Split (1)

train · 5.09k rows

pmcid stringlengths 6 6	title stringlengths 9 374	abstract stringlengths 2 4.62k ⌀	fulltext stringlengths 167 106k	file_path stringlengths 64 64
538274	Blood coagulation and the risk of atherothrombosis: a complex relationship	The principles of Virchov's triad appear to be operational in atherothrombosis or arterial thrombosis: local flow changes and particularly vacular wall damage are the main pathophysiological elements. Furthermore, alterations in arterial blood composition are also involved although the specific role and importance of blood coagulation is an ongoing matter of debate. In this review we provide support for the hypothesis that activated blood coagulation is an essential determinant of the risk of atherothrombotic complications. We distinguish two phases in atherosclerosis: In the first phase, atherosclerosis develops under influence of "classical" risk factors, i.e. both genetic and acquired forces. While fibrinogen/fibrin molecules participate in early plaque lesions, increased activity of systemic coagulation is of no major influence on the risk of arterial thrombosis, except in rare cases where a number of specific procoagulant forces collide. Despite the presence of tissue factor – factor VII complex it is unlikely that all fibrin in the atherosclerotic plaque is the direct result from local clotting activity. The dominant effect of coagulation in this phase is anticoagulant, i.e. thrombin enhances protein C activation through its binding to endothelial thrombomodulin. The second phase is characterized by advancing atherosclerosis, with greater impact of inflammation as indicated by an elevated level of plasma C-reactive protein, the result of increased production influenced by interleukin-6. Inflammation overwhelms protective anticoagulant forces, which in itself may have become less efficient due to down regulation of thrombomodulin and endothelial cell protein C receptor (EPCR) expression. In this phase, the inflammatory drive leads to recurrent induction of tissue factor and assembly of catalytic complexes on aggregated cells and on microparticles, maintaining a certain level of thrombin production and fibrin formation. In advanced atherosclerosis systemic and vascular wall driven coagulation becomes more important and elevated levels of D-dimer fragments should be interpreted as markers of this hypercoagulability.	Background The blood coagulation system comprises three basic elements: platelet adhesion, activation and aggregation, fibrin formation, and fibrinolysis. These elements interact with each other and with the blood vessel wall and under physiological conditions blood flow to tissues is unimpaired by clotting [ 1 ]. Under pathophysiological conditions, blood coagulation gets activated along the principles outlined by Virchov, which indicate that thrombosis (the formation of an intraluminal blood clot) always occurs through the interaction of three components: an altered vessel wall, an impaired or changed pattern of blood flow and an altered blood composition. The principles of Virchov's triad appear to be operational in each different type of thrombosis [ 2 , 3 ]. In venous thrombosis of the lower limbs, stasis, local inflammation on activated vascular endothelial cells induced by adhering leukocytes and platelets and in some cases direct vascular damage, promotes local thrombus formation. In a first episode of venous thrombosis the pre-existing composition of the blood is particularly important where congenital and acquired hypercoagulable factors such as factor V Leiden mutation and oral contraceptives, respectively, act in concert to accelerate clotting [ 4 ]. In disseminated intravascular coagulation (DIC) , widespread fibrin formation is the result of systemic inflammatory changes that induce cellular tissue factor dependent activated blood coagulation as well as local alterations in microcirculatory flow and enhanced activity and permeability of capillary endothelial cells [ 5 ]. Again, DIC follows Virchov's principles, i.e. interactions among all three elements occur which are all relevant determinants of outcome. In arterial thrombosis , local flow changes and particularly vascular wall damage are the main pathophysiological elements. Alterations in composition of the arterial blood are also involved but the specific role and importance of blood coagulation is an ongoing matter of debate [ 6 , 7 ]. While numerous studies have shown increased activity of the blood coagulation system in patients at risk of arterial thrombotic complications, Tracy concludes on the basis of genetic studies that there is no "compelling argument supporting the importance of a preexisting hypercoagulable state as a major risk factor for atherothrombotic disease" [ 8 ]. In a recent debate, Reitsma points out that in the context of atherosclerosis a hypercoagulable state is of minor importance for the risk of thrombosis and high levels of coagulation factors such as factor VIII are risk indicators rather than causal factors [ 6 ]. On the other hand, in the same debate Grant argues on the basis of biochemical, clinical and philosophical arguments that hypercoagulability is indeed an issue of importance in arterial thrombosis, illustrated on the basis of several observations in patients with diabetes and insulin resistance [ 7 ]. In spite of the apparent controversies regarding this topic, observational studies focused on activity of coagulation and fibrinolysis in patients with arterial vascular disease continue to be published. As an example of a "clotting" marker, measurement of fibrin D-dimer fragments by one of many commercial assays, has been implicated as a risk indicator since more than 15 years, in a range of patient studies related to severity of atherosclerosis and/or risk of (recurrent) thrombotic complications [ 9 - 25 ]. In general, these studies indicate that D-dimer, similar to C-reactive protein (CRP), is a moderate but consistent and independent marker of risk of cardiovascular disease, both in population studies and in patients at risk [ 22 , 24 , 26 ]. Given the actual debate on the relevance of coagulation in arterial vascular disease it is timely to consider whether D-dimer should be regarded a risk marker (or bystander ), or a marker of a causal process , i.e. hypercoagulability. More specifically, the question remains whether hypercoagulability, here defined as an increased potential to produce fibrin in plasma (indicated by elevated thrombin production, fibrin production or both), as compared to individuals of similar age and sex, should be seen as a cause or merely consequence of atherosclerosis and thrombosis. Thrombogenicity and atherosclerosis In the majority of patients atherothrombotic complications develop on the basis of atherosclerosis in one or more coronary, cerebrovascular or peripheral arteries [ 27 ]. Atherosclerosis, a multifactorial disease, is the consequence of many years of exposure to atherogenic influences that lead already at young age to early lesions, or so-called "fatty streaks". Under influence of age- and sex-related factors these early lesions advance and this process is accelerated by genetic determinants (such as related to lipid and glucose metabolism and blood pressure) and environmental influences, including smoking and diet [ 28 , 29 ]. Arterial thrombi form in the course of progression to complex lesions, where the combination of vascular remodeling, erosion of the vessel luminal surface or frank rupture of plaques triggers the blood coagulation system. Central to this process is chronic inflammation and proteolysis culminating in plaque damage and exposure to luminal blood flow [ 27 , 30 ]. In addition, angiogenesis related neovessels are prone to rupture resulting in increased intra-plaque hemorrhage [ 31 ]. Activation of blood coagulation occurs primary through interaction of platelets, vessel wall and plasma proteins (so-called primary haemostasis). When injury to the blood vessel wall causes disruption of its endothelial layer, the underlying extracellular matrix is exposed. In this matrix, both von Willebrand factor (vWF) and collagen are present and after exposure, they will bind to specific receptors, glycoproteins (GP), present on the platelets. Dependent on the flow within the vessel other glycoproteins are involved in the adhesion of the platelets to the vessel wall. During low shear stress, GP Ia-IIa, GP VI and GP IV are the primary receptors for collagen, and during high shear forces, the primary indirect receptor for collagen is GP Ib-IX-V in a vWF dependent interaction. After adhesion of the platelets, they become deformed due to cytoskeletal changes, thereby exposing activated integrins and secreting ADP, serotonin etc. One of the integrins, GP IIb-IIIa, binds vWF (in high shear areas) or fibrinogen (in low shear areas) to mediate platelet aggregation under shear conditions. Also other platelet receptors and lipid products i.e. arachidonic acid, contribute to platelet aggregation. In this review however we will focus on secondary haemostasis, in which the interaction between circulating factor VII(a) to tissue factor, exposed by the damaged vessel wall, leads to activation of the coagulation cascade. Several studies have shown that tissue factor is a prominent component of plaque lesions where it is localized in the outer membranes of infiltrating macrophages/foam cells and smooth muscle cells as well as on apoptotic cells and cell bodies (Figure 1 ) [ 30 ]. Unstable plaques contain most potent tissue factor activity; in addition, tissue factor-rich microparticles are being shed from activated and apoptotic cells and may contribute to acute thrombotic occlusion, particularly in the downstream microcirculation [ 30 , 31 ]. Formation of the tissue factor-factor VII(a) complex drives the intrinsic pathway of coagulation to form thrombin and fibrin. Platelet adhesion and activation, and interactions with leukocytes, accelerate the process of thrombin formation providing catalytic surfaces, expressing tissue factor and yielding coagulation proteases such as factor XIa that amplify thrombin generation [ 32 - 35 ]. According to Virchov's postulate acute arterial thrombosis occurs due to interaction among a damaged atherosclerotic vessel wall, an altered blood flow due to changes in shear stress related to atherosclerosis and blood elements, i.e. cells and coagulation proteins [ 3 ]. Whether the state of activity of the blood coagulation system (in other words "high risk blood") is really altered prior to thrombosis is the principal issue of controversy. Figure 1 The initiation of an atherosclerotic lesion is characterized by retention of LDL and subsequent oxidative modification (oxLDL) within the matrix of the vascular intima. Stimulation of the overlying endothelial cells by oxLDL recruits monocytes from the circulation to the vessel wall. Differentiation of monocytes into macrophages and scavenger receptor mediated uptake of oxLDL aggregates results in the formation of foam-cells. Upon stimulation vascular smooth muscle cells (VSMC) migrate and proliferate. Tissue factor is expressed on macrophages and VSMCs within the advanced lesion and is likely to be involved in the conversion of accumulated fibrinogen into fibrin, although fibrin polymerization can be facilitated by other enzymes than thrombin. Furthermore, VSMCs and macrophage derived apoptotic bodies exposing TF probably contribute in thrombin formation. Considering atherosclerosis as a chronic inflammation, the inflammatory drive leads to IL-6 induced TF expression of circulating monocytes and the formation of microparticle exposing TF in the circulation, maintaining a certain level of thrombin production and fibrin formation. Increased circulating D-dimer levels are thus the result of fibrin proteolysis in both circulation and the advanced atherosclerotic lesion. The contribution of blood coagulation to atherosclerosis: the role of fibrin/fibrinogen and its split products The involvement of coagulation in the pathological substrate of atherosclerosis is beyond dispute. For many years pathologists have noted the abundant presence of fibrin in advanced atherosclerosis and this finding has fueled part of the debate on the relevance of fibrin or fibrinogen for vessel wall lesions. Rokitansky and later Duguid proposed the encrustation theory as concept for the role of fibrin in atherosclerosis (reviewed in [ 36 ]). In this concept thrombosis was considered an etiological factor of importance in atherosclerosis, which was probably based on the presence of the end product of clotting, fibrin. Later work confirmed that fibrin is indeed an abundant protein in the arterial vessel wall, but not confined to atherosclerotic lesions. Schwartz and colleagues demonstrated that fibrin was also present (although at lower ratios of fibrin: fibrin/fibrinogen) in the non-sclerotic regions of the carotid artery where it did not co-localize with tissue factor in about 50% of the sections studied [ 37 ]. Thus, it seems unlikely that all fibrin in the vessel wall is the direct result from local clotting activation; alternatively, inflammatory influences that are characteristic of atherosclerosis [ 27 ] activate the coagulation system and also stimulate the transfer of fibrinogen and fibrin molecules to the intima where fibrin can be polymerized also by other enzymes than thrombin [ 38 ]. Autopsy data have indicated that fibrinogen accumulation in the vessel wall may be an early event in atherosclerosis, i.e. a small amount of fibrinogen in a thickened intima was demonstrated in a 4 year old boy [ 36 ]. The deposition of fibrinogen was apparently associated with the presence of LDL in the vessel wall and was related to age and intimal thickening. These authors suggested that intimal deposition of fibrin or fibrinogen preceded or facilitated LDL accumulation in the arterial vessel wall. Direct evidence for such a function of fibrin or fibrinogen, however, is still lacking. Fibrinogen knockout mice against an apoE -/- background did not have fewer arterial lesions ranging from early lesions to complex fibrous plaques, suggesting that fibrinogen is not an essential molecule for atherosclerosis [ 39 ]. However, a later study demonstrated that fibrinogen was an important mediator of atherogenesis in apo(a) transgenic mice where the accumulation of apo(a) in the vessel wall and average lesion area were markedly attenuated in the fibrinogen -/- x apo(a) crossbred animals [ 40 ]. The specific effect of fibrin and its split products in the vessel wall has also been studied. In general it appears that with increasing complexity of lesions there is an increase in the presence of intimal fibrinogen/fibrin and threads of fibrin, as well as an accumulation of various split products that may be involved in atherogenesis. The effect of fibrin and its split products on smooth muscle cells may be such that fibrin stimulates proliferation, while split products inhibit this process. Fibrin cleavage products may be detrimental for endothelial cell function, increasing permeability and promoting endothelial cell migration [ 36 , 41 , 42 ]. Degradation products also enhance chemotaxis of smooth muscle cells and monocytes. Extracellular accumulation of fibrin(ogen) on monocytes stimulates cholesterol transfer from platelets to monocytes/macrophages and each of these mechanisms may be relevant to the development of atherosclerosis. In addition, D-dimer fragments induce Il-6 production by monocytes in vitro [ 41 , 42 ]. The complex interactions of plasminogen, plasmin and its inhibitor, with regard to vessel wall function and remodeling, have recently been reviewed and its discussion is beyond the scope of this paper [ 43 ]. However, a few points need to be addressed. Plasmin, produced by activation of plasminogen, is the crucial enzyme in fibrin degradation and generation of split products. Deficiency of plasminogen in mice (Plg -/- ) results in markedly discrepant effects on atherosclerosis. While Plg -/- mice with an apoE -/- background showed an accelerated development and progression of intimal lesions, Plg -/- mice were protected against atherosclerosis in association with transplantation (reviewed in [ 43 ]). The origin of such apparently conflicting effects may lie in a dominant effect of the absence of plasminogen on lipid metabolism, including markedly lower HDL levels in the knockout mice in the first experiment, while an effect on leukocyte transport and migration was the major effect in the transplant experiments. Hence, as Plow et al conclude, "it may be the importance of the cellular migration as a rate-determining step that establishes the influence of plasminogen in either atherosclerosis or restenosis". Accumulating fibrin that polymerizes in the vessel wall triggers fibrinolysis. Fibrinolytic enzymes tissue plasminogen activating factor and urokinase plasminogen activating factor (tPA and uPA, respectively) are present in intima and are secreted by endothelial cells and likely play an important role in vascular remodeling [ 42 - 44 ]. However, their intrinsic capacity to generate plasmin cleaving fibrin may also contribute to increased local fibrinolysis. In addition, the main inhibitor of fibrinolysis plasminogen activator inhibitor-1 (PAI-1) is also more abundantly expressed in tissues of patients with atherosclerosis. Under influence of inflammation, vascular endothelium may produce increased amounts of PAI-1 that might inhibit fibrin cleavage. However, it is questionable to what extend endothelial cells contribute to PAI-1 production in patients with atherosclerosis, since at least in patients with the metabolic syndrome, adipocytes and hepatocytes are more prominent sites of PAI-1 synthesis in relation to plasma PAI-1 [ 45 ]. The net effect on fibrin cleavage and progression of atherosclerosis cannot be estimated. The above mentioned experiments with Plg -/- mice give important clues regarding the range of mechanisms that are influenced. Effective fibrinolysis may be important in limiting fibrin accumulation and atherosclerosis in the initial phases. However, upon stronger inflammatory stimulation the effect on cell trafficking into the vessel wall becomes more dominant and the outcome may reverse such that impaired fibrinolysis may limit atherosclerosis. The latter would imply that high levels of PAI-1 may even be protective against atherosclerosis under certain conditions. Consequently, high concentrations of D-dimers, reflecting active fibrinolysis, may indeed be regarded as a sign of progressive atherosclerosis under inflammatory conditions. All of these issues may have therapeutic consequences since several drugs that are routinely used in patients with atherosclerosis including statins, angiotensin converting enzyme inhibitors as well as angiotensin receptor blockers appear to influence the balance of coagulation and fibrinolysis, which may influence atherosclerosis on the long term by altering vascular properties [ 42 ]. Clinical studies of increased intravascular fibrin as indicator of severity of atherosclerosis; studies in peripheral arterial disease As mentioned above, a large number of clinical studies in different groups of patients with atherosclerotic disease have generally shown that increased levels of D-dimer fragments in plasma are associated with an increased risk of severe atherosclerosis and an increased risk of vascular complications. In population based studies the contributable risk of an increased D-dimer level is quite small but statistically significant. In specific cohorts of patients the risk association is more outspoken, but of course here selection bias may produce slightly stronger associations than may be found in "real life". Before addressing specific study findings a few general observations deserve attention. First, strong associations between age and sex on the one hand, and D-dimer levels on the other hand, are noted [ 9 , 11 , 42 ]. D-dimer levels increase with age, are higher in women and may be influenced by a number of additional factors that differ per study. Second, D-dimer levels are oftentimes associated with markers of inflammation, i.e. CRP and Il-6 [ 19 , 22 , 24 , 25 , 42 , 46 ]. In addition, D-dimers often correlate with fibrinogen levels, which may be related to inflammation, but fibrinogen is also the substrate for fibrin, thus a more straightforward substrate-enzyme-cleavage product relation may also play a significant role. Before addressing the mechanisms we will consider D-dimer as an independent entity, i.e. a marker of disease severity. The most striking associations with clinical disease come from patients with peripheral artery disease (PAD), a reflection of systemic and advanced atherosclerosis in the majority of individuals. The total risk of clinical complications or mortality reaches figures of up to 25% annually in patients with PAD (48). In patients with PAD, elevated D-dimer levels are independent predictors of complications and are associated with severity of atherosclerosis [ 12 - 14 , 16 , 25 ]. Functionally, patients with PAD and highest D-dimers had the worst walking distance [ 23 ] and venous occlusion resulted in impaired fibrinolytic response in patients with PAD versus those without PAD [ 47 ]. Significant and independent associations between D-dimer and clinically relevant endpoints were also found in several studies in patients with PAD [ 15 , 18 , 25 , 48 ], in line with observations in other groups of patients with atherosclerotic manifestations. Elevated levels of D-dimers are usually considered as a marker of increased clotting activity. This assumption is one of the key elements of the controversy regarding cause and consequence of hypercoagulability. Indeed, Herren et al observed increased levels of D-dimer in patients with PAD, correlating with severity of disease. They also noted an association between hypercoagulability (higher F1+2 and TAT) and occurrence of myocardial ischemia during exercise testing [ 13 ], suggesting a link between enhanced clotting activity, D-dimer levels and PAD. A similar link between activated clotting and higher D-dimer levels was also noted by van der Bom and colleagues showing that the association between D-dimers and severity of PAD was most apparent in those with highest thrombin cleavage fragment F1+2 and thrombin-antithrombin (TAT) levels [ 14 ]. However, a number of other studies in patients with atherosclerosis failed to reveal significant correlations between D-dimers and markers of thrombin generation [ 21 , 26 ]. This leads to the question whether D-dimer generation reflects hypercoagulability in blood, increased fibrin production and fibrinolysis in the arterial intima as part of advanced atherosclerosis, or an increased state of inflammation due to proteolytic cleavage of fibrin by neutrophilic enzymes such as elastase? In spite of the substantial observational data, application of D-dimer assays or other risk factor measurements such as for CRP have not gained acceptance in individual patients with PAD or other cardiovascular disease yet. Thus, secondary prevention of complications is not guided by any laboratory assay, but limited to general recommendations such as the advice to stop smoking and the prescription of a platelet inhibiting drug [ 49 ]. Three reasons explain this lack of implementation: one is the substantial overlap in D-dimer (or CRP) values between normals and patients in general; second, the low specificity and three the lack of understanding the cause of the D-dimer production and its interpretation. In the context of this paper we will focus on the third reason and discuss the mechanisms that lead to elevated fibrin cleavage products in plasma in patients with atherosclerosis. Theoretically, there are different options to explain increased D-dimer levels in plasma. If D-dimers indicate increased systemic clotting activity then a specific anticoagulant intervention may theoretically be the preferred intervention. Clinical studies with anticoagulants in patients randomized or stratified on the basis of D-dimer levels have, however, not been carried out yet. If however, D-dimers are a reflection of severity of atherosclerosis such an intervention may be inappropriate and potentially harmful because of the avoidable risk of bleeding and calcification of the arterial vessel wall upon long-term administration (at least with vitamin K antagonists). Alternatively, if D-dimer levels merely reflect inflammation , than therapy should preferably consist of anti-inflammatory agents including higher doses of aspirin, statins or ACE inhibitors. Thus, the interpretation of elevated D-dimer levels is quite important in order to guide decisions about individual therapy. Inflammation and fibrin formation in atherosclerosis Atherosclerosis is a chronic inflammatory disease [ 27 , 28 ]. This widely accepted concept is based on a body of evidence from experimental and human observational studies. An indication of systemic inflammation is an elevated level of plasma CRP, the result of increased production influenced by Il-6. Several meta-analyses have established that CRP is an independent predictor of mortality in patients with atherosclerosis [ 50 , 51 ]; thus, a systemically activated inflammatory system is probably involved in its pathogenesis, i.e. progression and extension of atherosclerosis, as well as plaque rupture in advanced atherosclerosis. Studies from the sepsis field including models of endotoxemia and sepsis in humans and primates, respectively, have shown that inflammatory stimulation leads to activation of blood coagulation [ 52 , 53 ]. Tissue factor synthesis is a rapid consequence of endotoxin infusion, which is a strong inflammatory stimulus, and this is followed by tissue factor expression on inflammatory cells and on microparticles, inducing thrombin and fibrin generation. Il-6 is a dominant cytokine in this process, but Il-1β and TNF-α are also involved. These experimental studies also exposed discordance in time of the coagulation activation steps, in which not a true cascade but a delayed and protracted course of activation occurred (1). If we consider atherosclerosis as a chronic (or recurrent) inflammatory condition, than the recurrent inflammatory drive leads to recurrent induction of tissue factor (with intermediate phases of hypo-responsiveness to stimulation) and assembly of catalytic complexes on aggregated cells and on microparticles, maintaining a certain level of thrombin production and fibrin formation [ 32 , 33 ]. The increased level of fibrinogen and fibrin monomers may enhance the uptake by the vessel wall of lipid-loaded particles and macrophages. In the vessel wall, further fibrin polymerization can occur due to local thrombin or other proteases activities. In this concept, there is an increased generation of thrombin and fibrin in the blood circulation due to increased presence of inflammatory cytokines and proteins, but this does not necessarily lead to increased free thrombin in plasma. One should realize that coagulation enzymes that are generated associate with any available "scavenger", which can be an inhibitor such as antithrombin, but could also be a protease activated receptor (PAR) on platelets or endothelial cells [ 54 - 56 ]. Thus, a lack of rise of TAT at a moment when an elevated D-dimer level is noted cannot be interpreted as proof of a lack of increased thrombin production. Similarly, a lack of rise in F1+2 does not necessarily imply lack of thrombin production, because the F1+2 fragment can associate with cell membranes and little is known about the influence of microparticles. In addition, there are issues of sensitivity of commercial laboratory tests, i.e. the F1+2 assay is not a very sensitive tool in general, for monitoring activated coagulation. In fact, a D-dimer assay is the tool of choice for excluding (venous) thrombosis because of its superior sensitivity as compared to other tests for activated clotting [ 57 ]. We would propose that an increased production of thrombin in atherosclerosis is associated with an altered distribution of thrombin over the available binding sites leaving a greater procoagulant fraction that converts fibrinogen to fibrin. In advanced atherosclerosis a diminution in natural anticoagulant mechanisms including antithrombin (reduced expression of glycosaminoglycans at endothelium) and activated protein C (by down regulation of thrombomodulin) contributes to a higher level of procoagulant thrombin in the absence of increased TAT levels. Due to lack in sensitivity and maybe redistribution of F1+2 fragments binding to cell membranes the increased thrombin production is mostly undetectable by commercial F1+2 assays. Finally, discordances in peak levels of thrombin and fibrin production and cleavage may obscure any mechanistic associations. The exact contribution of subendothelial fibrin formation and cleavage to D-dimer levels in blood cannot be estimated. While locally deposited tissue factor acts as a trigger of thrombin generation, it has not been shown that this is a source of ongoing subendothelial coagulation activity. The recent discovery of factor VII in plaque contents and the in vitro evidence for production of this Gla-protein by smooth muscle cells might form a basis for local thrombin production, but there is no indication yet that this might be quantitatively important as compared to hepatic production of factor VII [ 58 ]. The point to make is that high D-dimer levels in blood from patients with atherosclerosis should be primarily viewed as an indication of systemic hypercoagulability , a conclusion based on the above arguments and the experimental evidence indicating the intimate relationship between inflammation and coagulation [ 52 ]. Inherited or acquired hypercoagulability and atherosclerosis? Early studies from Rosendaal et al, suggested that thrombophilic traits including the prothrombin 20210 gene variant would be a risk factor for myocardial infarction in specific individuals such as heavy smoking young women [ 59 ]. These data led to speculations about the importance of inherited thrombophilia in arterial disease in general, but this association was refuted in a subsequent large study in young individuals [ 60 ]. Indeed, the prevailing opinion is that most known thrombophilic traits with an associated risk of venous thrombosis, do not influence the risk of arterial thrombosis [ 60 , 61 ]. This notion may lead to the erroneous assumption that a state of increased coagulation activity, irrespective the cause, would be of no significant influence for the risk to develop arterial thrombosis. In spite of discarding Rosendaal's data as being the result of bias due to lack of power and patient selection we should perhaps accept the possibility that in this specific group of individuals with an unhealthy lifestyle, hypercoagulable influences may have accumulated: smoking related inflammation (and tissue factor expression in arterial walls; [ 62 ]) in conjunction with estrogenic stimulation leading to an disproportionably high risk of arterial thrombosis even in the absence of overt atherosclerosis. In the majority of the population of young individuals (< 40 yrs) such scenarios do not play a major role and the risk of early atherosclerosis is influenced predominantly by "classical" risk factors. This concept matches with the observation that D-dimers levels are also no independent risk indicator in relatively healthy and younger populations including that of the Physicians Health Study [ 63 ]. In people of advanced age this situation may change considerably and the weight of risk factors may change over time. A study that specifically addressed this point is the Bruneck community study [ 64 ]. In this population study carotid artery atherosclerosis was monitored with duplex ultrasound, risk factors were recorded, baseline blood samples collected and individuals were prospectively followed in time. Conventional and clotting (candidate) risk factors were then linked to markers of disease. This study suggested a two-stage model of disease in which conventional risk factors such as dyslipidemia and smoking, influence early stages of atherosclerosis, while other factors including those linked to coagulation, influenced later stages of disease (Figure 1 ). In advanced atherosclerosis the influence of coagulation may indeed be more prominent than in early stages, but it should be realized that acquired rather than genetically determined forces are involved. In this regard, the similarities between arterial thrombosis and the risk of recurrent venous thrombosis was used by Reitsma to make the point that inflammation is a key player under such conditions, reducing the influence that genetic thrombophilic background might inflict. On the other hand the same argument could be used to illustrate that indeed inflammation plays a more prominent role in advanced atherosclerosis where it more strongly drives the risk of thrombosis. Let us consider this situation from the scope of venous thrombosis; this is not far edged because a recent study suggested similarities in risk factors between patients with previous venous thrombosis and atherosclerosis [ 65 ]. Recent studies clearly show that the risk of recurrent venous thrombosis depends on the one hand on persistent thrombus [ 66 ] as an inflammatory focus of disease as well as on persistent coagulation, i. e. elevated D-dimers, on the other hand [ 67 ]. Genetic influences such as factor V Leiden play no role of importance in the risk of recurrent venous thrombosis. In analogy with residual venous thrombus on the damaged venous vessel wall, advanced atherosclerosis represents a comparably damaged and inflammatory/thrombotic arterial vessel wall. Accordingly, similar plasma risk factors appear to be involved in recurrent venous thrombosis and arterial thrombosis (inflammatory markers and D-dimers). The "Bruneck" model of atherosclerosis and recommendations The influence of blood coagulation on atherosclerosis follows a two stage model in which variants may occur under exceptional conditions. In general in the first phase, roughly covering the first four decades of life, atherosclerosis develops under influence of "classical" risk factors, including hypercholesterolemia and smoking, i.e. both genetic and acquired forces. While fibrinogen/fibrin molecules participate in early plaque lesions, increased activity of systemic coagulation is of no major influence on the risk of arterial thrombosis, except in rare cases where a number of specific procoagulant forces collide. The dominant effect of coagulation is anticoagulant, i.e. thrombin enhances protein C activation through its binding to endothelial thrombomodulin. Defects in the protein C mechanism may indeed precipitate arterial thrombosis, but only under highly thrombogenic conditions. Fibrinolysis limits fibrin accumulation in the intima and herewith progression of plaque lesions. At this stage elevated PAI-1 levels may diminish fibrinolysis and may stimulate plaque progression, which may explain that in a large Japanese study the PAI 4G/5G polymorphisms appeared to be a risk factor for myocardial infarction in women [ 68 ]. The second phase is characterized by advancing atherosclerosis, with greater impact of inflammation and increased infiltration of fibrin in the arterial vessel wall, enforcing pro-inflammatory effects. The extensive interactions between inflammation and coagulation enzymes and inhibitors (in much greater detail than discussed here) amplify the chain of events that determine the risk of atherothrombosis. Inflammation overwhelms protective anticoagulant forces, which in itself may have become less efficient due to down regulation of thrombomodulin (TM) and endothelial cell protein C receptor (EPCR) expression. In this phase, evidence of activated coagulation is measurable in peripheral blood reflecting both the extent of atherosclerotic burden and the systemic clotting tendency, which poses a direct risk of thrombotic complications. This point of view deviates from Tracy's viewpoint and provides a more constructive model for integrating coagulation in arterial disease. We would also propose that D-dimer and other, novel assays such as for endogenous thrombin generation ("endogenous thrombin potential") [ 69 ] or for activated factor XII (suggested as a novel risk factor for cardiovascular disease [ 70 ]), be used to determine prospectively the risk of new atherothrombotic complications and to guide randomized intervention trials in patients stratified on the basis of such plasma markers.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538274.xml
529423	Diversity and Recognition Efficiency of T Cell Responses to Cancer	Background Melanoma patients vaccinated with tumor-associated antigens frequently develop measurable peptide-specific CD8+ T cell responses; however, such responses often do not confer clinical benefit. Understanding why vaccine-elicited responses are beneficial in some patients but not in others will be important to improve targeted cancer immunotherapies. Methods and Findings We analyzed peptide-specific CD8+ T cell responses in detail, by generating and characterizing over 200 cytotoxic T lymphocyte clones derived from T cell responses to heteroclitic peptide vaccination, and compared these responses to endogenous anti-tumor T cell responses elicited naturally (a heteroclitic peptide is a modification of a native peptide sequence involving substitution of an amino acid at an anchor residue to enhance the immunogenicity of the peptide). We found that vaccine-elicited T cells are diverse in T cell receptor variable chain beta expression and exhibit a different recognition profile for heteroclitic versus native peptide. In particular, vaccine-elicited T cells respond to native peptide with predominantly low recognition efficiency—a measure of the sensitivity of a T cell to different cognate peptide concentrations for stimulation—and, as a result, are inefficient in tumor lysis. In contrast, endogenous tumor-associated-antigen-specific T cells show a predominantly high recognition efficiency for native peptide and efficiently lyse tumor targets. Conclusions These results suggest that factors that shape the peptide-specific T cell repertoire after vaccination may be different from those that affect the endogenous response. Furthermore, our findings suggest that current heteroclitic peptide vaccination protocols drive expansion of peptide-specific T cells with a diverse range of recognition efficiencies, a significant proportion of which are unable to respond to melanoma cells. Therefore, it is critical that the recognition efficiency of vaccine-elicited T cells be measured, with the goal of advancing those modalities that elicit T cells with the greatest potential of tumor reactivity.	Introduction The immunotherapy of cancer holds promise in harnessing the host immune response to specifically target tumor cells without harming normal tissues. Strategies involve adoptive cellular therapy or active immune induction (commonly referred to as “cancer vaccination”). Cancer vaccines may consist of whole tumor cells or tumor lysates, but identification of tumor-associated antigens (TAAs) over the past decade has made possible the use of specific proteins or peptides as cancer vaccines. The anti-tumor potential of TAA-specific CD8+ T cells has been illustrated by the demonstrated capacity of adoptive T cell therapy to reduce tumor size [ 1 ]. While endogenous anti-tumor CD8+ T cell responses may already exist in some cancer patients [ 2 ], vaccination with TAA-derived peptides, and in particular heteroclitic peptide analogs, increases the frequency of TAA-specific T cell responses to detectable levels in many patients [ 3 , 4 , 5 , 6 , 7 , 8 , 9 ]. Heteroclitic peptide analogs are created by substitutions at anchor residues resulting in increased association of peptide with the major histocompatibility complex (MHC) [ 10 ]. Consequently, heteroclitic peptide analogs are predicted to be more immunogenic than their native counterparts because of more stable binding at the surface of antigen-presenting cells (APCs). Indeed, T cells capable of tumor lysis have been isolated from patients vaccinated with heteroclitic peptide [ 8 , 11 , 12 , 13 ]. However, the presence of TAA-specific T cells elicited by vaccination often does not correlate with clinical responses [ 3 , 14 , 15 , 16 , 17 ]. Various reasons for the paradoxical coexistence of cancer cells and TAA-specific T cells within patients have been proposed [ 18 , 19 ]. One possibility is that elicited TAA-specific T cells are not optimally functional in vivo [ 2 , 18 ]. Another possibility is that T cells inefficient in tumor recognition or lysis are induced by vaccination [ 20 ]. It is becoming recognized that antigen-specific T cells may have substantially different requirements for cognate peptide (the peptide that is recognizable to a specific T cell clone) for efficient target lysis [ 20 , 21 , 22 , 23 ]. “Recognition efficiency” (RE) (also known as “functional avidity”) is a measure of the sensitivity of a T cell to different peptide concentrations for stimulation [ 24 , 25 , 26 ]. We hypothesized that high antigen densities on APCs resulting from vaccination with heteroclitic peptide may paradoxically drive T cells of predominantly low RE, which are not efficiently activated by the endogenous expression levels of native peptides on tumor cells. Consequently, such T cells would be ineffective in tumor cell destruction. Support for this notion is emerging: T cells with low RE are predominantly expanded in vitro with high peptide concentration [ 22 ]. Moreover, in vitro stimulation of T cells from healthy donors with heteroclitic peptides results in expansion of cells with a wide range of RE [ 23 ]. A similar phenomenon may occur in vivo, leading to TAA-specific T cells of low RE depending on the nature of antigen stimulation [ 20 ]. While isolated T cell clones with low RE have indeed been generated from melanoma patients following heteroclitic peptide vaccination, the proportion of vaccine-elicited T cell responses these cells represent in vivo is not clear. If predominantly high-RE, tumor-cytolytic T cells are generated, then a small fraction of low-RE T cells generated would be of little consequence. However, if predominantly low-RE T cells are generated, then this low proportion of high-RE T cells may be an important factor in the observed lack of clinical effectiveness of current cancer vaccination strategies. To address this important issue, we undertook a systematic examination of the complexity of T cell responses induced by heteroclitic peptide vaccination, and compared these responses to endogenous anti-tumor T cell responses which develop in some patients. Typically, responses to vaccination are examined following in vitro expansion from patient samples, which may alter the composition of cells and consequently not reveal the proportion of cells in vivo having sufficiently high RE to lyse tumor targets. Although staining with peptide–MHC tetramers provides a direct estimate for the number of TAA-specific T cells present in vivo, and intensity of tetramer staining has been employed as a parameter for isolation of high-RE, tumor-lytic T cells [ 27 ], staining intensity does not correlate well with RE or tumor-lytic potential [ 28 , 29 ], and cannot be considered a reliable indicator for the functional status of TAA-specific T cells. To analyze and compare T cell responses in melanoma patients on a single-cell level, we generated and examined a large number of cytotoxic T lymphocyte (CTL) clones derived from post-vaccination or endogenous anti-tumor T cell responses. Each clone was analyzed for T cell receptor (TCR) variable chain beta (VB) expression, RE, and ability to lyse melanoma targets. Importantly, these clones were generated directly ex vivo through tetramer-guided sorting, which minimizes the selection bias that could be introduced by prior in vitro expansion. Therefore, data from these clones could be taken to estimate the complexity of the responses in vivo. Methods Patients and Samples All patients had resected stage III or IV melanoma, as determined by the 1988 modified American Joint Commission on Cancer staging system. They were required to have a magnetic resonance imaging or computed tomographic scan of the head and computed tomographic scans of the chest, abdomen, and pelvis showing no indication of disease within 4 wk of therapy to verify that they were clinically free of melanoma. Eligibility criteria included age 18 y or older, creatinine of less than 180 μmol/l, bilirubin of less than 110 μmol/l, platelet count of 100 × 10 9 /l or more, hemoglobin of 90 g/l or more, and total white blood cell count of 3.0 × 10 9 /l or greater. Tests for human immunodeficiency virus, hepatitis C antibody (Ab), and hepatitis B surface antigen were required to be negative, and all patients were HLA-A2 antigen positive by a microcytotoxicity assay. All patients were required to comprehend and sign an informed consent form approved by the National Cancer Institute (NCI; Bethesda, Maryland, United States) and the Los Angeles County/University of Southern California Institutional Review Board. Analysis of the patient samples was approved by Stanford University's Institutional Review Board. Peripheral blood mononuclear cell (PBMC) samples were isolated from patients after vaccination with the heteroclitic peptides MART 26–35 (27L) (ELAGIGILTV) and gp100 209–217 (210M) (IMDQVPSFV) at the University of Southern California Norris Cancer Center (Los Angeles, California, United States). Clinical-grade peptides used were provided by the Cancer Therapy Evaluation Program of the NCI under an Investigational New Drug application BB 6123 held by the NCI. Immunizations (1 mg of each peptide emulsified with incomplete Freund's adjuvant) were administered every 2 wk for 8 wk, then every 4 wk for 12 wk, and then once 8 wk later. PBMC samples were collected 4 wk after the final immunization and stored at −130 °C. Samples were thawed the day before an experiment for overnight culture in CTL medium. The following morning, viable cells were isolated by ficoll density centrifugation, washed, and resuspended to the appropriate concentration in a solution of 90% Iscove's Modified Dulbecco's Medium (IMDM) and 10% fetal bovine serum (FBS). Flow Cytometry Analysis For isolation and detection of peptide-specific T cells, patient PBMC samples were stained and analyzed by fluorescence-activated cell sorting (FACS) as previously described [ 2 ]. Briefly, cells were stained with anti-human CD8− fluorescein isothiocyanate (Caltag Laboratories, Burlingame, California, United States) and CD19-CyChrome (BD Biosciences, Palo Alto, California, United States) Abs, and HLA-A0201/peptide tetramer–phycoerythrin (PE). The final staining dilution of each Ab was 1/200 and 1/80, respectively. Tetramer–PE was titrated for optimal staining, usually between 1 and 10 μg/ml. For TCR VB typing, cells were divided in seven aliquots and stained with CD8 PerCP-Cy5.5 (BD Biosciences), tetramer–PE, and a panel of two or three different anti-VB monoclonal Abs labeled with fluorescein isothiocyanate, allophycocyanin (APC), or both. Cells were incubated at room temperature for 30 min, washed, then analyzed using a two-laser, four-color FACSCalibur (Becton Dickinson, Franklin Lakes, New Jersey, United States) or sorted using a FACSVantage flow cytometer (Becton Dickinson). Lymphocytes were identified by forward and side scatter signals, then selected for CD8+ and tetramer positive. Up to one million events were acquired and analyzed using FlowJo (TreeStar, San Carlos, California, United States). CD107 Mobilization Assays Target cells The HLA-A0201-positive melanoma lines Malme-3M and A375 and the T2 cell line were purchased from ATCC (Manassas, Virginia, United States) and maintained according to instructions provided by the ATCC. The HLA-A0201-positive melanoma line mel526 was obtained from the Surgery Branch of the NCI. While Malme-3M and mel526 express both MART and gp100, A375 does not express MART or gp100 and served as a negative control. Expression (or lack thereof) of these antigens by each cell line was further confirmed by immunohistochemical staining. Cells were trypsinized using Trypsin/EDTA solution (GIBCO, San Diego, California, United States) before use. T2 cells were HLA-A2.1+ and were pulsed prior to assays with peptides indicated in the text. Effector cells Effector cells, which include clones, cell line, and PBMC samples, were frozen and analyzed in batches. The cells were thawed the day before an experiment for overnight culture in CTL medium. The following morning, viable cells were isolated by ficoll density centrifugation, washed, and resuspended to the appropriate concentration (usually 10 7 /ml) in CTL medium. Experimental procedure All assays were done at least twice, with duplicates for each condition. The effector to target (E:T) ratio used was generally 1:2, with 2 × 10 5 for clones or 10 6 for the cell line and patient PBMC samples. To each well, the following was added in order: 1 μl of 2 mM monensin (Sigma, St. Louis, Missouri, United States) in 100% EtOH, 100 μl of target cells, 100 μl of effector cells, and 1 μl of CD107-APC Abs. The cells were mixed well using a multichannel pippetor. The plate was centrifuged at 300 g for 1 min to pellet cells, then placed into an incubator at 37 °C for 4 h. After the incubation, the plates were centrifuged to 500 g to pellet cells, and the supernatant was removed. Cell–cell conjugates were disrupted by washing the cells with PBS supplemented with 0.02% azide and 0.5 mM EDTA, and mixed vigorously, then stained with additional Abs. Generation of CTL Clones CD8+ T cell clones were derived by FACSorting individual tetramer-positive cells from PBMC samples prepared for flow cytometry as described above. CD8+ tetramer-positive T cells were sorted under sterile conditions into 96-well plates, one cell per well, using a FACS Vantage (Becton Dickinson). Wells contained 100 μl of CTL IMDM, with 10% FBS, 2% human AB sera, and penicillin, streptomycin, and L-glutamine, supplemented with 100 units/ml IL-2. Sorted cells were expanded in vitro using standard protocols. Briefly, irradiated feeder cells (JY cells and fresh PBMCs) were added to wells containing the sorted T cells, and the 96-well plates were incubated at 37 °C with 7% CO 2 to allow for growth. Potential clones became visible around day 14 and were then transferred to 24-well plates containing 1 ml of CTL medium with 100 units/ml IL-2. Wells were selected based on cell confluency for expansion and further analysis. Clones confirmed to be tetramer-positive were expanded in T-25 flasks containing irradiated JY cells and fresh PBMCs in 25 ml of CTL medium containing PHA. IL-2 was added to a final concentration of 50 units/ml on day 1 and then every 2 d thereafter for 2 wk. Cytotoxic Assays Target cells Target cells were as described above under CD107 Mobilization Assays, and were labeled overnight with 51 Chromium, washed, and resuspended to 10 5 cells/ml. One hundred microliters of target cells were incubated with 100 μl CTL clones at 10:1 E:T ratio for 4 h. Percent specific release of 51 Chromium from target cells was calculated from 40-μl cell-free supernatants. Determination of RE Chromium-labeled T2 targets were pulsed with a range of peptide concentrations, generally starting at 10 −7 M and decreasing by log steps to 10 −13 M. T cell clones were incubated with T2 targets at 10:1 E:T ratios for 4 h, then chromium release was measured and percentage cytotoxicity calculated by standard methods. Prior to each cytotoxicity assay, clones underwent ficoll-hypaque centrifugation to remove dead feeder cells and were determined to be greater than 80% CD8+ tetramer-positive T cells by FACS. The E:T ratio was based upon live T and target cells. For each T cell clone, percent cytotoxicity was plotted against peptide concentration. The peptide concentration at which the curve crossed 40% cytotoxicity was defined as the RE of that clone [ 30 ]. Microcytotoxic assay Cells were isolated directly from PBMCs from patient 422 by FACS as described above. Cells were collected in microfuge tubes containing 1 ml of ice-cold 90% IMDM with 10% FBS. Collected cells were washed and resuspended to 83,300 cells/ml in 90% IMDM with 10% FBS. Targets were prepared as described above and resuspended to 8,300 cells/ml in 90% IMDM with 10% FBS. A total of 2,500 sorted cells (30 μl) and 250 target cells (30 μl) were transferred to a microcentrifuge tube (VWR International, West Chester, Pennsylvania, United States), centrifuged 1 min at 200 g, and incubated 4 h at 37 °C. Percent specific release of 51 Chromium was calculated from 40 μl of cell-free supernatant. TCR VB Spectratyping RNA was extracted from clones and tetramer-positive cells using TRIzol (Invitrogen, Carlsbad, California, United States) and reverse-transcribed into cDNA using SuperScript II Reverse Transcriptase (Invitrogen). PCR was performed using 34 different 5′ primers that specifically amplify all functional TCR VB genes. Most of the 5′ primers used have been previously described [ 31 ]. These primers were used in combination with a common 3′ primer based in the beta chain constant region, BC63 (5′- GTGTGGCCTTTTGGGTGT-3′). As an internal control, PCR for a section of the beta chain constant region was performed in parallel with VB-specific PCRs using the following primers: UpBC (5′- CGCTGTGTTTGAGCCATC-3′) and LoBC (5′- TGCTCAGGCAGTATCTGGA-3′). All primer concentrations were 200 nM. PCR was performed using an iCycler iQ thermic cycler equipped with a real-time detection system (Bio-Rad, Hercules, California, United States) and a QuantiTect SYBR Green PCR kit (Qiagen, Valencia, California, United States). PCR reactions were performed as follows: 94 °C for 9 min, followed by 50 cycles of 94 °C for 30 s, 58 °C for 1 min, and 72 °C for 1 min, followed by 72 °C for 10 min. Specific amplification was determined relative to constant region control PCR. For spectratyping, PCRs were performed as described above with the following VB14- and VB17-specific 5′ primers: VB14m (5′- ACCCAAGATACCTCATCACAG-3′) and VB17 (5′- GACAGGACCCAGGGCAAG-3′), followed by a run-off PCR with downstream VB-specific primers: VB14 (5′- GGGCTTAAGGCAGATCTACT-3′) and VB17m (5′- TTTCAGAAAGGAGATATAGCT-3′), and FAM6-labeled BC63 3′ primer. Run-off PCR was performed as described above except that only five cycles of PCR were run with the 55 °C annealing temperature and QuantiTect Probe PCR kit (Qiagen). Labeled PCR fragments were run on an ABI Prism 377 DNA Sequencer (Applied Biosystems, Foster City, California, United States) and analyzed using GeneScan software (Applied Biosystems). Statistical Analysis A standard software package (SigmaPlot 5.0, Systat Software, Richmond, California, United States) was used to provide descriptive statistical plots. Barcharts were provided with standard errors on them. Linear plots were provided with standard errors computed at each point. A linear regression (using least squares) of percent specific lysis on recognition efficiency is shown in Figure 5 A and 5 B. Figure 4 High RE Recognition of Native G209n but Not G209–2M Peptide Correlates with Efficiency in Tumor Cell Lysis CTL clones 476.105 and 132.1 were assayed for lysis of T2 cells pulsed with 10-fold dilutions of (A) native or (C) heteroclitc peptide at concentrations ranging from 100 fg/ml to 100 ng/ml. (B) Lysis of Malme-3M melanoma cells by 476.105 and 132.1 CTLs. All assays were performed in triplicate, and each clone was assayed twice. Error bars reflect variation between two separate assays. Figure 5 Endogenous T-Cell Responses Have Higher RE Than Vaccine-Elicited Responses CTL clones representing different tetramer-positive populations in each patient expressing different VB were assayed for lysis of T2 cells pulsed with various dilutions of G209n, G209–2M, M27, or M26 peptides in 51 Chromium release cytotoxicity assays as described in Figure 4 legend. A RE score was attributed to each clone equal to the negative log 10 of the peptide concentration that resulted in 40% lysis of peptide-pulsed T2 cells. (A and B) RE scores for both (A) MART-specific and (B) gp100-specific clones from all patients were correlated with efficiency in lysing melanoma cells. Correlation coefficients were 0.66 for MART-specific clones and 0.81 for gp100-specific clones. (C–F) Comparison of RE scores for endogenous (patients 461 and 132) and vaccine-induced (patients 517, 520, 422 and 476) responses. (C and D) RE analysis with native peptides (C) M27 and (D) G209n. Mean RE (weighted) for each response is indicated with horizontal bars. Weighted means were based on all clones, not only those assayed, and were estimated by summing the RE of each analyzed clone multiplied by the number of total clones expressing the same VB, in each patient. Weighted means were as follows: patient 517, 5.7; patient 520, 7.0; patient 461, 7.9; patient 422, 9.7; patient 476, 9.9; and patient 132, 11.2. One-tailed T-tests demonstrated that endogenous responses had significantly higher RE than vaccine-induced responses: patient 461 versus patient 517, p = 1.8 × 10 −5 ; patient 461 versus patient 520, p = 1.1 × 10 −3 ; patient 132 versus patient 422, p = 6 × 10 −6 ; and patient 132 versus patient 476, p = 4.3 × 10 −4 . (E and F) RE analysis with heteroclitic peptides (E) M26 and (F) G209–2M. Weighted means were as follows: patient 517, 10.6; patient 520, 11.1; patient 461, 11.2; patient 422, 10.5; patient 476, 11.6; and patient 132, 11.3. Results T Cell Responses to TAAs in Patients with Melanoma To address the complexity of T cell responses against melanoma in vivo, patients with vaccine-induced or endogenous TAA-specific responses were selected. In recent cancer vaccine trials [ 3 , 4 , 5 ], many melanoma patients who received heteroclitic peptide vaccines gp100 209–217 (210M) (IMDQVPSFV; G209–2M) and MART 26–35 (27L) (ELAGIGILTV; M26) had measurable CD8+ peptide-specific T cell responses in PBMCs detected by peptide–MHC tetramer staining. In addition, TAA-specific T cell responses could be detected in some patients without vaccination, suggesting the existence of an endogenous anti-tumor T cell response in these patients. For the current study, we selected samples from six melanoma patients from these trials—four with vaccine-elicited responses (patients 422, 476, 517, and 520) and two with endogenous T cell responses (patients 132 and 461)—for detailed analyses of TCR VB usage, RE for the target peptide, and tumor cytotoxicity. The samples from these six patient had peptide-specific T cell populations detectable with G209–2M-tetramers (patients 422, 476, and 132) or M26-tetramers (patients 517, 520, and 461) ranging from 0.1% to 2.5% of total CD8+ T cells ( Figure 1 A). Figure 1 Melanoma Patient Samples Selected for Analysis of RE for Melanoma Cells (A) Six patients with T cell responses reactive with for M26 or G209–2M tetramers were selected for analysis. PBMCs from each patient were stained with PE-conjugated peptide–MHC tetramers, G209–2M-tet PE or M26-tet PE, and co-stained with anti-CD8 fluorescein isothiocyanate and anti-CD14, -CD19, and -CD4 Cy5PE. The plots shown are gated for CD8+, CD14−, CD19−, and CD4− cells. Tetramer-positive cells are boxed and estimated for percent of total CD8+ cells: patient 422, 2.5%; patient 476, 0.31%; patient 132, 0.22%; patient 517, 0.23%; patient 520, 0.12%, and patient 461, 0.50%. (B) Microcytotoxicity 51 Chromium release assay with tetramer-positive cells isolated by FACS from the CD8+ PBMC population from patient 422. Isolated cells were assayed for lysis of T2 cells treated with relevant or irrelevant peptide, or mel526 melanoma cells. Sorted cells were combined with 250 target cells at 13:1 E:T ratios for 4 h, and supernatants were assayed for percent specific release of radiolabel. Vaccine-Elicited T Cells Are Functional Directly Ex Vivo but of Variable Tumor Reactivity Patient 422 had the largest detectable TAA-specific CD8+ T cell response (2.5% G209–2M-tetramer-positive) and thus sufficient numbers for examination of lytic function immediately following isolation. To test whether peptide-vaccine-induced T cell responses were functionally active directly ex vivo, T cells isolated by G209–2M-tetramer-guided cell sorting from patient 422 were tested for lysis of peptide-pulsed and melanoma target cells in microcytotoxic assays ( Figure 1 B). The directly isolated tetramer-positive T cells from this patient specifically lysed T2 cells pulsed with high concentrations (1 μg/ml) of G209–2M and native (G209n) peptides, but not with T2 cells pulsed with a cytomegalovirus-derived, HLA-A0201-restricted peptide (NLVPMVATV) or melanoma targets. This suggests that while a significant portion of the vaccine-elicited T cells from patient 422 may be functional in vivo, they did not have significant tumor lysis activity. To assess the functional status of the smaller TAA-specific CD8+ T cell responses in the other five patients—which were too small for direct cytotoxicity assays after sorting—we utilized a novel FACS assay for degranulation based on CD107 mobilization [ 24 ]. All six TAA-specific populations exhibited robust functional responses ex vivo, as measured by percentage of G2090–2M- and M26-tetramer-positive cells that mobilized CD107 and/or downregulated the CD3 complex upon incubation with T2 cells pulsed with cognate peptides ( Table 1 ; 86%-99.6%). In response to melanoma targets mel526 and Malme-3M, which both express gp100 and MART-1 and are HLA-A0201 positive, the two endogenous TAA-specific responses (samples from patients 132 and 461) also exhibited robust functional responses directly ex vivo ( Table 1 ; 36.8%–87%), and these responses were specific as they had little response to A375, a HLA-A0201-positive melanoma cell that does not express gp100 or MART-1 and served as a negative control for antigen-specific killing ( Table 1 ; 2.7% and 3%). In contrast, the vaccine-elicited responses exhibited much lower reactivity to mel526 and Malme-3M ( Table 1 ; 23.8%–32.5%). These data demonstrate that all six TAA-specific CD8+ T cell responses were functional ex vivo, but there were significant differences in reactivity to melanoma targets between endogenous and vaccine-elicited responses. Table 1 Functional Status of TAA-Specific T Cell Response Functional response was determined by percent of G209–2M- and M26-tetramer-positive cells that mobilized CD107 and/or downregulated CD3 complex in response to incubation with T2 cells pulsed with 100 ng/ml of cognate peptide (G209–2M or M26), mel526, or Malme-3M melanoma cells. A375 melanoma cells served as negative control To substantiate the generality of these findings, we analyzed four additional patients with vaccine-elicited responses. One subject responded to G209–2M only (patient 722), one to M26 only (patient 713), and two to both G209–2M and M26 (patients 721 and 735). Similar to the first four vaccine-elicited patients, these four additional patients (six TAA-specific responses in total) exhibited variable reactivity to melanoma targets, ranging from 13% to 49.6% ( Table 1 ). Vaccine-Elicited T Cells Have Varied Capacity to Lyse Melanoma Targets To confirm and further investigate the differences in tumor reactivity between endogenous and vaccine-elicited responses, we reasoned that analysis of a set of clonal CTL lines that represented the tetramer-positive population would provide an accurate estimate of the complexity of the TAA-specific T cell response in each patient. A large number of clonal CTL lines (more than 200) were generated by FACS of individual G209–2M- and M26-tetramer-positive cells directly from PBMC samples ( Table 2 ). Up to 85% of sorted cells expanded in various sorts (data not shown). Randomly selected expanding clones and the tetramer-positive population from which they were derived were examined for TCR VB expression using TCR VB-specific monoclonal Abs and VB-specific primers in PCR. Diverse TAA-specific T cell responses were found in the four vaccinated patients, with multiple T cells expressing different TCR VB, while the two endogenous responses were less diverse. All but one clone derived from patient 132 expressed VB17, while two dominating T cell populations in patient 476 expressed VB14 and VB17 ( Table 2 ). The clonality of the dominant populations in these patients was evaluated by PCR fragment length analysis ( Table 3 ). Identical length fragments were demonstrated in the four selected clones from 476 BV14+ and 476 BV17+ populations. Identical length fragments were also demonstrated in all BV17+ clones from patient 132. Furthermore, analysis of sorted tetramer-positive cells from patient 476 demonstrated single fragment sizes for BV14 and BV17, which were identical to the fragment sizes generated from the selected clones, arguing for clonality of these dominant populations ( Table 3 ). Table 2 CTL Clones Established from Each Patient Represent a Random Selection from the Tetramer-Reactive CD8+ Parent Population a Clonal CTL lines were established from each patient b Number of clonal CTL lines from each patient expressing the same TCR VB chain c Percent of G209–2M- and M26-tetramer-reactive CD8+ T cells in each patient expressing the indicated TCR VB chain Table 3 PCR-Generated Fragment Length Analysis PCR fragment length was determined for selected clones and for sorted G209–2M- or M26-tetramer-positive populations from which the clones were derived. Numbers indicate length in base-pairs of fragments generated by PCR with VB14 or VB17 5′ primer and BC63 constant region 3′ primer followed by a run-off reaction with VB-specific nested primers and FAM-labeled BC63 primer. Fragments were analyzed using an Applied Biosystems 377 automated sequencer and GeneScan software Table 4 CTL Clones from Each Patient Selected for Functional Analysis The TCR VB usage of each CTL clone was determined using a panel of 19 anti-VB monoclonal Abs by flow cytometry or by PCR with 34 VB-specific primers. All clones selected for functional analysis were assayed for lysis of melanoma cells. Some clones were also subjected to RE analysis M, assay for lysis of melanoma cells; MR, RE analysis Peptide specificity and CD8 expression of each clone was confirmed by staining with G209–2M- and M26-tetramers and anti-CD8 monoclonal Ab (data not shown). To obtain an accurate reflection of the total T cell population detected with tetramer in each patient, we decided to rigorously examine at least one representative clone for each subpopulation expressing a different TCR VB ( Table 4 ). Multiple clones were analyzed to determine dominating populations. From patients 132, 517, and 461, for which fewer clones were generated, all clones were included in the analyses ( Table 4 ). To determine the effectiveness of tumor lysis by the different TAA-specific T cell clones that were propagated, clones were analyzed for their ability to lyse melanoma cell lines mel526 and Malme-3M. A375 cells served as a control for antigen-specific killing. In addition, each CTL clone was examined for antigen-specific lysis of T2 cells pulsed with high levels (1μg/ml) of G209–2M or M26 peptides. “Efficient lysis” in these experiments was defined as 40% or greater specific release of radiolabel from the target cells; 10% or less specific release was categorized as “low or no lysis,” and 10% to 40% was termed “intermediate lysis.” All but two of the CTL clones elicited from endogenous anti-tumor responses (from patients 132 and 461) exhibited “efficient lysis” of both the mel526 and Malme-3M melanoma cell lines ( Figure 2 ). In contrast, only a few clones from the vaccine-elicited responses (from patients 422, 476, 520, and 517) efficiently lysed melanoma cells. The majority of clones examined from these vaccine-elicited responses either failed to lyse melanoma targets altogether or lysed them with intermediate efficiency ( Figure 2 ). This lack of efficiency in melanoma cell lysis was not due to cellular dysfunction, since each clone efficiently lysed T2 cells pulsed with high levels of relevant, but not irrelevant, peptide ( Figure 2 ). Overall, the majority of clones derived from endogenous anti-tumor responses (patients 132 and 461) lysed both mel526 and Malme-3M melanoma target cells more efficiently than clones from vaccine-elicited responses (patients 422, 476, 520, and 517) ( Figure 3 ). These findings suggest that TAA-specific T cells elicited by heteroclitic peptide vaccination have different tumor-cytolytic potentials from those which develop endogenously to cancer. Figure 2 Endogenous T Cell Responses Are More Efficient in Melanoma Lysis Than Vaccine-Elicited Responses Cells from 87 clonal CTL lines were assayed for lysis of melanoma cells mel526, Malme-3M, and A375 in 51 Chromium release cytotoxicity assays. Mel526 and Malme-3M are HLA-A2.1+ and express both gp100 and MART-1. A375 cells are HLA-A2.1+ but do not express either gp100 or MART-1 and served as a negative control. T2 cells treated with 1 μg/ml G209–2M or M26 peptides served as controls for antigen-specific lysis. The CTL clones assayed were selected to represent different tetramer-positive subsets expressing different VB. Dominating tetramer-positive populations in each patient were represented with two or more clones. Each CTL clone was assayed in triplicate wells, and the data displayed are averages of two different experiments. Clones from the same patient expressing similar VB while exhibiting different lysis potential were viewed as separate subsets. Each assay was performed at 10:1 E:T ratio as detailed in Methods. The height of each bar represents percent specific lysis, while the width represents the relative size of the tetramer-positive subpopulations (defined by VB expression) in each patient. Population size was defined as the percent of clones from each patient expressing the same VB. Error bars show standard deviation between two experiments within each clone and/or between different clones where more than one clone was analyzed. Figure 3 Most CTL Clones Isolated from Endogenous Responses Are Efficient in Tumor Cell Lysis CTL clones derived from each patient were classified as “efficient” (greater than 40%), “intermediate” (between 10% and 40%), or “low/no” (less than 10%) in lysis of melanoma cells based on data displayed in Figure 2 . Each bar represents the portion of total clones from each patient with “efficient,” “intermediate,” or “low/no” melanoma lysis potential. RE for Native and Heteroclitic Peptides of T Cells from Endogenous or Vaccine-Elicited Responses We hypothesized that CTL clones that did not efficiently lyse melanoma targets may be incapable of recognizing the relatively low surface densities of native peptide present on tumor cells. CTL clones selected for analysis of tumor lysis were also assessed for RE for the native and heteroclitic peptides via a ten-log range of dilutions. This is illustrated with clones 132.1 and 476.105 ( Figure 4 A). There were considerable differences in killing of peptide-pulsed T2 cells by these two clones. The differences in RE for G209n native peptide displayed by the two clones highlighted in Figure 3 A correlated with their ability to lyse melanoma cells: the high-RE clone 132.1 efficiently lysed melanoma targets, whereas the low-RE clone 476.105 did not ( Figure 4 B). In contrast to the differences in RE for G209n peptide, similar assays revealed little difference in RE of the two clones for G209–2M heteroclitic peptide ( Figure 4 C), suggesting that these clones recognize the native and heteroclitic peptides differently, and that RE for the native, but not heteroclitic, peptide correlates with tumor-lytic potential. Similar RE assays were performed for the remaining clones from each patient selected for analysis. In order to compare REs of various CTL lines, each clone was assigned an RE score expressed as the negative log 10 value of the peptide concentration required for 40% specific lysis at an E:T ratio of 10:1. For clones 132.1 and 476.105, these scores were 11.1 and 8.3 for assays with G209n peptide ( Figure 4 A), and 11.2 and 11.2 for assays with G209–2M heteroclitic peptide ( Figure 4 C), respectively. We compiled the data on clones from all patients, which showed a correlation between tumor-lytic potential and RE for native peptide ( Figure 5 A and 5 B). Overall, clones generated from endogenous anti-tumor responses had higher RE for the native peptide than clones generated from post-vaccine responses ( Figure 5 C and 5 D). We estimated the composite RE of the overall TAA-specific response (composed of a heterogeneous population of T cells) in vivo by summing the RE of each clone multiplied by its representation in the original mixture (the representation was estimated based on the proportion of TAA-specific cells expressing the same VB as the clone). These composite RE values are represented in Figure 5 as horizontal bars for each response. Clearly, the endogenous responses (patients 461 and 132) had a higher overall, and more homogeneous, RE for the native peptide than the vaccine-elicited responses (patients 422, 476, 517, and 520) ( Figure 5 C and 5 D). Importantly, the vaccine-elicited clones also exhibited wide variations in RE even for the heteroclitic peptide, compared to the endogenous responses ( Figure 5 E and 5 F). This suggests that the variation in RE for native peptides, and hence ability to lyse tumor cells, for vaccine-elicited responses is not merely a reflection of differential recognition of native and heteroclitic peptides by many clones. Rather, variations in RE may be a function of the manner in which these cells were elicited in vivo via vaccination. Discussion To achieve maximal clinical responses, the majority of T cells elicited by vaccination in cancer patients should be capable of responding to tumor targets. We have undertaken the most detailed analysis to date, on a single-cell level, of T cell responses elicited by cancer vaccination and have compared these with endogenous anti-tumor responses. To evaluate the full spectrum of T cells elicited in each patient by vaccination, we utilized tetramers made with the vaccine peptides (heteroclitic M26 and G209–2M) to isolate such cells. CTL clones were selected directly from patient PBMC samples without enrichment in culture to closely reflect the composition of the antigen-specific T cell response in vivo at the time of isolation. Our data revealed that T cell populations induced by vaccination were significantly different from endogenous responses: while some CTLs elicited by vaccination could kill melanoma targets, most were inefficient in tumor cell lysis. In contrast, nearly all clones from endogenous responses were efficient at melanoma cell lysis. This difference was related to RE for the native peptide. Clones that did not lyse tumor cells required up to 10 3 -fold higher concentration of peptide for similar levels of lysis of targets compared to T cell clones that were tumor-lytic. Side-by-side comparison of endogenous responses and vaccine-induced responses suggests that low RE TAA-specific T cell responses may be preferentially driven by heteroclitic peptide vaccination. Thus, high doses of peptide and/or the higher levels of expression of heteroclitic peptide on APCs may induce and actively propagate predominantly T cells with RE too low for recognition of physiological levels of the native peptide present on tumor targets. These data suggest an inverse relationship between antigen density and the RE of T cells elicited. This would be an important consideration in design of future vaccine strategies. Differential recognition of native and heteroclitic peptides by many T cells may also account for the induction of non-tumor-lytic clones by heteroclitic peptide vaccines, which has been suggested previously [ 23 , 32 ]. However, our data suggest that epitope density may be the dominant driving factor for RE in vivo. In all of the vaccine-elicited T cell responses, many of the T cells generated were either of low or intermediate RE not only for the native peptide, but also for the heteroclitic peptide, and exhibited no or intermediate lysis of tumor targets. In contrast, nearly all of the clones generated from the endogenous responses were of high RE. This suggests that the high dosage of peptides administered in vaccinations and the increased binding capacity of heteroclitic peptides to MHC molecules—the very quality that provides them with increased immunogenicity—drive the induction of many T cells with low RE for both heteroclitic and native peptides. Another implication of this study is that the number of cells measured by current methods, including ELISPOT or staining with MHC tetramers, may not correlate directly with the RE or tumor reactivity of T cell responses to vaccination. For example, of the nine clones analyzed from patient 517, none were efficient in tumor cell lysis, yet these cells were detectable by MHC tetramer staining. T cells with low RE for native TAA do not efficiently lyse tumor, and therefore are unlikely to have an impact on clinical outcome. Furthermore, it may be possible that low-RE TAA-specific T cells may interfere with elicitation of high-RE T cells, either by direct competition for antigen on APC surface [ 33 , 34 ] or down-modulation of peptide–MHC complexes. Our data support the notion that not only quantity, but quality, of the T cell response elicited by vaccination may be important for clinical efficacy. There are a number of strategies to increase the magnitude of T cell responses to peptide vaccines. These include using various adjuvants, such as incomplete Freund's adjuvant and immunomodulatory agents, such as IL-12 [ 4 ], GM-CSF [ 5 ], anti-CTLA-4 Abs [ 35 ], or heat shock proteins [ 36 ]. Thus far, none of these approaches have produced improved clinical outcomes. Our data suggest that in addition to driving higher numbers of vaccine-elicited T cells, strategies to modulate the relative RE of T cell responses are also needed. While the selective activation of high- versus low-RE T cells is relatively easy to manipulate in vitro via stimulation with limiting amounts of peptides, this may be more difficult to control in vivo. It is important to bear in mind that signals needed to drive a de novo naïve T cell response may be different from those required to drive further expansion of an activated T cell population [ 37 ]. Thus, a complete vaccination strategy may involve an initial induction phase, followed by progressive shaping of the response to higher RE. Although heteroclitic peptide vaccination may drive T cells of mixed high and low RE, such a strong stimulus may be needed to induce an initial de novo T cell response. Studies in mice suggest that once activated, effector CD8+ T cells may have an increase in RE of up to 70-fold compared to naïve cells [ 38 , 39 , 40 ]. Thus, naïve TAA-specific T cells, with inadequate RE to become activated by low densities of native peptides present on tumor cells, may become efficient in tumor lysis upon vaccination with heteroclitic peptide. This notion has support from studies in tolerized mice: vaccination with a heteroclitic peptide analog recruited T cells, which were responsive to secondary stimulation with native peptide [ 41 , 42 ]. Therefore, optimized use of heteroclitic peptide to induce an initial peptide-specific T cell response, followed by selective expansion of the highest RE tumor-lytic T cells may be needed for an effective strategy with clear clinical application. In summary, we have demonstrated that vaccination with heteroclitic peptide at high concentrations may drive T cell responses of variable tumor-cytolytic potential in cancer patients—and that the ability to lyse tumor cells correlates with the T cell's RE for native peptides. This represents an important—but not sole—factor in explaining the lack of correlation between immunological and clinical responses after vaccination for cancer. Importantly, the situation is different in endogenous responses, in which cells are predominantly of high RE. This suggests that the manner in which T cells are elicited in vivo are different in these two settings and may underlie their differences in biology. Patient Summary Why Was This Study Done? Our immune system protects us against infectious diseases. It can also recognize and destroy early cancer cells before they form tumors. Researchers have been trying to find a way to boost the anti-cancer function of the immune system so that it can kill even established tumors. This is the idea behind developing vaccines for treating cancer—the vaccine alerts and boosts the patient's immune system and so helps to fight the cancer. The idea of enlisting the immune system against cancer has been around for a long time. There have been some spectacular successes, but it has proven difficult to find vaccines that work in more than just a few patients. And we don't yet understand why vaccines seem to work in some patients but not in others. What Did the Researchers Do? Peter Lee and colleagues are trying to find out why some patients respond to vaccines and others don't by looking at the immune response in vaccinated patients. In this study, using state-of-the art technology, they examined patients who received different vaccines against the skin cancer melanoma. They concentrated on the so-called killer T cells (cytotoxic T cells), which directly attack and kill tumor cells, and analyzed them in great detail. What Did They Find? Most cytotoxic T cells produced by patients after vaccination—including vaccination with so-called heteroclitic peptides that had been specifically designed to provoke a very strong immune response—did not kill tumor cells very well, but a few of them did. These results provide some explanation as to why cancer vaccines haven't been as successful as many had hoped, but also suggest that if it were possible to get more of the potent T cells or to expand the ones that are already produced with the current vaccines, there would be a stronger anti-tumor response. What Next? How to get effective cancer vaccines remains an open question. But at least technologies such as those used in this study now exist that allow researchers to analyze how the immune systems of different patients react to vaccination and hence can guide the development of better vaccines. Additional Information. The Cancer Research Institute: http://www.cancerresearch.org/ US Food and Drug Administration page on cancer vaccines: http://www.cancerresearch.org/ University of Michigan page on cancer vaccines: http://www.cancer.med.umich.edu/learn/cancervaccines.htm	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529423.xml
524488	Effect of splenectomy on type-1/type-2 cytokine gene expression in a patient with adult idiopathic thrombocytopenic purpura (ITP)	Background In view of clinical observations and laboratory results that support a central role of the spleen in idiopathic thrombocytopenic purpura (ITP) pathophysiology, we studied the effect of splenectomy on type-1 and type-2 cytokine gene expression in an adult ITP case, refractory to conservative treatment. Case presentation The patient was subjected to splenectomy 9 months after the diagnosis with complete response, attaining platelet counts over 150 × 10 6 /L within 10 days after the operation. Two consecutive blood samples were obtained from the patient, 3 and 7 months after the splenectomy for the purposes of this study. A control group consisted of 11 healthy adults. Peripheral blood mononuclear cells were prepared from each blood sample and cultured in vitro for 8 h with the addition of the mitogens phorbol myristate acetate and ionomycin. Total cellular RNA extracted from 10 6 cells was submitted to semiquantitave reverse transcriptase-polymerase chain reaction (RT-PCR) for the amplification of IL-2, IFN-γ, IL-4, IL-5, and IL-10 metagraphs. The PCR products were run on ethidium-stained agarose gels, photographed and quantified by densitometry. A steep decrease of type-1 cytokine expression (IL-2, IFN-γ) and their calculated sum expressing Th1 activity was observed at 7 months post-splenectomy compared to 3 months post-splenectomy, in parallel with a rise of platelet count from 190 × 10 6 /L to 265 × 10 6 /L. The change of type-2 cytokine expression (IL-4, IL-5, IL-10) was slight and the Th2 activity (IL-4+IL-5) remained largely unchanged. The Th1/Th2 ratio, that reflects the pathogenic disease-specific T-cell immune deviation, was accordingly reduced 7 months post-splenectomy (Th1/Th2 = 1.3) compared to 3 months (Th1/Th2 = 3.5). Conclusions The reduction of the Th1/Th2 cytokine ratio that was observed over time after splenectomy was accompanied by full clinical remission. Nevertheless, the persistence of a type-1 polarization, even after several months following spleen removal, is suggestive of a more basic abnormality of the immune function in these patients.	Background Adult autoimmune thrombocytopenic purpura (ITP) is a chronic acquired organ-specific autoimmune thrombocytopenic syndrome [ 1 ]. The low peripheral platelet concentration observed in ITP is the result of reduced platelet life span because of their early removal from the peripheral blood by the activated reticuloendothelial system of the spleen, liver or bone marrow, after their sensitization by autoantibodies that recognize their surface glycoprotein antigens [ 2 ]. Apart from phagocytosis, destruction mechanisms include complement activation [ 3 ] and direct cellular attack by T lymphocytes [ 4 ]. Ineffective thrombocytopoiesis because of autoimmune attack of megakaryocytes in the bone marrow contributes to the thrombocytopenia with varying degrees among cases [ 2 ]. The production of platelet autoantibodies by B-cells is driven by activated platelet-specific autoreactive T-cells [ 5 ]. The phenotype of the disease-specific T helper cells has been shown to be skewed towards type 1 cytokine production [ 6 - 8 ]. The spleen is considered to be the primary site of the autoimmune response where initiation, maintenance and regulation of the autoimmune attack take place. The spleen is the site of autoreactive T- and B-cell interaction and activation, and autoreactive anti-platelet antibody production [ 9 ]. Platelet destruction is also sited mainly in the spleen in most patients [ 2 , 10 , 11 ]. Splenectomy is followed by reduction of autoantibody peripheral blood titre [ 12 , 13 ]. Spleen cells isolated from ITP patients produce antiplatelet immunoglobulin in in vitro cultures [ 14 ]. The percentage of T- and B-cells with activated phenotype is far greater in the spleen than in the peripheral blood of ITP patients [ 9 ]. The number of circulating autoreactive anti-platelet T- and B-cells declines after splenectomy that leads to clinical remission, whereas the peripheral blood concentration of CD3 + CD4 + , CD3 + CD8 + , CD3 + HLADR + , and CD3 + CD25 + cells increases significantly in ITP patients refractory to splenectomy [ 15 ]. Changes in the histology of the spleen have been observed in ITP patients and include follicular hyperplasia, foam macrophages, and extramedullary hematopoiesis, among others [ 16 ]. Splenectomy is the most clinically effective therapeutic intervention in ITP patients, resulting in complete remission in two thirds of the patients with more than 60% maintaining the therapeutic effect in the long term [ 17 , 18 ]. Irrespective of clinical response, splenectomy seems to affect the natural history of the disease and to enhance the response of ITP patients to other treatments that follow [ 19 ]. Given the important immunoregulatory role of the spleen, we looked at the effects of splenectomy on immune activation and immune deviation indices in the peripheral blood of an ITP patient after splenectomy in association with peripheral platelet counts. Case presentation A 42 year old woman presented in October 1999 in Patras University Hospital (PUH) with lower limb purpura and low platelet count (7 × 10 6 /L). Following clinical exclusion of causes of secondary thrombocytopenia [ 20 ] the diagnosis of ITP was reached. The patient initially received glucocorticoid treatment to which she showed a temporary response until 6 months later when she relapsed. She was subsequently started on danazol without any clinical benefit. Intravenous immune globulin administration also proved ineffective after two 5-day cycles. As a result, the patient was subjected to splenectomy 9 months after the diagnosis with complete response, attaining platelet counts over 150 × 10 6 /L within 10 days after the operation. Five years later, she remains in clinical remission. Two consecutive blood samples were obtained from the patient, 3 and 7 months after splenectomy for the purposes of this study. A control group consisted of 11 adult healthy volunteers (6 women and 5 men, median age 40 years, range 18–65 years). Informed consent was obtained from the patient. PUH abides by the Helsinki declaration on ethical principles for medical research involving human subjects. Peripheral blood mononuclear cells (PBMC) were prepared from each blood sample by centrifugation over a Ficoll-Paque gradient (Pharmacia, Sweden). The cells were cultured in vitro for 8 h with the addition of 20 ng/ml phorbol myristate acetate (PMA) and 1 μM ionomycin (Sigma, St-Louis, MI). Total cellular RNA extracted from 10 6 cells was submitted to semiquantitative RT-PCR for the amplification of IL-2, IFN-γ, IL-4, IL-5, and IL-10 metagraphs [ 8 ]. Primers and conditions for the RT-PCR are summarized in Table 1 . The PCR products were run on ethidium-stained agarose gels, photographed and quantified [ 8 ]. Table 1 Primers and conditions for the RT-PCR experiments performed in this study. Gene Sequence (5'→3') T (°C) Product (bp) IL-2 GCAACTCCTGTCTTGCATTG AATGTGAGCATCCTGGTGAG 59 173 IFN-γ AGCTCTGCATCGTTTTGGGTTC CAAATATTGCAGGCAGGACAACC 64 492 IL-4 CTGTGCTCCGGCAGTTCTAC ACGTACTCTGGTTGGCTTCC 58 176 IL-5 GCTTCTGCATTTGAGTTTGCTAGCT TGGCCGTCAATGTATTTCTTTATTAAG 59 291 IL-10 ACCCAGTCTGAGAACAGCTGC GTTCACATGCGCCTTGATGTCT 61 260 β2m CCCCCACTGAAAAAGATGAG TCACTCAATCCAAATGCGGC 56 150 A sharp decrease in the expression of the type-1 cytokines IL-2 and IFN-γ and their calculated sum expressing Th1 activity was observed at 7 months after splenectomy compared to 3 months after splenectomy (Figure 1 ); this was accompanied by a parallel rise of platelet count from 190 × 10 6 /L to 265 × 10 6 /L. Regarding type-2 cytokine gene expression, IL-4 increased, IL-5 decreased, and IL-10 remained unchanged, whereas the change in Th2 activity (IL-4 units plus IL-5 units) was slight (Figure 1 ). The Th1/Th2 ratio {(IL-2+IFNγ)/(IL-4+IL-5)}, that reflects immune deviation, was accordingly greatly reduced 7 months post-splenectomy (Th1/Th2 = 1.3) compared to 3 months (Th1/Th2 = 3.5) (Figure 2 ). Mean Th1/Th2 ratio of the controls was 0.5 with 95% confidence intervals of the mean (0.15–0.85). The Th1/Th2 values at 3 months and at 7 months post-splenectomy lie at 6.25 and 1.6 standard deviations above the mean of the controls, respectively. Figure 1 Gene expression levels of individual cytokines and of calculated Th1 and Th2 activities at 3 and 7 months after splenectomy. Th1 equals with IL-2 units plus IFN-γ units and Th2 equals with IL-4 units plus IL-5 units. Figure 2 Th1 activity (IL-2+IFNγ) versus Th2 activity (IL-4+IL-5) scatter gram. Asterisks denote the patient's case values at 3 and at 7 months after splenectomy. Diamonds denote the control values. Solid line is the regression line for the controls (r 2 = 0.55, p = 0.014) and the two broken lines show the 95% confidence intervals. The above results show that in this patient type-1 polarization persists after removal of the spleen and attainment of clinical remission. This may mean that the spleen is not exclusively responsible for the coordination or the maintenance of the pathological immune response in this patient, provided that no accessory splenic tissue exists. Other disease centres may control the autoimmune reaction as well, such as the liver or the bone marrow. Alternatively, it is possible that ITP is the manifestation of a general immune system malfunction that pre-existed before the development of thrombocytopenia and persists after removal of what seems to be the effector of a manifestation of an autoimmune proclivity. Unfortunately, a pre-splenectomy sample was not available for analysis. As a result, no conclusions can be drawn about the effect of splenectomy on the direction of change of Th1 activity. Based on phenotypic studies showing increased presence of T lymphocytes with activated phenotype after splenectomy in ITP patients [ 15 ], it is plausible that peripheral Th1 activity may have increased after splenectomy. The clinical remission may be due to the removal of a major platelet destruction site, although the underlying immune activity that drives the destruction may remain unaffected. Complete remission does not mean that increased platelet destruction has stopped after splenectomy. Platelet life span may still be shortened in this patient and/or her normal platelet count may be even higher than what was achieved after splenectomy. The pathological immune activity seems to decrease over time after splenectomy, as reflected by the lower Th1/Th2 ratio that is indicative of the degree of immune polarization. This may be explained by reduced stimulation of the immune system by activated spleen reticuloendothelial cells that present platelet antigens to T helper lymphocytes. In this way, it may be hypothesized that removal of one vital component of the self-attacking immune process can break the vicious circle that culminates in even greater immune activation, polarization, and platelet destruction. Removal of a major site of autoimmune activity may have abrogated recruitment of naïve T-cells. As a result, overall autoimmune activity wears off, as existing activated Th1 effector cells perish leaving behind a much smaller population of peripheral memory cells that retain the initial Th phenotype. Another consideration that stems from the results of this case study is that immune polarization and immune deviation of the pathological response depend more on upregulation of type-1 mediators rather than on suppression of type-2 cytokines, or that type-2 response is inadequate to control excess type-1 response in active disease. Conclusions Clinical improvement after splenectomy is associated with reduced but not normalized immune activation and polarization in the patient studied. However, the spleen seems not to be absolutely necessary for the maintenance of the autoimmune reactivity. Competing interests The authors declare that they have no competing interests. Authors' contributions FPP prepared the case report, performed the experiments and drafted the manuscript. AM conceived of the study, participated in its design and coordination and co-wrote the manuscript. Both authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524488.xml
533876	Random allocation software for parallel group randomized trials	Background Typically, randomization software should allow users to exert control over the different aspects of randomization including block design, provision of unique identifiers and control over the format and type of program output. While some of these characteristics have been addressed by available software, none of them have all of these capabilities integrated into one package. The main objective of the Random Allocation Software project was to enhance the user's control over different aspects of randomization in parallel group trials, including output type and format, structure and ordering of generated unique identifiers and enabling users to specify group names for more than two groups. Results The program has different settings for: simple and blocked randomizations; length, format and ordering of generated unique identifiers; type and format of program output; and saving sessions for future use. A formatted random list generated by this program can be used directly (without further formatting) by the coordinator of the research team to prepare and encode different drugs or instruments necessary for the parallel group trial. Conclusions Random Allocation Software enables users to control different attributes of the random allocation sequence and produce qualified lists for parallel group trials.	Background An important aspect of any trial that should be clearly stated in the final report is the method used to assign treatments (or other interventions) to participants [ 1 ]. In the final report of the trial, authors should specify the method of sequence generation, i.e. whether they have used mechanical means, a computer generated random list or random number table. After preparing a random sequence, subjects will be allocated to the trial groups using an implementation method such as numbered containers, central telephone line, or allocation by a person who is not involved in the main research and patient care (the encoder). During the process of allocation each subject will be given a unique identification code (Unique Identifier, UI). This UI will be used as a label to uniquely identify the patient's group after completion of the study. During the study period this UI will be given to the main researchers together with the necessary treatment (e.g. drug or placebo). Usually these treatments are prepared by the encoder with the same physical characteristics (shape, color, size, etc), differing only by the UI labels to blind the main researchers about the actual patient group. Following enrollment of all subjects into the study, these UIs are decoded to determine the patient group. Depending on the preference of the researchers or facilities of the research environment, subjects are randomly allocated to intervention groups using either a random list prepared before the study (In Advance method) or a randomized allocation at the moment of intended intervention (Just In Time method; JIT). Both the JIT and In Advance methods produce acceptable allocations, and the actual choice depends on the availability of certain facilities for each method. Usually the randomization components of these two methods are produced by running a randomization software on either an Internet service provider or a local computer. The local encoder will obtain the next allocation (JIT method) or the entire random list (In Advance method) from the service provider or from the software on local computer and prepare the necessary blinded equipments. Without the use of computer software or Internet services the maintenance of the whole process of randomization and allocation is difficult. In addition, in the case of any necessary restrictions on the process of randomization (i.e. block randomization [ 2 ]) the complexity of the process will be increased even further and be prone to errors. Randomization software may run on a local computer or may be hosted by an Internet server. A complete list of these software and services can be found on Martin Bland's web site [ 3 ]. Most of randomization software are hosted by websites for both JIT and In Advance methods, which require access to the Internet [ 4 - 8 ]. However, most of these Internet services have restricted capabilities with respect to the block design specification, control over the output format and flexibility of UIs. Among these available services, the tool in Randomization.com [ 4 ] seems to be more advanced than the others. It allows users to specify the number of subjects per block, the number of blocks and up to 20 treatment labels. Therefore, this service produces simple and block randomization using fixed and equal block sizes. Unfortunately, this service does not allow further restriction on block design (e.g., multiple block lengths or random variation in block number or size). The generated random list is in the form of UI and group name pairs, formatted in a single column, which in cases of large sample sizes may require further work to fit it in multiple columns for fine printing. Moreover, the block borders are not visible to allow for easy visual inspection of block sizes and equality of cases. Although a minor problem, Randomization.com only produces sequential numeric UIs with variable lengths (e.g. 1, 10, 100). The variability in lengths of UIs may disturb the visual impact of the generated list compared to the fixed length UIs (e.g. 001, 010, 100). Some researchers prefer to use random UIs in mixed alphanumeric format to decrease the likelihood of memorization and to improve the blindness of the study and concealment of allocations. Available randomization software have more restrictions in their capabilities than the Internet randomization services. They have limitations in their output format [ 9 ] and users can not specify the number and naming of the treatment groups [ 10 , 11 ]. In addition, these software are not designed with the capability to produce flexible UIs for participants. Therefore, the main objective of Random Allocation Software was to construct a randomization software for parallel group trials with the following characteristics: 1. Independent running on a local computer without any need to access the Internet 2. Different types of program output: to file (html or text files), window and system clipboard 3. Provisions for different block design 4. Capability to deal with a larger number of groups 5. Specifying a name for each group 6. Control over the format, length and ordering of the generated UIs 7. Control over the format of generated sequence 8. Saving or loading the randomization settings 9. Viewing previously generated randomized sequences Implementation Random Allocation Software is a program created in Microsoft Visual Basic 6, and it installs in the same way as ordinary Windows software (i.e. running setup.exe and following on screen instructions). Once installed and run, there are some controls in the main window for specifying the number of groups (2 to 16), sample size and the name of each group. It also contains menu items to determine the program output and randomization settings. The default program output is saved into either html or text files, and it may also have output to a window or to the system clipboard. A variety of randomization options can be set in the options window. The length of generated UIs (named as Code in the program) can be between 3 to 10 characters and there are options for different alphanumeric structures. In addition, these UIs can appear in sequential or random order in the generated random list. The program can generate simple or block randomization in different types, including equal size blocks, multiple block lengths with random variation among the specified block sizes and complete randomized blocks (random number and size of blocks). The generated sequence will appear in a multiple columns format and the number of columns (1 to 10) can be changed in the options window. Output to html file will be formatted in the form of one block per table. Borders of the tables may be shown or hidden. By clicking the 'Generate' button in the main window the random sequence will be generated and opened by the default viewer for the output file (e.g. Internet Explorer for html files). Previously created output files can also be viewed from inside the program. Additional options include saving the current randomization settings, loading a previous setting and enabling the program to save the last setting upon program exit. During execution, the program produces a random sequence of allocation using the Rnd function that generates a floating point random number. The Rnd function uses the linear-congruent method for pseudo-random number generation as depicted by the following formula: x1 = (x0 * a + c) MOD (2^24) [12] where: x1 = new value x0 = previous value (an initial value of 327680 is used by Visual Basic unless the Randomize X function is used to specify a different seed as X) a = 1140671485 c = 12820163 The seed of the Rnd will be the Timer function, which will return the number of seconds elapsed since midnight. Although this version of the program does not produce repeatable lists, it is possible to revise the program in subsequent versions to save the value of the seed to reproduce the same random list. The output consisted of shuffled allocations each of which is a UI, group name pair. The program checks for the uniqueness of the UIs and generates an error message if the specified UI length is insufficient to hold the entire sample size. Runs test was used to check randomness of the output list with sample sizes from 10 to 190 (10, 30, 50, ..., 190) and from 200 to 3000 (200, 600, 1000, ..., 3000). Each runs test was carried out for the group number of 2, 3, 5 and 6. SPSS 10 software was used to perform the runs test. Results The program starts running with the default settings. Users may run the program with the default settings or set the number of groups, the name of each group and the sample size. Clicking the 'Generate' button (figure 1 ) produces the random sequence. Before generating the random sequence, the option window will be displayed and different randomization settings can be entered (figure 2 and 3 ). Consider, for example, that we want to produce a simple randomized list for a sample size of 30 subjects into three groups of Case, Control and Placebo with numeric sequential UIs. After setting different options and clicking the 'Generate' button, the generated list will appear in columns (Table 1 ). Each entry in the list consists of a UI, and a group name pair. Alternatively numeric UIs may appear in random order (Table 2 ). Table 3 shows the output of the program for a block randomization with blocks of equal sizes. Figure 1 Main window. The main window showing different options for number of groups, sample size, and group names. Figure 2 Options window: Blocks. Options window, settings for block design. Figure 3 Options window: Code. Options window, setting the format of unique identifier (UI) specified in the program as Code. Table 1 A simple randomized list produced by the software for a sample size of 30 subjects into three groups of Case, Control and Placebo with numeric sequential unique identifiers 001: Case 009: Placebo 017: Control 025: Placebo 002: Control 010: Control 018: Control 026: Case 003: Case 011: Placebo 019: Case 027: Case 004: Case 012: Control 020: Control 028: Placebo 005: Control 013: Case 021: Placebo 029: Placebo 006: Placebo 014: Case 022: Case 030: Control 007: Placebo 015: Placebo 023: Case 008: Control 016: Control 024: Placebo Table 2 The same setting as in table 1, but with the numeric UIs in random order 288: Case 200: Control 462: Placebo 775: Case 644: Control 437: Case 448: Case 622: Control 278: Placebo 364: Control 523: Control 327: Control 427: Case 525: Control 837: Case 514: Placebo 146: Placebo 796: Case 804: Placebo 610: Case 383: Placebo 208: Control 581: Control 167: Placebo 493: Placebo 862: Placebo 181: Control 484: Case 079: Case 254: Placebo Table 3 A block randomization list with four blocks of equal sizes 504: Placebo 671: Placebo 767: Case 442: Control 094: Placebo 256: Case 636: Case 200: Control 677: Control 669: Placebo 355: Control 334: Case 765: Control 073: Control 377: Case 537: Placebo 527: Placebo 485: Case 183: Placebo 658: Control 612: Case 875: Control 888: Case 733: Case 127: Placebo 864: Placebo 476: Control 552: Control 138: Case 548: Case 938: Placebo 592: Placebo 810: Control 213: Control 584: Placebo 438: Case Figure 4 is the printed output of the program for a block randomization with random block sizes. Figure 4 Sample output. Sample output for a block randomization with random block sizes. In block randomization the final sample size is usually larger than the specified one. A total of 18 runs test were performed to check the randomness of the program output, which resulted in P values of 0.22 to 0.81. Discussion The main use of Random Allocation Software is to produce simple or block randomized sequences for parallel group trials. Its use is restricted to parallel group randomized trials. Compared with similar software, it enables the user to control the length, order and format of the UIs; and the type and format of the output. It allows specifying up to 16 groups for parallel trials. Conclusions Random Allocation Software has been designed to produce random sequences consisting of UI, group name pairs with additional control over the output format and type. Available randomization software generally has limitations in the number of groups, naming each group, generating UIs and control over the output. Many of these problems have been addressed in the present software. As has been stated in previous sections, the main use of this software is for randomization in parallel group trials. The software can be revised to support crossover and other types of randomized trials. The experienced user may test the randomness of the program output by selecting numeric labels for group names and then exporting the generated list into a statistical software such as SPSS to execute a runs test on the exported data. Availability and requirements Project name: Random Allocation Software Public use access: (Latest version) Operating systems: Windows 98, Me, 2000, XP. It should be noted that on some Windows operating systems (especially Windows 2000) during installation of the program an error message like "Setup Cannot Continue... System Files Are Out of Date" may be displayed. If this happens, click OK and restart the system. Then run the setup.exe again. This is due to a known bug in the installation programs of Microsoft Visual Basic [ 13 ]. This problem has been removed from the newer versions of the program. Users are recommended to download the latest version from the first address. Programming Language: Visual Basic 6 Other requirements: Internet Browser (Internet Explorer 5 or higher is recommended) License: Free for academic use. Abbreviations JIT = Just in Time UI = Unique Identifier Competing interests The author declares that he has no competing interests. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533876.xml
551598	Prolonged conservative treatment or 'early' surgery in sciatica caused by a lumbar disc herniation: rationale and design of a randomized trial [ISRCT 26872154]	Background The design of a randomized multicenter trial is presented on the effectiveness of a prolonged conservative treatment strategy compared with surgery in patients with persisting intense sciatica (lumbosacral radicular syndrome). Methods/design Patients presenting themselves to their general practitioner with disabling sciatica lasting less than twelve weeks are referred to the neurology outpatient department of one of the participating hospitals. After confirmation of the diagnosis and surgical indication MRI scanning is performed. If a distinct disc herniation is discerned which in addition covers the clinically expected site the patient is eligible for randomization. Depending on the outcome of the randomization scheme the patient will either be submitted to prolonged conservative care or surgery. Surgery will be carried out according to the guidelines and between six and twelve weeks after onset of complaints. The experimental therapy consists of a prolonged conservative treatment under supervision of the general practitioner, which may be followed by surgical intervention in case of persisting or progressive disability. The main primary outcome measure is the disease specific disability of daily functioning. Other primary outcome measures are perceived recovery and intensity of legpain. Secondary outcome measures encompass severity of complaints, quality of life, medical consumption, absenteeism, costs and preference. The main research question will be answered at 12 months after randomization. The total follow-up period covers two years. Discussion Evidence is lacking concerning the optimal treatment of lumbar disc induced sciatica. This pragmatic randomized trial, focusses on the 'timing' of intervention, and will contribute to the decision of the general practictioner and neurologist, regarding referral of patients for surgery.	Background One of the greatest advantages of publishing the design of a randomized controlled trial (RCT) before results are available is the accessibility to criticism of the methodological quality irrespective of the results. Firstly the scientific reader must be enabled to search for epidemiological shortcomings when the results differ from the expected outcome as compared to results in line with one's expectations. Secondly, it is possible to more extensively elaborate the background and rationale of the research question, the study population, the chosen treatments and outcome measures, as compared to publications describing the trial results. Thirdly, but not less important, publishing the design of a RCT is instrumented in preventing publication bias in subsequent meta-analyses. Studies with non-significant results are less likely to be published than those with significant results [ 1 , 2 ]. It is a considerable loss for data pooling that unpublished trial results are omitted. After pre-publishing the study design even unpublished data can be used in a systematic review, since these can be required from the study group. This article describes the rationale and parallel group design of a RCT in which the optimal timing of disc surgery for sciatica will be investigated. The lumbosacral radicular syndrome (LSRS or LRS; also called sciatica) is typically characterized by radiating pain in the dermatome of a lumbar or sacral spinal nerve root. Occasionally more than one root is involved. Contained in the syndrome pain may be accompanied with lumbar fixation, reflex abnormalities motor and sensory disturbances. In diagnosis includes stenosis of the spinal and/or root canal, infection, multiple sclerosis, autoimmune or metabolic neuropathy, and tumour. This study will be restricted to herniations at the lowest three lumbar disc levels, since these represent the most common sites. In the vast majority of cases LSRS is the result of a herniated disc. In the Netherlands annually between 60,000 and 75,000 new cases of LSRS are diagnosed by the General Practitioner (GP) [ 3 ]. The presumed direct medical costs of treatment of LSRS are € 133 million each year [ 4 ]. Most of these costs are attributable to in-hospital treatment; only a small portion is incurred by GP's or physiotherapists (€ 3.2 million). In a study, performed in 1988, more than 11.000 patients were operated in the Netherlands and this frequency did not change in the past years [ 4 , 5 ]. The combined direct and indirect costs are estimated to be € 1,2 billion per year [ 6 ]. The indirect costs are considerable due to the high rate of production loss caused by sciatica. The natural history of LSRS is in general favourable. In 60–80 percent of patients, the leg pain decreased or disappeared within 6–12 weeks after onset [ 7 - 9 , 51 ]. These patients no longer experienced problems at work or in their private lives after three months. The minority with lasting complaints beyond three months further decreases with time. At one year only a small proportion of herniated discs continues to produce discomfort and disability. At present it is not possible to identify these latter groups of patients in an early stage of their disease by means of intensity of pain, neurological deficit, root irritation signs, or diagnostic imaging. For this reason it is not helpful to perform early diagnostic imaging (CT or MRI), unless a disease entity different from disc herniation is considered. After the indication for surgery has been set diagnostic imaging is helpful in defining the exact site of disc herniation and its anatomical relationship with the nerve root involved. Since the first publication on lumbar disc surgery by Mixter and Barr [ 17 ] many studies have demonstrated the success of surgery for the treatment of LSRS. Unfortunately only a few prospective studies investigated the difference in outcome between surgical and conservative care [ 7 , 8 , 18 - 22 ]. The published treatment results vary as much as the frequency of reported complications and the recurrence rate. The only study, which compared surgery with conservative care directly in a RCT, was performed by Weber more than 20 years ago [ 7 , 8 ]. He found better results for surgery at one-year follow-up. At four and ten years follow-up the results of surgical and conservative care no longer differed. Being the only published RCT comparing surgical and conservative care, this study regrettably carries some important methodological flaws in both design and outcome measures when compared to today's epidemiological standard rules [ 23 ]. One of the main shortcomings is the exclusion of patients, who do have an indication for surgery because of "intolerable" pain. Those are the current patients who ask for surgery and are not comparable to the randomized population of Weber. Therefore it is impossible to extrapolate and generalize these results to the treatment policy of today. Since 1983 a few cohort studies have been published on non-surgical treatment of patients with at least six weeks of leg pain with good short-term results at one-year follow-up [ 25 , 22 ]. These studies also suffer from methodological flaws. The only conclusion that can be drawn from these reports and the study of Weber is that the policy of prolonged conservative care can be effective, as a result of the favourable natural course of LSRS. Epidemiological and clinical studies have shown that most lumbar disc protrusions resolve spontaneously with the elapse of time [ 15 , 16 ]. Another finding is that prolonged conservative care appears safe and without complications if the patient remains active. Recent population based studies however state that the natural history is not favourable at all [ 50 ]. Whether particular demographic findings, symptoms, physical signs and/or MRI findings either separately or combined do have prognostic value has not been investigated scientifically yet. It would be of great value if one were able to identify early in the course of the disease those patients who will have an unfavourable outcome without surgery. In spite of the known favourable natural course the surgical rate in the Netherlands is quite high [ 10 ]. We perform six times as many lumbar discectomies compared to Scotland, four times the number in England and two times the number in Sweden. In the latter study comparing 12 Western countries the United States is the only country where more operations are performed for the indication LSRS. There are no substantial differences in the incidence of this disease in the countries mentioned that can explain the difference in surgical rates. There is no indication [ 6 ] that the surgical rate has changed under influence of the consensus reports [ 11 , 12 , 11 ]. Actually change was not likely to occur because the published guidelines were representative for daily practice and normal care before 1996 in the Netherlands. With respect to the indications for and timing of surgery no evidence in the literature is available to either support or contradict these guidelines. These guidelines were produced after agreement between all medical (sub-) disciplines involved in the care for patients with LSRS. Our high surgical rate, as contradictory as it may seem, may reflect good clinical practice. Because of the observation that most people recover from their complaints in the first 6–8 weeks [ 9 , 51 ] this period of persistent radicular leg pain is considered a good indication for surgery in the Netherlands. Although there is consensus that surgery is only offered in case of persistent pain, the timing of this treatment seems to depend on local production capacity and patient and doctor preferences rather than on evidence-based practice. This lack of evidence for the timing of surgery after the 6–8 week period explains the large variations in daily practice. Exact data on the problems associated with surgery, such as surgical failure, recurrent disc herniation and adverse effects are limited. This is one of the reasons that in some regions surgery will only be carried out after a period of 3–6 months of LSRS. [ 14 ]. It is not known whether the relative high rate of disc surgery in the Netherlands is cost-effective or not, compared to other countries [ 15 , 16 ]. In summary, consensus is missing on the preferred timing of disc surgery, due to insufficient evidence that a prolonged conservative care strategy is effective. More insight is needed into the potential short-term effects of a relative early surgery strategy, as compared to an extended wait-and-see period. In particular the effects on the return to work or resumption of previous daily activities as well as the complications of both strategies have not yet been investigated. The main goal of this comparative study is to investigate whether the completion of a 6–12 weeks period of lasting radicular pain constitutes a solid indication for surgery and is superior to prolonged conservative care. A secondary goal is to identify possible subgroups of patients who will substantially benefit from one of the proposed treatment strategies. The cost-effectiveness results will be a trade-off between a quicker relief of leg pain in the surgery group versus the advantage of lower costs and avoiding the negative effects of surgery in the conservatively treated group. The difference in disease related quality of life depends on the duration of persisting pain and disability after randomization in the prolonged conservative care group. This study to investigate this scientific gap in our understanding of the effectiveness of surgery for LSRS is in line with a recommendation by the Dutch Health Council in 1999 to the Minister of Health [ 4 ] and the current Cochrane Review [ 15 , 16 ]. The results of this trial will lead to a more rational use of the existing guidelines if the hypothesis is rejected. If the latter is accepted and prolongation of the conservative treatment policy is more cost-effective than surgery after 6–12 weeks, the current guidelines for the timing of surgery need correction. Methods/design To answer the main research question the investigators propose to conduct a multi-centre comparative randomized clinical trial with parallel group design. The main research question will be answered after a follow-up of six months (Figure 1 ). The complete follow-up will last two years. The multi-centre design is necessary to collect enough patients in two years. The Medical Ethics Committee of all participating hospitals approved the study protocol. Figure 1 Flow chart of the Sciatica Trial Patients All patients between 18 and 65 years with sciatica of less than 12 weeks duration are eligible for this study. Because of the multi-centre (15 hospitals) design the patients in a large region in the western part of the Netherlands can be included in this trial if they meet the in- and exclusion criteria (Table 1 ). Because these are the only hospitals, which treat lumbar disc herniations in this area, included patients will reflect a representative population treated in primary and secondary care. Inclusion of patients will be started after a visit to the neurological outpatient clinics. Randomization will start after at least 6 weeks persistent disabling pain in the dermatome of the leg served by the L4, L5 or S1 root. All 1100 GP's involved will be informed about this study and receive information about developments and the results of the trial. They will refer patients within the first 6–12 weeks after onset sciatica. Table 1 Selection criteria for trial eligibility Inclusion criteria: • Age 18–65 yr. • Persistent radicular pain in the L4, L5 or S1 dermatome with or without mild neurological deficit • Severe disabling leg pain of 6–12 weeks duration • Evidence of a unilateral disc herniation confirmed on MRI • Sufficient knowledge of Dutch language • Informed consent Exclusion criteria: • Cauda equina syndrome or severe paresis (MRC<3) • Complaints of a lumbosacral radicular syndrome in the same dermatome within the past 12 months • A history of unilateral disc surgery on the same level • Spinal canal stenosis • Degenerative or lytic spondylolisthesis • Pregnancy • "Severe life-threatening" or psychiatric illness • Planned (e)migration to another country in the year after randomization During the first visit to the neurological outpatient clinic the patient's history will be taken and a standardized neurological examination will be performed. During this visit the neurologist will inform the patient on the cause and course of a lumbosacral radicular syndrome and convey the doubt regarding the timing of surgery for this condition. The study will be explained to the patient and in case of a positive reaction an appointment is made to meet one of the research nurses as soon as possible. Preferably the study MRI scans will be performed after informed consent during the first visit to the research nurse. Because the patient needs some time to consider participation a second visit will be planned at least two days after the first visit to the outpatient clinic. The research nurse will give all extra information needed to understand the trial and will ask the patient if he/she agrees to be randomized. Informed by the radiologist and surgeon, the research nurse will only randomize the patient during the third visit if the MRI confirms the presence of unilateral disc herniation and the patient is eligible according to the inclusion and exclusion criteria. The patient will not be aware of detailed MRI data. The radiologist and neurosurgeon independently using a standardized Case Record Form (CRF) will register the MRI findings. The MRI will be performed according to a standardized protocol and including Gadolinium series for the intended subgroup analysis. Treatment allocation Patients will randomly be allocated to either surgery within 1–2 weeks or prolonged conservative treatment by their GP. Patients, their doctors and research nurses can obviously not be blinded for the allocated treatment. Blinding of the outcome measurements is not possible, due to the fact that mainly self-reported outcomes are used. A randomization list is prepared for every participating hospital. Permuted blocks of random number patients are formed to ensure near-equal distribution of patients over the two randomization arms in the hospitals. Using random number tables generates the random sequence of the permuted blocks. The data manager, who is not involved in the selection and allocation of patients will prepare coded, sealed envelopes containing the treatment allocation. During the second patient visit the research nurse will open the envelope together with the patient and appointments will be made for the allocated treatment, either surgery or referral back to the GP, to ensure that treatment is started as soon as possible after randomization. This will be done after checking all the criteria and especially the persistence of pain with disability in daily functioning. A letter about the allocated treatment arm informs all caregivers. Although the principal investigator will not include and operate upon trial patients he may be biased with a preference for surgery, which could theoretically influence analysis. Therefore the principal investigator is blinded for the allocated treatment. As he is not involved in treatment of the study population blinding during later analysis is only possible after blinding during the randomization and follow-up period. Interventions After randomization two groups of patients will exist. Group A ; the surgically treated patients and group B; the conservatively managed patients. Surgical treatment (A) will be performed in the conventional manner with microscope or loupe magnification. The investigators prefer the standard surgical approach because the other (minimally invasive) surgical approaches have limited indications, are not more cost-effective, and have a long learning curve. During the transflaval approach care is undertaken to minimize bony removal and on the other hand to prevent overstretching of the compromised nerve root. In addition to removal of herniated disc material as much as possible nuclear material will be removed with pituitary forceps, curettes and rongeurs in order to prevent recurrence. The participating treating doctors are 2 orthopaedic- and 12 neurosurgeons with large experience in the standard approach with loupe magnification or microscope. A standardized CRF will register the findings of the surgeon and the herniated disc material will be investigated histologically for granular infiltration. Surgery will take place as soon as possible and within a maximum of two weeks after randomization. Hospital admission will be 2–7 days, including the day of surgery. During the immediate post-operative period the patients will be mobilised with the help of a physiotherapist. At home guidance is confirmed by their own physiotherapist. The frequency will be 2 times a week for 8 weeks. Conservative management (B) will be conducted by the general practitioner (GP) or neurologist when necessary. The GP will provide ample information about the favourable prognosis of LSRS. The treatment of LSRS is aimed primarily at pain relief and maintenance/restoration of normal day-to-day activities. Unfortunately, the effect of giving information and counselling has not been studied specifically among LSRS patients. However, various studies have evaluated the effect of such support for people suffering from other pain syndromes [ 24 ]. Inferences can reasonably be made from the findings of these studies. Hence, it may be assumed that adequate and unambiguous information about what is wrong (the nature of the condition) and what the patient can expect (the prognosis), together with trustworthy counselling can reduce the anxiety and uncertainty felt by the patients and thus ease the pain [ 12 ]. The GP's will encourage the patients to continue with normal day-to-day activities in so far as possible. When necessary analgesic medication can be prescribed according to the guidelines. The GP will advise the patients to stay active and if possible return to work and/or their leisure activities. After the first consultation the GP will make a follow-up schedule. During the next visit the patient and doctor will look at the changes since the first visit to determine whether there is any improvement in the ability to perform normal activities. The doctor will check the efficacy of the prescribed pain medication and may adjust the dose or sort of analgesics according to the NHG guidelines. In these guidelines paracetamol is the first choice. If not effective, NSAID's (ibuprofen, diclofenac or naproxen) are to be prescribed. Only in the event of severe disabling pain morphine may be given for a restricted period of time. By preference all analgesics should be taken at fixed times of the day rather than on a 'if necessary' basis. If the GP and the patient conclude that there is considerable kinesiophobia because of the fear that the radicular or low back pain will increase, the help of a physiotherapist can be recommended. Guided by the GP (and physiotherapist) the patient will upgrade his or her activities according to the agreed time schedule [ 25 , 26 ]. The guide will be time, not the intensity of the pain. The GP will be free in her/his choice of prescription of medication and referral to physiotherapists. The research nurse will register the conservative management strategy after communication with the responsible GP. In case of progressive neurological deficit or worsening intolerable pain the GP can refer the patient back to the research nurse or neurosurgeon. If, six months after randomization, the patient has still not improved or suffers from intermittent LSRS, surgical treatment will be offered. Some patients will ask for surgery earlier because of worsening drug resistant leg pain. In these cases and in the case of a progressive neurological deficit, surgery will be performed in consultation with the patient. If after maximum conservative treatment and counselling the patient is still not able to cope with the functional disability surgery can be requested. If surgery in these cases is not offered by the study-group the patient does have the right to have a second opinion with an undependable neurosurgeon of another university hospital. Outcome assessment In the LSRS the most common complaints are pain and disability to perform normal daily activities. We will use below described validated outcome parameters, which will be assessed by means of questionnaires. Patients are not informed about their earlier scores. Follow-up examinations by the research nurse will take place 8, 26 and 52 weeks after randomization and the patients will keep a diary (table 2 ). In between at 2, 4, 12, 38, and 78 and after 104 weeks the main questionnaire (primary outcome measures) will be filled in at home and send to the data centre. Table 2 Data collection and outcome measures Time in weeks ? 0 2,4 8 12 26 38 52 78 104 Likert X X X X X X X X X X Neurological examination X X X X Severity of complaints (VAS) X X X X X X X X X X McGill X Health Status (SF 36) X X X X X Functional Status (RDQ) X X X X X X X X X X EuroQol/VAS Q-of-life X X X X X X X X X X MRI X X Costs X X X X X X X X X X Prolo X X X X Complications X X X X Surgery X X X X X SFBI X X X X X X Primary outcome measures 1) Roland Disability Questionnaire for Sciatica.' This illness-specific 23-item functional assessment questionnaire is frequently used for low back pain and sciatica [ 38 , 39 ]. Scores range from 0 to 23, reflecting a simple unweighted sums of items endorsed by the respondent. Patients with high scores at baseline do have a severe disabling LSRS. To define recovery a difference of at least 11 points from baseline has to be seen [ 38 , 17 ]. The Roland Questionnaire for Sciatica has a documented high level of internal consistency; construct validity, and responsiveness [ 38 , 39 ]. It is the main primary outcome measure in this trial. 2) Perceived recovery.' This is a seven-point Likert scale measuring the perceived recovery, varying from 'completely recovered' to 'worse than ever'. This outcome scale has been used in previous studies and appears to be valid and responsive to change [ 27 ]. Next to this global self-assessment a job and hobby specific Likert will be scored. During the intake of the study the patient will be asked to rank their five most important functional disabilities in daily live (work, hobby), which they can use in their own evaluation overall and in separate items. 3) VAS pain in the leg. This parameter will measure the experienced intensity of pain in the leg during the week before visiting the research nurse. Pain will be assessed on a horizontal 100 mm scale varying from 0 mm, 'no pain in the leg', to 100 mm, 'the worst pain ever'. Patients do not see the results of earlier assessments and will score the pain experienced at the visit. [ 28 - 32 ]. Secondary outcome measures 1) EuroQol classification system and VAS rating personal health. A cost-utility analysis will be performed using QALY's based on the EuroQol questionnaire, which has been validated in many studies and is easy to fill out [ 41 , 42 , 51 ]. The EuroQol will be measured twice a week during the first four weeks and at all follow-up moments. Patients describe their general health status using the EuroQol classification system, consisting of 5 questions on mobility, self care, usual activities, pain/discomfort, and anxiety/depression [ 44 ]. From the EQ-5D classification system, the EQ-5D utility index will be calculated [ 43 ]. This utility measure reflects how the general public values the health status described by the patient, which is preferred for economic evaluations from a societal perspective. Patients also rated their personal health using a visual analog scale (VAS) ranging from worst imaginable health to best imaginable health. 2) Short-Form 36 (SF-36). Quality of life was also assessed using the RAND-36 questionnaire. This is a generic health status questionnaire, which can easily be filled out at home. The questionnaire consists of 36 items on physical and social functioning has 8 domains; 1) physical functioning, 2) physical restrictions, 3) emotional restrictions, 4) social functioning, 5) somatic pain, 6) general mental health, 7) vitality, 8) general health perception. This questionnaire has been used frequently and was validated in studies on low back pathology and surgery [ 33 - 37 ]. From the RAND-36, the SF-6D utility index was calculated. Like the EQ-5D, this SF-6D reflects the general public's valuation of the health described by the patient. The SF-6D is a recent instrument that has not been used much yet, but it richer classification system could make it a more sensitive utility measure than the EuroQol measure. 3) Sciatica Frequency and Bothersome Index (SFBI). This is a scale from 0 to 6, which can assess the frequency (0 = not at all to 6 = always) and bothersomeness (0 = not bothersome to 6 = extreme bothersome) of back and leg symptoms. The sum of the results of four symptom questions yields both indexes, ranging from 0 to 24: leg pain; numbness and/or tingling in the leg; weakness in the leg or foot; pain in the back or leg while sitting. [ 17 ]. 5) PROLO-scale. This scale measures the evaluation of the research nurse of the functional-economic status of the patients. This parameter has been used in studies on the difference in functional outcome between different techniques of lumbar spine fusion [ 40 ]. 6) VAS pain in the back. This parameter measures the intensity of the pain in the back experienced during the week before visiting the research nurse. Assessment will be based on a horizontal 100 mm scale varying from 0 mm, 'no pain in the back', to 100 mm, 'the worst pain ever'. Patients do not see the results of earlier assessments and will score their pain during the visit. This parameter is included because a lot of patients with LSRS also have back pain in varying intensities, which can change after surgery or conservative treatment. Other outcome measures 1) Costs. The societal costs during the first year will be estimated in accordance with the recent pharmacoeconomic guideline [ 47 , 48 ]. The costs of hospital admission and surgery will be based on an integral top-down cost analysis in three large regional participating hospitals (aggregated according to the total number of patients per department). From this institutional analysis, the constant costs per admission and the variable costs per admission day will be estimated. From these constant and variable costs, the individual costs of hospital admission and surgery for all patients can be estimated, using the duration of the hospitalization. In the study an MRI is performed in all cases. The costs of this MRI will only be calculated for patients undergoing surgery, because in the normal situation MRI would only be performed when a surgical indication exists. Patients will register other health care needs in a diary (including physiotherapy, visits to GP's and specialists, nursing care and medication). Each diary covers a period of 3 months and will be discussed with the patient during the follow-up visits to the research nurse. The volume of health care will be assessed using standard prices [ 48 ]. In the diary the patient will also register direct non-medical costs (including time costs, travel expenses and domestic help). To estimate productivity costs the patients will also report absenteeism in the diary. At the follow-up visits, the research nurse will register the work situation, work efficiency and gross wages. Absenteeism will be valued according to the friction-cost method. 2) Incidence of (re-) surgery. One of the goals of the policy for group B is to avoid surgery while achieving at least the same effects. The surgical rate is therefore an indication of the success or failure of this policy. The incidence of re-operation at the same disc level in group A will be an indication of the failure rate for surgery. 3) Side-effects or complications that are ascribed to the treatment are recorded by the patients, their treating physicians and the research nurses. 4) MRI findings. The results of the differences between the baseline MRI and the MRI made 52 weeks after randomization are important secondary outcome measures. The difference in size of the disc herniation (in mm), nerve root compression, and amount of scar tissue will be registered. Failures of surgery can be recognized by inadequate disc removal or decompression of the nerve. The data will be gathered, using a standardized CRF, which will be filled out by the local radiologist, orthopaedic- or neurosurgeon and (neuro-) radiologist Sample size The result of this study is based on the short-term success of surgical intervention and will be a trade-off between a quicker relief of leg pain versus an advantage in cost-effectiveness for conservatively managed patients. The sample size is calculated on the basis of the Roland Disability Questionnaire for Sciatica averaged during the 12 months follow-up period. The numbers used for this sample-size are drawn from the Maine Lumbar Spine Study 1 year and recently published 5-year results [ 19 , 55 ]. The difference in the Roland score between the surgical- and non-surgical group in this study did not change between 3 and 12 months follow-up as shown in their study [ 19 ] and can be averaged over the first year. The main aim of this study is to measure the short-term functional difference at 12 months follow-up. Surgical treatment is considered better when the post treatment change is at least 4 points more when compared to the conservative treatment arm [ 38 ] and constant over time. Considering this constant difference and a mean standard deviation =10 over the first year [ 55 ] 140 patients per treatment arm are needed to reach a power (1-β) of 0,90 with α = 0.05 (two-sided). To answer the main research question 280 patients are needed for analysis with at least 12 months follow-up. The aim is to enrol 300 (150 per arm) patients in the study, including 8 % loss to follow-up after 1 year. The total number of operated patients each year in all participating hospitals exceeds 1400. With this number of patients also a clinically important difference in median time to recovery of two months can be detected by survival analysis. Although the time to recovery is the main issue, the problem of recurrent complaints is still not solved in the different approaches of survival and proportional hazard analysis. Statistical and cost analysis Baseline comparability will be investigated by descriptive statistics to examine if randomisation was successful. Differences in success rates between both groups are calculated, together with 95 per cent confidence intervals. In addition to an analysis of the difference in recovery between the two groups (as explained under the paragraph sample size) analyses of the difference in time to recovery will be carried out. Due to lack of data in the literature we could not base our sample size calculations on these differences. Survival-analysis is used to calculate differences in median time to recovery. Continuous outcomes are evaluated as change scores (differences between baseline measurement and each follow-up measurement). Multivariable analyses are performed to adjust for the eventual differences between the groups at baseline in prognostic indicators. All the analyses are performed according to the intent-to-treat principle. An additional per protocol analysis is performed comparing patients in the wait-and-see group who received surgery with patients in the same group who had not and with patients in the surgery group. To compare the actual treatment sec instead of strategies an explorative analysis will be performed in subgroups off all patients who actually received surgery and who did not receive surgery in both groups. All patients who withdraw from the study are included in the analysis until the time of withdrawal. The result of this study will be a trade off between the disadvantages of surgery (hospitalisation, reduced quality of life and costs) versus the possible advantages (earlier relief of pain and return to work). For that reason recovery, measured as an 11 point difference in score when compared to baseline (Roland Disability Questionnaire for Sciatica), is the clinically most relevant patient outcome. Quality of Life (SF-36) and perceived recovery are important to compare the reduced quality of life from surgery to the possibly prolonged pain from conservative therapy and also to be able to compare cost-effectiveness with that of other spine interventions. The EuroQol is important to obtain cost-utility ratio's that can be compared with those of a wide range of other interventions. Utilities are obtained from the descriptive classification system of the EuroQol, using the model described by Dolan [ 43 , 53 ]. Conservative treatment may decrease costs compared to surgery but possibly at the expense of delayed effectiveness. In an incremental cost-effectiveness analysis, societal costs during the first year will be compared to the primary outcome measure (Roland Disability Questionnaire for Sciatica, averaged over the first year), Quality of Life (SF-36, during the first year) and perceived recovery (7-points Likert scale). Cost-effectiveness analyses with these effectiveness measures have been conducted before, allowing comparison with other spine interventions. Finally, to answer the second research question explorative analyses are conducted to investigate whether the treatment effect after two, six and twelve months varies in specific subgroups of patients (Table 3 ). Table 3 Selected prognostic variables for subgroup analysis Demographic Variables • Age < 39 years versus > 39 years, • Intellectual versus physical demanding job, Anamnestic and Neurological Variables • Acute start LSRS versus slow start, • History of backpain versus no history, • Influence of coughing, sneezing on complaints versus no influence, • Difficulty to put on shoes and/or socks versus no difficulty, • Straight leg raising ≤ 30 degrees versus > 30 degrees, • Positive crossed straight leg raising sign versus negative sign, • VAS-pain > 70 versus < 69 mm, • Tingling/numbness in pain area versus no tingling (9), • Pain leg worse by sitting versus no worsening (9), • McGill affective high score versus low score, Radiological Variables • MRI disc sequester versus contained disc herniation, • MRI circumferential gadolinium enhancement versus no enhancement of disc herniation, • Mediolateral versus median and lateral disc herniation, • High versus low height of disc level (height 9 mm), Miscellaneous Variables • Preference for surgery versus no preference for surgery. • Disc Herniation at L5S1 vs. L4L5 Using logistic regression for success rate and linear regression for severity of the disability, each prognostic indicator is checked for interaction with treatment. If the interaction term is significant, a stratified analysis will be performed. Discussion In this article the rationale and design of a pragmatic RCT on the cost-effectiveness of timing of disc surgery for LSRS is described. The only randomized trial [ 7 ] so far on this subject only included patients where the caregiver was in doubt about the surgical indication. Patients with severe disabling pain were not randomized [ 8 ]. The Sciatica Trial is directed to those patients with a clear surgical indication according to current usual care. The study is pragmatic because it acknowledges that sometimes it may not be possible to postpone surgery for every conservative care patient until 6 months after allocation and that some patients will recover before surgery is performed in the surgical group. In these cases we consider it unethical to hold on to the randomized treatment. Because of the Intent-to-Treat analysis these cases will be analysed in their own allocated randomization arm and will not cause methodological problems because it is two healthcare strategies that are compared, as opposed to two treatments. The objective of this trial is to provide evidence on the preferred timing of disc surgery for sciatica. A prolonged conservative treatment strategy is compared to the international guideline advise of surgery after 6–8 weeks LSRS. The intended size of the study population is sufficiently large to detect short and long term differences between both strategies. Abbreviations GP = General Practitioner LSRS = Lumbosacral Radicular Syndrome RCT = Randomized Controlled Trial VAS = Visual Analogue Scale Competing Interests The author(s) declare that they have no competing interests. Author's contributions WP designed the study is responsible for the protocol. HH is responsible for the calculation of the sample size and contributed to the design of analysis. WH is responsible for the design of the cost-effectiveness analysis. RB has contributed in the case record forms and is responsible for the database ProMIse. JE contributed to the involvement of the GP's. JT structured the ideas about the diagnostics of LSRS and intake by neurologists. RT is the neurosurgical supervisor of WP. BK is the epidemiological supervisor of WP. All authors participated in the trial design and coordination. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC551598.xml
545076	Development of a questionnaire weighted scoring system to target diagnostic examinations for asthma in adults: a modelling study	Background Identification and treatment of unrecognised asthmatics in the community is important for improving the health of the individual and minimising cost and quality of life burden. It is not practical to offer clinical diagnostic assessment to whole communities, and a simple tool such as a questionnaire is required to identify a smaller target group. Conventional questionnaire screening methods which separate individuals into positive and negative categories have resulted in large numbers of individuals requiring clinical assessment. This study has therefore developed and tested a weighted scoring system that prioritises those most urgently in need, based on their questionnaire responses. Methods A stratified random sample of adult respondents to a general practice postal questionnaire survey were categorised 'asthmatic' or 'non-asthmatic' according to three expert physicians' opinions. Based on this categorisation, logistic regression was used to derive weights reflecting the relative importance of each question in predicting asthma, allowing calculation of weighted scores reflecting likelihood of asthma. Respondents scoring higher than a chosen threshold would be offered diagnostic examination. Results Age and presence of wheeze were most influential (weight 3) and overall weighted scores ranged from -1 to 13. Positive predictive values (PPV) were estimated. For example, setting the threshold score at nine gave an estimated PPV for asthma diagnosis of 93.5%, a threshold score of seven corresponded to PPV 78.8%. PPV estimates were supported by examining 145 individuals from a new survey. Conclusion Weighted scoring of questionnaire responses provides a method for evaluating the priority level of an individual 'at a glance', minimising the resource wastage of examining false positives.	Background There are individuals in the community who are asthmatic but are not receiving treatment because they are unknown to the medical services [ 1 - 5 ]. Detecting these unrecognised asthmatics may be important for short-term health, prevention of long term airway remodelling and minimisation of cost and quality of life burden [ 6 ]. The reliable diagnosis of asthma, however, requires full clinical assessment [ 7 ]. Since it is clearly not practical in terms of resource allocation to offer this to whole communities, a simple tool such as a questionnaire is needed to identify a smaller target group. Conventional targeting approaches which separate individuals into 'positive' and 'negative' categories have failed to reduce the number of potential examinations to a manageable level. For example, in the Wythenshawe Community Asthma Project (WYCAP) [ 8 - 10 ], respondents who had four or more 'yes' answers to six key questions on a postal respiratory questionnaire were deemed possible asthmatics in need of clinical assessment. However, this method identified approximately 350 individuals for examination per practice. A more refined screening method, which could identify more than two categories, would allow the screening threshold to be adjusted to take into account available resources. Defining additional categories according to the possible numbers (0 to 6) of 'yes' answers to the 6 key questions is too simplistic as it assumes equal weights for each question and takes no account of variables such as age and gender. The present paper describes the development of a more refined method, which includes additional questions and uses question weights to calculate a score for each respondent. This weighted score reflects probability of asthma, allowing clinical resources to be targeted to those most likely to have the disease. A score is chosen as a threshold and the questionnaire is given in an opportunistic fashion to patients consulting a medical practice for any reason. The questionnaire score is calculated 'at a glance' and those whose score is higher than the chosen threshold, and who are not already receiving asthma treatment, are offered diagnostic examination. The threshold can be adjusted to take resource limitations into account. Methods The study described in this paper was based on data collected as part of the WYCAP investigation. WYCAP is a long term prospective investigation of the natural history of asthma in two general practice populations. It is based on a postal respiratory questionnaire adapted from the European Community Respiratory Health Questionnaire (ECRHQ) [ 11 ]. Four postal questionnaire surveys of all adults (16 years and over) registered to the two practices were conducted in 1993, 1995, 1999 and 2001. Additional data were obtained in 1995 and 2001 by invitation of selected respondents for clinical review. Ethical approval for the present study was obtained from South Manchester Local Research Ethics Committee and statistical analysis was performed using SPSS for Windows [ 12 ]. The study uses questionnaire responses and additional data obtained as part of the 1995 and 2001 adult surveys and the investigations were conducted in three phases. First, a weighted scoring system was developed based on a subset of adult questionnaires and clinical review data collected in 1995. Next, this scoring system was applied to the questionnaires from the whole population of respondents to the 1995 survey in order to estimate its positive predictive value (PPV) in finding undiagnosed asthmatics in the community. Finally, the validity of the scoring system was tested in a new data set in the form of a subset of questionnaires and clinical review data selected from the 2001 WYCAP survey. Phase 1: Development of the weighted scoring system It was necessary to select an asthma-enriched sample in order to determine response patterns of both asthmatic and non-asthmatic respondents [ 13 ]. Respondents from the 1995 survey were stratified based on the number of 'yes' answers (0, 1–3, 4, 5 and 6) to six key questions on the postal questionnaire. These were questions which were believed to be important in the detection of an asthmatic, namely questions relating to symptoms in the last 12 months (wheezing, chest tightness, shortness of breath, night cough), family history of asthma and associated conditions of hayfever or eczema [see Additional file 1 ]. The target numbers selected from each stratum were based on clinical intuition and aimed to provide a stratified random sample with approximately equal numbers of asthmatics and non-asthmatics. Individuals were selected from one of the practices and invited for clinical review by the research clinician, including full history, physical examination, spirometry, a test for reversibility to beta-2 agonists, bronchial challenge with histamine, electronic peak flow diaries and skin prick testing to five common allergens. The clinical review results were sent separately to three expert respiratory physicians who had no knowledge of the questionnaire responses. Each physician was asked to rate the probability of asthma as <50%, 50–90%, or >90%. Bayesian methods were used to combine the physicians' opinions into a consensus estimate of probability of asthma for each individual [ 14 ]. In order to ensure the independence of the expert opinions required by these methods, the physicians did not confer with each other and were not given any diagnostic criteria or guidelines. Those reviewed individuals in whom the consensus estimate of probability of asthma was 50% or more were designated clinically asthmatic and the remainder were designated clinically non-asthmatic. By modelling the relationship between questionnaire responses and clinical asthma status, a weight was determined for each question, reflecting its importance in predicting asthma status of the respondent. For each questionnaire, the weights for the relevant responses ('yes' answers) were summed to produce an asthma score reflecting the probability of asthma in that individual. The weights were derived based on a linear logistic regression model, designed to take as input the questionnaire responses 1 to 11 and predict the probability of asthma for any individual from the wider community. The model associates each question on the questionnaire with a coefficient reflecting the relative importance of that question in determining probability of asthma. A positive value for a coefficient influences towards a decision that the individual is classified asthmatic, the higher the value the greater the effect. Negative coefficients influence away from a decision that the individual is classified asthmatic. In order to provide a model representative of the wider community, a cross-validation technique was used [ 15 ] in which several candidate logistic regression models were produced and their respective coefficients averaged. These average values were rounded to the nearest whole number to give the question weights prior to summing to produce the weighted scores. Phase 2: estimating positive predictive value The full set of postal responses to the 1995 questionnaire survey was used to assess effectiveness by estimating PPV of the scoring system in detecting undiagnosed asthmatics in the community. Weighted scores were calculated for all 1995 survey respondents and evidence of either 'asthma diagnosis ever' or 'prescription for asthma medication in the last 12 months' was obtained from their medical records. The percentage of undiagnosed asthmatics with a threshold score or higher, who could be expected to be confirmed clinically asthmatic, was estimated using a technique which referenced the probabilities of asthma of sample respondents, (based on the combined expert opinions), their population strata (based on number of 'yes' answers to the six key questions) and weighted scores (based on questionnaire responses) [see Additional file 2 ]. This estimation was repeated for various score thresholds. These expected percentages provided estimates of the positive predictive values of the score thresholds [ 16 ]. Phase 3: testing the validity of the weighted scoring system in a new data set The estimates of PPV for the various score thresholds depend on the fact that higher scores correspond to higher probability of asthma. The weighted scoring system is valid if it can be shown to retain this property when applied to a new data set, and this was tested by comparing scores with clinical diagnosis of asthma in a subset of data collected in 2001. Sophisticated modelling techniques had been used to rank the 2001 survey questionnaires in order of probability of asthma as part of another study [ 17 , 18 ]. The test subset for this study consisted of questionnaires from the top 10% of this ranked list (most likely asthmatics) and a random sample (approximately 1 in 5) of low probability individuals selected at random from the end 200 of the population list. As in 1995, each selected individual was invited for clinical review and the results were sent separately to another three respiratory physicians. The purpose of the low probability individuals was simply to illustrate the diagnostic discrimination of the physicians. The method of combining the three independent physician opinions differed from that used in phase 1. This time, the aim was not to produce diagnoses as the basis of a mathematical model, but rather to emulate as far as possible diagnosis in a clinical setting. Each physician was therefore asked, on the basis of the clinical review information, to classify each individual as 50% or greater probability of obstructive airways disease (OAD), or less than 50% probability. For those 50% or greater probability (more likely than not to have OAD), the physician was asked to state whether the disease was asthma, chronic obstructive pulmonary disease (COPD) or mixed disease (asthma and COPD). Individuals in whom at least two physicians diagnosed asthma or mixed disease were designated clinically asthmatic. As part of the 2001 population survey, each reviewed individual had already completed a postal questionnaire, and a weighted score was calculated based on questionnaire responses. For each possible score, the proportion of reviewed individuals with that score or higher, who had been designated clinically asthmatic was calculated. These percentages reflect the association between score and probability of asthma and were used as an indication of the validity of the postal questionnaire and scoring system. For comparison, unweighted scores (number of 'yes' answers to the six key questions) were also calculated for respondents from the top 10% of the ranked population list, and the percentage of asthmatics associated with each of the unweighted score thresholds was similarly calculated. Results In the 1995 WYCAP survey, 10429 individuals were sent a postal respiratory questionnaire with a 72.7% response rate, leaving 6825 after exclusion of incomplete questionnaires. Four hundred and twenty individuals were selected for the stratified random sample and invited for clinical review. Of these, 201 (48%) attended, leaving 180 after exclusion of those whose screening questionnaire was incomplete. The ages of the 180 respondents with fully completed questionnaires ranged from 16 to 83 years, with a median age of 50.5 years, and 41.7% were male. Eighty four individuals were designated clinically asthmatic and 96 non-asthmatic based on the combined opinions of the three physicians. Phase 1 The weights derived for each question are shown in table 1 . The features most influential towards classifying an individual as asthmatic (weight 3) were reporting of wheeze in the last 12 months and being in age group 1 (16–34 years). Reporting an asthma attack in the last 12 months and being in age group 2 (35 – 54 years) were next most influential with weights of 2. Cigarette smoking (weight -1) was a small influence away from a classification as asthmatic. Under this scoring system, the responses to questions 5 and 6 (woken by an attack of shortness of breath or coughing) were found to be zero weighted and therefore, when used in combination with this set of questions, did not contribute further to the overall weighted score. The questions relating to breathlessness whilst wheezing and wheeze in the absence of a cold (questions 3.1 and 3.2) are conditional on the reporting of wheeze at any time (question 3) and therefore although the model for the prediction of asthma probability requires them, it is difficult to place a clinical interpretation on their values. Table 1 Weights associated with each question on the questionnaire Question Number Variable based on question Scoring system weights 1 Age (16 – 34 years) 3 1 Age (35 – 54 years) 2 1 Age (55 – 74 years) 1 2 Sex male 1 3 Wheeze in last 12 months 3 3.1 Breathless whilst wheezing 0 3.2 Wheeze in absence of cold -1 4 Woken by chest tightness in last 12 months 1 5 Woken by shortness of breath in last 12 months 0 6 Woken by night cough in last 12 months 0 7 Asthma attack last 12 months 2 8 Currently taking asthma medication 1 9 Family history of asthma 1 10 Hayfever/eczema ever 1 11 Smoker -1 Age was aggregated into 4 categories. The 4th age category (>= 75) years is excluded from the model because it is dependent on the other 3 ** Key variables used in simple scoring of 'yes' answers to key questions Phase 2 Weighted scores were calculated for all postal questionnaires (n = 6825) in the 1995 WYCAP population and ranged from -1 to 13. Table 2 shows the percentage of previously undiagnosed asthmatics expected to be found when using the various scores as thresholds, that is, recommending a clinical review for all individuals with that score or higher. A score of nine or more was found in 335 individuals. Of these, 80 had no evidence of asthma diagnosis in their medical records and it would be expected that 75 (93.5%) would be diagnosed clinically asthmatic. Lowering the score threshold below nine progressively decreased the expected efficiency, so that examining those scoring eight or more increases the number of false positives with only 82.5% expected to be confirmed asthmatic. The 4537 individuals scoring four or less are much less likely to be asthmatic (expected percentage of asthmatics in this group was 5%). Table 2 Positive predictive value estimates for weighted scoring applied to all respondents to 1995 postal survey (n = 6825) Threshold Score Number of individuals above threshold Number eligible for review (no asthma diagnosis) Undiagnosed individuals expected to be diagnosed clinically asthmatic Number Percent (Positive predictive value) 11 76 13 12 90.5 10 178 33 31 94.2 9 335 80 75 93.5 8 577 195 161 82.5 7 891 394 310 78.8 6 1289 715 437 61.2 5 2098 1427 678 47.5 4 3332 2592 769 29.7 3 4667 3874 894 23.1 2 5769 4938 929 18.8 1 6524 5670 933 16.5 All respondents 6825 5964 934 15.7 Phase 3 The third phase confirmed that higher scores correspond to higher probabilities of asthma in the independent test subset selected from the 2001 survey respondents. Two hundred and eighty three individuals comprised the top 10% of the probability ranked list and were invited for review along with 39 low probability individuals selected from the end 200. In total, 145 individuals attended for review, their ages ranged from 16 to 91 years with a median age of 50 years, and 40.7% were male. One hundred and twenty six high probability and 19 low probability individuals attended. For the 126 high probability asthmatics from the top 10% of the ranked list, scores ranged from six to 12, compared with a maximum score of eight in those respondents not in the top 10%. The age range and median age associated with each score threshold is reported in table 3 , which shows that lower score thresholds are associated with increasing age. Table 3 Percentage of individuals with threshold scores or higher who were found clinically asthmatic in the 2001 validation set (n = 126) Threshold Score Number of individuals above threshold Median age (years) [range] Majority verdict asthma Majority verdict mixed disease Majority verdict asthma or mixed disease Number Percent 11 11 33 [21–54] 9 1 10 90.9 10 32 40.5 [21–74] 27 1 28 87.5 9 72 44.5 [18–75] 53 10 63 87.5 8 114 47 [17–88] 76 19 95 83.3 7 121 47 [17–88] 77 19 96 79.3 6 126 47 [17–88] 81 20 101 80.1 In general, agreement between the experts was good with pair-wise kappa statistics of 0.62, 0.64 and 0.65 respectively. All three experts agreed the same diagnostic category for 99 out of the 145 reviewed individuals (82 from the high probability group and 17 from the low probability group) and two out of three experts agreed for a further 38 (37 high probability and one low probability). The percentages of individuals diagnosed as having asthma or mixed disease associated with each score threshold are shown in table 3 . The higher the questionnaire score the higher the probability of asthma in the respondent. Of the 72 individuals scoring nine or more, 63 (88%) were clinically diagnosed as having asthma or mixed disease (53 asthma, 10 mixed). Lowering the threshold reduces the percentage of asthmatics diagnosed, such that of the 121 individuals scoring seven or more only 96 (79%) were clinically diagnosed as having asthma or mixed disease (77 asthma, 19 mixed). The 19 low probability individuals selected from the end 200 of the ranked list all achieved scores of zero or one and 18 were diagnosed clinically non-asthmatic based on the physician opinions. Table 4 shows the percentage of clinical asthmatics associated with each of the non-weighted score thresholds, illustrating that ranking according to non-weighted scores does not demonstrate the ranking in order of probability of asthma that is achieved by the weighted scoring system. Table 4 Percentage of individuals with non-weighted threshold scores or higher who were found clinically asthmatic in the 2001 validation set (n = 126) Number 'yes' answers out of 6 key questions (screening threshold) No. individuals above threshold Number ( percentage ) diagnosed asthma or mixed disease 6 30 25 ( 83.3% ) 5 74 64 ( 86.5% ) 4 103 86 ( 83.5% ) 3 119 98 ( 82.4% ) 2 124 100 ( 80.6% ) 1 126 101 ( 80.1% ) Discussion In general, questionnaire surveys for detecting individuals likely to have a disease partition respondents into 'positive' and 'negative' categories. This is appropriate when a community-wide survey can be undertaken and the medical services are able to provide clinical assessment for all those in the 'positive' category. This study has developed a method for optimising limited resources by targeting expensive clinical examinations to those most likely to be at risk, minimising the chance of needlessly examining a healthy individual and allowing resource availability to be taken into account. Developing a model which uses a questionnaire to identify likely asthmatics in the general population crucially depends on having a subset of questionnaires from respondents reliably classified as 'asthmatic' or 'non-asthmatic' on which to build the model. In a condition such as asthma there is inherent uncertainty in the diagnosis, and even detailed clinical review information inevitably produces disagreement between experts as to the diagnostic category for some of the reviewed individuals. Other studies have produced questionnaire models for predicting high risk asthmatics [ 4 , 9 , 10 , 19 ] and these studies have either used a single expert or resolved the problem of disagreement between experts using methods such as 'majority verdict' or designated diagnostic rules. However, where more than one expert is involved, even permitting the experts to discuss problem cases did not always result in complete agreement [ 4 ]. The importance of this present study lies in the rigorous methods used to capture the uncertainty inherent in asthma diagnosis by combining the opinions of three experts using standard statistical techniques. These aimed to produce a reliable diagnosis for each individual in the phase 1 subset of questionnaires on which the weighted scoring model was built. These methods are described in detail elsewhere [ 14 ] but briefly, the opinions of the three experts were combined using probabilistic techniques which took into account not only differences between responders but also differences in diagnostic judgement thresholds between the experts. The result of applying these techniques was a number between 0 and 1 for each questionnaire in the reviewed phase 1 subset which reflected probability of asthma of the respondent. Those with probability greater than 0.5 were designated 'asthmatic'. Since the aim of a screening system is generally to identify individuals with a high probability of the condition being tested for, the performance of the screening system on high probability individuals is most important. For this reason it was necessary in this study to identify a set of high probability asthmatics for clinical review in the phase 3 validation stage. It was also necessary that these high probability asthmatics were identified by a system which was independent of the weighted scoring system being tested. For this reason, a neural network was used to rank the 2001 survey questionnaires in order of probability of asthma. The neural network is a sophisticated statistical technique that was used to model complex relationships between questions on the questionnaire to predict probability of asthma of a respondent based on questionnaire responses. The neural network model was validated [ 20 ] and applied to the questionnaires from the 2001 population survey to produce a population ranking based on individual predicted probability of asthma according to the neural network model. The top 10% most likely asthmatics from this population ranking, along with some low probability individuals, comprised the independent subset used to test the weighted scoring system in phase 3. The question weights were consistent with clinical explanations. Wheeze in the last twelve months was the most important symptom and the weights for each age band decreased with increasing age which is consistent with the explanation that wheezing in older people can be explained by reasons other than asthma, for example, heart disease, malignant lung disease or one of the chronic obstructive lung conditions collectively referred to as COPD. The interpretation of the low weighting given to respiratory symptoms other than wheeze, for example, night cough or shortness of breath, is that these symptoms are highly correlated with wheezing, and as such provide no additional information over an above the high 'wheeze' weighting. Whereas any of these symptoms viewed in isolation would be highly correlated with a diagnosis of asthma, when viewed in the context of a multivariate model such as this weighted scoring system, it is the interaction between the variables which defines the final model. The small negative coefficient for smoking is consistent with the explanation that those with asthma tend to stop smoking or never to begin. Alternatively, asthma-like symptoms in some smokers may be caused by conditions other than asthma, for example malignant lung disease, COPD, or heart disease, or smoking itself may cause symptoms. The intended application of weighted scoring is to find individuals who are likely to have asthma but are not already receiving appropriate treatment. Superficially, it may appear that the questions relating to an attack of asthma in the last twelve months and current medication for asthma should be excluded from the model as they may produce higher scores in asthmatics who are already diagnosed and receiving treatment, effectively disadvantaging the undiagnosed individuals. However, the relationship between these two questions and practice records of diagnosis is far from definitive. For example in the 1995 survey, of the 6570 respondents who had no record of either diagnosis of asthma nor medication prescription in the last year, 175 individuals answered 'Yes' to the current medication question on the questionnaire and 128 reported an asthma attack in the last twelve months. The scoring system reflects an accurate ranking of the whole population and the questions relating to attack of asthma in the last 12 months and current medication for asthma were found to be important in the whole population model. In general, higher scores reflected higher PPV estimates, but the final column of table 2 illustrates that this is not a simple linear relationship, rather the scores can be considered associated into four priority levels, that is, 9 or more, 7–8, 5–6, and 4 or lower. One possible source of error in the weighted scoring system is under or over self-reporting of symptoms, such as the individual ranked as low probability who was found clinically asthmatic in the absence of reported symptoms on the questionnaire. Another possible source of error is differences between the development data sample and the intended target population due to non-response in the postal survey, exclusion of incomplete questionnaires and non-attendance at clinical review. Analyses revealed no evidence of sample bias from these sources in terms of gender or numbers of 'yes' answers to the six key questions on the questionnaire. Those who attended for review had a higher median age (50 years) than those invited but not attending (32 years). It is also important to acknowledge that the system here was tested using postal questionnaires, whereas in practice the patient may well complete the questionnaire in a surgery or clinic environment. The question weights were derived from a logistic regression model and commonly, scarcity of data means that this modelling technique is applied as a single model fitted to the entire data subset [ 21 ] with none reserved for independent validation and no allowance made for differences in prevalence between training data and the general population. However, this gives limited information about how well the model will generalise to the wider community. In this study three steps were taken to gain a realistic estimate of model performance. First, the regression coefficients were not based on a single model but were the average of several representative models [ 22 ]. Second, there was an independent validation set comprising examples from the higher scores likely to be of interest when targeting diagnostic examinations, along with a small number of respondents believed to have a low probability of asthma. Third, the consensus probability of asthma information available for the phase 1 reviewed individuals allowed the positive predictive values estimated in phase 2 to take account of differences in prevalence of asthma between the questionnaires used in developing the scoring system and the general population. However, the intended target population is practice attendees rather than a population-wide survey and prevalence of all diseases including asthma may be higher in practice attendees than in the general population. In this case the positive predictive values reported in phase 2 may be under-estimates. Assessing likelihood of asthma and ranking by simply counting the number of 'yes' answers to key questions is a simple but relatively crude method since it assumes equal weights for the questions. However, some responses may exert a stronger influence than others, there may be interactions between responses and features such as age and gender cannot be included at all. For example, an elderly person with four 'yes' answers to the key questions may be considered less likely to be asthmatic than a young person with three, since respiratory symptoms in the elderly may sometimes be explained by conditions other than asthma. Hence, ranking according to the number of 'yes' answers may not produce the required probability ordering. This is illustrated by the small differences in proportions of clinical asthmatics associated with the various simple score thresholds shown in table 4 . Where the availability of resources is a limiting factor, the relevant measure of effectiveness in targeting clinical review is the 'true positive' rate, that is, the percentage of individuals scoring above the threshold in whom the diagnosis is confirmed. The other commonly reported test measures of sensitivity, specificity and negative predictive value have no equivalent in a technique based on population ranking where the aim is to target expensive clinical examinations to those most at risk and minimise the 'wasted' examination of false positives. Conclusions When resources for diagnostic examination are limited prioritisation may be necessary. The use of weighted scoring and priority levels rather than the more conventional binary separation into positive and negative allows the score threshold to be adjusted to balance patient need with available resources Abbreviations COPD Chronic obstructive pulmonary disease OAD Obstructive airways disease PPV Positive predictive value WYCAP Wythenshawe Community Asthma Project Competing interests The author(s) declare that they have no competing interests. Authors' contributions SH performed the statistical analysis and developed the weights. TF and PF performed the clinical reviews. JS advised on statistical aspects of combining expert opinions. MH managed the data collection. All authors participated in study design and read and approved the final manuscript. Table 5 Information for estimating PPV for those scoring 9 or more in the 2001 population survey Stratum ( i ) (number of 'yes' answers to key questions) Score ( j ) Number of respondents ω i j proportion of population postal questionnaires in stratum i , ∏' ij prevalence of asthma for score j in stratum i ω i j × ∏' ij 6 12 2 0.0250 0.9914 0.0248 6 11 6 0.0750 0.9914 0.0744 6 10 5 0.0625 0.9867 0.0616 6 9 7 0.0875 0.7996 0.0700 5 12 1 0.0125 0.7181 0.0090 5 11 3 0.0375 0.7181 0.0269 5 10 9 0.1125 0.9777 0.1100 5 9 16 0.2000 0.8938 0.1788 4 10 3 0.0375 0.8908 0.0334 4 9 11 0.1375 0.9975 0.1371 1 – 3 11 1 0.0125 0.9633 0.0120 1 – 3 10 3 0.0375 0.9757 0.0366 1 – 3 9 13 0.1625 0.9880 0.1606 Total 80 1 Estimated PPV for those scoring 9 or more: Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 The Respiratory Questionnaire. The respiratory questionnaire used in the postal survey Click here for file Additional File 2 Estimating PPV for weighted scores. Details of method used for estimating PPV based on postal questionnaire responses Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545076.xml
517718	Glutamine-enriched enteral nutrition in very low birth weight infants. Design of a double-blind randomised controlled trial [ISRCTN73254583]	Background Enteral feeding of very low birth weight (VLBW) infants is a challenge, since metabolic demands are high and administration of enteral nutrition is limited by immaturity of the gastrointestinal tract. The amino acid glutamine plays an important role in maintaining functional integrity of the gut. In addition, glutamine is utilised at a high rate by cells of the immune system. In critically ill patients, glutamine is considered a conditionally essential amino acid. VLBW infants may be especially susceptible to glutamine depletion as nutritional supply of glutamine is limited in the first weeks after birth. Glutamine depletion has negative effects on functional integrity of the gut and leads to immunosuppression. This double-blind randomised controlled trial is designed to investigate the effect of glutamine-enriched enteral nutrition on feeding tolerance, infectious morbidity and short-term outcome in VLBW infants. Furthermore, an attempt is made to elucidate the role of glutamine in postnatal adaptation of the gut and modulation of the immune response. Methods VLBW infants (gestational age <32 weeks and/or birth weight <1500 g) are randomly allocated to receive enteral glutamine supplementation (0.3 g/kg/day) or isonitrogenous placebo supplementation between day 3 and 30 of life. Primary outcome is time to full enteral feeding (defined as a feeding volume ≥ 120 mL/kg/day). Furthermore, incidence of serious infections and short-term outcome are evaluated. The effect of glutamine on postnatal adaptation of the gut is investigated by measuring intestinal permeability and determining faecal microflora. The role of glutamine in modulation of the immune response is investigated by determining plasma Th1/Th2 cytokine concentrations following in vitro whole blood stimulation.	Background Enteral feeding of very low birth weight (VLBW) infants is a challenge, since metabolic demands are high and administration of enteral nutrition is limited by immaturity of the gastrointestinal tract. In particular, small for gestational age VLBW infants may have impaired gut function, as fetal blood flow to heart, brain and adrenals is compensatory increased, while other organs including the gastro-intestinal tract are relatively hypoperfused in intrauterine growth retardation [ 1 ]. Experimental studies have shown that the amino acid glutamine plays an important role in maintaining functional integrity of the gut [ 2 - 5 ]. Glutamine serves as fuel for enterocytes [ 2 ] and provides nitrogen for the synthesis of amino sugars, involved in maintenance of tight junctions [ 3 ] and mucin synthesis [ 4 ]. Moreover, glutamine has a stimulatory and regulatory effect on mucosal cell proliferation and differentiation [ 5 ]. Glutamine is not only utilised at a high rate by intestinal epithelium but also by cells of the immune system. In vitro studies have shown that increasing availability of glutamine stimulates proliferation of T-lymphocytes in response to T-cell mitogens [ 6 ], phagocytosis and antigen presentation by monocytes [ 7 ] and Th1 cytokine response [ 8 ]. In critically ill patients, endogenous glutamine synthesis cannot meet increased demand and for this reason glutamine is considered a conditionally essential amino acid [ 9 ]. VLBW infants may be especially susceptible to glutamine depletion as placental supply suddenly ceases at birth, tolerance of enteral nutrition is limited and parenteral nutrition does not contain glutamine for solubility and stability reasons. Glutamine depletion has negative effects on the functional integrity of the gut [ 10 ] and leads to immunosuppression [ 11 ]. Studies in adults have shown that glutamine supplementation decreases mortality in critically ill adults [ 12 ], infectious morbidity in recipients of bone marrow transplantation [ 13 ] and multiple trauma patients [ 14 ] and length of hospital stay in surgical patients [ 15 ]. In VLBW infants, only two studies have investigated efficacy of glutamine supplementation [ 16 , 17 ]. In the study of Lacey et al [ 16 ], 78 VLBW infants at high risk of developing necrotising enterocolitis (birth weight 530–1250 g) were randomised to receive either standard or glutamine-supplemented parenteral nutrition. After exclusion of 34 infants, 22 treated and 22 control infants were compared for length of stay, days on total parenteral nutrition, days on the ventilator and infectious morbidity. In infants with a birth weight ≥ 800 g no effect of glutamine supplementation was found. However, in infants with a birth weight <800 g glutamine supplementation was associated with shorter time to full enteral nutrition, fewer days on parenteral nutrition, fewer days on ventilatory support and reduced length of stay. Incidence of positive blood cultures and rate of weight gain were not different in glutamine and control groups. Neu et al [ 17 ] performed a randomised controlled trial of glutamine-enriched enteral nutrition in 68 VLBW infants with a gestational age of 24–32 weeks and a birth weight of 500–1250 g. Analysis was performed in 66 infants on an intention to treat basis. Feeding tolerance (as measured by number of days on which feeding had to be withheld) was better in the glutamine group compared to the control group. In addition, after adjusting for birth weight the odds ratio of developing sepsis was 3.8 for the control group compared to the glutamine group. Average weight at different time intervals and length of stay were not different between the groups. Although some methodological concerns can be raised (the sample size of both studies is small; Lacey et al [ 16 ] did not perform analysis on an intention to treat basis), these studies suggest that glutamine supplementation enhances feeding tolerance and decreases infectious morbidity in VLBW infants. The current double-blind randomised controlled trial is designed to determine the effect of glutamine-enriched enteral nutrition on feeding tolerance in a sufficient large population VLBW infants. We hypothesise that time to full enteral feeding is shorter in infants who receive glutamine-enriched nutrition compared to infants in the control group. Furthermore, infectious morbidity and short-term outcome are evaluated. To elucidate the effect of glutamine-enriched enteral nutrition on the functional integrity of the gut, intestinal permeability for macromolecules is measured. As part of the postnatal adaptation of the gut, the intestinal permeability decreases during the first days of life [ 18 ]. We hypothesise that glutamine-enriched enteral nutrition stimulates postnatal adaptation of the gut, reflected by a larger decrease in intestinal permeability. Another aspect of postnatal adaptation of the gut is the development of intestinal microflora. In VLBW infants, the colonisation by bacteria (including beneficial Bifidobacterium and Lactobacillus species [ 19 , 20 ]) commonly present in healthy breast fed infants is delayed [ 21 ]. Intestinal mucin is an important site for bacterial adhesion and colonisation [ 22 ]. Glutamine may improve mucin quality [ 4 ] and consequently influence bacterial colonisation. We hypothesise that glutamine-enriched enteral nutrition stimulates the presence of Bifidobacterium and Lactobacillus species in the intestinal microflora. Intestinal microflora is investigated by determining faecal microflora with a culture independent technique. The effect of glutamine-enriched enteral nutrition on the immune response is investigated by determining plasma T-helper type 1 (Th1) and T-helper type 2 (Th2) cytokine concentrations following in vitro whole blood stimulation. As pregnancy is associated with skewing towards Th2 immunity [ 23 ], Th2 cytokine responses dominate the neonatal immune response [ 24 ]. Exposure to microbes stimulates Th1 cytokine responses and deviates the neonatal immune response towards balanced Th1/Th2 cytokine responses [ 24 ]. We hypothesise that glutamine-enriched enteral nutrition contributes to balanced Th1/Th2 cytokine responses by stimulating the Th1 cytokine response [ 8 ]. To assess safety of glutamine-enriched enteral nutrition, plasma amino acid profiles are determined. We hypothesise that plasma amino acid profiles in glutamine and control groups will not differ during the study period. In addition, to exclude negative effects of glutamine-enriched enteral nutrition on neurodevelopmental outcome, neuromotor development at the corrected age of 1 and 2 years and mental/motor development at the corrected age of 2 years are assessed. In conclusion, this double-blind randomised controlled trial aims to determine the effect of glutamine-enriched enteral nutrition on feeding tolerance, infectious morbidity and short-term outcome in VLBW infants. In addition, an attempt is made to elucidate the role of glutamine in postnatal adaptation of the gut and modulation of the immune response. Methods The study is designed as a double-blind randomised clinical trial. The national central committee on research involving human subjects and the medical ethical review board of our hospital approved the study protocol. Study population Infants with a gestational age <32 weeks and/or birth weight <1500 g admitted to the level III neonatal intensive care unit (NICU) of the VU University Medical Center, Amsterdam, are eligible for participation in the study. Written informed consent is obtained from all parents. Exclusion criteria are: major congenital or chromosomal anomalies, death <48 h after birth, transfer to another hospital <48 h after birth and admission from an extraregional hospital. Treatment allocation and blinding To balance birth weight distribution into treatment groups, each infant is stratified to one of three birth weight groups (<799 g, 800–1199 g, ≥ 1200 g) and randomly allocated to treatment <48 hours after birth. An independent researcher uses a computer-generated randomisation table based on blocks of four (provided by Nutricia Nederland BV, Zoetermeer, The Netherlands) to assign infants to treatment A or B, which correspond to batch numbers on the nutrition products. Investigators, parents, medical and nursing staff are unaware of treatment allocation. The code for the batch numbers is broken after data analysis is performed. Treatment Glutamine powder contains 82% L-glutamine and 18% glucose (nitrogen 15.5 wt/wt%; 371 kcal/100 g), whereas the isonitrogenous control powder contains 100% L-alanine (nitrogen 15.7 wt/wt%, 435 kcal/100 g). The two powders are indistinguishable by appearance, colour and smell. During the study period, glutamine and control powder are monitored for stability and microbiological contamination. Between days 3 and 30 of life, supplementation is administered in increasing doses to a maximum of 0.3 g/kg glutamine per day in the glutamine group. Initially, the supplementation dose is based on birth weight. After 2 weeks the dose is adjusted to actual weight. Two members of the nursing staff daily add supplementation to breast milk or to preterm formula (Nenatal ® , Nutricia Nederland B.V., Zoetermeer, The Netherlands), according to the parents' choice. Per 100 ml, Nenatal ® provides 78 kcal, 2.1 g protein (casein-whey protein ratio 40:60), 4.4 g fat and 7.5 g carbohydrate. Nenatal ® does not contain free L-glutamine. When infants are transferred to other hospitals before the end of the study, the protocol is continued under supervision of the principal investigator. Nutritional support Protocol guidelines for the introduction of parenteral and enteral nutrition follow current practice at our NICU. Administration of parenteral nutrition starts at day 2 and will be advanced gradually until amino-acid intake reaches 3 g/kg/day at day 6. Parenteral nutrition is discontinued if enteral feeding reaches a volume of approximately 150 mL/kg/day. Parenteral nutrition, an all-in-one mixture provided by the hospital pharmacy, contains per 100 mL 54 kcal, 8.5 g glucose, 1.7 g amino acids and 1.7 g lipids. If necessary, glucose, amino acids and lipids are given in separate solutions. Guidelines for the introduction of enteral nutrition are as follows: 1. minimal enteral nutrition starts at day 1 (6–12 mL daily); 2. enteral nutrition is advanced either from day 3 or from day 5 in case of complications: BW <p10, GA <26 weeks, Apgar score at 5 minutes <6, umbilical artery pH <7.10 or base deficit >10 mmol/L; 3. feeding is advanced at a dose of 15–20 mL/kg/day to a maximum of 150 mL/kg/day (based on actual weight). Furthermore, guidelines for reduction/withholding of enteral feeding are: 1. enteral feeding is reduced/withheld in case of gastric residuals (> total volume of past 2 feedings), bilious residuals, emesis, ileus or necrotising enterocolitis Bell's stage ≥ II [ 25 ]; 2. when signs of feeding intolerance resolve, feeding is advanced in the volume given before reduction/withholding within 2 days. For each infant in the study a feeding schedule is proposed, based on birth weight and the guidelines as mentioned above. However, the staff of our NICU has final responsibility for the administration of parenteral nutrition and advancement of enteral feeding. Study outcome measures Study outcome measures Primary outcome of the study is time to full enteral feeding, defined as a feeding volume ≥ 120 mL/kg/day. Furthermore, other parameters of feeding tolerance, infectious morbidity, and short-term outcome are evaluated (Table 1 ). In addition to clinical outcome, intestinal permeability, faecal flora, plasma Th1/Th2 cytokine concentrations and plasma amino acid profiles are determined during the 30 day study period (Table 2 ). Table 1 Clinical outcome measures Remarks Feeding tolerance Enteral feeding >120 mL/kg/day Primary outcome Age at finishing parenteral nutrition Days of no enteral feeding during study period Necrotising enterocolitis Bell et al [25] Infectious morbidity Serious infections Number of infectious episodes Cultured micro-organisms Short-term outcome Weight z scores at birth, day 30 and at discharge Usher et al [33] Patent ductus arteriosus Ventilatory support Use of oxygen at postmenstrual age of 36 weeks Jobe et al [34] Intraventricular hemorrhage Papile et al [35] Retinopathy of prematurity Committee for ROP [36] Death Age at discharge from NICU and at discharge home ROP = retinopathy of prematurity; NICU = neonatal intensive care unit. Table 2 Study schedule < 48 h day 7 day 14 day 30 Amino acid profile x x x x Intestinal permeability x x x x Faecal flora x x x x Th1/Th2 cytokine profile x x x - Clinical outcome measures The following perinatal characteristics are registered to assess prognostic similarity: maternal age and race, obstetric diagnosis, administration of antenatal steroids and antibiotics, mode of delivery, sex, gestational age, birth weight, birth weight <p10, Apgar scores, pH of the umbilical artery, clinical risk index for babies [ 26 ] and administration of surfactant. During the study period actual intake of enteral and parenteral nutrition, powder supplementation and type of feeding (breast milk or preterm formula) are recorded daily. Evaluation of medical records for the presence of serious infections is performed by one investigator/neonatologist, unaware of treatment allocation. Serious infections include sepsis, meningitis, pyelonephritis, pneumonia, and arthritis. Sepsis work-up consists of blood, cerebrospinal fluid and urine (suprapubic bladder tap) culture. Sepsis is defined as the combination of a positive blood culture and the presence of at least two clinical signs (body temperature <36.5°C or >37.5°C, hypotension, tachycardia, apnoeic attacks, feeding problems, irritability or apathy). Meningitis is diagnosed when micro-organisms are cultured in the cerebrospinal fluid. Pyelonephritis is diagnosed when both urine culture and dimercaptosuccinic acid (DMSA) renal scan are positive. Pneumonia is defined as the combination of a positive culture of tracheal aspirate, bronchial secretion or sputum and the presence of at least one clinical sign in ventilated infants (purulent sputum, changed sputum characteristics or deterioration of ventilation settings) or at least two clinical signs in non-ventilated infants (tachypnea, cyanosis, wheezing/rales/crepitation or purulent sputum/changed sputum characteristics). Arthritis is defined as the combination of a positive culture of intra-articular fluid and the presence of signs of articular inflammation. Postnatal adaptation of the gut The effect of glutamine-enriched enteral nutrition on postnatal adaptation of the gut is studied by measuring intestinal permeability and by determining faecal flora. Intestinal permeability is measured by the sugar absorption test, as previously described [ 18 ]. After instillation of the test solution, 2 ml/kg by nasogastric tube, urine is collected for 6 hours. After collection, 0.5 ml chlorohexidine digluconate 20% (preservative) is added to the urine and samples are stored at -20°C until analysis. Lactulose and mannitol concentrations (mmol/mol creatinine) are measured by gas chromatography as previously described [ 27 ]. The lactulose/mannitol ratio is used as a measure of intestinal permeability. Faecal samples are stored at -20°C until analysis by fluorescent in situ hybridisation (FISH) using specific 16S rDNA-targeted probes as described by Harmsen et al [ 28 ]. Immune response The effect of glutamine enriched-enteral nutrition on the immune response is investigated by determining plasma Th1/Th2 cytokine concentrations following in vitro whole blood stimulation. Heparinized blood (0.5 mL), diluted 1:1 in sterile medium (RPMI 1640 without L-glutamine, Gibco, Paisley, United Kingdom) is stimulated for 24 h at 37°C in the presence of anti-CD3/anti-CD28 (Central Laboratory of the Netherlands Red Cross Blood Transfusion service, Amsterdam, the Netherlands) and Escherichia coli lipopolysaccharide (concentration 1:1000 both). After incubation, blood is centrifugated, supernatant is collected and stored at -20°C until analysis. Th1 cytokines IFN-γ, TNF-α, IL-2 and Th2 cytokines IL-4, IL-5 and IL-10 are measured by cytometric bead array (BD biosciences, Alphen aan den Rijn, the Netherlands). Safety Safety of enteral glutamine supplementation in a dosage of 0.3 g/kg/day is investigated by determination of plasma amino acid profiles. Immediately after withdrawal, heparinized blood (0.5 mL) is centrifuged at 10000 rpm for 4 minutes. Plasma is deproteinized by sulfosalicylic acid (2mg/100 μL) and stored at -70°C until analysis. Amino acid profiles are determined by high-performance liquid chromatography as described by Teerlink et al [ 29 ]. To investigate neurodevelopmental outcome, neuromotor development at the corrected age of 1 and 2 years [ 30 ] and mental/motor development at the corrected age of 2 years are assessed [ 31 ]. Sample size We have calculated that a sample size of 102 infants is necessary to detect a difference of at least 2.5 days in time to full enteral feeding, assuming a SD of 4.5 days (two-tailed α = 0.05, β = 0.20). The SD value is based on an retrospective analysis of time to full enteral feeding in infants with GA <32 weeks and/or BW <1500 g admitted to our NICU in 1998. Statistical analysis To determine whether randomisation is successful, prognostic similarity (perinatal and nutritional characteristics) between treatment groups is assessed. The Students' t-test, Mann-Whitney U test, and chi-square test or Fisher's exact test are used to compare continuous normally distributed data, nonparametric continuous data and dichotomous data respectively. Cox regression is performed to examine the effect of glutamine-enriched enteral nutrition on time to full enteral feeding. Logistic regression is performed to examine whether glutamine-enriched enteral nutrition influences the incidence of serious infections. In an additional analysis, adjustments are made for possible confounding factors as administration of antenatal corticosteroids, birth weight <p10, administration of breast milk and other prognostic factors that may be different between treatment groups. Analyses of secondary outcomes (only crude) is performed by Mann-Whitney U test, chi-square test or Fisher's exact test and log rank test for nonparametric continuous data, dichotomous data, and time-dependent data respectively. Generalised estimated equations for longitudinal analysis [ 32 ] are used to analyse changes over time in intestinal permeability, faecal microflora, plasma Th1/Th2 cytokine concentrations and plasma amino acid profiles. Distribution of optimal and non-optimal neuromotor development and normal and abnormal mental/motor development in glutamine and control groups is examined by logistic regression with adjustments for possible confounding factors as gestational age and birth weight. All statistical analyses are performed on an intention to treat basis. In addition, alternative per protocol analyses are performed, excluding all patients who are not treated according to protocol, defined as more than 3 consecutive days or a total of 5 days on minimal enteral feeding or without supplementation. A p value <0.05 is considered significant (two-tailed). SPSS 9.0 (SPSS Inc., Chicago, IL, USA) and STAT 7.0 (StatCorp LP, College Station, TX, USA) are used for data analysis. Competing interests Nutricia Nederland B.V. (Zoetermeer, the Netherlands) provided Nenatal ® , glutamine and placebo supplementation. Authors' contributions Ruurd van Elburg and Willem Fetter formulated the research question and wrote the study protocol. Anemone van den Berg, Ruurd van Elburg and Willem Fetter contributed to the development of the protocol. Jos Twisk gave advice on data analysis. Anemone van den Berg wrote the draft for this manuscript and the other authors reviewed the manuscript. All authors approved the final version of the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517718.xml
