \(igo1\Delta\) background were collected during nitrogen starvation and analysed by SDS- PAGE and western blot using anti- HA and anti- GST antibodies. Strains were grown with or without thiamine (+T or - T) to repress or induce the \(pab1\) gene, encoding the B55 regulatory subunit of PP2A. Ponceau staining was used as the loading control. d Immunoblot quantification of c was performed with Image Studio Lite software from at least two independent experiments. e Radial Profile Analysis of WT, \(igo1\Delta\) and \(igo1\Delta P_{nmt}41x:GST:pab1\) cells bearing Cut11 (in red), Sad1 (in cyan) or Taz1 (in yellow) in EMM2 (0 hours) and after 8 hours of nitrogen starvation. The \(igo1\Delta P_{nmt}41x:GST:pab1\) cells were grown with or without thiamine (+T or - T) to repress or induce the expression of \(pab1\) . Over 100 nuclei were projected to generate the images and graphics. f Co- localization between Cut11/Sad1 and Cut11/Taz1 signals of e was quantified as Pearson correlation coefficients using ImageJ software. Student's t- test p- values are indicated, significant differences are in red.  
+
+Fig. 3. Crucial proteins required for silencing subtelomeric gene expression are downregulated in \(igo1\Delta\) cells. a Extracts from \(sgo2:L:HA\) cells in a WT and \(igo1\Delta\) backgrounds were collected every 30 minutes during the first 2 hours and then at 4 hours of nitrogen starvation. These extracts were analysed by SDS- PAGE and western blotting using anti- HA antibodies. Ponceau staining was used as the loading control. Immunoblot quantification was performed using Image Studio Lite software from at least two independent experiments. b Extracts from strains bearing \(sgo2:L:HA\) or \(sgo2:L:HA P_{nmt}41x:GST:pab1\) , were analysed by SDS- PAGE and western blotting with anti- HA and anti- GST antibodies. The \(igo1\Delta P_{nmt}41x:GST:pab1\) cells were grown with or without thiamine (+T or - T) to repress or induce the expression of \(pab1\) . Ponceau staining was used as the loading control. Immunoblot quantification was performed with Image Studio Lite software from at least two independent experiments. c Representative micrographs of WT or \(igo1\Delta\) cells expressing \(sgo2:L:GFP\) during entry into quiescence. The overlay of fluorescence and DIC images is shown. Quantification was carried out using ImageJ software from two independent experiments involving more than 150 cells. Bar: 5 μm. d Similar to (a), \(clr2:L:HA\)
+
+<--- Page Split --->
+
+protein was analysed in both WT and \(igo1\Delta\) backgrounds. e Similar to (a), clr3:L:HA protein was analysed in both WT and \(igo1\Delta\) backgrounds. f ChIP-qPCR was performed with anti- H3K14- acetyl antibodies and quantified with primer pairs at the indicated ORFs. WT and \(igo1\Delta\) cells grown in nitrogen- rich media (EMM2) or after 4 hours of nitrogen starvation were analysed. The graphs represent normalized values, and error bars (SD) for all ChIP- qPCR experiments were calculated from biological triplicates.  
+
+Fig. 4. PP2A interacts with proteins involved in tRNA modification. a Interacting network resulting from mass- spectrometry analysis for \(paa1:L:YFP\) in EMM- N. Cellular processes or protein complexes with a significant enrichment are colour- coded. b Interaction between \(paa1:L:YFP\) and \(trm112:L:HA\) . Protein extracts from cells expressing \(paa1:L:YFP\) , \(trm112:L:HA\) or \(paa1:L:YFP\) \(trm112:L:HA\) were immunoprecipitated in nitrogen- rich or nitrogen- depleted media with anti- GFP beads and probed with anti- GFP and anti- HA antibodies. Extracts (WCE) were assayed for levels of \(paa1:L:YFP\) and \(trm112:L:HA\) by western blot. c Extracts from cells expressing \(trm112:L:HA\) or \(trm112:L:HA\) \(P_{nml}41x:GST:pab1\) , were analysed by SDS- PAGE followed by immunoblotting with anti- HA and anti- GST antibodies. The \(igo1\Delta P_{nml}41x:GST:pab1\) cells were grown in EMM2 with or without thiamine (+T or -T) to repress or induce the expression of \(pab1\) . Ponceau staining was used as the loading control. Immunoblot quantification was performed with Image Studio Lite software from at least two independent experiments. d Extracts from \(Ctu1:L:HA\) cells in a WT and \(igo1\Delta\) backgrounds were collected every 30 minutes during the first 2 hours and then at 4 hours of nitrogen starvation. These extracts were analysed by SDS- PAGE and immunoblotting using anti- HA antibodies. Ponceau staining was used as the loading control. Immunoblot quantification was performed using Image Studio Lite software from at least two independent experiments. e Serial dilutions from WT and \(igo1\Delta\) cultures were spotted onto EMM2 (Minimal Media containing NH4Cl) or MMPhe (Minimal Media containing Phenylalanine) plates without or with paromomycin (0.5 mg/ml), puromycin (0.5 mg/ml) or cycloheximide (CHX, 2.5 μg/ml).