545062	The Prader-Willi syndrome murine imprinting center is not involved in the spatio-temporal transcriptional regulation of the Necdin gene	Background The human Prader-Willi syndrome (PWS) domain and its mouse orthologue include a cluster of paternally expressed genes which imprinted expression is co-ordinately regulated by an imprinting center (IC) closely associated to the Snurf - Snrpn gene. Besides their co-regulated imprinted expression, two observations suggest that the spatio-temporal expression of these genes could also be co-regulated. First, the PWS genes have all been reported to be expressed in the mouse nervous system. Second, Snurf - Snrpn and its associated IC are the most ancient elements of the domain which later acquired additional functional genes by retrotransposition. Although located at least 1.5 megabases from the IC, these retroposons acquired the same imprinted regulation as Snurf - Snrpn . In this study, we ask whether the IC, in addition to its function in imprinting, could also be involved in the spatio-temporal regulation of genes in the PWS domain. Results We compared the expression pattern of Snurf - Snrpn and C/D-box small nucleolar RNAs (snoRNAs) MBII - 85 and MBII - 52 to the expression pattern of the two evolutionary related retroposons Ndn and Magel2 , in the developing mouse embryo. We show that these genes have highly similar expression patterns in the central nervous system, suggesting that they share a common central nervous system-specific regulatory element. Among these genes, Ndn and Magel2 display the most similar expression patterns. Using transgenic mice containing the Ndn and Magel2 genes, we show that the transgenic Ndn gene whereas not imprinted is correctly expressed. Search for DNase I hypersensitive sites in the Ndn - Magel2 genomic region and comparative genomic analyses were performed in order to identify potential transcriptional cis -regulatory elements. Conclusions These results strongly suggest that paternally expressed genes of the PWS domain share a common central nervous system-specific regulatory element. We proposed that this regulatory element could co-localize with the IC. However, we demonstrate that the IC, if required for imprinted regulation, is not involved in the spatio-temporal regulation of distantly located retrotransposed genes such as the Ndn gene in the PWS domain.	Background Genomic imprinting in mammals is a process that leads to the preferential mono-allelic expression of specific genes in diploid cells, depending on whether they are inherited from the sperm or from the oocyte. To date, approximately 70 mammalian imprinted genes have been identified which map to at least 11 regions of the mouse genome [ 1 ]. Most imprinted genes are therefore located in clusters, which are generally conserved between human and mouse. For some of these clusters, coordinate imprinted gene regulation has been shown to be controlled by imprinting centers (IC) [ 2 ]. Imprinted genomic regions from the two parents are differentially marked by heritable epigenetic modifications including DNA methylation and histone acetylation and/or methylation. These epigenetic modifications or imprints are established at least for some of them in the germ line of either parent [ 2 ]. The Prader-Willi syndrome (PWS) domain on human chromosome 15q11-q13 and its ortholog on mouse chromosome 7C-D1 are large chromosomal domains containing paternally expressed genes [ 3 ]. PWS results from the loss of expression of several of them including SNURF - SNRPN , NDN , MAGEL2 , MKRN3 and the C/D-box small nucleolar RNAs(snoRNAs). In humans, mini-deletions, upstream the SNURF - SNRPN transcriptional unit, have led to the characterization of an IC which coordinates their imprinted expression [ 4 ]. One sub-region of this IC defined as the shortest region of deletion overlap in PWS patients (PWS-SRO) and encompassing the SNURF - SNRPN promoter, has been shown to be required for maintaining the paternal imprint [ 5 , 6 ]. Its deletion in the germ line or post-zygotically on the paternal chromosome leads to silencing of all paternally expressed genes of the PWS domain [ 4 , 6 ]. In mice, although the Snrpn promoter is not required for genomic imprinting [ 7 ], deletion of a 35 kb region including 16 kb of sequences upstream Snurf - Snrpn exon 1 to Snurf - Snrpn exon 7 unit also leads to silencing of the paternally expressed genes, indicating that both the position of the IC and its role in the coordinate imprinted expression of genes is conserved between human and mouse [ 8 , 9 ]. Recently, examination of tissue-specific pattern of mRNA expression at a genomic scale allowed the identification of several chromosomal regions harbouring tissue-specific co-regulated genes defined as regions of correlated transcription (RCTs) [ 10 ]. Noticeable, some of these RCTs overlap with known imprinted loci and it suggests that transcriptional regulatory elements controlling imprinted expression might also regulate tissue-specific expression. Therefore, one might hypothesize that this type of regulation is applied at the PWS domain. Two arguments support this hypothesis. First, all the murine orthologues of the PWS genes have been reported to be expressed in the developing nervous system [ 11 ]. Second, Nicholls proposed an evolutionary model for the origin of the human and mouse PWS domains [ 12 , 13 ]. In this model, the Snurf - Snrpn locus and its associated IC would be the ancestral imprinted transcriptional unit of the domain. The other genes, which are intronless, would have later and sequentially been acquired by retrotransposition and adopted the same imprinted regulation. Altogether, these data suggest that a spatio-temporal co-regulation of PWS genes could exist and have been acquired through the evolution of the domain. Thus, it is tempting to speculate that the cis -regulatory element(s) initially involved in the spatio-temporal regulation of the Snurf - Snrpn locus might influence the spatio-temporal regulation of retrotransposed genes in the PWS region. The IC could be a good candidate to play this role. In this study, we investigate this hypothesis. First, we compared the expression pattern of the Snurf - Snrpn , MBII - 85 , MBII - 52 , Ndn and Magel2 in the developing mouse embryo. We show that their expression in the embryo is restricted to neural tissues and that these genes display strikingly similar expression patterns in the central nervous system. Second, we created transgenic mice with a BAC containing the Ndn and Magel2 genes. We showed that the Ndn transgene is not imprinted but is correctly expressed in the developing embryo. These results demonstrate that if the IC is required for imprinted regulation, it is not involved in the Ndn spatio-temporal regulation. Finally, since we have shown an almost identical spatio-temporal expression profile of Ndn and Magel 2 in the developing embryo, it suggests that these two genes might share common cis -acting regulatory elements which should be present in the BAC transgene. We therefore searched for DNaseI hypersensitive sites in the Ndn - Magel2 genomic region and performed comparative genomic analyses in order to identify potential transcriptional cis -regulatory elements. Results Nervous system tissue-specific expression of mouse PWS genes in the embryo In order to determine if mouse PWS genes might be transcriptionally co-regulated, we compared the expression profiles of Snurf - Snrpn , MBII -85, MBII -52, Ndn and Magel2 by in situ hybridization in the developing mouse embryo (Fig. 1 , 2 and 3 ). In the 10.5 and 13.5 mouse embryos, Snurf / Snrpn , MBII - 85 , MBII - 52 , Ndn and Magel2 are almost exclusively expressed in the developing nervous system, and although expressed at different levels, they display strikingly similar expression patterns in the central nervous system (brain and spinal chord). As previously reported for Ndn [ 14 ], they are predominantly expressed in all the ventral parts of the neural tube, mostly or even exclusively in marginal areas where differentiating neurons reside. In the peripheral nervous system, Ndn and Snurf / Snrpn are both expressed at high levels in cranial and dorsal root ganglia, sympathetic and parasympathetic ganglia, structures in which MBII - 85 , MBII - 52 and Magel2 trancripts are expressed at much lower levels or not at all (Fig. 2 and 3 ). Magel2 when compared to Ndn , Snurf - Snrpn and MBII - 85 / - 52 seems to have a more highly restricted expression domain. In the 12.5 embryo, Magel2 transcripts are detected at high levels and predominantly in the hypothalamus (Fig. 3 ). Careful examination shows that although found at low or very low levels, Magel2 transcripts are present in similar domains as the four other PWS genes studied. Ndn and Magel2 also share expression domains in some non-neuronal tissues such as in the muscles (skeletal muscles and tongue) and in some non-neuronal neural crest cells derived structures such as the branchial arches (Fig. 3 ; data not shown). Magel2 transcripts are detected at very low levels in the dorsal root ganglia and not detected in cranial ganglia. Figure 1 Imprinted cluster on mouse chromosome 7C. Upper and lower boxes represent genes expressed from the paternal and maternal alleles respectively. Small nucleolar RNAs (snoRNAs) MBII - 13 , MBII - 85 and MBII - 52 genes lie within introns of long primary transcripts initiated at the U exons (U9 to U1 exons). Arrows indicate the transcriptional orientation of the genes. Frat3 , Mkrn3 , Magel2 and Ndn lie within 120 kb and around 1.5 to 2 Mb from the Snurf - Snrpn gene. Location of BAC109 used for the transgenic study is represented. Figure 2 Comparison of Ndn and Snurf-Snrpn transcriptional unit expression patterns in the E13.5 mouse embryo. Comparison of Ndn (A, E, I), Snurf - Snrpn (B, F, J), MBII - 85 (C, G, K) and MBII - 52 (D, H, L) RNA expression on E13.5 mouse embryo adjacent sagittal sections, from lateral to more medial sections (rows I to II). Ndn as well as Snurf / Snrpn , MBII - 85 and MBII - 52 are almost exclusively expressed in the developing nervous system and display strikingly similar expression patterns in the developing brain and spinal chord. In the peripheral nervous system, Ndn and Snurf - Snrpn are expressed at similar levels in cranial ( Ndn : A; Snurf - Snrpn : B) and dorsal root ganglia ( Ndn : I; Snurf - Snrpn : J) whereas MBII - 85 (C, K) and MBII - 52 (D, L) are expressed at much lower levels or not at all. Note that Ndn is additionally expressed in muscle tissues such as the skeletal muscles (A) and the tongue (E), and in the adrenal primordium (I). ap, adrenal primordium; drg, dorsal root ganglia; gn, gonad; hp, hypothalamus; ht, heart; lg, lung; lv, liver; mb, midbrain; mo, medulla oblongata; mt, metanephros; ms, muscles; poa, post-optic area; pp, telencephalic epithelium preplate; ps, pons; sa, septal area; tg, tongue. Asterics indicate cranial and dorsal root ganglia in A, B, C, D. Figure 3 Comparison of Magel2 and Ndn expression patterns in the E10.5 and E12.5 mouse embryo. Comparison of Magel2 (A, C, E, G) and Ndn (B, D, F, H) RNA expression in E10.5 (A, B, C, D) and E12.5 (E, F) mouse embryos on adjacent sagittal sections, and in E12.5 (G, H) mouse embryos on adjacent transversal sections. In the E10.5 embryo, the predominant expression in the ventral parts of the neural tube is particularly evident (A, B, C, D) for both genes. Magel2 and Ndn display a strikingly similar profile of expression in the infundibulum recess (C, D). Asterics indicate expression of both genes in symetrical ventral stripes of cells in the neural tube (A, B). Note that Magel2 as Ndn are expressed in the mandibular component of the first branchial arch (C, D), in the tongue (E, F). In the E12.5 embryo, Magel2 and Ndn are expressed in identical structures such as the ganglionic eminence, the pons, the medulla oblongata, in the preplate of the telencephalic vesicle, the hypothalamus, the ventral thalamus, the zona limitans intrathalamica and in the basal plate (E, F, G, H). drg, dorsal root ganglia; ge, ganglionic eminence; hp, hypothalamus; mcba, mandibular component of the first branchial arch; mo, medulla oblongata; pp, telencephalic epithelium pre-plate; ps, pons; rp, rathke pouc;h so, somites; tg, tongue; vt, ventral thalamus; zi, zona limitans intrathalamica. In conclusion, all the PWS genes that we studied were expressed mainly ( Ndn and Magel2 ) or even exclusively ( Snurf - Snrpn , MBII - 85 /- 52 ) in the mouse developing nervous system at E12.5 and E13.5, with strikingly similar identical patterns in the central nervous system (brain and spinal chord). More divergent expression patterns were observed in the peripheral nervous system (cranial and dorsal root ganglia, sympathetic ganglia), some genes being expressed at high levels ( Ndn and Snurf - Snrpn ) and the others being expressed at low levels or not at all ( MBII - 85 , MBII - 52 and Magel2 ). Our in situ hybridization data suggests the presence in the PWS domain of a central nervous system-specific neural element which coordinates the PWS genes expression. Transgenic experiments In order to determine if the spatio-temporal transcriptional regulation of PWS genes is dependant upon the IC, we initiated BAC transgenic analyses and chose to investigate the transcriptional regulation of the Ndn gene outside its natural genomic context. Since Ndn and Magel2 are co-regulated and the intergenic region between these two genes is around 30 kb only, we chose a bacterial artificial chromosome (BAC 109) containing both genes to generate transgenic mice (Fig. 4A ). Three transgenic male founders were obtained and crossed with C57BL/6 females. Only one male transmitted the BAC109 transgene to its progeny and gave rise to the transgenic line 92 (Tg92) described in this study. The presence of BAC vector sequences allowed us to discriminate the transgenic from the wild type DNA. Genomic DNA from Tg92 mice was thoroughly analyzed by PFGE, Southern blot and PCR, to precisely determine the structure and the number of copy of the integrated transgene (Fig. 4A , data not shown). The transgene integrated in one full copy along with a truncated copy containing the whole 5' region upstream Ndn up to the Ndn promoter (Fig. 4A ), near the centromeric region of chromosome 2, as determined by DNA FISH analysis (data not shown). Figure 4 Transgenic analyses. (A) Structure and copy number of the BAC109 transgene in line Tg92. The BAC109 transgene, the relative positions of the Ndn and Magel2 genes, and the transgenic Eag I and Pme I restriction fragments detected with probes 5' and 3' are represented on the upper diagram. Genomic DNA was isolated from wild type (WT) or transgenic (TG) mice, digested by Eag I or Pme I, separated by PFGE, blotted and hybridized to the 5' or 3' probes. The presence of the 54 and 50 kb Eag1 transgenic fragments detected by the 5' and the 3' probes respectively demonstrate the integrity of the transgene. Eag 1 is a methylation sensitive enzyme which explains the presence of a higher hybridization band corresponding to undigested methylated DNA (Und). Hybridization of Pme I digested genomic DNA with either the 5' or 3' Ndn probes, indicated that the BAC integrated in one full copy along with a truncated copy containing the whole 5' region upstream Ndn up to the Ndn promoter.(B) Non imprinted brain-specific expression of transgenic Ndn . Ndn expression was tested by Northern blot analysis. RNAs were isolated from Tg92mat- Ndn +/-pat or wild type littermate tissues: Br, brain, Kd, kidney, Lv, liver, Sp, spleen, Th, thymus, Ht, heart, Mu, muscles. Note that Ndn transgenic mRNAs are not detected in adult muscles whereas Ndn endogenous mRNAs are. After being hybridized to an intragenic Ndn probe, Northern blots were stripped and rehybridized to a β-actin probe to check for the presence and integrity of RNAs. Transgenic Ndn expression analysis was performed by Northern blot hybridization, on adult tissues of mice inheriting the transgene either maternally or paternally, on an Ndn null background (Tg92mat- Ndn +/-pat and Tg92pat- Ndn +/-pat, respectively). Whether maternally or paternally inherited, the transgenic Ndn gene was expressed with the same tissue specificity as the endogenous Ndn gene (Fig. 4B ). Transgenic Ndn transcripts were only detected in the brain, although at a slightly lower level than endogenous transcripts. It should be noted that no transgenic Ndn gene expression was detected in muscle, a tissue in which endogenous Ndn transcripts are detected although at low levels as compared to the brain. In situ hybridization experiments were then performed on transgenic embryos carrying a paternal deletion of the Ndn allele (Tg92mat-Ndn+/-pat) and wild type mouse embryos, to compare the transgenic and endogenous profile of Ndn expression. Transgenic Ndn transcripts were detected in embryos from E10.5 as endogenous Ndn transcripts, and in the same structures, namely the central and peripheral nervous system (Fig. 5 ), at a slightly lower level in transgenic embryos as it was noted in transgenic adult tissues. No transgenic Ndn expression was detected in the dermo-myotome, the muscles or the tongue, although endogenous Ndn transcripts were detectable in these tissues (Fig. 5 : H and I). Since Magel2 null mice were not available, Magel2 imprinting and expression from the BAC109 transgene could not be analyzed. However, no ectopic site of Magel2 expression was noted during embryogenesis. Figure 5 Comparison of the wild type and transgenic Ndn expression profiles in E10.5 and E12.5 mouse embryos. In situ hybridization was performed with an Ndn riboprobe on sagittal sections of wild type (A, B, C, H), Tg92mat- Ndn +/-pat (D, E, F, I) and Ndn+/-pat (G) embryos. The absence of Ndn signal in the E10.5 Ndn +/-pat embryo (G) demonstrates the specificity of the Ndn riboprobe. Similar expression profiles are shown in the infundibulum recess (ir) of E10.5 embryos (A, D), in the preplate of the telencephalic epithelium (pp) and in the dorsal root ganglia (drg) of E12.5 embryos. Note that no hybridization signal is detected in the transgenic Tg92mat- Ndn +/-pat E12.5 muscle (ms) (I) whereas endogenous Ndn is expressed in this tissue (H). rp, rathke pouch; tv, telencephalic vesicle. Since the transgenic Ndn gene was not imprinted, methylation studies of the transgenic Ndn promoter region and CpG island were omitted. Search for Ndn transcriptional regulatory elements Since Ndn was correctly expressed at least in the nervous system from the BAC transgene, Ndn regulatory elements must be present in the BAC109 genomic sequences. With the increasing availability of genomic sequence and the recent development of powerful global alignment algorithms, interspecies genomic sequence comparisons are becoming an efficient mean to identify conserved non-coding sequences (CNS) which regulatory potential can further be assessed experimentally [ 15 , 16 ]. However, since the whole human and mouse PWS domains are rather highly conserved which might render difficult to discriminate between functional and non functional CNS, we first undertook an experimental approach. We analyzed the chromatin DNase I sensitivity of the genomic sequences present in the BAC transgene. In the adult brain, hybridization of Bgl II- or Bam HI-cut DNA with a series of probes described in Fig. 1 led to the characterization of several major DNaseI hypersensitive sites associated to the Ndn gene (HS1/2, HS3 and 4) and of one site upstream the Magel2 gene (HS5) (Fig. 6 ). The probes used allowed a systematic analysis of the Ndn / Magel2 genomic sequences present in the BAC transgene, from -17.5 kb upstream the Ndn gene to17 kb downstream the Magel2 gene. Additional upstream Ndn sequences which were present in the BAC could not be analyzed since their sequences are not yet available. Figure 6 Mapping of DNase I hypersensive sites in the Ndn and Magel2 genomic region. (A) The diagram shows a map of the sequenced genomic region present in the BAC109. Localisation of HS sites is indicated by arrows in relation to the Ndn and Magel2 genes, and to the AK086725 EST. Position of Bgl II and BamH 1 sites and probes 1 to 10 are indicated. Asterics indicate polymophic Bgl II restriction sites present on castaneus alleles detected with probes 1/2 and probes 7/8. A, B and C design three non-coding regions of homology between the mouse and the human genomes.(B) Southern blot analysis of Bgl II or BamH I cut genomic DNA isolated from brain nuclei treated with increasing concentrations of DNase I (from 0 to 100 U). Maternal (mat) and paternal (pat) alleles are indicated when discrimination is possible.(C) Sequence comparison between mouse and human Ndn / Magel2 genomic regions. Alignments were peformed with the genome VISTA program (window size 100 bp, homology threshold 70%). Conserved non-coding sequences (CNS) are depicted in pink, untranslated regions (UTR) and ORF in pale and dark blue respectively. Genes are indicated by blue arrows. Repeated elements are depicted above the sequences comparison. Mapped HS sites are indicated by vertical arrows. A, B and C design three non-coding regions of homology between the mouse and the human genomes. Although the C region is depicted as a non-gening sequence, recent isolation of ESTs corresponding to this region suggests that it might belong to the Magel2 transcription unit. The asterisk above the C region indicates the beginning of the ESTs. Hybridization of Bgl II-cut DNA with probe 1 revealed two adjacent (~50–70 bp apart) and strong DNase I hypersensitive sites (HS1 and HS2) in the Ndn promoter region of the paternal allele (Fig. 6B ). Hybridization with probe 2 revealed two additional but weaker DNase I hypersensitivity sites on the paternal allele (HS3 and HS4), localized in the Ndn 5'UTR and at the beginning of the coding region, respectively. In contrast to the paternal allele, the maternal allele was highly resistant to DNase I digestion. In the adult kidney, a tissue in which Ndn is not expressed, the two parental alleles were highly and almost equally resistant to DNase I digestion (data not shown). The presence of DNase I hypersensitive sites HS1/2, HS3 and HS4 is therefore closely linked to the transcriptional activity of the Ndn gene since they were present on the paternal allele and only in Ndn -expressing tissues. Hybridization of brain BamH 1-cut DNA with probe 6 or Bgl II-cut DNA with probes 7 and 8 revealed one relatively strong DNase I hypersensitive site (HS5) 3.5 kb upstream the Magel2 gene, which was present on both parental alleles and co-localized with a polymorphic Bgl II site present on the CAST/Ei allele. HS5 was localized 1 kb upstream newly isolated full length Magel2 ESTs (AK086725, AK082944, AK086725; see Fig. 6A ) which initiate around 2.5 kb 5' to the described Magel2 mRNA (NM013779) [ 17 ]. HS5 was present on both maternal and paternal alleles and in tissues in which Magel2 and/or Ndn are not expressed such as the adult kidney (data not shown) which suggests that this DNase I hypersensitive site is not linked to the transcriptional activity of these two genes. No prominent DNase I hypersensitive site were detected in the Ndn - Magel2 intergenic region (excepted HS5). Multiple faint hybridization signals could however be detected with probes 3 and 4 on Bgl II-cut DNA (data not shown) and with probe 5 on BamH 1-cut DNA, revealing regions of low DNase I hypersensitivity in the Ndn - Magel2 intergenic region (Fig. 6B ). Finally, hybridization of probe 10 on brain BamH 1-cut DNA allowed the detection of a lower hybridization band, revealing one or potentially two DNase I hypersensitive sites (HS6a or/and HS6b). We were not able to precisely localise the HS6 site(s) as well as its (their) parental origin. We next mapped HS1 to 6 DNase I hypersensitive sites on a mouse/human sequence alignment (Fig. 6-C ). The available mouse sequences used for this comparison start from -17.5 kb upstream the Ndn gene to the Magel2 extremity of BAC109. Several CNS defined as orthologous sequences greater than 100 bp and greater than 70% identity [ 15 ] were found and mainly gathered into three regions: upstream the Ndn gene (region A), in the intergenic Ndn - Magel2 region (region B) and in a 2.8 kb upstream the Magel2 gene (region C). None of the hypersensitive sites identified in this study mapped to region A, B or C. HS1/2 which are closely linked to the transcriptional activity of the Ndn gene are localized in a CNS immediately upstream the Ndn gene. HS5 co-localized to an isolated CNS localized 0.7 kb upstream the highly conserved C region and around 1 kb upstream the newly described full length Magel2 ESTs. However, the functional significance of HS5 in the Magel2 and/or Ndn gene transcriptional regulation remains to be determined since this site was neither allele nor tissue-specific. It should be noted that there was almost no conservation of sequence in the region downstream the Magel2 gene. In contrast, the intergenic Ndn - Magel2 region was relatively well conserved and regions of low DNase I hypersensitivity were detected in this region which might have functional importance. Discussion Coordinated central nervous system-specific expression of PWS genes Study and comparison of Snurf - Snrpn , MBII - 85 /- 52 snoRNAs, Ndn and Magel2 expression profiles brought new information on the transcriptional regulation of these genes. Snurf - Snrpn , the snoRNAs MBII - 52 and MBII - 85 and the distantly located Ndn and Magel2 genes were expressed in the mouse developing nervous system at E10.5 and E13.5 with strikingly similar patterns in the central nervous system. More divergent expression patterns were observed in the peripheral nervous system and non-neuronal tissues. Accordingly, Snurf - Snrpn , MBII - 85 and MBII - 52 snoRNAs were exclusively expressed in the developing nervous system, and in identical structures in the central nervous system. Marked differences of expression between Snurf - Snrpn and the snoRNAs were found in cranial and dorsal root ganglia, structures in which Snurf - Snrpn was highly expressed whereas the snoRNAs were expressed at very low levels ( MBII - 85 ) or not at all ( MBII - 52 ). These results reinforce the proposal that the snoRNAS derive from the processing of long primary transcripts initiated at the U exons, upstream Snurf - Snrpn [ 18 ] (Le Meur et al., submitted). It is interesting to note that Snurf - Snrpn was previously reported to be ubiquitously expressed in adult tissues [ 12 ], whereas we found it to be exclusively expressed in the nervous system at E10.5 and 13.5 of mouse embryogenesis. As previously reported [ 14 , 11 ], Ndn and Magel2 mRNAs were detected almost exclusively in the nervous system of the mouse embryo from E9.5/10.5 onwards, and although Magel2 transcripts were detected at much lower levels than Ndn transcripts, these two genes displayed almost identical expression patterns in the central nervous system (brain and spinal chord). Particularly striking was their identical pattern of expression in the E10.5 embryo. At later developmental stages, Magel2 expression domain became more restricted but remained included in Ndn expression domain. Magel2 transcripts were however detected at very low levels in almost all the structures in which Ndn was expressed, including non neuronal structures such as the mandibular component of the first branchial arch, the tongue and the muscles. Magel2 transcripts have been reported to be particularly instable due to the presence of AU-rich elements (ARE) in their 3' untranslated region [ 17 ] and this could explain why they were detected at such low levels. Our expression analysis in the mouse embryo indeed suggests the existence of a central nervous system-specific neural regulatory element coordinating the PWS genes expression. Ndn and Magel2 , specifically, could share additional regulatory elements. Co-expression of PWS genes and the IC We further emitted the hypothesis that the central nervous system-specific regulatory element could be physically associated to the PWS IC because of the proposed evolutionary history of the PWS domain and because the IC is involved in coordinating long range chromatin modifications leading to imprinted expression of genes in the whole domain. The Snurf - Snrpn locus with its associated IC has been proposed to be the most ancestral transcriptional unit of the domain [ 12 , 13 ]. Other PWS genes ( Ndn , Magel2 , Mkrn3 and Frat3 ) would have later been acquired by retroposition and/or local cis -duplication of retroposed genes and would have adopted the same transcriptional regulation as the Snurf - Snrpn transcriptional unit. Unless retroposons arise from reverse-transcriptase-mediated processing of aberrant or alternative transcripts including endogenous promoter elements, their functionality is dependent upon regulatory elements present at their site of insertion. Retroposons in the PWS region could have therefore adopted the Snurf - Snrpn regulation both for imprinted and spatio-temporal expression. Finally, the fact that paternal deletions of the IC abolish the expression of all the PWS genes [ 8 , 7 ] also suggests a possible intertwining between regulatory elements controlling both imprinting and spatio-temporal expression. Our study did not however confirm that the IC might coordinate the spatio-temporal expression of PWS genes. Our transgenic analysis clearly showed that Ndn does not rely on the IC for its spatio-temporal expression. Ndn was correctly regulated from the BAC transgene lacking the IC both in the developing embryo and in adult tissues. These results demonstrate that the cis -regulatory sequences involved in both the developmental and tissue-specific expression of Ndn were present in the transgene and were not associated to the IC. The fact that the IC was not involved in the spatio-temporal regulation of the Ndn gene does not however exclude its putative involvement in the spatio-temporal regulation of the Snurf - Snrpn transcriptional unit. No transcriptional regulatory elements or sequences excepted those involved in the imprinted expression [ 19 ] have yet been characterized in the Snurf - Snrpn transcriptional unit. Search for Ndn and Magel2 regulatory elements Since the cis -regulatory sequences involved in both the developmental and tissue-specific expression of Ndn were present in the transgene and since we showed that Ndn and Magel2 expression was coordinated both in the central nervous system and in some non neuronal tissues of the mouse embryo, we searched for Ndn and Magel2 regulatory elements in the BAC109 genomic sequences. Ndn and Magel2 are two evolutionary related retrotransposons, which belong to the MAGE D gene family [ 20 ]. Phylogenetic studies could not predict whether these two genes arose from two distinct retroposition events or whether they arose from the initial retroposition of one of these two genes followed by a cis -duplication event (Blanc M., unpublished observations). Their close genomic localisation, within 30 kb in the mouse genome, would however be in favour of this second hypothesis. The co-expression of Ndn and Magel2 could therefore result from regulatory elements sharing and/or duplication. Intraspecies genomic comparisons of mouse or human Ndn and Magel2 upstream sequences did not reveal any significant homology which does not exclude the possibility that one of these two genes arose by a cis - duplication. Our experimental search for regulatory elements elements in the Ndn / Magel2 genomic sequences was limited to sequences available in the databases and was therefore not exhaustive. However, we found several DNase I hypersensitive sites (HS1/2, HS3 and 4) linked to Ndn transcriptional activity and associated to the Ndn promoter. Previous transgenic studies in the zebrafish using a series of hybrid transgenes containing various mouse Ndn promoter sequences associated to the reporter LacZ gene suggested that the mouse Ndn promoter from -845 bp to +63 bp functioned in the zebrafish embryo in a temporal, spatial and tissue-specific manner [ 21 ]. In particular, a cis -acting element driving the neuronal-specific expression was located into an 87 bp sequence from -173 to -87 bp of the Ndn gene which exactly co-localizes with HS1/2. No identified transcriptional factor brain-specific binding site could however be identified in this sequence. Deletion of promoter sequences in Ndn -KO mutant mice [ 22 ] did not affect Magel2 transcriptional regulation (Watrin F., data not shown) which suggests that if the coordinated Magel2 and Ndn expression results from the sharing of a putative enhancer, this enhancer is not associated to Ndn promoter sequences. The human NDN gene has been shown to be expressed in a larger panel of tissues than the mouse Ndn gene [ 23 , 24 ] but our data suggest that these two genes might have similar expression profiles in the nervous system [ 23 ] (data not shown). An interspecific (mouse/human) comparison of available genomic sequences was therefore performed in order to identify genomic sequences involved in central nervous system expression. This comparison did not allow the identification of such sequences but nevertheless brought some new information on the Magel2 transcriptional unit, showing the existence of a 2.5 kb highly conserved region upstream the described Magel2 gene. Identification of full length embryonic and brain specific Magel2 ESTs including this region further confirms that this region belong to the Magel2 transcriptional unit. In view of our results, expression of the Ndn and Magel2 genes could be coordinated by a common enhancer distinct from Ndn promoter sequences and which could be localized in sequences upstream the Ndn gene that we could not analyze. It should be noted that the Ndn / Magel2 intergenic region was rather well conserved and harbored several regions of low DNase I hypersensitivity which might deserve further investigation. The transgene is not imprinted As expected, our transgene which was physically separated from the IC was not imprinted. When outside of its natural genomic context, Ndn was expressed from either parental allele. Whether particular genomic sequences around or associated to Ndn such as the 5' CpG island are necessary to respond to the primary imprinting signal established at the IC remains to be determined. An imprinted brain-specific Ndn non coding (nc) antisense RNA (PX00010K13; DDBJ accession no. AK14392) initiated in (or including) the 5' part of Ndn coding sequence and extending in Ndn upstream sequences has recently been described and according to the authors, this antisense transcript could be involved in the imprinted expression of Ndn mRNA [ 25 ]. This transcript could potentially be transcribed from our transgene. We did not detect this antisense nc transcript in transgenic brain RNA which is coherent with the fact that the transgene was not imprinted. However, we did not detect it in wild type brain either (data not shown). In one of the mouse models in which part of the Ndn CDS -and therefore the region in which the antisense transcript is initiated- was deleted [ 26 ], the LacZ gene which replaced the Ndn was monoallelically expressed, further strengthening an absence of role of this hypothetical antisense transcript in Ndn imprinting. Conclusions Our analysis strongly suggests the existence of a central nervous system-specific regulatory element which would coordinate expression of the PWS genes in the developing mouse embryo and we proposed that this element could be associated to the IC. However, our transgenic studies clearly demonstrate that when physically separated from the IC, a transgenic Ndn gene, although not imprinted anymore, can still be correctly expressed in the developing nervous system of the mouse embryo. The BAC transgene therefore contained the regulatory elements needed for the spatio-temporal expression of Ndn (and most likely Magel2 ) in the developing nervous system. Three hypotheses can be made: 1) the PWS genes co-expression that we observed is incidental, 2) the element(s) localised in genomic sequences surrounding the Ndn / Magel2 and which regulate(s) Ndn (and Magel2 ) expression could also regulate other PWS genes, 3) several central nervous system-specific regulatory elements resulting from duplication events might be present in the PWS domain. The evolutionary history of the PWS domain is more in favour of the third hypothesis but will need further investigations to be confirmed. Methods Mice Adult C57BL/6 and C57BL/6 X CBA mice were purchased from IFFA CREDO, and M. musculus castaneus (CAST/Ei) male mice from the Jackson Laboratory. Mice carrying an Ndn null mutation on a 129/Sv genetic background [ 22 ] and transgenic Tg92 mice were bred in-house. In situ hybridization All in situ hybridization experiments for the study of PWS genes expression were performed on C57BL/6 embryos. In situ hybridization was performed on 14 μM paraformaldehyde-fixed cryosections with antisens digoxigenin-labeled riboprobes. Sections were washed in 1X PBS, treated with RIPA buffer (150 mM NaCl, 1% NP-40, 0.5% Na deoxycholate, 0.1% SDS, 1 mM EDTA, 50 mM Tris pH 8), post-fixed in paraformaldehyde, acetylated and further washed in PBST before the hybridization step. All hybridization and post-hybridization washes were performed at 70°C. Hybridization was performed in 50% formamide, 5X SSC, 5X Denhart's solution (Sigma), 0.5 mg/ml herring sperm DNA, 0.25 mg/ml yeast RNA. Washes were performed in 50% formamide, 2X SSC, 0.1% Tween 20. Digoxigenin labelling was detected using anti-digoxigenin Fabs (Roche Biochemicals) coupled to alkaline phosphatase and NBT/BCIP (Sigma). The Ndn riboprobe (290 bp) hybridizes to the 3' UTR of the Ndn mRNA (nt 2130 to 2420; accession number D76440). The Magel2 riboprobe (318 bp) hybridizes to the 3' part of the Magel2 ORF (nt 3730 to 4048; accession number AK086725). The snoRNAs MBII - 85 and MBII - 52 riboprobes were synthesized from plasmids given by J. Cavaillé. The Snurf riboprobe (277 bp) hybridizes to U / Snurf - Snrpn transcripts and recognizes sequences from the 3' end of the Snurf exon 1 to the 3' end of the Snurf exon 3 (nt 3639 to nt 77687; accession number AF332579). Generation and breeding of transgenic mice BAC109 was isolated by hybridization of mouse bacterial artificial (BAC) high-density membranes (Research Genetics, Inc, USA) with a genomic Ndn probe, and contains a 103 kb Not I insert comprising the Ndn and Magel2 genes. The Ndn gene is localized in the center of BAC109, the Eag I intragenic site being situated 54 and 50 kb away from the vector Not I sites and the 3' extremity of the Magel2 gene is localized 17 kb from one of the vector Not I site (Fig. 3 ). Unlinearized cesium chloride gradient purified BAC109 DNA was resuspended in injection buffer (10 mM Tris-HCl pH 7.5, 0.1 mM EDTA pH 8.0, 100 mM NaCl, 30 μM spermine, 70 μM spermidine) at 3 ng/μl and microinjected in the pronucleus of C57BL/6 X CBA fertilized eggs. Founders were identified by Southern blot analysis using a PCR probe amplified from the pBeLoBAC11 vector and a Bgl II digestion, and subsequently bred on a C57BL/6 genetic background. Transgene copy number and integrity were determined by PFGE and Southern blot analysis. The 5' and 3' probes used in the Southern blot analysis correspond to the probes 1 and 2 described in the "DNase I Hypersensitivity Mapping" section. Transgenic Ndn expression analysis (Northern blot and in situ hybridization) was performed on embryos or adult mice inheriting the transgene either paternally or maternally, on a Ndn null background. These embryos or adult mice were obtained by mating transgenic males carrying the Ndn null mutation to wild type C57BL/6 females (Tg92pat- Ndn +/-pat) or transgenic females to males carrying the Ndn null mutation (TgBAC92mat- Ndn +/-pat). DNase I hypersensitivity mapping Tissues were dissected either from inter-specific F1 hybrids between M. musculus C57BL/6 females and M. musculus castaneus males (B6 x CAST/Ei) F1 or from inbred C56BL/6 mice (B6). Nuclei were isolated from adult mouse tissues as described [ 27 ]. Nuclei were resuspended in DNase I digestion buffer (0.3 M sucrose, 60 mM KCl, 15 mM NaCl, 5 mM MgCl 2 , 0.1 mM EGTA, 15 mM Tris-HCl pH 7.5, 1 mM DTT, 0.2 mM PMSF, 0.4 mM CaCl 2 , 5% glycerol) at a concentration of 5.10 7 nuclei/ml. Increasing amounts of DNase I (bovine pancreas grade I; Boehringer Mannheim) from 0 to 100 units diluted in 5 μl of DNase I digestion buffer were added to 95 μl aliquots of nuclei. After a 1 min incubation at 25°C, the digestion was stopped by addition 100 μl of stop solution (20 mM EDTA, 1% SDS, 0.1 mg/ml proteinase K) and samples were incubated overnight at 37°C. Genomic DNA was purified by multiple phenol/chloroform extractions and precipitated with ethanol. After resuspension, the DNA was digested with the appropriate restriction enzymes, separeted by gel electrophoresis, and transferred to Hybond N+ membranes. The membranes were hybridized by random hexamer- radiolabeled probes, in church solution (0.5 M NaPi pH 5.5, 7% SDS), at 65°C. Filters were washed three times in 0.2X SSC, 0.1% SDS at 65°C and exposed to X-ray film at -70°C. When necessary, blots were stripped by incubation in 0.1X SSC, O.1% SDS at 95°C. Most probes were prepared by PCR amplification of BAC109 sequences, using the following primers: Probe 1: S-5'-AGATCTGAAGACATAATG-3', AS-5'-GCTCTCCATTTCTAT TAGGTC-3'; Probe 2:S-5'-ATAAGTATTTGGTACTTTCAC-3', AS-5'-TGCTAAGTGCCTACACTGAG3'; Probe 3: S-5'-GAGCGAAACTATTCTGACAG3', AS-5'-AAGCTTCCTCCTCTATGGCAA-3'; Probe 4: S-5'-GATTTCTGCTAAGATTGG-3', AS-5'-ATGTTCCCTCTAGAAACC-3'; Probe 6: S-5'-AGTTAGAGACAAGCCTAG-3', AS-5'-TTCTGGGATGTCTCAGGA-3'; Probe 7: S-5'-GCATTTTGAGGAAGTACCCA-3', AS-5'-CATGGCCATTTCTAACTGTG-3'; Probe 8: S-5'-CCAAGGAGCTTGGAGGGC-3', AS-5'-CTCGTAGAGTGCGGCCAA-3'; probe 9: S-5'-ACATCAATAGTTTGATAC-3', AS-5'-GGGTGTGGCTGTGCATTGTT-3' Authors' contributions FW drafted the manuscript, designed the study, and carried out the PWS genes expression analysis, the transgenic analysis and the search for transcriptional regulatory elements. ELM contributed to the PWS gene expression analysis. NR carried out molecular studies on the transgenic lines and participated to the search for DNase I hypersensitive sites. LD was involved in the design of the transgenic studies. MR injected the transgene and established the transgenic lines. FM coordinated the project and participated to the manuscript draft.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545062.xml
524489	Vaccine candidates derived from a novel infectious cDNA clone of an American genotype dengue virus type 2	Background A dengue virus type 2 (DEN-2 Tonga/74) isolated from a 1974 epidemic was characterized by mild illness and belongs to the American genotype of DEN-2 viruses. To prepare a vaccine candidate, a previously described 30 nucleotide deletion (Δ30) in the 3' untranslated region of DEN-4 has been engineered into the DEN-2 isolate. Methods A full-length cDNA clone was generated from the DEN-2 virus and used to produce recombinant DEN-2 (rDEN-2) and rDEN2Δ30. Viruses were evaluated for replication in SCID mice transplanted with human hepatoma cells (SCID-HuH-7 mice), in mosquitoes, and in rhesus monkeys. Neutralizing antibody induction and protective efficacy were also assessed in rhesus monkeys. Results The rDEN2Δ30 virus was ten-fold reduced in replication in SCID-HuH-7 mice when compared to the parent virus. The rDEN-2 viruses were not infectious for Aedes mosquitoes, but both readily infected Toxorynchites mosquitoes. In rhesus monkeys, rDEN2Δ30 appeared to be slightly attenuated when compared to the parent virus as measured by duration and peak of viremia and neutralizing antibody induction. A derivative of rDEN2Δ30, designated rDEN2Δ30-4995, was generated by incorporation of a point mutation previously identified in the NS3 gene of DEN-4 and was found to be more attenuated than rDEN2Δ30 in SCID-HuH-7 mice. Conclusions The rDEN2Δ30 and rDEN2Δ30-4995 viruses can be considered for evaluation in humans and for inclusion in a tetravalent dengue vaccine.	Background The increased prevalence of disease caused by the mosquito-borne dengue (DEN) viruses (four serotypes; DEN-1 – DEN-4) has intensified the effort to generate a vaccine that would both confer protection and be economically feasible for use in countries with limited resources for healthcare [ 1 ]. Dengue fever and dengue hemorrhagic fever and shock (DHF/DSS) are a severe disease burden for tropical and semitropical countries inhabited by more than 2.5 billion people [ 2 ]. Risk factors for the more severe disease, DHF/DSS, include the strain of virus, age and genetic background of the host, and secondary infection by a DEN serotype different from that which caused the primary infection [ 2 ]. Increased risk associated with secondary infection by a different DEN serotype is believed to be caused both by increased virus replication resulting from antibody-dependent enhancement and by augmented immune activation induced by the secondary infection [ 3 , 4 ]. Typically, regions with DHF/DSS have all four DEN serotypes circulating simultaneously, and an effective DEN vaccine must contain a tetravalent formulation that confers protection against each of the four DEN serotypes. Immunity to the DEN viruses is primarily mediated by neutralizing antibodies directed against the envelope (E) glycoprotein, and most vaccine strategies aim to induce antibody against this major protective antigen. Live attenuated tetravalent vaccines appear to be the best vaccine candidates since they are economical to manufacture and they induce long-term immunity with the live attenuated yellow fever virus vaccine serving as a successful model flavivirus vaccine [ 5 ]. Several strategies to produce live attenuated tetravalent vaccines are being pursued including attenuation of viruses by conventional passage in tissue culture or introduction of defined attenuating mutations into recombinant DEN viruses [ 6 - 9 ]. In addition, chimeric dengue viruses are being evaluated that contain the E protein of a DEN virus on a background of either an attenuated DEN virus from a different serotype or a more distantly related, but attenuated, flavivirus [ 10 - 12 ]. We have previously described attenuated and immunogenic monovalent vaccine candidates for DEN-1, DEN-2, DEN-3, and DEN-4 that were generated by two distinct recombinant methodologies. Using the first methodology, nucleotides 10478–10507 were deleted from the 3' UTR (Δ30) of a wild type cDNA clone for DEN-4 to generate a vaccine candidate, rDEN4Δ30, which is safe, attenuated, and immunogenic in rhesus monkeys and humans [ 13 ]. Incorporation of the Δ30 mutation into an infectious cDNA clone of DEN-1 wild type virus at a site homologous to that in DEN-4 attenuated DEN-1 for rhesus monkeys and is currently being evaluated in humans [ 14 ]. The Δ30 mutation did not confer attenuation upon DEN-3 for reasons that have not been defined [ 15 ]. Thus, this approach has yielded live attenuated virus vaccine candidates for both DEN-1 and DEN-4. Using a second methodology, antigenic chimeric viruses have been generated by replacing the membrane protein (M) and E structural genes of rDEN4Δ30 with those from DEN-2 or DEN-3 [ 12 , 15 ]. These antigenic chimeric viruses were attenuated and immunogenic in rhesus monkeys and represent vaccine candidates for DEN-2 and DEN-3. We have also described a set of point mutations that can attenuate wild type rDEN-4 for SCID mice transplanted with human liver cells (SCID-HuH-7) or for rhesus monkeys [ 16 , 17 ]. Such mutations identified in rDEN-4 could be introduced into conserved sites of cDNA clones for other DEN serotypes to fine-tune the level of attenuation of vaccine candidates. We have found it prudent to pursue several strategies to develop a live attenuated virus vaccine for each dengue serotype recognizing that it has been a challenge to achieve a satisfactory balance between attenuation and immunogenicity [ 15 , 18 - 20 ]. Thus, in addition to the antigenic chimeric DEN-2 vaccine candidate described above, a second approach was pursued in the present study that involved the construction of an infectious cDNA clone of a wild type DEN-2 virus isolated in Tonga [ 21 ], and the generation of DEN-2 vaccine candidates by the sequential introduction of defined attenuating mutations into the recombinant version of the DEN-2 Tonga/74 wild type virus. The rDEN2Δ30 vaccine candidate was evaluated for replication in SCID-HuH-7 mice, mosquitoes, and rhesus monkeys. In addition, an attenuating point mutation, previously described in DEN-4, was introduced into the rDEN2Δ30 virus, and this rDEN2Δ30 derivative was characterized in SCID-HuH-7 mice. Methods Cells and viruses Vero cells (African green monkey kidney) were propagated in OptiPro SFM (Invitrogen, Grand Island, NY) supplemented with 4 mM L-glutamine (Invitrogen). HuH-7 cells (human hepatoma) were maintained in D-MEM/F-12 (Invitrogen) supplemented with 10% fetal bovine serum (FBS), 1 mM L-glutamine and 0.05 mg/ml gentamicin (Invitrogen). C6/36 cells (Aedes albopictus mosquito cells) were maintained at 32°C in Minimal Essential Medium (MEM) containing Earle's salts and 25 mM HEPES buffer (Invitrogen) and supplemented with 10% FBS, 2 mM L-glutamine, and 0.1 mM non-essential amino acids (Invitrogen). A dengue virus type 2 isolate, Tonga/74, was provided by Dr. Duane Gubler (CDC, Fort Collins, CO). The virus was isolated during a 1974 dengue outbreak in the South Pacific island of Tonga [ 21 ]. The virus was isolated by inoculation of patient sera into Aedes albopictus mosquitoes, and subsequent passage in C6/36 cells before determination of genomic sequence. Sequence analysis Viral RNA was isolated from DEN-2 Tonga/74 wild type virus using the QIAamp Viral RNA mini kit (Qiagen, Valencia, CA). Reverse transcription was performed using random hexamer primers and the SuperScript First-Strand Synthesis System for RT-PCR (Invitrogen). Overlapping PCR fragments of approximately 2000 base pairs were generated using DEN-2 specific primers and Advantage cDNA polymerase (ClonTech, Palo Alto, CA). Both strands of the resulting PCR fragments were sequenced directly on a 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA) using DEN-2 specific primers in BigDye terminator cycle sequencing reactions (Applied Biosystems) and the results were assembled into a consensus sequence. To determine the nucleotide sequence of the genomic 5' and 3' regions, the 5' cap nucleoside of the viral genome was removed with tobacco acid pyrophosphatase (Epicentre Technologies, Madison, WI), followed by circularization of the genome using RNA ligase (Epicentre Technologies). An RT-PCR fragment spanning the ligation junction was generated and sequenced using DEN-2 primers. For the DEN-2 Tonga/74 consensus sequence, GenBank accession number AY744147 was assigned. Genetic construction of rDEN-2 Tonga/74 cDNA clone cDNA fragments of DEN-2 Tonga/74 were generated by reverse-transcription of the genome as indicated in Figure 1 . Each fragment was subcloned into a plasmid vector and sequenced to verify that it matched the consensus sequence as determined for the virus. This yielded seven cloned cDNA fragments spanning the genome. Cloned fragments were modified as follows: Fragment X, representing the 5' end of the genome was abutted to the SP6 promoter; Fragment L was modified to contain a Spe I restriction site at genomic nucleotide 2353; Fragment R was modified to contain a Spe I restriction site also at genomic nucleotide 2353, and, to stabilize the eventual full-length clone, two additional mutations at nucleotides 2362 – 2364 and 2397 were created to ensure that translation stop codons were present in all reading frames other than that used to synthesize the virus polyprotein; Fragment A was modified at nucleotide 3582 to ablate a naturally occurring Spe I restriction site and at nucleotide 4497 to ablate a naturally occurring Kpn I restriction site; Fragment C was modified at nucleotide 9374 to ablate a naturally occurring Kpn I restriction site; and Fragment Y, representing the 3' end of the genome was abutted to a Kpn I restriction site. All mutations introduced into the cloned cDNA fragments were translationally-silent, thereby preserving the wild-type polyprotein sequence. Each fragment was added incrementally between the Asc I and Kpn I restriction sites of DEN-4 cDNA clone p4 (GenBank accession number: AY648301) to generate a full-length DEN-2 cDNA clone (p2) with the same vector background successfully used to generate rDEN-4 and rDEN4Δ30 virus [ 13 ]. cDNA clone p2 was sequenced to confirm that the virus genome region matched the DEN-2 Tonga/74 consensus amino acid and nucleotide sequence, with the exception of the translationally-silent modifications noted above. The Δ30 mutation which removes nucleotides 10541–10570 was introduced into Fragment Y to generate Fragment YΔ30. To create p2Δ30, the Fragment Y region of p2 was replaced with Fragment YΔ30 (Figure 1 ). The genomic region of each full-length cDNA was sequenced as described above and GenBank accessions were assigned as follows (cDNA clone: accession numbers): p2: AY744148, p2Δ30: AY744149. Using site-directed mutagenesis, an attenuating amino acid change characterized in the NS3 gene of DEN-4 (nt 4995–7; a.a. 158, Ser→Leu) was introduced into the p2Δ30 cDNA clone [ 17 ]. A mutagenic oligonucleotide was designed to change DEN-2 NS3 amino acid 158 from Ser (AGT) to Leu (CTA) and used to construct the cDNA clone, p2Δ30-4995 (accession number: AY744150), which was sequenced for confirmation of nucleotide changes. Recovery of rDEN-2 viruses cDNA clones were linearized with Acc 65I (isoschizomer of Kpn I which cleaves leaving only a single 3' nucleotide) and were transcribed in vitro using the AmpliCap SP6 Message Maker kit (Epicentre Technologies, Madison, WI). Purified transcripts were then transfected into Vero or C6/36 cells. Viruses recovered in C6/36 cells were passaged 3 times in Vero cells, and all viruses were biologically cloned by terminal dilution in Vero cells. The genomes of recombinant viruses used to infect rhesus monkeys were completely sequenced as described above to identify adventitious mutations that had accumulated during transfection and biological cloning. Replication in SCID-HuH-7 mice Four to six week-old SCID mice (Tac:Icr:Ha(ICR)- Prkdc scid ) (Taconic, Germantown, NY) were injected intraperitoneally with 10 7 HuH-7 cells suspended in 0.2 ml phosphate-buffered saline. Tumors were detected in the peritoneum, and mice were infected by direct inoculation of the tumor with 10 4 PFU of virus in 0.05 ml Opti-MEM (Invitrogen). On day 7 post-infection, serum was obtained from cardiac blood and stored at -70°C. Virus titer in serum samples was determined by plaque assay in Vero cells. Replication, immunogenicity, and protection in rhesus monkeys The DEN-2 viruses were evaluated in rhesus macaques using established methods [ 13 ]. DEN virus sero-negative monkeys were injected subcutaneously with 10 5 PFU virus diluted in L-15 medium (Invitrogen) or with a mock inoculum. Serum was collected on days 0–6, 8, 10, 12 and 28 after inoculation and stored at -70°C. Virus titer was determined for each serum sample by plaque assay in Vero cells and serum neutralizing antibody titer was determined for serum from days 0 and 28 by plaque reduction neutralization test. On day 28, monkeys were challenged with 10 5 PFU of DEN-2 Tonga/74, and serum was collected on days 29–34, 36, and 56. Virus titer was determined for serum from days 28–34 and 36 and serum neutralizing antibody titer was determined for serum from day 56. Virus replication in mosquitoes Replication in Aedes aegypti and Toxorynchites amboinensis mosquitoes was evaluated as previously described [ 22 ]. Briefly, A. aegypti were fed blood meals containing serial 10-fold dilutions of virus. After 21 days, viral antigen was detected in head and midgut preparations by immunoflourescence assay using DEN-2-specific hyperimmune mouse ascitic fluid and fluorescein isothyocyanate conjugated goat anti-mouse IgG (KPL, Gaithersburg, MD), and the mosquito infectious dose-50% (MID 50 ) was determined. T. amboinensis were inoculated intrathoracically with a 0.2 ul dose containing serial ten-fold dilutions of virus and incubated for 14 days. Head preparations were made and antigen visualized as described above. Results Generation and sequence analysis of recombinant DEN-2 Tonga/74 viruses A full-length cDNA clone, p2, was constructed that matched the genomic consensus sequence of the American genotype DEN-2 isolate, Tonga/74, with the exception of translationally-silent modifications made to facilitate cloning (Figure 1 ). The previously described Δ30 deletion mutation was incorporated into the p2 cDNA clone to form p2Δ30 [ 13 ]. The rDEN-2 virus was recovered in C6/36 and Vero cells, but the presence of the Δ30 mutation limited recovery to only C6/36 cells. After passage in Vero cells, adaptation mutations were identified by sequence analysis as had been described for other DEN viruses [ 23 ]. Both rDEN-2 and rDEN2Δ30 viruses accumulated a single nucleotide change in NS4B at nt 7169 encoding a Val→Ala change at amino acid position 115 as has been observed for rDEN-3 (Table 1 ) [ 15 ]. The same nucleotide change was previously reported to occur at the homologous site following passage of DEN-4 in Vero cells resulting in a Leu→Ser change (Table 1 ) [ 23 ]. Inclusion of the 7169 mutation into the p2Δ30 cDNA permitted recovery in both C6/36 and Vero cells (data not shown). The rDEN2Δ30 virus reached a virus titer of 6.6 log 10 PFU/ml in Vero cells. Replication of rDEN-2 viruses in SCID-HuH-7 mice As an initial evaluation of replication of the DEN-2 Tonga/74 virus and the rDEN-2 viruses, replication in SCID mice transplanted with HuH-7 human hepatoma cells (SCID-HuH-7 mice) was tested. Wild-type viruses from each DEN serotype have been shown to replicate to approximately 6.0 log 10 PFU/ml serum in SCID-HuH-7 mice, and an att phenotype in SCID-HuH-7 mice has been shown to be a predictor of reduced replication in rhesus monkeys [ 12 , 14 , 15 , 17 ]. The parent DEN-2 Tonga/74 virus replicated efficiently in SCID-HuH-7 mice and reached a mean titer in serum of 5.9 log 10 PFU/ml (Table 2 ) similar to that previously observed with the DEN-2 New Guinea C (NGC) prototype strain [ 12 ]. The rDEN-2 virus replicated to the same level as the wild-type isolate, while rDEN2Δ30 was 10-fold restricted in replication. This reduction was statistically significant (Tukey-Kramer post-hoc test; P < 0.05), and was similar to that observed for the well-characterized rDEN4Δ30 virus [ 17 ]. Replication of rDEN-2 viruses in mosquitoes The DEN-2 viruses were evaluated for infectivity of Aedes aegypti fed on an infectious bloodmeal (oral infectivity only) and for Toxorynchites amboinensis inoculated intrathoracically (Table 3 ). At the doses tested neither DEN-2, rDEN-2, or rDEN2Δ30 were detected in the midgut or head of A. aegypti mosquitoes which had fed on an infectious bloodmeal 21 days earlier. The inability to infect the midgut led to a lack of infection in the head tissue. This indicates that the DEN-2 Tonga/74 viruses are poorly infectious for A. aegypti mosquitoes by oral infectivity, as has been demonstrated for multiple DEN-2 viruses of the American genotype [ 24 , 25 ]. In contrast the DEN-2 NGC prototype strain, an Asian genotype member, was highly infectious in A. aegypti mosquitoes when tested previously but it was not included here as a concurrent control [ 12 ]. The defect in rDEN-2 infectivity for A. aegypti was further investigated by directly inoculating the same virus stocks intrathoracically into T. amboinensis and measuring the ability of the viruses to infect the head tissues. Both rDEN-2 and rDEN2Δ30 were highly infectious by intrathoracic inoculation (Table 3 ). The Δ30 mutation did not alter the infectivity of rDEN-2 following intrathoracic inoculation, a property also previously observed for DEN-1, -3 and -4 [ 14 , 15 , 22 ]. These results indicate that the lack of infectivity for A. aegypti was likely caused by the inability of the DEN-2 Tonga/74 viruses to establish a midgut infection and that the viruses retained the ability to infect head tissues. Replication, immunogenicity, and protective efficacy in rhesus monkeys The replication (viremia), immunogenicity, and protective efficacy of the DEN-2 viruses in monkeys were studied. Monkeys inoculated with the DEN-2 Tonga/74 wild-type isolate were viremic for an average of 4.5 days with a mean peak titer of 2.1 log 10 PFU/ml (Table 4 ). Inoculation with rDEN-2 resulted in detectable viremia for 4.0 days with a mean peak titer of 1.9 log 10 PFU/ml. While the levels of rDEN2Δ30 replication (2.8 days viremia; mean peak titer of 1.7 log 10 PFU/ml) were lower than DEN-2 and rDEN-2, the differences were not as dramatic as had been observed for rDEN1Δ30 and rDEN4Δ30 when compared to their parent viruses [ 13 , 14 ]. The level of neutralizing antibodies induced by the rDEN2Δ30 virus was also less than that induced by the wild-type DEN-2 viruses, a finding consistent with the decreased replication exhibited by this vaccine candidate. Therefore, by three quantitative measures, duration and peak titer of viremia and the level of neutralizing antibodies induced, rDEN2Δ30 appeared to be attenuated when compared to DEN-2 Tonga/74. When vaccinated monkeys were challenged with DEN-2 Tonga/74, all monkeys were protected, as indicated by the lack of viremia (Table 4 ). The 4995 mutation further attenuates rDEN2Δ30 in SCID-HuH-7 mice Based on the limited attenuation conferred upon rDEN-2 by the Δ30 mutation in rhesus monkeys, we sought to construct a further attenuated derivative of rDEN2Δ30. To further attenuate rDEN2Δ30, an att mutation that has been characterized in another DEN serotype was imported into a homologous region in DEN-2. One such mutation, the 4995 mutation in DEN-4 NS3 at amino acid 158 (Ser→Leu), was previously incorporated into the DEN-4 vaccine candidate, rDEN4Δ30, and found to further attenuate the virus for SCID-HuH-7 mice and rhesus monkeys [ 17 ]. Site directed mutagenesis was used to introduce a Ser→Leu mutation at amino acid 158 of NS3 in rDEN2Δ30-7169, and the rDEN2Δ30-4995 virus was recovered in C6/36 cells and propagated in Vero cells reaching a virus titer of 6.2 log 10 PFU/ml (Table 1 ). Importantly, the resulting Leu codon would require two nucleotide changes to revert to one of the six odons encoding a Ser residue. Replication in the SCID-HuH-7 mouse model was used as an initial assessment of the rDEN2Δ30-4995 virus phenotype. Table 5 includes results from three separate experiments (including those from Table 1 ) and confirms the approximate 10-fold reduction in replication conferred by the Δ30 mutation upon rDEN-2 replication in SCID-HuH-7 mice. The rDEN2Δ30-4995 virus had a mean peak virus titer of 4.6 log 10 PFU/ml which was only a modest reduction from that of rDEN2Δ30, 5.2 log 10 PFU/ml. However, comparison of a large number of samples indicated that the reduction in virus titer conferred by the NS3 4995 mutation upon rDEN2Δ30 was statistically significant (rDEN2Δ30-4995 versus rDEN2Δ30; Tukey-Kramer post-hoc test; P < 0.05). The virus titer of rDEN2Δ30-4995 virus in SCID-HuH-7 mice was over 60-fold reduced from that of the rDEN-2 parent virus. Discussion Development of a live-attenuated tetravalent dengue vaccine has been complicated by two major factors. First, monovalent vaccine candidates that exhibit a satisfactory balance between attenuation and immunogenicity have been difficult to identify [ 15 , 18 - 20 ]. Second, satisfactorily attenuated tetravalent vaccine formulations that induce a broad neutralizing antibody response against each of the four DEN serotypes have been difficult to develop [ 6 , 10 , 20 , 26 ]. For these reasons, we have sought to develop multiple vaccine candidates for each DEN serotype to increase the likelihood that a vaccine with a satisfactory balance between attenuation and immunogenicity will be identified. To produce a live-attenuated DEN-2 vaccine candidate, we previously generated an antigenic chimeric virus, rDEN2/4Δ30, expressing the M and E structural genes of the DEN-2 NGC strain on the attenuated rDEN4Δ30 background [ 12 ]. The vaccine candidates described in the present study, rDEN2Δ30 and rDEN2Δ30-4995, could serve as alternates to this antigenic chimeric virus if evaluation of the rDEN2/4Δ30 virus in humans, either as a monovalent vaccine or as a component of a tetravalent vaccine, indicates that it lacks a balance between attenuation and immunogenicity. It was hoped that each of the four components of a tetravalent vaccine, consisting of DEN-1, -2, -3, and -4 wild type viruses, each with the common 30 nucleotide deletion mutation in the 3' UTR, would exhibit a similar level of attenuation in animal models [ 13 - 15 ]. Unfortunately, the level of attenuation conferred by the Δ30 mutation upon each of the four serotypes has proven to be variable. In rhesus monkeys, the rDEN2Δ30 virus appears to have an intermediate attenuation phenotype in between that of the attenuated rDEN1Δ30 and rDEN4Δ30 and the non-attenuated rDEN3Δ30 [ 13 - 15 ]. Although rDEN2Δ30 was slightly attenuated compared to its DEN-2 parent virus in rhesus monkeys, the reduction in replication was less than that of rDEN1Δ30 and rDEN4Δ30. While the latter two viruses had detectable viremia in only 50% of monkeys, a mean number of viremic days of less than one day, and a mean peak viremia of less than 1.0 log 10 PFU/ml [ 13 , 14 ], the rDEN2Δ30 virus infected 100% of the rhesus monkeys and reached a peak virus titer of 1.7 log 10 PFU/ml. However, the 10-fold reduction of replication of rDEN4Δ30 and rDEN2Δ30 in SCID-HuH-7 mice, compared to that of their respective wild type parents, was similar. To date, rDEN4Δ30 is the only Δ30 vaccine candidate that has been tested in humans, and it was found to be both safe and immunogenic [ 13 ]. If the level of attenuation in SCID-HuH-7 mice serves as a better guide to attenuation in humans, rDEN2Δ30 might be satisfactorily attenuated in humans since its level of attenuation for SCID-HuH-7 mice and that of the rDEN4Δ30 vaccine candidate are comparable. To construct a further attenuated derivative of rDEN2Δ30, the 4995 mutation present in the NS3 gene of DEN-4 at amino acid 158 (Ser→Leu) was introduced into the homologous region of the NS3 protein of rDEN2Δ30 [ 17 ]. Although the 4995 mutation results in a single amino acid change and thus may be susceptible to reversion, the mutant leucine codon selected for insertion into rDEN2Δ30-4995 would require two nucleotide changes to revert to a serine codon. Introduction of the 4995 mutation into rDEN4Δ30 resulted in a 100-fold greater reduction of replication in SCID-HuH-7 mice [ 17 ]. In rDEN2Δ30, its introduction resulted in nearly a 10-fold reduction in virus titer, a smaller but still statistically significant reduction. These results provide a second example of the difficulty in predicting the precise level of attenuation following import of an attenuating mutation into a different DEN serotype. Nevertheless, the rDEN2Δ30-4995 vaccine candidate is more attenuated than its rDEN2Δ30 parent and warrants evaluation in rhesus monkeys and humans. Epidemiologic and molecular pathogenesis studies of DEN-2 strains support the concept that the DEN-2 Tonga/74 virus, from which the vaccine candidates were derived, may naturally have a lower level of virulence than other DEN-2 viruses. If the DEN-2 Tonga/74 parent virus is naturally attenuated to some degree, only a small incremental increase in attenuation might be required to satisfactorily attenuate it for humans. Gubler et al. investigated the 1974 outbreak of DEN-2 infection in the Pacific island of Tonga [ 21 ]. In comparison to a subsequent DEN-1 outbreak, the 1974 DEN-2 outbreak was distinguished by mild disease with few hemorrhagic sequelae, low viremia, and an overall slow spread of virus infection [ 21 ]. The weak DEN-2 outbreak was proposed to be a result of the circulation of a strain with an inherently low level of virulence [ 21 ]. Since the Tonga/74 outbreak, additional evidence has emerged that supports the suggestion that there are at least two circulating lineages of DEN-2 viruses that differ in virulence [ 27 - 29 ]. The DEN-2 Tonga/74 virus is a member of the DEN-2 American genotype, which as a group appear to possess lower virulence than that of the Asian genotype of DEN-2 viruses [ 28 , 29 ]. Despite the presence of the American DEN-2 genotype viruses and limited co-circulation of DEN-1 and DEN-3 viruses in the Americas in the 1960s and 1970s, the first major epidemic of DHF/DSS in the Americas occurred only after the introduction of a DEN-2 Asian genotype virus in 1981 [ 27 - 30 ]. It was thought that genetic differences might have contributed to this difference in virulence and evidence to this effect has been forthcoming. Rico-Hesse and colleagues have defined genetic elements within the genome of DEN-2 American genotype viruses which distinguish them from members of the Asian genotype [ 31 ]. In addition, using chimeric rDEN-2 American/Asian viruses, introduction of three genetic elements (a point mutation in the E gene, the 5' UTR, and the 3' UTR) of the American genotype was found to confer reduced virus replication in dendritic cells and monocytes upon an Asian genotype rDEN-2 [ 32 ]. The Tonga/74 virus shares each of these three attenuating genetic determinants specific to the American genotype [ 31 , 32 ], which provides a possible explanation for its lower virulence in humans. The rDEN2Δ30 vaccine candidate, whose parent is the DEN-2 Tonga/74 American genotype virus, thus contains naturally occurring and experimentally introduced attenuating mutations. Thus, the small incremental increase in attenuation provided by the Δ30, with or without the 4995 mutation, might prove to satisfactorily attenuate the DEN-2 Tonga/74 for humans. The American genotype DEN-2 viruses exhibit decreased infectivity for Aedes mosquitoes in comparison to Asian DEN-2 viruses [ 24 , 25 ]. Consistent with these observations, the wild type New Guinea C Asian DEN2 virus was highly infectious for Aedes mosquitoes in our laboratory [ 12 ] whereas the Tonga/74 American genotype virus was poorly infectious by the oral route (present study). In fact, the increased prevalence of DEN-2 viruses of the Asian genotype in the Americas has been suggested to be a result of their enhanced transmission [ 28 ]. However, the active circulation of American genotype viruses over many decades indicates that mosquito transmission does occur and large epidemics have been associated with viruses of this genotype [ 21 , 27 ]. At the doses tested, neither the DEN-2 Tonga/74 isolate nor the recombinant viruses were found to infect the midgut or head of Aedes aegypti mosquitoes fed an infectious blood meal. Since rDEN-2 and rDEN2Δ30 viruses were infectious by intrathoracic inoculation of Toxorynchites mosquitoes, the lack of infectivity for A. aegypti was likely caused solely by the inability of the DEN-2 Tonga/74 viruses to establish a midgut infection. Decreased infectivity for Aedes mosquitoes could serve to help limit transmission of the vaccine virus. Conclusions The live-attenuated DEN-2 virus candidates described here, rDEN2Δ30 and rDEN2Δ30-4995, have several properties desired in a live attenuated virus vaccine for humans. First, both viruses reached a titer over 6.0 log 10 PFU/ml in Vero cells that would permit economical manufacture. Second, the viruses are derived from the DEN-2 Tonga/74 strain, a member of the American genotype, which has been associated with decreased virulence. Third, rDEN2Δ30 was attenuated for replication in SCID-HuH-7 mice and slightly attenuated for rhesus monkeys while inducing a protective neutralizing antibody response. Fourth, rDEN2Δ30-4995 was more attenuated in SCID-HuH-7 mice than rDEN2Δ30. Fifth, the DEN-2 Tonga/74 strain, like other members of the American genotype, is poorly infectious for Aedes aegypti mosquitoes which would help to limit uncontrolled transmission of the vaccine virus. Competing interests The authors declare that they have no competing interests. Authors' contributions J.B. recovered viruses, conducted animal studies, and drafted the manuscript. C.H. and S.W. constructed the DEN-2 cDNA clone and C.H. performed sequencing. K.H. performed mosquito studies. B.M. and S.W. supervised the study and participated in planning and design. All authors read and approved the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524489.xml
520822	Proinflammatory role of inducible nitric oxide synthase in acute hyperoxic lung injury	Background Hyperoxic exposures are often found in clinical settings of respiratory insufficient patients, although oxygen therapy (>50% O 2 ) can result in the development of acute hyperoxic lung injury within a few days. Upon hyperoxic exposure, the inducible nitric oxide synthase (iNOS) is activated by a variety of proinflammatory cytokines both in vitro and in vivo . In the present study, we used a murine hyperoxic model to evaluate the effects of iNOS deficiency on the inflammatory response. Methods Wild-type and iNOS-deficient mice were exposed to normoxia, 60% O 2 or >95% O 2 for 72 h. Results Exposure to >95% O 2 resulted in an increased iNOS mRNA and protein expression in the lungs from wild-type mice. No significant effects of iNOS deficiency on cell differential in bronchoalveolar lavage fluid were observed. However, hyperoxia induced a significant increase in total cell count, protein concentration, LDH activity, lipid peroxidation, and TNF-α concentration in the bronchoalveolar lavage fluid compared to iNOS knockout mice. Moreover, binding activity of NF-κB and AP-1 appeared to be higher in wild-type than in iNOS-deficient mice. Conclusion Taken together, our results provide evidence to suggest that iNOS plays a proinflammatory role in acute hyperoxic lung injury.	Background Supplemental oxygen therapy is administered for the treatment of tissue hypoxia, most commonly in an intensive care setting of respiratory insufficient patients, though its potent toxicity is well described [ 1 ]. The pathophysiology of oxygen injury is characterized by lung inflammation including activation and recruitment of neutrophils and alveolar macrophages, tissue and alveolar edema, surfactant dysfunction, and excess production of free radicals and inflammatory cytokines [ 2 - 4 ]. Although the exact mechanisms of pulmonary oxygen toxicity are still unknown, compelling evidence suggests that reactive oxygen species such as superoxide anion, hydroxyl radical, and hydrogen peroxide are important mediators of lung injury [ 5 - 7 ]. High oxygen concentrations induce cellular damage by several mechanisms such as oxidation of proteins, peroxidation of membrane lipids, and breakage of DNA strands [ 8 - 10 ]. Moreover, hyperoxia also induces the release of a wide spectrum of proinflammatory cytokines such as tumor necrosis factor- [ 11 - 13 ]. However, the precise molecular mechanisms by which hyperoxia produces acute lung injury remain unclear. Reactive oxygen species can also react with other free radicals such as nitric oxide (NO) to yield more cytotoxic species including peroxynitrite anion [ 14 , 15 ]. Peroxynitrite is a strong oxidizing agent that can also initiate lipid peroxidation [ 16 , 17 ]. Since the discovery of NO as a potent vascular smooth muscle relaxant and regulator of blood pressure, NO generated by the inducible nitric oxide synthase (iNOS) has been identified in many cell types such as alveolar macrophages or epithelial cells and implicated in a variety of biological roles [ 15 , 18 , 19 ]. NO is synthesized from L-arginine by two main isoforms of the NO synthase: the constitutive and the inducible isoform [ 20 ]. The NOS enzymes are complex homodimeric proteins consisting of a N-terminal oxygenase domain, a central calmodulin binding sequence, and a C-terminal reductase domain [ 21 , 22 ]. Inducible NOS is expressed following induction by a variety of inflammatory cytokines such as TNF-α [ 22 ] or by lipopolysaccharide (LPS) [ 23 - 25 ]. Constitutive iNOS expression has been reported in the lung [ 26 , 27 ] and several inflammatory processes involving the lung, such as sepsis [ 23 - 25 ], asbestosis-induced lung injury [ 28 , 29 ] and hyperoxia [ 30 , 31 ] are associated with an elevated NO production. However, the effect of hyperoxia on endogenous NO production is a matter of controversial discussion, depending on the experimental conditions [ 30 , 32 - 34 ]. Inappropriate regulation of nuclear factor-κB (NF-κB)-and activator protein-1 (AP-1)-mediated transcription has also been associated with pathological conditions, including acute inflammation such as hyperoxic exposure [ 35 ]. Intracellular signaling pathways leading to an activation of transcriptional regulators such as NF-κB and AP-1 can be affected by reactive oxygen and nitrogen species [ 36 - 38 ]. Both transcription factors are activated in lung cells after short periods of hyperoxic exposure. Binding sites for NF-κB and AP-1 are present in the promotor of the iNOS gene and of proinflammatory cytokines such as TNF-α [ 35 , 39 , 40 ]. The objective of this study was to investigate the effect of iNOS deficiency on acute hyperoxic lung injury. As indicators of lung hyperpermeability, lavageable lung protein and LDH activity were measured; lung lipid oxidation was assessed based on the levels of thiobarbituric acid reactive substances. To characterize the inflammatory process, lavageable cell counts, cell differential, and TNF- concentration were analyzed. Binding activity of NF-κB and AP-1 was investigated in order to elucidate transcriptional mechanisms for iNOS and TNF-α expression. Methods Animals Inducible NOS-deficient mice (C57BL6/J × 129SvEv) were originally provided by Dr. J. Mudgett (Merck & Co., Rathway, New Jersey, USA), Dr. J. MacMicking, and Dr. C. Nathan (Cornell University Medical College, New York, USA) [ 41 ]. As controls, matching wild-type mice were used (C57BL6/J × 129SvEv). Animals were bred in the facilities of the Institute for Surgical Research (Munich, Germany). Protocols used in this study were approved by the appropriate government body. Hyperoxic exposure Male mice (12 – 16 weeks old, body weight between 26.1 g and 27.3 g) were kept in groups of seven in a sealed Plexiglas chamber (27 × 27 × 20 cm 3 ). Animals were randomized and exposed to 60% O 2 and >95% O 2 with a gas flow rate of 6 l/min at atmospheric pressure for 72 h. Mice exposed to room air in the same chamber served as controls. O 2 levels were monitored twice daily with an oxygen analyzer (Drägerwerk AG, Lübeck, Germany). The environmental temperature was maintained at 24°C ± 1, relative humidity was 73% ± 13, and air pressure was 947 mbar ± 5. Oxygen was humidified by bubbling through a water chamber. The Plexiglas chamber bottom was lined with soda lime for CO 2 absorption (Mallinkrodt Baker B. V., Deventer, Holland). Exposures were continuous for the time indicated except for 5 – 10 min daily when the chamber was opened for housekeeping purposes. Animals were kept on a 12 h light/dark cycle. Standard rodent food and water were available ad libitum . Bronchoalveolar lavage cell counts and cell differential Immediately following exposure, mice were anaesthetized by intraperitoneal injection of sodium pentobarbital (10 mg/kg body weight, Narcoren ® , Merial, Halbergmoos, Germany). Tracheotomy was performed and a 20 G × 32 mm needle (Abbocath ® -T, Venisystems, Sligo, Ireland) was inserted and secured. Bronchoalveolar lavage (BAL) was performed five times with 1 ml of sterile non-pyrogenic phosphate-buffered saline solution (PBS; Serva, Heidelberg, Germany) in each animal. After centrifugation at 300 × g for 10 min, the supernatant was collected and stored at -20°C and -80°C for later protein assays. The BAL cell pellet was resuspended in PBS and washed by centrifugation. Cells were stained with May-Grunwald-Giemsa (Varistain 4, Shandon Labortechnik GmbH, Frankfurt, Germany) to identify cellular populations. Total cell counts were assessed with a hemacytometer (Coulter Ac T 8, Coulter Electronics, Krefeld, Germany). Lavageable lung protein assay Cell free BAL fluid was evaluated for total protein content by the bicinchoninic acid assay using bovine serum albumin (PAA Laboratories, Linz, Austria) based on a method of Smith et al. [ 42 ]. Lactate dehydrogenase activity assay To evaluate lactate dehydrogenase (LDH) activity in cell free BAL fluids, a commercially available kit was used (LDH Optimiert, Roche Diagnostics, Mannheim, Germany). TNF-α assay Concentration of TNF-α in cell free BAL fluid was measured by an enzyme linked immunosorbent assay using a commercially available kit (EM-TNFA, Endogen, Woburn, Massachusetts, USA). Briefly, 50 ml biotinylated antibody reagent were added to 50 ml-samples in an anti-mouse TNF-α pre-coated strip well plate. After incubation for 2 h at room temperature, the plate was washed, a streptavidin horseradish peroxidase solution and the 3, 3',5, 5'-tetramethylbenzidine substrate solution were added and incubated in the dark. The absorbance was detected at 450 nm in a microplate reader (EAR 400 AT, Salzburger Labortechnik, Salzburg, Austria). A standard curve was used to determine the amount of TNF-α concentration in the samples. Reverse transcriptase-polymerase chain reaction Total RNA was isolated from non-lavaged lung homogenate of each mouse (RNeasy Mini Kit, Quiagen, Hilden, Germany), reverse transcribed into cDNA in a volume of 20 ml, containing 2 μg RNA, 1.5 μM Oligo-p(dT)15-primer, 5 × PCR-buffer, 0.1 M DTT, 10 nM dNTP-mix and 200 U/μl of Moloney murine leukemia virus reverse transcriptase. Reverse transcriptase-polymerase chain reaction (RT-PCR) amplifications were performed with aliquots of cDNA (3 μl) in total volume of 50 μl (5 μl 10 × PCR reaction-buffer, 1 μl dNTP-mix, 1 μl each of forward and reverse single strand DNA primers specific for mouse iNOS, 38.8 μl sterile deionized water, 0.2 μl Taq DNA-polymerase 1 U/ml). Oligonucleotide primers for iNOS were 5'-CAC AAG GCC ACA TCG GAT TTC-3' (sense) and 5'-TGC ATA CCA CTT CAA CCC AG-3' (antisense). Co-amplification of the housekeeping gene β-actin served as an internal control, using the following primers, 5'-GGA CTC CTA TGT GGG TGA CGA GG-3' (sense), 5'-GGG AGA GCA TAG CCC TCG TAG AT-3' (antisense). RT-PCR was started with 1 min incubation at 95°C followed by the steps of denaturation at 95°C for 45 sec, annealing at 55 – 64°C for 45 sec, elongation at 72°C for 1 min. The number of cycles (30 – 35 each) was chosen to ensure that the amplification product did not reach the level of saturation. Reactions were electrophoresed in 1% agarose gel and stained with ethidium bromide. The densitometry of each cDNA band was quantified using BIO-1D.V96 software and the ratio of iNOS cDNA to β-actin cDNA was determined. Electrophoretic mobility shift assay Nuclear protein extracts were prepared from pooled lung tissue as previously described [ 43 ]. Briefly, the oligonucleotides were incubated with a binding buffer (0.04 M Tris, 0.2 M NaCl, 2 mM EDTA, 8% glycerine, 2 μm Ficoll 400, 0.2 mM PMSF, 4 mM DTT). After 5 min incubation, 5 μ l of [γ 32 P]-dATP end-labeled double-stranded oligonucleotides containing an NF-κB-consensus sequence (5'-AGT TGA GGG GAC TTT CCC AGG C-3') or AP-1-consensus sequence (5'-CGC TTG ATG AGT CAG CCG GAA-3') were added to the reaction followed by an incubation for 1 h at 37°C. The mixture was subjected to electrophoresis on a 6% PAA-Gel (75% H 2 O, 45 mM Tris, 45 mM bore acid, 1 mM EDTA, pH 8. 6% APS, 60 μl TEMED) for 2 h at 250 V. Thiobarbituric acid reactive substances assay Concentration of thiobarbituric acid reactive substances (TBARS) was evaluated with an assay according to Thiery et al. [ 44 ]. BAL fluid was prepared to denaturate proteins with 50% trichloroacetic acid. The supernatants were transferred to a clean tube and 75 μl of 1.3% thiobarbituric acid (Sigma Chemie, Deisenhofen, Germany) in 0.3% NaOH were added. After incubation for 1 h at 90°C and subsequent cooling in ice water, samples were centrifuged for 6 min. Finally, 200 μl of sample were transferred to a 96-well plate and the absorbance at 530 nm was read in a microplate reader (Dynex Technologies, Denkendorf, Germany). TBARS were quantified by using a standard curve of malondialdehyde (Sigma Chemie, Germany). Statistical analysis Results are presented as the group mean ± standard error of the mean (SEM). Statistical comparison between values of the three oxygen concentrations was performed by using analysis of variance on ranks and Mann-Whitney rank sum test followed by Bonferroni's correction. Statistical comparison between wild-type and iNOS knockout mice was analyzed by using Mann-Whitney rank sum test. Significance was accepted at p < 0.05. Results General conditions of the animals and body weight Wild-type and iNOS knockout mice all survived hyperoxia the entire 72 h. After hyperoxic exposure >95% O 2 , animals showed signs of reduced general conditions and reactions. Hyperoxic exposure >95% O 2 also caused a significant reduction in body weight of wild-type mice compared to normoxic conditions and 60% oxygen exposure within the 72 h experimental period. In contrast, there was no significant change in body weight of iNOS knockout mice before and after normoxia and hyperoxia, respectively (table 1 ). Table 1 Body weight (g) of wild-type and iNOS knockout mice after 72 h exposure to 21%, 60%, and >95% O 2 * 21% O 2 60% O 2 >95% O 2 wild-type mice before exposure 27.1 ± 1.0 26.5 ± 0.8 27.8 ± 0.9 after exposure 28.1 ± 0.9 28.1 ± 0.3 23.2 ± 0.6 # iNOS knockout mice before exposure 24.9 ± 0.7 23.3 ± 0.5 25.2 ± 0.9 after exposure 26.1 ± 0.4 24.0 ± 0.5 24.1 ± 0.7 Each value represents mean ± SEM, n = 7. # p < 0.05 vs. before exposure. Differential and total cell counts BAL in wild-type and iNOS knockout mice was performed to assess cellular infiltration in the alveolar space upon 72 h hyperoxic exposure (60% and >95% O 2 ). Results presented in table 2 demonstrate no differences in baseline cell differentials between wild-type and iNOS knockout mice. Upon 72 h exposure to >95% O 2 , there was a significant decrease in the percentage of alveolar macrophages as well as a significant increase in the percentage of neutrophils and lymphocytes in both wild-type and iNOS knockout mice compared to normoxic conditions. No significant differences between wild-type- and iNOS knockout mice were found. However, hyperoxic exposure (>95% O 2 ) resulted in a significant increase in total BAL cell counts after 72 h in wild-type (0.54 ± 0.05 × 10 6 /ml) and in iNOS knockout mice (0.38 ± 0.04 × 10 6 /ml) compared to normoxia (0.20 ± 0.03 × 10 6 /ml and 0.16 ± 0.02 × 10 6 /ml, respectively) and 60% O 2 exposure (0.24 ± 0.04 × 10 6 /ml and 0.19 ± 0.04 × 10 6 /ml, respectively). This increase in BAL total cell counts under >95% O 2 was significantly higher in wild-type than in iNOS knockout mice (figure 1 ). Table 2 BAL cell differential in wild-type and iNOS knockout mice after 72 h exposure to 21%, 60%, and >95% O 2 alveolar macrophages (%) neutrophils (%) lymphocytes (%) wild-type mice 21% O 2 98.8 ± 0.5 0.0 ± 0.0 1.2 ± 0.5 60% O 2 97.7 ± 1.0 0.3 ± 0.2 2.0 ± 1.0 >95% O 2 86.2 ± 1.8#* 4.6 ± 1.0#* 9.2 ± 0.9#* iNOS knockout mice 21% O 2 99.0 ± 0.4 0.7 ± 0.2 0.3 ± 0.2 60% O 2 98.9 ± 0.5 0.4 ± 0.3 0.7 ± 0.4 >95% O 2 85.1 ± 2.0#* 4.9 ± 1.3#* 10.0 ± 1.2#* Each value represents mean ± SEM, n = 7. # p < 0.05 vs. normoxia; p < 0.05 vs. 60% O 2 Figure 1 Total cell counts in BAL from wild-type and iNOS knockout mice after 72 h exposure to 21%, 60%, and >95% O 2 . Data are mean ± SEM of seven mice for each group. # p < 0.05 vs. normoxia; p < 0.05 vs. 60% O 2 ; $ p < 0.05 vs. iNOS knockout mice. Lavageable lung protein Total protein concentration in the BAL fluid was determined as an indicator of lung hyperpermeability induced by hyperoxic exposure. Under normoxia and 60% O 2 , total protein concentration did not differ between wild-type mice (21% O 2 : 86.4 ± 37.3 μg/ml; 60% O 2 : 95.5 ± 22.8 μg/ml) and knockout mice (21% O 2 : 157.1 ± 23.7 μg/ml; 60% O 2 : 86.0 ± 26.6 μg/ml). Exposure to >95% O 2 resulted in a significant increase in protein concentration in wild-type mice (973.8 ± 95.7 μg/ml) and only in a modest increase in iNOS knockout mice (326.8 ± 90.4 μg/ml) that did not reach statistical significance. Lactate dehydrogenase activity As an indicator of cellular damage, LDH activity was measured in BAL fluid. Under normoxic conditions and 60% O 2 , LDH activity was comparable between wild-type (21% O 2 : 3.1 ± 1.0 U/min/ml; 60% O 2 : 1.6 ± 0.4 U/min/ml) and iNOS knockout mice (21% O 2 : 2.3 ± 0.5 U/min/ml; 60% O 2 : 6.5 ± 1.7 U/min/ml). Exposure to >95% O 2 resulted in a significant enhancement of LDH activity in wild-type mice (41.8 ± 10.8 U/min/ml) compared to iNOS knockout mice (9.6 ± 3.0 U/min/ml) and to normoxia (figure 3 ). Figure 3 Lactate dehydrogenase activity in BAL from wild-type and iNOS knockout mice after 72 h exposure to 21%, 60%, and >95% O 2 . Data are mean ± SEM of seven mice for each group. # p < 0.05 vs. normoxia; p < 0.05 vs. 60% O 2 ; $ p < 0.05 vs. iNOS knockout mice. TNF-α concentration TNF-α concentrations were determined in BAL fluid to investigate inflammatory cytokine release. Under normoxic conditions, TNF-α release did not differ between wild-type (28.5 ± 3.8 pg/ml) and iNOS knockout mice (35.0 ± 5.4 pg/ml), the same as upon 60% O 2 exposure (29.2 ± 2.6 pg/ml and 25.4 ± 6.0 pg/ml, respectively). However, there was a significantly enhanced TNF-α release measured upon >95% O 2 exposure in wild-type (83.0 ± 9.8 pg/ml) and iNOS knockout mice (54.9 ± 9.0 pg/ml) compared to normoxic conditions. TNF-α concentration was significantly higher in wild-type than in iNOS knockout animals (figure 4 ). Figure 4 TNF-α concentration in BAL from wild-type and iNOS knockout mice after 72 h exposure to 21%, 60%, and >95% O 2 . Data are mean ± SEM of seven mice for each group. # p < 0.05 vs. normoxia; p < 0.05 vs. 60% O 2 ; $ p < 0.05 vs. iNOS knockout mice. Concentration of thiobarbituric acid reactive substances Lung lipid peroxidation was assessed based on the levels of thiobarbituric acid reactive substances in BAL (figure 5 ). Wild-type mice exposed to >95% O 2 exhibited a pronounced increase in concentration of TBARS (146.0 ± 62.0 nmol/ml) compared to normoxia (35.0 ± 14.0 nmol/ml) and 60% O 2 (31.0 ± 17.0 nmol/ml). In iNOS knockout mice, concentrations of TBARS after >95% O 2 (52.0 ± 18.0 nmol/ml) did not differ from those after normoxic conditions (26.0 ± 0.0 nmol/ml) and 60% O 2 exposure (35.0 ± 26.0 nmol/ml), respectively. Figure 5 Concentration of thiobarbituric acid reactive substances in BAL from wild-type and iNOS knockout mice after 72 h exposure to 21%, 60%, and >95% O 2 . Data are mean ± SEM of seven mice for each group. $ p < 0.05 vs. iNOS knockout mice. Activation of NF-κB and AP-1 In an effort to elucidate transcriptional mechanisms for increased iNOS and TNF-α expression after hyperoxic exposure, electrophoretic mobility shift assays for NF-κB and AP-1 were performed (figure 6 ). NF-κB and AP-1 were weakly activated under normoxic conditions. Increased activation of both NF-κB and AP-1 was observed after >95% O 2 compared to normoxia and 60% O 2 . This enhancement of binding activity under hyperoxia appeared to be more prominent in the group of wild-type mice in comparison to iNOS knockout animals. Figure 6 Binding activity of NF-κB and AP-1 in lung tissue from wild-type and iNOS knockout mice after 72 h normoxia or hyperoxia. Figure shown is representative for seven experiments. iNOS mRNA expression To investigate the induction of the iNOS gene in lung tissue, expression of iNOS mRNA was analyzed (figure 7 ). As expected, there was no expression of iNOS mRNA in lung tissues from iNOS knockout mice. In wild-type mice, hyperoxic exposure (60% and >95% O 2 ) induced an increased expression of iNOS mRNA in lung samples compared to normoxic situation. Densitometric analysis was performed by determining the ratio of iNOS cDNA to β-actin cDNA. Results demonstrated a significant increase in iNOS mRNA expression upon >95% O 2 (1.2 ± 0.1) compared to 60% O 2 (0.8 ± 0.1). Figure 7 Ethidium bromide stained gels of β-actin and iNOS RT-PCR products in lung tissue from wild-type ( A ) and iNOS knockout mice ( B ) after 72 h normoxia or hyperoxia. Data shown are representative for seven experiments. Discussion Prolonged exposure to high concentrations of oxygen (>50% O 2 ) during an intensive care setting to maintain arterial pO 2 can lead to progressive lung injury. Several cellular systems including alveolar macrophages and leukocytes are involved in this process. Activation of inflammatory cells causes the release of reactive oxygen species and proinflammatory cytokines, resulting in endothelial dysfunction, tissue and alveolar edema formation, and surfactant inactivation. Furthermore, high levels of NO produced by inducible NO synthase may contribute to tissue damage. NO is directly cytotoxic or can combine with superoxide anions to form the more reactive oxidant peroxynitrite. Although a large amount of literature exists concerning the pulmonary response to oxidant exposure, some issues remain unresolved. Our findings confirm previous results showing that hyperoxia is able to upregulate iNOS expression in lung tissue [ 30 , 34 ]. As expected, there was no expression of iNOS mRNA in lungs of iNOS knockout mice. In wild-type mice, exposure to 60% and >95% O 2 induced a significant increase in iNOS mRNA expression. This enhanced iNOS mRNA expression during hyperoxic exposure seems to contradict findings reported in a study published by Arkovitz and colleagues, in which hyperoxia did not induce iNOS expression in lungs of mice [ 32 ]. This may be explained by the fact that the detection of iNOS mRNA by using northern blot technique is not as sensitive as RT-PCR. In accordance with results from others [ 28 - 30 ], we found little iNOS protein immunostaining under normoxic conditions and 60% oxygen exposure, while hyperoxic exposure >95% O 2 induced a prominent expression of iNOS protein in the lungs from wild-type mice (data not shown). The data from the present study demonstrate that in vivo oxygen exposure significantly elevated total BAL cell count after 72 h >95% O 2 both in wild-type and in iNOS knockout mice. According to this, oxygen exposure resulted in a significant enhancement in the number of neutrophils and lymphocytes in BAL fluid, combined with a significant reduction in the number of alveolar macrophages both in wild-type and iNOS knockout mice. Dedhia et al. also found elevated numbers of neutrophils and lymphocytes combined with decreased numbers of alveolar macrophages in rat lungs [ 45 ]. Recent studies report that, although iNOS deficiency does not affect leukocyte rolling and adhesion following treatment with thrombin [ 46 ], iNOS-deficient mice have significantly elevated leukocyte accumulation and enhanced leukocyte-endothelium interactions in endotoxinemia [ 24 ]. These results suggest that iNOS expression plays a potent role in regulation of leukocyte recruitment depending on the way of induction. Hyperoxia-induced inflammatory cell influx, particularly of neutrophils, can contribute to oxidant stress through formation of reactive oxygen species. Auten and collaborators demonstrated that DNA damage in hyperoxia-exposed rat lungs may be reduced by blocking neutrophil influx [ 47 ]. In our model of oxidant injury, no effect of iNOS deficiency on BAL cell differentials could be made out, whereas total BAL cell counts were significantly elevated in wild-type mice compared to iNOS knockout mice. The increase in the number of neutrophils and lymphocytes in BAL fluid may partially reflect the loss of integrity of the endothelium barrier. This damage is indicated by a significant elevation of total protein concentration and LDH activity after acute hyperoxia in wild-type mice in comparison to iNOS knockout animals. Kleeberger and colleagues previously reported that iNOS expression is involved in ozone-induced lung hyperpermeability showing reduced mean BAL fluid protein and leukocyte accumulation [ 48 ]. Recent studies indicate that iNOS also plays a proinflammatory role in the development of asbestosis-related pulmonary disorders, measured as a significantly decreased total protein count, LDH activity, and nitrotyrosine staining in iNOS-deficient mice [ 27 ]. In contrast, Kobayashi et al. reported that hyperoxia caused an increased accumulation of leukocytes, elevated LDH activity and albumin concentration, and a higher wet-dry-ratio in lungs from iNOS-deficient mice compared to wild-type animals [ 31 ]. Based on their findings, these authors suggest the presence of an iNOS-independent pathway of lung nitration and injury in hyperoxia. In our study, we found that nitrosylation of proteins in the lungs of mice exposed to >95% O 2 was attenuated in iNOS-deficient mice (data not shown). Formation of nitrotyrosine was proposed as a relatively specific marker for detecting endogenous generation of peroxynitrite. However, recent evidence indicates that alternate reactions are capable of inducing nitration of tyrosine in proteins, for example the reaction of myeloperoxidase with hydrogen peroxide. Therefore, increased nitrotyrosine staining is considered as an indicator of "increased nitrative stress" rather than a "footprint" for the formation of peroxynitrite [ 49 , 50 ]. Amplified formation of reactive oxygen and nitrogen species can be proved by determination of thiobarbituric acid reactive substances, a secondary product of lipid peroxidation indicating oxidative and/or nitrative stress [ 9 ]. In our study, significantly reduced formation of thiobarbituric acid reactive substances following >95% oxygen exposure was found in iNOS knockout mice, again suggesting a beneficial effect of iNOS deficiency on oxidant lung injury. Cytokines may also play a role in oxygen toxicity. Several studies point out that TNF-α is produced during hyperoxic exposure [ 51 , 52 ]. Furthermore, hyperoxia induces sequential formation of pulmonary TNF-α and IL-6, which corresponds to the severity of pathological findings [ 12 ]. In our study, iNOS deficiency resulted in a significant decrease in BAL TNF-α concentration during hyperoxic exposure. Findings of Sass et al. also demonstrate that iNOS-derived NO regulates proinflammatory genes in vivo resulting in inflammatory liver injury in mice by stimulation of TNF-α production [ 53 ]. To investigate whether hyperoxia-induced TNF-α expression was regulated on the level of protein or mRNA, activation of the redox-sensitive transcription factors NF-κB and AP-1 was analyzed. As recently described, NF-κB was activated following hyperoxia resulting in an increase in TNF-α and IFN-γ gene expression in murine pulmonary lymphocytes [ 35 ]. Moreover, we found that the activation of both factors seen in wild-type mice was weaker in iNOS knockout mice suggesting that induction of iNOS upon hyperoxia may in fact activate these transcription factors. These findings contrast the silencing effect of NO on NF-κB demonstrated upon stimulation with LPS or silica [ 54 ]. Data from Kupatt et al. [ 55 ] also indicate a negative feedback mechanism of eNOS-derived NO on activation of NF-κB following myocardial reoxygenation. In addition to isotype-specific differences in NO forming capacity, the synergistic NF-κB and AP-1 activation upon an reactive oxygen or nitrogen species challenge might diminish the inhibitory effect of NO. Recent studies indicate that exogenously administered NO causes increased c-fos and c-jun gene and protein expression combined with an evident AP-1 binding activity mediated by reactive oxygen and nitrogen species [ 56 ]. Conclusions Taken together, our data show that the absence of the iNOS gene does attenuate, but not fully abolish, oxidation, nitration, and cytotoxicity in response to acute hyperoxic exposure. The degree of transcriptional activation, inflammation, and oxidative lung injury caused by hyperoxia is significantly reduced in iNOS knockout mice compared to wild-type animals. In conclusion, these findings provide evidence to suggest that, upon hyperoxic exposure to >95% O 2 , proinflammatory effects of iNOS may be predominant, thereby contributing to the extent of acute hyperoxic lung injury. Authors' contributions AKH carried out the hyperoxic model, subsequent cytological and biochemical analyses, and writing and preparation of the manuscript. MD and FK participated in the direction of the study as well as in writing and preparation of the manuscript. CK carried out the electrophoretic mobility shift assays. The data presented in this paper are part of the doctoral thesis of AKH. All authors read and approved the final manuscript. List of abbreviations AP-1 activator protein-1 BAL bronchoalveolar lavage iNOS inducible nitric oxide synthase LDH lactate dehydrogenase LPS lipopolysaccharide NF-κB nuclear factor-kappa B NO nitric oxide TBARS thiobarbituric acid reactive substances TNF-α tumor necrosis factor-alpha Figure 2 Protein concentration in BAL from wild-type and iNOS knockout mice after 72 h exposure to 21%, 60%, and >95% O 2 . Data are mean ± SEM of seven mice for each group. # p < 0.05 vs. normoxia; p < 0.05 vs. 60% O 2 ; $ p < 0.05 vs. iNOS knockout mice.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC520822.xml
546219	The stability of life satisfaction in a 15-year follow-up of adult Finns healthy at baseline	Background While physical health has improved considerably over recent decades in Finland, the disease burden of mental health, especially that of depression, has become increasingly demanding. However, we lack long-term data on the natural course of subjective well-being in the general population. The aim of this study was to investigate the long-term course of self-reported life satisfaction. Methods This was a 15-year prospective cohort study on a nationwide sample of adult Finnish twins (N = 9679), aged 18–45 and healthy at baseline, who responded to postal questionnaires in 1975, 1981 and 1990 including a 4-item life satisfaction scale (happiness/easiness/interest in life and feelings of loneliness). Life satisfaction score (range: 4–20) was classified into three categories: satisfied (4–6), intermediate (7–11) and dissatisfied group (12–20). The associations between life satisfaction scores during the follow-up were studied with linear/logistic regression. Results Moderate stability and only a slight effect of age or birth-cohort on mean life satisfaction score (LS) were detected. In 1990, 56% of all and 31% of the dissatisfied remained in the same LS category as at baseline. Only 5.9% of the study subjects changed from being satisfied to dissatisfied or vice versa. Correlations between continuous scores (1975, 1981 and 1990) were 0.3–0.4. Baseline dissatisfaction (compared to satisfaction) predicted dissatisfaction in 1981 (OR = 10.4; 95%CI 8.3–13.1) and 1990 (5.6; 4.6–6.8). Multiple adjustments decreased the risk only slightly. Conclusions Life satisfaction in adult Finns was moderately stable during 15 years. Among an identifiable group (i.e. the dissatisfied) life dissatisfaction may become persistent, which places them at a greater risk of adverse health outcomes.	Background While physical health has improved considerably during recent decades in Finland, the disease burden of mental health – especially that of depression – has become increasingly demanding for the health care services and society. When the global disease burden – including both fatal and non-fatal outcomes – has been assessed, major depression has been shown to be one of its leading causes [ 1 ]. However, not only diagnosed depression but also incomplete recovery from depression and subthreshold depressive symptoms have adverse consequences and chronic courses [ 2 - 6 ]. In general, poor mental health affects somatic health and subjective well-being, but poor subjective well-being might also develop into mental disorder and create a loss of functional capacity, if not a sign of an undiagnosed mental disorder already at baseline. Life satisfaction and happiness are some of the concepts that have previously been viewed as indicators of subjective well-being [ 7 , 8 ]. Life dissatisfaction, even reported by seemingly healthy subjects, is associated with several indicators of poor health or health risk factors, but especially with depressive symptoms [ 9 - 11 ]. Longitudinally, it predicts poor health outcomes such as morbidity, mortality and premature work disability – due to both somatic as well as psychiatric causes – among the healthy but dissatisfied subjects [ 9 , 11 - 15 ]. When dissatisfaction is repeatedly reported over years the risk of a poor health outcome increases even more [ 13 - 15 ]. Due to these adverse health outcomes and shortened life expectancy among the identifiable group from the healthy general population (i.e. the dissatisfied), more attention should be paid to the natural course of life dissatisfaction in the general population. However, good mental health is also an area that should be studied in psychiatry. It is something more than the absence of symptoms. It is a mental state that is objectively desirable, indicating for example maturity, emotional and social intelligence, resilience and subjective well-being according to Vaillant [ 16 ]. While the possibilities to directly assess mental health at the level of the general population are limited, subjective well-being at the population level can be measured. Subjective well-being has mainly been investigated in cross-sectional settings and among the elderly, while follow-ups may have been brief and data from the general population have been sparse. Even if subjective well-being in general population has been suggested to be quite stable [ 9 , 17 - 23 ], it has also been pointed out that the apparent stability should not be due to the insensitivity of measurements to change or due to fact that most people report satisfaction with life [ 17 , 20 ]. However, this is not the case with psychiatric patients and life satisfaction, among whom life satisfaction has been shown to be lower than in any other patient group [ 9 ] and to improve markedly concurrently with their recovery from depression [ 10 ]. Mental health policy plays an increasingly recognized role in society, but it needs both epidemiological data as well as experts on the field of mental health to monitor the populations and trends [ 24 ]. Thus, in psychiatry, we need information on the natural long-term course of subjective well-being in the general population. This study aimed to examine the long-term course of life satisfaction in healthy adults and to determine how strongly self-reported life dissatisfaction predicts future life dissatisfaction. Methods This prospective cohort study with a follow-up from 1975 to 1990 was based on the Finnish Twin Cohort, a nationwide sample of all Finnish same-sex twin pairs born before 1958 with both members alive in 1975. A baseline health questionnaire was sent in 1975 to twin candidates [ 25 ]. The follow-up questionnaires in 1981 and 1990 were sent only to verified twins. Furthermore, the 1990 questionnaire was sent only to twins from pairs born in 1930–1957 with both co-twins alive and residing in Finland. The overall response rates to the questionnaires were 89% in 1975, 84% in 1981 and 77% in 1990. The study procedure has been presented in detail elsewhere [ 12 , 25 ]. The questionnaires included a four-question scale for life satisfaction, which was modified from a questionnaire developed for measuring the quality of life for research purposes in Nordic countries [ 26 ]. It has been used among all adult age-groups [ 9 , 12 ] as well as among psychiatric patients [ 9 , 10 , 27 ]. The study subjects were asked to rate aspects of life satisfaction: interest in life, happiness, ease of living and loneliness (very interesting/happy/easy/not at all lonely = 1, fairly interesting/happy/easy = 2, fairly boring/unhappy/hard/lonely = 4, very boring/unhappy/hard/lonely = 5). Missing data and the response 'cannot say' were scored as 3. If three or four items were missing, the sum score was recorded as 'missing'. Thus, the total score (LS) ranged from 4–20, with increasing scores indicating a decrease in life satisfaction . On the basis of the distribution of the sum score (LS), subjects were categorized into the satisfied (LS:4–6), the intermediate group (LS:7–11) and the dissatisfied (LS:12–20) [ 12 ]. The intermediate group consisted of those with an LS score within one standard deviation from the mean [ 9 ]. At baseline, responses to all four items were provided by 95.8% (N = 22,416) and at least two items, enabling LS to be calculated, by 99.2% (N = 23,212) of all respondents aged 18–45 years. Thus, one or two missing values were recoded as '3' values for 3.4% of subjects. The criteria for inclusion in the present study were the availability of baseline life satisfaction data, an age of 18–45 years on 1 January 1976 and being a twin (N = 19,973), since only twins were eligible to receive follow-up questionnaires, as well as being healthy at baseline (N = 16,496, see below for criteria). Moreover, the questionnaire was sent in 1990 only to those whose twin partner was alive. Thus, from these eligible subjects, study subjects were those with all three life satisfaction scores available (N = 9679). They consisted of 4466 (46.1%) male and 5213 (53.9%) female twins (Table 1 ). The mean age (SD) at baseline was 28.8 years (7.5) for men and 28.1 years (7.6) for women. Those subjects who had incomplete follow-up data (N = 6817) were compared with study subjects. Their life satisfaction data was available as follows: 1) LS 1975 and 1981 (n = 4930); 2) LS 1975 and 1990 (n = 385); 3) LS 1975 only (n = 1502). Table 1 Baseline characteristics of the study subjects and those with incomplete follow-up data on life satisfaction. Study subjects Subjects with incomplete follow-up p-value Baseline characteristics N (9679) column % N (6817) column % Sex <0.001 1) Men 4466 46.1 3932 57.7 Women 5213 53.9 2885 42.3 Age-group <0.001 2) 18 – 25 4121 42.6 3100 45.5 16 – 35 3513 36.3 2431 35.6 36 – 45 2045 21.1 1286 18.9 Social class <0.001 3) Upper 565 5.8 361 5.3 Intermediate 2695 27.8 2121 31.1 Lower 6419 66.4 4335 63.6 Marital status <0.001 4) Cohabiting 5386 55.7 3443 50.6 Living alone 4291 44.3 3366 49.4 Smoking cigarettes daily <0.001 5) Non-smoker 6698 69.3 3991 58.7 1 – 19 2344 24.2 2976 30.5 > 19 627 6.5 737 10.8 Pure alcohol g/month <0.001 6) None 1361 14.1 807 11.9 1 – 99 4077 42.2 2334 34.3 100 – 399 2626 27.1 2076 30.5 400 – 799 1083 11.2 939 13.8 ≥ 800 524 5.4 648 9.3 Physical activity/month <0.01 7) < 1 1138 12.2 888 13.7 1 – 5 4585 49.3 3045 46.9 ≥ 6 3577 38.5 2563 39.4 Life satisfaction in 1975 <0.001 8) 4 – 6 2214 22.9 1370 20.1 7 – 11 6239 64.4 4303 63.1 12 – 20 1226 12.7 1144 16.8 Life satisfaction data not available from 1981 and/or 1990. 1) F(1, 10007) = 165; 2) F(2, 10006) = 7.02; 3) F(2, 10006) = 9.17; 4) F(1, 10002) = 36.1; 5) F(2, 9995) = 94.0; 6) F(4, 9995) = 45.0; 7) F(2, 9786) = 5.50; 8) F(2, 10006) = 29.0 The criteria for baseline health were based on a health questionnaire (Q) and three nationwide registries: the Hospital Discharge Registry (H), the Registry of Specially Refunded Medication (M) and the Cancer Registry (C). Thus, those with symptoms or diseases covering cardiovascular disease, diabetes, chronic obstructive pulmonary disease or malignant cancer, those who used medications for 37 selected chronic somatic or psychiatric diseases, as well as those who were on a work disability pension due to any cause or had an inpatient admission between 1972 and April 1976, were excluded [ 12 ]. The specific exclusion criteria for psychiatric disorders covered work disability (Q), inpatient treatment due to psychiatric causes (ICD-8: 290–309) (H), the right to free medication for psychosis before 1977 (M) and use of hypnotics/tranquilizers for more than 10 days in the preceding year (Q). It has previously been reported that 4-item life satisfaction is associated with a lower age, female sex, cohabiting, an upper social class, non-smoking, lower alcohol consumption and physical activity [ 9 , 12 , 13 ]. Thus, the multivariate model included baseline variables such as age (18–24/25–34/35–45), sex, marital status (married or cohabiting/single, divorced or widowed), social class (lower/intermediate/upper group), physical activity (at least 30 minutes of exercise < 1/1–5/ ≥ 6 times a month), current smoking status (non-smoker/1–19/ > 19 cigarettes daily) and alcohol consumption (none/1–99/100–399/400–799/ ≥ 800 g pure alcohol/month) [ 12 ]. The upper social class consisted of those with at least 13 years of education and sedentary work, while the lower social class consisted of those with less than 10 years of education and work involving at least standing and walking. Data analysis was carried out using STATA (version 7.0). Since a study subject could be an age- and sex-matched twin sibling of another study subject, not all the observations were necessarily independent. Therefore, correct standard errors were computed by treating each pair of twins as a single unit (i.e. cluster sampling). The statistical significance of differences was tested by estimates of means (SVYMEAN and SVYLC procedure) for continuous variables and by the chi-squared test for categorical variables (SVYTAB procedure), corrected for clustered data and converted into F-statistics. The stability of life satisfaction over time was examined by computing Pearsonian correlation coefficients between continuous variables. To study how former life dissatisfaction predicts later life dissatisfaction, linear and logistic regression for clustered data was used. Results The baseline characteristics of the study population and those whose follow-up data on life satisfaction was not available at all three data collection times is presented in Table 1 . Study subjects were more often women, cohabiting, non-smokers and used less alcohol than those whose follow-up data on life satisfaction was incomplete. There were also slight differences in social class and physical activity. Furthermore, study subjects were somewhat more satisfied (mean LS 8.23; 95%CI 8.18–8.29 vs. LS 8.61; 8.54–8.68) and slightly older (mean age 28.4; 28.2–28.6 vs. 27.9; 27.7–28.1) than those with incomplete follow-up data (Table 1 ). In the study population with complete follow-up data on life satisfaction, no marked differences were observed in mean LS scores between age or gender groups or measurement times (Table 2 ). There was only a slight decrease in mean life satisfaction during the follow-up of 15 years. The main decrease took place in women during 1981–1990. When birth cohorts were studied, only those born during 1940–49 showed a trend of decreasing satisfaction throughout the follow-up, regardless of gender, but they were also the most satisfied group at baseline, being then 26–35 years of age. Young men aged 18–25 at baseline were and remained the most dissatisfied group throughout the follow-up. Table 2 Mean (95% CI) life satisfaction (LS) according to sex, birth cohort and current age among 9679 Finnish adults in 1975, 1981 and 1990. Subjects (n) LS 1975 (n) LS 1981 (n) LS 1990 All (9679) 8.23 (8.18 – 8.29) (9679) 8.26 (8.20 – 8.31) (9679) 8.35 (8.29 – 8.41) Men (4466) 8.30 (8.22 – 8.38) (4466) 8.33 (8.25 – 8.41) (4466) 8.38 (8.30 – 8.46) Women (5213) 8.18 (8.10 – 8.25) (5213) 8.19 (8.12 – 8.27) (5213) 8.32 (8.24 – 8.40) Birth cohorts, all 1950 – 57 (4121) 8.46 (8.36 – 8.55) (4121) 8.31 (8.23 – 8.40) (4121) 8.45 (8.36 – 8.54) 1940 – 49 (3513) 7.98 (7.90 – 8.07) (3513) 8.13 (8.05 – 8.22) (3513) 8.32 (8.22 – 8.41) 1930 – 39 (2045) 8.22 (8.11 – 8.32) (2045) 8.35 (8.23 – 8.46) (2045) 8.19 (8.08 – 8.31) Current age, all 18 – 23 (3254) 8.50 (8.40 – 8.61) - - - - 24 – 32 (3608) 8.04 (7.96 – 8.13) (4520) 8.28 (8.20 – 8.36) - - 33 – 45 (2817) 8.17 (8.08 – 8.26) (4101) 8.19 (8.11 – 8.27) (6236) 8.39 (8.31 – 8.46) 46 – 51 (1058) 8.39 (8.23 – 8.55) (1679) 8.37 (8.23 – 8.50) 52 – 60 - - (1764) 8.18 -(8.06 – 8.30) In terms of 3-category LS scores (Table 3 ), more than half (56%) of the study subjects in 1990 scored in the same category as they did in 1975, but the relationship was less stable for the satisfied (36%) and the dissatisfied (31%) than the intermediate group (69%). However, only 5.9% of the study subjects were satisfied (LS 4–6) at one of the three data collection times but dissatisfied (LS12-20) at one of the other time points. Table 3 The distribution of study subjects according to their self-reported life satisfaction (LS) in 1975 and 1990. LS 1990 Satisfied Intermediate Dissatisfied TOTAL N row % N row % N row % N row % LS 1975 Satisfied 785 35.5 1266 57.2 163 7.4 2214 100 Intermediate 1088 17.4 4293 68.8 858 13.8 6239 100 Dissatisfied 122 10.0 728 59.4 376 30.7 1226 100 TOTAL 1995 20.6 6287 65.0 1397 14.4 9679 100 Life satisfaction score: satisfied (LS 4–6); intermediate (LS 7–11); dissatisfied (LS 12–20). The correlation was 0.30 between continuous life satisfaction scores in 1975 and 1990, 0.38 between 1975 and 1981 scores and 0.40 between 1981 and 1990 scores. These coefficients were similar for men and women, but lowest among those aged 18–25 (0.26, 0.34 and 0.39, respectively) and highest among those aged 36–45 (0.36, 0.47 and 0.42, respectively). For the total study population the annual auto-correlation was estimated as 0.92 during 1975–90. Baseline life dissatisfaction predicted future life dissatisfaction (Table 4 ). This was also true after adjusting for all the covariates as well as when the categories of each covariate were separately investigated. The same pattern was shown both with categorical and continuous life satisfaction scores. The predictive ability was expectedly stronger for the shorter follow-up (1975–1981) than for the total follow-up (1975–1990). When the odds ratios were compared with those which could be calculated for the subjects with incomplete life satisfaction follow-up data, no significant differences were found. Table 4 Prediction of future dissatisfaction (LS 12–20) according to baseline dissatisfaction. Risk (OR with 95%CI) for the dissatisfied (LS 12–20) compared to the satisfied (LS 4–6) at baseline. COMPARISON OF LS SCORES BETWEEN 1975 & 1981 1975 & 1990 Subjects (n) OR (95 % CI) OR (95 % CI) All † 9679 10.42 (8.28 – 13.10) 5.56 (4.55 – 6.80) Adjusted ‡ 9283 9.79 (7.68 – 12.47) 5.23 (4.24 – 6.46) Men † 4466 14.81 (10.19 – 21.51) 6.11 (4.47 – 8.35) Adjusted ‡ 4298 12.86 (8.74 – 18.91) 5.62 (4.05 – 7.80) Women † 5213 7.93 (5.89 – 10.68) 5.29 (4.06 – 6.90) Adjusted ‡ 4985 7.94 (5.78 – 10.92) 4.98 (3.78 – 6.56) Risk of future dissatisfaction : LS 12–20 vs. LS 4–12 † Adjusted for age ‡ Adjusted simultaneously for age, sex, marital status, social class, alcohol consumption, current smoking and physical activity (cf. method section). Discussion Life satisfaction was moderately stable in healthy adult Finns during a 15 year period. Age or the birth-cohort had only a slight effect on mean life satisfaction. One third of those dissatisfied at baseline remained the same after the 15-year follow-up. The ability of baseline life satisfaction to predict future life satisfaction was strong, but decreased during the follow-up period, which is a trend that has also been suggested previously [ 23 ]. Previous studies have indicated that depression and depressive symptoms may have a chronic course [ 2 - 6 ]. On the other hand, in non-patient samples subjective well-being has also been suggested to be stable [ 17 - 22 ]. Concomitant anxiety or personality traits might play a role in this [ 28 - 32 ]. However, our results concerning the possible chronic course of life dissatisfaction are now based on a very long follow-up and a large sample of adults who reported or were found to have no indication of sickness at baseline. Although regression towards the mean in life satisfaction score was shown in the follow-up and greater instability among the dissatisfied and the satisfied than in the intermediate group, a complete shift from one extreme to another was rare. In Finland the number of new work disability pensions due to depression has strongly increased [ 33 ]. However, at the population level, subjective well-being seems not to have decreased correspondingly according to a comparison of two separate cross-sectional national surveys in 1980 and 2000 using the 12-item General Health Questionnaire [ 34 ]. Our cohort study with the 4-item life satisfaction scale measured three times during a 15-year follow-up on the same population strengthens these findings. On the other hand, during these years the physical health of the Finnish population has improved in many objectively assessed ways [ 34 ], but an improvement has not been seen in mental health indicators or in life satisfaction at population level. Thus, objectively assessed better somatic health or strong national economic growth (an increase of 48% in inflation-adjusted gross national income per capita from 1975 to 1990), which has also taken place during these years, seems not to guarantee better subjective well-being in a population that is globally speaking already quite well-off. This kind of trend has also been suggested previously [ 16 , 32 , 35 ]. On the contrary, the mean level of subjective well-being, which was previously sufficient to maintain work ability, seems not to meet the requirements of today's working life. Our results with respect to predictions seemed not to be overestimations. Those subjects whose dissatisfaction might have led to the most adverse result, i.e. death, were excluded from our study. The response to follow-up surveys was somewhat lower among the dissatisfied, but the observed risks among the study subjects did not differ statistically significantly from those obtained from the subjects with incomplete follow-up data. Similarly, when adjusting for follow-up health behavior instead of baseline health behavior variables, these predictions strengthened slightly, but not significantly. The 4-item life satisfaction scale is easily administered and well accepted. Its sum score was available for 96% (with imputed scores for 99%) of all respondents at baseline [ 12 , 13 ]. This might be due to the low number of items and its ability to tap the positive pole of subjective well-being [ 7 ], even if its sum score is also strongly associated with scores obtained by the 21-item Beck Depression Inventory [ 10 , 11 , 13 , 36 ]. In the general population, life dissatisfaction predicts both fatal and non-fatal poor long-term health outcomes [ 11 - 15 ]. Thus, it is worthwhile to assess subjective well-being. In general, if only poor subjective well-being is detected, our long follow-up suggests that there seems to be time to intervene. To prevent a process from leading to more adverse outcomes, subjects should acknowledge their situation and use their own personal resources, if available. According to a panel of experts, the concept of mental health can be regarded as a developmental process providing an individual or a group with the necessary resources to cope with the demands of life without the simultaneous appearance of negatively experienced moods of longer duration [ 37 ]. Society and health care services should support the growth of personal resources and start to intervene when these resources are inadequate and poor subjective well-being persists. In psychiatry, however, according to Vaillant, "since primary prevention is clearly superior to treating disease once it has occurred, we need to study also individuals with positive mental health the way that agronomists study wheat that is resistant to drought and blight" [ 16 ]. Our large nationwide sample with a high response rate and a long follow-up period enabled an examination of the long-term course of life satisfaction. The exclusion criteria for baseline health disorders were comprehensive and based on both self-reports and several national registries with high coverage and validity [ 38 - 41 ]. Although these analyses were performed on individuals drawn from a twin cohort, the results should be applicable to the general population. Being a twin does not affect the predictive ability of life satisfaction for mortality or suicide [ 12 , 13 ], and there is at most only a modest contribution of genetics to inter-individual differences in life satisfaction [ 42 ]. However, the potential influence of twinship was taken into account in the statistical analysis. The arbitrary exclusion of those twins who did not have a living twin partner in 1990, required in the composition of the Twin Finnish Cohort data, enabled us to control for the loss of a twin sibling. Conclusions Life satisfaction among healthy adult Finns was moderately stable in a 15-year follow-up. Since the dissatisfied, one third of whom consistently rated themselves as dissatisfied, can be identified from the general population, and since dissatisfaction places them at a risk, and repeatedly reported dissatisfaction at even greater risk, of adverse health outcomes, assessing subjective well-being should be encouraged both in surveys and in clinical practice in order to identify those in need of further evaluation of their mental health. Competing interest The author(s) declare that they have no competing interest. Authors' contributions KJ participated in composing the Finnish Twin Cohort data, in planning, commenting on, revising and approving the final manuscript. KM participated in composing the Finnish Twin Cohort data, in planning, commenting on, revising and approving the final manuscript. HR participated in planning, commenting on, revising and approving the final manuscript. HV participated in planning, commenting on, revising and approving the final manuscript. KHH planned the study, performed the statistical analyses and was the main author. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546219.xml
546231	Differential gene expression in recombinant Pichia pastoris analysed by heterologous DNA microarray hybridisation	Background Pichia pastoris is a well established yeast host for heterologous protein expression, however, the physiological and genetic information about this yeast remains scanty. The lack of a published genome sequence renders DNA arrays unavailable, thereby hampering more global investigations of P. pastoris from the beginning. Here, we examine the suitability of Saccharomyces cerevisiae DNA microarrays for heterologous hybridisation with P. pastoris cDNA. Results We could show that it is possible to obtain new and valuable information about transcriptomic regulation in P. pastoris by probing S. cerevisiae DNA microarrays. The number of positive signals was about 66 % as compared to homologous S. cerevisiae hybridisation, and both the signal intensities and gene regulations correlated with high significance between data obtained from P. pastoris and S. cerevisiae samples. The differential gene expression patterns upon shift from glycerol to methanol as carbon source were investigated in more detail. Downregulation of TCA cycle genes and a decrease of genes related to ribonucleotide and ribosome synthesis were among the major effects identified. Conclusions We could successfully demonstrate that heterologous microarray hybridisations allow deep insights into the transcriptomic regulation processes of P. pastoris . The observed downregulation of TCA cycle and ribosomal synthesis genes correlates to a significantly lower specific growth rate during the methanol feed phase.	Background The methylotrophic yeast Pichia pastoris is well established as expression host for heterologous proteins (reviewed by [ 1 ] and [ 2 ]). However, despite the high technological impact of P. pastoris , the physiological and genetic information is still rather scarce. The genome sequence has not been published, and in fact less than 100 complete gene sequences have been deposited with GenBank by the time of writing. Consequently, as for most other non-model species, no DNA microarrays are being manufactured. Hence, one of the most powerful tools for the investigation of changes in expression patterns is not available for this yeast species. To circumvent this problem, heterologous hybridisation to commercially available DNA microarrays might be conceivable. In fact, the successful non-homologous probing to microarrays has been reported recently. These studies cover a wide variety of organisms, including bacteria [ 3 ], a yeast [ 4 ], but also plants [ 5 ] and metazoan organisms [ 6 - 9 ]. The highest number of cross-hybridisation experiments has been performed with human microarrays. Chismar et al. [ 10 ] report, for instance, that heterologous probing of human cDNA arrays allows to gain useful information about gene expression in various primates. Moody et al. [ 11 ] compared, furthermore, the reproducibility of species-specific and cross-species hybridisations by evaluation of microarray hybridisations of porcine and human samples to human cDNA arrays. They reported that results generated by heterologous hybridisation were as reproducible as by homologous hybridisation, and the correlation between data derived from porcine and human hybridisations was strong. As judged from EST sequences of the porcine genome, the authors propose, that stretches of at least 100 bp with high similarity to the human homologue are sufficient for hybridisation. Renn et al. [ 9 ] compared the performance of cDNA microarrays from an African cichlid fish ( Astatotilapia burtoni ) for the heterologous hybridisation with cDNAs from eight different fish species, some of them closely related (other cichlids) and others more distantly related (among them Atlantic salmon and guppy). They conclude that significant results are obtained even with distantly related species, however, the number of positive spots declines with the phylogenetic distance, as strikingly does the degree of measured regulation. While yeasts share many morphological and physiological similarities, they represent a very heterogeneous group of fungal organisms, and a high degree of gene sequence similarity cannot be assumed a priori . When cDNA of the non-conventional yeast Zygosaccharomyces rouxii was probed by cross-hybridisation to Saccharomyces cerevisiae GeneFilters, only 155 ORFs out of the Z. rouxii genome could be reproducibly detected [ 4 ]. Anyhow, 86 genes out of these showed altered expression patterns between non-stressed and salinity-stressed Z. rouxii cells and 38 genes behaved differently than the S. cerevisiae homologues, indicating that the information gained is limited but nevertheless useful. As judged from 26S ribosomal DNA sequences, Z. rouxii is assumed to be more closely related to S. cerevisiae than P. pastoris , but still, all three belong to the hemiascomycetes [ 12 ]. As there are not many genes characterised for P. pastoris , there is no simple way to assess the degree of gene sequence similarity between P. pastoris and S. cerevisiae . However, in many of the few genes sequenced, stretches of high similarity (score >75% over a length of at least 250 bp and more) can be identified. Most of the known genes belong to the carbon and energy metabolism or contribute to amino acid or protein synthesis. Another group of P. pastoris genes with known sequence belongs to pathways specific for methylotrophic yeasts. For these genes there are no homologues present in S. cerevisiae . Evidently, this respective fraction of the P. pastoris genome would remain unevaluated by heterologous hybridisation. The main aims of this work are to verify whether a heterologous DNA array approach allows to obtain useful information for P. pastoris , and to identify genes that are specifically regulated upon a shift from glycerol to methanol as carbon and energy source. This shift is of particular interest since one of the specific features of methylotrophic yeasts is the tightly regulated methanol metabolism, which has been utilised for the construction of strong and tightly controlled expression vectors. The methanol induced promoter of the alcohol oxidase 1 ( AOX1 ) gene, which is repressed by many carbon sources such as glucose, glycerol or ethanol, is widely used for heterologous gene expression in P. pastoris . Accordingly, methanol is often used as the carbon source that induces the production of heterologous proteins. In addition to heterologous protein induction, the shift of the carbon source to methanol causes major structural and physiological changes within the cell. The enzymes for methanol metabolism are synthesised de novo and some of them are translocated into peroxisomes. Strikingly, peroxisomes can fill most of the cellular volume and AOX1 alone can account for up to 35% of the total soluble protein [ 13 ]. Additionally, heterologous protein production and environmental conditions like low fermentation pH have been proven to exert stress in recombinant P. pastoris [ 14 ]. Hence, for a first study of the transcriptomic regulations of recombinant P. pastoris , we used a strain expressing human trypsinogen under control of the AOX1 promoter, under conditions that strongly influence the physiology of the host cells, as previously described [ 15 , 16 ]. A series of microarray hybridisations was performed as depicted in table 1 , first to qualify the feasibility of cross-species hybridisation, and secondly to analyse the effects of the substrate change in fed-batch fermentations. Results and discussion 1. Qualification of heterologous hybridisation Before analysing differential gene expression data, it was our intention to verify whether the heterologous hybridisation of S. cerevisiae DNA microarrays with P. pastoris cDNA results in significant data. Obviously, the intensity of a signal will depend both on the amount of the specific mRNA in the sample, and the sequence similarity with the respective gene of S. cerevisiae . Therefore, we compared the signals obtained from four microarrays hybridised with P. pastoris cDNA with four microarrays hybridised with S. cerevisiae cDNA (as a control), obtained from shake flask cultures. To estimate the overall potential to obtain data, and the degree of loss of information, the total number of genes giving significant values, and those determined to be under a given threshold were compared (table 2 ). In average, 66 % of all genes present on the microarray were either only weakly transcribed, not similar enough to produce a significant signal or not present at all in P. pastoris . In contrast, by hybridisation with S. cerevisiae cDNA about 46 % of all genes remained undetected. We analysed those genes of P. pastoris for which sequences were deposited in the GeneBank database for sequence similarities to the S. cerevisiae genome, and determined the number of significant spots on 6 microarrays. 66 % of the signals derived from genes with high similarity (score > 75 % along stretches longer than at least 250 bp) were significant, while of the moderately to weakly similar genes only 28 % of the signals were significant. This indicates that a high sequence similarity of a sub-sequence within a gene is sufficient for efficient hybridisation. Considering that the signal intensity will depend on sequence similarity and length, but also on mRNA abundance, it becomes obvious, however, that a distinct minimum threshold of similarity cannot be defined. It was expected that the number of positive spots would be lower for heterologous hybridisation as compared to homologous hybridisation, but the relatively high number of significant values obtained in our experiment is very promising to achieve useful and new information from this technique. Nevertheless, we sought statistical evidence for the biological significance of hybridisation signals obtained with P. pastoris cDNA. First of all, data obtained from microarrays that were hybridised with identical but differentially labelled samples (yellow experiment) were evaluated, showing very high correlation coefficients of 0.97 and 0.98, for S. cerevisiae and P. pastoris respectively (table 3 and Fig. 1A ). The correlation of data from identical samples on different microarrays is somewhat lower (r = 0.86 – 0.92), due to different relative intensities on different chips. Since usually the data of two samples on one microarray are to be compared, it was important to evaluate the reproducibility, expressed as the standard deviations of both values of each spot on microarrays hybridised with identical samples. Fig. 1B displays the standard deviations plotted against the mean relative intensities of all significant spots of such a P. pastoris experiment, showing that the standard deviations do not vary over a wide range of signal intensities, which is in concordance with the results of Moody et al. [ 11 ]. 96 % of the signals have a relative standard deviation (s. d. divided by mean) below 0.2. Furthermore, we evaluated the correlation between the signal intensities obtained from P. pastoris cDNA with that from S. cerevisiae cDNA (cells grown under the same conditions). A highly significant correlation (r = 0.72) was observed (Fig 1C ). Considering expectable differences in gene expression, different sequence similarities and the fact that signals from two microarrays were compared, such a high correlation is remarkable and suggests that the data obtained are biologically meaningful. The results obtained are in line with the data published by Moody et al. [ 11 ] who showed a strong correlation between porcine and human samples on human microarrays. Our data also support Renn et al. [ 9 ] who demonstrated that both the number of significant spots and signal correlation of different fish samples on A. burtoni microarrays depended on the phylogenetic distances between sample species and test species. To evaluate the significance level for up- and downregulation, the signals from yellow experiments were plotted one against the other (Fig. 1A ). These values should obviously fall in the unregulated range. A threshold regulation factor of 1.5 (illustrated by the dotted lines) includes 99.1 % of all significant spots as not significantly different, which means that such a threshold would yield false positives for less than 1 % of the significant genes. Finally, the global regulation pattern of P. pastoris and S. cerevisiae between pH 5.0 and 3.5 was compared by correlating all genes downregulated under acidic conditions (Fig. 1D ). The correlation is highly significant (p < 0.01) with a correlation coefficient r = 0.51. Clearly, different gene regulation, as well as different degrees of sequence similarities contribute to a reduction of this correlation. Interestingly, the average fold signal change of regulated genes is lower for P. pastoris than for S. cerevisiae (correlation slope = 0.8). A similar observation was made for more distantly related fish species [ 9 ]. 2. Analysis of gene regulation in P. pastoris As an example for differential gene expression in this study, the level of transcripts during the methanol induction phase of a lab scale fermentation was compared to that during the glycerol feeding phase, both at pH 5.0, and at pH 3.0. Figure 2 shows the development of biomass over time along with the pH and indications of feed changes, and furthermore depicts the time points when the analysed samples were taken. Of all genes that gave significant signals, we report those that showed significant differences upon the shift from glycerol to methanol under at least one condition (pH), and that have a defined function in S. cerevisiae . Unsurprisingly, genes involved in the core carbon metabolism show a significantly different expression during methanol and glycerol metabolism (table 4 ). Almost no differences between the samples obtained at pH 5.0 and 3.0 were found. Fructose-1, 6-bisphosphatase was significantly upregulated at least in one sample, whereas fructose 1, 6-bisphosphate aldolase, glyceraldehyde-3-phosphate dehydrogenase, enolase, and pyruvate kinase were found to be downregulated on methanol. Transketolase expression was enhanced while the transaldolase transcript level was relatively reduced. Of these enzymes, fructose-1, 6-bisphosphatase and transketolase (among others) are needed for biomass synthesis on methanol. Some genes related to ATP production were found to be downregulated which appears plausible as the rate of energy consumption is lower at the significantly lower specific growth rate when methanol is the carbon source. The beta subunit of pyruvate dehydrogenase was significantly downregulated at pH 5.0, further indicating that methanol metabolism decreases the TCA cycle flux. The upregulation of pyruvate decarboxylase under at least one condition comes somewhat unexpected, because an increase of the flux to ethanol production appears questionable for cells growing on methanol. Table 5 displays a list of ribosomal genes that are essentially all downregulated on methanol. Considering the decreased specific growth rate, a decrease of total RNA is generally expected due to a decreased overall protein synthesis. Interestingly, four histone genes that gave significant signals were not regulated at pH 5.0, but induced at pH 3.0. Only a few amino acid biosynthesis genes appeared to be regulated (table 6 ), indicating that amino acid synthesis is turned on, both on glycerol and methanol, as mineral media were used throughout the experiment. A major exception was the significant regulation of the proteins involved in methionine metabolism. While the homologues to MET3 , MET16 and MET17 were upregulated on methanol, the homologues to MET6 , SAM1 , SAM2 and SAH1 were downregulated. As shown in figure 3 , the first group of enzymes catalyses the reduction and fixation of sulphur, while the second group drives the cycle responsible for methyl group donation. The reduction of this pathway in cells grown on methanol would imply a decrease of the flux from the C1-pool to methionine by MET6 (5-methyltetrahydrofolate-homocysteine S-methyltransferase), which catalyses the transport of activated methyl groups from 5-methyl-THF to homocysteine. These methyl groups are then passed on via S-adenosyl-methionine by a variety of S-adenosyl-L-methionine-dependent methyl transferases, many of them being involved in ribosomal subunit biogenesis, rRNA and tRNA-processing, mRNA capping and nuclear export – once again stressing the lower demand for protein synthesis rate upon growth on methanol. Other significantly regulated groups of genes (table 7 ) belong to the thiamine biosynthesis, all being upregulated, and the so-called snooze genes related to the stationary phase. Two of these genes, SNZ1 and SNZ2 , were differently regulated at pH 5 and pH 3, being repressed on methanol at pH 5 and induced at pH 3, while SNZ3 was repressed at pH 5, but did not yield a significant value at pH 3. However, the reported high sequence similarity of the S. cerevisiae SNZ genes will not enable a reliable differentiation of their regulation on microarrays. For S. cerevisiae it is reported that the highly homologous products of the SNZ gene family are involved in vitamin B6 (pyridoxal) synthesis [ 17 ]. Zeidler et al. [ 18 ] postulated that pyridoxal is a precursor of thiamine in yeast. Accordingly, the SNZ genes have been reported to be induced both by thiamine and pyridoxal depletion [ 17 ]. However, with the data obtained in our experiment we cannot interpret the differential behaviour of the SNZ and THI genes, since among the genes utilising thiamine-pyrophosphat (TPP) as cofactor, transketolase ( TKL1 ) and pyruvate decarboxylase ( PDC1 , PDC5 ) are upregulated while PDB1 (pyruvate dehydrogenase beta-subunit) and presumably also α-ketoglutarate dehydrogenase (belonging to the TCA cyle) and DHAS (belonging to the methanol utilisation pathway) are downregulated. Thioredoxin-related genes appeared to be regulated upon shift from glycerol to methanol, too (table 7 ). Those being downregulated at least in one sample (IMP cyclohydrolase, ribonucleotide reductase, and thioredoxin reductase) are involved in ribonucleotide synthesis, again indicating a decreased demand of RNA precursors. Thioredoxin peroxidase, on the other hand, which is involved in the regulation of cell redox homeostasis and response to oxidative stress by reducing H 2 O 2 and peroxide radicals, tended to be upregulated. This was expected due to the higher amount of oxidative stress during methanol utilisation. To verify the data obtained with microarrays with an independent method, a northern blot analysis of the total RNA samples from the culture grown at pH 5.0 was performed, using the respective P. pastoris homologous sequences to produce the probes (Fig. 4 ). The actin mRNA level was unchanged whereas the human trypsinogen mRNA level (as a positive control) was strongly induced on methanol, both as expected. MET17 and SAH1 exemplify differentially regulated genes identified by the microarray experiment (table 6 ). The data from the northern blot confirmed the results derived by microarray analysis, further underlining the reliability of the method. Surprisingly, only minor differences in transcriptional regulation were observed between cultures grown at pH 5.0 and pH 3.0. Of course it has to be considered that not the direct effects of a shift in external pH was observed. Still, one could expect a different set of genes to be significantly regulated upon a shift to methanol metabolism at the different pH values. At pH 3.0 a decreased yield in biomass was detected as compared to cultures at pH 5.0, which is consistent with the observation of Hohenblum et al. [ 16 ] stating that low fermentation pH decreases the viability of P. pastoris . In order to compare the general behaviour of P. pastoris at different external pH to that of S. cerevisiae , we measured the intracellular pH (pH i ) of the cells. The cells of the culture at pH 5.0 showed a pH i of 7.1 at the end of the glycerol phase, and of 7.2 the end of the methanol phase, whereas the pH i of both samples of the cultures at pH 3.0 was 7.3. Thus, no changes of the pH i were observed between cultures grown at the chosen pH values, which is in contrast to the behaviour of S. cerevisiae , where the pH i appears to be more dependent on the external pH [ 19 , 20 ]. Conclusions We could successfully demonstrate that it is possible to obtain new and valuable information about transcriptomic regulation in P. pastoris by probing S. cerevisiae DNA microarrays. Specific regulation upon a shift from glycerol mineral medium to methanol mineral medium under conditions similar to a production process were analysed. As major effects we recognised a downregulation of TCA cycle genes, and a transcriptional decrease of genes related to ribonucleotide and ribosome synthesis. Furthermore, the supply of activated methyl groups via adenosyl methionine was reduced, indicating a decreased ribosome and tRNA synthesis, which is not surprising since the specific growth rate is significantly decreased during the methanol feed phase in comparison to the glycerol feed phase. Correspondingly, a downregulation of the energy metabolism upon methanol induction appears reasonable. Only a few genes were differentially regulated when comparing expression differences between growth on glycerol and growth on methanol of a culture grown at pH 5.0 to one at pH 3.0. Among the few genes found are the SNZ genes and the histone genes, but a plausible hypothesis for the differential pH dependent regulation was not found. Interestingly, also the intracellular pH did not change between the different external conditions, indicating a major difference in pH regulation between P. pastoris and S. cerevisiae . Materials and Methods Unless stated otherwise, all chemicals were purchased from Merck Eurolab, and all enzymes for DNA manipulation were purchased from MBI Fermentas. 1. Strains The expression strain used in this study was P. pastoris strain X33 (Invitrogen), a wild type strain which can grow on minimal media without supplements. The identity of the strain in use was verified by partial 26S ribosomal DNA sequencing (data not shown). The selection mechanism was based on the Zeocin™ resistance of the transformation vector. Transformation of the strain was carried out with a plasmid derived from pPICZαB (Invitrogen), containing the gene for human trypsinogen 1 [ 21 ]. pPICZαB utilises the AOX1 promoter of P. pastoris and the α-factor leader sequence of S. cerevisiae for product secretion. The selected strain was of the methanol utilisation positive (mut + ) phenotype, which means that it is fully capable to metabolise methanol as the sole carbon source. As a control strain we used S. cerevisiae CEN.PK 113-5D ( MATa , ura3 ) [ 22 ]. 2. Shake flask cultivation of P. pastoris and S. cerevisiae Shake flask cultures were performed at 28°C in YPD medium (2% peptone, 2% glucose, 1% yeast extract). The cells were inoculated to an OD 660 of 0.3 from a pre-culture grown over night. For the yellow experiment the samples were taken in exponential phase after 5 h of growth (OD 660 for S. cerevisiae : 0.89, P. pastoris : 1.2). To assess differentially regulated genes the respective cultures were divided after 5 h of growth. One half was incubated as before, whereas the other half was supplemented with 250 mM acetic acid. Samples were collected after shaking for 1.5 h at 28°C. The pH of the untreated culture was 5. The pH of the acid treated culture was 3.5. 3. Fermentation of P. pastoris Fed batch fermentations were performed with a MBR mini bioreactor with a final working volume of 2 l, essentially as described by Hohenblum et al. [ 16 ]. The media were as follows PTM 1 trace salts stock solution contained per litre 6.0 g CuSO 4• 5H 2 O, 0.08 g NaI, 3.0 g MnSO 4• H 2 O, 0.2 g Na 2 MoO 4• 2H 2 O, 0.02 g H 3 BO 3 , 0.5 g CoCl 2 , 20.0 g ZnCl 2 , 65.0 g FeSO 4• 7H 2 O, 0.2 g biotin and 5.0 ml H 2 SO 4 (95 %-98 %). All chemicals for PTM 1 trace salts stock solution were from Riedel-de Haën, except for biotin (Sigma), and H 2 SO 4 (Merck Eurolab). Batch medium contained per litre 23.7 ml H 3 PO 4 (85 %), 0.6 g CaSO 4• 2H 2 O, 9.5 g K 2 SO 4 , 7.8 g MgSO 4• 7H 2 O, 2.6 g KOH, 40 g glycerol, 4.4 ml PTM 1 trace salts stock solution. Glycerol fed-batch solution contained per litre 632 g glycerol (100 %) and 12 ml PTM 1 trace salts stock solution Methanol fed-batch solution contained per litre 988 ml methanol (100 %) and 12 ml PTM 1 trace salts stock solution The dissolved oxygen was controlled at DO = 30 % with the stirrer speed (600 – 1200 rpm). Aeration rate was 100 l h -1 air, which was supplemented with oxygen (up to 25 %) after the begin of the fed batch. The temperature was 25°C, and the pH was controlled with NH 3 (25 %). Before starting the fermentation, the pH of 1.2 l batch medium was set to 5.0 with NH 3 (25 %). The batch phase of approximately 32 h was followed by a 4 h fed batch with glycerol medium (feed rate 15.6 ml h -1 ), leading to a dry biomass concentration of approximately 40 g l -1 . Then, the feed with methanol medium was started with a feed rate of 6.4 ml h -1 . The fermentation was terminated 14 h after the methanol feed start. The pH was 5.0 during batch, and either kept at 5.0 throughout the fermentation, or decreased to 3.0 at the beginning of the glycerol fed batch. The final dry biomass concentration was 51.4 g l -1 at pH 5.0, and 46.7 g l -1 at pH 3.0. Samples were taken at the end of the glycerol fed batch phase and at the end of the methanol fed batch phase, respectively, as depicted in figure 2 . 4. mRNA preparation The cell pellets were re-suspended in 10 × the volume of TRI-reagent (Sigma) and frozen. The samples were thawed on ice and after addition of acid washed glass beads the cells were homogenised in a Ribolyser (Hybaid Ltd.) for 2 × 20 sec, in between cooling on ice. After addition of chloroform, the samples were centrifuged and the total RNA was precipitated from the aqueous phase adding isopropanol. The pellet was washed 2 × with 70% ethanol, dried and re-suspended in RNAse free water. mRNA was isolated using the MicroPoly(A)Purist mRNA purification Kit (Ambion) according to the manufacturers protocol. 5. Synthesis and labelling of cDNA 5 μg of mRNA and 0.5 μg of oligo dT primer were mixed in 7 μl of water, incubated for 5 min at 70°C and subsequently at 42°C for about 3 min. The following components were added to 5 μl of said reaction mixture: 4 μl reaction buffer (5 x) for SuperScript II reverse transcriptase (Invitrogen), 2 μl dTTP (2 mM), 2 μl dATP, dGTP, dCTP (5 mM), 2 μl DTT (100 mM), 2.5 μl RNasin (40 U, Promega) and 2 μl FluoriLink Cy3-dUTP (1 mM) or 2 μl FluoriLink Cy5-dUTP (1 mM, Amersham Biosciences) respectively, and 1 μl SuperScript II reverse transcriptase (200 U, Invitrogen) to result in a total of 19.5 μl. The mixture was incubated for 1 h at 42°C. After addition of further 200 U SuperScript II reverse transcriptase the mixture was incubated for another 1 h at 42°C. 7 μl of 0.5 M NaOH/50 mM EDTA were added and the mixture was incubated at 70°C for 15 min. The reaction mixture was neutralised by addition of 10 μl Tris-HCl pH 7.5 (1 M). The labelled cDNA of the two corresponding samples were pooled and purified with Qiaquick purification columns (Qiagen) according to the manufacturer's protocol. 6. Chip hybridisation and set-up of microarrays The cDNA microarrays used for this study were Hyper Gene Yeast Chips from Hitachi Software Engineering Europe AG. According to the manufacturer, about 0.1 to 0.3 ng of PCR amplified cDNA (approximately 200 bp to 8000 bp) were spotted onto a poly-L-lysine coated glass slide and fixed by baking, succinic anhydride blocking and heat denaturation. Labelled cDNA was resuspended in about 70 μl of 5 × SSC/0.05% SDS, heat denatured at 95°C for 3 min and cooled on ice. SDS crystals appearing were dissolved by short and slight warming and the mixture was gently applied to a Yeast Chip according to the scheme presented in table 1 . The spotted area was covered with a cover glass and the chips were placed in an airtight container with a humidified atmosphere at 60°C for 16 h. The cover glasses were removed in 2 × SSC/0.1% SDS and the chips were washed consecutively for 5–10 min each in 2 × SSC/0.1% SDS, 0.5 × SSC/0.1% SDS, and 0.2 × SSC/0.1% SDS at RT. The chips were centrifuged at 600 rpm for 3 min in order to dry them. The washing conditions were chosen according to the manufacturer's manual. We have tested less stringent washing conditions which led to higher background without increasing the number of positive signals. 7. Data acquisition and processing Images were scanned at a resolution of 50 μm with a G2565AA Microarray scanner (Agilent) and were imported into the GenePix Pro 4.1 (Axon Instruments) microarray analysis software. GenePix Pro 4.1 was used for the quantification of the spot intensities. Each appearing gene spot was averaged. The data set was then imported into GeneSpring 6.1 (Silicon Genetics) for further normalisation and data analysis. All of the values of each channel on each chip were divided by their respective median for normalisation. Subsequently, the median intensity of all 84 TE spots (spotted with buffer, no DNA) deduced from each value, and all spot values less then the standard deviation of said 84 threshold values were considered to be not significant and were set to the value of the standard deviation. To determine induction or repression of gene activity, the normalised signals on each spot were compared, and all genes showing a signal difference exceeding the threshold (2 fold for S. cerevisiae , and 1.5 fold for P. pastoris , see results) on both parallel independent microarrays were judged as significantly regulated. 8. Statistical evaluation of microarray data After normalisation and background deduction, pairwise correlations of all significant values were calculated using Pearson's correlation coefficient. To evaluate the variability of data derived from both dyes on one chip, the standard deviations of all significant spots hybridised with two identical samples were plotted against the respective normalised mean intensity value. To judge the correlation of gene regulation between S. cerevisiae and P. pastoris , the regulation factors of all of genes that were significantly downregulated in S. cerevisiae upon a difference of pH 5.0 to 3.5, were correlated to the respective P. pastoris regulation factor upon the same media difference, using the Pearson's correlation coefficient. The significance of all correlations against randomly distributed values were evaluated by a t-test, applying a significance level p < 0.01. To exclude an effect of data clustering (as the majority of the values are rather low) on correlation, Spearman's correlation coefficients were calculated as well. As these differed only slightly from Pearson's coefficients, they are not shown. Linear regression analysis was performed with the regulation intensities of P. pastoris against S. cerevisiae in order to compare the average fold change observed for both yeasts. 9. Determination of the intracellular pH (pH i ) The pH i was determined as described by Valli et al. [ 19 ]. Essentially, samples were centrifuged and resuspended in McIlvaine buffer [ 23 ] at pH 3.0, containing 20 μM carboxy SNARF-4F AM (Molecular Probes). Loaded cells were analysed on a FACS Calibur (Becton Dickinson, Franklin Lakes, NJ USA) with a 488 nm argon-ion laser. 10 4 cells were measured per analysis, using PBS as the sheath fluid. Carboxy SNARF-4F fluorescence emission was measured through a 585/21 BP filter (FL2) and a 670 LP filter (FL3). Threshold settings were adjusted so that cell debris were excluded from data acquisition. The ratio of the two fluorescence intensities is a measure for the internal pH. Calibration was performed with amphotericin B (Sigma) perforated cells as described in Valli et al. [ 19 ]. 10. Northern blot analysis Northern blot analysis was essentially performed as described by Sambrook et al. [ 24 ]. In short, total RNA prepared as described above was fractionated on a denaturing formaldehyde containing gel, capillary blotted onto a nylon membrane (Nytran Supercharge, Schleicher & Schuell) and fixed by baking. The membrane was stained with methylene blue (0.04% in 0.5 M NaOAc, pH 5.2) for quality control and to ensure that equal amounts of RNA had been loaded. The probes were PCR amplified from genomic P. pastoris DNA using the following primers: actin: gttccagccttctacgtttctattca and acggagtactttctttctggtggag ; SAH : agctgaacttgattttggacgac and acttgaggcttgatgttgctgac ; Met17 : tgcatcaatggtcacggtaaca and tggtgagtagagtagtaaggagcaatga . The probe for human trypsinogen was prepared as described in [ 15 ]. The probes were DIG labelled using the PCR DIG Labeling Mix (Roche) according to the manufacturer's protocol. Pre-hybridisation and hybridisation were performed in high SDS hybridisation buffer at 42°C. The blots were washed twice at RT with 2 × SSC/0.1 % SDS and two times at 68°C with 0.5 × SSC/0.1 % SDS. Staining of the blots was performed using anti-Digoxigenin-alkaline phosphatase Fab Fragments (Roche) and the CDP Star chemiluminescent reagent (Tropix) according to the manufacturer's protocol. The images were taken with a Lumi imager F1 (Boehringer Mannheim). Authors' contributions MS and PB designed and performed the microarray hybridisation experiments. MS and BG analysed and interpreted the microarray data, and drafted the manuscript. MV performed the pH i determinations. MM ran and analysed the fed batch fermentations. DP participated in the design of this study, and in data interpretation. DM participated in the design of this study, and in data interpretation, performed the statistical analyses, and drafted part of the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546231.xml
514563	Medical Students' and Residents' preferred site characteristics and preceptor behaviours for learning in the ambulatory setting: a cross-sectional survey	Background Medical training is increasingly occurring in the ambulatory setting for final year medical students and residents. This study looks to identify if gender, school, level of training, or speciality affects learner's (final year medical students and residents) preferred site characteristics and preceptor behaviours for learning in the ambulatory setting. Methods All final year medical students and residents at the five medical schools in Ontario (N = 3471) were surveyed about the site characteristics and preceptor behaviours most enhancing their learning in the ambulatory setting. Preferred site characteristics and preceptor behaviours were rank ordered. Factor analysis grouped the site characteristics and preceptor behaviours into themes which were then correlated with gender, school, level of training, and speciality. Results Having an adequate number and variety of patients while being supervised by enthusiastic preceptors who give feedback and are willing to discuss their reasoning processes and delegate responsibility are site characteristics and preceptor behaviours valued by almost all learners. Some teaching strategies recently suggested to improve efficiency in the ambulatory teaching setting, such as structuring the interview for the student and teaching and reviewing the case in front of the patient, were found not to be valued by learners. There was a striking degree of similarity in what was valued by all learners but there were also some educationally significant differences, particularly between learners at different levels and in different specialities. Key findings between the different levels include preceptor interaction being most important for medical students as opposed to residents who most value issues pertaining to patient logistics. Learning resources are less valued early and late in training. Teaching and having the case reviewed in front of the patient becomes increasingly less valued as learners advance in their training. As one approaches the end of ones' training office management instruction becomes increasingly valued. Differences between specialities pertain most to the type of practice residents will ultimately end up in (ie: office based specialties particularly valuing instruction in office management and health care system interaction). Conclusions Preceptors need to be aware of, and make efforts to provide, teaching strategies such as feedback and discussing clinical reasoning, that learners have identified as being helpful for learning. If strategies identified as not being valued for learning, such as teaching in front of the patient, must continue it will be important to explore the barriers they present to learning. Although what all learners want from their preceptors and clinic settings to enhance their learning is remarkably similar, being aware of the educationally significant differences, particularly for learners at different levels and in different specialities, will enhance teaching in the ambulatory setting.	Background "The ideal preceptor should be like Captain Picard from Star Trek, who has a good grasp of situations but lets his subordinates push themselves to their limits without interfering/imposing his views and methods"! (survey comment) Medical care is being delivered primarily in the ambulatory setting in an increasing number of specialties. Since learning is best done contextually [ 1 , 2 ] it is appropriate and necessary that medical training also increasingly occur in the ambulatory setting. Theory suggests that trainees at different levels [ 3 - 8 ] and in different specialties [ 3 , 6 , 9 - 11 ] may have different learning needs. Students early in their training may be looking to be taught certainties about facts and concepts, corresponding to Perry's concept of simple dualism, ie; right versus wrong, and to not find it helpful to be engaged in discussions about "softer" emotional and social issues [ 5 ]. Stritter found first year residents preferred being told what to do, whereas higher-level residents preferred more autonomy and more explanations from their preceptors [ 7 ]. Work looking at learning styles in different specialities has been mainly based on Kolb's work, who outlined four different learning styles, and suggests those in different specialities learn differently (ie: surgeons learn best by hands-on practical application of ideas [ 11 ], while pathologists learn best using abstract theoretical models [ 10 ]). An article by Kernan [ 12 ] outlined site accommodations and preceptor behaviours that third year medical students felt facilitated their learning during a one-month ambulatory internal medicine rotation. A pilot project at our institution asking first year family medicine residents to rank Kernan's study items found differences between the two groups. What was not clear was if these differences were due to school attended, level of training or specialty. Given that trainees at all levels and in all specialties are increasingly being trained in the ambulatory setting, it seemed important to understand if there truly are differences between different types of students in what is perceived as being most helpful for learning. If differences are identified it will then be important to study whether adjusting to these differences actually improves learning. We surveyed all final year medical students and residents in Ontario about the site characteristics and preceptor behaviours that they find most enhance learning in the ambulatory setting and determined if these were related to demographic factors, level of training or residency program. Implications for teaching in the ambulatory setting are explored based on these results. Methods All medical students (n = 532) and residents (n = 2939) at the five medical schools in Ontario were surveyed using a four part questionnaire which collected information on demographics, preferred site characteristics, preferred preceptor behaviours, and approaches to learning and perceptions of learning climate. Questions for the site characteristics and preceptor behaviours included previously validated questions [ 12 - 15 ] and questions believed to be important by study group consensus. The approaches to learning and perceptions of learning climate questionnaire was validated by Kirby et al [ 16 ] and is not reported here. Students rated 24 site characteristics and 38 preceptor behaviours on a Likert scale from 1 (very important for learning) to 5 (not at all important for learning) or D (detrimental for learning). Within each section they indicated the five most important and 5 most unimportant or detrimental items for learning. A section for general comments was included at the end of the survey. The survey was piloted with a group of Queen's University residents and final year medical students checking for ambiguity and content. Ethical approval was granted by the Queen's University General Research Ethics Board. To ensure privacy for their students schools requested that the questionnaires be addressed by their own undergraduate and postgraduate offices. Coded questionnaires were thus sent with student's names and bulk mailed to the undergraduate and postgraduate medical schools who then addressed and forwarded the questionnaires to their final year medical students and residents. Entry into a draw for a Personal Digital Assistant or equivalent monetary prize was offered for completed surveys. Non-responders were identified by a lack of a returned coded questionnaire. Two subsequent mailings were sent to the non-respondents through their schools' undergraduate or postgraduate office. In addition an email reminder was sent to everyone between the second and third mailing. Data were analyzed using SPSS for Windows, version 11.0 [ 17 ]. A systematic effort to look for out-of-range data was conducted by doing frequency distributions for each of the variables, identifying out-of-range entries and correcting the errors by going back to the original data sheets. Double entry data assessment was not done. Frequency distributions for demographic factors, valued site characteristics, and preceptor behaviours were compiled. Counts were derived for each site characteristic or preceptor behaviour by calculating percentages of respondents giving the item a score of 1 or 2 on the Likert scale. Detrimental items were tallied from the frequency data. Factor analysis of the site characteristics and preceptor behaviours was carried out. Cronbach alpha coefficients were calculated for the identified factors. (Factor analysis is a means of reducing a large number of items to a smaller, more manageable number of dimensions, based on the ways in which the items correlate with each other. Cronbach alpha coefficients can then be calculated for internal consistency of the scales based on the identified factors. The resulting factors/scales need to be interpreted, but may provide a view of underlying constructs that are responsible for the observed variables and their correlations. Both the choice of the number of factors to extract and the interpretation of the factors/scales are matters of interpretation [ 18 ].) Counts (derived by calculating percentages for each item ranked 1 or 2 on the Likert scale) were generated for gender, school, level of training and residency for each factor. The Post-Graduate Year 2 (PGY2) group had an additional, possibly confounding feature, containing a large number of family medicine residents, who would be at the end of their training, instead of half way through their training like the remainder of the group. A subanalysis was done on the level of training data removing family medicine residents from the PGY2 data to analyse the impact of this on preferred site characteristics and preceptor behaviours. Initial data interpretation for residency used 11 residencies. Residencies were then collapsed into five groups (medicine; family medicine, paediatrics, psychiatry; lab/path, radiology; surgery, emergency, ob/gyn; and intensivists,anaesthetists) based on similarity of practice patterns. Logistic regression analysis[ 17 ] was used to compare gender, school, level of training and residency with respect to the site characteristic and preceptor behaviour factors. Each independent variable (ie: gender, school, level of training and residency grouping) was entered individually into a regression procedure as a categorical variable and the proportion of positive responses (1 or 2 on the Likert scale) for each level of the variable was compared to the proportion of positive responses in the full sample. Results Survey response was 48% (1642/3430). Of these 44 had not worked in an ambulatory setting and so were eliminated from further analysis (N = 1598). The demographics of the five medical schools are listed in Table 1 . Demographic characteristics of responders are shown in Table 2 . Comparisons to all Ontario and Canadian clerks and residents revealed more women, junior residents, McMaster and Family Medicine residents and fewer PGY6-fellows and Toronto trainees responded. Table 1 Demographics of the five medical schools School City Size # final year medical students (2001–02)* # residents (2001–02)* Queen's University 113,000 71 248 University of Toronto 4,700,000 167 1268 University of Western Ontario 432,000 98 353 University of Ottawa 823,000 85 398 McMaster University 662,000 103 396 * Data from Association of Canadian Medical Colleges Table 2 Demographics of Study Group Demographic Study Numbers N (%) Gender N = 1642 Male 805 (49) Female 837 (51) Level of Training N = 1641 Clerks 279 (17.0) First year residents 377 (23.0) Second year residents 366 (22.3) Third year residents 231 (14.1) Fourth year residents 165 (10.1) Fifth year residents 185 (11.3) Sixth and above year residents including Fellows 38 (2.3) University N = 1642 Queen's U. 172 (10.5) U. of Toronto 611 (37.3) U. of Western Ontario 243 (14.8) Ottawa U. 296 (18.0) McMaster U. 317 (19.3) Mean Age Residents 29.9 Training Program N = 1356,* Medicine 298 (22.0) Family Medicine 351 (25.9) Paediatrics 100 (7.4) Surgery 226 (16.7) Psychiatry 104 (7.7) Radiology 56 (4.1) Intensivists 7 (0.5) Anaesthesia 102 (7.5) Laboratory 24 (1.8) Obstetrics/Gynaecology 65 (4.8) Emergency 23 (1.7) N = total number-clerks 7 not specified The rank ordering of site characteristics and preceptor behaviours, including missing data and number judging an item to not only be unhelpful but detrimental for learning are shown in Tables 3 and 4 . The five most and five least important items for learning essentially matched the rank ordering and thus are not separately reported. Table 3 Ranking of Site Characteristics Rank Question Number saying important to learning (%) Number not answering question Number saying detrimental for learning (%) 1 Effective teachers 1569 (98.4) 4 3 (0.2) 2 Opportunity to see patients independently 1592 (97.3) 6 0 3 Opportunity to see a large variety of patients 1511 (94.8) 4 1 (0.1) 4 Opportunity to see an adequate number of patients 1496 (93.9) 5 2 (0.1) 5 Preceptors readily available 1490 (93.5) 4 2 (0.1) 6 Opportunity to do procedures 1357 (85.3) 8 3 (0.2) 7 Readily available examination room 1348 (84.8) 9 1 (0.1) 8 Opportunity to see patients in follow-up visits 1275 (80.1) 6 1 (0.1) 9 Opportunity to observe preceptor if desired 1239 (77.8) 6 0 10 Opportunity to interact with consultants and/or referring doctors 1227 (77.1) 7 0 11 Block rotation 1094 (68.8) 7 3 (0.2) 12 Efforts to meet objectives made by preceptor 1059 (66.5) 5 1 (0.1) 13 Teaching of medical record keeping skills 956 (60.0) 4 0 14 Computer learning resources available in the clinic 947 (59.4) 3 0 15 Orientation to the practice 937 (59.0) 11 1 (0.1) 16 Teaching of time management skills 904 (56.7) 3 3 (0.2) 17 Teaching of office management skills 872 (54.7) 4 2 (0.1) 18 Clearly defined site objectives for the rotation 843 (52.9) 5 4 (0.2) 19 Library resources available in the clinic 778 (48.8) 4 0 20 Existence of a site-coordinator 762 (48.4) 22 2 (0.1) 21 Longitudinal/horizontal rotation 603 (38.4) 29 42 (2.7) 22 Limited number of preceptors 443 (27.9) 13 233 (14.7) 23 Presence of other trainees in the clinic 432 (27.1) 4 74 (4.6) 24 Close proximity of clinic to campus 366 (23.0) 8 5 (0.3) Table 4 Preceptor Behaviours Ranking Rank Question Number saying important for learning (%) Number not answering question Number saying detrimental for learning (%) 1 Is open to questions 1540 (96.7) 5 1 (0.1) 2 Gives constructive feedback 1522 (95.6) 6 1 (0.1) 3 Demonstrates enthusiasm for teaching 1515 (95.1) 5 1 (0.1) 4 Reviews differential diagnoses 1507 (94.6) 5 0 5 Delegates appropriate responsibility for patient care 1491 (93.7) 7 1 (0.1) 6 Gives timely feedback 1445 (90.7) 5 0 7 Has a strong command of his or her specialty 1433 (90.1) 7 2 (0.1) 8 Discusses clinical topics in an organized way 1416 (88.9) 5 0 9 Makes student feel like a valued member of the practice 1407 (88.3) 5 1 (0.1) 10 Identifies and responds to student's specific learning needs 1398(87.9) 8 1 (0.1) 11 Discusses own clinical reasoning processes 1396 (87.3) 8 1 (0.1) 12 Asks for students' ideas before giving own 1372 (86.1) 5 0 13 Discusses clinical topics concisely 1361 (85.5) 6 1 (0.1) 14 Demonstrates a caring attitude towards students 1347 (84.6) 5 1 (0.1) 15 Sets time aside to discuss topics unable to be discussed during busy clinics 1340 (84.3) 10 4 (0.3) 16 Provides a role model of professional behaviour 1327 (83.5) 8 0 17 Asks students differing complexities of questions 1302 (81.8) 6 3 (0.2) 18 Welcomes differing points of view 1294 (81.3) 7 2 (0.1) 19 Demonstrates a caring attitude towards patients 1279 (80.3) 5 0 20 Facilitates student's participation in follow-up care 1263 (79.4) 7 0 21 Teaches physical examination 1218 (76.8) 13 3 (0.2) 22 Monitors quality of the rotation 1216 (76.4) 6 1 (0.1) 23 Seeks to understand student's ideas 1186 (74.5) 6 0 24 Suggests relevant reading 1172 (73.6) 5 0 25 Connects new ideas to existing knowledge 1149 (72.4) 10 0 26 Defines student's role 1087 (68.4) 8 3 (0.2) 27 Provides a role model of a balance between personal and professional life 1079 (67.9) 9 0 28 Teaches appropriate use of health care resources 1072 (67.4) 8 0 29 Teaches use of community resources 1005 (63.2) 7 0 29 Demonstrates effective interactions with support staff 1005 (63.2) 9 0 30 Observes clinical interactions directly 966 (60.8) 8 7 (0.4) 31 Teaches communication skills 940 (59.2) 10 2 (0.1) 32 Discusses limitations of his or her own knowledge 899 (56.5) 7 1 (0.1) 33 Provides background on patients before students sees patient 602 (37.8) 5 36 (2.3) 34 Outlines specific task(s) to be done during a clinical encounter 595 (37.5) 12 30 (1.9) 35 Teaches in the patient's presence 429 (27.1) 14 116 (7.3) 36 Focuses on one teaching theme per clinic 348 (21.9) 9 71 (4.4) 37 Reviews case in the patient's presence 281 (17.7) 10 242 (15.1) Six factors, accounting for 55 % of the variance for the 24 site characteristics, and 7 factors, accounting for 54% of the variance for the 38 preceptor behaviours, were identified (Tables 5 , 6 ). Labels describing the factors were decided by group consensus among the researchers. 7 items that failed to load on any factor were eliminated from the analysis. The site characteristic factors were office management, patient logistics, objectives, learning resources, clinic set-up and preceptor interaction. The preceptor behaviour factors were professional role modeling, teaching, learning climate, feedback, direction, patient presence and health care system interaction. Cronbach alpha coefficients for the factors identified in the factor analysis ranged from 0.52 to 0.83. Table 5 Factor Analysis makeup for Site Characteristics Factor Items making up factor Factor Loading Alpha Analysis Office Management Teaching of time management skills .832 .62 Teaching of medical record keeping skills .760 Teaching of office management skills .746 Patient Logistics Opportunity to see an adequate number of patients .766 .69 Opportunity to see a large variety of patients .542 Opportunity to see patients independently .538 Readily available examination room .473 Opportunity to see patients in follow-up visits .442 Objectives Clearly defined site objectives for the rotation .806 .53 Efforts to meet objectives made by preceptor .776 Learning Resources Library resources available in the clinic .794 .60 Computer learning resources available in the clinic .756 Clinic Set-up Close proximity of clinic to campus .442 .55 Presence of other trainees in the clinic .418 Existence of a site co-coordinator .386 Longitudinal/horizontal rotation .364 Orientation to the practice .342 Preceptor Interaction Effective teachers .514 .55 Preceptors readily available .506 Opportunity to observe preceptor if desired .491 Table 6 Factor analysis for Preceptor Behaviours Factor Items Making Up Factor Factor Loading Alpha Analysis Professional Role Modeling Provides a role model of professional behaviour .681 .79 Demonstrates effective interactions with support staff .565 Provides a role model of a balance between personal and professional life .557 Teaches communication skills .526 Discusses limitations of his or her own knowledge .500 Discusses own clinical reasoning processes .426 Teaching Discusses clinical topics in an organized way .739 .82 Discusses clinical topics concisely .650 Suggests relevant reading .462 Identifies and responds to student's specific learning needs .390 Is open to questions .365 Asks students differing complexities of questions .362 Has a strong command of his or her area of specialty .340 Asks for students' ideas before giving own .334 Sets time aside to discuss topics unable to be discussed during busy clinics .323 Monitors quality of the rotation .301 Learning Climate Makes student feel like a valued member of the practice .613 .83 Demonstrates a caring attitude towards students .591 Seeks to understand student's ideas .563 Demonstrates a caring attitude towards patients .512 Demonstrates enthusiasm for teaching .365 Welcomes differing points of view .328 Facilitates student's participation in follow-up care .301 Feedback Gives constructive feedback .730 .73 Gives timely feedback .709 Reviews differential diagnosis .473 Direction Outlines specific task(s) to be done during a clinical encounter .588 .69 Focuses on one teaching theme per clinic .507 Provides background on patients before student sees patient .447 Teaches physical examination .432 Defines student's role .404 Patient Presence Teaches in the patient's presence .759 .77 Reviews case in the patient's presence .720 Health Care System Interaction Teaches use of community resources .531 .82 Teaches appropriate use of health care resources .516 Logistic regression analysis of the independent variables revealed striking similarities, but some significant differences, in valued site characteristics and preceptor behaviours for male and female students and those in different schools, at different levels of training and in different residencies (Figures 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 ). Similarities are evident by mainly flat, nonintersecting lines on the graphs indicating similar percentages of subgroups of learners valuing a factor and similar relative valuing of factors respectively. Differences are apparent where lines intersect within a graph and/or percentages are statistically higher (indicated by , , ) or lower (#, ##, ###) than the means. Figure 1 Gender and site characteristics Figure 2 Gender and preceptor behaviours Figure 3 School and site characteristics Figure 4 School and preceptor behaviours Figure 5 Training level and site characteristics Figure 6 Training level and preceptor behaviours Figure 7 Residency and site characteristics Figure 8 Residency and preceptor behaviours Male and female residents rank ordered all site characteristics and preceptor behaviours identically. Women ranked all factors, with the exception of teaching in the patient's presence, higher than men, usually significantly so (Figures 1 , 2 ). Across schools the rank ordering of factors was identical with the exception of Toronto and Ottawa ranking office management instruction higher than the other schools. Toronto valued six factors significantly more than the other schools, Queen's and Western ranked three and two items respectively significantly less than the other schools (Figures 3 , 4 ). Across levels the only difference in rank ordering was clerks ranking preceptor interaction as the most important site characteristic whereas all other groups ranked patient logistics as most important. Those at the beginning and end of their training valued having learning resources available less than all other levels. Clerks were most different from all the other levels in what they valued or did not value (indicated by the number of and # for this group) (Figures 5 , 6 ). Subanalysis of the PGY2 data removing family medicine residents significantly decreased the importance of office management and health care system interaction instruction (52.2% of all PGY2's rated office management instruction important versus 44.8% removing family medicine residents, and 60.1% of all PGY2's rated health care system interaction instruction important versus 50.5% removing family medicine residents). Residency groups again showed mainly similarities in rank ordering, the exceptions being the family medicine/paediatrics/psychiatry group ranking office management and learning climate higher, the lab/path/radiology group patient logistics and learning climate lower and the surgery/emergency/ob/gyn group health care system interaction lower than the rest. There were a large number of responses significantly different from the group averages throughout all the residency groups (Figures 7 , 8 ). Combining residencies into five groups lost only two pieces of information, that of ob/gyn residents being similar to the family medicine, paediatrics, psychiatry group in relatively highly valuing office management instruction and that of anaesthesia residents being similar to lab/path, radiology residents in relatively less valuing feedback, teaching, and learning climate than other groups. Discussion The ambulatory teaching site characteristics most valued by clerks and residents are having an adequate number and variety of patients while being supervised by enthusiastic and available preceptors. These characteristics have been identified before and are well summarized by Bowen and Irby [ 19 ]. Little value is placed on having other trainees in the clinic despite social learning theory that suggests this enhances learning. Bowen [ 20 ] and Lesky [ 21 ] suggest that students learn by teaching and may feel less threatened asking questions that reveal a lack of knowledge of a fellow student than of a preceptor. Although what students value may not translate into effective learning, it is still important to understand why something is valued or not valued. Without reliable learning outcome measures perceived learning value is a proxy measure of learning effectiveness. Further studies should assess what learners do not like about having other trainees present. Computer resources were more valued than books, likely reflecting a generation of learners who are comfortable accessing electronic information. Proximity of the clinic to university campus was unimportant. In contrast to other studies [ 22 - 24 ] we found block rotations were valued more than longitudinal rotations. Some programs, particularly Canadian Family Medicine programs, encourage longitudinal rotations to enhance the continuity of care experience. Merenstein et al [ 25 ] however recently reported there to be no difference in continuity of care provided by residents in longitudinal rotations. Exploration of the value of block versus longitudinal rotations is an area for further research. Valued preceptor behaviours identified in this study are feedback by enthusiastic, open preceptors who are willing to discuss their reasoning processes and delegate responsibility. Recent studies report 3 rd year medical students to also value these preceptor behaviours[ 26 , 27 ]. Lesky and Borkan [ 21 ] suggest that pathogenesis and natural histories of disease can be learned from a variety of resources, including books and computers but problem solving, decision making and dealing with uncertainty are learned mainly from preceptors and practice. This study supports students' perceived value of these aspects and suggests them as priorities for teachers in ambulatory settings. We have confirmed the value of feedback found in most studies [ 26 , 28 - 31 ] (a study by O'Malley [ 32 ] being the exception). As one respondent commented "constructive and honest feedback in a timely manner is by far the most important (item)". Feedback leads to positive learning outcomes. Cope [ 33 ] demonstrated that giving feedback to residents improved their patient satisfaction scores, which in turn has been correlated with improved patient outcomes[ 34 ]. Unfortunately this teaching behaviour is underutilized. Irby [ 1 ], in a review of studies, reports that feedback is given only 3–6% of the time (range 0–16%). This is an effective teaching behaviour that is valued by students and deserves high priority. Meaningful feedback about many aspects of students' patient care is best based on direct observation[ 35 ]. Only 61% of our respondents actually value direct observation by their preceptors. Some of the reasons for more not valuing this may be similar to why students do not want to be taught in front of the patient (see next section). Since direct observation is a necessary component of good teaching it will be important to explore further why more students do not value this important preceptor behaviour. A number of strategies have been suggested to improve efficiency in the ambulatory teaching setting including teaching in the patient's presence and preceptors directing tasks to be covered in the interview [ 36 - 38 ]. A significant proportion of our respondents rated reviewing the case and teaching in the patient's presence, structuring the interview by providing patient information background, outlining tasks to be done during the interview and focusing on one teaching theme per clinic not only to be unimportant for learning but detrimental. Kernan similarly found 3 rd year medical students to not value being taught in front of the patient[ 26 ]. Comments from students in this study give some indication why teaching in front of the patient is disliked ("it would undermine a therapeutic alliance with the student", "it gives a tense atmosphere more often than not", "...impairs free thinking of student because student feels inhibition in front of patients", "makes it difficult for students to ask questions, not wanting to scare/worry the patient"). Teaching however occurs within a larger context where providing background information on patients may be necessary for ongoing patient care and safety and to model continuity of care. Teaching in the patient's presence may be necessary for efficiency and maintaining a relationship between the preceptor and the patient.Further studies are needed to determine if explanation or teaching methods can overcome this aversion. Analysis of the impact of gender, school, level of training or residency on valued site characteristics and preceptor behaviours revealed striking uniformity between the groups. There were some statistically significant differences between the groups, many of which do not appear to be educationally relevant, others which likely are important. Male and female students rank ordered site and preceptor behaviour factors identically. It is of interest that female students ranked all factors, with the exception of teaching in the patient's presence, as being more important for learning than male students. The literature [ 39 - 41 ] suggests that women predominantly emphasize relationship issues, which may partially explain this finding. It would appear however with respect to the items surveyed that there are no gender-based educationally important differences in valued site characteristics and preceptor behaviours. The five schools also essentially rank ordered the factors identically. One school did stand out from the others in frequently ranking factors significantly higher than the rest. This school is the largest of the five schools with the most trainees and teaching sites. It would be valuable to know the ratio of students to preceptors at the different schools. If this were high at the larger school, perhaps resulting in residents feeling relatively anonymous, it may partially explain why these students there particularly value factors like learning climate, professional role modeling and clinic set up. Within level of training preceptor interaction is most important for clerks. This is the only group to rank this item more important than patient logistics. This may reflect the clerks' developmental stage of being eager to go beyond textbook lists and start to put clinical decisions into patient context–skills best learned by preceptor interaction. Learning resources are significantly less valued by those at either end of their training–clerks for perhaps the above reason and PGY6's/fellows presumably because they are confident in their theoretical knowledge. Beyond clerkship patient logistic factors usurp preceptor interaction as the highest ranked site characteristic. Becoming an expert clinician involves, in part, connecting disparate units of knowledge into networks [ 3 , 5 , 42 ]. This encapsulating of knowledge occurs when students learn with patients. The residents in this study recognize this, ranking seeing an adequate and large variety of patients independently as the most important site characteristic for their learning. Having objectives defined with efforts made to meet them was third in importance for most levels, superceding available learning resources, office management skills instruction, and clinic setup items. Office management instruction is relatively more important for PGY2's and those at the end of their training. Subanalysis of the PGY2 data removing family medicine residents who would be at the end of their training and leaving those in the middle of their training significantly decreased the importance of office management and health care system interaction instruction. Teaching these aspects thus seems most important for those at the end of their training. Directing the clinical encounter and teaching in the patient's presence is valued less as residents gain seniority and presumably identify themselves more as the patients' physicians. Increasing desire for autonomy and decreasing potential for undermining their relationship with the patient may be reasons for these trends. Within almost all residencies patient logistics and preceptor interaction are the most valued site characteristics; feedback, teaching and learning climate the most important preceptor behaviours. Lab/path, radiology and anaesthesia residents value all these preceptor behaviours less than other residents. Arguably these are areas of medicine where decision making is more clear-cut without as much patient input, which may explain these results. Other significant differences between the specialties seem best explained by considering future practice ie: office-based specialties (paediatrics, psychiatry, family medicine) most valuing office management and health care system interaction instruction. Strengths of this study are the large multi-institutional sample size (n = 1642) encompassing students at multiple levels in all specialties. The response rate (48%) limits the external validity of the result. A confounding factor within the level of training data set may be the variability in residency lengths as suggested by the subanalysis of the PGY2 data. Rather than years from graduation from medical school what seems to influence valued site characteristics and preceptor behaviours more are years from independent practice. Conclusions "Software" (patient encounters and enthusiastic preceptors who delegate, give feedback and explain clinical reasoning) is valued more than "hardware" (clinic set-up, learning resources). All learners value the above preceptor behaviours; most do not value, and a significant number consider detrimental, having the structure of the patient encounter dictated to them and having the patient present during review and teaching. Future work is needed to explain why learners do not value these practices. Learners at all levels and in all specialties are strikingly similar in what they value from their preceptors and clinic sites for their learning. There are some differences between levels and residencies however that require consideration when teaching these different groups. Educationally significant differences within levels include preceptor interaction being paramount for medical students; patient logistics (adequate number and variety of patients seen independently and in follow-up) being second. The reverse is true for residents. Proportioning time accordingly deserves attention. The more senior the learner the more being taught or having the case reviewed in the patients' presence is not valued. Sensitivity to the patient-learner relationship is required if these practices are utilized but particularly so for more senior learners. Finally relevance not surprisingly dictates importance. Office management instruction is valued by those at the end of their training and those primarily in office-based specialties. Similarly office-based specialties appreciate instruction in health care system interaction. This study identifies preceptor behaviours and site characteristics valued by medical students and residents for their learning in the ambulatory setting. Further studies are needed to determine the effect of providing these valued site characteristics and preceptor behaviours on learning outcomes. Competing interests None declared. Authors' contributions KS conceived of the study, prepared the manuscript and participated in the conceptual planning and design of the study and data interpretation. JK, DD, and MG participated in the conceptual planning and design of the study, statistical analysis and data interpretation and manuscript revision. SV and RB contributed to the design of the study and manuscript revision. CK contributed to the design of the study. All authors read and approved the final manuscript. RS participated in the statistical analysis. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514563.xml