+
+<--- Page Split --->
+
+Fig. 5. Igo1 regulates some tRNA modifications. a Heatmap analysis of changes in the relative levels of tRNA ribonucleoside modifications in the WT and the \(igo1\Delta\) mutant. The side colour bar displays the range of z- score change values. The z- score was calculated as the value for each time point minus the average value for the modification, and the resulting value was divided by the standard deviation. b Fold- change of \(\mathsf{cm}^5\mathsf{S}^2\mathsf{U}_{34}\) modification in WT and \(igo1\Delta\) mutant cells. Student's t- test p- values were calculated from biological triplicates. c Extracts from strains bearing rap1:L:HA protein transformed with episomal plasmids \(\mathsf{ptRNA}_{\mathsf{CuULys}}\) , \(\mathsf{ptRNA}_{\mathsf{uuULys}}\) or the empty vector pREP42x were analysed by SDS- PAGE and western blotting with anti- HA antibodies during nitrogen starvation. Ponceau stain was used as a loading control. d Immunoblot quantification performed with Image Studio Lite software from at least three independent experiments. e Extracts from \(igo1\Delta\) mutant transformed with episomal plasmids \(P_{nmt}41x:HA:rap1\) (AAA/AAG) or mutated version \(P_{nmt}41x:HA:rap1\) (all AAG) were analysed by SDS- PAGE followed by immunoblotting with anti- HA antibodies during nitrogen starvation. Ponceau staining was used as a loading control. Immunoblot quantification were performed with Image Studio Lite software from at least two independent experiments.  
+
+Fig. 6. Endosulfine, Gad8 and Elongator are required for efficient translation of certain mRNAs during quiescence entry. a Extracts from WT and \(igo1\Delta\) cells expressing gad8:L:HA, were analysed by SDS- PAGE and immunoblotting with anti- HA antibodies during nitrogen starvation. Ponceau stain was used as a loading control. Immunoblot quantification were performed with Image Studio Lite software from at least two independent experiments. b Same as in (a), Gad8 phosphorylation state was analysed in WT and \(igo1\Delta\) cell extracts. c Same as in (a), sgo2:L:HA protein was analysed in a WT and gad8Δ cell extracts. d Same as in (a), sgo2:L:HA protein was analysed in a WT and elp3Δ cell extracts. e Serial dilutions from cultures of WT, \(igo1\Delta\) , gsk3Δ and \(igo1\Delta\) gsk3Δ were spotted onto MMPhe (Minimal Media with Phenylalanine) plates without or with paromomycin (0.5 mg/ml).
+
+<--- Page Split --->
+
+Fig. 7. Activation of TORC2- Gad8 signalling in quiescent cells promotes the translation of mRNAs with a high \(\mathrm{AAA}_{\mathrm{Lys}}\) codon usage. This model is based on previous work in fission yeast, demonstrating that nitrogen starvation induces the inactivation of TORC1 and the activation of TORC2 signalling through the Greatwall- Endosulfine- PP2A/B55 pathway \(^{4 - 7}\) . Phosphorylation of Gad8 at S546 leads to the inhibition of Gsk3 and the activation of Elongator, which promotes \(\mathrm{U}_{34}\) tRNA modification and translation of Tsc1, an inhibitor of TORC1, as well as activators of TORC2, such as Tor1 and Rictor (depicted by blue arrows) \(^{8}\) . In this study, we present additional feedback loops (indicated by orange arrows) that enhance the translation of Gad8, Trm112, Ctu1 and Cgi121, further increasing the \(\mathrm{U}_{34}\) and A37 tRNA modifications necessary for the efficient translation of mRNAs enriched in AAA codons. Such mRNAs include rap1, clr2, clr3 and sgo2, which encode proteins required for the correct attachment of telomeres to the NE.