546225	Effects of the cyclooxygenase-2 inhibitor nimesulide on cerebral infarction and neurological deficits induced by permanent middle cerebral artery occlusion in the rat	Background Previous studies suggest that the cyclooxygenase-2 (COX-2) inhibitor nimesulide has a remarkable protective effect against different types of brain injury including ischemia. Since there are no reports on the effects of nimesulide on permanent ischemic stroke and because most cases of human stroke are caused by permanent occlusion of cerebral arteries, the present study was conducted to assess the neuroprotective efficacy of nimesulide on the cerebral infarction and neurological deficits induced by permanent middle cerebral artery occlusion (pMCAO) in the rat. Methods Ischemia was induced by permanent occlusion of the middle cerebral artery in rats, via surgical insertion of a nylon filament into the internal carotid artery. Infarct volumes (cortical, subcortical and total) and functional recovery, assessed by neurological score evaluation and rotarod performance test, were performed 24 h after pMCAO. In initial experiments, different doses of nimesulide (3, 6 and 12 mg/kg; i.p) or vehicle were administered 30 min before pMCAO and again at 6, 12 and 18 h after stroke. In later experiments we investigated the therapeutic time window of protection of nimesulide by delaying its first administration 0.5–4 h after the ischemic insult. Results Repeated treatments with nimesulide dose-dependently reduced cortical, subcortical and total infarct volumes as well as the neurological deficits and motor impairment resulting from permanent ischemic stroke, but only the administration of the highest dose (12 mg/kg) was able to significantly (P < 0.01) diminish infarct volume. The lower doses failed to significantly reduce infarction but showed a beneficial effect on neurological function. Nimesulide (12 mg/kg) not only reduced infarct volume but also enhanced functional recovery when the first treatment was given up to 2 h after stroke. Conclusions These data show that nimesulide protects against permanent focal cerebral ischemia, even with a 2 h post-treatment delay. These findings have important implications for the therapeutic potential of using COX-2 inhibitors in the treatment of stroke.	Background The brain is highly sensitive to disturbance of its blood supply. Stroke is a devastating disease and is the third most common cause of death, and the most common cause of motor and mental disability in adults, in developing countries [ 1 ]. Complex pathophysiological events occur in brain during ischemic processes, and these are considered responsible for cell damage leading to neuronal death (for review see [ 2 , 3 ]). However, it is now generally accepted that the mammalian brain may be more resistant to ischemia than previously thought. This raises the possibility of therapeutic intervention before brain damage has become irreversible. A number of interacting and sequentially evoked events tend to reinforce the initial ischemic insult. A key role in these processes is played by post-ischemic inflammation. The Ca 2+ -related activation of intracellular second messenger systems, the increase in reactive oxygen species formation, as well as hypoxia itself triggers the expression of a large number of pro-inflammatory genes following cerebral ischemia. Thus, mediators of inflammation such as platelet-activating factor (PAF), tumor necrosis factor α (TNFα), interleukin 1β (IL-1β), chemokines (IL-8, monocyte chemoattractant protein-1) and other pro-inflammatory factors are produced by the ischemic brain tissue [ 3 ]. In addition, the expression of adhesion molecules with the subsequent infiltration of polymorphonuclear leukocytes occurs following ischemic stroke. Results from several studies also suggest that the marked and sustained expression of inflammation-related enzymes such as inducible nitric oxide synthase (iNOS) and cyclooxygenase-2 (COX-2) plays an important role in the secondary events that amplify cerebral injury after ischemia [ 4 - 12 ]. Nimesulide (N-(4-nitro-2-phenoxyphenyl)-methanesulfonamide) is a non-steroidal anti-inflammatory drug with potent effects. It shows a high affinity and selectivity for COX-2 with a COX-2/COX-1 IC 50 selectivity ratio of 0.06 (whole blood assay) [ 13 ]. Nimesulide readily crosses the intact blood-brain barrier in both humans and rodents [ 13 , 14 ]. Several recent studies have demonstrated a marked neuroprotective effect of nimesulide on chronic cerebral hypoperfusion [ 15 ], kainate-induced excitotoxicity [ 16 ], quisqualic acid-induced neurodegeneration [ 17 ], diffuse traumatic brain injury [ 18 , 19 ], glutamate-mediated apoptotic damage [ 20 ] and induction of the expression of the B subunit of endogenous complement component C1q (C1qB) in transgenic mice with neuronal overexpression of human COX-2 [ 21 ]. Recently, we have found a significant neuroprotective effect of nimesulide both in global cerebral ischemia [ 10 , 22 ], a type of injury that mimics the clinical situation of cardio-respiratory arrest, and in a rat model of ischemic stroke induced by the transient (1 h) occlusion of the middle cerebral artery [ 12 ]. Since most cases of human ischemic stroke are caused by permanent occlusion of cerebral arteries [ 23 - 26 ], the present study was conducted to assess whether nimesulide would also show neuroprotective efficacy on the cerebral infarction induced by permanent middle cerebral artery occlusion (pMCAO) in the rat, a clinically relevant model of ischemic stroke. The effects of the COX-2 inhibitor nimesulide had not been previously investigated in a model of permanent ischemic stroke. Methods Animals Male Sprague-Dawley rats (CENPALAB, Havana, Cuba) weighing 280–340 g at the time of surgery were used in the present study. Our institutional animal care and use committee approved the experimental protocol (No. 02/67). The animals were quarantined for at least 7 days before the experiment. Animals were housed in groups in a room whose environment was maintained at 21–25°C, 45–50 % humidity and 12-h light/dark cycle. They had free access to pellet chow and water. Animal housing, care, and application of experimental procedures were in accordance with institutional guidelines under approved protocols. Induction of permanent focal cerebral ischemia in the rat Rats were anesthetized with chloral hydrate (300 mg/kg body weight, i.p.). Once surgical levels of anesthesia were attained (assessed by absence of hind leg withdrawal to pinch), ischemia was induced by using an occluding intraluminal suture as described previously [ 27 - 29 ]. Briefly, the right common carotid artery (CCA) was exposed by a ventral midline neck incision and ligated with a 3-0 silk suture. The pterygopalatine branch of the internal carotid artery was clipped to prevent incorrect insertion of the occluder filament. Arteriotomy was performed in the CCA approximately 3 mm proximal to the bifurcation and a 3-0 monofilament nylon suture, whose tip had been rounded by being heated near a flame was introduced into the internal carotid artery (ICA) until a mild resistance was felt (18–19 mm). Mild resistance to this advancement indicated that the intraluminal occluder had entered the anterior cerebral artery and occluded the origin of the anterior cerebral artery, the middle cerebral artery (MCA) and posterior communicating arteries [ 27 ]. After the advancement of the nylon suture, the ICA was firmly ligated with a 3-0 silk suture. The incision was closed and the occluding suture was left in place until sacrificing the animals. The duration of surgery did not exceed 12 min in any case. The animals were allowed to recover from anesthesia and to eat and drink freely. The body temperature was strictly controlled during and after ischemia. To allow for better postoperative recovery, we chose not to monitor physiological parameters in the present study because additional surgical procedures are needed for this monitoring. Nevertheless, we performed separate experiments to investigate the effects of nimesulide on major physiological variables such as mean arterial blood pressure, blood glucose, rectal temperature, hematocrit, blood pH and blood gases (pO 2 and pCO 2 ). The effects observed with nimesulide in the present study were not related to modification of physiological variables since these parameters did not differ between nimesulide-treated and vehicle-treated rats (data not shown). These findings are in agreement with our previous results [ 10 , 12 ], suggesting that nimesulide does not significantly change major physiological variables. Neurological evaluation An unaware independent observer performed the neurological evaluations prior to the sacrifice of the animals according to a six-point scale: 0= no neurological deficits, 1= failure to extend left forepaw fully, 2= circling to the left, 3= falling to left, 4= no spontaneous walking with a depressed level of consciousness, 5= death [ 30 , 31 ]. Assessment of functional outcome Motor impairment in this study was assessed with the use of the accelerating rotarod (Ugo Basile, Varese, Italy, Model 7750). Rats were given 2 training sessions 10 minutes apart before surgery. Rats were first habituated to the stationary rod. After habituation they were exposed to the rotating rod. The rod was started at 2 rpm and accelerated linearly to 20 rpm within 300 sec. Latency to fall off the rotarod was then determined before ischemia (presurgery) and before sacrificing the animals. Animals were required to stay on the accelerating rod for a minimum of 30 sec. If they were unable to reach this criterion, the trial was repeated for a maximum of five times. The two best (largest) fall latency values a rat could achieve then were averaged and used for data analysis. Rats not falling off within 5 min were given a maximum score of 300 seconds [ 32 , 33 ]. A sham-operated group was also included (n = 8). The investigator performing the rotarod test did not know the identity of the experimental groups until completion of data analysis. Quantification of brain infarct volume The method for quantification of infarct volume was performed exactly as reported by others [ 34 , 35 ]. Briefly, the animals were sacrificed under deep anesthesia and brains were removed, frozen, and coronally sectioned into six 2-mm-thick slices (from rostral to caudal, first to sixth). The brain slices were incubated for 30 min in a 2% solution of 2,3,5-triphenyltetrazolium chloride (TTC) (Sigma Chemical Co.) at 37°C and fixed by immersion in a 10% phosphate-buffered formalin solution. Six TTC-stained brain sections per animal were placed directly on the scanning screen of a color flatbed scanner (Hewlett Packard HP Scanjet 5370 C) within 7 days. Following image acquisition, the image were analyzed blindly using a commercial image processing software program (Photoshop, version 7.0, Adobe Systems; Mountain View, CA). Measurements were made by manually outlining the margins of infarcted areas. The unstained area of the fixed brain section was defined as infarcted. Cortical and subcortical uncorrected infarcted areas and total hemispheric areas were calculated separately for each coronal slice. Total cortical and subcortical uncorrected infarct volumes were calculated by multiplying the infarcted area by the slice thickness and summing the volume of the six slices. A corrected infarct volume was calculated to compensate for the effect of brain edema. An edema index was calculated by dividing the total volume of the hemisphere ipsilateral to pMCAO by the total volume of the contralateral hemisphere. The actual infarct volume adjusted for edema was calculated by dividing the infarct volume by the edema index [ 36 - 38 ]. Infarct volumes are expressed as a percentage of the contralateral (control) hemisphere. The investigators who performed the image analysis were blinded to the study groups. Experimental design Time course of lesion development after pMCAO At various times after pMCAO (4, 8, 12, 24 and 48 h, n = 6–8 per group) the animals were sacrificed and the brains were quickly removed, sectioned and stained as previously described in order to calculate the infarct volume. Evaluation of nimesulide's effects: dose-response experiment In order to evaluate the effect of nimesulide administration on rat focal cerebral ischemia, three different doses of nimesulide (3, 6 and 12 mg/kg) were given to rats by intraperitoneal administration 30 min before the onset of pMCAO (n = 7–9 animals per group). Additional doses were given at 6, 12 and 18 h after stroke. This treatment schedule and dosage range was based on the pharmacokinetic profile of nimesulide [ 39 ] and on our previous experience with this compound in models of cerebral ischemia [ 10 , 12 ]. We also studied the effect of a single dose of nimesulide (12 mg/kg; i.p.) given 30 min before ischemia (n = 8). A single injection vehicle-treated group was also included (n = 7). Assessment of the therapeutic time window for the neuroprotective effect of nimesulide in pMCAO After investigating the dose-response relationship, we studied the effect of nimesulide (12 mg/kg; i.p.) when administered 0.5, 1, 2, 3 or 4 h after ischemia (n = 8–11 animals per group). The corresponding vehicle-treated groups were included as controls (n = 7–10 rats per group). Three additional doses were given every 6 h after the first treatment with nimesulide or vehicle exactly as described before for the repeated treatment schedule in the dose-response experiment. After completing the neurological evaluation and rotarod performance at 24 h after permanent focal cerebral ischemia, animals were sacrificed and the brains were removed to calculate the infarct size. Data analysis Data are presented as means ± S.D. Values were compared using t-test (two groups) or one-way ANOVA with post-hoc Student-Newman-Keuls test (multiple comparison). Neurological deficit scores were analyzed by Kruskal-Wallis non-parametric ANOVA followed by the Dunn test (multiple comparison) or Mann-Whitney test for analysis of individual differences. Rotarod performance was expressed as a percentage of pre-surgery values for each rat and analyzed by ANOVA for repeated measures followed by the Student-Newman-Keuls test. Differences were considered significant when p < 0.05. Results Time course of the development of cerebral infarction and neurological deficits after pMCAO The temporal evolution of the lesion volumes is presented in Fig. 1A as the cortical and subcortical components of the infarction. Subcortical injury was evident in TTC-stained coronal sections as early as 4 h after permanent stroke (see insets of TTC-stained sections at different times after stroke in Fig. 1A ). Subcortical lesion was maximal between 8 and 12 h after pMCAO, although there was a slight but significant increase between 8 and 24 h when the overall comparison was performed (one-way ANOVA, followed by Student-Newman-Keuls test). Nevertheless, the Student's t-test analysis failed to detect any significant increase between 12 and 24 or 48 h post-injury, thus indicating that the subcortical damage reached maximal values by 12 h after the insertion of the occluding filament (Fig. 1A ). On the other hand, cortical damage progressed more slowly; it was detected at 4 h after pMCAO, and by 8 h there was an increase of the infarct but this was not statistically significant as compared to that at 4 h. On the contrary, there was a significant (p < 0.05) increase in the lesion when the infarction at 12 h is compared with that at 4 or 8 h, and a more dramatic increase of damage is seen at 24–48 h after stroke, a time at which the cortical infarct volume is maximal in this model as shown in Fig. 1A . Figure 1 Temporal development of focal cerebral infarction induced by permanent middle cerebral artery occlusion (pMCAO). ( A ): Evolution of cortical and subcortical infarct volumes after pMCAO in rats. Representative TTC-stained sections at different times after stroke are shown in the insets. ( B ) and ( C ): Time course of the increase of neurological deficits and motor impairment induced by pMCAO. Infarct volumes are expressed as a percentage of the contralateral (control) hemisphere. Bars represent the group mean ± SD. * p < 0.05 with respect to subcortical infarct volume at 4 h. & p < 0.05 with respect to subcortical infarct volume at 8 h. # p < 0.05 with respect to cortical infarct volume at 8 h. p < 0.05 with respect to cortical infarct volume at 12 h. § p < 0.05 with respect to 4 and 8 h. The horizontal bar in Panel B shows the median neurological score. With regard to the neurological deficits and motor impairment induced by pMCAO (assessed by the neurological score and accelerating rotarod test), it is important to emphasize the fact that these parameters were maximal by 12 h after stroke and the animals did not show any further increase in the neurological deficits or motor impairment after 24 or 48 h of the occlusion, as depicted in Fig. 1B and Fig. 1C . Based on these results, we decided to evaluate the effects of nimesulide after 24 h of pMCAO. Effects of different doses of nimesulide on infarct volume and functional outcome after pMCAO Repeated treatments with nimesulide dose-dependently reduced cortical, subcortical and total infarct volumes in the permanent model of stroke, although only the administration of the highest dose (12 mg/kg) was able to significantly (P < 0.01) diminish brain damage (Table 1 ). There was a trend towards a reduction in lesion volumes in animals treated with nimesulide 6 mg/kg, but this effect was not confirmed by the statistical analysis of the data. Unlike the long-term treatment paradigm, the administration of a single dose of nimesulide (12 mg/kg) 30 min before pMCAO failed to significantly reduce total infarct volume, though a modest neuroprotective effect was seen in the subcortical areas as shown in Table 1 . Table 1 Effect of different doses of the cyclooxygenase-2 inhibitor nimesulide on total, cortical and subcortical infarct volumes in a rat model of permanent focal cerebral ischemia. Treatment Total infarct volume (%) Cortical infarct volume (%) Subcortical infarct volume (%) Repeated doses Vehicle (n = 9) 56.1 ± 11.4 41.6 ± 10.3 12.6 ± 4.5 Nimesulide 3 mg/kg (n = 7) 54.9 ± 14.9 38.1 ± 17.2 14.3 ± 2.4 Nimesulide 6 mg/kg (n = 8) 41.4 ± 12.3 31.8 ± 9.3 9.7 ± 3.5 Nimesulide 12 mg/kg (n = 9) 34.1 ± 13.8 24.6 ± 11.2 ** 7.1 ± 3.9 * Single dose Vehicle, single dose (n = 7) 55.2 ± 15.5 43.9 ± 10.9 13.2 ± 4.2 Nimesulide 12 mg/kg, single dose (n = 8) 49.5 ± 11.7 39.4 ± 11.8 9.1 ± 3.1 & Data are mean ± S.D. * P < 0.05 and ** P < 0.01 compared to vehicle. One-way ANOVA followed by Student-Newman-Keuls post-hoc test. & P < 0.05 compared to vehicle single dose (Student's t-test). Interestingly, repeated treatments with 6 and 12 mg/kg of nimesulide were similarly effective in reducing the neurological deficits and the motor impairment resulting from pMCAO (Table 2 ). This effect was not accompanied by a significant reduction in infarct volume in the case of the dose of 6 mg/kg (Table 1 ). No neuroprotective effect of nimesulide was observed on the neurological score or rotarod performance when this COX-2 inhibitor was administered as a single dose (12 mg/kg) before the onset of ischemia (Table 2 ). Table 2 Effect of different doses of nimesulide on neurological deficits and functional outcome (evaluated using the rotarod test) following permanent middle cerebral artery occlusion in the rat. Treatment Neurological Score Rotarod performance (% of presurgery levels) Sham-operated control (n = 8) 0 128 ± 21 Repeated doses Vehicle (n = 9) 3 (3–5) 49 ± 18 Nimesulide 3 mg/kg (n = 7) 3 (2–4) 64 ± 13 Nimesulide 6 mg/kg (n = 8) 2 (1–5) * 89 ± 20 Nimesulide 12 mg/kg (n = 9) 2 (1–4) 84 ± 14 ** Single dose Vehicle, single dose (n = 7) 3 (3–5) 43 ± 21 Nimesulide 12 mg/kg, single dose (n = 8) 3.5 (2–5) 52 ± 19 Values show the median and range (neurological score) and means ± S.D. (rotarod performance). For the analysis of neurological score data, the Kruskal-Wallis nonparametric ANOVA followed by Dunn test (multiple comparison) or Mann-Whitney test for analysis of individual differences were used. For the statistical analysis of rotarod performance results, ANOVA followed by Student-Newman-Keuls post-hoc test was employed. * P < 0.05 and ** P < 0.01 compared to vehicle. Therapeutic time window for nimesulide protection in rats subjected to pMCAO In this experiment we investigated the effect of nimesulide (12 mg/kg) in a situation in which its first administration was delayed for 0.5–4 h after the ischemic challenge. A significant reduction in subcortical infarct volume was observed when the treatment was delayed until 0.5–1 h after pMCAO, but this protective effect of nimesulide was not evident when administered after 2–4 h of the onset of permanent occlusion (Fig. 2A ). In the case of cortical infarction, nimesulide diminished lesion volume when treatment was delayed until 2 h after the ischemic insult (Fig. 2B ). Similar results were found for total infarct volume as shown in Fig. 2C , though as expected, an overall decline of the neuroprotective effect with post-treatment time was observed. Figure 2 Reduction of subcortical ( A ), cortical ( B ) and total ( C ) infarct volumes by the cyclooxygenase-2 inhibitor nimesulide (12 mg/kg; i.p.) when its first administration was delayed for several hours after the onset of permanent stroke. Nimesulide reduced the infarct size in animals treated at 0.5 (n = 8), 1 (n = 9) and 2 h (n = 9), but not at 3 (n = 11) and 4 h (n = 9) after pMCAO, compared to vehicle-treated and time-comparable control groups (n = 7–9 per group). Infarct volumes are expressed as a percentage of the contralateral (control) hemisphere and the data are represented as the mean ± SD. * p < 0.05 and p < 0.01 with respect to vehicle (Student's t-test). Of interest is the finding that nimesulide not only reduced infarct volume but also enhanced functional recovery when the first treatment is given 2 h after permanent ischemic stroke. Post-ischemic treatment with nimesulide significantly reduced neurological deficits and increased the fall latencies to remain on the accelerating rotarod as compared to those rats given only the vehicle (Table 3 ). However, this protective effect was lost when the first administration is delayed until 3–4 h after the occlusion of the middle cerebral artery as presented in Table 3 . Table 3 Effect of delayed administration of nimesulide (12 mg/kg; i.p.) on neurological deficit score and rotarod performance after permanent middle cerebral artery occlusion (pMCAO) in rats. Vehicle or nimesulide was administered 0.5, 1, 2, 3, or 4 h after stroke. Neurological Score Rotarod performance (%) Time after stroke (h) Vehicle Nimesulide Vehicle Nimesulide 0.5 3 (2–5) 2 (1–3) 44 ± 17 81 ± 18 1 3 (2–5) 2 (1–4) 40 ± 21 85 ± 22 ** 2 3 (3–5) 2 (1–5) * 50 ± 11 73 ± 13 * 3 3 (2–5) 3 (2–5) 47 ± 16 60 ± 15 4 3.5 (2–5) 3 (2–5) 52 ± 23 59 ± 14 Values represent the median and range (neurological score) and means ± S.D. (rotarod performance). P < 0.05 and *P < 0.01 compared with the corresponding vehicle-treated group. Discussion The present study was prompted by our previous encouraging results with nimesulide in a model of transient focal cerebral ischemia, which show that this COX-2 inhibitor is able to potently reduce infarct volume and improve functional recovery [ 12 ]. These neuroprotective effects are also observed when treatment is delayed until even 24 h after the onset of ischemia [ 12 ]. Since we believe that it is very important to perform thorough, multifactorial and well-designed pre-clinical studies before assuming definitive conclusions on the neuroprotective effect of any compound, and considering that in stroke patients a very early spontaneous recanalization of an obstructed brain vessel is, unfortunately, only rarely found, we conducted the present investigation to shed more light into the effects of nimesulide on ischemic damage using a permanent stroke model in the rat considering that this model might be more relevant to the clinical situation of stroke, as suggested previously [ 23 - 26 ]. The core findings of this study are: (i) administration of clinically relevant doses of nimesulide confers protection against the damage induced by permanent focal cerebral ischemia in two modalities (reduction of infarct size, and improvement of functional outcome) and (ii) nimesulide's neuroprotection is still evident when the first administration is delayed until 2 h after the onset of stroke. Depending on the experimental conditions, the temporal evolution of ischemic damage may vary considerably [ 30 , 40 , 41 ]. Thus, it is very important to characterize the time course of brain damage, especially if one wants to interpret correctly the effects of a given compound using delayed treatment schedules. Our results showed that in the permanent model of stroke induced by the occlusion of the middle cerebral artery using an intraluminal suture, the infarct size progresses very fast in the subcortical areas (mainly striatum) and much slower in cortical areas, but in general the evolution of damage is relatively quick, reaching maximal values by 24 h after the insertion of the filament (Fig. 1A ). These findings are in line with those published previously in this model of stroke [ 42 , 43 ]. Although infarct size continues to increase between 12 and 24–48 h of ischemia (Fig. 1A ), the neurological deficits and motor impairment reached their maximum by 12 h, and the animals did not showed any further deterioration of their neurological functions (Fig. 1B and 1C ). This might reflect the fact that unlike ischemic injury to many other tissues, the severity of disability is not predicted well by the amount of brain tissue lost. For example, damage to a small area in the medial temporal lobe may lead to severe disability, while damage to a greater volume elsewhere has little effect on function [ 2 ]. There is not always a direct correlation between the lesion size and the severity of neurological deficits as demonstrated before in animal models [ 29 , 44 ] and in stroke patients [ 45 ]. For that reason, it is essential to evaluate the neuroprotective effects of agents by combining both histological and functional measures. The present study offers a good example of this: even when the lowest doses of nimesulide did not reduce infarct volume in pMCAO (Table 1 ), one can not minimize the beneficial effects of these doses since a significant reduction in neurological deficits and an improvement of rotarod performance were observed (Table 2 ). Thus, further studies would be required to better characterize the effects of the lowest doses of nimesulide (3 and 6 mg/kg) in models of cerebral ischemia. Repeated treatments with nimesulide afforded a more remarkable neuroprotection than the administration of a single dose given before the insult (Tables 1 and 2 ). These data show the importance of continuous long-term administration after ischemic damage in clinical trials to achieve the maximal beneficial effects of neuroprotection by nimesulide. Unfortunately, a large number of promising neuroprotective compounds identified from preclinical experiments have failed in clinical trials in stroke patients [ 3 , 45 - 47 ]. Although several factors may contribute to these disappointing results, an important issue is the 'therapeutic time window of protection', defined as the time period after the onset of ischemia during which administration of treatment is effective [ 48 , 49 ]. Most of the agents that confer protection in experimental animal models of stroke when given before or a short period after cerebral ischemia have failed in clinical studies [ 45 , 50 ]. Thus, the assessment of the therapeutic time window of protection is of paramount importance in pursuing future therapies to treat stroke victims. Therefore, our next experiments were conducted to evaluate the effects of nimesulide when administered in a delayed treatment schedule in order to establish the therapeutic time window of protection of this COX-2 inhibitor in pMCAO, thus increasing predictive outcome in the clinic. Interestingly, reduction in infarct size and neurological deficits and improvement of rotarod performance were still observed when nimesulide treatment was delayed until 2 h after ischemia (Fig. 2A , Table 3 ). It is important to compare our present results in pMCAO with those previously obtained in transient ischemia [ 12 ]. In the model of transient focal ischemia, the time window of nimesulide's neuroprotection extends over a 24 h period [ 12 ], and in other models of cerebral ischemia, the time window of protection of nimesulide is similarly wide [ 10 , 22 , 51 ]. These results have been also obtained with other COX-2 inhibitors (e.g., NS-398, SC58125 and rofecoxib) in models of transient ischemic stroke [ 4 , 52 ] and global cerebral ischemia [ 53 , 54 ]. These studies suggest that although the protective effects of COX-2 inhibitors are more beneficial when administered early after the ischemic insult, COX-2 selective inhibitors show a wide therapeutic window for the prevention of neuronal death in both focal and global ischemia. However, our present results suggest that in permanent stroke, COX-2 inhibition by nimesulide is not as protective as in transient models (39 % of infarct reduction with pre-treatment in pMCAO vs. 60 % lesion reduction in transient ischemia with immediate treatment) and the therapeutic time window is narrower as compared to temporary occlusion models (2 h in pMCAO vs. 24 h in transient ischemia) as the present results (Tables 1 and 3 , Fig. 2 ) and our recent studies indicate [ 12 ]. The vascular inaccessibility of nimesulide into the ischemic/infarcted region could be a plausible explanation for these findings considering that, unlike transient ischemia, in permanent ischemic stroke the protective effects of any drug/agent depend, in large part, on the ability of the compound to reach the ischemic areas mainly through passive diffusion. Although these findings on nimesulide's effects on stroke tempted us to conclude that COX-2 selective inhibitors are less protective in permanent than in transient stroke models, these results should be interpreted with caution since a structurally similar COX-2 inhibitor (NS-398) reduced permanent stroke damage in mice when the treatment started 24 h after MCA occlusion [ 55 ] and the same COX-2 inhibitor also reduced lesion size when administered starting 6 h after pMCAO in another previous study [ 5 ]. Apparent discrepancies between our present results and these two reports [ 5 , 55 ] might be due to different methods to induce pMCAO (intraluminal vs. distal MCAO involving craniectomy), the specific COX-2 inhibitor used, treatment paradigm or animal species. This emphasizes the importance of conducting more preclinical studies with COX-2 selective inhibitors before these agents could be used in clinical trials in stroke patients. Another issue that needs urgent consideration in future studies with COX-2 inhibitors in cerebral ischemia is the effect of long-term treatment since anti-inflammatory interventions could interfere with nervous regeneration/plasticity and recovery as demonstrated in some types of neuronal injury [ 56 , 57 ]. Conclusion In summary, the present study has evaluated for the first time the neuroprotective effects of the COX-2 inhibitor nimesulide in permanent focal cerebral ischemia, showing beneficial effects on reduction of infarct volume and improvement of functional recovery. This ability of nimesulide to diminish permanent ischemic damage is observed even when the first treatment was delayed 2 h after the ischemic episode. Taken together, these results have important implications for the therapeutic potential of using the COX-2 selective inhibitor nimesulide in the treatment of cerebral ischemia. List of abbreviations used COX-2, cyclooxygenase-2; pMCAO, permanent middle cerebral artery occlusion; MCA, middle cerebral artery; TTC, 2,3,5-triphenyltetrazolium chloride; ANOVA, analysis of variance Competing interests The author(s) declare that they have no competing interests. Authors' contributions ECJ carried out the surgical procedures to induce stroke, participated in the design of the study and in the statistical analysis, reviewed the data and drafted the manuscript. NHM performed the evaluation of neurological deficits and rotarod performance. AGF and MGC performed the calculation of the infarct volumes and participated in the statistical analysis of the data. OSL, EM and BLF participated in the design and coordination of the study, reviewed the data, provided consultation and helped to draft the manuscript. OSL and BLF share senior authorship. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546225.xml
535552	Use of a highly sensitive two-dimensional luminescence imaging system to monitor endogenous bioluminescence in plant leaves	Background All living organisms emit spontaneous low-level bioluminescence, which can be increased in response to stress. Methods for imaging this ultra-weak luminescence have previously been limited by the sensitivity of the detection systems used. Results We developed a novel configuration of a cooled charge-coupled device (CCD) for 2-dimensional imaging of light emission from biological material. In this study, we imaged photon emission from plant leaves. The equipment allowed short integration times for image acquisition, providing high resolution spatial and temporal information on bioluminescence. We were able to carry out time course imaging of both delayed chlorophyll fluorescence from whole leaves, and of low level wound-induced luminescence that we showed to be localised to sites of tissue damage. We found that wound-induced luminescence was chlorophyll-dependent and was enhanced at higher temperatures. Conclusions The data gathered on plant bioluminescence illustrate that the equipment described here represents an improvement in 2-dimensional luminescence imaging technology. Using this system, we identify chlorophyll as the origin of wound-induced luminescence from leaves.	Background It is well documented that essentially all living systems spontaneously generate and emit very low levels of light (reviewed in [ 1 ]). This ultra-weak bioluminescence is generally characterized by emission of photons (sometimes termed "biophotons") at an intensity less than 10 -14 W.cm -2 (< 1000 photons.sec -1 .cm -2 ). This is in contrast to the more widely known bioluminescence from animals that visibly glow, such as certain species of jellyfish, fireflies and beetles, which employ fluorescent proteins and/or luciferase enzymes to catalyse reactions that result in chemiluminescence. Ultra-weak bioluminescence is generally considered to result from oxidative chemistry occurring within cells [ 1 , 2 ], though in most cases, the source of observed emissions has not been identified. Nevertheless, changes in emission levels in response to stress – particularly oxidative stress – have been observed in many systems, including bacteria [ 3 ], plants ( e.g . [ 2 , 4 - 6 ]) and animals [ 7 , 8 ]. For this reason, measurement of ultra-weak bioluminescence may be a useful non-invasive technique to monitor rapid perturbations to cellular activity and for early detection of diseased and damaged cells. In plants, increases in spontaneous low-level luminescence have been observed in response to pathogen infection [ 5 ], salt stress [ 9 ], osmotic stress [ 10 ] and mechanical damage or wounding [ 2 , 4 , 6 , 11 ]. It is suggested that luminescence is produced by singlet oxygen and excited carbonyl species generated as a result of lipid peroxidation reactions [ 2 , 11 ]. Lipid peroxidation in wounded and pathogen-infected plant tissues is a common consequence of the generation of ROS, which also act as signals to induce plant defence responses [ 12 ]. Quantitative measurements of such ultra-weak photon emissions are normally obtained using sensitive photomultiplier tubes as photon counting devices. However, more recently, 2-dimensional photomultiplier tubes and cooled charge-coupled device (CCD) cameras have also been employed as a means of imaging the spatial distribution of light emission from diseased and damaged plant tissues [ 2 - 6 , 9 ]. The clearest images were obtained by Chen et al ., [ 2 ], using a micro-channel plate coupled to a cooled CCD to image light captured through a lens. However, the acquisition time required for what is a relatively weak signal, was 1 hour, preventing a detailed temporal investigation of stress-induced bioluminescence. Here, we present a novel configuration of a cooled CCD that enables high sensitivity, high resolution 2-dimensional imaging with short integration times. We demonstrate the utility of the system to image and quantify delayed chlorophyll fluorescence and wound-induced luminescence from plant leaves. Results and Discussion Construction of a luminescence imager The experimental imaging set-up, shown in Figure 1 , essentially consists of a glass sample stage coupled to a cooled CCD with fibre optics. This configuration enables maximal capture of light emitted from one side of a flat sample placed on the stage by avoiding the inevitable light dispersal associated with lens-based systems. The system is useful for imaging bioluminescence from flat samples such as plant leaves or cells grown transparent culture vessels or on solid supports such as microscope slides or membranes. By analysing the data captured from the CCD, it is possible to estimate the number of photons detected per unit area over a very wide dynamic range. We routinely measured emission levels in the region of 1000 photons/sec/mm 2 from wounded leaves (see below). In addition to its high sensitivity, the imager is significantly less expensive than equipment based on photomultiplier tubes. Imaging delayed chlorophyll fluorescence Delayed chlorophyll fluorescence is a well known phenomenon that is the result of emission of photons from excited chlorophyll molecules following transfer of leaves to the dark. It is generally measured using devices such as photomultiplier tubes and in principle, can be detected by existing lens-based CCD imaging systems. We first tested the spatial and temporal resolution of our system by imaging delayed chlorophyll fluorescence from leaves of Arabidopsis thaliana and Tradescantia albiflora (Figure 2 ). The image of a leaf from a variegated T. albiflora plant clearly indicates a dependence on chlorophyll for the emission to occur, since white, non-pigmented regions of the leaf do not emit light (compare Figure 2B and 2C ). Bright images could be obtained with relatively short exposure times of 1–5 minutes. Delayed chlorophyll fluorescence rapidly subsided between 5 and 10 minutes after transfer to dark. The absolute level of delayed chlorophyll fluorescence in Arabidopsis leaves varied greatly, generally appearing to be lower later in the day. Spatial and temporal resolution of ultra-weak luminescence in wounded leaves Delayed chlorophyll fluorescence produces relatively strong light emission compared to typical levels of ultra-weak bioluminescence. To test the sensitivity of our equipment, we wanted to try to image these lower level emissions. Several previous reports had indicated that mechanical wounding increases ultra-weak luminescence from plant leaves, but studies on the temporal and spatial nature of this phenomenon are limited. We used a haemostat to inflict crushing wounds on Arabidopsis leaves and imaged them over a time course by making successive 5-minute exposures. In many experiments, no wound-induced luminescence could be seen in the first 5 minutes after transfer to the imager, since it was masked by delayed chlorophyll fluorescence. After this initial period however, photon emission was much more intense around the wounded tissue. In other experiments, where delayed chlorophyll fluorescence was lower, signals could be seen even in the first exposure. Results from such an experiment are shown in Figure 3 and in an accompanying extended animation (Additional File 1 ). In the first image, which includes an unwounded control leaf and two leaves each wounded twice across the lamina, striking wound-induced luminescence can be seen above a general background of delayed chlorophyll fluorescence (Figure 3A ). In the two subsequent 5 min exposures, chlorophyll fluorescence has subsided, whilst the wound-induced luminescence remains high (Figure 3B,3C ). The serrated pattern of the haemostat surface is clearly reflected in the pattern of luminescence from the leaf, with areas of greater damage exhibiting maximal photon emission. In general, we observed that induced luminescence is strongest immediately after wounding and then gradually decays over a period of an hour or more. Origin of wound-induced luminescence In order to understand more about the origin of the luminescence emitted by wounded leaves, we used coloured filters to investigate its spectral characteristics. The results are shown in Figure 4 . A red filter (LEE 182), which transmits visible light of wavelengths greater than 600 nm, was effectively transparent with respect to the light emitted from wound sites, whereas blue and green filters (LEE 183 and LEE 735), which absorb light in the region of 550 – 750 nm, caused a significant reduction in signal reaching the detector when placed between the wounded leaf and the sample stage. Interestingly, a different blue filter (LEE 118), which transmits light above 700 nm, did not significantly affect luminescence (Figure 4 ). These results suggest that light emitted from wounded Arabidopsis leaves has a wavelength between 700 and 750 nm, i.e . the red region of the visible spectrum. This is consistent with typical fluorescence from chlorophyll (730 nm). To test whether chlorophyll might be the source of wound-induced photon emission, we used the carotenoid biosynthesis inhibitor, norflurazon, to generate photobleached leaves lacking chlorophyll [ 13 ]. Compared with equivalent leaves from control plants, these white leaves failed to luminesce following wounding, except in remaining chlorophyll-containing areas (Figure 5 ). Together, these results demonstrate that light emission from wounded leaves is chlorophyll-dependent. Since luminescence occurs in the dark, and extends for times well beyond delayed chlorophyll fluorescence, we conclude that excitation energy produced by some wound-induced process is transferred to chlorophyll and emitted as light. One possible source of such energy might be products of oxidative damage, such as excited triplet carbonyls produced during lipid peroxidation [ 14 ]. Temperature-dependence of wound-induced luminescence That delayed chlorophyll fluorescence exhibits temperature-dependence has been shown previously in several systems [ 15 ]. Since we determined that chlorophyll is the major source of light emission from wounded leaves, we investigated whether wound-induced luminescence might also be affected by temperature. Application of heat to the sample via a foil heater in the lid of the sample chamber significantly increased the intensity of luminescence over the tested temperature range. Without applied heat, the temperature of the sample stage was approximately 7°C, when only low levels of luminescence could be detected. Luminescence increased significantly upon activation of the heater, resulting in an increase in the temperature of the sample stage to something in the region of 20°C. This phenomenon is illustrated in Figure 6 , and in an accompanying extended animation (see Additional file 2 ). Furthermore, we found that heating could activate luminescence in leaves wounded up to 3 hours before imaging. For example, Figure 7 shows a series of images from an experiment in which leaves were wounded at different times before imaging. Without heating, only the leaf wounded immediately before imaging showed detectable wound-induced luminescence, and this emission decayed to very low levels within 20 min. After 45 mins, heating was applied to the sample chamber, and within 10 minutes, wounds applied immediately, 15 mins and 1 hour before the start of imaging were clearly visible, with a faint emission also detectable from a wound applied to the leaf 2 hours before the start of imaging (therefore the image shown here (t = 65) was captured over 3 hours after wounding this particular leaf). This phenomenon is reminiscent of the light emission measured during thermoluminometry [ 14 ]. Thermoluminescence is a chlorophyll-dependent light emission observed at high temperatures (maxima at 70–90°C and 120–140°C), which, like ultra-weak bioluminescence, is attributed to singlet oxygen and triplet carbonyls produced by lipid peroxidation during oxidative stress [ 14 ]. Coupled with the results above, these data suggest that temperature increases the rate of decay of excited triplet carbonyls, and that the energy released is transferred to chlorophyll and emitted as light. There may also be an additional temperature-dependent component in the rate of emission of light from excited chlorophyll, as observed for delayed fluorescence [ 15 ]. Conclusions The novel configuration of cooled CCD that we have used in this study to capture 2-dimensional images of plant leaves provides excellent temporal and spatial resolution of bioluminescence/chemiluminescence processes in biological materials, which are important markers of various forms of stress and disease. Using this system, we are able to define the origin of wound-induced luminescence from plant leaves as chlorophyll, and suggest that this arises via the temperature-dependent release of energy from excited triplet carbonyls produced by oxidative stress. Methods Plant material Plants of Arabidopsis thaliana , Columbia ecotype, were grown in soil in a glasshouse at 22°C with a 16-hour photoperiod. Wounding was performed by crushing leaves with a haemostat. To generate leaves lacking chlorophyll, Arabidopsis plants grown for 3 weeks under 16 h light/8 h dark, were watered with 0.1 mM norflurazon (Sigma Aldrich PS1044) and grown under constant illumination at 120 μmol/m 2 /sec. Under these conditions, newly emerging leaf tissue was white. Leaves with white regions were taken 15 days after the start of norflurazon treatment, and used for luminescence imaging. Two dimensional luminescence imaging apparatus The imager used is a two-dimensional imaging system based on a slow-scan 'scientific grade' Silicon Photo Area-Detector [SPAD] (Integrated SPAD Imaging System – PiXxell 1A, Biolumonics Ltd), which consists of a cooled ultra-sensitive area photo-detector within a stainless steel vacuum vessel and associated image acquisition electronics. The input image is coupled to the detector with optical fibres (numerical aperture = 1). The detector contains 222,336 pixels, and in the experiments presented here, ran at an operating temperature of -58°C. The combination of the proprietary low read-out noise electronics, cooling of the detector and the optical fibre coupling results in a sensitive imaging system capable of recording the faintest luminescences at high resolution, (40 μm to 200 μm depending on geometry of sample and integration time). The spectral absorption response of the SPAD ranges from the near UV to IR, thus including the entire visible spectrum up to the band-gap energy barrier of Silicon (1.1 eV) i.e . at approx 1100 nm; however, for ultimate sensitivity the detector output is monochrome, i.e . there is no wavelength selectivity in the present instrument. Further information is available on request from Michel Flor-Henry, Biolumonics Ltd. Filters Coloured filters were obtained from Lee Filters (Andover, UK). We used lighting effect filters LEE 182 – "light red," LEE 118 – "light blue," LEE 183 – "moonlight blue" and LEE 735 – "velvet green." Spectral characteristics of the filters can be viewed at , though readers should note that the transmittance spectra provided with the filters covers the range 300–800 nm, rather than the more limited 400–700 nm spectra shown on the web site. Filters were placed between the leaf and the sample stage prior to imaging. Authors' contributions MF-H designed the PiXxell 1A luminescence imager. MF-H and MRR conceived of the current study, participated in its design and coordination, carried out experiments and wrote the manuscript. MF-H performed data analysis. TCM performed the norflurazon experiment, and both TCM and GLdB participated in some experimental design and execution. All authors read and approved the final manuscript. Supplementary Material Additional File 1 Animated gif file showing images from five sequential 5-minute exposures of one control leaf (bottom) and four wounded Arabidopsis leaves. Click here for file Additional File 2 Animated gif file showing images from seven sequential 5-minute exposures of two control leaves (bottom) and four wounded Arabidopsis leaves. Images were initially captured from samples on a cold imaging stage, but the heater in the lid of the sample stage was switched on when indicated. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535552.xml
539340	Separating Wheat from Chaff in Plant Genomes	null	Plant genome sizes span the modest—54 million base pairs (Mb) in the bitter cress Cardamine amara —to the enormous—124,000 Mb in the lily Fritillaria assyriaca . By comparison, fruitfly and human genomes have 180 Mb and 3,200 Mb, respectively. Genomes of important crops such as sorghum, soybean, maize, and wheat hover between 735 Mb and 16,900 Mb, and determining their complete sequences is daunting and costly. Wide size variations do not necessarily reflect differences in gene content, but rather reflect the presence of repetitive sequence elements that do not generally code for genes. Repetitive elements account for at least 75% of the maize and sorghum genomes. In a new study, Joseph Bedell and his colleagues describe a way to filter away repetitive elements when sequencing the genome of sorghum ( Sorghum bicolor ), a staple crop in much of the developing world because of its resilience in arid climates. The authors use an approach known as methylation filtration that has been employed before for pilot plant genome analyses. Here they present compelling evidence of the method's reliability when applied to large-scale genome sequencing. The approach is built on the observation that in plants, methylation—a chemical tagging of DNA with methyl groups—occurs at repetitive sequences to a much greater degree than at gene sequences. This provides an opportunity to concentrate sequencing efforts on the coding portion of the genome. To eliminate repetitive sequences, the authors introduced small pieces of sorghum chromosomes into bacteria strains designed to specifically destroy DNA sequences that carry methyl groups. Using two independent assessments, they estimated that methylation filtration reduced the amount of sorghum DNA they would need to sequence by two thirds, from 735 Mb to approximately 250 Mb. But were any genes lost in the filtration step? The authors compared their results to partial sequence information generated previously from bacterial artificial chromosomes (BACs). BACs offer the most comprehensive representation of the genome because they contain large pieces of unmodified sorghum chromosomal DNA. Of the 148 genes identified on 14 sorghum BACs, 133 appeared in the filtered set. This means that the methylation filtration method captured at least 90% of the genes in the sorghum genome and 96% (131/137) if a repeat cluster of 11 known methylated genes is removed from the analysis. A field of hybrid sorghum Methylation filtration also compared favorably to shotgun sequencing, a method that reads the whole genome in small fragments that are progressively assembled into larger pieces by computer analysis. The authors reported that after sequencing 285 Mb of filtered sorghum DNA—approximately 1.15 times the length of the sorghum coding regions—they obtained on average 65% of the length of 96% of the genes. Theoretical calculations and simulation based on the genome of Arabidopsis —a plant model organism—predicted that shotgun approach would yield similar results (67% of the length of 96% of the genes) after sequencing the equivalent of 1.15 times its total length (rather than 1.15 times the length of just the coding regions). Thus, methylation filtration can provide as much information on coding sequences as the shotgun approach, with less investment in sequencing. Methylation filtration does not yield a complete genome map, but it offers quicker, more affordable access to genes than most commonly used sequencing approaches. Sorghum is closely related to maize and sugar cane, and more distantly to rice. The availability of its genome sequence offers the chance for more in-depth experiments into the evolution of the grass family, and promises important insights into the genetic control of drought resistance.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539340.xml
534102	Mycobacterium tuberculosis from chronic murine infections that grows in liquid but not on solid medium	Background Old, stationary cultures of Mycobacterium tuberculosis contain a majority of bacteria that can grow in broth cultures but cannot grow on solid medium plates. These may be in a non-replicating, dormant growth phase. We hypothesised that a similar population might be present in chronic, murine tuberculosis. Methods Estimates of the numbers of viable M. tuberculosis , strain H37Rv, in the spleens and lungs of mice in a 7-day acute infection and in a 10-month chronic infection were made by conventional plate counts and, as broth counts, by noting presence or absence of growth in serial replicate dilutions in liquid medium. Results Plate and broth counts in 6 mice gave similar mean values in the acute infection, 7 days after infection. However, the broth counts were much higher in 36 mice with a chronic infection at 10 months. Broth counts averaged 5.290 log 10 cfu /organ from spleens and 5.523 log 10 cfu/organ from lungs, while plate counts were 3.858 log 10 cfu/organ from spleens and 3.662 log 10 cfu/organ from lungs, indicating that the total bacterial population contained only 3.7% bacilli in spleens and 1.4% bacilli in lungs, capable of growth on plates. Conclusion The proportion growing on plates might be a measure of the "dormancy" of the bacilli equally applicable to cultural and animal models.	Background The organisms in a log phase, actively multiplying culture of Mycobacterium tuberculosis all grow well on plates and are estimated as colony forming units (cfu). However, cultures that have been grown undisturbed in the depths of liquid medium for 100 days contain a majority population which grows in liquid medium but is not able to form colonies on solid medium [ 1 , 2 ]. Since the bacilli in such cultures are hardly multiplying and have an uptake of [ 3 H] uridine of only 15% of log phase cultures [ 1 ], they may be considered as dormant. A smaller population, that could grow on plates as well as in broth, possibly the survivors of the log phase bacilli, has also been demonstrated in these cultures. At the end of a 10-month mouse experiment on vaccines, we questioned whether the same two populations would be found, particularly in view of the evidence that there are similarities in gene expression patterns following adaptation to micro-aerophilic conditions in stationary cultures and exposure to NO in macrophages [ 3 ]. We therefore estimated populations present in the chronic infections in the organs of the 10-month mice both by conventional plate counts and by counts of the probable number of viable organisms obtained from serial dilutions in liquid medium. Similar counts were also set up, as a control, in a short-term acute infection in mice. Methods Culture media and bacteria The media used were 7H9 liquid medium with 10% albumin, dextrose, catalase supplement and 0.05% Tween 80, and 7H11 agar medium with 10% oleic acid, albumin, dextrose, catalase supplement (Becton Dickinson, Oxford, UK). They were made selective by the addition of 100 μg carbenicillin, 200 U polymyxin B, 20 μg trimethoprim and 10 μg amphotericin B per ml (Mast, Bootle, UK) [ 4 ]. The two experiments both used 6-week, female Balbc mice, which were infected, with a mouse passaged H37Rv strain of M. tuberculosis suspended in 0.1% gelatin. The spleens and lungs were obtained at sterile autopsy and were homogenised as described elsewhere [ 5 ] in 5 ml water. From this suspension 100 μl amountsfrom the neat suspension and from serial 10-fold dilutions in 1 ml, were inoculated onto duplicate thirds of selective 7H11 medium plates. The number of colonies was counted after 3–4 weeks incubation at 37°C to give the plate count. A negative plate therefore had <25 cfu / organ. For the broth counts, serial 10-fold dilutions of the organ homogenate were made, in triplicate, in 1 ml amounts to 9 ml amounts of selective 7H9 broth with 0.05% Tween 80 in plastic 28 ml screw-capped bottle. In the acute infection, 10 serial dilution were set up, so as to obtain 30 tubes in all, while in the chronic infection 6 serial dilutions were set up yielding 18 tubes in all. These were incubated at 37°C and examined at 3 and 6 weeks and finally at 9 weeks for the characteristic growth of M. tuberculosis , namely a clear supernatant in undisturbed cultures with an upwards swirl of white growth on shaking,. Probable numbers of bacilli (pnb) per organ were obtained from a table of densities of organisms estimated by the dilution method [ 6 ]. Samples of the positive growth from 18 broth cultures were plated out on 7H11 medium. Acute infection experiment Each of 6 mice was infected by the intravenous route with 200 μl of a suspension of a containing 2.6 × 10 6 cfu of the H37Rv strain. Plate and broth counts were carried out 7 days later. Chronic infection experiment Each of 88 mice was infected by the intra-peritoneal route with 200 μl of a suspension containing 3.1 × 10 3 cfu of the H37Rv strain. Mice in our experiments are usually housed in an isolator, connected by a tunnel port to a Class 1 safety cabinet through which air from the environment is sucked. The intra-peritoneal route was chosen so that mice could be kept throughout the experiment in the isolator to prevent cross infection with mouse pathogens during exposure to the outside air in the Class 1 cabinet. One day after infection, samples of 6 mice yielded scanty or no colonies in plate counts of spleens and lungs. After a further 4 weeks these organ counts in 6 mice had risen to 2.24 × 10 4 in spleens and 1.15 × 10 4 in lungs. The mice were then divided into 6 experimental groups, 4 of which were vaccinated with various DNA vaccines over a 4-week period and 2 were unvaccinated controls. At 12 weeks after infection only 9 of 36 lungs and 18 of 36 spleens yielded positive growth on plates. At 10 months after infection, the 36 remaining mice were sacrificed, and plate and broth counts were set up on all, using dilutions estimated from a sample of 4 mice sacrificed 3 weeks earlier. Statistics The results of the plate and broth counts were examined by 2-way analysis of variance using the Stata package, release 8 (Stata Corp., College Station, TX) Results As the experimental vaccines appeared to have only small effects, which will be reported elsewhere, in the chronic infection model, the results in all 36 mice at 10 months are considered together. A typical broth count is shown diagrammatically in Table 1 . Note that there were never any sporadic positive tubes in the no growth zones (inoculated in the example with 10 -4 or 10 -5 dilutions) of any set of broth cultures. The results in the 6 mice in the acute infection experiment and in 27 of the 36 mice in the chronic infection experiment that had assessable numbers of bacilli estimated as cfu by the plate method and as pnb by the broth method are set out in Table 2 . In the analysis of variance of the acute infection counts, neither the variation between individual mice nor the difference between the counting methods was statistically significant. However, in the chronic infection, the broth counts were higher than the plate counts. In the spleens, the mean broth count was 5.290 log 10 cfu / organ and the plate count was 3.858 log 10 cfu /spleen. The difference between these counts is 1.432 log 10 cfu / spleen (27-fold) so that, on the assumption that all bacilli that grew on plates also grew in broth, the bacilli capable of growing on plates comprised 3.7% (100/27) of the total count. In the lungs the mean broth count was 5.523 log 10 cfu /lungs, about 73-fold higher than the plate count for the lungs. Thus the bacilli able to grow on plates comprised about 1.4% (100/73) of the total. The differences between individual mice were significant (p = 0.01) and highly significant between the counting methods (p < 0.001). In the remaining 9 mice no colonies were obtained on the neat dilution plates in either the lungs alone or in both lungs and spleen (Table 3 ). However, broth counts were obtainable from both organs in all 9 mice, though their mean values were appreciably lower than those in Table 1 . Where a comparison could be made between counting methods in the spleens of the 7 mice with colonies in plate counts, the means of the broth counts (4.347 log 10 cfu / ml) were 19-fold higher than the corresponding plate counts (3.079 log 10 cfu / ml, giving 5.3% of the total), in approximate agreement with the 27-fold (3.7%) estimate of the difference between broth and plate counts obtained from Table 1 . The following changes occurred in the broth counts during incubation. The counts between the 3-week and the 6-week readings increased on average in the lungs of each acute mice by 2.3 tubes and by 4.7 tubes in the spleens, and in each of the chronic mice by 2.0 tubes in lungs and 1.6 tubes in spleens. Thereafter, the increase from 6 weeks to 9 weeks was 1.5 tubes in the lungs of acute mice and 0.83 tubes in their spleens, while the increases in the lungs of chronic mice were 1.3 tubes and 1.0 tube in the spleens. Since an increase of 1 tube indicates a rise of about log 10 0.8, that is about 12%, in the count, it is evident that counts increased during incubation, rapidly between 3 and 6 weeks and slowly between 6 and 9 weeks. Discussion The chronic infection experiment was unusual in that the intraperitoneal infection in vaccinated mice led to trapping of the bacilli in the peritoneal cavity, so that few bacilli reached the organs, and thus the plate counts were sometimes negative, with a count of less than 25 cfu / organ. Variation, considerably greater than after intravenous or airborne infection, was also evident, the SD of the 4-week spleen counts, expressed as log10 cfu/organ, being 0.51 as compared to 0.23 for intravenous infection [ 5 ]. Evidence that growth in the broth cultures was M. tuberculosis was provided by the growth of typical colonies on 7H11 plates. That it was not due to sporadic contamination was shown by the usual complete absence of contamination in selective media, by a clear supernatant in the unshaken cultures and by the absence of any growth in the "no growth" zones of the broth cultures. In log phase cultures (Hu Y-M, personal communication) and in the acute infection of mice, similar estimates of viable organisms were obtained in plate and broth counts However, our best estimate indicated that the bacilli in the chronic infections that would grow on plates was about 3.7% of the total population in spleens and 1.4% in lungs. This can be compared to the findings on a culture in 7H9 broth grown undisturbed for 100 days in the depths of liquid medium with caps screwed tightly on. In such a 100-day culture, population A, capable of growth in broth but not on plates, was estimated by broth dilution counts to be 7.60 log 10 pnb / ml, while a smaller population B, that grew on plates, was estimated from parallel plate counts as 5.85 log 10 cfu / ml.[ 1 ] Thus, population A was 1.75% of the total population. The corresponding estimates for a 30-day static culture were population A = 9.983 log10 pnb / ml and population B = 8.013 log10 cfu /ml, so that population B comprised 1.07% of the total population (Hu Y-M, personal communication). These estimates suggest that the extent to which bacilli have gone into ill-defined dormancy might be measured as the proportion of a total bacterial population that can grow on plates. The lower this proportion, the greater the overall degree of "dormancy". Whatever, the theoretical significance of this simple technique for measuring dormancy, there is a practical implication for those undertaking long-term mouse experiments. Some end such an experiment with plate counts and others with culture in liquid medium. Those using only plate counts may be seriously underestimating the residual populations. It is also clear that there is much work to be done in seeing how various experimental conditions, such as the duration of the infection and the immune state of the mice, affect the ratio between broth and plate counts. Those exploring the development of new drugs need to know how these two populations respond to current anti-tuberculosis drugs and to novel drugs. Competing interests The author(s) declare that they have no competing interests. Authors contributions All three authors took part in the running of the experiments with JD contributing the most. DAM contributed the concept of parallel broth and plate counts. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC534102.xml
535546	Assessment of FIV-C infection of cats as a function of treatment with the protease inhibitor, TL-3	Background The protease inhibitor, TL-3, demonstrated broad efficacy in vitro against FIV, HIV and SIV (simian immunodeficiency virus), and exhibited very strong protective effects on early neurologic alterations in the CNS of FIV-PPR infected cats. In this study, we analyzed TL-3 efficacy using a highly pathogenic FIV-C isolate, which causes a severe acute phase immunodeficiency syndrome, with high early mortality rates. Results Twenty cats were infected with uncloned FIV-C and half were treated with TL-3 while the other half were left untreated. Two uninfected cats were used as controls. The general health and the immunological and virological status of the animals was monitored for eight weeks following infection. All infected animals became viremic independent of TL-3 treatment and seven of 20 FIV-C infected animals developed severe immunodepletive disease in conjunction with significantly ( p ≤ 0.05) higher viral RNA loads as compared to asymptomatic animals. A marked and progressive increase in CD8 + T lymphocytes in animals surviving acute phase infection was noted, which was not evident in symptomatic animals ( p ≤ 0.05). Average viral loads were lower in TL-3 treated animals and of the 6 animals requiring euthanasia, four were from the untreated cohort. At eight weeks post infection, half of the TL-3 treated animals and only one of six untreated animals had viral loads below detection limits. Analysis of protease genes in TL-3 treated animals with higher than average viral loads revealed sequence variations relative to wild type protease. In particular, one mutant, D105G, imparted 5-fold resistance against TL-3 relative to wild type protease. Conclusions The findings indicate that the protease inhibitor, TL-3, when administered orally as a monotherapy, did not prevent viremia in cats infected with high dose FIV-C. However, the modest lowering of viral loads with TL-3 treatment, the greater survival rate in symptomatic animals of the treated cohort, and the lower average viral load in TL-3 treated animals at eight weeks post infection is indicative of a therapeutic effect of the compound on virus infection.	Background Feline immunodeficiency virus (FIV) is a lentivirus that infects domestic and feral cat populations worldwide. Infected cats exhibit similar disease patterns as human immunodeficiency virus (HIV) infected patients by developing multiple immuno-depletive symptoms collectively referred to as acquired immunodeficiency syndrome (AIDS). As with HIV, differences in virulence among the different FIV subgroups are evident [ 1 - 4 ]. Thus, the cat represents an amenable animal model for testing certain anti-HIV-1 drug modalities in vivo . One of the major breakthroughs in HIV-1 treatment has been the use of specific inhibitors of the viral aspartic protease family as part of a drug cocktail, called highly active anti-retroviral therapy (HAART), with the ultimate goal of suppressing HIV-1 replication in patients to low or undetectable levels [ 5 - 8 ]. Effective HAART therapy continues to be dependent on the development of new drug modalities due to the rapid mutation rate of HIV-1, leading to drug resistance development [ 9 ]. Therefore, an effective small animal model for evaluating new drugs and treatments for HIV is of paramount importance. Experimental testing of new protease inhibitors in cats has been of limited success due to the ineffectiveness of HIV-1 specific protease inhibitors against FIV [ 10 , 11 ]. The promising development of the protease inhibitor TL-3, which inhibits FIV, HIV-1 and SIV (simian immunodeficiency virus) infections in vitro with similar effectiveness [ 12 ] led us to analyze its efficacy in the cat model. Initial in vivo studies using the predominantly neurotropic FIV-PPR strain, showed that TL-3 treatment lowered plasma viral loads and resulted in a significant protective effect against neurologic alterations in the CNS in FIV infected cats [ 13 ]. In the present study, we employed the highly pathogenic FIV-C isolate (CABCpady00C), which causes a fulminant acute phase disease in the periphery, with high death rates from acute phase immunodeficiency disease [ 1 ]. Results In Vivo Infection Twenty-two female specific pathogen-free (SPF) cats were randomly divided into five groups. Group 0 consisted of two cats, which received TL-3 treatment without viral infection and were considered controls. Group 1 (n = 5) received 0.1 ml (10 5 RNA copies/ml) of FIV-C-infected plasma I.V. with TL-3 drug treatment. Group 2 (n = 5) received 0.1 ml FIV-C-infected plasma without TL-3 treatment. Group 3 (n = 5) received 0.5 ml FIV-C-infected plasma with TL-3 and Group 4 (n = 5) received 0.5 ml FIV-C-infected plasma without TL-3 treatment. Blood (1 ml) was drawn from all cats prior to the start of the experiment, at weekly intervals for the first four weeks after infection, and at bi-monthly intervals from week 4 until the end of the study. Complete blood counts were assessed as a function of infection and TL-3 treatment. In addition, quantitative reverse transcription PCR (QRT-PCR) analyses were performed to assess plasma viral load. All animals were continuously observed for any changes in general health. No significant differences were noted between viral load or disease phenotype between the two plasma dosages used in infection and subsequent discussion will not distinguish between these two groups. By week 6 post infection, seven animals (221, 222, 220, 234, 229, 219, 215) were showing clinical signs of debilitating acute phase disease. Four of the seven affected animals (215, 219, 220, 234) were from the untreated groups, and three animals (222, 221, 229) were from the TL-3 treated groups. Symptoms in all seven symptomatic cats varied from conjunctivitis, anorexia, corneal ulcerations, and gingivitis to increases in temperature, dermatitis and marked lethargy. Despite intensive antibiotic treatment, the general state of health of 6 of the cats did not improve (221, 220, 234, 229, 215, 219) and they received mandated euthanasia between six to seven weeks post-infection. Cat 222 (TL-3 treated) responded to antibiotic therapy and recovered from acute phase symptoms. Control and infected cats gained weight at approximately the same rate during the first 4 weeks post infection, regardless of drug treatment status (Figure 1 , data expressed as a ratio to starting weight for each animal). However, at 6 weeks post infection, three animals in the + TL-3 cohort (Figure 1 , upper panel) and three animals in the -TL-3 cohort (lower panel) had lost weight. All three of the animals in the -TL-3 cohort (bottom panel, shown in red; 215, 220, and 234) required mandated euthanasia prior to the next weighing at week 8. Cat 229 in the TL-3 treated cohort also required mandated euthanasia prior to week eight. Cat 221 in the + TL-3 cohort was euthanized on the same day as cat 229, but a final weight was not recorded. Cat 219 of the -TL-3 cohort had a normal weight at week 6, but required euthanasia at week 7, with an approx. 10% weight loss relative to week 6 (data not shown in figure). Thus, weight loss at week 6 occurred with onset of severe acute phase disease. Cat 222 in the TL-3 treated group (upper panel) responded to aggressive rehydration and antibiotic therapy, gained weight, and survived the acute phase. Cat 223 was never noticeably symptomatic and it is unclear why this animal showed a dip in weight at week 6 which it recovered by week 8. Figure 1 Cat weight ratios as a function of FIV infection. Cats infected with FIV-C in the presence and absence of TL-3 treatment were weighed regularly throughout the course of infection. Graphs depict weight of individual cats as a ratio to respective starting weight. Control cats were 213 and 214; green symbols. Weight ratios of cats that required euthanasia as a result of acute phase feline AIDS are shown in red. Brainstem auditory evoked potential changes (BAEPs) Previous studies using FIV-PPR showed that the isolate induces marked and consistent delays in BAEPs of infected cats and that TL-3 could reverse this effect [ 13 ]. We, therefore, analyzed the FIV-C infected animals for similar BAEP delays with or without TL-3 treatment. Animals were analyzed at two-week intervals for the first eight weeks of the experiment. Interestingly, no delays in BAEPs were noted with FIV-C infection (data not shown) in spite of high viral loads in the periphery (see below). Viral load quantification Plasma viral RNA loads were measured at regular intervals throughout the experimental period to evaluate viral load changes in cats treated with or without TL-3. Viral loads ranged from undetectable to approximately 9 × 10 9 RNA copies/ml plasma (Figure 2 ). Examination of plasma viremia at 2 weeks (i.e. before an active immune response) revealed little differences in viral loads between cats in the TL-3 treated and untreated cohorts (Figure 2A ). In contrast, after 2 weeks of infection, the highest viremias were found in cats not receiving TL-3 (compare + TL-3 to -TL-3, Figure 2B ). Moreover, of the 6 FIV symptomatic cats that were euthanized due to the severity of clinical disease, 4 of these animals had received no TL-3 and 3 of these cats presented with the highest viral load (compare open symbols to filled symbols, Figure 2C ). Figure 2 Plasma viral loads of FIV infected cats as a function of TL-3 treatment and disease. (A), normalized viral load (copies/ml × 10 6 ) at week 2; (B), and at peak viremia between weeks 0 – 6.5 in FIV infected cats treated with TL-3 (+TL-3, solid symbols) or untreated (-TL-3, open symbols). (C), peak viremia between weeks 0 to 6.5 post infection, in healthy (asymptomatic) and symptomatic (euthanized) cats. The average value (largest horizontal bar, –) is plotted for each group. Equal volumes of plasma were normalized using an external RNA spike and analyzed as detailed in Materials and Methods for the presence of FIV RNA by reverse-transcription real-time quantitative PCR. * indicate average values in (C) differ significantly ( p ≤ 0.05). Analysis of viremia patterns in individual cats during the initial 8 week evaluation period yielded additional differences in the TL-3 treated vs. untreated groups. Although infected cats showed an initial viral peak in the second week post-infection, by week 4, viral loads decreased for each cat and the average viral load for symptomatic and asymptomatic cats was similar (Figures 3A and 3B ). However, by week 6 the symptomatic group had an average viral load of 2 × 10 8 copies/ml, 27 times more virus than the average viral load of asymptomatic cats (compare Figures 3A and 3B ). Half of the 6 symptomatic cats were euthanized at week 6 due to severe illness. The remaining three symptomatic cats (all from the untreated group) had an average viral load of 3.2 × 10 8 copies/ml at 6.5 weeks (Figure 3A ) and required euthanasia by week 7. Within the asymptomatic cohort, three (223, 224, 230) out of eight TL-3 treated cats had reduced viral loads by week 4. Their viral loads remained low during week 6 and at week 8, four of 8 surviving TL-3 treated animals had viral load levels below detection limits. Of the six surviving animals that had not received TL-3, only one cat (233) had a viral load below detection limits at week 8 (Figure 3B ). Figure 3 Individual plasma viral loads as a function of TL-3 treatment over an 8 week period. Normalized viral loads in (A) symptomatic (euthanized) and (B) asymptomatic cats. Each value corresponds to the volume normalized viral load (copies /ml) between weeks 2–8. Control cats, 213 and 214 (not shown) had viral levels below detection. Cats are grouped for the presence (■ solid bars) or absence (□ open bars) of TL-3 treatment, and then in numerical order from left to right. Specific bars are marked for the approximate week of euthanasia (†) and viral levels below detection are denoted as (). Leukocyte Changes Peripheral blood mononuclear cells (PBMC) were isolated from whole EDTA-treated blood from each animal and CD4 + and CD8 + T cell quantitations were performed by flow cytometry and leukocyte counts. Absolute values for CD4 + T cells showed a general decreasing trend in both the TL-3-treated and untreated cohorts when all animals were averaged in each group and CD8 + T cell counts and neutrophils did not show significant variance from control values (data not shown). However, analysis of CD4 + and CD8 + T cells counts of symptomatic cats as compared to treatment-matched asymptomatic animals showed interesting differences (Figure 4A and 4B , resp.). The CD4 + T cell population of the symptomatic cats decreased progressively over the first 6-week period (Figure 4A ). CD4 + T cell counts of identically treated cats that showed no signs of disease also decreased, but less precipitously than those of animals eventually requiring euthanasia due to severity of illness ( p ≤ 0.05). Uninfected control animals maintained their T cell values during the same time frame. The CD8 + population of T cells showed an even more drastic decrease in the symptomatic animals compared to the asymptomatic infected animals (Figure 4B ). Between weeks two and four, the CD8 + T cell population in the symptomatic animals showed a small rebound from week 2, then declined through week 6 post infection. In contrast, CD8 + T cell counts in treatment-matched asymptomatic animals increased significantly ( p ≤ 0.05) from week 2 onward, consistent with a strong CD8 + T cell response in the protected cats. Figure 4 Peripheral CD4 + and CD8 + lymphocyte levels as a function of clinical outcome over 8 weeks. A) Total CD4 + cells per ml average and standard error of the mean; B) Total CD8 + cells per ml average and standard error of the mean. Symbols: ●: uninfected cats (213 and 214); □ Symptomatic cats (215, 219–221, 229 and 234); ◇ asymptomatic cats (216–218, 222–228, 230–233). indicate values between symptomatic and asymptomatic cats differ significantly ( p ≤ 0.05). Consistent with previous reports [ 14 - 16 ], we also observed changes in the total neutrophil counts in the symptomatic FIV-C infected cats. Within one week post-infection, neutrophil numbers increased markedly in cats 220 (10, 962 cells/μl) and 234 (11, 926 cells/μl) as compared to other cats with identical TL-3 treatment (average = 6880 ± 2847 cells/μl). By week 4, neutrophil values fell drastically in the same two cats (220: 558 cells/μl; 234: 416 cells/μl) as well as in cats 215 (420 cells/μl) and 221 (400 cells/μl). By week 6, four of the symptomatic cats had neutrophil counts near zero. Only cat 220 slightly recovered its neutrophil cell count (3675 cells/μl) prior to mandated euthanasia. Protease escape variants Drug pressure induced viral resistance to protease inhibitors represents one of the major hurdles of HIV HAART treatment regimens [ 17 , 18 ]. HIV and FIV proteases share only 23–28% overall identity at the protein level, yet the enzymatic active site residues are virtually identical [ 19 ], allowing a convenient side-by-side comparison. When analyzing the surviving cohort of animals that had received TL-3 treatment, we noted high viral loads in some cats (Figure 3B ; 222, 228, 231, 232) that remained elevated (except 228) between weeks 6 and 8, relative to other cats (Figure 3B ; 223, 224, 230) that had significantly lowered viral loads. We, therefore, analyzed plasma samples from symptomatic and asymptomatic TL-3 treated animals to look for potential drug-resistance. The protease gene was cloned from week 6 plasma of the two symptomatic TL-3 treated cats (221 and 229) and sequenced. Twenty individual protease clone sequences revealed a wild-type protease gene sequence (data not shown). We then chose cats 231 and 228, two TL-3-treated asymptomatic cats with relatively high viral loads in week 6 (Figure 3B ), as well as cat 222, which overcame its disease syndrome with drug treatment. Cloning and sequencing of the protease gene revealed a number of interesting mutations (Table 1 ). The aspartic acid to glycine point mutation at FIV protease residue 105 (D105G) occurred in cats 231 and 222. The equivalent site in HIV protease is residue 88 (HIV88N), which is associated with development of resistance to some HIV protease inhibitors [ 20 ]. Other potentially relevant point mutants cloned from cat 231 were the H72R (HIV63L) and N55D (HIV46M) from week 6 plasma and M107R (HIV90L) from week 8 plasma. The equivalent HIV residue escape mutants exhibit various degrees of resistance against current protease inhibitors [ 20 ]. Additional multiple point mutations, lying outside of the FIV protease active site or having no apparent important HIV equivalents were also observed (Table 1 ). The protease genes were cloned into expression vectors, expressed and purified for enzymatic analysis. The findings revealed that most of the mutants could not be distinguished from wild-type FIV-C protease as to TL-3 susceptibility. However, mutant D105G exhibited a K i of 47.3 nM as compared to 9.9 nM for wild-type, consistent with a 5-fold increase in resistance to TL-3. Table 1 FIV Protease Escape Mutants (HIV equivalent residue) PR mutations K i vs TL-3 (nM) Wildtype PR ---- 9.9 Cat 231 D105G (HIV88N) 47.3 D94G () N.D. H72R (HIV63L) 10.9 N55D (HIV46M) 9.1 G52R (HIV43K) N.D. 1 M107R (HIV90L) inactive 1 N51Y (HIV42W) N.D. Cat 228 C69R (HIV60D) N.D. Cat 222 D105G (HIV88N) 47.3 1 isolated from 8 th week plasma no HIV equivalent site Bold type, FIV-C point mutant of interest N.D. not determined Discussion The protease inhibitor, TL-3, demonstrated broad efficacy against FIV, SIV and HIV in tissue culture [ 12 ], as well as against drug-resistant HIV isolates [ 21 ]. Furthermore, TL-3 treatment had a very strong protective effect on early neurologic alterations in the CNS of FIV-PPR infected cats [ 13 ]. However, molecularly cloned FIV-PPR causes little acute phase disease in the periphery. We, therefore, sought to test TL-3 efficacy in vivo in the context of the highly pathogenic, uncloned CABCpady00C species (FIV-C) [ 1 ], which causes a severe acute phase immunodeficiency syndrome, with high early mortality rates. Although only partial protection was afforded by TL-3 in our studies, the results are promising in that average peak viral loads in some cats were lower in the presence of drug, even in the face of a highly aggressive infection (Figure 2B ). Of 20 cats infected with uncloned FIV-C, seven animals showed signs of immunodepletive disease early on (Figure 4 ) and developed full-fledged acute phase AIDS symptoms with anorexia, conjunctivitis, corneal ulcerations, gingivitis and marked lethargy by week 6, mandating euthanasia of six animals. The symptomatic cats had viral RNA loads significantly higher (>10 8 RNA copies/ml, p ≤ 0.05) than asymptomatic infected animals independent of drug treatment. This finding suggests that the intense viral infection severely compromised the immune system leading to immunodeficiency and the development of concomitant AIDS, as evidenced by the rapid loss of CD4 + T cells as well as neutrophils in the affected cats during the first few weeks (Figure 4 ). Cats receiving TL-3 treatment had lower peak viral loads compared to cats not receiving TL-3 at weeks 4 and 8, indicating that the protease inhibitor reduced systemic expansion of viral infection. Previous studies have correlated disease progression with high initial peak viral loads [ 22 ]. Of the five out of eight cats treated with TL-3 and having higher viral loads at weeks 4 and 8 compared to non-TL-3-treated cats, three (222, 228, 231) were evaluated for FIV resistance to TL-3. None of the protease genes recovered from cat 228, whose viral levels fell below detection at week 8, showed evidence of TL-3 resistance. However, cats 222 and 231 were found to harbor virus in the plasma that encoded TL-3 resistant protease. In particular, a D105G mutant demonstrated a 5-fold resistance to TL-3 relative to wild type protease, which may indicate the onset of drug resistance development and explain the higher viral levels in some of the TL-3 treated cats. Preparation of isogenic virus containing the D105G point mutation will allow the direct determination of potential drug resistance. Interestingly, TL-3 treated cats with highest viral loads (221 and 229) developed severe disease syndromes, which suggests that TL-3 efficacy was limited to a specific viral threshold in this study. Once the threshold has been crossed, TL-3 may not be able to overcome the full-blown, acute, viral infection, resulting in rapid onset of immune suppression. We were unable to show any correlation between humoral antibody responses and clinical outcome (data not shown). However, a consistent observation was a marked and progressive increase of CD8 + T cells in animals surviving the acute phase infection and a lack of such responses in animals that required euthanasia within the first 8 weeks following infection. Although not formally tested, the findings imply a strong cell-mediated response in the surviving animals contributed to controlling the viral infection. The above analyses underscore the natural variation in the response to FIV challenge in outbred cats, similar to the observed variation in responses to HIV infection in humans. However, more consistent responses, both in peak viral loads (Figure 2 ) and in lymphocyte counts (Figure 4 ) were seen when animals were analyzed as a function of disease severity. Thus, the question arises as to whether there is a genotypic link to susceptibility. Upon close scrutiny of the parental heritage pattern, we observed that one male in particular (96AGQ1) had sired eight of the experimental animals with three different females (Table 2 ). Two of his offspring (213, 214) had been randomly placed into the uninfected control group, while of the remaining six offspring, four (215, 219, 220, 221) succumbed to FIV induced disease. One of the surviving siblings (222) was in the TL-3 treatment group and exhibited conjunctivitis and possible corneal ulceration and was mildly to moderately lethargic. Cat 222 received Baytril treatment and fluid replacement therapy and eventually recovered from her symptoms. The last sibling (216) never showed any detectable signs of disease throughout the experiment. Although complicating the analyses and statistical treatments, the variable responses with FIV infection of cats parallels those observed with HIV infection in humans and thus affords a relevant model for study of infection and treatment modalities. As with humans, cats are outbred, which complicates defining susceptibility markers. However, identifying the genetic basis for susceptibility in the cat may yield important clues to similar phenomena in humans. Table 2 Feline Lineages Dam Sire Siblings 99AIE3 96AGQ1‡ 213* 214* 215† 216◇ 00AIB4 96AAK4 217 218 98AXS5 96AGQ1‡ 219† 220† 00ANU5 96AGQ1‡ 221† 222◇ 00AJT2 96ACJ2 224 01TBK5 99IAS1 225 226 227 01QAL4 00XAZ4 228 01QEJ3 98ATY2 229† 01QBJ3 00IRX4 230 95PAA3 98IUU1 231 01IQP4 98IUG3 232 233 00XBA1 00XAG1 234† * control animals † sacrificed animals Bold type, offspring from same sire (‡) ◇ survivor offspring of sire Viral protease inhibitors are of paramount importance for HIV treatment and successful tempering of viral infection. However, drug resistant escape variants are an important consideration in treatment protocols [ 17 , 18 ]. Although we previously failed to isolate TL-3-resistant FIV in vitro , the findings here suggest that in vivo , drug resistance to the compound may develop. The results were not unexpected in that we had been able to develop TL-3 resistant HIV variants [ 21 ] and it seemed unlikely that FIV would prove an exception. The finding of drug resistant mutants, in fact, strongly indicates that the feline/FIV model is valuable in the assessment of the ability of other protease drugs and drug cocktails to suppress virus infection and limit drug resistance development. Conclusions The findings indicate that the protease inhibitor TL-3, when given orally as a monotherapy, did not prevent viremia in cats infected with a high dose challenge with FIV-C and substantial virus loads were evident in circulation throughout the acute phase (between 2–6 weeks post infection) in all infected animals, regardless of drug regimen. Average peak viral loads in the acute phase were lower in TL-3 treated animals, but variability was such that the numbers did not reach statistical significance. However, of six animals that required euthanasia, four were from the untreated cohort and two were from the TL-3 treated group. Additionally, at eight weeks post infection, half the surviving TL-3 treated animals had viral loads below the detection limits, whereas only one of six untreated animals had markedly reduced viral loads. Thus, therapeutic benefit was noted with TL-3 treatment, even in the face of an aggressive FIV infection. The findings also show clear differences in the lymphocyte responses of animals that succumb to acute phase illness versus those that survive to the asymptomatic phase. The most pronounced difference was in the lack of an increase in CD8 + cell numbers starting around three weeks post infection in animals that eventually required humane euthanasia versus a pronounced and significant increase in CD8 + T cell numbers in animals that survived the acute phase. Certain animals that received TL-3 had higher than average viral loads after the acute phase. Analyses of the protease genes of FIV quasi-species prevalent in these animals revealed sequence variations relative to protease of wild type FIV-C. One particular protease, cloned and expressed from two TL-3-treated animals, contained the mutation D105G, which imparted 5-fold resistance against TL-3 relative to wild type protease. This may represent the initial stages of drug resistance development and preparation of this mutation in the context of isogenic virus will address this issue. The findings suggest that the cat model will serve as a valuable animal model for study of resistance development against lentivirus infections. Materials and methods Animals 22 female purpose-bred 8–9 week old kittens purchased from Liberty Laboratories were inspected upon arrival for signs of illness, examined by a veterinarian and weighed. Animals were maintained in a 2-week quarantine and observed for any signs of illness prior to the beginning of the study. IACUC number ARC 61 JAN 3. Viral Infection Plasma samples (10 5 RNA copies/ml) from a cat, that had died from an acute infection with CABCpady00C (FIV-C), were kindly provided by Dr. E. Hoover, of Colorado State University. Cats were injected I.V. with either 0.1 ml (10 5 RNA copies/ml) or 0.5 ml of plasma. Drug Dosing All procedures for care of cats during dosing as well as dosing procedures were mandated by TSRI's IACUC. Oral TL-3 (L-Iditol,1,2,5,6-tetradeoxy-1,6-diphenyl-2,5-bis [N-[(phenylmethoxy)carbonyl]-L-alanyl-L-valyl]amino]) [ 12 ] treatment was initiated in 12 cats, three days prior to infection of ten of the twelve animals with FIV-C. All TL-3 treated animals received 20 mg TL-3 by capsules at eight hour intervals, for approx. the first 7.5 weeks of the experiment. Dosage was then doubled to 40 mg TL-3 per dose at eight hour intervals for an additional week for the two control animals and the eight surviving animals in the TL-3 treated, infected cohort. No adverse effects were noted in the uninfected, TL-3 treated controls. Evoked Potentials Uninfected and FIV-infected animals were intermittently scheduled for analyses of evoked potentials in conjunction with the blood sampling, including testing for both auditory and visual evoked potentials as previously described [ 23 ]. Once recordings were complete, a blood sample was collected and animals returned to the vivarium. Clinical Evaluation Animals were examined daily and in case of health concerns further therapy/diagnostics were initiated. Animals with abnormal weight, or on antibiotics were placed on supplemental feeding with moist food and Nutrical. Dehydrated animals received subcutaneous fluid therapy. More affected animals received supportive care and medications, consisting of BID administration of antibiotics (Baytril), BID subcutaneous fluid therapy, BID temperature evaluation, BID application of antibiotic ophthalmic ointment supportive care. Any animals that required extensive supportive care (TID fluid therapy, TID force feeding) were euthanized. Euthanasia of research animals was conducted with strict adherence to NIH Office of Laboratory Animal Welfare protocols. Blood Collection and Peripheral Blood Separation Blood samples (1 ml/animal) were collected weekly during the first month of the study and then every two weeks thereafter, as described [ 13 ]. Samples were placed in EDTA blood tubes (1 cc/tube) for further use. Plasma was separated from blood by centrifugation at 3000 rpm for 5 minutes at room temperature. Blood cells were resuspended in 3 ml PBS and PBMC were separated from buffy coats by density gradient centrifugation using Ficoll-Hypaque Plus (Amersham Biosciences, Sweden). PBMC were washed once in PBS and twice in PBS/2% FBS for flow cytometry analyses. Statistical Analysis Statistical p values for the FIV-C viral load were determined by the Student's two-tailed t-test (paired, two-tailed distribution between the treated and non-treated group and the symptomatic vs asymptomatic group). Statistical p values for the weekly total CD4 and CD8 cell counts were also determined by the Student's two-tailed t-test (paired, two-tailed distribution compared to base line levels at week 0). Flow Cytometry Analysis Two-color flow cytometry analysis was performed on cells stained with mouse α-feline CD4 FITC and mouse α-feline CD8 PE (Southern Biotech, Birmingham, AL). Anti-mouse IgG 1κ FITC and PE (BD PharMingen, San Diego, CA) were used as isotype controls. Cells were fixed with 2% PFA prior to analysis performed on a FACScan flow cytometer (Beckton Dickenson Immunocytometry Systems) using the Cell Quest Software program. RNA Isolation and Reverse Transcription Plasma for weeks 0, 2, 4, 6 (terminal points for cats 215, 221, and 229), 6.5 (terminal points for cats 219, 220, and 234), and 8, were isolated from whole blood by centrifugation and stored at -20°C. Viral RNA was extracted using the QiaAmp Viral RNA Isolation Kit (Qiagen, Valencia, CA) according to manufacturer's instructions with slight modifications: Plasma samples (280 μl) were lysed in buffer AVL (1,120 μl) (Qiagen) for 10 minutes at room temperature in the presence of carrier RNA (10 μg/ml) and an external Kanamycin (KAN) RNA spike. Equal amounts of the external RNA spike (10 9 copies RNA /280 μl plasma), corresponding to the 1.2 kb KAN gene (Promega, Madison, WI), was used to normalize plasma volumes between samples and to correct for sample loss from viral RNA extraction and cDNA synthesis. An on-column DNase/ RNase free (Qiagen) incubation step for 10 minutes at room temperature was added to remove residual cellular DNA. Complimentary DNA (cDNA) was generated in a 20 μl total reaction using 13 μl of sample RNA, 0.5 μl (2 μM stock) each of KAN and FIV specific reverse primers (sequences below) and StrataScript reverse transcriptase, following the manufacturer's protocol (Stratagene, La Jolla, CA). After incubation, each cDNA sample was diluted in water to 30 μl and stored at -80°C for use in real-time PCR. Real-Time Quantitative PCR 25 μl real-time PCR reactions were set up containing 2X Platinum Quantitative PCR SuperMix-UDG (12.5 μl) (Invitrogen, Carlsbad, CA), forward, reverse primers and probe mix (7.5 μl), and cDNA target (5 μl). The mixture was incubated at 50°C for 2 minutes, 95°C for 10 minutes, then cycled at 95°C for 15 seconds and 60°C for 60 seconds 55 times, using the ABI Prism 7700 Sequence Detection System (Applied Biosystems, Foster City, CA). Data were analyzed using the ABI 7700 Sequence Detection Software. Forward and reverse primers (10 nM, 100 nM, final concentration respectively) used in real-time PCR were made at IDT, (Coralville, IA), while the fluorescein-dabsyl Amplifluor UniPrimer (100 nM final concentration) was purchased from Serologicals (Norcross, GA). The primer sequences used for real-time PCR are as follows: FIV reverse-transcriptase forward: 5'-ACTGAACCTGACCGTACAGATAAATTACAGGAA GAACCCCCATA-3' FIV reverse-transcriptase reverse: 5'-TGTTAATGGATGTAATTCA TAACCCATC-3' KAN forward: 5'-ACTGAACCTGACCGTACACGCTCAGGCGCAATCAC-3' KAN reverse: 5'-CCAGCCATTACGCTCGTCAT-3' Standard Curves and Background Detection To determine the relative copy numbers of KAN and FIV from plasma samples, a linear standard curve was generated by plotting 10-fold dilutions (5 × 10 8 to 5 × 10 2 copies per well) of dsDNA plasmids of known copy number (log scale), against the cycle threshold (Ct) determined for that value. The pET28 vector (Novagen) was used as the target plasmid for the KAN gene, while a plasmid containing the molecular clone of FIV-C was used for the FIV reverse-transcriptase gene. Calculated values for each plasma sample represent relative copy numbers for the purposes of evaluation between individual samples. Cloning, Purification and Analysis of Protease Complementary DNA (cDNA) was synthesized from isolated plasma viral RNA of infected cats. The cDNA pool was used as a template for PCR reactions using 5' primer MFIVCPL5' (5'-GATTTATAAATCATATG GCATATAATAAAGTGGGTACCACTACAACATTAG-3'), which adds an NdeI restriction site, methionine, and alkaline to the N-terminal of the protease and 3' primer MFIVCPL33' (5'-CTGAGATCTGAGCAAGCTTTTACATTACTAATCT AATATTAAATTTAACCATG TTATC-3'), which adds a stop codon and a Hind III restriction site to the C-terminus of the protease. The amplified PCR product was gel purified and cloned into pCR-TOPOII vector (Invitrogen, Carlsbad, CA) for sequencing. Selected DNA of mutants, N55D, H72R, D105G, and M107R were digested with Nde I/Hind III and ligated into pET21a expression vector (Novagen). The mutant FIV-C proteases were expressed in Rosetta pLysS cells (Novagen) and purified as previously described [ 24 ]. Enzyme kinetics of FIV-C protease were assayed on the flourogenic substrate Arg-Ala-Leu-Thr-Lys(Abz)-Val-Gln/Phe(NO 2 )-Val-Gln-Ser-Lys-Gly-Arg-NH 2 . The concentration was determined by active-site titration with inhibitor TL-3. Inhibitor constant (K i ) of TL-3 against mutant FIV-C protease was analyzed as described [ 25 ]. Competing Interests None of the authors have commercial interests or direct association with a company that is marketing TL-3. J.H.E. is a co-author of a patent application regarding use of protease inhibitors like TL-3 reported here, containing small P3 residues, as inhibitors of HIV and FIV. Authors' Contributions S.R. and C.H.S contributed equally to this work and carried out all tissue culture analyses, purification and characterization of cell populations isolated from PBMC, statistical analyses, and preparation of results for publication. Quantitative PCR analyses was carried out by D.A.S. K.J.C. was responsible for all in vivo animal work, including TL-3 administration and oversight of veterinary care for the animals. S.H-R. carried out measurements of brainstem auditory evoked potential changes. S.H., B.E.T, and J.H.E. acted as mentors for the various facets of the project, oversaw the writing of the manuscript, and provided space and funding to carry out the work.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535546.xml
526260	Pimecrolimus 1% cream for anogenital lichen sclerosus in childhood	Background Lichen sclerosus is a chronic inflammatory disease with a predilection of the anogenital region. Because of the potential side effects of repeated local application of potent glucocorticosteroids, equally-effective, safer therapeutic options are required, especially in the treatment of children. Case presentations We report on the efficacy of twice-daily application of pimecrolimus 1% cream in four prepubertal girls (range of age: 4 to 9 years) who suffered from anogenital lichen sclerosus. After three to four-month treatment, all patients had almost complete clinical remission including relief from itch, pain and inflammation. Only minor improvement was observed for the white sclerotic lesions. No significant side effects have been observed. Conclusions Topical pimecrolimus appears to be an effective and safe treatment for children with anogenital lichen sclerosus. The clinical benefits observed in the four patient presented particularly include relief of pruritus, pain and inflammation. Vehicle-controlled studies on a larger number of patients are now warranted to substantiate our promising findings, and to investigate long-term efficacy and safety of topical pimecrolimus in anogenital lichen sclerosus.	Background Lichen sclerosus (LS) is an inflammatory sclerotic skin disease of unknown origin. LS affects all age groups and occurs in about 15% in female children mostly involving the anogenital region [ 1 ]. Major subjective complains are severe pruritus, dysuria, painful defecation and vaginism. Clinically, LS is characterized by porcelain-white sclerotic plaques [ 1 ]. Histological features include orthokeratotic hyperkeratosis, vacuolar degeneration of the basal layer, oedematous and sclerotic papillary dermis as well as lymphohistiocytic infiltrates in the mid-dermis [ 2 , 3 ]. The disease runs a relapsing course indicating an ongoing inflammatory process. Although the exact pathogenesis of LS is still unclear, the recognized active involvement of skin immune system (e.g., activated T cells and CD1a+/HLA-DR+ dendritic cells) and the association with autoimmune disease and human leukocyte antigen DQ7 in women and girls with LS suggests an immunogenetic component to the disease [ 2 , 4 ]. Topical glucocorticosteroids (GCS) are the first-line therapeutic option for genital LS. Oestrogen or testosterone containing ointments are usually of limited efficacy. Surgery including cryotherapy and laser treatment should be reserved for patients with adhesions or symptomatic patients who fail to respond to multiple medical treatments, since there is a high recurrence rate following surgery [ 1 , 5 ]. Because of the chronic course of genital LS and the potential side effects associated with potent topical GCS or aggressive surgical treatments, alternative well-tolerated therapies are required, especially in the treatment of childhood LS. We therefore investigated the efficacy and safety of topical pimecrolimus in anogenital LS in four prepubertal girls. Case presentations We describe four prepubertal girls between the age of 4 and 9 years who suffered from LS of the anogenital region (Tab. 1 ). Because of the unequivocal clinical appearance and the patients' age, we renounced a biopsy proven diagnosis. All patients had extensive pruritus, two of them (patient no. 1 and 3) showed additional burning pain, dysuria and increased vulnerability with bleeding. Patient 1 suffered also from painful defecation. Three patients (patient no. 1, 2 and 4) showed partly deep fissuring in the affected areas. All patients had characteristic whitish, sclerotic skin changes in the vulvar and perianal region, three had additional involvement of the perineal region (patient no. 1, 3 and 4). Previous therapies of the patients included emollients and topical antifungal agents (patient 2 and 4). Six months before pimecrolimus treatment patient no. 4 had underwent cryotherapy and steroid instillation with a relapse after six weeks (Table 1 ). Table 1 Clinical data of patients treated with pimecrolimus 1% cream twice daily Patient (No)/age (years) Duration of disease (years) Previous therapies Clinical features pre-therapy Involvement Duration of therapy (weeks) Clinical features post-therapy 1/5 2.5 emollients a, b, c, d, e, f, g genital, perineal, perianal 12 (f) 2/4 0.3 nystatin cream a, f, g genital, perianal 12 f 3/9 0.3 polidocanol cream a, f genital, perineal, perianal 12 (f) 4/6 2.5 nystatin cream/cryotherapy/steroid instillation a, b, c, e, f, g genital, perineal, perianal 16 f a = pruritus; b = burning pain; c = dysuria; d = painful defecation; e = bleeding; f = whitish sclerotic lesions: g, = fissuring Pimecrolimus 1% cream (Elidel ® , Novartis Pharma, Basel, Switzerland) was applied twice daily in a thin layer to the affected areas. Clinical examination and recording of patients symptoms was performed before, after six weeks, and after three and four months of therapy, respectively. On the six-week follow-up visit, substantial improvement of pruritus, dysuria, and painful defecation has been reported by the patients which had already occurred during the first weeks of treatment. Accordingly skin lesions had improved as well. The patients were then encouraged to continue the treatment twice daily for further six and eight weeks, respectively (Table 1 ). At the end of therapy, almost complete remission of symptoms was achieved in all four patients, except for the white sclerotic skin changes that showed only minor improvement (patient no. 1 and 3). Side effects observed included transitory mild burning at the initiation of treatment. Two patients (no. 2 and 4) showed no recurrence of disease activity three months after discontinuation of therapy, two patients (no. 1 and 3) were lost on post-treatment follow-up. Conclusions Pimecrolimus belongs to the ascomycin class of macrolactam immunosuppressives, acting by the inhibition of T-cell activation via the calcineurin pathway and inhibition of the release of numerous inflammatory cytokines, thereby preventing the cascade of immune and inflammatory signals. In contrast to GCS, there is no potential to induce skin atrophy [ 6 ]. Pimecrolimus 1% cream has been approved for the treatment of atopic dermatitis. It has been proven to be effective in various inflammatory skin diseases, e.g., cutaneous lupus erythematosus, vitiligo and psoriasis [ 7 ]. In large studies it has been demonstrated that treatment is well tolerated in paediatric patients and even infants with atopic dermatitis. Three weeks of therapy regardless of the skin areas treated resulted in low blood concentrations without any accumulation at which no systemic effect is expected [ 6 , 8 ]. We therefore refrained from blood monitoring. Successful treatment of genital LS with calcineurin inhibitors has been reported recently. Böhm et al. [ 9 ] used topical tacrolimus 0.1% ointment in three prepubertal girls and three adults with anogenital LS. A complete remission was obtained in all patients and therapy was well tolerated. Additionally, two other case reports of genital LS who responded to tacrolimus ointment have been published [ 10 , 11 ]. Recently Goldstein et al. [ 12 ] reported for the first time a 10-year old girl with genital LS who was successfully treated with pimecrolimus ointment. The efficacy of calcineurin inhibitors in LS are mainly due to their immunosuppressive and anti-inflammatory effects. Moreover recent studies indicate a release of neuropeptides from sensory nerve fibres during tacrolimus and pimecrolimus treatment [ 13 ]. Accordingly we observed that subjective symptoms such as pruritus and pain completely resolved after a few weeks of treatment and clinical features such as fissuring, purpura, inflammatory erythema and genital bleeding almost completely resolved at the end of therapy (Table 1 ). The white sclerotic lesions could however not be changed significantly. On the basis of our case observations, pimecrolimus ointment seems to be effective in LS. However, we cannot fully exclude that the benefit observed was due to emollient effect of the ointment. For example, it has recently been demonstrated in a vehicle-controlled study that the beneficial effect of testosterone ointment claimed in the past was probably due to the emollient effect only [ 14 ]. Nevertheless all of our patients had quite severe disease and were previously treated with different emollients without significant success. Hence pimecrolimus was very likely the active part of the treatment. Furthermore in view of the natural course of LS and the severity of patients' disease spontaneous resolution can widely be excluded. In conclusion, pimecrolimus 1% seems to be a safe and effective treatment modality in pre-pubertal children with anogenital LS. Our case observations provide further evidence for the beneficial effects of topical pimecrolimus on pruritus and inflammatory features. As the recurrence rate of active LS in prepuberty is relatively high and many of the patients have continuing symptoms after menarche as well, a long-term treatment regime not based on GCS is needed to avoid the well-known side effects linked to GCS. Vehicle-controlled studies on a larger number of patients are now warranted to substantiate our promising findings, and to investigate long-term efficacy and safety of topical pimecrolimus in anogenital LS. List of abbreviations Lichen sclerosus: LS; glucocorticosteroids: GCS Competing interests The authors declare that they have no competing interests. Authors' contributions SB conceived of this case study including its conduction and drafting of the manuscript. TG and AK participated in the literature search and critically revised the manuscript. All authors read and approved the final manuscript. Figure 1 Vulvar region before (a) and after (b) 12 weeks of pimecrolimus 1% cream twice daily (Patient no. 1) Figure 2 Perianal region before (a) and after (b) 12 weeks of pimecrolimus 1% cream twice daily (Patient no. 3) Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526260.xml
522748	Allermatch™, a webtool for the prediction of potential allergenicity according to current FAO/WHO Codex alimentarius guidelines	Background Novel proteins entering the food chain, for example by genetic modification of plants, have to be tested for allergenicity. Allermatch™ is a webtool for the efficient and standardized prediction of potential allergenicity of proteins and peptides according to the current recommendations of the FAO/WHO Expert Consultation, as outlined in the Codex alimentarius. Description A query amino acid sequence is compared with all known allergenic proteins retrieved from the protein databases using a sliding window approach. This identifies stretches of 80 amino acids with more than 35% similarity or small identical stretches of at least six amino acids. The outcome of the analysis is presented in a concise format. The predictive performance of the FAO/WHO criteria is evaluated by screening sets of allergens and non-allergens against the Allermatch databases. Besides correct predictions, both methods are shown to generate false positive and false negative hits and the outcomes should therefore be combined with other methods of allergenicity assessment, as advised by the FAO/WHO. Conclusions Allermatch™ provides an accessible, efficient, and useful webtool for analysis of potential allergenicity of proteins introduced in genetically modified food prior to market release that complies with current FAO/WHO guidelines.	Background The safety of genetically engineered foods must be assessed before authorities in most nations will consider granting market approval. An important issue in current food safety assessment is the evaluation of the potential allergenicity of food derived from biotechnology. Since many food allergens are proteins, introduction of a new ("foreign") protein in food by genetic engineering can in theory cause allergic reactions. Therefore the allergenicity of novel proteins needs to be assessed. Potential allergenicity of a protein is a complex issue and various tests can be used for prediction, including bioinformatics, in vitro digestibility and binding of antisera of allergic patients. A step-by-step procedure to assess allergenicity is described by the Codex alimentarius and the FAO/WHO consultation group [ 1 , 2 ]. An important step in this procedure is to use bioinformatics to determine whether the primary structure (amino acid sequence) of a given transgenic protein is sufficiently similar to sequences of known allergenic proteins. The recommended procedure [ 1 ] to establish the possibility of allergenicity is to: (1) Obtain the amino acids sequences of known allergens in protein databases in FASTA format (using the amino acids from the mature proteins only, disregarding the leader sequences, if any). (2) Prepare the complete set of 80-amino acid length sequences derived from the query protein (again disregarding the leader sequence, if any). (3) Compare each of the sequences of (2) with all sequences of (1), using the program FASTA [ 3 ] with default settings for gap penalty and extension. According to the Codex alimentarius [ 2 ], potential allergenicity should be considered, when there is either: (a) More than 35 % similarity over a window of 80 amino acids of the query protein with a known allergen. (b) A stretch of identity of 6 to 8 contiguous amino acids. This procedure is described in more detail by the expert consultation and the Codex Alimentarius. Potential allergenicity requires further testing of the protein with panels of patient sera and possibly animal exposure tests [ 1 , 2 ]. Construction and content Three allergen databases were created, one derived from SwissProt [ 4 ] and one from the WHO-IUIS allergen list [ 5 ]. A third database is a non-redundant combination of the other two. The databases were created by extracting all proteins from public databases; SwissProt (version 44.1, July 5 2004, [ 4 ]), PIR [ 6 ] and GenPept . Leader sequences were, if annotated, trimmed from the sequence. The SwissProt allergen list contains 334 mature protein sequences, while the WHO-IUIS allergen list (version June 7, 2004) contains 632 sequences (correcting for three internal duplications). These two databases contain 236 duplicate entries. The non-redundant combined database contains 730 sequences (Figure 1 ). Allermatch™ is build around the FASTA package (version 3.4t21; , [ 3 ]) running with default parameters (ktup = 2, matrix = Blosum50, Gap open = -10, Gap extend = -2). The Allermatch™ analysis tool and the web interface are written in Python and run on a Suse L Linux Enterprise server with an Apache web server (version 1.3.26). Allermatch™ provides two search methods (mode 1 & 2) corresponding with the FAO/WHO guidelines described above and a third method (mode 3) is provided as an extra tool. The outline of the application is schematically presented in Figure 2 . Mode 1: Sliding window approach The query protein sequence is divided into 80 amino acid (aa) windows using a sliding window with steps of a single residue. Each of these windows is compared with all sequences in the allergen database of choice. All database entries showing a similarity higher than a configurable threshold percentage (default is 35%) to any of the 80 aa query sequence windows are flagged. Upon completion of the analysis, a table is shown with all flagged database entries. Per entry, the highest similarity score is given, as well as the number of windows having a similarity above the cut-off percentage. For each allergen database entry identified, more detailed information on the similarity between the allergen and query sequence can be retrieved, such as those areas of both proteins within all 80 aa windows scoring above the cut-off percentage. The similarity score calculated by FASTA can apply to stretches smaller than 80 aa, Allermatch™ converts such a similarity score to an 80 aa window. For example, 40% similarity on a stretch of 40 aa converts to 20% similarity on an 80 aa window. Mode 2: Wordmatch This method looks for short sub-sequences (words), which have a perfect identity with a database entry. The wordsize is configurable (default is 6 aa). The output given is similar to the output given by Mode 1. All database entries with at least one hit are listed and for each of these, more detailed information can be retrieved upon request. Mode 3: full FASTA alignment with an Allermatch™ allergen database The Allermatch™ webtool also offers a full alignment of the query sequence with either of the allergen databases using FASTA. Although this full alignment is currently not required by the FAO/WHO guidelines, the full alignment of protein sequences helps positioning of regions of potential allergenicity in the whole primary structure of the protein. The FASTA output is parsed and information from the allergen database is added and presented. Utility and discussion To examine the predictive performance of the FAO/WHO criteria for potential allergenicity, we have performed two tests. The first test determines the percentage of false negative and the second test assesses the amount of false positives. Both tests are performed with standard settings; for the sliding window approach an 80 amino acid window with a 35% similarity cutoff is used and for the wordmatch approach 6, 7 and 8 aa word sizes are tested. The false negative error-rate is estimated by a leave-one-out method, testing all sequences in each Allermatch™ database against that database with the tested sequence excluded. Each sequence not resulting in a hit is considered a false negative. The results of each method/database combination are summarized in Table 1 , column 1. The results show that the number of false negatives decreases when a larger database of allergen sequences is used. This may (partly) be explained by an increased proportion of similar, but not equal, sequences in the larger databases, such as isoallergens listed by WHO-IUIS. In examining the results, various sequences were observed that were not able to produce a hit (data not shown) due to their short length, since a perfect hit on a sequence shorter than 28 amino acids cannot convert to a 35% hit on an 80 amino acid window. Column 2 of the same table shows the corrected false negative rate after exclusion of these sequences. Also after this correction the wordmatch with 6 amino acids method shows lower numbers of false negatives than the sliding window approach. It is clear, however, that in case of short protein sequences the sensitivity of the sliding window methods is reduced. In the second test, we assess the odds of a false positive by testing 12 protein sequences known to be non allergenic. This is based on non-reactivity of these proteins towards IgE-sera of allergy patients or on the inability to cause IgE-responses in experimental animals (Table 2 ). It should be noted that such data are only available for a limited number of proteins, which accounts for the size of this dataset. Each of these 12 sequences was tested against all databases with all methods. Each non-allergenic sequence resulting in a hit is considered a false positive (Table 1 , column 3). The number of false positives grows with the database size, as is to be expected: the chance of a random hit increases with a larger database. In contrast to the false negative hit rates the sliding window method gives the lower error rate. This test might, however, overestimate the number of false positives. A number of these non-allergens are related to and display similarities with their allergenic counterparts, i.e. T1 (related to Bet v 1), human serum albumin (related to animal serum albumins), and human heat shock protein 70 (similar to heat shock proteins from fungi and other allergens). A selection of unrelated, non-allergenic proteins is therefore likely to give a lower false positive rate. Caution should be taken in interpreting these false hit rates. The used methods might perform differently with other sets of proteins. For example, a member of a completely novel group of valid allergens is likely to generate a false negative result. The imperfect results show here agree with literature where the FAO/WHO methods for sequence comparisons are also shown to lack full predictive capability [ 7 - 9 ]. Interestingly, the results show that there is a balance between false positives and negatives when increasing the threshold level for short exact matches from 6 to 8 amino acids, with the number of false positives sharply decreasing at 8 amino acids (Table 1 ). The outcomes of these tests therefore need to be further refined by checking for the presence of potential IgE-epitopes as recommended by Kleter and Peijnenburg [ 7 ], as well as combined with results of other assays as recommended by the Codex. Other methods to decrease false hit rates may also be considered [ 8 , 9 ]. We plan to implement such supplementary methods in the future to support the Codex based predictions of potential allergenicity. The prediction of potential allergenicity by primary sequence comparison depends on the quality of the data used for comparison. Addition of a non-allergenic or poorly annotated protein to any of the Allermatch™ allergen databases would obviously result in undesired false positives and should be prevented. A workable strategy could be to use multiple databases, i.e. a database based on SwissProt's list of allergens, which contains well-annotated sequences from SwissProt, simultaneously with a larger database based on the WHO-IUIS list, which contains possibly less well annotated sequences from other protein databases, such as GenPept. For example, a number of protein accessions in the WHO-IUIS database do not mention the presence of signal- and/or pro-peptides, where removal of such peptides is essential to prevent false positives. Users of Allermatch™ should, at all times, take into account the possibility of a false positive or negative, for example by checking original data (accessions, clinical literature) and confirm results, before arriving at conclusions. To prevent false positives as much as possible, one should choose for the well-annotated SwissProt database. To prevent false negatives, the combination of the larger WHO-IUIS database with that of SwissProt is more appropriate. Updates to the SwissProt and WHO-IUIS allergen lists will be incorporated in the Allermatch™ databases on a regular basis. Several other websites in the public domain offer sequence alignment facilities that support the prediction of potential allergenicity, such as SDAP [ 10 , 11 ], AllerPredict [ 12 ] and Farrp [ 13 ]. These websites offer search algorithms that find contiguous similar amino acids between a query sequence and database sequences (SDAP, AllerPredict) and more than 35% identity in alignments (SDAP, AllerPredict), as well as a general FASTA of a query protein sequence against the database (SDAP, Farrp). Conclusions Allermatch™ is an efficient and comprehensive webtool that combines all bioinformatics approaches required to assess the allergenicity of protein sequences according to the current guidelines in the Codex. The application will be kept up to date with the FAO/WHO criteria and the SwissProt and WHO-IUIS allergen lists. It will be extended with other, supplementary methods to support and refine the prediction of allergenicity. Availability and requirements Allermatch™ is platform independent and accessible using any Netscape 4+ compatible webbrowser at . Authors' contributions MF developed and implemented the Allermatch™ webtool. HN provided the domain name registration and advised in the web site development. GK and AP provided the scientific background and constructed the sequence databases. JPN and RvH provided time, resources and ample discussion. All authors have read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC522748.xml
509283	Autoimmune progesterone dermatitis in a patient with endometriosis: case report and review of the literature	Autoimmune progesterone dermatitis (APD) is a condition in which the menstrual cycle is associated with a number of skin findings such as urticaria, eczema, angioedema, and others. In affected women, it occurs 3–10 days prior to the onset of menstrual flow, and resolves 2 days into menses. Women with irregular menses may not have this clear correlation, and therefore may be missed. We present a case of APD in a woman with irregular menses and urticaria/angioedema for over 20 years, who had not been diagnosed or correctly treated due to the variable timing of skin manifestations and menses. In addition, we review the medical literature in regards to clinical features, pathogenesis, diagnosis, and treatment options.	Introduction While many women complain of worsening acne and water retention during their menstrual cycle, there exist a small number in whom the menstrual cycle is associated with a variety of other skin manifestations such as urticaria, eczema, folliculitis, and angioedema. This condition is known as autoimmune progesterone dermatitis (APD) due to the fact that progesterone is most frequently identified as the etiologic agent. In women with irregular menses, the diagnosis may remain elusive for years. We present a case of APD, and review the current literature in regards to clinical features, pathogenesis, diagnosis, and treatment options. Case A 33y/o woman with a history of endometriosis presented with complaints of chronic urticaria. The patient noted that the urticaria began at the age of 12, and did not seem to have any obvious trigger. Each individual lesion would last from 12–24 hours, and the entire episode would last 5–10 days. Lesions would usually start on the chest and then spread over the entire body. She had seen multiple physicians, including allergists and dermatologists, and had been treated with a variety of medications including certirizine, desloratadine, hydroxyzine, montelukast, ranitidine, and diphenhydramine without relief. Prednisone at high doses would provide temporary relief, and she had required multiple courses of prednisone over the past 20 years. In addition, she complained of occasional angioedema, usually at the same time as the hives but occasionally occurring when hives were not present. The patient also had acne that had been very difficult to control since her teenage years, and she noted that the acne would also respond to prednisone. Multiple lab tests over the years had been unremarkable. These included SSA/SSB, anti-Smith, ACE level, C3/C4, hepatitis B, ANA, anti double-stranded DNA, immunoglobulins, SPEP, C1 esterase inhibitor level and function, chemistry panel, liver tests, TSH, T4, thyroid antibodies, rheumatoid factor, ESR, and CBC. Skin biopsy of a lesion had been read as "chronic urticaria". Upon further questioning, it was learned that due to the patient's endometriosis, she had very irregular menstrual cycles in terms of length and timing. It was determined that the hives and/or angioedema would begin approximately 4 days prior to the onset of menses, and would last about 2 days into menses. The symptoms would not occur with every episode of menses. The patient's acne would often occur on her face during the urticarial episodes. Of note, the patient had 2 children, and during each pregnancy her hives, acne, and angioedema had been markedly improved. Because of her endometriosis, she had been started on Depo-Provera (medroxyprogesterone acetate) in her twenties. After 1 injection, she developed severe hives that lasted over 2 months and required multiple courses of prednisone. Due to the urticaria, Depo-Provera was discontinued after one injection. As the patient complained of acne, Ortho Tri-Cyclen (norgestimate/ethinyl estradiol) was initiated by her dermatologist. This treatment modality did not have any effect on the urticaria, angioedema, or acne. The patient was evaluated in our clinic. Physical examination was essentially normal, and no hives were noted. Allergy skin testing was performed with progesterone 50 mg/mL in normal saline. Prick test was normal, but a full strength intradermal test revealed a 7 mm wheal with erythema. The histamine control showed a 9 mm wheal with erythema, and saline control was negative for wheal and erythema. Two healthy controls also underwent intradermal testing to exclude irritant reaction, and were found to be negative Based on the above results, the patient was diagnosed with autoimmune progesterone dermatitis. The patient was started on a GnRH agonist (nafarelin acetate nasal spray, 200 mcg twice a day). Within one month, she noted dramatic improvement in her urticaria and angioedema. Acne was still occasionally present, but much improved. She did complain of mild hot flashes, but felt these were tolerable. Discussion In a small group of women, the menstrual cycle has been associated with a spectrum of dermatologic diseases including eczema, erythema multiforme, stomatitis, papulopustular lesions, folliculitis, angioedema, urticaria, and others (Table 1 ) [ 1 - 8 ]. As progesterone sensitivity has been the most commonly identified cause, dermatologic diseases associated with the menstrual cycle have been labeled autoimmune progesterone dermatitis (APD) [ 4 ]. The first documented case of APD was in 1921, in which a patient's premenstrual serum caused acute urticarial lesions. In addition, it was shown that the patient's premenstrual serum could be used to desensitize and improve her symptoms [ 9 ]. Since 1921, approximately 50 cases of APD have been published in the medical literature. Table 1 Dermatologic manifestations of autoimmune progesterone dermatitis - Urticaria - Angioedema - Eczema - Erythema multiforme - Stomatitis - Folliculitis - Papulopustular/papulovesicular lesions - Stephens-Johnson syndrome - Vesiculobullous reactions - Dermatitis herpetiformis-like rash - Mucosal lesions Clinical Features The clinical symptoms of APD (eczema, urticaria, angioedema, etc.) usually begin 3–10 days prior to the onset of menstrual flow, and end 1–2 days into menses. Severity of symptoms can vary from nearly undetectable to anaphylactic in nature, and symptoms can be progressive [ 10 , 11 ]. There are no specific histological features on biopsy in APD [ 12 ]. The age of onset is variable, with the earliest age reported at menarche [ 13 ]. Some studies have noted that a majority of patients had taken an oral contraceptive (OCP) prior to the onset of APD [ 14 ], but multiple cases exist in which women have never been exposed to exogenous progesterone [ 15 - 17 ]. The symptoms of APD correlate with progesterone levels during the luteal phase of the menstrual cycle. Progesterone begins to rise 14 days prior to the onset of menses, peaks 7 days prior to menses, and returns to a low baseline level 1–2 days after menses begins. In studies where an etiologic agent has been sought, progesterone has been found most frequently. However, estrogen, prostacyclin, and gonadotropin levels have correlated with symptoms in some cases [ 18 - 21 ]. Symptoms may first appear, improve, or worsen during pregnancy and the peripartum period [ 2 , 22 - 24 ]. In addition, APD during pregnancy has been associated with spontaneous abortions [ 2 , 25 ]. Pregnancy is associated with an increase of maternal progesterone levels, which may explain the initiation or worsening of symptoms. In regards to an improvement of symptoms during pregnancy, a number of theories have emerged. Explanations include a slow rise of progesterone during pregnancy that acts as a method of desensitization, a decrease in maternal immune response during pregnancy, or an increased production of anti-inflammatory glucocorticoids [ 13 , 25 , 26 ]. Pathogenesis The exact pathogenesis of APD is unknown. If exogenous progesterones (i.e. OCPs) are initially used, it is conceivable that uptake by antigen presenting cells and presentation to T H 2 cells could result in subsequent IgE synthesis; however this mechanism would not explain the pathogenesis in patients such as ours who have the onset of APD prior to exogenous progesterone exposure. Some authors have suggested that hydrocortisone or 17-α-hydroxyprogesterone have cross-sensitivity with progesterone and may cause initial sensitization, but this has not been observed in all studies [ 27 , 28 ]. To further delineate the pathogenesis, antibodies against progesterone have been investigated. Using immunofluorescent techniques and basophil degranulation tests, studies have found that such antibodies do exist in certain patients with APD [ 1 , 13 , 29 ]. However, negative results looking for antibodies have also been reported [ 24 ]. In addition, skin test results with progesterone have shown immediate reactions (within 30 minutes), delayed reactions (24–48 hours later), and reactions with features of both immediate and delayed features [ 13 , 14 , 30 , 31 ]. This presumably indicates both type I and type IV hypersensitivity reactions. Progesterone has also been reported to have a direct histamine releasing effect on mast cells, yet very little research has been done to support this hypothesis [ 32 ]. Additionally, one study found an in vitro increase of an interferon-γ release assay, possibly implying a role for T H 1-type cytokines in APD [ 33 ]. Eosinophils may also be involved in the pathogenesis of APD. Eosinophilia has been correlated with cutaneous symptoms in some cases, and studies have found a decrease in total eosinophil count after therapy [ 13 , 29 , 34 ]. Whether increased eosinophils are a response to cytokines from lymphocytes or play a primary mechanistic role in APD remains to be determined. Diagnosis The diagnosis of APD requires an appropriate clinical history accompanied by an intradermal injection test with progesterone. An aqueous suspension or aqueous alcohol solution of progesterone is the preferable vehicle of testing as progesterone in oil can cause an irritant reaction [ 35 ], though many published case reports have used progesterone in oil for testing. Various authors have advocated different amounts of progesterone or medroxyprogesterone to be used for testing [ 12 , 33 , 36 ]. As had been done in some prior studies, the patient presented here was tested with progesterone in aqueous solution at a concentration of 50 mg/mL. As mentioned above, APD may be due to an immediate or delayed hypersensitivity reaction. Therefore, intradermal testing may not become positive until 24–48 hours later [ 14 , 24 ]. In addition, some authors have advocated patch testing with progesterone to further evaluate for a hypersensitivity reaction [ 33 ]. Of note, intradermal testing has been negative in some patients with typical clinical symptoms of APD and who improved after APD treatment [ 2 , 3 , 24 ]. Some authors have recommended further tests to evaluate the immunologic evidence in APD. These include circulating antibodies to progesterone, basophil granulation tests, direct and indirect immunofluorescence to luteinizing cells of the corpus luteum, in vitro interferon-γ release, and circulating antibodies to 17-α-hydroxyprogesterone [ 1 , 7 , 13 , 29 , 33 , 36 ]. However, most case reports in the medical literature do not routinely check for serologic evidence of APD, and when checked these markers have not always been found to be reliable. This is most likely due to the fact that, as mentioned above, the pathogenesis of APD is incompletely understood. Treatment Autoimmune progesterone dermatitis is usually resistant to conventional therapy such as antihistamines. The use of systemic glucocorticoids, usually in high doses, has been reported to control the cutaneous lesions of APD is some studies, but not in others [ 3 , 10 , 37 ]. Early reports of APD describe attempts of progesterone desensitization, and some authors even attempted injections derived from the corpus luteum [ 18 , 24 , 38 ]. However, results were usually temporary, and such methods of treatment have now fallen out of favor. Current therapeutic modalities often attempt to inhibit the secretion of endogenous progesterone by the suppression of ovulation. Table 2 lists some of the pharmacologic strategies used in APD. Oral contraceptives (OCPs) are often tried as initial therapy, but have had limited success, possibly due to the fact that virtually all OCPs have a progesterone component. Conjugated estrogens have also been used in the treatment of APD. These did show improvement in many of the patients, but often required high doses [ 2 , 16 , 22 ]. However, due to the increased risk of endometrial carcinoma with unopposed conjugated estrogens, this treatment is not commonly used today [ 39 ]. Table 2 Treatment options used in autoimmune progesterone dermatitis Treatment Option Advantages Disadvantages Oral Contraceptives (OCPs) - Usually tried as initial therapy - Limited success due to the progesterone component of OCPs - Fewer side effects than other most other therapies Antihistamines - Well tolerated, few side effects - Rarely effective as monotherapy - Does not address underlying mechanism Conjugated Estrogens - Avoids progesterone component of OCPs - Increased risk of endometrial cancer, not commonly used today - Often require high doses Glucocorticoids - Able to suppress multiple components of the immune system - Usually not effective alone - Can be combined with other therapies - Often require high doses GnRH Agonists - Often used if OCPs and glucocorticoids are not effective - Can cause symptoms of estrogen deficiency (hot flashes, decreased bone mineral density) Alkaylated Steroids - Can be combined with low dose steroids - Can cause symptoms of excess androgens (facial hair, hepatic dysfunction, mood disorders) - Interferes with gonadal hormone receptors Tamoxifen - Has been used successfully in patients unresponsive to conjugated estrogen - Can cause symptoms of estrogen deficiency - Increased risk of venous thrombosis and cataract formation Bilateral oopherectomy - Definitive treatment, used if medical options unsuccessful - Surgical procedure, associated morbidity - Symptoms of estrogen deficiency Various other therapy modalities are currently used in APD, and there is no clear treatment of choice. GnRH agonists, such as buserelin and triptorelin, have been used to induce remission of symptoms by causing ovarian suppression [ 7 , 11 , 15 ]. However, side effects include symptoms of estrogen deficiency (hot flashes, vaginal dryness, decreased bone mineral density), and estrogen supplementation may be needed [ 40 ]. Alkaylated steroids such as stanozol have been used to successfully suppress ovulation, sometimes in combination with chronic low doses of corticosteroids [ 37 ]. Side effects of alkaylated steroids include abnormal facial or body hair growth, hepatic dysfunction, and mood disorders, any of which may limit their use. To decrease the risk of side effects, some authors have recommended using the alkaylated steroid only in the perimenstrual period [ 37 ]. Another therapeutic option used in APD has been the antiestrogen tamoxifen, which also can suppress ovulation [ 3 , 5 ]. As with GnRH agonists, patients on tamoxifen may experience symptoms of estrogen deficiency. In addition, tamoxifen has been associated with an increased risk of venous thrombosis and cataract formation. In some patients with unremitting symptoms of APD, bilateral oopherectomy has been required [ 10 , 15 , 24 ]. While this definitive treatment has been successful in controlling symptoms, today it is rarely used before all medical options have been exhausted. Conclusion Autoimmune progesterone dermatitis is a condition seen in a small number of women who present with eczema, erythema multiforme, stomatitis, papulopustular lesions, folliculitis, angioedema, urticaria, and other skin manifestations in relation to the menstrual cycle. It is usually seen 3–10 days prior to the onset of menstrual flow, but may be difficult to recognize in women with irregular menses. The exact pathogenesis is unknown, and is thought to involve a hypersensitivity reaction to progesterone. The diagnosis of APD is made by an appropriate clinical history accompanied by an intradermal injection test with progesterone. Current treatment modalities often attempt to inhibit the secretion of endogenous progesterone, but may be unsuccessful. More research is needed into the pathogenesis of APD to most appropriately care for these patients.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509283.xml
539342	Dispersal or Drift? More to Plant Biodiversity Than Meets the Eye	null	Over 250 million years ago (mya), all the continents of Earth formed a single land mass called Pangaea. Some 50 million years later, this supercontinent began to split in two, forming Laurasia—now North America, Asia, and Europe—and Gondwana—present-day Antarctica, Australia, South America, Africa, and India. After another 50 million years, Gondwana, too, broke up. At the end of the Cretaceous period, New Zealand split off (about 80 mya), then South America and Australia separated from Antarctica (about 35 mya). Fairy-tale quality aside, the story of continental drift fits comfortably with the geological and fossil record and feeds our understanding of current distributions of plant biodiversity. Although we know how and when Pangaea broke apart, the distribution of fossils of the same species on many different continents, separated by vast ocean waters, challenges us to explain how they got there. Plant life on New Zealand, for example, shares striking similarities to that on other Southern Hemisphere land masses, but scientists have yet to agree on how this came to pass. In particular, one genus, Nothofagus —the southern beech tree, a plant whose 80-million-year-old fossil history goes back to the days of Gondwana—has polarized views on the nature of Southern Hemisphere biogeography. One theory suggests that geographic barriers (New Zealand and Australia are separated by the Tasman Sea) would have prevented species expansion after the break-up of the continents, so similar contemporary species must have already existed in both places before New Zealand broke away from Gondwana. In this scenario, called vicariance, ancestors of existing lineages drifted with the repositioned land masses. Another hypothesis, born of existing distributions and fossil data, suggests that long-range oceanic dispersal is more likely. But since Nothofagus seeds are not considered ocean-worthy vessels, many believe vicariance is the only possible explanation. Peter Lockhart and colleagues argue that a clear picture of the divergence dates of various southern beech species could help clarify the relative contributions of vicariance versus dispersal. But they would need significant lengths of DNA sequences to reliably characterize the evolutionary history of each species. Consequently, Lockhart and colleagues analyzed a 7.2-kilobase fragment of the chloroplast genome (which typically ranges from 110,000 bp to 160,000 bp) for 11 species of three Nothofagus subgenera— Lophozonia , Fuscospora , and Nothofagus —from South America, Australia, and New Zealand. Reconstructing the trees' evolutionary relationships (phylogeny) based on analyses of their chloroplast sequences, the authors discovered a nuanced evolutionary history that supports vicariance for some species and dispersal for others. Nothofagus , the southern beech, on the slopes of Mt. Ruapehu in New Zealand (Photo: Peter Lockhart) Assuming that beech was present throughout Gondwana (which fossil data support), the sequence of the Gondwana breakup should be reflected in the beech's phylogeny. New Zealand beeches should be more distantly related to both Australian and South American species, because of the greater period of separation—65 million years compared to 30 million years. Yet Australian and New Zealand beeches are more closely related to each other than to South American species, which reflects more recent relationships. Given that fossils of all beech subgenera extend back to the New Zealand Cretaceous period, the dating of splits and the nature of the relationships indicate extinction of beech lineages within current subgenera in New Zealand, and possibly in Australia and South America. Lockhart and colleagues' analyses suggest that the relationships of the Australian and New Zealand Lophozonia and Fuscospora species are too recent to have roots in Gondwana, indicating a role for transoceanic dispersal. The evolutionary relationship between the Australasian and South American Fuscospora lineages, however, is consistent with vicariance. These divergence results, the authors conclude, indicate that current distributions of Nothofagus cannot be explained solely by continental drift (followed by extinction of some species) and that contemporary New Zealand Nothofagus species are not direct descendants of the beeches thought to have reached the island after the split from Antarctica. Taken together, the results highlight the need for caution in evaluating fossil evidence. The fossil record doesn't necessarily capture when a species first appeared, and a continuous fossil presence can mask extinctions and reinvasions. The authors conclude that their molecular data make the case for investigating possible mechanisms of long-range dispersal—especially the dispersal properties of Nothofagus seeds—and stresses the need to consider more complex hypotheses to explain something as dynamic and complex as the evolutionary history of biodiversity.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539342.xml
544354	Educational and economic determinants of food intake in Portuguese adults: a cross-sectional survey	Background Understanding the influences of educational and economic variables on food consumption may be useful to explain food behaviour and nutrition policymaking. The aim of this study was to evaluate the importance of educational and economic factors in determining food pattern in Portuguese adults. Methods A cross-sectional study in a representative sample of Portuguese adults (20977 women and 18663 men). Participants were distributed in four categories according to years of education (≤4, 5–9, 10–12, and >12) and income (≤314 euros, 315–547 euros, 548–815 euros, and >815 euros). Logistic regression models were fitted to estimate the magnitude of the association between food groups and education/income, adjusting for confounders. Results In both genders, the odds favouring milk, vegetable soup, vegetables, fruit, and fish consumption, increased significantly with education, for those having >12 years of education compared to those with ≤4 years; the odds favouring wine, and spirits consumption decreased significantly with education, for those having >12 years of education compared to those with ≤4 years. In males, the odds favouring starchy foods and meat consumption decreased significantly with income, while for milk, the odds increased with higher income (those having >815 euros compared to those with ≤314 euros). Conclusions The low and high income groups are or tend to be similar in regard to several food groups consumption, and access to education/information appears to be the key element to a better food pattern as indicated by higher frequency of milk, vegetable soup, vegetables, fruit, and fish consumption.	Background There is a large published literature on associations between socio-economic position and chronic disease, with socioeconomically disadvantaged groups experiencing higher mortality and morbidity rates for coronary heart disease, noninsulin dependent diabetes mellitus and some cancers [ 1 - 4 ]. Chronic diseases are largely preventable diseases, and diet has been known fore many years to play a key role as a risk factor for chronic diseases. While age, sex and genetic susceptibility are non-modifiable, many of the risks associated with age and sex are modifiable. Such risks include a complex mixture of interacting socio-economic, cultural and other environmental factors [ 5 - 8 ]. The relationship between socio-economic factors and diet has been examined on the basis of food and nutrient intake and the results are mixed, sometimes contradictory, and often the observed differences are small [ 9 , 10 ]. When differences are found, it is usually the case that persons from socioeconomically disadvantage backgrounds have food intakes consistent with their higher rates of chronic diseases [ 11 - 17 ]. In the context of European countries, economic development and increased purchasing power have recently changed the food availability situation [ 18 ]. On the other hand, socio-cultural influences may contribute, along with economic constraints, to particular food choices, which may explain the still substantial differences in food consumption across European countries [ 19 ]. Understanding the influences of socio-economic variables on food consumption may be useful to predict the outcome of interventions, to change food behaviour, and generate hypotheses concerning food consumption in diverse circumstances, as well as to explain observations in epidemiological studies. The aim of this study was to evaluate the importance of educational and economic factors in determining food choice in a representative sample of the Portuguese general adult population. Methods Subjects and general characteristics Data for this study were drawn from the Portuguese third National Health Survey (National Health Systems Observatory, National Institute of Health – Dr. Ricardo Jorge, Ministry of Health) carried out in 1998–1999. The study sample included all subjects (20977 women and 18663 men) older than 18 years, who reported their education level, income, physical activity, smoking habits, weight, height, and food intake when participating in the National Health Survey. Subjects were selected from 21808 households distributed according to the five regions of Portugal (there are five regions in mainland Portugal, namely Norte, Centro, Lisboa/Vale do Tejo, Alentejo, and Algarve; these regions are the portuguese NUTS II subdivisions), using a multi-stage random probability design. This probabilistic sample is representative of the Portuguese population from the Continental area (Azores and Madeira islands were not included). The survey response rate was 82%. Trained interviewers conducted face-to-face interviews with the person in each household and inquired participants on social and demographic characteristics, smoking status (non-smokers, ex-smokers, smoking less than one cigarette per day, and smoking one or more cigarettes per day), weight, height (those anthropometric measures were self-reported and body mass index – BMI – was then calculated), food and beverages intake, and daily physical activity (occupational and leisure-time physical activity). Physical activity Occupational physical activity was measured using the respondent's own occupation at the time of the survey. Respondents were asked about what best characterized their daily occupational activity, namely: usually seated and walking during short periods of time; standing activities or walking during long periods of time without carrying loads to often; carrying light objects or walking upstairs/downstairs several times; heavy physical work or carrying heavy objects; or don't know. Respondents were asked to describe their leisure-activity using the following classification: heavy training and competitive sports more than once a week; running or practicing recreational sports or gardening activities ≥4 hours per week; walking for pleasure, bicycling (light effort) or doing other light activities ≥4 hours per week; reading, watching television or other sedentary activities; and don't know. Respondents were also asked to provide information about whether they had regular activities (once or more per week) such as running or bicycling (enough to make them feel tired). Food and beverages intake Respondents were asked twelve questions related to their intake of central food groups and beverages, namely milk, vegetable soup, meat, fish, vegetables, fruit, bread, starchy foods (pasta/rice/potatoes), beer, spirits, Port Wine, and wine, and the consumption was recorded as a yes (when the respondent indicated the consumption of the food) or no answer. Because the data were collected by interviewers within the framework of an epidemiological study that was not specifically designed to assess quantitative aspects of nutritional and food intake, the dietary assessment method employed generic classifications of food groups, rather than specific varieties or species (fish rather than fatty fish or salmon, etc.), or quantitative measures. Consumption of these food items was determined by asking "For each of the listed food items please indicate those consumed": "during the day before the interview" (vegetable soup, meat, fish, vegetables, fruit, bread, and starchy foods – pasta, rice and potatoes); "during the week before interview" (beer, spirits, and Port Wine); and "daily consumed" (milk and wine). Education Respondents were asked to provide information about whether they had attained further education since leaving school and if so, the highest qualification completed. Respondent's education was subsequently classified in four levels of education: less than 4 years, 5–9 years, 10–12 years, and more than 12 years. Income Respondent's were asked to estimate the total income (including pensions, allowances and investments) received by all household members in the last month and to indicate this using a single measure comprising ten narrow-ranged income categories. This measure was subsequently re-coded into four categories according the number of salaries: less than 315 euros, 315–547 euros, 548–815 euros, and >815 euros. Statistical analysis Separate logistic regression models were fitted for male and female to estimate the magnitude of the association between food groups consumption and education or income categories, adjusting for age, BMI, smoking habits, physical activity and income/education. An exploratory approach was chosen in the selection of explanatory variables in order to control for as many potentially significant variables as possible in the regression model. The choice of variables (age, BMI, smoking habits and physical activity) was based on findings reported in the literature, our own experience with specifically Portuguese factors associated with food consumption and their associations with the variables of interest; education was also adjusted for income and vice versa. Student's t-tests, ANOVA, Spearman rank correlation analyses and chi-squared tests were used to compare BMI, age, frequency of smoking, physical activity categories between genders to determine the degree to which those variables correlated with education and income. A p -value of less than 0.05 was considered statistically significant. Statistics were performed using SPSS 12.0. Results The study sample comprised 20977 women (52,9%) and 18663 men, with mean ages of 50.3 (±18.88) and 47.7 (±18.51) years, respectively; BMI was significantly lower in women than in men (25.1 ± 4.53 Kg/m 2 versus 25.6 ± 3.83, p < 0.001). There was a lower proportion of smokers among women compared to men (8.2% versus 30.5%, p < 0.001). General characteristics (gender, age, BMI, smoking status, and physical activity) by education and income categories are presented in Tables 1 and 2 . Table 1 Characteristics of Portuguese adults by education categories Education ≤4 years 5–9 years 10–12 years >12 years Gender Female 54.0% 22.8% 12.6% 10.6% Male 49.7% 28.7% 13.2% 8.4% Age (years) Female 55.6 ± 14.47 37.8 ± 14.44 31.7 ± 13.63 34.7 ± 13.50 Male 55.3 ± 14.72 37.0 ± 14.99 33.0 ± 14.44 38.7 ± 16.19 BMI (Kg/m 2 ) Female 26.4 ± 4.45 24.3 ± 4.03 22.5 ± 3.51 22.3 ± 3.45 Male 26.3 ± 3.91 25.1 ± 3.63 24.5 ± 3.52 24.6 ± 3.19 Smokers Female 3.1% 16.1% 19.1% 18.7% Male 26.0% 42.9% 32.6% 29.2% Physical activity in females Daily occupational activity best characterized by Usually seated, walking short periods of time 29.2% 28.8% 30.2% 30.4% Standing, walking long periods of time 44.5% 46.1% 41.7% 42.0% Carrying light objects, walking up/downstairs 11.3% 11.0% 12.2% 11.1% Heavy physical work or carrying heavy objects 14.9% 14.1% 15.8% 16.5% Don't know 0.1% 0% 0.1% 0.1% Leisure-activity best characterized by Heavy training/competitive sports (>1x/wk) 3.4% 3.3% 3.6% 3.1% Running/recreational sports/gardening (≥4 h/wk) 9.3% 7.8% 9.0% 9.7% Walking for pleasure, bicycling light (≥4 h/wk) 20.4% 21.2% 19.3% 21.2% Reading, watching TV, sedentary activities 66.8% 67.7% 68.0% 66.0% Don't know 0.1% 0% 0.1% 0% Regular activity such as running or bicycling, enough to feel tired Yes 13.1% 11.9% 13.2% 14.2% No 86.8% 88.1% 86.7% 85.8% Don't know 0.1% 0% 0.1% 0% Physical activity in males Daily occupational activity best characterized by Usually seated, walking short periods of time 36.4% 37.3% 37.5% 36.2% Standing, walking long periods of time 51.6% 51.5% 50.9% 50.0% Carrying light objects, walking up/downstairs 8.3% 8.2% 7.5% 10.5% Heavy physical work or carrying heavy objects 3.6% 3.0% 3.9% 3.3% Don't know 0.1% 0% 0.2% 0% Leisure-activity best characterized by Heavy training/competitive sports (>1x/wk) 0.9% 0.5% 1.1% 1.2% Running/recreational sports/gardening (≥4 h/wk) 5.1% 4.6% 4.1% 4.9% Walking for pleasure, bicycling light (≥4 h/wk) 15.9% 15.7% 16.8% 16.7% Reading, watching TV, sedentary activities 78.1% 79.1% 77.8% 77.2% Don't know 0.1% 0.1% 0.2% 0% Regular activity such as running or bicycling, enough to feel tired Yes 7.0% 6.0% 6.7% 7.2% No 92.9% 94.0% 93.2% 92.8% Don't know 0.1% 0.0% 0.2% 0% Table 2 Characteristics of Portuguese adults by income categories Income ≤314 euros 315–547 euros 548–815 euros >815 euros Gender Female 16.3% 25.4% 24.5% 33.8% Male 20.8% 24.3% 22.8% 32.2% Age (years) Female 50.7 ± 18.97 50.3 ± 18.90 50.1 ± 18.68 50.2 ± 18.94 Male 47.9 ± 18.54 47.2 ± 18.45 47.6 ± 18.44 48.1 ± 18.55 BMI (Kg/m 2 ) Female 25.3 ± 4,55 25.0 ± 4.51 25.1 ± 4.52 25.1 ± 4.55 Male 25.6 ± 3.70 25.4 ± 3.95 25.6 ± 3.74 25.7 ± 3.88 Smokers Female 7.4% 8.0% 8.6% 8.2% Male 28.5% 31.9% 30.3% 30.7% Physical activity in females Daily occupational activity best characterized by Usually seated, walking short periods of time 30.3% 26.0% 24.6% 35.7% Standing, walking long periods of time 46.4% 43.8% 42.8% 43.6% Carrying light objects, walking up/downstairs 10.3% 11.4% 13.5% 10.0% Heavy physical work or carrying heavy objects 12.7% 18.7% 19.0% 10.6% Don't know 0.2% 0% 0.1% 0.1%% Leisure-activity best characterized by Heavy training/competitive sports (>1x/wk) 1.2% 2.3% 3.1% 5.4% Running/recreational sports/gardening (≥4 h/wk) 4.0% 7.3% 11.1% 11.1% Walking for pleasure, bicycling light (≥4 h/wk) 16.7% 19.2% 19.0% 24.0% Reading, watching TV, sedentary activities 77.9% 71.2% 66.7% 59.5% Don't know 0.1% 0% 0.1% 0.1% Regular activity such as running or bicycling, enough to feel tired Yes 5.8% 9.1% 12.0% 19.1% No 94.1% 90.9% 87.9% 80.8% Don't know 0.1% 0% 0.1% 0.1% Physical activity in males Daily occupational activity best characterized by Usually seated, walking short periods of time 41.8% 33.5% 32.2% 39.3% Standing, walking long periods of time 48.1% 53.2% 53.1% 50.8% Carrying light objects, walking up/downstairs 7.6% 9.6% 10.0% 7.0% Heavy physical work or carrying heavy objects 2.4% 3.6% 4.7% 2.8% Don't know 0.1% 0.1% 0% 0% Leisure-activity best characterized by Heavy training/competitive sports (>1x/wk) 0.3% 0.5% 0.8% 1.4% Running/recreational sports/gardening (≥4 h/wk) 2.9% 4.4% 5.5% 5.4% Walking for pleasure, bicycling light (≥4 h/wk) 10.3% 13.1% 16.3% 21.5% Reading, watching TV, sedentary activities 86.4% 81.9% 77.4% 71.7% Don't know 0.1% 0.2% 0% 0% Regular activity such as running or bicycling, enough to feel tired Yes 2.2% 4.7% 5.4% 11.7% No 97.6% 95.2% 94.6% 88.3% Don't know 0.1% 0.1% 0% 0% In women, the odds favouring milk, vegetable soup, vegetables, fruit, and fish consumption, increased with increasing education (p-values for trends were always ≤0.046), being the odds ratios, respectively, 2.60 (2.24–3.01), 1.20 (1.05–1.38), 1.75 (1.44–2.13), 1.92 (1.49–2.49), and 1.40 (1.23–1.60) for those having >12 years of education compared to those with ≤4 years, after adjusting for age, BMI, smoking habits, physical activity and income (Tables 3 and 4 ). The odds favouring bread, starchy foods (other than bread), wine, and spirits consumption in women decreased with increasing education (p trend ≤ 0.002), being the odds ratios, respectively, 0.44 (0.34–0.56), 0.68 (0.53–0.87), 0.51 (0.41–0.62), and 0.13 (0.03–0.53) for those having >12 years of education compared to those with ≤4 years (Tables 3 and 4 ). Table 3 Odds ratios for food consumption according level of education, adjusted for age, BMI, smoking habits, physical activity and income Women Men OR IC(95%) P trend OR IC(95%) P trend Vegetable soup Vegetable soup Education Education ≤4 years (reference) ≤4 years (reference) 5–9 years 0.90 0.82–1.00 5–9 years 0.99 0.90–1.09 10–12 years 0.94 0.83–1.07 10–12 years 1.07 0.94–1.21 >12 years 1.20 1.05–1.38 0.046 >12 years 1.15 1.00–1.32 0.045 Vegetables Vegetables Education Education ≤4 years (reference) ≤4 years (reference) 5–9 years 1.05 0.92–1.20 5–9 years 1.09 0.97–1.23 10–12 years 1.17 0.99–1.39 10–12 years 1.38 1.17–1.62 >12 years 1.75 1.44–2.13 <0.001 >12 years 1.44 1.19–1.74 <0.001 Fruit Fruit Education Education ≤4 years (reference) ≤4 years (reference) 5–9 years 1.30 1.09–1.55 5–9 years 1.30 1.13–1.49 10–12 years 1.63 1.29–2.06 10–12 years 1.75 1.45–2.13 >12 years 1.92 1.49–2.49 <0.001 >12 years 1.68 1.35–2.10 <0.001 Bread Bread Education Education ≤4 years (reference) ≤4 years (reference) 5–9 years 0.78 0.63–0.96 5–9 years 0.97 0.75–1.25 10–12 years 0.50 0.39–0.64 10–12 years 0.73 0.54–1.00 >12 years 0.44 0.34–0.56 <0.001 >12 years 0.44 0.33–0.59 <0.001 Other starchy Other starchy Education Education ≤4 years (reference) ≤4 years (reference) 5–9 years 0.99 0.81–1.20 5–9 years 1.09 0.88–1.35 10–12 years 0.72 0.57-0.92 10-12 years 1.07 0.80-1.43 >12 years 0.68 0.53-0.87 <0.001 >12 years 1.15 0.83-1.60 0.355 Fish Fish Education Education ≤4 years (reference) ≤4 years (reference) 5-9 years 1.06 0.96-1.17 5-9 years 1.14 1.04-1.25 10-12 years 1.24 1.09-1.40 10-12 years 1.36 1.20-1.54 >12 years 1.40 1.23-1.60 <0.001 >12 years 1.50 1.31-1.72 <0.001 Meat Meat Education Education ≤4 years (reference) ≤4 years (reference) 5-9 years 1.01 0.89-1.14 5-9 years 1.09 0.96-1.23 10-12 years 1.02 0.86-1.21 10-12 years 1.27 1.06-1.52 >12 years 0.95 0.80-1.13 0.693 >12 years 1.16 0.96-1.41 0.014 Table 4 Odds ratios for beverage consumption according level of education, adjusted for age, BMI, smoking habits, physical activity and income Women Men OR IC(95%) P trend OR IC(95%) P trend Milk Milk Education Education ≤4 years (reference) ≤4 years (reference) 5-9 years 1.54 1.38-1.70 5-9 years 1.53 1.39-1.68 10-12 years 2.24 1.95-2.57 10-12 years 3.00 2.62-3.44 >12 years 2.60 2.24-3.01 <0.001 >12 years 3.07 2.62-3.59 <0.001 Wine Wine Education Education ≤4 years (reference) ≤4 years (reference) 5-9 years 0.79 0.68-0.93 5-9 years 0.73 0.63-0.84 10-12 years 0.45 0.36-0.55 10-12 years 0.41 0.35-0.49 >12 years 0.51 0.41-0.62 <0.001 >12 years 0.46 0.38-0.56 <0.001 Beer Beer Education Education ≤4 years (reference) ≤4 years (reference) 5-9 years 0.92 0.72-1.18 5-9 years 0.86 0.77-0.97 10-12 years 0.76 0.56-1.03 10-12 years 0.61 0.52-0.71 >12 years 0.82 0.61-1.10 0.097 >12 years 0.57 0.48-0.68 <0.001 Spirits Spirits Education Education ≤4 years (reference) ≤4 years (reference) 5-9 years 0.66 0.27-1.58 5-9 years 0.72 0.60-0.87 10-12 years 0.15 0.03-0.74 10-12 years 0.46 0.34-0.62 >12 years 0.13 0.03-0.53 0.002 >12 years 0.27 0.19-0.40 <0.001 Port Wine Port Wine Education Education ≤4 years (reference) ≤4 years (reference) 5-9 years 1.44 1.03-2.01 5-9 years 1.00 0.81-1.24 10-12 years 1.22 0.80-1.88 10-12 years 0.97 0.73-1.30 >12 years 1.34 0.90-2.00 0.265 >12 years 1.42 1.07-1.89 0.093 In men, similar odds ratios were observed for milk, vegetable soup, vegetables, fruit, fish, bread, wine, and spirits (Tables 3 and 4 ). However, in men but not in women, odds favouring meat consumption increased with increasing education (OR = 1.16 (0.96–1.41) for those having >12 years of education compared to those with ≤4 years), while for beer consumption, odds decreased with increasing education (OR = 0.57 (0.48–0.68) for those having >12 years of education compared to those with ≤4 years). No such significant trends were observed for these food groups and income with the exceptions of meat and starchy foods (other than bread) consumption, in men, which decreased with increasing income (p trend ≤ 0.022), and milk consumption which increased with increasing income (Tables 3 and 4 ). Discussion The main finding of the present study is that educational attainment was more frequently associated with food choices than income. There is general agreement among researchers [ 20 - 23 ] that education and income are conceptually distinct, and that they are likely to make separate and unique contributions to health-related outcomes [ 24 ]. In our study, the most educated consumed more frequently fruit, vegetables, milk and fish, and less wine and spirits, than their counterparts from less educated groups. Over the last years, several studies have attempted to identify the influence of socioeconomic factors on individual's dietary intake [ 25 - 28 ]. Our interest in educational and economic determinants of food choice in Portuguese adults relate to these particular characteristics in the population. Portugal, according European standards, is a small and relatively poor country, exhibiting the highest level of social inequalities in the European Union [ 29 ]. Nevertheless, Portugal had significantly and positively changed in the last four decades, in several domains such as the economy and culture, although the census of 1991 revealed that 15.3% of the Portuguese were illiterate. That of 2000 showed that, despite the improvements and changes in the education of adults, 7% can still not read or write [ 29 ]. This is a reality that classifies Portugal as the country with the higher percentage of individuals with low level of education in all the European Union [ 29 ]. From the employment perspective, Portugal's unemployment rates in the last 25 years never surpassed 10% of the active population, which is a better indicator than the observed levels in the majority of the European countries. However, the percentage of individuals with low-remuneration in Portugal is much higher than the EU average [ 29 ]. Several studies have concluded that a strong relationship exists between countries' per capita national incomes and nutrition [ 30 - 32 ]. The economic issue is of considerable significance, and it is sometimes suggested that this is probably the key variable of all in influencing food choice [ 30 ]. Household income is expected to influence food choices, especially for relatively high-priced food items such as fish, fresh fruit and vegetables [ 33 ]. Nevertheless, this not seems to be the case when we compared income and education levels as determinants of intake of significant food groups in Portuguese adults. Our data shows, in both genders, a significant positive trend in the consumption of vegetables, vegetable soup, fruits, milk and fish, with higher levels of education, which did not occurred in relation to income with the exception of milk. In our study, education was adjusted for income and vice versa. While the majority of investigators use two or more indicators of socio-economic position, several [ 34 - 37 ] do not simultaneously adjust for the unmeasured effects of each indicator on the other. Two types of bias may result from this practice: (1) using a single indicator such as education may bias the point estimate (food choice) because the education variable is allowed to account for some of the variation that is actually the product of unmeasured socio-economic influences; as a result, if we did not simultaneously adjusted education for income and vice versa, our claims about the influence of education level on food choice probably would have been overestimated; (2) the use of a single indicator may result in the overall or total socio-economic effect being underestimated. Data from the Portuguese Household Budget Surveys (using the DAta Food Networking – DAFNE – classification system), shows similar results to ours in relation to the positive association between education attainment and the availability of fruits, fish, milk and alcoholic beverages but some different data in regard to other foods (availability of vegetables and cereal products is fairly stable or tends to decline with education) [ 18 ]. Curiously, we found that meat consumption in men was positively associated with level of education, as in the DAFNE study [ 18 ], although our study showed a significant reduction of meat consumption with higher categories of income, in men. In Portugal [ 18 ] fish is more available among the trend-leading educated individuals which may be more advantageous to their cardiovascular health [ 38 ]. In our study, there was also a significant trend in the consumption of milk in men, being more frequently ingested with increasing income. As suggested by Axelson, [ 39 ] positive health relationships between dietary patterns and income may reflect a growing concern about health in the higher socio-economic groups. The association between milk consumption and socio-economic position is sometimes contradictory. Cristofar and Basiotis [ 40 ], for example, reported lower intake of milk among low-income women, while Roos et al. [ 41 ] found that higher educational and income groups from both genders consumed less milk. Consumption of alcoholic beverages, such as wine and spirits, in both genders, and beer in men, exhibited significant decreases in their frequency of intake, with increasing education levels. In Portugal, alcoholic beverages consumption is a major public health problem [ 42 ]. In DAFNE study, [ 18 ] using Portuguese Household Budget Surveys, alcoholic beverages availability was also higher in the lower educated households. Interestingly, in our study, the consumption of bread (in both genders) and other starchy foods (in women), decreased with increasing number of years of education; men seem to abandon starchy foods (other than bread) consumption under condition of higher incomes. It is possible that higher educated individuals tend to avoid foods that are considered as being more fattening or rich in energy, such as bread and other starchy foods [ 43 , 44 ]. Research has demonstrated that for a given body size, higher educated women are more dissatisfied with or concerned about their bodies and are more likely to have dieted in the past than lower educated women [ 45 , 46 ]. One of the most interesting findings in terms of economic constraints and food consumption relationship in our study, is the few significant associations between income and food choices, even though the well established links between economic and material resources, food availability and dietary quality [ 47 ]. By contrast to our results, Turrell et al. [ 23 ] showed household income to be the strongest and most robust independent predictor of food purchasing behaviour, and the effects of education to be substantially attenuated (to non-significance or marginal significance). In our study, the specificity of the relationship between education and food choice probably reflect each respondent's individual contribution to food choice, whereas household income was possibly capturing the combined contextual effects of numerous individuals, as well as many other within-household processes [ 23 ], and thus showed a weaker relationship with food choice. Our results may also reflect lesser difficulties faced by low-income groups when selecting the food groups that we studied. In several urban and rural areas of Portugal, there are many people who own plots of land that are too small to make a living, but allow them to work the land and produce foods (e.g., fruit, vegetables and poultry) for their own consumption. Although they produced a limited range of foods that is not accounted in official agricultural statistics, probably, if they stopped working the land they would experience greater difficulties in obtaining access to those particular foods. A potential limitation of our study and most nationwide population surveys is that the poor are usually not well presented. We know from previous research into survey participation that population-based samples typically under-represent the most socio-economically disadvantaged and over-represent the advantaged [ 48 , 49 ], because homeless and unemployed may be difficult to reach, and this may debilitates the interpretation of our results. In our study, it remains to be explained the different pattern of associations between income and important food groups (milk versus fruit and vegetables, for example) and the different pattern of associations between food choices and income in each gender (e.g., milk and starchy foods). Several reasons may explain specific differences in the findings of our study compared with those of previous mentioned studies, including differences in populations sampled (e.g., both genders versus women only, different cultural backgrounds ranges), differences in assessment of education or income, differences in dietary assessment (e.g., qualitative food data versus 24-h dietary recalls or food frequency questionnaires) and differences in analytic methods (e.g., covariates included in statistical models). Nonetheless, results from our study indicate that the associations between food choice were stronger in relation to educational attainment than income categories. Differences in food choices according the level of education reflect that more knowledge may influence the perceived relationship between diet and health as well as the perceived outcomes of following a healthy diet [ 33 ]. Despite differences in food consumption according education and income, in our study we could not assess if these differences were also evident on the energy and nutrient level, which was a limitation. British data point to micronutrient and antioxidant intakes as the most likely nutritional influences on health inequalities [ 50 ]. Nevertheless, according Galobardes et al., [ 10 ] it is also possible that despite differences in food consumption, nutrient intake is similar among socio-economic groups, as these may not be substantial enough to translate into differences in nutrient intake. If a country like Portugal wants to change the adult food choice behaviour, or in other words, wants to reach certain dietary goals, the support of applied research like ours is needed in order to plan the right strategies for promoting healthy diets. Confidence in a significant positive causal link between per capita national income and individual nutrition reinforces the importance of economic growth [ 51 ] but also implies that public policy should stress education as a mean for improving healthy food choices. Education might influence food choice by facilitating or constraining one's ability to understand the information communicated in nutrition education or in food labels [ 52 , 53 ]. Whereas income-related dietary differences suggest ameliorative responses through the potential of the economic system, differences based on education point to initiatives such as nutrition education programmes [ 22 , 54 ]. According to Geraldes,[ 55 ]in Portugal it may sometimes be more appropriate to correct inequalities in the domains of education or nutrition than that of health. Given the poor education level of the majority of Portuguese adults, a move towards an increased acquisition of general knowledge and personal development through compulsory and higher education, lifelong learning and improved qualifications of the population, is desirable to promote the development of a knowledge society and improve the level and quality of national education which, in turn, may relate to healthier food choices. It is well recognized that changes in dietary behaviour may be brought about, not by direct modification of food habits, but by alteration or manipulation of the education and culture [ 8 ]. Conclusions Regardless of the reasons explaining the complex and diversified patterns of economic and educational associations of food consumption found in Portugal, the findings of this study suggest that education and income have distinct associations with food choice. The low and high income groups are or tend to be similar in regard to the majority of food choices, and access to education appears to be the key element to a better food pattern as indicated by higher frequency of milk, vegetable soup, vegetables, fruit, and fish consumption. Competing interests The author(s) declare that they have no competing interests. Authors' contributions PM and PP designed the study. PM and PP did the statistical analysis, and PM wrote the paper. PM and PP reviewed the final version of the paper. Table 5 Odds ratios for food consumption according level of income, adjusted for age, BMI, smoking habits, physical activity and education Women Men OR IC(95%) P trend OR IC(95%) P trend Vegetable soup Vegetable soup Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 1.02 0.91-1.14 315-547 euros 0.95 0.85-1.05 548-815 euros 1.01 0.90-1.14 548-815 euros 0.95 0.85-1.06 >815 euros 0.96 0.86-1.07 0.281 >815 euros 0.92 0.83-1.02 0.195 Vegetables Vegetables Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 1.11 0.96-1.29 315-547 euros 1.22 1.06-1.39 548-815 euros 1.07 0.92-1.24 548-815 euros 1.16 1.01-1.33 >815 euros 1.05 0.92-1.21 0.768 >815 euros 1.06 0.93-1.20 0.796 Fruit Fruit Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 0.94 0.81-1.21 315-547 euros 0.99 0.84-1.15 548-815 euros 0.98 0.80-1.19 548-815 euros 0.95 0.81-1.12 >815 euros 1.01 0.84-1.23 0.983 >815 euros 1.03 0.89-1.20 0.769 Bread Bread Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 1.16 0.92-1.47 315-547 euros 1.05 0.80-1.39 548-815 euros 1.06 0.84-1.34 548-815 euros 1.05 0.79-1.40 >815 euros 1.00 0.81-1.25 0.502 >815 euros 0.85 0.66-1.10 0.139 Other starchy Other starchy Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 0.86 0.69-1.06 315-547 euros 0.91 0.71-1.17 548-815 euros 0.95 0.77-1.19 548-815 euros 0.84 0.65-1.07 >815 euros 0.88 0.72-1.08 0.466 >815 euros 0.72 0.57-0.90 0.002 Fish Fish Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 0.86 0.77-0.96 315-547 euros 0.99 0.89-1.10 548-815 euros 0.84 0.75-0.93 548-815 euros 0.95 0.85-1.06 >815 euros 0.92 0.83-1.03 0.409 >815 euros 0.98 0.88-1.08 0.457 Meat Meat Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 0.98 0.86-1.12 315-547 euros 0.95 0.82-1.09 548-815 euros 1.09 0.95-1.25 548-815 euros 0.92 0.80-1.06 >815 euros 1.04 0.92-1.19 0.299 >815 euros 0.87 0.76-0.99 0.026 Table 6 Odds ratios for beverage consumption according level of income, adjusted for age, BMI, smoking habits, physical activity and education Women Men OR IC(95%) P trend OR IC(95%) P trend Milk Milk Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 0.94 0.83-1.05 315-547 euros 1.01 0.90-1.12 548-815 euros 1.02 0.91-1.15 548-815 euros 1.05 0.94-1.18 >815 euros 0.99 0.89-1.11 0.575 >815 euros 1.12 1.01-1.24 0.017 Wine Wine Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 0.97 0.81-1.17 315-547 euros 1.07 0.93-1.22 548-815 euros 0.92 0.76-1.11 548-815 euros 1.00 0.87-1.14 >815 euros 0.86 0.72-1.03 0.065 >815 euros 0.89 0.79-1.01 0.757 Beer Beer Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 0.89 0.67-1.19 315-547 euros 0.95 0.83-1.09 548-815 euros 0.89 0.66-1.20 548-815 euros 1.05 0.92-1.21 >815 euros 1.00 0.76-1.31 0.671 >815 euros 1.06 0.93-1.21 0.139 Spirits Spirits Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 2.18 0.74-6.43 315-547 euros 1.14 0.92-1.41 548-815 euros 2.52 0.86-7.43 548-815 euros 1.12 0.89-1.40 >815 euros 1.28 0.44-3.78 0.968 >815 euros 1.09 0.88-1.34 0.229 Port Wine Port Wine Income Income ≤314 euros (reference) ≤314 euros (reference) 315-547 euros 1.15 0.78-1.70 315-547 euros 0.98 0.77-1.24 548-815 euros 1.28 0.87-1.89 548-815 euros 1.20 0.94-1.53 >815 euros 1.09 0.75-1.59 0.704 >815 euros 1.05 0.84-1.33 0.483 Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544354.xml
545937	Web GIS in practice II: interactive SVG maps of diagnoses of sexually transmitted diseases by Primary Care Trust in London, 1997 – 2003	Background The rates of Sexually transmitted diseases (STDs) in England have been rising steadily since the mid 1990s, making them a major public health concern. In 2003, 672,718 people were diagnosed with an STD in England, and around one third of those cases were diagnosed in London. Results Using GeoReveal v1.1 for Windows, we produced Web-based interactive choropleth maps of diagnoses of STDs by Primary Care Trust (PCT) in London for the years from 1997 to 2003 . These maps are in Scalable Vector Graphics (SVG) format and require a freely available Adobe SVG browser plug-in to be displayed. They are based on data obtained from the House of Commons Hansard Written Answers for 15 October 2004. They show steadily rising rates of STDs in London over the covered seven-year period. Also, one can clearly see on the maps that PCTs located in central London had the highest numbers of STD diagnoses throughout the mapped seven years. A companion bar chart allows users to instantly compare the STD figure of a given PCT for a given year against the average figure for all 25 mapped PCTs for the same year, and also compare those figures across all seven years. The maps offer users a rich set of useful features and functions, including the ability to change the classification method in use, the number of ranges in the map, and the colour theme, among others. Conclusions Wizard-driven tools like GeoReveal have made it very easy to transform complex raw data into valuable decision support information products (interactive Web maps) in very little time and without requiring much expertise. The resultant interactive maps have the potential of further supporting health planners and decision makers in their planning and management tasks by allowing them to graphically interrogate data, instantly spot trends, and make quick and effective visual comparisons of geographically differentiated phenomena between different geographical areas and over time. SVG makes an ideal format for such maps. SVG is a World Wide Web Consortium non-proprietary, XML-based vector graphics format, and is an extremely powerful alternative to Macromedia ® Flash and bitmap graphics.	Background Sexually transmitted diseases (STDs) have become a major public health concern in the UK during recent years. The rates of STDs in England have been rising steadily since the mid 1990s. In 2003, the number of STDs in England rose by 4% compared to 2002. Overall, 672,718 people were diagnosed with an STD in England in 2003, and around one third of those cases were diagnosed in the London area alone [ 1 , 2 ]. The House of Commons Hansard Written Answers for 15 October 2004 included an answer by Melanie Johnson MP, Minister for Public Health at the Department of Health, to a question by Sarah Teather MP on "how many cases of diagnosed STDs there were in each Primary Care Trust (PCT) in London in each year since 1997". The answer was provided in the form of a long table showing the figures for 25 PCTs in London (Table 1 – [ 2 ]). Table 1 Diagnoses of STDs by PCT in London, 1997 – 2003 PCT Name 1997 1998 1999 2000 2001 2002 2003 Barking and Dagenham 2,454 2,797 3,020 2,764 2,457 3,461 3,292 Barnet 1,542 1,591 1,546 1,993 2,001 1,899 923 Brent 8,753 8,851 8,724 9,258 9,479 10,156 10,113 Bromley 1,386 1,899 2,287 2,468 3,296 3,683 3,575 Camden 15,714 17,117 17,478 18,750 20,962 22,257 26,987 City and Hackney 14,998 16,595 16,601 16,169 19,171 22,673 21,072 Croydon 4,089 5,546 5,707 6,968 7,805 7,995 7,587 Ealing 1,514 585 1,518 1,687 2,062 2,751 1,832 Enfield 581 1,513 1,168 1,481 1,957 2,129 1,659 Greenwich 4,361 4,314 4,901 6,012 5,170 5,385 6,607 Hammersmith and Fulham 7,838 7,889 8,281 9,556 9,992 4,996 5,723 Haringey 5,029 5,229 5,568 6,344 7,183 6,617 6,379 Hillingdon 1,931 2,387 3,038 3,096 4,131 4,000 3,110 Hounslow 3,270 3,185 3,029 4,101 4,637 5,861 6,473 Islington 5,339 5,552 6,259 5,781 5,309 4,988 6,186 Kensington and Chelsea 11,995 11,636 11,040 13,149 13,301 11,739 12,243 Kingston 1,926 2,443 2,646 3,129 3,474 3,872 4,481 Lambeth 17,865 19,998 21,327 19,971 19,754 22,003 22,209 Lewisham 412 353 346 482 21 - - Newham 8,948 9,033 11,214 11,023 12,769 13,200 15,236 Southwark 16,481 17,073 15,232 13,836 16,699 19,618 19,249 Sutton and Merton 10,870 9,110 11,569 13,464 14,229 16,589 17,784 Walthamstow, Leyton and Leytonstone 1,157 1,355 1,726 1,713 2,311 2,433 2,988 Wandsworth 2,629 2,712 3,072 3,004 3,519 4,151 4,199 Westminster 18,639 21,490 21,359 20,503 20,870 18,462 19,657 Diagnoses of sexually transmitted diseases by Primary Care Trust in London, 1997 – 2003 (Source: [2]). Overall figures may be lower than stated for the London region in the annual report because the data presented here have not been imputed. One clinic in each of these PCTs (Barnet; City and Hackney) did not submit all the KC60 returns for 2003. The Alexis Clinic of Lewisham PCT closed in June 2001; no GUM clinics currently open in this PCT. Though the table presents all the requested data, it remains very difficult for the reader to fully appreciate the patterns and trends buried in them, or make quick and effective comparisons between the figures for different PCTs or between the seven data sets for the years from 1997 to 2003. Such data patterns, trends and comparisons derived from this Hansard table and other sources, e.g., demographic, deprivation/social exclusion, transport and existing GUM (Genito-Urinary Medicine) clinic data sets, are crucial for the decision maker wanting, for example, to: - improve access to GUM clinics and make decisions regarding the expansion or closure of existing clinics, or the creation of new ones; - channel resources and target STD prevention programmes to areas with the most need, or scale such programmes according to the magnitude of the problem in different areas (this is especially important in a climate of finite resources); and/or - monitor the impact of such programmes in a given area over time. In this paper, we describe a much better way of presenting the same Hansard table data in the form of interactive Web maps in Scalable Vector Graphics (SVG) format to further support health planners and decision makers in their planning and management tasks. Results Using GeoReveal, a tool from Graphical Data Capture Ltd ( – see 'Methods' section below), we produced Web-based interactive choropleth maps of diagnoses of STDs by PCT in London for the years from 1997 to 2003, which readers can browse at (Figure 1 ). These maps and companion bar chart ('Chart Panel') are based on the data in Table 1 , and show steadily rising rates of STDs in London over the covered seven-year period. Also, one can clearly see on the maps that PCTs located in central London, e.g., Camden PCT and Lambeth PCT, had the highest numbers of STD diagnoses throughout the seven years from 1997 to 2003. Figure 1 Screenshot from our interactive Web maps of STD diagnoses by London PCT, 1997 – 2003. Screenshot from our Web-based interactive choropleth maps of diagnoses of STDs by PCT in London for the years from 1997 to 2003 . The map shown in this screenshot is for the year 2002, with Camden PCT highlighted in yellow. The bar chart ('Chart Panel') on the left shows steadily rising STD rates in Camden PCT over the covered seven-year period. Camden's rates are well above the average for all 25 mapped PCTs over the same period (the purple portions of the bars represent the average for all PCTs). The maps require the free Adobe SVG Viewer . Visitors will be automatically prompted to download it on their first visit to the site, if they don't already have it installed on their machine. Scripting must also be enabled in Internet Explorer. As the mouse cursor is moved around the main map window, the 'Chart Panel' changes to display statistical data about the currently highlighted PCT. Each row of the bar chart represents data for one year and has two bars; a red bar that shows the value for the highlighted PCT, and a transparent blue bar which shows a mean value of this piece of data (or year) for all 25 mapped PCTs (when both bars overlap, a purple colour is produced – Figure 1 ). Additionally, clicking a PCT area on the map will display an information box with all available data for that PCT (Figure 2 ). Figure 2 Screenshots of two information boxes from our interactive Web map interface. Upper information box: clicking a PCT area on the map ('Camden PCT' for this screenshot) will display this pop-up box with all available data for that PCT. Lower information box: clicking the 'Info' button (shown in Figure 1 – bottom right 'Navigation Panel') will open this pop-up box with extra information about the maps, links to the Web sites of London PCTs, and detailed help about the map interface. An 'Options Panel' controls the data that are shown in the map window. Users can tick the 'Background Map' box to add a raster background map of London to the main map. They can use the 'Choose Classification Method' list to select how the data are mapped; with equal ranges, equal counts or by highlighting the highest values. A 'Number of Colours' list allows users to select how many classes or ranges are used in the choropleth map (two to five ranges). Using the 'Select Map Topic' list, users can select the topic that is shown on the main map (a year from 1997 to 2003). Finally, users can select from the 'Colour of Theme' list the colour theme that is used in the main map (orange, green, or blue). Map zooming, panning, MapTips (displaying PCT names), and a dynamic legend are available. An overview map shows a miniature version of the full extent map. When the user zooms in, a rectangle on the overview map highlights the area that is currently being displayed in the main map window. The user can click and drag this rectangle to change the view in the main map window. After zooming into the main map, users can use the 'Reset' button to return to the full extent of the main map. An 'Info' button is also available. Clicking this button will open a pop-up window with extra information about the maps, links to the Web sites of London PCTs, and detailed help about the map interface (Figure 2 ). Discussion From complex raw data to valuable decision support information Turning raw tabular data into much more useful and accessible visual information in the form of interactive Web maps is much needed to support and empower decision makers, and even members of the general public. Such maps help us understand the relationships, patterns and trends buried in the original data sets and also enable instant visual comparisons to be made between different geographical areas and over time (when data sets for successive periods of time are available) [ 3 ]. We believe this transformation of raw data into valuable decision support information is very evident in the London STD example described in this paper. Readers only have to compare the original Hansard table (Table 1 ) with the corresponding interactive Web maps we have produced to see the difference for themselves and appreciate the value of interactive maps. SVG: an ideal format for interactive Web maps SVG is a non-proprietary language for describing rich, stylable two-dimensional graphics and graphical applications in XML (eXtensible Markup Language). SVG is fully endorsed by the W3C (World Wide Web Consortium – ). It is rapidly becoming a popular choice for delivering interactive Web maps, being designed to work effectively across platforms, output resolutions, colour spaces, and a range of available bandwidths. It offers a rich modern graphics format providing the ability for better map display, and advanced graphical features such as transparency, arbitrary geometry, filter effects (shadows, lighting effects, etc.), scripting, and animation [ 4 ]. All these features have made SVG a direct competitor to the proprietary Macromedia ® Flash format [ 5 ]. Vector-based images (describing shapes and paths), such as those in SVG and SWF (Macromedia ® Shockwave/Flash File) formats, will keep their sharp character when enlarged, while raster-based images (storing information about each and every pixel in the image), such as those saved in GIF (Graphics Interchange Format) or JPEG (Joint Photographic Experts Group) formats, will show jagged edges. A free SVG Web browser plug-in is available from Adobe for different platforms (Adobe SVG Viewer – ), in the same way the free Adobe Reader software is available for rendering PDF (Portable Document Format) files. Besides the example described in this paper, other examples of SVG interactive Web maps in the health arena include the Office for National Statistics' England and Wales 2001 Census Key Statistics maps and 2001 Area Classification for Health Areas maps , and Leeds Interactive Health Atlas . Tools for producing interactive SVG and Flash maps from desktop GIS projects Besides GeoReveal , the tool we have used to produce the maps described in this paper (see 'Methods' section below), other SVG/Flash mapping tools available today for publishing maps created in desktop GIS (Geographic Information Systems) include GéoClip , SVGMapMaker , MapViewSVG , and SVGMapper . The latter two tools (MapViewSVG and SVGMapper) are specific to ESRI ArcView GIS. Table 2 provides an overview of the features of GeoReveal, GéoClip, and SVGMapMaker. Table 2 Overview of the features of GeoReveal, GéoClip, and SVGMapMaker GeoReveal GéoClip SVGMapMaker GeoReveal is a powerful .NET application that combines geography and statistics to produce revealing interactive graphics for the Web. It produces fully interactive geo-statistical presentations. There is no requirement for a GIS. GeoReveal uses MID/MIF files that can be exported from practically any GIS. Other information: - Unlimited data fields and includes a comprehensive set of pre-defined templates - Features a simple wizard driven interface and an advanced set of menus for users who need to fine tune the finished output - Generates SVG data, overlay boundaries, overview and background maps. For UK Local Government users, SVG is also an e-GIF (e-Goverment Interoperability Framework) compliant format - Presentations can include interactive charts that change to display statistics about the currently selected map region or area - Map display can be changed after generation, i.e., ability to change map topic, map colour, number of ranges, and classification method - Includes pre-prepared UK census boundary data and sample projects GéoClip is a MapBasic program that takes TAB file information and turns it into interactive Macromedia ® Flash presentations. There is also a version for ESRI ArcGIS. GéoClip does not currently have an SVG component. Other information: - Written in MapBasic and therefore dependent on having a MapInfo Pro licence - Does not offer a very intuitive interface - Limited to ten data fields and to only one page layout - Users cannot remove the GéoClip logo/branding from the map page - Can only produce Flash (proprietary format, not e-GIF compliant) and not SVG (open format) - Some of its unique features like adjustable ranges and multiple themes are planned for the next release of GeoReveal SVGMapMaker is a MapBasic program that takes TAB file information and uses it to replicate limited GIS functionality in a browser window. It cannot be considered as a geo-statistical tool, as the level of interactivity is limited. For example, there are no interactive charts as provided in GeoReveal. Other information: - Written in MapBasic and therefore dependent on having a MapInfo Pro licence - Limited interface – no ability to change map topic and no statistical delivery - Map display cannot be changed after generation, i.e., no ability to change map colour, number of ranges, and classification method A quick comparison of three tools available today for producing interactive SVG and Flash maps from desktop GIS projects: GeoReveal (from Graphical Data Capture Ltd, London, UK – ), GéoClip (from eMc3, a young company based in Toulouse, France – ), and SVGMapMaker (from TETRAD Computer Applications Inc, Bellingham, WA, USA – ). Conclusions Using GeoReveal v1.1 for Windows, we produced Web-based interactive choropleth maps of diagnoses of STDs by PCT in London for the years from 1997 to 2003 . These maps are in SVG format and require a freely available Adobe SVG browser plug-in to be displayed. They are based on data obtained from the House of Commons Hansard Written Answers for 15 October 2004. They show steadily rising rates of STDs in London over the covered seven-year period. Also, one can clearly see on the maps that PCTs located in central London had the highest numbers of STD diagnoses throughout the mapped seven years. A companion bar chart allows users to instantly compare the STD figure of a given PCT for a given year against the average figure for all 25 mapped PCTs for the same year, and also compare those figures across all seven years. The maps offer users a rich set of useful features and functions, including the ability to change the classification method in use, the number of ranges in the map, and the colour theme, among others. We also presented a quick review of some of the tools available today for creating interactive vector graphics maps from desktop GIS projects. Wizard-driven tools like GeoReveal have made it very easy to transform complex raw data into valuable decision support information products (interactive Web maps) in very little time and without requiring much expertise. The resultant interactive maps have the potential of further supporting health planners and decision makers in their planning and management tasks by allowing them to graphically interrogate data, instantly spot trends, and make quick and effective visual comparisons of geographically differentiated phenomena between different geographical areas and over time. SVG makes an ideal format for such maps. SVG is a W3C non-proprietary, XML-based vector graphics format, and is an extremely powerful alternative to Macromedia ® Flash and bitmap graphics. Methods We used GeoReveal v1.1 for Windows to create the interactive SVG maps described in this paper. GeoReveal runs under Windows 98/NT/2000/XP and requires Microsoft ® .NET framework v1.1 to be installed on the production machine. We started by extracting London PCT boundaries from a larger data set of all England (2001 Census PCT – post April 2002 change), which is the copyright of the Crown/Ordnance Survey , and is freely available in both ArcView and MapInfo formats to the UK academic community from EDINA UKBORDERS service with the support of the ESRC and JISC . We also prepared a spreadsheet containing data about the number of STDs recorded in each London PCT between 1997 and 2003 using data from [ 2 ]. The two files were merged using MapInfo Professional v7.5 , creating a MapInfo .TAB file. This file was exported to MID/MIF format, the format that files need to be in, in order to be used by GeoReveal. We created our presentation (the interactive SVG Web maps) using the GeoReveal Wizard; an eight-step process that allows users to create a fully interactive SVG page. Introduction: setting the output directory The first Wizard dialog that must be completed is the 'Introduction' dialog, which is used to specify the directory to which the final GeoReveal output files will be saved (Figure 3 ). Figure 3 Screenshot of the 'Introduction' dialog. Screenshot of the 'Introduction' dialog in GeoReveal Wizard. For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Introduction: setting the output directory'. To select the output directory, we clicked the 'Browse' button and in the resulting file browse dialog, we selected the required directory and clicked 'OK'. To move to the next step, we clicked 'Next'. Wizard step 1: choosing the template The first Wizard step is the 'Choose Template' dialog (Figure 4 ). This dialog is used to specify general settings for a GeoReveal presentation. Figure 4 Screenshot of the 'Choose Template' dialog. Screenshot of the 'Choose Template' dialog in GeoReveal Wizard (Wizard Step 1 of 8). For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Wizard step 1: choosing the template'. We selected the template to be used. Ten templates are provided; four that include a bar chart, four that include a pie chart and two that have no chart. With our London PCT STD data, the most appropriate is a bar chart template. We selected a logo image to be added to the top-left corner of our GeoReveal page. We clicked the 'Browse' button and selected the required file – in this instance, a University of Bath logo has been used. We then specified the title of the GeoReveal page. This title is placed next to the logo at the top of the output page (see the logo and title at . Finally, we selected the title colour and page background colour. To move to the next step, we clicked 'Next'. Wizard step 2: selecting the main map data The second step is the 'Select Main Map Data' dialog (Figure 5 ). This dialog is used to specify settings for the main GeoReveal map. Figure 5 Screenshot of the 'Select Main Map Data' dialog. Screenshot of the 'Select Main Map Data' dialog in GeoReveal Wizard (Wizard Step 2 of 8). For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Wizard step 2: selecting the main map data'. We selected the MID/MIF file that contains the statistics that will form the main map in our GeoReveal page by clicking the 'Browse' button and choosing the required file. We selected the field from this MID/MIF file that will be used for ToolTips. A ToolTip (or MapTip) is the piece of text that is displayed when the mouse cursor is hovered over a region in the map. In this instance, the PCT Name field has been selected. Users can also choose to turn the information box on or off. When it is turned on, a user can click a region in the map to display a dialog containing all information held within the MID/MIF file about that region (Figure 2 ). Finally, we selected the background colour and highlight colour for the map. The highlight colour is the colour a map region (an individual PCT area in our case) will be displayed in when the mouse cursor hovers over it. To move to the next step, we clicked 'Next'. Wizard step 3: selecting the overview map data The third step is the 'Select Overview Map Data' dialog (Figure 6 ). This dialog is used to specify settings for the GeoReveal overview map. Figure 6 Screenshot of the 'Select Overview Map Data' dialog. Screenshot of the 'Select Overview Map Data' dialog in GeoReveal Wizard (Wizard Step 3 of 8). For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Wizard step 3: selecting the overview map data'. We selected the MID/MIF file to be used for the overview map by clicking the 'Browse' button and selecting the required file. It is possible to use the same MID/MIF file for both the main and overview maps. We selected the overview map colour and the overview rectangle colour (this rectangle shows a user where they are currently zoomed in on the main map). Additionally, users can select the background colour and border colour for the overview map panel. To move to the next step, we clicked 'Next'. Wizard step 4: selecting options controls The fourth step is the 'Select Options Controls' dialog (Figure 7 ). This dialog is used to specify settings for the 'Options Panel' in the presentation. Figure 7 Screenshot of the 'Select Options Controls' dialog. Screenshot of the 'Select Options Controls' dialog in GeoReveal Wizard (Wizard Step 4 of 8). For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Wizard step 4: selecting options controls'. First, we selected the background map image that is to be used. We ticked the top 'Checkbox Visible' box to activate the background map option and then clicked the top 'Browse' button to select the file that will be used. This file must be a JPEG or GIF image and information is needed about its width and height in real terms, and also the bounding coordinates. Once the required file is selected, the 'Background Image Settings' dialog is displayed (Figure 8 ). In this dialog, we entered the bounding coordinates of our selected image, the height and width of the image in real terms, and specified how opaque the background map will be in the presentation. We then clicked 'OK' to confirm and return to the previous dialog (Figure 7 ). Figure 8 Screenshot of the 'Background Image Settings' dialog. Screenshot of the 'Background Image Settings' dialog in GeoReveal Wizard. For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Wizard step 4: selecting options controls'. In the 'Select Options Controls' dialog, it is also possible to: - enable a vector map option; and - select the text colour, border colour, background colour and control colour (the colour in which the fields are rendered) for the 'Options Panel'. To move to the next step, we clicked 'Next'. Wizard step 5: setting the bar chart The fifth step is the 'Set Bar Chart' dialog (Figure 9 ). This dialog is used to specify settings for the bar chart in the presentation. This dialog will differ depending on the template that was selected. In this dialog, one can: Figure 9 Screenshot of the 'Set Bar Chart' dialog. Screenshot of the 'Set Bar Chart' dialog in GeoReveal Wizard (Wizard Step 5 of 8). For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Wizard step 5: setting the bar chart'. - specify the bar colour, grid colour (if the grid is enabled), text colour, background colour and border colour; - add or remove a % sign to the figures that are shown at the end of each bar. As the data being used in this presentation are absolute, the 'Show %' box should be unchecked; and - specify the opacity of the bars in the bar chart. After specifying our settings for the bar chart, we clicked 'Next' to move to the next step. Wizard step 6: setting the legend The sixth step is the 'Set Legend' dialog (Figure 10 ). This dialog is used to specify settings for the legend panel in the presentation. Users can: Figure 10 Screenshot of the 'Set Legend' dialog. Screenshot of the 'Set Legend' dialog in GeoReveal Wizard (Wizard Step 6 of 8). For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Wizard step 6: setting the legend'. - specify a title and subtitle for the legend. Rather than specifying a main title, we checked the 'Use Dynamic Legend' box. When this box is checked, the title will be determined by the topic that the GeoReveal map is based on (a year from 1997 to 2003 in our case); - enable or disable the 'Highlight Range' option. If this option is enabled (as is the case in our presentation), when the mouse is hovered over a map region (PCT), the legend range that this region falls into will be highlighted; and - specify the text colour, background colour and border colour for the legend panel. After specifying our settings for the legend, we clicked 'Next' to move to the next step. Wizard step 7: setting the 'Information Panel' The seventh step is the 'Set Information Panel' dialog (Figure 11 ). This dialog is used to specify settings for the 'Information Panel' in the presentation. In the presentation, this panel displays information about the region in the map (PCT) that the mouse cursor is currently over. Users can: Figure 11 Screenshot of the 'Set Information Panel' dialog. Screenshot of the 'Set Information Panel' dialog in GeoReveal Wizard (Wizard Step 7 of 8). For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Wizard step 7: setting the 'Information Panel". - enable or disable each of the information panels and select the fields from the MID/MIF file that will be used to populate each panel. A title can also be entered for each panel (in our instance, Panel 1 was enabled, has been given the title 'PCT Name' and will display the PCT Name field from the MID/MIF file); and - specify the text colour, background colour and border colour for the 'Information Panel'. We then clicked 'Next' to move to the next step. Wizard step 8: setting the 'Navigation Panel' The eighth step is the 'Set Navigation Panel' dialog (Figure 12 ). This dialog is used to specify settings for the 'Navigation Panel' in the presentation: the text colour, background colour, border colour and button colour. Figure 12 Screenshot of the 'Set Navigation Panel' dialog. Screenshot of the 'Set Navigation Panel' dialog in GeoReveal Wizard (Wizard Step 8 of 8). For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Wizard step 8: setting the 'Navigation Panel". We then clicked 'Next' to move to the final step. Generating the presentation Lastly, we saved the settings for our presentation and generated an initial presentation page (Figure 13 ). The 'Wizard Final Step' dialog enables users to: Figure 13 Screenshots of the 'Wizard Final Step' dialog and final 'Build Successful' message box. Screenshots of the 'Wizard Final Step' dialog and final 'Build Successful' message box in GeoReveal Wizard. For a detailed description of the functions available in this dialog, please refer to the 'Methods' section > 'Generating the presentation'. - save the settings they have just made by clicking 'Save Settings', then in the resulting dialog, browsing to the required directory and clicking 'Save'. The saved settings file enables users to restore their settings and edit the presentation at a later date; and - generate the presentation by clicking 'Finish'. A message is displayed to confirm that the presentation has been successfully generated (Figure 13 ). Clicking 'Yes' on the message box will open the page in Internet Explorer. When the presentation is created, the 'Advanced View' window is also opened. This can be used to add the finishing touches to the presentation. It contains all the options that the wizard does, along with a few advanced options. 'Advanced View': 'Legend' tab The 'Legend' tab can be used to select the legend colour schemes that will be available (Figure 14 ). In this tab, users can add and remove colour schemes using the 'Add' and 'Remove' buttons, and ensure that the legend to be first loaded is at the top. Legend schemes are provided with the GeoReveal installation, and additional ones can be created using this tab. Moreover, the number of decimal places used in the legend can be changed. For this presentation, it has been set to 0. Figure 14 Screenshot of the 'Legend' tab in the 'Advanced View' window. Screenshot of the 'Legend' tab in the 'Advanced View' window in GeoReveal Wizard. For a detailed description of the functions available in this tab, please refer to the 'Methods' section > "Advanced View': 'Legend' tab'. 'Advanced View': 'Bar Chart' tab The 'Bar Chart' tab can be used to edit the bar chart (Figure 15 ). It is possible to add average value bars to the bar chart. When enabled, a second bar is added to each row of the chart which shows the mean value of all the data in that field, as opposed to the first set of bars showing values for the currently selected map region (PCT) alone. To enable this: Figure 15 Screenshot of the 'Bar Chart' tab in the 'Advanced View' window. Screenshot of the 'Bar Chart' tab in the 'Advanced View' window in GeoReveal Wizard. For a detailed description of the functions available in this tab, please refer to the 'Methods' section > "Advanced View': 'Bar Chart' tab'. - we ticked the 'Show Average Value Bars' box to add average value bars to the bar chart; and - set the average bar colour and specified the opacity of the bars. 'Advanced View': 'Navigation Toolbar' tab Finally, we used the 'Navigation Toolbar' tab to edit the 'Navigation Panel' (Figure 16 ). In this tab, it is possible to add a button to the 'Navigation Panel', which when clicked, will open a simple HTML (HyperText Markup Language) page that contains information on how the GeoReveal presentation can be used and additional information about what it shows. To do this: Figure 16 Screenshot of the 'Navigation Toolbar' tab in the 'Advanced View' window. Screenshot of the 'Navigation Toolbar' tab in the 'Advanced View' window in GeoReveal Wizard. For a detailed description of the functions available in this tab, please refer to the 'Methods' section > "Advanced View': 'Navigation Toolbar' tab'. - we enabled the help button by ticking the 'Display Help Button' box; - entered the text that will be shown on the button. In this instance, it will show 'Info'; and - selected the help file that will be used by clicking the 'Browse' button and selecting the required file. In this instance, an HTML document was created that contains hyperlinks to the PCT Web sites, and information about using the presentation (see lower information box in Figure 2 ). Regenerating the presentation We then resaved our settings by going to 'File' > 'Save Settings'. In the resulting dialog, we browsed to the required directory and clicked 'Save'. The presentation was regenerated by clicking 'File' > 'Generate SVG Page'. Finally, we uploaded all the generated final page files to our Web server ( – Figure 1 ). Competing interests CR and MS work for GDC (Graphical Data Capture Ltd), the company that produces GeoReveal. Authors' contributions All three authors contributed equally to this paper.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545937.xml
193604	Drosophila Free-Running Rhythms Require Intercellular Communication	Robust self-sustained oscillations are a ubiquitous characteristic of circadian rhythms. These include Drosophila locomotor activity rhythms, which persist for weeks in constant darkness (DD). Yet the molecular oscillations that underlie circadian rhythms damp rapidly in many Drosophila tissues. Although much progress has been made in understanding the biochemical and cellular basis of circadian rhythms, the mechanisms that underlie the differences between damped and self-sustaining oscillations remain largely unknown. A small cluster of neurons in adult Drosophila brain, the ventral lateral neurons (LN v s), is essential for self-sustained behavioral rhythms and has been proposed to be the primary pacemaker for locomotor activity rhythms. With an LN v -specific driver, we restricted functional clocks to these neurons and showed that they are not sufficient to drive circadian locomotor activity rhythms. Also contrary to expectation, we found that all brain clock neurons manifest robust circadian oscillations of timeless and cryptochrome RNA for many days in DD. This persistent molecular rhythm requires pigment-dispersing factor (PDF), an LN v -specific neuropeptide, because the molecular oscillations are gradually lost when Pdf 01 mutant flies are exposed to free-running conditions. This observation precisely parallels the previously reported effect on behavioral rhythms of the Pdf 01 mutant. PDF is likely to affect some clock neurons directly, since the peptide appears to bind to the surface of many clock neurons, including the LN v s themselves. We showed that the brain circadian clock in Drosophila is clearly distinguishable from the eyes and other rapidly damping peripheral tissues, as it sustains robust molecular oscillations in DD. At the same time, different clock neurons are likely to work cooperatively within the brain, because the LN v s alone are insufficient to support the circadian program. Based on the damping results with Pdf 01 mutant flies, we propose that LN v s, and specifically the PDF neuropeptide that it synthesizes, are important in coordinating a circadian cellular network within the brain. The cooperative function of this network appears to be necessary for maintaining robust molecular oscillations in DD and is the basis of sustained circadian locomotor activity rhythms.	Introduction Circadian rhythms of diverse organisms are based on similar intracellular molecular feedback loops ( Dunlap 1999 ; Allada et al. 2001 ; Panda et al. 2002 ). Based on this view, it is believed that one or a small number of clock cells are sufficient for self-sustained rhythms ( Dunlap 1999 ). This is despite the complex cellular organizations of many tissues, organisms, and systems ( Kaneko and Hall 2000 ; Schibler and Sassone-Corsi 2002 ). In Drosophila , circadian clocks have been identified in a diverse range of cell types throughout the head and the body ( Glossop and Hardin 2002 ; Hall 2003 ). However, the clocks in different cells are considered nonidentical ( Krishnan et al. 2001 ; Glossop and Hardin 2002 ; Levine et al. 2002a ; Schibler and Sassone-Corsi 2002 ). In many tissues, molecular oscillations undergo rapid damping without environmental timing cues ( Hardin 1994 ; Plautz et al. 1997 ; Stanewsky et al. 1997 ; Giebultowicz et al. 2000 ). This is similar to the damping of in vitro rhythms in some mammalian tissues ( Balsalobre et al. 1998 ; Schibler and Sassone-Corsi 2002 ). In contrast, the Drosophila “core pacemaker” is believed to maintain robust oscillations for a long time in constant darkness (DD) with little or no damping, such that circadian behaviors can persist under such conditions ( Dowse et al. 1987 ). Indeed, self-sustaining oscillations are a defining characteristic of true circadian rhythms and are believed to be required of a fully functional rhythmic cell. The differences between the “core pacemaker” and the clock machinery within damping cells or systems are unknown. The six clusters of approximately 100 clock neurons in the adult Drosophila brain are well characterized ( Kaneko and Hall 2000 ). Recent studies have focused principally on one of these groups, the small ventral lateral neurons (s-LN v s), as the best “core pacemaker” candidate for the following reasons: (1) in the developmental mutant disco , the presence of LN v s correlates with the maintenance of behavior rhythmicity ( Helfrich-Förster 1997 ); (2) LN v s specifically express the neuropeptide pigment-dispersing factor (PDF), and the Pdf 01 -null mutant loses behavioral rhythmicity under DD conditions ( Renn et al. 1999 ); (3) genetic ablation of the LN v s by expressing proapoptotic genes causes the loss of rhythmicity in DD ( Renn et al. 1999 ); and (4) the s-LN v s maintain robust molecular oscillations for at least for 2 days in DD ( Yang and Sehgal 2001 ; Shafer et al. 2002 ), in contrast to at least some other brain neurons and nonneuronal tissues. This final property suggests that these cells might fulfill the self-sustaining criterion for the “core pacemaker.” Indeed, the s-LN v s have been proposed to the primary pacemaker cells that generate locomotor activity rhythms ( Helfrich-Förster 1997 ; Renn et al. 1999 ; Emery et al. 2000 ). Consistent with this cell-autonomous view of circadian rhythmicity, it has been shown that the LN v s possess all components of a fully functional, independent circadian clock: the photoreceptor cryptochrome, the rhythm-generating feedback loops, and a putative output factor, the neuropeptide PDF ( Emery et al. 2000 ). Our pursuit of the self-sustaining “core pacemaker” of the Drosophila circadian system began with a test of the s-LN v cell-autonomous clock hypothesis. Results LN v s Cannot Support Circadian Behavior Independently To test whether the LN v s can support free-running circadian locomotor activity rhythms independently of other functional clock cells, we restricted pacemaker activity to these few PDF-expressing cells. CYCLE (CYC) is a bHLH–PAS protein ( Rutila et al. 1998 ) and forms a heterodimeric transcription factor with CLOCK (CLK), another bHLH–PAS protein ( Allada et al. 1998 ). CYC is an essential component of the Drosophila circadian oscillator transcriptional feedback loop ( Glossop et al. 1999 ). The cyc 01 nonsense mutation completely eliminates molecular oscillations, and the direct target genes period (per) and timeless (tim) mRNAs are essentially undetectable ( Rutila et al. 1998 ). Behavioral rhythms are also absent in the cyc 01 homozygous mutant strain ( Rutila et al. 1998 ). We rescued cyc 01 specifically in the LN v s, by using a well-characterized pdf–GAL4 driver ( Renn et al. 1999 ) in combination with a UAS–CYC transgene to express ectopically wild-type CYC. Since CYC is apparently not a rate-limiting component of active dCLK–CYC complexes ( Bae et al. 2000 ) and does not undergo molecular oscillations itself ( Rutila et al. 1998 ), we expected that CYC overexpression would not cause circadian oscillator dysfunction. Indeed, the presence of the two transgenes did not affect locomotor activity rhythms in a wild-type background ( Figure 1 C, right panel). Figure 1 Rescuing Molecular Oscillations within the LN v s Is Not Sufficient to Rescue Locomotor Activity Rhythms The rescued mutant genotype is y w ; pdf–GAL4;UAS–CYC , cyc 01 / cyc 01 . The flies were entrained in standard LD conditions and timepoints taken. Molecular oscillations were examined by whole-mount in situ hybridization of the tim gene. Double staining with a Pdf probe was used to label the LN v s neuronal group. (A and B) These show representative duplicate experiments. No tim mRNA signal is detectable in the dorsal region of the brain. The lower arrows point to the s-LN v s and the upper arrows to the l-LN v s. (A) Brain taken at timepoint ZT3. Panels shown from left to right are Pdf (green, FITC labeled), tim (red, Cy3 labeled), and an image overlay. (B) Brain taken at timepoint ZT15. Panels shown from left to right are Pdf (green, FITC labeled), tim (red, Cy3 labeled), and an image overlay. (C) The double-plotted actograms of rescue mutant and control flies in a standard LD:DD behavior assay. The colors on the background indicate the lighting conditions of the behavior monitors (white, lights on; light blue, lights off). In the actogram, the average locomotor activity of the group of flies is plotted as a function of time. The left panel shows the actogram of the rescued mutant flies ( y w;pdf–GAL4/+;UAS–CYC,cyc 01 / cyc 01 , n = 30). RI (rhythm index; Levine et al. 2002a ) = 0.14. The right panel shows the actogram for the rescued wild-type (control) flies ( y w;pdf–GAL4/+;UAS–CYC/+ , n = 32, RI = 0.61). The rescued mutant flies ( pdf–GAL4;UAS–CYC,cyc 01 /cyc 01 ) were examined by two independent criteria. First, molecular oscillations were assayed by in situ hybridization with a tim probe ( Figure 1 A and 1B). tim RNA levels undergo robust cycling in wild-type flies, with a trough at ZT3 and a peak at ZT15 ( Sehgal et al. 1994 ). This is also true within all individual clock neurons ( Zhao et al. 2003 ). tim mRNA cycled in the LN v s ( Figure 1 A and B), indicating successful rescue of the molecular oscillator within these cells. The fact that other clock neurons were still tim mRNA-negative ( Figure 1 A and B) suggests that CYC and the rest of the molecular machinery can function cell autonomously, at least in the LN v s under these light–dark (LD) conditions. The observed oscillations are also not passively driven by light, since they persisted in DD, at least in the s-LN v s ( Figure S1 ). Second, locomotor activity rhythms were examined by standard behavioral criteria. The transgenic flies were completely arrhythmic in DD. They were also arrhythmic under LD conditions, as the flies failed to anticipate the discontinuous transitions from light to dark or from dark to light (see Figure 1 C, left panel; Rutila et al. 1998 ). In summary, the behavioral phenotypes were indistinguishable from those of the parental cyc 01 mutant strain. Brain Clock Neurons Manifest Robust Molecular Oscillations in DD The insufficiency of LN v molecular rhythmicity indicates that one or more additional groups of rhythmic clock neurons are required for behavioral rhythmicity. We considered that robust molecular cycling under extended constant darkness conditions might be a good criterion for identifying these cell groups, because prior biochemical studies showed that some head and brain locations undergo damping of molecular oscillations under free-running conditions ( Hardin 1994 ; Stanewsky et al. 1997 ). This conclusion has been extended by more recent immunohistochemical observations ( Yang and Sehgal 2001 ; Shafer et al. 2002 ). The criterion of maintaining persistent and robust molecular rhythms in DD therefore suggests that only a limited set of brain locations are likely to be free-running pacemaker candidates. In order to identify these neurons, we assayed fly brains by tim in situ hybridization after 8 days in DD. To our surprise, we found that all tim- expressing brain cell groups (including both large ventral lateral neurons [l-LN v s] and small ventral lateral neurons [s-LN v s], doral lateral neurons [LN d s], and all three groups of dorsal neurons [DNs]) still cycle robustly at this time ( Figure 2 ). Previous studies have reported that the l-LN v s fail to maintain oscillations at the beginning of DD ( Yang and Sehgal 2001 ; Shafer et al. 2002 ). We have reproduced these observations, but noticed that the l-LN v s “adapt” to constant conditions by becoming rhythmic once again after about 2 days in DD (data not shown). These results clearly distinguish the brain from the eyes and other peripheral tissues, which rapidly lose coherent molecular oscillations under free-running conditions ( Hardin 1994 ; Plautz et al. 1997 ; Stanewsky et al. 1997 ; Giebultowicz et al. 2000 ). Although this approach failed to identify the additional neuronal groups necessary for behavioral rhythms, it suggests that many of these brain neuronal groups might act together in a network to support robust rhythms. Figure 2 All Brain Clock Neuronal Groups Maintain Robust Oscillations of tim RNA Levels in DD Wild-type flies were entrained for at least 3 days and then released into DD. tim RNA was assayed at trough (left panels) and peak (right panels) timepoints by whole-mount in situ hybridization. Wild-type flies in LD (A) were compared with the eighth day of DD (B). On the eighth day of DD, the locomotor activities of the fly population were still in close synchrony, without any obvious phase spreading (data not shown). Left panels, brains at ZT3 (A) or CT3 (B); right panels, brains from ZT15 (A) or CT15 (B). Both (A) and (B) are representative of three replicate experiments. Sustained Molecular Oscillation in Constant Darkness Requires PDF This association between robust molecular oscillations in all brain clock cells and behavioral rhythms in DD also made us consider the role of the neuropeptide PDF. The Pdf 01 mutant strain is unique among identified Drosophila circadian mutants, as it has little effect under LD conditions, but loses behavioral rhythmicity gradually and specifically in DD ( Renn et al. 1999 ). This phenotype might reflect a disassociation between behavioral rhythmicity and the underlying molecular oscillations, as predicted from the role of PDF as a circadian output signal; it is proposed to connect the molecular oscillation in the LN v s to locomotor activity ( Renn et al. 1999 ). We considered a completely different interpretation, namely, that PDF contributes to the functional integration of several brain clock neuronal groups, which is necessary to sustain molecular as well as behavioral rhythmicity under constant conditions. This fits well with previous studies of PDF in other organisms ( Rao and Riehm 1993 ; Petri and Stengl 1997 ). In contrast to the canonical output model, this possibility suggests that the Pdf 01 mutant might manifest unusual molecular oscillations within clock neurons, especially under DD conditions. To address this issue experimentally, we examined Pdf 01 mutant flies by tim in situ hybridization. In Pdf 01 flies, all clock neurons had robust tim RNA oscillations in LD, and the cycling phase and amplitude were comparable to those of wide-type flies ( Figure 3 A). The mutant flies were then released into DD and assayed at various times thereafter. In the first day of DD, cycling was similar to that observed in LD ( Figure 3 B). By the fourth day of DD, however, the cycling amplitude was much reduced in all clock neurons ( Figure 3 C and 3D). This was most evident from the unusually high signal in the CT2 sample; in wild-type flies, no tim signal was detected in any clock neuron at this timepoint ( Figure 3 C, left panels). There was also a reduced signal strength at the peak time, CT14 ( Figure 3C , fourth panel from the left). The result parallels the damping of behavioral rhythms in the Pdf 01 mutant strain ( Renn et al. 1999 ). Figure 3 Molecular Oscillations of tim RNA Damp in DD in the Pdf 01 Mutant tim RNA oscillations were examined in the Pdf 01 mutant under both LD (A) and different days in DD ([B] and [C]), by whole-mount in situ hybridization. (A), (B), and (C) are representative images from replicas of three experiments. (A) The left panel is from ZT3, and the right panel is from ZT15. A normal tim oscillation profile is observed compared to that of wild-type (see Figure 2 A). (B) Brains from the Pdf 01 mutant in the first day of DD. Left panel, CT3; right panel, CT15. Oscillations are comparable to those in LD. (C) Brains taken in the fourth day of DD. Six timepoints were taken throughout the circadian day. The sequence of panels from left to right is CT2, 6, 10, 14, 18, and 20, respectively. Wild-type brains (top row) were assayed in parallel with those from the Pdf 01 mutant (bottom row). See text for details. (D) Quantification of (C). Relative intensities are taken from normalized mean pixel intensities. Different clock neuronal groups were quantified independently and compared between wild-type (blue curves) and Pdf 01 mutant (purple curves). The panels from left to right are quantification of tim RNA oscillation in the DNs, in the LN d s, and in the LN v s. Reduced cycling amplitude and a significant advanced phase were observed in the fourth day of DD. See text for details. Despite the gradual fading of locomotor activity rhythms in DD, a significant fraction of Pdf 01 mutant flies is still weakly rhythmic after 4 d of DD ( Renn et al. 1999 ). By tracking their locomotor activity phases, we observed that most of them had accumulated an approximately 4-hour phase advance relative to wild-type flies by the fourth day in DD. This is consistent with the measured ca. 23-hour periods of these weakly rhythmic flies (1-hour phase advanced per day for 4 days) as well as their advanced evening activity peak in LD ( Renn et al. 1999 ). Quantitation of the tim in situ hybridization signal showed that there was a comparable one-point (4 h) advance in the peak of tim RNA and also confirmed the reduced cycling amplitude ( Figure 3 D). In order to eliminate the possibility that the observed damping is caused by the asynchrony of the Pdf 01 fly population, locomotor activities were tracked in real time. Individual flies were then removed from the monitors to assay tim RNA levels. Identical damped molecular oscillations were also observed in this case (data not shown). Taken together, the results indicate an excellent quantitative correspondence in phase and amplitude between the tim RNA rhythms and the behavioral rhythms in all clock neurons of the Pdf 01 strain. To extend these observations, we also assayed cryptochrome (cry) mRNA oscillations by in situ hybridization. cry is expressed in a similar clock neuron pattern to tim , but it has a peak expression at ZT2 and a trough at ZT14 ( Emery et al. 1998 ; Zhao et al. 2003 ). This phase is opposite to that of tim and other CLK–CYC direct target genes and reflects the fact that cry is only indirectly regulated by this heterodimeric transcription factor; CLK–CYC directly regulates the transcription factors PDP1 and VRILLE, which then regulate cry ( Cyran et al. 2003 ; Glossop et al. 2003 ). Despite these differences between tim and cry , a similar result was obtained for cry in the Pdf 01 strain in the fourth day of DD ( Figure 4 ), i.e., a reduced cycling amplitude compared to the fourth day of DD in a wild-type strain. This is suggested by the in situ pictures and is strongly indicated by the quantitation ( Figure 4 ). The correspondence between the tim and cry mRNA patterns indicates that the entire circadian transcriptional program damps in the mutant strain in DD, which underlies the behavioral damping. Figure 4 cry RNA Oscillation Amplitude Is Also Reduced by the Fourth Day of DD in the Pdf 01 Mutant cry RNA expression in the brain was examined at the fourth day of DD by whole-mount in situ hybridization using a cry probe. Timepoints were taken every 4 hours throughout the circadian day. The sequence of panels from left to right is CT2, 6, 10, 14, 18, and 20, respectively. Wild-type brains (top row) were analyzed in parallel with those from the Pdf 01 mutant (bottom row). Shown are representative images from duplicate experiments. Quantification of cry RNA oscillations in different cell groups is as shown in Figure 3 . Ubiquitous damping of the cycling amplitude in the different cell groups was observed in the Pdf 01 mutant. PDF Is Likely to Act upon Clock Neurons Directly It is noteworthy that the mRNA oscillations damp uniformly in the Pdf 01 mutant strain, including the PDF-expressing LN v s (see Figures 3 and 4 ). Since PDF is a neuropeptide ( Rao and Riehm 1993 ), it is unlikely to exert a direct intracellular effect on the LN v transcriptional machinery. A more conservative interpretation is that PDF maintains intercellular communication between individual LN v neurons ( Petri and Stengl 1997 ) and/or between the LN v s and other cells; the communication is essential for self-sustained molecular rhythms within the LN v s. Although this “feedback” could be quite indirect, the l-LN v s project to the contralateral LN v s through the posterior optic tract. Moreover, the s-LN v s project dorsally to the superior protocerebrum, the location of the DNs. ( Helfrich-Förster 1995 ). These anatomic features suggest that PDF might bind directly to clock neurons. To test this hypothesis, in vitro biotinylated PDF peptide was incubated with fixed adult brains under near physiological conditions. The bound peptide was then detected in situ with a streptavidin-conjugated enzymatic amplification reaction. The vast majority of the signal localized with numerous cells at the periphery of medulla ( Figure 5 A). This is exactly where the l-LN v s send large arborizations as their centrifugal projections ( Helfrich-Förster 1995 ). Importantly, signal was also detected coincident with the LN v s ( Figure 5 B) and likely DN3 clock neurons ( Figure 5 C) within the superior protocerebrum region, i.e., the bound peptide colocalized with GFP when the brains were from a strain with GFP-labeled clock neurons. Staining intensity was temporally constant; i.e., there was no systematic variation in signal intensity with circadian time. Although we obtained identical results with two differently biotinylated PDF peptides and there was no staining with two other biotinylated control peptides, we had difficulty to compete specifically the signal with nonbiotinylated PDF (see Materials and Methods). Moreover, PDF peptide staining of clock neurons was not reliably detected in every brain, in contrast to optic lobe staining. Nonetheless, we never detected peptide staining of other neurons in the vicinity of the LN v s; i.e., signal in this region of the brain was always coincident with the GFP-labeled LN v s. The peptide staining therefore suggests that PDF acts on the LN v s in an autocrine or paracrine fashion as well as on other clock neurons, but the results do not exclude additional, more indirect modes of action. Figure 5 A PDF Peptide Binds to Many Cells, Including Several Clock Neuronal Groups In vitro biontinylated PDF peptide was used to visualize the peptide binding locations (middle panels, with Cy3) in the brain (see Materials and Methods for details). We used membrane-bound GFP (green panels on the left) to label specific circadian neurons as well as their projections (right panels show the overlay of both channels). (A) The brain is from flies with labeled LN v s (y w,UAS–mCD8iGFP;pdf–GAL4) . Numerous cells at the periphery of the medulla have the vast majority of the bound PDF peptide signal within the brain. This region receives widespread dendritic arborizations from the l-LN v s. (B) Bound PDF peptide was also detected on the surface of LN v s at a lower intensity. LN v cell bodies were labeled using UAS–mCD8iGFP;pdf–GAL4 . Since the signal from the Cy3 channel was much weaker than the GFP signal, we reduced the output gain from the GFP channel. Sequential scanning was used to prevent cross-talk between the two channels. (C) y w,UAS–mCD8iGFP;tim–GAL4/+ flies were used to label all circadian neurons. In the dorsal region shown in this series, the arrow points to a group of DN3 neurons. Discussion The strong behavioral phenotype of the Pdf 01 mutant strain in DD indicates that PDF makes an important contribution to free-running circadian rhythms. It was, however, unanticipated that the Pdf 01 mutant would have an additional effect on transcriptional oscillations within most if not all clock neurons. This observation extends the tight parallel between strong behavioral rhythms and robust transcriptional rhythms and suggests that the behavioral damping is due to the transcriptional damping ( Marrus et al. 1996 ). In contrast to this strong effect of the Pdf 01 mutation on free-running rhythms, the molecular as well as behavioral rhythms of these mutant flies are nearly normal under LD conditions. We now interpret this difference to indicate that intercellular communication among different clock cells and neuronal groups is less important when they can independently receive photic information via cryptochrome. This probably serves not only to synchronize clock neurons but also to reinforce and strengthen the molecular oscillation ( Emery et al. 1998 ; Stanewsky et al. 1998 ). The damping phenotype includes the LN v s, which have been proposed to be the principal pacemaker neurons in Drosophila ( Helfrich-Förster 1997 ; Renn et al. 1999 ). Their counterparts in mammals, the suprachiasmatic nucleus (SCN) neurons, can support circadian rhythms independently (e.g., Sujino et al. 2003 ). However, our data indicate that the LN v s cannot support locomotor activity rhythms without other clock cell groups (see Figure 1 ). A similar attempt to rescue behavioral rhythms of an arrhythmic Clk mutant also failed ( Allada et al. 2003 ). Although the negative result shown here might be due to developmental defects of the cyc 01 mutation ( Park et al. 2000 ), the conclusion fits well with a role for PDF in functional cooperation between individual neuronal groups. Indeed, it appears that PDF secretion comprises much of what the LN v s contribute to rhythms, as the phenotype of flies missing the LN v s is virtually identical to that of the Pdf 01 strain ( Renn et al. 1999 ). There is less known about the roles of other clock neurons, although they do have specific wiring properties ( Kaneko and Hall 2000 ) as well as specific sets of gene expression profiles (unpublished data). An additional indication that other clock neurons contribute to locomotor activity rhythms is that LD behavioral rhythms do not require the LN v s ( Hardin et al. 1992 ; Renn et al. 1999 ). As the Pdf 01 strain also has a strong effect on geotaxis ( Toma et al. 2002 ), clock neurons may even contribute to other behavioral modalities. The staining pattern suggests that the PDF ligand contacts a receptor on the surface of clock neurons, including the LN v s themselves. This is consistent with the notion that PDF acts as an important intercellular cell communication molecule within the Drosophila circadian system. The dorsal projections of the s-LN v s stain rhythmically with anti-PDF antibodies, and it has been suggested that released PDF affects dorsal clock neurons ( Helfrich-Förster et al. 2000 ). Indeed, ectopic expression of PDF in neurons that project to the dorsal brain region causes severe rhythm defects, suggesting that misregulation of this signaling causes circadian system dysfunction ( Helfrich-Förster et al. 2000 ). Our staining with a PDF peptide indicates that the PDF signaling to the DNs may be direct. Although rhythmic PDF staining is restricted to the s-LN v terminals ( Park et al. 2000 ), this could be because a smaller fraction of PDF is released from the l-LN v terminals. Some of these processes follow the posterior optic track to the opposite side of the brain. Taken together with the LN v peptide staining, it is likely that PDF from the l-LN v s signals contralaterally and positively influences clock cells on the opposite side of the brain. A very recent study of the Drosophila prothoracic gland (PG) clock and eclosion rhythms suggests that the LN v s also control the PG clock via PDF signaling ( Myers et al. 2003 ). This raises the possibility that PDF not only synchronizes brain clock neurons, but also keeps peripheral clocks in pace with the core brain network. The Pdf 01 molecular phenotype implies that the wild-type organization of the system normally supports the individual clock cells as well as the entire circadian program in DD. Although we do not know that all molecular aspects of rhythms damp in DD in Pdf 01 flies, we suggest that damped transcriptional rhythms are the intracellular default state in Drosophila and are manifest without the driving and entraining LD cycle or without a functionally integrated clock network. This view is also consistent with recent studies showing that electrical silencing of clock neurons eliminates free-running molecular as well as behavioral rhythms ( Nitabach et al. 2002 ). It will be interesting to learn how PDF signaling connects to the intracellular transcriptional machinery. We note that communication among clock neurons is likely to be important in other organisms. The ability of PDF to phase-shift the cockroach circadian clock ( Petri and Stengl 1997 ) is more consistent with our proposal than with a simple role in clock output. A recent study of VPAC(2) receptor knock-out mice ( Harmar et al. 2002 ) showed that these mice fail to sustain behavioral rhythms and have molecular rhythms defects within the SCN. This raises the intriguing possibility that SCN neurons as well as Drosophila clock neurons may require network integration to sustain free-running intracellular oscillations. Materials and Methods Drosophila genetics. Full-length cyc cDNA was obtained from BDGP cDNA clone GM02625 and was tagged with hemagglutinin (HA) epitope by PCR cloning. CYC–HA was subsequently cloned into pUAST to generate pUAS–CYC–HA. The transformation plasmid was used to generate transgenetic flies. A third chromosome insertion line (UAS–CYC–HA15) was used subsequently. All wild-type flies and specimens were taken from a Canton-S stock. The circadian driver lines pdf–GAL4 ( Renn et al. 1999 ), tim–GAL4 ( Kaneko and Hall 2000 ), as well as the cyc 01 ( Rutila et al. 1998 ) and Pdf 01 ( Renn et al. 1999 ) mutant strains have been previously described. All molecular and behavioral analyses were conducted on flies entrained at 25°C. GFP expression analysis. To visualize the axon projections from circadian neurons, a UAS–mCD8GFP line labeling the cell membrane was crossed with various circadian GAL4 drivers. The progeny brains were dissected in PBS and fixed in 3.7% paraformaldehyde in PEM. After rinses in PBS plus 0.3% Triton and PBS, brains were mounted in Vectashield mounting medium (Vector Laboratories, Burlingame, California, United States) and imaged on a Leica laser scanning confocal microscope. Optical sections were taken at 1–2 μm intervals and used to construct a maximum projection image for each brain. In situ mRNA hybridization on adult brain whole mounts. In situ hybridization of tim and cry was done as described previously ( Zhao et al. 2003 ). The maximum projection images taken from a Leica laser scanning confocal microscope were used for the quantification. The quantification was done using three brain images per sample with Leica confocal software. The mean pixel intensities of cell groups were normalized by subtracting the average of two general background areas in the brain. Behavioral analysis. Flies were entrained for 3–5 d in 12 h light:12 h dark (LD) conditions before release into DD. Locomotor activities of individual flies were monitored using Trikinetics Drosophila Activity Monitors (Waltham, Massachusetts, United States). The analysis was done by using a signal processing toolbox ( Levine et al. 2002b ). Autocorrelation and spectral analysis were used to assess rhythmicity and to estimate the period. The phase information was extracted using circular statistics ( Levine et al. 2002b ). In some cases, the phases of individual Pdf 01 flies were also examined by inspection. In vitro peptide binding assay. Biotinylation of the PDF peptide was with EZ-Link Sulfo–NHS–LC–Biotin reagent (Pierce Biotechnology, Rockford, Illinois, United States), following the manufacturer's instruction. Excess biotinylation reagent was removed by prolonged incubation in Tris–HCl buffer (1 M [pH 7.5]) followed by protein purification through a Polyacrylamide 1800 desalting column (Pierce Biotechnology). A control neuropeptide, allatostatin I (Sigma-Aldrich, St. Louis, Missouri, United States), was biotinylated using the same method. A second control was a synthetic, biotinylated peptide derived from the Drosophila PER protein (a gift from P. Nawathean). In addition, a new N-terminus biotinylated PDF peptide was chemically synthesized de novo (Sigma-Aldrich). Identical results were obtained with the two PDF peptides, and no specific signal was obtained with the two control peptides. To detect the binding of the neuropeptide in the CNS of Drosophila , brains were dissected in PBS and fixed in 3.7% paraformaldehyde in PEM for 30 min. After they were rinsed in PBS plus 0.3% Triton and blocked using 1% FBS or BSA, biotinylated peptide was incubated with the brains at a final concentration of 0.2 μg/ml. The brains were washed thoroughly with TNT (0.1 M Tris–HCl [pH 7.5], 0.15 M NaCl, 0.05% Tween 20). The bound peptide was subsequently detected through the biotin label using streptavidin–HRP (NEN LifeScience, now Perkin-Elmer, Torrance, California, United States) and fluorescent tyramides (NEN LifeScience). A detailed protocol is provided as Protocol S1 . For the competition assay, unlabeled peptide was added at a 200- to 5000-fold concentration increase in the blocking step; subsequent steps were as described above. Supporting Information Figure S1 Rescued Molecular Oscillations Persist during DD in the s-LN v s The “rescued” mutant y w; pdf–GAL4;UAS–CYC , cyc 01 / cyc 01 was released into DD after entrainment and assayed by tim whole-mount in situ hybridization on the fourth day of DD. A Pdf probe was used to label the LN v group. Brains were taken at two opposite timepoints, CT3 (top panels) and CT15 (bottom panels). From left to right are Pdf (green, FITC labeled), tim (red, Cy3 labeled), and an image overlay. The lower arrows point to the s-LN v s and the upper arrows to l-LN v s. Whereas the l-LN v s show barely visible tim RNA oscillations under these conditions, the s-LN v s are obviously cycling. This difference suggests that the l-LN v s might damp more rapidly or be more light-dependent than the s-LN v s in this unusual genotype. (7.1 MB PDF). Click here for additional data file. Protocol S1 Short Protocol for Neuropeptide Biotinylation and Receptor Detection (23 KB DOC). Click here for additional data file.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC193604.xml
539354	Morphological brain differences between adult stutterers and non-stutterers	Background The neurophysiological and neuroanatomical foundations of persistent developmental stuttering (PDS) are still a matter of dispute. A main argument is that stutterers show atypical anatomical asymmetries of speech-relevant brain areas, which possibly affect speech fluency. The major aim of this study was to determine whether adults with PDS have anomalous anatomy in cortical speech-language areas. Methods Adults with PDS (n = 10) and controls (n = 10) matched for age, sex, hand preference, and education were studied using high-resolution MRI scans. Using a new variant of the voxel-based morphometry technique (augmented VBM) the brains of stutterers and non-stutterers were compared with respect to white matter (WM) and grey matter (GM) differences. Results We found increased WM volumes in a right-hemispheric network comprising the superior temporal gyrus (including the planum temporale), the inferior frontal gyrus (including the pars triangularis), the precentral gyrus in the vicinity of the face and mouth representation, and the anterior middle frontal gyrus. In addition, we detected a leftward WM asymmetry in the auditory cortex in non-stutterers, while stutterers showed symmetric WM volumes. Conclusions These results provide strong evidence that adults with PDS have anomalous anatomy not only in perisylvian speech and language areas but also in prefrontal and sensorimotor areas. Whether this atypical asymmetry of WM is the cause or the consequence of stuttering is still an unanswered question.	Background Persistent developmental stuttering (PDS) is a relatively severe disturbance characterized by involuntary, audible or silent, repetitions or prolongations of sounds or syllables. These are not readily controllable and often are accompanied by other movements and by negative emotions [ 1 , 2 ]. Developmental stuttering evolves before puberty without apparent brain damage or other known cause. Several authors suppose a hereditary component of PDS because of the relatively high concordance rate in family members of PDS subjects (70% for monozygotic twins, about 30% for dizygotic twins, and 18% for siblings of the same sex) [ 3 - 5 ]. Because of this hereditary component and the early onset of stuttering it has repeatedly been suggested that some kind of anatomical or neurophysiological predetermination increases the vulnerability for stuttering [for a summary of theories and findings related to stuttering research [ 6 ]]. Several experimental studies have shown that stutterers reveal prolonged manual and vocal reaction times to simple and complex verbal and nonverbal stimuli [ 7 - 10 ], reduced bimanual coordination measures [ 11 - 13 ], atypical functional lateralizations [ 14 - 18 ], abnormalities in the auditory system [ 19 - 21 ], or increased variability of time-critical speech parameters [ 9 , 22 , 23 ]. More recent neuroimaging studies have shown atypical hemodynamic responses in speech-related brain areas even during fluent utterances suggesting a dysfunctionally operating speech control circuit in stutterers [ 24 - 29 ]. However, although much research has been invested to understand the neurophysiological mechanisms and underpinnings of this disorder, none of the aforementioned studies provide a substantial breakthrough in understanding stuttering. Several researchers have hypothesized subtle but nevertheless crucial deficiencies in the anatomical and neurophysiological underpinnings of the speech and language system. One popular hypothesis is that stutterers would show an atypical lateralisation of the speech system (reversed or reduced laterality) thought to make the system more vulnerable to speech dysfluencies [ 17 ]. Although recent neuroimaging studies have shown atypical activation and deactivation of brain regions in adults with PDS [ 24 , 30 - 32 ] the anatomical underpinnings of stuttering have not been examined in detail so far. According to the present literature four anatomical studies revealed brain abnormalities in stutterers compared to controls. The earliest study examined two left-handed stuttering siblings using CT and revealed an atypical (reduced) anatomical asymmetry of the occipital poles [ 33 ]. The first high-resolution MRI study investigating stutterers revealed a reduced volumetric asymmetry of the planum temporale (a brain area which is involved in higher order auditory processing) and other anatomical peculiarities in speech-related areas [ 34 ]. A more recent paper of the same group revealed that PDS is also associated with atypical (mostly reduced) prefrontal and occipital lobe asymmetries [ 35 ]. In addition, deficits in language processing were associated with some anatomic measures in the adults who stutter. Using a new MRI technique (diffusion tensor imaging: DTI), that allows the assessment of white matter ultrastructure, Sommer et al. [ 36 ] found an area of decreased white matter tract coherence in the left Rolandic operculum. This structure is adjacent to the primary motor representation of tongue, larynx, and pharynx and the inferior arcuate fascicle linking temporal and frontal language areas, which both form a temporo-frontal language system involved in word perception and production. Thus, there are indeed first strong hints that the brain of stutterers differ from non-stuttering subjects on a macroanatomical level suggesting that morphological predispositions determine stuttering. Although the aforementioned anatomical studies have focussed on the anatomical foundations of stuttering, several questions are not answered yet. Therefore we re-examined the hypothesis of anatomical differences between stutterers and non-stutterers using voxel-based morphometry (VBM). This approach circumvents the problem of analyzing predetermined regions of interests by analyzing stereotactically normalized brains on a voxel-by-voxels basis with respect to differences in the volume of white matter (WM) or grey matter (GM) [ 37 , 38 ]. This approach has successfully been used in the last 8 years for several clinical populations and allows studying the morphology separately for GM and WM looking at the entire brain [ 39 - 41 ]. A further advantage of this method is the objectivity and thus rater independence. We hypothesized that beside the previously reported atypical anatomical asymmetries in perisylvian and frontal areas there should be additional differences in further brain areas also involved in speech motor control. Because several studies report that the auditory system in stutterers is dysfunctional [ 19 , 21 , 42 - 47 ] especially during speaking (thus, emphasizing the role of auditory feedback in the context of stuttering), we anticipated structural peculiarities in the auditory cortex (Heschl's gyrus and the planum temporale) in this group. In addition, we also anticipated anatomical peculiarities in frontal brain areas and in the somatosensory and motor system controlling the speech muscles. Methods Subjects The sample included adults with PDS (n = 10) and controls (n = 10) matched according to sex, age and education. All subjects were consistent right-handers (CRH) according to the Annett handedness questionnaire (AHQ) [66]. Our sample contained the approximate sex distribution as those reported in population studies of adults who stutter; thus, there were more men (n = 8) than women (n = 2) in this sample. The dysfluent sample was limited to adults with PDS who had been diagnosed with developmental stuttering before the age of 8 years and had undergone treatment at some point, but continued to be dysfluent. None of the subjects was taking centrally acting medications that could have resulted in PDS, and all met the clinical criteria of developmental stuttering – not acquired stuttering. Of the adults who stuttered, 50% had a family history of stuttering; none of the controls had a family history of stuttering. Stuttering severity was determined using the Stuttering Severity Inventory (SSI) [67] with individuals in the sample ranging from mild (2), moderate (7) to severe (1). All participants were native German speakers with no reported history of dyslexia, specific language impairment, attention deficit disorder, traumatic brain injury, substance abuse, or other neuropsychiatric conditions. All participants gave informed consent before participating. MRI scanning protocol and data analysis We used a Siemens 1.5 T magnet and a 22-min fast-low-angle-shot MR sequence yielding 128 contiguous sagittal slices with 1 × 1 × 1.17 mm image voxel size [68–70]. Data were analyzed on a PC workstation using MATLAB 5.3 (MathWorks, Natick, MA) and SPM 99 (Wellcome Dept. Cogn. Neurol, London; ) [71]. Preprocessing were guided by the VBM method proposed by Godd et al. [ 40 , 41 ]. In short, the following steps were conducted: (1) Spatial normalization of each brain to the MNI space using the MNI template; (2) spatial smoothing with an 8-mm full-width at half-maximum (FWHM) isotropic Gaussian kernel; (3) creating of a mean anatomical image from these normalized and smoothed scans; (4) stereotactic normalisation of all MRI scans (in native space) using the newly developed template and non-linear smooth spatial basis functions; (5) these spatially normalized images were resliced with a final voxel size of 2 × 2 × 2 mm 3 . The normalized scans were then segmented into grey (GM) and white matter (WM), cerebro spinal fluid (CSF), and other non-brain partitions applying the algorithmus implemented in SPM99 based on the algorithms developed by Ashburner and Friston [ 37 , 38 ]. In order to sensitize our subsequent statistical analysis not only to differences in the GM (WM) proportions but also to differences in the true GM (WM) volumes a further processing step – known as the 'Jacobian Modulation – was incorporated. The partitioned images (GM and WM) were multiplied by the Jacobian determinants of the deformation field transforming the GM and WM density values into volume equivalents [ 38 , 40 ]. The normalized, segmented (and modulated) images are smoothed using a 10-mm FWHM isotropic Gaussian kernel to improve statistical quality of the data (e.g., normal distribution). Statistical analysis of VBM data The normalized, smoothed, segmented (and modulated) data were analyzed using statistical parametric mapping (SPM99) employing the framework of the General Linear Model. Regionally specific differences in GM (and WM) (both for the density and the volume equivalents) between groups were assessed statistically using a two-tailed contrast. Corrections for the search volume (and implicit multiple comparisons) in terms of the P values were made using Gaussian random field theory, which accommodates spatial correlations inherent in the data and is now established as the conventional approach to inference in smooth spatially extended data. We restricted the search volume to the GM or WM volume enabling us to increase the statistical power of statistical testing. Significance levels for two-sided T statistics were set at T = 5 (corrected for multiple comparisons across the WM or GM volumes) and a spatial extend criterion of k = 50. The spatial extend of k = 50 was introduced because this volume size roughly corresponds to the size of a meaningful anatomical area (0.4 cm 3 ). Results Because there was no substantial difference between the results of our statistical tests for the density and volume equivalents, we only report the findings based on the analysis of the volume equivalents. We found increased WM volumes in stutterers within four clusters on the right hemisphere. The clusters are located in the superior temporal gyrus (STG) including the planum temporale, the precentral gyrus (PrCG), the inferior frontal gyrus (IFG) comprising the pars opercularis (POP), and the middle frontal gyrus (MFD) (Table 1 ). There was no significant difference between stutterers and non-stutterers with respect to the GM volumes. Table 1 Regions of increased WM volumes in stutterers. Indicated are the peak differences (in t-values), their stereotactically coordinates, and the associated anatomical labels derived from the MNI standard brain. Please note, there were no areas with increased WM volumes in controls compared to stutterers. Anatomic region Coordinates (X, Y, Z) t-Value R Superior temporal gyrus (STG) 64 -34 21 7.45 R Inferior frontal gyrus (IFG) 66 8 21 6.58 R Middle frontal gyrus (MFG) 44 48 11 6.23 R Precentral gyrus (PrCG) 30 -28 63 7.25 R Precentral gyrus (PrCG) 62 -12 37 6.53 In order to understand the differences between stutterers and non-stutterers with respect to the WM volumes in these anatomical areas more precisely, we placed regions of interest (ROI) in these anatomical areas and the homotopic areas on the left hemisphere. For the auditory cortex we used a rectangular ROI including Heschl's gyrus (HG) and the planum temporale (PT) (size of the ROI on both hemispheres: 8.4 cm 3 ). The placement pf these ROIs were guided by anatomical landmarks and published probability atlases of the HG and PT [ 48 , 49 ]. The other ROIs were defined according to the stereotactic coordinates found in the VBM analysis. For these ROIs, rectangular volumes (10 mm edge length resulting in a volume of 10 × 10 × 10 mm) were used. The mean WM measures were calculated for each ROI and subjected two-way ANOVAs with one repeated measurement factor (Hemisphere: left vs. right) and one grouping factor (Group: stutterers vs. non-stutterers). Because we found significant interaction effects for all ROIs we will only interpret these interactions. For the auditory cortex we found a strong main effect for the factor Hemisphere (F(1, 18) = 29.2, p <= 0.001, ETA 2 = 0.62) and a significant interaction between both factors (F(1, 18) = 31.6, p <= 0.001, ETA 2 = 0.64). Subsequent Scheffé contrasts and Figure 2 show that there is a strong between-hemisphere difference for non-stutterers (larger WM volume on the left hemisphere, p < 0.01) but not for stutterers (p > 0.4). For the IFG there were strong main effects (Hemisphere: F(1, 18) = 9.2, p = 0.007, ETA 2 = 0.34; Group: F(1, 18) = 23.6, p <= 0.001, ETA 2 = 0.58) and a significant interaction (F(1, 18) = 23.9, p <= 0.001, ETA 2 = 0.57). The strong interaction is qualified by a between-hemisphere difference found for stutterers (with larger WM volumes on the right compared to the left IFG) while there is no between-hemisphere difference in non-stutterers. For the PrCG we found a significant between-group difference (F(1, 18) = 11.1, p = 0.004, ETA 2 = 0.38) and a significant interaction (F(1, 18) = 31.0, p <= 0.001, ETA 2 = 0.64). The pattern of this interaction resembles the interaction found for the IFG with larger WM volumes on the right hemisphere for stutterers than on the left while non-stutterers show similar values for both hemispheres. For the MFG all two main effects as well as the interaction were strongly significant (Hemisphere: F(1, 18) = 23.5, p <= 0.001, ETA 2 = 0.56; Group: (F(1, 18) = 13.2, p = 0.002, ETA 2 = 0.42; interaction: (F(1, 18) = 18.4, p <= 0.001, ETA 2 = 0.50). The interaction is due to the fact that stutterers revealed larger WM volumes on the right compared to the left hemisphere. Figure 2 ROI analysis Mean WM volumes (and standard errors of the mean as vertical bars) in the precentral gyrus (PrCG), middle frontal gyrus (MFG), inferior frontal gyrus (IFG), and the superior temporal gyrus (STG) broken down for the left (open bars, LH) and right (filled bars, RH) hemisphere. The STG comprises Heschl's gyrus and the planum temporale. The volume measures are expressed as arbitrary values because these measures were obtained from brains transformed into the MNI space. In addition, we did not find any correlation between the stuttering severity measures (SSI measures) and the anatomical peculiarities neither in the context of the VBM nor the ROI analysis. Discussion This study was motivated by the question whether stutterers reveal morphological brain anomalies compared to non-stuttering controls. In fact, we found prominent increases of WM in stutterers within a right-hemispheric network including brain structures relevant for language and speech. These areas comprise the STG (including the auditory areas PT and HG), the IFG (including the pars opercularis which is part of Broca's right-sided homologue), the somatosensory area (including the face and mouth representation, as well as the mesial part of the hand representation), and the middle frontal gyrus (MFG). Our findings of regionally increased right-hemispheric WM in stutterers might suggest an increased and possibly atypical intrahemispheric communication within these areas via association fibres [ 50 , 51 ] possibly accompanying different processing strategies in the right hemisphere in stutterers. Three of the brain areas with different WM composition in stutterers are known to be involved in different speech and language functions. For example, the right IFG (including the pars opercularis) is involved in the perception and generation of phonological or prosodic speech features [ 52 - 55 ] while the ventral part of the precentral gyrus is part of the somatosensory representation of the mouth and tongue. The MFG has been shown to be involved during rhyme and tone perception [ 34 ]. The core region of this network is the auditory cortex with neurons specialized for tone, pitch, and prosody perception [ 56 , 57 ]. Within this circuit the auditory cortex plays a pivotal role because speech production follows the ultimate goal to generate speech sounds others can understand. The auditory cues in speech production are either phonetic cues (such as voice onset times or formant transitions) or specific suprasegmental features like duration, intensity, linguistic or emotional stress. During speech production the auditory system controls whether the appropriate auditory cues have been generated by means of auditory feedback control of the own speech. Several studies have shown that the auditory cortex is strongly involved in the continuous control of self-generated suprasegmental speech features (duration, intensity, stress pattern) and that this auditory feedback control is detrimental in stutterers [ 19 - 21 , 45 , 46 , 58 - 60 ]. The auditory cortices in the two hemispheres are relatively specialized in normal subjects [ 56 , 57 ]. Thus, temporal resolution is better in left auditory cortical areas and spectral resolution as well as processing of prolonged auditory information is better in right auditory cortical areas. It is thought that this functional specialisation is based on cytoarchitectonic peculiarities (more heavily myelinated axons and greater interconnectivity) and the relative composition of WM and GM in this area. In fact, in addition to the present findings, two previous studies [ 61 , 62 ] found a leftward asymmetry of WM volume in the auditory cortex in healthy subjects. However, our findings show that stutterers do not reveal the typical leftward asymmetry; they rather show symmetry with an atypically enlarged WM volume in the right auditory cortex. This atypical symmetry of WM volume in the auditory cortex in stutterers might suggest different and perhaps deficient processing of slowly changing auditory cues necessary to control suprasegmental features. In fact several studies have shown that stutterers reveal substantial peculiarities with respect to various aspects of the auditory feedback of their own speech especially when they control suprasegmental speech features [ 19 , 60 , 63 ]. The reported morphological features complement previous morphological studies comparing stutterers and non-stutterers. Firstly, this study shows again that stutterers reveal atypical anatomical lateralisation in speech-relevant areas. In three areas (PrCG, MFG, and IFG) stutterers reveal more WM volumes on the right than on the left. For the auditory cortex (STG) we found symmetric WM volumes while non-stutterers typically show a leftward asymmetry for this measure. Thus, some kind of hemispheric imbalance seems to be related to persistent developing stuttering. Secondly, using a different method than Foundas et al. [ 34 ], we also found an atypical anatomical lateralisation in the auditory cortex expressed as an increased symmetry of WM volume. Thus, the different "hardware" composition of the auditory cortex in stutterers is a crucial peculiarity possibly determining the processing mode of the right auditory cortex and the interaction between both auditory cortices. Taken together, the present results and findings of previous behavioural and neuroimaging studies emphasize a specific role of the auditory cortex in stuttering. Thirdly, while Foundas et al. [ 35 ] found atypical anatomical lateralisation in prefrontal areas we detected increased WM volumes in the right anterior MFG which is part of the prefrontal cortex. Thus, the atypical prefrontal lateralization may be due to atypical lateralisation of the WM volume of the right MFG. Finally, our analysis also revealed an atypical asymmetry with respect to the WM volume in vicinity of the right sensorimotor cortex (including the face and mouth representation as well as parts of the hand representation) possibly suggesting that these areas use different processing strategies as compared to non-stuttering subjects. However, although we and others found large morphological differences between stutterers and non-stuttereres we cannot rule out the possibility that the anatomical differences are the consequence of stuttering rather than the cause. Persistent developmental stuttering commences early in life forcing the affected subject to cope with this annoying and detrimental situation. Thus, some kind of adaptation or cortical reorganisation might accompany this process. Indeed, several studies indicate that intensive practise of various skills might affect the brain even on the macroanatomical level [ 64 , 65 ]. Future studies, however, are clearly needed to disentangle whether the anatomical peculiarities in stutterers are the cause or the consequence of stuttering. Conclusions These results provide strong evidence that adults with PDS have anomalous anatomy not only in perisylvian speech and language areas but also in prefrontal and sensorimotor areas. These anatomical features might indicate a deficiently working speech system. Whether this atypical asymmetry of WM is the cause or the consequence of stuttering is still an unanswered question. Competing interests The author(s) declare that they have no competing interests. Authors' contributions L.J., J.H. and H.S. conceived the experiment and drafted the manuscript. L.J. and J.H. prepared the exact experimental setup. H.S. supervised data acquisition. J.H. and L.J. performed all data and statistical analyses. All authors read and approved the final manuscript. Figure 1 VBM results Areas where stutterers show increased relative white matter (WM) volume superimposed onto the standard MNI template. The brain outline on the right indicates the four different anatomical regions showing increased WM volume in stutterers. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539354.xml
533862	Role of tyrosine phosphorylation in sperm capacitation / acrosome reaction	Capacitation is an important physiological pre-requisite before the sperm cell can acrosome react and fertilize the oocyte. Recent reports from several laboratories have amply documented that the protein phosphorylation especially at tyrosine residues is one of the most important events that occur during capacitation. In this article, we have reviewed the data from our and other laboratories, and have constructed a heuristic model for the mechanisms and molecules involved in capacitation/acrosome reaction.	Introduction The process of fertilization is characterized by a series of complex set of events. It involves a species-specific interaction between egg and sperm activating a chain of events that leads to formation of zygote, fetus and finally a baby. However, before a spermatozoon can fertilize an oocyte, it must undergo a cascade of biochemical and physiological changes that facilitates its binding and penetration into the oocyte [ 1 , 2 ]. This time-dependent acquisition of fertilizing competence has been defined as "Capacitation" [ 3 , 4 ]. Capacitation confers upon the spermatozoon an ability to gain hyperactive motility, interact with oocyte zona pellucida (ZP), undergo acrosome reaction and initiate oocyte plasma membrane fusion [ 1 ]. Capacitation normally occurs in the female genital tract, however, it can also be achieved in vitro . In fact, most of the information about various aspects of sperm capacitation has emanated from in vitro studies. Sperm cells can be capacitated in vitro by using chemically defined media containing appropriate concentrations of electrolytes, metabolic energy sources, and serum albumin (cholesterol acceptor). Although minor variations exist between these media depending on the mammalian species, most of these media contain bicarbonate, calcium and a macromolecule predominantly serum albumin. Although capacitation of a sperm cell is required before fertilization virtually in every mammalian species studied, the molecular mechanisms and signal transduction pathways involved in this process are not clearly understood. Capacitation involves an increase in membrane fluidity, cholesterol efflux, ion fluxes resulting in alteration of sperm membrane potential, increased tyrosine phosphorylation of proteins, induction of hyperactivation and the acrosome reaction. Protein phosphorylation represents a very important aspect of capacitation. Recently, many advances have been made in the study of phosphorylation of proteins during capacitation, which we will review in this article. Protein phoshphorylation Phosphorylation of proteins is a post-translational modification event that acts as one of the cell's regulatory mechanisms to control various processes such as cellular growth, cell cycle control, cytoskeleton assembly, ionic current modulation, and receptor regulation [ 5 , 6 ]. In fact in eukaryotic cells, one of the most common mechanisms for regulating protein activity is the addition and/or removal of phosphate groups from serine, threonine, or tyrosine residues of protein moieties. Addition or removal of phosphate groups can induce allosteric modifications resulting in conformational changes in proteins leading either to their activation or inactivation. Mature spermatozoa are terminally differentiated and specialized cells. They are highly compartmentalized but are devoid of any major transcriptional and translational activity. Therefore one can justify the importance of post-translational modifications such as protein phosphorylation/dephosphorylation in regulating important phenomena such as sperm capacitation, hyperactive motility and acrosome reaction, which are required for the spermatozoon to reach, bind, penetrate and fuse with the oocyte. The phosphorylation/dephosphorylation state of phosphoproteins is controlled by the activity of protein kinases and phosphatases, and the counteracting activities of these kinases and phosphatases provide cells with a "switch" that can turn on or turn off the function of various proteins. Earlier studies reported the presence of various phosphoproteins, protein kinases and protein phosphatases in mammalian spermatozoa and implicated their role in sperm motility acquisition, capacitation and acrosome reaction [ 7 , 8 ]. Phosphorylation can occur at serine, threonine, and tyrosine residues in proteins. Although both serine/threonine phosphorylation and tyrosine phosphorylation of proteins have been reported in spermatozoa (discussed below), the tyrosine phosphorylation is very important and may be the primary or even the exclusive indicator of a signal transduction pathway in a cell. Protein tyrosine phosphorylation in spermatozoa 1.Tyrosine phophorylated proteins Historically, in 1989, Leyton and Saling provided the first evidence for the presence of tyrosine phosphorylation in mammalian spermatozoa namely the mouse sperm [ 9 ]. Using anti-phosphotyrosine antibody they identified three proteins of 52, 75, and 95 kDa respectively, in mouse sperm. The 95 kDa protein showed enhancement in immunoreactivity with the antibody after sperm capacitation and interaction with oocyte ZP proteins [ 9 ]. In 1991, the second study by Naz and associates identified tyrosine phosphorylation in sperm of several mammalian species including human, rat, rabbit, and mouse. They reported four sets of tyrosine phosphorylated proteins in the molecular weight range of 95 kDa/94 ± 3 kDa (FA-2 antigen), 46 ± 3 kDa, 25 ± 7 kDa and 12 ± 2 kDa, respectively, in human sperm [ 10 ] and also identified a protein of molecular identity of 94 ± 3 kDa in mouse sperm, that was reported earlier by Leyton and Saling [ 9 ]. However, this protein of 94 ± 3 kDa was not identified in rat and rabbit sperm. Although it needs to be confirmed using molecular cloning and sequencing studies, it seems that 94 ± 3 kDa is not an evolutionarily conserved protein. Using 32 P metabolic labeling and in vitro kinase assays, human sperm was found to have at least seven proteins (200, 112, 104, 48, 42, 31 and 25 kDa) that are phosphorylated and fourteen proteins (122, 105, 95, 89, 73, 62, 48, 46, 40, 33, 30, 28, 25 and 22 kDa) that are autophosphorylated [ 11 ]. Further studies showed that the 94 ± 3 kDa and 46 ± 3 kDa proteins are also phosphorylated at ser/thr residues besides phosphorylation at tyrosine residues [ 12 ]. The 46 ± 3 kDa protein was found out to be the FA-1 antigen, which has been known to play an important role in sperm-ZP binding [ 13 ]. FA-1 antigen also plays an important role in capacitation [ 14 , 15 ]. Treatment of human spermatozoa with an anti-FA-1 monoclonal antibody (mAb) during capacitation reduces tyrosine phosphorylation of both 94 ± 3 kDa and 46 ± 3 kDa (FA-1 antigen) proteins, which indicates cross-talk between these two proteins [ 11 , 15 ]. It has been observed that different compartments of human spermatozoa undergo a specific sequence of phosphorylation during capacitation and upon binding to zona pellucida [ 16 ]. In order to establish the link between the different phosphorylated proteins and a specific sperm function, it is necessary to differentially localize the tyrosine phosphorylated proteins in various regions of spermatozoon. The flagellum seems to be the major component of sperm cell that undergoes tyrosine phosphorylation in most species except boar [ 17 ]. Immunocytochemistry has been used to localize tyrosine phosphorylated proteins in flagellum of human [ 10 , 18 , 19 ], monkey [ 20 ], hamster [ 21 ], rat [ 22 ], and mouse [ 23 ] spermatozoa. In a study using immunofluorescence, Urner et al [ 23 ] localized the phosphotyrosine proteins in the flagellum during capacitation, zona pellucida binding and gamete fusion in mouse sperm. They observed that during capacitation there is an increase in the proportion of spermatozoa with phosphorylated proteins in the whole flagellum [ 23 ]. The increase in the phosphorylation in the principal-piece precedes that in the mid-piece, and phosphorylation in the principal- piece is the pre-requisite for phosphorylation in the mid-piece region. Upon binding to the zona pellucida, nearly all mouse sperm became progressively phosphorylated in both the principal-piece and the mid-piece regions [ 23 ]. A significant increase in phosphorylation with capacitation has also been observed in human spermatozoa but it is localized mainly in the principal piece [ 18 , 19 ]. A study from our laboratory using human sperm revealed that the capacitating conditions and zona exposure increases the degree of tyrosine phosphorylation per sperm cell as well as the number of sperm cells that were phosphorylated, especially in the acrosomal regions of the sperm head [ 10 ]. Interestingly, with these changes, there was also a shift in the site of phosphotyrosine-specific fluorescence from the tail regions of non-capacitated sperm to the acrosomal regions of capacitated/zona-exposed sperm cells. In other cellular systems, there are reports indicating a shift in subcellular localization of various proteins after phosphorylation. In human epidermoid carcinoma A431 cells, it has been shown that the binding of epidermal growth factor (EGF) to its receptor rapidly triggers redistribution of phospholipase C-r 1 from a predominantly cytosolic localization to the membrane-bound activity, followed by phosphorylation at tyrosine residues. Since the acrosomal region of the sperm cell is involved in sperm-zona interaction, the shift in phosphotyrosine-specific fluorescence seems to have a physiological significance [ 10 ]. Tyrosine phosphorylation of the sperm flagellar proteins has shown to be related to the acquisition of the hyperactive motility [ 20 , 24 ], which is required for the spermatozoa to penetrate the cumulus and the zona pellucida of the oocyte. In human spermatozoa, the protein A-kinase anchoring proteins (AKAPs) localized on the fibrous sheath, namely AKAP82, its precursor pro-AKAP82, and FSP95 are the most prominent tyrosine phosphorylated proteins during capacitation [ 18 , 25 ]. In hamster sperm, the homolog of mouse AKAP4 has been identified as the major tyrosine phosphorylated protein in the capacitated spermatozoa [ 26 ]. In contrast, in the mouse sperm, AKAP4 is phosphorylated at ser/thr residues not at tyrosine residues. So there are species-specific variations in the pattern of tyrosine phosphorylation even for the same protein. Another protein that gets tyrosine phosphorylated during capacitation is CABYR (calcium-binding and tyrosine phosphorylation-regulated protein). It has been localized on the principal- piece of human spermatozoa [ 27 ]. It is speculated that the CABYR is involved in cross-talk between tyrosine phosphorylation and Ca 2+ in the signal transduction pathway. A 55 kDa tyrosine phosphorylated protein has been linked to motility in bovine spermatozoa [ 28 ]. During capacitation of mouse spermatozoa, heat shock protein (HSP)-90, a highly evolutionary conserved molecular chaperone protein, becomes tyrosine phosphorylated [ 29 ]. HSP-90 is also tyrosine phosphorylated in human and rat spermatozoa when incubated under conditions that induce capacitation [ 29 ]. In boar spermatozoa, capacitation induces tyrosine phosphorylation of plasma membrane proteins, which are believed to initiate binding to the zona pellucida and induce acrosome reaction [ 30 ]. It has been shown that capacitation induces tyrosine phosphorylation of three major (27 kDa, 37 kDa and 40 kDa) and three minor (34 kDa, 47 kDa and 55 kDa) plasma membrane proteins [ 30 ]. In a later study, two plasma membrane proteins isolated from capacitated boar sperm cells (35 kDa and 46 kDa) showed high binding affinity with zona pellucida [ 31 ]. These two proteins are most likely the 34 and 47 kDa proteins identified earlier. Although, phosphorylation of sperm proteins is a key feature of capacitation, it is not clear how tyrosine phosphorylation of these proteins is involved in sperm-zona recognition or interaction, and acrosomal exocytosis. However in a recent study, Asquith et al [ 32 ] have examined the relationship between protein phosphorylation and the ability of mouse spermatozoa to interact with zona pellucida. They have identified two chaperone proteins namely endoplasmin (erp99) and heat shock protein 60 (hsp60) expressed on the surface of mouse spermatozoa. Both erp99 and hsp60 proteins are tyrosine phosphorylated and are localized on the plasma membrane of sperm head, the region that participates in zona binding. They proposed that "activation" of erp99 and hsp60 proteins by tyrosine phosphorylation during capacitation may trigger conformational changes facilitating the formation of a functional zona pellucida receptor complex on the surface of spermatozoa [ 32 ]. On a similar line, two sperm proteins namely, ERK-1 and ERK-2 (extracellular signal-regulated kinases) have been identified to be tyrosine phosphorylated and "activated" during capacitation in human spermatozoa [ 33 ]. It is interesting to know that sperm-oviductal epithelial cell interaction in vitro modifies both the sperm tyrosine phosphorylation and capacitation. The selective sperm binding to oviductal epithelial cells [ 34 ] suppresses tyrosine phosphorylation of sperm proteins in boar [ 17 ] and canine [ 35 ] spermatozoa delaying capacitation. This oviductal modulation of tyrosine phosphorulation/capacitation may help to synchronize sperm function with the time of ovulation. In order to understand the molecular basis underlying capacitation, it is very important to characterize phosphoproteins involved in the signal transduction pathways. Although a large number of proteins have been reported to be tyrosine phosphorylated, very few have been characterized so far. Most of the studies at the beginning used specific inhibitors and/or phospho-specific antibodies to delineate phosphoproteins in sperm cell. Mass-spectrometric analysis provides another powerful approach to identify and characterize these phosphorylated proteins. Ficarro et al [ 36 ] used this approach to identify proteins phosphorylated during capacitation of human sperm. They also investigated the phosphorylation sites in these proteins. They are able to map more than 60 phosphorylated sequences in sperm cell. They also provided evidence for tyrosine phosphorylation of two proteins namely, valosin-containing protein (VCP) and AKAP3 (sperm tail protein) during capacitation [ 36 ]. Also, the gene knockout technology is very helpful in delineating phosphoproteins that have a role in the tyrosine phosphorylation cascade. The studies have shown that the targeted disruption of the Akap4 gene causes defects in sperm flagellum and motility. In the mice lacking AKAP4 protein, that undergoes phosphorylation during capacitation, sperm numbers were not reduced but the spermatozoa failed to show progressive motility, rendering the male mice infertile [ 37 ]. The mice showed defect in fibrous sheath formation indicating that the AKAP4 is a scaffold protein required for the organization and integrity of the fibrous sheath. The sperm motility is lost in the absence of AKAP4 protein because signal transduction and glycolytic enzymes do not associate with the fibrous sheath [ 37 ]. It would be interesting to examine the gene knockouts of other proteins to delineate the missing links in signal transduction pathways involved in capacitation/acrosomal exocytosis. 2.Molecular mechanisms of signal transduction in sperm capacitation / acrosome reaction Several studies have correlated the degree of tyrosine phosphorylation with the capacitative state of spermatozoa. Visconti et al [ 38 , 39 ] examined the correlation between the capacitative state and protein tyrosine phosphorylation in mouse spermatozoa. They observed a time-dependent increase in the protein tyrosine phosphorylation of a set of specific proteins in the molecular range of 40–120 kDa, which was correlated with the capacitation state of spermatozoa [ 38 ]. Later studies reported that the protein tyrosine phosphorylation increases in spermatozoa during capacitation in various species, including hamsters [ 40 , 41 ], cats [ 42 ], pigs [ 43 ], boar [ 44 ], bovine [ 45 , 46 ], equine [ 47 ], cynomolgus monkey [ 20 ], tammar wallaby and brushtail possum (marsupial species) [ 48 ] and human [ 49 , 50 ]. All these studies provide evidence that the protein tyrosine phosphorylation is an important regulatory pathway in modulating the events associated with capacitation. The increase in protein tyrosine phosphorylation during capacitation has been shown to be regulated by a cAMP-dependent pathway involving protein kinase A (PKA) in sperm of various species including mouse [ 39 ], hamster [ 40 ], boar [ 44 ], bovine [ 45 , 46 ], equine [ 47 ], cynomolgus monkey [ 20 ] and human [ 49 , 50 ]. cAMP is a ubiquitous and "central" second messenger in all cell types. It may target cyclic nucleotide gated ion-channels, cAMP-activated guanine nucleotide exchange protein, and PKA in different signaling pathways. In sperm, cAMP has been shown to activate PKA, which regulates protein tyrosine phosphorylation. This signaling pathway is unique to sperm. It has been observed that addition of H89, a protein kinase A inhibitor, during capacitation reduces/blocks and addition of cell permeable analog of cAMP, dibutyryl cAMP, increases tyrosine phosphorylation of sperm proteins [ 51 ]. PKA is a tetrameric enzyme composed of two regulatory and two catalytic subunits. The activity of PKA is dependent on the activities of adenylate cyclase and phosphodiesterase. In sperm cell, PKA is compartmentalized thus ensuring specificity of function through binding of its regulatory subunit to the AKAP family of proteins. Different types of AKAPs have been characterized from sperm of different species [ 18 , 52 , 53 ]. PKA although a serine/threonine kinase, induces and causes an increase in tyrosine phosphorylation of sperm proteins, by indirect activation of tyrosine kinases [ 49 ]. Such kinases must be highly specific to spermatozoa, since cAMP-dependent tyrosine phosphorylation has not been reported in any other cell type examined to date. Tyrosine kinases can be divided into two classes namely the receptor tyrosine kinases (RTKs) and non-receptor protein tyrosine kinases (PTKs). RTKs are transmembrane proteins having an extracellular ligand binding domain and an intracellular tyrosine kinase domain. The localization of phospholipase C-γ(PLCγ) on the plasma membrane and its tyrosine phosphorylation-dependent activation in mouse spermatozoa [ 54 ] and phosphoinositide 3 kinase (PI-3 kinase) activity operating downstream of tyrosine phosphorylation in human spermatozoa [ 55 ], indirectly indicate the presence of RTKs. Upon extracellular ligand binding, a RTK is activated and then phosphorylate itself (autophosphorylation) or other proteins. In contrast, PTKs are located in the cytoplasm, nucleus or anchored to the inner leaflet of the plasma membrane. Presence of various tyrosine kinases (RTKs and PTKs) has been demonstrated in spermatozoa of several mammalian species. These include c-ras in human sperm cells [ 56 ], EGF receptor tyrosine kinase in human, mouse, rabbit and rat sperm cells [ 57 ] and bovine sperm [ 58 ], c-Abl in human sperm cells [ 59 ] and p190 c-met tyrosine kinase in human sperm cells [ 60 ]. Recently another tyrosine kinase TK-32 have been identified in pig spermatozoa and it has been demonstrated that activation of TK-32 occur concomitant with capacitation [ 61 ]. Insulin-like growth factor I (IGF-I) receptor has been identified in human [ 62 ], and bovine [ 63 ] sperm. IGF-1 receptor has tyrosine kinase activity and its ligand IGF-1 is present in the seminal plasma. The IGF-1 system (IGF-1/IGF-1R and IGF-1 binding protein) may be involved in the signal transduction pathway leading to sperm capacitation and acrosome exocytosis [ 62 ]. The src -related protein tyrosine kinase, c- yes (cellular-yamaguchi sarcoma viral oncogene) has been detected in the human sperm head [ 64 ]. The c- yes belongs to PTK family. The activity of the c- yes kinase depends on cAMP, indicating again that tyrosine phosphorylation of proteins in sperm cell is a result of the cross-talk between the cAMP pathway and tyrosine kinase(s). The protein kinase C (PKC) is also present in mammalian spermatozoa and its role has been implicated in sperm motility and acrosome reaction [ 65 ], but its function in capacitation is poorly understood. a. PKA pathway The sperm proteins also undergo phosphorylation at ser/thr residues. Studies have shown capacitation-associated increase in phosphorylated ser/thr residues in human [ 12 ] and hamster [ 66 ] spermatozoa. Besides PKA (a major ser/thr kinase), the extracellular signal-regulated protein kinase (ERK1/2), which is also a ser/thr kinase, is present in sperm cell [ 67 ]. Earlier studies using anti-phosphoserine and anti-phosphothreonine antibodies indicated an increase in ser/thr phosphorylation of 18, 35, 43, 55, 94, 110 and 190 kDa proteins during capacitation of human sperm [ 12 ]. A recent study used antibody against the Arg-X-X- (Phospho-ser/thr) motif to study the PKA-dependent ser/thr phosphorylation in capacitating human sperm cell in combination with the agents that stimulates (dibutyryl cAMP, dbcAMP and 3-isobutyl-1-methylxanthine, IBMX) or inhibit (H89 and Rp-adenosine-3', 5'-cyclic monophosphorothionate, Rp-cAMPS) PKA [ 68 ]. A significant increase in phosphorylation of proteins designated as P80 and P105, respectively, was observed. This study showed for the first time that the phosphorylation of Arg-X-X- (ser/thr) motif that is characteristic of PKA substrates increases during capacitation. These ser/thr-phosphorylated proteins could provide a link between the early increase in PKA activity that is followed by protein tyrosine phosphorylation. In another report, two 36 kDa proteins (36 k-A and 36 k-B) were phosphorylated in a cAMP-dependant manner at serine residues, in hamster sperm flagella during an increase in the hyperactivated motility [ 69 ]. A ser/thr protein kinase, ecto-cyclic AMP-independent protein kinase (ecto-CIK) has been localized on the outer surface of mature goat spermatozoa [ 70 ]. Its substrate MPS (major physiological substrate) has been characterized and localized in sperm cell and is speculated to play an important role in sperm-egg interaction in this species [ 70 ]. b. MAP Kinase Pathway There is evidence that the mitogen-activated protein kinases (MAPKs), also known as extracellular signal-regulated kinases (ERKs), are present in spermatozoa and are involved in sperm motility and capacitation [ 56 , 67 ] and acrosome reaction [ 71 , 72 ]. The MAP kinases are ser/thr kinases that are involved in signal transduction pathways in several cell types. Their activity is regulated by a cascade of events, which starts with the activation of p21 Ras that stimulates a ser/thr kinase Raf (MAPK kinase). Raf phosphorylates and activates MEK. MEK (MAPK kinase), which is a dual specificity kinase, phosphorylates ERK1 and ERK2 (p44/p42 MAPK, respectively). The ERK pathway has been shown to be present in fowl [ 73 ] and human spermatozoa [ 71 ]. In fowl sperm, it is involved in the phosphorylation of axonemal and/or accessory cytoskeletal proteins and in the regulation of flagellar motility [ 73 ]. In human sperm, progesterone stimulates the ERK2 (p42ERK) and thus this kinase may have a role in capacitation and acrosomal exocytosis [ 71 ]. The MAPK isoform ERK2 [ 71 ], the adaptor protein Shc [ 74 ] and Ras [ 56 ] have been localized in human sperm head indicating that this pathway may be required for regulating protein phosphorylation in sperm. It has been speculated that MAP kinase may phosphorylate proteins that influence protein tyrosine phosphorylation indirectly. Inhibition of MAPK blocks protein tyrosine phosphorylation associated with capacitation. c. Factors That Affect Tyrosine Phosphorylation during Sperm Capacitation a. Cholesterol The lipid composition of sperm plasma membrane is different from that of somatic cell plasma membrane. The plasma membrane of sperm head has high amounts of cholesterol, which regulates the membrane fluidity and plays an important role of capacitation [ 75 ]. The cholesterol from sperm plasma membrane is transferred to high molecular weight proteins such as albumin and high-density lipoproteins present in the oviductal fluid as the sperm cell traverses through the female genital tract. Cholesterol efflux from sperm plasma membrane is associated with the activation of membrane signal transduction pathways related to capacitation [ 76 ]. β-Cyclodextrins (cyclic heptasaccharides) promote cholesterol efflux from mouse sperm plasma membrane that results in an increase in capacitation and protein tyrosine phosphorylation through the cAMP/PKA pathway [ 76 ]. cAMP analogues can induce protein tyrosine phosphorylation in the absence of bovine serum albumin (BSA) and can also overcome the inhibitory effect of cholesterol-3-sulphate in medium containing BSA. Rp-cAMPS, an inhibitor of PKA, decreases the degree of tyrosine phosphorylation induced by either BSA or cyclodextrins [ 50 ]. These findings strongly indicate an important role of cholesterol efflux in tyrosine phosphorylation, mediated through the cAMP/PKA pathway. Cholesterol efflux can also have an indirect influence on the signaling pathways. An increase in membrane fluidity after Ca 2+ efflux can increase the permeability of sperm cell plasma membrane to various ions such as Ca 2+ and HCO 3 - , which can then affect the downstream signaling molecules. b. Calcium Ca 2+ influx is one of the crucial biochemical events that occur during capacitation. There is abundant evidence to support requirement of Ca 2+ for capacitation [ 77 , 78 ] and induction of acrosome reaction [ 79 ]. It has been demonstrated that there is an increase in concentration of Ca 2+ in sperm cell during capacitation in several mammalian species [ 80 - 85 ]. It has been observed that different species have different requirements for Ca 2+ during capacitation. There is also evidence that different aspects of human sperm function such as capacitation, acrosome reaction and ZP binding in vitro has distinctive Ca 2+ requirements. Experiments have shown that in humans 0.22 mM of Ca 2+ ions are needed for capacitation while ≥ 0.58 mM is required for AR and ZP binding [ 86 ]. Although micromolar concentration of extracellular Ca 2+ is needed to achieve sperm capacitation in the mouse [ 77 ], millimolar concentration is required for capacitation in man [ 78 , 87 ]. Contrasting reports exist on the impact of extracellular Ca 2+ on tyrosine phosphorylation in spermatozoa. The reports on mouse [ 38 ] and human spermatozoa [ 88 ] have documented that increasing amounts of extracellular Ca 2+ increase tyrosine phosphorylation. In contrast, other studies have demonstrated the opposite effect [ 18 , 89 ], indicating that Ca 2+ negatively regulates tyrosine phosphorylation during sperm capacitation. A recent study was conducted by Baker et al [ 90 ] to investigate the impact of extracellular Ca 2+ on tyrosine phosphorylation of human and mouse spermatozoa and delineate the mechanisms by which this cation exerts its regulatory effect. They reported that Ca 2+ suppresses tyrosine phosphorylation by decreasing the availability of intracellular ATP [ 90 ]. Since there is an increase in the intracellular concentration of Ca 2+ during capacitation of sperm of several mammalian species [ 33 ], it could be speculated that such changes modulate protein tyrosine phosphorylation to regulate capacitation and acrosomal exocytosis. c. Bicorbonate (HCO 3 - ) The requirement of for HCO 3 - in capacitation is well elucidated [ 91 , 92 ]. The exact mechanism by which HCO 3 - regulates capacitation is not clear. However, the HCO 3 - influx has been associated with an increase in intracellular pH observed during capacitation, regulation of cAMP levels, reversible change in the lipid architecture of plasma membrane, and hyperpolarization of sperm plasma membrane. The presence of a Na + /HCO 3 - cotransporter has been demonstrated in the mouse sperm and it plays a significant role in capacitation [ 93 ]. An increase in intracellular pH during capacitation is attributed to HCO 3 - influx [ 94 ]. However, the role of pH is not clear since an increase in pH in sperm cell does not induce capacitation [ 95 ]. HCO 3 - may be involved in stimulation of adenylate cyclase activity rather than modulation of pH [ 39 , 96 ]. HCO 3 - induces rapid changes in plasma membrane lipid architecture and in motility by a cAMP/PKA-dependent pathway in boar spermatozoa [ 97 ]. Harrison [ 97 ] and Da Ros et al [ 98 ] have shown the role of HCO 3 - in protein tyrosine phosphorylation in sperm during capacitation and fertilization. It is interesting that HCO 3 - levels are low in epididymis and high in seminal plasma and oviduct [ 99 ]. These changes in the concentration of HCO 3 - in the male and female reproductive tracts could play an important role in the suppression of capacitation in the epididymis and the promotion of this process in vivo in the female reproductive tract. It is observed that a protein, designated as CRISP-1, that is added to the sperm surface in epididymis is lost during capacitation [ 100 ]. Addition of exogenous CRISP-1 to the incubation medium inhibits tyrosine phosphorylation in a concentration-dependent manner, thus inhibiting capacitation and subsequently the acrosome reaction. A recent study investigated the effect of in vitro addition of HCO 3 - on intracellular cAMP production and protein tyrosine phosphorylation in human spermatozoa [ 101 ]. The addition of HCO 3 - in the medium resulted in a significant increase in sperm motility as well as several hyperactivation parameters, mediated by an increase in cAMP production and tyrosine phosphorylation of AKAP3. The effects disappeared with the addition of 4,4'-diisothiocyanostilbene-2, 2'-disulfonic acid, which is a known inhibitor of bicarbonate transport. It was observed that both tyrosine phosphorylation of AKAP3 and sperm motility were blunted by the soluble adenylate cyclase (sAC) inhibitor 2OH-estradiol. These findings indicate that HCO 3 - stimulates human sperm motility and hyperactivation through activation of sAC and tyrosine phosphorylation of AKAP3, causing an increased recruitment of PKA to AKAP3 [ 101 ]. d. Reactive Oxygen species (ROS) Studies have suggested that ROS such as H 2 O 2 and superoxide anion are involved in the regulation of human sperm capacitation and protein tyrosine phosphorylation [ 19 , 102 ]. Some reports suggest that the ROS action lie upstream from cAMP site in the signal transduction cascade [ 103 ], while others believe that ROS is localized downstream from cAMP [ 88 ]. In a recent study, Rivlin et al [ 104 ] suggested that H 2 O 2 activates adenylate cyclase to produce cAMP, leading to PKA-dependent protein tyrosine phosphorylation. Since adenylate cyclase present in sperm is activated by HCO 3 - , they proposed that H 2 O 2 and HCO 3 - can act in concert or partially substitute each other [ 104 ]. However, ROS also has detrimental effects on spermatozoa; while low concentrations of H 2 O 2 are beneficial, high concentrations can lead to sperm immobilization and death. It has been proposed that the accurate balance of amount of ROS produced and scavenged at any moment determines whether a sperm function will be promoted or jeopardized. Nicotinamide adenine dinucleotide (NADH) and NADPH are also known to promote sperm capacitation [ 104 ]. It has been found that exogenous NADPH enhances protein tyrosine phosphorylation in bovine sperm [ 104 ] and promote capacitation. NADPH acts as a coenzyme for sperm oxidase that generates superoxide anion, which is later dismutated to H 2 O 2 by superoxide dismutase. The inhibition of lactate dehydrogenase (LDH)-C4 blocks capacitation of mouse spermatozoa in vitro [ 105 ]. LDH-C4 is a mammalian testis-specific enzyme and is the only lactate dehydrogenase isozyme present in sperm. LDH-C4 seems to play an important metabolic role in sperm capacitation. The oxidation of NADH with the conversion of pyruvate to lactate by LDH provides ATP necessary for PKA activity [ 105 ]. NADH-diaphorase (enzyme that transfer electrons from NADH to an electron acceptor such as 2,6-dichlorophenol-indophenol) plays a role in spermatozoal function through ROS generation. Overproduction of ROS due to high diaphorase activity has been observed in some infertile men [ 106 ]. The free radical nitric oxide (NO) generated by spermatozoa has been implicated in sperm function [ 107 ]. Non-capacitated human spermatozoa produce low levels of NO, whereas under capacitating conditions, a time dependent increase in NO synthesis has been observed [ 108 ]. It has been reported that the increased levels of NO during capacitation modulate cAMP pathway that regulates the downstream protein tyrosine phosphorylation [ 109 ]. In vitro studies have shown that low concentration of NO enhances the acrosome reaction of mouse [ 110 ] and bull [ 111 ] spermatozoa, as well as the zona pellucida binding ability of human sperm [ 112 ]. e. Progesterone Progesterone has been reported to affect several sperm functions especially capacitation, motility and acrosome reaction [ 113 ]. The effects of progesterone on spermatozoa are mediated via progesterone binding sites/progesterone receptor (PR) on the acrosomal membrane [ 114 ]. Different types of PRs have been shown to be present on the sperm plasma membrane. These are: plasma membrane Ca 2+ channel (PR1), a membrane-associated protein tyrosine kinase (PTK; PR2), and a plasma membrane chloride channel (PR3) [ 115 ]. The progesterone stimulates Ca 2+ influx in the human spermatozoa through PR1 [ 114 ] and voltage-dependent calcium channels (VDCC) [ 116 ]. In human spermatozoa, progesterone stimulates tyrosine phosphorylation of sperm proteins [ 117 , 118 ] causing hyperactivation [ 119 ] with an increase in cAMP levels [ 117 ]. The tyrosine kinase-associated PR (PR2) is responsible for the effect of progesterone on the hyperactive motility and acrosome reaction. PR3 mediates the chloride influx, which takes place during the acrosome reaction [ 115 ]. Progesterone also increases the membrane fluidity of human sperm plasma membrane, which is an important event in sperm capacitation and tyrosine phosphorylation [ 115 ]. f. Gamma-Aminobutyric Acid (GABA) Recently GABA has emerged as a putative modulator of sperm function. GABA is the most widely distributed inhibitory neurotransmitter in vertebrate central nervous system. Three types of membrane receptors (A, B and C) mediate the inhibitory action of GABA. GABA A-receptor has been identified in human spermatozoa. In a recent study, the in vitro effect of GABA was studied on bovine spermatozoa capacitation [ 120 ]. It was observed that addition of GABA to the incubation medium results in a concentration-dependent increase in the percentage of capacitated spermatozoa. A significant increase in intracellular Ca 2+ and cAMP levels was induced by GABA, and the GABA A-R antagonists, bicuculline or picrotoxin abolished the effect [ 120 ]. Thus, GABA seems to induce sperm capacitation through a signal transduction pathway involving Ca 2+ , cAMP and tyrosine phosphorylation. g. Angiotensin II (AII) AII is known to be present in seminal plasma and it modulates the adenylate cyclase (AC)/cAMP signal transduction pathway [ 121 ]. The existence of a class of angiotensin receptors (AT1) has been shown in bovine spermatozoa. In the capacitated sperm, AII AT1 receptors are localized in the head and tail, whereas in non-capacitated cells the receptors are localized only in the tail region [ 122 ]. AII significantly stimulates cAMP production in capacitated mouse spermatozoa with an associated increase in protein tyrosine phosphorylation [ 123 ]. h. Cytokines Cytokines are a family of polypeptide hormones produced primarily by the cells of the immune system in response to various stimuli including foreign antigens. Although originally identified from the immune cells, the non-immune cells also secrete them. The cytokines can have both positive and negative effects on a variety of cell types and tissues. Cytokines are present in circulation and local genital tract secretions in both men and women [ 124 , 125 ]. The effects of various cytokines on sperm cell have also been studied. It has been observed that cytokines can affect sperm cell motility, capacitation, acrosome reaction, zona-pellucida binding and penetration and embryonic development in positive or negative manner [ 126 ]. It has been observed that cytokines namely, interferon-α (IFN-α), interferon-γ (IFN-γ) and tumor necrosis factor-α (TNF-α) have negative effect on sperm motility [ 126 ], while interleukin-6 (IL-6) enhances sperm cell capacitation and/or acrosome reaction [ 127 ]. However, the mechanisms by which cytokines affect sperm functions are not clear. Identification of cytokine receptors in the sperm cell is an important step in elucidating the mechanism of cytokine action. Very few cytokine receptors have been identified on sperm cells till date. A study in our laboratory, demonstrated the presence of IFN-α and IFN-γ receptors in mouse, rabbit, pig and human sperm [ 128 ]. The presence of interleukin-2α (IL-2α) [ 129 ], interleukin-2β (IL-2β) [ 129 ], insulin like growth factor-1 (IGF-1) [ 62 , 63 ] and c-met (hepatocyte growth factor receptor) [ 60 ] receptors on human sperm surface has also been documented. A recent study identified the presence of CX(3)CR1 receptors on human sperm [ 130 ]. Fractalkine, a CX(3)C chemokine, is present in fallopian tubes. It might play an important role in maintaining the motility of spermatozoa and their ability to undergo the acrosome reaction when sperm transit through fallopian tubes [ 130 ]. However, the effect of various cytokines on tyrosine phosphorylation during capacitation has not been extensively studied. Since many of the cytokines affect capacitation/acrosome reaction and tyrosine phosphorylation is required for these processes, it can be envisaged that these cytokines modulate tyrosine phosphorylation in sperm cell. Conclusions Before fertilization can occur, spermatozoon undergoes a series of changes to acquire the ability to bind and penetrate the oocyte. These events are regulated by the activation of intracellular signaling pathways involving various molecules such as cAMP, protein kinase A, receptor tyrosine kinases, and non-receptor tyrosine kinases. A number of molecules have been identified, that regulate these pathways such as calcium, bicarbonate, ROS, GABA, progesterone, angiotensin, and cytokines. Studies have shown that phosphorylation of sperm proteins is an important aspect of capacitation and has been shown to be associated with hyperactivated motility, zona pellucida binding and acrosome reaction. Extensive research has started to elucidate various pathways involved in protein phosphorylation during sperm capacitation. Presence of three major pathways involving cAMP/PKA, receptor tyrosine kinases, and non-receptor protein tyrosine kinases have been shown. These pathways are not mutually exclusive and may involve cross-talk among several molecules. The exact pathways have not been clearly delineated. Several questions such as how does the stimulation of cAMP/PKA pathway upregulate tyrosine phosphorylation are still to be answered. Many components of these pathways and several phosphoproteins remain to be identified. The novel technologies such as gene knockout technique will help us to elucidate the key molecules in these pathways. Based upon the present data, we have constructed a heuristic model that is shown in figure 1 . Figure 1 Heuristic model showing the tyrosine phosphorylation signaling pathways in sperm cell involved in capacitation . There seems to be three major signaling pathways operating in sperm cell, namely cAMP/PKA-dependent pathway (pathway I), receptor tyrosine kinase pathway (pathway II), and non-receptor protein tyrosine kinase pathway (pathway III). The pathway I, that is exclusive only to sperm cells, seems to be the major pathway among the three pathways. These cascades may not be mutually exclusive and include cross-talk among several molecules. Many key molecules and receptors are still to be identified to completely elucidate the molecular mechanism and signal transduction cascade involved in capacitation. G-protein coupled receptor pathway has not been included in this model.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533862.xml
526262	Patients' perspectives on high-tech home care: a qualitative inquiry into the user-friendliness of four technologies	Background The delivery of technology-enhanced home care is growing in most industrialized countries. The objective of our study was to document, from the patient's perspective, how the level of user-friendliness of medical technology influences its integration into the private and social lives of patients. Understanding what makes a technology user-friendly should help improve the design of home care services. Methods Four home care interventions that are frequently used and vary in their technical and clinical features were selected: Antibiotic intravenous therapy, parenteral nutrition, peritoneal dialysis and oxygen therapy. Our qualitative study relied on the triangulation of three sources of data: 1) interviews with patients (n = 16); 2) interviews with carers (n = 6); and 3) direct observation of nursing visits of a different set of patients (n = 16). Participants of varying socioeconomic status were recruited through primary care organizations and hospitals that deliver home care within 100 km of Montreal, the largest urban area in the province of Quebec, Canada. Results The four interventions have both a negative and positive effect on patients' lives. These technologies were rarely perceived as user-friendly, and user-acceptance was closely linked to user-competence. Compared with acute I.V. patients, who tended to be passive, chronic patients seemed keener to master technical aspects. While some of the technical and human barriers were managed well in the home setting, engaging in the social world was more problematic. Most patients found it difficult to maintain a regular job because of the high frequency of treatment, while some carers found their autonomy and social lives restricted. Patients also tended to withdraw from social activities because of social stigmatization and technical barriers. Conclusions While technology contributes to improving the patients' health, it also imposes significant constraints on their lives. Policies aimed at developing home care must clearly integrate principles and resources supporting the appropriate use of technology. Close monitoring of patients should be part of all technology-enhanced home care programs.	Background The possibility of managing patients outside of the hospital has rarely been so widely considered. Indeed, home care is often seen as less costly and more patient-friendly [ 1 ]. However, the effectiveness and safety of home care have yet to be subjected to rigorous study [ 2 , 3 ]. The transfer of care from the hospital to the home raises economic and organizational issues [ 2 ]. For example, low levels of physician involvement [ 4 ] and significant increases in private spending [ 5 ] have been observed. Furthermore, the impact of the increasing use of technology in home care has not been examined [ 6 ]. A recent survey showed that 98% of Quebec (Canada) primary care organizations are using programmable pumps to deliver intravenous antibiotics therapy at home, and 84% are providing home oxygen therapy [ 7 ]. The growth of technology in home care means that lay people with varying levels of technical skills and education become direct users of health technology [ 6 ]. Therefore, the objective of our study was to document, from the patient's perspective, how the level of user-friendliness of medical technology influences its integration into the private and social lives of patients. More specifically, we adopted a "technology-in-practice perspective", which relies on qualitative in-depth investigation, by looking at what technologies do and help accomplish in the daily practices of technology users and in the organization of health care [ 8 ]. Sullivan [ 9 ] showed how growing interest in the patient's perspective is the result of convergent trends in health and social scientific research. There is a growing appreciation of how the patients' values affect their experience of a chronic health state. Medical sociologists have shown that shifting family and social relationships shape patients' perceptions and coping strategies [ 10 ]. For instance, Lowton and Gabe [ 11 ] observed that adults with cystic fibrosis, who were not expected to live for long, deployed diverse strategies to downplay the importance of their illness and compare themselves favourably to "normal, healthy" people. Although clinicians are often concerned about patient compliance, the technical and human dimensions at the root of this problem remain understudied [ 10 , 12 ]. Observers of high-tech home care have stressed that certain health technologies are, by their very design, unfriendly [ 6 ]. Nevertheless, very little is known about the characteristics that facilitate or impede the use of medical devices and whether patients perceive them as user-friendly or not. The term user-friendly is used to characterize an object – often a computer system – as "easy to operate or understand; not needing special training" [ 13 ]. This notion has gained impetus over the last 20 years with the growth of information technology and research into Human-Computer Interface (HCI). In general, the human-machine interface is seen as key in enabling a smooth fit between the user, the task and the technology [ 14 , 15 ]. The need to design interfaces that users can rapidly understand and interact with was recognized several decades ago as it impacted workers' efficiency. Although the "overriding ethos within the community of system designers has been to try to ensure that the system is user-friendly," it remains difficult for designers to grasp all of the subtleties that shape users' needs and practices [[ 15 ]: p.126]. According to Norman, "There is a big difference between the expertise required to be a designer and that required to be a user. In their work, designers often become expert with the device they are designing. Users are often expert at the task they are trying to perform with the device" [[ 16 ]: p.156]. In the field of health care, this specialization often adds to the complexity of the work of designers, who, in addition to not being end users of the device, are simply unfamiliar with the complicated tasks health care providers are achieving through the use of technology. When moving health technologies away from the hospital and into the patient's home, the design characteristics of medical devices become even more salient, since patients have to learn how to operate them safely and with confidence. According to Norman [ 16 ], the use of any device is learned more readily if the user has a good conceptual model. "This requires that the principles of operation be observable, that all actions be consistent with the conceptual model, and that the visible parts of the device reflect the current state of the device in a way consistent with that model" [[ 16 ]: p.189]. Hence, the designer must create a conceptual model that is understandable for the user and that captures the important steps of the operation of the device. In a similar perspective, when assessing the level of user-friendliness, Lun [ 14 ] has suggested paying attention to two components: 1) user-acceptance – the extent to which the user is favourable to using the technology; and 2) user-competence – the abilities required to use the technology effectively. These two components interact with each other – the more technically complex a technology, the more elaborate the user training required. This author also underlined three principles for designing user-friendly interfaces: 1) human-machine interaction is pivotal; 2) this interaction evolves through use; and 3) user should be the key informant. Methodologically speaking, these principles imply that designers should compile users' perspectives, directly observe how technology is being used, and identify the learning curve by which technology is appropriated by users. Because most computer interfaces are used in a somewhat confined environment, the work of scholars who have studied technical aids for the disabled and the elderly brings another dimension to the definition of user-friendliness. Conceptualizing disability as a social phenomenon, the user-friendliness of technical aids has to be gauged with respect to their ability to assist users' move freely in their social environment [ 17 ]. From this perspective, autonomy and mobility become prominent, as well as the impact of the technology on users' social identity. For instance, Pippin and Fernie [ 18 ] conducted focus groups and interviews with elderly patients in order to explore the following issues: Acceptance of dependence, experience of social stigma, recognition of one's own physical loss, appearance of technical aids, autonomy and perceptions of alternatives. This work shows that users have to cope not only with the technology but, more importantly, with the limitations they themselves experience (i.e., architectural barriers, growing old) and that may become the object of others' gazes as soon as they engage in environments outside of their private sphere. Thus, the importance of user-acceptance and user-competence will vary depending on where the technology is used. The level of user-friendliness would then be a function the type of settings in which users circulate, which affects how they succeed in (re)constructing their identity as a medical technology-user. Social scientific work on patients learning how to cope with chronic illnesses and life-sustaining technology has offered similar observations, highlighting that each patient tends to go through a personal trajectory, or engage in "biographical work" whose aim is to give meaning to a constellation of unfolding events [ 10 ]. Here also, the setting where services are provided is likely to influence patients' perceptions and coping strategies. Although the hospital enables the patient to adopt the sick role rather straightforwardly, the home setting may force him/her to be more active and show optimism [ 5 , 6 ]. Family members and caregivers are also affected by the use of high-tech home care. They might be asked to provide technical and moral assistance, while coping with a profoundly modified family dynamic. In certain cases, providing assistance implies inflicting pain and discomfort [ 6 ]. To summarize, the literature underscores the many technical (e.g. weight, functionality, complexity) and human (e.g. self-image, cognitive resources, social stigma, pain, etc.) variables that influence the use of technology and which are affected by the setting (institutional, private or public) where technology use takes place. The user-friendliness of a technology therefore results from a smooth fit between human and technical features, with the fit varying between and within settings and individuals. Figure 1 illustrates the relationships between these variables. This framework posits that technical dimensions largely influence user-acceptance, and human dimensions will affect user-competence. In addition, the level of autonomy that a technology can provide in private and social settings is both shaped by its technical features and the human factors associated with its use. Accordingly, this study sought to define, from the patient's perspective, the extent to which different home care interventions could be considered user-friendly and how they were integrated into patients' private and social lives. Methods We selected four interventions: Antibiotic intravenous therapy (IV), parenteral nutrition (PN), oxygen therapy (O2) and peritoneal dialysis (PD). These were chosen because they are frequently used [ 7 ] and vary in their technical and clinical features, and so are likely to differentially influence how users interact on a daily basis with them (see Table 1 ). Our study relied on the triangulation of three sources of data. We conducted semi-directed, individual interviews with patients (n = 16) and carers (n = 6), and directly observed nursing visits, involving a different set of patients (n = 16). This strategy enabled us to gather data on a broader set of patients. The carers we interviewed were not necessarily related to a patient participating to our study (to reduce pressures on interviewees) and were spouses or family members (often mother or daughter) of a person receiving high-tech home care. Our purposeful sampling strategy was to diversify viewpoints, by including participants of varying socioeconomic status, gender and age (Table 2 ). These variables are all likely to affect how patients and their carers adapt to the use of technology (e.g. contracting out private home care services, adapting the home, understanding written instructions, etc.). All participants were recruited through primary care organizations (for IV and O2) or hospital-based home care programs (for PN and PD) located within 100 km of Montreal, the largest urban centre in the province of Quebec (Canada). A member of the nursing staff in these organizations was asked to give a brochure to all eligible patients explaining the objectives and procedures of the study. After contact had been established between patients and our research team, we constructed the sample according to our diversification variables. We obtained approval from the organizations' ethics committees. Our approach was structured according to symbolic interactionism, which focuses on how individuals, through regular interactions, develop shared meanings and conceptualize, perceive and understand the role of technology [[ 19 ]: p.201]. This approach was particularly helpful in identifying how patients, formal caregivers and informal carers, through their experience in interacting together, anticipated and defined the contributions and responsibilities of each other. Interviews were biographical, relying on Lafaille and Lebeer's technique for examining coping strategies [ 20 ]. Interview questionnaires were structured to systematically explore the themes highlighted by our framework while allowing the interviewee to develop or introduce issues he/she felt were important (questionnaire available upon request). Carers' interviews were key to eliciting how they, themselves, perceive the technology – they often have to intervene when the patient is tired or not feeling well – and how the patients' social lives were transformed because of the use of technology. Interviews lasted 60–120 minutes, and were audio-taped with the written consent of the interviewee, then transcribed into electronic format. Direct observations enabled a better understanding of how patients were educated about, and supported in the use of technology [ 21 ]. An observation guide was used to rapidly record descriptive notes during the visit, while a structured summary of key events was written up subsequent to the visit [ 22 ]. The NUD*IST software was used to code and selectively retrieve verbatim extracts [ 23 ]. A mixed strategy was applied: Codes were either derived from our framework or created when their recurrence across interviews became significant. Our analyses were designed to compare and contrast the participants' experiences with using technology, both inside and outside the patient's home. We drew up tables [ 24 ], summarizing the observations stemming from the three sources of data, to identify the main technical and human factors at play in private and social settings for each of the four technologies. Most verbatim extracts were translated from French to English, then slightly edited for the purposes of this paper. Results Home care technology transforms the patient's life both inside and outside the home In the patient's home User-acceptance was shaped by different types of anxiety (see Table 3 ). In the case of IV and PN, the catheter access site must be protected to avoid potential infections and dislodging of the catheter. The alarm system of the programmable pump, used for both interventions, tends to go off too easily (e.g. occlusions when the tube gets twisted). These false alarms were initially perceived as very stressful but, over time, they became a "normal" disturbance: "I don't really sleep at night. I'm afraid the catheter will get dislodged and the alarm will go off" (Interview, PN, w3). O2 patients were concerned about the risk of fire when cooking over a gas stove or being in a room with smokers. PD patients were preoccupied by a demanding regimen that required them to balance treatment with meals and other daily activities. Nevertheless, some enjoyed being empowered through greater involvement: "You do all the follow-up yourself: Why is my blood pressure high today? Do I have edema?" (Interview, PD, m2). User-competence was affected by the relationship between patient and carer. One PD patient felt annoyed by his wife who, at the beginning, was checking whether he was applying the aseptic procedures rigorously (Obs., PD, m1). The opposite was observed for a PN patient, whose partner felt useless and avoided her during the treatments (Obs., PN, w3). The IV patients' perceptions of the technical aspects of their technology were striking. With use of this technology being, in most cases, temporary, patients were generally passive or even submissive: "You're always a slave to it, having to carry it everywhere" (Interview, IV therapy, w2). User-acceptance is in fact closely linked to competence. Older patients on IV did not feel comfortable with the electronic components of the programmable pump, which they associated with the "computer age" and about which they felt ignorant. Chronic patients seemed, in general, keener to master the technical aspects. "You can't live without air! So you have to be careful and do it right" (Interview, O2, carer, m1). One PN patient was technically confident and had developed her own technique for preventing air bubbles from forming in the tube (Obs., PN, w1). A similar confidence was shown by the carer of a PD patient: "When you see all that stuff – the reservoir, the wires – you wonder if you'll be able to do it! But, once you know how, it's easy" (Interview, PD, carer, w1). For all four interventions, manual dexterity was required to properly manipulate the different components. "If my eyes were okay, I'd have been able to do it. But I was frightened of not doing it properly, of not seeing the needle, which is so tiny" (Interview, IV, carer, w1). We also observed patients who were not able to read messages on the digital screen due to poor eyesight (due to old age or co-morbidity), limited English linguistic skills, or illiteracy. They relied on their memory or made informed guesses when operating the device. Finally, the technology did not always fit neatly in the home setting: "Well, you wouldn't believe how hot [the room gets] when the door is closed!" (Interview, O2, caregiver, m3). Some O2 patients liked to have an extra set of tubing so they could use a second floor or sit outside on a patio (Obs., O2, w1). One PD patient planned to have an evacuation system installed so he would not have to dispose of the solution exiting the tubing from his peritoneal cavity through the toilet anymore (Obs., DP, m1). In the patient's social life While some of the barriers described above can be managed fairly well in the home setting, problems arise in the unpredictable "outside world": "It's great when you're at home where you're all set up. But when you're out, you're always worried about people lighting up a cigarette" (Interview, O2, carer, m3) (see Table 4 ). In addition, O2 patients did not like to be seen with nasal tubes, and less often invited friends over or ate in restaurants (Obs., O2, w1). In the case of PN patients, who rarely or never eat food, it was relatives who felt uncomfortable and tended to invite them over less often. Carers sometimes curtailed their social activities because they felt needed by the patient: "I didn't dare go out, absolutely not" (Interview, IV, carer, w1). The mother of a woman with PN was "always worried" and "always available" (Interview, PN, carer, w1). A wife "found the manual PD a burden – four times a day... It's like being in jail, you can't go anywhere" (Interview, PD, carer, w1). The non-retired patients experienced major obstacles in continuing with employment because of the frequency and/or duration of treatments. Few of these patients had a full-time job. Being "hooked up" to a fixed O2 concentrator for up to 18 hours/day is not compatible with many types of work, and portable cylinders also restrict autonomy (2–4 hours). One PN patient said it was "impossible to work because of being connected for 12 hours", and receiving lots of fluid at night compromised her sleep (Interview, PN, w2). On the other hand, one PD patient chose a nocturnal exchange regulator, refusing hospital-based hemodialysis, in order to work: "You can't work if you go to the hospital three times a week, and work is very important to me" (Interview, PD, m2). In the latter case, the comparison with an alternative makes PD more acceptable. More crucial was the "black-and-white" definition of ability to work that governs disability pension plans, which is incompatible with what patients experience: "You know, sometimes you feel okay, and sometimes you don't; sometimes you're disabled and sometimes you're not..." (Interview, PN, w3). Overall, due to compromised health and technical barriers, most patients agreed to apply for a disability plan (according to which they cannot accept paid work). A number of patients did volunteer work, such as helping neighbourhood kids with their homework (Obs., PN, w1), doing clerical work for the family business (Obs., O2, w1), or volunteer work for a patient association (Obs., O2, w3). Discussion This study, adopting a "technology-in-practice" perspective [ 8 ], shows that the four interventions have both a negative and a positive effect on the lives of patients and carers. Indeed, our analyses sought to provide more detail on how technology simultaneously improves and constrains patients' lives. This is compatible with Pierret's observation that "Medicine gives the chronically ill reason to hope, even as it produces limitations with which these persons have to live by making adjustments to meet everyday requirements" [[ 12 ]: p.14]. Although each technology provided patients with relative autonomy from the hospital, none of them were seen by patients as truly user-friendly. IV patients remained passive and accepting, knowing that the constraints were temporary, and although O2 does not require a high level of competence, user-acceptance remained very low, especially in public places. Patients seemed more likely to develop competence in using both PN and PD because alternatives in these cases are limited (e.g. hospital-based services or death) and acceptance becomes the only way to make sense of this whole (life-long) experience. Such findings highlight the need to increase the fit between users and technology through a better design of high-tech home care devices and through effective patient education strategies. Indeed, competence and acceptance are likely to be mutually reinforced, especially if patients are supported and their know-how re-assessed over time. Although this is already part of the nursing staff duties, the experiences shared by our interviewees and the literature [ 1 , 2 ] indicate that the level of support they receive may be insufficient. As observed in a recent U.K. survey on the quality of primary health care provision, crucial factors include educational training, patient education programs and improved communication and teamwork [ 25 ]. In the case of home care, manifesting a greater concern with supporting patient education is particularly relevant in the current health policy context [ 1 , 5 ], where high-tech home care is increasingly seen as an "easy solution" to budgetary constraints and a growing elderly population. However, this study pinpoints the possibility that using high-tech home care without a proper patient support system might create more problems than it would solve. As stressed by Sinding, "It is only in more collectively oriented social action that higher standards of care can be established (or re-established) as within the purview of health professionals' duties, and thus confirmed as patients', and carers', entitlements" [[ 26 ]: p.1384]. It is in this perspective that this study's results take on significance. In particular, two issues related to policy-making and clinical practice require prompt consideration. First, technology is often designed as though patients all possess similar abilities, and are neither ageing or incapacitated by other illnesses or physical disabilities [ 18 ]. Since the development of home care is largely industry-driven [ 6 ], technology designers should be asked to gather and more explicitly integrate feedback from different groups of users with varying cognitive and physical capacities [ 27 , 28 ]. Norman [ 16 ] suggests paying attention to two components that affect the use of technology – the design model and the user's model. Both are more or less implicit explanations (or "road maps") about how to operate the device. When the gap between these two models becomes too significant, misuse – which can lead to ineffective treatments or harmful consequences – is likely to happen. Since the biomedical equipment designer and the patient rarely, if ever, interact, it is critical that the proper use of the technology be "communicated" through its physical appearance, by the way it responds (visual and/or audio feedback), and by its fit with the private and social settings where patients are evolving. Second, competence of patients and their carers should be reinforced through adequate education and support from physicians and nurses. Medical specialists who manage hospital-based home care programs for O2, PN, PD patients could emphasize, when enrolling patients, the need for a smooth fit between the technical and human barriers that affect patient compliance. Training programs for nurses could also focus on skills and routines that help increase the user-friendliness of technology [ 29 , 30 ]. Finally, both patients and clinicians need to be involved in redesigning home care services so they meet the diverse and changing needs of chronic patients [[ 28 ]: p.877]. This study has sought to better understand how the level of user-friendliness of medical technology influences its integration into the private and social lives of patients. In this regard, qualitative methods are particularly well suited for uncovering patients' views. Nevertheless, the limitations of our study should be acknowledged [ 21 , 31 ]. The reason for including four different interventions was to define which technical and human dimensions make health technology user-friendly (or not). This study design characteristic increased the complexity of the sampling strategy. For instance, we could not explore the specific role of variables such as gender, age and ethnicity [ 28 ]. The study design, however, put a broader perspective on the research problem. Indeed, redundancy from one interview to another and the growing saturation of our analytical categories suggest that we have captured the key elements associated with the introduction of high-tech home care into patients' lives. Overall, the triangulation of three data sources increased the credibility (or internal validity) of our findings by sharpening our understanding of how technologies are integrated into patients' lives. Our findings should, therefore, be applicable in countries and regions where similar devices are used and in similar home settings [ 21 , 31 ]. Finally, although lower levels of criticism were to be expected, thereby reflecting positive functional avoidance [ 26 ], overall, the participants we interviewed expressed several grievances. This is similar to the findings of a recent qualitative study where patients, in retrospect, "regretted accepting, in hope, the offer of 'active treatment' because of reduced quality of life" [[ 28 ]: p.4]. Because home care patients may, over time, adopt less than optimal routines, the concept of "acceptance" thus requires careful analysis and further investigation. There is a danger of forgetting about the "social and political factors that sustain perceptions of health system constraints" as being unavoidable, and therefore acceptable [[ 26 ]: p.1384]. Conclusion This paper shows that the barriers facing home care patients can easily be "taken for granted," as though nothing can be done to improve the situation. With a growing elderly population and limited health care resources, this will become a major issue in most industrialized countries. According to McGarry, "The home environment as a location of care provision is largely beyond the public-professional gaze, and therefore, remains potentially hidden from scrutiny" [[ 32 ]: p.429]. In the future, the delivery of high-tech home care is likely to grow [ 1 ]. Nonetheless, this study indicates that patients who are asked to become users of medical technology face major challenges. As stressed by Sullivan, moving beyond a discussion of the benefits of technology to patients' health, to a consideration of both the positive and negative impact of technology on patients' lives, not only brings medicine closer to issues that really matter to patients, but also generates "greater scientific, ethical, and social complexity" [[ 9 ]: p.1602]. Home care involves more than simply transferring a particular technology from the hospital to the home – it requires transferring knowledge and skills to lay people, and making sure that the home and social environments enable a safe, effective, appropriate and personally satisfying use of technology. Otherwise, ineffective, potentially hazardous and socially compromising treatments may be disseminated. Policies aimed at increasing the provision of home care must carefully integrate principles and resources that support the appropriate use of technology, and close monitoring of patients must be part of all technology-enhanced home care programs [ 3 ]. Competing interests The author declares that she has no competing interests. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526262.xml
509295	Deconstructing Genetic Contributions to Autoimmunity in Mouse Models	null	Given the overwhelming complexity of the immune system, it's no wonder that unraveling the mechanisms responsible for immunological disease has proved so difficult. The factors that trigger autoimmunity—which involves a breakdown in the body's ability to tolerate its own molecules—are not well understood, though animal studies show that genetic predisposition greatly increases risk. And that's where the real challenge begins. For certain diseases, individuals with defects in both copies of a specific gene invariably develop the disease. But more often, diseases with an inherited component result from the complex interplay of a variety of genes, each contributing a small effect that is typically dependent both on the expression of other genes and on both random and environmental factors. Researchers have increasingly turned to mice to model autoimmunity in humans and have found the same genetic complexity at work, with different strains of lupus-prone mice having different genes predisposing them to autoimmune disease. One model involves targeted disruption of candidate immune system genes to study their role in disease. Gene targeting experiments modify or remove a gene of interest and then watch for corresponding effects on the organism's physiology. Interpretations of results from these experiments have traditionally been predicated on the assumption that the “background,” or nontargeted, genes do not contribute to observed physiological changes. Yet in some studies, mice without targeted mutations develop an unexpected susceptibility to the autoimmune disease under study. Genetic background can induce autoimmunity in knock-out mice To investigate what effects background genes might be having in these mouse models, Marina Botto and colleagues compared the genomes of three hybrid strains of the most commonly used genetic background—the 129 and C57BL/6 hybrid mice. One of the hybrids, which carries a mutation in both copies of the Apcs gene, was chosen as an example of a gene-targeted model that develops a lupus-like disease, offering an opportunity to examine the relative contributions of the targeted versus background genes. Apcs is a candidate gene for human systemic lupus erythematosus (SLE), a form of autoimmunity marked by chronic inflammation resulting from a sustained attack on antibodies throughout the body. Botto and colleagues found that several genomic regions from both the 129 and C57BL/6 mice contributed to autoimmunity, even in the absence of gene-targeted mutations. All of the hybrid strains developed autoimmunity, though disease was more severe in mice with Apcs mutations. Disease outcome in gene-targeted mice, it turns out, can be influenced not just by disease-susceptible (or disease-resistant) gene variants near the targeted gene (in this case, Apcs ) but by the random heterogeneity of the hybrid genetic background: multiple combinations of genes in the hybrids can produce the same result. These results fall in line with mounting evidence that background genes are not silent partners in gene-targeted disease models, but can themselves facilitate expression of the disease. This finding underscores the notion that genes are not solitary, static entities; their expression often depends on context. With genetically complex diseases, having the requisite combination of susceptibility genes does not always lead to disease. Much work remains to be done to identify the triggers that cause the immune system to turn on itself. For more on mouse models and autoimmunity, see the primer by Morel in this issue.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509295.xml
509297	Forgetting, Reminding, and Remembering: The Retrieval of Lost Spatial Memory	Retrograde amnesia can occur after brain damage because this disrupts sites of storage, interrupts memory consolidation, or interferes with memory retrieval. While the retrieval failure account has been considered in several animal studies, recent work has focused mainly on memory consolidation, and the neural mechanisms responsible for reactivating memory from stored traces remain poorly understood. We now describe a new retrieval phenomenon in which rats' memory for a spatial location in a watermaze was first weakened by partial lesions of the hippocampus to a level at which it could not be detected. The animals were then reminded by the provision of incomplete and potentially misleading information—an escape platform in a novel location. Paradoxically, both incorrect and correct place information reactivated dormant memory traces equally, such that the previously trained spatial memory was now expressed. It was also established that the reminding procedure could not itself generate new learning in either the original environment, or in a new training situation. The key finding is the development of a protocol that definitively distinguishes reminding from new place learning and thereby reveals that a failure of memory during watermaze testing can arise, at least in part, from a disruption of memory retrieval.	Introduction For more than a century, the phenomenon of retrograde amnesia (RA)—the loss of memory for events that occur prior to a variety of precipitating brain insults—has provided the foundation for theories of memory consolidation and the locus of trace storage ( McGaugh 1966 ; Davis and Squire 1984 ; Dudai and Morris 2000 ). However, RA may also reflect the inability of a memory system to access a trace—a failure of memory retrieval ( Warrington and Weiskrantz 1968 ). This very dilemma was noted by Ribot (1883 , p. 475) in his seminal discussion of RA: “Two suppositions are equally warranted, viz., that either the registration of the prior states has been effaced; or that the retention of the anterior states persisting, their aptitude for being revived by associations with the present is destroyed. We are not in a position to decide between these two hypotheses.” Studies of RA have favoured a memory-consolidation interpretation in instances in which systematic variation of the time interval between experience or training and the subsequent brain insult has revealed a temporal gradation of RA ( Squire 1992 ). Computational models also point to the need for a rapid encoding and storage system, together with a slower interleaving mechanism that is thought to underlie systems-level consolidation and long-term storage in the cortex (e.g., McClelland et al. 1995 ). However, the existence of some amnesic patients with long, flat gradients of RA extending for years or decades into periods of their life when memory function was normal provided some of the first evidence that RA might be due to retrieval failure ( Sanders and Warrington 1971 ). This perspective on RA was initially supported by studies indicating that, in the anterograde domain, impaired memory could be alleviated by partial cues ( Warrington and Weiskrantz 1968 ). However, these observations were later construed as reflecting the operation of a separate memory phenomenon called priming ( Graf et al. 1984 ). Several animal studies have also indicated that a variety of ‘reminder' treatments delivered prior to retention testing can realize the expression of lost memories ( Gold et al. 1973 ; Miller and Springer 1973 ; Spear 1973 ; Gold and King 1974 ; Riccio and Richardson 1984 ; Sara 1999 ), but it is not easy to distinguish priming-induced memory from explicit recall and recognition in animal studies. Experimental resolution of the consolidation-versus-retrieval controversy has been notoriously difficult, and no consensus has been achieved. A key methodological issue, and the focus of the new technique described here, concerns the need to demonstrate that the memory observed after a reminder treatment results from the reactivation of an existing memory ( Miller and Springer 1972 ), rather than a facilitation of new learning ( Gold et al. 1973 ). In studies of spatial memory using the watermaze, amnesia for the location of the escape platform in posttraining probe trials (PTs) has generally been interpreted as a failure of learning, consolidation, or storage ( D'Hooge and De Deyn 2001 ). To investigate the alternative possibility of retrieval failure, we deliberately created conditions that should maximize the possibility of seeing such an effect. This involved training rats to find an escape platform in a specific location followed by partial lesioning of the hippocampus. We reasoned that this would weaken but not completely disrupt the memory of the correct location by damaging a subset of the ensemble of stored traces. The animals' memory was tested and observed to be undetectable. This same memory test provided, however, the opportunity to remind animals that escape from the water was possible via an escape platform in the correct or incorrect location. One hour later, the animals' memory was tested again. We observed that memory was now detectably above chance and was equally strong when the animals had previously been given correct or potentially misleading information about the current location of the platform. Additional control procedures, and the performance of other groups with sham or complete hippocampal lesions, established that the earlier failure of memory must have been due, at least in part, to retrieval failure. Results A summary of the experimental design is provided in Figure 1 (see Materials and Methods ). Figure 1 Experimental Design Outline of the different phases of testing. The platform position used during training is indicated by a red circle in the NE quadrant of the pool (large blue circle), although in practice platform locations were counterbalanced between NE and SW locations. The novel location, to which a subset of rats was exposed during reminding, is indicated by a black circle in the SW quadrant. This position was always opposite to that used during training. PT1 and PT2: probe test 1 and 2. The hatched areas represent the original training quadrant irrespective of the position of the platform (i.e., original or novel) during retention testing. PT n 1 and PT n 2: PTs during new context learning in the second pool. Training Prior to the Lesions During cued pretraining, the rats quickly learned to search for, and climb onto, the visually cued escape platform. In the main spatial training phase, the animals rapidly learned to locate and raise the platform in order to escape from the pool ( Figure 2 ), as indicated by the highly significant reduction in latencies over trials ( F [7.78, 412] = 30.4, p < 0.001). Only animals that reached the acquisition criterion received lesions (69 out of 73 rats trained). The prospective lesion groups, trained as normal animals, did not differ ( F < 1, n = 59; see Surgery below). Figure 2 Training Mean latencies to escape from the water and climb onto the hidden platform during task acquisition. Data are averaged in blocks of five trials and grouped according to the lesion made at the end of training; note that all animals were unoperated during acquisition. Only rats that reached criterion (mean escape latency less than 15 s over the last ten trials) and whose lesions were considered acceptable (see Results: Surgery) are presented. Animals rapidly learned to locate the escape platform, and prospective lesion groups did not differ. Surgery Of the 69 animals that received lesions, one died after surgery and nine were excluded based on strict histological criteria, leaving a total of 59 animals (22 sham lesions, 19 complete hippocampal lesions, and 18 partial hippocampal lesions; see Figure 3 ). Figure 3 Lesion Analysis Representative photomicrographs of cresyl-violet-stained coronal brain sections taken from subjects belonging to each of the three lesion groups—partial hippocampal lesion (A), sham lesion (B), and complete hippocampal lesion (C). In each case, sections corresponding to anterior, middle, and posterior levels of the hippocampus are displayed. The mean area of spared hippocampal tissue in each group (see Materials and Methods for calculation) is plotted below in (D). Note that the volumes of spared tissue in the septal and temporal halves of the hippocampus are plotted separately, but these values are still expressed as percentages of the entire hippocampal volume—hence the value of 50% per half in shams. The cartoon hippocampi accompanying the graph indicate lesioned tissue in dark grey, and spared tissue in light cream. As intended, partially lesioned rats exhibited substantial sparing only in the septal (dorsal) half of the hippocampus, and rats with complete hippocampal lesions exhibited minimal sparing (less than 5% at either pole). Retention Testing The key new findings are shown in Figures 4 and 5 using two separate but related measures of memory retrieval: percentage time in quadrant ( Figure 4 ) and a more sensitive measure, percentage time in a zone centred on the platform location ( Figure 5 ; see Materials and Methods ). An overall analysis of variance (ANOVA) of percentage time in the training (where the platform was located during training) and the opposite quadrants of the pool revealed a significant quadruple interaction ( F [2, 53] = 7.66, p < 0.01) involving two between-subject factors: lesion group and platform location during the reminder treatment (original versus novel), and two within-subject factors: PT (PT1 and PT2) and quadrant (training versus opposite). In both figures, the initial memory expressed during PT1 is shown in the left lane. This reveals that the partially lesioned rats were at chance, whereas the sham-lesioned rats could remember the location of the platform ( t = 6.15, df = 21, p < 0.005, paired-sample t-test, training versus opposite quadrant). The complete-lesioned animals were at chance. Analysis of percentage time in zone ( Figure 5 ) likewise confirmed that memory was detectable in the sham lesion group ( t = 4.18, df = 21, p < 0.005, one-sample t-test, comparison with chance = 50%), but not in the two lesion groups. Figure 4 Retention Testing: Quadrant Analysis Percentage time during PT1 and PT2 spent in the training and opposite quadrants of the pool (left and right lanes) and the reminder treatment (grey central lane). The training location is represented as a red circle in the NE quadrant, and the novel location (novel subgroups only) as a black circle in the SW quadrant. In practice, NE and SW quadrants were counterbalanced. Rats with partial hippocampal lesions were unable to remember the platform location on PT1 but could be reminded of the training location by exposure, at the end of PT1, to a platform in the original or a novel location. (Note that the ‘reminder' lane simply refers to this exposure to a platform—PT1 is itself the ‘reminder trial.') The key finding is that the improvement in PT2 occurred irrespective of the platform location during reminding. In contrast, sham-lesioned animals exhibited some reversal learning upon exposure to the platform in a novel location. Complete-lesioned rats did not remember the platform location during either PT1 or PT2. * p < 0.05; ** p < 0.01; n.s. = nonsignificant; comparisons with chance = 50%; one-sample t-tests. Representative swim paths are included. Figure 5 Retention Testing: Zone Analysis Percentage time in PT1 (left) and PT2 (right) spent within a zone, 20 cm in radius, centred on either the original training location (broken circle; grey) or an equivalent location in the opposite quadrant (broken circle; yellow), expressed as a percentage of the total time spent in both of the zones. The reminder treatment is again shown as the grey central lane and as the location where the hidden platform became available at the end of PT1 within these zones (original = red; novel = black). Consistent with Figure 4 , rats with partial hippocampal lesions were amnesic in PT1 but could be reminded of the correct location, even by exposure to the platform in a novel location. * p < 0.05; ** p < 0.01; n.s. = nonsignificant; comparisons with chance = 50%; one-sample t-tests. PT1 ended with the animals finding the platform in the original training location, or in a novel location in the ‘opposite' quadrant of the pool (middle lane in Figures 4 and 5 ; see Materials and Methods for explanation of terminology). These different events at the end of the swim trial potentially served both as a reminder of what happens in a watermaze, namely, escape from the water at a particular location, and/or as an opportunity for new learning. We reasoned that if the reward of escaping from the water served only to support new learning, animals capable of learning would show an enhanced bias towards the training location after finding the platform in the original location, but a reduced bias after finding it in the opposite novel location. Conversely, if these events served only as reminder cues, they might be equally effective in reminding the rats of the original training location. The key new finding is that the partial lesion group displayed a bias for the training quadrant that was equivalent whether the animals had found the platform in the original training location or in the novel opposite location, at the end of PT1. The overall ANOVA of the PT2 quadrant data revealed a triple interaction of lesion group × quadrant (training versus opposite) × platform location during reminding (novel versus original) ( F [2, 53] = 19.28, p < 0.001). With respect to the performance of the partial lesion group alone on this quadrant measure (see Figure 4 , right lane), there was a significant improvement between PT1 and PT2 ( F [1, 16] = 7.98, p < 0.02) and no difference between novel and original reminding locations ( F < 1). The partial lesion group also showed a highly significant preference for the training quadrant versus the opposite quadrant on PT2 ( F [1, 16] = 16.83, p < 0.001). The same pattern of results is apparent in the zone data (see Figure 5 ) where, overall, the partial lesion group displayed a significant improvement between PT1 and PT2 ( F [1, 16] = 7.64, p < 0.02) that also did not differ between ‘novel' and ‘original' groups ( F < 1). Because a bias for the training location appeared even in the animals that were exposed to a novel platform position, memory on PT2 cannot be attributed to relearning of the platform location. In contrast, sham-lesioned animals behaved quite differently in PT2 as a function of whether the platform was presented in the original or the novel location during the reminder treatment. Performance showed a further bias towards the training location between PT1 and PT2 following the event of climbing onto the escape platform in its original location, but exposure to the novel location resulted in a reduction in time spent in the training zone—a partial reversal. Supported by significant interactions in the overall ANOVA, analysis of time spent in the training quadrant revealed that, as expected, sham-lesioned animals reexposed to the original location increased their time there between PT1 and PT2 ( F [1, 11] = 12.41, p < 0.005). Conversely, sham-lesioned animals exposed to the novel location exhibited modest reversal learning, increasing their time in the opposite quadrant ( F [1, 9] = 9.35, p < 0.02). The same pattern of results was obtained from the analysis of time in the training zone ( Figure 5 ), for which a significant interaction between PT (PT1 or PT2) and platform location during reminding (original versus novel) was observed ( F [1, 20] = 5.46, p < 0.05). Complete-lesioned rats performed at chance during all PTs (see Figures 4 and 5 , left and right lanes). That is, their behaviour during the retention tests before and after the reminder treatment showed no impact of that treatment. Novel Context Learning As an independent test of whether the reminder treatment of escape onto a platform could support new learning, all animals were taken to a second (‘downstairs') watermaze and given two PTs ( Figure 6 ). This was a novel environment, and, therefore, there was no reason to expect the animals to perform at better than chance levels in the first of these PTs in a novel environment (PT n 1). However, escape from the water at the end of this PT might be sufficient to support new one-trial learning. Such learning was absent in the partial hippocampal lesion group ( F < 1). The sham lesion group, in contrast, did learn ( F [1, 21] = 4.51, p < 0.05), performing significantly better than the lesioned groups on PT n 2 (post hoc Ryan–Einot–Gabriel–Welsch range test, p < 0.005). The complete lesion group again showed no evidence of learning in a new environment ( F < 1). Figure 6 Novel Context Learning Percentage time spent in the target quadrant containing the escape platform during one-trial new learning in a different pool. * p < 0.05; n.s. = nonsignificant; comparison of percentage time spent in training zone during PT n 1 and PT n 2; paired-sample t-tests. New learning was observed only in sham-lesioned rats. Discussion The key finding of this study is that rats with partial lesions of the hippocampus can be reminded of a preoperatively learned escape location in a watermaze by both correct and potentially misleading information. Whereas sham-lesioned rats showed new one-trial learning towards or away from the originally trained quadrant as a function of the type of reminder treatment to which they were exposed, partially lesioned animals were unable to learn. Instead, the first PT served only as a reminder of the original platform location irrespective of where in the pool the platform was raised at the end of this trial. Rats with complete hippocampal lesions showed neither new learning nor reminding. There is an extensive classic literature on the nature and effectiveness of reminder treatments ( Riccio and Richardson 1984 ). Exposure to the training context, noncontingent stimuli, or additional training trials are just some examples of methods successfully used to remind animals of a prior training experience ( Zinkin and Miller 1967 ; Miller and Springer 1973 ; Mactutus et al. 1979 ; Gisquet Verrier and Schenk 1994 ; Przybyslawski and Sara 1997 ). Controversy did, however, surround studies that interpreted memory following a reminder treatment as evidence that the original amnesia was the result of a retrieval deficit ( Zinkin and Miller 1967 ; Miller and Springer 1973 ). It was argued that a reminding trial simply strengthens a weak memory that is behaviourally unobservable, similar to what happens during initial learning ( Cherkin 1972 ; Gold et al. 1973 ; Haycock et al. 1973 ; Gold and King 1974 ), or that, when amnesia is complete, it results in one-trial learning or response generalization. However, manipulations that are unlikely to produce new learning can also serve as effective reminders. Examples include pharmacological manipulations of the internal state ( Mactutus et al. 1980 ; Concannon and Carr 1982 ) and reexposure to the amnestic agent prior to retention testing ( Thompson and Neely 1970 ; Hinderliter et al. 1975 ). In many such studies, however, the use of inhibitory avoidance as a memory test makes it difficult to determine the cognitive ‘content' (cf. Riccio and Richardson 1984 ) of the behaviour expressed during retention testing. Although memory reactivation may have occurred when a rat inhibits movement that previously led to electric shock, an alternative interpretation is that a generalized fear state has been induced. The issue of whether and when amnesia reflects a storage or retrieval deficit was, thus, left unresolved. Two features are distinctive about our study. First, unlike in many previous studies, the reactivated memory involves the recall and expression of highly specific information—a discriminable position in space, and not just a faster escape latency, or greater freezing. Second, despite exposure to a novel platform location leading to reversal learning in the sham lesion group, the partial lesion group displayed only reminding of the original platform location. This distinction is important because, with the current revival of interest in memory retrieval, our protocol circumvents the ambiguities involved in the use of relearning as an index of retention. One example of a study that used a reacquisition rather than a true reminding protocol ( Land et al. 2000 ) revealed that a reminder prior to retention testing could alleviate amnesia in animals with hippocampal lesions. However, it is difficult to distinguish between ‘pure' reminding and the facilitation of new learning using reacquisition alone. Nonetheless, the watermaze task is deceptively complex, and successful performance depends on the operation of several distinct memory systems ( Bannerman et al. 1995 ; Whishaw and Jarrard 1996 ; Warburton and Aggleton 1999 ; Eichenbaum 2000 ; White and McDonald 2002 ). Accordingly, while no new learning of the platform location occurs in the partial and complete lesion groups, some ‘procedural' learning may take place during PT1; this may enhance a weak, subthreshold spatial memory to a point at which it can be expressed in PT2. However, for this argument to be plausible, one would expect there to be minimal retention of the procedural components in PT1. This was clearly not the case, as rats with both partial and complete hippocampal lesions did not behave like naïve animals during PT1. They searched at an appropriate distance from the pool walls and readily climbed onto the escape platform when it was eventually made available. Procedural learning is also generally well retained over time and, being slow, unlikely to change much in one trial. We also doubt that the recovery of memory on PT2 reflects the emergence of latent memory mediated solely by an extrahippocampal structure, but not expressed during PT1. For example, rats with complete hippocampal lesions have been shown to learn a spatial conditioned-cued preference mediated by the amygdala ( White and McDonald 1993 ), a form of memory that is partially masked by hippocampus-dependent learning in normal rats ( McDonald and White 1995 ). However, seeing reminding in partial but not complete hippocampus-lesioned animals argues against this possibility in this case. Finally, the recovery of a simple stimulus–response strategy based on approaching single cues is unlikely, as novel start locations were always used during retention testing (cf. Eichenbaum et al. 1990 ; see Materials and Methods ). Under these circumstances, it is reasonable to interpret the apparently complete amnesia observed in PT1 as, at least in part, a failure of spatial memory retrieval. Our use of partial hippocampal lesioning introduces several other issues. First, it is a technique that is arguably more relevant to human amnesia, in which damage to a structure is typically incomplete. Second, it is also relevant to the many studies in which a pharmacological intervention is applied at a single site within a brain region—microinfusion into the dorsal hippocampus, for instance, is likely to have minimal effects on ventral hippocampal tissue (see Steele and Morris 1999 ). Third, and perhaps most interesting, is the question of where memory traces are located. Given that reminding only occurs in partially lesioned rats, it is reasonable to suppose that spatial memory traces are either located (and reactivated) within the hippocampus, or that the hippocampus is required for the process of reactivation or expression of a reactivated memory stored elsewhere. According to the latter hypothesis, spatial memory traces might be stored in cortex but require fast synaptic transmission in the hippocampus to be retrieved (cf. Teyler and DiScenna 1986 )—at least during the period after training and before the completion of systems-level consolidation. Alternatively, some hippocampal tissue might be required for cortically expressed memory to gain access to striatal motor planning and executive systems. Findings reported by Virley et al. (1999) suggest that this retrieval hypothesis might not be implausible. In this study, monkeys with CA1 pyramidal cell lesions were amnesic for a preoperatively acquired visuospatial discrimination. Subsequent grafting of CA1 pyramidal cells resulted in the recovery of memory for a second preoperatively acquired discrimination. As the grafted tissue cannot contain specific memory traces, the implication is that the recovery of some aspects of CA1 cellular function is sufficient for the information processing mediating the retrieval of memories stored elsewhere. In raising many more questions than they answer, the present findings open a potential avenue of research into the neural dynamics of memory reactivation and retrieval. Specific interventions such as local AMPA receptor blockade (cf. Riedel et al. 1999 ) might be directed at the hippocampus or cortex during PT1 or PT2. Such a study could provide information about the role of these structures—and their network interactions—in the reactivation of apparently lost memories, and in their subsequent retrieval. For example, hippocampal neural activity may be necessary for effective retrieval, but perhaps not for the reminding-induced reactivation of memory, even for an ostensibly hippocampus-dependent task (cf. Land et al. 2000 ). Similarly, the necessity for hippocampal neural activity during retrieval might vary as a function of time after memory consolidation. In addition, the determinants of the reminder phenomenon itself remain unclear. It would be useful to establish whether reinforcement in the form of an escape platform is, in fact, necessary during PT1, or indeed whether a reminder trial in a separate pool would have been effective. Experiments involving partial versus complete sets of cues might also provide valuable insights into the reminding process (cf. O'Keefe and Conway 1978 ). These and related analyses will be the subject of future studies. Dissociating the storage and retrieval functions of the hippocampus in memory is central to our understanding of the role of hippocampo–cortical connections. Many theories of hippocampal function are based on the idea that the hippocampus acts as a mediating link between different cortical regions during the interval before systems consolidation is complete ( Teyler and DiScenna 1986 ; Squire and Alvarez 1995 ). Paradoxically, the same features that point to the alternative possibility—that the hippocampal formation is a site of encoding and long-term storage of complex multimodal memories within its distributed intrinsic circuitry ( Moscovitch and Nadel 1998 )—also place this group of structures in an ideal position to help reactivate memories from traces distributed over several cortical structures, perhaps via a mechanism such as pattern completion (see Marr 1971 ; Nakazawa et al. 2002 ). It is possible that, when the hippocampus is partially damaged and the cortico–hippocampal network is therefore degraded, retrieval is only possible once a more complete recreation of the training situation, possibly including reexposure to a platform, is provided. Although comparisons across different species and forms of memory should be viewed with caution, this scenario is reminiscent of Tulving's encoding specificity principle ( Tulving and Pearlstone 1966 ; Thomson and Tulving 1970 ) in that exposure to similar cues during encoding and retrieval phases permits the recovery of the original memory, despite the provision of incorrect information about the target location itself. Paradoxically, the poor learning abilities of partially lesioned rats might explain why a trial ending with exposure to a novel spatial location can serve as a reminder for the original location—by limiting new learning of the new location, a reactivated memory for the old location is unmasked. Materials and Methods Subjects We used a total of 73 male Lister Hooded rats obtained from a commercial supplier (Charles River Laboratories, United Kingdom). They were pair-housed in plastic cages with sawdust bedding and ad libitum access to food and water. Their care and maintenance and all experimental procedures were carried out in accordance with United Kingdom Home Office Regulations. Behavioural testing was conducted using two separate circular pools, 2.0 m in diameter and 60 cm high, each located in well-lit rooms with numerous distal visual cues. One pool was used for training and retention (‘upstairs') and the other for new context learning (‘downstairs'). The pools were filled with water at 25 °C ± 1 °C made opaque by the addition of 200 ml of latex liquid (Cementone-Beaver, Buckingham, United Kingdom). We used the ‘Atlantis platform' ( Spooner et al. 1994 ), a polystyrene platform that becomes available by rising from the bottom of the pool only if the animals swim to and stay within a specified ‘dwell radius' centred on the correct location for a predetermined ‘dwell time.' When risen, the top of the platform remained 1.5 cm below the water surface. The animals' swimming was monitored by an overhead video camera connected to a video recorder and an online data acquisition system (Watermaze, Watermaze Software, Edinburgh, United Kingdom; Spooner et al. 1994 ) located in an adjacent room. This system digitizes the path taken by an animal and computes various parameters such as escape latency, time spent in a zone overlying the platform, and other conventional measures of watermaze performance. Training protocol Testing was carried out according to the schedule illustrated in Figure 1 . Cued pretraining This phase consisted of a single day of nonspatial cued training in the ‘upstairs' watermaze (curtains drawn around the pool to occlude extramaze cues, with ten trials in two sessions of five trials each (intertrial interval ≈ 20 min; intersession interval ≈ 3 h). The visible cue was suspended approximately 25 cm above the platform, which was moved every two trials to one of four possible locations, according to a pseudorandom schedule; the dwell radius was set at 20 cm, and the dwell time was 1 s. Training Training on a spatial reference memory task began 3 d later in the same watermaze. Rats received ten trials/day, in two sessions of five consecutive trials each (intersession interval ≈ 2 h), for 4 d. The dwell time was set to 0.5 s throughout training, but the dwell radius was gradually reduced over days (day 1: 20 cm; day 2: 15 cm; days 3 and 4: 13 cm). This schedule was intended to promote accurate and focused searching, but without generating the highly perseverative strategy that typically results from the use of long dwell times ( Riedel et al. 1999 ). Rats were given a maximum of 120 s to find an escape platform located at the centre of either the NE or SW quadrant, after which they remained on the platform for 30 s On the rare trials in which a rat failed to escape within 2 min, the experimenter placed a hand above the correct location in order to guide the animal to the platform. For each animal, the platform position remained constant throughout training, but start locations (N, S, E, or W) were varied pseudorandomly across trials. Only those animals achieving the acquisition criterion of mean escape latencies of 15 s or less on day 4 of training proceeded to the next phase of testing. Surgery Surgery took place 1–2 d after the end of training. Rats were given either partial or complete bilateral neurotoxic lesions of the hippocampal formation (DG and CA fields), or sham surgery. Complete lesions were intended to remove 85% or more of the total hippocampal volume. Partial lesions targeted the temporal two-thirds of the hippocampus, sparing the septal (dorsal) third of the structure. The rats were assigned to groups of equivalent mean performance on the basis of their escape latencies during the final day of training. Lesions were made with ibotenic acid (Biosearch Technologies, Novato, California, United States; dissolved in 0.1 M phosphate-buffered saline [pH 7.4] at 10 mg/ml) following the protocol of Jarrard (1989) . The animals were anaesthetized with an intraperitoneal injection of tribromoethanol (avertin) and placed in a Kopf Instruments (Tujunga, California, United States) stereotaxic frame such that Bregma and Lambda lay on the same horizontal plane. Rats received nine or 13 injections of ibotenic acid (partial and complete lesion groups, respectively; 0.05 μ1, 0.08 μ1, or 0.1 μ1 per injection) at different rostrocaudal and dorsoventral levels via an SGE syringe secured to the stereotaxic frame (see de Hoz et al. 2003 ). The injection rate was 0.1 μ1/min, and the needle was removed very slowly 90 s after the injection. A total of 0.65 μ1 or 0.91μ1 per hemisphere was necessary for the partial and complete lesions, respectively. The coordinates were modified from Jarrard (1989) to suit the slightly different brain size of Lister Hooded rats and to achieve the desired amount of partial hippocampal damage (see de Hoz et al. 2003 ). Sham lesions were made in the same way, with the injections replaced by a piercing of the dura (intended to cause comparable neocortical damage). Retention testing This phase began 14 d after the end of training. It consisted of two PTs (PT1and PT2) spaced 1 h apart, with a reminder treatment occurring at the end of PT1. Each PT (PT1 and PT2) began with a standard 60-s swim with the platform unavailable. In each PT, the rats were placed into the pool in either the adjacent right or the adjacent left quadrants with respect to the training quadrant. Start positions were counterbalanced across PTs and across rats. At the end of the 60 s the platform was raised and the animals were allowed to find and climb onto it. The rats were allowed a further 60 s to locate the platform once risen (but still hidden just below the water surface); if unsuccessful within this period, they were guided to the platform. They then remained on the platform for 30 s. The raising of the platform at the end of PT1 constituted the reminder treatment; thus PT1 is sometimes referred to as the ‘reminder trial.' A key variable was that the platform was raised in either the original training location (half the animals) or in a novel location in the centre of the opposite quadrant of the pool (the other half). Note that reminding using the original location always occurred in the training quadrant, and reminding using the novel location always occurred in the opposite quadrant. However, whereas the terms ‘training' and ‘opposite' are used to refer to physical areas of the pool, ‘novel' and ‘original' refer also to separate groups that received each type of reminder. For analysis of the different behavioural phases, several measures of performance were assessed, including escape latency, swim speed, and time spent within defined regions of the pool. Memory retention during PTs is inferred from the time spent in each quadrant of the pool as a percentage of the 60-s duration of the PT. A more sensitive measure can be obtained by analysing percentage time spent within a specified radius (zone) centred on the platform location ( Moser and Moser 1998 ). When time in zone is presented, it is expressed as a percentage of the total time spent in both the original training zone and the novel opposite zone. Statistical analysis (SPSS, Chicago, Illinois, United States) began with an ANOVA followed by appropriate post hoc comparisons. Numerical data are reported as mean ± standard error (s.e.m.) throughout. Novel context learning New learning was assessed the next day in a separate ‘downstairs' watermaze that constituted a novel context. The protocol was identical to that used during ‘upstairs' retention testing, i.e., two rewarded PTs (PT n 1 and PT n 2) spaced 1 h apart. Lesion analysis At the end of behavioural testing, rats were perfused intracardially with saline followed by 10% formalin under terminal pentobarbitone anaesthesia (Euthatal, 1 ml). Their brains were removed and stored in 10% formalin for 24 h before being blocked and embedded in egg yolk. The embedding procedure is described in de Hoz et al. (2003) . Coronal, 30-μm sections through the hippocampus and other structures were cut using a cryostat: every fifth section was recovered, mounted on a slide, and stained with cresyl violet (see Figure 3 A– 3 C). The relative volume of spared tissue was calculated by measuring the area of hippocampus spared in each section of a particular brain according to the following protocol: Each coronal section containing hippocampus was placed under a photomacroscope (Wild, Heerbrugg, Switzerland), and the image taken by a mounted video camera was imported into NIH Image 1.63 (National Institutes of Health, Bethesda, Maryland, United States). The area of spared hippocampal tissue in each section was then outlined and automatically calculated. Surrounding fibres such as the fimbria were excluded on the grounds that they would not be considered in a section were all the hippocampal cells dead. The sections were spaced 150 μm apart, yielding up to 32 sections in a sham lesion animal, and fewer in animals with acceptable partial lesions. For each rat, the total hippocampal ‘volume' was calculated by adding the area of hippocampal tissue spared in each successive section. The proportion of hippocampus spared for each lesioned animal was expressed as a percentage of the mean hippocampal ‘volume' for sham-lesioned animals. Values for the left and right hippocampi were initially calculated separately and then averaged (see Figure 3 D). Strict criteria for acceptance of a lesion were used. The lesion had to be confined to the hippocampus in all cases, and leave intact tissue volumes of 25%–50% in the septal hippocampus with minimal sparing (less than 10%) elsewhere in the structure in the case of partial lesions, or less than 15% total hippocampal sparing in the case of complete lesions. Animals with minimal subicular damage, typically located at medial levels of the structure, were accepted.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509297.xml
539356	Seroprevalence of hepatitis C and associated risk factors among an urban population in Haiti	Background The seroprevalence of hepatitis C varies substantially between countries and geographic regions. A better understanding of the seroprevalence of this disease, and the risk factors associated with seropositive status, supply data for the development of screening programs and provide insight into the transmission of the disease. The purpose of this investigation was to determine the seroprevalence of hepatitis C and associated risk factors in an urban population in Haiti. Methods A prospective survey for hepatitis C antibodies was conducted among an urban outpatient population in Cap-Haïtien, Haiti, with a sample size of 500 subjects. An anonymous 12 question survey, with inquiries related to demographic characteristics and risk factors for HCV acquisition, was concomitantly administered with testing. These demographic and behavioral risk factors were correlated with HCV antibody status using univariate and multivariate tests. Results The prevalence of positive HCV antibody was 22/500 (4.4%). Subjects that were anti-HCV positive had an average of 7 ± 8.6 lifetime sexual partners, compared to average of 2.5 ± 3.5 lifetime sexual partners among HCV-negative subjects (p = 0.02). In a multiple logistic regression model, intravenous drug use (OR 3.7, 1.52–9.03 95% CI) and number of sexual partners (OR 1.1, 1.04–1.20 95% CI) were independently associated with a positive HCV antibody result. Conclusions A substantial number of subjects with HCV antibodies were detected in this population in Haiti. Further investigation into the correlation between the number of sexual partners and testing positive for hepatitis C antibodies is indicated.	Background The seroprevalence of hepatitis C virus (HCV) varies substantially in different geographic regions throughout the world [ 1 ]. Prior studies have suggested a low prevalence of HCV antibodies among a sample of patients in rural Haiti [ 2 ]. No cases of positive HCV antibody were detected among 485 patients in a sexually transmitted infections clinic in Jamaica [ 3 ], but 41% of hemophiliacs in Jamaica were HCV antibody positive [ 4 ]. Our hypothesis was that a higher prevalence of HCV antibodies would be detected in an urban population in Haiti. Risk factors associated with HCV serologic status may be specific to a country or region. In particular, the role of sexual contact in the transmission of HCV appears to be influenced by characteristics and location of the population studied [ 5 ]. Therefore, examination of the risk factors associated with the presence of HCV antibodies in this population can be utilized to guide screening procedures as well as provide insight into the transmission of HCV in the context of Haitian society. A better understanding of the transmission of HCV could enhance the effectiveness of prevention efforts. This research study provides analysis of risk factors associated with hepatitis C from a population that had not been previously studied. We have observed a seroprevalence rate of 4.4% (22/500) of HCV antibodies, with intravenous drug usage and the number of sexual partners being associated with positive HCV antibodies. Methods The study was approved by the Institutional Review Board at Brooke Army Medical Center (Fort Sam Houston, Texas). Subjects were recruited in the year 2000 from a healthy population utilizing hospital and clinic services in Cap-Haïtien, Haiti, which is the second largest city in the country. Subjects were recruited on presentation to the hospital laboratory for blood draws for routine laboratory tests. This laboratory was the only location in which subjects were recruited and samples obtained. Subject participation was not limited to a particular medical condition. The first 500 subjects that agreed to participate in the study were enrolled. These subjects were presenting to the hospital for services, and were not patients in a specific clinic. After informed consent was obtained, subjects completed a written 12 question survey in Creole, at the same location as where the blood was drawn. The subjects completed the survey by themselves. The questionnaire focused on demographic information and topic areas possibly associated with transmission of hepatitis C (intravenous drug use, intranasal cocaine use, blood transfusions, sexual history, number of tattoos). Intravenous drug use and cocaine use was measured on a 0–3 scale (never-rare-frequent-daily). Blood transfusions, number of sexual partners, age of first sexual intercourse and number of tattoos were quantified. Serum was obtained for testing for HCV antibody, utilizing the Abbott HCV EIA 3.0 kit™ (Abbott Laboratories, Abbott Park, Illinois). The survey information and serum results were identified only by a subject identification number, with all other identifying information removed. Data were analyzed using univariate correlations with a Pearson's correlation coefficient. The number of sexual partners was compared between HCV-positive and negative subjects using an independent-sample t-test. Intravenous drug use between HCV-seropositive and seronegative subjects was compared using a Fisher's exact test. A multivariate logistic regression model was employed with stepwise backward elimination of non-significant variables, with HCV antibody status as the dependent variable. All of the variables from the survey were included in the model. Results A total of 500 subjects were recruited and had serum tested for HCV antibody. Only two of these subjects did not complete the survey. Most subjects who were informed of the study agreed to participate, but an exact number of subjects who refused participation in the study was not determined. The background characteristics of the patient population are displayed in Table 1 . Of note, few subjects (12/496) admitted intravenous drug use, and only one subject noted having a tattoo. Table 1 Characteristics and survey responses of recruited subjects Characteristic Frequency Sex- % males 332/498 (67%) Age (mean ± SD in years) 33.7 ± 15.9 Years of Education (mean ± SD) 7.8 ± 5.7 Marital Status- % married 152/495 (31%) Age of First Sexual Intercourse (mean ± SD in years) 19.1 ± 5.7 Number of Lifetime Sexual Partners (mean ± SD) 2.7 ± 4.0 Intravenous Drug Use- % with any history of usage 12/498 (2%) Intranasal Cocaine Use- % with any history of usage 12/496 (2%) Subjects with tattoos 1/498 The prevalence of positive HCV antibodies was 22/500 (4.4%, 95% CI 2.6–6.2%). Subjects that were anti-HCV positive had an average of 7 ± 8.6 lifetime sexual partners, compared to average of 2.5 ± 3.5 lifetime sexual partners among HCV-negative subjects (p = 0.02). There were no other statistically significant differences between the HCV antibody-positive and HCV antibody-negative groups in terms of demographic characteristics or topic areas associated with HCV infection/transmission. Among subjects with HCV antibodies, 4/22 subjects admitted intravenous drug use compared to 8/476 among HCV-antibody negative subjects (p < 0.001). A similar result was observed for rates of intranasal cocaine use among HCV-antibody positive (4/22) vs. HCV-antibody negative (8/474) subjects (p < 0.001). Of the 22 subjects with positive HCV antibodies, 12 subjects denied intravenous drug use, had no tattoos, and had either 1 or 2 lifetime sexual partners. Five subjects with positive HCV antibodies denied intravenous drug use and had >10 lifetime sexual partners. Intravenous drug use (r = 0.26, p < 0.001), intranasal cocaine use (r = 0.29, p < 0.001) and the number of lifetime sexual partners (r = 0.24, p < 0.001) were the only three variables with a statistically significant correlation with the presence of HCV antibodies on univariate analysis. The number of lifetime sexual partners had some correlation with intravenous drug use (r = 0.12, p = 0.009) and intranasal cocaine use (r = 0.13, p = 0.004). There was a very close correlation between intravenous drug use and intranasal cocaine use (r = 0.99, p < 0.001). Therefore, intranasal cocaine use was not incorporated into the multivariate model due to concerns about co-linearity. In the multiple logistic regression model with backwards elimination, intravenous drug use (OR 3.7, 1.52–9.03 95% CI) and number of sexual partners (OR 1.1, 1.04–1.20 95% CI) were independently associated with a positive HCV antibody result. Discussion We observed a prevalence of 4.4% of HCV antibodies in an urban population in Haiti. Additionally, we found that HCV antibodies were associated with intravenous drug use as would be expected, but also with increasing number of lifetime sexual partners. It is conceivable that subjects who reported higher number of sexual partners were more likely not to admit past intravenous drug usage. However, it is also possible that this finding represents sexual transmission of hepatitis C among subjects with multiple prior sexual partners. Finally, we observed HCV antibodies were present is some subjects who denied intravenous drug use and had less than 3 lifetime sexual partners. A recent survey of the influences on HIV preventive behaviors among youth in Haiti observed that 80% of males and 42% of females self-disclosed sexual activity [ 6 ]. This survey noted a mean age of first intercourse of 13.1 years, a number lower than our observations (19.1 years). This study also noted that condom use was infrequent in the surveyed population (18% of subjects reported always or sometimes using a condom), while 43% reported 3 or more lifetime sexual partners. The applicability of these data to our study population is limited because the age difference between this population and the population of our study. Although the results of this investigation do not directly describe the population of our study, we might extrapolate that our population would be unlikely to have a high rate of condom usage. Only a single prior study in Haiti suggested that HCV antibodies were rare [ 2 ]. The most obvious difference between our results and this prior study was that the previous study was performed on rural subjects, while our data was collected in an urban setting. It is possible that intravenous drug abuse would be more prevalent in an urban environment, which could partially explain the observed differences in seroprevalence. The previous study was also conducted more than 10 years before our research. Intravenous drug use does not exclusively explain the prevalence of HCV antibodies in the studied population. Although intravenous drug use was associated with an increased likelihood of having HCV antibodies (as has been thoroughly documented [ 1 ]), only four of the 22 subjects with HCV antibodies admitted drug use. Additionally, none of the HCV positive subjects had tattoos. These findings suggest either that self-reported drug use underestimates the prevalence of drug use in the population, or another mode of transmission of HCV has occurred. The degree to which sexual transmission of HCV occurs is exceedingly controversial [ 5 , 7 ]. Studies in monogamous relationships suggest that sexual transmission of hepatitis C occurs very rarely [ 8 ]. Seroprevalence studies from sexually-transmitted diseases clinics describe a variable amount of HCV positive subjects, which tends to be low when injection drug users are excluded [ 9 , 10 ]. The number of sexual partners has been previously associated with increasing risk of HCV exposure [ 11 ]. We also observed the association between increasing number of sexual partners and the likelihood of having HCV antibodies. This association was observed even when controlling for other variables in a multivariate model. Our data suggest that sexual transmission of hepatitis C may occur more frequently in persons with multiple sexual partners. However, additional larger studies directed at evaluating HCV-infected persons with multiple sexual partners are needed. All studies on the sexual transmission of hepatitis C (including our study) are limited by the potential of the confounding variable of shared toothbrushes, razor blades and other items among sexual partners. Our findings are limited by the lack of information regarding active HCV infection. The presence of HCV antibodies only indicates prior exposure, and definitive documentation of active HCV infection requires detection of virus in the bloodstream utilizing HCV RNA polymerase chain reaction testing. However, since most patients exposed to hepatitis C develop chronic infection [ 12 ], HCV antibody testing provides a reasonable estimate of the amount of HCV infection in a population. Other limitations of the study are linked to the difficulties inherent in self-reporting of behaviors such as sexual activity and drug use. Additionally, we were unable to obtain information on subjects that refused participation in the study, which may limit the representativeness of this population. It is also possible that the patients receiving blood draws in the Cap-Haïtien health clinic may not be representative of the population of Haiti, or even the urban population of Haiti. Finally, there are other possible alternative sources of percutaneous exposure to HCV, such as medical injections by alternative practitioners or other medical, surgical or dental procedures. In conclusion, we provide seroprevalence data of HCV antibodies in an urban population in Haiti. These data are useful for understanding the risks of transfusion in Haiti if the blood has not been previously screened for HCV. This information contributes to our understanding of the worldwide prevalence of hepatitis C, which allows for informed decisions regarding the priorities of funding for the treatment and prevention of this infection. Additionally, we observed the number of sexual partners may be related to a greater likelihood of having HCV antibodies, but some subjects who denied intravenous drug use and had few lifetime sexual partners still had HCV antibodies. These findings could be utilized to foster consideration of new studies into some of the risk factors that are not clearly understood (such as procedures by medical, dental or alternate practitioners). These results suggest that further study into the mode of transmission of hepatitis C should focus on patients with a high number of lifetime sexual partners but no evidence of intravenous drug use. Competing interests Matthew J. Hepburn, MD: no competing interests to declare. Eric J. Lawitz, MD: Dr. Lawitz has received research grants to conduct investigator-initiated research from Schering-Plough Corporation (Kenilworth, New Jersey). Author's contributions MH was involved in study design, data analysis, and manuscript preparation. EL was involved in study design, data collection, data analysis, and manuscript preparation. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539356.xml