+
+<--- Page Split --->
+![](images/Figure_unknown_0.jpg)
+
+
+<--- Page Split --->
+![](images/Figure_unknown_1.jpg)
+
+
+<--- Page Split --->
+![](images/Figure_unknown_2.jpg)
+
+ 
+
+![PLACEHOLDER_43_1]
+
+
+<--- Page Split --->
+![PLACEHOLDER_44_0]
+
+ 
+
+b 
+
+<table><tr><td></td><td>WCE</td><td>IP (GFP)</td></tr><tr><td>Paa1-L-YFP</td><td>+</td><td>+</td></tr><tr><td>Trm112-L-HA</td><td>+</td><td>+</td></tr><tr><td>Nitrogen</td><td>+</td><td>+</td></tr></table>
+
+![PLACEHOLDER_44_1]
+
+ 
+
+d 
+
+<table><tr><td></td><td>Ctu1-L-HA</td></tr><tr><td>WT</td><td>\(igo1\Delta\)</td></tr><tr><td>40 kDa</td><td></td></tr><tr><td>70 kDa</td><td></td></tr></table>
+
+![PLACEHOLDER_44_2]
+
+ 
+
+Translation 
+
+initiation factors 
+
+<table><tr><td colspan="5">Molecular Function (Gene Ontology)</td></tr><tr><td>GO-term</td><td>description</td><td>count in network</td><td>strength</td><td>false discovery rate</td></tr><tr><td>GO:0030743</td><td>Translation initiation factor activity</td><td>19 of 36</td><td>0.75</td><td>3.36e-06</td></tr><tr><td>GO:0054812</td><td>aminoacyl-tRNA ligase activity</td><td>18 of 35</td><td>0.74</td><td>8.98e-06</td></tr><tr><td>GO:0051082</td><td>Unfolded protein binding</td><td>20 of 58</td><td>0.57</td><td>0.00023</td></tr></table>
+
+<table><tr><td colspan="5">Cellular Component (Gene Ontology)</td></tr><tr><td>GO-term</td><td>description</td><td>count in network</td><td>strength</td><td>false discovery rate</td></tr><tr><td>GO:0000159</td><td>Protein phosphatase type 2A complex</td><td>6 of 6</td><td>1.03</td><td>0.0120</td></tr><tr><td>GO:0090443</td><td>FAR/SIN/STRIPAK complex</td><td>6 of 6</td><td>1.03</td><td>0.0120</td></tr><tr><td>GO:0070993</td><td>Translation preinitiation complex</td><td>12 of 16</td><td>0.91</td><td>0.00011</td></tr><tr><td>GO:0022626</td><td>Cytosolic ribosome</td><td>63 of 140</td><td>0.68</td><td>2.15e-18</td></tr><tr><td>GO:1990904</td><td>Ribonucleoprotein complex</td><td>95 of 528</td><td>0.29</td><td>3.81e-07</td></tr></table>
+
+![PLACEHOLDER_44_3]
+
+ 
+
+![PLACEHOLDER_44_4]
+
+ 
+
+![PLACEHOLDER_44_5]
+
+ 
+
+![PLACEHOLDER_44_6]
+
+
+<--- Page Split --->
+![PLACEHOLDER_45_0]
+
+ 
+
+<center>b</center> 
+
+![PLACEHOLDER_45_1]
+
+ 
+
+![PLACEHOLDER_45_2]
+
+ 
+
+<center>d</center> 
+
+![PLACEHOLDER_45_3]
+
+ 
+
+<center>e</center> 
+
+![PLACEHOLDER_45_4]
+
+
+<--- Page Split --->
+![PLACEHOLDER_46_0]
+
+
+<--- Page Split --->
+![PLACEHOLDER_47_0]
+
+
+<--- Page Split --->
+
+## Supplementary Files  
+
+This is a list of supplementary files associated with this preprint. Click to download.  
+
+SupplementaryTable1. xlsxSupplementaryTable2. xlsxSupplementaryTable3. xlsxSupplementaryTable4. xlsxSupplementaryTable5. xlsxSupplementaryTable6. docxSupplementaryTable7. docxEncindarDelDedoSupplementaryFigures.pdf
+
+<--- Page Split --->