544397	Dexamethasone protected human glioblastoma U87MG cells from temozolomide induced apoptosis by maintaining Bax:Bcl-2 ratio and preventing proteolytic activities	Background Glioblastoma is the deadliest and most prevalent brain tumor. Dexamethasone (DXM) is a commonly used steroid for treating glioblastoma patients for alleviation of vasogenic edema and pain prior to treatment with chemotherapeutic drugs. Temozolomide (TMZ), an alkylating agent, has recently been introduced in clinical trials for treating glioblastoma. Here, we evaluated the modulatory effect of DXM on TMZ induced apoptosis in human glioblastoma U87MG cells. Results Freshly grown cells were treated with different doses of DXM or TMZ for 6 h followed by incubation in a drug-free medium for 48 h. Wright staining and ApopTag assay showed no apoptosis in cells treated with 40 μM DXM but considerable amounts of apoptosis in cells treated with 100 μM TMZ. Apoptosis in TMZ treated cells was associated with an increase in intracellular free [Ca 2+ ], as determined by fura-2 assay. Western blot analyses showed alternations in the levels of Bax (pro-apoptotic) and Bcl-2 (anti-apoptotic) proteins resulting in increased Bax:Bcl-2 ratio in TMZ treated cells. Western blot analyses also detected overexpression of calpain and caspase-3, which cleaved 270 kD α-spectrin at specific sites for generation of 145 and 120 kD spectrin break down products (SBDPs), respectively. However, 1-h pretreatment of cells with 40 μM DXM dramatically decreased TMZ induced apoptosis, decreasing Bax:Bcl-2 ratio and SBDPs. Conclusion Our results revealed an antagonistic effect of DXM on TMZ induced apoptosis in human glioblastoma U87MG cells, implying that treatment of glioblastoma patients with DXM prior to chemotherapy with TMZ might result in an undesirable clinical outcome.	Background Glioblastoma patients usually receive steroids for alleviation of vasogenic edema and pain prior to treatment with chemotherapeutic drugs. Steroids, however, may modulate the sensitivity of tumor cells to chemotherapeutic drugs. Dexamethasone (DXM), a synthetic glucocorticoid, is commonly used to reduce inflammation and pain associated with glioblastoma [ 1 ]. However, DXM has been reported to make human glioblastoma cells resistant to ionizing radiation and chemotherapeutic agents that otherwise cause DNA damage [ 2 - 5 ]. Execution of cells by apoptosis usually requires the activation of cysteine proteases such as calpains and caspases [ 6 ]. Diverse stimuli may cause an increase in intracellular free [Ca 2+ ], which is absolutely required for activation of calpain [ 7 ]. Activation of caspases may occur via different mechanisms [ 8 , 9 ]. Mitochondria mediated pathway of apoptosis may be activated in course of cell death. This involves the regulation of apoptosis by the Bcl-2 family proteins via controlling the release of cytochrome c from mitochondria [ 10 , 11 ], and subsequent formation of the cytosolic 'apoptosome' complex [ 12 , 13 ], which ultimately activates caspase-3 for execution of cells. Thus, the members of the Bcl-2 family modulate the mitochondrial pathway of apoptosis [ 14 ]. The pro-apoptotic (e.g., Bax, Bcl-xS) and anti-apoptotic (e.g., Bcl-2, Bcl-xL) members of this family, respectively, promote and inhibit the translocation of cytochrome c from mitochondria to cytosol [ 15 ]. Glucocorticoids are steroid hormones, which are secreted in response to stress and can modulate the ability of cells to undergo apoptosis [ 16 ]. For example, glucocorticoids induce apoptosis in thymocytes [ 17 ] and also increase the sensitivity of hippocampal neurons to cell death [ 18 ]. In contrast, DXM has been reported to induce resistance to certain drugs in glioblastoma cell lines [ 3 - 5 ]. Although an association with p21 WAF1/CIP1 protein accumulation has been reported [ 19 ], the exact mechanism of DXM mediated protection of glioblastoma cells from apoptosis is still largely unclarified. Exposure of human astrocytoma D384 and rat glioblastoma C6 cells to staurosporine induced apoptosis but pretreatment of those cells with DXM caused reduction in staurosporine mediated apoptosis [ 20 ]. In addition, DXM also conferred protection against the induction of apoptosis by anti-cancer agents including camtothecin and etoposide [ 20 ]. It has also been shown that exposure of glioblastoma cells to glucocorticoids induces partial resistance to anti-cancer agents such as cisplatinum, methotrexate, vincristine, cytarabine, adriamycin, and teniposide [ 3 - 5 ]. DXM appears to interfere with p53-dependent pathways of drug toxicity since the glioblastoma cell lines (LN-229 and U87MG) with wild-type p53 status were protected from drug toxicity by DXM to a greater extent than the cell lines (LN-18, LN-308, and T98G) with mutant p53 [ 3 - 5 ]. It has been reported earlier that DXM mediated protection from cancer chemotherapy occurs via a p53-independent pathway of regulating p21 WAF1/CIP1 expression in glioblastoma cells but this effect appears to be cell-type specific [ 19 ]. Thus, there remains a concern of modulatory effects of DXM on the mechanism of action of any chemotherapeutic agent for treatment of glioblastoma. Therefore, we have initiated this investigation to examine the modulatory effect of DXM on temozolomide (TMZ) induced apoptosis of glioblastoma cells. In an in vitro model using the human glioblastoma U87MG cells, we have investigated whether DXM confers resistance to TMZ action via inhibition of apoptosis. TMZ is an alkylating chemotherapeutic drug that readily crosses the blood-brain-barrier in glioblastoma patients [ 21 ]. It is chemically related to decarbazine and is the 3-methyl derivative of the experimental anti-cancer drug mitozolomide. It has shown anti-tumor activity and relatively low toxicity in Phase I and Phase II clinical trials in patients with various advanced cancers, including malignant glioblastomas [ 21 ]. TMZ is spontaneously hydrolyzed under physiological conditions to its active metabolite 5-(3-methyltriazen-1-yl) imidazole-4-carboxamide (MTIC) [ 22 ]. The mechanism of action of MTIC is proposed to be methylation of DNA at the O 6 position of guanine, with an additional methylation at its N 7 position [ 23 , 24 ]. However, O 6 -methylguanine (O 6 -meG) may be removed by O 6 -methylguanine methyl transferase (MGMT) [ 25 ]. Cells deficient in MGMT do not repair O 6 -meG. Replication of DNA introduces a T instead of C opposite to O 6 -meG, resulting in GT mismatches [ 26 ]. Activation of mismatch repair system (MMRS) may remove T during DNA repair synthesis. However, ineffective MMRS causes growth arrest and eventually apoptotic death [ 27 ]. These studies have helped define the action of TMZ in cancer cells of mostly hemopoietic origin. However, the action of TMZ in glioblastoma cells remains largely undefined. Glioblastomas are relatively resistant to anti-cancer agents that cause apoptosis via DNA alkylation. Therefore, we have investigated the mechanism of TMZ induced apoptosis in human glioblastoma U87MG cells. We found that TMZ caused apoptosis in U87MG cells as detected by morphological and biochemical assays. Alterations in the levels of pro-apoptotic Bax and anti-apoptotic Bcl-2 proteins are known to regulate the commitment to apoptosis [ 14 , 15 ]. Therefore, we investigated the levels of these apoptosis regulatory proteins following treatment of U87MG cells with TMZ. The initiation of apoptosis in U87MG cells following exposure to TMZ requires activation of calpain, a Ca 2+ -dependent cysteine protease, which plays a role in the mechanism of cell death in human malignant brain tumors including glioblastoma [ 28 ]. Besides, caspase-3 activity was also increased in TMZ induced apoptosis in U87MG cells. Pretreatment of U87MG cells with DXM blocked TMZ induced apoptosis, indicating that DXM worked as an antagonistic agent in TMZ induced apoptosis in human glioblastoma cells. The knowledge gained from our investigation implies that the combination of DXM and TMZ for the treatment of human glioblastoma patient may result in an undesirable clinical outcome. Preliminary results of this investigation have previously been presented [ 29 ]. Results Evaluation of viability and apoptotic death both morphologically and biochemically Exclusion of trypan blue dye by viable U87MG cells was evaluated under a light microscope using a hemocytometer after all treatments. A pretreatment with DXM prevented decrease in cell viability (panel A, Fig. 1 ). Morphological features of apoptosis were detected following Wright staining (panel B, Fig. 1 ) and counted to determine the amount of apoptotic cell death (panel C, Fig. 1 ) based on characteristic morphological features such as condensation of the nucleus and cytoplasm, cytoplasmic blebbing, and the formation of apoptotic bodies. All treatment groups were examined under the light microscopy and cells were counted to determine the percentage of apoptotic cells (panel C, Fig. 1 ). Compared to control (CTL) cells, cells treated with 100 μM TMZ showed an increase in the percentage of apoptotic cells ( P < 0.001). A pretreatment of cells with 40 μM DXM decreased TMZ induced apoptosis by three-fold, compared to treatment of cells with TMZ only. Figure 1 Determination of apoptosis based on morphological features. Four treatment groups: control (CTL); 40 μM dexamethasone (DXM) for 8 h; 100 μM temozolomide (TMZ) for 6 h; pretreatment with 40 μM DXM for 2 h followed by 100 μM TMZ for 6 h. (A) DXM prevented TMZ mediated decrease in U87MG cell viability. The trypan blue exclusion assay was used to assess cell viability in U87MG cells. (B) Photomicrographs showing representative cells from each treatment group. The arrows indicate apoptotic cells. (C) Bar graphs indicating the percentage of apoptotic cells counted from each group. Significant difference between CTL and TMZ treated cells was indicated by * ( P ≤ 0.05) and significant difference between TMZ treated cells and DXM plus TMZ treated cells was indicated by # ( P ≤ 0.05). Results obtained from Wright staining were further supported by the ApopTag assay (panel A, Fig. 2 ). Both CTL and DXM treated cells showed little or no brown color, confirming almost absence of ApopTag positive cells or apoptosis. The percentage of ApopTag positive cells was calculated (panel B, Fig. 2 ) and found to be highly significant ( P < 0.001) in TMZ treated cells, compared to CTL cells. A pretreatment of cells with DXM considerably attenuated apoptotic DNA fragmentation in TMZ treated cells. Figure 2 ApopTag assay for detection and determination of DNA fragmentation in U87MG cells. Four treatment groups: control (CTL); 40 μM dexamethasone (DXM) for 8 h; 100 μM temozolomide (TMZ) for 6 h; pretreatment with 40 μM DXM for 2 h followed by 100 μM TMZ for 6 h. (A) The photomicrographs showing representative cells from each treatment group. The arrows indicate apoptotic cells. (B) Bar graphs indicating the average percentage of apoptotic cells counted from each group. Significant difference between CTL and TMZ treated cells was indicated by * ( P ≤ 0.05) and significant difference between TMZ treated cells and DXM plus TMZ treated cells was indicated by # ( P ≤ 0.05). Treatment with TMZ increased intracellular free [Ca 2+ ] Using fura-2 assay, intracellular free [Ca 2+ ] was determined in all treatment groups (Fig. 3 ). No significant difference ( P = 0.928) was seen between CTL cells and cells treated with DXM alone. Cells treated with TMZ showed a significant increase ( P = 0.007) in intracellular free [Ca 2+ ], compared to CTL cells. This increase was attenuated almost 100% by a pretreatment of the cells with DXM. There was no significant difference ( P = 0.999) between intracellular free [Ca 2+ ] in CTL cells and those treated with DXM plus TMZ. Figure 3 Bar graphs indicating percentage of increase of intracellular free [Ca 2+ ] using fura-2. These data were generated from U87MG cells grown in phenol red-free medium for 24 h prior to treatments with the drugs in freshly prepared phenol red-free medium. Four treatment groups: control (CTL); 40 μM dexamethasone (DXM) for 8 h; 100 μM temozolomide (TMZ) for 6 h; pretreatment with 40 μM DXM for 2 h followed by 100 μM TMZ for 6 h. Percent changes in intracellular free [Ca 2+ ] were shown at nM levels. Significant difference between CTL and TMZ treated cells was indicated by * ( P ≤ 0.05) and significant difference between TMZ treated cells and DXM plus TMZ treated cells was indicated by # ( P ≤ 0.05). TMZ induced apoptosis with an increase in Bax:Bcl-2 ratio A commitment to apoptosis was measured by examining any increase in the ratio of Bax (pro-apoptotic protein) expression to Bcl-2 (anti-apoptotic protein) expression. The bax gene encodes different isoforms. The antibody we used in this investigation could recognize 21 kD Baxα and 24 kD Baxβ bands (panel A, Fig. 4 ). Here, we considered both bands in our estimation of total Bax expression. We also examined the level of Bcl-2 expression in all treatment groups (panel B, Fig. 4 ). Almost same level of β-actin expression in each treatment ensured that equal amount of protein was loaded in each lane (panel C, Fig. 4 ). Based on the Western blot experiments (panels A and B, Fig. 4 ), the Bax:Bcl-2 ratios were measured in all treatment groups (panel D, Fig. 4 ). There was no significant difference ( P = 0.983) in Bax:Bcl-2 ratio between CTL and DXM treated cells (panel D, Fig. 4 ). Compared to CTL cells, a rise in Bax:Bcl-2 ratio (panel D, Fig. 4 ) in cells exposed to TMZ was influenced more by a change in Bax expression (panel A, Fig. 4 ) than a change in Bcl-2 expression (panel B, Fig. 4 ). Compared to CTL cells, cells treated with TMZ showed a significant increase ( P = 0.007) in the Bax:Bcl-2 ratio (panel D, Fig. 4 ). There was a significant difference ( P = 0.019) in Bax:Bcl-2 ratio between cells treated with TMZ alone and those treated with DXM plus TMZ, indicating a loss of commitment to apoptosis due to a pretreatment with DXM. There was no significant difference ( P = 0.871) in Bax:Bcl-2 ratio between CTL cells and cells pretreated with DXM and then treated with TMZ. Figure 4 The Bax:Bcl-2 ratio measured by Western blot analysis. Four treatment groups: control (CTL); 40 μM dexamethasone (DXM) for 8 h; 100 μM temozolomide (TMZ) for 6 h; pretreatment with 40 μM DXM for 2 h followed by 100 μM TMZ for 6 h. (A) A representative gel picture showing level of expression of Bax. (B) A representative gel picture showing level of expression of Bcl-2. (C) A representative gel picture showing level of expression of β-actin. (D) Densitometric analysis showing the Bax:Bcl-2 ratio in all treatment groups. Significant difference between CTL and TMZ treated cells was indicated by * ( P ≤ 0.05) and significant difference between TMZ treated cells and DXM plus TMZ treated cells was indicated by # ( P ≤ 0.05). Calpain and caspase-3 activities as determined by α-spectrin degradation Calpain and caspase-3 activities were assessed by Western blot analysis of the calpain-specific 145 kD SBDP and the caspase-3-specific 120 kD SBDP, respectively (panel A, Fig. 5 ). Level of β-actin expression, which was almost uniform in all treatments, was used as a loading control (panel B, Fig. 5 ). There was no significant difference ( P = 0.911) between CTL cells and DXM treated cells in generation of 145 kD SBDP, indicating similar levels of calpain activity in these two cases (panel C, Fig. 5 ). The generation of 145 kD SBDP in cells treated with TMZ was about 1.5-fold more intense ( P = 0.003) than CTL cells, indicating that the level of calpain activity was increased in cells due to treatment with TMZ. Cells pretreated with DXM and then treated with TMZ showed a significant decrease in the generation of 145 kD SBDP, indicating an inhibitory effect of DXM on TMZ mediated increase in calpain activity (panel C, Fig. 5 ). Figure 5 Determination of calpain and caspase-3 activities using Western blot analysis of α-spectrin breakdown products (SBDPs). Four treatment groups: control (CTL); 40 μM dexamethasone (DXM) for 8 h; 100 μM temozolomide (TMZ) for 6 h; pretreatment with 40 μM DXM for 2 h followed by 100 μM TMZ for 6 h. (A) A representative gel picture showing generation of 145 kD and 120 kD SBDPs. (B) A representative gel picture showing level of expression of β-actin. (C) Densitometric analysis showing percent changes in optical density of the calpain-specific 145 kD SBDP over CTL. (D) Densitometric analysis showing percent change of optical density of the caspase-3-specific 120 kD SBDP over CTL. Significant difference between CTL and TMZ treated cells was indicated by * ( P ≤ 0.05) and significant difference between TMZ treated cells and DXM plus TMZ treated cells was indicated by # ( P ≤ 0.05). Caspase-3 activity was also measured by Western blot analysis in the generation of caspase-3-specific 120 kD SBDP (panel D, Fig. 5 ). Compared to CTL cells, treatment of cells with DXM alone did not cause a significant change ( P = 0.983) in caspase-3 activity. Caspase-3 activity in cells treated with TMZ was almost 1.5 times more ( P = 0.001) than CTL cells (panel D, Fig. 5 ). Thus, pretreatment of cells with DXM prior to treatment with TMZ appeared to decrease the upregulation of caspase-3 activity. Furthermore, there was no significant difference ( P = 0.785) between CTL cells and cells that were pretreated with DXM and then treated with TMZ (panel D, Fig. 5 ). Caspase-3 activation as determined by generation of caspase-3-p20 fragment Caspase-3 activation was also measured by Western blot analysis of the production of active 20 kD caspase-3 fragment (panel A, Fig. 6 ). Again, almost uniform expression of β-actin in all treatments served as an internal standard and indicated equal amounts of protein loadings in all lanes (panel B, Fig. 6 ). The intensities of active 20 kD caspase-3 band were almost similar in CTL cells and cells treated with DXM alone (panel C, Fig. 6 ). Treatment of cells with DXM alone did not cause a significant change ( P = 0.613) in caspase-3 activation over CTL cells. But there was a significant increase ( P = 0.001) in production of active 20 kD caspase-3 fragment in cells treated with TMZ, compared to CTL cells. Treatment of cells with DXM prior to TMZ appeared to significantly decrease the activation of caspase-3 (panel C, Fig. 6 ), indicating an inhibitory effect of DXM on TMZ induced caspase-3 activation. Figure 6 Determination of caspase-3 activation using Western blot analysis of caspase-3-p20 active band. Four treatment groups: CTL; 40 μM DXM for 8 h; 100 μM TMZ for 6 h; pretreatment with 40 μM DXM for 2 h followed by 100 μM TMZ for 6 h. (A) A representative gel picture showing caspase-3 activation. (B) A representative gel picture showing level of expression of β-actin. (C) Densitometric analysis showing percent change in optical density of the caspase-3-p20 active band over CTL. Significant difference between CTL and TMZ treated cells was indicated by * ( P ≤ 0.05) and significant difference between TMZ treated cells and DXM plus TMZ treated cells was indicated by # ( P ≤ 0.05). Discussion Our studies indicated that pretreatment of human glioblastoma U87MG cells with DXM did not support chemotherapeutic action of TMZ. Treatment of U87MG cells with TMZ induced apoptosis (Figs. 1 and 2 ) to a significant extent by increasing intracellular free [Ca 2+ ] (Fig. 3 ), interfering with the expression of apoptosis regulatory proteins of the Bcl-2 family resulting in upregulation of Bax:Bcl-2 ratio (Fig. 4 ), and increasing the activities of calpain and caspase-3 (Figs. 5 and 6 ). But a pretreatment of the cells with DXM prevented all these pro-apoptotic mechanisms (Figs. 3 , 4 , 5 , 6 ). Our data also suggested that pretreatment with DXM can play a critical role in inhibiting Ca 2+ influx into the cells due to treatment with TMZ, and thus preventing the progression of apoptotic process. Several in vitro studies documented a role for calpain in apoptosis of neuronal [ 30 , 31 ] as well as non-neuronal cells [ 32 ]. However, the mechanisms of calpain mediated cell death are not yet fully understood. Pro-apoptotic Bax is translocated to mitochondria and has shown to be activated by calpain [ 33 ]. Increased expression of calpain concurs with elevated expression of Bax relative to Bcl-2, suggesting that calpain overexpression plays an important role during cell death [ 34 , 35 ]. Because changes in expression of pro-apoptotic Bax and anti-apoptotic Bcl-2 control the mitochondrial pathway of apoptosis [ 14 , 15 ], we examined the levels of expression of Bax and Bcl-2 proteins in U87MG cells following treatment with TMZ (Fig. 4 ). Our findings support a relationship between an increase in intracellular free [Ca 2+ ] (Fig. 3 ) and cell death with an elevation of calpain activity (Fig. 5 ) following exposure of U87MG cells to TMZ. Pretreatment of cells with DXM showed a significant decrease in both intracellular free [Ca 2+ ] and calpain activity in a subsequent exposure to TMZ. Increased intracellular free [Ca 2+ ] causes activation of calpain and degradation of cytoskeletal proteins [ 36 ] with destabilization of the cellular integrity leading to cell death [ 37 ]. Our results indicate that DXM plays an important role in the prevention of calpain activation following treatment of glioblastoma cells with TMZ. DXM has been shown to inhibit apoptosis by induction of transcriptional expression of anti-apoptotic proteins Bcl-2 and Bcl-xL [ 37 , 38 ]. Upregulation of Bcl-2 either directly or indirectly can repress the Ca 2+ flux across the membrane of endoplasmic reticulum, thereby abrogating apoptosis via Ca 2+ signaling [ 39 ]. Treatment of U87MG cells with TMZ caused increase in calpain and caspase-3 activities as evidenced from the cleavage of α-spectrin at specific sites generating 145 kD SBDP and 120 kD SBDP, respectively (Fig 5 ). A pretreatment with DXM decreased calpain and caspase-3 activities in U87MG cells. Overall, the results from this investigation showed that DXM pretreatment interfered with proteolytic activities and apoptotic death in U87MG cells exposed to TMZ. A previous report from our laboratory indicated that corticosteroids could inhibit the proteolytic activity of calpain [ 40 ]. Our study suggests that pretreatment of glioblastoma with DXM should be avoided if there is a plan to treat the glioblastoma patients subsequently with TMZ. Some recurrent glioblastomas remain resistant to almost all current therapeutic endeavors, with low response rates and survival rarely exceeding six months. As there are no clearly established chemotherapeutic regimens for drug resistant glioblastomas, obviously the only aim of therapy is palliation with improvement in the quality of life. In such cases, use of DXM or other glucocorticoids may not be controversial. However, promising therapeutic activity of TMZ against newly diagnosed anaplastic astrocytomas and glioblastomas warrants continued evaluation of this agent in combination settings [ 41 ]. Delaying disease progression by treatment with TMZ is beneficial to the patients with recurrent glioblastomas [ 42 ]. Therefore, the use of this drug should be explored further in an adjuvant setting and in combination with other agents [ 43 ]. We showed that a pretreatment of human glioblastoma U87MG cells with a low dose of DXM abolished the chemotherapeutic action of TMZ (Figs. 1 , 2 , 3 , 4 , 5 , 6 ), raising a renewed concern about the validity of DXM as a supportive therapy in the treatment of glioblastomas. We acknowledge that pharmacological studies with a glioblastoma cell line may not always yield results that are easily transferred to the in vivo situation for cancer therapy. Also, clinical recommendations should not be based on in vitro data alone. Nevertheless, our data strongly suggested that DXM treatment could well interfere with therapeutic efficacy of chemotherapy in human glioblastoma patients. In fact, this hypothesis is in line with the results from a 1983 clinical trial where the combination of bis(chloroethyl) nitrosourea (BCNU) plus high dose methylprednisolone, a steroid, tended to be less effective than BCNU alone in patients with poor prognosis [ 44 ]. The data reported here and the previous reports by others [ 45 ], taken together, provide enough reason to call for steroid withdrawal during investigative clinical trials of chemotherapeutic agents in glioblastoma patients. This should be a serious concern at the initial clinical situation when glioblastoma patients are enrolled. In the case of tumor progression during chemotherapy, steroids may still be life-saving agents, and individual decisions concerning a continuation of chemotherapy with concurrent steroid treatment must be made. We also suggest limiting steroid treatment in glioblastoma patients who are receiving chemotherapy outside a controlled clinical trial, because the benefit of chemotherapy for glioblastoma patients is still with limited efficacy and should not be further compromised by co-medication with steroid. Further investigations in xenografted and allografted animal models of glioblastoma as well as in human glioblastoma patients may shed new light in the controversial use of DXM in palliation of human glioblastoma patients. Materials and methods Cell culture and treatments Human glioblastoma U87MG cells were purchased from the American Type Culture Collection (Manassas, VA). Cells were grown in 75-cm 2 flasks containing 10 ml of 1 × RPMI 1640 supplemented with 10% fetal bovine serum (FBS) and 1% penicillin and streptomycin in a fully-humidified incubator containing 5% CO 2 at 37°C. Prior to drug treatments, the cells were starved in 1 × RPMI 1640 supplemented with 0.5% FBS for 24 h. Dose-response studies were conducted to determine the suitable doses of the drugs for using in the experiments. Cells were pretreated with 40 μM DXM for 1 h. The DXM pretreated or untreated cells were subsequently treated with 100 μM TMZ for 6 h. Cells were washed with drug-free medium and allowed to grow for 48 h. Then, cells were collected for determination of viability, apoptosis, or Western blot analysis. DXM and TMZ were obtained from Sigma Chemical (St. Louis, MO) and Schering Corporation (Kenilworth, NJ), respectively. The drugs were dissolved in dimethyl sulfoxide (DMSO) to make stock solutions, which were then stored at -20°C until used for treating cells. Trypan blue dye exclusion test for cell viability Following all treatments the viability of attached and detached cell populations was evaluated by trypan blue dye exclusion test [ 46 ]. Viable cells maintained membrane integrity and did not take up trypan blue. Cells with compromised cell membranes took up trypan blue, and were counted as dead. At least 800 cells were counted in four different fields and the number of viable cells was calculated as percentage of the total cell population. Wright staining for morphological analysis of apoptosis The cells from each treatment were washed with PBS, pH 7.4, and sedimented onto the microscopic slides using Cytobucket and Centra CL2 centrifuge (IEC) at 1200 rpm for 5 min. Cells were fixed in 95% (v/v) ethanol before examination of morphology with Wright staining [ 46 ]. The morphology of the apoptotic cells as detected by light microscopy included such characteristic features as chromatin condensation, cell-volume shrinkage, and membrane-bound apoptotic bodies. Four randomly selected fields were counted for at least 800 cells. The percentage of apoptotic cells was calculated from three separate experiments. ApopTag assay for biochemical detection of apoptotic DNA fragmentation ApopTag Peroxidase kit (Intergen, Purchase, NY) was used to assess the extent of cell death following drug treatments. Briefly, cells for each treatment were grown on six-well cell culture plates (Corning Corporation., Corning, NY) and were treated as described above. Following treatments, cells were washed with PBS and then centrifuged to sediment onto the microscopic slides. Residual PBS was then removed and cells were fixed using 95% (v/v) ethanol and allowed to dry overnight. Slides were pretreated with a protein-digesting enzyme for 15 min and then washed with distilled water for 2 min. Cells were quenched with 3% (v/v) hydrogen peroxide for 5 min followed by washing with PBS. Terminal deoxynucleotidyl transferase (TdT) enzyme was added to the pre-equilibrated cells and incubated for 1 h at 37°C. Stop-buffer was added to the slide and agitated for 15 sec followed by 10 min incubation at room temperature. After washing three times with PBS for 1 min each, anti-digoxigenin peroxidase conjugate was added to the slides and incubated for 30 min. After slides were washed twice with PBS, freshly prepared peroxidase substrate 3,3'-diaminobenzidine was added to the slides and kept for 6 min and then slides were washed with water two times. Slides were counterstained with 0.5% (w/v) methyl green for 10 min followed by washing with water and then 100% n-butanol. After 10 min, cells were dehydrated in xylene for 2 min and then mounted with glass coverslip. Experiments were conducted in triplicates and the percentage of ApopTag-positive cells was determined by counting cells under light microscopy. Determination of intracellular free [Ca 2+ ] using fura-2 We recently reported this method [ 47 ], which was modified for determination of intracellular free [Ca 2+ ] in U87MG cells. Briefly, cells were grown to 80% confluency in phenol red-free medium for 72 h, suspended in the culture medium, centrifuged at 2000 rpm for 5 min to obtain pellet, and washed twice in phosphate buffered saline (PBS, pH 7.4). Cells were resuspended in culture medium, and incubated at 37°C for 2 h with gentle shaking. Following incubation, cells wee washed twice in Ca 2+ -free Locke's buffer [ 48 ] and then counted on a hemocytometer. Cells (2 × 10 7 cells/ml) were dispersed in Locke's buffer with 10% FBS. Fura-2 (Molecular Probes, Eugene, OR) was dissolved in DMSO and diluted in Ca 2+ -free Locke's buffer containing 10% FBS. Cells were mixed with 5 μM fura-2, incubated at 37°C for 30 min, washed twice and diluted to 1 × 10 6 cells/ml in Ca 2+ -free Locke's buffer. The intracellular free [Ca 2+ ] was calculated spectrofluorometrically using the equation [Ca 2+ ] = K d β(R-R min )/(R max -R) , where β is the ratio of F 380 max , fluorescence intensity exciting at 380 nM for zero free Ca 2+ to F 380 min , and fluorescence intensity at saturating free [Ca 2+ ] as reported previously [ 49 ]. The determination of fluorescence ratio (R) was performed using an SLM 8000 fluorometer (Thermospectronic, Shelton, CT) at 340 and 380 nm wavelengths. The maximal (R max ) and minimal (R min ) ratios were determined using 200 μl of 250 μM digitonin (Sigma) and 500 mM EGTA (Sigma, St. Louis, MO), respectively. The value of K d , a cell-specific constant, was determined experimentally to be 0.476 μM using standards of the Calcium Calibration Buffer Kit with Magnesium (Molecular Probes, Eugene, OR). Antibodies Monoclonal antibody against α-spectrin (Affiniti, Exeter, UK) was used to measure calpain activity as well as caspase-3 activity. Caspase-3 polyclonal antibody (MBL International, Woburn, MA) was used to determine caspase-3 activation. Bax and Bcl-2 monoclonal antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) were used to assess apoptotic threshold by determining the Bax:Bcl-2 ratio. Antibody to β-actin (monoclonal clone AC-15, Sigma) was used to standardize protein loading in Western blot experiments. The secondary antibody was horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG (ICN Biomedicals, Aurora, OH), except in case of calpain and α-spectrin where HRP-conjugated goat anti-rabbit IgG was used (ICN Biomedicals). Western blotting and ECL detection Cells were washed in culture flasks using Hank's balanced salt solution without Ca 2+ (GIBCO, Grand Island, NY). The cells were then washed twice in phosphate buffered saline (PBS, pH 7.4) and centrifuged in Eppendorf 5804R (Brinkmann Instruments, Westbury, NY) at 106 × g for 10 min. Cells were resuspended in a homogenizing buffer composed of 50 mM Tris-HCl (pH 7.4), 1 mM PMSF (Bethesda Research Laboratories, Gaithersburg, MD), and 5 mM EGTA (Sigma). A polypropylene pestle (Kontes Glass, Vineland, NJ) was placed inside the microcentrifuge tube (1.5 ml) containing cells and the tube was placed inside an ice bucket for 2 min to let the pestle cool to 4°C. Cells were completely homogenized for 30 sec (the homogenization was performed while holding the microcentrifuge tube in ice with one hand). Following homogenization, protein concentration was determined using Coomassie Plus Protein Assay Reagent (Pierce, Rockford, IL) and spectrophotometric measurement at 595 nm (Spectronic, Rochester, NY). Samples were then diluted (1:1) in sample buffer (62.5 mM, Tris pH 6.8, 2% SDS, 5 mM β-mercaptoethanol, 10% glycerol) and boiled for 5 min. Samples were then loaded onto the 4–20% gradient gels for electrophoresis at 200 V for 30 min (Bio-Rad, Hercules, CA). For detection of α-spectrin bands, a 5% gel was used for electrophoresis at 100 V for 2 h. Following electrophoresis, gels with the resolved proteins were electroblotted to nylon membranes (Millipore, Bedford, MA) in an electroblotting Genie apparatus (Idea Scientific, Minneapolis, MN). The membranes were blocked for 1 h in blocking buffer (5% powdered non-fat milk, 20 mM Tris pH 7.6, 0.1% Tween 20 in saline). Primary antibody was diluted (1:100 for Bax, Bcl-2, caspase-3, and 1:500 for calpain, 1:2,000 for α-spectrin, and 1:15,000 for β-actin) in blocking solution and then added to the blots for 1 h. The blots were washed three times with a wash buffer (Tris/Tween solution) and covered with secondary antibody (goat anti-rabbit for calpain and α-spectrin and goat anti-mouse for all others) at a 1:2000 dilution for 1 h. Blots were incubated with enhanced chemiluminescence (ECL) detection system (Amersham Pharmacia, Buckinghamshire, UK) and exposed to X-OMAT AR films (Eastman Kodak, Rochester, NY). The films were scanned on a UMAX PowerLook Scanner (UMAX Technologies, Fremont, CA) using Photoshop software (Adobe Systems, Seattle, WA), and optical density of each band was determined using Quantity One software (Bio-Rad, Hercules, CA). Estimation of the degradation products of α-spectrin indicated calpain and caspase-3 activities. The 145 kD spectrin breakdown product (SBDP) is specific for calpain activation [ 30 ], and the 120 kD SBDP is specific for caspase-3 activation [ 50 ]. Also, Western blot analysis was performed to examine the generation of active 20 kD caspase-3 fragment from 32 kD caspase-3, indicating activation of caspase-3. Statistical analysis Data from various experiments were analyzed using StatView software (Abacus Concepts, Berkeley, CA). Results were compared using one-way analysis of variance (ANOVA) with Fisher's protected least significant difference (PLSD) post hoc test at a 95% confidence interval. All results were presented as mean ± standard error of mean (SEM) of separate experiments (n ≥ 3). A difference between two values was considered significant at p ≤ 0.05. Authors' contributions AD performed the experiments and participated in writing the manuscript. NLB contributed to the interpretation of results. SJP participated in the discussion of the study and provided comments on clinical importance of this study. SKR conceived the study, planned experimental design, supervised the study, and helped writing the manuscript. All authors read and approved the final version of the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544397.xml
509281	Activity of telithromycin and comparators against bacterial pathogens isolated from 1,336 patients with clinically diagnosed acute sinusitis	Background Increasing antimicrobial resistance among the key pathogens responsible for community-acquired respiratory tract infections has the potential to limit the effectiveness of antibiotics available to treat these infections. Since there are regional differences in the susceptibility patterns observed and treatment is frequently empirical, the selection of antibiotic therapy may be challenging. PROTEKT, a global, longitudinal multicentre surveillance study, tracks the activity of telithromycin and comparator antibacterial agents against key respiratory tract pathogens. Methods In this analysis, we examine the prevalence of antibacterial resistance in 1,336 bacterial pathogens, isolated from adult and paediatric patients clinically diagnosed with acute bacterial sinusitis (ABS). Results and discussion In total, 58.0%, 66.1%, and 55.8% of S. pneumoniae isolates were susceptible to penicillin, cefuroxime, and clarithromycin respectively. Combined macrolide resistance and reduced susceptibility to penicillin was present in 200/640 (31.3 %) of S. pneumoniae isolates (128 isolates were resistant to penicillin [MIC >= 2 mg/L], 72 intermediate [MIC 0.12–1 mg/L]) while 99.5% and 95.5% of isolates were susceptible to telithromycin and amoxicillin-clavulanate, respectively. In total, 88.2%, 87.5%, 99.4%, 100%, and 100% of H. influenzae isolates were susceptible to ampicillin, clarithromycin, cefuroxime, telithromycin, and amoxicillin-clavulanate, respectively. In vitro, telithromycin demonstrated the highest activity against M. catarrhalis (MIC 50 = 0.06 mg/L, MIC 90 = 0.12 mg/L). Conclusion The high in vitro activity of against pathogens commonly isolated in ABS, together with a once daily dosing regimen and clinical efficacy with 5-day course of therapy, suggest that telithromycin may play a role in the empiric treatment of ABS.	Introduction The incidences of both the acute and chronic forms of sinusitis have been increasing, and between 10 and 15% of the population of central Europe are affected annually [ 1 ]. There are an estimated 30 million cases of ABS in the USA each year [ 2 - 4 ]. Acute sinusitis accounts for 0.5–2.0% of all upper respiratory tract infections in adults and between 5–10% in children and therefore is a common reason for visits to primary care physicians [ 5 ]. Although usually mild in severity, complications can be life threatening, including meningitis, brain abscess, orbital cellulitis and abscess, subempyema, osteomyelitis, and nasal polyposis [ 6 - 9 ]. S. pneumoniae is the most common pathogenic bacterium responsible for ABS, isolated in 30–50% of cases, followed by H. influenzae , isolated in 20–40% of cases. Moraxella catarrhalis is isolated in 5–10% of cases, beta haemolytic streptococci in less than 5%, and Staphylococcus aureus in less than 10% although it is often found co-infecting with other bacteria [ 10 ]. Treatment options for ABS are controversial as up to 40% of patients recover spontaneously, however, studies have shown that treatment with an antibacterial reduces the time to recovery from sinusitis, improves symptoms, and helps to prevent complications [ 11 , 12 ]. Guidelines on antibacterial use for ABS vary, possibly because of different regulations, antibacterial resistance patterns, and etiology in different countries, however, the choice of first-line antibacterial is similar across treatment guidelines [ 3 , 13 - 17 ]. Nearly all recommend amoxicillin, as it is active against the major causative pathogens of AMS and is generally well tolerated. For patients with penicillin allergy, the recommended first-line agents vary in different countries. Trimethoprim or trimethoprim-sulfamethoxazole is commonly recommended [ 3 , 14 ]. In addition, the French guidelines have recently been reviewed and telithromycin has been included as an alternative first-line agent [ 15 ]. Macrolides are not included in the French guidelines due to the high macrolide resistance prevalence in France [ 15 , 18 ]. Second line, or alternative antibacterial agents of choice, are clarithromycin or second-generation cephalosporins such as cefuroxime and cefpodoxime, third-generation cephalosporins such as cefdinir, and trimethoprim-sulphamethoxazole [ 3 , 5 , 16 , 17 ]. Telithromycin has been recommended as an alternative agent in Germany [ 13 , 16 ]. High-dose amoxicillin-clavulanate should be used if the patient does not improve [ 3 , 16 , 17 ]. In France, anti-pneumococcal fluoroquinolones are recommended after bacterial confirmation or if the patient is at high risk of complications [ 15 ]. A single dose of ceftriaxone can be used in a child who cannot be treated orally, i.e. vomiting [ 17 ]. However, due to increasing levels of resistance in bacterial respiratory tract pathogens to these commonly used antibacterials (particularly the rapid emergence of penicillin-and macrolide-resistant strains of pneumococci), new agents are required that have high in vitro activity and demonstrated clinical efficacy against bacterial pathogens causing community-acquired respiratory tract infections (RTI's) [ 19 - 22 ]. Telithromycin is the first ketolide approved for clinical use. The ketolides are semisynthetic derivatives of the 14-membered ring macrolide erythromycin and have high in vitro activity against the common community-acquired RTI pathogens [ 23 ]. Clinical trials have demonstrated the efficacy and tolerability of telithromycin therapy in ABS [ 24 - 26 ]. The PROTEKT (Prospective Resistant Organism Tracking and Epidemiology for the Ketolide Telithromycin) study is a longitudinal, global multicentre surveillance study designed in part to determine the activity of telithromycin against community-acquired RTI isolates, in relation to the frequency of prescribing, in the regions where the study is conducted [ 27 ]. The aim of this paper is to focus on the data gathered in the PROTEKT surveillance study to determine the in vitro efficacy of the new ketolide telithromycin and comparator agents against bacterial pathogens isolated from the subset of patients with clinically diagnosed ABS collected in PROTEKT (2000–2001, and 2001–2002). Materials and Methods Patients and bacterial isolates Detailed study design, including patient selection and methodology for isolate identification and storage in the PROTEKT study has been described previously [ 27 ]. The isolates in this sub-study of PROTEKT were selected from those patients presenting with clinically diagnosed ABS in which the isolates were determined clinically to be the pathogenic organism and the specimen type was sinus aspirate or nasopharyngeal swab/aspirate only. Methodology for sinus aspiration was that used routinely by the investigator. Antimicrobial testing MIC susceptibility status was determined, using the National Committee of Clinical and Laboratory Standards (NCCLS) breakpoints, at a central laboratory (GR Micro Ltd, London, UK) from a panel of existing and new antibacterials, using the NCCLS broth microdilution method and lyophilised microtitre plates (Sensititre, Trek Diagnostics) [ 28 ]. NCCLS breakpoints were used for interpretation of MIC's [ 29 ]. Tentative NCCLS breakpoints for telithromycin are: S. pneumoniae and S. aureus , ≤ 1 μg/ml is susceptible, 2 μg/ml is intermediate, and ≥ 4 μg/ml is resistant; for Haemophilus influenzae , ≤ 4 μg/ml is susceptible, 8 μg/ml is intermediate, and ≥ 16 μg/ml is resistant [ 29 ]. Statistical analysis Statistical analysis was performed using a χ 2 test. Results A total of 1,336 bacterial pathogens in all were collected from 25 countries within Western Europe ( n = 652), North America ( n = 14), Latin America ( n = 207), Asia ( n = 464), Eastern Europe ( n = 68), Australia ( n = 2), and South Africa ( n = 126) in the PROTEKT study from years 2000–2001 and 2001–2002 for analysis of the susceptibility of bacterial pathogens isolated from patients with acute sinusitis. Gender distribution was 52.2% male (695 patients), 46.7% female (624 patients); gender was not provided for 1.3% of patients. Almost two thirds (66.3%) of patients were in the 0–12 year age group, one third (29.7%) in the 13–65 year age group, 2.7% in the over 65 year age group and age was not specified in 1.3% of patients. S. pneumoniae was the pathogen most frequently isolated (47.9% of isolates) followed by H. influenzae (24.6% of isolates) (Table 1 ). Table 1 Distribution of species by specimen type for the 1336 bacterial pathogens causing acute sinusitis [n (%)] Specimen S. pneumoniae H. influenzae M. catarrhalis S. aureus S. pyogenes Total Sinus 272 (47.5) 148 (25.9) 67 (11.7) 64 (11.2) 21 (3.7) 572 (42.8) Nasopharynx 1 368 (48.1) 181 (23.7) 145 (19.0) 52 (6.8) 18 (2.4) 764 (57.2) Total 640 (47.9) 329 (24.6) 212 (15.9) 116 (8.7) 39 (2.9) 1336 (100) 1 Aspirate or swab MIC data for isolates from patients with ABS demonstrated that the in vitro activity of telithromycin against gram-positive cocci was similar to amoxicillin-clavulanate and was higher and more potent than clarithromycin and beta-lactams tested such as cefuroxime (Table 2 ). In total, 99.5 % of streptococcal isolates were susceptible to telithromycin. With the exception of S. aureus isolates more than 90% of gram-positive cocci were inhibited at a telithromycin MIC of 0.25 mg/L (Table 2 ). Table 2 In vitro activity of antibacterial agents and percent susceptible against 1336 bacterial pathogens isolated from patients with clinically diagnosed acute sinusitis Organism N (total, SA 1 , NP 2 ) Antibiotic MIC (mg/L) Percent susceptible (Total, SA, NP ) Range 50 90 S. pneumoniae 640, 272, 368 Penicillin 0.008 – 8 0.06 2 58.0, 64.7, 53.0 Amoxicillin-clavulanate 0.015 – 8 0.03 2 95.5, 95.2, 95.7 Cefuroxime 0.015 – 16 0.12 8 66.1, 73.2, 60.9 Cefpodoxime 0.12 – 32 0.12 2 65.0, 71.7, 60.1 Trimethoprim-sulphamethoxazole 0.12 – 32 0.5 8 56.3, 58.1, 54.9 Erythromycin 0.03 - >64 0.06 >64 55.9, 60.7, 52.5 Clarithromycin 0.015 - >32 0.06 >32 55.8, 60.7, 52.2 Azithromycin 0.03 - >64 0.12 >64 55.8, 60.7, 52.2 Telithromycin 0.008 – 8 0.015 0.12 99.5, 98.9, 100 H. influenzae 329, 148, 181 Ampicillin 0.12 – 32 0.25 8 88.2, 91.2, 85.6 Amoxicillin-clavulanate 0.12 – 4 0.5 1 100, 100,100 Cefuroxime 0.12 – 16 1 2 99.4, 99.3, 99.5 Cefpodoxime 0.015 – 4 0.06 0.25 99.4, 100, 98.9 Cefdinir 0.06 – 4 0.25 0.5 97.3, 96.0, 98.3 Trimethoprim-sulphamethoxazole 0.03 – 16 0.06 4 84.5, 82.4, 86.2 Erythromycin 0.25 – 16 4 8 - 3 Clarithromycin 0.25 – 32 8 16 87.5, 87.2, 87.9 Azithromycin 0.06 – 4 1 2 100, 100,100 Telithromycin 0.06 – 4 1 2 100, 100,100 M. catarrhalis 212, 67, 145 Ampicillin 0.12 – 32 4 16 - Amoxicillin-clavulanate 0.12 – 0.5 0.12 0.25 - Cefuroxime 0.12 – 16 1 4 - Cefpodoxime 0.06 – 4 0.5 1 - Cefdinir 0.06 – 1 0.12 0.25 - Trimethoprim-sulphamethoxazole 0.03 – 2 0.12 0.25 - Erythromycin 0.25 – 1 0.25 0.25 - Clarithromycin 0.25 – 0.5 0.25 0.25 - Azithromycin 0.06 – 0.25 0.06 0.06 - Telithromycin 0.004 – 0.5 0.06 0.12 - S. aureus 116, 64, 52 Methicillin - - - 90.5, 92.2, 88.5 Amoxicillin-clavulanate 0.06 – 8 0.5 4 90.5, 92.2, 88.5 Cefuroxime 0.5 – 16 1 2 90.5, 92.2, 88.5 Cefpodoxime 1 – 32 2 4 88.8, 92.2, 84.6 Trimethoprim-sulphamethoxazole 0.12 – 32 0.12 0.12 96.6, 95.3, 98.1 Erythromycin 0.03 - >64 0.25 >64 69.0, 73.4, 63.5 Clarithromycin 0.015 - >32 0.25 >32 69.8, 75.0, 63.5 Azithromycin 0.03 - >64 0.5 >64 69.0, 75.0, 61.5 Telithromycin 0.015 - >32 0.06 2 89.7, 89.1, 90.4 S. pyogenes 39, 21, 18 Penicillin 0.008 – 0.008 0.008 0.008 100, 100,100 Amoxicillin-clavulanate 0.008 – 0.03 0.015 0.015 100, 100,100 Cefuroxime 0.015 – 0.015 0.015 0.015 100, 100,100 Cefpodoxime 0.12 – 0.12 0.12 0.12 100, 100,100 Trimethoprim-sulphamethoxazole 0.12 – 0.5 0.12 0.25 - Erythromycin 0.03 – 4 0.06 0.25 92.3, 100, 83.3 Clarithromycin 0.015 – 2 0.03 0.25 92.3, 100, 83.3 Azithromycin 0.03 – 16 0.12 0.25 92.3, 100, 83.3 Telithromycin 0.008 – 0.12 0.015 0.015 - 1 Sinus aspirate 2 Nasopharyngeal aspirate or swab 3 No NCCLS interpretive guidelines available or pending Resistance to most antibiotics was slightly greater in nasopharyngeal specimens than sinus aspirates (Table 2 ). Considerable variation in in vitro antibiotic activity was apparent between geographical regions as observed in the key examples shown in Table 3 . Insufficient data were available for analysis by country. Table 3 Key example of regional variation in in vitro antibiotic activity Streptococcus pneumoniae Haemophilus influenzae REGION 1 N Penicillin susceptible Erythromycin susceptible N Beta-lactamase positive Eastern Europe 40 70.0% 82.5% 13 0.0% Far East 185 24.9% 15.7% 96 13.5% Latin America 108 63.0% 73.1% 31 6.5% South Africa 61 37.7% 59.0% 34 2.9% Western Europe 246 83.7% 73.6% 153 11.8% Grand Total 640 58.0% 55.9% 329 10.3% 1 Australasia and North America not included due to insufficient data Combined macrolide resistance and reduced susceptibility to penicillin was present in 200/640 (31.3 %) of S. pneumoniae isolates (128 isolates were resistant to penicillin [MIC >= 2 mg/L], 72 intermediate [MIC 0.12–1 mg/L]). Of note, 3 isolates of S. pneumoniae were non-susceptible to telithromycin (2 isolates intermediate with an MIC of 2 mg/L, 1 isolate resistant with an MIC of 8 mg/L). This represented 0.5% of isolates, a value that is significantly (p < 0.001) lower than those obtained by erythromycin (44.1%), clarithromycin (44.2%) and cefuroxime (33.9%). Of the 329 H. influenzae isolates, 34 (10.3 %) were positive for β-lactamase production. All isolates of H. influenzae were susceptible to amoxycillin-clavulanate and telithromycin with an MIC 90 of 1 and 2 mg/L, respectively. Amoxycillin-clavulanate and telithromycin were more potent and had greater activity than clarithromycin (MIC 90 = 16 mg/L, 87.5% susceptible). This activity was comparable to azithromycin (MIC 90 = 2 mg/l, 100% susceptible). Although the number of S. aureus isolated from the total number of specimens was small (116/1366 isolates), telithromycin was as efficacious as comparators. Of the 116 isolates, 11 were resistant to methicillin (MRSA) and 105 were methicillin susceptible (MSSA). Ninety-nine (94.3%) of the MSSA isolates and 5 of the 11 MRSA isolates were susceptible to telithromycin. Of note, all of the S. pyogenes isolates were inhibited by ≤ 1 mg/L telithromycin, despite 17.7% resistance to erythromycin and clarithromycin. Telithromycin was the most potent antimicrobial against M. catarrhalis with an MIC 50 of 0.06 mg/L and MIC 90 of 0.12 mg/L. β-lactamase production was detected in 97.6% of these isolates. Discussion The data in this analysis demonstrates that telithromycin has high in vitro activity against bacterial pathogens isolated from a large, globally distributed population of patients diagnosed with ABS. Telithromycin was the most active and potent agent against all isolates of the pathogens isolated from patients with ABS with 99.4% of isolates susceptible. Not surprisingly, high levels of penicillin resistance, macrolide resistance, and combined penicillin and macrolides resistance were prevalent in S. pneumoniae although prevalence varied widely between geographical regions. Amoxicillin has been the treatment of choice in ABS because of its general effectiveness, safety, tolerability, low cost and narrow spectrum [ 17 ]. The high prevalence of beta-lactamase in H. influenzae and M. catarrhalis found in the present study demonstrate compromised in vitro efficacy of amoxicillin against these isolates. Although the cephalosporins (cefuroxime, cefpodoxime and cefdinir) showed high activity against H. influenzae (including beta-lactamase positive strains), resistance to these agents was high in S. pneumoniae : >30% for cefuroxime and cefpodoxime – cefdinir was not tested against pneumococci in PROTEKT, however susceptibility is usually similar to the other cephalosporins reported here [ 30 ]. Similarly, macrolides are prescribed in various countries for ABS and an overall resistance rate for S. pneumoniae of 44.1% to erythromycin, azithromycin, and clarithromycin was found. Trimethoprim-sulfamethoxazole activity was low for S. pneumoniae (56.3% susceptible) and decreased for H. influenzae (84.5% susceptible). Respiratory fluoroquinolones are recommended second-line treatment options in some countries (references needed to support this statement). However, recent evidence suggests that resistance to fluoroquinolones is rapidly developing in pneumococci and other pathogens (including gram-positive and gram-negative [ 31 - 34 ]. To preserve the long-term utility of fluoroquinolones, including their use in the treatment of serious non-respiratory infections, it has been recommended that respiratory fluoroquinolones be reserved for treating severe (e.g. hospitalized) community-acquired RTIs only [ 35 , 36 ]. The high prevalence of beta-lactam, macrolide, TMP-SMX resistance demonstrated in the large number of isolates from patients with clinically diagnosed sinusitis in our study demonstrates the need to be exploring new therapeutic options, especially in geographical regions of high prevalence such as the Far East. The high in vitro activity of telithromycin against ABS pathogens reported in this study, regardeless of geographical region, also demonstrates its potential as an empiric therapeutic option for ABS. There are several other reasons to consider this option – 1) High rates of clinical cure and bacteriological eradication have been demonstrated using telithromycin against sinus isolates of S. pneumoniae , H. influenzae , M. catarrhalis and S. aureus [ 24 - 26 ]. 2) Telithromycin has been shown to have a targeted spectrum of activity against the major bacterial respiratory tract pathogens and has less effect on normal bacterial ecology [ 37 - 39 ]. 3) The pharmacokinetic profile of oral telithromycin allows it to be prescribed with a dosing regime of 800 mg once daily for 5 days [ 40 , 41 ]. This contrasts favourably with its comparators, where a 10 – 14 day course with administration either 2 or 3 times daily, depending on the chosen antibacterial, is generally prescribed. Studies have shown that the once daily dosing regime affords greater patient treatment compliance, thereby avoiding clinical failure and the ensuing development of antibacterial resistance [ 41 - 43 ]. 5) Telithromycin has been shown to have high penetration levels in paranasal sinuses, and it is preferentially absorbed by polymorphonuclear neutrophils (PMNs) within the azurophil granules allowing effective delivery to phagocytized intracellular bacteria [ 44 , 45 ]. Although this study provides valuable information on the overall antimicrobial profile of bacteria causing ABS care should be taken when interpreting data related to specific demographics. A major limitation of this study, inherent to most surveillance studies, is the requirement for collecting centers to fulfill a specified quota of isolates over a defined time period (1 year). If, for instance, 1 center manages to fulfill their quota for S. pneumoniae isolates from patients with community-acquired pneumonia, they may then only send H. influenzae from patients with ABS to fulfill their quota for this organism. Thus, the potential exists to over or under estimate the prevalence of a species in a particular disease. A further limitation of this study is it is restricted to the major bacterial pathogens causing sinusitis and does not therefore assess anaerobic bacteria, which are also known to be involved in this disease. However, a recent study of sinus puncture specimens demonstrated that telithromycin had good in vitro activity against anaerobes involved in sinusitis [ 46 ]. The inclusion of nasopharyngeal specimens is a potential limitation of this study and the higher rate of resistance compared to sinus aspirates may indicate some isolates were nasopharyngeal flora rather than pathogens. However, the difference in resistance prevalence between nasopharyngeal specimens and sinus aspirates was not great for any species/antibiotic combination, and assuming the majority of isolates were the responsible pathogen, significant bias of resistance patterns is unlikely. The treatment of ABS is complicated by a difficulty in establishing the causative pathogen(s). Sampling of infected fluid using sinus puncture is a painful and rare procedure [ 47 ]. Nasopharyngeal culture is a painless and reliable method that can help identify patients that may benefit from antibacterial therapy [ 48 ] and hence, could be useful in determining antibiotic resistance implicated in sinusitis – particularly Streptococcus pneumoniae and Haemophilus influenzae . Additionally, there are regional differences in the susceptibility patterns observed and, as therapy is usually empirical, choosing an effective therapy can be challenging [ 18 , 20 , 49 ]. In summary, the data presented here demonstrates that telithromycin has good in vitro activity against S. pneumoniae , H. influenzae, M. catarrhalis and S. aureus respiratory pathogens commonly isolated in ABS. It is as active as or more active than antibacterial agents that are currently used in this clinical setting. The development of resistance will always be a threat to the usefulness of antibacterial compounds, however surveillance studies such as PROTEKT allow the rapid detection and characterization of resistance mechanisms and highlight the need for and examine the in vitro efficacy of newer antibacterial agents. Providing careful surveillance for the development of resistance is maintained telithromycin currently offers a useful therapeutic option in the treatment of AS. Authors' contributions JD, RCa and RCo reviewed the data and provided clinical and microbiological interpretation and discussion. DF and DJF participated in the design of the study, supervised the scientific testing, and provided data analysis, microbiological interpretation and discussion. All authors drafted the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509281.xml
193606	Biological Clock Depends on Many Parts Working Together	null	How do people subjected to the endless dark days of winter in the far northern latitudes maintain normal daily rhythms? Though many might feel like hibernating, a highly regulated internal system keeps such impractical yearnings in check. From fruit flies to humans, nearly every living organism depends on an internal clock to regulate basic biological cycles such as sleep patterns, metabolism, and body temperature. And that clock runs on similar molecular mechanisms. Specific clusters of neurons in the brain are known to control the biological clock. Scientists believed these brain “clock cells” function as independent units. But new research described in this issue shows that the neurons do not act in isolation; rather, they collaborate with other neurons in a cell-communication network to sustain the repeating circadian rhythm cycles. Clock cells within the brain maintain an organism's circadian rhythms, even in the absence of cyclical environmental signals like light, in a state scientists call “free running.” Though it has long been clear that the circadian rhythms of an organism persist under such free-running conditions (for example, constant darkness), it was thought that the gene-expression patterns within the cells governing these biorhythms did not require any external, or extracellular, signals to continue ticking. In experiments described here, Michael Rosbash and his colleagues show that the key brain clock cells in fruit flies ( Drosophila ), called ventral lateral neurons , do indeed support the fly's circadian rhythms during periods of constant darkness and that the molecular expression patterns associated with these rhythms continue to cycle as well within other clock cells. These sustained expression patterns, however, require intercellular communication between different groups of brain clock cells. In other words, the ventral lateral neurons do not act alone. When the molecular clock machinery was manipulated so that only the ventral lateral neurons were active, the fly's circadian rhythms were not sustained, suggesting the rhythms depend on other neuronal groups as well. The researchers also demonstrate that the persistence of normal cycling during constant darkness depends on a protein (called PDF) secreted by the ventral lateral cells. The PDF neuropeptide protein was thought to connect the molecular expression pattern of the ventral lateral neurons with the manifestation of circadian rhythms, but the researchers found evidence of a larger influence. When mutant flies lacking a functional PDF gene were exposed to constant darkness, the molecular expression patterns gradually stopped. The scientists say this suggests that the ventral lateral neurons and the PDF protein it produces help coordinate the entire neural network that underlies circadian rhythms. Drosophila lateral neuron (green)	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC193606.xml
534114	The docking protein Gab1 is the primary mediator of EGF-stimulated activation of the PI-3K/Akt cell survival pathway	Background Gab1 is a docking protein that recruits phosphatidylinositol-3 kinase (PI-3 kinase) and other effector proteins in response to the activation of many receptor tyrosine kinases (RTKs). As the autophosphorylation sites on EGF-receptor (EGFR) do not include canonical PI-3 kinase binding sites, it is thought that EGF stimulation of PI-3 kinase and its downstream effector Akt is mediated by an indirect mechanism. Results We used fibroblasts isolated from Gab1-/- mouse embryos to explore the mechanism of EGF stimulation of the PI-3 kinase/Akt anti-apoptotic cell signaling pathway. We demonstrate that Gab1 is essential for EGF stimulation of PI-3 kinase and Akt in these cells and that these responses are mediated by complex formation between p85, the regulatory subunit of PI-3 kinase, and three canonical tyrosine phosphorylation sites on Gab1. Furthermore, complex formation between Gab1 and the protein tyrosine phosphatase Shp2 negatively regulates Gab1 mediated PI-3 kinase and Akt activation following EGF-receptor stimulation. We also demonstrate that tyrosine phosphorylation of ErbB3 may lead to recruitment and activation of PI-3 kinase and Akt in Gab1-/- MEFs. Conclusions The primary mechanism of EGF-induced stimulation of the PI-3 kinase/Akt anti-apoptotic pathway occurs via the docking protein Gab1. However, in cells expressing ErbB3, EGF and neuroregulin can stimulate PI-3 kinase and Akt activation in a Gab1-dependent or Gab1-independent manner.	Background Ligand stimulation of the epidermal growth factor receptor (EGFR) and the three other members of the EGFR family of receptor tyrosine kinases (ErbB2, ErbB3 and ErbB4) results in tyrosine autophosphorylation, recruitment of signaling proteins, and activation of distinct complement of signaling pathways that regulate a great variety of cellular responses [ 1 , 2 ]. One of the signaling pathways that is activated by the EGFR is the phosphatidylinositol-3 kinase (PI-3 kinase)/Akt anti-apoptic signaling pathway [ 3 ]. The PI-3 kinase holoenzyme consists of a regulatory subunit (p85) and a catalytic p110 subunit. The regulatory subunit contains two SH2 domains that bind specifically to pYXXM motifs in a variety of cellular proteins, including receptor tyrosine kinases such as the PDGF (platelet-derived growth factor) receptor, and c-kit and docking proteins such as IRS (insulin receptor substrate) 1, IRS2 and Gab1. Although the cytoplasmic domain of the EGFR does not contain any canonical p85 binding motifs, EGF stimulation leads to PI-3 kinase activation by an indirect mechanism. It has been proposed that the PI-3 kinase is activated in response to EGF stimulation as a consequence of formation of EGFR/ErbB3 heterodimers [ 4 ]. Unlike EGFR, the cytoplasmic region of ErbB3 contains at least six pYXXM motifs [ 5 , 6 ]. Indeed, EGF stimulation of cells co-expressing the EGFR and ErbB3 results in recruitment and activation of PI-3 kinase by tyrosine phosphorylated ErbB3 [ 4 ]. EGF stimulation of PI-3 kinase may also be mediated by the docking protein Gab1 (Grb2-associated binder-1). EGF stimulation leads to tyrosine phosphorylation of Gab1 enabling recruitment and activation of PI-3 kinase by the three canonical pYXXM motifs on Gab1 [ 7 ]. Gab1 was originally identified as a Grb2 binding protein, and was shown to be tyrosine phosphorylated in response to treatment with a variety of growth factors [ 7 - 9 ]. Gab1 contains a number of tyrosine residues that could serve as potential binding sites for the SH2 domain containing proteins Grb2, PI-3 kinase, and the protein tyrosine phosphatase Shp2 [ 10 ]. While there have been reports that Gab1 binds directly to the EGFR via an 83-amino acid stretch termed the Met-binding-domain or MBD [ 9 ], the majority of Gab1 is believed to be indirectly associated with the EGFR via the adaptor protein Grb2, which binds to a proline rich region on Gab1 via its C-terminal SH3 domain [ 10 - 13 ]. Cells over-expressing a mutant Gab1 protein containing tyrosine to phenylalanine mutations at the three p85 binding sites have been shown to be defective in EGF-induced JNK activation, and treatment of cells over-expressing wild type Gab1 with PI-3 kinase inhibitors interferes with MAPK signaling in response to EGF treatment, thus revealing a link between Gab1 and PI-3 kinase in EGF-induced mitogenic signaling [ 9 , 14 ]. Furthermore, the PI-3 kinase product phosphatidylinositol (3,4,5) triphosphate (PIP3) has been shown to bind to the pleckstrin homology (PH) domain of Gab1 resulting in membrane-association of Gab1, suggesting a positive feedback loop in which PI-3 kinase acts as both an upstream regulator and a downstream effector of Gab1 signaling via the EGFR [ 9 ]. Gab1 thus acts as a docking protein facilitating the recruitment of a multi-protein signaling complex including the EGFR, p85 and Shp2 in response to EGF treatment. Although the role of the Shp2 protein in the control of EGFR/Gab1 interactions is not well understood, several studies have suggested that Gab1-associated Shp2 may influence EGF-induced PI-3 kinase signaling. Previous work has shown that Gab1 is not a global substrate of Shp2, as complex formation between Gab1 and Shp2 does not reduce the total EGF-induced tyrosine phosphorylation levels of Gab1 [ 15 ]. However there have been several reports suggesting that Shp2 may specifically de-phosphorylate the tyrosine phosphorylation sites on Gab1 that bind to p85, thus terminating recruitment of PI-3 kinase and EGF-induced activation of the PI-3 kinase pathway [ 16 - 18 ]. It has been shown that cells devoid of Shp2 show an increase in PI-3 kinase activity, as well as elevated and sustained levels of Akt activation in response to EGF treatment [ 18 ]. It was reported that treatment of cells with PI-3 kinase inhibitors or with the phosphatidylinositol (3,4,5) triphosphate (PIP3) phosphatase PTEN interferes with the association between Gab1 and Shp2 in response to EGF treatment, suggesting PI-3 kinase may be required for Shp2 recruitment of Gab1 following EGF stimulation [ 14 ]. However, the mechanism for this postulated recruitment is unknown. In the experiments presented here we utilized fibroblasts isolated from Gab1-/- mouse embryos in order to examine the role of Gab1 in EGF-mediated activation of the PI-3 kinase/Akt cell survival pathway. We also address the question of whether ErbB3 recruitment of PI-3 kinase is dependent on or independent of Gab1. Our results demonstrate a clear requirement for Gab1 in recruitment and activation of PI-3 kinase in response to EGF stimulation. Additionally, we show that while Shp2 does not mediate global dephosphorylation of Gab1, it does appear to negatively regulate the EGF-induced activation of PI-3 kinase through an undefined mechanism. Finally we demonstrate that ErbB3 is capable of recruiting PI-3 kinase in the absence of Gab1, but Gab1 functions as the major mediator of PI-3 kinase activation in response to EGF stimulation. Results and Discussion Previous studies have suggested that ErbB3 and Gab1 can function as links between EGFR and PI-3 kinase. In this report we use MEFs derived from Gab1 -/- embryos [ 19 ] to explore the contribution of Gab1 and ErbB3 to EGF stimulation of PI-3 kinase and Akt in these cells. Gab1 is essential for EGF stimulation of PI-3 kinase and Akt MEFs derived from Gab1-/- or wild type (WT) embryos were stimulated with EGF and assayed for Gab1 tyrosine phosphorylation, for activation of PI-3 kinase and for Akt stimulation. As shown in Figure 1A , the endogenous Gab1 present in WT MEFs is tyrosine phosphorylated in response to EGF treatment. As shown in Figure 1B , Gab1 -/- MEFs displayed very low levels of EGF-induced PI-3 kinase activity relative to cells expressing Gab1. We did observe an approximate 2-fold increase in this low-level basal PI-3 kinase activity in Gab1 -/- MEFs, which represents a Gab1-independent signaling pathway. Gab1 -/- and WT MEFs were additionally stimulated with EGF and the activation of Akt was analyzed by immunoblotting with antibodies which recognize specifically the activated form of Akt. As shown in Figure 1C (top and middle panels), Gab1 -/- cells display no activation of Akt in response to EGF, while WT MEFs show EGF-stimulated Akt activation within two minutes of EGF treatment. The cDNA encoding the wild type murine Gab1 sequence was cloned into a retroviral vector, and the virus was used to infect Gab1 -/- MEFs. Stable cell lines were selected for co-transduction of a puromycin resistance gene and pools of selected cells were used for further analysis. As shown in Figure 1A , the ectopic Gab1 protein was expressed at slightly lower levels in the Gab1 -/- MEFs relative to endogenous Gab1 expression seen in the wild type MEFs. Quantitation by densitometry reveals Gab1 expression in wild type MEFs to be 1.4-fold higher than ectopic Gab1 expression in Gab1 -/- MEFs. Treatment with EGF induced tyrosine phosphorylation of the exogenous Gab1 protein expressed in the Gab1 -/- MEFs at levels similar to endogenous Gab1 in wild type MEFs. As shown in Figure 1B , expression of exogenous Gab1 in the Gab1 -/- MEFs results in Gab1-associated PI-3 kinase activity that is augmented following EGF treatment. The low level of EGF-induced PI-3 kinase activity observed in the Gab1 -/- cells may be due to signaling via an alternate, Gab1-independent mechanism. These cells were additionally treated with EGF over a period of time and the activation of Akt was assayed by immunoblotting with antibodies specific for the Ser473 phosphorylated form of Akt. The experiment presented in Fig 1C shows that ectopic expression of Gab1 in the Gab1 deficient cells rescues EGF-induced Akt activation to levels similar to those observed in EGF-treated wild type MEFs. Taken together, these results demonstrate that Gab1 is required for EGF-stimulation of PI-3 kinase and Akt. The canonical p85 binding sites on Gab1 are essential for PI-3 kinase and Akt activation in response to EGF stimulation The cDNA encoding a mutant Gab1 protein, containing tyrosine to phenylalanine point mutations at the three binding sites for the p85 regulatory subunit of PI-3 kinase (Y446F/Y472F/Y589F) (Gab1 F446/472/589 ), was cloned into a retroviral vector and used to generate pools of stable MEF cell lines as described above. We first assayed the cells for Gab1 expression, and for the ability of the mutant Gab1 F446/472/589 protein to become tyrosine phosphorylated in response to EGF treatment. As shown in Figure 2A , both wild type Gab1 and Gab1 F446/472/589 undergo tyrosine phosphorylation in response to EGF treatment (upper left panel). Quantitation of multiple experiments by densitometry reproducibly demonstrates that Gab1 F446/472/589 is tyrosine phosphorylated following EGF treatment to similar levels when normalized for Gab1 expression levels. We next subjected lysates from unstimulated or EGF stimulated cells to immunoprecipitation with anti-Gab1 antibodies followed by immunoblotting with anti-p85 antibodies. As has been demonstrated previously [ 20 ], wild type Gab1 readily coimmunoprecipitated p85 following EGF treatment, while the Gab1 F446/472/589 mutant protein failed to show an association with p85, confirming that the Gab1 F446/472/589 protein is deficient in p85 binding. Similar levels of Gab1 expression in these cells were confirmed by reprobing the Gab1 phosphotyrosine blot with anti-Gab1 antibodies (Figure 2A , bottom left panel). Additionally, total cell lysates of all Gab1 expressing cell lines described in this study were subjected to anti-Gab1 immunoblotting, providing independent evidence for similar levels of Gab1 expression across all cell lines (Figure 2A , right panel). Because the substrates of Shp2 are for the most part unknown, we were additionally interested in examining the state of EGFR tyrosine phosphorylation following treatment with EGF in order to determine if the failure of Gab1 to bind p85, and potentially recruit Shp2, would influence levels of EGFR autophosphorylation. However, stimulation with EGF for varying time intervals revealed no significant differences in the levels of autophosphorylation of EGFR in cells expressing wild type Gab1 versus the Gab1 F446/472/589 mutant (Figure 2B ). A linear representation of the EGF-induced EGFR tyrosine phosphorylation following normalization for EGFR expression levels is shown in Figure 2B (bottom). These results are consistent with our finding that p85 binding to Gab1 does not influence the recruitment of Shp2 to the Gab1 signaling complex, and are inconsistent with the conclusion that Gab1 mediates a PI-3 kinase-dependent recruitment of Shp2 [ 14 ]. We next explored the role of Gab1 in EGF-induced activation of the PI-3 kinase/Akt cell survival pathway utilizing the Gab1 F446/472/589 expressing cells. We first assayed the Gab1-associated PI-3 kinase activity directly through a PI-3 kinase assay. As shown in Figure 2C , immunoprecipitation of wild type Gab1 following EGF treatment brings down associated PI-3 kinase activity. However immunoprecipitation of Gab1 F446/472/589 is not associated with significant levels of PI-3 kinase in the presence or absence of EGF stimulation. In order to assay the effects of EGF stimulation on signaling downstream of PI-3 kinase, Gab1 -/- MEFs expressing no Gab1, wild type Gab1 or Gab1 F446/472/589 were treated with EGF over varying times and cell lysates were immunoblotted for serine-phosphorylated Akt. Mutation of the p85 binding sites on Gab1 essentially eliminated all EGF-induced Akt activation relative to cells expressing wild type Gab1 (Figure 2D ). The binding of p85 is absolutely required for Gab1-mediated activation of PI-3 kinase and Akt following EGF treatment [ 9 ]. Previous work has indicated that treatment of cells with PI-3 kinase inhibitors reduces levels of EGF-induced complex formation between Gab1 and the protein tyrosine phosphatase Shp2 [ 14 ]. This finding suggests a role for Gab1 in the PI-3 kinase-dependent recruitment of Shp2 following EGF stimulation. We have examined the possibility that mutation of the p85 binding sites on Gab1, which prevents PI-3 kinase activation, altered EGF-induced recruitment of Shp2 as compared to the recruitment of Shp2 by wild type Gab1. For this purpose, lysates from unstimulated or EGF-stimulated cells were subjected to immunoprecipitation with anti-Gab1 antibodies following immunoblotting with anti-Shp2 antibodies. The experiment presented in Fig 2A shows that mutation of the p85 binding sites on Gab1 did not affect recruitment of Shp2 by Gab1 following EGF stimulation (Figure 2A , fourth panel from the top). As has been previously observed [ 20 ], we noted a low level of basal association between Shp2 and both Gab1 and Gab1 F446/472/589 , which may be due to incomplete growth factor starvation prior to EGF stimulation in these experiments. Expression of Gab1 mutant protein deficient in Shp2 binding rescues EGF-induced PI-3 kinase/Akt activation in Gab1 -/- MEFs To assess the role of the protein tyrosine phosphatase Shp2 in Gab1-mediated signaling induced by EGF, two Gab1 mutants were generated and expressed in pools of Gab1 -/- MEFs. The first contained tyrosine to phenylalanine point mutations at the two binding sites for the Shp2 protein tyrosine phosphatase (Y627F/Y659F) (designated Gab1 F627/659 ) and the second contained mutations at the Shp2 binding sites as well as at the three PI-3 kinase binding sites described above (designated Gab1 F446/472/589/627/659 ). We first assayed the ability of the mutant Gab1 proteins to become tyrosine phosphorylated in response to EGF. As shown in Figure 3A (top panel) both wild type Gab1 protein and the Gab1 F627/659 protein readily undergo tyrosine phosphorylation when stimulated with EGF. Quantitation following densitometry indicates that Gab1 F627/659 is reproducibly tyrosine phosphorylated to levels approximately 1.5-fold higher than Gab1. This result suggests that Gab1 may be a substrate of Shp2, and that blocking Shp2 binding thereby increases EGF-induced Gab1 tyrosine phosphorylation. The Gab1 F446/472/589/627/659 mutant reproducibly displayed lower levels of tyrosine phosphorylation following treatment with EGF suggesting that these five tyrosines are the main phosphorylation sites on Gab1. Immunoprecipitation of cell lysates with anti-Gab1 antibodies followed by immunoblotting with anti-Shp2 antibodies demonstrates that wild type Gab1 forms a complex with Shp2 following EGF treatment, while the Gab1 F627/659 mutant proteins fail to show an association with Shp2 thus confirming that phosphorylation of Tyr627 and 659 is required for Shp2 binding (Fig 3A ). The basal interaction we observed between Gab1 and Shp2 in the absence of EGF stimulation (Figure 2A ) is absent in the Gab1 F627/659 mutant, even following prolonged exposures of the western blot. We did not detect a change in the tyrosine phosphorylation of EGFR in cells expressing Gab1 proteins that are deficient in recruitment of Shp2. The experiment presented in Fig 3B shows cells stimulated with EGF over varying periods of time and cell extracts assayed for levels of EGFR tyrosine autophosphorylation. As has been previously reported [ 15 , 18 , 20 ], recruitment of Shp2 by Gab1 does not alter the magnitude or kinetics of tyrosine autophosphorylation of EGFR (Figure 3B , left panels). Levels of EGFR autophosphorylation are represented linearly following quantitation by densitometry and normalization for protein expression levels (Figure 3B , bottom). Previous work with Shp2 -/- cells demonstrated an elevated and sustained activation of PI-3 kinase and Akt in response to EGF treatment, and it was proposed that Shp2 may act to dephosphorylate Gab1 at one or both of the p85 binding sites [ 18 ]. We utilized the Gab1 proteins deficient in Shp2 binding to assay more directly the role of the Shp2-Gab1 complex in mediating activation of PI-3 kinase and Akt in response to EGF stimulation. As shown in Figure 3C , immunoprecipitation of Gab1 F627/659 brings down 1.6-fold higher basal levels of PI-3 kinase activity relative to wild type Gab1 as assayed by PIP3 production. Importantly, Gab1 mutants defective for Shp2 binding show approximately 2-fold higher Gab1-associated PI-3 kinase activity in response to EGF treatment. Consistent with these findings, previous studies have demonstrated that cells transiently over-expressing Gab1 F627/659 bound more p85 [ 18 ]. In both Gab1 and Gab1 F627/659 expressing cells the Gab1-associated PI-3 kinase activity is augmented by EGF treatment. As expected, the additional mutation of the p85 binding sites eliminates Gab1-associated PI-3 kinase activity. In order to assay the effects of EGF stimulation on signaling downstream of PI-3 kinase, cells were treated with EGF for varying periods of time and cell lysates were assayed for Akt activation by immunoblotting with P-Ser473 Akt antibodies. Interestingly, cells expressing the Gab1 F627/659 protein reproducibly showed activation of Akt with significantly sustained kinetics relative to cells expressing wild type Gab1 (Figure 3D , left panels). As expected, the additional mutation of the p85 binding sites (Gab1 F446/472/589/627/659 ) limited Akt activation to levels similar to those observed in the Gab1 -/- cells, confirming the requirement for PI-3 kinase association with Gab1 to induce EGF-mediated activation of the Akt pathway. Taken together, these results suggest a role for Shp2 in negatively regulating the EGF induced activation of the PI-3 kinase pathway via Gab1, possibly by dephosphorylating Gab1 at p85 binding sites. Expression of ErbB3 in Gab1 -/- MEFs enhances activation of the PI-3 kinase signaling pathway As described above, PI-3 kinase is recruited to the EGFR via the adaptor protein Gab1. The results presented here demonstrate that Gab1 is required for EGF-induced activation of the PI-3 kinase pathway via the EGFR, presumably because this receptor does not contain binding sites for the p85 regulatory subunit of PI-3 kinase. The catalytically inactive ErbB3 receptor, however, contains at least six binding sites for p85 [ 5 ], and thus may bypass the requirement for Gab1 in response to EGF by heterodimerizing with the catalytically active EGFR. In order to test this hypothesis, retroviral vectors were used to introduce either the Gab1 or ErbB3 genes into Gab1 -/- MEFs that endogenously express the EGFR but not ErbB3, and pools of stable cell lines were selected for further analysis. We first assayed the ability of the endogenous EGFR to be tyrosine autophosphorylated in response to EGF, as well as the ability of the exogenous ErbB3 receptor to be tyrosine phosphorylated in response to stimulation with either EGF or neuregulin (NRG). The experiment presented in Fig 4A shows that all cell lines exhibit EGFR autophosphorylation in response to EGF treatment (Figure 4A , upper left panel), while only cells expressing the ectopically introduced ErbB3 protein show ErbB3 tyrosine phosphorylation in response to EGF stimulation. Interestingly, ErbB3 reproducibly shows constitutive low-level tyrosine phosphorylation that is augmented only 1.3-fold in response to EGF treatment. Cells expressing ErbB3 show tyrosine phosphorylation in response to treatment with NRG (Figure 4A , upper and middle right panels). Expression of EGFR and ErbB3 in the appropriate cell lines was confirmed by immunoblotting with antibodies specific for EGFR and ErbB3, respectively (Figure 4A , lower panels). The apparent decrease in EGFR expression in cells co-expressing EGFR and ErbB3 following EGF treatment was not observed in repetitions of this experiment, and is likely due to a stripping anomaly. Additionally, we demonstrated that Gab1 -/- MEFs that express wild type Gab1 display Gab1 tyrosine phosphorylation in response to EGF treatment, while Gab1 -/- control cells or those expressing ErbB3 do not show Gab1 phosphorylation (Figure 4B , upper panel). In order to test the ability of ErbB3 to rescue the EGF-induced activation of the PI-3 kinase/Akt signaling pathway in Gab1 -/- MEFs, we first assayed these cells for EGF-induced PI-3 kinase activity. Cells were either left unstimulated or were stimulated with EGF and cell lysates were immunoprecipitated with anti-phosphotyrosine antibody. Phosphotyrosine-associated PI-3 kinase activity was then assayed by analysis of PIP3 production. As shown in Figure 4C , both cells expressing wild type Gab1 and ErbB3 show PI-3 kinase activity, while Gab1 -/- cells do not. Interestingly, EGF induces PI-3 kinase activity to a greater degree in cells expressing Gab1 relative to cells expressing ErbB3. In order to assay the effects of EGF stimulation on signaling downstream of PI-3 kinase, cells were treated with EGF over varying periods of time and assayed for the presence of Ser473-phosphorylated Akt. Treatment of cells expressing either wild type Gab1 or ErbB3 with EGF induced rapid activation of Akt, although cells expressing wild type Gab1 reproducibly displaed higher levels of phosphorylated Akt with significantly sustained kinetics relative to Gab1 -/- cells expressing ErbB3 (Figure 4D , left panels). Equal loading and expression levels of Akt were confirmed by immunoblotting (Figure 4D , right panels). Cells expressing ErbB3 displayed activation of Akt in response to treatment with NRG at levels similar to or greater than that seen in Gab1-expressing cells following EGF treatment, with activation reproducibly observed at longer time points (Figure 4E , left panels). Again, equal loading and expression levels of Akt were confirmed by immunoblotting (Figure 4E , right panels). The phosphorylation of AKT in cells expressing Gab1, which shows modest enhancement following treatment with NRG, may be attributed to alternate signaling pathways including those mediated by ErbB2 and ErbB4. Involvement of these receptors was not explored in this study. Taken together, these results demonstrate that ErbB3 can partially compensate for Gab1 deficiency in EGF-induced activation of the PI-3 kinase/Akt signaling pathway, although Gab1-mediated activation appears to be more robust, and likely represents the primary mechanism by which EGF stimulates PI-3 kinase and Akt. While ErbB3 is relatively ineffective at mediating EGF-stimulation of PI-3 kinase activation, it is an efficient mediator of PI-3 kinase stimulation in response to NRG stimulation. Thus, EGF and NRG can stimulate PI-3 kinase activation in normal and transfected cells by means of Gab1 and ErbB3, respectively. Conclusions The results presented here demonstrate an absolute requirement for Gab1 in EGF-induced activation of the PI-3 kinase/Akt signaling pathway. Using this approach we demonstrated a strict requirement for association between Gab1 and p85 in EGF-induced PI-3 kinase activation, suggesting that Gab1 indeed provides an essential link between the EGFR and PI-3 kinase. Additionally, p85 binding does not play a significant role in Shp2 recruitment to the EGFR-Gab1 signaling complex, in contrast to previous studies [ 14 ]. We further demonstrated that the Gab1-Shp2 complex is responsible for the negative regulation of the strength and duration of PI-3 kinase/Akt signaling in response to EGF previously observed in Shp2 -/- cells. ErbB3 expression can bypass the requirement for Gab1 in EGFR signaling and can partially rescue EGF-induced activation of the PI-3 kinase/Akt cell survival pathway. This alternate pathway to PI-3 kinase activation may provide cells with a means of controlling either the strength or duration of PI-3 kinase signaling through differential expression of Gab1 and ErbB3, since the ErbB3-mediated response appears to be weaker. Thus Gab1 plays an essential role in bringing together a multi-protein signaling complex in response to EGF that modulates a critical aspect of cellular survival. Methods Expression constructs Expression vectors for wild type Gab1 and for Gab1Δp85 (Y446F/Y472F/Y589F) were previously described [ 21 ]. The cDNAs of Gab1 (pcDNA3-Gab1-WT), Gab1 F446/472/589 (pcDNA3-Gab1-3YF), and ErbB3 (pcDNA3-ErbB3) were subcloned into the mammalian retroviral vector pBabe containing a gene for puromycin resistance. The Gab1 F627/659 and Gab1 F446/472/589/627/659 mutants were generated by site-directed mutagenesis (Strategene) carried out on pBabe-Gab1-WT (to generate Gab1 F627/659 ) or pBabe-Gab1 F446/472/589 (to generate Gab1 F446/472/589/627/659 ) according to the manufacturers' specifications. Cell lines and culture Gab1-deficient (Gab1 -/-) mouse embryonic fibroblasts (MEFs) were obtained from Walter Birchmeier [ 22 ]. Wild type or Gab1-/- MEFs were maintained in DMEM supplemented with 10% fetal bovine serum (Gibco), 2 mM L-glutamine (Gibco), and 100 μg each of penicillin and streptomycin. The retroviral vectors described above (pBabe, pBabe-Gab1, pBabe-Gab1-Gab1 F446/472/589 , pBabe-Gab1-Gab1 F627/659 , pBabe-Gab1-Gab1 F446/472/589/627/659 , and pBabe-ErbB3) were used to transfect the amphitropic retroviral packaging cell line GPG (obtained from Joan Brugge) and high-titer viral stocks were used to infect Gab1 -/- MEFs. Cells were selected in medium supplemented with puromycin and pools of selected cells were used for further experiments. Prior to stimulation with EGF, cells were starved in serum-free medium. Immunoprecipitation and immunoblotting Cells were grown on 15-cm plates as described above to approximately 80% confluence and then starved overnight in DMEM without serum. Cells were left unstimulated or were stimulated with recombinant human EGF (Invitrogen) as indicated. Stimulations were halted by the addition of ice-cold PBS. Cells were washed in PBS and lysed in buffer containing 1% Triton X-100 as previously described [ 23 ]. For Gab1 immunoprecipitations, a mixture of polyclonal antibodies directed against both the N- and C-termini of Gab1 and cross-linked to Protein A Sepharose (Zymed) was incubated with cell extracts for 2–4 hours at 4°C. For p85 immunoprecipitations, a mixture of polyclonal antibodies directed against the N-terminal SH2 domain of p85 and the full-length p85 protein (Upstate) was incubated with cell extracts for 2–4 hours at 4°C. For EGFR immunoprecipitations, a polyclonal antibody directed against the C-terminus of the EGFR was incubated with cell extracts overnight at 4°C. For ErbB3 immunoprecipitations, a mixture of polyclonal antibodies directed against both the N- and C-termini of ErbB3 was incubated with cell extracts overnight at 4°C. Protein A Sepharose was added to immunoprecipitates (except anti-Gab1) and incubated for 1 hour at 4°C. Immunoprecipitates were then washed in buffer containing 0.1% Triton X-100, separated by SDS/PAGE, and transferred to nitrocellulose membranes (Bio-Rad). For Akt immunoblotting, total cell extracts were separated directly by SDS/PAGE and transferred to nitrocellulose membranes. Membranes were blocked for 1 hour or overnight in 5% BSA/TBS and immunoblotted as indicated. The anti-phosphotyrosine blotting was carried out with antibody 4G10 (Upstate). Anti-Gab1, anti-EGFR and anti-ErbB3 blotting was performed with the indicated polyclonal antibodies (Upstate). Anti-Shp2 blotting was carried out with polyclonal antibody (Santa Cruz). Anti-phosphoSer473-Akt and anti-Akt blotting were performed with the respective polyclonal antibodies (Cell Signaling Technology). Proteins were visualized by incubation with Enhanced Chemiluminescence (Amersham Pharmacia) according to the manufacturer's specifications. PI-3 kinase assay The PI-3 kinase assay was performed essentially as previously described [ 22 - 24 ]. Briefly, cells were serum-starved for 24 hours and stimulated with EGF as indicated, and cell extracts were prepared as described above. The lysates were immunoprecipitated using anti-Gab1 polyclonal antibodies or the monoclonal anti-phosphotyrosine antibody PY20 (Santa Cruz) for 2 hours at 4°C. Protein A Sepharose was incubated with immunoprecipitates for 1 hour at 4°C. The immunoprecipitates were washed three times with Buffer 1 (1X PBS, 1% NP-40), twice with Buffer 2 (0.5 M LiCl, 0.1 M Tris pH 7.5), twice with TNE (10 mM Tris pH 7.5, 100 mM NaCl, 1 mM EDTA pH 8.0), and twice with Buffer 4 (20 mM Hepes pH 7.5, 50 mM NaCl, 5 mM EDTA pH 8.0, 0.03% NP-40, 30 mM Tatrasodium Pyrophosphate (Sigma)). L-α-Phosphatidylinositol (Sigma) was added (10 μl of a sonicated solution at 10 mg/ml in 20 mM HEPES pH 7.5) and the reaction was initiated by the addition of 50 μl of the kinase buffer (20 mM Tris pH 7.5, 75 mM NaCl, 10 mM MgCl 2 10 μM ATP, 100 mM adenosine, 10 μCi [γ- 32 P] ATP) per sample. Samples were incubated at 30°C for 15 minutes, and the reaction was stopped by addition of 100 μl 1N HCl. Samples were extracted by addition of 200 μl CHCl 3 /CH 3 OH (1:1). The samples were vortexed and centrifuged, and the lower organic phases containing phospholipids were dried at 27°C for two hours. Samples were resuspended in 10 μl of PI-4-P standard (0.5 ml CHCl 3 , 0.5 ml CH 3 OH, 2.5 μl HCl, 1 mg L-α-Phosphatidylinositol 4-monophosphate (Sigma)) and subjected to thin layer chromatography (TLC plates – VWR) in CHCl 3 /CH 3 OH/NH 4 OH/H 2 O (45:35:7:3). The TLC plates were exposed in a PhosphorImager cassette for four days. Authors' contributions BL carried on the initial biochemical experiments before leaving the laboratory. DM continued this study and carried out the biochemical studies, the PI-3 kinase assays, and drafted the manuscript. IL participated in the PI-3 kinase assays. JS initiated and supervised the project. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC534114.xml
535544	Large bilateral adrenal metastases in non-small cell lung cancer	Background The adrenal gland is one of the common sites of metastasis from primary lung cancer. Adrenal metastases are usually unilateral however bilateral adrenal metastases are seen in 10% of all lung cancer patients; of these 2–3% occurs at the initial presentation of non-small cell lung cancer. Secondary tumors can disrupt the structure and function of the adrenal. This can lead to adrenal hemorrhage, which constitutes a life threatening hazard for the patient. Case presentation A 59-year-old male presented with persisting abdominal pain. His initial work-up revealed significant anemia, an invasive process in the right upper lobe of the lung and large masses of heterogeneous texture, with hemorrhagic and necrotic elements in both adrenal glands. A biopsy confirmed it to be a large-cell carcinoma of the lungs. The patient developed severe leukocytosis akin to the paraneoplastic syndrome and died suddenly five days after the administration of chemotherapy. Conclusion Intratumoral hemorrhage is a rare but life threatening complication of adrenal metastases and should be treated as soon as it has been diagnosed. If adrenalectomy is not feasible, combination chemotherapy should be applied as in metastatic disease. For choosing the appropriate chemotherapeutic regimen it is important to accurately achieve the diagnosis.	Background The adrenal gland is a common site for metastases in breast, lung and renal cell carcinomas, melanoma, and lymphoma [ 1 ]. Adrenal metastasis, at the initial diagnosis of non-small cell lung cancer, occurs in less than 10% of lung cancer patients [ 2 ]. Most cases involve solitary, unilateral, small asymptomatic lesions, incidentally discovered by CT-scan of the upper abdomen during a staging evaluation. Bilateral adrenal metastases are observed in less than 3% of patients with lung cancer. Again, most cases involve small, asymptomatic lesions. Intratumoral hemorrhage is a rare but serious complication of adrenal metastases. We report a case of non-small cell lung cancer with large bilateral adrenal metastases complicated with intratumoral hemorrhage and paraneoplastic leukemoid reaction. Case presentation A 59-year-old male, working in a quarry, was admitted to the hospital complaining of abdominal pain persisting for 45 days, afternoon fever, fatigue and 8 kg weight loss during the last month. Previous medical history was non contributory. He was a heavy smoker for 45 years. At first assessment, the patient presented with disseminated dry rales and decrease of vesicular murmur in the right upper lung field. Following abdominal examination, palpable smooth surface masses were discovered in both renal areas. Laboratory tests demonstrated a significant anemia with normocytic features (hematocrit 24%, Hb 7.2 mg/dl). Chest X-ray revealed a consolidated area in the right upper lobe of the lung. Abdomen ultrasound demonstrated large masses with foci of hemorrhagic necrosis in both adrenal glands (Figure 1 ). Figure 1 Abdomen ultrasound showing large masses with foci of hemorrhagic necrosis in both adrenal glands. Computed tomographic (CT) -scan revealed an invasive process in the right upper lobe of the lung (Figure 2 ) and large masses of heterogeneous texture, with hemorrhagic and necrotic elements within both adrenal glands (Figure 3 ). Figure 2 CT-scan of the chest showing a 60 × 53 mm invasive process in the right upper lobe of the lung. Figure 3 CT-scan of the abdomen showing large masses (98 × 85 mm right and 90 × 65 mm left) of heterogeneous texture, with hemorrhagic and necrotic elements within both adrenal glands. Magnetic resonance imaging (MRI) was performed which showed the masses with disseminated foci of high, heterogeneous signal intensity and pathologic enhancement after intravenous (i.v.) administration of contrast agent in both T 1 and T 2 weighted sequences. Radiological features were suggestive of adrenal metastasis from lung primary (Figures 4 , 5 ). Histological examination of the biopsy specimen confirmed the diagnosis of a large-cell carcinoma of the lung. The patient showed significant deterioration and was hospitalized again for altered level of consciousness, blurring of vision, tinnitus, lability and intense fatigue. Laboratory examination demonstrated severe leukocytosis with 98,000 polymorphonuclear leukocytes, compatible with the leukemoid reaction within the framework of a paraneoplastic syndrome. Hematocrit (Ht) levels dropped significantly a second time (Ht: 21%, Hb: 6.9 mg/dl), the biochemical parameters were normal, while the adrenocorticotropic hormone (ACTH) stimulation test excluded the possibility of cortical adrenal failure (serum cortisol basal value: 9μg/dl and 34μg/dl six hours following intramuscular (i.m.) administration of 1 mg of ACTH). A repeat abdominal ultrasound revealed numerous hemorrhagic elements inside the adrenal masses. The patient received urgent chemotherapy with paclitaxel and carboplatin, with simultaneous supportive treatment. Following chemotherapy, leukocytes decreased progressively, while symptoms related to the central nervous system (CNS) improved, however, the patient died suddenly on the fifth day following chemotherapy. Autopsy was not performed. The physician's on-call medical report documented a typical cardiac arrest, without evidence of acute hemorrhage in adrenal mass. Figure 4 T 2 weighted MR image showing disseminated foci of high heterogeneous signal intensity. Figure 5 T 1 weighted MR image showing enhancement after i.v. administration of contrast agent. Discussion Adrenal metastases in the natural history of a malignant neoplasm occur in 20 – 45% of cancer patients, depending on the localization of the primary site [ 3 ]. Up to 40% of patients with non-small cell lung cancer develop unilateral or bilateral adrenal metastases, as the carcinoma progresses [ 4 ]. The differential diagnosis of an adrenal mass incidentally discovered during imaging examination of the abdomen in patients with non-small cell lung cancer, usually comprises one or more of the characteristics detailed in Table 1 . Table 1 Differential diagnosis of incidentally discovered adrenal mass on Imaging • Nonfunctional adenoma • Adrenal metastasis • Primary carcinoma in adrenal glands • Adrenal cyst • Nonfunctional pheochromocytoma • Other causes (myelolipoma, lymphoma, aldosteronoma, neuroblastoma, pheochromocytoma) More than 90% of the cases comprising this involve a nonfunctional adenoma or an adrenal metastasis. Nonfunctional adenomas can be distinguished from adrenal metastasis with precision using modern imaging techniques, especially with MR imaging, where adenoma presents a low intensity signals in T2 sequences compared to hepatic parenchyma. In contrast, the adrenal metastasis presents higher intensity T2 sequence signals than that of the adenoma, relative to hepatic parenchyma. In CT-scan adenoma presents low intensity signals, below 10 Hounsfield units, compared to higher intensity signals in adrenal metastasis. Porte et al , observed that CT-scan technology has low specificity and sensitivity, which culminates in false positive and negative readings, given that 21% of true adenomas can be mistaken for metastases (density > 10 Hounsfield units) and 11% of true metastases may be mistaken for adenomas (density < 10 Hounsfield units) [ 5 ]. CT-scan has 100% sensitivity in detecting the metastatic nature of adrenal masses; however, 50% of adenomas may be mistaken as metastasis. Diagnostic accuracy of adrenal metastasis typically increased with detection of a primary tumor, as well as changes in the size and architecture of the parenchyma over time or during treatment. In equivocal cases, CT-scan-guided percutaneous biopsy of the adrenal mass is a safe and reliable method for the diagnosis of the lesion (sensitivity 81%, specificity 99%, potential complications 2.8%) [ 6 ]. In presence of adrenal metastasis the disease is staged as stage IV, which demonstrates the aggressive biological behavior and systemic dissemination of secondary lesions. These are treated as any other metastatic neoplasm and the appropriate chemotherapy is started. In select patients, where the primary lung site is surgically resectable (T1, T2 and maybe T3), with no involvement or the involvement of only the peribronchial and portal lymph nodes (N0, N1) and where the adrenal metastasis constitutes the unique indication of the disease, simultaneous adrenalectomy along with resection of the primary can increase overall survival [ 7 ]. Technically, this can be achieved with thoracoabdominal or transabdominal approaches, or laparoscopically in cases of small tumors (< 5 cm), [ 8 ]. A recent meta-analysis suggests that the latter technique, when it is feasible, is associated with fewer complications than open adrenalectomy [ 9 ]. However, adrenal metastases are usually larger than 5 cm, as in our case, which makes this approach inappropriate for general application. There is also a perceived increased risk of local tumor recurrence and intraperitoneal tumor dissemination occurring after laparoscopic resection of malignant adrenal tumors [ 10 ]. The most important complication of adrenal metastases is intratumoral hemorrhage and only nine cases have been reported previously [ 11 ]. When hemorrhage risk is high, as in the case of large metastasis, timely adrenalectomy is a necessary intervention. For resecting larger adrenal tumors in presence of bilateral involvement the transabdominal approach with extended subcostal incision (Kocher) is preferred. When this is considered technically impossible, radiation therapy can be an alternative palliative approach. During the course of the disease, our patient developed a leukemoid reaction and the performance status dropped remarkably, making it difficult to proceed with the planned surgical resection. Thus, the paclitaxel and carboplatin chemotherapeutic regimen was administered in an attempt to decrease the leucocytes count and improve the CNS symptoms. Even though the infusion of paclitaxel has been associated with cardiovascular events, there is no clear relation between the drug administration and the sudden death of our patient. A less common complication of bilateral adrenal metastases is primary adrenal insufficiency, caused by the destruction of adrenal gland architecture and function due to the development of the tumor. The main symptoms resulting from the decreased production of steroid hormones often overlap with the general indicators caused by the primary site, hindering the proper treatment of these patients. Primary adrenal insufficiency is diagnosed when serum cortisol basal values are low (<5μg/dl), and particularly when the values fail to increase within six hours following the i.m. administration of 1 mg of tetracosactride, the synthetic analog of ACTH. The treatment comprises daily administration of glucocorticoid and mineralocorticoid for life. Conclusion Bilateral adrenal metastases can be found in a small number of patients with non-small cell lung cancer. Early identification of these lesions is necessary, since adrenalectomy may increase survival in select patients. Intratumoral hemorrhage is a rare but life threatening complication and should be treated as soon as it has been diagnosed. If adrenalectomy is not feasible, combination chemotherapy should be applied. Competing interests The authors declare that they have no competing interests. Authors' contributions CK – conceived the study, did the literature search and coordinated the preparation of the manuscript for submission AT – helped in literature search, reviewed and edited the manuscript for final submission SM and KV – helped in literature search and drafted the manuscript for submission All authors read and approved the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535544.xml
544395	The relevance of genetic analysis to dairy bacteria: building upon our heritage	Lactic acid bacteria (LAB) are essential for the manufacture of fermented dairy products. Studies on the physiology, biochemistry and genetics of these microorganisms over the last century have contributed considerably to the improvement of fermentation processes and have resulted in better and safer products. Nevertheless, the potential of LAB is far from being maximized. The sophistication of biotechnologies and the availability of complete genome sequences have opened the door to the metabolic engineering of LAB. In this regard, the recent publication of the complete genome sequences of two Streptococcus thermophilus strains will provide a key tool to facilitate the genetic manipulation of this important dairy species.	Introduction The souring of milk by microorganisms has been used for thousand of years as a natural preservation procedure [ 1 ]. The method was domesticated to manufacture man-made edible fermented products such as cheese, which appeared in the human diet some 8,000–10,000 years ago. However, up to the beginning of the 20th century, milk fermentation was by no means a controlled process, and the search for improvements was strictly empiric and based on trial and error. The discovery and characterization of lactic acid bacteria (LAB) has tremendously modified the way fermented milk products are made. Considerable efforts have been devoted over the last fifty years to increase our knowledge about the genetics, biochemistry, and physiology of LAB. In addition to enhancing our understanding of microbial life, these studies have allowed dairy microbiologists and cheesemakers to select better strains and improve the productivity, quality, and safety of the final products. The characterization of LAB has promoted the rational development of blends of defined bacterial strains, now known as starter cultures, which are increasingly replacing the undefined blends traditionally used by the dairy industry. This strategy makes it easier to control acid production and, to some extent, phage problems. Unfortunately, it is also widely perceived that this approach reduces the much sought-after rich, complex flavour of fermented products made with undefined cultures. Thus, the search for LAB strains that produce characteristic aroma and flavours remains an attractive challenge for the dairy industry. The increasing awareness of the population of the importance of a healthy diet has led scientists to revisit the century-old hypothesis that some specific fermented dairy products may provide health benefits. A plethora of studies are underway to evaluate this concept and to improve the health benefits of LAB. Also, the long history of the safe use of these bacteria has made them very attractive candidates as vaccine delivery vehicles [ 2 , 3 ]. Lastly, thanks to their relatively small genome and simple metabolism, LAB can be exploited as food-grade cell factories to produce molecules of industrial and therapeutic interest. To meet some of these challenges and promises, it became apparent that metabolic engineering was the best approach. Owing to the industrial interest, the sound knowledge of its genetic and the availability of genetic tools, metabolic engineering of LAB has been carried out mainly with Lactococcus lactis . The publication of the complete genome sequence of L. lactis in 2001 [ 4 ] was a boost to speed up achievements in rerouting metabolic fluxes with this LAB. For instances, it became possible to increase the production of diacetyl and to transform this mesophilic bacterium from homolactic fermenter to homoalanine fermenter. The innovative production of vitamins by L. lactis via genetic manipulation [ 5 , 6 ] is a convincing example of how metabolic engineering of LAB can be beneficial for consumers. A number of reviews [ 7 - 12 ] provide a more comprehensive discussion of the metabolic engineering of L. lactis . Recently, the complete genome sequence of two Streptococcus thermophilus strains were published [ 13 ] and a third one is nearing completion [ 14 ]. S. thermophilus is a thermophilic LAB species widely used for the manufacture of yogurt and cheeses that require elevated cooking temperatures such as Swiss and Italian types. The availability of these genome sequences will make it possible to apply genomic-related analytic methods such as functional and comparative genomics, microarray technologies, proteomics, and bioinformatics to this species. These genomic-based technologies will undoubtedly accelerate the metabolic engineering of S. thermophilus , which is still at its infancy. We will briefly discuss current studies that will greatly benefit from these S. thermophilus genomic sequences: food-grade vectors, optimization of lactose metabolism and exopolysaccharide (EPS) production, and resistance to bacteriophages. Discussion Food-grade vectors Metabolic engineering often involves inactivation or overexpression of relevant chromosomal genes or stringent control of an extrachromosomal foreign gene borne by a plasmid artificially introduced into the host. Plasmids thus play a central role in these studies because they are the primary vehicles used to manipulate target DNA sequences. However, these extrachromosomal genetic elements are naturally but infrequently observed in S. thermophilus strains[ 15 ]. Understandably, very few cloning tools are at hand, and those that are available are based on similar replication machinery. This lack of diversity poses significant problems for many genetic studies. While several types of transformation methods have been developed to introduce foreign DNA into S. thermophilus , an efficient, stable, food grade expression system is still lacking. The recent discovery of a novel theta-type replicating plasmid [ 15 ] as well as suitable selection markers effective in S. thermophilus , such as galactokinase [ 16 ] and alpha-galactosidase [ 17 ], suggest that help is on the way. The complete S. thermophilus genome sequences also provide crucial information on transcription and translation initiation signals and the codon usage of this bacterium. This may help in the design of efficient expression systems. Lactose metabolism The ability of S. thermophilus to rapidly take up and metabolize lactose is crucial in several fermentation processes. While S. thermophilus readily metabolizes the glucose moiety of lactose, it is unable to metabolize galactose, which is expelled into the external medium. The presence of galactose in yogurt and other dairy products might be unwanted for different reasons. Notably, galactose is poorly metabolized by humans and may cause, under certain conditions, health problems [ 18 , 19 ]. The release of galactose by S. thermophilus results either from poor expression of the gal operon [ 20 ] or inefficient translation of the galK gene coding for galactokinase [ 16 , 21 ]. The finding that the inability to grow on galactose was a consequence of poor galK translation was first suggested by a comparative analysis of the S. thermophilus gal - lac operons with the homologous genes from the phylogenetically related oral bacterium Streptococcus salivarius . This is a convincing example of how comparative genome analysis may help to decipher specific metabolic pathways of LAB. While the introduction of a functional extrachromosomal S. salivarius galK allele in S. thermophilus allows it to grow on galactose, it does not prevent the expulsion of galactose during growth on lactose [ 16 ]. In this context, it is noteworthy that 10% of S. thermophilus genes are not functional (pseudogenes) and that one third of these pseudogenes are dedicated to sugar metabolism [ 13 ]. It would be interesting to determine whether reactivation of one or several pseudogenes that were originally involved in sugar metabolism could prevent or decrease galactose expulsion during growth on lactose. In this context, the availability of the complete sequence of the S. salivarius genome would be welcome since many S. thermophilus pseudogenes involved in sugar transport are functional in S. salivarius [ 13 ]. Exopolysacharides Some strains of S. thermophilus are widely used for the commercial manufacture of yogurt because they produce exopolysaccharides (EPS), which give a viscous texture to the fermented dairy product. Other EPS-producing S. thermophilus strains can also enhance the functional properties of some cheeses such as Mozzarella. The organoleptic properties of these products are largely due to the amount and types of exopolysaccharide produced during the fermentation process. The yield and sugar constituents of EPS are influenced by several factors and vary from strain to strain [ 22 ]. S. thermophilus does not naturally produce large amounts of EPS, which explains why considerable efforts are being directed toward understanding the cellular mechanisms of EPS biosynthesis. Since low levels of sugar precursors may limit EPS synthesis, a better understanding of S. thermophilus metabolic flux during sugar metabolism may lead to new strategies to enhance EPS production [ 22 ]. Levander et al. [ 23 ] showed that such knowledge can be used to enhance EPS production in S. thermophilus by metabolic engineering of central carbon metabolism. Moreover, because EPS yields are growth associated, efforts to increase production levels are likely to require novel strategies to enhance biomass production. This task clearly requires a comprehensive view of the cell machinery. The sequencing of a whole genome is a mandatory step to achieve this goal. As our knowledge of S. thermophilus continues to improve, novel EPS and/or applications for EPS + cultures are likely to emerge. Phage resistance Bacteriophages are the most significant cause of fermentation failures in the dairy industry worldwide. Dairy microbiologists have attempted for more than 70 years to eliminate, or at least bring under better control, the bacteriophages that interfere with the manufacture of many fermented milk products [ 24 ]. The publication of the complete genomes of two S. thermophilus hosts should provide new insights in several areas of phage research. The development of bacteriophage-insensitive S. thermophilus mutants is generally the first approach used to transform a phage-sensitive strain into a phage-resistant mutant, most likely following spontaneous chromosomal mutation in the gene encoding the phage receptor. Although progress has been made in identifying phage proteins involved in S. thermophilus host recognition [ 25 ], the identification of the phage receptors on the cell surface has remained elusive. Based on the results of genome data mining [ 26 ], a number of potential receptors can now be experimentally verified. At least eight complete S. thermophilus phage genomes are now available [ 27 , 28 ]. Phage research has thus already entered into the post-genomic era. Microarrays covering the two main groups of S. thermophilus phages are already available[ 29 ], meaning that it is now possible to design a complete array containing host and phage genes to study phage-host interactions on a novel and global scale during the infection process. Conclusions A thorough understanding of LAB metabolism and how it is regulated by external stimuli is a prerequisite for maximizing the potential of LAB. The availability of complete S. thermophilus genome sequences will obviously facilitate our understanding of the metabolic potential of S. thermophilus . It will also make it easier to design rational genetic manipulations of this important dairy bacterium in order to produce added value cheeses and yogurts and to use it as a cell factory. In addition, knowing the complete genome sequences should lead to the development of new genetic tools that will provide insights into the evolution of microbial communities, shifts in metabolism, and how each member adapts to the environmental changes that occur during complex fermentation processes. The exciting biotechnological developments in LAB and studies that are already underway will benefit both consumers and the dairy industry. The availability of the complete S. thermophilus genome sequence have opened up exciting, new possibilities that will build on an already rich heritage.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544395.xml
534100	Thymus-derived glucocorticoids are insufficient for normal thymus homeostasis in the adult mouse	Background It is unclear if thymus-derived glucocorticoids reach sufficient local concentrations to support normal thymus homeostasis, or if adrenal-derived glucocorticoids from the circulation are required. Modern approaches to this issue (transgenic mice that under or over express glucocorticoid receptor in the thymus) have yielded irreconcilably contradictory results, suggesting fundamental problems with one or more the transgenic mouse strains used. In the present study, a more direct approach was used, in which mice were adrenalectomized with or without restoration of circulating corticosterone using timed release pellets. Reversal of the increased number of thymocytes caused by adrenalectomy following restoration of physiological corticosterone concentrations would indicate that corticosterone is the major adrenal product involved in thymic homeostasis. Results A clear relationship was observed between systemic corticosterone concentration, thymus cell number, and percentage of apoptotic thymocytes. Physiological concentrations of corticosterone in adrenalectomized mice restored thymus cell number to normal values and revealed differential sensitivity of thymocyte subpopulations to physiological and stress-inducible corticosterone concentrations. Conclusion This indicates that thymus-derived glucocorticoids are not sufficient to maintain normal levels of death by neglect in the thymus, but that apoptosis and possibly other mechanisms induced by physiological, non stress-induced levels of adrenal-derived corticosterone are responsible for keeping the total number of thymocytes within the normal range.	Background Although it is clear that elevated concentrations of endogenous glucocorticoids can cause apoptosis in the thymus [ 1 - 3 ], the role of normal concentrations of glucocorticoids in thymic homeostasis remains controversial [ 4 - 7 ]. Results reported by Ashwell and colleagues suggest glucocorticoids are essential at very low concentrations for early development and survival of thymocytes and that glucocorticoids can alter the sensitivity of more mature thymocytes to positive selection, thereby influencing the T cell receptor repertoire [ 5 ]. In addition, there is convincing evidence that corticosterone is produced in the thymus and that it acts locally to affect thymocyte development [ 8 , 9 ]. Therefore, it was surprising when normal cellular development (including repertoire) was observed until the time of birth in glucocorticoid receptor knockout mice [ 10 ]. This raised serious questions about the necessity of glucocorticoids as a required or permissive agent in thymic development. It has also been suggested that glucocorticoids play a role in homeostasis in the adult thymus by inducing death by neglect of thymocytes that are neither positively nor negatively selected. This idea has been based on the observation that the predominant cell type subjected to death by neglect, CD4 + CD8 + non-mature thymocytes, is most susceptible to elevated concentrations of glucocorticoids [ 11 ]. Recent results indicate that overexpression of glucocorticoid receptors (GR) in developing and mature T cells leads to decreased cell number in the thymus and a decreased number of T cells in the periphery in adult mice. In addition, decreased expression of GR is associated with increased cell number in the thymus [ 4 ]. However, results obtained with knockout or transgenic mice have been contradictory [ 4 , 7 , 12 - 14 ]. For example, one group using transgenic mice that express anti-sense GR mRNA in the thymus found increased thymus cellularity [ 4 ], whereas another group using a cre-lox conditional knockout system to eliminate glucocorticoid receptor in cells that express CD4 (including double positive cells) reported no increase in cellularity [ 14 ]. Both groups verified that the expected decrease in sensitivity to high concentrations of glucocorticoids occurred in the transgenic mice. Until the basis for such differences can be determined, it seems reasonable to use an alternate approach that does not alter the glucocorticoid receptor (except by natural mechanisms relating to glucocorticoid concentration). In addition, transgenic approaches cannot provide the concentration-response information for corticosterone that would be needed to distinguish normal physiological effects and stress-related effects. A small number of studies have been reported in which systemic glucocorticoid concentrations were reduced by adrenalectomy, leading to increased numbers of cells in the thymus [ 15 - 17 ]. However, this observation has not been universal, with one report indicating no increase in thymus cellularity in adrenalectomized mice [ 18 ]. Thus, confirmation of an adrenalectomy-induced increase in thymus cellularity would be useful. Even if confirmation is obtained, it would still be possible that an adrenal product other than corticosterone was responsible for increased cell number in the thymus. However, if corticosterone was the major regulator of thymus homeostasis, restoring corticosterone to physiological levels in adrenalectomized mice should return thymus cell number and subpopulation ratios to normal values. Therefore, this approach was used in the present study to determine the role of systemic corticosterone in thymus homeostasis. The study was designed so the results would also indicate whether thymus-derived corticosterone is sufficient to permit normal maintenance of number of cells in each major subpopulation in the thymus. If physiological (adrenal-derived) concentrations of corticosterone are important in the induction of death by neglect of thymocytes, it would seem likely that any increase in cell number in the thymus of adrenalectomized mice would be explained mostly by an increase of CD4 + CD8 + cells, which are the predominant cell type subjected to death by neglect [ 11 ]. In addition, it has also been proposed that physiological concentrations of corticosterone increase the sensitivity of thymocytes to negative selection. Preventing this would presumably cause fewer single positive thymocytes to die, thus increasing the percentages of these cells in the thymus. It might also be expected that immature single positive thymocytes (CD3 low, CD4 - CD8 + ) would be increased, as these cells have been reported to be particularly sensitive to glucocorticoids [ 19 ]. If physiological (non-stress) concentrations of corticosterone contribute to the induction of death by neglect or negative selection, it would be expected that sub-physiological concentrations of corticosterone would decrease apoptosis in the thymus. Failure to observe these changes in mice with sub-physiological concentrations of corticosterone would suggest either that thymus-derived corticosterone [ 8 ] is sufficient to compensate for loss of systemic (adrenal-derived) corticosterone or that corticosterone is not directly involved in these processes under physiological, non-stress conditions. The studies described here were designed to directly test these predictions and thus to indirectly evaluate the role of endogenous glucocorticoids in death by neglect in the thymus. In addition, this study was designed to distinguish the relative contributions of systemic (mostly adrenal-derived) glucocorticoids and those produced in the thymus [ 8 ]. The use of a dose-response approach permitted identification of the point on the corticosterone concentration vs. thymocyte subpopulation plot that corresponds to a physiological corticosterone concentration, and it permitted identification of a distinction between the effects of sub-physiological concentrations of corticosterone and stress-inducible concentrations. Results and discussion Adrenalectomy increases cell number and alters subpopulation percentages in the thymus, and this effect is inhibited 24 hr after restoration of corticosterone The results shown in Figure 1 demonstrate that adrenalectomy significantly increases the total number of cells in the thymus as well as the number of CD4 + CD8 - and CD4 + CD8 + cells. The number of CD4 - CD8 + and CD4 - CD8 - cells was also greater in adrenalectomized mice than in the naive control group, but the difference was not significant. This indicates that the number of CD4 + CD8 - and CD4 + CD8 + cells is diminished to a greater extent than CD4 - CD8 + and CD4 - CD8 - cells by normal, physiological concentrations of corticosterone. Results for adrenalectomized mice were comparable whether a placebo pellet was implanted or not, indicating that non-adrenal-derived stress mediators induced by pellet implantation did not affect cell number. Timed release corticosterone pellets were used to restore corticosterone in adrenalectomized mice, and 0.5 and 1.5 mg pellets had only minimal effects on any subpopulation 24 hours after implantation of pellets (Figure 1 ). Pellets containing 2.5 mg of corticosterone returned the number of CD4 + CD8 - and CD4 + CD8 + cells to near normal values. A pellet containing 5.0 mg of corticosterone significantly decreased the number of CD4 - CD8 - and CD4 + CD8 + cells as compared to the naive control group, suggesting that CD4 - CD8 - and CD4 + CD8 + are more sensitive than CD4 + CD8 - and CD4 - CD8 + cells to stress-inducible corticosterone concentrations. The greater sensitivity of CD4 + CD8 + cells as compared to the other subpopulations at high corticosterone concentrations has prompted an assumption that these cells are more sensitive to physiological (unstressed) concentrations of corticosterone. However, this was not supported by the results presented here. The results in Figure 1 do indicate that CD4 + CD8 + cells increase in number to a greater extent than CD4 - CD8 + or CD4 - CD8 - cells in ADX mice, but CD4 + CD8 - cells increase in number proportionally more than all these other subpopulations in mice with sub-physiological concentrations of corticosterone (ADX groups). Figure 1 Effect of adrenalectomy and restoration of various concentrations of corticosterone on the number of nucleated cells in the thymus 24 hours after implantation of corticosterone pellets Treatment groups were: Naive, untreated; ADX-N, adrenalectomized naive; ADX-P, adrenalectomized with placebo pellet implanted; other groups, adrenalectomized with corticosterone pellets of the indicated size implanted. The thymus was evaluated 24 hours after pellet implantation (~3–4 weeks after ADX). Group size was 5–12. Values shown are mean ± S.E. obtained by normalizing groups to the mean value for the Naive group (defined as 100%). Results shown are pooled from two independent experiments, and groups significantly different from the naive group (by ANOVA followed by Dunnett's test) are shown by * ( p < 0.05) or ** ( p < 0.01). The changes in subpopulation percentages (as opposed to cell numbers, which are shown in Figure 1 ) in the thymus 24 hours after implantation of pellets are shown in Figure 2 . These results confirm the greater sensitivity of CD4 + CD8 + cells than the other subpopulations to stress-inducible concentrations of corticosterone (in mice with a 5 mg pellet). The greater sensitivity of CD4 - CD8 - cells than CD4 - CD8 + and CD4 + CD8 - cells noted in Figure 1 , was also evident in Figure 2 in terms of a lesser increase in percentage of the former cell type as compared to the latter cell types in mice treated with a 5 mg pellet. However, it should be emphasized that all subpopulations decreased in absolute number in these mice, so these percentage values do not reflect increases in the number of cells in these subpopulations but increases relative to CD4 + CD8 + cells, which are the most abundant and were diminished to the greatest extent. No significant change in subpopulation percentages was caused by adrenalectomy with or without a placebo pellet. However, it should be noted that the results shown in Figure 1 indicate that the number of CD4 + CD8 + and CD4 + CD8 - cells increased to a greater extent than the number of CD4 - CD8 - or CD4 - CD8 + cells. This did not result in a substantial change in percentages (Figure 2 ), because the cells that increased the most (CD4 + CD8 + cells) account for over 90% of the total thymocyte population, and changes in this subpopulation were reflected in the denominator (total cell number) of the equation that determines the percentage of cells in each population. Figure 2 Effect of adrenalectomy and restoration of various concentrations of corticosterone on the percentage of cells in the 4 major subpopulations in the thymus 24 hours after implantation of corticosterone pellets Data from the experiments noted in Figure 1 were analyzed for changes in the percentage of each cell type as compared to the total number of nucleated cells per thymus. The values did not differ significantly between experiments for any subpopulation, so the data were pooled (without normalizing). Groups significantly different from the naive group (by ANOVA followed by Dunnett's test) are shown by * ( p < 0.05) or ** ( p < 0.01). Adrenalectomy increases cell number and alters subpopulation percentages in the thymus, and this effect is inhibited 72 hr after restoration of corticosterone Changes in the thymus were first evaluated 24 hours after implantation of pellets to allow determination of the role of apoptosis in loss of cells at a time point at which cell numbers were still decreasing. To determine if greater or different effects were evident after a longer period of corticosterone exposure, thymuses were evaluated 72 hours after implantation of pellets. As shown in Figure 3 , the effects on cell number and on the number of cells in most subpopulations were greater than observed after 24 hours (Figure 1 ). For example, mice with a 2.5 mg corticosterone pellet had significantly fewer total thymocytes and significantly fewer cells in all but one of the major subpopulations than the naive group, whereas such decreases were only observed in the group with a 5 mg pellet after 24 hours (Figure 1 ). Similarly, the changes in subpopulation percentages were more pronounced after 72 hours (Figure 4 ) than after 24 hours (Figure 2 ). Some of the increases in cell number in adrenalectomized naive mice (ADX-N) or adrenalectomized mice with a placebo pellet implanted (ADX-P) compared to naive control that were observed in Figure 1 were not significant in the 72 hour experiment (Figure 3 ). It should be noted that these groups were essentially equivalent in all experiments, because adrenalectomy occurred 3 weeks before analysis in all mice. Evaluation of pooled, normalized data from 4 independent experiments indicates that compared to naive mice (100 ± 4.3%), ADX-N (143 ± 5.2%) and ADX-P (144 ± 8.3%) groups had significantly more total thymocytes. The mean number of nucleated cells per thymus in the naive (non-adrenalectomized) groups was 8.1 × 10 7 in these 4 experiments. Figure 3 Effect of adrenalectomy and restoration of various concentrations of corticosterone on the number of nucleated cells in the thymus 72 hours after implantation of corticosterone pellets Treatment groups were: Naive, untreated; ADX-N, adrenalectomized naive; ADX-P, adrenalectomized with placebo pellet implanted; other groups, adrenalectomized with corticosterone pellets of the indicated size implanted. The thymus was evaluated 72 hours after pellet implantation (~3–4 weeks after ADX). Group size was 5–12. Values shown are mean ± S.E. obtained by normalizing groups to the mean value for the Naive group (defined as 100%). Results shown are pooled from two independent experiments, and groups significantly different from the naive group (by ANOVA followed by Dunnett's test) are shown by * ( p < 0.05) or ** ( p < 0.01). Figure 4 Effect of adrenalectomy and restoration of various concentrations of corticosterone on the percentage of cells in the 4 major subpopulations in the thymus 72 hours after implantation of corticosterone pellets Data from the experiments noted in Figure 3 were analyzed for changes in the percentage of each cell type as compared to the total number of nucleated cells per thymus. Groups significantly different from the naive group (by ANOVA followed by Dunnett's test) are shown by * ( p < 0.05) or ** ( p < 0.01). Serum corticosterone concentrations indicate that corticosterone replacement with pellets yields appropriate corticosterone concentrations and that sub-physiological corticosterone concentrations (in ADX mice) are not sufficient to maintain thymus homeostasis The serum corticosterone concentration measured in naive mice was 185 ng/ml (Figure 5A ). Normal corticosterone values in female B6C3F1 mice vary from less than 100 ng/ml to approximately 300 ng/ml in a circadian pattern [ 20 ]. Adrenalectomized mice had barely detectable concentrations of corticosterone in the serum, and corticosterone pellets containing increasing amounts of corticosterone produced increasing corticosterone concentrations in the serum. These results were obtained 24 hours after implantation of pellets. These serum corticosterone concentrations range from values that can be found in normal (unstressed) mice at certain times of the day [ 20 ] (observed in groups treated with 0.5 and 1.5 mg pellets) to values that are comparable to those measured in mice exposed to moderately or highly stressful conditions [ 20 , 21 ] (observed in mice treated with 2.5 or 5.0 mg pellets). Non-linear curve fitting was used to extrapolate the size of pellet that would be required to produce the same corticosterone concentration measured in naive mice. The extrapolated value (1.75 mg) is shown in Figure 5B . Thus, the effects of adrenalectomy on the thymus should not quite be reversed by a 1.5 mg pellet and should be more than reversed by a 2.5 mg pellet. The results shown in Figure 3 are reasonably consistent with this expectation. It seems appropriate to examine the 72-hour data in this regard (Figure 3 ), because this exposure period most likely represents the time required to achieve the maximum effects of corticosterone on the thymus. The slight differences from expected effects for some subpopulations may be related to the fact that pellets do not reproduce the circadian changes in corticosterone that occur in normal animals. These changes likely contribute to thymus homeostasis, and restoring corticosterone to a constant concentration may not mimic this effect precisely. Nevertheless, these results clearly demonstrate that corticosterone alone is able to reverse the effects of adrenalectomy and that this reversal occurs at concentrations within physiological levels. This suggests that corticosterone is the only adrenal product required to regulate thymocyte number. More importantly, these results conclusively demonstrate that corticosterone produced in the thymus is not sufficient to maintain normal thymocyte numbers. Figure 5 Serum corticosterone concentration 24 hours after implantation of corticosterone pellets Mice from one of the experiments noted in Figures 1 and 2 were bled prior to removal of the thymus, and serum corticosterone concentrations were determined by radioimmunoassay. In panel A, values shown are means ± SE (n = 5 mice/group), and values significantly different from the naive control are indicated by * ( p < 0.05) or ** ( p < 0.01). The treatments are described in the legend for Figure 1. In panel B, non-linear regression was used to determine a best-fit line through points from 0 (ADX-N) to 5 mg corticosterone pellets (open squares). The extrapolated pellet size required to produce the same corticosterone concentration shown for Naive mice in panel A is indicated by an open circle. Dotted lines indicate the 95% confidence interval for the regression line. The R-squared value for this relationship was 0.987. Differential sensitivity of thymic subpopulations as indicated by linear regression analysis It is clear that thymic subpopulations differ in their sensitivity to high (stress-inducible) corticosterone concentrations. However, linear regression analysis (Figure 6 ) indicates that the sensitivities of the various cell populations are not as different as might be expected on the basis of the decreased percentage of CD4 + CD8 + cells and the increased percentages of the other subpopulations. The results demonstrate that increasing concentrations of corticosterone affect CD4 - CD8 - and CD4 + CD8 + similarly, but the slope for CD4 + CD8 + cells is significantly greater (by the method of Zar as implemented by Prism software) than for CD4 - CD8 - cells, indicating slightly greater sensitivity to high concentrations of corticosterone. The effect of increasing concentrations of corticosterone on CD4 + CD8 - cells was significantly less than the effect on CD4 + CD8 + cells with regard to the slopes of the respective lines. The slope for CD4 - CD8 + cells was less than the slope for all other subpopulations, suggesting a lower sensitivity of these cells to corticosterone across the whole range of concentrations. However, the decrease for CD4 + CD8 - cells and CD4 - CD8 + cells relative to naive control was essentially the same in mice treated with a 5 mg pellet, suggesting that the difference in sensitivity is minimal as the corticosterone concentration increases. Although non-linear models would likely have given better correlation coefficients than linear ones for some of these data, linear models facilitate comparison. In addition, the Runs Test was conducted for all linear regression analyses, and the non-linear component was not significant for any of these data. Figure 6 Linear regression analysis of corticosterone pellet size (mg) and the number of cells per thymus for each of the 4 major subpopulations The thymus was evaluated 24 hours after pellet implantation (~3–4 weeks after ADX), and all groups shown were ADX. Group size was 5–12. Values shown are mean ± S.E. obtained by normalizing groups to the mean value for the Naive group (defined as 100%). Results shown are pooled from two independent experiments, and statistically significant differences in the slope and intercept of each pair-wise combination are described in the text. Role of changes in the rate of apoptosis in decreased and increased cell number in the thymus of ADX mice with or without a corticosterone pellet There are several mechanisms by which corticosterone might act to alter the number of cells in each subpopulation in the thymus. Although induction of apoptosis is generally regarded to be the major mechanism [ 1 - 3 , 22 ], altering differentiation or proliferation of thymocytes, altering the development of pro-thymocytes in the bone marrow, or altering pro-thymocyte or thymocyte trafficking to or from the thymus are all possible mechanisms. To determine if changes in apoptosis may play a role in the effects noted in this study, apoptosis was evaluated using two criteria: cell size (indicated by forward scatter) and TUNEL labelling for DNA fragmentation. As shown in Figure 7 , the results demonstrate that the percentage of apoptotic cells in the thymus was decreased in mice that had sub-physiological concentrations of corticosterone in the serum (ADX-naive, ADX-placebo, and ADX mice with a 0.5 mg corticosterone pellet). The percentage of apoptotic cells was substantially increased in mice with a high (stress-inducible) concentration of corticosterone in the serum (caused by a 5 mg pellet). Figure 7 Apoptosis in the thymus in response to sub-physiological and stress-inducible concentrations of corticosterone Apoptosis was evaluated using the TUNEL technique with fluorescein-labelled dUTP. The cells were also surface labelled with fluorescent-labelled antibodies to CD4 and CD8. Values shown represent means ± SE for groups of 5 mice. The values in the upper panel represent the percentage of apoptotic cells in the thymus. The other values represent the percentage of apoptotic cells that are within each of the major 4 subpopulations in the thymus. Groups significantly different from the naive group (by ANOVA followed by Dunnett's test) are shown by * ( p < 0.05) or ( p < 0.01). Initially, changes in the percentage of apoptotic cells in the various subpopulations did not seem entirely consistent with the changes in the percentages of each subpopulation in the thymus (compare Figure 2 and Figure 7 ). Three-color flow cytometry allowed determination of the percentage of apoptotic cells in each subpopulation. In naive mice, most apoptotic cells were CD4 + CD8 + , as expected, with a substantial percentage in the CD4 - CD8 - category and lesser percentages of the mature single positive categories. The percentages of apoptosis in all populations increased in mice treated with a 5.0 mg corticosterone pellet, except CD4 + CD8 + cells, for which the percentage decreased. This almost certainly reflects the fact that this subpopulation was depleted by 24 hours of elevated corticosterone concentrations (Figure 1 ), and at least some of the remaining CD4 + CD8 + cells were likely glucocorticoid resistant [ 23 ]. However, it should also be noted that an increase in apoptosis was only observed in mice treated with a 5 mg pellet, not in mice treated with a 0.5 mg pellet. In contrast, a 0.5 mg pellet diminished the significant increase in cell numbers caused by adrenalectomy (Figure 3 ). This suggests the possibility that mechanisms other than apoptosis may also be involved glucocorticoid-mediated homeostasis in the thymus. The large percentage of CD4 - CD8 - cells in the apoptotic population was not entirely unexpected, because there is a developmental checkpoint that can lead to death in CD4 - CD8 - cells that do not productively rearrange a TCR β chain [ 24 ]. In the human thymus, the percentage of apoptotic CD4 - CD8 - is lower than we noted (13% in humans vs. 34% in the present study) [ 25 ]. However, the human thymuses in that study were obtained from newborns, and the differences could thus reflect age as well as species differences. Effect of ADX with or without corticosterone pellets on immature single positive cells in the thymus Single positive CD8 cells that express low levels of CD3 (and TCR) are apparently the immediate precursors for CD4 + CD8 + cells in the thymus [ 24 ]. One report indicates that these cells are more susceptible than mature single positive cells to glucocorticoids [ 19 ]. Results shown in Figure 8 are consistent, at least in part, with this finding. The increase in the percentage of this subpopulation in mice treated with a 5 mg pellet (2.4 fold) was substantially less than the increase in mature single positive cells (~5-fold) (Figure 2 ). In addition, the number of immature CD8 + cells was decreased to a greater extent than for mature single positive cells at this time point (Figure 1 ). This suggests that immature single positive cells are more sensitive to high levels of corticosterone than mature single positive cells, though they are apparently less sensitive than CD4 + CD8 + and CD4 - CD8 - cells. The number of immature single positive cells did not increase significantly in ADX mice that did not receive a corticosterone pellet, indicating that homeostasis of this cell type is not affected by physiological levels of corticosterone. This further illustrates that it is not appropriate to infer the effects of physiological concentrations of glucocorticoids on various cells types on the basis of the action of pharmacological concentrations of glucocorticoids. The percentage of immature single positive cells in the thymus of untreated mice in our study was comparable to values noted by others [ 19 , 26 ]. Figure 8 Effect of sub-physiological and stress-inducible concentrations of corticosterone on immature single positive (CD8 + ) thymocytes The histograms at the top of this figure illustrate the region containing cells that express low levels of CD3 (the larger purple peak on the histogram to the left, in which the isotope control is indicated by a green line), and the subsequent analysis of those cells for CD4 and CD8. The cells in the lower right quadrant (outlined with bold lines) are CD3 low CD8 + , immature single positive cells. The values shown in the graphs are means ± SE for groups of 5 mice representing either the percentage of immature single positive cells in the thymus or the total number of these cells in the thymus (obtained by multiplying the percentage by the total thymus cell number for each mouse). Groups significantly different from the naive group (by ANOVA followed by Dunnett's test) are shown by ( p < 0.01). Relationship of present results and results from other studies The findings reported here are consistent with some, but not all, results from other laboratories. The study most relevant to the present one involved the use of transgenic mice that express glucocorticoid receptor at higher than normal or lower than normal levels [ 4 ]. Comparing changes in the number of thymocytes in various subpopulations in mice expressing twice the normal level of glucocorticoid receptor in the thymus with our results at moderately elevated corticosterone concentrations (2.5 mg pellet, figure 3 ) indicates some similarities and some differences. In both studies, the total number of thymocytes was decreased significantly. However, expression of higher levels of glucocorticoid receptor caused significant suppression of cell numbers for CD4 + CD8 + , CD4 - CD8 - , and CD4 + CD8 - cells, but not CD4 - CD8 + cells [ 4 ]. In contrast, the 2.5 mg pellet in adrenalectomized mice (a situation that should be analogous to higher levels of glucocorticoid receptors with normal corticosterone concentrations) caused significant decreases in cell number in CD4 + CD8 + , CD4 - CD8 + , and CD4 - CD8 - cells, but not in CD4 + CD8 - cells (Figure 3 ). In mice expressing lower than normal levels of glucocorticoid receptor in the thymus (due to incorporation of anti-sense DNA under the control of the lck promoter) [ 4 ], the pattern of change was very similar to that which we observed in adrenalectomized mice (with no corticosterone pellet). In both cases, the significant increases in cell number were noted only for the CD4 + CD8 - and CD4 + CD8 + subpopulations. Cell number in the other two major subpopulations was increased slightly, but not significantly. The overall relationships between these results might be explained by a report indicating that normal expression of glucocorticoid receptor in the thymus is a very dynamic process, with substantial changes in expression in different subpopulations of cells [ 19 ]. In addition, the evidence suggests that sensitivity of the various subpopulations to glucocorticoids is not always strictly dependent on the amount of glucocorticoid receptor expressed. Thus, other factors that change during the development of T cells play an important role in sensitivity to glucocorticoids. Causing excess production of glucocorticoid receptor in all cells of the thymus (as in transgenic mice with an extra glucocorticoid receptor gene, transcribed in all thymocytes under the control of the lck promoter) [ 4 ] would not be likely to produce the same differences in glucocorticoid receptor levels among cellular subpopulations in the thymus as noted in normal animals (in which glucocorticoid receptor levels vary in different subpopulations). A portion of the glucocorticoid receptor production would be subject to the normal dynamic regulatory process, but a portion of production (the portion under the control of the lck promoter) would not. This may explain the differences in the results obtained using transgenic mice with elevated levels of glucocorticoid receptor [ 4 ] as compared to our results using elevated corticosterone concentrations. However, expression of glucocorticoid receptor anti-sense RNA in the thymus in a uniform manner [ 4 ] seems to have produced similar results as decreased corticosterone concentrations (in adrenalectomized animals) (Figs. 1 , 2 , 3 , 4 ,). This may reflect the fact that the action of anti-sense RNA in a particular cell type would be expected to be proportional to the amount of glucocorticoid receptor expressed in that cell. Thus, the normal differences between subpopulations with regard to glucocorticoid sensitivity might be retained. Thus, it is not surprising that the results with anti-sense glucocorticoid receptor transgenic mice are comparable to those for adrenalectomized mice in our study with regard to the differential increase in cell number for different subpopulations. It is not clear why no increase in thymus cellularity or changes in subpopulations were noted by other investigators using a conditional knockout system to produce mice in which the thymus contains little glucocorticoid receptor [ 14 ]. However, glucocorticoid receptor knockout mice can express portions of the glucocorticoid receptor, which may have unexpected functions [ 27 ]. Such contradictory findings with transgenic approaches have been common in this field of research (see Introduction), indicating a useful role for classical pharmacological approaches such as those in the present study. The relationship between the results reported here and results from studies on the interactions between glucocorticoids and self-antigen in positive selection is not clear. A recent study indicates that activation through the TCR down regulates SRG3, a protein that associates with the glucocorticoid receptor and increases sensitivity to glucocorticoids [ 28 ]. This has been proposed as an explanation for the decreased sensitivity of mature single positive thymocytes as compared to non-mature double positive thymocytes to high concentrations of glucocorticoids. However, as already noted, this pattern did not seem to apply when comparing the effects of sub-physiological and physiological concentrations of glucocorticoids, i.e., the CD4 + CD8 - cells increased to a greater extent than CD4 + CD8 + cells in mice with sub-physiological corticosterone concentrations. This suggests that the observed changes in SRG3 may not account for differences in sensitivity of cells in different subpopulations to physiological concentrations of corticosterone. This leaves open the question of what does mediate those differences and the role (if any) of TCR signalling. It would be of interest to explore this with TCR transgenic mice. One study in which adrenalectomy has been used to evaluate the effects of glucocorticoids on cellular subpopulations in the thymus yielded different results than those reported here. In that study, there was no increase in total cell number in the thymus, and there were no changes in subpopulation percentages in adrenalectomized mice two weeks after adrenalectomy [ 18 ]. The basis for the difference in this result and the results of other studies, which indicate increased numbers of thymocytes in adrenalectomized mice or rats [ 15 - 17 ] is not clear. In a study in which one of the authors of the present report (E. L. P.) was involved, there was a greater increase in cell number in the thymus in adrenalectomized mice than in the present study [ 16 ]. Perhaps because the overall increase in cell number was greater, the increases in all subpopulations were significant [ 16 ]. The age and housing conditions (and resulting environmental stress levels) of the control group probably plays an important role in this regard, and it would be very difficult to assure precisely the same conditions for the control group in every case. Nevertheless, the results reported here along with results from most other adrenalectomy studies and results from one study using transgenic mice [ 4 ] indicate that physiological (non-stress) concentrations of corticosterone normally decrease the number of cells in the thymus. Conclusions The results presented here do not directly demonstrate the extent to which corticosterone contributes to death by neglect or the extent to which it contributes to the death of negatively selected thymocytes. The fact that both CD4 + CD8 + and CD4 + CD8 - cells are increased in number in ADX mice is consistent with the idea that death by neglect of CD4 + CD8 + cells is mediated by corticosterone. The increase in CD4 + CD8 - cells could be explained by the failure of CD4 + CD8 + cells to die before reaching maturity, as they would have done in the presence of corticosterone. The observation that CD4 - CD8 + cells are not increased to the same extent as CD4 + CD8 - cells is exactly what would have been predicted if sub-physiological concentrations of corticosterone allow survival of cells that would normally die by neglect. Whereas maturation of CD4 + CD8 + cells to CD4 - CD8 + cells requires MCH class I-dependent signals, maturation to CD4 + CD8 - status can be MHC-independent and apparently occurs by default [ 29 ]. Thus, cells that are not selected positively may preferentially mature to CD4 + CD8 - cells before dying by corticosterone-mediated apoptosis. However, sub-physiological concentrations of corticosterone apparently allow these non-selected cells to survive. Methods Animals and animal care Female C57BL/6 × C3H F1 (B6C3F1) mice were used in this study. Normal, adrenalectomized (ADX), and sham adrenalectomized mice were purchased from Charles River Labs (Wilmington, MA). The mice were allowed to recover from shipping stress for at least two weeks before use in experiments, and surgery was performed approximately one week before shipping. Thus, mice were evaluated at least three weeks after surgery at an age of 8–12 weeks. Mice were maintained on a 12 hour light/dark cycle, with free access to lab chow and water, except that ADX mice were given water with 0.9% sodium chloride. Sentinel mice housed periodically in the same room as the mice used in this study were negative for common adventitious agents and pathogens of mice. Animal care and use was in accord with the regulation of LSU Health Sciences Center and the NIH Guide for Care and Use of Laboratory Animals. The animal facility in which the mice were maintained is approved by the American Association for Accreditation of Laboratory Animal Care. Implantation of timed release corticosterone pellets Timed-release corticosterone pellets and placebo pellets were purchased from Innovative Research of America (Sarasota, FL). The pellets are designed to yield constant blood levels of corticosterone for 3 weeks. Pellets were implanted subcutaneously in the scapular area of mice that were anesthetized with sodium pentobarbital (55 mg/kg) and inhalation of methoxyflurane, as described in a previous study [ 2 ]. The incision was closed with a surgical staple. The entire process was conducted aseptically. Mice were allowed to recover on a heating pad prior to being returned to their home cages. Parameters were evaluated after 24 hours in two experiments and after 72 hours in a third experiment. Preparation of cells and flow cytometry In two experiments, mice were euthanized by CO 2 inhalation and the thymus was removed for analysis. In one experiment mice were euthanized by decapitation, trunk blood was obtained and allowed to clot, and serum was isolated after centrifugation. The serum was used to determine corticosterone concentration, using a radioimmunoassay kit (Diagnostic Products Corporation, Los Angeles, CA) as described previously [ 30 , 31 ]. The thymus was then removed from each mouse. Single cell suspensions were prepared in 3 ml of RPMI 1640 by pressing the organs between the frosted ends of sterile glass microscope slides, as in previous studies [ 1 , 2 ]. After centrifugation, the cells were resuspended in 3 ml of RPMI 1640, 20 μl samples were taken, diluted in 10 ml of Isoton II isotonic buffered saline (Coulter Corp., Miami, FL), and counted using an electronic cell counter (Coulter Model Z1). Cells were adjusted to 2 × 10 7 per ml, and 50 μl was added to the wells of a 96-well V-bottom microplate. Antibodies diluted in 50 μl of FACS buffer (phosphate buffered saline without calcium and magnesium plus 0.1% bovine serum albumin and 0.1% sodium azide) were added to appropriate wells. In each experiment in which multiple antibodies were used, controls included cells labeled with each antibody singly, cells labeled with each isotype control antibody singly, cells labeled with all isotype control antibodies together, and unlabeled cells. The following antibodies were used: anti-CD4 (GK1.5) labeled with phycoerythrin (PE), anti-CD8a labeled with Cychrome, and anti-CD3 labeled with fluorescein isothiocyanate (FITC). These antibodies and matching isotype controls were obtained from BD Pharmingen. Titration of the antibodies indicated that a 1:8 dilution of anti-CD4 and anti-CD8 and a 1:5 dilution of anti-CD3 were appropriate for this study. After labeling for 30 min at 4°C, the cells were washed, fixed with 1% paraformaldehyde (EM Sciences, Ft. Washington, PA), washed again, and resuspended in FACS buffer. Samples were diluted in Isoton II (0.4 ml) for analysis. Cells were analyzed using a FACScan flow cytometer (Becton-Dickinson). A gate was set using forward scatter and side scatter to exclude debris, erythrocytes, and clumps of cells. All cells within this gate were then analyzed for CD3, CD4, and CD8. In some experiments, cells were labeled to detect DNA fragmentation instead of CD3. Cells were first labeled with anti-CD4 (phycoerythrin) and anti-CD8 (cychrome) as described above, then the cells were fixed with 4% paraformaldehyde (in phosphate buffered saline). A terminal dUTP nick end labeling (TUNEL) kit from Boehringer-Mannheim (Indianapolis, IN) with fluorescein-labeled dUTP was used to label apoptotic cells. Flow cytometry was used to identify apoptotic cells by two criteria. Cells that were small (as indicated by forward scatter) and labeled with fluorescein (indicating DNA fragmentation) were regarded to be apoptotic. Statistical analysis Values significantly different from the naive (untreated) control group were determined by analysis of variance (ANOVA) followed by Dunnett's post hoc test. Statistical analysis, linear regression, and non-linear regression analysis were performed using Prism 4.0 software (GraphPad, Inc., San Diego, CA). Comparison of slope or intercept of pairs of regression lines was done using the method of Zar [ 32 ] as implemented by Prism software. Authors' contributions The authors contributed approximately equally to conceiving, designing, and conducting these experiments. S.B.P. wrote the manuscript, and E.L.P. revised it.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC534100.xml
526276	Reduced inhibitory action of a GABAB receptor agonist on [3H]-dopamine release from rat ventral tegmental area in vitro after chronic nicotine administration	Background The activation of GABA B receptors in the ventral tegmental area (VTA) has been suggested to attenuate the rewarding properties of psychostimulants, including nicotine. However, the neurochemical mechanism that underlie this effect remains unknown. Since GABA B receptors modulate the release of several neurotransmitters in the mammalian brain, we have characterised the effect of the GABA B receptor agonist baclofen on the release of [ 3 H]-dopamine ([ 3 H]-DA) from VTA slices of naïve rats and of rats pre-treated with nicotine. Results In naïve rats, baclofen concentration-dependently inhibited the electrically evoked release of [ 3 H]-DA from the isolated VTA (EC 50 = 0.103 μM, 95% CI = 0.043–0.249), without affecting the basal [ 3 H]-monoamine overflow. This effect was mediated by activation of GABA B receptors as it was blocked by the selective receptor antagonist CGP55845A. Chronic administration of nicotine (0.4 mg kg -1 , s.c., for 14 days) affected neither the basal nor the electrically evoked release of [ 3 H]-DA from VTA slices. However, the inhibitory effect of baclofen (10 μM) on the stimulated [ 3 H]-monoamine overflow was abolished in rats pre-treated with nicotine as compared to saline-injected controls. Conclusions Our results demonstrate that GABA B receptor activation reduces the release of DA from the rat VTA. In addition, a reduced sensitivity of VTA GABA B receptors appears to develop after chronic exposure to nicotine. The resulting disinhibition of VTA DA neurones might therefore contribute to the sensitised dopaminergic responses observed in the rat mesocorticolimbic system following repeated administration of nicotine.	Background The ventral tegmental area (VTA) represents the site of origin of the mesocorticolimbic dopaminergic pathway that has been implicated in mediating the reinforcing properties of drugs of abuse, including nicotine [ 1 - 3 ]. The majority of cells within the ventral tegmental area consist of dopaminergic, tyrosine-hydroxylase containing neurones, which send axon projections to forebrain structures such as the nucleus accumbens and the prefrontal cortex. The non-tyrosine hydroxylase containing neurones are mainly GABAergic and function either as local interneurones to modulate the activity of the principal dopaminergic cells or as projection neurones providing inhibitory input to the cortex and the nucleus accumbens [ 4 - 8 ]. In addition to classical release from the axon terminals located in the forebrain, midbrain dopaminergic neurones release dopamine (DA) from their soma and dendrites [ 9 - 12 ]. The somatodendritic release of DA provides a primary modulation of dopamine cell function. Activation of D 2 autoreceptors inhibits excitability and firing rate of VTA dopaminergic neurones [ 13 , 14 ] and decreases the release of dopamine from their axon terminals in the forebrain [ 15 , 16 ]. Furthermore, somatodendritic dopamine can indirectly modulate the activity of midbrain dopamine cells by acting on D 1 receptors, which are found on GABA- and excitatory amino acids-containing terminals in the VTA [ 17 - 20 ]. Besides the short-loop feedback inhibition exerted by dopamine, the activity of VTA dopaminergic neurones is strongly modulated by glutamatergic and GABAergic inputs. Activation of NMDA receptors by excitatory afferents arising from the medial prefrontal cortex [ 21 , 22 ] induces burst firing in VTA DA neurones [ 23 - 25 ], which is associated with increased dopamine release from the nerve terminals in the nucleus accumbens [ 26 , 27 ]. The VTA also receives an extensive inhibitory influence arising from GABA interneurones and descending projections from the basal forebrain, innervating GABA A and GABA B receptors, respectively [ 4 ]. Activation of GABA B receptors has been reported to inhibit the spontaneous pacemaker-like activity of VTA DA neurones in slice preparations [ 4 , 28 , 29 ] and to decrease the firing rate and burst firing of these cells in vivo [ 30 , 31 ]. In addition, in vivo microdialysis studies have demonstrated that intra-VTA administration of the GABA B receptor agonist baclofen decreases extracellular dopamine levels in both the somatodendritic [ 32 , 33 ] and the axon-terminal regions of the mesocorticolimbic system [ 15 , 16 ]. The GABA B receptor-mediated inhibition of the activity of the mesocorticolimbic neurones might explain the effectiveness of baclofen to suppress nicotine self-administration when microinjected into the VTA [ 34 ]. In fact, the rewarding properties of nicotine have been ascribed to its ability to stimulate VTA dopamine neurones that project to the nucleus accumbens [ 2 , 35 ]. Acute administration of nicotine to drug-naïve rats increases extracellular levels of dopamine in the nucleus accumbens shell, while repeated exposure to the drug results in sensitisation of its effect on dopamine overflow in the nucleus accumbens core [ 36 ]. Sensitisation of the mesoaccumbens dopamine response to nicotine appears to be closely related to the dependence-liability of the drug and has been suggested to reflect an altered control of DA release, including reduced inhibitory influence by DA autoreceptors, and co-stimulation of NMDA receptors [ 37 - 39 ]. Since GABA B receptors have a prominent role in regulating the activity of VTA DA neurones and they appear to be involved in the modulation of nicotine reinforcing properties, we have characterised the effect GABA B receptor stimulation on the somatodendritic release of [ 3 H]-DA from VTA slices of naïve rats and used this model to determine whether chronic exposure to nicotine results in altered GABA B receptor-mediated modulation of VTA DA cells. Results Tetrodotoxin and calcium dependence of the stimulated [ 3 H]-DA release from VTA slices The influence of tetrodotoxin (TTX) and calcium on the electrically evoked release of [ 3 H]-DA from ventral tegmental slices of naïve rats was evaluated in order to determine its physiological significance under the experimental conditions used here. Removal of calcium with the inclusion of EGTA (1 mM), or addition of TTX (1 μM) to the superfusion buffer abolished the stimulated release of [ 3 H]-DA from the tissue, without significantly affecting the spontaneous monoamine overflow. Recovery of the release occurred during the second stimulation (S2), after removal of TTX or replacement of calcium to the superfusion buffer (Figure 2 ). Figure 2 Calcium-dependence and tetrodotoxin (TTX) sensitivity of the electrically evoked release of [ 3 H]-DA from ventral tegmental area slices. Electrical stimulation (20 mA, 2 Hz, 4 min) occurred 14 min (S1) and 59 min (S2) after the beginning of sample collection. Slices were perfused with buffer alone (control), calcium-free medium containing 1 mM EGTA (Ca 2+ -free), or buffer containing 1 μM TTX for 15 min before and during S1 (black bar). After 18 min from the beginning of sample collection, all samples were superfused with normal buffer. Values represent mean ± s.e.mean ( n = 3 rats). Effect of GABA B receptor activation on VTA [ 3 H]-DA release Baclofen (0.1–100 μM), added to the superfusion buffer 36 min before the second stimulation, dose-dependently reduced the electrically evoked [ 3 H]-DA release from VTA slices (EC 50 = 0.103 μM, 95% CI = 0.043–0.249, n = 4–7 rats, Figure 3 ), without having any significant effect on the basal monoamine overflow (data not shown). Figure 3 Baclofen-mediated inhibition of the electrically evoked [ 3 H]-DA release from ventral tegmental area slices of naïve rats (EC 50 = 0.103 μM, 95% CI = 0.043–0.249). Baclofen (0.01–100 μM) was added to the superfusion buffer 36 min before and during S2. Values represent mean ± s.e.mean ( n = 4–7 rats). The inhibitory effect of baclofen (10 μM) on the release of [ 3 H]-DA was abolished when the selective GABA B receptor antagonist CGP55845A (1 μM), which had no effect on its own, was added to the superfusion buffer concomitantly with baclofen (Figure 4 ). Release of dopamine, expressed as S2/S1 ratio, was 1.09 ± 0.10 for control slices perfused with buffer alone; 0.73 ± 0.08 for baclofen added before S2 (P < 0.05 vs control, one-way ANOVA, Dunnet's post hoc test, n = 6–10 rats); 0.99 ± 0.13 for CGP55845A alone; and 0.93 ± 0.14 for slices perfused with both baclofen and CGP55845A. Figure 4 Reversal of baclofen-mediated inhibition of the evoked [ 3 H]-DA release from VTA slices by the GABA B antagonist CGP55845A. Baclofen (Bacl, 10 μM) and/or CGP55845A (CGP, 1 μM) were added to the superfusion buffer 36 min before and during S2. Drug effects were assessed by comparing the ratio S2/S1 in the presence and absence of the drug, respectively. Values represent mean ± s.e.mean ( n = 6–10 rats). * p < 0.05 vs control (one-way ANOVA, Dunnett's post hoc test). Effect of nicotine pre-treatment on the release of [ 3 H]-DA from VTA The release of [ 3 H]-DA from the VTA slices of rats that received a chronic nicotine treatment was compared with the overflow observed in saline-control animals. Neither the basal nor the electrically stimulated release of [ 3 H]-DA was significantly affected by the drug pre-treatment (Figure 5 ). Figure 5 Effect of nicotine pre-treatment on basal and electrically evoked release of [ 3 H]-DA from VTA slices. The rats received daily subcutaneous injections of nicotine (0.4 mg kg -1 , n = 6) or saline ( n = 6) for 14 consecutive days and the experiments were performed 24 hours after the last injection. An electrical stimulation (20 mA, 2 Hz, 4 min, black bar) was applied to the VTA slices 14 min after the beginning of sample collection. The data are expressed as fractional [ 3 H]-DA release and represented as mean ± s.e.mean. Effect of baclofen on the release of [3H]-DA from VTA of nicotine-treated rats The effect of baclofen on the VTA [ 3 H]-DA release in rats pre-treated with nicotine was assessed and compared with the effect of the drug in saline-control animals (Figure 6 ). The addition of baclofen (10 μM) to the superfusion buffer for 36 min before and during S2, significantly reduced the evoked release of [ 3 H]-DA from VTA slices of rats pre-treated with saline (S2/S1: control, 1.06 ± 0.08; baclofen, 0.73 ± 0.06; P < 0.01, ANOVA for repeated measures, Bonferroni post-test, n = 5 rats). By contrast, in the rats pre-treated with nicotine, the addition of baclofen to the superfusion buffer did not produce any significant effect on the evoked release of [ 3 H]-DA (S2/S1: control, 1.02 ± 0.06; baclofen, 0.92 ± 0.05) (Figure 6 ). Figure 6 Effect of baclofen on the electrically evoked release of [ 3 H]-DA from VTA slices of rats pretreated with nicotine or saline. The rats received daily subcutaneous injections of nicotine (0.4 mg kg -1 , n = 5) or saline ( n = 5) for 14 consecutive days and the experiments were performed 24 hours after the last injection. Two electrical stimulations (20 mA, 2 Hz, 4 min) were applied to the slices 14 min (S1) and 59 min (S2) after the beginning of sample collection. Baclofen (10 μM) was added to the superfusion buffer 36 min before and during S2. Control tissue was perfused with buffer alone during both S1 and S2. Data are expressed as mean ± s.e.mean and compared by ANOVA for repeated measures. ** p < 0.01 vs saline-control (Bonferroni post-test). Discussion In the present study we have characterised the effect of baclofen on the release of [ 3 H]-DA from ventral tegmental area slices of naïve rats, and used this model for studying the functional status of local GABA B receptors after chronic exposure to nicotine. The electrically induced [ 3 H]-DA release from VTA somatodendrites was calcium-dependent and tetrodotoxin-sensitive. This suggests that the release is tightly coupled with voltage-sensitive calcium influx and that it depends on the propagation of action potentials by voltage-dependent sodium channels. Our findings are consistent with the calcium sensitivity and the partial block by tetrodotoxin of the release of endogenous dopamine observed in DA cell body areas using microdialysis [ 10 , 40 ]. They are also consistent with the calcium and tetrodotoxin dependency of the electrically evoked [ 3 H]-DA release from slices of the ventral tegmentum [ 41 ]. Therefore, under the experimental conditions used in the present study, release of preloaded [ 3 H]-dopamine appears to be of neuronal origin and to have physiological relevance. Although the radioactivity measured in the collected effluent may consist of a mixture of neurotransmitters and metabolites, the amount of tritium released from rat brain slices after electrical stimulation has been previously shown to represent a close estimation of the release of labelled or endogenous DA release [ 42 , 43 ]. Furthermore, the release of metabolites during the superfusion was inhibited by the presence of the monoamine oxidase inhibitor pargyline in the superfusion buffer [ 44 ]. Therefore, under our experimental conditions, the overflow of tritium from VTA slices seems likely to represent, predominantly, [ 3 H]-DA and closely resembles exocytotic release of DA, as it was dependent on the presence of calcium in the superfusion buffer. The VTA receives GABAergic input from both interneurones and descending projections from the basal forebrain, innervating GABA A and GABA B receptors, respectively [ 4 , 45 ]. Activation of GABA B receptors, located both postsynaptically on dopaminergic neurones and presynaptically on glutamatergic nerve terminals [ 29 ], decreases the spontaneous pacemaker-like activity and the burst firing of VTA DA cells [ 4 , 31 ]. In addition, baclofen microinjected into the VTA has been shown to decrease somatodendritic release of dopamine in this midbrain region as monitored by in vivo microdialysis [ 32 , 33 ]. The activation of GABA B receptors has also been reported to inhibit the potassium- or electrically evoked release of dopamine from various regions of the mammalian brain in vitro [ 46 , 47 ]. However, to date, there is a lack of information regarding the effect of GABA B receptor activation on the release of somatodendritic dopamine from the isolated VTA. Therefore, in the present study we have demonstrated that baclofen dose-dependently reduces the electrically evoked release of [ 3 H]-DA from ventral tegmental area slices of naïve rats. This effect appears to be mediated by activation of GABA B receptors, since it was abolished by superfusion of the tissue with the selective receptor antagonist CGP55845A [ 48 ]. The existence of a tonic GABA B receptor-mediated inhibition of somatodendritic DA release has been suggested by the evidence that, in vivo , the administration of CGP55845A into the ventral tegmental area produces a dose-dependent increase in VTA dopamine levels [ 49 ]. By contrast, our results demonstrate that the GABA B receptor antagonist CGP55845A does not have any significant effect on the release of [ 3 H]-DA when applied on its own to the VTA slices. The apparent discrepancy between the two studies may be ascribed to the fact that we have used isolated tissue, which is deprived of the tonic GABA input arising from the forebrain, whereas in the work of Giorgetti et al. [ 49 ] the long-loop projections to the VTA are indeed intact. In fact, the innervation of VTA GABA B receptors originates primarily from projection neurones located in the nucleus accumbens and the ventral pallidum, while the GABA interneurones appear to innervate mainly GABA A receptors [ 4 , 50 ]. Interestingly, previous neurochemical studies have shown that activation of GABA B receptors in the ventral tegmental area of naïve rats inhibits the release of dopamine not only in the cell body area, but also in mesocorticolimbic terminal regions, such as the nucleus accumbens [ 33 ] and the prefrontal cortex [ 15 ]. The reduction of mesocorticolimbic dopamine release might represent a likely mechanism by which baclofen attenuates nicotine self-administration in rats when microinjected into the ventral tegmental area [ 34 ]. However, to date, this hypothesis has not been confirmed and little information is known about the neurochemical mechanisms underlying the GABA B receptor-mediated modulation of nicotine reinforcement, as well as about the pharmacological interaction between nicotine and GABA B receptors. With the present study we demonstrate that while baclofen significantly reduces the electrically induced release of [ 3 H]-DA from the VTA of saline-control rats, it has no effect on the evoked monoamine release from VTA slices of nicotine pre-treated rats. This finding represents the first demonstration that chronic exposure to nicotine might result in reduced GABA B -mediated inhibition of VTA dopaminergic neurones. This hypothesis is also consistent with preliminary in vivo studies performed in our laboratory showing that, after chronic pre-treatment of the rats with nicotine, microinfusions of baclofen into the VTA failed to reduce both the spontaneous and the nicotine-evoked overflow of dopamine in this midbrain region (D. Amantea, unpublished observation). Taken together, our findings suggest that after chronic nicotine the GABA B receptor may be desensitised, and this would result in a reduced inhibitory control of VTA dopaminergic cells, thereby facilitating a more sustained increase in the responses of mesolimbic neurones to nicotine. Interestingly, the inhibitory action of baclofen on VTA DA cells may be accounted for by stimulation of GABA B receptors located on dopaminergic and/or glutamatergic neurones [ 28 , 29 ]. This suggests that the effect observed after chronic administration of nicotine may be ascribed to a reduced sensitivity of GABA B receptors located either on DA cell bodies or, presynaptically, on glutamatergic terminals impinging onto DA neurones. The latter hypothesis would lead to the speculation that a chronic treatment with nicotine might result in increased excitatory input to the VTA, due to a reduced GABA B -mediated inhibitory control. Therefore, disinhibition or increased excitatory input to midbrain DA cells might contribute to the augmented dopamine output observed in the nucleus accumbens after a challenge injection of nicotine in rats pretreated with the drug [ 36 , 37 ]. Desensitisation of GABA B receptors located in the VTA has also been reported to occur after chronic cocaine administration, as the drug treatment resulted in reduced functional coupling of the receptor to G-proteins [ 51 ]. However, we have previously demonstrated that GABA B receptor expression and coupling to G-proteins in the ventral tegmental area of the rat are not altered after chronic exposure to nicotine [ 52 ], suggesting that desensitisation might occur at other levels, perhaps on downstream effector mechanisms. Nevertheless, it cannot be ignored that the use of in vitro autoradiography did not allow us to discriminate between receptor subpopulations, including receptor subtypes involved in different functions. Thus, if the GABA B receptors directly implicated in the control of dopamine release from the VTA represent only a small fraction of the overall receptor population in this area, autoradiographic analysis would not be sufficiently sensitive to evaluate receptor modifications occurring after chronic exposure to nicotine. Further studies aimed at characterising the mechanisms involved in GABA B receptor desensitisation following chronic nicotine treatment are required to clarify this issue. Not surprisingly, basal and evoked release of dopamine from VTA slices of rats chronically injected with nicotine did not differ from release obtained from saline-control tissue. This confirms that sensitised dopaminergic responses to the drug depend on the activation of nicotinic receptors (nAChRs) and are not the result of a generic increase in neuronal activity. Similar results have been obtained in striatal synaptosomes from rats pretreated with the nAChR agonist anatoxin-a for 7 days: while the drug pre-exposure increased the nicotine-stimulated release of [ 3 H]-DA from the in vitro preparation, no difference was found in the K + -evoked release between the drug pretreated animals and the saline-injected controls [ 53 ]. Thus, although the electrical stimulation used in the present study does not provide information about desensitisation or up-regulation of nAChRs, it nevertheless served as a reliable model to study the functional status of the GABA B receptor in the ventral tegmental area of rats pretreated with nicotine. Conclusions In conclusion, we have demonstrated that activation of GABA B receptors inhibits the release of preloaded [ 3 H]-DA from somatodendritic fields of VTA neurones in naïve rats. This effect appears to be reduced in rats chronically treated with nicotine, suggesting that subsensitivity of GABA B receptors in the VTA might occur as a result of the drug treatment. This, in turn, would lead to disinhibition of VTA dopaminergic cells, which might contribute to the increased activity of mesocorticolimbic neurones following repeated exposure to nicotine. Methods Drugs [ 3 H]-Dopamine (specific activity 47 Ci/mmol) was obtained from Amersham (Buckinghamshire, UK). (-)-Baclofen (CGP11973A) and CGP55845A were generous gifts from Novartis Pharma (Basel, Switzerland). (-)-Nicotine hydrogen tartrate salt, nomifensine, pargyline, tetrodotoxin and ethylene glycol-bis (b-amino ethyl ether) tetraacetic acid (EGTA) were purchased from Sigma-Aldrich (Dorset, UK). All the other chemicals used in this study were obtained from Fisher Scientific (Leicestershire, UK). Subjects Male Wistar rats (weight 250–280 g) were maintained on a 12-h light/dark schedule (on 6:00–18:00), with free access to food and water. For the drug treatments, rats, initially weighing 150–180 g, received subcutaneous injections of nicotine (0.4 mg kg -1 , expressed as free-base) or vehicle (0.9% NaCl, 1 ml kg -1 ) for 14 consecutive days, once a day, between 9:30 and 10:30 a.m. Experiments were performed 24 hours after the last injection. All procedures were carried out in accordance to the UK Animals (Scientific Procedures) Act, 1986. Tissue preparation Animals were sacrificed by stunning followed by decapitation. The brains were rapidly removed from the skull and cooled for 2–3 min in ice-cold superfusion buffer of the following composition (in mM): NaCl, 118; KCl, 4.7; CaCl 2 , 1.3; MgCl 2 , 1.2; NaH 2 PO 4 , 1; NaHCO 3 , 25; glucose, 11.1; Na 2 EDTA, 0.004 and ascorbic acid, 0.3 (pH 7.4), saturated with 95% O 2 /5% CO 2 . Each brain was dissected according to visual anatomical landmarks and the atlas of Paxinos & Watson [ 54 ]. To obtain the coronal section containing the VTA, the brain was placed ventral side up and two parallel cuts were made at the level of the mammillary body and the medial interpeduncular nucleus (approximately 5–6 mm posterior to bregma), using as landmark the basal cerebral peduncle. Two transversal cuts were made in correspondence of the medial lemniscus, to remove the substantia nigra from the slice. Finally the ventral part of the section was dissected out by making a horizontal cut 1.5 mm dorsal to the ventral edge of the section and deprived of the interpeduncular and mammillary nuclei by cutting 0.5 mm above the same edge. The procedure used for the dissection of the VTA is illustrated in Figure 1 , which shows a photomicrogaph of a representative coronal slice immunohistochemically stained for tyrosine hydroxylase. Figure 1 Photomicrograph of a representative coronal section of rat brain labelled with a polyclonal antibody directed against tyrosine hydroxylase. The picture shows the orientation of the cuts (dark lines) for the dissection of the ventral tegmental area (VTA). The basal cerebral peduncle (Cp) was used as a visual landmark to obtain the coronal section containing the VTA. To remove the substantia nigra (pars compacta, SNC, and reticulata, SNR) from this slice, two transversal cuts were made along the medial lemniscus (Ml) on either side. Finally, the VTA was dissected out by making a horizontal cut 1.5 mm dorsal to the ventral edge of the section and stripped of the interpeduncular nucleus (IP) by cutting 0.5 mm above the same edge. Scale bar is 1 mm. The dissected VTA was cross-chopped (250 μm × 250 μm) with a McIlwain tissue chopper and the tissue slices were washed and resuspended in ice-cold superfusion buffer. Tissue superfusion The VTA slices were pre-incubated in oxygenated superfusion medium at 37°C for 15 min, followed by incubation with 0.1 μM [ 3 H]-DA for 30 min, in the dark and in the presence of 10 μM pargyline. The incubation was terminated by washing the slices three times with buffer containing 2.5 μM nomifensine. 150 μl aliquots of slice suspension (1.5 to 2.0 mg of tissue) were transferred to each chamber of a Brandel 2000 superfusion apparatus. The tissue was superfused at a rate of 0.5 ml min -1 with oxygenated buffer maintained at 35°C and containing 2.5 μM nomifensine, to block dopamine uptake, and 10 μM pargyline to ensure that [ 3 H] overflow represented primarily [ 3 H]-DA rather than its metabolites [ 44 ]. After 36-min pre-superfusion, the effluent was collected in consecutive fractions of 4 min 30 sec each. The release of [ 3 H]-DA was induced by electrical field stimulation (20 mA, 2 Hz for 4 min) using a Brandel constant current stimulator. Two stimulations occurred 14 min (S1) and 59 min (S2) after the beginning of sample collection. Test drugs were added to the superfusion medium 36 min before the second stimulation. To determine the effects of calcium and tetrodotoxin (TTX), slices were perfused with Ca 2+ -free buffer (in the presence of 1 mM EGTA), or buffer containing 1 μM TTX, for 15 min before and during S1. The superfusion medium was then replaced with normal buffer until the end of superfusion and during the second stimulation. Liquid scintillation counting At the end of the superfusion, the slices with their filters were removed from the chamber, suspended in 1 ml of buffer, and sonicated. OptiPhase 'HiSafe' 3 scintillation fluid (Perkin Elmer, UK) was added to each vial and the radioactivity content in the superfusion samples and in the tissue slices was assayed by liquid scintillation counting using a Tri-Carb ® 1500 liquid scintillation analyser (Packard Bioscience Company), programmed to count for tritium, 3 minutes per vial, at an efficiency of 60%. The number of disintegrations per minute (d.p.m.) was measured in order to determine the concentration of tritium in each sample. Data analysis For each time point (4 min 30 sec), release of radioactivity was expressed as fractional release, i.e. , as a percentage of the amount of radioactivity in the tissue at the beginning of that collection. The electrically evoked release was expressed as the mean of the increased fractional release above baseline in the two fractions after the beginning of stimulations. Basal release, in turn, was calculated as the mean of the amount of radioactivity present in the three samples just before each period of electrical stimulation. Results were expressed as mean ± s.e.mean of n independent experiments conducted in either triplicate or quadruplicate. For statistical analysis one-way ANOVA with Dunnett's post hoc test was used to compare values of S2/S1 in the presence of a drug versus values of S2/S1 from control slices superfused with buffer alone. When nicotine or saline were administered to the animals, values of S2/S1 were analysed by ANOVA with repeated measures (with or without baclofen) with pre-treatment as factor analysed, and post hoc comparisons were made using the Bonferroni test. The accepted level of significance was p < 0.05. Histology Immunohistochemistry was performed on naïve rat brain during preliminary experiments aimed at characterising the exact orientation of the cutting for the dissection of the VTA (Figure 1 ). Animals were sacrificed by stunning and decapitation and their brains rapidly removed from the skull. A coronal section containing the VTA was obtained as described above and the tissue was post-fixed in 4% paraformaldehyde (BDH) for 48 hours at 4°C. Serial coronal sections (30 μm thick) were cut using a vibratome and washed in 0.01 M phosphate buffered saline (PBS, Sigma-Aldrich), pH 7.4. Slices were incubated with 3% hydrogen peroxide (H 2 O 2 ) for 30 min, followed by incubation with 0.2% Triton X-100 (Sigma-Aldrich) for 20 min at room temperature. After a 1-hour pre-incubation in 10% normal goat serum (NGS, Vector), the primary antibody (rabbit polyclonal antibody to tyrosine-hydroxylase, Affiniti) was applied at a final concentration of 1:1000 (in 0.01 M PBS) and the sections were allowed to incubate overnight at 4°C. After three washes with fresh buffer, the slices were incubated for 90 minutes at room temperature with a biotinylated secondary goat anti-rabbit antibody (1:200 dilution, Chemicon). Immunoreactivity was visualised by the avidin-biotin complex method of detection (Vectastain Elite ABC Kit, Vector) using 3,3'diaminobenzidine (DAB, peroxidase substrate kit, Vector) as the chromogen. Authors' contributions DA carried out the experiments, participated in the design of the study and drafted the manuscript. NGB conceived the study, and participated in its design and coordination. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526276.xml
526512	Highly Conserved Non-Coding Sequences Are Associated with Vertebrate Development	In addition to protein coding sequence, the human genome contains a significant amount of regulatory DNA, the identification of which is proving somewhat recalcitrant to both in silico and functional methods. An approach that has been used with some success is comparative sequence analysis, whereby equivalent genomic regions from different organisms are compared in order to identify both similarities and differences. In general, similarities in sequence between highly divergent organisms imply functional constraint. We have used a whole-genome comparison between humans and the pufferfish, Fugu rubripes, to identify nearly 1,400 highly conserved non-coding sequences. Given the evolutionary divergence between these species, it is likely that these sequences are found in, and furthermore are essential to, all vertebrates. Most, and possibly all, of these sequences are located in and around genes that act as developmental regulators. Some of these sequences are over 90% identical across more than 500 bases, being more highly conserved than coding sequence between these two species. Despite this, we cannot find any similar sequences in invertebrate genomes. In order to begin to functionally test this set of sequences, we have used a rapid in vivo assay system using zebrafish embryos that allows tissue-specific enhancer activity to be identified. Functional data is presented for highly conserved non-coding sequences associated with four unrelated developmental regulators (SOX21, PAX6, HLXB9, and SHH), in order to demonstrate the suitability of this screen to a wide range of genes and expression patterns. Of 25 sequence elements tested around these four genes, 23 show significant enhancer activity in one or more tissues. We have identified a set of non-coding sequences that are highly conserved throughout vertebrates. They are found in clusters across the human genome, principally around genes that are implicated in the regulation of development, including many transcription factors. These highly conserved non-coding sequences are likely to form part of the genomic circuitry that uniquely defines vertebrate development.	Introduction Identification and characterisation of cis -regulatory regions within the non-coding DNA of vertebrate genomes remain a challenge for the post-genomic era. The idea that animal development is controlled by cis- regulatory DNA elements (such as enhancers and silencers) is well established and has been elegantly described in invertebrates such as Drosophila and the sea urchin [ 1 , 2 , 3 , 4 ]. These elements are thought to comprise clustered target sites for large numbers of transcription factors and collectively form the genomic instructions for developmental gene regulatory networks (GRNs). However, relatively little is known about GRNs in vertebrates. Any approach to elucidate such networks necessitates the discovery of all constituent cis- regulatory elements and their genomic locations. Unfortunately, initial in silico identification of such sequences is difficult, as current knowledge of their syntax or grammar is limited. By contrast, computational approaches for protein-coding exon prediction are well established, based on their characteristic sequence features, evolutionary conservation across distant species, and the availability of cDNAs and expressed sequence tags (ESTs), which greatly facilitate their annotation. The completion of a number of vertebrate genome sequences [ 5 , 6 , 7 , 8 , 9 ], as well as the concurrent development of genomic alignment, visualisation, and analytical bioinformatics tools (for an overview see [ 10 ]), has made large genomic comparisons not only possible but an increasingly popular approach for the discovery of putative cis -regulatory elements. Comparing DNA sequences from different organisms provides a means of identifying common signatures that may have functional significance. Alignment algorithms optimise these comparisons so that slowly evolving regions can be anchored together and highlighted against a background of more rapidly evolving DNA that is free of any functional constraints. One of the key decisions inherent in comparative genomics is the choice of organisms for which the comparison will be made. A number of successful pairwise and multiple-species sequence comparisons have already been carried out to identify novel enhancer elements in mammalian genomes [ 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 ]. Unfortunately, owing to differences in mutation rates across the genome, many slower-evolving regions show a significant degree of non-coding sequence conservation that reflects the short evolutionary distance between mammals and the slow rate of neutral divergence in vertebrates [ 20 ]. Consequently, although approximately 40% of the human and mouse genomes is alignable, only approximately 5% is estimated to be under selection, making it difficult to identify functionally relevant sequences [ 8 ]. One approach has recently been described [ 21 ] that identifies only those sequences that are identical over at least 200 bp between human and mouse genomes (termed ultra-conserved elements) and examines their distribution in the genome. Around half of the 481 elements identified showed no evidence of transcription and are therefore likely to be regulatory. Another highly successful approach to increasing the resolving power of comparative analyses is to use multi-species alignments combining both closely related and highly divergent organisms [ 14 , 22 , 23 , 24 ]. By using large evolutionary distances, even the slowest-evolving neutral DNA has reached equilibrium, thereby significantly improving the signal to noise ratio in genomic alignments. Although non-coding sequences generally lack sequence conservation between highly divergent species [ 22 ], there are a number of striking examples where comparison between human and pufferfish (Fugu rubripes) gene regions has readily identified highly conserved non-coding sequences that have been shown to have some function in vivo [ 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 ]. Humans and Fugu last shared a common ancestor around 450 million years ago [ 35 ], predating the emergence of the majority of all extant vertebrates, implying that any non-coding sequences conserved between these two species are likely to be fundamental to vertebrate life. The Fugu genome has the added advantage of being highly compact, reducing intronic and intergenic distances almost 10-fold [ 7 , 36 ]. Without exception, all reported examples of non-coding conservation between these two species have been associated with genes that play critical roles in development. This suggests that some aspects of developmental regulation are common to all vertebrates and that whole-genome comparisons may be particularly powerful in identifying regulatory networks of this kind. As a first step towards identifying such networks in humans, we have used comparative genomics to identify and begin to characterise non-coding sequences that are highly conserved between human and Fugu . A general strategy for testing whether non-coding regulatory sequences of this type are functionally relevant involves assaying their ability to up-regulate reporter gene expression in vivo. “Enhancer” assays using transgenic animals, in particular mouse, are both slow and laborious, but have provided some exciting data on the function of non-coding DNA around developmental genes (e.g., [ 31 ]). An alternative approach uses transient expression in zebrafish (Danio rerio) embryos [ 37 , 38 , 39 ], which are particularly suited to this form of analysis. Whilst transient expression is highly mosaic, the availability of large numbers of fertilised eggs, ease of micro-injection, and transparency of the developing embryos means that hundreds of individuals may be screened at a time. This provides a rare opportunity for increasing the throughput of this kind of functional assay. We have adopted a medium-throughput strategy to test DNA sequences for enhancer activity in zebrafish embryos, whereby results may be cross-referenced and compared through a generalised scheme. We present functional data for 25 highly conserved non-coding sequences that are located around four unrelated developmental regulators, SOX21, PAX6, HLXB9, and SHH in order to demonstrate the suitability of this screen to a wide range of genes and expression patterns. Results Identification of Highly Conserved Non-Coding Sequences in Vertebrate Genomes To locate conserved non-coding sequences, we masked the majority of the coding and tRNA content of the Fugu genome assembly [ 7 ] and compared the remaining regions using MegaBLAST [ 40 ] with the human genome sequence contained in Ensembl release v18.34.1 [ 41 ]. From this analysis we identified 19,744 sequences with similarity between the two genomes. By only including alignments of at least 100 bp in length, the number of sequences was reduced to 4,400. We then removed telomere-like sequences and transposons, and excluded any known protein-coding sequence or non-coding RNA species that may have been missed (see Materials and Methods ). Sixty-five unique human sequences had matches to two independent locations in the Fugu genome. This is due to additional gene or genome duplications in the teleost lineage [ 42 ], where regulatory elements have been retained together with both copies of the fish gene [ 43 ]. To avoid redundancy in the human set, the longest matching sequence was retained and the duplicate removed. Finally, from the 1,373 sequences that remained, we determined whether any formed part of untranslated regions (UTRs) of mRNA molecules. Eighty sequences (approximately 6%) are situated in the 5′ or 3′ UTRs of known mRNA molecules. In addition, a similar number match one or more EST sequences, although most of these appear to be unspliced genomic contamination within EST libraries or incompletely spliced pre-mRNA. We did not remove these potentially transcribed sequences as, unlike vertebrate UTRs in general, they demonstrate a remarkable degree of conservation, and it is not clear whether they might be functional at the genomic or the transcript level. The remainder had no match against any expressed sequence in any database. This core set of 1,373 highly conserved non-coding elements (CNEs) forms the basis of this study. The set of CNEs comprise a total of 273 kb of sequence, with a maximum length of 736 bp (average = 199 bp) and identity ranging from 74% to 98% (average = 84.3%). This is considerably higher than the level of identity seen between coding regions in these two organisms. Unsurprisingly, virtually all of the CNEs are conserved in rodent and chicken genomes, as well as a majority in the zebrafish genome. Of the 1,373 CNEs, 1,365 are conserved in the mouse, 1,316 in rat, and 1,310 in chicken, aligning to the human sequence with average identities of 97% for mouse and rat and 96% for chicken; 1,093 are also found to be conserved in the zebrafish genome, aligning with an average identity of 87.6% to the Fugu sequence. The zebrafish, chicken, mouse, and rat genomes are at different stages of completeness, and therefore missing sequence information may account for the missing CNEs (as well as the lower percent identity in zebrafish), although it may also reflect regulatory differences between the lineages. Although CNEs are found throughout the human genome in all chromosomes except 21 and Y, their distribution is not uniform; in fact, they appear highly clustered. To examine their distribution in more detail, we plotted the position of each CNE on its respective chromosome in the human genome ( Figure 1 A). We then calculated the percentage of CNEs that were located in close proximity to another. We found that 90% of CNEs are less than 1 Mb apart, 85% of CNEs have a neighbouring CNE within 370 kb, and 75% are located within 158 kb of another CNE. The probability that over 85% of CNEs would be within 370 kb of another in a random distribution is less than 10 −76 ( Figure 1 B). By carefully examining the distribution of CNEs across the genome, we generated a total of 165 clusters, including 19 singletons ( Table S1 ). Over 85% of the CNEs (1,172/1,373) are located in clusters containing five or more CNEs. The 20 largest clusters each contain 20 or more CNEs, comprising 43% (594/1,373) of the total number of elements. Figure 1 Distribution of CNEs along the Human Genome (A) Each CNE is plotted relative to its position along each of human Chromosomes 1 to 9 (data for other chromosomes not shown). The y-axis represents length along the chromosome (in megabases). (B) Distribution of the fraction of CNEs that are within certain distances of each other; e.g., 85% of the distances between CNEs are less than or equal to 370 kb. χ 2 tests were carried out by comparing observed cluster sizes with those generated randomly for each chromosome (see Materials and Methods ). We then looked at the type of genes that are associated with CNEs in the human genome. For each CNE, we extracted the closest gene from Ensembl and submitted the resulting set of genes to GOstat [ 44 ] in order to identify the most statistically over-represented Gene Ontology (GO) terms [ 45 ]. Critically, 12 of the most over-represented 13 terms ( p < 0.001) relate to transcriptional regulation and development ( Table S2 ). We examined each cluster in turn to see how many were situated close to genes implicated in transcriptional regulation or development (we have termed these trans-dev genes). Over 93% of the clusters (154/165) have a trans-dev gene located within 500 kb of one or more of its CNEs ( Figure 2 ; Materials and Methods ; Table S1 ). Of the remaining 11 clusters, five are closest to genes with zinc finger domains as identified by InterPro [ 46 ], one is in a gene desert, one maps to the AUTS2 gene region [ 47 ], and four are located adjacent to uncharacterised genes. Figure 2 CNE Clusters Are Found Close to Trans-Dev Genes in the Human Genome Chromosomal locations of trans-dev genes that are within 500 kb of CNE clusters in the human genome (each cluster is represented by a green arrowhead). Genes in bold script are located next to clusters of ten or more CNEs. Gene names are taken from Ensembl v23.34e.1. Graph inset shows distribution of CNE cluster sizes in the human genome. Whilst most of the clusters can be associated with one trans-dev gene, there are 15 clusters in which CNEs are located close to two or more trans-dev genes. In nine of these cases, the CNEs associate with a group of paralogous genes, including the HOX, IRX, Nkx2–2/2–4, and DLX clusters, although there are three instances where a pair of unrelated trans-dev genes are located next to a CNE cluster (SHH and HLXB9, PBX3 and LMX1B, and PAX1 and FOXA2). Finally, there are three clusters that associate with two or more zinc finger genes. Trans-dev genes associated with CNE clusters tend to be located in regions of low gene density. We counted the number of genes situated within 500 kb upstream and downstream of a trans-dev gene, and compared this with the average for all human genes. Whereas the average for all human genes is 17, it is just six for the trans-dev genes. This is similar to the “gene desert” phenomenon described around the DACH gene [ 31 ]. Interestingly, the CNEs themselves are generally located large distances from their nearest gene. The average distance between a CNE and the 5′ end of the closest human gene is 182 kb (median = 120 kb), with 93 CNEs more than 500 kb, and 12 CNEs more than 1Mb, from any known gene. A number of the trans-dev genes that we identified have previously been shown to have highly conserved cis -regulatory elements associated with them, including the Hox clusters [ 24 , 33 ], PAX6 [ 48 ], PAX9 [ 32 ], SOX9 [ 28 ], OTX2 [ 34 ], SHH [ 30 ], DLX genes [ 29 ], and DACH [ 31 ]. Five CNEs do not appear to cluster with any known genes in either the human or Fugu genomes and are located in a large gene desert on human Chromosome 22. Given that gene annotation and genomic sequencing of parts of the human genome are not yet fully complete, the discovery of CNEs here may well point to the existence in this region of an important transcriptional or developmental regulation gene with which they are associated. Indeed we find the largest number of CNEs (48) clustered around a relatively uncharacterised gene with zinc finger domains, ZNF503 on human Chromosome 10, the rat orthologue of which was recently characterised as a probable transcriptional regulator in brain development [ 49 ]. All CNEs were compared with each other to look for local similarities. Forty-three elements show significant similarity to at least one other CNE, and in each case are situated close to genes with clear paralogous relationships, e.g., the HOX and IRX clusters. The remainder of the sequences appear to be unique in the human genome. In order to identify additional conserved sequences around specific genes for further functional assay, localised multiple-alignment comparisons were performed using the multiple LAGAN (MLAGAN) alignment tool kit [ 50 ]. This tool kit provides the opportunity to introduce genomic sequence from additional species, in this case mouse and rat, which significantly enhances the signal to noise ratio. For a random subset of 25 of the trans-dev genes associated with CNE clusters, stringent whole-genome alignment located 408 CNEs, whereas MLAGAN identified over twice as many conserved regions (871) of at least 100 bp in length. The whole-genome analysis was more stringent in that we used a minimum exact word match of 20 bp, whereas MLAGAN uses short inexact words to create anchors from which a more sensitive (Needleman–Wunsch) alignment is carried out. It is important to note that similar alignments on genes that are not implicated in developmental regulation do not identify conserved non-coding sequence (e.g., [ 22 , 51 ]). The alignment of a known transcription factor, SOX21, identifies a large number of conserved non-coding sequence elements in addition to the CNEs found in the whole-genome analysis. We have called these “regionally defined CNEs” (rCNEs) ( Figure 3 A). In mammalian genomes, the distance between the first and last element around SOX21 is over 450 kb. As is the case for a number of the larger CNEs throughout the genome, some of the CNEs around the SOX21 gene are more highly conserved than the gene's coding exon. For example, in multiple alignments of mouse, rat, human, and Fugu sequence, one CNE (SOX21_19) has 90% identity over 558 bp whilst another (SOX21_1) contains a 112-bp region of 100% identity ( Figure 3 B), demonstrating an extraordinary level of conservation for genomes separated by 900 million years of divergent evolution. Figure 3 Comparative Sequence Analysis of the SOX21 Gene SOX21 genomic regions for mouse, human, and rat were extracted from Ensembl to include all flanking DNA up to the nearest neighbouring genes (ABCC4 and NM_180989 in the human genome and their orthologues in the rodent genomes). The region covering Fugu SOX21 (138–178 kb of Fugu Scaffold_293 [M000293]) was extracted from the Fugu Genome Server at http://fugu.rfcgr.mrc.ac.uk/fugu-bin/clonesearch . (A) MLAGAN alignment of the SOX21 gene using Fugu DNA as the base sequence compared with mouse, rat, and human genomic DNA. Coloured peaks represent regions of sequence conservation above 60% over at least 40 bp. The SOX21 coding region (SOX21 is a single exon gene) is annotated, and sequence identity is shaded in blue. Non-coding regions of sequence identity are shaded in pink. The eight elements that have been functionally assayed are labelled. Six of these are identified in the global analysis as seven CNEs (SOX21_8–10 covers two CNEs). SOX21_7 and SOX21_18 are rCNEs. (B) Multiple DNA sequence alignments of CNE SOX21_1 and CNE SOX21_19 between mouse, rat, human, and Fugu . Finally we searched invertebrate sequence databases, including the whole-genome sequences of Ciona intestinalis, Drosophila melanogaster, and Caenorhabitis elegans, to see whether we could identify any of these highly conserved vertebrate sequences within the invertebrate lineage. Although many of the genes identified in our analysis have clear homologues within these genomes, we found no significant matches to any CNEs. More sensitive alignment using MLAGAN also failed to identify any conserved non-coding sequence similarity between vertebrates and non-vertebrates (including C. elegans , D. melanogaster and A. gambiae ), whilst in each case the coding sequences were identified. This is surprising, given that the degree of identity between CNEs in vertebrates is higher than that of the coding regions for these genes. Thus, it is unlikely that the same set of sequences that appear to regulate important vertebrate trans-dev genes are found in invertebrates. Functional Assay We have assayed the ability of conserved non-coding sequences identified both from the whole-genome MegaBLAST analysis (CNEs) and from regional MLAGAN alignments (rCNEs) to up-regulate green fluorescent protein (GFP) reporter expression in zebrafish embryos (see Materials and Methods ). We chose four cluster regions that contain different types of developmental genes: SOX21, PAX6, HLXB9, and SHH. Elements are co-injected with a minimal promoter–GFP reporter construct into early zebrafish embryos. This co-injection strategy [ 37 , 38 ] is an efficient, yet simple and rapid method for identifying enhancer activity; indeed enhancer activity of elements is more striking when tested in a co-injection assay than when ligated directly to a promoter–reporter construct [ 37 ]. A total of 25 conserved non-coding regions were selected ( Figures 3 , 4 , and S1 ), of which ten were CNEs and 15 were rCNEs ( Table 1 ). GFP expression was analysed in live embryos on the second day of development and recorded both schematically and in tabular form. A mean of 188 embryos were screened for each element, compared with a mean of just over 200 embryos per control ( Table 1 ). Figure 4 MLAGAN Alignments of Regions Encompassing the PAX6, HLXB9, and SHH Genes PAX6 (A), HLXB9 (B), and SHH (C). In each panel, human (top), mouse (middle), and rat (bottom) genomic DNA from Ensembl is aligned with Fugu genomic DNA from orthologous regions. Alignment parameters are the same as in Figure 2 . Seventeen elements that have been functionally assayed from these regions have been labelled. The following were identified as CNEs: PAX6_6, PAX6_9–10, KIAA0010_1, and KIAA0010_3. Table 1 Elements Used in Functional Assay a In some cases, two conserved regions that are very close together have been included in one PCR. In this case, the length of each element is given with the region they span in parenthesis N/A, not applicable Controls in which no element was injected (GFP reporter construct injected alone), in which non-conserved, non-coding genomic DNA from the PAX6 or SOX21 regions was co-injected with the GFP reporter, or in which conserved, coding DNA from PAX6, SOX21, or SHH exons was co-injected with the GFP reporter produce essentially no up-regulation of GFP expression ( Table 1 ; Figure S1 ). When conserved non-coding sequences were injected, up-regulation of GFP expression was observed with all but two of the elements tested, with between 4% and 44% of embryos screened being positive ( Table 1 ). Furthermore, GFP expression was generally observed in consistent patterns, specific to the element injected ( Figure 5 ). Figure 5 Composite Overviews of GFP Expression Patterns Induced by Different Elements Tested in the Functional Assay Cumulative GFP expression data, from SOX21 -associated elements (A), PAX6 -associated elements (B), HLXB9 -associated elements (C), and SHH -associated elements (D). Cumulative data pooled from multiple embryos per element on day 2 of development (approximately 26–33 hpf) are displayed schematically overlayed on camera lucida drawings of a 31-hpf zebrafish embryo. Categories of cell type are colour-coded: key is at bottom of figure. Bar graphs encompass the same dataset as the schematics and use the same colour code for tissue types. Bar graphs display the percentage of GFP-expressing embryos that show expression in each tissue category for a given element. The total number of expressing embryos analysed per element is displayed in the top left corner of each graph. Legend for the bar graph columns accompanies the bottom graph in each panel; “blood+” refers to circulating blood cells plus blood island region, “heart+” refers to heart and pericardial region (Please note: Some cells categorised as heart/pericardial region may be circulating blood cells), and “skin” refers to cells of the epidermis or EVL. s. cord, spinal cord. In order to build up a comprehensive picture of the GFP expression pattern induced by each of the elements, the expression profiles from multiple embryos positive for a given element were overlaid onto a schematic diagram, so providing a composite overview for each element ( Figure 5 ). This also provided a convenient format for data storage and comparison between elements. SOX21-associated elements Of the eight SOX21-associated elements tested in our functional assay, seven enhance GFP expression ( Table 1 ). Three of these enhancing elements direct reporter gene expression most prominently to the central nervous system (CNS) (SOX21_4 and SOX21_19 [ Figures 5 A, 6 A, and 6 B] and SOX21_7). SOX21_19 strongly directs remarkably widespread GFP expression throughout the brain and rostral spinal cord (88% of expressing embryos show GFP-positive cells in the CNS; Figures 5 A and 6 B). SOX21, a member of the SRY-related HMG-box (SOX) gene family of DNA-binding proteins, acts as a transcriptional repressor during early development [ 52 ], and is expressed in a complex, dynamic pattern in the developing vertebrate CNS [ 53 , 54 , 55 ]. Figure 6 Different Elements Enhance GFP Expression in Specific Tissue and Cell Types GFP expression is shown in fixed tissue following wholemount anti-GFP immunostaining, bright-field views (A–D, F, J, K, and N), or in live embryos as GFP fluorescence, merged bright-field and fluorescent views (E, G–I, L, M, and O). Lateral views, anterior to the left, dorsal to the top (A, B, and D–O) or dorsal view, anterior to the top (C). Embryos approximately 28–33 hpf (A, D–I, L, and O), approximately 48 hpf (B, C, J, K, and N), or approximately 26 hpf (M). The identity of the element co-injected with the GFP reporter construct is shown at the bottom of each panel. Black arrows indicate the approximate position of the midbrain–hindbrain boundary; black and white arrowheads indicate GFP-expressing cells. Scale bars approximately 100 μm (A–E, G–I, and L–O) and 50 μm (F, J, and K). b, blood island; d, diencephalon; e, eye; f, fin fold; hb, hindbrain; l, lens; n, notochord; ov, otic vesicle; r, retina; s, somite; sc, spinal cord; t, telencephalon; te, tectum; y, yolk. (A) SOX21_4. Head region (eyes removed): neurons in the telencephalon and diencephalon are GFP-positive (arrowheads). (B) SOX21_19. Head region: numerous GFP-expressing neurons are visible in the forebrain, midbrain, and hindbrain. Retinal expression is also apparent. (C) SOX21_5–6. Hindbrain region: white arrowheads indicate GFP expression by several cells in the epithelium of the right developing ear (ov). GFP-expressing cells in left deveoping ear are in slightly different focal plane. (D) SOX21_1. Trunk region: two individual notochord cells express GFP (arrowheads). (E) PAX6_6. Head region of live embryo: GFP is expressed in several retinal cells. (F) PAX6_9–10. Anterior trunk region (at the level of somites 1–3): three spinal cord neurons with ventrally projecting axons express GFP (arrowheads). (G) PAX6_1. Tail region of live embryo: arrowhead indicates GFP expression in the developing median fin fold. (H) KIAA0010_1. Trunk region, three notochord cells express GFP (arrowheads). (I) KIAA0010_2. Anterior end of embryo: arrowheads point to circulating blood cells expressing GFP. (J) HLXB9_3. Trunk region: GFP-expressing muscle fibres in somite 5 (arrowheads) lie immediately dorsal and ventral to the horizontal myoseptum. (K) HLXB9_3. Trunk region (at the level of somites 13–15): arrowheads mark GFP expression in six cells forming the epidermis or EVL. (L) SHH_6. Whole live embryo: numerous GFP-expressing muscle fibres can be seen in the trunk. (M) SHH_1. Tail region of live embryo: GFP is expressed in a single bipolar neuron near the caudal end of the spinal cord (arrowhead marks cell body). (N) SHH_4. Head region (dorsolateral view): cells labelled with anti-GFP include midbrain and hindbrain neurons and cells in the retina (slightly out of focal plane). Arrowheads indicate cell bodies of hindbrain neurons, from which axons can be seen projecting ventrally. (O) SHH_2. Trunk region of live embryo: GFP-positive cells in the region of the blood islands (caudal to the urogenital opening; arrowheads) show a slightly elongated morphology, suggesting they may be blood vessel precursors rather than blood cells. Three elements strongly enhance GFP expression in the sense organs: SOX21_4 and SOX21_19 direct GFP expression to the developing eye (in 52% and 27% of expressing embryos, respectively; Figures 5 A and 6 B), and SOX21_5–6 strongly enhances reporter expression in the developing ear (75% of expressing embryos; Figures 5 A and 6 C). These observations draw parallels with prominent regions of endogenous SOX21 expression in the sense organs: i.e., the nasal epithelium, the lens and retina of the eye, and the sensory epithelia of the developing inner ear [ 55 ]. SOX21_1 strongly enhances expression in the notochord (62% of expressing embryos; Figures 5 A and 6 D), a domain not normally associated with SOX21 expression. PAX6-associated elements Six out of seven PAX6-associated elements tested in our functional assay enhance GFP expression ( Table 1 ). Four of these six functional elements direct GFP expression most frequently to the developing eye (PAX6_6, 90% of expressing embryos; PAX6_19, 59% of expressing embryos [ Figures 5 B and 6 E]; PAX6_2, 92% of expressing embryos; and PAX6_4, 100% of expressing embryos). A fifth element, PAX6_9–10, also directs reporter gene expression to the eye in a significant proportion (25%) of expressing embryos ( Figure 5 B) as well as to neurons most frequently in the hindbrain and spinal cord ( Figures 5 B and 6 F). Significantly, PAX6 is a paired-box-containing transcription factor, expressed in and playing essential roles in the developing eye; it is also expressed in the forebrain, hindbrain, and spinal cord (data from the Zebrafish Information Network; http://zfin.org ). PAX6 is associated with the loss-of-function disorder aniridia. Some aniridia cases show chromosomal rearrangements downstream of an intact PAX6 gene, indicating that cis -acting elements can influence PAX6 gene expression in the eye at a significant distance from the coding region [ 56 ]. Indeed, PAX6 expression is known to be influenced by cis -acting elements in upstream, intronic, and downstream positions. For example, 5′ elements drive expression in the lens, pancreas, and parts of the neural tube [ 27 ], intronic elements drive expression in the retina, forebrain, and hindbrain [ 27 , 57 ], and several 3′ regions direct expression to the developing pretectum, neural retina, and olfactory region [ 58 ]. In addition to the eye and CNS, other tissues to which GFP expression is directed by our PAX6-associated elements include the blood islands (PAX6_9–10, 36% of expressing embryos; PAX6_1, 16% of expressing embryos [ Figure 5 B]) and the median fin fold (PAX6_1, 55% of expressing embryos; Figures 5 B and 6 G); these tissues have not been associated with endogenous expression of PAX6. HLXB9-associated elements We assayed six elements associated with a genomic region containing the HLXB9 and KIAA0010 genes ( Table 1 ). Each of these elements induces GFP expression in a variety of tissues (data from four elements are shown in Figure 5 C). Most notably, KIAA0010_1 directs GFP expression to the notochord in more than 87% of expressing embryos ( Figures 5 C and 6 H), KIAA0010_2 directs expression to the blood (38% of expressing embryos; Figures 5 C and 6 I) and the pericardial region (36% of expressing embryos; Figure 5 C), HLXB9_1 directs expression to the skin/enveloping layer (EVL; 52% of expressing embryos) and skeletal muscle (40% of expressing embryos; Figure 5 C), HLXB9_3 directs expression to skeletal muscle (48% of expressing embryos; Figures 5 C and 6 J) and to skin/EVL (33% of expressing embryos; Figures 5 C and 6 K), and HLXB9_2 directs expression to the spinal cord (87% of expressing embryos). HLXB9 is a Mnx-class homeobox gene associated with autosomal dominant caudal defects [ 59 ]. The zebrafish orthologue, hb9, is expressed in the notochord, hypochord, tail mesoderm, and tailbud [ 60 ], paralleling some of the domains of GFP expression induced by HLXB9/KIAA0010-associated elements. SHH-associated elements Two of the four SHH-associated elements tested in this study ( Table 1 ) direct GFP expression most frequently to muscle cells (SHH_1, 46% of expressing embryos; SHH_6, 83% of expressing embryos [ Figures 5 D and 6 L]). All four elements also prominently direct GFP expression to the CNS (SHH_1, 64% of expressing embryos; SHH_2, 42%; SHH_4, 57%; and SHH_6, 48% [ Figures 5 D, 6 M, and 6 N]). The SHH signalling molecule is crucial for a number of developmental processes, and is extensively implicated in disease (reviewed in [ 61 ]). In zebrafish, shh and its co-orthologue twhh are both expressed predominantly in midline structures, i.e., floorplate and notochord. Later expression domains include the branchial arches, pectoral fin buds, and the retina [ 62 , 63 ]. GFP expression directed by SHH-associated elements and shh/twhh expression overlap in the floorplate; however, most of the other domains of GFP expression (e.g., muscle and blood islands; Figure 6 O) are not reflected by endogenous expression of hedgehog genes. Discussion Understanding the intricate and finely tuned process of gene regulation in vertebrate development remains a major challenge facing post-genomic research. In order to begin to understand how genomic information can coordinate regulatory processes, we have adopted an approach integrating comparative genomics and a medium-throughput functional assay. Nearly 1,400 non-coding DNA sequence elements were identified that exhibit extreme conservation throughout the vertebrate lineage. Despite a degree of overlap, less than half of the non-coding ultra-conserved regions (109 out of 256) identified from the mouse and human genomes [ 21 ] are present in this set. Most, if not all, of the CNE sequences appear to be associated with genes involved in the control of development, many of them transcription factors. A significant proportion of genes identified in this study are homologous to those identified in the sea urchin and other invertebrates as master regulators of early development, leading us to believe that they interact in GRNs. Consequently, it is extremely likely that the CNEs identified compose at least part of the genomic component of GRNs in vertebrates, acting as critical regions of regulatory control for their associated genes. Such regions would mediate up- or down-regulation of expression, effecting a cascade of downstream events. In agreement with current GRN models, and given the function of many of the genes we have identified in our analysis, it is logical to speculate that CNEs consist of modules of binding sites for transcription factors. However, the model of CNEs as transcription factor binding sites, even for large numbers of transcription factors, does not fully explain their high sequence identity across vertebrates, given that transcription factor binding sites are generally rather short and exhibit a level of redundancy. Consequently, we have not ruled out the possibility that the CNEs may have a completely different mode of action or act in numerous different ways. The relative positions and order of CNEs within a cluster is completely conserved in all vertebrate genomes we have analysed (generally mouse, rat, human, and Fugu ) together with some degree of proportional compaction in the Fugu genome. This suggests that the CNEs might play a role in structuring the genomic architecture around trans-dev genes, which in turn may lead to an additional level of transcriptional control. Further evidence that genomic architecture may be important comes from the fact the trans-dev genes are generally located in regions of low gene density. Alternatively, despite the lack of EST data, it is possible that CNEs are transcribed and work at the RNA level. A number of other ideas on the evolutionary mechanisms responsible for “ultra-conservation” have been suggested [ 21 , 64 ], involving decreased mutation rate, increased DNA repair, and multiply-overlapping transcription factor binding sites, but without more functional studies such hypotheses remain speculative. Whatever their mode of action, the striking degree of conservation displayed by this set of CNEs suggests they play critically important functional roles. Having established a “map” of the major locations of CNEs in the genome, we were able to take a more sensitive alignment approach in a number of these regions in order to identify additional CNEs (rCNEs). The distinction between CNEs and rCNEs is purely a bioinformatics one, based on our search parameters, and we have no reason to believe that there is any functional distinction between the two sets of elements. We selected a number of elements (both CNEs and rCNEs) as candidates for functional analysis. Data from our functional assay of 25 elements from four different developmental genes demonstrate that a significant proportion can act as enhancers, inducing expression of a GFP reporter gene in a tissue-specific manner. The observed expression patterns differ among elements, but are reproducible for individual elements. Enhanced GFP expression domains frequently coincide with endogenous expression domains of the trans-dev gene most closely associated with a particular element, although in several instances, expression of GFP was induced in a tissue in which the most closely associated developmental gene is not normally expressed . This is not surprising because we are assaying elements out of context and individually. Thus, in our assay, we may have excluded another regulatory sequence in the region that under normal circumstances acts to silence the enhancer activity of an element in a specific tissue. Indeed GRN models would predict that a number of different regulatory regions must interact in order to precisely effect a particular spatiotemporal pattern of expression. One of our future directions will therefore be to assay the combinatorial effects of injecting a number of elements together. Alternatively, we may have associated a CNE with the wrong gene, particularly where there are two or more trans-dev genes in the same region (see below). Whilst it is straightforward to assign CNEs unequivocally to the SOX21 and PAX6 genes based on their location in the genome, the situation is more complex for elements in the vicinity of the SHH and HLXB9 genes, which are situated in close proximity to each other in the human, rodent, and Fugu genomes. This is exacerbated by the fact that some CNEs may also be found within or around neighbouring genes. This phenomenon has been described for both the PAX6 [ 65 ] and PAX9 [ 32 ] genes, as well as for the SHH gene [ 30 ], where a long-range enhancer in the intron of a neighbouring, unrelated gene regulates SHH expression in developing limb buds and demonstrates the large genomic distances over which regulatory regions may act. This enhancer is identified as a CNE in our dataset and, despite its established mode of action, is located much closer to the HLXB9 gene (200 kb in human and 12 kb in Fugu ) than to SHH (1,000 kb in human and 60 kb in Fugu ). Furthermore, a number of elements are located directly 5′ of the HLXB9 gene, whilst others are found located further upstream, in introns of the next gene, KIAA0010. Although we strongly suspect that all these elements are associated functionally with the HLXB9 gene (e.g., KIAA0010_1 directs expression prominently to the notochord, an expression domain of the zebrafish HLXB9 orthologue), we cannot rule out the possibility that they may associate with the SHH gene (also expressed in the notochord), which lies a few genes downstream. There are a number of cases where a CNE cluster is located close to more than one trans-dev gene, illustrating the value of correlating endogenous expression pattern with CNE enhancer activity. However, it should be noted that in order to build GRN maps for the elements, it is desirable but not essential to know which element associates functionally with which gene. Our confidence in the correctness of our gene assignment for the elements tested in this study is borne out by the results of our functional analysis. For the elements that we have associated with PAX6 and SOX21, there is a good correlation between tissues that express the gene endogenously and tissues induced by the associated co-injected elements to express GFP, i.e., the major sites of endogenous gene expression are highly represented in our mosaically expressing embryos (e.g., eye, hindbrain, and spinal cord for PAX6; forebrain, midbrain, hindbrain, and spinal cord for SOX21; see Figure 5 ). However, for elements in the vicinity of the HLXB9, KIAA0010, and SHH genes, GFP expression overlaps less often with expression domains of the associated gene to which the element has been assigned. As mentioned above, this reduced correlation with endogenous expression of their “associated” genes is probably due to the difficulty of assigning genes to elements in this region of relatively high trans-dev gene density. It is likely that we have missed some developmental regulators in our whole-genome analysis owing to the stringency of our search parameters. Both the RUNX2 [ 66 ] and WNT1 [ 26 ] genes, for instance, share conserved non-coding sequences in humans and fish but were excluded because they failed to satisfy our stringent whole-genome search parameters. We may also have missed some elements because they were inadvertently hidden during the process used to mask coding sequence. Nevertheless, this is the first comprehensive attempt to identify the most highly conserved non-coding sequences common to all vertebrates. The use of the compact Fugu genome sequence, with its large evolutionary divergence from mammals, was critical in providing an exceptionally low degree of background noise in comparisons at the level of whole-genome and genomic regions. As with any high-throughput approach, our functional screen has limitations. Since there is a negligible background level of GFP expression from our reporter construct alone, as well as from our other negative controls (see Table 1 ), the expression we see is most likely to be directly attributable to the enhancer properties of the CNEs. However, since GFP is a relatively stable protein [ 67 ], down-regulation of expression will not be detected during the time course of this screen; instead, expression of GFP by a particular cell indicates that expression was stimulated at some previous point in that cell's lineage. False negatives are a further limitation of the assay, e.g., tissues that develop from few cells will be under-represented and late-developing tissues or cell types (after 24 h) will be missed completely in this screen, since there is a delay between the time of onset of GFP transcription and the time when GFP fluorescence is detectable. The proportion of screened embryos that showed GFP expression varied from around 4% (SOX21_21) to around 44% (SHH_6); this is probably due to many factors, e.g., variations in the embryonic stage at the time of injection and stochastic variations from embryo to embryo with regard to which cells the injected DNA is segregated into during cleavage. However, by combining expression data from a number of expressing embryos (an average of 30 embryos per positive element), we can gain insight into the overall pattern of reporter gene expression prescribed by each element. In addition to seeing GFP expression in “expected” domains (with respect to the associated gene), GFP expression was also often detected in tissues in which the associated gene is not normally expressed (e.g., muscle cells for SHH_6 and notochord for SOX21_1; see Figure 5 ). This might be due to incorrect association of gene to element (see above); alternatively, it might reflect the importance of genomic context for function of CNEs and rCNEs. It is possible that certain regions of the genome function as silencers or suppressors, repressing the transcription-stimulating activity of enhancer elements. In our assay we are testing the autonomous enhancing function of our CNEs independent of their normal genomic context. Whilst this enables us to screen rapidly for function in an unconstrained context, it might also result in a loss of the endogenous negative constraints. It will be interesting to determine the combinatorial language of CNEs, and to uncover the importance of genomic context for their function. Conserved non-coding sequences are likely to function as negative as well as positive regulatory elements. Indeed, it is possible for a conserved non-coding element to act as either an enhancer or repressor of transcription depending on what factors are bound to it [ 68 ]. Whether any of our CNEs can function as negative regulatory elements is an interesting question that is beyond the scope of the present study. Zebrafish are the ideal model vertebrate for this screen. These sequences are, by definition, highly similar between mammals and fish, and the data generated are therefore relevant to any vertebrate. Given that CNE DNA can easily be generated from any vertebrate species (given its high degree of sequence identity), subtle differences between CNE sequences may be tested functionally in this system. Zebrafish embryos are both readily produced and easily visualised, allowing convenient live screening throughout development. Their transparency makes the embryos ideally suited to GFP analysis and the problems associated with mosaicism in this screen are relatively easily overcome by injecting large numbers of embryos. Technical advances, such as the use of meganuclease injection, may facilitate this further. The combination of a comparative genomics approach together with functional screening of conserved elements produces a large and complex dataset. Efficient access, integration, and interrogation of this bioinformatics and functional data is crucial, and of increasing interest to the scientific community, to begin to characterise GRNs in vertebrates. To this end, we have submitted all CNE DNA sequences from Fugu to the EMBL nucleotide database and are developing a publicly available relational database in order to store, curate, and analyse data from this study as well as data generated from ongoing identification and characterisation of rCNEs surrounding trans-dev genes. We have identified an important set of highly conserved non-coding vertebrate sequences that associate with developmental regulators and have provided evidence that at least some of them demonstrate regulatory function. They are likely to be implicated in genetic disease, as has already been shown for the SHH gene [ 30 ]. Their distal location from coding sequence, often megabases away, makes them strong candidates as causative agents in position effect and breakpoint disorders [ 69 , 33 ]. They are amongst the most highly conserved of all sequences in vertebrate genomes yet they are completely unrecognisable in invertebrates. Given their strong association with genes involved in developmental regulation, they are most likely to contain the essential heritable information for the coordination of vertebrate development. Materials and Methods Similarity searching of non-coding sequence between Fugu and human genomes GENSCAN [ 70 ] (using a suboptimal exon probability cutoff of 0.1) and tRNA-scan-SE (release 1.1) [ 71 ] were used to predict coding exons and tRNA genes within the Fugu draft genome assembly (release 3.0; Rosalind Franklin Centre for Genomics Research Comparative Genomics Group; http://fugu.rfcgr.mrc.ac.uk/ ). These predicted sequences were then masked in the Fugu sequence by supplying them as a “repeat library” to Repeatmasker35. The masked sequence was similarity searched against human genomic sequence from the Ensembl [ 41 ] database v18.34.1 in 1-Mb sections using MegaBLAST [ 40 ] version 2.2.6 (word size 20 and mismatch penalty –2). Human and Fugu sequences with alignments of 100 bp or over were selected to form the initial CNE sequence dataset. All CNEs with a significant similarity to an expressed transcript in the EMBL database or protein sequence in Swiss-Prot/TrEMBL were removed from the dataset unless located within a UTR. CNEs with significant similarity to non-coding RNAs were also removed. These were located by comparing the CNEs to the microRNA Registry [ 72 ] and the Rfam database (version 5.0) [ 73 ] using BLASTn [ 74 ]. CNEs were also searched against Rfam using the INFERNAL software. This resulted in the detection of 1 microRNA, four U1 snoRNAs, six U2 snoRNAs, three U5 snoRNAs, one U6atac RNA, three 7S RNAs, one 7Sk RNA, and one 5S RNA. The CNEs were also searched against the UTRdb ( http://www.ba.itb.cnr.it/BIG/UTRScan/ , which is a collection of functional sequence patterns located in 5′ or 3′ UTR sequences, but no significant hits were found. We used the program QRNA [ 75 ] to see whether any of the BLAST matches had a pattern of mutation consistent with RNA secondary structure. However, the known RNAs detected above had the most significant hits from this analysis. QRNA uses the mutational pattern in a pairwise alignment to detect non-coding RNAs, but in general the sequence identity of the CNEs is too high for this to be of use. Analysis of the distribution of CNEs in the human genome In order to test whether CNEs were randomly distributed, a new random location was allocated uniformly for each CNE within its chromosome. This process was repeated 1,000 times for each chromosome, and the average cluster sizes were calculated for the different distances given in Figure 1 B. These cluster sizes were then compared to the cluster sizes of the CNEs. χ 2 tests were carried out comparing the number of clusters containing five or fewer CNEs with the number of clusters containing six or more CNEs. The p -values obtained from the χ 2 test statistics on one degree of freedom are also shown in Figure 1 B. They give very strong evidence against the CNEs being randomly distributed. Identification of genes associated with CNEs The closest gene (using the transcription start site as defined in Ensembl) to the start of each CNE was determined from a list of all human genes supported by external evidence (“known” genes) downloaded using EnsMart, available from the Ensembl Web site (release 24.34e.1; http://www.ensembl.org/ ). The GOstat program was used to find statistically over-represented GOs in this group of genes [ 44 ], using the “goa_human” GO gene association database as a comparator. The minimum length of a considered GO path was five. The false discovery rate option was used to adjust for multiple comparisons. MLAGAN alignments More sensitive global alignment of the CNE regions surrounding 25 orthologous genes in human, Fugu, and other vertebrate species was carried out using the MLAGAN alignment tool kit [ 50 ]. To locate the orthologous regions in mouse and rat, local similarity searches with BLASTn were carried out using the most outlying CNE associated with each gene. The relevant genomic regions were extracted from Ensembl for human, mouse, and rat. For Fugu the genomic regions were extracted from the Medical Research Council Rosalind Franklin Centre for Genomics Research Fugu Genomics Project Web site ( http://fugu.rfcgr.mrc.ac.uk/ ) (where there is additional mapping information for scaffolds. All sequences were orientated prior to alignment so that the coding sequence of the gene was in positive orientation in all sequences. The MLAGAN alignment was visualised using the VISTA program [ 76 ], enabling the identification of conserved sequences. Because of the larger evolutionary distance between fish and mammals, conservation was measured using a 40-bp window and a cutoff score of 60% identity. Fugu was always used as the baseline sequence. Similarity searching of human CNEs against other vertebrate and invertebrate genomes To look for the presence of CNEs in other available vertebrate genomes, CNEs were similarity searched against Ensembl mouse (v19.32.2), rat (v21.3.2), chicken (v22.1.1), and zebrafish (v21.3.2) genome sequences using BLASTn with default parameters. All invertebrate sequences in the EMBL database were searched in the same way using BLASTn with non-stringent parameters (mismatch penalty –1, gap open penalty 1, word size 9, and soft masking). More sensitive alignment of flanking orthologous sequence around the SOX21 gene (up to the coding sequence of the genes on either side) from Ensembl C. elegans (v21.25), D. melanogaster (v21.3.1), and Anopheles gambiae (v21.2.2) was carried out using MLAGAN as above. Fish care Zebrafish were raised and bred and embryos staged following standard protocols [ 77 , 78 ]; stages are described as the approximate number of hours post-fertilisation (hpf) when embryos are raised at 28.5 °C. To prevent pigment formation, some embryos were raised in 0.003% phenylthiocarbamide in embryo medium from tailbud stage. Functional Assay We assayed for enhancer activity in embryos co-injected with candidate enhancer elements or control DNA and a minimal promoter–reporter construct in a method adapted from Muller and colleagues [ 37 ] as described below: For the preparation of DNA and micro-injection, CNEs, rCNEs, and negative controls were PCR-amplified from Fugu genomic DNA (see Figure S1 for PCR primer sequences; primers are represented by the first and last 20 bp of each sequence). The reporter construct consisting of EGFP (Clontech, Palo Alto, California, United States) under the control of a minimal promoter from the mouse β-globin gene, was PCR-amplified from a plasmid vector (available upon request). Amplified DNA was purified using the GFX PCR purification kit (#27–9602-01; Amersham Biosciences, Amersham, United Kingdom) or the QIAquick PCR purification kit (#28106; Qiagen, Valencia, California, United States). Element DNA or control DNA (at 150–300 ng/μl), reporter construct DNA (at 25 ng/μl), and phenol red (at 0.1%, used as a tracer) were combined and co-injected into embryos produced from natural matings between the one-cell stage and early cleavage stages, using an Eppendorf (Hamburg, Germany) FemtoJet pressure injection system. Any embryos developing abnormally were discarded before screening. For screening of embryos and data collection, on the second day of development (approximately 26–33 hpf), injected embryos were anaesthetised in Tricaine [ 77 ] and analysed for GFP expression by observation under fluorescence illumination using an Olympus (Tokyo, Japan) IX81 motorised inverted microscope. Images were captured using an FVII CCD monochrome digital camera and analySIS image-processing software. GFP-expressing cells were classified according to the following tissue categories: forebrain, midbrain, hindbrain, spinal cord, eye, ear, notochord, muscle, blood (circulating)/blood islands, heart/pericardial region (Please note: Some cells classified in this category may be circulating blood cells), epidermis/EVL, or fins. Cells that did not fall into one of these major expression categories (or that were not possible to unequivocally identify from morphology or localisation) were categorised as “other”. The location and tissue category of each GFP-expressing cell for each embryo was recorded schematically using Adobe Photoshop software (Adobe Systems, San Jose, California, United States), by manually drawing colour-coded schematised cells in appropriate positions onto an overlay of a camera lucida drawing of a 31-hpf embryo (from staging series by C. Kimmel, downloaded from “Zebrafish: The Living Laboratory”, courtesy of the Zebrafish CD Exchange Project; contact Mark Cooper at E-mail: mscooper@uwashington.edu ;data relating to tissue category was also recorded on a spreadsheet. GFP expression data were collected from between 25 and 55 expressing embryos per element injected. Cumulative overlaid schematised expression data for each element were compressed into a single JPEG file (displayed in Figure 5 ). Thus, the JPEG image for each element is designed to give an overall impression of the spatial pattern to which the element directs expression. Coupled with the accompanying graphs, the data present an overview of the spatial localisation of GFP expression as well as an idea of the number of cells per tissue in which GFP expression was detected, indicating the strength of the element's enhancing properties or the size of the cell population to which expression is directed. Anti-GFP immunostaining. Embryos were fixed in 4% paraformaldehyde and stained with rabbit polyclonal anti-GFP (#TP401 at 1/1,000 dilution; AMS Biotechnology, Abingdon Oxon, United Kingdom) using standard protocols [ 79 ] and the ABC amplification system (Vectastain; Vector Laboratories, Burlingame, California, United States). Stained embryos were cleared in glycerol, flatmounted, and observed/imaged as above. Supporting Information Figure S1 DNA Sequence Data for Functionally Assayed Regions Each sequence represents the PCR product used in the functional assay. Sequence in bold type represents the position of the conserved element or elements within the PCR product. All PCR products were generated from Fugu DNA. (61 KB DOC). Click here for additional data file. Table S1 Chromosomal Locations of Genes Associated with CNE Clusters in the Human Genome (from Ensembl) (40 KB XLS). Click here for additional data file. Table S2 Statistically Over-Represented GO Terms for Genes Located Closest to the CNEs (67 KB DOC). Click here for additional data file. Accession Numbers All 1,373 CNEs (CR846105 to CR847477) and 16 rCNEs (CR847478 to CR847493) have been submitted to the EMBL database.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526512.xml
545935	Cell surface heparan sulfate proteoglycans contribute to intracellular lipid accumulation in adipocytes	Background Transport of fatty acids within the cytosol of adipocytes and their subsequent assimilation into lipid droplets has been thoroughly investigated; however, the mechanism by which fatty acids are transported across the plasma membrane from the extracellular environment remains unclear. Since triacylglycerol-rich lipoproteins represent an abundant source of fatty acids for adipocyte utilization, we have investigated the expression levels of cell surface lipoprotein receptors and their functional contributions toward intracellular lipid accumulation; these include very low density lipoprotein receptor (VLDL-R), low density lipoprotein receptor-related protein (LRP), and heparan sulfate proteoglycans (HSPG). Results We found that expression of these three lipoprotein receptors increased 5-fold, 2-fold, and 2.5-fold, respectively, during adipocyte differentiation. The major proteoglycans expressed by mature adipocytes are of high molecular weight (>500 kD) and contain both heparan and chondroitin sulfate moieties. Using ligand binding antagonists, we observed that HSPG, rather than VLDL-R or LRP, play a primary role in the uptake of DiI-lableled apoE-VLDL by mature adipocytes. In addition, inhibitors of HSPG maturation resulted in a significant reduction (>85%) in intracellular lipid accumulation. Conclusions These results suggest that cell surface HSPG is required for fatty acid transport across the plasma membrane of adipocytes.	Background The adipocyte plays a central role in overall metabolic regulation serving as a storage depot for fatty acids and as an endocrine cell to regulate energy utilization and feeding behavior [ 1 , 2 ]. The mass of adipose tissue is maintained by a well-controlled balance of cell proliferation (hyperplasia) and increase in fat cell size (hypertrophy). Contributing to adipocyte hypertrophy is the assimilation of fatty acids into cytosolic triacylglycerol-rich lipid droplets. Fatty acids enter the adipocyte through the plasma membrane, are converted to their acyl-CoA derivatives and transported through the cytosol with the assistance of fatty acid binding proteins due to the lipophilic nature of the fatty acid hydrocarbon chain [ 3 , 4 ]. They are then reassembled into triacylglycerol units by acyltransferases. The intracellular lipid droplet that forms from the coalescence of triacylglycerols has recently been shown to associate with regulators of membrane trafficking in addition to enzymes needed for fatty acid storage and utilization, suggesting a complex and dynamic role deserving of the name adiposome [ 5 ]. Extracellular fatty acids that are available for adipocyte uptake are either 1) associated with circulating albumin, 2) hydrolyzed from triacylglycerol-rich lipoprotein particles by lipoprotein lipase, or 3) in the form of VLDL particles which can be directly internalized by adipocyte lipoprotein receptors. In the circulation, VLDL represents the major source of fatty acids for peripheral tissues in the form of triacylglycerols and provides a concentrated source of esterified fatty acids. It is interesting that in light of the well studied processes of cytosolic transport and assimilation of free fatty acids into triacylglycerol-rich storage droplets, the mechanism of transport of fatty acids across the adipocyte plasma membrane remains controversial. Two mechanisms, which are not mutually exclusive, have been proposed: one involves passive diffusion across the plasma membrane [ 6 , 7 ], the other requires protein-mediated transport [ 8 , 9 ]. Passive diffusion, which requires protonation of the fatty acid prior to entering the bilayer, has long been regarded as the major pathway for uptake of fatty acids by cells. However, recent kinetic data suggest that passive diffusion, while sufficient for cells with relatively low metabolic rates, is likely to be insufficient for cells with high fatty acid utilization such as skeletal muscle and adipose [ 10 - 12 ]. Moreover, the role of fatty acid-albumin complexes as a significant source of diffusible free fatty acids has recently been questioned, as evidence indicates that a significant transfer of fatty acids from albumin occurs only at very high and non-physiological fatty acid to albumin ratios [ 13 , 14 ]. Protein-mediated transport of fatty acids has been investigated using fatty acid binding and uptake studies [ 15 , 16 ]. These results show that fatty acid permeation demonstrates concentration-dependent, nonlinear saturation kinetics with a Km of transport of ~7 nM [ 9 ]. Moreover, uptake of long-chain fatty acids (>18 carbons) was competable [ 17 , 18 ], further suggesting a receptor-mediated process. Several cell surface proteins are expressed by adipocytes which potentially contribute to receptor-mediated uptake of extracellular fatty acids; these include CD36, fatty acid transport protein-1 (FATP1), very low density lipoprotein receptor (VLDL-R), low density lipoprotein receptor-related protein (LRP), and heparan sulfate proteoglycans (HSPG). CD36 is a cell surface glycoprotein that binds long-chain fatty acids with high affinity and demonstrates a subcellular distribution that is consistent with a role in fatty acid transport across the plasma membrane [ 19 - 23 ]. Following the induction of pre-adipocyte 3T3-L1 cells, CD36 expression increases [ 24 ] with a concomitant increase in fatty acid uptake. In vivo studies corroborate these observations as uptake of long-chain fatty acids is impaired in CD36 knock-out mice [ 25 ] or when CD36 expression is reduced by anti-sense RNA treatment [ 22 ]. Although these data indicate a significant role for CD36 in long-chain fatty acid uptake, the data further suggest that it is not sufficient to account for the entirety of fatty acid uptake by adipocytes since the absence of CD36 reduces uptake by only 50% [ 25 ]. FATP1 is a 71 kD transmembrane protein and the major FATP family member expressed in adipocytes [ 26 ]. In adipocytes, insulin is known to induce the translocation of FATP1 from a perinuclear compartment to the plasma membrane [ 27 ]. Similar to CD36, FATP1 expression enhances the uptake of fatty acids [ 26 , 28 ]. Studies using a VLDL-R knock-out animal model have suggested a role for this receptor in the accumulation of fatty acids by adipose tissue; mice lacking VLDL-R demonstrate only modest weight gain and a reduction in adipose tissue mass when placed on a high-fat diet [ 29 ]. VLDL-R mRNA is known to increase ~3–5-fold during adipocyte differentiation [ 30 ], suggesting that it may play a critical role in adipocyte physiology. The LRP is also expressed by adipocytes [ 31 - 33 ] and like VLDL-R can bind and internalize apoE-enriched VLDL particles [ 34 , 35 ]. However, its role in lipid accumulation by adipocytes has not been investigated. HSPG have been well characterized in their ability to bind and internalize apoE-enriched VLDL [ 36 - 38 ], as well as localize lipases to the cell surface through high affinity binding [ 39 , 40 ]. HSPG have also been shown to play a significant role in the hepatic clearance of lipoprotein, however, like LRP, no studies to date have investigated their role in intracellular lipid accumulation by adipocytes. To better understand the process of fatty acid transport across the plasma membrane of adipocytes, we have examined the expression levels and functional contributions of lipoprotein receptors, VLDL-R, LRP and HSPG toward intracellular lipid accumulation. Our findings suggest that cell surface HSPG play an essential role in fatty acid uptake and intracellular lipid accumulation in adipocytes. Results VLDL-R and LRP protein expression increases during adipocyte differentiation Since members of the LDL receptor family are known to play a predominant role in the uptake of lipoproteins, we chose to examine the protein expression levels of VLDL-R and LRP during adipocyte differentiation using 3T3-L1 cells [ 41 ], which are a well-established in vitro model for adipocyte differentiation [ 42 , 43 ]. 3T3-L1 cells were incubated with or without differentiation agents (dexamethasone, 3-isobutyl-1-methylxanthine, and insulin) and total protein extracts were immunoblotted with antibodies specific for VLDL-R, LRP, or the LDL receptor family-specific chaperone, receptor associated protein (RAP) (Fig. 1 ). Densitometric analysis of the resulting immunoblots indicated that VLDL-R expression increased by ~5-fold over non-differentiated, control cells. LRP expression demonstrated a more modest increase of ~2-fold over control cells. RAP showed no difference in expression between treated and non-treated cells, thus serving as a useful internal loading control. These results are consistent with previous studies which have shown an increase in mRNA levels during adipocyte differentiation for VLDL-R (3–5-fold) [ 30 ] and LRP (1.5 to 2-fold) [ 31 ]. Figure 1 Expression of VLDL-R and LRP increases following adipocyte differentiation. Pre-adipocyte 3T3-L1 cells were incubated for 4 d in the presence (+) or absence (-) of 100 ng/ml dexamethasone, 100 μg/ml 3-isobutyl-1-methylxanthine, and 1 μg/ml insulin, followed by 1 μg/ml insulin for an additional 4–8 d. Total cellular protein was obtained by detergent extraction, equal amounts of protein (20 μg/lane) were separated by SDS-PAGE and immunoblotted with anti-VLDL-R polyclonal IgG (4 μg/ml), or anti-LRP (1:2000) or anti-RAP (1:2000) antisera. The chemiluminescence image was quantitated by densitometry. Values obtained for non-treated cells were assigned as 100% for comparison with treated cells. Data shown is representative of 3 separate experiments. RAP, besides being an exocytic chaperone for members of the LDL receptor family, is also a high affinity ligand for most, if not all, members of the family [ 44 , 45 ]. This property is useful for evaluating specific receptor binding and internalization activities by cells. Cell surface expression of LDL receptor family members was compared between control, non-treated 3T3-L1 cells and differentiated cells by incubating with 125 I-RAP at 4°C (Fig. 2A ). Differentiated cells were found to bind ~7-fold more 125 I-RAP than non-treated control cells. This increase in receptor expression, as measured by ligand binding, is quantitatively consistent with that measured by immunoblotting. When 125 I-RAP was incubated with cells at 37°C to evaluate receptor internalization, we found that differentiated adipocytes internalized ~2-fold more 125 I-RAP than control 3T3-L1 cells (Fig. 2B ). Together, these results indicate that levels of both VLDL-R and LRP are increased at the cell surface resulting in increased receptor internalization activity in differentiated adipocytes. Figure 2 Differentiated adipocytes bind and internalize more RAP than pre-adipocytes. Pre-adipocyte 3T3-L1 cells and differentiated adipocytes were incubated with 125 I-RAP (500 ng/ml) in the presence (hatched bars) or absence (solid bars) of unlabeled RAP (20 μg/ml) at 4°C for 3 h (A) or 37°C for 3 h (B). For 4°C incubations, bound ligand was solubilized and directly quantitated by scintillation counting. For 37°C incubations, culture media was processed for trichloroacetic acid (TCA) precipitation and soluble radioactivity, representing degraded ligand, was quantitated by scintillation counting. Sulfated proteoglycan levels are increased in differentiated adipocytes Although HSPG are known to play a critical role in binding of lipoproteins to the cell surface [ 46 , 47 ] and additionally serve as a primary interaction site for LPL [ 48 , 49 ], little evidence has been reported as to their function in lipid accumulation by adipocytes. To begin to characterize their role in adipocyte growth, we first examined changes in overall proteoglycan synthesis during adipocyte differentiation. 3T3-L1 preadipocytes and differentiated adipocytes were incubated with 35 SO 4 to label glycosaminoglycan moieties. Cell lysates were prepared by extracting with urea and radiolabeled proteoglycans were analyzed by anion exchange chromatography. As shown in Fig. 3A , the amount of 35 SO 4 -labeled proteoglycan recovered from mature adipocytes was ~2.5-fold greater than that extracted from pre-adipocytes indicating that proteoglycan synthesis is augmented during differentiation. When the chromatographic fractions were separated by SDS-PAGE, interestingly, we found primarily high molecular weight species of proteoglycans in extracts from either pre-adipocytes or mature adipocytes (Fig. 3B ). Figure 3 Adipocyte differentiation results in increased synthesis of a high molecular weight proteoglycan that consists of a mixture of heparan and chondroitin sulfate glycosaminoglycan moieties. Pre-adipocyte 3T3-L1 cells and differentiated adipocytes were incubated for 20 h with 35 SO 4 (125 μCi/ml). (A), cells were lysed with 8 M urea and proteins were fractionated by anion exchange chromatography as described in Methods. Radioactivity in each fraction was determined by scintillation counting. (B), fractions from (A) were separated by 7% SDS-PAGE and subjected to phosphorimager analysis (upper panel, pre-adipocytes; lower panel, adipocytes). (C), 35 SO 4 -labeled proteoglycans, enriched by anion exchange chromatography, were incubated with or without heparinase I (5 Units/ml), chondroitinase ABC (5 Units/ml), or both enzymes for 20 h at 37°C. Labeled material was then fractionated by size exclusion chromatography and fractions were assessed by scintillation counting. To analyze the composition of the high molecular weight proteoglycan species synthesized by mature adipocytes, we incubated the chromatographically enriched material with heparinase I, chondroitinase ABC, or a mixture of both enzymes followed by size fractionation (Fig. 3C ). With no enzymatic digestion, the material eluted at the exclusion limit of the column. Treatment with heparinase I reduced peak fractions 6–8 by ~15% and yielded a lower molecular weight product that eluted between fractions 10–14. By contrast, chondroitinase ABC treatment reduced the high molecular weight, peak fractions 6–8 by ~65–70% and generated mostly low molecular weight products eluting between fractions 11–15. The high molecular weight material remaining after chondroitinase ABC treatment likely consists primarily of heparan sulfate moieties. This was confirmed by treating the enriched proteoglycans with both heparinase I and chondroitinase ABC which resulted in essentially a quantitative shift of labeled material from high molecular weight fractions to fractions consisting of primarily lower molecular weight products. These data provide us with two important observations; 1) major proteoglycan species synthesized by mature adipocytes are of very high molecular weight, and 2) this proteoglycan structure consists mostly of chondroitin sulfate glycosaminoglycans with a smaller percentage of heparan sulfate moieties. HSPG participate in the uptake of DiI-labeled apolipoprotein E-enriched VLDL by adipocytes The extent to which extracellular lipoproteins contribute to lipid accumulation by adipocytes has not been fully established. A recent study suggests that uptake of VLDL by adipocytes stimulates intracellular lipid accumulation [ 50 ]. To identify if either HSPG or LDL receptor family members play a role in this process, we incubated mature adipocytes with DiI-labeled apoE-enriched VLDL in the presence or absence of either heparin (to inhibit HSPG activity) or RAP (to inhibit VLDL-R and LRP function) and visualized the cells by fluorescence microscopy (Fig. 4 ). In the absence of potential competitors, mature adipocytes readily internalized DiI-apoE-VLDL into small cytosolic vesicles (upper panel). Incubation with RAP showed little or no effect on DiI-apoE-VLDL uptake (lower panel). However, incubation with heparin significantly inhibited intracellular accumulation of DiI-apoE-VLDL (middle panel) indicating that HSPG, rather than VLDL-R or LRP, play a major role in apoE-VLDL internalization. Figure 4 Heparin, but not RAP, competes for DiI-apoE-VLDL uptake by adipocytes. Differentiated adipocytes were cultured on glass coverslips and incubated with DiI-labeled apoE-VLDL (4 μg/ml) in the presence or absence (upper panel) of either heparin (500 μg/ml, middle panel) or RAP-GST (50 μg/ml, lower panel) at 37°C for 3 h. Cells were then fixed and processed for fluorescence microscopy. Left panels, phase contrast image; right panels, rhodamine filter set (550 nm excitation-573 nm emission). Magnification, 630×. HSPG are necessary for intracellular lipid accumulation in mature adipocytes 4-methylumbelliferyl-β-D-xylopyranoside (4-MUmb) and p-nitrophenyl-β-D-xylopyranoside (pNP-Xyl) are reagents that can serve as alternative acceptors within cells for heparan sulfate moieties and are thus able to compete for heparan sulfate chain addition to proteoglycan core proteins [ 51 , 52 ]. The bare proteoglycan core proteins that result from this treatment traverse to the plasma membrane but are devoid of any heparan sulfate glycosaminoglycan modifications. This in effect removes heparan sulfate proteoglycan functionality from the surface of treated cells and permits an examination of HSPG contributions to cell function. To ensure that these reagents do not interfere with the events of adipocyte differentiation, we treated 3T3-L1 pre-adipocytes with 4-MUmb or pNP-Xyl concurrently with the addition of differentiation reagents and examined the expression levels of proteins known to increase during adipocyte formation, namely VLDL-R, LRP, CD36/FAT [ 21 ], and leptin [ 53 ]. By immunoblot analysis, we found that the expression levels of these proteins were comparable between treated and non-treated cells (Fig. 5A ). Since lipoprotein lipase (LPL) synthesis is also known to increase following 3T3-L1 differentiation [ 54 , 55 ], we measured enzymatic activity of LPL in the culture supernatants of induced and non-induced 3T3-L1 cells and compared these results with those obtained from induced 3T3-L1 cells treated with either 4-MUmb or pNP-Xyl (Fig. 5B ). As expected, conversion of 3T3-L1 pre-adipocytes to mature adipocytes resulted in ~5–6-fold increase in LPL activity. Importantly, no significant difference in LPL activity was noted between normal mature adipocytes and those treated with either 4-MUmb or pNP-Xyl. These data indicate that expression of these proteins is unaffected by inhibited proteoglycan maturation and, importantly for our purposes, that addition of 4-MUmb or pNP-Xyl to cells does not prevent adipocyte differentiation. Figure 5 Inhibitors of HSPG maturation have no effect on adipocyte differentiation. 3T3-L1 pre-adipocytes were incubated with or without 3 mM pNP-Xyl or 4-MUmb concurrently with differentiation reagents for 4 d. Media was then changed to include just insulin with or without 3 mM pNP-Xyl or 4-MUmb for an additional 4–8 d. (A), cell lysates were immunoblotted with anti-VLDL-R mAb (1:50, culture supernatant), anti-LRP antisera (1:2000), or anti-CD36 IgG (5 μg/ml). For anti-leptin, proteins in culture supernatant were concentrated 10-fold by acetone precipitation prior to SDS-PAGE fractionation and immunoblotted with anti-leptin IgG (2 μg/ml). (B), following differentiation and xyloside treatment, cells were cultured for 16 h in serum-free, phenol red-free media. Culture supernatants were then assayed for LPL enzymatic activity as described in Methods. No significant difference was found in LPL levels between adipocytes treated with xylosides and untreated adipocytes (asterisk, p-value <0.01). To determine if HSPG are involved in intracellular lipid accumulation by mature adipocytes, we assessed cytosolic lipid droplet formation in induced 3T3-L1 cells following treatment with either 4-MUmb or pNP-Xyl. Again 3T3-L1 pre-adipocytes were incubated with 4-MUmb or pNP-Xyl concurrently with the addition of differentiation reagents and intracellular lipid accumulation was assessed by Oil Red O staining. Upon Oil Red staining, untreated mature adipocytes demonstrated large intracellular droplets; the characteristic morphologic feature of cytosolic lipid accumulation (Fig. 6A , left panel). By contrast, adipocytes treated with either 4-MUmb or pNP-Xyl showed little staining by the lipophilic dye (Fig. 6A , middle and right panels, respectively). To quantitate this effect, cell-associated lipophilic dye was extracted and measured by spectrophotometry (Fig. 6B ). Treatment of adipocytes with 4-MUmb or pNP-Xyl resulted in ~6–7-fold decrease in lipid accumulation. Moreover, adipocytes treated with varying concentrations of 4-MUmb or pNP-Xyl demonstrated a concentration-dependent effect of these HSPG inhibitors on intracellular lipid accumulation (Fig. 7 ). Together, these data indicate that HSPG play an essential role in lipid accumulation by adipocytes. Figure 6 Reduction of cell surface heparan sulfate glycosaminoglycans significantly reduces intracellular lipid accumulation in adipocytes. 3 mM 4-MUmb or pNP-Xyl was added to 3T3-L1 cells concurrently with differentiation reagents as in Fig. 5. Cells were then fixed with 10% formaldehyde, stained with 0.1% Oil Red O and photographed with phase contrast optics (A; magnification, 400×). (B), after image capture, cell-associated Oil Red O was extracted with 0.1 N HCl and dye was quantitated by spectrophotometry. Asterisk indicates a statistically significant difference compared to untreated adipocytes (p < 0.01). Figure 7 Reduction in intracellular lipid accumulation in adipocytes by heparan sulfate inhibitors is concentration-dependent. 4-MUmb (solid bars) or pNP-Xyl (hatched bars) was added to 3T3-L1 cells at the indicated concentrations concurrently with differentiation reagents as in Fig. 5. Cells were then fixed with 10% formaldehyde, stained with 0.1% Oil Red O, and cell-associated lipophilic dye was extracted with 0.1 N HCl and quantitated by spectrophotometry. Data shown represents an average of three experiments. Discussion Intracellular lipid accumulation is a hallmark event of adipocyte development and the major factor in adipocyte hypertrophy. This prompted us to investigate the molecular mechanism by which adipocytes take up extracellular lipid components. In the present study, we provide new information on the expression levels and functional activities of certain cell surface receptors in mature adipocytes that are known to play primary roles in lipoprotein processing and clearance. These receptors have been extensively studied in liver, the major organ for dietary lipoprotein clearance [ 56 ], however little information is available regarding their function in lipid accumulation by adipocytes. Using 3T3-L1 cells, which is a well established model of adipocyte differentiation, we show that protein expression of the lipoprotein receptors, VLDL-R and LRP, is increased during adipocyte differentiation, 5-fold and 2-fold, respectively. This increase is consistent with previous studies reporting increases in mRNA levels for these proteins [ 30 , 31 ]. The comparable increase between mRNA and protein levels also suggests that the overall increase in receptor expression following differentiation is primarily due to an increase in transcription with little or no post-transcriptional regulation. We also demonstrate that both pre-adipocytes and mature adipocytes express one major form of sulfated proteoglycan with a relative molecular mass of >500 kD. Expression of this species of proteoglycan increases ~2.5-fold during differentiation and appears to contain both heparan and chondroitin sulfate glycosaminoglycans. Furthermore, we found that uptake of DiI-labeled VLDL by adipocytes is inhibited by heparin, but not RAP, suggesting that in adipocytes HSPG play a greater role in lipoprotein uptake than do members of the LDL receptor family. Since VLDL is a major carrier of fatty acids in the form of triacylglycerols, these data suggest to us that HSPG may provide a mechanism for facilitating the uptake of fatty acids and significantly contribute to intracellular lipid accumulation. To address this hypothesis, we treated adipocytes with 4-methylumbelliferyl-β-D-xylopyranoside and p-nitrophenyl-β-D-xylopyranoside, which are competitive inhibitors of heparan sulfate chain addition, to prevent the synthesis of functional cell surface HSPG molecules. We found this treatment to effectively block intracellular lipid accumulation in adipocytes without affecting adipocyte differentiation. These findings offer a novel observation that cell surface HSPG appear to be essential for lipid accumulation and lipid droplet formation in adipocytes. However, this observation also raises the question as to the exact role of HSPG in the transport of fatty acids across the adipocyte plasma membrane. Based on the results presented in this study and those reported elsewhere, we have drawn a testable model to define how HSPG contribute to intracellular lipid accumulation by adipocytes (Fig. 8 ). HSPG are known to serve as binding sites for apoE-enriched lipoproteins [ 36 - 38 ]. In liver, HSPG are thought to either directly internalize bound lipoproteins by hepatocytes or, alternatively, localize lipoproteins to the cell surface and subsequently transfer particles to LDL receptor family members for endocytosis [ 57 ]. In the present study, we show that heparin effectively competes for uptake of DiI-labeled apoE-VLDL while RAP has little or no effect suggesting that, in adipocytes, HSPG are able to internalize apoE-VLDL independent of LDL receptor family members. A direct internalization function by HSPG is also supported by recent findings from our laboratory [ 47 ] and others [ 36 , 37 ]. Alternatively, HSPG may serve as a reaction center for triacylglycerol hydrolysis. In addition to apoE-rich lipoproteins, HSPG are also capable of binding LPL [ 39 , 40 ]. The binding function of HSPG for both apoE-rich lipoproteins and LPL can serve to co-localize enzyme and substrate and facilitate release of fatty acids in proximity of the adipocyte cell surface. Uptake of these liberated fatty acids can then be accomplished by fatty acid transport proteins such as CD36 [ 58 ] or FATP1 [ 59 ]. Figure 8 Model for HSPG activity contributing to fatty acid uptake and intracellular lipid accumulation in adipocytes. Cell surface heparan sulfate proteoglycans (HSPG) serve as primary binding sites for apoE-enriched VLDL (apoE-VLDL) and lipoprotein lipase (LPL) on adipocytes. Localization of apoE-VLDL and LPL to the cell surface can create a focal reaction center for triacylglycerol hydrolysis thereby releasing fatty acids for cellular uptake by fatty acid transporters such as FATP1 or CD36. Alternatively, similar to that proposed for hepatic clearance of apoE-VLDL and chylomicron remnants [57], initial binding to HSPG serves to concentrate lipoprotein particles at the cell surface and their uptake is mediated either by direct HSPG internalization or following their transfer to VLDL-R or LRP. The identity of this HSPG is currently unknown; however, its large relative molecular mass and composition containing both heparan and chondroitin sulfates are consistent with the structural properties of the syndecan [ 60 , 61 ] and perlecan [ 62 ] families of proteoglycans. There are four members to the syndecan family, all of which are type I transmembrane cell surface molecules. Syndecan-1 is found primarily in epithelial and plasma cells, syndecan-2 is found in endothelial cells and fibroblasts, syndecan-3 is expressed in cells of neural crest origin, and syndecan-4 demonstrates a more ubiquitous distribution, including adipose tissue [ 63 ]. bFGF treatment increases expression of syndecan-1 in 3T3 cells [ 64 ]; however its presence in adipocytes has not been investigated. Although the core proteins of the syndecan family are of modest molecular weight, they are typically modified with the addition of heparan sulfate glycosaminoglycan chains near their N-termini and chondroitin sulfate moieties attached more proximal to their membrane spanning domains, which results in very high molecular masses when resolved by SDS-PAGE, often >500 kD. Perlecan has a large complex modular core protein with a molecular weight of ~400 kD [ 62 , 65 ]. It typically contains three heparan sulfate chains near its N-terminus, but can also have chondroitin sulfate substitutions. Perlecan is a secreted proteoglycan that demonstrates a widespread distribution as a basement membrane component [ 66 ]. As both syndecan and perlecan contain heparan sulfate glycosaminoglycan chains, they are able to bind LPL with high affinity [ 67 ] and localize its enzymatic activity near the cell surface thereby serving as a reaction center for triacylglycerol hydrolysis. Moreover, studies have shown that they can also bind and internalize lipoprotein particles independent of the classical lipoprotein receptors [ 37 , 38 ]. We are currently in the process of accurately quantitating the amounts of heparan and chondroitin sulfate moieties present on these high molecular weight proteoglycans and identifying their core protein structure. Once this information is available, we will be able to confirm their role in fatty acid transport and adipocyte hypertrophy using expression inhibition procedures and obtain additional information on their role in adipocyte physiology. The availability of circulating triacylglycerol-rich lipoproteins for adipocyte utilization and storage relies on transport of these particles across the endothelial barrier. Transcytosis of albumin across endothelium has been shown to occur by a vesicular transport pathway [ 68 - 70 ] and is a likely mechanism for transport of fatty acid-albumin complexes. More recently, members of the LDL receptor family, including VLDL-R [ 71 ] and megalin [ 72 , 73 ], have also been shown to undergo transcytosis across endothelium and thus provide a mechanism for transport of triacylglycerol-rich lipoproteins into the tissues. Transport of VLDL across the endothelium is also likely to be assisted by hydrostatic and osmotic pressures within the capillary lumen. VLDL and related remnant lipoprotein particles represent the richest source of triacylglycerols in the body. The high metabolic requirements of adipocytes for fatty acids for both storage and utilization make these triacylglycerol-rich particles a suitable source of available fatty acids. The presence of a transendothelial transport mechanism via lipoprotein receptors makes VLDL particles a logical source of fatty acids for adipose growth. How HSPG might coordinate its activity with CD36 or FATP1 for fatty acid uptake or possibly with members of the LDL receptor family for assisted internalization is under current investigation. Presently, our results provide novel findings indicating that cell surface HSPG activity is necessary for lipid accumulation by adipocytes. It is anticipated that these observations will aid in our understanding of the complex mechanism leading to adipose hypertrophy and obesity, and provide a novel avenue to explore for targeted reduction of intracellular lipid accumulation Methods Materials Anti-LRP polyclonal antibody was raised against an 18 amino acid peptide from the cytoplasmic tail of human LRP [ 74 ] and anti-RAP polyclonal antibody was raised against a recombinant RAP-GST fusion protein as described [ 75 ]. Anti-VLDL-R polyclonal antibody was a generous gift from Dr. Dudley Strickland (School of Medicine, University of Maryland, Baltimore). Anti-VLDL-R monoclonal antibody-producing hybridoma was purchased from American Type Culture Collection (Manassas, VA) anti-leptin polyclonal antibody was purchased from Chemicon (Temecula, CA), and anti-CD36 polyclonal antibody (H-300) was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Dexamethasone, 3-isobutyl-1-methylxanthine and insulin were obtained from Sigma-Aldrich (St. Louis, MO). 35 SO 4 were purchased from MP Biomedicals (Irvine, CA). p-nitrophenyl-β-D-xylopyranoside and 4-methylumbelliferyl-β-D-xylopyranoside were from Calbiochem (La Jolla, CA). Heparin, heparinase I, chondroitinase ABC, p-nitrophenylbutyrate, lipoprotein lipase, and Oil Red O were purchased from Sigma-Aldrich. Optiprep was from Greiner Bio-One (Longwood, FL). DiI (1,1'-dictadecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate) was purchased from Molecular Probes (Eugene, OR). Apolipoprotein E was obtained from Calbiochem. RAP-GST fusion protein was purified as previously described [ 76 ]. Tissue culture plastics were purchased from Corning (Corning, NY) or Greiner Bio-One. Buffers, salts, and detergents were obtained from either Sigma-Aldrich or Calbiochem. Cell culture and adipocyte differentiation 3T3-L1 cells were obtained from American Type Culture Collection (Manassas, VA) and grown in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen, Carlsbad, CA) supplemented with 10% (v/v) fetal calf serum (Irvine Scientific, Santa Ana, CA), 1 mM sodium pyruvate, 100 μg/ml streptomycin sulfate, and 100 units/ml penicillin. Cells were cultured at 37°C with 5% CO 2 and passaged twice weekly. To differentiate 3T3-L1 cells into adipocytes, cells were incubated with 100 ng/ml dexamethasone, 100 μg/ml 3-isobutyl-1-methylxanthine, and 1 μg/ml insulin for 4 days, followed by 1 μg/ml insulin for an additional 4–8 days. Immunoblotting Cell lysates were prepared with 20 mM Tris pH 7.4, 150 mM NaCl (TBS) containing 1% (v/v) Triton-X100. For leptin, culture supernatants were concentrated by mixing with 3 volumes ice cold acetone, incubating at -20°C for 1 h, followed by centrifugation at 10,000 × g at 4°C for 15 min. Protein pellets were resuspended with SDS-PAGE sample buffer supplemented with 2% (v/v) β-mercaptoethanol. Proteins were then separated by SDS-PAGE and transferred to Immobilon-P (Millipore, Billerica, MA) using a wet tank transfer system (BioRad, Hercules, CA). Membranes were blocked with TBS, 0.1% (v/v) Tween-20, 5% (w/v) non-fat dry milk for 20 minutes at 23°C and incubated with the indicated antibody for 2 h at 23°C. Membranes were washed three times (10 min each) with TBS, 0.1% (v/v) Tween-20, and bound antibodies were detected with species-specific HRP-conjugated secondary antibodies (1:3000, BioRad) followed by chemiluminescence detection according to the manufacturer's instructions (Pierce, Rockford, IL). Images were captured using a Syngene GeneGnome system equipped with a Peltier-cooled 16-bit CCD camera and saturation detection. Densitometric analysis was performed using Scion Image, version 4.0.2. 125 I-RAP cell surface 4°C binding assay RAP-GST was purified [ 76 ] and labeled with Na 125 I as previously described [ 77 ]. Differentiated 3T3-L1 cells were grown on tissue culture plates precoated with 1% (w/v) gelatin and incubated with 125 I-RAP-GST (500 ng/ml) diluted into 20 mM Hepes, pH 7.4, 150 mM NaCl, 2 mM CaCl 2 , 1% (w/v) bovine serum albumin (buffer A) at 4°C for 3 h in the presence or absence of a 50-fold molar excess of unlabeled RAP-GST. Unbound 125 I-RAP-GST was removed by rinsing cells three times with cold buffer A after which cells with bound ligand were solubilized with 0.1 N NaOH. Solubilized proteins were added to EcoLume (ICN Biomedicals, Costa Mesa, CA) and subjected to scintillation counting (Packard Tri-Carb 1600CA, 73% efficiency for 125 I). Results were normalized to total cellular protein (BCA Protein Assay, Pierce, Rockford, IL). Specificity was determined as the difference between total binding (without competition) and non-specific binding (non-competable) [ 78 ]. The actual amount of ligand bound to cells was calculated as cpm ÷ the specific activity of 125 I-labeled ligand. All data points represent averages of duplicates or triplicates with standard errors of <5%. 37°C ligand degradation assay Differentiated 3T3-L1 cells were incubated at 37°C/5% CO 2 for 3 h with 500 ng/ml 125 I-RAP-GST diluted into DMEM containing 1% (w/v) bovine serum albumin in the presence or absence of a 50-fold molar excess of unlabeled RAP-GST. Media was then removed and processed for trichloroacetic acid (TCA) precipitation [ 75 ]. TCA-soluble material was added to EcoLume and subjected to scintillation counting. Degradation was calculated as TCA-soluble cpm ÷ specific activity of the radioiodinated ligand. 35 SO 4 incorporation and proteoglycan analysis 3T3-L1 pre-adipocytes and differentiated adipocytes grown in 100 mm tissue culture dishes were incubated with sulfate-free DME containing 10% (v/v) fetal calf serum and 125 μCi/ml Na 2 35 SO 4 for 20 h at 37°C/5% CO 2 . Cells were detached by incubating with phosphate buffered saline containing 5 mM EDTA for 20 min at 23°C. Following centrifugation to pellet cells (1000 × g, 4°C, 10 min), they were lysed by incubation with solubilization buffer (100 mM Tris, pH 7.5, 150 mM NaCl, 8 M urea) for 1 h at 4°C. Insoluble material was removed by centrifugation (10,000 × g, 4°C, 5 min) and supernatant was batch adsorbed onto 0.5 ml Macro-Prep DEAE Support (BioRad) (pre-washed with solubilization buffer) with gentle agitation for 1 h at 4°C. DEAE resin was applied to a 1.0 cm × 5.0 cm column and washed with 10 column volumes solubilization buffer followed by 10 column volumes 100 mM Tris, pH 7.5, 150 mM NaCl. Bound proteins were eluted with 100 mM Tris, pH 7.5, 1.0 M NaCl and 120 μl fractions were collected. Fifteen μl from each fraction was mixed with 3 ml EcoLume and subjected to scintillation counting. Peak fractions from DEAE enrichment of proteoglycans obtained from labeled differentiated adipocytes (fraction no. 3 and 4, Fig. 3A ) were pooled and diluted with dH 2 O to adjust to 150 mM NaCl. This preparation was then incubated with or without either heparinase I (5 units/ml) or chondroitinase ABC (5 units/ml) or both for 20 h at 37°C. The material was applied to a 1.0 cm × 12 cm Sephacryl S-200 HR (Amersham Biosciences, Piscataway, NJ) gel filtration column and eluted with 20 mM Tris, pH 7.4, 150 mM NaCl. 350 μl fractions were collected of which 200 μl was mixed with 3 ml EcoLume and analyzed by scintillation counting. Preparation of DiI-labeled apoE-VLDL New Zealand White rabbits were placed on a high-fat chow diet (10% peanut oil/1% cholesterol) for a minimum of 4 d, blood was drawn into in 1 mM EDTA and centrifuged at 2000 × g, 15 min to remove cells. Chylomicrons were floated by centrifuging plasma at 100,000 × g for 10 min and removed by pipetting. Plasma was then mixed with OptiPrep™ (12% iodixanol final concentration) and centrifuged at 350,000 × g for 3 h (SW55Ti rotor) with slow acceleration and deceleration. VLDL particles (density of 1.006 g/ml) were removed from the top of the gradient by pipetting. Purified VLDL was analyzed by SDS-PAGE and Coomassie R staining to confirm the presence of apoB100 (515 kD) and apoE (35 kD). Animal protocol (#2032) was approved by the University of New Mexico, Health Sciences Center Laboratory Animal Care and Use Committee. For DiI labeling, a working stock of 3 mg/ml was made in dimethylsulfoxide and 0.15 mg was slowly added to 1.67 mg VLDL (in 1.9 ml) with vortexing to rapidly mix. The mixture was then wrapped in foil and incubated for 8 h at 37°C. Unbound DiI was removed from DiI-labeled VLDL by OptiPrep gradient centrifugation as described above. DiI-labeled apoE-VLDL uptake assay 3T3-L1 cells were plated on glass coverslips and incubated with differentiation reagents as described above. After conversion to mature adipocytes, cells were rinsed twice with DMEM and incubated with DiI-VLDL (4 μg/ml) and apolipoprotein E (3 μg/ml) diluted into DMEM in the presence or absence of either heparin (500 μg/ml) or RAP-GST (50 μg/ml). After 3 h at 37°C, cells were rinsed with phosphate buffered saline, fixed with 1.5% (w/v) paraformaldehyde for 30 min, and mounted in Gelvatol (Air Products, Allentown, PA) containing 1 mg/ml p-phenylenediamine. Cells were observed with a Zeiss Axioskop microscope equipped for epifluorescence. Images were capture with a Hamamatsu digital/video camera and AxioVision software. LPL activity Treated and non-treated 3T3-L1 cells (as indicated in figure legends) were cultured overnight in serum-free DMEM without phenol red. Media was removed and combined with an equal volume of 100 mM sodium phosphate buffer, pH 7.2, 150 mM NaCl, 0.5% (v/v) Triton-X100. p-nitrophenyl butyrate in acetonitrile was added to a final concentration of 0.5 mM and absorbance at 400 nm was recorded every 10 s for 5 min using a Genesys UV Spectrophotometer. A standard curve for LPL activity was generated by plotting absorbance values obtained with varying concentration of purified LPL (from bovine milk). Quantitation of LPL activity in the individual media samples was determined from the standard curve. Oil Red O staining and quantitation Cells were fixed with 10% (v/v) formaldehyde in phosphate buffered saline for 1 h at 23°C, rinsed twice with water, then stained for 2 h at 23°C with 0.1% (w/v) Oil Red O in 75% (v/v) isopropanol, followed by rinsing twice with water to remove unincorporated dye. Stained cells were viewed and photographed using a Zeiss AxioVert microscope with phase contrast optics and Hamamatsu digital/video camera. For quantitation, cells were dried for 2 h at 37°C, followed by incubation with 100% isopropanol for 15 m at 23°C to extract bound dye. Solubilized dye was then quantitated by spectrophotometry in a BioRad Model 680 Microplate Reader by measuring absorbance at 510 nm. Statistical significance was determined by performing a paired t-test. List of abbreviations ApoE = apolipoprotein E HSPG = heparan sulfate proteoglycan VLDL = very low density lipoprotein VLDL-R = VLDL receptor LRP = low density lipoprotein receptor-related protein Syn-1 = Syndecan-1 FATP1 = fatty acid transport protein-1 RAP = receptor associated protein LPL = lipoprotein lipase DiI = 1,1'-dictadecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate 4-MUmb = 4-methylumbelliferyl-β-D-xylopyranoside pNP-Xyl = p-nitrophenyl-β-D-xylopyranoside bFGF = basic fibroblast growth factor Authors' contributions LCW carried out the majority of studies and drafted the manuscript. SC performed the xyloside titration study and DN performed the LPL assays. RAO provided the original conceptual framework for the study, carried out pilot experiments, participated in the experimental design and finalized the manuscript for submission. All authors read and approved the final version.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545935.xml
533874	Sample size for detecting differentially expressed genes in microarray experiments	Background Microarray experiments are often performed with a small number of biological replicates, resulting in low statistical power for detecting differentially expressed genes and concomitant high false positive rates. While increasing sample size can increase statistical power and decrease error rates, with too many samples, valuable resources are not used efficiently. The issue of how many replicates are required in a typical experimental system needs to be addressed. Of particular interest is the difference in required sample sizes for similar experiments in inbred vs. outbred populations (e.g. mouse and rat vs. human). Results We hypothesize that if all other factors (assay protocol, microarray platform, data pre-processing) were equal, fewer individuals would be needed for the same statistical power using inbred animals as opposed to unrelated human subjects, as genetic effects on gene expression will be removed in the inbred populations. We apply the same normalization algorithm and estimate the variance of gene expression for a variety of cDNA data sets (humans, inbred mice and rats) comparing two conditions. Using one sample, paired sample or two independent sample t-tests, we calculate the sample sizes required to detect a 1.5-, 2-, and 4-fold changes in expression level as a function of false positive rate, power and percentage of genes that have a standard deviation below a given percentile. Conclusions Factors that affect power and sample size calculations include variability of the population, the desired detectable differences, the power to detect the differences, and an acceptable error rate. In addition, experimental design, technical variability and data pre-processing play a role in the power of the statistical tests in microarrays. We show that the number of samples required for detecting a 2-fold change with 90% probability and a p-value of 0.01 in humans is much larger than the number of samples commonly used in present day studies, and that far fewer individuals are needed for the same statistical power when using inbred animals rather than unrelated human subjects.	Background Microarray technology has become an important tool for studying gene expression levels on the whole genome scale [ 1 ]. One important objective of many microarray studies is to identify differentially expressed genes between different conditions. Despite the effectiveness of the technology, microarray experiments are usually done with very few replicates due to budgetary constrains, which often results in high false positive (Type I error) and false negative rates (Type II error). For many microarray experiments, once a list of genes has been identified, intensive follow-up investigations of these genes using traditional molecular tools are often pursued. Hence, valuable resources can be wasted in pursuing genes from experiments with a high false positive rate. Increasing the sample size increases the statistical power to detect expression differences in microarray analysis while also decreasing the error rate. However, it is important to balance sample size with other experimental goals so as not to waste resources. An important issue that concerns many biologists is, therefore "how many replicates are needed to obtain a given type of result?" In general, the required sample size depends on the magnitude of the variability of the population, the magnitude of the expression change that is biologically meaningful (or desirable to detect), the power to detect the expression change, and the P-value/significance level/false positive rate. However, power and sample size have been viewed as complicated and difficult issues for microarray studies due to the large number of genes being investigated and little knowledge of the degree of natural expression variation within a population. To date, very few studies have assessed power and sample size requirements in microarray experiments. Pan et al. [ 2 ] proposed a normal mixture model to calculate the number of replicates required. In this study, the parameters were estimated using a subset of a real data set generated by cDNA arrays. This paper assumed that the replicates were independent of each other, whether they were drawn from the same individual or multiple individuals. Lee and Whitmore [ 3 ] discussed conceptual issues and presented computational methods (Analysis of Variance) for statistical power and sample size for different types of experimental designs, taking multiple testing into account. However, the data sets used to demonstrate these models contained a single pooled sample for each treatment/time point but not true biological replicates. Zien et al. [ 4 ] proposed a complex model that applies only to Affymetrix data to estimate biological variation and measurement error (normal additive and multiplication measurement error) for two sample comparisons. This study was based on 5 real Affymetrix data sets where the minimum required sample size was estimated based on a simulation study. Zien et al. [ 4 ] assumed a normal distributed additive measurement error and lognormal distributed measurement error. However, in real data, the functional form of the distribution of gene expression levels is generally unknown. The most common practice in microarray experiments is to assume normality of log transformed intensities or ratios. Pavlidis et al. [ 5 ] used a random sampling approach to evaluate the stability of the genes found to be differentially expressed between two groups from 16 published data sets; they found that the stability of some of the smaller data sets with fewer than 10 replicates was inconclusive. This approach is sound for the purpose of planning a study when pilot data is available with a large number of replicates. However, a pilot study is usually done with a small number of replicates where this is not feasible. Pair-wise comparisons between conditions/groups/treatments are frequently used in microarray studies. Parametric and nonparametric statistical methods have been proposed to identify differentially expressed genes, among which t-tests are most commonly used. This paper is intended to provide some guidelines for sample size planning for pair-wise comparisons. Normalization is an essential and important pre-processing step in microarray data analysis. To our knowledge, no previous studies using multiple data sets have pre-processed the data sets in a comparable way. In addition, previous studies did not look at the effect of inbred vs outbred populations on the variation of gene expression. In order to make the results more comparable, we make use of 7 cDNA microarray data sets and apply the same normalization method (spatial lowess). We estimate the variance by one sample t-test, paired t-tests or two sample t-tests on a gene-by-gene basis using several large expression data sets from both human, rats and mice. We then calculate the sample size required to detect a 1.5-, 2-, and 4-fold change in expression levels for the 90 th , 75 th , 50 th and 25 th percentile of genes ranked by variability at fixed settings for false positive and false negative rates. The sample size calculation provides the approximate but not exact number of replicates required for a given set of criteria. Results Data sets We estimate the standard deviation and required sample size from 1 unpublished and 6 published cDNA data sets (Table 1 ). Data set A-C and E are from human samples. Data sets D and F-G are from mouse or rat samples. Data set F is from a study that is not yet published. The raw and/or pre-processed data for the unpublished dataset with gene order randomized and without the original gene/probe identifications can be downloaded from . Removing the gene IDs does not affect the analysis in this paper in any way. The same website also provides the links for the published data sets. In addition, all the control genes were removed from the data analysis for the cDNA data sets. Table 1 cDNA microarray data sets used in the study Data set Reference # Rep # Genes Tissue type Description Hybridization A Smith et al. 2003 20 15,592 Human liver Paired HCC tumor vs adjacent non-tumor Direct hyb between tumor and non tumor B Lapointe et al. 2004 41 38627 Human prostate Paired prostate tumor vs adjacent non-tumor Indirect hyb using common reference C Chen et al 2002 48 22618 Human liver Paired HCC+HBV vs HBV Indirect hyb using common reference D Pritchard et al. 2001 6 5281 Mouse liver and kidney Paired liver vs kidney Indirect hyb using common reference E Zhao et al. 2004 36 ductal + 21 lobular 44549 Human breast lobular and ductal tumor tissue Indirect hyb using common reference F NA 6 13, 056 Mouse liver One third vs two thirds hepatectomy Indirect hyb using individual baseline G Callow et al. 2000 8 5548 Mouse liver ApoAI knock-out vs normal Indirect hyb using common reference (pool) Data set A comprises data generated from 40 liver RNA samples isolated from paired liver hepatocellular carcinoma (HCC) tumor and adjacent cirrhotic non-tumor tissue from 20 HCV infected Caucasian patients [ 6 ]. The objective of this study was to identify potential hepatocellular carcinoma markers. The microarray analysis was performed using in-house spotted human cDNA arrays containing 15,592 genes split between and spotted in duplicate on arrays HHD1 and HHD2. The duplicate sets of cDNAs within each slide were spotted side by side (panels A, B). RNA from tumor and adjacent non-tumor tissue from individual patients was co-hybridized to 2 slides (one was a dye flip of the other). Data set B was generated from 41 matched pairs of prostate tumor and non-tumor tissue hybridized to arrays spotted with 38627 cDNAs. All samples were labeled with Cy5 and co-hybridized with a common reference labeled in Cy3 [ 7 ]. The purpose of this study was to identify the difference in expression levels between normal and prostate tumor tissues. Data set C was generated from RNA isolated from paired HCC tumor and adjacent non tumor liver from 41 HBV infected patients [ 8 , 9 ] hybridized to cDNA arrays. Each RNA sample was labeled with either Cy5 and co-hybridized with Cy3-labeled reference RNA. The purpose of this study was to identify gene expression differences between HCC tumor and non-tumor liver tissue. Data set D was generated from RNA isolated from paired liver and kidney tissue from 6 male C57BL6 mice [ 9 ] hybridized to cDNA arrays. Each RNA sample was labeled with Cy5 and co-hybridized with Cy3-labeled amplified RNA from Universal Human Reference to total RNA (Stratagene). The purpose of this study was to identify gene expression differences between tissue types. Data set E was generated from RNA isolated from 36 breast ductal tumor and 16 lobular tumor tissues [ 10 ] hybridized to cDNA arrays. Each RNA sample was split in two and labeled with either Cy5 or Cy3, and co-hybridized with a common reference RNA as color flips with two replicates (4 arrays/tissue/mouse). The purpose of this study was to identify gene expression differences between ductal carcinoma and lobular carcinoma. Data set F consists of data generated from 24 liver tissue samples from 12 inbred mice (unpublished data). One third or two thirds of the liver was removed from each mouse and used as the baseline samples. At 12 hours post operation, the mice were sacrificed and the remaining liver tissue was used as the experimental sample. The aim of this study was to screen for genes potentially related to liver regeneration after hepatectomy. RNA samples from the 12 hour post-operation livers were co-hybridized with their own baseline liver samples. A total of four DNA arrays were used for each sample comparison. Two sets of arrays (MOD1 and MOD2), each containing 6528 different cDNAs spotted in duplicate (A and B) on each array were used. In addition, each comparison was done with a dye flip pair of slides. This data set made use of arrays generated at the University of Washington Center for Expression Arrays. The goal of data set G was to identify genes with altered expression in the liver tissues of two mouse models with very low HDL cholesterol levels (treatment groups) as compared to inbred control mice. The mouse model considered in this study is the Apolipoprotein AI (ApoAI) knock-out, where ApoAI is a gene known to play a pivotal roles in HDL metabolism [ 11 , 12 ]. Each cDNA array contained 5548 non-control genes or ESTs. A pool of normal RNA samples labeled with Cy3 served as the reference for all the arrays. In summary, three of the data sets (D, F-G) are from inbred mouse and rat strains respectively, and the other four data sets (A-C, and E) are from large scale studies of gene expression in humans. If all other factors (assay protocol microarray platform, data pre-processing) were equal, one might anticipate that fewer individuals would be needed for the same statistical power using inbred animals as opposed to unrelated human subjects. Background adjustment and normalization Background adjustment and normalization is necessary to remove systematic biases of non-biological origin in microarray studies. A number of methods of background correction and normalization have been proposed [ 13 , 14 ]. We used the locally written program "spot-on Image" to analyze the cDNA array data for data sets A and F. Spot-on uses a local background for each spot. The background subtracted intensity of all cDNA data sets were normalized by the spatial lowess method using the R add on package MAANOVA written by the Jackson Lab, which is available at . Estimates of standard deviation and sample size calculation The distribution of the standard deviations estimated from these 7 data sets are presented in Figure 1 . All data are log 2 transformed prior to data analysis. Figure 1A shows the standard deviation of the log ratio of the 4 paired cDNA data sets (A-D). The standard deviations of data sets E-G in Figure 1B are the common standard deviation of the log 2 ratio (sample/reference) of two independent groups. Figure 1 Histogram of standard deviation The X axis is the standard deviation, and the Y axis is the percentage of genes that has standard deviation below the value of X. All data sets were normalized by spatial lowess; (A) Data set A-standard deviation of log ratio of two groups (direct hybridization); data set B-D standard deviation of the difference of log (sample/reference) of the two groups (indirect hybridization); (B) Data sets E-G common standard deviation of (sample/reference) of the two independent groups (indirection hybridization). The required sample size of an experiment depends on the variance component ( σ ), the desired detectable fold change ( δ ), the power to detect this change (1- β , the likelihood of detecting the change or the true positive rate), and a chosen type I error rate ( α ). For microarrays, a combination of fold-change and test p-value is commonly used for selecting differentially expressed genes between two groups or conditions. In this study, sample sizes were calculated in R using the function of power.t.test. The required input parameters are the log scale fold change of interest δ ( δ = 1 in log transformed data translates into a 2 fold change in expression level, δ = 2 in log transformed data translates into a 4 fold change in expression level, etc), significance level, power, and the standard deviation (common standard deviation for two sample t-test, the standard deviation of the difference within subject for paired t-test, or the standard deviation of one sample t-test), the type of t-tests (one sample, two sample, or paired t-test), and the type of test (two sided or one sided). For example, in the case of data set A (one sample t-test), if we wish to find out the approximate sample size to detect a 2 fold change ( δ = 1) in expression level between tumor and non-tumor tissue in the 75% least variable genes ( σ <= 0.5884) with a two sided 0.001 significance level test with 90% power, we could use the following R function power.t.test(n = NULL, delta = 1, sd = 0.5584, sig.level = 0.001, power = 0.9, type = "one.sample", alternative = "two.sided") Where sd = 0.5584 is the 75 th percentile of the standard deviation of log ratio. In the case of data set G (two sample t-test), if we wish to find the approximate sample size to detect a 2 fold change ( δ = 1) in expression level between knock-out and control mice in the 75% least variable genes ( σ <= 0.3102) with a two sided 0.001 significance level test with 90% power, we could use the following R function power.t.test(n = NULL, delta = 1, sd = 0.3102, sig.level = 0.001, power = 0.9, type = "two.sample", alternative = "two.sided") Where sd = 0.3102 is the 75 th percentile of the common standard deviation of log (sample/reference). In R, for a one sample t-test or a paired t-test to have power 1- β to reject for a two sided testing and strict interpretation of tail probability with significance level α for detecting a difference of δ , the minimum number of samples or pairs is obtained by solving the following equation iteratively Power = Pr(t v, ncp < t v, α /2 ) + Pr(t v, ncp > t v, 1- α /2 ) Where ncp is the noncentrality parameter of the non-central t-distribution, and is estimated by t v, α /2 is the α /2 quantile of a central t-distribution with v degrees of freedom and v = n-1. t v, ncp follows a non-central t-distribution with v degrees of freedom and a non-centrality parameter of ncp. For a two sample t-test with equal sample sizes, if we wish to have a large enough sample to detect a difference δ (with a two-sided test and strict interpretation of tail probability with α significance level test with 1- β power), then the sample size (n) for each group is obtained by solving the following equation iteratively Power = Pr(t v, ncp < t v, α /2 ) + Pr(t v, ncp > t v, 1- α /2 ) Where ncp is the noncentrality parameter of non-central t-distribution, and is estimated by t v, α /2 is the α /2 quantile of a central t-distribution with v degrees of freedom and v = 2n - 2. t v, ncp follows a non-central t-distribution with v degrees of freedom and a non-centrality parameter of ncp. Microarray experiments usually involve a large number of genes, with variance components varying greatly across the genes. In general, the variance is higher for low expressors which make up of a large percentage of the genes (Figure 2 ). Approximately 50% of genes are called absent on the Affymetrix full genome GeneChips. It is reasonable to choose a value of variance, e.g. the median or the upper 75 th percentile of variance across all genes, and to use this as the value in the power calculations. For example, if we use the variance for the 50 th percentile, then the sample size calculations will assure us of having the desired power to detect a chosen n-fold change for all but the 50% most variable genes. In Figure 1 , we show horizontal lines at the 25 th , 50 th , 75 th and 90 th percentiles. The intersection of these lines with the "cumulative percentage of genes" provides the value of α for each data set. Additional file 1 shows the estimated sample size required to detect a 1.5-, 2-, and 4-fold change in expression level for the 90 th , 75 th , 50 th , and 25 th percentile genes for a given setting of false positive rate and power. As is expected, the required sample size increases with increasing variance, increasing power, and decreasing fold-change and false positive rate. Figure 2 Standard deviation versus log intensity Standard deviations are based on one sample t-test (data set A), paired t-test (data sets B-D), or two independent t-test (data sets E-G). A significance level (the probability of making a type I error, that is getting a false positive) of 0.05 is often employed in hypothesis testing. Thousands of genes are usually studied in microarray experiments. When more than 10,000 genes are tested independently, we would expect more than 500 genes to appear as false positives when the 0.05 significance level is applied. Hence, a smaller cut-off p-value should be used in order to reduce the number of false positives. Many multiple testing correction methods have been proposed. The simplest one is the Bonferroni correction (family wise error control) where the nominal significance level is divided by the number of tests. The Bonferroni correction is very stringent. False discovery rate (FDR) [ 15 ], the proportion of false positives among the genes that are identified as differentially expressed, is a post-data measure for controlling false positive. For the purpose of sample size planning, we suggest using the family wise type of error control. A higher false positive rate (lower false negative rate) can be employed for studies aiming to eliminating non-significant differentially expressed genes, and a smaller false positive rate can be used for those studies involving costly follow-up research. For reference, Table 2 lists the number of genes/ESTs/probes found to be differentially expressed between two conditions by different significance levels and fold changes. The number of significant genes in data sets F and G are small, possibly due to the relatively fewer number of genes included in the study and the homogeneity between conditions (data set F compare in-bred mice with two different volumes of hepatectomy; data set G compares ApoAI knock-out versus normal mice). Table 2 Significant genes/ESTs/probes called by methods used in the studies using different criteria (combination of significance level and fold changes) Data set Reference P <= 0.001 and the estimated fold change >=2 P <= 0.001 only P <= 0.01 only A Smith et al. 2003 183 1783 3590 B Lapointe et al. 2004 609 6549 10153 C Chen et al 2002 1253 4187 6197 D Pritchard et al. 2001 479 1557 1845 E Zhao et al. 2004 270 1050 3821 F NA 16 145 723 G Callow et al. 2000 6 11 77 Discussion Factors that affect sample size calculation include the magnitude of the variability of the population, the magnitude of the desired detectable expression change, the chosen power to detect the expression change, and the cut-off P-value/significance level/false positive rate. For a given study, the variability of the population being studied is fixed, and once researchers have identified the desired detectable expression change, the required sample size depends on the chosen false positive and false negative rates. The variability of human subject data is typically larger than that seen with laboratory animals and cell lines due to genetic influences on gene expression. Hence, more replicates are needed for studies that involve human subjects (or any other outbred population) than for studies with samples from an inbred population. This is readily apparent in the cDNA data in Additional file 1 (data sets A-C, and E are human samples while D and F-G are from mice). With the cDNA array data, one needs roughly 5 times as many human samples relative to mouse to detect the same magnitude of change with the same statistical power at the same significance level. This increase in the required number of samples for an outbred population has not been discussed before and has practical implications for those wishing to translate gene expression work from animal models to studies in human populations. Multiple levels of replicates are common in two color microarray experiments. Multiple arrays probed with RNA samples isolated from multiple individuals of a population/treatment/group are referred to as biological replicates. Multiple arrays hybridized using the same RNA or multiple replicates of the same gene within an array are referred to as technical replicates. Although technical replicates can improve the precision and the reliability of the measurement and provide information for quality control, biological replicates are most effective in reducing the variance of the estimate of mean difference. Biological replicates therefore increase the power to detect biologically significant gene expression differences. More importantly, when trying to identify differences between a treatment and a control group, accurate estimates of the biological variability within the groups is essential to determine if the between group differences are meaningful (by a t-test, Analysis of Variance (ANOVA) or other method). Careful experimental design is necessary to maximize the statistical power of the test [ 16 - 18 ] while balancing resource allocation. For example, dye swapping (in which pairs of RNA samples are hybridized twice with reverse dye labeling) is common in two color array experiments and is a great help in removing dye bias. However, if the experiments involve a common reference sample, which is not of biological interest and the goal is to identify gene expression differences between two groups (both of which are co-hybridized again the common reference), using twice as many independent biological replicates is preferable to dye swapped technical replicates. Caveats This paper is intended to give some guidance to those planning microarray experiments. The sample size calculations we performed provide an approximate number of replicates for a given set of criteria. Our studies were limited to a small number of published microarray studies for which the following criteria were true: 1) A reasonably large number of biological replicates were analyzed. 2) Raw data was readily available so that we could reprocess all data with the same algorithms. 3) Other potentially large sources of variability such as flow sorting, laser micro-dissection and/or multiple rounds of amplification were not present. We have only analyzed date from a limited number of tissue types – liver, prostate, breast and blood in human, liver and kidney in mouse, and mammary gland in rat. It is entirely possible that different tissue types will have larger or smaller degrees of biological variation and hence will require more or fewer samples to reach a given conclusion. In addition, lab or experiment specific methods of obtaining and processing samples may induce greater degrees of expression variation than seen in our sample data. As more large data sets become available, it will be useful to extend these studies to better define the magnitude of gene expression variation in purebred animals and in outbred humans across a variety of tissues. However, the data shown in Additional file 1 however should be sobering to those planning or reviewing an experimental protocol for microarray analysis. In these limited data sets with human samples hybridized to cDNA arrays with a common reference (B-C, E), we show that 32 samples for each group are required to detect a 2-fold change in the 75% least variable genes with 90% power and a p-value of 0.001. With a p-value of 0.01 and 90% power, at least 20 samples are required to detect a 2-fold change in the 75% least variable genes. This is a much larger number of samples than is frequently used in human microarray case-control studies designed to identify gene expression differences between two groups. Methods Data set selection, pre-processing and normalization Background adjustment and normalization is needed in microarray data analysis in order to remove non-biological variation. Intensity based normalization methods such as locally weighted least square polynomial regression (lowess) is commonly used in cDNA microarray experiments. The background subtracted intensities were normalized by the spatial lowess method using the R add on package MAANOVA written by the Jackson Lab. For the two cDNA experiments with replicate panels within each array, we normalized the two panels separately. All control genes were excluded from data analysis for data sets A-G. Estimate of variance components Pair-wise comparisons among conditions/groups/treatments of gene expression levels are common goals of microarray studies. Simultaneous comparison of more than two treatments/conditions using one way ANOVA can be advantageous. However, a significant F for a comparison of several treatments does not provide information about which particular groups differ from each other. In addition, one way ANOVA is not sensitive to treatment effects when only one or two samples out of many are quite different. T-tests are commonly used to compare individual treatments in pairs. In order to calculate power and plan sample size, one must first estimate the variance. We applied paired or two sample t-tests in this study based on the correlation between the two groups. For data set A, as the pairs of tumor and adjacent non-tumor tissue are highly correlated, we used two tailed one sample t-tests with the normalized log 2 ratio of tumor/non-tumor as the response variable. Data sets B-D were generated from paired samples using a reference design on cDNA arrays; Paired t-tests are appropriate for these three data sets. We performed two tailed, two sample (or independent) t-tests on data sets E-G with normalized log ratio as the response variable. The two sample t-tests are based on unequal variances for the two groups of samples. The variances of the data sets with paired samples are the variance of the difference. The common variance of the datasets with independent samples was estimated by the following formula: Where n 1 , n 2 are the number of observations for group 1, and group 2, respectively; and S 1 and S 2 are the standard deviation for group 1, and group 2, respectively. To simplify power and sample size calculation, and to focus our calculation on biological variance, the log ratios of the 4 technical replicates of data sets A, E, and G were averaged for each RNA pair (data set A) or sample (data sets E and G). The standard deviations across biological replicates were estimated on a gene-by-gene basis. Sample sizes were calculated using R for detecting a 1.5-, 2- or 4-fold change for the 90%, 75%, 50%, and 25% least variable genes with a range of power (0.70, 0.80, 0.90) and confidence level (0.01, 0.001, 0.0001), assuming equal sample size for the two groups. The numbers for sample size are rounded to integers. Authors' contributions JL provided the mouse liver microarray data set F prior to publication. CW performed all the analysis in this paper. RB supervised JL and CW and contributed to the design, coordination and writing. All authors read and approved of the final manuscript. Supplementary Material Additional File 1 Sample size required to detect a 1.5-, 2-, and 4-fold changes of expression level for the 90%, 75%, 50%, and 25% least variable genes for a given settings of false positive rates ( α ) and power (1- β ). This additional file shows the estimated sample size to a 1.5-, 2, and 4-fold changes of expression level for the 90%, 75%, 50%, and 25% least variable genes for a given settings of false positive rates ( α ) and power (1- β ) for all of the data sets referred in Table 1. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533874.xml
517929	Performance evaluation of commercial short-oligonucleotide microarrays and the impact of noise in making cross-platform correlations	Background Despite the widespread use of microarrays, much ambiguity regarding data analysis, interpretation and correlation of the different technologies exists. There is a considerable amount of interest in correlating results obtained between different microarray platforms. To date, only a few cross-platform evaluations have been published and unfortunately, no guidelines have been established on the best methods of making such correlations. To address this issue we conducted a thorough evaluation of two commercial microarray platforms to determine an appropriate methodology for making cross-platform correlations. Results In this study, expression measurements for 10,763 genes uniquely represented on Affymetrix U133A/B GeneChips ® and Amersham CodeLink™ UniSet Human 20 K microarrays were compared. For each microarray platform, five technical replicates, derived from the same total RNA samples, were labeled, hybridized, and quantified according to each manufacturers' standard protocols. The correlation coefficient (r) of differential expression ratios for the entire set of 10,763 overlapping genes was 0.62 between platforms. However, the correlation improved significantly (r = 0.79) when genes within noise were excluded. In addition to levels of inter-platform correlation, we evaluated precision, statistical-significance profiles, power, and noise levels for each microarray platform. Accuracy of differential expression was measured against real-time PCR for 25 genes and both platforms correlated well with r values of 0.92 and 0.79 for CodeLink and GeneChip, respectively. Conclusions As a result of this study, we recommend using only genes called 'present' in cross-platform correlations. However, as in this study, a large number of genes may be lost from the correlation due to differing levels of noise between platforms. This is an important consideration given the apparent difference in sensitivity of the two platforms. Data from microarray analysis need to be interpreted cautiously and therefore, we provide guidelines for making cross-platform correlations. In all, this study represents the most comprehensive and specifically designed comparison of short-oligonucleotide microarray platforms to date using the largest set of overlapping genes.	Background There are several commercial microarray systems currently available on the market for genome-scale gene expression analysis. Different microarray manufacturers provide distinct underlying technologies, protocols and reagents specific to each system [ 1 ]. Despite the widespread use of microarrays, much ambiguity regarding data analysis, interpretation and correlation of the different technologies exists. There is a need for standardization that will facilitate comparison of microarray data from different platforms [ 2 ]. Comparison and cross-validation between microarray platforms would greatly increase the understanding and value of the wealth of data generated from each microarray experiment [ 3 ]. A number of cross platform comparisons have reported a failure to demonstrate an acceptable level of correlation between different microarray technologies [ 4 - 7 ]. Some of the difficulties in correlating data can be attributed to fundamental differences between cDNA and oligonucleotide based microarray technologies. For example, target preparation differences and single vs. dual labeling techniques complicate the comparisons. Furthermore, cDNA arrays have difficulty in distinguishing between splice variants and highly homologous genes, while oligonucleotide arrays can make these distinctions if designed appropriately. However, when considering oligonucleotide platforms, which have widespread popularity, direct comparisons between different platforms should be less complex and more direct. We assert that differences in platform sensitivity, reproducibility and annotation cross-referencing accuracy account for a majority of the irreconcilable differences previously reported between different platforms [ 4 - 7 ]. When considering these factors we demonstrate a strong correlation between expression ratio data from two different commercially available short oligonucleotide based microarray technologies. This paper provides a comprehensive guideline for microarray analysis, interpretation and cross-platform correlation. There are two commercially available high-density microarray platforms that use short oligonucleotides for expression profiling. CodeLink (GE Healthcare formerly Amersham Biosciences, Chandler, AZ) and GeneChip (Affymetrix, Santa Clara, CA) microarray platforms utilize oligonucleotide gene target probes of 30 and 25 bases, respectively. Some of the notable differences between the GeneChip and CodeLink systems are, respectively, multiple probes vs. one pre-validated probe per gene target, two-dimensional surface vs. three-dimensional array matrix, and in situ synthesized oligonucleotides vs. pre-synthesized, non-contact oligonucleotide deposition. We restricted our comparative analysis to these two platforms because these systems are most similar with respect to oligonucleotide length, target preparation, and single color indirect labeling methodology. Since these commercial assays are similar, and systematic variables were isolated by using the same total RNA starting material for all target preparations, we expected disparity in performance to reflect differences inherent to the microarray platforms. To provide data for comparison of the platforms, five technical replicates of brain and pancreas were processed on each platform and the results were compared for reproducibility, sensitivity, and similarity of results. Standard manufacturer-recommended protocols and settings were employed to obtain the raw data from each platform. In the case of Affymetrix GeneChip, a recent cross-platform microarray evaluation [ 7 ] used two additional algorithms [ 8 , 9 ] for analysis of the GeneChip data and found the same level of discordance across the three analysis algorithms as was observed in the cross-platform microarray comparisons [ 7 ]. We therefore restricted our analysis of the GeneChip data to the Affymetrix recommended MAS 5.0 software [ 10 ]. This methodology was followed to simulate the results users would achieve by following current protocols supplied with each microarray system. Results Two different tissue types were compared in this study to ensure a large number of differentially expressed genes, and provide expression ratios across a wide dynamic range for derivation of the correlation coefficient between the two platforms. The array-to-array precision of each microarray platform was calculated from the five replicates within each tissue sample. The pair-wise array-to-array precision of each microarray platform is illustrated in Figure 1 with respective noise levels for both CodeLink and GeneChip. In these graphs all 10,763 uniquely represented genes, common between both microarray platforms, are illustrated. The GeneChip comparisons display a wider distribution relative to CodeLink at the lower end of the fluorescence detection range. While this wider distribution could be interpreted as indicating a lower level of precision relative to CodeLink, precision should only be assessed for the population of genes with expression values above the noise calculation (i.e. 'present' on the arrays being considered). Due to the variation in noise and specificity level between expression detection systems, each system must individually define its own threshold level cutoff for resultant confidence in signals above technical noise. In addition, in a multi-oligonucleotide detection system, a defined algorithm must be set to determine the method for combining individual probe data to yield a final gene expression level. Therefore, we used each manufacturer's indications for gene signals that should be considered confidently above system noise. The wider distribution observed in the GeneChip platform is within the noise population and therefore should not penalize the overall precision measurements. Qualitatively, CodeLink and GeneChip showed similar levels of precision when concordantly 'absent' genes were excluded within each platform, as illustrated by the blue data points representing the true signal range (Figure 1 ). Figure 1 Pair-wise array precision of CodeLink and GeneChip with illustration of respective noise levels. The representative scatter plots show precision of normalized expression values relative to noise. All 10,763 overlapping gene probes are represented in these plots. Values highlighted in red were concordantly 'absent' (noise) calls on both arrays compared. Orange lines show two-fold limits, while the black line represents equality. Precision measurements were calculated from signals above noise across the arrays being compared (Tables 1 and 2 ) to obtain a quantitative assessment. For the five array replicates, within each tissue, a total of 10 pair-wise combinations were made for all genes above noise (i.e. 'present'). Ratios were made in cases where the gene was called 'present' on both arrays being compared. False-change rates of CodeLink and GeneChip were calculated from each pair-wise array comparison between arrays processed with the same starting material. The percentage of ratios derived from the population of concordantly 'present' genes, which fall outside 2-fold (i.e. \|log 2 ratio\| > 1), is defined as the false-change rate. Table 1 shows the average and standard deviation of the false-change rate that was calculated for each of the 10 pair-wise array combinations within a sample. The false-change rates between microarray platforms were very similar, however the performance of CodeLink was slightly better with only 0.32% and 0.20% of ratios falling outside 2-fold for brain and pancreas, respectively. GeneChip showed 0.69% and 1.28% ratios outside 2-fold for brain and pancreas, respectively. To assess the level of tightness in the intensity distribution for each platform, we calculated the pair-wise ratio range within which 95% of all ratios fall for each platform (Table 2 ). For CodeLink, 95% of ratios are below 1.36 and 1.27 for brain and pancreas, respectively. On the other hand, 95% of GeneChip ratios are below 1.49 and 1.64 for brain and pancreas, respectively. Taken together, this data illustrates the precision for CodeLink is slightly higher than GeneChip for both samples tested. Table 1 False-change rate for GeneChip and CodeLink microarray platforms. The false-change rate is defined as the percentage of ratios, derived from the population of concordantly 'present' genes, which fall outside 2-fold (i.e. \|log 2 ratio\| > 1). The table above contains the average and standard deviation of the false-change rate, calculated across the 10 pair-wise array combinations within a sample. False-change rate was calculated from signals above noise across the arrays being compared. Array Platform Tissue AVG STDEV CodeLink Brain 0.32% 0.13% Pancreas 0.20% 0.14% GeneChip Brain 0.69% 0.27% Pancreas 1.28% 0.17% Table 2 Precision ratio summary for GeneChip and CodeLink microarray platforms. Precision measurements were calculated from signals above noise across the arrays being compared. For CodeLink, there were 7,882 and 6,603 ratios, on average, for each pair-wise array-to-array comparison, within brain and pancreas respectively. For GeneChip, there were 6,734 and 5,137 ratios, on average, for each pair-wise array-to-array comparison, within brain and pancreas respectively. For each of the 10 pair-wise combinations, the ratio range within 95% of the ratios fall was calculated. This table contains the average and standard deviation, in which 95% of ratios fall within, across all 10 pair-wise array combinations within a sample. Array Platform Tissue AVG STDEV CodeLink Brain 1.36 0.09 Pancreas 1.27 0.01 GeneChip Brain 1.49 0.05 Pancreas 1.64 0.03 In addition to pair-wise array precision, we calculated coefficients of variation (CV) for each platform as a function of intensity, across all replicates. In Figure 2 , CV is represented as a percentage calculated as the gene's signal standard deviation divided by mean signal across all array replicates. Genes that are concordantly 'absent' are shown in red. Concordantly 'absent' refers to genes called 'absent' by the manufacturer's software on all 5 replicate arrays tested. The black line represents the 100-probe moving average of all data points. The precision of all 'present' signals is similar between CodeLink and GeneChip, as illustrated by the moving-average level within the blue region. The median percent CV for the population of 'present' genes was 8% for both platforms. However, as gene intensity decreases, the average variance increases earlier in the distribution for GeneChip relative to CodeLink, as illustrated by the 100-probe moving average, at the boundary between red and blue data points. It is expected that variance would naturally increase at this boundary and since the rise in variance coincides with the level of concordantly 'absent' signals, demonstrating that noise is more than likely being identified correctly by each platform's image quantification software. Notably, Figures 1 and 2 illustrate a higher level of noise for GeneChip relative to CodeLink. Figure 2 Coefficients of variation for each platform as a function of intensity, across all replicates. Genes which are concordantly 'absent' are shown in red. The black line represents the 100-probe moving average. Differential expression ratios were compared between platforms to determine the cross-platform correlation. As shown in Figure 3 , when all 10,763 uniquely overlapping genes are compared between platforms, the correlation is weak (r = 0.62, where 'r' represents the Pearson correlation coefficient). However, when removing the population of concordantly 'absent' signals, the correlation is r = 0.70 between microarray platforms. When limiting the comparison to those values which are called 'present' on at least 3 of the 5 replicates across tissues and platforms, the correlation improves further to 0.74. If we further limit our comparison to only genes called concordantly 'present' (i.e. 'present' on all 5 replicates across both tissues and platforms) the correlation r = 0.79. Figure 3 Correlation of differential expression ratios between CodeLink and GeneChip. Pearson correlation coefficients (r) are shown for each comparison. (A.) When all 10,763 overlapping genes are compared between platforms the correlation is 0.62. (B.) All values for genes concordantly 'absent' were removed prior to making the cross-platform correlation. In this case, 3,362 genes are called 'present' on at least 1 of the 5 replicates across both tissues and platforms. (C.) 2,569 genes called 'present' on at least 3 of the 5 replicates across both tissues and platforms. The correlation improves further to 0.74. (D.) Genes called present on all 5 replicates across both tissues and platforms. For these 1,760 genes the correlation is 0.79. The improvement in the correlation coefficient from 0.62 to 0.79 achieved by excluding noise underscores the value in identifying the population of signals above noise for cross-platform comparisons. The 'volcano plots' in Figure 4 further confirm this point. Each data point represents a probe from the uniquely common set of 10,763 gene probes between the platforms relative to ratio and significance value. Data points highlighted in blue represent genes that are concordantly 'present' in both tissues. Hence, these blue data points are the genes called 'present' on all replicates across both tissues (n = 10). The mean log 10 ratio of expression (brain/pancreas) is shown on the x-axis and the p-value, from a two-tailed Student's t-tests on normalized log-transformed intensities, is shown on the y-axis. The vertical dashed lines represent 2-fold change ratios, which are commonly used in the field as significance levels for non-replicated array experiments. The horizontal dashed line represents the statistical-significance level where p = 0.01 (an uncorrected lenient level, used to error on the side of inclusion). The lower right- and left-hand corners of each graph contain the genes that showed a large fold-change but fail to achieve statistical significance (p > 0.01). GeneChip results show a larger number of genes in these regions as compared to the CodeLink data. The data points located in the upper-central region of each graph represent genes that were statistically significant (p < 0.01) despite modest fold-changes (< 2-fold). The minimal-detectable statistically significant fold-change was tighter for CodeLink relative to GeneChip as illustrated by the distance across the 'volcano' plot at the 0.01 significance level. In addition, the number of genes above the 0.01 significance level was greater for CodeLink relative to GeneChip. The distribution difference between the red and blue data points demonstrates the advantage of identifying signals above noise for making ratio calculations. Figure 4 'Volcano plots' for CodeLink and GeneChip. Each point represents a gene from the uniquely common set of 10,763 genes between platforms. Data points highlighted in blue represent genes which are concordantly 'present' in both tissues. The log 10 ratio of expression (brain/pancreas) is shown on the x-axis and the p-value, from a two-tailed Student's t-tests on normalized log-transformed intensities, is shown on the y-axis. The vertical dashed lines represent 2-fold change ratios and the horizontal dashed line represents the statistical-significance level where p = 0.01. The 'volcano' plots are translated into Venn diagrams of statistically significant differentially expressed genes for each platform in Figure 5 . Statistically significant (p < 0.01) expression ratios were determined using the entire set of 10,763 uniquely common genes between platforms. The total number of statistically significant differentially expressed genes detected by both platforms from this common set was 8,393. The intersection of the two platforms represents 50% of the total number of significantly differentially expressed genes. It is important to note that using the method described here, only probes considered above system noise are utilized for the correlation calculation. This leaves a set of probes which are discrepant calls and require further analysis to determine the accuracy of detection. The CodeLink platform called 5,322 genes concordantly present across the two tissues while the GeneChip platform called 3,691 genes (figure 5B , top panel). The union represents 2,569 concordantly present calls common to both platforms, where n = 3 or more. In addition, the set of 1,760 concordantly 'present' gene probes, across both platforms and tissues, was used to create a Venn diagram of ratios derived from signals concordantly above noise. The intersection of the two platforms represents 69% of the total number of differentially expressed genes. There are a higher percentage of commonly significantly changed genes between platforms when noise is excluded from ratio calculations. In both cases CodeLink shows a larger percentage of statistically differentially expressed genes at a p value less than 0.01. Figure 5 Venn diagrams of differential expression calls and statistical significance across both microarray platforms. A two-sample two-tailed t-test on normalized log-transformed intensities was performed for each microarray platform. (A) The entire set of 10,763 uniquely common genes between platforms was used to determine the number of statistically significant (p < 0.01) expression ratios. Genes above and below noise were included in the analysis. (B) Statistical significance determined from the set of 2,569 genes which are 'present' on at least 3 arrays in both tissues. Expression values below noise ('absent') were not included in the analysis. (C) Statistical significance determined from the set of 1,760 genes which are 'present' on all 5 arrays in both tissues (i.e. concordantly 'present'). A power analysis was conducted on each microarray platform to estimate the number of technical replicates needed to achieve a reasonable level of statistical confidence when noise was either included or excluded from the dataset (Figure 6 ). Evaluating the power of each platform at the level of technical replication allows researchers to gauge the underlying system variance before introducing biological variance in their studies. From our analysis, to achieve a power of 0.90 using all 10,763 genes, 3 array replicates are minimally necessary for CodeLink while 8 replicates are required for GeneChip. However, when noise is excluded, both CodeLink and GeneChip require only 1 array to achieve this same level of power. In fact, when noise is excluded, 1 array for both GeneChip and CodeLink has a 0.99 level of power in detecting two-fold differences in expression. The significant improvement in power by excluding noise provides considerable value to microarray users since fewer arrays are required to resolve desired differences in expression. By identifying and removing noise both systems can detect differential expression ratios less than 2-fold with a high level of power. However, more genes are lost on the GeneChip platform as a result of the higher level of noise relative to CodeLink. Additionally, when noise is excluded, 1.5-fold changes in expression can be detected, at a 0.90 power, using 2 CodeLink or 3 GeneChip technical replicates. Figure 6 Power analysis estimating the number of technical array replicates needed to achieve a reasonable level of statistical power or confidence for CodeLink (blue) and GeneChip (red) when noise was included (solid diamonds) or excluded (open diamonds). For both graphs the alpha was set at 0.01. (A) Relationship between power and arrays necessary to statistically discriminate two-fold changes in expression. To achieve a power of 0.90 using all 10,763 genes, 3 arrays are minimally necessary for CodeLink while 8 are required for GeneChip. However, when noise is excluded, both GeneChip and CodeLink require only 1 array to achieve this same level of power. In fact, when noise is excluded, 1 array for both GeneChip and CodeLink has a power of 0.99 to detect two-fold changes in expression. B.) In order to detect 1.5 fold changes in expression, at a 0.90 power, when noise is excluded, CodeLink minimally requires 2 arrays while GeneChip requires 3. The accuracy of CodeLink and GeneChip differential-expression ratios were compared to quantitative real-time PCR (qrtPCR). Microarray expression ratios were measured against results from qrtPCR for a randomly selected subset of 25 genes (Table 3 ) and plotted in Figure 7 . Both microarray platforms correlated well to this alternative expression-profiling technology with Pearson correlation coefficients of 0.92 and 0.79 for CodeLink and GeneChip, respectively. Figure 7 Accuracy of CodeLink and GeneChip differential-expression ratios relative to qrtPCR. Expression ratios for each microarray platform were measured against results from qrtPCR for a randomly selected subset of 25 genes. Pearson correlation coefficients (r) are shown for each comparison. Discussion Increased access and utilization of microarray data through core facilities and affordable commercial microarray systems is driving the need for direct comparisons of data between the different available platform technologies. The ability to exchange data across different platforms gives the research community the ability to cross-validate results and extend understanding of biological processes through integration of published data collected with different technologies. The results presented here demonstrate that we are closer to reaching this goal than previously reported [ 4 - 7 ]. We have compared two commercial platforms and in doing so present several steps required for making comparisons between short oligonucleotide microarray data sets. First, one must normalize annotation. Unfortunately, despite the completion of the human, rat and mouse genome sequencing projects, accurate and stable gene annotation information is not available. The existence of inaccurate sequence information, absence of an exact gene count, incomplete understanding of splicing variations, and the complexity of highly homologous gene sequences all contribute to the challenges of generating a controlled vocabulary for uniquely and constantly annotating genes at the present time. In addition, when considering commercially available arrays, the consumer is left to rely on the manufacturer to provide a probe with a one to one correlation to the intended gene target. Furthermore, until recently manufacturers have withheld the release of the exact probe sequences to researchers [ 7 ]. Now that with a simple disclosure agreement probe sequences from the major manufacturers are readily available to the users, discrepancies in some results will be explained by differences in actual probe and probe sets targets as defined by sequence homology. Some probes target different or multiple splice variants and some probes are not specific to a single gene, but instead, target multiple homologous genes. Since the use of GeneChip probe sequences for deriving inter-platform overlap is currently prohibited by Affymetrix for publication purposes, we needed to rely upon public annotation to determine the overlap between products rather than more informative sequence-based comparisons. We believe that the use of probe sequences will help to further refine the accuracy of the gene overlap set, and increase the already strong correlation between platforms demonstrated here. In addition, without the use of sequence information, we filtered the data to include only those probes and probe sets that identify a specific gene target or common regions of splice variants of a single gene target. Both manufacturers in some cases carry multiple probes or probe sets per target gene. Trying to determine which probes to compare in this case without the use of sequence information is nearly impossible. Therefore, only uniquely represented gene probes by both manufacturers were used for comparisons. By employing this conservative methodology, we reduce the risk of inappropriately comparing data from probes designed to detect different transcripts or genes despite having a similar annotation. Importantly, we used a common build of UniGene cluster IDs to find unique gene probes which overlap between the two products. When comparing between the two platforms using tissue ratio data without regard for noise, the correlation between platforms is not very strong (r = 0.62, Figure 3A ), similar to what was reported by Tan et al. 2003 [ 7 ]. This brings us to the second step, removing background signals. Considering background noise has random sources and sources that are different in nature for the two platforms, one would not expect to find a strong correlation when using noise values in platform comparisons. Each manufacturer warns users to be critical of confidence in calls that are below the defined threshold or considered 'absent'. Therefore, we removed noise and made correlations based only on calls that were 'present' in both tissue samples and microarray systems. Kuo et al. made a limited but similar attempt to reduce noise by using what they termed a "variance filter" [ 4 ]. Our process of filtering noise reduced the overlap of 10,763 genes to 3,362, 2,569 or 1,760 genes if one accepts 'present' calls on at least 1, 3 or all 5 of the array replicates, respectively, across both tissues and platforms. Using this methodology, however, we found a stronger ratio correlation between the two platforms (r = 0.70, 0.74 or 0.79, Figure 3B,3C,3D ). We have found that when limiting the comparison set to those probes which are uniquely represented, specific for their targets of interest, and called 'present' in the samples tested on each platform, the correlation between technologies is very reasonable for data sharing. Supporting this methodology, a recent study found a substantial improvement in the correlation between spotted long-oligo arrays and the Affymetrix platform with data filtering by removing low intensity signals below the median [ 11 ]. Interestingly, when Barczak and colleagues removed low intensity signals, the Pearson correlation coefficient improved from 0.60 to 0.80, which is in the same range as in our study [ 11 ]. Rather than removing all low intensity signals below the median, we recommend data filtering by using each manufacturer's standard software package to identify those genes which are within noise. This approach to filtering noise offers great value to microarray users since our recommendation does not require the immediate loss of 50% of the data in making cross-platform comparisons. Finally, an alternative expression-profiling technology, qrtPCR, was used to follow up on a smaller subset of the concordantly correlated set to demonstrate that the data generated here was not merely an anomaly specific to oligonucleotide arrays (Figure 7 ). Both platforms correlated well to this alternative expression-profiling technology with r values of 0.92 and 0.79 for CodeLink and GeneChip, respectively. Previous studies have found agreement between genes screened with microarray technology and subsequent qrtPCR verification of those expression measurements [ 12 , 13 ]. We are in the experimental process of using qrtPCR with a larger set of genes as an independent method to resolve discordant gene expression results between the two microarray platforms. The comparison described here parses the data into three sets: (1.) Concordantly 'present' which was used to calculate the correlation comparisons; (2.) Concordantly 'absent', where both platforms agree that the transcript is not 'present' in the samples tested; and (3.) 'Present' on one microarray platform but not the other, which are considered a separate set of discrepant results. In the studies presented here, the CodeLink platform generates a higher percentage of detectable signals above noise (Figures 1 , 2 , and 5 ). This finding is consistent for all replicate arrays across both tissues analyzed. Previously, Ramakrishnan et al. 2002 reported detection down to an estimated sensitivity level between 1:750,000 and 1:900,000 for the CodeLink platform [ 14 ]. However, biological validity of these low level calls by qrtPCR or other method have not been confirmed the results. In addition, a significant number of signals were detected by the GeneChip platform and were not detected by the CodeLink platform. Therefore, follow up studies are necessary to definitively determine which of the discordant calls are biologically relevant and which may be potential false positive calls. It would be informative to understand the underlining basis of the discordant calls. Assigning cause such as differences in sensitivity, analysis algorithms, or characteristics of the two platforms would be of great values to furthering comparative studies. Discrepant calls between the two platforms may derive from differences in the GeneChip and CodeLink platform technologies. The platforms differ in the oligodeoxyribonucleotide probe length and number of probes per gene. A microarray study, using covalently attached oligodeoxyribonucleotides, found that 30- and 35-mer oligodeoxyribonucleotides generated signals two- to five-fold higher than 25-mers [ 15 ]. Relogio et al. suggested that 30-mers offered the best compromise between sensitivity and specificity [ 15 ]. However, the GeneChip platform offers multiple probes per gene, potentially offsetting the need for longer probes through multiple hybridization points. The CodeLink platform contains one pre-validated probe per gene that was screened for performance from an original panel of three probes per gene. Previous research has demonstrated that one probe per gene is sufficient to accurately measure differential expression [ 16 ]. Having one pre-validated probe per gene rather than a panel of probes per gene on a microarray platform may be advantageous towards improving sensitivity since there is no requirement that many probes within a gene must agree for expression to be detected and called. A single probe must, however, be very accurately designed to cover the range of splice variants feasible, and must reside in an area accessible to the RNA or DNA fragments hybridizing. Variation in signals may also derive from the nature of the substrate for probe attachment. Previous publications have indicated that the use of a three-dimensional matrix coated slide results in a larger number of potential attachment sites than modified glass [ 17 - 19 ]. Stillman and Tonkinson [ 20 ] have shown higher specific hybridization signals on a three-dimensional matrix compared with glass. In addition, it has been demonstrated that the CodeLink three-dimensional matrix allows for reduced steric hindrance and increased availability of the entire oligonucleotide for hybridization with its intended target [ 21 ]. Side-by-side comparisons of the performance of the same probe set and analysis technique would be required to confirm any contribution to discrepant results observed in this study. Discrepant calls between the two platforms may also likely derive from differences in the GeneChip and CodeLink analysis algorithms. The use of mismatches on the GeneChip platform may limit detection since others have reported that, in general, one third of GeneChip mismatches are higher in signal than their perfect match counterparts [ 9 , 22 , 23 ]. Alternative analysis methodologies that do not utilize the mismatch controls may alter the discordant set, but as described earlier, there is a large potential variation in the different methodologies and a lack of a single majority method. Therefore, we chose to analyze the dataset in this study with the MAS 5.0 algorithms, as recommended by Affymetrix. It is likely that each of the aforementioned factors, in addition to annotation differences, contribute to variable results, and taken together account for the set of discrepant calls observed between the GeneChip and CodeLink platforms (Figure 5 ). Conclusions This paper highlights the value of separating signal from noise in order to improve microarray cross-platform correlations. We also demonstrate a stronger correlation between platforms than previously reported based on our data filtering and parsing methodology. We believe there is strong similarity in calls by each system and differences in sensitivity and levels of noise are largely responsible for lower levels of correlation. Furthermore, as a standardized annotation system develops and freely open access to the use of microarray probe sequences is realized, it will help clear up discrepancies on a case by case basis. Methods Array design and fabrication CodeLink UniSet Human 20 K Bioarray (Amersham Biosciences, Chandler, AZ) contains a collection of approximately 20,289 probes within a single reaction chamber on each individual slide. All oligonucleotide probes are 30 bases in length. The core of the CodeLink platform is a glass slide coated with a polyacrylamide gel matrix to create a three-dimensional aqueous hybridization environment. Modified 5'-amine-terminated oligonucleotides are deposited onto the polymer using piezoelectric dispensing robots and then covalently attached to activated functional groups within the gel matrix. Oligonucleotides are co-dispensed with a fluorescein-derivative dye, which enables scanning and inspection of every feature element on every slide after the dispensing. Additional sites are then blocked and slides are washed, rinsed and dried prior to attachment of an integrated, proprietary, polypropylene hybridization chamber. All probes appearing on the final product have been pre-validated for performance and screened from an original panel of up to three probes per gene. The HG-U133 GeneChip Set from Affymetrix (Santa Clara, CA, USA) contains 44,928 probes, on 2 chips, that represent 42,676 unique sequences from the GenBank database corresponding to 28,036 unique UniGene clusters. The GeneChip technology is based on a photolithographic in situ synthesis. Individual probes consist of 25 base DNA sequences. Target preparation and array hybridization One lot of human brain and pancreas total RNA (brain lot#033P010402009A and pancreas lot#022P0102B from Ambion) was assessed for quality using the Agilent 2100 Bioanalyzer and split equally between Amersham Biosciences in Chandler, Arizona and the Genomics Shared Service at the Arizona Cancer Center. The Affymetrix target preparations and hybridizations were performed entirely at the Arizona Cancer Center to ensure that these microarrays were run by an independent party with GeneChip expertise. In addition, an aliquot from these lots of total RNA was saved and subsequently used in qrtPCR reactions for verifying the expression profiles obtained by each microarray platform. For each Affymetrix GeneChip, double-stranded cDNA was synthesized from 5 ug of total RNA with the SuperScript Double-Stranded cDNA Synthesis Kit (Invitrogen) and dT24-T7 primer (Operon) according to the manufacturer's instructions. Biotin-labeled cRNA was prepared by in vitro transcription using the BioArray High Yield RNA Transcript Labeling Kit (Enzo). The dsDNA was mixed with 1× HY reaction buffer, 1× biotin labeled ribonucleotides (NTPs with Bio-UTP and Bio-CTP), 1× DTT, 1× RNase inhibitor mix and 1× T7 RNA polymerase. The mixture was incubated at 37°C for 5 hours. The labeled cRNA was then purified using an RNeasy mini kit (Qiagen) according to the manufacturer's protocol and ethanol precipitated. Fragmentation of cRNA, hybridization, washing, staining, and scanning were performed as described in the Affymetrix GeneChip Expression Analysis Technical Manual [ 24 ]. Briefly, the purified cRNA was fragmented in 1× fragmentation buffer (40 mM Tris-acetate, 100 mM KOAc, 30 mM MgOAc) at 94°C for 35 minutes. For hybridization with GeneChip cartridge (Affymetrix), 15 ug of fragmented cRNA was incubated with 50 pM control oligonucleotide B2, 1× eukaryotic hybridization control (1.5 pM BioB , 5 pM BioC , 25 pM BioD , and 100 pM cre ), 0.1 mg/ml herring sperm DNA, 0.5 mg/ml acetylated BSA and 1× manufacturer recommended hybridization buffer, and hybridization was performed with a GeneChip Fluidic Station (Affymetrix) using the appropriate antibody amplification, washing and staining protocol. The phycoerythin-stained array was scanned, resulting in a digital image file. In all, 5 replicates of U133A and U133B were processed for each total RNA sample. Therefore, 10 target preparation reactions were performed for each of the two tissues to generate the necessary cRNA for this study. For each CodeLink Bioarray, double-stranded cDNA and subsequent cRNA was synthesized from 5 ug of total RNA using the CodeLink Expression Assay Kit (Amersham Biosciences) according to manufacturer's instructions [ 25 ]. Briefly, cRNA was prepared by in vitro transcription using a single, labeled nucleotide, biotin-11-UTP in the IVT reaction at a concentration of 1.25 mM. Unlabeled UTP was present at 3.75 mM, while GTP, ATP, and CTP were at 5 mM. The mixture was incubated at 37°C overnight for 14 hours. The labeled cRNA was then purified using an RNeasy ® mini kit (Qiagen). Fragmentation of cRNA, hybridization, washing, staining, and scanning were performed as described [ 26 ]. Briefly, the purified cRNA was fragmented in 1× fragmentation buffer (40 mM Tris-acetate pH 7.9, 100 mM KOAc, 31.5 mM MgOAc) at 94°C for 20 minutes. For hybridization with CodeLink bioarrays (Amersham Biosciences), 10 ug of fragmented cRNA in 260 ul of hybridization solution was added to each bioarray via the Flex Chamber port and incubated for 18 hours at 37°C, while shaking at 300 r.p.m. in a New Brunswick Innova™ 4080 shaking incubator. The 10 bioarrays, in this study, were processed in parallel using the CodeLink Shaker Kit and CodeLink Parallel Processing Kit (Amersham Biosciences). Bioarrays were stained with Cy5™-streptavadin (Amersham Biosciences) and scanned using a GenePix ® 4000 B scanner (Axon Instruments). Deriving expression values and classifying probes within noise ('absent') for each platform For the U133 GeneChip technology, each gene is represented by 11 probe pairs containing both a perfect match probe (PM) and a mismatch probe (MM) where the middle (13 th ) base of each 25-mer probe is incorrect. The MM probe is designed to give an indication of the degree of nonspecific hybridization [ 26 ]. The MAS 5.0 software uses both PM and MM values for the expression calculation, one that avoids the production of negative values. MAS 5.0 employs a scenario-based approach to expression calculations and in general hypothesizes that MM probes should show lower hybridization signal than the corresponding PM probes. A decision process is used when this PM > MM assumption is broken. When all MM values are less than their PM counterparts, an expression value is calculated using a one-step bi-weight estimate of the log(PM – MM) values for each probe pair. However, when the MM value for a probe pair is greater than the PM value, two differing scenarios are applied. 1.) If the values of the PM probes are sufficiently large and separable from the background and MM signals, then the MM value is replaced with a value calculated as typical for the probe set. 2.) If it is difficult to separate the probe signals from background then the MM signal is substituted with a value slightly less than the PM signal. Once an expression value is calculated for each probe set the next step is the calculation of a Detection p -value and the comparison of each Discrimination score to the user-definable threshold (Tau). Tau is a small positive number that can be adjusted to increase or decrease sensitivity and/or specificity of the analysis (default value = 0.015). The One-sided Wilcoxon's Signed Rank test is the statistical method employed to generate the Detection p -value. It assigns each probe pair a rank based on how far the probe pair discrimination score is from Tau. The user-modifiable Detection p -value cut-offs, Alpha 1 (α1) and Alpha 2 (α2) provide boundaries for defining 'Present', 'Marginal' or 'Absent' calls. At the default settings (α1 = 0.04 and α2 = 0.06), any p-value that falls below α1 is assigned a 'Present' call, and above α2 is assigned an 'Absent' call. 'Marginal' calls are given to probe sets which have p -values between α1 and α2. In our study, the MAS 5.0 default parameters were retained. For a complete description of the MAS 5.0 algorithms and statistical tests please refer to the Affymetrix manuals [ 10 , 27 , 28 ]. For the CodeLink bioarrays, spot signals are quantified using ImaGene 5.5 software (BioDiscovery, Marina Del Ray, CA). The mean intensity is taken for each spot and background corrected by subtracting the surrounding median local background intensity. A spot is considered 'absent' (within noise) if the spot's signal mean is less than its corresponding local background mean plus one standard deviation of local background pixels. For each probe the local background is comprised of a circular area of pixels surrounding the segmented signal. The image segmentation and quantification process is outlined in the ImaGene 5.5 user's manual [ 29 ]. Cross-platform comparisons of expression data To facilitate comparisons between data sets, CodeLink probes and GeneChip probe sets were mapped to specific sequence clusters according to the NCBI Human UniGene build #166 relative to the manufacturer's provided NCBI accession numbers. Multiple probe or probe sets targeting a single UniGene cluster or single probe or probe sets targeting multiple clusters were removed from consideration. The overlapping and uniquely represented UniGene clusters were used to identify 10,763 gene probes for comparison between platforms. Gene-expression values were global linearly normalized according to manufacturers' standard normalization procedure [ 9 , 26 ]. The 96% trim-mean of the entire GeneChip array was used for Affymetrix normalization while CodeLink values were normalized against the array median. The globally normalized data from both platforms were scaled to 1.0 in order to bring both platforms to the same intensity range for comparative purposes. The analysis was performed using SAS statistical software and Microsoft Excel. Power analysis of CodeLink and GeneChip platforms A power analysis is a computational tool used to determine the replication needed to achieve a desired level of confidence in results from a particular experiment [ 30 - 32 ]. Determining the number of microarray replicates necessary for classification of expression profiles has been presented as an important issue [ 33 , 34 ] and should be one of the first things to consider when designing any experiment. Fore each tissue we hybridized the same target on each of five microarrays; therefore the expected fluorescence values for each independent probe should be the same from each array to array replicate, making the expected fold change equal to 1 (i.e. μ 1 = μ 2 ). The power analysis was modeled from log 2 transformed ratios derived from all pair-wise array-to-array combinations across the five replicates within the brain sample, since this tissue had the greatest similarity in performance between microarray platforms. Expression profiling of the pancreas sample showed many more genes within noise ('absent') for the GeneChip platform relative to CodeLink. The power analysis was conducted as previously described [ 35 , 36 ] for the population of all 10,763 genes within each platform and the population of genes above noise ('present'). Real-time PCR The TaqMan ® One-Step RT-PCR Master Mix Reagent Kit (Applied Biosystems, Foster City, CA, USA) was used with each custom designed, gene-specific primer/probe set to amplify and quantify each transcript of interest. Optimal primer/probe sets were selected using Primer Express software version 1.0 B6 (Applied Biosystems). Reactions (25 ul) contained 100 ng of total RNA, 300 nM forward and reverse primers, 200 nM TaqMan probe, 12.5 uL 2X Master Mix without the enzyme uracil DNA glycosylase (UNG), 0.625 mL MultiScribe™ and RNAase Inhibitor Mix, and 6.875 uL RNAse-free water. RT-PCR amplification and real-time detection were performed using an ABI PRISM 7700 Sequence Detection System (Applied Biosystems) for 30 min at 48°C (reverse transcription), 10 min at 95°C (AmpliTaq Gold activation), 38 cycles of denaturation (15 s at 95°C), and annealing/extension (60 s at 60°C). Data were analyzed using ABI PRISM Sequence Detection Software version 1.6.3 and then further processed using Microsoft ® Excel (Microsoft, Redmond, WA). Cyclophilin (PPIE) served as the endogenous control for the normalization of input target RNA. Raw C T values, qrtPCR primer/probe sequences, and corresponding array probe names are available in supplementary material [see additional files 1 , 2 , and 3 , respectively]. Competing interests RS, TJS, RL, TK-K and CP are employees of GE Healthcare. Authors' Contributions RS planned and designed the study, conducted the micoarray experiments with the CodeLink platform, analyzed the data, generated all of the figures, and drafted the paper. TJS helped with writing of the paper, deriving the overlap between platforms, provided overall technical guidance and coordination. RL designed, conducted, and analyzed the quantitative PCR experiments, performed bioinformatics support, and edited the manuscript. CP read the manuscript and provided comments. TK-K edited the manuscript. GW helped with the experimental design and was responsible for generating all of the GeneChip data as well as editing the manuscript. JA provided guidance with the statistical power analysis and additions to the manuscript. All authors read and approved the final manuscript. Table 3 List of genes evaluated using qrtPCR. For each gene, the microarray and qrtPCR brain/pancreas log 2 ratios are listed. Raw C T values, qrtPCR primer/probe sequences, and corresponding array probe names are available in supplementary material [see additional files 1, 2, and 3, respectively]. Gene NCBI Acc Description qrtPCR CodeLink GeneChip MLP NM_023009.1 MARCKS-like protein 2.33 2.08 4.20 COX7A2L NM_004718.1 cytochrome c oxidase subunit VIIa polypeptide 2 like 1.19 0.10 -0.31 COL6A3 NM_004369.1 collagen, type VI, alpha 3, transcript variant 1 -4.80 -4.75 -5.02 PRDX3 NM_006793.1 peroxiredoxin 3, nuclear gene encoding mitochondrial protein 0.82 0.81 -0.42 CDKN1A NM_000389.1 cyclin-dependent kinase inhibitor 1A -3.25 -2.89 -3.47 NUTF2 NM_005796.1 nuclear transport factor 2 1.25 1.25 -0.12 CEBPD NM_005195.1 CCAAT/enhancer binding protein, delta -0.54 0.41 -0.07 COL9A3 NM_001853.1 collagen, type IX, alpha 3 2.62 2.35 2.22 GLDC NM_000170.1 glycine dehydrogenase 5.93 3.33 2.85 TGFA NM_003236.1 transforming growth factor, alpha 2.44 1.54 1.43 GALK2 NM_002044.1 galactokinase 2 0.64 1.73 -0.39 ESR1 NM_000125.1 estrogen receptor 1 0.03 0.39 -0.28 FMO3 NM_006894.2 flavin containing monooxygenase 3 -1.06 -0.49 1.22 AKT1 NM_005163.1 v-akt murine thymoma viral oncogene homolog 1 0.14 0.40 0.34 PRPS1 D00860.1 phosphoribosyl pyrophosphate synthetase subunit I 1.53 0.62 -0.04 RPA3 NM_002947.1 replication protein A3 1.30 0.57 -0.97 SLIT2 NM_004787.1 slit homolog 2 (Drosophila) 2.81 1.64 0.51 HIC AF054589.1 HIC protein isoform p40 and HIC protein isoform p32 0.70 1.83 -0.22 HSA275986 NM_018403.1 transcription factor SMIF 1.51 1.41 -0.06 TFCP2L1 NM_014553.1 transcription factor CP2-like 1 -1.55 -1.23 -0.66 PPIE NM_006112.1 peptidylprolyl isomerase E (cyclophilin E) 0.00 -0.29 -1.10 FLJ14800 NM_032840.1 hypothetical protein FLJ14800 0.97 1.41 1.20 MGC24039 AL137364.1 cDNA DKFZp434E0626 2.70 1.92 2.29 USF1 X55666.1 late upstream transcription factor 2.18 0.81 -0.86 B4GALT7 NM_007255.1 xylosylprotein beta 1,4-galactosyltransferase, polypeptide 7 0.45 -0.41 -0.44 Supplementary Material Additional File 1 This file contains gene names, sample designations, C T values for triplicates, mean C T values, median C T values, standard deviation of replicates, and coefficient of variation of replicates for each gene evaluated by qrtPCR. Click here for file Additional File 2 This file contains gene names, oligonucleotide sequences, and oligonucleotide type (FORWARD PRIMER, HYB OLIGO, and REVERSE PRIMER) for each gene evaluated by qrtPCR. Click here for file Additional File 3 This file contains gene names, descriptions, GenBank accessions, log2 ratios (qrtPCR, CodeLink, and GeneChip), CodeLink probe names, and GeneChip probe set names for each gene examined by all three technologies. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517929.xml
546233	T cell responses against tumor associated antigens and prognosis in colorectal cancer patients	Introduction Spontaneous T cell responses against specific tumor-associated antigens (TAA) are frequently detected in peripheral blood of tumor patients of various histiotypes. However, little is known about whether these circulating, spontaneously occurring, TAA-reactive T cells influence the clinical course of disease. Methods Fifty-four HLA-A2 positive colorectal cancer patients had been analyzed for the presence of T cell responses against epitopes derived from the TAA Ep-CAM, her-2/ neu , and CEA either by ELISPOT assay or by intracellular cytokine staining. Then, Kaplan-Meier survival analysis was performed comparing T-cell-responders and T-cell-non-responders. For comparison, a group of T-cell-non-responders was compiled stringently matched to T-cell-responders based on clinical criteria and also analyzed for survival. Results Sixteen out of 54 patients had a detectable T cell response against at least one of the three tested TAA. Two out of 21 patients (9.5%) with limited stage of disease (UICC I and II) and 14 out of 33 patients (42.4%) with advanced disease (UICC III and IV) were T cell response positive. Comparing all T-cell-responders (n = 16) and all T-cell-non-responders (n = 38), no survival difference was found. In an attempt to reduce the influence of confounding clinical factors, we then compared 16 responders and 16 non-responders in a matched group survival analysis; and again no survival difference was found (p = 0.7). Conclusion In summary, we found no evidence that spontaneous peripheral T cell responses against HLA-A2-binding epitopes of CEA, her-2/neu and Ep-CAM are a strong prognostic factor for survival.	Introduction The importance of the immune system in containing tumor growth is supported by animal studies and various observations in humans [ 1 , 2 ]. These include increased prevalence of certain tumors following immunosuppression as well as the demonstration, that the presence of intralesional T cells is correlated with improved clinical outcome in various solid tumors [ 1 , 3 - 6 ]. In particular in CRC, the presence of CD8+ T cells within the tumor microenvironment was significantly associated with a better survival in several studies [ 3 , 7 - 9 ]. However, the antigen-specificity of these cells was not determined. T cell responses against specific tumor-associated antigens (TAA) are frequently detected in the peripheral blood of tumor patients [reviewed in [ 10 ]] of various histiotypes including colorectal cancer [ 11 ], melanoma [ 12 , 13 ], acute myeloid leukemia [ 14 ], breast cancer [ 15 ], neuroblastoma [ 16 ], and head and neck cancer [ 17 ]. Data from selected single patients suggest a favorable clinical course in patients with peripheral, spontaneous TAA-directed T cells [ 18 , 19 ]. However, this type of analysis does not allow firm conclusions. The only study comparing clinical outcome of patients with presence of antigen-specific immune responses including natural as well as vaccine-induced antibodies against melanoma antigen GM2 showed an improved survival in favor of immune responders [ 20 ]. TAA-directed T cell responses can reliably be induced using various vaccination approaches [reviewed in [ 21 ]]. Several recent reports have found a correlation between induction of a TAA-directed T cell response by vaccination and clinical response [ 22 - 25 ]. Preliminary data also suggest a possibly favorable clinical effect of vaccine-induced T cells in adjuvant vaccination [ 26 - 29 ]. Taken together, these data lead to the question, whether the presence of spontaneous TAA-specific T cells might be a positive prognostic factor. So far, however, no study has systematically compared survival data of patients with and without presence of a spontaneous TAA-directed T cell response. In previous studies, we have demonstrated spontaneous T cell responses against the TAA CEA, Ep-CAM, or her-2/neu in peripheral blood of approximately 25% of colorectal cancer patients [ 11 , 30 ]. These cells were identified in functional T cell assays by antigen-induced IFNγ production. More detailed analyses in some samples revealed a CD3+ CD8+ IFNγ+ CD69+ CD45RA+ phenotype [ 11 ], indicative of an effector T cell subset that is able to directly mediate tumor cell lysis [ 19 ]. Spontaneous TAA-specific T cells with the potential of effector cells should, theoretically, be capable of destroying tumor cells and thereby lead to elimination of residual disease or prevent tumor progression. To investigate whether a peripheral, spontaneous T cell response has an effect on the clinical outcome of tumor patients, we analyzed survival data of CRC patients with a TAA-directed T cell response and compared these data with the clinical course of CRC patients without detectable T cell response. Patients, materials, and methods Patient selection and T cell assays After institutional review board approval and informed consent, peripheral blood mononuclear cells from 132 patients with CRC in all stages of disease had been prospectively collected and frozen for T cell analysis. All analyses have been performed in compliance with the Helsinki Declaration. Fifty-four patients were tested positive for HLA-A2 and were subsequently analyzed for the presence of T cell responses against the HLA-A0201 presented T cell epitopes Ep-CAM p263–271 [ 31 ], her-2/ neu p654–662 [ 32 , 33 ], and CEA p571–579 [ 34 ] either by ELISPOT assay or by intracellular cytokine staining. HLA analysis, ELISPOT, and intracellular cytokine staining were performed as previously described [ 11 , 30 ]. Positive responses were defined as previously described [ 11 , 30 ]. Survival analysis First, we performed a Kaplan-Meier survival analysis comparing all T-cell-responders and all T-cell-non-responders. Additionally, a two-sided log rank test was used to test statistical significance. Then, in an attempt to reduce the influence of external factors, we compiled a patient group from non-responders matching them to responders according to the following criteria: UICC stage of disease, gender, presence of clinically detectable tumor at time of blood draw, duration of disease until blood draw, age at first diagnosis, and previous therapy. Survival in both groups was compared using Kaplan-Meier analysis and by a two-sided log rank test. A level of p < 0.05 was considered significant. SPSS (11.5) software was used. Results Patients, T cell response, survival of responders and non-responders Fifty-four HLA-A2 positive CRC patients who had been analyzed for T cell responses were included in this study [ 11 , 30 ] and retrospectively analyzed for survival. The overall survival rate was 66.7% at a median of 27.5 months follow-up after blood draw for T cell analysis. In 16 out of 54 patients a total number of 26 T cell responses (between 10 and 1110 specific T cell per 10 6 PBMC) against one of the three TAA Ep-CAM, her-2/ neu , and CEA had been detected. Comparing all responders (n = 16) and all non-responders (n = 38), no survival difference was found in a two-sided log rank test (p = 0.4; Fig. 1A ). A very slight trend toward better survival was found in favor of non-responders, which is, however, far from being statistically significant. Since we have previously observed that T cell responses occur more frequently in patients with advanced stages of disease, we compared clinical data for all patients; and found that among non-responders only 50% (19 out of 38) had clinical stage III or IV disease while 87.5% (14 out of 16) responding patients had stage III or IV disease. Thus, the data on survival are possibly based on a strong stage-related bias. Therefore, we subsequently used an approach matching non-responders to responders. Figure 1 Kaplan-Meier survival analyses of colorectal cancer patients based on their T cell response state. Dashed lines refer to T-cell-responders, full lines refer to T-cell-non-responders, crosses mark censored cases. Time point 0 refers to the time of blood draw for T cell analysis. A . Two groups of CRC patients (total n = 54) were analyzed for survival after T cell analysis. One group had a spontaneous T cell response against the tumor associated antigens CEA, Ep-CAM, or her-2/neu (n = 16, dashed line). The other group had no T cell response against these antigens (n = 38, full line). No survival difference between the groups was found (log rank, p = 0.4). B . Two matched groups of CRC patients (total n = 32) were analyzed for survival after T cell analysis. One group had a spontaneous T cell response against tumor antigens CEA, Ep-CAM, or her-2/neu (n = 16, dashed line). The other group had no T cell response against these antigens and was selected by having similar clinical characteristics for stage, gender, presence of clinically detectable tumor at time of blood draw, duration of disease, age, and prior therapy (n = 16, full line). No survival difference was found (log rank, p = 0.7). Survival in matched patient groups Sixteen patients from the above group of 38 non-responders were matched with 16 responders to obtain two groups comparable for potentially confounding clinical factors, in particular stage of disease (see table 1 ). At the time of the survival analysis 13 of the total of 32 (40.6%) patients had died: seven T-cell-responders (n = 1 stage III, n = 6 stage IV) and six T-cell-non-responders (n = 1 stage III, n = 5 stage IV). All deaths were CRC-related. Median time to death among T-cell-responders was 11 months, among T-cell-non-responders 14.5 months. The calculated mean survival time after blood draw for patients without T cell response was 37.0 months (± 4.8 SEM) with a 95% confidence interval 27.5–46.5. Mean survival of T-cell-responders was 40.2 months (± 6.5 SEM) with 95% confidence interval of 27.5–52.9 (Fig. 1B ). In a two-sided log rank test, survival did not show a statistically significant difference between responders and non-responders (p = 0.7). Of note, one to two years after blood draw, patients without T cell response had an up to 20% higher survival rate (approx. 80% vs. approx 60%). These results were, however, not significant. In a two-sided test with β = 0.2, a survival difference of 70% could have been considered significant at a level of α = 0.05 in a population of this size. Table 1 Patient characteristics Peripheral spontaneous T cell response against TAA Positive Negative (matched) Negative (total) Number of patients 16 16 38 Age at first diagnosis, mean ± SD 61.2 ± 12.1 59.9 ± 9.1 60.8 ± 11.1 Male/Female 11/5 11/5 19/19 Duration of disease until T cell analysis (mean) 31.8 30.2 27.4 Stage (UICC) I 0 0 1 (2.6%) II 2 (12.5%) 2 (12.5%) 18 (47.4%) III 3 (18.8%) 3 (18.8%) 5 (13.2%) IV 11 (68.8%) 11 (68.8%) 14 (36.8%) Previous therapy CT 9 (56.3%) 12 (75%) 18 (47.4%) RT 3 (18.8%) 4 (25%) 7 (18.4%) Surgery 12 (75%) 14 (87.5%) 30 (78.9%) Patients without evidence of tumor at time of blood draw 4 (25%, 2 II, 2 III) 4 (25%, 2 II, 2 III) 17 (45%, 13 II, 3 III, 1 IV) Time from blood draw until death or last contact, mean ± SD 28.1 ± 19.2 27.4 ± 15.0 25.7 ± 14.0 Abbreviations: SD standard deviation, CT chemotherapy (5-FU + folic acid, in few cases plus Irinotecan), RT radiotherapy Discussion In the present study, we analyzed the clinical course of colorectal cancer patients with or without T cells reactive against HLA-A2-binding epitopes of Ep-Cam, her-2/neu, and CEA. No survival difference between T-cell-responders and T-cell-non-responders was found. This result has to be interpreted cautiously due to the small number of HLA-A2+ patients responding to the above antigens. Obviously, these small numbers cause a high beta error. Thus, this study is only a first indication that the tested spontaneous T cell responses are not important prognostic factors for survival. The second limitation of our study is that the known repertoire of TAA as potentially important T cell targets in CRC grows every year; and our T cell analysis included only a fraction of potential epitopes. Various other CRC-associated antigens, such as MUC1 or p53, and additional MHC class I antigenic epitopes of CEA and her2-neu have been described [summarized in [ 35 ]]. T cell responses against antigens in addition to the ones tested here could potentially play a role in immune surveillance of CRC. Immune surveillance is understood as a complex process in which T cells and tumor cells influence each other in several ways ["immunoediting", [ 1 ]]. There are various potential factors related to tumor cells as well as T cells which may explain a lack of survival effect by TAA-specific T cells. The frequency of TAA-specific T cells detected in most patients was quite low in the range of 10 to 100 T cells per 10 6 PBMC. These numbers may be too low to control tumor growth especially in patients with a higher tumor burden. Furthermore, a general T cell dysfunction including anergic T cells and T cells with downregulated CD3-zeta chains has been described in CRC patients [ 36 , 37 ]. It is possible that the specific T cells detected in the present study are functionally unable to destroy tumor cells. This assumption, however, is not supported by our previous finding that TAA-specific T cells have an effector potential as analyzed in selected patients [ 11 ]. Since we have analyzed peripheral blood, we do not know if the circulating T cells have the potential to migrate to the tumor site or compartments where CRC cells frequently migrate to including lymph node, liver and bone marrow. Furthermore, tumor cells may not be recognizable by TAA-specific T cells. It has been shown that CRC cell lines secrete immunosuppressive cytokines and that development of T cell responses is impeded due to low HLA expression and lack of intercellular adhesion molecule-1 (ICAM-1) and HLA-DR [ 38 , 39 ]. This is especially relevant considering the fact that TAA-specific T cell responses in peripheral blood are more frequently detectable in advanced stages of CRC [ 11 , 40 ], as well as other tumors [ 40 , 41 ]. These data led to the hypothesis that metastasizing of tumor cells to lymph nodes is a prerequisite for the development of circulating T cell responses [ 11 , 42 ]. Furthermore, the presence of TAA-directed T cell responses may have selected immune escape tumor variants. A broad variety of tumor escape mechanisms, such as antigen loss or loss of HLA expression, is described in various clinical conditions [ 43 ]. It is possible that we encounter similar mechanisms in the present study since malignant cells had grown in vivo during the presence of specific T cell responses. Finally, the role of suppressor and regulatory T cells in this specific context is unknown. Taken together, no evidence was found that peripheral, spontaneous T cell responses against HLA-A0201-binding epitopes of CEA, Ep-CAM, or her-2/neu influence survival of CRC patients. Since the low patient number limits the conclusion, further studies should investigate more patients, more detailed function and migratory pattern of spontaneous T cell responses as well as the genetic profile of the tumor; and consider a broader antigen and epitope repertoire. These studies could have implications for vaccination therapy as we learn more about why immune surveillance may fail to control tumors and if the presence of a natural T cell response may impact on the efficacy of a vaccine. Abbreviations CEA carcinoembryonic antigen, CRC, colorectal carcinoma; ELISPOT, enzyme-linked immunospot; IFNγ, Interferon-γ; HLA, human leukocyte antigen; PBMC, peripheral blood mononuclear cells, TAA, tumor associated antigen. Competing interests The author(s) declare that they have no competing interests.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546233.xml
546227	Article processing charges, funding, and open access publishing at Journal of Experimental & Clinical Assisted Reproduction	Journal of Experimental & Clinical Assisted Reproduction is an Open Access, online, electronic journal published by BioMed Central with full contents available to the scientific and medical community free of charge to all readers. Authors maintain the copyright to their own work, a policy facilitating dissemination of data to the widest possible audience without requiring permission from the publisher. This Open Access publishing model is subsidized by authors (or their institutions/funding agencies) in the form of a single £330 article processing charge (APC), due at the time of manuscript acceptance for publication. Payment of the APC is not a condition for formal peer review and does not apply to articles rejected after review. Additionally, this fee is waived for authors whose institutions are BioMed Central members or where genuine financial hardship exists. Considering ordinary publication fees related to page charges and reprints, the APC at Journal of Experimental & Clinical Assisted Reproduction is comparable to costs associated with publishing in some traditional print journals, and is less expensive than many. Implementation of the APC within this Open Access framework is envisioned as a modern research-friendly policy that supports networking among investigators, brings new research into reach rapidly, and empowers authors with greater control over their own scholarly publications.	Introduction Journal of Experimental & Clinical Assisted Reproduction is a scientific and clinical journal established in September 2004, offering rapid peer review of research of the advanced reproductive technologies. Content is administered by two chief editors with offices in New York and Atlanta, with peer review supported by an international editorial board. Importantly, all published manuscripts are freely available to a global audience in full text format to facilitate sharing of investigative insights, laboratory methods, and surgical techniques. No other journal presently has these objectives. The journal considers submissions of the following types: Original research, Reviews, Book reviews, Case reports, Commentaries, Debate articles, Hypotheses, Methodology articles and Short reports. All manuscripts published by Journal of Experimental & Clinical Assisted Reproduction are included by PubMed at the National Library of Medicine (United States) [ 1 ], as well as by the national libraries of the Netherlands [ 2 ], Germany [ 3 ], and France [ 4 ]. The evolution of modern academic publishing Creation of the "Open Access" publishing model represented a watershed moment in how research manuscripts are processed, paid for, provided, and protected. Historically, it was the end consumer (the library or reader) who was charged for access to medical, scientific and technical literature. Such payment was traditionally in the form of subscriptions or by access fees levied in exchange for controlled access to particular articles via internet. In contrast, Open Access publishing asks the author to offset the cost of access. The advent of internet technology facilitated the indexing of medical literature with a speed and degree of sophistication impossible before the world wide web. Nevertheless, once an article was located via computer database search, generally only the 250 word abstract could be viewed – even though full text was often required. As no library can offer each issue of every journal, the literature collection/review process increasingly has become a multi-institutional endeavor. Unfortunately the high cost of maintaining multiple subscriptions has actually caused some libraries to provide fewer journals for users [ 5 , 6 ]. More to know, but less to read? With journal holdings at some libraries trending down, smaller academic publishers were early casualties of the reduced demand. Many were lost to consolidation. This resulted in a once diverse publishing community being controlled by only a few organizations with little incentive to change. Meanwhile the number of specialty journals multiplied to keep pace with the growing complexities of medical research. Even as modern research became more multidisciplinary, academic publishers did not make it easy to share findings across dissimilar journals without the reader paying a price. Thus, a paradoxical albeit unintended contraction of access to published medical research resulted. Within the world of academic publishing, limited competition may have registered profits for some but the worldwide impact on readers' access to scholarly literature has been distinctly negative [ 7 ]. It was against this "restricted access" background that the Open Access model of academic publishing was conceived, with a view to maximize rapid collaboration among researchers using a relatively new resource with tremendous potential – the world wide web. Open Access publishing incorporates internet technology which frees the investigator from the limitations of a particular institution's journal holdings as well as the delay of the interlibrary loan. Open Access and the future of academic publishing Implementation of the Open Access publishing model has several important benefits for the scientific community. First, published works are available full text via internet thus making them rapidly accessible to a global audience. Since authors maintain the copyright to their own work, they are free to reproduce it on their own website, link it to related sites, or distribute it according to their own needs without obtaining the publisher's permission first. The only requirement is to acknowledge the article's source. "Reprint requests" are obviated since full text manuscripts are available for free via internet. Such networking has already been shown to boost article citations and impact since these manuscripts are easy to find [ 8 , 9 ]. Second, investigators need only an internet connection and a computer to access every published article with open access – no longer are they limited by their library's journal subscription list. The Open Access publishing model is subsidized by authors (or their institutions/funding agencies) in the form of a single £330 article processing charge (APC), collected at the time of manuscript acceptance for publication. Payment of the APC is never a condition for any manuscript's formal peer review and does not apply to articles rejected after peer review. Additionally, this fee is not assessed for authors whose institutions are BioMed Central members or where genuine financial hardship exists. Such waiver requests are considered by the chief editors on a case-by-case basis. Since the journal exists as an internet resource there is no page limit to the number of color photographs, diagrams, or figures attached to a particular article. Movie files (<10 MB) accompanying a manuscript may also be published using the journal's electronic publishing platform, a feature particularly suited for articles demonstrating new operative or laboratory techniques. Given the full range of publishing possibilities at Journal of Experimental & Clinical Assisted Reproduction , it is hoped that the £330 charge will be affirmed as a good value, especially since subsequent costs associated with reprint fees or page charges assessed by many traditional print journals can easily exceed this amount. Conclusion While several journals have moved to offer free internet access to selected articles, this should not be confused with Open Access [ 10 ]. Free access is typically associated with a 6–12 month delay between publication and actual availability of a manuscript, and even when full-text articles are provided copyright restrictions limit the reader's distribution and reproduction of the work. In contrast, the APC assessed by Journal of Experimental & Clinical Assisted Reproduction guarantees Open Access to all published work and permits the author to keep the copyright to their own data, thus benefiting all who are interested in advancing research in the assisted reproductive technologies. Abbreviations APC = article-processing charge. Competing interests The author(s) declare that they have no competing interests.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546227.xml
535550	Biochemical prevention and treatment of viral infections – A new paradigm in medicine for infectious diseases	For two centuries, vaccination has been the dominating approach to develop prophylaxis against viral infections through immunological prevention. However, vaccines are not always possible to make, are ineffective for many viral infections, and also carry certain risk for a small, yet significant portion of the population. In the recent years, FDA's approval and subsequent market acceptance of Synagis, a monoclonal antibody indicated for prevention and treatment of respiratory syncytial virus (RSV) has heralded a new era for viral infection prevention and treatment. This emerging paradigm, herein designated "Biochemical Prevention and Treatment", currently involves two aspects: (1) preventing viral entry via passive transfer of specific protein-based anti-viral molecules or host cell receptor blockers; (2) inhibiting viral amplification by targeting the viral mRNA with anti-sense DNA, ribozyme, or RNA interference (RNAi). This article summarizes the current status of this field.	Introduction A landmark in the battle against viral infectious diseases was made in 1798 when Jenner first inoculated humans against smallpox with the less virulent cowpox. For about two centuries since then, humans relied almost exclusively on vaccines for protection against viruses. Only in the recent years, new strategies for controlling viral infectious diseases have emerged, which have so far led to a couple of viral prophylaxis/therapeutics on the market. These strategies are fundamentally different from vaccines in that they attempt to directly interrupt viral infectious life cycle at molecular level by using proteins or oligonucleotides. To differentiate them from the conventional vaccines that prevent viral infection by boosting immune system, we refer the new antiviral approaches as "Biochemical Prevention and Treatment" (see figure 1 ). Biochemical Prevention and Treatment, as an alternative to vaccines and chemical compound based antiviral drugs, may prove to be particularly valuable in the areas where vaccines and/or chemical drugs can not be generated or have not been successful in human, including diseases caused by some common pathogenic viruses, such as HIV, hepatitis C virus (HCV), RSV and human rhinovirus (HRV). In this review, we will discuss various molecular intervention approaches. Figure 1 Targets of different Biochemical Prevention and Treatment strategies . Antibodies (Ab) or soluble receptors (Rc) can inhibit the viral entry. Antisense oligonucleotides (AS-ONs), ribozymes (Rz) or siRNA (SI) pair with their complementary target genomic DNA, RNA or mRNA. AS-ONs can block recombination, transcription, translation of the mRNA or induce its degradation by RNaseH. Rz possess catalytic activity and cleave their targets. SiRNAs (SI) induce degradation of the target mRNA via RNA-induced silencing complex (RISC). 1. Biochemical Prevention and Treatment via Protein targeting Among the biochemical therapeutics currently in clinical trials, the majority consists of monoclonal antibodies (MAbs). Soluble receptor drug candidates have gradually lost favor over the past several years due to issues relating to low potency and cost. Peptide-based drug candidates are limited by insufficient efficacy and unfavorable pharmacokinetics. MAbs have increasingly gained favor in large part because of the development of chimeric, humanized, and human antibodies have reduced the immunogenicity of antibody therapies. The MAbs that are currently in clinical trials for viral infection prophylaxis and treatment are listed in Table 1 . Table 1 Monoclonal Antibodies in Clinical Trials Product Company Disease Status MEDI-501 MedImmune Genital Warts HPV II Nabi-HB Nabi Biopharmaceuticals Hepatitis B Market Ostavir Protein Design Labs Hepatitis B II XTL-002 XTL Biopharmaceuticals Ltd. Hepatitis C I Civacir Nabi Biopharmaceuticals Hepatitis C I/II 1F7 Antibody Immune Network Ltd. Hepatitis C, HIV/AIDS Preclinical PRO 140 Progenics Pharmaceuticals HIV/AIDS Preclinical hNM01 AbNovo Inc., Immune Network Ltd. HIV/AIDS I PRO 367 Roche Holding Progenics Pharmaceuticals HIV/AIDS I/II TNX-355 Tanox, Inc., Biogen, Inc. (Massachusetts) HIV/AIDS I OraQuick HIV-1 OraSure Technologies, Inc. HIV/AIDS Market Cytolin CytoDyn Amerimmune Pharmaceuticals, Inc. HIV/AIDS I/II Tipranavir TIPRANAVIR HIV/AIDS III HXB AAI International, AnaaiPharma Company Herpes Simplex Virus type 2 Preclinical MEDI-491 MedImmune Human B19 parvovirus I Synagis™ (Palivizumab) MedImmune Respiratory Syncytial Virus Approved in 1998 Numax MedImmune Respiratory Syncytial Virus Preclinical INS37217 Intranasal Inspire Pharmaceuticals Rhinovirus (common cold) II Biochemical Prevention and Treatment of Respiratory Syncytial Virus Infection The respiratory syncytial virus (RSV) is a major cause of lower respiratory tract infection in infants and young children producing bronchiolitis and pneumonia worldwide. RSV infection leads to more than 90,000 hospitalizations and a 2% mortality rate among infants nationwide [ 2 - 5 ]. Approximately two-thirds of infants are infected with RSV during the first year of life and approximately 95% of children test seropositive for RSV by the age of two [ 6 ]. Unfortunately, even natural RSV infection produces limited immunity at best. In fact, an inactivated RSV vaccine paradoxically resulted in more severe disease instead of protection [ 7 ]. The most successful approach to date has been Biochemical Prevention and Treatment with anti-viral antibodies. In 1996, RespiGam™ (respiratory syncytial virus immune globulin or RSV-IG) became available for use in children less than two years of age with high-risk factors [ 8 - 10 ]. The use of RespiGam™ was largely supplanted with the approval of Synagis™ (Palivizumab) in 1998. Palivizumab is an IgG1 MAb administered IM monthly that selectively binds to the RSV surface glycoprotein F [ 1 , 51 ]. The drug specifically inhibits RSV replication by preventing the virus from fusing with the respiratory endothelial cell membrane. Palivizumab has been shown to reduce the rate of hospitalization of at-risk infants by about 55% in clinical studies and now serves as the primary medical means of RSV prevention [ 11 - 13 ]. Prevention of Human Rhinovirus infections Human rhinovirus (HRV) causes over 80% of the common cold in the fall [ 14 ]. Developing vaccines against HRV is unfeasible because HRVs have at least 115 antigenically distinct serotypes [ 15 , 16 ]. One of the proven methods to prevent and inhibit viral infections is to block host cell receptors that are used by viruses to gain cell entry. Receptor blockage is commonly achieved via application of MAbs that bind to specific epitopes on the receptor molecules. A plethora of in vitro studies have reported effective viral inhibition by receptor-blocking MAbs. However, these works have not yielded yet any approved drug on the market. In HRV infection, about 90% of HRV serotypes utilize a single cell surface receptor exclusively, which is the intercellular adhesion molecule-1 (ICAM-1), for viral attachment and subsequent viral entry [ 17 , 18 ]. As such, ICAM-1 has become a very promising target for biochemical prevention. A receptor blocking approach has shown that the soluble ICAM-1 and an anti-ICAM-1 monoclonal antibody, Mab 1A6, could prevent infections by a broad spectrum of rhinovirus serotypes in human cells in vitro [ 19 - 21 ]. Administration of soluble ICAM-1 and MAbs in human clinical trials had indeed achieved reduction in symptoms, but did not prevent the incidence of the disease [ 22 - 24 ]. For the MAbs, the limited efficacy is most likely due to its low functional affinity (or avidity) for ICAM-1 when compared to that of the multivalent HRV particles [ 25 ]. High avidity is achieved by multivalency. To improve avidity of HRV receptor blocking antibody, a novel tetravalent recombinant antibody, CFY196, has been generated against ICAM-1 [ 26 ]. CFY196 is composed of Fab fragment of a humanized version of MAb 1A6 fused with a linker derived from human immunoglobulin D (IgD) hinge and a tetramerization domain derived from the coiled-coil sequence of human transcription factor ATFα. CFY196 is expressed in bacteria and purified as a homogenous tetrameric molecular complex. CFY196 exhibited almost two-orders-of-magnitude improvement in functional affinity compared with its bivalent counterpart based on the kinetic parameters measured by BIAcore analysis. Such kinetic improvement also directly leads to functional superiorities of CFY196. In in vitro assays, CFY196 consistently and significantly outpaced the best commercial anti-ICAM-1 MAbs in preventing HRV infection as measured by reduction of cytopathic effects and HRV viral titers [ 26 ]. The preclinical findings of CFY196 bode well its efficacy in human since MAb 1A6, from which CFY196 is derived, has already exhibited positive effects in a human trial. Moreover, to prevent possible immunogenicity, CFY196 is humanized [ 27 ]. Further pre-clinical and clinical development of CFY196 is warranted to fully evaluate its potential as a prophylaxis and therapeutics for the HRV induced common colds. 2. Biochemical Prevention and Treatment via targeting on viral mRNA Targeting viral mRNA is one of the most active areas of research and development. Several strategies have emerged over the years and are being tested pre-clinically and clinically. They include: antisense-oligonucleotides (AS-ONs), ribozymes, and recently, RNA interference (RNAi). All these strategies share the features of conceptual simplicity, straightforward drug design and quick route to identify drug leads. However, the challenges have been to improve potency, pharmacokinetics and, most importantly, intracellular delivery of the drug candidates. As the oldest strategy, AS-ON technology has produced to date one drug in the market place, Vitravene ® . A number of clinical trials of drug candidates from these technologies are currently ongoing. Antisense-oligonucleotides Antisense-oligonucleotides (AS-ONs) are short synthetic oligonucleotides that form complementary pair with specific viral mRNA targets. AS-ONs inhibit viral protein production by both blocking viral mRNA translation and triggering its degradation. Since the discovery of viral inhibition effect of AS-ONs by Zamecnik and Stephenson in 1978 [ 28 ], antisense technology has been developed as a powerful tool for target validation and therapeutic purposes. Vitravene is the first AS-ON based drug approved by FDA. Vitravene, or fomivirsen sodium, is a 21-base phosphorothioate oligodeoxynucleotide complementary to the messenger RNA of the major immediate-early region proteins of human cytomegalovirus, and is a potent and selective antiviral agent for cytomegalovirus retinitis, a herpes-like eye disease that afflicts the immune-suppressed [ 29 , 30 ]. A number of clinical trials as well as one approved therapy based on AS-ON technologies are summarized in Table 2 . Table 2 Clinical trials and an approved therapy based on AS-ON technologies [31-33]. Product Company Target Disease Chemistry Status Vitravene (Fomivirsen) ISIS Pharmaceuticals CMV IE2 CMV retinitis PS DNA Approved in 1998 Affinitac (ISIS 3521) ISIS PKC-α Cancer PS DNA Phase III Genasense Genta Bcl2 Cancer PS DNA Phase III Alicaforsen (ISIS 2302) ISIS ICAM-1 Psoriasis, Crohn's disease, Ulcerative colitis PS DNA Phase II/III ISIS 14803 ISIS Antiviral Hepatitis C PS DNA Phase II ISIS 2503 ISIS H-ras Cancer PS DNA Phase II MG98 Methylgene DNA methyl transferase Solid tumors PS DNA Phase II EPI-2010 EpiGenesis Pharmaceuticals Adenosine A1 receptor Asthma PS DNA Phase II GTI 2040 Lorus Therapeutics Ribonucleotide reductase (R2) Cancer PS DNA Phase II ISIS 104838 ISIS TNFα Rheumatoid Arthritis, Psoriasis 2nd generation Phase II Avi4126 AVI BioPharma c-myc Restenosis, cancer, Polycystic kidney disease 3rd generation Phase I/II Gem231 Hybridon PKA RIα Solid tumors 2nd generation Phase I/II Gem92 Hybridon HIV gag AIDS 2nd generation Phase I GTI 2051 Lorus Therapeutics Ribonucleotide reductase (R1) Cancer PS DNA Phase I Avi4557 AVI BioPharma CYP3A4 Metabolic redirection of approved drugs 3rd generation Phase I Phosphorothioate (PS) oligodeoxynucleotides are the ' first generation ' DNA analogs. The ' second generation ' ONs contain nucleotides with alkyl modifications at the 2' position of the ribose. They are less toxic than PS-DNAs and have a slightly enhanced affinity. DNA and RNA analogs with modified phosphate linkages, or different sugar residues substituting the furanose ring have been referred as ' third generation ' [ 34 ]. For instance, peptide nucleic acids and their analogs display superior sequence specificity and are resistant to nuclease degradation. These third generation AS-ON have limited non-specific interactions with other genes and, therefore, have shown great potentials in clinical trials. Ribozymes Ribozymes (Rz) are catalytically active ONs that both bind and cleave target RNAs. They were discovered after the AS-ON technology. Initial findings on ribozymes raised the hope that they may offer a more potent alternative to AS-ONs. Many cell based and animal tests have performed on anti-viral effects of ribozymes, including HIV, hepatitis B, hepatitis C, influenza, etc. Results from these tests have shown that ribozymes are promising viral inhibitors [ 35 - 38 ]. However, further progress in the field has been hampered by difficulties to achieve satisfactory potency and efficient intracellular delivery of ribozymes in vivo. HEPTAZYME is a modified ribozyme that cleaves the internal ribosome entry site of the Hepatitis C virus. The Rz was demonstrated to inhibit viral replication up to 90% in cell culture [ 39 ]. HEPTAZYME was tested in a Phase II clinical trial, but was later withdrawn from further clinical trials due to insufficient efficacy. So far, there is no anti-viral ribozymes that are being actively tested in advanced clinical trials. RNA Interference (RNAi) RNA interference, or RNAi, is the inhibition of expression of specific genes by double-stranded RNAs (dsRNAs). It is becoming the method of choice to knockdown gene expression rapidly and robustly in mammalian cells. Comparing to the traditional antisense method, RNAi technology has the advantage of significantly enhanced potency; therefore, only lower concentrations may be needed to achieve same level of gene knockdown. RNAi gained rapid acceptance by researchers after Tuschl and coworkers discovered that in vitro synthesized small interfering RNAs (siRNAs) of 21 to 23 nucleotides in length can effectively silence targeted genes in mammalian cells without triggering interferon production [ 40 , 41 ]. In mammalian cells, the level of gene inhibition mediated by siRNA routinely reaches an impressive 90% [ 42 ]. Several initial studies, which test the potential application of synthetic siRNAs as antiviral agents, have shown very promising results. To date, RNAi has been used effectively to inhibit the replication of several different pathogenic viruses in culture, including: RSV (respiratory syncytial virus) [ 43 ], influenza virus [ 44 ], poliovirus [ 45 ] and HIV-1 [ 46 - 48 ]. In the case of HIV-1, several specific mRNAs have been successfully targeted for siRNA-mediated silencing, including those that encode Gag, Pol, Vif and the small regulatory proteins Tat and Rev. These studies show that RNAi can effectively trigger the degradation of not only viral mRNAs, but also genomic RNAs at both the pre- and post-integration stages of the viral lifecycle. In addition to targeting viruses directly, alternative strategies have employed siRNAs that silence the expression of essential host factors including Tsg101, required for vacuolar sorting and efficient budding of HIV-1 progeny [ 49 ], and the chemokine receptor CCR5, required as a co-receptor for HIV-1 cell entry [ 50 ]. Conclusions Currently, our understanding of the biological mechanisms underlying RNAi lags behind the movement to apply this technology to human diseases such as viral infections. Some major technical hurdles need to be overcome before siRNA-based anti-viral prophylaxis and treatments move into the clinics. Especially, intracellular delivery of siRNA needs to be greatly improved. The next few years of research will indicate whether RNAi technology will realize its potential as the next wave of Biochemical Prevention and Treatment. Competing Interests Dr. Hervé Le Calvez declares that he has no competing interest. Dr. Mang Yu and Dr. Fang Fang are the co-founders and current share holders of Perlan Therapeutics who has developed CFY196. Figure 2 3D model of the tetrameric Fab anti-ICAM-1 molecule CFY196 [26]. Each identical subunit is represented by a different color.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535550.xml
544356	Delayed bowel perforation following suprapubic catheter insertion	Background Complications of suprapubic catheter insertion are rare but can be significant. We describe an unusual complication of a delayed bowel perforation following suprapubic catheter insertion. Case presentation A gentleman presented with features of peritonitis and feculent discharge along a suprapubic catheter two months after insertion of the catheter. Conclusion Bowel perforation is the most feared complication of suprapubic catheter insertion especially in patients with lower abdominal scar. The risk may be reduced with the use of ultrasound scan guidance.	Background Suprapubic catheterization is a common urological procedure. Complications of catheter insertion are uncommon but can be serious including bowel perforation or obstruction. We describe an unusual complication of delayed bowel perforation after suprapubic catheter insertion. Case presentation An 86 year old gentleman had a suprapubic catheter inserted for bladder outlet obstruction. This was done under a local anaesthetic using the standard Lawrence Add-a-cath ® trocar with ultra sound guidance to measure the bladder volume which was estimated as 500 ml. He had been diagnosed with poorly differentiated carcinoma of the prostate six years earlier and had undergone radical radiotherapy, bilateral sub-capsular orchidectomy and transurethral resection of prostate gland during the interim period. He had bilateral ureteric stents inserted for obstructive uropathy six months earlier. He had a past history of abdomino-perineal resection for rectal carcinoma fourteen years earlier as a curative procedure. He returned for the first change of supra pubic catheter to the Urology suite in two months time. The catheter was changed easily by the specialist nurse and the patient was discharged home. He returned about ten hours later with features of peritonitis and feculent discharge along the supra pubic catheter. He underwent an emergency explorative laparotomy. A loop of small bowel – adhered to the scar – was placed between the anterior abdominal wall and the bladder. The supra pubic tract was seen to pass through and penetrate the loop in two places before going into the urinary bladder [figure 1 ]. There was excessive fibrosis of the bowel segment in the area surrounding the perforation sites. Resection of the affected bowel segment and end-to-end anastomosis was undertaken. An indwelling urethral catheter was left in situ. He made a complete recovery and has been left with the urethral catheter. Conclusions Perforation of the abdominal viscera is well documented as a rare but important major complication of suprapubic cystostomy [ 1 , 2 ]. To our knowledge, only one case of delayed bowel perforation has been reported [ 3 ] three months after the actual catheter insertion. The likely mechanism is the injury occurred during the original insertion. The catheter and the ensuing inflammatory fibrosis sealed the perforation. On removal of the catheter during the change, the sealed perforation opened up. Our case explains the increased risk of bowel damage during suprapubic catheterization in patients with history of previous lower abdominal surgery as the bowel frequently adheres to the scar. In one study, it was found that 59% of patients with midline laparotomy incision have anterior abdominal wall adhesions [ 4 ]. Therefore, patients with lower abdominal scar should only have suprapubic catheter placement under ideal conditions to reduce risk of bowel perforation. Patients must have adequately distended bladder and placed in Trendelenburg position. We do recommend that the procedure to be performed by a skilled operator using ultrasound scan to look for bowel loops between the bladder and anterior abdominal wall. If bowel loops are present or if ultrasound facilities are not available, then open cystostomy method should be considered. The first change of the catheter should be done in the urology department rather than in the community. Patients returning after having their first catheter change with features of localised peritonitis (lower abdominal pain, high temperature and raised White Blood Cell count) should alert the urologist for the possibility of bowel perforation. Authors' competing interests The author(s) declare that they have no competing interests. Author's contributions SA Collected the data; drafted and revised the manuscript and drew the illustration. AM helped to draft and revise the manuscript and PR was the main surgeon and helped to revise the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544356.xml
509242	Age-associated alterations in CXCL1 chemokine expression by murine B cells	Background The CXCL1 chemokines, macrophage inflammatory protein-2 (MIP-2) and cytokine-induced neutrophil chemoattractant (KC), have been shown to play a role in a number of pathophysiological disease states including endotoxin-induced inflammation and bacterial meningitis. While the expression of these chemokines has been identified in a variety of cell types in the mouse, little is known about their expression with murine B-lymphocytes. Results Here, we demonstrate that highly purified murine splenic B cells are capable of expressing both MIP-2 and KC protein and mRNA upon activation with lipopolysaccharide (LPS) but not in response to anti-μ and anti-CD40 in combination with interleukin-4 (IL-4) stimulation. Moreover, these chemokines are expressed at higher levels in B cells derived from young (4 m) compared to old (24–29 m) mice. Upon fractionation into distinct B-cell subsets, we found that the expression of MIP-2 and KC by aged follicular (FO) B cells is significantly decreased when compared to the same cells from younger mice, while only MIP-2 production was found to be diminished in aged marginal zone (MZ) B cells. Interestingly, MIP-2 and KC production by newly formed (NF) B cells did not significantly differ with age. Moreover, the potential relevance of these findings is supported by the poor ability of LPS-activated aged B cells to specifically mediate CXCL1-dependent leukocyte recruitment when compared to younger B cells. Conclusion Overall, the decreased expression of CXCL1 chemokines by aged B cells in response to LPS may have potential implications on the secondary recruitment of leukocytes to sites of microbial infections and inflammation possibly contributing to the increased susceptibility of older subjects to pathogen challenge.	Background Chemokines are a superfamily of small chemotactic proteins that have been classified into four major subfamilies, namely CXC, CC, C, and CX3C, based on the presence or absence and positional arrangement of N-terminal cysteine (C) residues [ 1 - 3 ]. One of the hallmarks of chemokine function is to facilitate trafficking and recirculation of immune cells from the circulation and tissues into secondary lymphatic organs and various peripheral tissues to maintain immune homeostasis in vivo [ 4 ]. These ligands also control the selective recruitment of specific leukocyte subsets to sites of inflammation and immune reactions. Besides migration, chemokines also induce the rapid activation of integrin molecules. The two CXC chemokines, macrophage inflammatory protein-2 (MIP-2) and cytokine-induced neutrophil chemoattractant (KC), are members of the CXCL1 subfamily containing a glutamate-leucine-arginine (ELR) motif that are well known for their ability to induce the activation and recruitment of neutrophils in vitro and in vivo [ 5 - 7 ]. These chemokines are also believed to be the murine structural and functional homologues of human CXCL8, IL-8 and chemokine growth-related oncogene (GRO) [ 8 , 9 ]. MIP-2 and KC display high affinity binding and signaling through the murine CXCR2, a 7-transmembrane G protein-coupled receptor [ 10 ]. It has been well established that endotoxin and various proinflammatory cytokines (e.g., TNF-α and IL-1) stimulate the expression of MIP-2 by macrophages, neutrophils and epithelial cells [ 11 - 13 ]. Numerous studies have also demonstrated a pathophysiological role for MIP-2 and KC in several inflammatory disease states including endotoxemia-induced lung injury [ 14 ], 1996), glomerulonephritis [ 15 ], bacterial meningitis [ 16 ] and herpes simplex virus type 1 (HSV-1) infection [ 17 ]. In mice, the cellular sources of MIP-2 have been confirmed to be macrophages [ 6 ], epithelial cells [ 12 ], bone marrow endothelial cells [ 18 ], astrocytes [ 19 ] and mast cells [ 20 ]. In humans, in addition to macrophages, monocytes, T, NK and B cells have also been shown to produce and respond to CXCL8 [ 21 - 24 ]. Despite all of these reports, few studies have focused on chemokine production by B cells and the relevance of such production in cell-mediated immune responses. Age-related dysfunction of the immune system has often been attributed to a variety of measurable changes in the functional activity of many immunomodulatory factors. This immune deterioration with age is believed to contribute to the morbidity and mortality in humans, possibly due to the greater incidence of infection, autoimmunity and cancer in the elderly. Dysregulation of lymphocyte function is thought to play a critical part in these processes. Many factors are believed to contribute to age-associated immunodeficiencies including defects in cellular signaling, stem cell and bone marrow defects, thymus involution, alterations in hormone and growth factor production, and replicative senescence. Chemokines are believed to play a pivotal role in the complex communication network between different cell types that enable the selective trafficking of many immune effector cells to the necessary sites at the appropriate times. Several reports have also demonstrated increased expression of inflammatory cytokines and chemokines in the circulation and by peripheral blood cells with age suggesting that uncontrolled inflammation may contribute to the increased susceptibility to infection and injury in certain aged cohorts. For example, IL-8 chemokine production by human T, NK and monocytes is altered with age in response to bacteria products [ 25 - 27 ]. Alterations in chemokine production and responsiveness may have a significant impact on the capacity of aged subjects to control or even mount an immune response. The objective of the current study is to explore both young and aged murine splenic B cells and B cell subsets for their ability to express the CXC chemokines, MIP-2 and KC, in response to LPS and other stimuli. Three B cell subpopulations were studied here including: (1) Marginal zone (MZ) B cells, which are uniquely positioned in the marginal sinus where they interact with efficient Ag trapping circulating cells and are thought to be involved in the early response against thymus-independent (TI) blood-borne Ags [ 28 , 29 ]; (2) Follicular (FO) B cells, which are long-lived recruiting B cells located at the border of the B cells follicle and the T cell containing PALS zone that facilitate responses to thymus-dependent (TD) Ags and give rise to both germinal center (GC) B cells and plasma cells [ 30 ]; and (3) Newly formed (NF) B cells, which correspond to the recently immigrated B cells from the bone marrow and have been described as immature/transitional B cells and the precursors of FO and MZ B cells [ 31 ]. We also investigated the effect of these B cell-derived chemokines on splenocyte migration and the influence of aging on this event. The relevance of these findings to the generation of humoral and cellular immune responses shall be discussed. Results and Discussion As expression of MIP-2 and KC by murine splenic B cells has not been previously defined, we initially examined whether murine B cells are capable of producing the CXCL1 chemokines, MIP-2 and KC, upon culture or post cellular activation with LPS, anti-IgM or CD40 mAb in combination with IL-4 for 24 h. Culture supernatants were subsequently examined for MIP-2 and KC expression using ELISA analysis. The results shown in Figure 1A demonstrate that while cultured non-stimulated splenic B cells failed to spontaneously produce detectable levels of MIP-2 and KC (<10 pg/ml), stimulation of B cells with LPS, but not anti-IgM or CD40 mAb in combination with IL-4, resulted in the significant increase in the expressions of both of these chemokines. Interestingly, the quantity of MIP-2 produced in response to LPS was significantly greater than KC. Fig. 1B shows representative images of MIP-2 expression patterns in the cytoplasm of LPS-stimulated splenic B cells. Visually, accumulation of cytoplasmic MIP-2 was observed in splenic B cells (IgM + ) following 24 h incubation with LPS, but was less evident in non-stimulated splenic B cells. Similar results were also obtained examining cytoplasmic KC (data not shown). In agreement with the above findings, MIP-2 and KC mRNA was only detected in LPS- (Fig. 1C ) but not anti-μ- or anti-CD40 mAb/IL-4-stimulated (data not shown) B cells. Consistent with protein expression, the mRNA levels for MIP-2 were expressed at a greater level than KC. As the engagement of the B cell receptor (BCR) initiates signaling pathways mediated through nonreceptor protein tyrosine kinases, including Fyn, Lyn, Syk, and Bruton's tyrosine kinase (BTK), while LPS activate B cells by stimulating signaling through toll-like receptors, more specifically TLR4. The CC chemokines, MIP-1α and MIP-1β, have been shown to be induced and secreted by human B cells in response to BCR signaling [ 22 ]. In contrast, the lack of CXCL1 chemokine induction in response to cross-linking of the BCR with soluble anti-IgM antibody here suggests that the BCR signal(s) may not be optimal and/or essential for MIP-2 and KC production. Perhaps additional costimulatory signals or activation pathways may be necessary for CXCL1 production in combination with BCR signaling. These data demonstrate that murine splenic B cells are capable of secreting the CXCL1 ligands upon activation with LPS but not via anti-μ or anti-CD40 stimulation. Figure 1 Splenic B cells produce MIP-2 and KC in response to LPS. (A) Purified splenic B cells (1 × 10 6 cells/ml) from 4 month-old C57BL/6 mice were stimulated with 10 μg/ml of LPS, anti-IgM or anti-CD40 plus IL-4 for 24 h. Then, cell-free supernatants were collected and assayed by ELISA for MIP-2 and KC secretion. Data are mean of triplicate ± SD of one representative of two experiments. The value was significantly different from non-stimulated control. (, P < 0.01; , p < 0.005) (B) Immunofluorescence visualization of MIP-2 expression in the cytoplasm of 24 h LPS-stimulated splenic B cells. Cells were incubated with biotinylated goat anti-mouse MIP-2 antibody followed by SA-Oregon green-488 (green), PE-conjugated anti-IgM antibody (red) and the DNA dye DAPI (blue not shown) and subjected to cytospin (5 × 10 5 cells/microscope slide). Magnification: +100. (C) Total RNA was isolated from non- and LPS-stimulated cells and levels of MIP-2 and KC mRNA were examined by RT-PCR. The housekeeping gene β-actin was amplified as an internal control. The data are representative of two experiments. Several reports have demonstrated that the proliferative response of whole spleen cells to LPS declines with age [ 32 , 33 ]. The results in Figure 2A demonstrate that purified splenic B cells derived from aged mice also display a significantly diminished proliferative response when cultured with LPS for 72 h when compared to B cells derived from younger mice. However, no significant differences in cellular proliferation were observed between young and aged mice after 24 h of culture, which is in accordance with the previous report [ 34 ]. Figure 2 Effect of aging on LPS-induced B cell proliferation and MIP-2 and KC production by splenic B cells. (A) In vitro proliferation of splenic B cells stimulated with LPS. Purified splenic B cells (1.25 × 10 5 /ml) were cultured with or without LPS. Proliferation was measured by [ 3 H] thymidine uptake after 24 and 72 h of culture. Data are means ± SD of three mice in each group. The value was significantly different from that of aged mice. (* P < 0.01) (B) Splenic B cells from three to five young and aged mice were stimulated with 10 μg/ml of LPS for 4 and 24 h. After stimulation, cell-free supernatants were collected and assayed by ELISA for MIP-2 and KC secretion. One representative experiment out of three is shown. (C) 24 h LPS-stimulated splenic B cells were subjected to immunofluorescence staining with anti-MIP-2 and anti-KC antibodies and DNA dye DAPI as described in Figure 1. One representative experiment out of three is shown. Control anti-goat IgG staining in both young and aged B cells are shown as inserts in this panel. (D) After stimulation, cells were harvested, and RNA was prepared for MIP-2 and KC specific RT-PCR. The housekeeping gene β-actin was amplified as an internal control. Data shown are representative of two independent experiments. (E) MIP-2 and KC mRNA levels were measured by real time RT-PCR and normalized to threshold cycle ( Ct ) values of the co-amplified house-keeping gene GAPDH. Normalized values were calibrated to the value derived from non-stimulated controls and shown as fold change of mRNA expression. Data shown are representative of two independent experiments. Value were significantly different from those in aged mice ( P < 0.01; * P < 0.005) Despite no significant differences in LPS proliferative responses between young an aged mice at 24 h, LPS-stimulated B cells derived from young mice demonstrated approximately 4 times greater levels of MIP-2 in the culture supernatants than aged B cells during this same time period. Similarly, reduced levels of KC expression were also detected in the 24 h culture supernatants LPS-stimulated B cells derived from aged mice when compared to younger animals (Fig. 2B ). Moreover, as shown in Fig. 2C , splenic B cells from aged mice demonstrate significantly less expression of intracytoplasmic MIP-2 and KC upon stimulation with LPS when compared to younger B cells. These results were supported by conventional RT-PCR analysis where MIP-2 and KC mRNA signals were barely detectable in both freshly isolated and non-stimulated splenic B cells from young or aged mice. The mRNA levels of MIP-2 and KC chemokines in LPS-stimulated B cells from aged mice were dramatically lower (> 2-fold) than those in younger mice (Fig. 2D ). Unlike MIP-2, the KC mRNA signal was only detected after 24 h of stimulation with LPS and demonstrated similar patterns of age-associated expression as MIP-2. Furthermore, real time RT-PCR analysis also demonstrated similar age-related alterations in B cell-derived MIP-2 and KC mRNA expression, albeit after 24 h of LPS stimulation (Fig. 2E ). Similar results were obtained using an in vitro B cell culture system in which murine naïve B cells can be activated and induced to proliferate and differentiate into Ab-forming cells. The EL-4 system has been described as a potent in vitro culture system for B activation, proliferation and differentiation [ 35 - 37 ]. Naive B cells cultured in the EL-4 system differentiate to Ig-secreting cells expressing switched isotypes and a plasma cell phenotype (Fig. 5 ; [ 38 , 39 ]). The presence of both LPS-dependent and T cell-dependent stimulation (e.g., CD40-CD40L) contribute to the plasma cell differentiation in this model system. Here, splenic B cells derived from young and aged mice were co-cultured with irradiated EL-4 thymoma cells in presence of LPS and a macrophage-derived culture supernatant. MIP-2 and KC levels were assessed in the supernatants of those cultures via ELISA on days 3 or 5. As shown in Fig. 3A , splenic B cells cultured in this EL-4 culture system yielded similar results to those observed in Fig. 2 . Briefly, a considerable reduction of MIP-2 production by aged splenic B cells was observed in comparison to B cells derived from younger animals. Interestingly, no KC expression was detected in any of the aged B cell cultures while significant KC production was observed in the 3 and 5-day culture supernatant of young B cells. Furthermore, real time RT-PCR also demonstrated a significant reduction in the mRNA expression of MIP-2 and KC in LPS-activated aged B cells when compared to younger B cell cultures (Fig. 3B ). Moreover, given our data in Figure 1 demonstrating that cross-linking of CD40 by anti-CD40 antibody and IL-4 failed to elicit CXCL1 chemokine secretion by B cells, the B cell-derived MIP-2/KC secretion in EL-4 system may be attributed to the effect of LPS-dependent signals rather than CD40 cross-linking signals and/or other cytokines/lymphokines presented in the culture. Taken together, these results demonstrate that MIP-2 and KC expression by murine splenic B cells is significantly altered with increasing age. Figure 3 Effect of aging on MIP-2 and KC production by splenic B cells co-cultured with irradiated EL-4 cells. (A) Decreased amounts of MIP-2 and KC proteins in supernatants of splenic B cells cultured in EL-4 culture system. Splenic B cells (3 × 10 4 /well) from three to five young and aged mice were cultured with irradiated EL-4 thymoma cells in the presence of LPS and macrophage supernatant for 3 or 5 days. Culture supernatants were collected and analyzed by ELISA for MIP-2 and KC secretion. These values were significantly different from those in aged mice (** P < 0.01). (B) Real-time RT-PCR analysis of MIP-2 and KC mRNA expression in activated splenic B cells from young and aged mice. After 3 days of culture, total RNA was isolated from cultured cells and levels of MIP-2 and KC mRNA were measured by real time RT-PCR and normalized to threshold cycle ( Ct ) values of the co-amplified housekeeping gene GAPDH. Normalized values were calibrated to the value derived from EL-4 only controls and expressed as fold induction of mRNA. One representative experiment out of two is shown. Splenic B cells derived from adult mice are comprised of several distinctive subpopulations based on their surface marker expression [ 40 ]. To assess if the CXCL1 ligand expression differences observed between young and aged mice are due to differences in B cell subsets, MZ, FO, and NF B cells as well as total IgM + B cells were sorted by FACS based on their surface IgM, CD21, and CD23 markers (Fig. 4A ). Sorted individual B cell subpopulations were co-cultured with EL-4 cells for 5 days. Similar to our above results, total IgM + splenic B cells derived from aged mice expressed significantly reduced levels of MIP-2 compared to young B cells. Moreover, although MIP-2 secretion by NF B cells derived from aged mice did not significantly differ from young NF B cells, the levels of MIP-2 expressed by aged FO B cells were substantially diminished in comparison to B cells derived from their younger counterparts. Furthermore, unlike MZ B cells derived from young mice, MIP-2 production was almost undetectable in MZ B cells obtained from aged mice (Fig. 4B ). In support of these data, MIP-2 mRNA expression was significantly reduced in aged FO and MZ B cells but not in aged NF B cells when compared to B cells derived from younger mice (Fig. 4C ). Similar results were observed on day 3 (data not shown). Unfortunately, in several repeated subset studies, we were unable to detect KC in the culture supernatants of B cell subsets derived from young and aged mice. We believe this may be due to the low level of KC produced by the splenic B cell subsets compared to non-sorted, non-fractionated splenic B cells, which were also quite low albeit detectable. B cells stimulated with EL-4 T cells behave differently than LPS-stimulated primary B cells. B cells co-cultured with EL-4 cells demonstrate better viability and more plasmocytic differentiation than LPS alone. Thus, B cells are in different states of activation in these two culture systems and may account for the differences in KC expression. Figure 4 Ability of distinct splenic B cell subpopulations from young and aged mice to produce MIP-2 and KC chemokines. (A) Spleen cells were isolated from three young and aged mice, and then stained with anti-IgM, anti-CD23 and anti-CD21 Abs. Subsequently, NF, FO and MZ B cell subpopulations were sorted from IgM + gated cell population. (B) Amounts of MIP-2 protein in supernatants of total IgM + B cells and distinct splenic B cell subpopulations cultured in EL-4 system. NF, FO and MZ B cells within the respective gates shown were directly sorted into individual wells of 96 well plates (1,000 cells/well). Each well had 200 μl of medium containing irradiated EL-4 cells, LPS and macrophage supernatant. After 5 days of culture, supernatants were collected and analyzed by ELISA for MIP-2 and KC secretion. These values were significantly different from those in aged mice (* P < 0.05; P < 0.01; * P < 0.005). (C) Real time RT-PCR analysis of MIP-2 and KC mRNA expression in 5 days-cultured B cells from young and aged mice. MIP-2 and KC mRNA levels in total IgM + , NF, FO and MZ B cells were measured by real time RT-PCR and normalized to threshold cycle ( C t ) values of the co-amplified housekeeping gene GAPDH. Normalized values were calibrated to the value derived from EL-4 only controls and expressed as fold induction of mRNA. One representative experiment out of two is shown. These values were significantly different from those in aged mice (* P < 0.05; P < 0.01; * P < 0.005). Given the possibility that the reduced MIP-2 and KC expression by aged B cells in these EL-4 cultures might reflect a diminished capacity of these cells to proliferate and/or differentiate, we examined the proliferative response as well as the generation of class switched B cells in the culture of those B cells. By culturing MZ, FO and NF B cell subpopulations as well as IgM + B cells on EL-4, no defect in the cell proliferations was found among those distinct B cell subsets in aged mice on day 3 (Fig. 5A ) and day 5 (data not shown) of culture. It should be noted that the EL-4 cells utilized in these assays were thymidine kinase deficient [ 41 ] and thus, the background thymidine incorporation was quite low (<200 cpm). On the other hand, analysis of both percentages (Fig. 5B ) and the absolute number of IgG1 + B cells (Fig. 5C ) revealed that the frequency of IgG1 + B cells presented in the aged B cell culture were comparable to those generated in the young B cell culture in response to LPS. Furthermore, ELISA analysis of Ig levels in culture supernatants demonstrated no difference in the level of IgG produced by young and aged splenic B cells in response to LPS (Fig. 5D ). These studies suggest that the reduced MIP-2 and KC production observed in EL-4 B cell subset cultures are not due to significant alterations in cellular proliferation and/or differentiation. Figure 5 Similar proliferation and differentiation between young and aged splenic B cells cultured in EL-4. FACS-sorted distinct B cell subsets (1000/well) from three young and aged mice were cultured in EL-4 culture system. Proliferation was measured by [3H] thymidine uptake after 3 days culture. Data represent the mean and variations (SD) from triplicate cultures. The data presented are representative of two independent experiments (A). Purified splenic B cells (4 × 10 4 /well) from three young and aged mice were cultured with irradiated EL-4 thymoma cells in the presence of LPS and macrophage supernatant for 3 and 5 days. The cultured cells were stained with anti-IgG1 and anti IgM Abs. Percentages of IgG1 + IgM + B cells are indicated (B). Absolute numbers of IgG1 + B cells in culture. Total numbers of IgG1 + B cells per well were calculated from a mixture of 20 wells in each group (C). Culture supernatants were measured by ELISA for IgG production. Data are representative of two independent experiments (D). To assess the potential functional relevance of MIP-2 and KC expression by LPS-stimulated splenic B cells, splenocytes derived from young mice were labeled with Hoechst and examined for their ability to migrate in response to culture supernatants derived from LPS-stimulated young and aged cells. The results in Fig. 6A show that culture supernatants of LPS-stimulated splenic B cells from either young and aged mice induced significant responder cell migration, as compared with non stimulated B cells-derived supernatants. More importantly, the chemotactic activity of culture supernatants derived from LPS-activated younger B cells was significantly higher than the activity observed in response to supernatants derived from LPS-stimulated aged B cells. Interestingly, the addition of anti-MIP-2 Ab to the B cell cultures resulted in an approximate 5% reduction of the migratory capacity of the young responder cells to migrate in response to the LPS-derived supernatants of young B cells but this antibody addition failed to exhibit any significant inhibition of migration induced by aged B cell supernatants (Fig. 6B ). Neutralization of responder cell migration was also significantly blocked with anti-KC Ab (~20%) in the young but not aged B cell supernatants. Additional neutralization studies using a panel of anti-chemokine antibodies revealed that several CC chemokines are also being made by B cells and are playing a role in the remaining chemotaxis observed using these young responder cell populations (data not shown). Overall, these results suggest that B cell-derived MIP-2 and KC may play a possible role in leukocyte trafficking and that the MIP-2 and KC derived from these B cell cultures are biologically active. Figure 6 B cell-derived MIP-2 and KC induce splenocyte migration. (A). Purified splenic B cells (1.5 × 10 6 /ml) from three young and aged mice were added to the lower chamber of Transwell plates and stimulated with and without LPS. 24 hr after LPS stimulation, spleen cells from 4 month old C57BL/6 mice were preincubated with Hoechst and subjected to chemotaxis through 5-μm pore size Transwell filters (upper chamber) to the supernatants in the lower chambers. Hoechst fluorescence of accumulated cells in the lower chamber was measured. Data are the mean of triplicate cultures ± SD of one representative of two experiments. The value was significantly different from that of control. (* P < 0.05; ** P < 0.01) (B). For the neutralization, Abs against MIP-2 and KC were added to the supernatants in the lower chambers at the beginning of culture. The percent inhibition was calculated as follows: 100 - 100 × (chemotaxis with neutralizing Ab/chemotaxis without neutralizing Ab). Data are the mean of triplicate cultures ± SD of one representative of two experiments. The value was significantly different from that of control (* P < 0.05; ** P < 0.01). In the current study, we report that endotoxin-activated murine B cells express and secrete the CXC chemokines, MIP-2 and KC. These results are in accordance with the previous studies, which demonstrate that activated human B cells express and secrete CXCL8 [ 21 - 23 ]. We also demonstrate that the expression of CXCL1 chemokines, MIP-2 and KC, decline with age in murine splenocytes and B cells, particularly evident in MZ and FO B cell subsets (Fig. 2 , 3 and 4 ). This could not be attributed to impaired numbers of MZ and FO B cells as judged by flow cytometric analysis as young and aged B cells demonstrated comparable numbers of these populations (data not shown). In addition, these differences could also not be attributed to the alterations in proliferation (Fig. 2A and 5A ; [ 34 ]) or terminal differentiation of these cells to IgG plasma cells (Fig. 5B,5C and 5D ). Thus, the downregulation of MIP-2/KC chemokine production in response to LPS by MZ and FO B cells from aged in relation to young mice appears to be due to an age-dependent signaling difference in response to LPS. In this respect, the expression and function of the LPS receptor TLR4 has been shown impaired in aged animals [ 42 ]. In addition, aging can affect gene regulation and some transcription factors, such as nuclear factor-kappa B (NF-kappa B) are required for induction by LPS of MIP-2/KC expression through TLR4 [ 43 ]. Thus, further studies will be necessary to elucidate the relevance of age-related alterations in TLR4 expression and/or function as well as NFκB-dependent transcriptional control in the age-related decline of CXCL1 expression in murine B cells. The accumulation of MIP-2 and KC in tissues is known as an important event in early host defense against bacteria infection. Moreover, evidence indicates the MZ B cells are involved in the early stages of immune response against TI type 2 (TI-2) Ags derived from a number of encapsulated bacteria, including Streptococcus pneumoniae , Neisseria meningitides , and Haemophilus influenzae [ 44 , 45 ]. MZ B cells generate an early IgM producing plasma response after in vivo stimulation with TI antigen [ 29 ]. The remarkable correlation between the ability of MZ B cells to mount an immune response against bacterial-associated antigens and our observation showing that these cells produce neutrophil-attracting chemokines MIP-2/KC in response to LPS indicate a potential dual role for MZ B cells in preventing host from bacterial infection. Of particular significance for the functionality of MZ B cells, we found that MZ B cell-derived MIP-2 and KC expression was impaired in aged mice. Although TI-2-specific Ab immune responses were not found to be significantly altered with age [ 46 ], the diminished capacity of aging immune system to mount an optimal antibody response to encapsulated microbes could be attributed, at least in part, to the diminished capacity of lymphocytes to express inflammatory cytokines and chemokines, such as MIP-2 and KC. Alterations in CXCL1 chemokine production by aged B cells may also have implications in the secondary recruitment of granulocytes to local immune responses at peripheral sites or even within secondary lymphatic organs such as the mucosal immune system in the aged host. Murine MIP-2 and KC exhibit similar expression patterns and functional activities to that of human IL-8 in inflammatory response. For example, it has been previously shown that a significant increase of IL-8 or MIP-2 in human and mice, respectively, occurs in the grain dust-induced inflammation of the lower respiratory tract [ 47 ]. Although age-related alterations in IL-8 production by human B cells have not yet been described, diminished expression of IL-8 has been observed in cultured IL-2-stimulated NK cells [ 25 ] and LPS-stimulated monocytes [ 27 ] derived from elderly subjects. These data suggest that an age-related MIP-2 or IL-8 production could be consequence of a defective functional activity of B cells in aging. It should also be noticed that human IL-8 has been shown to be a potent chemoattractant for human B cells [ 23 , 24 ]. In the present study, we demonstrate that chemotactic activity of culture supernatant from young B cells was dramatically higher than that from aged B cells (Fig. 6A ). Neutralization of these chemokines effects with addition of anti-MIP-2 and anti-KC Abs to the cultures resulted in a significant reduction of the migratory capacity of spleen cells to the culture supernatant from young, but not from aged B cells (Fig. 6B ). This may be due to the low levels of MIP-2 and KC present in the culture supernatants of aged B cells that were not sufficient to induce significant migration of spleen cells to them. In addition, antibodies specific to KC and MIP-2 partially blocked the young splenic B cell-derived chemotactic activity toward splenocytes for approximately 20% and 5%, respectively, although high level of MIP-2 was present in the culture supernatants. The low ability of neutralizing anti-MIP-2 Ab to alter splenocyte migration suggests that MIP-2 produced by B cells may be a weaker migratory factor for murine splenocytes in relation to KC. As MZ B cells in spleen produce both MIP-2 and KC, one could hypothesize that these chemokines may facilitate B cell migration into and within the MZ area and/or amplify their activity and thus contributing to host defense. Conclusion In summary, we demonstrate for the first time that murine splenic B cells are highly efficient in producing ELR-positive CXC inflammatory chemokines, in particular MIP-2, upon activation by LPS stimulation. Moreover, MIP-2 production, particularly by MZ B cells, was found to decline with age. Our finding suggests a possible linkage between functional activity of MZ B cells in production of neutrophil-attracting inflammatory chemokines and host defense. However, detailed and well-controlled in vivo studies will be necessary to assess these various possibilities and the significance of this CXCL1 production defect by aged B cells. Methods Mice Specific pathogen-free 3–5 months (young) and 24–29 months (aged) C57BL/6 mice were purchased through the Office of Biological Resources and Resource Development of the National Institute on Aging (Bethesda, MD). All mice were maintained in an AAALAC-certified barrier facility and were acclimated for 2 weeks prior to use. All mice were fed autoclaved food and water ad libitum . All mice with evidence of disease (e.g., enlarged spleen, gross tumors) were not utilized in these studies. Preparation of splenic B cells Splenic B cells were negatively selected via depletion of non-B lineage cells from spleen cells using a MACS system. Briefly, spleen cells were incubated with magnetic microbeads coated with anti-CD43 antibody and anti-CD11b antibody (Miltenyi Biotec, Bergisch Gladbach, Germany) at 4°C for 15 min, after which the cells were passed over a MACS apparatus. The purity of splenic B cells was consistently >95% as routinely checked by FACS analysis. ELISA analysis Supernatants were collected from cultures and were then frozen at -80°C. The frozen supernatants were thawed at room temperature and chemokine levels were measured with commercial ELISA assay kits for MIP-2 and KC (R and D Systems, Minneapolis, MN) and immunoglobulins (Igs) (Bethyl, Montgomery, TX) according to the manufacturers' instructions. Immunofluorescence staining Purified splenic B cells derived from young and aged C57BL/6 mice were cultured in the presence of absence of LPS (10 μg/ml) for 24 h at 37°C in 5% CO 2 . After culture, the cells were harvested, washed, fixed and permeabilized using 3.7% paraformaldehyde and 0.1% Triton X-100 for 15 min. After thoroughly washing these cells, non-specific binding sites were blocked using a 2% BSA solution containing 1% goat, rabbit serum and normal mouse IgG for 15 min at room temperature. Post incubation, these cells were incubated overnight at 4°C in presence of biotinylated mouse anti-MIP-2 and -KC antibodies (1 μg/ml) (R & D Biosystems, Minneapolis, MN). Streptavidin-conjugated Oregon green-488 (Molecular Probes, Eugene, OR) was utilized to label these cells the following day at a concentration of 1:250 for a period of 45 min at room temperature. After washing, cells were then labeled with PE-conjugated anti-IgM antibody (331,12; PharMingen) for 30 min., followed by the stains of cellular nuclei with DAPI (Molecular Probes, Eugene, OR) at concentration of 1 μg/ml for 10 min. These cells were subsequently placed into cytospin funnels and spun onto glass slides using a cytospin centrifuge (Shandon, Pittsburgh, PA) at 1200 rpm for 5 minutes. After being mounted in Immuno Fluor medium (ICN Biomedicals, Aurora, OH), images were acquired by Spot Advanced software on a Zeiss Axiovert S100 microscope under 100X objective (Carl Zeiss, Thornwood, NY). RT-PCR analysis For conventional RT-PCR analysis, total RNA was extracted from cells using the RNeasy Mini kit (Qiagen, Valencia, CA) and cDNA was prepared from 1 μg of total RNA transcribed by the SuperScript First-strand Synthesis system for the RT-PCR procedure (Invitrogen, Carlsbad, CA) according to the manufacture's instructions. The mouse MIP-2 and KC primers (Sigma Genosys, Woodlands, TX) utilized in these studies were: MIP-2 sense 5'-TGCCGGCTCCTCAGTGCTG-3' and MIP-2 antisense 5'-AAACTTTTTGACCGCCCTTGA-3'; KC sense 5'-CGCTCGCTTCTCTGTGCA-3'and KC antisense 5'-ATTTTCTGAACCAAGGGAGCT-3' as described previously [ 48 ]. The cycling conditions for PCR were 95°C for 4 min for denaturation, followed by 30 cycles at 95°C for 30 s, annealing at 57°C for 45 s plus extension at 72°C for 45 s and a final 10 min at 72°C. After 30 cycles of the PCR, 10 μl of the PCR products were separated on a 1.8% agarose gel, stained with ethidium bromide, and photographs were taken. Their densities were quantified by using the image-analysis system FluorChem (Alpha Innotech Corporation, San Leandro, CA) and normalized using β-Actin housekeeping gene in the same sample. Real-time PCR analysis Approximately 1 μg of total RNA was reverse transcribed by using SuperScript First-strand Synthesis system (Invitrogen, Carlsbad, CA) as described above. Real-time primers for murine MIP-2 and KC were designed using Primer Express software (Applied Biosystems) using the sequences from GenBank (MIP-2, accession no. X53798; KC, accession no. J04596; and GAPDH, accession no. NM_008084). Primers were constructed as follows: MIP-2 (forward primer, AGTGAACTGCGCTGTCAATGC; reverse primer, AGGCAAACTTTTTGACCGCC), KC (forward primer, TGCACCCAAACCGAAGTCAT; reverse primer, TTGTCAGAAGCCAGCGTTCAC), and GAPDH (forward primer, TGCATGGCCGTTCTTAGTTG; reverse primer, AGTTAGCATGCCAGAGTCTCGTT). Reverse-transcribed cDNA was amplified with primer sets for murine MIP-2, KC and GAPDH as indicated above using SYBR Green PCR core reagents and the GeneAmp 5700 Sequence Detection System (PE Applied Biosystems) following the manufacturer's instructions. No PCR products were generated from genomic versus cDNA template. Fold induction of mRNA was determined from the threshold cycle ( Ct ) values normalized for GAPDH expression and then normalized to the value derived from controls. EL-4-based B cell culture The splenic B cell culture was performed using the EL-4-based B cell culture system as previously described [ 38 ]. Briefly, individual wells of 96-well flat-bottom plates were loaded with 5 × 10 4 irradiated (5,000 cGy) murine EL-4 thymoma cells (clone B5) in 200 μl of RPMI 1640 medium supplemented with 10% FCS, 10 -5 M 2-mercaptoethanol (2-ME), 25 mM HEPES buffer, penicillin (100 U/ml), streptomycin (100 μg/ml), 10 μg/ml LPS, and 10% supernatant from culture of J774A.1 macrophage cells (no. TIB-67; American Type Culture Collection, Manassas, VA). MACS purified splenic B cells or FACS sorted B cells were seeded directly onto a feeder layer of irradiated EL-4 cells and cultured at 37°C in 5% CO2. Moreover, we have previously shown that approximately 97% of wells contained sorted single cells using the same outfitted FACStar Plus. These results were verified based on the resulting sequencing histograms demonstrating evidence of only one V6 light chain sequence in the amplified cDNA [ 38 ]. Therefore, any differences in the levels of CXCL1 chemokine secretion among B cell subsets and age should not be a consequence of unequal number of cells sorted in the culture plates. The culture supernatants were collected on day 3 or day 5 and tested for chemokine secretion by ELISA. These time points were selected based on optimal cell viabilities and time to permit cellular activation and proliferation. Flow cytometric analysis and cell sorting The monoclonal Abs utilized for the cell surface staining were FITC-anti-CD21 (clone 7G6; PharMingen, San Diego, CA), -anti-IgM (331,12; PharMingen), PE-anti-CD23 (PharMingen), APC-anti-IgM (clone II/41; PharMingen), and/or biotinylated anti-IgG1 (PharMingen). Biotin-labeled antibody binding was visualized using UltraAvidin-R-Phycoerythrin (Linco Technologies, St. Louis, MO). For each staining, 10 6 cells were pre-incubated with the blocking antibodies, the anti-Fc receptor (24G2), for 30 minutes on ice and then incubated with a mixture of mAbs for an additional 15 min on ice. Post washing, the cells were subsequently incubated with PE-streptavidin for 15 min on ice after which the cells were washed with 5% FCS/HBSS. These stained cells were subsequently analyzed on a Becton Dickinson FACScan flow cytometer using CellQuest software. The B cell subsets, NF, FO and MZ, were isolated through the use of cell sorting using the combination of anti-IgM-APC, anti-CD23-FITC and anti-CD21-PE mAbs. Single spleen cell suspensions were stained with the aforementioned Abs and the MZ, FO and NF B cells within the gate of IgM + cell population were sorted based on their differential expression of CD21 and CD23 using a FACStar Plus (Becton Dickinson, San Jose, CA). The purity of each sorted population was consistently >95%. Proliferation assay MACS-purified splenic B cells (1.25 × 10 5 /well) were cultured with 10 μg/ml of LPS for 1 and 3 days or FACS-sorted distinctive B cell subpopulations (1000/well) were cultured in EL-4 culture system for 3 and 5 days at 37°C in a 5% CO2. Cultures were pulsed with 1 μCi [ 3 H] thymidine (NEN, Boston, MA) for the final 18 h. Cells were harvested on fiberglass paper. [ 3 H] thymidine uptake was measured in a liquid scintillation counter (Beckman, Fullerton, CA). Chemotaxis assay In the lower chambers of Transwell plates (Costar, Cambridge, MA), 1.5 × 10 6 /ml purified splenic B cells derived from young and aged mice suspended in 600 μl of RPMI 1640 medium supplemented with 2% FCS, 10- 5 M 2-mercaptoethanol (2-ME), penicillin (100 U/ml), streptomycin (100 μg/ml) were cultured in the presence and in the absence of LPS (10 μg/ml) and incubated at 37°C in 5% in CO 2 for 24 hr. After incubation, the cells and supernatants in the lower chambers were assessed for chemotactic activity using young splenocyte responder cells. These splenocytes were isolated from 4 month old C57Bl/6 mice after which they were preincubated with 10 μM Hoechst 33342 (Molecular Probes, Eugene, OR) in RPMI 1640 supplemented with 10% FCS for 30 min at 37°C. Subsequently, the Hoechst-stained splenocytes (1 × 10 6 ) were washed with RPMI 1640 supplemented with 2% FCS, resuspended in 100 mL of medium, and then added to the upper chamber, containing a 6.5-mm diameter polycarbonate Transwell culture insert with 5 μm size pore. Each expected group was performed in duplicates with supernatants in the lower chambers for 8 to 10 h at 37°C. In certain assays, neutralizing antibodies specific for MIP-2 and KC at (1 μg/ml) were added to the low chamber containing the LPS-stimulated B cell supernatants. The transmigration of the Hoechst-labeled cells into the lower chamber were measured in a fluorescent spectrophotometer at 350 nm (excitation)/460 nm (emission). The results are expressed as fluorescent units or as the percentage inhibition of Hoechst-labeled splenocyte migration. Statistical analysis Statistical evaluation of significance between the experimental groups was determined by Student's t test. List of abbreviations Ag, antigen; MIP-2, Macrophage inflammatory protein-2; ELR, glutamate-leucine-arginine; GRO, growth-related oncogene; HSV-1, herpes simplex virus type 1; KC, cytokine-induced neutrophil chemoattractant; MZ, marginal zone; FO, follicular; NF, newly formed; TI, Thymus-independent; TD, Thymus-dependent Authors' contributions LH, VMC, and VDD performed the experiments. LH and DDT prepared the figures and wrote the paper. DDT also supervised the work and edited the manuscript. All authors have read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509242.xml
533860	Clinical and inheritance profiles of Kallmann syndrome in Jordan	Background Proper management of patients with Kallmann syndrome (KS) allows them to attain a normal reproductive health. The purpose of this study is to demonstrate the presentation modalities, phenotypes and the modes of inheritance among 32 patients with Kallmann syndrome in Jordan. Recognition of the syndrome allows for prompt proper management and provision of genetic counselling. Subjects Over a period of five years (1999–2004), the clinical and inheritance profiles of 26 male and 6 female patients with Kallmann syndrome from 12 families were evaluated at the National Center for Diabetes, Endocrinology and Genetics in Jordan. Results The patients belonged to twelve Jordanian and Palestinian families and their age at presentation ranged from 4 – 46 years. Nine boys aged 4–14 years presented with cryptorchidism and microphallus, all other males presented with delayed puberty, hypogonadism and/or infertility. The main presentation among six female patients was primary amenorrhea. Intrafamilial variability in clinical phenotype was specifically evident for renal abnormalities and sensorineural hearing impairment. Familial KS was diagnosed in 27 patients belonging to five families with the X-linked mode of inheritance and two families with the autosomal recessive mode of inheritance. Conclusions (1) the majority of cases in this study represented the X-linked form of KS, which might point to a high prevalence of Kal 1 gene in the population. (2) Genetic counselling helps these families to reach a diagnosis at an early age and to decide about their reproductive options. (3) Children presenting with cryptorchidism and microphallus in our population should be investigated for KS.	Background One of the most common causes of hypogonadotropic hypogonadism is Kallmann syndrome (KS). KS is a genetically heterogeneous condition that affects approximately one in 8000 males and one in 40,000–70,000 females [ 1 - 3 ]. This recent estimate is much higher than the previously estimated prevalence of Kallmann syndrome among males of 1:80,000 [ 4 ]. To our knowledge, no data is available on the incidence of this syndrome in Jordan or in the Arab world. In addition to the sporadic form which is the most common [ 2 ], KS has three modes of inheritance, X-linked, autosomal recessive and autosomal dominant [ 5 ]. The gene responsible for the X-linked form of the disease is KAL1 gene [ 6 - 8 ], and encodes the protein anosmin that is directly responsible for the migration of GnRH neurons and the olfactory nerves from the olfactory system to the hypothalamus [ 5 , 9 - 11 ]. Males usually present in the second decade with delayed puberty and females present with primary amenorrhea. Prepubertal boys may present with microphallus and cryptorchidism [ 12 , 13 ]. Proper management of patients with Kallmann syndrome usually allows them to attain normal reproductive health. The purpose of this paper is to demonstrate the presentation modalities, phenotypes and the modes of inheritance among 32 patients with Kallmann syndrome in Jordan, which may contribute to the recognition of the syndrome and the provision of adequate management and genetic counselling. Subjects & Methods Over a period of five years, thirty-two male and female patients from twelve Jordanian and Palestinian families were referred to the National Center for Diabetes, Endocrinology and Genetics (NCDEG) in Amman, Jordan for evaluation of hypogonadism among adults, or microphallus among children. Prospective evaluation was performed including pedigree construction (fig 1 ), and complete clinical examination with special emphasis on assessment of anosmia, the presence of mirror image movements (synkinesia) and examination of external genitalia. Figure 1 Pedigrees of 12 families with KS Olfactory testing was performed through a smell identification test designed by our department using odours that can be easily identified among all social classes in the local population. These odours included cinnamon, coffee, camomile, thyme, soap, oriental perfume and tobacco. The specific odoriferous substance was approximated to the nostrils while the patient had his eyes closed. The patient is then asked to name the olfactory stimulus. According to the number of substances identified by smelling, patients were divided into anosmic and hyposmic. Those who failed to identify all odours were categorized as anosmic, while those who identified three or less odours were categorized as hyposmic. Those who identified four to all seven odours were considered normosmic. First-degree relatives of the probands were also questioned and tested for the smell sensation in a similar fashion. The same investigator carried out olfactory testing for all patients and family members. In addition to the assessment of smell sensation among relatives of probands, their answers to specific questions were recorded. These questions addressed the issues of surgery to correct cryptorchidism, administration of testosterone and whether there were any complaints of delayed puberty or infertility. The specific questions and the olfactory testing led to the identification of further cases of KS in the family. Patients were asked to perform a screwdriver motion in one hand while the examiner watches for any similar non-voluntary movement in the other hand to assess mirror image movements. Subjects were also asked to perform alternate supination and pronation of one forearm while the examiner watches for a similar movement in the other arm [ 14 ]. Phallus length was measured and compared to age-matched controls [ 15 ], and the testicle size was estimated using an orchidometer. Neurosensory hearing impairment was assessed by audiometry, renal abnormalities by renal ultrasound and congenital heart defects by echocardiography. All patients were tested for color vision and abnormal eye movements. Complete hormonal evaluation including basal gonadotrophins, sex hormones and gonadotrophin releasing hormone (GnRH) stimulation test were done for 20 patients 16 years and above. Three adult patients were not tested due to loss of contact. GnRH test involved the administration of 100 mcg of GnRH, followed by the assessment of gonadotrophin levels at 0, 15, 30, 45, 60 and 120 minutes. For individuals with no or low response to the test, priming was performed, where GnRH was given for five days, with repetition of assessment of gonadotrophin levels. Seminal fluid analysis was done for male patients aged 16 years and above. Radiological studies included brain, pituitary and olfactory tract magnetic resonance imaging (MRI). The criteria for diagnosis of KS among adults included the presence of anosmia or hyposmia with clinical signs and symptoms of hypogonadism and a testosterone level <100 ng/dl among males 16 years and older, and estradiol level <20 pg/dl among adult females, together with basal low gonadotrophin level. In family II, loss of contact of three brothers hindered testing their hormonal levels. They were, however, included among our patients because clinical examination revealed hypogonadism, anosmia and micropenis. Among prepubertal males, the criteria for diagnosis of KS included presence of microphallus with anosmia/hyposmia, and/or absent olfactory bulbs on MRI. Pedigree analysis was used to establish the modes of inheritance of KS in the familial cases. Inheritance in a family was classified as X-linked if only males were affected in more than one sibship connected by females, or if two or more males were affected in the sibship with a negative family history and with associated synkinesia. Inheritance was classified as autosomal recessive if all affected individuals were members of the same generation and included at least one female. Results Over a period of five years (1999–2004), thirty-two patients were prospectively diagnosed with Kallmann syndrome at the National Center for Diabetes, Endocrinology and Genetics (NCDEG) in Amman, Jordan. The patients belonged to twelve Jordanian and Palestinian families and their ages at presentation ranged between 4 – 46 years. They included 26 males and 6 females with a male/female ratio of 4.34/1. Nine male patients were aged 14 years and younger. The clinical features among male patients are presented in table (1) , and among female patients in table (2) . Table 1 Clinical features of 26 males with Kallmann syndrome FAMILY No Age Anosmia/hyposmia Synkinesia Hearing Impairment Renal abnormalities Azoospermia cryptorchidism Micropenis 1 1 46 +/H - + - + + - 2 14 + + + + + + - 3 27 + + - + + + - 4 20 + + - + + + - 5 20 + + - + + + - 6 19 + + - + + + - 7 16 + + - + + + - 8 14 + + - + NA + - 9 9 + + - + NA + + 10 6 + + - - NA - + 11 4 + + - - NA + + II 1 37 + - - - + + + 2 24 + - ND* ND* ND* + + 3 22 + - ND* ND* ND* - + 4 20 + - ND* ND* ND* + + III 1 14 + + - - NA - + 2 10 + + - + NA - + 3 8 + + - + NA - + IV 1 6 + + - - NA + + 2 5 + + - - NA + + V 1 20 +/H - + - + - + 2 19 +/H + + + + - + VI 1 19 + - - + ND + - VII 1 37 + - - - + + + VIII 1 22 + - - + + + + IX 1 20 + - - - oligospermia + + NA: not applicable. ND: not done. *: lost contact H: hyposmia Table 2 Clinical Features of 6 females with Kallmann syndrome FAMILY No Age Anosmia/hyposmia Synkinesia Hearing loss Renal abnormalities Primary Amenorrhea X 1 23 + - - - + 2 21 + - - - + 3 18 + - - - + IX 1 18 +/H - - - + XI 1 30 + - - - + XII 1 18 +/H - - - + H: hyposmia Table 3 Hormonal profile and GnRH stimulation test in adult male patients with KS. FAMILY Age Inheritance Base line LH (mIU/ml) Peak LH (mIU/ml) after GnRH testing Peak LH (mIU/ml) after priming Baseline FSH (mIU/ml) Peak FSH (mIU/ml) after GnRH testing Peak FSH (mIU/ml) after priming Testosterone ng/dl 1 1 46 XR 0.5 2.26 1.09 2.22 60 2 14 XR Undetected 0.35 2.50 0.64 2.39 4.25 Undetected 3 27 XR Undetected 0.90 2.85 0.59 1.58 5.24 Undetected 4 20 XR 0.6 2.90 1.66 3.43 80 5 20 XR 0.8 3.10 2.1 5.6 50 6 19 XR Undetected 0.40 2.88 0.8 2.3 6.4 Undetected 7 16 XR 0.6 3..50 1.35 3.21 50 II 1 37 XR Undetected 0.50 3.25 0.43 2.14 6.34 Undetected 2 24 XR ND ND ND ND ND ND ND 3 22 XR ND ND ND ND ND ND ND 4 20 XR ND ND ND ND ND ND ND V 1 20 XR Undetected 0.45 2.1 0.47 1.65 3.2 Undetected 2 19 XR 0.42 4.36 0.90 5.22 40 VI 1 19 sporadic 0.5 6.2 0.9 10.1 70 VII 1 37 sporadic Undetected 1.9 0.46 3.2 Undetected VIII 1 22 sporadic 0.53 3.25 0.85 4.50 Undetected IX 1 20 AR 0.85 4.25 1.21 6.30 50 ND: not done (lost contact) XR :X-linked recessive AR :autosomal recessive Table 4 Hormonal profile and GnRH stimulation test in adult female patients with KS. FAMILY Age Inheritance Base line LH (mIU/ml) N= Peak LH (mIU/ml) after GnRH testing Peak LH (mIU/ml) after priming Base line FSH (mIU/ml) N= Peak FSH (mIU/ml) After GnRH testing Peak FSH (mIU/ml) after priming Estradiol Pg/ml X 1 23 AR 0.6 7.25 2.6 19.33 5 2 21 AR 0.5 3.5 1.95 6.60 7 3 18 AR 0.6 5.2 1.2 9.5 10 IX 1 18 sporadic 0.8 5.10 2.1 5.6 15 XI 1 30 sporadic 0.7 4.2 1.5 6.2 10 XII 1 18 sporadic Undetected 2.10 2.1 5.8 Undetected AR :autosomal recessive Twenty seven patients were anosmic and five patients were hyposmic (table 1 ). In family V, both affected brothers were hyposmic. Cryptorchidism was found or previously operated on in 19/26 (73%) and microphallus in 17/26 (65%) male patients respectively. All patients included in this study manifested high-arched palate. Renal abnormalities including unilateral renal agenesis, malrotated kidney, and horseshoe kidney were detected in 11/19 (58%)cases with the X-linked form of KS. Two sporadic cases showed renal anomalies. A variable degree of sensorineural hearing impairment was found in 4/19 patients with X-linked KS, and in none of the other mode of inheritance or the sporadic cases. Olfactory MRI revealed olfactory tract agenesis among 19/24 cases for which the investigation was done in the series. Among females diagnosed as KS in this series primary amenorrhea was the main presenting feature. Pedigrees were constructed for all families (figure 1 ). Five cases were sporadic and 27 cases were familial belonging to seven families. Pedigree analysis assigned an X-linked mode of inheritance to 3 families with affected males linked through normal females (families I, II and IV). Family I is the largest family in our series with 11 affected males. Two further families were designated as having the X-linked form of KS because the affected males among these siblings displayed synkinesia with a negative family history (families III and V). Synkinesia has been reported to be associated only with the X-linked form of KS [ 16 ]. However, a recent report identifying the specific gene mutated in autosomal dominant KS pointed out that synkinesia may occur in the autosomal forms of KS [ 17 ]. In families III and V in this report, pedigree analysis strongly pointed to the X-linked form of KS although autosomal recessive inheritance cannot be definitely excluded. Two families were designated as having the autosomal recessive mode of inheritance. One family included affected brother and sister with normal consanguineous parents (family IX), while the other family had 3 affected sisters with normal consanguineous parents (family X). Consanguinity rate among parents of all patients was 83%, with 50% of all marriages being between first cousins. Discussion This study points to the higher proportion of the X-linked form of Kallmann syndrome among all KS cases seen at an endocrine/genetic clinic in Jordan over a period of 5 years. Among 7 families with inherited KS, the X-linked form was the mode of inheritance in 5 families (71% of familial KS). None of the pedigrees was consistent with autosomal dominant inheritance in this series. In the two families with autosomal recessive inheritance, the probability of a non-penetrant autosomal dominant gene in either parent was considered remote because of absence of any relevant family history. Autosomal dominant inheritance was also considered a very remote possibility in the X-linked families because of absence of affected individuals in earlier generations and the pattern of inheritance as revealed by pedigree construction (fig 1 ) was compatible with X-linked inheritance. The X-linked form of KS has at times been reported to account for only one third of inherited cases [ 1 ], and at other times to be the most frequent form [ 2 ]. The high proportion of the X-linked form among our cases may represent a high prevalence of Kal1 gene among Jordanians and Palestinians. Consanguineous marriages in Jordan are favored culturally. Among two thousand marriages in the general population, 32% have been reported to be between first cousins [ 18 ]. The figure of 50% first cousin marriages among parents of our patients would thus reflect the high consanguinity rate among the population in general. Among the XR form of KS, 75% of patients were reported to show synkinesia [ 9 ]. Synkinesia in X-linked KS has been attributed to an abnormal projection of the corticospinal tract [ 19 ]. In our experience, synkinesia was present in 16/22 (73%) patients with the XR form of KS and was characteristically more pronounced in the younger age group. Intrafamilial clinical heterogeneity has been reported among family members carrying the same mutation in Kal 1 gene [ 9 ]. In this series, intrafamilial variability in renal anomalies was exemplified in family I, the largest family in our series, where eight out of eleven patients had renal abnormalities. Intrafamilial variability was also seen in family III, in which one patient had a malrotated kidney, his brother had a horseshoe kidney while the third brother had no renal anomalies. Sensorineural hearing loss has also been reported to be associated mainly with the X-linked form of KS [ 20 ]. The KAL1 gene is expressed in the inner ear from early developmental stages suggesting that the defect underlying the hearing loss in X-linked Kallmann syndrome occurs during the organogenesis period [ 9 ]. In our study, sensorineural hearing impairment was only diagnosed among patients with the X linked form of Kallmann syndrome. Four of nineteen tested males (21%) showed sensorineural hearing impairment with evident intrafamilial variability. In this series cryptorchidism or a history of cryptorchidism was present in 73% of patients (19/26), and was not related to a specific mode of inheritance or etiology. Microphallus was present among 17/26 (65%) patients in this study, with several other patients reporting a history of treatment with testosterone; the exact number of treated patients or the treatment profile could not be precisely determined. Seminal fluid analysis was done for all patients of 16 years and above. The test showed azoospermia for all tested males except patient 1 in family IX, where oligospermia was reported with a sperm count of 10,000,000 per ml. None of the patients with the X-linked form manifested ichthyosis, mental retardation, short stature or ocular albinism, pointing to the underlying etiology being a mutation in the Kal 1 gene rather than a contiguous-gene deletion syndrome [ 21 ]. Olfactory MRI revealed olfactory tract agenesis in 80% of cases for which the investigation was done in the series (19/24). Quinton et al, 1996, indicate that KS may be present with no pathology detected in the olfactory tract on MRI, and that phenotypic characterization of KS was effectively achieved by accurate estimation of olfactory sensation [ 22 ]. Kallmann syndrome has a favorable prognosis under proper management. Its investigation should thus be considered in any child presenting with cryptorchidism and microphallus. Since the gonad state is still dormant in childhood, gonadotrophin levels are not helpful. Olfactory MRI may be a more useful tool for the diagnosis [ 23 ]. Nevertheless, a normal MRI does not rule out Kallmann syndrome as normal olfactory bulbs can be present in up to 25% of cases [ 22 ]. The presence of anosmia/hyposmia and history of delayed puberty or infertility in the family are helpful in establishing the diagnosis. Where diagnosis remains difficult, it is indicated to follow up these children till they reach puberty. The majority of Kallmann syndrome cases in our study showed the X-linked mode of inheritance, which might indicate a high prevalence of Kal1 gene in the population. However, molecular studies for the Kal1 gene were not performed in this study. Patients in our series manifested a wide range of phenotypic heterogeneity with intrafamilial variability of clinical manifestations. We recommend an evaluation for Kallmann syndrome in our population in any child presenting with microphallus and cryptorchidism. Further studies are needed to establish the prevalence rate of Kallmann syndrome in Jordan and to define the causative mutations. Competing interests The authors declare that they have no competing interests. Authors' contributions MAJ and KA were the main researchers. HH and MAJ drafted the manuscript. All authors were part of the team that evaluated the patients	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533860.xml
514575	Expression of PEG11 and PEG11AS transcripts in normal and callipyge sheep	Background The callipyge mutation is located within an imprinted gene cluster on ovine chromosome 18. The callipyge trait exhibits polar overdominant inheritance due to the fact that only heterozygotes inheriting a mutant paternal allele (paternal heterozygotes) have a phenotype of muscle hypertrophy, reduced fat and a more compact skeleton. The mutation is a single A to G transition in an intergenic region that results in the increased expression of several genes within the imprinted cluster without changing their parent-of-origin allele-specific expression. Results There was a significant effect of genotype (p < 0.0001) on the transcript abundance of DLK1, PEG11 , and MEG8 in the muscles of lambs with the callipyge allele. DLK1 and PEG11 transcript levels were elevated in the hypertrophied muscles of paternal heterozygous animals relative to animals of the other three genotypes. The PEG11 locus produces a single 6.5 kb transcript and two smaller antisense strand transcripts, referred to as PEG11AS , in skeletal muscle. PEG11AS transcripts were detectable over a 5.5 kb region beginning 1.2 kb upstream of the PEG11 start codon and spanning the entire open reading frame. Analysis of PEG11 expression by quantitative PCR shows a 200-fold induction in the hypertrophied muscles of paternal heterozygous animals and a 13-fold induction in homozygous callipyge animals. PEG11 transcripts were 14-fold more abundant than PEG11AS transcripts in the gluteus medius of paternal heterozygous animals. PEG11AS transcripts were expressed at higher levels than PEG11 transcripts in the gluteus medius of animals of the other three genotypes. Conclusions The effect of the callipyge mutation has been to alter the expression of DLK1 , GTL2 , PEG11 and MEG8 in the hypertrophied skeletal muscles. Transcript abundance of DLK1 and PEG11 was highest in paternal heterozygous animals and exhibited polar overdominant gene expression patterns; therefore, both genes are candidates for causing skeletal muscle hypertrophy. There was unique relationship of PEG11 and PEG11AS transcript abundance in the paternal heterozygous animals that suggests a RNA interference mechanism may have a role in PEG11 gene regulation and polar overdominance in callipyge sheep.	Background The mutation responsible for the callipyge trait is located within an imprinted gene cluster on the distal end of ovine chromosome 18 [ 1 - 3 ]. The callipyge phenotype is associated with an altered carcass composition including a 30–40% increase in muscle mass, a 6–7% decrease in carcass fat, decreased organ weights, and a more compact skeleton, all without a net affect on animal growth [ 4 - 7 ]. There is a pronounced hypertrophy of muscles in the loin and pelvic limbs and a lesser degree of hypertrophy in muscles of the thoracic limbs [ 6 - 8 ]. The callipyge phenotype is inherited in a non-Mendelian mode termed polar overdominance [ 9 , 10 ] in which only animals that inherit a normal allele (wild type; +) from the dam and the mutant callipyge allele from the sire ( CLPG Pat ) exhibit the callipyge phenotype. Maternal heterozygotes ( CLPG Mat /+ Pat ) and callipyge allele homozygotes ( CLPG Mat / CLPG Pat ) have muscling and carcass compositions that are similar to normal sheep (wild type homozygotes; +/+). A physical contig spanning the region containing the callipyge mutation was constructed using overlapping ovine bacterial artificial chromosomes [ 11 ] and 215 kb of sequence was obtained from the contig [ 3 ]. Comparisons of the ovine sequence to the human genome sequence and to expressed sequence databases indicated the presence of at least six transcribed genes with allele-specific expression [ 3 ] (Figure 1 ). The gene order along the contig was found to be Delta-like 1 ( DLK1 ), DLK associated transcript ( DAT ), g ene-trap locus 2 ( GTL2 ), paternal expressed gene 11 ( PEG11 / PEG11AS ) and maternal expressed gene 8 ( MEG8 ). The same conserved gene order was also found as an imprinted domain on human chromosome 14 and mouse chromosome 12 [ 12 - 15 ]. The DLK1 locus (also referred to as PREF-1 , Zog-1 and pG2 ) encodes a transmembrane protein that contains epidermal-growth factor repeats [ 16 - 19 ]. Cleavage of the extracellular domain of DLK1 produces the circulating protein fetal antigen-1 [ 20 ]. DAT is a short non-coding RNA that has been proposed to be a cleavage product of extended DLK1 transcripts [ 21 ]. Both DLK1 and DAT are expressed from the paternal allele [ 3 , 13 - 15 ]. GTL2 (also referred to as MEG3 ) and MEG8 genes express non-coding RNA from the maternal allele [ 3 , 13 - 15 ]. The PEG11 gene contains an intronless open reading frame of 1333 amino acids in sheep [ 3 ]. The human and mouse orthologues known as retrotransposon-like 1 ( RTL1 / rtl1 ) encode 1358 and 1745 amino acids, respectively. RNA transcripts were detected from the opposite strand of the same gene, referred to as PEG11AS (formerly known as antiPEG11 ) [ 3 ]. In the mouse, two maternally expressed microRNAs have been identified with perfect complementarity to mouse Rtl1 [ 22 ]. The causative mutation for callipyge is a single base transition of A (wild type; +) to G ( CLPG ) in the intergenic region located between DLK1 and GTL2 [ 23 ] (Figure 1 ). This mutation has been shown to be 100% concordant with all animals of the + Mat / CLPG Pat genotype based on haplotype analysis [ 23 ]. Analysis of sheep from 19 different breeds as well as 13 mammalian species revealed a highly conserved 12 base sequence that includes the single nucleotide polymorphism [ 24 ]. The G polymorphism is unique to direct descendents of the first known callipyge animal, a ram named "Solid Gold". This animal was mosaic for the mutation [ 24 ], providing strong evidence that this single nucleotide polymorphism is the causative mutation. Initial results indicate that the mutation alters the expression of several of the genes within the imprinted cluster [ 25 , 26 ] when they are inherited in cis without altering their parent-of-origin-specific expression [ 25 ]. In this study, we analyzed the expression of five genes within the callipyge cluster in the muscles of lambs of all four genotypes. Quantitative analysis of gene expression using a series of orthogonal contrasts showed that DLK1 , PEG11 and MEG8 exhibited a polar overdominant pattern of gene expression. The expression of PEG11 and PEG11AS transcripts in the muscles of paternal heterozygous callipyge lambs (+ Mat / CLPG Pat ) was different from the other three genotypes. A sense/antisense interaction of PEG11 and PEG11AS , such as an RNA interference mechanism, would be consistent with a trans interaction between reciprocally imprinted genes that has been previously proposed as a mechanism for polar overdominance [ 27 , 28 ]. Results Northern blot analysis Muscle samples were collected from 12- and 8-week-old lambs, when muscle hypertrophy is well established in animals with the callipyge phenotype. Total RNA was extracted from three muscles that undergo hypertrophy including longissimus dorsi (loin), semimembranosus, and gluteus medius (pelvic limb), and one muscle that does not undergo hypertrophy, the supraspinatus, (thoracic limb). Strand specific probes were used to analyze PEG11 and PEG11AS expression. Hybridization of longissimus dorsi, semimembranosus and gluteus medius northern blots with a PEG11 probe indicated expression of a 6.5 kb PEG11 transcript in paternal heterozygotes (+ Mat / CLPG Pat ) that was not readily detectable in the other three genotypes (Figure 2 ). Two smaller PEG11AS transcripts of 1.7 kb and 0.8 kb were detected in the two genotypes with maternally inherited callipyge alleles ( CLPG Mat /+ Pat and CLPG Mat / CLPG Pat ) by a probe from the complementary strand. PEG11 and PEG11AS transcripts were not detected in the supraspinatus. The expression of DLK1 was detectable at various levels in each of the four muscles and all four genotypes, although DLK1 appeared to be up-regulated in the loin and pelvic limb muscles of + Mat / CLPG Pat and CLPG Mat / CLPG Pat animals (Figure 2 ). GTL2 transcripts of around 2.4 kb were evident in the CLPG Mat /+ Pat and CLPG Mat / CLPG Pat genotypes. A distinct 1.9 kb GTL2 transcript was consistently detected in the + Mat / CLPG Pat genotype and no GTL2 transcripts were detected in the +/+ genotype. A 1.8 kb MEG8 transcript was expressed to a lesser degree than GTL2 in the CLPG Mat /+ Pat and CLPG Mat / CLPG Pat genotypes and was also detectable in the + Mat / CLPG Pat genotype. The expression pattern of MEG8 was the least consistent in the northern blot analysis of individuals of the four genotypes. PEG11/PEG11AS Ribonuclease protection assays were performed using five riboprobes to map the PEG11 and PEG11AS transcripts to the contig sequence (Figure 3A ). The results for riboprobe P confirm the expression of PEG11 transcripts in the gluteus medius of + Mat / CLPG Pat animals and showed a very low level of PEG11 in the gluteus medius of CLPG Mat /CLPG Pat animals. PEG11AS transcripts were readily detectable in animals of the +/+, CLPG Mat /+ Pat and CLPG Mat / CLPG Pat genotypes but were slightly above background levels in the + Mat / CLPG Pat animals (Figure 3B ). Two other PEG11 probes from the open reading frame (E) and from the 3'UTR (C) showed equivalent results to riboprobe P in the gluteus medius (Figure 3B ). The same expression pattern was seen for PEG11 and PEG11AS transcripts in the supraspinatus, but signal from the protected riboprobes was just above background level. Two probes from upstream of the PEG11 coding sequence (riboprobes F and G) did not detect any PEG11 transcripts, indicating that transcription of the 6.5 kb PEG11 was initiated within 350 bp of the start codon of the open reading frame. Riboprobe F detected PEG11AS transcripts in the gluteus medius and supraspinatus of all four genotypes, with the lowest expression in the paternal heterozygous (+ Mat / CLPG Pat ) animals. The PEG11AS riboprobe G shows three protected fragments, suggesting variable splice junctions or transcription termination sites for PEG11AS transcripts. Quantitative analysis of the effect of genotype on gene expression Gene expression was measured in the gluteus medius and supraspinatus of 8 week-old animals (Figure 4 ) using quantitative PCR. The expression of glyceraldehyde-3-phosphate dehydrogenase was not significantly different across the four genotypes in the gluteus medius and supraspinatus (Table 1 ), indicating that equivalent amounts of RNA were used for cDNA synthesis and quantitative PCR. The effect of the callipyge mutation on genotype-specific expression of DLK1 and PEG11 in gluteus medius was the same (Table 1 ) although the magnitude of the response was greater for PEG11 than DLK1 (Figure 4A and 4B ). The paternal heterozygous (+ Mat / CLPG Pat ) animals had the highest transcript abundance (p < 0.05), followed by CLPG Mat / CLPG Pat animals, which had significantly greater transcript abundance (p < 0.05) than CLPG Mat /+ Pat or +/+ animals. The mRNA abundance of DLK1 in the gluteus medius was 6-fold and 2.5-fold greater in + Mat / CLPG Pat lambs and CLPG Mat /CLPG Pat lambs respectively, relative to normal lambs (+/+; Figure 4A ). PEG11 mRNA abundance in the gluteus medius was 200-fold and 13-fold greater in + Mat / CLPG Pat lambs and CLPG Mat / CLPG Pat lambs respectively, relative to normal lambs (+/+; Figure 4B ). The effect of the callipyge mutation on DLK1 and PEG11 expression in the supraspinatus was different (Table 1 ). No differences in DLK1 transcript abundance in the supraspinatus were found among the four genotypes (Figure 4C ). Although not statistically analyzed, the level of DLK1 expression was similar between the gluteus medius and supraspinatus (Figure 4A and 4C ). PEG11 expression in the supraspinatus had a different genotype-specific pattern than in the gluteus medius. PEG11 expression was elevated 150-fold in + Mat / CLPG Pat animals (p < 0.05) but was not significantly changed in the other three genotypes (Figure 4D ). The transcript abundance of PEG11 in supraspinatus was generally much lower than in gluteus medius of the same genotype. PEG11 expression in the supraspinatus was 12-fold lower than in the gluteus medius of paternal heterozygous animals (Figure 4B and 4D ). Maternal inheritance of the callipyge allele significantly altered the expression of MEG8 but not PEG11AS in the gluteus medius (Table 1 ). Transcript abundance of MEG8 in gluteus medius was 6-fold greater (p < 0.05) in the CLPG Mat /+ Pat and CLPG Mat / CLPG Pat animals relative to the + Mat / CLPG Pat and +/+ animals (Figure 4A ). Expression of MEG8 in the supraspinatus was also affected by the callipyge mutation but to a lesser degree (Figure 4C ). The homozygous callipyge animals had significantly higher MEG8 transcript abundance than paternal heterozygotes, but those two genotypes were not significantly different from maternal heterozygotes and homozygous wild type lambs. Orthogonal contrasts were used to analyze different models of gene action for the genes that had a significant effect for genotype (Table 1 ). The polar overdominance contrast was significant for transcript abundances of DLK1 , PEG11 and MEG8 in gluteus medius and for PEG11 and MEG8 in the supraspinatus. The additive contrasts were also significant for DLK1 , PEG11 and MEG8 in gluteus medius but were only significant for PEG11 in supraspinatus. The maternal dominance contrast was significant for DLK1 and PEG11 gluteus medius. Discussion Our results show a clear pattern of increased expression for genes within the imprinted callipyge cluster in muscles that become hypertrophied, whereas expression of these genes was either reduced or absent in a muscle that does not become hypertrophied. The increased gene expression occurred when the mutation was inherited in cis and was dependent on each gene's imprinting status, consistent with previous reports [ 25 , 26 ]. These results support the hypothesis that the mutation has disrupted a long range control element [ 27 ]. Two paternally expressed genes, DLK1 and PEG11 , had significantly increased transcript abundance when a callipyge allele was inherited from the sire. One maternally expressed gene, MEG8 , showed significantly increased transcript abundance when the callipyge allele was inherited from the dam. The changes in gene expression were sustained until 12 weeks of age, when the differences in growth and body composition between callipyge and normal lamb are established and are subsequently maintained [ 7 ]. Northern blot analysis suggests that expression of GTL2 and PEG11AS was increased by maternal inheritance of the callipyge allele. Quantitative analysis was not done for GTL2 in this study due to the expression of numerous alternatively spliced transcripts that have been reported for mice and sheep [ 26 , 29 ]. The 2.4 kb GTL2 mRNA seen in this study consisted of a heterogeneous population of alternatively spliced mRNAs. Similarly, two PEG11AS transcripts were detected over a 5.6 kb area extending from beyond the 5' end of the PEG11 transcript to the 3'UTR. The 1.7 and 0.8 kb PEG11AS transcripts detected by northern blot analysis using probe P would not be protected by probes C or G unless the PEG11AS transcripts undergo intron splicing or there are other transcripts that were not detected in the northern blots. Therefore, the effect of the callipyge mutation on GTL2 and PEG11AS will require a more extensive analysis to fully elucidate their expression patterns in the four genotypes and determine their role in the callipyge model. Due to their paternal allele-specific expression, the DLK1 and PEG11 genes are both candidates for an effector gene that is responsible for the skeletal muscle hypertrophy exhibited by paternal heterozygous (+ Mat / CLPG Pat ) animals. In this study, both genes showed a polar overdominant expression pattern in that paternal heterozygotes had significantly higher levels of gene expression than the other three genotypes. The major differences between DLK1 and PEG11 were the magnitude and the muscle specificity of the up-regulation. DLK1 was readily detectable in all muscles and genotypes by northern blot, but the up-regulation in paternal heterozygous animals was restricted to muscles of the loin and pelvic limb. The quantitative results showed a 6-fold increase in DLK1 transcript abundance in the gluteus medius and no change of DLK1 abundance in the supraspinatus. This pattern of gene expression is consistent with studies on individual muscle growth that show significant muscle hypertrophy in the gluteus medius but not in the supraspinatus [ 6 , 8 ]. DLK1 transcripts were significantly increased (2.5-fold) in the gluteus medius of CLPG Mat / CLPG Pat animals, which do not exhibit muscle hypertrophy. The lack of a phenotype in CLPG Mat / CLPG Pat animals could be due to a threshold effect that requires more than a 2.5-fold increase in DLK1 transcript abundance to change muscle growth. The expression of PEG11 was very low in the gluteus medius and supraspinatus of normal sheep and was induced in both muscles of callipyge lambs. Northern blot analysis, ribonuclease protection assay and quantitative PCR results all show that the expression of PEG11 and PEG11AS transcripts was much lower in the supraspinatus than the other muscles. The high level of expression of PEG11 in the gluteus medius relative to the supraspinatus of callipyge lambs indicates that PEG11 could also be the gene responsible for muscle hypertrophy if there is a threshold level required to change muscle growth. The PEG11 gene has a long intronless open reading frame but it is not known if a protein is produced or what function it may have. The overdominant nature of the callipyge phenotype through the lack of muscle hypertrophy in animals with the CLPG Mat / CLPG Pat genotype is one of the more intriguing aspects of the trait and has led to a hypothesis of trans effects by other reciprocally imprinted genes in the callipyge region [ 25 , 27 ]. If the PEG11 gene has a direct role in muscle hypertrophy, either alone or in concert with DLK1 , then PEG11AS may have a role in a trans effect on PEG11 expression. In + Mat / CLPG Pat animals, PEG11 transcripts were 14-fold and 4-fold more abundant than PEG11AS transcripts in the gluteus medius and supraspinatus respectively. PEG11AS transcripts were more abundant than PEG11 transcripts in the other three genotypes for both muscles. Therefore, the relative abundance of PEG11 transcripts to PEG11AS transcripts was unique in paternal heterozygous animals. MicroRNAs are a central component of RNA interference mechanisms. In the mouse, two antisense microRNA have been identified for the orthologous rtl1 locus [ 22 ]. RNA interference mechanisms have been shown both to repress transcription by inducing heterochromatin formation [ 30 - 32 ] and to cause post-transcriptional silencing through nuclear retention or targeted degradation [ 33 - 35 ]. MicroRNAs may be produced from post-transcriptional processing of PEG11AS RNA and be involved in normal regulation of the locus and in generating overdominance [ 28 ]. Expression of PEG11AS may normally cause repression of the paternal PEG11 locus since very little PEG11 mRNA was detectable in the muscles of normal animals. In the animals with a paternally inherited callipyge allele, (+ Mat / CLPG Pat and CLPG Mat / CLPG Pat ), the normal repression of the PEG11 locus has been disrupted by the mutation in a putative long range control element [ 27 ]. PEG11 transcripts only accumulate in significant excess of PEG11AS in the paternal heterozygous animals, whereas in the CLPG Mat / CLPG Pat animals the PEG11AS transcripts remain in excess and may prevent the accumulation of the threshold level of PEG11 mRNA required to produce a muscle hypertrophy phenotype. Although the expression of PEG11AS was not affected by genotype in quantitative PCR, the northern blots and transcript mapping indicate there are multiple transcripts that are likely to undergo different intron splicing. Further analysis of PEG11AS expression will be necessary to determine its role in the PEG11 locus regulation. Conclusions The effect of the callipyge mutation has been to increase transcript abundance of four genes, DLK1 , GTL2 , PEG11 and MEG8 , within the imprinted cluster in skeletal muscles that become hypertrophied. The increase in transcript abundance was consistent with each gene's parental allele-specific expression. The DLK1 and PEG11 genes were both expressed at their highest levels in paternal heterozygous animals and exhibited polar overdominant gene expression patterns. Therefore, both genes are candidates for causing muscle hypertrophy. DLK1 expression was only elevated in muscles that undergo hypertrophy, so its muscle-specific increase was consistent with the callipyge phenotype. PEG11 was 12-fold more abundant in hypertrophied muscle than non-hypertrophied muscle and only paternal heterozygous animals had PEG11 transcript levels in excess of PEG11AS transcript levels. The unique relationship of PEG11 and PEG11AS in paternal heterozygous animals suggests that an RNA interference mechanism may have a role in regulating the PEG11 locus and polar overdominance in callipyge sheep. Methods Sample collection A series of planned matings were conducted to produce the four possible callipyge genotypes. The genotypes of all lambs were verified using the single nucleotide polymorphism [ 23 , 24 ] and several markers that flank the callipyge region on chromosome 18. Lambs were slaughtered in accordance with humane practices approved by the Utah State University Institutional Animal Care and Use Committee. Samples were collected from the longissimus dorsi, semimembranosus, gluteus medius and supraspinatus, and preserved in RNAlater (Ambion Inc., Woodlands, TX USA). The tissue samples were homogenized in 4 M guanidinium thiocyanate, 25 mM sodium citrate, 50 mM EDTA, 1% sodium-N-lauroyl-sarcosine and total RNA was sedimented by ultracentrifugation of the homogenate on a cushion of 5.7 M CsCl, 50 mM EDTA [ 36 ]. Purified RNA was quantified by spectrophotometry and the use of a constant mass of RNA for each quantitative assay was based on absorbance at 260 nm. Total RNA was treated with DNase I using DNA free ™ reagents (Ambion Inc.) to remove trace genomic DNA prior to use in the ribonuclease protection assay and quantitative PCR. Northern blot analysis Northern blots were prepared using denaturing formaldehyde gel electrophoresis (NorthernMax™; Ambion Inc) of 10 μg of total RNA and transferred to positively charged nylon membranes using standard methods [ 37 ]. Primer sequences used to amplify probes from the callipyge region are given in Charlier et al . [ 25 ], and the PCR products were verified by DNA sequencing. Strand specific DNA probes were synthesized by 40 cycles of asymmetric PCR with Strip-EZ™ nucleotides (Ambion Inc.) and 50 μCi of α-[ 32 P]-dATP (Amersham-Pharmacia, Piscataway, NJ USA). Unincorporated nucleotides were removed by spin column chromatography (BioSpin P30; Bio-Rad Inc., Hercules, CA USA). The probes were hybridized to the membranes without denaturation using Ultrahyb™ (Ambion Inc.) at 42°C overnight. After hybridization, the membranes were washed in 2X SSC (0.3 M sodium chloride, 0.03 M sodium citrate)/0.5% SDS followed by 3 washes in 1X SSC/ 0.1% SDS at 65°C for 30 min and a final high stringency wash in 0.1X SSC/ 0.1%SDS at 65°C for 30 min. The northern blots were exposed to Kodak XAR autoradiography film for 18 to 72 h at -80°C. After autoradiography, the probe was degraded and removed from the membranes using Strip-EZ™ reagents (Ambion Inc.). Ribonuclease protection assay Templates for synthesizing strand specific RNA probes for the ribonuclease protection assay (RPA) were generated from PCR products (Table 2 ) by ligation of double stranded oligonucleotides containing T7 or SP6 promoter sequences and re-amplification with an adapter and gene specific primer (Lig'n Scribe™, Ambion Inc.). Labeled RNA probes were synthesized using MAXIscript™ reagents (Ambion, Inc.) and 50 μCi of α-[ 32 P]-UTP (Amersham-Pharmacia). The RPA were conducted using RPA III™ reagents (Ambion Inc.) and standard urea/acrylamide gel electrophoresis methods [ 37 ]. Dried gels were exposed to phosphorimaging screens and images were collected using a Cyclone Storage Phosphorimager (Packard Instrument Co., Meriden, CT USA). Quantitative PCR Complementary DNA was synthesized from 3.2 μg of total RNA using random hexamer priming and MMLV reverse transcriptase reagents (Invitrogen, Carlsbad, CA USA) with RNase inhibitor supplementation (Superase Inhibitor, Ambion Inc). The cDNA samples were diluted with water and aliquoted so that the quantification was based on 213 ng of total RNA for analysis of glyceraldehyde-3-phosphate dehydrogenase, DLK1 and MEG8 transcripts. PEG11 and PEG11AS transcripts were measured using gene-specific priming of cDNA synthesis from 1.6 μg of total RNA. The cDNA was diluted with water and aliquoted for quantification based on 400 ng of total RNA. Each cDNA sample was amplified in triplicate using the SYBR Green Jump Start™ system (Sigma-Aldridge, St. Louis, MO USA). Quantification standards were composed of aliquots of plasmids containing target PCR products in 10-fold serial dilutions ranging from 10 8 to 10 2 molecules. The standards were used to calculate a regression of threshold cycle on molecule copy number to determine a log value of starting abundance for each of the cDNA samples based on their threshold cycle. The PCR reactions were run for 40 cycles in an iCycler Real-Time PCR Detection System (Bio-Rad Inc.). The log value of starting abundance for each gene was analyzed by analysis of variance using the PROC MIXED procedure of SAS [ 38 ]. The analysis model included genotype as a fixed effect and animal within genotype as a random effect. The number of animals representing each genotype (6 to 8) is given in Table 3 . Orthogonal contrasts were used to evaluate different models of gene action if the effect of genotype on log value of starting abundance was significant. Initially, traditional additive, dominance and reciprocal heterozygote effects were evaluated (see Table 3 for contrasts). If the reciprocal heterozygote effect was significant (p < 0.05), a second set of orthogonal contrasts was used to test additive, maternal dominance and polar overdominance effects as previously described by Freking et al . [ 10 ] (Table 3 ). Authors' contributions CB participated in planning the study and developing the experimental design, conducted the northern blot analysis and ribonuclease protection assays, and wrote the drafts of the manuscript. LK and AP isolated RNA and performed the quantitative PCR assays. TH collected the muscle samples and genotyped the animals used in this study. DM participated in experimental design and performed the statistical analysis. NC participated in planning the study, set up a series of matings to generate all four genotypes and provided the experimental animals. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514575.xml
449855	HIV Infection of Naturally Occurring and Genetically Reprogrammed Human Regulatory T-cells	A T-cell subset, defined as CD4 + CD25 hi (regulatory T-cells [Treg cells]), was recently shown to suppress T-cell activation. We demonstrate that human Treg cells isolated from healthy donors express the HIV-coreceptor CCR5 and are highly susceptible to HIV infection and replication. Because Treg cells are present in very few numbers and are difficult to expand in vitro, we genetically modified conventional human T-cells to generate Treg cells in vitro by ectopic expression of FoxP3, a transcription factor associated with reprogramming T-cells into a Treg subset. Overexpression of FoxP3 in naïve human CD4 + T-cells recapitulated the hyporesponsiveness and suppressive function of naturally occurring Treg cells. However, FoxP3 was less efficient in reprogramming memory T-cell subset into regulatory cells. In addition, FoxP3-transduced T-cells also became more susceptible to HIV infection. Remarkably, a portion of HIV-positive individuals with a low percentage of CD4 + and higher levels of activated T-cells have greatly reduced levels of FoxP3 + CD4 + CD25 hi T-cells, suggesting disruption of the Treg cells during HIV infection. Targeting and disruption of the T-cell regulatory system by HIV may contribute to hyperactivation of conventional T-cells, a characteristic of HIV disease progression. Moreover, the ability to reprogram human T-cells into Treg cells in vitro will greatly aid in decoding their mechanism of suppression, their enhanced susceptibility to HIV infection, and the unique markers expressed by this subset.	Introduction There is now compelling evidence that a subset of T-cells with regulatory activity suppresses T-cell activation in both mice and humans ( Sakaguchi et al. 1995 ; Asano et al. 1996 ; Suri-Payer et al. 1998 ; Takahashi et al. 1998 ; Thornton and Shevach 1998 ; Baecher-Allan et al. 2001 ; Dieckmann et al. 2001 ; Jonuleit et al. 2001 , 2002 ; Levings et al. 2001 ; Ng et al. 2001 ; Taams et al. 2001 ). Regulatory T-cells (Treg cells) have been shown to inhibit various autoimmune and allergic diseases ( Shevach 2000 ; Furtado et al. 2001 ; Curotto de Lafaille and Lafaille 2002 ; Green et al. 2002 , 2003 ; McHugh and Shevach 2002 ), mediate transplantation and self-tolerance ( Sakaguchi et al. 1995 ; Hara et al. 2001 ; Taylor et al. 2001 , 2002; Sanchez-Fueyo et al. 2002 ), and block the activation and proliferation of T-cells both in vitro and in vivo ( Takahashi et al. 1998 ; Thornton and Shevach 1998 ; Annacker et al. 2000 , 2001 ). These findings strongly suggest that Treg cells play a key role in immune regulation. Human and murine Treg cells are functionally characterized by a decrease in both proliferation and IL-2 secretion in response to T-cell receptor (TCR) stimulation and by their ability to suppress activation of conventional T-cells ( Asano et al. 1996 ; Takahashi et al. 1998 ; Thornton and Shevach 1998 ; Baecher-Allan et al. 2001 ; Dieckmann et al. 2001 ; Jonuleit et al. 2001 ; Levings et al. 2001 ; Ng et al. 2001 ; Taams et al. 2001 , 2002 ). Treg cells mediate their suppressive effects only when stimulated via their TCRs ( Takahashi et al. 1998 ; Thornton and Shevach 1998 ), although their suppressive effector function is antigen nonspecific ( Thornton and Shevach 2000 ). Treg cells are clearly enriched within peripheral CD4 + T-cells that also express the α subunit of the IL-2 receptor (CD25), which is currently the best marker for identifying these cells ( Shevach 2002 ). However, CD25 is also expressed on activated effector T-cells, and not all CD4 + Treg cells express CD25 ( Annacker et al. 2001 ; Stephens et al. 2001 ). In adults, Treg cells are exclusively found in the CD45RO + memory subset, and a sizable portion of these cells express the activation marker HLA-DR and the recently identified molecule glucocorticoid-induced tumor necrosis factor receptor (GITR, also known as TNFRSF18) ( Gumperz et al. 2002 ; Lee et al. 2002 ). Upon activation, Treg cells express the inhibitory receptor CTLA-4 at a higher level and for a longer period of time than conventional T-cells ( Read et al. 2000 ; Salomon et al. 2000 ; Takahashi et al. 2000 ). Interestingly, Treg cells have also been shown to express high levels of certain chemokine receptors such as CCR4 and CCR8 ( Iellem et al. 2001 ). The forkhead transcription factor FOXP3 was recently shown to be specifically expressed in mouse Treg cells and is required for their development ( O'Garra and Vieira 2003 ; Ramsdell 2003 ). A mutation in the FOXP3 gene carried by the scurfy mouse strain or a knockout of this gene causes a CD4 + T-cell-mediated lymphoproliferative disease characterized by cachexia and multiorgan lymphocytic infiltrates ( Lyon et al. 1990 ; Brunkow et al. 2001 ). The human genetic disease immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome (also called X-linked autoimmunity-allergic disregulation syndrome) is caused by mutations in the human homolog of FoxP3 and is characterized by hyperactivation of T-cells with autoimmune endocrinopathy, early-onset type 1 diabetes and thyroiditis, and in some cases manifestations of severe atopy ( Chatila et al. 2000 ; Bennett and Ochs 2001 ; Bennett et al. 2001 ; Wildin et al. 2001 ; Gambineri et al. 2003 ). In addition, expression of FOXP3 in conventional T-cells either in transgenic mice or by retroviral transduction is sufficient to confer a Treg cell phenotype ( Fontenot et al. 2003 ; Hori et al. 2003 ; Khattri et al. 2003 ). However, the role of FoxP3 in the development of human Treg cells has not been examined. The role of Treg cells in controlling T-cell activation during immune responses to pathogens such as chronic viral infections is currently a subject of great interest. Recently it was shown that Treg cells can regulate virus-specific or memory CD8 + T-cell responses, thus diminishing the magnitude of the immune response ( Kursar et al. 2002 , 2004 ; Murakami et al. 2002 ; Suvas et al. 2003 ; Aandahl et al. 2004 ). Because Treg cells express CD4 , they are also potential targets of HIV in vivo. HIV entry into target cells also requires cellular expression of the chemokine receptors CCR5 or CXCR4 in conjunction with CD4 ( Boshoff et al. 1997 ). However, the ability of HIV to establish a persistent infection is also critically dependent on activation signals that regulate HIV replication within target T-cells. Quiescent T-cells are resistant to infection unless TCR or cytokine activation signals are provided ( Unutmaz 2001 ). Indeed, chronic states of T-cell hyperactivation, viral persistence, and T-cell depletion are all hallmarks of HIV infection ( Grossman et al. 2002 ). Consequently, this state of chronic immune activation combined with the direct destruction of CD4 + T-cells by HIV leads to a profound immunodeficiency characterized by progressive deterioration of immune function ( Fauci 1993 ). If Treg cells are lost because of HIV infection, this could potentially result in hyperactivation of conventional T-cells due to lack of immunoregulation. In contrast, if Treg cells are activated to expand during certain stages of the infection, this could have a suppressive effect on protective immune responses against the virus. Thus, in both scenarios dysregulation of Treg subset during HIV infection could have a profound impact on anti-HIV immune responses and pathogenesis of the infection. We tested the susceptibility of both naturally occurring and in vitro genetically reprogrammed Treg cells to HIV infection. We found that Treg cells isolated from healthy donors express CCR5 and are highly susceptible to HIV infection. Ectopic expression of FoxP3 in conventional human T-cells genetically reprogrammed them into a Treg phenotype and enhanced their susceptibility to HIV infection. Remarkably, we also found a profound defect of FoxP3 + CD4 + CD25 hi T-cells in HIV-infected patients with low CD4 + and a high percentage of activated T-cells. Our findings have important implications in understanding the role of Treg cells and the chronic activated state of T-cells during HIV infection. Furthermore, reprogramming of T-cells in vitro into Treg cells establishes a novel system to understand the mechanism of T-cell suppression and enhanced susceptibility of this subset to HIV infection. Results Isolation and Characterization of Human Treg Cells To analyze susceptibility of Treg cells to HIV infection, we first developed a method to isolate these cells from peripheral blood. A sizable portion of human CD4 + T-cells (between 10%–20%) express CD25 ( Figure 1 A). However, approximately 1%–2% of CD4 + T-cells within the memory subset (CD45RO + ) express high levels of CD25 (CD25 hi ) ( Figure 1 A). Previous studies suggested that human Treg cells resided within the CD45RO + CD25 hi subset ( Baecher-Allan et al. 2001 ; Taams et al. 2001 ). We first performed a phenotypic analysis of CD45RO + CD25 hi (referred to as Treg), CD45RO + CD25 low/neg (memory T) and CD45RO − CD25 neg (naïve T) cells. Treg cells expressed higher levels of GITR and HLA-DR ( Figure 1 B), consistent with previous reports ( Baecher-Allan et al. 2001 ; McHugh et al. 2002 ; Shimizu et al. 2002 ). Treg cells also expressed high levels of CCR5 and CCR4 compared to memory and naïve T-cells, while expression of CXCR4 and CCR7 was lower and CXCR3 expression was similar as compared to memory T-cells ( Figure 1 B). Figure 1 Identification and Phenotype of Treg Cells (A) Purified CD4 + T-cells were stained with anti-CD45RO-FITC and anti-CD25-PE antibodies. The naïve, memory, and Treg subsets were identified as shown in boxes. (B) Purified CD4 + T-cells were first stained with a pure antibody against the cell surface molecule shown in the figure, followed by antimouse IgG conjugated with allophycocyanin, followed by CD25-PE and CD45RO-FITC. Gates were set for Treg, memory, and naïve T-cells as shown in (A). These results are representative of one out of five donors analyzed. A low proliferative response and reduced IL-2 secretion are characteristics of Treg cells ( Asano et al. 1996 ; Takahashi et al. 1998 ; Thornton and Shevach 1998 ; Baecher-Allan et al. 2001 ; Dieckmann et al. 2001 ; Jonuleit et al. 2001 ; Levings et al. 2001 ; Ng et al. 2001 ; Taams et al. 2001 ). To analyze their capacity to proliferate and secrete IL-2 upon TCR triggering, Treg and memory T-cells were sorted into highly purified populations by flow cytometry. Purified cells were then labeled with carboxy-fluorescein diacetate succinimidyl ester (CFSE) to monitor cell division in a quantitative manner and stimulated through the TCR using plate-bound anti-CD3 and soluble anti-CD28 antibodies. The secretion of IL-2 by TCR-stimulated Treg cells was about 10-fold lower as compared to memory T-cells ( Figure 2 ). Treg cells also secreted lower levels of IL-4, IL-5, and IFNγ as compared to memory T-cells ( Figure 2 ). The CFSE-labeled cells were analyzed 6 d after stimulation. Treg cells exhibited little proliferation, whereas most of the memory T-cells had divided four to five times ( Figure 3 A). In order to demonstrate that purified Treg cells also displayed suppressive activity, both naïve and CD25 low/neg memory CD4 + T-cells were labeled with CFSE and stimulated under suboptimal T-cell activation conditions in the presence of unlabeled purified autologous Treg, naïve, or memory T-cells. Coculture with Treg cells significantly slowed the proliferation of TCR-stimulated resting naïve and memory CD4 + T-cells as compared to the coculture with either unlabeled naïve or memory T-cells ( Figure 3 B and Figure3 C). Taken together, these results confirm that human Treg cells are part of the CD4 + CD25 hi subset of T-cells. Figure 2 Cytokine Secretion by Treg T-cells Sorted Treg and memory T-cells were activated using plate-bound anti-CD3 (3 μg/ml) and soluble anti-CD28 (1 μg/ml) antibodies. Supernatants were collected 18–24 h postactivation and analyzed for cytokines using the CBA assay. Results are representative of cytokine secretion from Treg and memory T-cells from three different donors. Figure 3 Proliferation and Suppressive Capacity of Treg Cells (A) Sorted Treg and memory T-cells were labeled with CFSE and then activated through suboptimal anti-CD3 (100 ng/ml) and anti-CD28 (1 μg/ml) antibodies. Day 6 postactivation, cells were fixed and CFSE expression was analyzed by flow cytometry. (B) Resting naïve or memory CD4 + T-cells (1.5 × 10 5 T-cells) were labeled with CFSE and cocultured with either unlabeled purified Treg, naïve, or memory T-cells at 1:1 ratio in 96-well plates coated with suboptimal anti-CD3 (100 ng/ml) and anti-CD28 (1 μg/ml) antibodies. At day 4 postactivation, cells were fixed and analyzed for CFSE expression and cell size by flow cytometry. (C) Regions were set based on 2-fold reduction in CFSE mean intensity of naïve or memory T-cells as gated on (B), and plotted as number of cell divisions. Results represent three separate experiments from three different donors. Treg Cells Are Highly Susceptible to HIV Infection The ability to obtain a pure population of functional human Treg cells provides an excellent model to study their role in HIV pathogenesis. To determine whether Treg cells were susceptible to HIV infection, purified Treg cells were first activated through the TCR and were infected with either replication-competent HIV, which uses CCR5 as a coreceptor (R5.HIV), or replication-defective viruses pseudotyped with vesicular stomatitis virus glycoprotein (VSV-G.HIV) that encode green fluorescent protein (GFP) as a marker of infection ( Motsinger et al. 2002 ). Treg and memory T-cells challenged with VSV-G.HIV resulted in an equivalent infection rate, while in some experiments R5.HIV resulted in about a 2-fold higher infection rate of the Treg cells as compared to the memory T-cells ( Figure 4 A). To determine the level of HIV replication in the Treg cells as compared to activated memory T-cells, both subsets were infected with R5.HIV for 2 d and washed to remove input virus. Supernatants were collected from the infected cultures daily and p24 levels were quantified by enzyme-linked immunosorbent assay (ELISA). HIV replicated in Treg cells as efficiently as in memory T-cells ( Figure 4 B). To assess whether the virus produced by Treg cells was infectious, supernatants from infected cells were added to Hut78/CCR5 cells, which are highly susceptible to HIV infection, and the titer of infectious virus was determined by GFP expression. Treg cells produced levels of infectious virus similar to those of the memory T subset (data not shown). The viability of infected cultures was also determined at days 3 and 7 to determine if HIV infection killed Treg cells. Indeed, infection with replication-competent HIV was highly cytotoxic to both Treg and memory T-cells ( Figure 4 C). We conclude that Treg cells are highly susceptible to HIV infection and are killed by viral replication. Figure 4 HIV Infection of Treg Cells (A) Sorted Treg and memory T-cells were activated using plate-bound anti-CD3 (3 μg/ml) and soluble anti-CD28 (1 μg/ml) antibodies and concurrently infected with R5.HIV or VSV-G.HIV at a MOI of 5 (based on prior virus titration using Hut78/CCR5 cells). The percentage of infected cells was determined by GFP expression at 3 d postinfection by flow cytometry. (B) Supernatants from Treg and memory T-cells infected with R5.HIV cultures were collected at different time points and HIV p24 levels were measured by ELISA. (C) Treg cell death was assessed by analyzing infected cells at days 3 and 7 postinfection based on forward- and side-scatter analysis and in some experiments using propidium iodine staining analysis with flow cytometry. All of the infection results are representative of one out of five separate experiments with reproducible results. Statistical significance was determined using the Student's two-tailed t test. * p < 0.05. Genetic Reprogramming of Conventional Human T-cells into Treg Cells by Ectopic Expression of FoxP3 Treg cells constitute less than 1%–2% of total human T-cells (see Baecher-Allan et al. 2001 ; see Figure 1 ). Although we could purify several hundred thousand Treg cells from 300 ml of blood as described below, isolation of sufficient numbers of Treg cells is clearly an obstacle to studying their function and susceptibility to HIV infection. Recently, a transcription factor called FOXP3 was shown to program murine T-cells into a Treg subset ( Fontenot et al. 2003 ; Hori et al. 2003 ; Khattri et al. 2003 ). Therefore we hypothesized that ectopic expression of FoxP3 in naïve T-cells could facilitate the generation of large numbers of human Treg cells. Accordingly, we subcloned FoxP3 cDNA into a HIV-derived vector (HDV) that encodes murine CD24 (mCD24) as a marker ( Sundrud et al. 2003 ). CD4 + T-cells were purified from both neonatal cord blood (CB) and adult blood (AB), activated through the TCR and transduced with FoxP3-expressing HDV (HDV.FoxP3) or control HDV as described previously ( Sundrud et al. 2003 ). Expression of FoxP3 mRNA in transduced cells was confirmed by real-time PCR analysis and was found to be about 50- to 100-fold higher in FoxP3 transduced primary CD4 + T-cells as compared to control HDV-transduced cells (data not shown). FoxP3-transduced and control cells were expanded for 14 d in IL-2-containing medium. Cells were then stained for mCD24 and also for CD25, GITR, and CCR4 markers that are expressed at higher levels on naturally occurring Treg cells (see Figure 1 ). Naïve T-cells ectopically expressing FoxP3 displayed higher levels of CD25, GITR ( Figure 5 A), and CCR4 ( Figure 5 B) as compared to control transduced T-cells. FoxP3-transduced memory T-cells displayed much less upregulation of these markers (data not shown). However, while the majority of these transduced T-cells also expressed CCR5, its expression levels on FoxP3-transduced and control T-cells were similar ( Figure 5 C). To determine whether FoxP3-transduced T-cells display functional properties such as hyporesponsiveness to TCR triggering, similar to freshly isolated Treg cells, transduced cells were purified by sorting mCD24 + cells as described previously ( Sundrud et al. 2003 ). Equal numbers of purified cells were then stimulated using plate-bound anti-CD3 and soluble anti-CD28 antibodies, and cytokine secretion was monitored. Secretion of IL-2 from both CB and AB naïve T-cells was reduced between 8- and 10-fold in FoxP3-expressing cells as compared to control cultures ( Figure 6 A). Secretion of IL-4, IL-5, and IFNγ was also reduced in FoxP3-expressing naïve T-cells ( Figure 6 B). In contrast, FoxP3-transduced memory T-cells secreted similar levels of IL-2 as compared to cells transduced with HDV alone ( Figure 6 A). To assess the proliferative capacity of FoxP3-expressing cells, transduced cells were labeled with CFSE and stimulated through the TCR. After 4 d of activation, very few FoxP3-transduced naïve T-cells had divided as compared to control lines ( Figure 7 A). Although, FoxP3-expressing memory T-cells also divided fewer times as compared to HDV-transduced cells, the effect of FoxP3 was greatly diminished in this subset ( Figure 7 A). Figure 5 Phenotype of FoxP3 Transduced T-cells Purified CB naïve CD4 + T-cells were activated through the TCR and transduced with either HDV.FoxP3 or HDV. Cells were expanded for 14 d in IL-2-containing medium and stained with (A) anti-mCD24, anti-GITR, and anti-CD25, (B) anti-mCD24 and anti-CCR4, or (C) anti-mCD24 and anti-CCR5 antibodies. Gates were set on the mCD24-positive population (transduced), and expression of surface molecule was analyzed. The expression of these markers in the CD24-negative portion of both cultures was identical (data not shown). These results are representative of T-cells isolated from five different donors and transduced independently. Figure 6 Cytokine Production by FoxP3-Transduced T-cells CD4 + naïve T-cells isolated from CB (CB-naïve) and AB (AB-naïve) and memory T-cells from AB (AB-memory) were transduced with HDV or HDV.FoxP3 as described in Figure 5 . Transduced T-cells were purified through magnetic sorting of mCD24 + cells and activated using plate-bound anti-CD3 and soluble anti-CD28 antibodies. Supernatants were collected at 18–24 h postactivation and analyzed for (A) IL-2 production or (B) IFNγ, IL-4, and IL-5 production from HDV or HDV.FoxP3-tranduced naïve T-cells, using CBA assay. The results represent five separate experiments from different donors with similar relative differences in cytokine production. Figure 7 Proliferation and Suppression by FoxP3-Expressing Cells (A) Purified CD4 + naïve and memory T-cells were transduced with either HDV.FoxP3 or HDV as described. The transduced T-cells were labeled with CFSE and activated with anti-CD3 (100 ng/ml) and anti-CD28 (1 μg/ml) antibodies. Day 6 postactivation, cells were fixed and analyzed for CFSE expression by flow cytometry. (B) Resting CD4 + T-cells (1.5 × 10 5 ) were labeled with CFSE and cocultured at 1:1 with either unlabeled sorted HDV.FoxP3-expressing or HDV-transduced CB naïve T-cells and activated with anti-CD3 (100 ng/ml) and anti-CD28 (1 μg/ml ) antibodies. At 4 d postactivation, cells were stained with mCD24-PE as a marker for infection. (C) Naïve and memory T-cells isolated from adult blood were transduced with HDV.FoxP3 or HDV. A coculture suppression experiment was set up with resting purified autologous CD4 + T-cells as described above. Region was set on mCD24 negative CFSE + cells (target resting CD4 + T-cells) as shown in (B), and CFSE expression was analyzed 6 d poststimulation by flow cytometry. The results are representative of three separate experiments. The key characteristic of Treg cells is suppression of conventional T-cells activated through the TCR ( Takahashi et al. 1998 ; Thornton and Shevach 1998 ). Thus, purified resting CD4 + T-cells were labeled with CFSE and cocultured with unlabeled naïve or memory T-cells that were transduced either with HDV.FoxP3 or HDV and then stimulated through the TCR as described for freshly isolated Treg cells (see Figure 3 ). FoxP3-expressing naïve T-cells, both from CB or AB, completely suppressed proliferation of target resting CD4 + T-cells ( Figure 7 B and Figure7 C). A significant but lower level of suppression was apparent with memory T-cells transduced with FoxP3 ( Figure 7 C). We conclude that ectopic expression of FoxP3 in naïve human T-cells recapitulates key phenotypic and functional features of naturally occurring Treg cells. T-cells Ectopically Expressing FoxP3 Are More Susceptible to HIV Infection We next determined the susceptibility of FoxP3-expressing cells to HIV infection. FoxP3-transduced cells were purified by flow cytometry sorting based on mCD24 expression and activated through their TCR or cultured in IL-2-containing medium. Subsequently, activated cells were challenged with either VSV-G.HIV or R5.HIV. Remarkably, FoxP3-expressing cells were infected at a level about 2- to 3-fold higher than control cells at different concentrations of the virus, in both activated and nonactivated conditions ( Figure 8 A). FoxP3-expressing cells stimulated at suboptimal levels of anti-CD3 antibody also displayed a very similar enhancement of infection compared to HDV-transduced cells (data not shown). Figure 8 HIV Infection of FoxP3-Expressing T-cells (A) HDV.FoxP3 and HDV-transduced T-cells were activated using plate-bound anti-CD3 (100 ng/ml) and soluble anti-CD28 (1 μg/ml) antibodies or maintained in IL-2-containing medium. Cells were concurrently infected at different MOI of VSV-G.HIV, and infection was determined by GFP expression at 72 h postinfection by flow cytometry. (B) Supernatants were collected at different time points from R5.HIV-infected HDV.FoxP3-expressing or HDV-transduced cell cultures, and HIV p24 levels were measured by ELISA. The percentages of infected cells at days 3, 9, and 16 for HDV.FoxP3 were 2, 10, and 26, and for HDV were 0.8, 10, and 18, respectively. We next analyzed the level of HIV replication and cell death in FoxP3-expressing cells as compared to HDV-transduced T-cells. Activated FoxP3-expressing and control cells were infected with R5.HIV, and culture supernatants were collected daily from day 3 postinfection. FoxP3-expressing cells showed increased HIV-infection and replication ( Figure 8 B). Infectivity of virus produced by FoxP3-expressing cells, as assessed on Hut78/CCR5 cells, as well as cell death in these cultures, was also proportionately higher (data not shown). These findings demonstrate that the expression of FoxP3 renders CD4 + primary T-cells more susceptible to HIV infection. HIV-Infected Patients Have Greatly Decreased Levels of FoxP3-Expressing CD4 + CD25 hi T-cells Our findings that Treg cells are highly susceptible to HIV infection prompted us to determine if this subset was disturbed within HIV-infected individuals. However, a major difficulty in such analysis is that many of the cell surface markers that define Treg cells are also expressed on activated T-cells (CD25, HLA-DR, GITR). Because a portion of HIV-positive individuals contain high levels of activated T-cells, simple cell surface analysis would not be sufficiently reliable to quantify Treg cells in these donors. Therefore, we utilized FoxP3 expression as the most reliable marker that defines Treg cells. To accomplish this, we sorted CD4 + CD25 hi , naïve, and memory T-cells from 11 HIV-negative, healthy donors (median age, 31; 64% male) and 24 HIV-infected individuals (median age, 38; 85% male; 88% receiving antiretroviral therapy). Total RNA was then isolated from each subset and FoxP3 mRNA expression was quantified using real-time PCR analysis. In order to normalize for experimental variability, FoxP3 expression of the CD4 + CD25 hi cells was normalized to GAPDH levels from the same samples and compared to the naïve T-cell subset isolated from the same donor. We found that within HIV-negative subjects there was on average a 49-fold higher level of expression of FoxP3 in CD4 + CD25 hi cells as compared to naïve T-cells ( Figure 9 A). The lowest FoxP3-expressor in the healthy subject group had 16-fold higher FoxP3 expression as compared to naïve T-cells from the same donor ( Figure 9 A). There was a similar increase in FoxP3 expression as compared to memory T-cells (data not shown). In HIV-positive subjects FoxP3 expression was only increased a mean of 25-fold in CD4 + CD25 hi cells. In contrast to healthy donors, we also observed that in about half of the HIV-positive subjects, CD4 + CD25 hi cells expressed very low to undetectable levels of FoxP3 ( Figure 9 A). FoxP3 expression in memory T-cells was similar in HIV-positive and HIV-negative subjects ( Figure 9 A). Figure 9 FoxP3 Expression in Purified CD4 + CD25 hi (Treg), Naïve, and Memory T-cells from HIV-Infected and Healthy Individuals (A) RNA was isolated from sorted Treg, naïve, and memory T-cells from HIV-positive ( n = 24) and HIV-negative ( n = 11) subjects, followed by cDNA synthesis. FoxP3 expression was quantified by TaqMan real-time PCR. The FoxP3 -fold difference expression was calculated for CD4 + CD25 hi (Treg) versus naïve (N), and memory (Mem) versus naïve (N) T-cells. Treg cells sorted from HIV-positive subjects were further subdivided into two groups based on FoxP3 expression of Treg compared to naïve T-cells (FoxP3-high, n = 13, FoxP3 difference >10-fold; FoxP3-low, n = 11, FoxP3 difference <10-fold; HIV-negative, n = 9). These groups were stained with anti-CD3, anti-CD4, anti-CD45RO, anti-CD25, and anti-HLA-DR and analyzed by flow cytometry for (B) CD4 + T-cell percentage, (C) activated T-cell percentage (CD4 + HLA-DR + ), and (D) CD4 + CD25 hi percentage. Horizontal lines identify means. Statistical significance between groups was determined by Mann–Whitney U test and shown on top of each figure. Progressive HIV disease is associated with decreased CD4 + T-cell percentages and increased levels of activated T-cells. We hypothesized that this hyperactivation may be due to a loss of Treg cells. Therefore, to further evaluate relationships between FoxP3 expression and these parameters in HIV-infected individuals, samples were divided into low FoxP3 expressors (less than 10-fold higher expression in CD4 + CD25 hi T-cells compared to naïve T-cells, designated FoxP3-low) versus high FoxP3 expressors (greater than 10-fold higher expression compared to naïve T-cells, designated FoxP3-high). HIV-positive subjects with a FoxP3-low profile had significantly lower CD4 + T-cell percentages, while FoxP3-high HIV-positive subjects had CD4 + T-cell percentages comparable to HIV-seronegative subjects ( Figure 9 B). Similarly, HIV-positive FoxP3-low subjects had significantly greater activated CD4 + T-cells (CD4 + HLA-DR + ) than either HIV-positive subjects with FoxP3-high profiles or HIV-seronegative subjects ( Figure 9 C). Interestingly, the CD4 + CD25 hi T-cells are also significantly increased in FoxP3-low expressors as compared to HIV-negative or HIV-positive FoxP3-high expressors ( Figure 9 D). These differences are most likely due to recently activated T-cells that also express high levels of CD25, as shown by higher HLA-DR expression on T-cells from the same subset of subjects ( Figure 9 C). Among HIV-positive subjects there was no significant association between FoxP3 expression and plasma HIV-1 RNA concentration, age, race, sex, or whether the subject was receiving antiretroviral therapy ( P > 0.05 for each comparison). These findings demonstrate that a decrease in Treg cells is associated with HIV disease progression and suggest that loss of Treg cells may contribute to increased T-cell hyperactivation. Discussion In this study we demonstrated that human Treg cells are highly susceptible to HIV infection and that ectopic expression of FoxP3 genetically reprograms conventional naïve T-cells, phenotypically and functionally, into Treg cells. Remarkably, overexpression of FoxP3 also greatly enhances the susceptibility of activated T-cells to HIV infection. Although Treg cells constitute a small subset of the total T-cells in humans (less than 1%–2%) and thus may not be a significant target population for HIV, they appear to have very potent suppressive activity against activation of T-cells ( Takahashi et al. 1998 ; Thornton and Shevach 1998 ; Baecher-Allan et al. 2001 ; Curotto de Lafaille and Lafaille 2002 ). Here we demonstrate that FoxP3-expressing CD4 + CD25 hi T-cells are greatly decreased in a portion of HIV-infected individuals with low CD4 and high activated T-cells, suggesting a loss of Treg cells. We therefore propose that infection and disruption of Treg cells during HIV infection could have a major influence on T-cell homeostasis and immune regulation. Similar to previous reports, our findings demonstrate that human Treg cells appear to be enriched within the CD25 hi subset of CD4 + T-cells. However, it is not clear if Treg cells are the only population represented in the CD25 hi subset since they share this phenotype with recently activated T-cells. Indeed, a portion of purified CD25 hi cells proliferated and their suppressive function was less efficient as compared to FoxP3-expressing cells (see Figure 3 ). Because the purification of Treg cells is rather arbitrary (brightest 1%–2% of antibody-stained CD25 + memory T-cells), it is conceivable that there is sizable contamination of non-Treg cells in these sorted preparations. In addition, the differences seen in the infection susceptibility of Tregs as compared to FoxP3-expressing cells may also be partly due to presence of non-Treg activated T-cells within the purified cells. Identification of the Treg subset in disease conditions with chronic T-cell activation, such as HIV, is even more problematic because a large proportion of CD4 + CD25 hi cells possibly represent recently activated T-cells. Availability of large numbers of genetically reprogrammed Treg cells should facilitate the identification of novel markers that can reliably detect human Treg cells. Our results clearly demonstrate that, similar to the mouse system, ectopic expression of FoxP3 is sufficient to recapitulate all of the characteristics of Treg cells, including lower cytokine secretion, higher expression of CD25 and GITR, and their suppressive functions. This system has allowed us to generate large numbers of Treg cells, which will be invaluable in characterizing their suppressive function as well as mechanisms of enhanced HIV susceptibility. The ability to genetically manipulate primary T-cells to reprogram them into the Treg phenotype also could have profound implications for preventing graft-versus-host disease, a serious clinical condition that can be manifested following hematopoietic cell transplantation ( Hoffmann et al. 2002 ; Taylor et al. 2002 ). It is interesting to note that naïve T-cells are more prone to reprogramming into a Treg phenotype than memory T-cells are. This loss in flexibility of reprogramming with a master transcription factor is reminiscent of Th1- and Th2-type T-cell reprogramming with ectopic expression of lineage-specific transcription factors T-bet and GATA-3, respectively ( Sundrud et al. 2003 ). The loss of flexibility in genetic modification of effector/memory T-cells could reflect heritable epigenetic changes at effector gene loci that might otherwise be responsive to FoxP3-mediated transcription. A recent study supports this hypothesis by demonstrating that lineage-committed human memory cells failed to modify their histone acetylation patterns of cytokine genes, unlike naïve T-cells, to differentiate into Th1- or Th2-type cells ( Messi et al. 2003 ). The cause of progressive depletion of CD4 + cells and the reason for high T-cell activation or turnover during HIV infection remains controversial ( Hazenberg et al. 2000 ; Grossman et al. 2002 ). It is thought that HIV-mediated destruction of CD4 + T-cells results in decline of this subset and that to maintain homeostasis the immune system attempts to replenish this subset, resulting in a massive turnover of T-cells ( Ho et al. 1995 ; Wei et al. 1995 ; Perelson et al. 1996 ; Mohri et al. 1998 , 2001 ). This excessive turnover rate eventually compromises proper function of homeostatic responses. Alternatively, T-cell depletion could result from disrupted thymic and peripheral homeostatic mechanisms by virus-induced generalized T-cell activation and gradual wasting of T-cell supplies, eventually leading to T-cell depletion ( Hazenberg et al. 2000 ; Grossman et al. 2002 ). Indeed, several mechanisms control unwanted activation of T-cells, including thymic deletion of autoreactive T-cells and induction of anergy in the periphery. In addition to these passive mechanisms, recent evidence clearly demonstrates that Treg cells exert an active suppression of T-cell activation. We postulate that the high susceptibility of Treg cells to HIV in vivo, as demonstrated by our in vitro studies, could result in gradual elimination of this subset. This Treg cell decline during HIV infection would in turn reduce active suppression of conventional T-cells and, hence, contribute to hyperactivation of T-cells. Our analyses of peripheral blood T-cells from HIV-infected subjects support this hypothesis that loss of Treg cells during HIV infection contributes to HIV disease progression. Indeed, we found that in a subset of HIV-infected subjects, the CD4 + CD25 hi T-cell subset had greatly reduced FoxP3 expression, suggesting that these cells represent recently activated T-cells rather then Treg cells. The presence of a higher percent of activated T-cells in this FoxP3-low profile supports this conjecture. The HIV-infected subjects with lower levels of FoxP3 + T-cells also contained a lower percentage of CD4 + T-cells. It is conceivable that the loss of Treg cells may be a correlative factor for disease progression; however, more detailed prospective studies will be required to address this important implication of our findings. Our findings show a 2- to 3-fold enhancement of HIV infection in FoxP3-expressing T-cells. While freshly purified Treg and memory T-cells are similar in their susceptibility to HIV infection, activated effector T-cells become gradually more resistant to infection (unpublished results). Indeed, preactivated T-cells require reactivation to render them susceptible to infection after about 1–2 weeks in culture (unpublished results). We do not yet know the mechanisms by which FoxP3 renders activated T-cells more susceptible to infection; however, two possibilities can be considered: (1) FoxP3 may be overcoming innate resistance factor(s) that accumulate in activated T-cells that block HIV infection, or (2) FoxP3 expression may be inducing critical host factors that are required for efficient completion of the HIV life cycle in primary T-cells. It is important to note that since FoxP3 expression enhances VSV-G.HIV single-round infections, it likely affects an early, postentry step in viral replication. Determining the mechanisms by which FoxP3 enhances HIV infection could reveal host factors involved in this process. In summary, our results indicate that both naturally occurring and genetically reprogrammed Treg cells are susceptible to HIV infection and that ectopic FoxP3 expression greatly increases the susceptibility of T-cells to HIV infection. Our finding that FoxP3-expressing CD4 + CD25 + T-cells are greatly reduced in HIV patients with low CD4 + T-cell percentages and increased T-cell activation suggests that loss of Treg cells may contribute to HIV disease progression. Further prospective studies will be required to unravel the role of this important subset in HIV infection. The ability to genetically reprogram conventional human T-cells to generate Treg cells will also lead the way to identifying unique markers expressed on this population in order to further investigate their status in HIV-infected individuals. Moreover, understanding how FoxP3 enhances HIV infection and programs T-cells into the Treg subset could help in the identification of novel host factors that mediate HIV infection in primary T-cells and decoding the molecular mechanisms by which Treg cells mediate their suppressive function. Materials and Methods Study subjects and statistical analysis Healthy subjects ( n = 11) were adults who were HIV-negative and with no history of chronic viral infections such as Hepatitis B or C. Blood samples from adults with HIV infection ( n = 24) were obtained during routine primary care visits at the Comprehensive Care Center, Vanderbilt University Medical School, Nashville, Tennessee, United States. There were no selection criteria based on race or sex. All subjects provided written consent, and the study was approved by the Vanderbilt Institutional Review Board. Continuous variables were compared by a Mann–Whitney U test, and categorical variables by an χ 2 test. All significance levels were based on two-tailed tests. Statistical analyses were performed using SPSS, version 12.0 (SPSS, Chicago, Illinois, United States). Cell isolation and culture Peripheral blood mononuclear cells (PBMCs) were separated from buffy coats of healthy and HIV-positive donors through Ficoll–Hypaque separation (Pharmacia-LKB Technology, Uppsala, Sweden). Resting CD4 + T-cells were purified as previously described ( Unutmaz et al. 1999 ). This purification protocol typically resulted in 99.5% purity of positively selected cells, as determined by postpurification fluorescence-activated cell sorting (FACS) analysis. To isolate Treg cells, PBMCs, or purified CD4 + cells, were stained with CD45RO, CD25, CD4, and HLA-DR antibodies (BD Biosciences Pharmingen, San Diego, California, United States), and CD4 + CD45RO + CD25 hi and CD4 + CD45RO − CD25 low/neg cells were sorted using flow cytometry (FACS Aria; BD Biosciences Pharmingen). For some experiments, adult CD4 + T-cells were sorted into CD45RO + (memory) and CD45RO − (naïve) T-cells with anti-CD45RO conjugated magnetic beads (Miltenyi Biotec, Bergisch Gladbach, Germany) using AutoMACS (Miltenyi Biotec). Purified resting T-cells were activated by cross-linking with plate-bound anti-CD3 antibody (OKT-3; American Type Culture Collection, Manassas, Virginia, United States) and soluble anti-CD28 antibody (1 μg/ml, BD Biosciences Pharmingen). The plates were first coated with goat antimouse IgG (10 μg/ml, CalTag Laboratories, Burlingame, California, United States) followed by either 3 μg/ml anti-CD3 for optimal TCR stimulation or 100 ng/ml anti-CD3 for suboptimal stimulation. The culture medium used in all experiments was RPMI (Life Technologies, Carlsbad, California, United States) and was prepared as previously described ( Motsinger et al. 2002 ). All cytokines were purchased from R & D Systems (Minneapolis, Minnesota, United States). Monocyte-derived dendritic cells were generated as previously described ( Motsinger et al. 2002 ). Superantigen, staphylococcal enterotoxin B (Sigma, St. Louis, Missouri, United States) was used to stimulate resting T-cells in the presence of dendritic cells ( Motsinger et al. 2002 ). Virus production and infections VSV-G.HIV particles were generated as previously described ( Unutmaz et al. 1999 ). R5-HIV was prepared similarly by transfecting 293T-cells with HIV that encodes R5-tropic (BAL) envelope and enhanced GFP (Clontech, Palo Alto, California, United States) in place of the nef gene as previously described ( Unutmaz et al. 1999 ). Typically viral titers ranged from 1–5 × 10 6 ifu/ml for replication-competent viruses and 10–30 × 10 6 for VSV-G.HIV. T-cells were infected at varying multiplicities of infection (MOI), and infection was quantified by GFP expression using flow cytometry. In some experiments, cells inoculated with virus were centrifuged for 1 h at 2,000 rpm to enhance infectivity, as described by O'Doherty et al. (2000) . Viral replication in T-cell cultures was determined by measuring p24 levels within supernatants by an ELISA ( Motsinger et al. 2003 ), and infectious virus production by infected T-cells was determined by culturing Hut78 cells expressing CCR5 (Hut78/CCR5) ( Wu et al. 2002 ) in infected T-cell supernatants. CFSE labeling Cell division was measured by labeling the T-cells with CFSE (Molecular Probes, Eugene, Oregon, United States). Purified cells were first washed and resuspended in PBS. While vortexing the cells, CFSE was added at a final concentration of 5 μM. The mixture was vortexed for an additional 15 s and incubated at 37 °C for 3 min. Labeling was quenched by addition of 50% fetal calf serum in PBS. Cells were washed one more time with 50% serum PBS, followed by two washes with RPMI-supplemented medium. All CFSE labeling and culturing were performed under dark conditions. FACS analysis and cytokine assay T-cells were stained with the relevant antibody on ice for 30 min in PBS buffer containing 2% fetal calf serum and 0.1% sodium azide. Cells were then washed twice, fixed with 1% paraformaldehyde, and analyzed with a FACSCalibur four-color cytometer. Live cells were gated based on forward- and side-scatter properties, and analysis was performed using FlowJo software (Tree Star, Ashland, Oregon, United States). The following antihuman antibodies were used for staining: CD3, CD4, CD45RO, CD45RA, CD25, GITR, HLA-DR, CCR5, CCR4, CCR7, CXCR4, CXCR3, and antimouse CD24, all obtained from PharMingen (San Diego, California, United States). Cytokines (IL-2, IL-4, IL-5, and IFNγ) in the supernatants were assayed using a commercially available cytometric bead array (CBA) (BD Biosciences Pharmingen) ( Cook et al. 2001 ) and analyzed using CBA 6-bead analysis software (BD Biosciences Pharmingen). Cloning of human FoxP3 RNA was isolated from activated human T-cells using an RNeasy kit (Qiagen, Valencia, California, United States). To synthesize cDNA 100 ng RNA was used (Superscript II Reverse Transcriptase; Invitrogen, Carlsbad, California, United States). FoxP3 was PCR amplified from T-cell cDNA with the following primers: FoxP3 forward, 5′-AGATATCCCAGCCATGCCCAACCCCAGGCCTGGCAAG-3′; FoxP3 reverse, 5′-TCAGGGGCCAGGTGTAGGGTTGGAACACCT-3′. The forward primer included an EcoRV restriction site to facilitate cloning. The FoxP3 PCR product was subcloned into a TOPO shuttle vector (pcDNA3.1/CT-GFP-TOPO; Invitrogen). Following an EcoRV digest, FoxP3 was ligated into an HDV-encoding mCD24 down stream of an internal ribosome entry site ( Sundrud et al. 2003 ). The FoxP3 coding sequence was confirmed by DNA sequencing. Real-time PCR protocol RNA was extracted, as described above, from cells transduced with HDV or HDV.FoxP3 or sorted Treg, naïve, and memory T-cells from HIV-positive and HIV-negative subjects. RNA (100 ng) was used to synthesize cDNA (as described above). Taqman Assays-on-Demand Gene Expression Primers (Applied Biosystems, Foster City, California, United States) were used in real-time PCR analyses: GAPDH primer mix assay ID Hs99999905_m1; FoxP3 primer mix assay ID Hs00203958_m1. Real-time PCR was performed using the ABI 7700 apparatus (PE Applied Biosystems, Weiterstadt, Germany). The reaction mixtures (20-μl total volume) contained 2 μl of serially diluted cDNA, 10 μl of Taqman Universal PCR Master Mix (PE Applied Biosystems), and 1 μl of either FoxP3 or GAPDH primer mix. The reactions were amplified as follows: 50 °C for 2 min and 95 °C for 10 min, followed by 40 cycles of 95 °C for 1 min and 65 °C for 1 min. Expression of FoxP3 was normalized to GAPDH expression in each sample. Supporting Information Accession Number The GenBank ( http://www.ncbi.nlm.nih.gov/ ) accession number for the gene FoxP3 discussed in this paper is AF277993.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC449855.xml
514549	Misoprostol for treating postpartum haemorrhage: a randomized controlled trial [ISRCTN72263357]	Background Postpartum haemorrhage remains an important cause of maternal death despite treatment with conventional therapy. Uncontrolled studies and one randomised comparison with conventional oxytocics have reported dramatic effects with high-dose misoprostol, usually given rectally, for treatment of postpartum haemorrhage, but this has not been evaluated in a placebo-controlled trial. Methods The study was conducted at East London Hospital Complex, Tembisa and Chris Hani Baragwanath Hospitals, South Africa. Routine active management of the third stage of labour was practised. Women with more than usual postpartum bleeding thought to be related to inadequate uterine contraction were invited to participate, and to sign informed consent. All routine treatment was given from a special 'Postpartum Haemorrhage Trolley'. In addition, participants who consented were enrolled by drawing the next in a series of randomised treatment packs containing either misoprostol 5 × 200 μg or similar placebo, which were given 1 orally, 2 sublingually and 2 rectally. Results With misoprostol there was a trend to reduced blood loss ≥500 ml in 1 hour after enrolment measured in a flat plastic 'fracture bedpan', the primary outcome (6/117 vs 11/120, relative risk 0.56; 95% confidence interval 0.21 to 1.46). There was no difference in mean blood loss or haemoglobin level on day 1 after birth < 6 g/dl or blood transfusion. Side-effects were increased, namely shivering (63/116 vs 30/118; 2.14, 1.50 to 3.04) and pyrexia > 38.5°C (11/114 vs 2/118; 5.69, 1.29 to 25). In the misoprostol group 3 women underwent hysterectomy of whom 1 died, and there were 2 further maternal deaths. Conclusions Because of a lower than expected incidence of the primary outcome in the placebo group, the study was underpowered. We could not confirm the dramatic effect of misoprostol reported in several unblinded studies, but the results do not exclude a clinically important effect. Larger studies are needed to assess substantive outcomes and risks before misoprostol enters routine use.	Background Excessive bleeding from the genital tract after birth, or postpartum haemorrhage (PPH) is the major cause of maternal deaths in many low-income countries. The global estimate is 125,000 deaths per year [ 1 ]. In South Africa, 240 of 2,445 maternal deaths reported between 1999 and 2001 were due to postpartum haemorrhage, the third most common cause [ 2 ]. In a community-based study in Senegal, estimates of maternal mortality ratio in three regions ranged from 436 to 852 per 100,000 live births. Two-thirds were due to direct obstetric causes, the commonest being haemorrhage [ 3 ]. In the United Kingdom, the risk of maternal death from haemorrhage is about 1 in 100 000 births [ 4 ]. The potential to save mothers' lives with medical interventions for haemorrhage is thus considerable. There is good evidence that a policy of active management of the third stage of labour [ 5 ], and one component of active management, the routine administration of uterotonics (drugs to contract the uerus) such as oxytocin [ 6 ] or oxytocin and ergometrine [ 7 ], are effective in reducing the risk of postpartum haemorrhage. When postpartum haemorrhage occurs, a number of medical and surgical interventions are used to control the bleeding [ 8 ]. A crucial aspect of both prevention and treatment of postpartum haemorrhage is uterotonic therapy. The most commonly used agents are injectable oxytocin and/or ergometrine. Misoprostol is a methyl ester (a synthetic analogue) of natural prostaglandin E1. It is marketed and registered for use in the prevention and treatment of peptic ulcer disease caused by prostanglandin synthetase inhibitors. Administered orally or sublingually, peak plasma concentrations of misoprostolic acid are achieved in less than 30 minutes [ 9 ]. It is a thermostable drug [ 9 ] and is relatively inexpensive. Administered orally or vaginally, it is an effective agent for the induction of abortion [ 10 ] and of labour [ 11 ]. Misoprostol for routine prevention of postpartum haemorrhage Oral and rectal misoprostol have been used for routine management of the third stage of labour (after the birth). Several small trials have given conflicting results. The main side-effects have been shivering and pyrexia, which are dose-dependent [ 12 ]. In the large WHO Collaborative Trial of Misoprostol in the Management of the Third Stage of Labour [ 13 ] and a systematic review of the topic [ 14 ], blood loss >1,000 ml and use of additional oxytocics (the primary outcomes in the WHO trial) were more frequent with misoprostol than with injectable oxytocics, indicating that injectable oxytocics should remain the drug of choice for routine prophylaxis. On the other hand, blood transfusion was used less frequently in the misoprostol group. This may have been a chance occurrence (it was not a primary outcome for the trial). Secondly, there could be a synergistic pharmacological effect between misoprostol and conventional uterotonics, which were given to most women with increased blood loss. Many women in the misoprostol group thus received misoprostol as well as conventional uterotonics, whereas those in the oxytocin group received only conventional oxytocics. Thirdly, we have suggested in the report of a pharmacokinetic study linked to the WHO trial that the longer time to peak levels of misoprostol (20–30 minutes) than syntocinon (3 minutes) may account for more early blood loss with misoprostol [ 15 ]. This does not exclude the possibility that misoprostol may have an effect on more persistent bleeding. Physiological studies have also shown a more rapid onset of uterine contractions following syntometrine than misoprostol after delivery [ 16 ]. Misoprostol has been widely recommended for the prevention of postpartum haemorrhage when other methods are not available [ 17 ]. Misoprostol for treatment of postpartum haemorrhage Apart from the prophylactic use of misoprostol in the third stage of labour, the therapeutic use of misoprostol for the treatment of postpartum haemorrhage has been promoted on the basis of unblinded studies. In a systematic review [ 18 ] we located 6 uncontrolled reports (41 women) [ 19 - 24 ] and one small, unblinded randomised trial [ 25 ]. This method has entered clinical use, particularly in developing countries, without systematic research to document the optimal route and dosage, effectiveness or risks of this treatment. The urgent need for randomised trials of this new intervention has been emphasised [ 26 ]. If effective it could have a major impact on maternal mortality, particulary in under-resourced countries. If not, its use should be discouraged because of the dangers of side-effects, and the risks associated with widespread introduction of misoprostol into health systems in which it might be used for labour induction in inappropriate doses, with the risk of fetal compromise and uterine rupture [ 27 , 28 ]. Route and dosage of misoprostol The most common regimens reported for the treatment of postpartum haemorrhage are 800 [ 23 ] or 1,000 μg [ 19 , 21 ] rectally. We reviewed the literature on the pharmacokinetics of misoprostol administered by various routes [ 18 ]. The oral route has the most rapid uptake, but the shortest duration. The rectal route has slow uptake but prolonged duration. The buccal or sublingual route has rapid uptake, prolonged duration and greatest total bioavailability. In order to improve the rapidity of onset of action and the overall bioavailability, we modified the reported practice of using 1,000 μg rectally, by administering 200 μg orally, 400 μg sublingually and 400 μg rectally. Objective The objective of this study was to determine the side-effects and effectiveness of high-dose misoprostol for the treatment of postpartum haemorrhage by means of a double-blind, placebo-controlled, randomised clinical trial. The hypothesis was that measured blood loss of 500 ml or more in the hour after enrolment would be significantly less frequent in the misoprostol than in the placebo group. Methods The study was conducted at the East London Hospital Complex (Frere Maternity and Cecilia Makiwane Hospitals, Eastern Cape) and Tembisa and Chris Hani Baragwanath Hospitals, Gauteng, South Africa, between Jan 2002 and Dec 2003. All are busy referral hospitals. Treatment packs were prepared independently, ordered in computer-generated random sequence and numbered consecutively. The packs contained either misoprostol 5 × 200 μg or inactive placebo. The treatment sequence was kept sealed, and the code broken only after complete entry and checking of all trial data. A confidential interim analysis, blind to which group was which, was performed by AMG after the first 100 enrolments. Evidence of clear benefit or harm for either group would have prompted stopping the trial, which was not the case. Women in labour were given basic information about the trial in the form of notices in the labour rooms (Table 2 [see additional file 1 ]). Management of labour 3 rd stage was routine active management with oxytocin 10 units or syntometrine one ampoule soon after the birth. Women aged 18 or more with bleeding judged by the clinician to be more than expected at least 10 minutes after delivery, and thought to be due to uterine atony and requiring additional uterotonic therapy, were given all the routine treatment for PPH (intravenous infusion, uterotonics, etc) from a special 'PPH Trolley' kept in the labour ward. The uterotonics used were oxytocin by intravenous infusion, and/or oxytocin/ergometrine, at the discretion of the attending clinician. Prostaglandin preparations such as sulprostone were not routinely available in the labour wards, and were not part of routine management of postpartum haemorrhage in the participating units. Once all emergency treatment was instituted, and if the women were in a position to give fully informed consent, they were given detailed information about the trial and asked whether they wished to participate. Information about the trial was given verbally and in writing in English or Xhosa (Table 3 [see additional file 2 ]). Those who gave written consent were enrolled in the trial. Baseline data were documented. The next treatment pack containing either misoprostol 5 × 200 μg or inactive placebo was drawn and the number recorded. In addition to routine management, misoprostol or placebo (an inactive base) were given, 1 orally, 2 buccally/sublingually and 2 rectally, and the time recorded. A low-profile plastic 'fracture bedpan' was placed under the woman's buttocks. In previous studies we have shown this to be an efficient method of collecting ongoing blood loss with very little spillage [ 29 ]. Any small swabs soaked with blood were dropped into the bedpan. After 1 hour, the blood collected in the bedpan was measured in a graduated measuring jug, the temperature was measured sublingually with a mercury thermometer, and any shivering was subjectively recorded as 'mild' (not causing any distress), 'moderate' or 'severe'. All other management was by hospital staff according to the hospital routine for the management of postpartum haemorrhage. After 24 hours, blood was collected for haemoglobin estimation and the hospital records checked for use of additional uterotonics and any other complications such as blood transfusion. The primary outcome measures were specified prior to commencing the study: (1) measured blood loss ≥500 ml in 1 hour after enrolment; (2) mean measured blood loss in 1 hour after enrolment; (3) haemoglobin level day 1 after birth <6 g/dl or blood transfusion; (4) Side-effects (pyrexia 38.5°C or more, moderate or severe shivering 1 hour after enrolment). Secondary outcome measures were: (1) blood loss ≥1,000 ml in 1 hour after enrolment; (2) blood transfusion; (3) haemoglobin level 1 day after birth <8 g/dl or blood transfusion; (4) additional uterotonic given after enrolment; (5) manual removal of the placenta; (6) evacuation of retained products of conception; (7) hysterectomy; (8) maternal death. Sample size calculation In the WHO misoprostol trial control group (routine syntocinon), significant postpartum haemorrhage (blood loss >1000 ml) occurred in 2.8% of women. Of these 25% lost >1,500 ml. To reduce blood loss >500 ml after enrolment from 25% to 10%, would require 112 per group (α = 95%, β = 80%). Data were collected on a data form, entered onto an excel spreadsheet, checked for accuracy, then analysed using Epi-info 2002 (United States Department of Health and Human Services, Centres for Diseases Control and Prevention, Epidemiology Program Office) and Review Manager (RevMan) [Computer program] version 4.2 for Windows. Oxford, England: The Cochrane Collaboration, 2003. Results were expressed as relative risks or mean difference with 95% confidence intervals. Analysis was by intention to treat. Ethical considerations All women enrolled in the trial received all the conventional management available for postpartum haemorrhage, in addition to the trial medication (misoprostol or placebo). Rapid conventional therapy was facilitated by the ready availability of the 'PPH trolley' with all the necessary materials. This trolley was also available for women not enrolled in the trial. Participation was limited to women who were in a position to give fully informed consent. Ethical approval was given by the Committee for Research on Human Subjects, University of the Witwatersrand, and the Ethics Committee, East London Hospital Complex. The trial complied with the Helsinki Declaration for research on human subjects. Results Altogether 244 women were enrolled in the trial. The pack numbers for 6 women were incompletely filled in on the data sheets. The group allocation of these women was therefore unknown, and they could not be included in the analysis. Of these 6 women, the highest measured blood loss after enrolment was 220 ml; none had pyrexia >37.5°C, one had moderate shivering, one had a blood transfusion and none had day 1 haemoglobin level below 8 g/dl or other complications. There was no reason to expect that these exclusions occurred in any selective way which would introduce bias into the final samples, nor could their results have materially altered the conclusions of the study. All the remaining 238 women were included in the analyses. The baseline characteristics are shown in table 1 . As would be expected in a randomised trial, the groups were well matched. Parity was not recorded at all hospitals, but was similar between groups where recorded (misoprostol mean 1.61, SD 1.14; placebo 1.75, SD 1.26). Table 1 Comparison of outcomes between women who received misoprostol or placebo, expressed as proportions (percent) or mean values [standard deviation]. Differences are expressed as relative risks or mean differences, with 95% confidence intervals. Misoprostol (117) Placebo (121) Baseline characteristics n n Age (years) 116 27.1 [6.0] 119 26.2 [6.2] Ergometrine before enrolment 106 36 (34%) 104 34 (33%) Oxytocin ≥20 U before enrolment 116 82 (71%) 117 78 (67%) Primary outcomes: RR/ MD 95% Conf. interval Blood loss ≥500 ml* 117 6 (5.1%) 120 11 (9.2%) 0.56 0.21 to 1.46 Mean blood loss* [SD] (ml) 117 168 [163] 120 176 [173] -8 -51 to 35 24 hour Haemoglobin <6 g/dL or blood transfusion 110 20 (18.2%) 116 17 (14.7%) 1.24 0.69 to 2.24 Pyrexia at 1 hour ≥38, 5°C 114 11 (9.6%) 118 2 (1.7%) 5.69 1.29 to 25 Shivering at 1 hour ≥ moderate 116 63 (54%) 118 30 (25%) 2.14 1.50 to 3.04 Secondary outcomes: Blood loss ≥ 1,000 ml* 117 1 (0.85%) 120 0 (0%) Blood transfusion 115 19 (17%) 119 15 (13%) 1.31 0.70 to 2.45 24 hour haemoglobin <8 g/dL or blood transfusion 110 43 (39%) 116 37 (32%) 1.23 0.86 to 1.75 Additional uterotonic after enrolment 111 63 (57%) 112 63 (56%) 1.01 0.80 to 1.27 Manual removal of the placenta 117 1 (0.85%) 121 4 (3.3%) 0.26 0.03 to 2.28 Evacuation of retained products of conception 117 2 (1.7%) 121 0 (0%) Hysterectomy 117 3 (2.6%) 121 0 (0%) Maternal death 117 3 (2.6%) 121 0 (0%) RR = relative risk; MD = mean difference; Conf = confidence; SD = standard deviation Measured, within 1 hour after enrolment * One woman died after hysterectomy and is counted in both outcomes The outcomes are shown in table 1 . In the placebo group, the primary outcome (measured blood loss within 1 hour after enrolment of 500 ml or more) was less frequent (9.2%) than the 25% prediction on which the sample size calculation was based. This may have been because the criteria for enrolment were loosely defined as 'more than expected blood loss', resulting in the enrolment of women with less severe blood loss than anticipated. The fact that enrolment was limited to women who were able to received detailed information about the study, may have caused selection of less severe cases. Although the intention was to enrol women who had not responded to conventional therapy, the time taken for counselling of the women and obtaining informed consent may have resulted in some women responding to the primary treatment, and therefore the low rate of ongoing blood loss in the placebo group. The study was thus underpowered to detect a statistically significant reduction in the primary outcome. With misoprostol there was a trend to reduced blood loss ≥500 ml in 1 hour after enrolment (6/117 vs 11/120, relative risk 0.56; 95% confidence interval 0.21 to 1.46). Other proxy estimations of blood loss showed no significant differences between the groups. The well-known side-effects of misoprostol, pyrexia and shivering, were significantly more frequent in the misoprostol than the placebo group. Serious morbidity or mortality occurred in 5 women, all from the misoprostol group. Two women who continued to bleed despite conservative therapy underwent abdominal hysterectomy and recovered well. There were 3 maternal deaths: Case 1 A 22-year old woman with one previous pregnancy developed postpartum haemorrhage after vaginal delivery, managed with a 40-unit oxytocin infusion. She was enrolled and received the trial medication 85 minutes after delivery. Thereafter she received a second 40-unit oxytocin drip, oxytocin/ergometrine (5 units/0.5 mg) and intravenous cyclokapron. In the hour after randomisation, measured blood loss was 125 ml. Subsequently bleeding continued and a sub-total hysterectomy was performed. Coagulopathy developed, bleeding continued through an abdominal drain, and she died 2 days after delivery despite re-laparotomy and multiple blood product transfusions. Case 2 A 32-year-old woman, para 2, gravida 4, delivered normally after labour was induced with misoprostol, 25 μg vaginally and 4 doses of 50 μg orally. Before enrolment she received oxytocin10 units, ergometrine 0.5 mg and a 20-unit oxytocin infusion. She was enrolled and received the trial medication 140 minutes after delivery. After enrolment she received a further 40-unit oxytocin infusion. Measured blood loss in the hour after enrolment was 380 ml. Bleeding continued and a blood transfusion was commenced. Six and a half hours after delivery she suffered a cardiac arrest and was resuscitated. Examination in theatre found the uterus to be empty and intact. The main source of bleeding appeared to be a torn cervix, which was sutured. At 7.5 hours after delivery she suffered a second cardiac arrest and could not be resuscitated. The estimated total blood loss was 3,000 ml. Case 3 A woman with one previous delivery by caesarean section, developed postpartum haemorrhage of about 500 ml after a vaginal delivery. She was enrolled 85 minutes after delivery. Measured blood loss over the next hour was 425 ml, after which she developed hypotension and cardiorespiratory arrest. As no post-mortem examination was performed, the possibility of internal bleeding from a dehisced caesarean section scar could neither be confirmed nor excluded. Unresponsive uterine atony was therefore not confirmed as the main cause of any of the deaths. Discussion Pyrexia and shivering were common side-effects of misoprostol, as found in previous studies. Three women, all in the misoprostol group, had severe pyrexia >40°C. Previous studies of high-dose misoprostol for the treatment of postpartum haemorrhage have mostly used the rectal route, and none has had sufficient numbers to be likely to detect rare adverse outcomes such as severe pyrexia. These side-effects may be related to the rapid absorption of misoprostol given orally, and the rapid absorption and high bioavailability when given sublingually. Because of these serious side-effects, we would recommend that for future trials the dose be reduced. Several previous unblinded studies have reported dramatic effects of misoprostol when used to treat postpartum haemorrhage. In 6 observational studies [ 19 - 24 ], 41 women with postpartum haemorrhage were treated with misoprostol 1,000 μg rectally (32 women), 200 μg rectally (5 women), 200 μg orally 2-hourly (2 women) or 800 μg intrauterine (2 women). In all but 2 of the women (who received 1,000 μg rectally), a prompt response to the treatment was reported. In a single blind randomised trial [ 25 ] subjective prompt cessation of bleeding was reported in 30/32 women who received misoprostol 800 μg rectally compared with only 21/32 who received oxytocin 5 units plus ergometrine 0.5 mg intramuscularly. Conclusions In our double blind trial we have been unable to confirm the dramatic effects of misoprostol reported previously in unblinded studies. Our results, however, are consistent with anything from a large beneficial effect to a smaller adverse effect (relative risk of additional blood loss of 500 ml or more anywhere between a reduction of 79% and an increase of 46% with misoprostol). It is important that this possible benefivial effect be assessed by sufficiently powered randomised trials before the treatment enters routine clinical use. List of abbreviations PPH: postpartum haemorrhage; SD: standard deviation Competing interests None known. Authors' contributions GJH prepared the first draft of the protocol and the final manuscript, and oversaw the clinical work CN assisted with the design of the protocol and data collection sheet, initiated enrolment at two sites, supervised some of the data collection and contributed to the manuscript SF, ZM, LM, MS and ZJ commented on the protocol, conducted the clinical procedures and collected data. SF entered and collated the data BM undertook the literature search and commented on the protocol. GW contributed to the protocol development AMG contributed to the protocol development and undertook the blinded interim analysis All authors read and approved the final manuscript Additional files Additional file 1 - Table 2. Notice informing women about the trial, MS Word file: HofmeyrFile1.doc Additional file 2- Table 3: Information and consent form. MS Word file: HofmeyrFile2.doc Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Table 2. Notice informing women about the trial Click here for file Additional File 2 Table 3: Information and consent form Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514549.xml
514561	The effect of Fucus vesiculosus, an edible brown seaweed, upon menstrual cycle length and hormonal status in three pre-menopausal women: a case report	Background Rates of estrogen-dependent cancers are among the highest in Western countries and lower in the East. These variations may be attributable to differences in dietary exposures such as higher seaweed consumption among Asian populations. The edible brown kelp, Fucus vesiculosus (bladderwrack), as well as other brown kelp species, lower plasma cholesterol levels. Since cholesterol is a precursor to sex hormone biosynthesis, kelp consumption may alter circulating sex hormone levels and menstrual cycling patterns. In particular, dietary kelp may be beneficial to women with or at high risk for estrogen-dependent diseases. To test this, bladderwrack was administered to three pre-menopausal women with abnormal menstrual cycling patterns and/or menstrual-related disease histories. Case Presentation Intake of bladderwrack was associated with significant increases in menstrual cycle lengths, ranging from an increase of 5.5 to 14 days. In addition, hormone measurements ascertained for one woman revealed significant anti-estrogenic and progestagenic effects following kelp administration. Mean baseline 17β-estradiol levels were reduced from 626 ± 91 to 164 ± 30 pg/ml ( P = 0.04) following 700 mg/d, which decreased further to 92.5.0 ± 3.5pg/ml ( P = 0.03) with the1.4 g/d dose. Mean baseline progesterone levels rose from 0.58 ± 0.14 to 8.4 ± 2.6 ng/ml with the 700 mg/d dose ( P = 0.1), which increased further to 16.8 ± 0.7 ng/ml with the 1.4 g/d dose ( P = 0.002). Conclusions These pilot data suggest that dietary bladderwrack may prolong the length of the menstrual cycle and exert anti-estrogenic effects in pre-menopausal women. Further, these studies also suggest that seaweed may be another important dietary component apart from soy that is responsible for the reduced risk of estrogen-related cancers observed in Japanese populations. However, these studies will need to be performed in well-controlled clinical trials to confirm these preliminary findings.	Background Epidemiological studies show that incidence rates of estrogen-dependent diseases such as cancers of the breast, endometrium and ovary are among the highest in Western, industrialized countries, while rates are much lower in China and Japan [ 1 , 2 ]. These disparities may be attributable, in part, to differences in dietary and environmental exposures associated with affluent and modern lifestyles that promote estrogenic stimulation and hormone imbalances [ 3 - 5 ]. Although the mechanisms are not fully understood, epidemiological and experimental data suggest that exposure to estrogens, through endogenous production and exogenous exposures resulting in an imbalance in the estrogen/progesterone ratio, may be the most critical determinants in disease risk [ 6 - 8 ]. In estrogen-sensitive tissues, estrogen triggers cell proliferation, and through prolonged stimulation, hyperplasia [ 9 ] and possibly neoplasia can occur. Reproductive factors associated with increased exposure to menstruation resulting in persistent and sustained estrogenic stimulation, such as shorter menstrual cycles, reduced parity, early menarche, and late menopause, are known to increase risk of endometriosis and estrogen-dependent cancers [ 10 , 11 ], while post-menopausal obesity, hormone replacement therapy and alcohol consumption may be associated with increased breast cancer risk [ 12 - 14 ]. Therefore, limiting exposure to estrogens and reducing the overall number of menstrual cycles in one's lifetime through dietary and lifestyle changes may be the simplest means to reduce disease risk. In particular, the identification of dietary compounds that have estrogen- reducing effects holds great promise in developing chemopreventive strategies to abrogate risk of these diseases. Studies show that Japanese women have longer menstrual cycle lengths (greater than the 28 day average) and lower circulating estrogen levels compared to Western populations [ 15 - 17 ], which until now has been at least partly attributed to the increased intake of soy protein among Asian populations [ 18 - 20 ]. Another less explored component but main staple of the Japanese diet is seaweed, which accounts for approximately 10–25% of their food intake [ 21 , 22 ]. Other reported estimated daily intakes are as high as 3–13 g/day [ 23 ]. A major source of dietary seaweed among Japanese populations is the edible brown kelp, wakame ( Undaria pinnatifida ) and kombu ( Laminaria japonica ). These species and the Atlantic brown kelp, bladderwrack ( Fucus vesiculosus) , have been shown to exert powerful anti-hypertensive activity related to angiotensin-I-converting enzyme inhibition [ 24 ], to possess antibacterial and antioxidant properties related to their high polyphenolic content [ 25 ], and to prevent dioxin absorption and accelerate dioxin excretion in rats [ 26 ]. Other chemopreventive properties such as antiviral activity [ 27 , 28 ], immunostimulatory effects [ 29 ], anti-proliferative effects on 7,12-dimethylbenz(a)-anthracene-induced rat mammary tumors [ 30 , 31 ], and anti-tumor and anti-metastatic activities in xenograft mouse models [ 32 ], have been associated with the high level of sulfated polysaccharides, also known as fucoidans, found in brown seaweed. Intake of bladderwrack, as well as other brown kelp species, also has been shown to alter cholesterol metabolism and to significantly lower plasma cholesterol levels [ 33 , 34 ]. A possible mechanism of action involves competitive inhibition by fucosterols found in kelp. Since cholesterol is the precursor involved in steroid hormone biosynthesis, a reduction in cholesterol bioavailability could lower circulating plasma 17β-estradiol levels that may lead to alterations in menstrual cycling patterns in pre-menopausal women. Until now, no studies have been performed in humans to determine the effects of brown kelp on menstrual cycling patterns and sex hormone status in pre-menopausal women, particularly in women with or at risk for estrogen-dependent diseases. To explore the hypothesis that kelp consumption could reduce circulating17β-estradiol levels and attenuate menstrual cycle irregularities, bladderwrack was administered to three pre-menopausal women with abnormal menstrual cycling patterns and/or menstrual-related disease histories. Case presentation Three pre-menopausal women with abnormal menstrual cycling histories volunteered for the present study. An abnormal menstrual cycle was defined as one or more of the following: menstrual cycles of <26 or >32 days in length; menstrual cycles consisting of >8 menstruating days; or anovulatory menstrual cycling. Study subject characteristics are outlined in Table 1 . Subject 1 had a history of hypermenorrhea (excessive blood loss during menstruation), polymenorrhea (shorter than average menstrual cycle length of 28 days), anovulatory menstrual cycles, and was diagnosed with luteal phase deficiency and endometriosis (through laparoscopy). Subject 2 suffered from hypermenorrhea and polymenorrhea. Subject 3 suffered from hypermenorrhea and was diagnosed with endometriosis. All three women reported a history of dysmenorrhea (painful menses). Otherwise, all women were in general good health and free of any chronic diseases. All women were active and exercised approximately three times per week. No hormones or other medications were taken for >3 months prior to the inception of the study. No soy protein products were consumed during the study period. Table 1 Study Subject Characteristics Subject Age Body weight (lb) Menstrual cycle history Medical conditions/health status Medications 1 43 142 hypermenorrhea, polymenorrhea, dysmenorrhea, luteal phase deficiency endometriosis; otherwise healthy none 2 42 138 hypermenorrhea, polymenorrhea, dysmenorrhea general good health none 3 21 126 hypermenorrhea, dysmenorrhea endometriosis; otherwise healthy none The protocol was approved by the Committee for the Protection of Human Subjects of the University of California at Berkeley. The nature of the study was explained, and written informed consent was obtained from all study subjects. Source and dose of bladderwrack (Fucus vesiculosus) Dried, powdered bladderwrack was obtained from Maine Coast Sea Vegetables (Franklin, ME) and encapsulated in 350 mg capsules. Two capsules were administered daily for the low dose treatment (700 mg) and four capsules were administered daily for the high dose treatment (1.4 g). Bladderwrack dosage levels were chosen to fall within the range of reported dietary seaweed intakes (10–25%) of the total diet reported for Japanese populations [ 21 , 22 ]. This was calculated by assuming a total 500 g/d total dietary intake and a range between 50–125 g/d (wet weight) or 0.5–1.25 g/d (dry weight) seaweed intake. This calculation falls below the estimated 3–13 g/d intake reported by Teas et al. [ 23 ]. Experimental protocols Details of the study protocol are outlined in Table 2 . All women provided self-reported menstrual cycling histories for the three months prior to the treatment period. In addition, 17β-estradiol and progesterone serum measurements were taken for Subject 1 throughout the course of the study as outlined in Table 2 . Ovulation was monitored through body basal temperature. Since the average length of her cycle was 16 days prior to treatment and she was not ovulating at the inception of the study, baseline hormone levels were ascertained on a set day (menstrual cycle day 12) for two consecutive cycles prior to the administration of 700 mg/d bladderwrack for two additional cycles. During the treatment period, serum hormone levels were measured on days 12 and 21 for the first cycle (which was another anovulatory cycle) and on day 21 thereafter during the treatment period. Subjects 2 and 3 were administered 700 mg/d of bladderwrack beginning on day 21 of their menstrual cycles and followed for two consecutive cycles. Subsequently, Subjects 1 and 3 agreed to continue the experiment for two additional cycles at which time they received a daily dose of 1.4 g/d kelp. Menstrual cycling logs were maintained on all subjects during the entire course of the experiment. Subjects were monitored at least weekly to insure compliance to the supplement regimen. Table 2 Treatment protocol and timeline Subject Pretreatment menstrual cycling history obtained Pretreatment serum 17β-estradiol and progesterone levels ascertained (Cycle and day) Treatment period/dose Treatment serum 17β-estradiol and progesterone levels ascertained (Cycle and day) Treatment period/dose Treatment serum 17β-estradiol and progesterone levels ascertained (Cycle and day) 1 Cycles 1–3 Cycle 2, day 12; Cycle 3, day 12 Cycle 4–5/ 700 mg/d Cycle 4, day 12; Cycle 4, day 21; Cycle 5, day 21 Cycle 6–7/ 1.4 g/d Cycle 6, day 21; Cycle 7, day 21 2 Cycles 1–3 NA Cycle 4–5/ 700 mg/d NA NA NA 3 Cycles 1–3 NA Cycle 4–5/ 700 mg/d NA Cycle 6–7/ 1.4 g/d NA Hormone assays All hormone assays were performed by Quest Laboratories (San Diego, CA), an outside-certified clinical laboratory. Serum 17β-estradiol and progesterone levels were measured in duplicate by radioimmunoassays. Blank and control sera were run with each assay. The coefficient of variation (a measure of laboratory error) was consistently low (<15%) for 17β-estradiol and progesterone. Statistical methods Statistical analyses were performed by unpaired t-tests (2-sided) with a commercially available statistical software package (Stata, College Station, Texas). All statistical tests were considered significant for p ≤ 0.05. Results are referred to as borderline significant for 0.05 < p ≤ 0.10. Clinical findings There were no adverse side effects reported and bladderwrack was well tolerated by all three women. Effects of treatment on length of menstrual cycle and total days of menstruation Following treatment, all women exhibited a significant increase in menstrual cycle lengths (Figure 1 ). Specifically, in Subject 1, who had a 30-year history of irregular menses, the menstrual cycle length increased from an average of 16.3 ± 0.6 days to 26.0 ± 1.4 days with the low dose treatment ( P < 0.002), which increased by approximately 5 additional days to 31.2 ± 1.1 days following administration of the higher dose ( P < 0.001). In Subject 2, the average cycle length increased 5.5 days, from 23.0 ± 1.7 to 28.5 ± 0.7 days ( P = 0.03). Subject 3 exhibited a 4-day increase in menstrual cycle length from 27.3 ± 0.6 to 31.5 ± 0.7 days with the 700 mg dose ( P = 0.005) that increased by approximately 6 more days to 36.0 ± 2.8 days with the 1.4 g dose ( P = 0.01). Figure 1 Average menstrual cycle length in days for Subjects 1–3 at baseline (black bars) and following bladderwrack administration of 700 mg/d (diagonal striped bars) and 1.4 g/d (white bars). Black bars indicate the averages of 3 menstrual cycles; diagonal striped and white bars indicate the averages of 2 menstrual cycles; and whiskers indicate standard deviations. * P value <0.05. Along with increased menstrual cycle lengths, all women reported marked reductions in blood flow and average number of days of menstruation following bladderwrack treatment (Figure 2 ). Subject 1 reported the most significant reduction in total days of menstruation, changing from an average 9.3 ± 0.6 to 6.3 ± 1.8 days ( P = 0.06) with the low dose and to 4.5 ± 0.7 average days with the high dose ( P < 0.003). Subject 2, who only took the low dose, also experienced a marked reduction in number of days of menstruation, from 8.0 ± 1.0 to 5.3 ± 2.5 days ( P = 0.06). Subject 3 exhibited a decrease in total menstruating days averaging from 6.3 ± 1.5 to 5.8 ± 0.4 days ( P = 0.65) with the low dose, and to 3.5 ± 0.7 days ( P = 0.10) with the 1.4 g/d dose. Subjects 1 and 3, who both suffered from endometriosis, reported substantial alleviation from pain during menstruation and throughout the menstrual cycle following bladderwrack treatment. Figure 2 Average number of days of menstruation per cycle for Subjects 1–3 at baseline (black bars) and following bladderwrack administration of 700 mg/d (diagonal striped bars) and 1.4 g/d (white bars). Each bar indicates averages from two menstrual cycles; whiskers indicate standard deviations. * P value <0.05. Subject 1 also reported an ovulatory cycle following 2 months on the 700 mg/d kelp intervention, and continued to have ovulatory cycles while on the 1.4 g/d dose. Effects of treatment on serum estradiol and progesterone levels A significant anti-estrogenic and progestagenic dose response was observed in plasma estradiol and progesterone levels in Subject 1 (Table 3 ). Specifically, the mean baseline 17β-estradiol levels were reduced from 626 ± 91 to 164 ± 30 pg/ml ( P = 0.04) with the low dose (700 mg/d), which decreased further to 92.5 ± 3.5 pg/ml ( P = 0.03) with the higher dose (1.4 g/d). Furthermore, mean baseline progesterone level rose from 0.58 ± 0.14 to 8.4 ± 2.6 ng/ml with the lower 700 mg/d dose ( P = 0.1), which increased further to 16.8 ± 0.7 ng/ml with the 1.4 g/d dose ( P = 0.002). Table 3 Average circulating plasma 17β-estradiol and progesterone levels prior to and during kelp intervention for Subject 1 Pre-treatment baseline levels 700 mg/d dose P -value 1.4 g/d dose P-value 17β-estradiol (pg/ml) 626 ± 91 164 ± 30 0.04 92.5 ± 3.5 0.03 Progesterone (ng/ml) 0.58 ± 0.14 8.4 ± 2.6 0.10 16.8 ± 0.7 0.002 All measures are from 2 month averages ± S.D. Discussion of clinical findings The results of this preliminary pilot study suggest that bladderwrack consumption can effectively increase the length of the menstrual cycle and reduce the total number of days of menstruation in pre-menopausal women. These effects were most marked in the two women that had shorter than average cycles (16 and 23 days) versus the normal range of 26 to 32 days seen in women in Western populations. Menstrual cycles were further lengthened with increasing dose, which may suggest a linear dose-response effect. However, since there was not a sufficient washout period between the 700 and 1400 mg/d doses, an effect of length of time of dosing rather than a dose response effect cannot be ruled out. Nonetheless, these marked increases in menstrual cycle length may have beneficial health effects in lowering risk of estrogen-dependent diseases such as endometriosis and ovarian, endometrial, and breast cancers as reported in a number of studies [ 16 , 35 - 38 ]. Menstrual characteristics are surrogate markers that may reflect a woman's overall exposure to and production of endogenous hormones. Shorter menstrual cycle lengths and prolonged menstruation confer longer follicular and luteal phases where estrogen and progesterone levels and endometrial and breast cell proliferation rates are at their highest. A nearly fourfold increase in mitotic activity in the breast lobules occurs during the luteal phase of the menstrual cycle [ 39 ], while the highest proliferation rates (nearly 100-fold) in the endometrium occur during the follicular phase [ 40 ]. Therefore, fewer menstrual cycles over a woman's lifetime would decrease the amount of time during which the breast and endometrial epithelia would be exposed to high levels of proliferation, which may decrease overall disease risk. Bladderwrack consumption also led to a marked reduction in circulating 17β-estradiol levels and an increase in progesterone levels in a subject who exhibited high serum estrogen levels and progesterone deficiency prior to the intervention. While estrogen's proliferative effects on mammary gland development and endometrial and breast tumorigenesis are well documented, progesterone's role in these processes is not as well defined. Studies show that progesterone deficiency is associated with increased endometrial cancer incidence [ 41 ], and that progesterone inhibits estrogen-induced luminal epithelial proliferation in the uterus [ 42 ]. However, progesterone has been shown to both stimulate and inhibit the growth of experimental mammary tumors [ 43 ], and the use of synthetic progestins in hormone replacement therapy has been associated with an increased risk of breast cancer [ 43 ]. Experimental rat models have elucidated progesterone's vital role in pregnancy-induced morphological changes in the breast, which confer protection against breast cancer [ 44 ]. Further, epidemiological studies suggest that it is not pregnancy alone but early first parity and increasing number of pregnancies that are associated with reduced breast cancer risk [ 45 , 46 ]. These studies suggest that the effects of progesterone in breast cancer risk may be dependent on timing and the type and level of progesterone/progestin exposure. Thus, the anti-estrogenic/progestagenic activity of kelp observed in this study warrants further investigation in its role in breast cancer and other hormone-dependent diseases. Study limitations Due to the small number of subjects and the lack of a control group, this study will need to be repeated in a larger, randomized population of women with placebo controls. Other weaknesses of the present study are the lack of data on luteinizing hormone and follicular stimulating hormone levels which would provide pertinent information regarding the effects of dietary bladderwrack on ovulation and the luteal and follicular phases of the menstrual cycle. The potential beneficial impact that dietary bladderwrack may have on abrogating symptoms of endometriosis warrants a closer look at a larger population of women suffering from this disease. However, studies should also be performed in women with normal menstrual cycles who have sex hormone levels within clinically normal ranges to determine the impact of dietary kelp on menstrual cycling patterns and hormone levels in the general population. Conclusions The observed responses to bladderwrack consumption in this study suggest that dietary modification may lead to significant changes in the regulation of the menstrual cycle by increasing the length of the cycle, stimulating ovulation, and lowering the estrogen/progesterone ratio in pre-menopausal women. Such changes may be beneficial particularly with regard to women at high risk of estrogen-dependent diseases or who are experiencing fertility problems. Results from these preliminary experiments also suggest that bladderwrack administration may alleviate hypermenorrhea and dysmenorrhea, which may provide some relief in the treatment of endometriosis. Although these reported effects are generally in a beneficial direction, their clinical significance is yet to be determined in a well-controlled study. Future Directions The critical role of hormones in breast, endometrial, and ovarian cancers in women and prostate cancer in men has long been recognized. Given the vast rise of these cancers in the U.S. and our limited success with prevention and treatment, there is clearly a need for the identification of novel, non-cytotoxic chemopreventive agents. Future investigations should clarify the role of bladderwrack and other seaweed species on estrogen and progesterone metabolism, to evaluate its potential binding affinity to estrogen and progesterone receptors, and to determine its effects on proliferation in hormone-sensitive tissues. These investigations should also be expanded to include effects of bladderwrack on other sex hormones including the androgens and gonadotropins. In this regard, animal and in vitro studies are currently underway in our laboratory to elucidate the potential mechanisms and clinical relevance of bladderwrack bioactivity, and to identify and isolate the active components involved. Competing Interests None declared. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514561.xml
545048	Modulating HIV-1 replication by RNA interference directed against human transcription elongation factor SPT5	Background Several cellular positive and negative elongation factors are involved in regulating RNA polymerase II processivity during transcription elongation in human cells. In recruiting several of these regulatory factors to the 5' long terminal repeat (LTR) promoter during transcription elongation, HIV-1 modulates replication of its genome in a process mediated by the virus-encoded transactivator Tat. One particular cellular regulatory factor, DSIF subunit human SPT5 (hSpt5), has been implicated in both positively and negatively regulating transcriptional elongation but its role in Tat transactivation in vivo and in HIV-1 replication has not been completely elucidated. Results To understand the in vivo function of hSpt5 and define its role in Tat transactivation and HIV-1 replication, we used RNA interference (RNAi) to specifically knockdown hSpt5 expression by degrading hSpt5 mRNA. Short-interfering RNA (siRNA) designed to target hSpt5 for RNAi successfully resulted in knockdown of both hSpt5 mRNA and protein levels, and did not significantly affect cell viability. In contrast to hSpt5 knockdown, siRNA-mediated silencing of human mRNA capping enzyme, a functionally important hSpt5-interacting cellular protein, was lethal and showed a significant increase in cell death over the course of the knockdown experiment. In addition, hSpt5 knockdown led to significant decreases in Tat transactivation and inhibited HIV-1 replication, indicating that hSpt5 was required for mediating Tat transactivation and HIV-1 replication. Conclusions The findings presented here showed that hSpt5 is a bona fide positive regulator of Tat transactivation and HIV-1 replication in vivo . These results also suggest that hSpt5 function in transcription regulation and mRNA capping is essential for a subset of cellular and viral genes and may not be required for global gene expression.	Background The elongation phase of transcription is often a critical juncture for regulating gene expression [ 1 , 2 ] and a number of genes including c-myc, c-fms, hsp70, and those encoded by HIV-1 are regulated at this stage of transcription [ 3 - 6 ]. During transcription elongation, shortly after successful initiation of RNA synthesis, RNA polymerase II (RNA pol II) can pause, arrest, pass through terminator sequences, or terminate transcription. The varying processivity of RNA pol II prior to entering productive elongation is controlled by the action of both negative and positive transcription elongation factors (N-TEFs and P-TEFs, respectively). The function of P-TEFs is to reduce the barrier of N-TEFs and promote the release of RNA pol II from the transition state that can cause termination of transcription [ 7 ]. Three elongation regulatory factors, P-TEFb ( p ositive t ranscription e longation f actor b ), DSIF ( D RB (5,6- d ichloro-1-β-D- r ibofuranosyl b enzimidazole) s ensitivity- i nducing f actor) and NELF ( n egative e longation f actor), have been identified using DRB as a transcription inhibitor [ 8 - 10 ] and function together to regulate transcription elongation. Modulation of HIV-1 gene expression provides one fundamental example of how transcription elongation can be controlled by such regulatory factors [ 11 - 14 ]. Tat, an HIV-1 regulatory protein, is required for synthesis of viral mRNA and increases the efficiency of transcription elongation from the HIV-1 promoter. In the presence of Tat, the processivity of RNA Pol II complexes that initiate transcription in the HIV-1 5' long terminal repeat (5' LTR) region becomes greatly enhanced. For this increased processivity to occur, Tat binds with a nascent leader RNA element, trans -activation responsive (TAR) RNA, located at the 5' end of all HIV-1 transcripts [ 15 ]. Cellular factors in association with Tat and TAR are then recruited to the 5' LTR, stimulating RNA pol II processivity during elongation. More specifically, the C-terminal domain (CTD) of RNA pol II is proposed to be hyperphosphorylated by P-TEFb during Tat transactivation to promote elongation [ 12 - 14 ]. Composed of cyclin-dependent kinase CDK9 and Cyclin T1, P-TEFb has been shown to bind the activation domain of Tat and TAR RNA loop sequence and phosphorylate the CTD of RNA pol II [ 16 - 18 ]. Tat transactivation is postulated to involve Tat-TAR interactions that then give rise to the recruitment of P-TEFb to RNA pol II complexes at the 5' LTR. This recruitment is necessary to enhance the processivity of RNA Pol II from the HIV-1 5' LTR promoter [ 7 , 14 , 17 , 19 ]. Thus, TAR RNA provides a scaffold for Tat and P-TEFb to bind and assemble a regulatory switch during HIV replication [ 20 ]. Human DSIF consists of subunits hSpt5 and hSPT4 and was originally discovered as a negative elongation factor that binds to RNA pol II [ 9 ]. In conjunction with NELF, DSIF represses transcriptional elongation at positions proximal to promoters [ 9 , 10 ]. Escape from transcriptional repression imposed by DSIF and NELF requires P-TEFb, which has been shown in vitro to phosphorylate both hSpt5 and CTD [ 7 , 10 , 21 - 29 ]. Interestingly, hSpt5 is conserved among eukaryotes and is a dual transcriptional regulator that can function as both a negative and positive elongation factor [ 30 - 32 ]. Currently, it is postulated that phosphorylation of hSpt5 and RNA pol II by P-TEFb is the key event during which hSpt5 functionally switches from a negative barrier to a positive elongation factor during transcription in human cells. Methylation of SPT5 also has been shown to regulate its interaction with RNA pol II and this posttranslational modification of SPT5 may alter transcriptional elongation functions in response to viral and cellular factors [ 33 ]. Although hSpt5's role in transcription regulation in association with P-TEFb has been established, its involvement in Tat transactivation and HIV-1 replication continues to be elucidated. Several in vitro studies have shown that hSpt5 is required for Tat transactivation and that both hSpt5 and RNA pol II phosphorylation is stimulated after recruitment of P-TEFb by Tat [ 25 , 29 , 34 ]. hSpt5 may also play a positive role in Tat transactivation through its association with human mRNA capping enzyme (HCE), which is a bifunctional triphosphatase-guanylyltransferase required for capping mRNA (reviewed in [ 1 , 35 ]), since SPT5, Tat, and CTD associate with the capping apparatus to stimulate capping [ 36 - 43 ]. However, studies in a recent report have suggested that only P-TEFb hyperphosphorylation of the RNA pol II CTD is directly required for Tat transactivation, precluding a direct role for hSpt5 in RNA pol II processivity during HIV-1 replication [ 26 ]. Therefore, hSpt5 role in Tat transactivation and HIV-1 replication in vivo remains unclear. Here, we used RNA interference (RNAi) to address whether hSpt5 is required for Tat transactivation and thus HIV-1 replication in vivo and further defined hSpt5 cellular functions. RNAi is a remarkably efficient process whereby double-stranded RNA (dsRNA) induces sequence-specific degradation of homologous mRNA in animal and plant cells (reviewed in ref. [ 44 ]). In mammalian cells, RNAi can be triggered by 21-nucleotide (nt) small interfering RNA (siRNA) duplexes and a few dsRNA molecules are sufficient to inactivate a continuously transcribed target mRNA for an observable period of time [ 45 , 46 ]. RNAi has recently been used to successfully knockdown the expression of a number of HIV genes, including p24, reverse transcriptase, vif, nef, tat, and rev, and has led to pre- and post-integrative HIV-1 RNA degradation and reduced HIV infectivity [ 47 - 52 ]. These results suggested that targeting viral factors required for the HIV life cycle with siRNAs including those required for HIV replication is a viable method for treating HIV infections. Other groups have targeted cellular factors implicated in supporting the HIV life cycle, including T-cell co-receptors CD4, CXCR4, CCR5, and CD8 [ 50 , 52 - 54 ] and transcription factor NF-κB [ 51 ], which has a role in HIV transcription initiation. Knockdown of the co-receptors reduced HIV infectivity, effectively blocking HIV entry into cells [ 55 ]. RNAi has become one of the leading methodologies for studying gene knockdown in human cells. During RNAi, a double-stranded 21-nucleotide (nt) short-interfering RNA (siRNA) targets a specific, complementary mRNA for degradation, resulting in significantly decreased expression, or knockdown, of the targeted gene (reviewed in [ 56 , 57 ]). In this report, siRNA designed to target hSpt5 successfully silenced hSpt5 as observed by decreased hSpt5 mRNA and protein expression. In addition, RNAi directed against hSpt5 did not significantly affect cell viability. hSpt5 knockdown led to significant decreases in Tat transactivation and inhibited HIV replication, indicating that hSpt5 was required for Tat transactivation and HIV replication in vivo . Taken together, silencing of hSpt5 by RNAi firmly established that the regulation of HIV-1 gene expression requires both Tat-TAR-P-TEFb interactions and interactions between RNA pol II transcription complexes and hSpt5. Results Specific silencing of hSpt5 expression by siRNA in HeLa cells To inhibit hSpt5 expression in a cultured human cell line using RNAi, siRNA targeting an hSpt5 sequence from position 407 to 427 relative to the start codon was designed (Figure 1A ). Magi cells were transfected with this hSpt5 duplex siRNA using Lipofectamine (Invitrogen). To evaluate the effects of hSpt5 RNAi, total cell lysates were prepared from siRNA-treated cells harvested at various time points after transfection. hSpt5 mRNA or protein levels were analyzed by RT-PCR or western blot using anti-hSpt5 antibodies, respectively. Cells transfected with hSpt5 siRNA had significantly lowered hSpt5 mRNA (Figure 1B , lane 3) and protein expression (Figure 1C , lane 3), indicating that siRNA-mediated silencing of hSpt5 had occurred successfully. hSpt5 knockdown was consistently between ~85–90%. This knockdown effect was dependent on the presence of the 21-nt siRNA duplex harboring a sequence complementary to the mRNA target. As shown in Figures 1B and 1C , mock-treated (no siRNA) (lane 1), single-stranded antisense hSpt5 siRNA (lane 2), or mismatched hSpt5 duplex siRNA (lane 4) containing two nucleotide mismatches between the target mRNA and siRNA antisense strand at the putative cleavage site of the target mRNA (Figure 1A ) did not affect hSpt5 mRNA or protein levels. These results showed that hSpt5 knockdown was specific to duplex siRNA exactly complementary to the hSpt5 mRNA target. In evaluating either mRNA or protein levels, human Cyclin T1 (hCycT1) was used as an internal control, showing that the effects of hSpt5 siRNA were specific to hSpt5 and did not effect hCycT1 mRNA or protein expression (Figure 1B and 1C , lower panel). Taken together, these results demonstrated that hSpt5 knockdown was sequence specific and led to significantly decreased hSpt5 mRNA and protein levels. Figure 1 Specific silencing of hSpt5 expression by RNAi . (A) hSpt5 mRNA is 3261 nucleotides in length. siRNA targeting sequence for hSpt5 was selected from position 407 to 427 relative to the start codon. As a specific control, mutant siRNA containing 2 nucleotide mismatches (underline) between the target mRNA and the antisense of siRNA at the hypothetical cleavage sites of the mRNA was generated. (B) Evaluation of specific hSpt5 siRNA activity by RT-PCR. Total cellular mRNA was prepared from HeLa cells transfected without siRNA or with hSpt5 duplex or control siRNAs and was followed by RT-PCR, as described in Material and Methods. Each RT-PCR reaction included 100 ng total cellular mRNA, gene-specific primer sets for hSpt5 and hCycT1 amplification (0.5 μM for each primer), 200 μM dNTP, 1.2 mM MgSO 4 and 1U of RT/platinum Taq mix. Primer sets for hSpt5 produced 2.6 kb products while hCycT1 produced 1.8 kb products. RT-PCR products were resolved on a 1% agarose gel and viewed by ethidium bromide staining. RT-PCR products are shown from cells that were not transfected with siRNA (lane 1), or cells transfected with single-stranded antisense hSpt5 siRNA (hSpt5 (AS), lane 2), hSpt5 duplex siRNA (hSpt5 (DS), lane 3), or mismatch hSpt5 duplex siRNA (hSpt5-mm (DS), lane 4). Lane M is a marker lane. (C) Analysis of specific hSpt5 siRNA activity by western blotting. Cell lysates were prepared from HeLa cells mock-transfected without siRNA (lane 1), or transfected with single-stranded antisense hSpt5 siRNA (hSpt5 (AS), lane 2), hSpt5 duplex siRNA (hSpt5 (DS), lane 3), or mismatch hSpt5 duplex siRNA (hSpt5-mm (DS), lane 4). Cell lysates were analyzed by 10% SDS-PAGE. Protein contents were detected by immunoblotting assay with antibodies against hSpt5 (top panel) and hCycT1 (lower panel). Kinetics of hSpt5 knockdown by RNAi Having established that hSpt5 could be knocked down using RNAi, the kinetics of hSpt5 knockdown were examined. To perform kinetic experiments, hSpt5 duplex siRNA, single-stranded antisense hSpt5 siRNA, or mismatch duplex hSpt5 siRNA were transfected into Magi cells. Cell lysates were collected at various time points to assay for protein levels during hSpt5 knockdown. Immunoblot analysis using anti-hSpt5 antibodies revealed the timing of gene suppression and persistence of hSpt5 RNAi effects in Magi cells during the time course experiment (Figure 2 ). hSpt5 knockdown was first observed between 30–42 h post-transfection, with maximum knockdown (~85–90% knockdown) occurring at 42–66 h post transfection (Figure 2 , lane 8–14). Protein levels gradually recovered to normal levels between 66–90 h (data not shown), indicating that the effects of hSpt5 siRNA did not last indefinitely. Neither single-stranded antisense siRNA (Figure 2 , lanes 1–7) nor mismatched duplex siRNA (Figure 2 , lanes 15–21) affected hSpt5 protein levels throughout the duration of the time course. These results indicated that hSpt5 knockdown by RNAi occurred after 30 h and these knockdown effects were specific to duplex siRNA targeting hSpt5. Figure 2 Kinetics of specific hSpt5 siRNA activity by Western blotting . HeLa cells were transfected with single-stranded antisense hSpt5 siRNA (hSpt5 (AS), lanes 1–7), hSpt5 duplex siRNA (hSpt5 (DS), lanes 8–14), or mismatch hSpt5 duplex siRNA (hSpt5-mm (DS), lanes 15–21) having 2 nucleotide mismatches between the target mRNA and the antisense strand of siRNA at the hypothetical cleavage site of the mRNA. Cells were harvested at various times post transfection. Protein content was resolved on 10% SDS-PAGE, transferred onto PVDF membranes, and immunoblotted with antibodies against hSpt5 (top bands) and hCycT1 as an internal control (lower bands). Knockdown of hSpt5 is not lethal to human cells Knowing that the kinetics of hSpt5 peaked at 42–54 h post-transfection, we were able to evaluate the viability of cells during hSpt5 knockdown experiments over varied time intervals. Cell viability was assessed using trypan blue exclusion at various times after a single transfection of various siRNAs. As shown in Figure 3 , during the 66 h time course experiment, the number of non-viable hSpt5 knockdown cells (yellow line) observed was comparable to mock-treated cells (no siRNA; dark blue line). Cells transfected with single-stranded antisense hSpt5 siRNA (purple line) or mismatched hSpt5 duplex siRNA (light blue line) that did not show hSpt5 knockdown also showed minimal changes in cell viability. Figure 3 Analysis of cell viability by counting trypan blue-stained cells . HeLa cells were transfected with Lipofectamine with various siRNAs or no siRNA. Three siRNA duplexes, including hSpt5 siRNA (yellow), mismatch hSpt5 siRNA (light blue) and siRNA targeting human capping enzyme (HCE, red), were used in these experiments. Controls for viability included cells mock-transfected with no siRNA (dark blue) or cells transfected with single-stranded antisense hSpt5 siRNA (purple). At various times after transfection, cells floating in the medium were collected and counted in the presence of 0.2% trypan blue. Cells that took up dye (stained blue) were counted as not viable. hSpt5 has been shown to interact with the human mRNA capping enzyme (HCE) and this interaction enhances capping enzyme guanylylation and mRNA capping several fold [ 40 ]. Since Spt5, Tat, and CTD associate with the capping apparatus to stimulate capping [ 36 , 37 , 39 - 42 ], we planned to separately define the role of HCE and hSpt5 in Tat transactivation by using RNAi to specifically knockdown HCE expression. HCE knockdown was confirmed by RT-PCR (data not shown). In contrast to hSpt5 knockdown cells, HCE knockdown cells showed a significant increase in cell death (Figure 3 , red line) over the course of the knockdown experiment. These results indicated that HCE is an essential enzyme for cell viability and growth. Simialr findings showing that RNA capping was essential for metazoan viability have also been previously reported using RNAi in C. elegans [ 58 ]. These results indicated that hSpt5 knockdown was not lethal to human cells, while a much more stringent requirement for HCE expression was essential for cell viability. Role of hSpt5 on HIV-1 Tat Transactivation To examine whether hSpt5 was required for HIV-1 Tat transactivation in vivo , Tat transactivation during hSpt5 knockdown in Magi cells was monitored. Magi cells are a HeLa cell line harboring a stably integrated single copy of the HIV-1 5' LTR-β-galactosidase gene. These cells are also genetically programmed to express the CD4 receptor for HIV-1 infection ([ 59 ]; see below). In this experiment, Magi cells were co-transfected with Tat expression plasmid pTat-RFP and hSpt5 duplex siRNA. Co-transfection with Tat siRNA was used as a positive control for inhibition of Tat transactivation while single-stranded antisense hSpt5 siRNA and mismatched siRNA were used as negative controls. Tat transactivation and protein levels were evaluated by harvesting cells 48 h post transfection, which was within the timeframe that hSpt5 knockdown peaked. Expression of HIV-1 Tat-RFP under the control of the CMV early promoter was confirmed by western blot using anti-RFP antibody and by measuring RFP fluorescence using a fluorescence spectrophotometer (data not shown). In addition, immunoblot analysis confirmed that hSpt5 siRNA specifically inhibited hSpt5 protein expression in the absence and presence of HIV-1 Tat protein in Magi cells (data not shown). Tat-RFP enhances the expression of genes that are under the control of the HIV-1 5' LTR promoter. In this experiment, Tat transactivation was measured by assaying the β-galactosidase activity resulting from expression of the β-galactosidase gene under the HIV-1 5' LTR promoter. To quantify the effects of various siRNAs on HIV-1 Tat transactivation, the ratio between β-galactosidase activity in cells transfected with pTat-RFP (with or without siRNAs) and mock-treated cells not transfected with pTat-RFP was determined. The results of this quantitation are shown in Figure 4 . In Magi cells, Tat-RFP strongly stimulates the expression of β-galactosidase, represented by a 13-fold increase in Tat transactivation (Figure 4 , lane 1). On the other hand, Tat transactivation was strongly inhibited in cells transfected with Tat siRNA (~90% knockdown; Figure 4 , lane 5), as previously shown [ 51 ]. Tat transactivation was similarly inhibited when cells were transfected with hSpt5 duplex siRNA, exhibiting only ~30% of the Tat transactivation observed with Tat-RFP alone (Figure 4 , lane 3). Neither antisense hSpt5 siRNA nor mismatched hSpt5 siRNA (Figure 4 , lane 4) showed any effect on Tat transactivation. These results indicated hSpt5 knockdown caused by siRNA specifically targeting hSpt5 mRNA inhibited HIV-1 Tat transactivation in human cells. These results strongly supported an important role for hSpt5 in Tat transactivation in vivo and suggested that RNAi of hSpt5 had the potential to inhibit HIV-1 replication. Figure 4 Effect of hSpt5 siRNA on HIV-1 Tat transactivation in Magi cells . Quantified effect of siRNA on HIV-1 Tat transactivation was determined by measuring β-galactosidase activity. Magi cells were co-transfected with pTat-RFP plasmid and various siRNAs targeting hSpt5 or Tat and harvested at 48 h post-transfection. Activity of β-galactosidase was measured using the β-Galactosidase Enzyme Assay System (Promega). Tat transactivation was determined by the ratio of β-galactosidase activity in pTat-RFP transfected cells to activity measured in cells without pTat-RFP. The inhibitory effect of siRNA was determined by normalizing Tat transactivation activity to the amount of Tat-RFP protein. Tat transactivation was measured for Magi cells transfected with pTat-RFP only (lane 1), or Tat-RFP transfected with single-stranded antisense hSpt5 siRNA (hSpt5-AS, lane 2), hSpt5 duplex siRNA (hSpt5-DS, lane 3), mismatch hSpt5 duplex siRNA (hSpt5-mm-DS, lane 4), or Tat siRNA duplex (Tat-DS, lane 5). Results are representative of three independent experiments. hSpt5 knockdown inhibits HIV-1 replication To evaluate the effect of hSpt5 knockdown on HIV-1 replication, a double siRNA transfection protocol was used to maximize the knockdown efficiency of hSpt5 during HIV-1 infection. Magi cells were transfected with siRNA directed against hSpt5. Cells mock transfected without siRNA, or transfected with single-stranded antisense hSpt5 siRNA or mismatch hSpt5 siRNA were used as negative controls. Transfection with Vif or Nef siRNAs was used as a positive control [ 20 ]. 24 h after the first transfection, a second siRNA transfection identical to the first was performed. 24 h later, doubly transfected cells were infected with various concentrations of HIV NL-GFP , an infectious molecular clone of HIV-1. Knockdown of hSpt5 protein levels was then evaluated 48 h post infection in doubly transfected cells. An even larger decrease in hSpt5 protein levels was observed in doubly transfected cells (~95% knockdown) as compared to singly transfected cells (~85–90% knockdown; Supplementary Figure 1 , compare lanes 4 and 10), suggesting that more robust knockdown of gene expression can be achieved using this double transfection method. HIV-1 Tat-mediated transactivation of the 5' LTR occurring in cells infected with virus led to β-galactosidase production, which was also quantified 48 h post-infection. In this single-cycle replication assay for evaluating HIV-1 replication, β-gal activity reflected the activity of reverse transcriptase and viral replication of varying amounts of viral inoculum. Therefore, changes in β-gal activity could be directly correlated to changes in the efficacy of HIV-1 replication. The positive siRNA control targeting HIV-1 Vif showed decreased levels of β-gal activity and HIV-1 replication, as shown previously (Figure 5 ; [ 47 ]). Double-stranded siRNA directed against hSpt5 showed a similar decrease in β-gal activity when compared with Vif knockdown. This observed decrease was equivalent to the β-gal activity measured when using 32 times less viral inoculum with mock-treated cells (Figure 5 ), indicating that hSpt5 knockdown had significantly reduced HIV-1 replication. p24 levels were also monitored during these experiments and decreased in the context of hSpt5 knockdown (data not shown), supporting the conclusion that hSpt5 knockdown has a negative effect on the HIV-1 life cycle. Control experiments using hSpt5 single-stranded antisense or mismatched duplex siRNA duplexes showed no antiviral activities. In addition, no significant toxicity or cell death was observed during these experiments, suggesting that hSpt5 knockdown was not lethal even in the context of HIV-1 infection. These results demonstrated that hSpt5 silencing using RNAi modulated HIV-1 replication and firmly established an important role for hSpt5 in Tat transactivation and HIV-1 replication in vivo . Figure 5 siRNA targeting hSpt5 modulate HIV-1 replication . HeLa-CD4-LTR/β-galactosidase (Magi) cells were mock-transfected (mock), or transfected with single-stranded antisense hSpt5 siRNA (AS), hSpt5 duplex siRNA (siRNA), mismatched hSpt5 duplex siRNA (MM) or Vif duplex siRNA (T98). 24 h after the first transfection, a second siRNA transfection was performed. 24 h later, cells were infected with HIV NL-GFP , an infectious molecular clone of HIV-1. Cells infected with virus and not treated with oligofectamine are shown (mock). HIV-1 Tat-mediated transactivation of the 5' LTR led to β-galactosidase production, which was quantified 48 h post-infection. Cells treated with duplex siRNA targeting Vif (lanes marked T98 [47]) served as a positive control. Serial double dilutions of the viral inoculum (in cpm of RT activity) are consistent with 32-fold decreases in viral replication. Discussion hSpt5, as part of the DSIF complex, was originally discovered as a negative elongation factor required for conferring DRB sensitivity to transcription elongation complexes thereby inhibiting transcription [ 9 ]. This negative barrier provided by hSpt5 was thought to be relieved through P-TEFb phosphorylation of both hSpt5 and RNA pol II CTD, which results in increased processivity of RNA pol II complexes [ 7 , 10 , 21 - 29 ]. Increased processivity has also been linked to the phosphorylated form of hSpt5 conferring a positive effect on transcription elongation [ 25 , 29 , 34 ]. Recently, however, it has been shown that Tat is able to enhance transcription elongation in vitro in the absence of hSpt5 [ 26 ]. These results appeared to indicate that P-TEFb phosphorylation of RNA pol II was the sole event that directly led to Tat transactivation and increased RNA pol II processivity [ 26 ]. Thus, from the results of all of these in vitro studies collectively, the requirement for hSpt5 in positively regulating transcription elongation during Tat transactivation has remained unclear. Here, we studied the role of hSpt5 in vivo using RNAi and established that hSpt5 played a positive role in Tat transactivation and HIV-1 replication. Knockdown of hSpt5 provided insight into several functional aspects of the hSpt5 protein. First, knockdown of hSpt5 was not lethal in Magi cells, indicating that hSpt5 was not required for cell viability. This was an interesting result because studies of SPT5 mutants in yeast and zebrafish and RNAi of SPT5 in C. elegans have shown that SPT5 was essential for growth and/or embryonic development in those organisms [ 30 , 31 , 60 ]. It seems likely that hSpt5 holds similar essential functions in human cells during embryonic development but may not be absolutely required in adult cells. Alternatively, hSpt5 knockdown may have led to decreased levels of expression that were still sufficient for hSpt5 to carry out its essential functions. Our results support the notion of using RNAi against hSpt5 as a potential therapeutic strategy for fighting HIV-1 infection since there is the potential that HIV-1 functions could be targeted for inhibition without significantly interfering with host cell functions. The key finding of this study was that hSpt5 knockdown significantly inhibited both Tat transactivation and HIV-1 replication. These results indicated that hSpt5 was a bona fide regulator of Tat transactivation that is required for HIV-1 replication in vivo . Our in vivo results strongly support previous in vitro results recapitulating Tat transactivation that showed immunodepletion of hSpt5 significantly inhibited Tat transactivation [ 29 , 34 ]. However, it is difficult to reconcile our in vivo results with recently published in vitro experiments showing that P-TEFb hyperphosphorylation of the CTD in the absence of hSpt5 still enhanced RNA pol II processivity during Tat transactivation [ 26 ]. In reconciling whether P-TEFb hyperphosphorylation was directly required for Tat transactivation to the exclusion of hSpt5, we would like to propose that the required function of P-TEFb hyperphosphorylation may be distinct from the role hSpt5 plays in enhancing RNA pol II processivity during Tat transactivation. In our model (Figure 6 ), P-TEFb hyperphosphorylation would occur first, triggering enhanced processivity of RNA pol II. hSpt5 presumably is phosphorylated at around the same time as RNA pol II, stimulating hSpt5 to switch from a negative regulator to a positive elongation factor [ 25 ]. Phosphorylated hSpt5 may then be important for positively regulating an initial step in Tat transactivation. Figure 6 Model for Tat transactivation in absence or presence of SPT5 . See text for details. Conceivably, hSpt5 functions in transcription elongation as a stabilization factor that enhances the stability of RNA pol II elongation complexes formed after P-TEFb hyperphosphorylation of the CTD. This type of role would also support hSpt5 function as an antiterminator factor as described previously [ 61 ]. Another important positive function for hSpt5 during Tat transactivation may involve hSpt5 and Tat interactions with the capping machinery [ 40 - 42 ]. Phosphorylation of hSpt5 by P-TEFb may stabilize hSpt5 interactions with HCE thereby stabilizing Tat and CTD interactions with the capping machinery to promote capping and successful production of stable HIV transcripts (see model in Figure 6 ). Due to the highly structured nature of TAR, capping of the 5' end of HIV transcripts is not very efficient in the absence of Tat [ 41 , 42 ] and Tat stimulated capping may require the presence of hSpt5 for greater access to the 5' end or to stabilize and kinetically arrest the elongation complex. Capping of HIV transcripts has also been shown to occur more proficiently when elongation is paused and not continuous [ 42 ], suggesting that DSIF/NELF-dependent pausing of early stage elongation complexes is representative of an elongation checkpoint. One function of this checkpoint may be to allow time for the recruitment of capping machinery and subsequent capping of HIV RNA to stabilize nascent transcripts prior to further elongation. In the absence of hSpt5, pausing may no longer occur during elongation since neither NELF nor hSPT4 binds to RNA pol II without hSpt5 [ 62 , 63 ]. Thus, the window for the capping apparatus to be recruited by Tat and/or stimulate capping may be severely shortened or lost altogether without hSpt5. Any resulting uncapped HIV transcripts would be prone to degradation, accounting for the lower level of Tat transactivation and HIV replication observed during hSpt5 knockdown. The potential roles for phosphorylated hSpt5 in stabilizing RNA pol II processive elongation complexes or with respect to capping during Tat transactivation are not mutually exclusive as shown in Figure 6 . hSpt5 may indeed have multi-functional roles as a positive regulator during HIV-1 replication. Conclusions The in vitro and in vivo approaches taken to address the importance of hSpt5 function all shed light on the multi-faceted nature of Tat transactivation. Accordingly, these studies altogether support important roles for both P-TEFb and hSpt5 in mediating transcription elongation during HIV-1 replication in vivo . The dual function of hSpt5 as a negative and positive transcription elongation factor also demonstrates the complexity associated with transcriptional regulation during transcription elongation and HIV-1 Tat transactivation. It is likely that posttranslational modifications of hSpt5 dictate functions of Spt5 at various promoters. Further studies will be required to elucidate how various modifications of hSpt5 such as phosphorylation and methylation control transcription elongation of both cellular and viral genes. Methods siRNA preparation Twenty-one nucleotide siRNAs were chemically synthesized as 2' bis(acetoxyethoxy)-methyl ether-protected oligos (Dharmacon, Lafayette, CO). Synthetic RNAs were deprotected, annealed and purified using standard protocols provided by the manufacturer. Formation of duplex RNA was confirmed by 20% non-denaturing polyacrylamide gel electrophoresis (PAGE). Sequences of siRNA duplexes were designed as described previously [ 46 ] and subjected to a BLAST search against the NCBI EST library to ensure that only the desired genes were targeted. Culture and transfection of cells Magi ( m ultinucleate a ctivation of g alactosidase i ndicator) cells carrying the endogenous HIV-5'LTR β-galactosidase gene were maintained at 37°C in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) supplemented with 10% fetal bovine serum (FBS), 0.2 mg/ml of Geneticin (G418) and 0.1 mg/ml Hygromycin B (Roche Molecular Biochemicals). Cells were passaged at sub-confluence and plated at 70% confluency for transfection. Reporter plasmids and siRNA were co-transfected into Magi cells using Lipofectamine (Invitrogen) in duplicate 6-well plates (Falcon). A transfection mixture containing 25–150 nM siRNA and 9 μl of Lipofectamine in 1 ml of serum-reduced OPTI-MEM (Invitrogen) was added to each well. For high efficiency knockdown experiments, 150 nM siRNA was used. After incubating at 37°C for 6 h, cells were cultured in antibiotic-free DMEM. For further analysis, transfected cells were harvested at various time intervals, washed twice with phosphate buffered saline (PBS, Invitrogen), flash frozen in liquid nitrogen, and stored at -80°C. RT-PCR for amplification of hSpt5 and hCycT1 mRNA Total cellular mRNA was prepared from HeLa cells transfected without siRNA or with hSpt5 or control siRNAs using a Qiagen RNA mini kit, followed by an oligotex mRNA mini kit (Qiagen). RT-PCR was performed using a SuperScript One-Step RT-PCR kit with platinum Taq (Invitrogen) and 40 cycles of amplification. Each RT-PCR reaction included 100 ng total cellular mRNA, gene-specific primer sets for hSpt5 and hCycT1 amplification (0.5 μM for each primer), 200 μM dNTP, 1.2 mM MgSO 4 and 1U of RT/platinum Taq mix. Primer sets for hSpt5 produced 2.6 kb products while hCycT1 produced 1.8 kb products. RT-PCR products were resolved on a 1% agarose gel and viewed by ethidium bromide staining. Forward and reverse primer sequence for amplifying SPT5 were 5'-ATGTCGGACAGCGAGGACAGC-3' (nts 1–21) and 5'-TGTACATGGCCGGCGTCCC-3' (nts 2638–2656), respectively. Forward and reverse primer sequences for amplifying hCycT1 are 5'-GCAACAAGTTCAAGATCTGGTCAT-3' (nts 381–404) and 5'- CCCGGGGGATCC TTACTTAGGAAGGGGTGGAAGTGG-3' (nts 2158–2200); underlined sequences represent restriction enzyme sites), respectively. Western blotting siRNA treated cells were harvested as described above and lysed in 1X reporter lysis buffer (Promega). After centrifugation to remove cellular debris, concentrations of proteins in lysates were determined using a Dc protein assay kit (Bio-Rad). Proteins in 30 μg of total cell lysates were fractionated by 10% SDS-PAGE, transferred onto a polyvinylidene difluoride membrane (PVDF membrane, Bio-Rad), and immunoblotted with antibodies against hSpt5 (Pharmingen) and hCycT1 (Santa Cruz Biotech). Protein content was visualized by a BM chemiluminescence Blotting Kit (Roche Molecular Biochemicals). The blots were exposed to X-ray film (Kodak MR-1) for various times (between 1 s and 5 min). β-galactosidase enzyme assay Magi cells were harvested 48 h after transfection with Tat-RFP plasmids in the absence or presence of siRNAs. Cell lysates were prepared and quantified as described above. To perform standard β-galactosidase assays, 120 μg of cell lysates were mixed in 150 μl of 1X reporter lysis buffer and 150 μl of 2X β-galactosidase assay buffer (Promega), and incubated at 37°C for 30 min. To stop the reaction, 500 μl of 1 M sodium carbonate was added to the mixture and mixed well by vortexing briefly. Absorbance of the reaction mixture was read immediately at 420 nm. The amount of Tat-RFP protein was determined using a fluorescence spectrophotometer (Photon Technology International). 300 μg of cell lysates was subjected to the spectrophotometer with slit widths set at 4 nm for both excitation and emission wavelengths as described previously [ 46 , 64 ]. Fluorescence of Tat-RFP in the cell lysate was detected by exciting at 558 nm and recording the emission spectrum from 568 nm to 650 nm; the spectrum peak at 583 nm represents the maximum fluorescence intensity of Tat-RFP. Tat transactivation was determined by calculating the ratio of β-galactosidase activity (absorbance at 420 nm) of the pTat-RFP transfected cells to that of cells without pTat-RFP plasmid treatment. The inhibitory effect of siRNA treatment was determined by normalizing Tat-transactivation activity to the amount of Tat-RFP protein (represented by RFP fluorescence intensity) in the presence and absence of siRNA. Magi infectivity assay HeLa-CD4-LTR/β-gal indicator (Magi) cells [ 59 ] were plated in 24-well plates (7.5 × 10 5 cells per well) and transfected with siRNAs as previously described [ 47 ]. siRNA (60 pmol) was transfected into cells using oligofectamine (2 μl, Invitrogen) for 3 h in serum-free DMEM (GIBCO). Cells were rinsed twice and top-layered in 500 μl of DMEM-10% FBS. 24 h after the first transfection, a second identical siRNA transfection was performed. 24 h after the second transfection, cells were trypsinized and seeded in 96-well microtiter plates (4 × 10 4 cells per well), incubated 3 h and infected with HIV NL-GFP , an infectious molecular clone of HIV-1. HIV-1 virions (normalized to RT activity in cpm) were added in doubling dilutions to duplicate wells. 48 h post infection, cells were harvested to quantify β-galactosidase activity and protein levels. Competing Interests The author(s) declare that they have no competing interests. Authors' contribution Y-HP carried out all the Spt5 silencing experiments, C-yC performed capping enzyme knockdown experiments, HC performed quantitative analysis of the data, J-MJ performed HIV-1 replication assays, and MS analyzed and interpreted HIV inhibition results. TMR conceived the ideas, and participated in the experimental design and in drafting the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545048.xml
449896	Honeybee Odometry: Performance in Varying Natural Terrain	Recent studies have shown that honeybees flying through short, narrow tunnels with visually textured walls perform waggle dances that indicate a much greater flight distance than that actually flown. These studies suggest that the bee's “odometer” is driven by the optic flow (image motion) that is experienced during flight. One might therefore expect that, when bees fly to a food source through a varying outdoor landscape, their waggle dances would depend upon the nature of the terrain experienced en route. We trained honeybees to visit feeders positioned along two routes, each 580 m long. One route was exclusively over land. The other was initially over land, then over water and, finally, again over land. Flight over water resulted in a significantly flatter slope of the waggle-duration versus distance regression, compared to flight over land. The mean visual contrast of the scenes was significantly greater over land than over water. The results reveal that, in outdoor flight, the honeybee's odometer does not run at a constant rate; rather, the rate depends upon the properties of the terrain. The bee's perception of distance flown is therefore not absolute, but scene-dependent. These findings raise important and interesting questions about how these animals navigate reliably.	Introduction When a scout honeybee discovers an attractive patch of flowers, she returns to the hive and performs the famous “waggle dance” to advertise the location of the food source to her nestmates ( von Frisch 1993 ). The dance consists of a series of alternating left-hand and right-hand loops, interspersed by a segment in which the bee waggles her abdomen from side to side. The duration of this “waggle phase” conveys to the potential recruits the distance of the food source from the hive: the longer the duration of the waggle, the greater the distance ( von Frisch 1993 ). This information is used by the recruited bees to locate the food source. Clearly, then, the scout as well as the recruits are able to gauge how far they have flown in search of food. Early studies concluded that bees estimate distance flown by gauging the amount of energy they expend to reach the destination ( von Frisch 1993 ; Neese 1988 ). More recent studies, however, are providing increasing evidence that this “energy hypothesis” is incorrect, at least for moderate distances of a few hundred meters ( Esch et al. 1994 , 2001 ; Esch and Burns 1995 ; Srinivasan et al. 1996 , 1997 , 2000 ). Over these distances, bees appear to gauge distance flown by measuring how much the image of the world appears to move in the eye en route to the food source ( Esch and Burns 1995 , 1996 ; Srinivasan et al. 1996 , 1997 , 2000 ; Esch et al. 2001 ). There are two kinds of experimental evidence that support this “optic flow” hypothesis. First, bees that fly a given distance close to the ground signal a much larger distance in their dances than do bees that fly the same distance at a considerable height above the ground and therefore experience a smaller rate of image motion ( Esch and Burns 1996 ). Second, bees trained to fly to a feeder placed inside a short, narrow tunnel, the walls and floor of which are lined with a random visual texture, indicate a hugely exaggerated flight distance in their waggle dances ( Srinivasan et al. 2000 ; Esch et al. 2001 ). Evidently, the proximity of the walls and floor of the tunnel greatly amplifies the magnitude of the optic flow that the bees experience, in comparison with the situation during outdoor flight in a natural environment. On the other hand, when the same tunnel was lined with axial stripes—so that a bee flying through it would experience very little optic flow because the stripes were parallel to the flight direction—the bees signaled a very small distance, even though they had flown the same physical distance as in the previous condition ( Srinivasan et al. 2000 ). This experiment indicated that distance flown was being measured in terms of integrated optic flow, and not in terms of physical distance flown or energy consumed. If bees do indeed gauge distance traveled by measuring optic flow and integrating it over time, it is pertinent to enquire into the properties of their visually driven “odometer.” Given that the environment through which a bee flies can vary substantially in terms of its visual properties, such as color, contrast, texture, and the distribution of objects, it is important to know whether, and to what extent, the bee's perception of distance flown is affected by these environmental variables. In other words, how “robust” is the honeybee's odometer? In a recent study, this question was explored by training bees to fly into a short, narrow tunnel (as described above), and analyzing the waggle dances of the returning bees as the texture and the contrast of the patterns lining the walls and the floor were systematically varied ( Si et al. 2003 ). The patterns used were black-and-white stripes and sinusoidal gratings of various spatial frequencies and contrasts. This study revealed that the honeybee's odometer is indeed rather robust to changes in the visual environment. For a flight of a given distance into the tunnel, the odometric signal is relatively unaffected by changes in the spatial frequency or contrast of the gratings, as long as the contrast of the grating is above 20%. (Contrast is defined here as 100 (I max −I min )/(I max +I min ), where I max and I min are the intensities of the bright and the dark bars of the grating, respectively.) At contrasts below 20%, the strength of the odometric signal starts to decrease. Another recent study revealed that the visual odometer of the honeybee is “color-blind” and driven exclusively by the green receptor ( Chittka and Tautz 2003 ). Here we examine the robustness of the honeybee's odometer when it is performing in natural conditions, during flight through varying landscapes. Specifically, we compare the strength of the odometric signal during flight over land, where contrasts tend to be relatively high and textures are relatively rich, to that during flight over water, where contrast tends to be low and texture is sparse. The experiments described here involve training bees to fly over stretches of water or land, and comparing their waggle dances. These experiments also provide us with the opportunity to address two further questions that are somewhat controversial and not yet completely resolved. One question relates to whether bees can, and do, fly safely over large bodies of water, such as lakes. Heran and Lindauer (1963) , for instance, suggested that bees experience difficulty in flying across lakes, often losing altitude and plunging into the water. The other question relates to whether bees, having discovered an attractive food source positioned in the middle of a large expanse of water, can successfully recruit their nestmates, through their dances, to visit the food source ( Gould and Gould 1988 ). Results Experiment 1 In this experiment the feeder route was initially over land (segment 1), then over water (segment 2), and finally again over land (segment 3) ( Figure 1 ). As the feeder was moved away from the hive in stepwise fashion, most of the marked bees followed. Virtually all of the marked bees continued to visit the feeder even when it was over water. Thus, the trained bees had no difficulty in flying over the water and finding the feeder in the boat. Figure 1 Aerial Photograph of Experimental Site In experiment 1 the bees were trained to fly due southwest from the hive initially over a stretch of land, then over water, and finally across the island. In experiment 2 the bees were trained to fly due northwest along a route that was entirely over land. The white dots depict successive locations of the feeding station. The red dots along the route of experiment 1 depict the shoreline stations, which represent the boundaries between segments 1 and 2, and between segments 2 and 3. The red dots along the route of experiment 2 represent the same distances from the hive as their counterparts in experiment 1. The situation with unmarked honeybees recruited to the feeder by the marked scouts was rather different. A few recruits were observed at the feeder as long as the feeder was on land. Once the feeder was on the water, however, very few recruits were observed. Recruits reappeared when the boat reached the island. In fact, they were observed searching at the shore of the island shortly before the boat reached that point. Recruits were present at all feeder positions on the island. For each feeder position along the route, the dances of the returning bees were filmed and analyzed as described in Materials and Methods . An example of a dance is given in Video S1 . The variation of the mean waggle duration with feeder position is shown in Figure 2 A. Waggle duration increased with feeder distance. Figure 2 Variation of Waggle Duration with Feeder Distance in Experiment 1 In this experiment, flight was over land (segment 1), water (segment 2), and again over land (segment 3) (see Figure 1 ). Symbols depict mean waggle durations. Bars represent standard error of the mean. R, correlation coefficient. (A) shows a global linear approximation of the data, using a single regression for the entire data set (solid line). Broken curves depict 95% CIs for regression slope. (B) shows a piecewise linear approximation of the data, using separate regressions for the data over land, water, and the island. Equations represent regression lines. Figure 2 A also shows the best-fitting straight line (a linear regression) through the data. The slope of this regression line is 1.303 ± 0.150, where the limits denote the 95% confidence interval (CI). The root mean square (rms) error between the data and this linear approximation is 42.3 msec. Is this a good fit to the data? Visual inspection of the data in Figure 2 suggests that the rate at which the waggle duration increases with feeder distance (the slope of the curve) is higher in the first and third segments—where the bees fly over land—than in the second segment, where they fly over water. This impression is confirmed when linear regressions of the data are carried out separately for each of the three segments, as shown in Figure 2 B. The results then reveal that the slope of the regression in segment 2 (water) is 0.546 ± 0.483, which is significantly lower than the slope in segment 1 (1.970 ± 0.348) and segment 3 (2.243 ± 0.500) ( p < 0.02, pairwise comparison, Student t -test). On the other hand, there is no significant difference between the slopes for the land (segment 1) and the island (segment 3) segments ( p > 0.2, Student t -test). Within each segment, the data fit a straight line quite well; the rms error between all of the data points and the piecewise linear approximation is 14.7 msec. This value is considerably lower than the rms error produced by the single straight line in Figure 2 A (42.3 msec). Thus, the data are not well represented by a single linear approximation. They are better approximated by a piecewise linear relationship in which the slope over water is lower than that over land. Experiment 2 In experiment 2 the feeder route was of the same total length as in experiment 1, but was entirely over land (see Figure 1 ). In this experiment, unmarked recruits were observed at each feeder position. The variation of the mean waggle duration with feeder position for this experiment is shown in Figure 3 A. Here, again, waggle duration increased with feeder distance. The best-fitting straight line through the entire data set is shown in Figure 3 A; the slope of the resulting regression line is 1.431 ± 0.695. The rms error between this line and the data is 37.9 msec. In addition, to enable a comparison of the data from experiment 2 with those from experiment 1, the same data set was artificially divided into three segments, of lengths 190 m, 240 m, and 150 m, as it was for experiment 1 (see Figure 2 B), and a linear regression was performed separately on each segment. The results of this piecewise linear approximation are shown in Figure 3 B. In this case, the slopes of the regression lines were 1.712 ± 0.526 (segment 1), 1.925 ± 0.358 (segment 2), and 1.110 ± 1.500 (segment 3). There was no significant difference between any of these three slopes ( p > 0.10, pairwise comparison, Student t -test). The rms error between the piecewise linear approximation and the data was 41.6 msec, a value similar to that obtained with the single, best-fitting straight line (37.9 msec, Figure 3 A). Thus, in this case, a piecewise linear approximation does not improve the fit. This analysis indicates that the data of experiment 2, in which the bees flew exclusively over land, is well approximated by a single line of constant slope. The slope of this best-fitting line is 1.431 ± 0.695, which is not significantly different from the slopes calculated for either of the land segments in Figure 2 A ( p > 0.05, Student t -test). Figure 3 Variation of Waggle Duration with Feeder Distance in Experiment 2 For the purposes of comparative analysis, the data were artificially divided into segments corresponding to the land, water, and island segments of experiment 1 (see Figure 2 ). (A) shows a global linear approximation of the data, using a single regression for the entire data set. (B) shows a piecewise linear approximation of the data as they correspond to the segments of experiment 1. Other details are the same as in Figure 2 . It is worth noting, however, that, in experiment 2, the three data points in the middle of segment 3 ( Figure 3 A and 3 B) exhibit a much lower slope than the rest of the curve. A possible reason for this local variation will be discussed later. We may summarize the findings of experiments 1 and 2 by saying that the mean waggle duration increases more rapidly with distance flown when bees fly over land, than when they fly over water. The slope of the mean waggle duration versus distance curve is three to four times greater over land. In other words, land provides a stronger odometric signal than does water. Similar results were obtained when experiment 1 was repeated 3 weeks later using a different colony of bees. The new colony was placed in the same shed and allowed 10 d to acclimatize itself to the new location and familiarize itself with the surrounding terrain before the experiment commenced. The feeder was moved along exactly the same route as the original experiment 1. (Stakes had been placed on the ground to mark the locations of the feeding stations during the first experiment, to ensure that the same locations were used in the repetition. The feeder positions over the water were reproduced by using a marked rope, as described in Materials and Methods .) In this case, the slopes of the curve were 1.071 ± 0.127 (segment 1), 0.283 ± 0.693 (segment 2), and 1.990 ± 0.381 (segment 3). Here, again, the slope of the curve in segment 2 is substantially and significantly lower, as indicated by Student t -test results, than the slope in segment 1 ( p < 0.01) and the slope in segment 3 ( p < 0.001). Scene Analysis As described in Materials and Methods , a digital camera was used to acquire samples of images of the visual scenes that the bees would have encountered during the flights over land and water. Two such samples, one over land and the other over water, are shown in Figure 4 . The mean contrasts of these and a few other scenes, as measured within windows of various sizes ( Figure 4 ) are given in Tables 1 and 2 . In each case, contrasts were measured for images that were obtained with and without a green filter placed in front of the camera lens. The green filter was chosen to mimic the spectral sensitivity of the honeybee's green photoreceptor channel, the channel involved in the sensing of image motion ( Lehrer 1987 ; Chittka and Tautz 2003 ). Contrasts tend to be slightly higher when scenes—over land as well as over water—are viewed through the green filter. It is clear from Tables 1 and 2 , however, that, in general, contrasts tend to be substantially greater over land than over water, regardless of whether the scenes are viewed directly or through a green filter. The contrast over water tends to be lowest when there is no wind, i.e., when the water is still and there are no ripples. However, we did not investigate this latter phenomenon systematically. Figure 4 Land and Water Terrain Examples of scenes over land (A) and water (B) that were photographed and analyzed for mean contrast within windows of various sizes, as illustrated by the red boxes in the center of the photographs. Results are given in Tables 1 and 2 . Table 1 Mean Contrasts of Scenes over Land Table 2 Mean Contrasts of Scenes over Water We have assumed in this analysis that the image motion experienced by the bee is dominated by flow in the ventral visual field. This is not an unreasonable assumption, because the lateral structures (such as bushes and trees) were sparse and usually farther away than the ground, except in certain rare circumstances (see below). Discussion Our results demonstrate, first of all, that bees can be trained to fly reliably and without accident over stretches of water that span a few hundred meters. This has also been reported in earlier studies ( Heran and Lindauer 1963 ; von Frisch 1993 ; Gould and Gould 1988 ), although some studies (e.g., Heran and Lindauer 1963 ) mention that bees experience difficulty in flying across lakes, often losing altitude and plunging into the water. As in earlier studies, we observed that individually marked bees trafficked regularly between the hive and the feeder, regardless of whether the feeder was on land or water. The short flight times (of a minute or two across the whole stretch of water, as established by radio communication) indicated that they had no difficulty in flying the direct route between origin and destination. We found, however, that the feeder attracted far fewer recruits when it was positioned over water. Although the scout bees that were trained to visit the feeder on land readily followed the feeder when it was moved over the water, recruits failed to appear once the water was reached. This was not because the trained scout bees ceased to dance once the feeder was over the water; they continued to dance with high vigor, and their waggles encoded a position over the water. Recruits reappeared when the feeder arrived at the island; in fact, they seemed to “anticipate” its arrival by patrolling the shore of the island even before the boat had landed there. Recruits were observed all along the route on the island, as well as along the entire stretch of the route used in experiment 2, which was exclusively over land. One possible explanation for the lack of recruitment when the feeder was over water could be the relatively low slope of the distance indication curve in this region, which would lead to a less precise indication of the position of the feeder, making it harder to pinpoint. This explanation seems unlikely, however, since the boat in which the feeder was carried was large and conspicuous, and was the only object on the water. Another possible reason for the lack of recruitment could have to do with the lack of assistance from experienced foragers. A recent study showed that experienced foragers can sometimes aid recruits in pinpointing a feeder by synchronizing their flights and performing buzzing flights around an unscented feeder ( Tautz and Sandeman 2003 ). However, such assistance was not apparent in our study when the feeder was over the water. Why this assistance was not provided remains unexplained. A third possibility is that, when inexperienced recruits fly over water, they fly at a higher or lower altitude than the trained scouts, and thus experience a different magnitude of optic flow. Consequently, they may search for the feeder at the wrong location. This possibility remains to be explored. What our observations do suggest, however, is that the trained bees fly at about the same height over land as they do over water (see below). A fourth possibility relates to the controversial hypothesis, advanced by Gould and Gould (1988) , that experienced bees have a “knowledge” of the surrounding landscape, including information about the existence and topography of bodies of water; and that they do not fly to locations signaled by other dancing bees that correspond to positions that are on the water, because such locations would be unlikely to bear food under natural circumstances. The primary contribution of the present study, however, is the demonstration that the honeybee's odometer does not run at a constant rate in outdoor flight. The results shown in Figures 2 and 3 reveal that the mean waggle duration in the dance increases at a slower rate when bees fly over water than when they fly over land. In other words, for routes of the same length over land and water, the bees' perception of distance flown (as indicated by their dances) is smaller for flights over water. Thus, the honeybee's odometer runs at a slower pace when flight is over water. There are a number of possible explanations for this finding. One reason could be that while flight over water is likely to stimulate only the ventral fields of view of the eyes with image motion, flight over land is likely to provide image motion signals in the lateral fields as well. In three earlier studies, honeybee odometry was investigated by training bees to fly through short, narrow tunnels lined with visual textures on the walls and/or or the floor. One of these studies ( Srinivasan et al. 1997 ) suggested that the odometric signal is driven primarily by image motion in the lateral fields of view. However, the two others found that motion in the lateral as well as the ventral fields of view is important ( Hrncir et al. 2003 ; Si et al. 2003 ). While the reasons for this discrepancy remain to be resolved, both studies suggest that lateral flow, if present, can contribute to the odometric signal. It must be noted, however, that in our experiments, the image motion experienced by the bee is likely to have been dominated by flow in the ventral visual field. This is because the lateral structures (such as bushes and trees) on both the water and land routes were sparse and were usually further away than the ground, except in rare circumstances. A second reason for the different sensitivities of the odometer to flight over land and water might be that bees increase their altitude when flying over water, thus reducing the extent of image motion that a given forward motion of the bee would elicit in the eye. Although we do not have precise information on flight heights in this study, visual observation of bees approaching the feeder over the water suggested that they flew at heights of between 1 m and 2 m above the water surface, similar to the heights at which they are reported to cruise over land ( Heran and Lindauer 1963 ). In fact, Heran and Lindauer (1963) reported that bees tend to fly lower over water, and seem to experience difficulty in maintaining the same altitude as they do over land. While we did not observe this phenomenon, their evidence as well as ours seems to argue against the possibility that the reduced sensitivity of the odometer over water is caused by flight at a higher altitude. A third possible explanation has to do with the fact that the visual spatial texture of land could be considerably different from that of water. Investigation of this possibility would require a detailed analysis of the spatial frequency spectra of samples of land and water images, which we have not undertaken in the present study. However, a recent study examined the effects of varying visual texture on perceived distance flown in honeybees ( Si et al. 2003 ). In that study, bees were trained to visit a feeder placed at the far end of a short, narrow tunnel, and their dances were analyzed as the texture and contrast of the patterns lining the floor and walls of the tunnel were systematically varied. The results revealed that perceived distance was almost invariant to changes in visual texture (i.e., changes in the spatial frequency content of the patterns). The same study, however, also found that the odometric signal dropped substantially when the contrast of the pattern was reduced to a level below 20% ( Si et al. 2003 ). This finding is consistent with our present field data, which suggest that the odometric signal is strong when the bees fly over land (which possesses a mean contrast of about 20%) but weak when they fly over water (which possesses a mean contrast of about 9%). Thus, a fourth explanation—and the most likely one—is that water surfaces exhibit a substantially lower visual contrast than do land surfaces. Regardless of which of the above explanations is the valid one, our results indicate that the differences in the odometric signal between flight over land and over water are due to differences in the visual environment. Of course, the visual properties of land terrain can also vary considerably, depending upon the nature of the vegetation and on the existence of manmade structures. We suggest that this is the reason for some of the local fluctuations in slope that are evident in the data over land. In particular, we noted in the Results section that, in experiment 2, the three data points in the middle of segment 3 ( Figures 3 A and 3 B) exhibit a much lower slope than the rest of the curve. This section of the land terrain was one in which the bees' flight took them along a paved bicycle path for a stretch of about 200 m. Two views of this section of the terrain, as would be experienced by a bee flying 1.7 m above the ground, are shown in Figure 5 . These images were acquired without using any color filter (see Materials and Methods ). It is evident that the surface of the bicycle path provides rather low contrast. (The mean contrast within the rectangle in Figure 5 A was measured to be 14.30%.) The contrast of the surface is particularly low on a cloudy day ( Figure 5 B), when the surrounding vegetation does not cast any sharp shadows on the path. (The mean contrast within the rectangle in Figure 5 B was measured to be 6.60%, which is even lower than that of most of the water surfaces that were measured.) The weather was indeed cloudy on the day the data for section 3 were obtained. This was confirmed by the weather entries in the experimental log book for that day, as well as by records from the Canberra Meteorological Station. Thus, the honeybee's odometer can run at different rates even on land, depending upon the nature of the local terrain. The three middle points in segment 3 show a progressively decreasing waggle duration, rather than a progressively increasing one. However, the decrease is not significant: A linear regression over these three data points reveals a slope of −0.369 ± 0.638, which is not significantly different from zero ( p > 0.08). Thus, we interpret this as implying that the bees experienced very little optic flow during flight over this region. The final data point in section 3 shows an abrupt increase in waggle duration, compared to the previous three points. Interestingly, as we see from Figure 5 , this is one of the rare segments of the bees' flight in which rows of trees appear close to the trajectory in the left and right lateral visual fields, potentially providing strong lateral flow. Figure 5 Terrain Along Which the Bees Were Trained to Fly in Segment 3 of Experiment 2 These two photographs show the terrain along which the bees were trained to fly in segment 3 of experiment 2. In (A) the sun was shining clearly, while in (B) it was behind a cloud. The rectangles depict the area within which mean contrast was measured (details in text). It is instructive to compare our results with the findings from an earlier study by von Frisch and Lindauer ( von Frisch 1993 ). In 1962–1963, von Frisch and Lindauer compared the dance tempos of bees returning from a 340-m flight over water with those of bees returning from a flight of the same distance over land ( von Frisch 1993 ). (The “dance tempo” is the number of dance circuits completed in 15 s [ von Frisch 1993 ].) They found no significant difference in the dance tempos under the two conditions. Based on this observation, von Frisch and Lindauer concluded that land and water drive the odometer at the same rate, and suggested, therefore, that the honeybee's odometer is driven largely by nonvisual signals. However, they did not measure the duration of the waggle phase, which is now considered to be the true representation of distance traveled ( Seeley et al. 2000 ). To investigate this issue more closely, we have measured four parameters of our dance data: mean circuit duration, mean waggle duration, mean return duration (the duration of the return phase of the circuit made during the dance), and mean dance tempo. The mean circuit duration is equal to the sum of the mean waggle duration and the mean return duration, and is proportional to the reciprocal of the dance tempo. The results of this analysis, for flights of the same distance over water and over land, are shown in Tables 3 and 4 , respectively. These numbers were obtained by analyzing the dance data from experiments 1 and 2, respectively, for feeder distances ranging from 190 m to 390 m. Our results concur with those of von Frisch and Lindauer: As the feeder distance increased from 190 m to 390 m, the mean circuit duration exhibited a similar increase over land as it did over water (compare the first columns of Tables 3 and 4 ). Although the increase over land is somewhat larger than that over water, the difference is not statistically significant ( p > 0.3, two-way ANOVA). The mean waggle duration increased much more rapidly on land than it does over water, just as Figures 2 and 3 indicate. However, the mean return duration increased much more rapidly on water than it did on land. As a consequence, the mean circuit duration showed a similar variation with distance over land as over water. This reconciles the present findings with those of von Frisch and Lindauer. The mean return duration is considered to be a measure of the “attractiveness” of the food source: the longer the duration, the lower the attractiveness ( Seeley et al. 2000 ). Our data therefore suggest that (a) the attractiveness of a feeder diminishes as its distance from the hive is increased, and (b) the attractiveness decreases more rapidly with distance when flight is over water than when it is over land. The mean waggle duration, on the other hand, is a measure of the perceived distance flown; this quantity increases more rapidly on land than it does over water. Thus, in general, the odometer indeed runs at a faster rate over land than over water. Table 3 Analysis of Dances for Flights over Water (Experiment 1) Table 4 Analysis of Dances for Flights over Land (Experiment 2) Our present findings confirm the suggestions from earlier work ( Esch and Burns 1995 , 1996; Srinivasan et al. 1996 , 1997, 2000; Esch et al. 2001 ; Si et al. 2003 ) that the honeybee dance does not convey information about distance traveled with absolute accuracy. Rather, the distance that is indicated (measured in terms of the mean waggle duration) depends upon the optical environment in which the bee flies. The odometer runs faster in terrain that presents a high contrast and rich texture (such as land with dense vegetation) than in terrain that carries low contrast and sparse texture (such as a water surface). Although the visual movement detection system that drives the odometer is impressively robust to variations in visual texture and contrast—as revealed by an earlier study in which bees were trained to fly through tunnels in which the textures lining the walls and floor were systematically varied ( Si et al. 2003 )—this robustness is not perfect. The study by Si et al. (2003) indicated that, for a flight of a given distance, the odometric reading is largely independent of visual contrast, as long as the contrast is above 20%. Below this value, however, the odometric reading begins to decline. This critical value of contrast is in approximate agreement with the findings of the present study, where we have investigated the properties of the odometric signal in a natural, rather than an artificial, environment. Given that the honeybee dance does not convey accurate information on distance flown, how do scout bees returning from a new food source usually manage to recruit other bees to visit it so quickly and so effectively? One explanation might be that the waggle dance conveys information on the direction as well as the distance of the food source. Thus, potential recruits that are persuaded by the scout's dance to seek out the source would fly in the direction signaled by the scout, and therefore experience approximately the same visual environment as the scout. Consequently, any terrain-induced variations in the odometric signal would be the same for the recruits as well as the scout, so that such variations would not compromise accurate pinpointing of the destination. The recruit would find the goal simply by flying in the specified direction until her odometric signal matched that indicated by the scout's dance. Thus, even though the scout's dance does not indicate distance in absolute terms, the recruits end up close to or at the correct location because their odometric signals evolve in the same way as that of the scout during the flight toward the food. Once the recruits are in the vicinity of the food source, the experienced foragers can assist them by providing visual and/or olfactory cues in the vicinity of the feeder to guide recruits to it ( Tautz and Sandeman 2003 ). Materials and Methods Experimental site The experiments were conducted on the shore of Lake Burley Griffin in Canberra. A two-frame observation hive was set up 190 m from the shore of the lake, with the entrance facing the shore (see Figure 1 ). The ground sloped gently downwards toward the shore. The terrain was grassy and interspersed with shrubs and eucalyptus trees. Beyond the lake shore, 240 m into the water, was an island (Springbank Island) measuring 150 m across along the line of sight from the hive. The island contained dense tree vegetation around its circumference, with grass and a few trees in the middle. Beyond the island was a further 1000-m stretch of water extending to the opposite shore of the lake. Experiments Two experiments were carried out, both using honeybees (Apis mellifera) from the same hive. In experiment 1, bees were trained to fly a route due southwest toward the island. This route comprised an initial stretch over land, followed by a stretch over water, and then again over land (across the island), as shown in Figure 1 . The total length of this route was 580 m. In experiment 2, bees were trained to fly a route of the same total distance due northwest that was entirely over land (see Figure 1 ). In experiment 1, about 20 bees were individually marked and trained to visit a feeder containing 1.0 M nonscented sucrose solution, initially placed 5 m from the hive. When the marked bees started to visit the feeder regularly, the feeder was moved step by step toward the shore. At each position, the dances of marked bees returning from the feeder were recorded at the hive. The distance from the hive to the lake shore was 190 m. The feeder was then placed in a rowboat and taken across the water, again stopping at several locations along the way to record dances back at the hive. In order to accustom the bees visually to the boat, we introduced them to the boat on land well before the water was reached. From about 50 m before the water's edge, the feeder was placed inside the boat. From this point on, the feeder was always in the boat, regardless of whether the location was on land or water. The stretch across the water was 240 m long. Upon reaching the shore of the island, the feeder (still placed in the boat) was moved stepwise across the island to the opposite shore, again recording the dances of bees returning from each location. The stretch across the island was 150 m long. The boat was an inflatable dinghy capable of carrying four adults in addition to the feeder. It was colored bright yellow to facilitate visual detection from a distance by the bees. A view of the boat on the lake is given in Figure S1 . A buoyant rope, carrying brightly colored markers at 10 m intervals, was used to measure out distances on land as well as water. During periods of strong wind or water currents, a stable position over water was maintained with the aid of an anchor, supplemented when necessary by compensatory paddling. The weather was sunny and calm through most of the study, except for a few days (see below). The wind speed rarely exceeded 20 km/h. Most of the time, there were only a few ripples on the water surface. Temperature and weather conditions were recorded through the course of the experiment and were supplemented by records from the Canberra Meteorological Station. There were always at least two experimenters at the feeder (regardless of whether the feeder was on land or water), and two experimenters at the hive. The visit of each marked bee at the feeder, and its subsequent dance in the hive, were followed through radio communication between the feeder and the hive. At each feeder position, recording of a given bee's dances was commenced after it had made three visits to the feeder. This was done to allow adequate time for the dances to adjust to each new feeder position. Data were collected from between seven and 14 different individually marked bees at each feeder position. This procedure required a stay of about 30–60 min at each feeder position. In experiment 2, about 20 bees were trained by moving a feeder step by step from the hive, as in experiment 1, but along a route that was entirely over land, as described above. The total length of this route (580 m) was identical to that in experiment 1, thus enabling a direct comparison of the of the bees' dances along the two routes. Recording and analysis of bee dances The observation hive was housed in a specially constructed shed that afforded a weatherproof environment for observing and filming bee dances. Dances of marked bees were filmed at 25 frames/s, using a Sony (Tokyo, Japan) DCR-TRV310E video camera mounted on a tripod placed adjacent to one face of the observation hive, near the entrance. A mechanical gate at the entrance to the hive ensured that bees entered (and left) the hive only on the side facing the camera, thus facilitating the filming of dances. Bee dances were analyzed frame by frame to measure the mean duration of the waggle phase. The waggle duration was considered to be a measure of the bees' perception of the distance flown from the hive to the feeder: the longer the waggle duration, the greater the perceived distance ( Esch et al. 2001 ). A total of between 68 and 217 dances were analyzed for each feeder position, from between seven and 14 individually marked bees. A total of over 6,000 dances were evaluated in the study. The dances were evaluated as follows. For each dance, the mean waggle duration was estimated by averaging the waggle durations over all loops. Then, the mean waggle duration for each bee was obtained by averaging the mean waggle durations over all of its dances at that feeder position. Finally, the mean waggle duration of all bees was calculated from the mean waggle durations for the individuals. The standard error of the mean was also calculated and displayed in the graphs (see Figures 2 and 3 ). Linear regressions of the data, and 95% CIs for the slopes of the fitted regression lines, were computed using the GraphPad Prism (GraphPad Software, San Diego, California, United States) statistical analysis package. Regression slopes of different data sets were also compared using the same package, which implemented the slope comparison test described in Sokal and Rohlf (1995) . Scene photography and analysis Samples of the visual scenes that the bees would have experienced while flying over land and water were acquired by digital camera (Coolpix 950, Nikon, Tokyo, Japan), which produced color images of 1200 × 1600 pixel resolution. Sections of these images were analyzed to compare the mean visual contrast over land with that over water. The intensity of the image at each pixel was taken to be the average of the values of the three color subpixels at that location. These images were taken on a calm day with weather conditions similar to those on which the experiments were conducted. The surface of the lake was smooth, with relatively few ripples. The mean contrast in an image section was computed as the ratio of the standard deviation to the mean value of the intensities of all the pixels within that section, and was expressed as a percentage. It was recently demonstrated that the odometer is “color-blind” and is driven primarily by the green receptor channel ( Chittka and Tautz 2003 ), as are other motion-sensitive pathways in the bee ( Lehrer et al. 1988 ; Zhang et al. 1990 ; Zhang and Srinivasan 1993 ). Therefore, each scene was photographed twice: once without any color filter, and once through a color filter with a spectral transmission that approximated the spectral sensitivity of the honeybee's green receptor (B+W 30061 3X MRC [Schneider-Kreuznach, Bad Kreuznach, Germany]; with peak transmission at 530 nm and a bandwidth of 120 nm at half sensitivity). Supporting Information Figure S1 View of Boat at One of the Training Positions on the Lake (1.8 MB JPG). Click here for additional data file. Video S1 Marked Bee Dancing Upon Return from the Feeder When the Feeder Is Positioned 60 m into the Lake (250 m from the Hive) (27.8 MB AVI). Click here for additional data file.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC449896.xml
520834	Production of infectious human immunodeficiency virus type 1 does not require depletion of APOBEC3G from virus-producing cells	Background The human immunodeficiency virus Vif protein overcomes the inhibitory activity of the APOBEC3G cytidine deaminase by prohibiting its packaging into virions. Inhibition of APOBEC3G encapsidation is paralleled by a reduction of its intracellular level presumably caused by the Vif-induced proteasome-dependent degradation of APOBEC3G. Results In this report we employed confocal microscopy to study the effects of Vif on the expression of APOBEC3G on a single cell level. HeLa cells dually transfected with Vif and APOBEC3G expression vectors revealed efficient co-expression of the two proteins. Under optimal staining conditions approximately 80% of the transfected cells scored double-positive for Vif and APOBEC3G. However, the proportion of double-positive cells observed in identical cultures varied dependent on the fixation protocol and on the choice of antibodies used ranging from as low as 40% to as high as 80% of transfected cells. Importantly, single-positive cells expressing either Vif or APOBEC3G were observed both with wild type Vif and a biologically inactive Vif variant. Thus, the lack of APOBEC3G in some Vif-expressing cells cannot be attributed to Vif-induced degradation of APOBEC3G. These findings are consistent with our results from immunoblot analyses that revealed only moderate effects of Vif on the APOBEC3G steady state levels. Of note, viruses produced under such conditions were fully infectious demonstrating that the Vif protein used in our analyses was both functional and expressed at saturating levels. Conclusions Our results suggest that Vif and APOBEC3G can be efficiently co-expressed. Thus, depletion of APOBEC3G from Vif expressing cells as suggested previously is not a universal property of Vif and thus is not imperative for the production of infectious virions.	Background Replication of human immunodeficiency virus type 1 (HIV-1) in most primary cells and some immortalized T cell lines is dependent on the expression of a functional Vif protein. In the absence of Vif, virus replication is restricted by a host factor that was recently identified as CEM15 (now referred to as APOBEC3G) [ 1 ], a host cytidine deaminase targeting DNA substrates in vitro [ 2 ] but whose role in normal cells is unknown. In the absence of Vif, APOBEC3G is efficiently incorporated into virus particles [ 3 - 9 ] where it causes extensive cytidine to uracil changes on the viral minus-strand cDNA during reverse transcription [ 5 , 10 - 12 ]. The conversion of cytidine to deoxyuridine on the minus-strand cDNA either results in guanine to adenine changes on the viral plus-strand cDNA to yield highly mutated viral genomes or triggers the degradation of the deaminated minus strand cDNA through the action of a DNA repair mechanism that involves removal of the uracil base by uracil DNA glycosylase and subsequent endonucleolytic cleavage at the abasic sites by apyrimidinic endonuclease (reviewed in [ 13 , 14 ]). While both mechanisms are detrimental to virus replication, the reported inability of vif-defective viruses grown in restrictive cells to reverse transcribe the viral genome into full-length cDNA is more consistent with the latter mechanism involving the degradation of deaminated viral cDNA [ 15 - 19 ]. Vif is a 23-kDa basic protein that is expressed late during infection in a Rev-dependent manner [ 20 ]. Immunocytochemical analyses revealed a largely cytoplasmic localization of Vif [ 21 - 23 ]. However, Vif is efficiently incorporated into HIV particles during productive infection through an interaction with viral genomic RNA and associates with viral nucleoprotein complexes [ 22 , 24 - 26 ]. In the presence of Vif, the steady-state levels of cell-associated APOBEC3G – as judged by immunoblot analysis – are reduced by 3–10 fold [ 3 - 8 , 27 , 28 ]. This Vif-dependent reduction in APOBEC3G levels has been attributed to proteasome-dependent degradation of the protein and requires a direct interaction of Vif with APOBEC3G [ 3 , 6 - 8 ]. Like Vif, APOBEC3G is a cytoplasmic protein. In fact, co-immunoprecipitation analyses demonstrated an interaction of Vif and APOBEC3G in transiently transfected cells [ 3 , 5 , 6 , 27 , 29 - 32 ]. The formation of stable Vif:APOBEC3G complexes seemed to be at odds with the reported proteasome-dependent degradation of APOBEC3G in Vif-expressing cells [ 3 , 6 - 9 , 27 , 28 ]. Indeed, the identification of Vif:APOBEC3G complexes in mixtures of cell extracts that had been individually transfected to express either Vif or APOBEC3G suggested that the stable interaction of Vif and APOBEC3G during co-immunoprecipitation may occur after cell lysis [ 6 ]. Thus, the co-immunoprecipitation of Vif and APOBEC3G from cell extracts is not necessarily an indication of the existence of stable intracellular complexes. Quite to the contrary, Marin et al reported a profound effect of Vif on the expression of APOBEC3G on a single cell level. They found that expression of Vif in transiently transfected COS7 cultures resulted in an almost complete segregation of cells expressing either APOBEC3G or Vif [ 6 ]. Interestingly, this segregation of Vif and APOBEC3G into separate cells was seen only for wild type Vif. In fact, only 10% of cells expressing wild type Vif were double-positive while 95% of cells expressing an inactive Vif variant also contained APOBEC3G [ 6 ]. The current study aims at characterizing in more detail the effects of Vif on the expression of human APOBEC3G on a single cell level. The study was initiated because of the apparent discrepancy between the drastic effects of Vif on APOBEC3G reported by Marin et al and our own finding of only moderate effects of Vif on APOBEC3G expression in transiently transfected cells. In our study, Vif was expressed from a subviral construct in a Tat- and Rev-dependent manner while APOBEC3G was expressed either in a Tat-dependent manner from an HIV-1-LTR-based vector or independently from a CMV-promoter-based expression vector. The Tat-dependent APOBEC3G expression vector was used to restrict APOBEC3G expression to cells also expressing Tat (and thus Vif). Confocal microscopic analysis of HeLa cells transiently transfected with Vif and APOBEC3G expression vectors revealed significant variations in the number of double-positive cells in identical samples ranging from as low as 40% to as high as 80% of transfected cells depending on fixation method and antibodies employed. Importantly, the appearance of cells expressing only Vif or APOBEC3G was observed both with wild type Vif and a biologically inactive variant and thus cannot be explained by Vif-induced degradation of APOBEC3G. Finally, despite the efficient co-expression of Vif and APOBEC3G, viruses produced in these cultures were fully infectious. We therefore conclude that the Vif-induced exclusion of APOBEC3G from virus-producing cells reported by Marin et al [ 6 ] does not apply to our system and because of that is not a universal property of all Vif proteins. This implies that elimination of APOBEC3G is not an obligate requirement for the production of infectious viruses from APOBEC3G-expressing cells. Results Expression of Vif in the context of a proviral vector only moderately reduces cellular APOBEC3G levels A number of previous studies reported the efficient Vif-dependent degradation of APOBEC3G by cellular proteasomes [ 3 , 6 , 8 , 28 ]. However, we and others noted only a moderate reduction of the cellular APOBEC3G levels in response to Vif expression [ 4 , 5 ]. This is exemplified in figure 1 where APOBEC3G was expressed either in the presence or absence of Vif. Specifically, HeLa cells were transfected with pcDNA-APO3G together either with wild type pNL-A1 (Fig. 1A , lane 2) or its vif-defective variant, pNL-A1vif(-) (lane 3). Mock transfected cells were included as a control (lane 1). Cells were harvested 24 hr post-transfection and whole-cell lysates were subjected to immunoblot analysis as described in Methods using an APOBEC3G-specific polyclonal antibody (Fig. 1A , top panel) or a Vif-specific monoclonal antibody (Fig. 1A , middle panel). To control for loading errors, the filters were re-probed with an antibody to α-tubulin (Fig. 1A , bottom panel). Consistent with our previous results, expression of Vif from pNL-A1 only moderately reduced the steady-state levels of APOBEC3G in HeLa cells. Quantitation of the data confirmed that expression of APOBEC3G in the presence of Vif was reduced by only about 20% (Fig. 1B ). Figure 1 Vif has a moderate effect on APOBEC3G steady-state levels. (A) HeLa cells were transfected with pNL-A1 and pcDNA-APO3G vector DNA at a 4:1 ratio. As control, mock-transfected cells (lane 1) and cells transfected with the Vif-deficient pNL-A1ΔVif construct and pcDNA-APO3G vector DNA at a 4:1 ratio (lane 3) were included. Cell lysates were processed for immunoblotting as described in Methods and APOBEC3G and Vif-specific proteins were identified using an APOBEC3G-specific polyclonal antibody (α-APO3G) or a Vif-specific monoclonal antibody #319 (α-Vif). Tubulin was identified using an antibody to α-tubulin. (B) APOBEC3G-specific bands were acquired by densitometric scanning of the film and were quantified using the Fuji ImageGauge 4.0 software (Fuji Photofilm Co, LTD). Results are expressed as percent of the Vif-negative control, which was defined as 100%. Co-expression of APOBEC3G and Vif in HeLa cells An earlier study investigating the coexpression of Vif and APOBEC3G by confocal microscopy concluded that APOBEC3G was virtually excluded from Vif-expressing COS7 cells [ 6 ]. To verify this observation, we investigated the effects of Vif on APOBEC3G expression on a single cell level by performing a series of immunocytochemical analyses. For that purpose, HeLa cells were transfected with the Vif expression vector pNL-A1 together with pcDNA-APO3G for the expression of human APOBEC3G. Cells were grown on cover slips, fixed 24 hr later with cold methanol, and stained with antibodies to APOBEC3G (Fig. 2 , panels A & D) and Vif (panels B & E). The results of this experiment show that APOBEC3G can be expressed in Vif-positive cells (white arrow heads) without a dramatic reduction in its expression level when compared to Vif-negative cells (yellow arrow heads). Furthermore, these data confirm that APOBEC3G is localized to the cytoplasm while Vif was observed in this experiment in some cells both in the cytoplasm and the nuclei of cells (red arrow heads). Finally, we also observed cells expressing Vif that were APOBEC3G-negative (blue arrow heads). Overlay of the Vif and APOBEC3G channels revealed a partial co-localization of Vif and APOBEC3G in the cytoplasm apparent by the yellow staining in panels C & F of figure 2 . Interestingly, a significant number of cells in this experiment were single-positive expressing either APOBEC3G or Vif alone. The appearance of Vif-positive, APOBEC3G-negative cells could be explained by a Vif-dependent restriction of APOBEC3G expression as proposed by Marin et al [ 6 ]. However, cells expressing Vif only were rare when compared to cells expressing APOBEC3G alone (data not shown). The preponderance of APOBEC3G single positive cells cannot be explained by a Vif-dependent restriction but more likely represents a technical, albeit reproducible, artifact. Figure 2 Co-expression of Vif and APOBEC3G in HeLa cells. HeLa cells were transfected with pNL-A1 and pcDNA-APO3G at a 1:1 molar ratio. Transfected cells were grown on cover slips, fixed in methanol and processed for confocal microscopic analysis as described in Methods. Cells were stained with a rabbit polyclonal antibody to APOBEC3G (A & D) and a monoclonal Vif antibody (B & E ). APOBEC3G was visualized using a Texas red-conjugated secondary antibody; Vif was visualized with a Cy2-conjugated secondary antibody. Panels C and F are merged images of panels A & B and D & E , respectively. Arrow heads are defined as follows: white = APOBEC3G:Vif-double-positive cells; yellow = Vif-negative cells; red = cells exhibiting nuclear and cytoplasmic staining for Vif. Tat-dependent expression of APOBEC3G reduces the fraction of single-positive cells In the experiment shown in figure 2 , APOBEC3G was expressed under the control of a CMV promoter while Vif was expressed from the HIV-LTR promoter under the regulatory control of Tat and Rev. Because of the independent expression of Vif and APOBEC3G it cannot be ruled out that the large number of single-positive cells in that experiment – while statistically improbable – was caused by the selective transfection of cells with either the Vif or the APOBEC3G expression vector. To check this possibility, we expressed APOBEC3G from the HIV-1 long terminal repeat (LTR) promoter driven vector pHIV-APO3G [ 4 ]. Because of its dependence on Tat, APOBEC3G expression from pHIV-APO3G is restricted to cells also expressing Vif. Indeed, transfection of pHIV-APO3G into cells in the absence of pNL-A1 or any other Tat expression vector did not reveal any APOBEC3G expression as judged by immunofluorescence analysis or immunoblotting attesting to the strict Tat-dependence of this APOBEC3G expression vector (data not shown). In addition of measuring the context-dependent expression of APOBEC3G, we also wanted to determine the influence of the fixation procedure on the efficiency of Vif:APOBEC3G co-staining. It is well known that the choice of fixative can affect the ability of a given antibody to recognize a specific epitope on its target protein. Frequently, epitopes are masked because of the folding properties of a protein in vivo or because of pre-existing protein-protein interactions that may compete for antibody binding. To test this possibility we compared a formaldehyde fixation procedure employed previously [ 6 ] with the methanol fixation procedure employed in our own studies [ 22 ]. HeLa cells were transfected with pNL-A1 and pHIV-APO3G at a 1:1 molar ratio. Cells were fixed 24 hr later either with methanol (MeOH) as in figure 2 or with formaldehyde (FA) as described in Methods. Cells were stained with Vif- and APOBEC3G-specific antibodies as described in figure 2 . The results of this experiment show that expression of APOBEC3G under the control of the HIV-1 LTR indeed increased the proportion of double-positive cells both in methanol-fixed samples (Fig. 3 , panels A-C) and formaldehyde-fixed specimens (Fig. 3 , panels D-F). This suggests that the high proportion of single-positive cells observed in figure 2 was not the result of a Vif-dependent restriction of APOBEC3G but was caused by the independent expression of APOBEC3G from a Tat-independent promoter. Again, in methanol-fixed samples APOBEC3G expression levels in Vif-positive cells (Fig. 3A , white arrow heads) were indistinguishable from those observed for neighboring Vif-negative cells (Fig. 3A , yellow arrow head). Interestingly, the APOBEC3G fluorescent intensity appeared to be reduced in Vif-positive formaldehyde-fixed specimens when compared to Vif-negative cells or cells expressing low levels of Vif (Fig. 3D ; compare white and yellow arrow heads). Because the methanol-fixed samples did not show a Vif-dependent reduction in APOBEC3G signals in these experiments, we conclude that the reduction in APOBEC3G signals observed in formaldehyde-fixed samples is not the result of Vif-induced degradation of APOBEC3G but is a technical artifact. Figure 3 Tat-dependent expression of APOBEC3G. HeLa cells were transfected with pNL-A1 and pHIV-APO3G at a 1:1 molar ratio. Transfected cells were grown on cover slips for 24 hr and then either fixed with ice-cold methanol (panels A-C) or with formaldehyde buffer as described in Methods (panels D-F). Cells were stained with an APOBEC3G-specific antibody (A & D) and a Vif monoclonal antibody (B & E) as in figure 2 and analyzed on a confocal microscope. Panels C & F are overlays of panels A & B and D & E, respectively. Arrow heads are defined as follows: white = APOBEC3G:Vif-double-positive cells; yellow = Vif-negative cells; blue = APOBEC3G negative cells. Co-expression of Vif and APOBEC3G: Protein degradation or epitope masking? For a more quantitative analysis and to determine possible effects that arise from the use of different antibodies, we extended the experiment shown in figure 3 to include three different antibodies for the identification of APOBEC3G. As before, HeLa cells were co-transfected with a 1:1 ratio of pNL-A1 and pHIV-APO3G plasmid DNAs. Cells were grown on cover slips and fixed 24 hr later either with formaldehyde (Fig. 4 , panels A-C) or methanol (panels D-F) as described in figure 3 . Cells were then stained with either a monoclonal antibody to the C-terminal Myc-epitope in APOBEC3G together with a polyclonal Vif antibody (Fig. 4 , panels A & D), or a polyclonal Myc antibody together with a monoclonal Vif antiserum (Fig. 4 , panels B & E). A third set of cells was stained with a polyclonal APOBEC3G-specific antiserum together with the monoclonal Vif antibody (Fig. 4 , panels C & F). Representative fields are shown for each combination. Figure 4 Effect of fixation method and antibody choice on co-expression of Vif and APOBEC3G. HeLa cells were transfected with pNL-A1 and pHIV-APO3G as described in figure 3. Cells were grown on cover slips for 24 hr and then either fixed with ice-cold methanol (panels A-C) or with formaldehyde buffer as in figure 3 (panels D-F) and stained with the following combinations of antibodies: (A & D) polyclonal Vif + anti-Myc MAb 9E10; (B & E) anti-Vif MAb #319 + anti-Myc polyclonal antibody; (C & F) anti-Vif MAb #319 + anti-APO3G polyclonal antibody. Vif was visualized using Cy2-conjugated secondary antibodies (green) and APOBEC3G was visualized with Texas red-conjugated antibodies (red). Areas of overlap appear as yellow. To quantify the results, multiple optical fields were analyzed (n = 5–10) with a total of at least 100 transfected cells for each parameter. As can be seen in figure 5 , methanol-fixed samples showed a relatively modest variation among the three antibodies used. All three antibodies identified between 45% to 60% of the cells as double-positive for Vif and APOBEC3G. In contrast, formaldehyde-fixed samples exhibited a larger antibody-dependent variation. Staining with the 9E10 monoclonal antibody to the Myc-epitope in APOBEC3G yielded the lowest efficiency of staining and identified little more than 40% of the transfected cells as double-positive for Vif and APOBEC3G. In contrast, staining of APOBEC3G was more efficient with the polyclonal Myc antibody, which identified approximately 80% of the transfected cells as double-positive in formaldehyde-fixed samples. Finally, the polyclonal APOBEC3G-specific antibody was slightly less efficient for the staining of FA-fixed samples than methanol-fixed samples and identified about 40% of the formaldehyde-fixed samples as double-positive. Since all samples were derived from the same transfected culture, variations in the co-expression of Vif and APOBEC3G in the individual samples can only be explained by the differential sensitivity of the antibodies to the fixation procedure. Figure 5 Quantitative analysis of Vif and APOBEC3G co-expression. Samples from figure 4 were analyzed for the expression of Vif (grey bars) or APOBEC3G (white bars) or for double-positive cells (black bars). Between 5 and 10 independent optical fields were analyzed to yield at least 100 transfected cells. Error bars reflect standard deviations calculated from multiple optical fields. The results obtained with methanol-fixed samples (MeOH) are on the left; results from formaldehyde-fixed samples (FA) are on the right. Exclusion of APOBEC3G from cells expressing biologically inactive Vif protein Under optimal conditions, wild type Vif and APOBEC3G were coexpressed in about 80% of transfected cells (see figure 5 ). Thus, 20% of the transfected cell population either was expressing Vif but not APOBEC3G or was single-positive for APOBEC3G. To investigate whether the presence of such single-positive cells is due to an activity of Vif or is a general characteristic of transiently transfected cells, we studied the effects of a biologically inactive Vif variant. For this purpose, we employed a Vif mutant carrying a deletion of residues 23–45 in Vif. We previously showed that this mutant is unable to rescue viral infectivity in APOBEC3G-expressing cells [ 4 ]. Like wild type Vif, VifΔ23–43 was expressed in the context of the subviral expression vector pNL-A1. HeLa cells were cotransfected with pNL-A1/VifΔ23–43 and pHIV-APO3G, fixed with methanol and processed for confocal microscopy as described for figure 2 . As shown in figure 6 , coexpression of VifΔ23–43 and APOBEC3G yielded a significant number of double-positive cells (white arrow heads). However, as observed before with wild type Vif, we also identified cells that were Vif-positive but had significantly reduced APOBEC3G levels (Fig. 6 , blue arrow heads) or cells that were APOBEC3G positive but did not express Vif (yellow arrow heads). As with wild type Vif, overlay of the Vif and APOBEC3G image channels suggested a partial colocalization of the two proteins. In cells, in which Vif had spontaneously collapsed into a perinuclear aggregate (green arrow head), APOBEC3G did not exhibit a similar change in subcellular distribution. This is in contrast to the Vif-induced reorganization of vimentin reported previously [ 22 ]. Thus, VifΔ23–43 is either unable to interact with APOBEC3G or forms complexes that are unstable. These results also imply that the partial colocalization of APOBEC3G and Vif noted in this study may not reflect a true physical interaction of the two proteins. Figure 6 Co-expression of APOBEC3G and a biologically inactive Vif variant. HeLa cells were transfected with pHIV-APO3G and pNL-A1/VifΔ23–43, encoding a biologically inactive Vif variant. Cells were fixed in methanol and stained with the monoclonal Vif antibody (MAb #319; green) and a rabbit polyclonal antibody to APOBEC3G (red) as described above. APOBEC3G is shown in panel A; panel B depicts samples stained for Vif; panel C is the merged image of panels A & B. White and yellow arrow heads depict APOBEC3G:Vif double-positive and Vif-negative cells, respectively. Blue arrow heads point to double-positive cells that show reduced levels of APOBEC3G; the green arrow head depicts a cell where Vif is concentrated around the microtubule organizing center without a similar effect on APOBEC3G. Rescue of viral infectivity and Vif-induced reduction of cellular APOBEC3G levels are not directly linked The combined results from the experiments shown in figures 2 , 3 , 4 , 5 , 6 , do not support the notion that Vif expression leads to the elimination of APOBEC3G from Vif-positive cells. It can be argued, however, that under the experimental conditions employed in our experiments, the Vif expression levels were insufficient or ineffective. To control for this possibility, we compared the infectivity of viruses produced in the presence of various ratios of Vif and APOBEC3G. To allow a direct comparison with the experiments shown in figures 2 to 6 , Vif and APOBEC3G were expressed in trans from pNL-A1 and pHIV-APO3G respectively in the presence of a Vif-defective NL4-3 proviral vector. The ratios of Vif to APOBEC3G expression vector were 1:1, 2:1, and 5:1, respectively. Note that the Vif to APOBEC3G ratio in the experiments shown in figures 2 to 4 was 1:1 throughout. A Vif-negative sample was analyzed as negative control. Virus-containing supernatants were harvested 24 hr after transfection, normalized for equal reverse transcriptase activity and used for the infection of LuSIV indicator cells. Relative virus infectivity was determined by comparing the Tat-dependent expression of luciferase in the target cells (Fig. 7 ). Interestingly, viruses produced at the lowest Vif:APOBEC3G ratio were virtually as infectious as viruses produced in the presence of higher levels of Vif. In fact, increasing the Vif:APOBEC3G ratio to 2:1 or 5:1 did not significantly increase viral infectivity. Instead, at the 5:1 ratio viral infectivity was slightly reduced, presumably due to the inhibitory effect of Vif at high concentrations as reported previously [ 39 ]. Taken together, our data suggest that the inability of Vif to prevent co-expression of APOBEC3G in transiently transfected HeLa cells is not caused by sub-optimal levels of Vif or a lack of Vif activity in our system. Figure 7 Vif efficiently rescues viral infectivity. HeLa cells were transfected with the vif-defective proviral vector pNL4-3vif(-) together with pNL-A1 and pHIV-APO3G at 1:1, 2:1, or 5:1 molar ratios. Cell lysates and purified, concentrated viral extracts were analyzed by immunoblotting using antibodies to APOBEC3G (APO3G), Vif (MAb #319), or an HIV-positive human serum for the identification of viral capsid protein (CA). Virus-containing, filtered supernatants were normalized for equal reverse transcriptase activity and used for the infection of the LuSIV indicator cell line [38]. Virus-induced luciferase activity was measured 24 hr after infection as described in Methods. Relative light units (RLU), which are directly proportional to the infectivity of the viruses, are shown. Error bars reflect standard deviations from duplicate experiments. Discussion APOBEC3G is able to deaminate cytidine residues on the HIV minus-strand cDNA and cause hypermutation of the viral genome. Nevertheless, HIV-1 is able to efficiently replicate in APOBEC3G expressing cells thanks to the activity of the accessory protein Vif. One of the prerequisites for the antiviral activity of APOBEC3G is that it is packaged into the virions where it selectively targets the viral minus-strand cDNA [ 5 , 10 - 12 , 40 , 41 ] and there is convincing evidence that HIV-1 Vif plays an important role in inhibiting the encapsidation of APOBEC3G. The question of how Vif accomplishes this remains under investigation. A number of groups have reported on the rapid Vif-induced degradation of APOBEC3G by cellular proteasomes [ 3 , 6 - 9 , 27 , 28 ]. Consistent with this, treatment of cells with proteasome inhibitors was found to increase APOBEC3G expression levels despite the presence of Vif [ 6 , 7 , 9 , 28 ]. This is contrasted by other reports that noted only a moderate effect of Vif on cellular APOBEC3G levels [ 4 , 5 ]. In fact, our own studies with proteasome inhibitors did not yield a significant increase in APOBEC3G levels in the presence of Vif (manuscript in preparation). Nevertheless, the currently prevailing opinion is that Vif inhibits the encapsidation of APOBEC3G by inducing its rapid degradation in virus-producing cells. While the results from our own study argue against a depletion of APOBEC3G in Vif-expressing cells – thus implying that Vif can rescue viral infectivity despite the presence of APOBEC3G in virus-producing cells – it is important to point out that our data do not rule out the possibility that Vif – under different experimental conditions – can indeed mediate the proteasome dependent degradation of APOBEC3G. In fact, expression of Vif from a codon-optimized vector consistently had a more pronounced effect on APOBEC3G steady-state levels than Vif expressed from pNL-A1 even though the Vif expression levels from the codon-optimized construct were consistently several-fold lower than those from pNL-A1 (manuscript in preparation). Experiments are ongoing to study the differential effects of Vif expressed from pNL-A1 and Vif expressed from a codon-optimized vector on APOBEC3G stability. However, these results could suggest that the effect of Vif on APOBEC3G steady-state levels may be influenced by the context in which Vif is expressed. At any rate, despite our inability to observe Vif-dependent cellular depletion of APOBEC3G, we were invariably able to recover fully infectious HIV under conditions were the intracellular levels of APOBEC3G were only moderately affected. We therefore conclude that (i) Vif has the ability to rescue viral infectivity even in the presence of APOBEC3G and (ii) that intracellular depletion of APOBEC3G and rescue of viral infectivity may be functionally separable activities of Vif. For now, the reason for the differences in the sensitivity of APOBEC3G to Vif noted by us versus other research groups remains unexplained. APOBEC3G can form oligomeric structures and is able to interact with Vif. It is therefore possible that such complexes undergo conformational changes that can mask epitopes thus limiting the access of antibodies used in the experiments. Thus, the discrepancy between our findings of the coexpression of Vif and APOBEC3G in the majority of cells and the virtual exclusion of APOBEC3G from Vif-expressing cells reported by Marin et al. [ 6 ] may be attributed at least in part to differences in the experimental protocols. It is unlikely that the observed discrepancies are strain-specific variations. To this end we have compared the activities of two HIV-1 Vif isolates, HXB2 and NL4-3, which differ by 18 amino acids (9.4%), and found them to be equally active against APOBEC3G (manuscript in preparation). It is unclear why cotransfection of pHIV-APO3G with pNL-A1 produces a fraction of cells that are single-positive for Vif or for APOBEC3G. Since APOBEC3G expression from the pHIV-APO3G vector is strictly Tat-dependent, the results cannot be explained by differential transfection of cells with individual plasmids. Also, this phenomenon is clearly not a consequence of Vif function, since similar results were observed in the presence of a biologically inactive Vif variant (Fig. 6 ) or when APOBEC3G was co-expressed with HIV-1 Gag in the absence of Vif (data not shown). The inability of Vif expressed from pNL-A1 to deplete APOBEC3G is consistent with our previous inability to observe APOBEC3G degradation in kinetic studies [ 4 ]. More recent in-depth kinetic analyses of APOBEC3G employing multiple epitope tags and various antibodies confirm these initial findings and suggest that – instead of inducing APOBEC3G degradation – Vif induces conformational changes in APOBEC3G that affect the ability of antibodies to interact with the protein (manuscript in preparation). Experiments are ongoing to study the nature of the APOBEC3G/Vif complexes and to further decipher the mechanism(s) by which Vif inhibits the encapsidation of APOBEC3G under conditions of no or low intracellular degradation. Conclusions Expression of Vif and APOBEC3G in our experimental setup does not lead to the elimination of APOBEC3G from Vif expressing cells. In fact, more than 80% of successfully transfected cells efficiently co-expressed both proteins. Similar results were observed when a biologically inactive Vif variant was co-expressed with APOBEC3G suggesting that the absence of APOBEC3G in some of the Vif-positive cells is not due to Vif-mediated APOBEC3G degradation but reflects a general characteristic of the transient expression system. Moreover, APOBEC3G expression levels were very similar for Vif-positive and Vif-negative cells as judged from the immunostaining consistent with the only modest reduction in APOBEC3G steady-state levels observed in our immunoblot analyses. Nevertheless, viruses produced under such conditions were fully infectious in the presence but not in the absence of Vif attesting to the biological activity of all the proteins involved and demonstrating that Vif was expressed at saturating levels. We conclude that production of infectious viruses from APOBEC3G expressing cells is dependent on Vif but does not necessitate APOBEC3G exclusion from virus-producing cells. Methods Plasmids The full-length molecular clone pNL4-3 [ 33 ] was used for the production of wild type infectious virus. For transient expression of Vif, the subgenomic expression vector pNL-A1 [ 34 ] was employed. This plasmid expresses all HIV-1 proteins except for gag and pol products. A vif-defective variant of pNL-A1, pNL-A1vif(-) was constructed by deletion of an NdeI/PflMI fragment [ 4 ]. Plasmid pNL-A1/VifΔ23–43 expresses a Vif variant carrying a 21 amino acid deletion (residues 23 to 43) in its vif gene as reported elsewhere [ 4 ]. This Vif variant is inactive and does not target APOBEC3G [ 4 ]. Plasmids pcDNA-APO3G and pHIV-APO3G are vectors for the expression of human APOBEC3G under the control of the CMV immediate early promoter or the HIV promoter, respectively, and were constructed as described elsewhere [ 4 ]. Antisera Serum from an HIV-positive patient (APS) was used to detect HIV-1-specific capsid (CA) proteins. A monoclonal antibody to Vif (MAb #319) was used for all immunoblot analyses and some of the immunohistochemical analyses as indicated in the text and was obtained from Michael Malim through the NIH AIDS Research and Reference Reagent Program [ 23 , 35 , 37 ] For all other immunocytochemical analyses our Vif-specific polyclonal antibody (Vif93) was employed. APOBEC3G, carrying a C-terminal Myc epitope tag was identified either with the Myc-specific 9E10 monoclonal antibody or a polyclonal antibodies to the Myc epitope tag (both antibodies were obtained from Sigma-Aldrich, St. Louis). Alternatively, APOBEC3G was identified using a polyclonal rabbit serum against recombinant APOBEC3G [ 4 ]. Tubulin was identified using a monoclonal antibody to α-tubulin (Sigma-Aldrich, St. Louis). Tissue culture and transfections HeLa cells were propagated in Dulbecco's modified Eagles medium (DMEM) containing 10% fetal bovine serum (FBS). LuSIV cells are derived from CEMx174 cells and contain a luciferase indicator gene under the control of the SIVmac239 LTR [ 38 ]. These cells were obtained through the NIH AIDS Research and Reference Reagent Program and were maintained in complete RPMI 1640 medium supplemented with 10% FBS and hygromycin B (300 μg/ml). For transfection of HeLa cells, cells were grown in 25 cm 2 flasks to about 80% confluency. Cells were transfected using LipofectAMINE PLUS™ (Invitrogen Corp, Carlsbad CA) following the manufacturer's recommendations. A total of 5–6 μg of plasmid DNA per 25 cm 2 flask was used. Cells were harvested 24 hr post-transfection. Transfection efficiency in our analyses was generally 30% to 40%. Preparation of virus stocks Virus stocks were prepared by transfecting HeLa cells with appropriate plasmid DNAs. Virus-containing supernatants were harvested 24 hr after transfection. Cellular debris was removed by centrifugation (3 min, 3000 × g) and clarified supernatants were filtered (0.45 μm) to remove residual cellular contaminants. For determination of viral infectivity, unconcentrated filtered viral supernatants were used for the infection of indicator cells. For immunoblot analysis of viral proteins, virus particles (7 ml) were concentrated by ultracentrifugation through 4 ml of 20% sucrose in PBS as described before [ 4 ]. Immunoblotting For immunoblot analysis of intracellular proteins, whole cell lysates were prepared as follows: Cells were washed once with PBS, suspended in PBS (400 μl/10 7 cells), and mixed with an equal volume of sample buffer (4% sodium dodecyl sulfate, 125 mM Tris-HCl, pH 6.8, 10% 2-mercaptoethanol, 10% glycerol, and 0.002% bromphenol blue). Proteins were solubilized by boiling for 10 to 15 min at 95°C with occasional vortexing of the samples to shear chromosomal DNA. Residual insoluble material was removed by centrifugation (2 min, 15000 rpm in Eppendorf Minifuge). Viral proteins were obtained by boiling concentrated viral pellets in a 1:1 mixture of PBS and sample buffer. Cell lysates and viral extracts were subjected to SDS-PAGE; proteins were transferred to PVDF membranes and reacted with appropriate antibodies as described in the text. Membranes were then incubated with horseradish peroxidase-conjugated secondary antibodies (Amersham Biosciences, Piscataway NJ) and visualized by enhanced chemiluminescence (ECL, Amersham Biosciences). Infectivity assay To determine viral infectivity, virus stocks were normalized for equal reverse transcriptase activity and used to infect LuSIV cells (5 × 10 5 ) in a 24-well plate total volume 1.2 to 1.4 ml. Cells were incubated for 24 hours at 37°C. Cells were then harvested and lysed in 150 μl of Promega 1x reporter lysis buffer (Promega Corp., Madison WI). To determine the luciferase activity in the lysates, 50 μl of each lysate were combined with luciferase substrate (Promega Corp., Madison WI) by automatic injection and light emission was measured for 10 seconds at room temperature in a luminometer (Optocomp II, MGM Instruments, Hamden CT). Immunocytochemistry For the analysis of transfected HeLa cells, cells were scraped off the flasks 3 hr after transfection and reseeded into 12 well plates containing 0.13 mm cover slips. Cells were grown for 15 to 24 hrs at 37°C in DMEM containing 10% FBS. Cells were then fixed at -20°C in precooled methanol (-20°C) for 10 min followed by two washes in PBS or fixed in FA buffer (5% formaldehyde + 2% sucrose in PBS) for 20 min at room temperature followed by two washes in PBS. Coverslips were stored in PBS at 4°C until use. FA-fixed samples were permeabilized for 30 min at room temperature in permeabilization buffer (1% Triton X-100, 10% sucrose in PBS) prior to incubation with antibodies. For antibody staining, cover slips were incubated in a humid chamber at 37°C for 30 min with primary antibodies at appropriate dilutions in 1% BSA in PBS. Cover slips were washed once in PBS (5 min, room temp) and incubated with Texas-Red- or Cy2-conjugated secondary antibodies (diluted in 1% BSA in PBS) for 30 min at 37°C in a humid chamber. Cover slips were then washed twice with PBS and mounted onto microscope slides with glycerol gelatin (Sigma-Aldrich, St. Louis) containing 0.1M N-propyl gallate (Sigma-Aldrich, St. Louis) to prevent photo bleaching and were stored at 4°C in the dark until analyzed by confocal microscopy. Confocal Microscopy For confocal microscopy, a Zeiss LSM410 inverted laser scanning microscope was employed. The microscope was equipped with a krypton/argon mixed-gas laser and was operated by the Microcosm Renaissance 410 (v2.3.4) software package. Images were acquired with a Plan-Apochromat 63x/1.4 oil immersion objective (Carl Zeiss, Thornwood). Additional optical magnification (up to 5-fold) was achieved using the zoom feature of the image acquisition software. For two-color analysis, objects were excited using 488/568 nm laser lines. Green and red emissions were recorded through appropriate filters (515–540 nm band pass filter for Cy2 and 590 nm long pass filter for Texas-Red) and stored in separate (red and green) image channels. At the same time, bright field images (Nomarski optics) were collected and stored in a third (blue) channel. Image quality was enhanced during data acquisition using the Renaissance 410 line average feature (8 or 16x). Competing Interests None declared. Authors' Contributions S. K. carried out immunoblot analyses, infectivity assays, and was involved in the construction of plasmids and the production of antibodies. E.M., M.A.K., H.T., S.O., and R.G. participated in immunoblot analyses, infectivity studies, sample preparations, data validation, and overall experimental design. K.S. conceived of the study, performed IFA analyses, and coordinated the study. S.K. and K.S. participated in the writing of the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC520834.xml
545060	An international comparative study of blood pressure in populations of European vs. African descent	Background The consistent finding of higher prevalence of hypertension in US blacks compared to whites has led to speculation that African-origin populations are particularly susceptible to this condition. Large surveys now provide new information on this issue. Methods Using a standardized analysis strategy we examined prevalence estimates for 8 white and 3 black populations (N = 85,000 participants). Results The range in hypertension prevalence was from 27 to 55% for whites and 14 to 44% for blacks. Conclusions These data demonstrate that not only is there a wide variation in hypertension prevalence among both racial groups, the rates among blacks are not unusually high when viewed internationally. These data suggest that the impact of environmental factors among both populations may have been under-appreciated.	Background Population surveys in the US from early in the last century have consistently documented higher blood pressures and related cardiovascular sequelae in blacks compared to whites [ 1 , 2 ]. The enormous attention focused on this observation has resulted in a dichotomous view of hypertension risk: whereby populations of African origin are considered more susceptible than all other continental groupings and a strong genetic hypothesis of inherent predisposition to hypertension among blacks has become the conventional wisdom [ 3 - 5 ]. Since this research has been limited primarily to the US, the generalizability of these conclusions is open to question. Data on the prevalence of hypertension in other genetically-related populations of African and European descent constitute important evidence but have so far not been considered in the debate. International comparative studies on hypertension have been seriously limited by the absence of a valid method of standardization. In the last decade, however, high quality population surveys have been conducted in a wide range of populations that used either careful internal standardization or sufficiently comparable methods [ 6 - 15 ]. We report here on the patterns of hypertension prevalence in a sample of 3 such surveys among blacks from Africa, the Caribbean and the US and 8 surveys among whites from the US, Canada and Europe. Methods Study design Black populations were drawn from the International Collaborative Study on Hypertension (ICSHIB) and the National Health and Nutrition Survey III [ 6 , 16 ]. A primary report of ICSHIB demonstrated a gradient in hypertension risk from east to west, parallel to the gradient in socioeconomic development and associated lifestyle [ 6 ]. An extensive process of cross-standardization was incorporated into ICSHIB to ensure that measurement technique did not bias the survey results [ 7 ]. We subsequently identified surveys on hypertension conducted since 1986 that were national in scope in North America and Europe. Two North American and six European surveys were included, viz: US [ 8 ] and Canada, [ 9 ], England [ 10 ], Finland [ 11 ], Germany [ 12 ], Italy [ 13 ], Spain [ 14 ] and Sweden [ 15 ]. The US data from NHANES-III are available for public use through the National Center for Health Statistics [ 8 ]. Investigators in Canada and Europe were contacted and invited to join this project. More detailed methods for this component of the study were reported earlier [ 16 ]. In brief, after achieving consensus on the main goals and resolving the methodological issues, data collection forms were distributed. Each collaborator provided average gender- and age-specific data by 5-year age groups for BPs, body mass index (BMI), and counts of hypertensives by treatment and control status. A description of the key aspects of each survey, including the BP measurement procedure, was collected in a standardized format. The surveys that formed the basis of ICSHIB were conducted in localized communities by door-to-door screening [ 6 , 7 ]. In summary, individual communities were chosen on the basis of apparent representativeness and census data were obtained. Sampling was based on probability proportional to size and was structured to lead to a sample equally balanced by gender and age group across the 10-year age. The studies of the European-origin populations and African Americans were larger in scope [ 16 ]. Some were based on a random probability sample of the entire nation, while others were a series of regional samples; none were restricted mainly to a single province or sub-region within the country (Table 3 ). Collectively the studies enrolled 85,000 participants and the number of subjects in individual studies ranged from 1,800 to 23,000. Participation rates varied from 61% to 88%. Sampling was conducted mainly on population registries. Table 3 Hypertension Prevalence (%) among Persons 35–64 Years, in African- and European-Origin Populations * Total (%) Men (%) Women (%) African-Origin Populations Nigeria 13.5 13.9 13.1 Jamaica 28.6 23.4 31.8 US – Black 44.0 43.1 44.8 European-Origin Populations US – White 26.8 29.7 23.9 Canada 27.4 31.0 23.8 Italy 41.5 48.0 35.1 Sweden 38.4 44.8 32.0 England 41.7 46.9 36.5 Spain 46.8 49.0 44.6 Finland 48.6 55.7 41.6 Germany 55.3 60.2 50.4 * Age-adjusted Data collection methods The examination methods have been reported in detail previously [ 6 , 7 , 16 ]. In brief, the mercury sphygmomanometer was used for BP measurements in every country except England, where the Dinamap 8100 oscillometric device was used. All studies had at least 2 measurements and the 2 nd BP from the clinic visit was used to create the mean for the age-gender groups, except for England where the 2 nd home BP was used. Hypertension was defined as BP ≥ 140/90 mmHg or current use of antihypertensive medication. Data analysis BP, body mass index (BMI), and hypertension prevalence were calculated for 5-year age-gender groups and aggregated as the primary data file. To achieve maximum overlap we restricted the analysis to 35–74 years for age-specific estimates of BP and hypertension prevalence, and 35–64 years for age-adjusted results. In the US NHANES whites and blacks were analyzed separately with the appropriate weighting for population size. As previously reported, the prevalence estimates obtained for US blacks from ICSHIB were virtually identical to those from NHANES [ 6 ]; to enhance generalizability, however, we used the NHANES data to represent the US black population. Hypertension prevalence and control was age-adjusted by age-averaging the 5-year age groups combining the data for men and women. For comparison of all white vs. all black populations the mean BP's and prevalences were averaged, considering each country as a single unit (i.e., without weighting by population size). Results Patterns of blood pressure The age-averaged BPs and BMIs are presented for each survey, by gender, in Table 2 . It must be recognized that where treatment is common these data may understate the true values, although this effect is likely to be small when the population is considered as a whole. Trends in BP with age showed considerable heterogeneity within population groups of both continental ancestry (i.e. African and European) (Figures 1 , 2 ). In rural Nigeria, mean BPs were low and rose only modestly with age (Figure 1 ). Intermediate levels of BP were observed in Jamaica, while the US blacks had higher BPs at all ages. As previously reported, whites in the US and Canada had substantially lower BPs over the entire life span than did the Europeans (Figure 2 ). For greater clarity, the age-specific patterns are presented for all black and all white groups combined (Figure 3 ). Table 2 Mean Systolic and Diastolic Blood Pressure and Body Mass Index among Persons 35–74 Years, in African- and European-Origin Populations* Total Sys / Dias Men Sys / Dias Women Sys / Dias BMI, All (mmHg) (mmHg) (mmHg) (kg/m2) African-Origin Populations Nigeria 121.5/72.4 122.2/73.0 121.0/71.9 22.9 Jamaica 122.9/71.7 122.5/72.0 123.2/71.5 27.0 US – Black 129.7/78.5 130.3/80.8 129.1/76.3 28.5 European-Origin Populations US – White 120.9/75.2 123.4/78.2 118.3/72.2 27.3 Canada 128.2/80.8 131.2/83.2 125.1/78.5 26.8 Italy 129.8/83.1 132.4/85.4 127.2/80.7 26.4 Sweden 130.6/80.9 133.0/83.4 128.3/78.4 26.5 England 135.0/77.2 137.3/80.3 132.7/74.2 27.1 Spain 131.4/83.2 132.3/83.9 130.5/82.5 27.4 Finland 134.3/83.8 136.9/86.0 131.6/81.5 27.1 Germany 138.0/86.4 139.5/88.5 137.3/84.3 27.3 * Age-adjusted Figure 1 Mean Systolic Blood Pressure, African Descent Populations; By Age Group Figure 2 Mean Systolic Blood Pressure, European Descent Populations; By Age Group Figure 3 Mean Systolic Blood Pressure, African and European Descent Populations; By Age Group Hypertension prevalence Hypertension prevalence, which accounts for the effect of treatment, follows a similar pattern although the east-west gradient among the African-origin groups is more consistent (Table 3 , Figure 4 ). Among the 14 populations, US blacks fall near the middle in terms of prevalence (mean prevalence = 37%, U.S. blacks = 44%). Among those above the mean, all but one is of European origin. Important differences are apparent in the gender-specific prevalence hypertension in these groups. Among Jamaican women hypertension was substantially more common than among Jamaican men (32% vs. 23%), and relative gender equality existed for US blacks. In Europe, however, the prevalence of hypertension was higher among men in every country (range 5–10%). Figure 4 Hypertension Prevalence (140/90 mmHg or Treatment), African and European Descent Populations; Ages 35–64, Age Adjusted Hypertension prevalence and obesity The only etiological factor on which standardized information was available was obesity, measured by its proxy BMI. The correlation between average BMI and hypertension prevalence was 0.6 (p < 0.01), all populations combined. Within the black populations the same correlation was observed between mean BMIs and hypertension prevalence (r = 0.6). Among whites, however, the relationship was weaker (r = 0.3). Of course, since obesity will be correlated with many other aspects of lifestyle, it is difficult to infer whether weight gain itself is playing a less important role in determining the variation among white populations. The contrasts noted above in hypertension prevalence by gender are consistent with the relative excess of obesity in women compared to men among Jamaicans and US blacks [ 2 , 6 ]. Discussion Comparisons of BP distributions across populations are made difficult by the requirement of comparability of the survey methods. In the last two decades, however, adoption of standardized protocols along with rigorous training have greatly improved the quality of epidemiological studies of hypertension [ 6 , 17 - 19 ]. A number of countries now conduct recurring national surveys that monitor both secular trends and regional variation within the country [ 10 - 12 , 19 , 20 ]. While independent surveys from the same base population had given divergent results in the past, at least two recent single-community studies conducted in the US provided estimates virtually identical to NHANES [ 21 , 22 ]. Although this evidence does not diminish the requirement of careful assessment of survey methodology before making comparisons, it does demonstrate that reliable information can be obtained from independent studies. The data presented here demonstrate a two-fold variation in prevalence of hypertension in both European- and African-origin populations. The prevalences are similar in blacks in the US and whites in Europe, although important gender differences are apparent. Although not a systematic sample, the populations that are included generally reflect the characteristic social setting in which these groups are found around the world. Summed across all groups, the white populations on average have a substantially higher burden of hypertension. This result can be attributed in large part to the inclusion of several black samples from developing countries where risk factors for hypertension are presently at a lower level. In the only head-to-head comparison within the same survey, US blacks have a prevalence that is 50% higher than among whites. Data from the UK, including the national survey, also demonstrate higher BPs and more hypertension among blacks of Caribbean and African descent [ 23 - 27 ]. On the whole, however, the published literature on racial disparities in hypertension from the UK is less consistent than in the US, where essentially every study has reported higher rates among blacks [ 28 ]. Surveys from Cuba, Trinidad and Brazil have also shown a smaller black-white gradient in BP than found in North America [ 29 - 31 ]. Are these findings merely artifactual, reflecting either methodological error or the sampling process? The most unexpected features of the data presented here are the high rates of hypertension in Europe, when contrasted to whites in Canada and the US. These results have been reported in greater detail in an earlier publication [ 16 ]. It is beyond the usual standard of statistical significance for the six European surveys to be higher by chance than both of those in North America (p < 0.05). As previously demonstrated, mortality rates for stroke – the most sensitive vital statistics indicator of uncontrolled high BP – are strongly correlated with the prevalence of hypertension among these countries ('r' = 0.8) [ 16 ]. Although the data are more limited, hypertension appears to be even more common in Eastern Europe [ 32 - 34 ]. In a comparison of Pol-MONICA with the US-based ARIC study, systolic BPs in Poland were 20 mmHg higher than in the US [ 3 ]. The primary purpose of this analysis was to provide descriptive results and very limited information was available on factors that might explain the findings we observed. The gradient among the black populations is consistent with the transition to an industrialized lifestyle and is thereby collinear with most known risk factors [ 6 ]. BMI is serving as an effective proxy for this relationship, although its independent contribution cannot be quantified. The explanation of the European-North American contrasts among the white populations is not as apparent. As we have discussed elsewhere, either known risk factors other than obesity are having a larger impact at the population level than usually appreciated, or unknown factors are at work [ 16 ]. In either case, further examination of this question seems justified. Treatment guidelines and practice patterns vary widely among these countries [ 16 - 19 ]. Widespread treatment could, of course, alter the mean BPs in a population, although this effect would be confined to persons over 55 where hypertension is common. The US has the highest rate of treatment, with about 25% of hypertensives controlled, compared to 10% in Europe and less than 1% in Africa (with hypertension defined as 140/90 mmHg)[ 16 ]. Any biases that would be introduced into the cross-national comparisons by differential treatment and control are insufficient to alter the primary conclusions, however. The virtual absence of treatment in rural Africa would mean that the natural distribution has essentially been observed unaltered. The effect of treatment in the US or Canada would not be apparent in younger individuals, where contrasts in BPs with Europe and Africa are equally large. If the North American-European contrasts are occurring in genetically homogeneous populations, large environmental influences must be at work that are not apparent on the surface. A similar process could be taking place across the social environments into which persons of African origin are assorted within societies such as the US and the UK. The debate over inherent susceptibility cannot be resolved with these data since neither the genetic nor the environmental influences can be held constant, allowing a test of the relative influence of the other factor. In fact, the question of inherent susceptibility is probably non-testable under any circumstances [ 35 - 37 ]. While the assumption is often made that contrasting environmental influences between blacks and whites can be adjusted by using proxy measures such as education, that assumption does not hold up under close examination [ 38 ]. Perhaps more to the point, however, these data demonstrate that the consistent emphasis given to the genetic elements of the racial contrasts may be a distraction from the more relevant issue of defining and intervening on the preventable causes of hypertension, which are likely to have a similar impact regardless of ethnic and racial background [ 39 ]. Once the problem of ethnic/racial contrasts is characterized more closely as a special instance of environmental influences at the population level, it could become more tractable in both the realms of research and practice. Competing interests The author(s) declare that they have no competing interests. Authors' contributions RC, KWM, AL and JB were responsible for study concept, design and supervision. RC, KWM, JB, SG, MJ, AA, TF MK, PP, BS and MT were involved in data acquisition. RC, KWM, JB, MJ, PP and AL were responsible for analysis and interpretation of data. RC and KWM drafted the manuscript. RC, JB, SG, AL, AA, TF, MJ, MK, PP, BS and MT were involved in critical revision of the manuscript for important intellectual content. Statistical expertise was provided by MJ. Administrative, technical and material support was provided by KWM, MK, and BS. Table 1 Characteristics of the Surveys Country Survey Yr(s) Population N Male (%) Participation Rate (%) Age Range Sampling Method* Nigeria 1991–93 Local 1931 45 NA 25–74 Multistage, address Jamaica 1993–95 Local 2573 41 65 25–74 Multistage, address USA, NHANES Black 1988–94 National 5283 45 82 18–80+ Multistage, population registry Canada 1986–92 National 23129 49 77.5 18–74 Multistage, medical insurance registries England 1998 National 11884 45 87.5 16–80+ Multistage, post code address Finland 1997 National 7064 47 72 25–64 Population registry Germany 1997–99 National 7047 49 61.4 18–79 Population registry Italy 1998 National 8233 50 - 35–74 Multistage, population registry Spain 1990 National 2021 40 73 35–65 Multistage, national registry Sweden 1999 Regional 1823 49 72 25–74 Population registry USA NHANES White 1988–94 National 7252 46 82 18–80+ Multistage, population registry * All stratified sampling methods Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545060.xml
359389	Reason as Our Guide	Two scientist members of the President's Council on Bioethics express their concerns about two recently issued reports by the Council in which the science is presented incompletely and myths are perpetuated	We are two of the scientist members of the President's Council on Bioethics. In late 2001, we were invited by the President of the United States to serve on this Council. The Bioethics Council was appointed by the President to “monitor stem-cell research, to recommend appropriate guidelines and regulations, and to consider all of the medical and ethical ramifications of biomedical innovation…. This council will keep us apprised of new developments and give our nation a forum to continue to discuss and evaluate these important issues.” This was a difficult invitation to accept. On the one hand, the President's views on the use of human embryonic stem cell research and somatic cell nuclear transfer techniques were well-known and in conflict with our own beliefs about the costs and benefits of the use of progressive technologies to advance biomedical research. On the other hand, we were grateful that the President, despite his views in opposition to these therapies, was willing to invite serious biomedical scientists to help formulate advice to him—and ultimately to contribute to the development of national policy—on these critically important advances. We knew that on this originally 18-member (but for most of the past two years a 17-member) Council, as scientists we would be in the minority in our belief of the good to be gained through these and other areas of biomedical research. We were also aware that some others on the Council had strong opposing views. Thus, it was only with the assurances of the Council chairman, Leon Kass of the University of Chicago, and of the President of the United States himself that we were persuaded that our voices would be heard and integrated into the statements of the Council. Furthermore, we felt, and continue to feel, that bioethical issues are important not only to all biologists, but also to society at large, and thus especially worthy of engaging debate and discussion. Two recently issued reports of the Council, “Beyond Therapy: Biotechnology and the Pursuit of Happiness” ( http://bioethics.gov/reports/beyondtherapy/index.html ) and “Monitoring Stem Cell Research” ( http://bioethics.gov/reports/stemcell/index.html ), are therefore of deep concern to us. We discuss them in turn below. Concerns about the “Beyond Therapy” Report The “Beyond Therapy” report deals with issues of direct concern for every thoughtful person. However, in the interests of setting straight the record of our views, as Council members and scientists, on the content of this report and for a proper assessment of the scientific content of the “Beyond Therapy” report, we feel it is important to point out aspects of the report for which we had requested revisions and for which those requests were declined. In the discussions of preimplantation genetic diagnosis, the specter of designer babies is raised by implying that selecting embryos for intelligence and other traits, such as temperament is a possibility. Scientifically, this simply is highly unlikely and indeed may not even be feasible. While such scientific unlikelihood is mentioned in passing in the report, it is easy to take away from the report the feeling that such genetic manipulation will happen and is even imminent. The report also claims that “the underlying impulse driving age-retardation research is, at least implicitly, limitless, the equivalent of a desire for immortality.” Furthermore, the title of Chapter 4 of the report, “Ageless Bodies,” implies that immortality is the goal of this research, despite all reliable scientific evidence to the contrary. Such a title is not consistent with the knowledge, stated in that chapter, that there is no scientific basis for immortality and implies that, by seeking to maintain and extend “youth,” research into aging, including stem cell research, is predominantly to serve vanity. Also, without presenting scientific or reliable evidence, the report presents the opinion that research into prolonging healthy life may result in a lifetime obsession with immortality. Hence, this chapter in the report falls short of explaining the serious challenge of preventing and curing age-related disease to extend health—very different from attempting immortality. The same chapter offers a sensational quote from a researcher that “the real goal [of aging research] is to keep people alive forever.” The request that quotes from researchers more representative of the biomedical research community also be included was declined. This leads to a misleading misrepresentation of the motivation of reputable researchers in the field of aging. In suggesting that slowing biological aging may increase the disjunction between “social aging” (the age at which children are exposed to “adult” images and concepts) and “biological aging” (expected lifespan), only one view, a conservative one, of the supposed “best” way to raise children is presented. The report also suggests, with no clear reasoning behind it, that longer lives will somehow undermine human determination to contribute as much as one can during a lifetime. Despite requests for inclusion of material that would allow for a balanced treatment of these topics, the report minimized discussion of potential positive aspects of slowing biological aging, such as prolonged good health. Finally, the report repeatedly emphasizes a “profound and mysterious” link between longevity and fertility, thereby leaving the reader with the distinct but erroneous impression that anything done to extend healthy life will be traded for decreased fertility, despite the fact that current scientific literature, which was made available for inclusion in the report, shows a lack of any necessary mechanistic linkage of the two. Concerns about the “Monitoring Stem Cell Research” Report With respect to the “Monitoring Stem Cell Research” report, we feel that some facts that would help the public and scientists better assess the content of the report were not brought out clearly or were omitted entirely. First, from the published scientific literature in peer-reviewed journals on stem cells, a major message can be distilled: namely, the vast difference that currently exists in our understanding of, and the potential utility of, embryonic versus adult stem cells as sources of material for research and clinical purposes. In brief, human stem cells have been isolated from a variety of embryonic, fetal, and adult tissue sources. However, enormous differences exist in purity, properties, data reproducibility, and understanding of cells from these different sources. Much of our ignorance is related to the relative paucity of funding for research using embryonic stem cells. Years of rigorous and careful research in animal models have documented that embryonic stem cells have great utility for scientific studies. This work has also rigorously and reproducibly established the great plasticity of these cells and supports the opinion that human embryonic stem cells possess the greatest broadest potential and promise for clinical applications. As well as therapeutic uses, important potential applications include studies of embryonic stem cells bearing complex genotypes susceptible to poorly understood common human diseases and testing and screening drug efficacy. The report does not make clear that the best-characterized adult stem cells are hematopoietic stem cells. Currently, major difficulties and inadequate understanding exist with most other types of adult stem cells reported to date. In addition, many experiments suggesting that adult stem cells have broad plasticity may be incorrectly interpreted owing to an error caused by an experimental artifact of cell fusion present in some unknown proportion of the experiments. Research on some of the reported adult stem cell preparations may conceivably in the future demonstrate that they, too, like hematopoietic stem cells, can also be prospectively identified, “single cell cloned,” expanded considerably by growth in vitro with retention of normal chromosome structure and number, and preserved by freezing and storage at low temperatures. But it should be strongly cautioned that this has not been done for most adult stem cell preparations, and, even if possible, it is not clear that any of the just-mentioned procedures will be accomplished in the near future, owing to the technically very demanding nature of such experiments. We feel it is important to emphasize a point that the report mentioned, that the reported isolation and properties of multipotent adult progenitor cells (MAPCs) must be reproduced in additional laboratories for any reliable interpretation of the results reported with these cells. After considerable effort, this has still not been achieved. Thus, in the reported results, the possible significance of the reported isolation and properties of human MAPCs is left unclear, as is their potential as a source of stem cells for clinical purposes. Hence, a strong overall caution is that many of the reports on the properties of cells differentiated from adult stem cell preparations are to date preliminary and incomplete. If results with any isolated and characterized adult stem cells are validated, it will then be very important to compare their properties—and those of any more differentiated cells that can be derived from them—with other stem cell sources, such as the well-characterized hematopoietic stem cells, and with human embryonic stem cell preparations. Two major considerations argue strongly for non-commercial, federal, peer-reviewed funding to be made available for this work. The first is the sustained effort this work will require. The second is the importance of reliable and unbiased design of experiments and of open, public availability of the complete findings. Reasons for Our Concern In being concerned about the content of these reports, neither of which makes any recommendations for legislative or policy actions, are we worrying too much? We think not. Indeed, already, sadly as a result of the way the sections on aging research in the report were written, the myth that longevity has an inevitable tradeoff of diminished fertility is now gaining a further foothold: witness the January 26, 2004, issue of the The New Republic . In it, an article about this report of the Council falls right into the trap: it states, “But changes come with longer life. Worms and mice that are altered for extended lifespans become sterile, or barely reproduce.” The public is done a disservice when science is presented incompletely; myths are then perpetuated. This is but one example of the dangers that three of the Council members who are scientists (the two of us along with Michael Gazzaniga of Dartmouth College) pointed out, in a Commentary within the edition of the “Beyond Therapy” report published by the Dana Foundation in November 2003. In that Commentary, we stated that “Our concern … is that, moving forward, the debate carry on with all of the scientific evidence—or as much as such a widespread public discussion can include—and take care not to leave an erroneous impression as to the nature of the potential problems at hand.” We ended the Commentary by saying “We urge both good reading and critical reading!” (our italics). These reports had as their premise the aim of neutrality in the scientific analysis of the issues addressed. But our concern is that some of their contents, as in the few examples outlined above, may have ended up distorting the potential of biomedical research and the motivation of some of its researchers. Continuing discussions will form the basis for future decisions on these topics; keeping such discussion open and balanced is of paramount importance. Box 1. President's Council on Bioethics The President's Council on Bioethics was created on November 28, 2001. Its mission includes: to “advise the President on bioethical issues that may emerge as a consequence of advances in biomedical science and technology. In connection with its advisory role, the mission of the Council includes the following functions: to undertake fundamental inquiry into the human and moral significance of developments in biomedical and behavioral science and technology; to explore specific ethical and policy questions related to these developments; to provide a forum for a national discussion of bioethical issues; to facilitate a greater understanding of bioethical issues; and to explore possibilities for useful international collaboration on bioethical issues.” From Executive Order 13237 George W. Bush The White House, November 28, 2001 Federal Register date: November 30, 2001 Federal Register page: 66 FR 59851 The members of the President's Council on Bioethics at the time these reports were written included Leon R. Kass, M.D., Ph.D. (Chair), American Enterprise Institute; Elizabeth H. Blackburn, Ph.D., University of California, San Francisco; Rebecca S. Dresser, J.D., M.S., Washington University School of Law; Daniel W. Foster, M.D., University of Texas, Southwestern Medical School; Francis Fukuyama, Ph.D., Johns Hopkins University; Michael S. Gazzaniga, Ph.D., Dartmouth College; Robert P. George, J.D., D.Phil., Princeton University; Mary Ann Glendon, J.D., M.Comp.L., Harvard University; Alfonso Gómez-Lobo, Dr. Phil., Georgetown University; William B. Hurlbut, M.D., Stanford University; Charles Krauthammer, M.D., syndicated columnist; William F. May, Ph.D., Southern Methodist University; Paul McHugh, M.D., Johns Hopkins Hospital; Gilbert C. Meilaender, Ph.D., Valparaiso University; Janet D. Rowley, M.D., University of Chicago; Michael J. Sandel, D.Phil., Harvard University; and James Q. Wilson, Ph.D., University of California, Los Angeles. * These members had their Council terms terminated by White House directive on February 27, 2004.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC359389.xml
368168	Protein Nanomachines	At the interface of biology and nanotechnology lies an area of research that aims to construct molecular-scale machines based on protein and nucleic acid	In 1959 Richard Feynman delivered what many consider the first lecture on nanotechnology. This lecture, presented to the American Physical Society at the California Institute of Technology, prompted intense discussion about the possibilities, or impossibilities, of manipulating materials at the molecular level. Although at the time of his presentation, the manipulation of single molecules and single atoms seemed improbable, if not impossible, Feynman challenged his audience to consider a new field of physics, one in which individual molecules and atoms would be manipulated and controlled at the molecular level ( Feynman 1960 ). As an example of highly successful machines at the “small scale,” Feynman prompted his audience to consider the inherent properties of biological cells. He colorfully noted that although cells are “very tiny,” they are “very active, they manufacture various substances, they walk around, they wiggle, and they do all kinds of wonderful things on a very small scale” ( Feynman 1960 ). Of course, many of these “wonderful things” that he was referring to are a result of the activities of proteins and protein complexes within each cell. The field of nanotechnology has indeed emerged and blossomed since Feynman's 1959 lecture, and scientists from many disciplines are now taking a careful look at the protein “machines” that power biological cells ( Drexler 1986 ). These “machines” are inherently nanoscale, ranging in width from a few nanometers (nm) to over 20 nm, and have been carefully refined by millions of years of evolution. As a graduate student in molecular biology, I have been especially interested in creative approaches to bridging the fields of biology and nanotechnology. Both DNA and protein molecules possess a number of intrinsic characteristics that make them excellent candidates for the assembly of dynamic nanostructures and nanodevices. Properties such as the site-specific molecular recognition among interacting protein molecules, the template-directed self assembly of complementary DNA strands, and the mechanical properties of certain protein complexes have enabled bionanotechnologists to envision a molecular world built “from the bottom up” using biological-based starting materials. In my own research, I have been very interested in investigating protein interactions and protein pathways on a genome-wide scale. In many ways, protein pathways are analogous to nanoscale “assembly lines,” since protein pathways often involve a series of proteins that act in successive order to yield a particular molecular “product” or perform a particular molecular function. While these protein-based “assembly lines” are commonplace within biological cells, they prompt two interesting questions with respect to the field of nanotechnology. First, can we mimic these multicomponent protein-based “assembly lines” on nanofabricated surfaces? And, second, can we tailor these “nanoscale assembly lines” to perform new and unique tasks? Nanomechanical protein complexes, such as the rotary ATP synthase complex, have also generated much interest from a nanotechnology standpoint ( Soong et al. 2000 ). These protein complexes enable highly controlled mechanical motion at the nanoscale and may some day lead to novel rotary machines that function as molecular motors for a variety of nanoscale applications. In order to fully exploit these nanoscale protein machines, it is of prime importance to be able to position individual proteins and protein complexes at the nanoscale. Progress in this area has recently been reported by Yan et al. (2003) , who developed a method to construct two-dimensional protein arrays using DNA-directed templates. Building on work pioneered by Nadrian Seeman ( Seeman 2003 ), Yan et al. constructed two-dimensional DNA “nanogrids” by exploiting the pairing that occurs between complementary DNA strands ( Figure 1 ). The two-dimensional DNA nanogrid exhibits a repeating periodic structure ( Figure 1B ) due to the inherent qualities of the individual DNA tiles that make up the nanogrid ( Figure 1A ). The distance between adjacent tile centers is approximately 19 nm (approximately 4.5 turns of the DNA double helix plus the diameter of two DNA helices). Figure 1 Self-Assembled DNA Nanostructures (A) DNA “tile” structure consisting of four branched junctions oriented at 90° intervals. These tiles serve as the primary “building block” for the assembly of the DNA nanogrids shown in (B). Each tile consists of nine DNA oligonucleotides as shown. (B) An atomic force microscope image of a self-assembled DNA nanogrid. Individual DNA tiles self-assemble into a highly ordered periodic two-dimensional DNA nanogrid. (Images were kindly provided by Thomas H. LaBean and Hao Yan.) Yan et al. utilized these DNA nanogrids to assemble periodic protein nanoarrays. The DNA nanogrid, in this case, served as a molecular scaffold for the self assembly of protein molecules into ordered arrays. In order to control the location of protein assembly, Yan et al. first tethered a covalently linked biotin moiety to the central region of each DNA tile. The biotin was covalently linked to one of the DNA strands at the position corresponding to the center of the tile. This design resulted in a uniform array of biotinylated tiles, with each biotin moiety separated by about 19 nm. The authors then added streptavidin, a protein that has a strong binding affinity for biotin, to form a periodic streptavidin protein array on top of the biotinylated DNA lattice. The resulting array represents the first periodic, self-assembled DNA lattice in which individual protein molecules are precisely positioned into a periodic array with nanometer dimensions. It is interesting to consider some of the applications of self-assembled protein arrays. Soong et al. (2000) demonstrated that the ATP synthase protein complex could be used to power the rotation of an inorganic nickel “nanopropeller.” ATP synthase is a multisubunit protein complex with a domain that rotates about its membrane-bound axis during the natural hydrolysis of ATP within a cell. Soong et al. attached a nanoscale inorganic “propeller” to the rotary stalk of ATP synthase, creating a “rotary biomolecular motor.” It is intriguing to consider the construction of an ordered array of ATP synthase driven nanomachines, each positioned precisely along a DNA scaffold, similar to that described by Yan et al. Such an assembly, combined with proposed “nanogears” ( Han et al. 1997 ), may one day enable the construction of nanoscale variations of the traditional “gear-train” and “rack-and-pinion” gearing systems. Construction of such systems may facilitate the design of machines that can transmit and transform rotary motion at the nanoscale. In addition to rotary biomolecular motors, proteins that undergo substantial conformational changes in response to external stimuli might also find some interesting uses in nanoarrays. Dubey et al. (2003) are working on methods to exploit the pH dependent conformational changes of the hemagglutinin (HA) viral protein to construct what they term viral protein linear (VPL) motors. Proteins that undergo substantial conformational changes in response to environmental stimuli may facilitate the design of nanoscale machines that produce linear motion ( Drexler 1981 ), as opposed to rotary motion. At neutral pH, the HA 2 polypeptide forms a compact structure composed of two α-helices folded back onto each other. At low pH, HA 2 undergoes a substantial conformational change, which results in a single “extended” helix. This conformational change results in a linear mechanical motion, with a linear movement of approximately 10 nm ( Dubey et al. 2003 ). It would be interesting to investigate the applications of ordered arrays of dynamic VPL motors, since an array of such “hinge” structures may enable the coordinated linear movement of hundreds of tethered macromolecules in a synchronous manner. The work of Yan et al. (2003) has opened up exciting new avenues in the field of nanotechnology and has provided the molecular framework for the construction of dynamic protein-based assemblies. It is foreseeable that variations of these same DNA scaffolds will eventually be used for the design and construction of more complex protein-based assemblies, such as nanoscale “assembly lines” or periodic arrays of dynamic motor proteins. This work is important to me because it demonstrates not only that it is possible to create uniform arrays of protein biomolecules using biomolecular scaffolds, but the study also emphasizes the important role that molecular biology will undoubtedly play as the field of nanotechnology matures. As the field of nanotechnology continues to evolve, it is likely that we will see many more nanotechnology applications utilizing biological macromolecules. Toward the end of Richard Feynman's 1959 lecture, he quipped, “What are the possibilities of small but movable machines? They may or may not be useful, but they surely would be fun to make.”	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC368168.xml
406398	Learning to Change	A paper published over 20 years ago by Susan Iversen and Mortimer Mishkin on reversal learning continues to inform cognitive neuroscience today	One of the hallmarks of human nature is our remarkably flexible behaviour, especially in the social domain, which is perhaps also a major reason for our relative evolutionary success. Our social skills are already being honed in childhood and early adolescence, when we quickly become very adept at forming and breaking alliances within and between groups and spend much of our time engaged in complex social interactions. At best, these interactions enrich our society; at worst, they become ‘Machiavellian’ and exploitative. While science might appear removed from such politics, many scientists would probably agree that science is in fact a social enterprise, sharing many characteristics with other human pursuits, and that any claim to greater scientific truth can only be accorded over decades, even centuries. I have always been fascinated by social intelligence, particularly of the ‘Machiavellian’ kind, and found myself wondering at the start of my doctoral research how one might use neuroimaging to study social intelligence in the human brain. I was also interested by the fact that some of this flexible behaviour is shared with other primates such as chimps, bonobos, and even monkeys, who also spend inordinate amounts of time in social interactions, working out social hierarchies. However, it was not immediately obvious how one might go about designing experiments that would address these somewhat intangible issues of social behaviour. Trawling the scientific literature, I came across the concept of reversal learning. While it is obviously important that we can learn arbitrary associations between stimuli and actions, it is also extremely important that we can relatively easily break these associations and learn others. If we learn that choosing a certain object leads to a reward, it would be rather maladaptive to keep choosing this object when it was no longer associated with a reward but, say, a punishment instead. In order to accommodate complex behaviour, we need to be able to adapt or reverse the learning patterns when things change. For a long time, it was thought that complex behaviour depended crucially on the prefrontal cortex of the brain, but it was not clear which parts were important for reversal learning. This was investigated in a classic paper by the eminent neuroscientists Susan Iversen and Mortimer Mishkin (1970) , who studied lesions in monkeys, with elegant and important results. The authors lesioned discrete parts of the prefrontal cortex in different monkeys and showed convincingly that these lesions had differential effects on the animals' ability to reverse rewarding associations in an object reversal task. When the inferior prefrontal convexity and parts of the lateral orbitofrontal cortex (which is the ventral part of the prefrontal cortex over the orbits) (see Figure 1 ) were lesioned, the monkeys became significantly impaired with respect to object reversal learning. Specifically, they continued to respond much longer than controls to an object that was no longer rewarded on the first reversal trial. Figure 1 Reversal Learning and the Orbitofrontal Cortex (A) Lateral and ventral views of the surface reconstructions of the lateral and medial orbitofrontal cortex lesions in monkeys (adapted from Iversen and Mishkin 1970 ), with the former monkeys having difficulty with the reversal task. (B) A ventral view of the human brain, with the cerebellum removed. Red activations in the lateral orbitofrontal cortex indicate the maximal activation for reversal compared to stable acquisition events. Blue activations indicate the main effects of facial expression (adapted from Kringelbach and Rolls 2003 ). This was not the case for monkeys who had had the medial parts of the orbitofrontal cortex lesioned. These monkeys were not completely unaffected by the lesion, but showed moderate impairment on all but the first of the object discrimination reversals. Furthermore, they had moderate difficulty withholding response between trials on an auditory differentiation task. These results strongly suggested a differential role for the lateral and medial parts of the orbitofrontal cortex. Although the paper was not published in a high-profile journal, this elegant and very significant result has had a huge influence on subsequent research. The paper, like many other great papers, was ahead of its time, and it took almost a decade before the citations started to pick up (at last online count, on February 1, 2004, of the ISI database, the paper had generated 229 citations since 1981). Iversen and Mishkin (1970) persuasively demonstrated the importance of the orbitofrontal cortex in reversal learning, and other studies have since extended this result in nonhuman primates. One study demonstrated that single neurons in the macaque orbitofrontal cortex change their responses to a visual cue after a single trial in which the reward association of the visual cue is reversed ( Thorpe et al. 1983 ). Another lesion study in marmosets by Dias et al. (1996) found that the orbitofrontal cortex is essential for the performance of emotion-related reversal learning tasks. There was also some evidence that humans with lesions to the orbitofrontal cortex have problems with reversal learning, but the lesions, caused by neurological insult, were not very clean or focal ( Rolls et al. 1994 ). In addition, it had also become clear that lesions to the orbitofrontal cortex were associated with impairments in emotional and social behaviour, characterised by disinhibition, social inappropriateness, and irresponsibility ( Anderson et al. 1999 ). These interesting but nonconclusive results in humans spurred us on to use neuroimaging on a modified version of a probabilistic reversal learning task designed by Julia Hornak and John O'Doherty ( Hornak et al. 2004 ), whose preliminary data suggested that patients with surgical lesions to the orbitofrontal cortex were impaired. The subjects' task was to determine, by trial and error, which of two stimuli was the more profitable to choose and to keep track of this, reversing their choice when a reversal occurred. By design, the actual reversal event was not easy to determine, since ‘money’ could be won or lost on both stimuli, but a choice of the rewarding stimulus would in general give larger rewards and smaller punishments. The converse was true of the punishing stimulus; losing a large amount of money would often (but not always) signal that a reversal had occurred. We used functional magnetic resonance imaging to show that dissociable activity in the medial orbitofrontal cortex was correlated with the magnitude of the monetary gains received, while activity in the lateral orbitofrontal cortex was correlated with the monetary losses incurred ( O'Doherty et al. 2001 ). This dissociation between the functions of medial and lateral orbitofrontal cortex seemed to mirror Iversen and Mishkin's initial dissociation in monkeys, in which the lateral orbitofrontal cortex was linked, in both cases, to the reversal trials. However, owing to the probabilistic nature of the task, in which receiving a monetary punishment did not always signal reversal, our imaging study did not reveal the cortical localisation of reversal trials. In addition, our task used money as the secondary reinforcer, which might be a powerful influence on humans but has little biological relevance for other animals, and certainly none in the social domain that I was interested in. One way to solve these problems was to use facial expressions rather than money as the reinforcing stimuli. This made sense, given that the key to social intelligence is the ability to detect subtle changes in communication and act upon these changes rapidly as they occur. Such changes in social behaviour are often based on facial expression and come so naturally to humans (and are in place so early in child development) that some might argue that this functionality is essentially innate. However, our human social behaviour is sufficiently flexible that we can easily learn to adapt our behaviour to most facial expressions. For example, other people's neutral expressions do not normally indicate that our behaviour should change, but it is easy to think of social contexts in which a neutral expression does indeed imply that our current behaviour is inappropriate and should change. I designed a reversal task in which the subject's overall goal was to keep track of the mood of two people presented in a pair and, as much as possible, to select the ‘happy’ person, who would then smile. Over time, the person with the ‘happy’ mood (who would smile when selected), changed his/her mood to ‘angry’. This person thus no longer smiled when selected, but instead changed to a facial expression that signalled that he/she should no longer be selected. In the main reversal task, the facial expression used to cue reversal was an angry expression (the most natural facial expression to cue reversal), while in the second, control, version of the reversal task, a neutral expression was used. By using two different reversal tasks in which different facial expressions signalled that behaviour must change, we were able to determine which brain areas were specific to general reversal learning, rather than just to reversal following a particular expression, such as anger. We used functional magnetic resonance imaging to show that the ability to change behaviour based on facial expression is not reflected in the activity of the fusiform face area (which invariably appears to reflect only identity and not valence), but that general reversal learning is specifically correlated with activity in the lateral orbitofrontal and anterior cingulate/paracingulate cortices (as well as other brain areas, including the ventral striatum and the inferior precentral sulcus) ( Kringelbach and Rolls 2003 ). This result confirmed and extended the results from Iversen and Mishkin's original paper. Further confirmation came from the neuropsychological testing, carried out by Julia Hornak on human patients with surgical lesions to the orbitofrontal cortex, which showed that bilateral (but not unilateral) lesions to the lateral orbitofrontal cortex produce significant impairments in reversal learning ( Hornak et al. 2004 ). Yet, as always, these results are not conclusive and raise many new issues. It is, for instance, not presently clear what other areas of the brain are necessary and sufficient for reversal learning. Among the other brain areas we found relating to general reversal learning in our study, the ventral striatum is, for instance, an obvious candidate ( Cools et al. 2002 ). In addition, functional magnetic resonance imaging is essentially a correlative technique, with poor temporal information, which makes it very difficult to infer causal relations between brain regions. Thus, further investigations, e.g., with magnetoencephalography, will still be required to gain temporal information on the milliseconds scale. I take heart from a friend, a very distinguished scientist, who states that the price for having spent a lifetime in cutting-edge research is that 99% of his (and other scientists') research is wrong—perhaps not completely wrong, but certainly wrong in the details. I would like to think that the original result from the Iversen and Mishkin paper is among the rare 1%, but the trouble with such foresight is that it lacks the vantage point of true hindsight. In his masterpiece, The Prince , Niccolò Machiavelli offers a rather pessimistic view on human nature, in which ‘love is held by a chain of obligation which, since men are bad, is broken at every opportunity for personal gain’. It may be that our capacity for rapid reversal learning is sometimes used for less than noble pursuits, both in science and in interpersonal relations in general, but we would be in real trouble if we couldn't learn to change.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC406398.xml
529421	Enzyme Replacement in Gaucher Disease	The development of enzyme replacement therapy for Gaucher disease was a triumph of translational medicine. What were the key steps in its development? What are the controversies surrounding its use?	Gaucher disease is the most common lysosomal storage disorder ( Box 1 ). A deficiency of the enzyme glucocerebrosidase ( Figure 1 ) causes accumulation of the glycolipid glucocerebroside in macrophages throughout the body. In the viscera, glucocerebroside arises mainly from the biodegradation of red and white blood cells. In the brain, glucocerebroside arises from the turnover of complex lipids during brain development and the formation of the myelin sheath of nerves. The disease may be discovered as an incidental finding in the elderly because of mild thrombocytopenia or splenomegaly, or it may present early in life with hepatosplenomegaly, thrombocytopenia, anemia, and bone lesions. Box 1. What Is Gaucher Disease? Gaucher disease is an inherited metabolic disorder in which harmful quantities of a fatty substance called glucocerebroside accumulate in the spleen, liver, lungs, bone marrow, and, in rare cases, the brain. There are three common forms. Type 1 is the most common. Clinical features include easy bruising, anemia, low blood platelets, enlargement of the liver and spleen, bone disease, and, in some instances, lung impairment. There are no signs of brain involvement. Problems may begin early in life, be delayed until adulthood, or not occur at all. In type 2, liver and spleen enlargement are apparent by three months of age, and there is extensive and progressive brain damage. These patients usually die by two years of age. In type 3, liver and spleen enlargement is variable, and signs of brain involvement, such as seizures, become apparent gradually. Figure 1 Glucocerebrosidase Cleaves a Linkage within Glucosylceramide, a Normal Intermediate in Glycolipid Metabolism Until 1990, treatment consisted only of palliative measures such as splenectomy and hip replacement. The development of enzyme replacement therapy for Gaucher disease, that is, exogenous administration of the missing enzyme, is a triumph of translational medicine. At the same time, powerful commercial interests may have been influential in physicians adopting a high-dose rather than a low-dose treatment schedule. Moreover, the high cost of enzyme replacement therapy forces us to consider what society can afford in the way of palliative treatments for very rare diseases. The History of Enzyme Replacement Therapy The possibility that the therapeutic replacement of enzymes missing from lysosomes could be achieved was first raised by de Duve forty years ago when he wrote: “Any substance that is taken up intracellularly by an endocytic process is likely to end up within lysosomes. This obviously opens up many possibilities for interaction, including replacement therapy” [ 1 ]. Type 1 Gaucher disease, the most common type, seems a particularly suitable target for enzyme replacement therapy because of the lack of central nervous system involvement (visceral damage in Gaucher disease is reversible whereas the brain damage usually is not). By the 1970s, the underlying enzyme deficiency had been identified, and methods had been developed to purify the enzyme from human placenta in a high state of purity. Three groups of investigators then attempted to treat the disease by infusing exogenous enzyme. In the United States, at the National Institutes of Health in Bethesda, Maryland, the unaltered enzyme was infused directly into the venous circulation [ 2 ]; at City of Hope in Duarte, California, it was entrapped in red cell membranes coated with antibody in an effort to direct it to macrophages [ 3 ]. In Harrow, United Kingdom, the enzyme was delivered entrapped in liposomes [ 4 ]. Although some mildly encouraging results were achieved, it was clear that none of these approaches was likely to be translated into a useful treatment. The needed conceptual breakthrough was provided by the identification of a mannose receptor on macrophages and the suggestion that this might prove useful in replacement therapy for Gaucher disease [ 5 ]. This led to the development of a modified enzyme, processed to expose mannose, and to its production on an industrial scale from placentas. After the gene encoding the enzyme was cloned [ 6 ], a recombinant product became available. The Pivotal Study The first study of commercially produced mannose-enriched glucocerebrosidase was carried out in Bethesda, Maryland, on only 12 patients, presumably because of a limited supply of the enzyme [ 7 ]. Given this small cohort of patients, only a single dose (60 units/kg) was administered. This dose was given every two weeks to ten of the patients, while two patients received it weekly. This is manifestly an unusual dose schedule for a preparation with a circulating half-life of only about 12 min that is being targeted to a relatively small number of receptors. Many of the patients studied did not live near Bethesda, and it is likely that the dose schedule that was chosen was based on convenience rather than on sound pharmacokinetic principles. Since it was unlikely that a second study would be launched if the first failed, the investigators wisely used a very generous dose of enzyme to maximize the probability that the trial would be successful. Intravenous administration of the enzyme produced objective clinical improvement (such as reduced liver and spleen size and increased hemoglobin levels and platelet counts). The enzyme was promptly approved and marketed. Since only a single dose had been tested, this was the dose that most physicians administered in clinical practice. But the preparation was extremely costly—about US$4.00 per unit. At the dose used in the pivotal trial, a 70-kg patient would receive enzyme costing US$16,800 every two weeks. Dosage Considerations Visceral organ responses. But was the large dose given actually the dose required? There were no data, and many physicians were unwilling to give less than the dose that had been used in the pivotal trial. Moreover, since most physicians took care of only one or at most two patients with the disease, they were not in a position to perform a dose-ranging study. And industry had no interest in supporting studies to show that a lower dose yielded equivalent results. But clinical trials carried out in our National Institutes of Health–sponsored General Clinical Research Center quickly established that a quarter of the dose given at more frequent intervals was fully effective [ 8 ]. By 2000, a considerable body of data had accumulated, making it possible to perform meta-analyses of the relationship between the total monthly dose, the interval at which the dose is administered, and the decrease in the size of the liver. The results were clear ( Figure 2 ) [ 9 ]. Even a dose of only 15 units/kg/mo, one-ninth of the dose given in the pivotal trial, resulted in an excellent clinical response. Most patients were receiving a substantial overdose of an extremely costly preparation. The data indicate that when very large doses are administered, the two-week time interval is adequate to give an optimal response, but when more modest doses are administered, more frequent infusions greatly improve the response [ 9 ]. Figure 2 A Meta-Analysis of the Decrease in Liver Size in Patients with Gaucher Disease Documented for Various Doses of Enzyme Given As Replacement Therapy The individual studies included in the meta-analysis are listed in [ 9 ]. Recent “consensus recommendations,” which were supported in part by the Genzyme Corporation, the manufacturers of recombinant human glucocerebrosidase (imiglucerase, brand name Cerezyme), suggest that children be given an initial dose of 30 to 60 units every two weeks [ 10 ]. But there is no high-quality evidence that such a costly treatment regimen provides results superior to those achieved with smaller doses. The only support for recommending this high dosage comes from uncontrolled studies showing that in some children bone lesions may progress at low dosages. However, we know from our own published observations that skeletal progression and even fractures also occur in some individuals receiving high-dose therapy [ 11 ]. Thus, I would caution against any recommendations to give high-dose therapy that have not been based on well-designed randomized, controlled trials. Having said this, I recognize that most, but not all, of the patients that were included in our meta-analyses were adults, whereas the company-sponsored consensus recommendations refer to children. However, in the absence of any evidence-based rationale for administering large, costly doses of enzyme, I believe that the use of smaller, more frequent doses is the most prudent treatment approach. It is often assumed that patients with severe disease require larger doses of enzyme than those with mild disease, but a meta-analysis based on liver size or spleen size made it clear that this is not the case ( Figure 3 ) [ 12 ]. Large organs shrink more rapidly than smaller ones, and this is true regardless of the dose that is used [ 13 ]. Figure 3 A Meta-Analysis of the Decrease in Liver and Spleen Size in Patients with Gaucher Disease As a Function of the Initial Organ Size BW, body weight. Redrawn from [ 11 ]. Skeletal response. The response of enlarged viscera to enzyme infusion is much more rapid than the response of bones. In one early study, the large dose used in the pivotal trial was given for up to four years to patients with bone disease, and although the response was slow, gradual improvement occurred [ 14 ]. Strangely, the authors concluded that large doses were required—“strangely,” because they did not give smaller doses to any patient. Subsequently, it was shown that less than a quarter of the dose (only 30 units/kg/mo) produced an equivalent response [ 15 ]. Whom to Treat The severity of Gaucher disease is very variable. We have estimated that some 60% of patients homozygous for the common c.1226 C → G (N370S) mutation never come to medical attention [ 16 ]. Accordingly, many—possibly most—patients with Gaucher disease require no treatment. In adults, the disease is rarely progressive [ 11 , 17 ]. What you see is what you have, more or less. Bone fractures, of course, are not gradual events but sudden ones. But almost invariably they occur in patients who already have very substantial, demonstrable bone disease. In children, the situation is different, and progression is common. It is only with proper awareness of the natural history of the disease that one can make rational judgments regarding who needs treatment. Individualized Treatment Evaluating dose–response relationships in patients with Gaucher disease has been difficult for several reasons. The number of new patients requiring therapy is relatively small, and the Genzyme Corporation has done little to encourage the performance of dose–response studies, making it difficult to enroll patients. But beyond that, the response of patients to any dose is variable. Some authors have suggested that this may be due to individual differences in dose requirements—that some patients are relatively resistant and require a large dose, while others do well on a small dose [ 18 ]. This is an attractive concept, but is it correct? Another meta-analysis indicates that it is not. Rather, there are patients who respond poorly to any dose and others who respond well to any dose [ 19 ]. Moreover, quadrupling the dose does not increase the rate of response [ 11 ]. What Does the Future Hold? The quality of life for patients with Gaucher disease has been greatly improved by the development of enzyme replacement therapy. Manufacturing and selling the enzyme has also been enormously profitable for industry. This profitability has served as a stimulus for the development of enzyme replacement treatments for diseases less common and generally less responsive to treatment than Gaucher disease. Given the small target population, these treatments are enormously costly on a per-patient basis. Treatments for Fabry disease and Hurler-Scheie disease (also called mucopolysaccharidosis I) are already licensed, and others are on the way [ 20 , 21 , 22 ]. This brings us face-to-face with a major ethical dilemma. We do not put a price on human life. Yet health-care resources are a zero-sum game. What is spent on one disease cannot be spent on another. Is it better to treat one child with Hurler-Scheie disease [ 22 ] or to provide good prenatal care to 100 women who might not otherwise obtain it, or for that matter, to feed 1,000 malnourished children? These are difficult decisions that will be forced on us as enzyme replacement and other high-technology therapies come of age.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529421.xml
521727	Breaking Down the Stereotypes of Science by Recruiting Young Scientists	Thomas Jefferson University Science Outreach Program brings the scientific method into the classroom	If you ask the average ten year old in America what a scientist looks like, they almost always describe an older man with crazy white hair and a lab coat. If you ask a group of adolescents how many have looked through a microscope, few raise their hands. If you discuss the implications of genetic research with a group of high school students, they're likely to cut your next class. The reason why these students have such profound stereotypes of scientists and are less than enthusiastic about science's impact on society is simple—the lack of exposure they receive during their pre-college education. According to a preliminary study conducted at Leicester University in England, students are often repeatedly confronted with stereotypes of science and scientists via television, cartoon, and comic book characters as well as uninformed adults or peers ( McDuffie 2001 ). A university set in a major city has the resources to change the mindset of urban students and engage them in the exciting field of science. At Thomas Jefferson University (TJU), located in Philadelphia, Pennsylvania, a team of scientists and educators has developed a program that breaks down the stereotypes of the science field and allows students to engage in real, live experiments at their own schools. Why Are Such Programs Necessary? Around the world, educators face difficult choices in focusing educational goals with limited resources. In the United States, the No Child Left Behind Act , which guides school funding policy, currently places an emphasis on literacy and math in schools, with the result that best programs and practices in education are increasingly directed toward these two areas. Unfortunately, science education has become a lesser priority. Teachers are not given adequate resources to allow students to “get their hands dirty” during science lessons. If we want a society that is interested and knowledgeable about the need for scientific research, the basic principles of the life sciences need to be integrated early in the pre-college curriculum ( Sylwester 2001 ). We developed the Thomas Jefferson University Science Outreach Program (TJUSOP) to address this key issue and provide inquiry-based educational strategies through collaborative efforts between the university's faculty and partnering school districts. How Are We Making a Difference? Using the TJU facilities and laboratories, this innovative program integrates life science into the education of students between the ages of eight and eighteen from Pennsylvania, New Jersey, and Delaware. The mission of the TJUSOP is to foster an enthusiasm for science education, promote interest in future participation in biology-related fields, and allow all students the opportunity to learn life science through a hands-on, student-centered approach to instruction. The program is a supplement to the established curriculum, developed to support the content knowledge that is taught at each grade level. Teachers are invited to attend a professional development workshop held at the beginning of the school year where they receive training and resources for the units. Then, TJUSOP educators assist the teachers and students in their own classroom in running a weeklong experiment. This allows a large amount of group work to be completed simultaneously, even when teachers are faced with time constraints and large class sizes. This program is at no cost to the districts participating and is funded through the Jefferson Medical College and the Kimmel Cancer Center, as well as through the generosity of several local and national groups including Glaxo Smith Kline, the Christopher Ludwick Foundation, the Joan and Joseph Fernandez Family Foundation, the Brook J. Lenfest Foundation, the Foerderer Foundation, Drinker Biddle and Reath, and the Pennsylvania Department of Agriculture. Since its inception in August of 2002, this program has reached over 2,000 students and 75 teachers through our one-week zebrafish classroom experiments, our hands-on zebrafish and Drosophila facility tours in conjunction with Dr. James Jaynes (a Drosophila scientist at the Kimmel Cancer Center), and our High School Mentorship program held each summer. One of the main goals of TJUSOP is to reach students from ethnic and economic groups that are underrepresented in the scientific community. We have successfully partnered with the School District of Philadelphia, where 84.9% of the students come from ethnic backgrounds other than Northern European and 80% of the students are eligible for free or reduced-cost meals. This district, among many others, receives pre- and post-instruction for all teachers and students at no cost to the district, allowing it to improve the quality of its science education. Although we target these school districts, it is important to note that the US is facing a problematic decrease in the number of Americans, of any background, entering the science and engineering workforce. According to a National Science Foundation report, “If action is not taken now to change these trends, we could reach 2020 and find that the ability of U.S. research and education institutions to regenerate has been damaged and that their preeminence has been lost to other areas of the world” ( National Science Board Committee 2004 ). In regard to this unsettling discovery, TJUSOP welcomes all districts to participate and hopes to secure funding to double the number of students reached per year. Anyone Can Be a Scientist Our pedagogical approach to experiments allows students and teachers to become scientists, following the scientific process from beginning to end. Our live, one-week classroom experiments for the fourth, seventh, and tenth grades use zebrafish, a popular model organism for genetic research. A curriculum sample is as follows: in the seventh-grade unit, the students mate albino (recessive trait) male and a wild-type (dominant trait) female zebrafish in order to observe what the offspring will look like. Students form hypotheses, such as that the young offspring will look like the mother and the older offspring will be striped. Throughout the week, students observe and record embryos developing a head, tail, and notochord and pigment development. By the end of the experiment, a live heartbeat can be seen as well as the individual blood cells flowing throughout the larvae using a stereomicroscope TJUSOP provides ( Figure 1 ). Figure 1 A Philadelphia Student Uses the Zeiss Stereomicroscope during the Weeklong Experiment Grade-specific scientific journals are given to the students. The journals contain an introduction to TJUSOP and the experiment, background information about zebrafish in research, scientific vocabulary words used throughout the unit, and a word search activity. Students are given the title of “Junior Scientists” in grades 4 and 7 and “Student Scientists” in grade 10 and are asked to record the research question, a hypothesis, daily observations, and the conclusion of the experiment ( Figure 2 ). Figure 2 An Example of a Seventh-Grade Journal Entry How to Break the Stereotypes of What Science Is TJUSOP allows student participants to use scientific tools, talk with real scientists, and gain scientific knowledge so they can become informed members of their communities. Upon asking a fourth-grade student why she thought it was important to learn about science using zebrafish and the microscope, the student wrote, “I think it is important because we can find facts about oursefs [sic].” This sounds like a good start. For more information about the program, or if you would like to get involved in the initiative, please contact Jamie Schaefer, at jamie. E-mail: Schaefer@mail.jci.tju.edu or visit http://www.kimmelcancercenter.org/scienceoutreachprogram .	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521727.xml
545074	The two neutrophil members of the formylpeptide receptor family activate the NADPH-oxidase through signals that differ in sensitivity to a gelsolin derived phosphoinositide-binding peptide	Background The formylpeptide receptor family members FPR and FPRL1, expressed in myeloid phagocytes, belong to the G-protein coupled seven transmembrane receptor family (GPCRs). They share a high degree of sequence similarity, particularly in the cytoplasmic domains involved in intracellular signaling. The established model of cell activation through GPCRs states that the receptors isomerize from an inactive to an active state upon ligand binding, and this receptor transformation subsequently activates the signal transducing G-protein. Accordingly, the activation of human neutrophil FPR and FPRL1 induces identical, pertussis toxin-sensitive functional responses and a transient increase in intracellular calcium is followed by a secretory response leading to mobilization of receptors from intracellular stores, as well as a release of reactive oxygen metabolites. Results We report that a cell permeable ten amino acid peptide (PBP10) derived from the phosphatidylinositol 4,5-bisphosphate (PIP 2 ) binding region of gelsolin (an uncapper of actin filaments) blocks granule mobilization as well as secretion of oxygen radicals. The inhibitory effect of PBP10 is, however, receptor specific and affects the FPRL1-, but not the FPR-, induced cellular response. The transient rise in intracellular calcium induced by the active receptors is not affected by PBP10, suggesting that the blockage occurs in a parallel, novel signaling pathway used by FPRL1 to induce oxygen radical production and secretion. Also the FPR can activate neutrophils through a PBP10-sensitive signaling pathway, but this signal is normally blocked by the cytoskeleton. Conclusions This study demonstrates that the two very closely related chemoattractant receptors, FPR and FPRL1, use distinct signaling pathways in activation of human neutrophils. The PIP 2 -binding peptide PBP10 selectively inhibits FPRL1-mediated superoxide production and granule mobilization. Furthermore, the activity of this novel PBP10 sensitive pathway in neutrophils is modulated by the actin cytoskeleton network.	Background The molecular basis for cellular recognition of signal molecules is their binding to specific cell surface receptors [ 1 ]. Despite large structural differences between the huge numbers of extracellular ligands, many bind to (and activate) specific receptors belonging to a large family of pertussis toxin-sensitive G-protein linked receptors (GPCRs) [ 2 ]. These receptors possess a high degree of similarity and although activated by different agonists, they transduce downstream signals that have many common features [ 3 ]. Nevertheless, it is clear that there are also important differences between receptor-ligand pairs regarding their functional repertoire [ 4 ]. The pattern recognition, formyl peptide receptor (FPR) family, belongs to the larger GPCR group of chemoattractant receptors [ 5 ]. The FPR gene family has a complex evolutionary history and the number, function, and specific cell expression of genes comprising the family vary considerably between different mammalian species [ 5 , 6 ]. Human neutrophil granulocytes express two FPR members [ 7 ], the FPR and the FPRL1 (the formyl peptide receptor like 1). FPRL1 was originally defined as an orphan receptor, cloned from an HL-60 cell cDNA library by low-stringency hybridization with the FPR sequence [ 7 ]. In the past few years several neutrophil activating ligands specific for FPRL1 have been identified [ 8 ], but knowledge on the precise functional activities and the signal transduction pathways utilized by FPRL1 are still somewhat limited. In contrast to FPRL1, a large number of studies on FPR induced cell function and signaling have been performed in neutrophils and in receptor-expressing transfected cell lines. These studies reveal that FPR signaling shows all the basic characteristic of a GPCR. Binding of the prototype FPR agonist fMLF to its receptor initiates a chain of events starting with dissociation of the G-protein α subunit from its βγ subunit. These subunits directly or indirectly activate downstream signaling molecules such as protein kinase C (PKC), mitogen-activated protein kinase (MAPK), and phosphoinositide 3-kinase (PI3K) that uses the membrane phosphoinositide phosphatidylinositol 4,5-bisphosphate (PIP 2 ) as substrate [ 9 ]. The dissociated subunits of the G-protein also activate the phosphoinositide-specific phospholipase C (PLC) that upon cleavage of PIP 2 produces the second messengers responsible for an elevation of intracellular free calcium [ 8 ]. It has been assumed that FPRL1 shares signal transduction features with FPR, since both receptors are sensitive to pertussis toxin and possess a high degree of amino acid identity in the signaling cytoplasmic domains [ 7 ]. Further, the functional responses induced by the FPRL1 specific hexapeptide agonist WKYMVM is in most respects similar to (or even indistinguishable from) those induced by the prototype FPR agonist fMLF [ 10 - 15 ]. However, in this study we have used a membrane permeant polyphosphoinositide-binding peptide (PBP10 [ 16 ]) derived from the cytoskeletal protein gelsolin, and we show that with respect to messengers generated by the these receptors leading to mobilization of secretory granules and NADPH-oxidase activation, two totally different signaling routes are used by FPR and FPRL1, one being sensitive (the FPRL1 route) the other insensitive (the FRP route) to the PIP 2 -binding peptide. Results NADPH-oxidase activity induced by fMLF and WKYMVM and effects of PBP10 The peptides fMLF and WKYMVM, agonists for FPR and FPRL1, respectively, both induced a robust oxidative burst measured as a release of superoxide anions (Fig 1 ). In accordance with the known receptor specificity [ 10 , 17 ], the WKYMVM response was totally inhibited by the FPRL1 specific antagonist WRWWWW whereas the FPR specific antagonist (cyclosporine H)was without effect. The effects were reversed for fMLF-triggered activity, that is, this response was totally inhibited by cyclosporine H but not affected by WKWWWW (data not shown). The time-courses of the responses were very similar, as were the EC 50 values (Fig. 1A,B ). Both the FPR and FPRL1 mediated response was inhibited by pertussis toxin (whereas no reduction in oxidase activity was seen with PMA, a PKC activator that bypasses the G-protein), showing that a heterotrimeric G-protein is involved in the signal transduction of both receptors (Fig. 1C ). These indistinguishable responses were expected, based on the fact that the two receptors are very similar in the regions suggested to be of importance for intracellular signaling. Figure 1 Oxidase activity induced by peptide agonists and the effect of PBP 10. Neutrophils were activated by fMLF or WKYMVM and the extracellular release of superoxide anions was recorded by chemiluminescence (expressed in Mcpm). (A) The figure shows the kinetics of the neutrophil response to two concentrations (100 nM and 20 nM) of fMLF or WKYMVM. (B) Dose dependent oxidase activation induced by the two agonists. The peak values were measured and the responses are given as percent of the maximal response. (C) Neutrophils were incubated for 120 minutes in the absence (control) or presence of pertussis toxin (PTX; 500 ng/ml) and the cells were then activated with fMLF (100 nM), WKYMVM (100 nM) or the receptor independent PKC activator PMA (100 nM) (inset). For comparision, the response induced by the two peptides after 90 minutes long incubation time with pertussis toxin is also included. (D) Effect of different concentrations of PBP10 on the neutrophil NADPH-oxidase response induced by fMLF or WYMVM, respectively. Data are expressed as percent of control (without PBP10; mean ± SD of three independent experiments). Despite the indistinguishable activation by FPR and FPRL1 shown in Fig. 1A–C , functional differences between these two highly homologous receptors emerge when they are treated with the membrane permeant polyphosphoinositide-binding peptide rhodamine-B-QRLFQVKGRR (PBP10; Fig. 1D ) prior to activation. The neutrophil NADPH-oxidase activity was totally inhibited by PBP10 when FPRL1 was stimulated by WKYMVM, whereas there was no effect of the peptide on the fMLF-induced, FPR-mediated neutrophil response. The IC 50 value for the WKYMVM induced response was around 0.05 μM whereas no effect was seen on the fMLF induced activity even at concentrations up to 10 μM. There was no effect on WKYMVM or fMLF induced superoxide production of rhodamine alone (data not shown). The neutrophil response is inhibited by Wortmannin The amino acid sequence of PBP 10 peptide corresponds to the phosphatidylinositol 4,5-bisphosphate (PIP 2 ) binding region segment 2 of the cytoskeletal protein gelsolin [ 16 ]. Following G-protein coupled receptor (GPCR) activation, the dissociated G-protein subunits activate the downstream phosphoinositide remodeling enzyme phosphatidylinositol 3-kinase (PI3K) [ 18 ], an enzyme that uses the membrane phosphoinositide PIP 2 as substrate to generate the signaling molecule phosphatidylinositol 3,4,5-trisphosphate (PIP 3 ) [ 9 ]. The oxidase activity induced by fMLF as well as by WKYMVM was largely inhibited by the specific PI3K inhibitor Wortmannin (Fig. 2A ) suggesting that this signaling pathway is of importance in the cellular response. In contrast to the effect of PBP10, the effects of the PI3K inhibitor showed no specificity in the inhibition of the WKYMVM/FPRL1 triggered response. Figure 2 Effect of different inhibitors on the neutrophil NADPH-oxidase activity. (A) Neutrophils were incubated with different concentrations of the PI3K specific inhibitor Wortmannin at 37°C for 30 minutes followed by stimulation with fMLF (100 nM) or WKYMVM (100 nM). A representative experiment is shown. (B) Neutrophils were incubated with the cell impermeable PIP 2 binding peptide QRLFQVKGRR (1 μM final concentration) at 37°C for 5 minutes followed by stimulation with WKYMVM (100 nM). The effect of PBP10 (1 μM) is included for comparison. PBP 10 inhibits the response to FPRL1 but not to FPR The inhibition induced by PBP10 on neutrophil oxidase activity was linked to the receptor FPRL1 rather than to an FPRL1-specific agonist, since the same inhibition pattern was obtained when WKYMVM was replaced by serum amyloid A (SAA) (Table 1 ), an agonist that activates cells through binding to FPRL1 [ 19 ]. Moreover, no effect was induced by PBP10 when fMLF was replaced by annexinI 9–25 (Table 1 ), a 17 amino acid peptide derived from the N-terminus of annexin I that has been shown to be a FPR agonist ([ 20 ] and our own observation). Moreover, the cellular responses mediated through other receptors, such as the chemokine receptor CXC or the C5a-receptor were not affected by PBP10 (Table 1 ). Table 1 Effect of PBP10 on neutrophil NADPH-oxidase activity induced by receptor agonists # . Neutrophil NADPH-oxidase activator (Concentration) Neutrophil receptor(s) Inhibition of NADPH-oxidase by PBP10 (1 μM) WKYMVM (100 nM) FPRL1 + SAA (5 μM) FPRL1 + fMLF (100 nM) FPR - Annexin 9–25 (50 μM) FPR - C5a (100 ng/ml) C5aR - IL-8 (100 ng/ml) CXCR1, CXCR2 - Abbreviations: C5a, complement 5a; FPR, formyl peptide receptor; IL-8, interleukin 8; SAA, serum amyloid A. The membrane permeability of PBP10 is necessary for FPRL1 inhibition Interaction of PBP10 with PIP 2 depends on the peptide sequence, as truncation of the peptide either at the N terminus or C terminus reduces the PIP 2 binding affinity [ 21 ]. Accordingly, Rh-VKGRRG affected the oxidase activity induced by WKYMVM only when higher concentrations of peptide were used (IC 50 = 1 μM compared to 0.05 μM for PBP10) suggesting that PIP 2 binding is of importance for inhibition of the neutrophil oxidase response induced by FPRL1. The PBP10 peptide possesses not only PIP 2 binding capacity but is also membrane permeable. This feature has been shown to be coupled to the rhodamine B part of the molecule, i.e., the unconjugated peptide QRLFQVKGRR exhibits the same PIP 2 -binding characteristics as PBP10 but lacks the ability to cross the plasma membrane [ 16 ]. The unconjugated peptide had no effect on the WKYMVM induced neutrophil oxidase activity (Fig. 2B ) suggesting that membrane permeation is required for peptide-induced inhibition. No effect of PBP10 on the transient calcium response PIP 2 is utilized for generation of IP 3 that induces a transient increase in intracellular calcium [ 22 ], we thus investigated the effect of PBP10 on the neutrophil calcium response. Stimulating the cells with fMLF or WKYMVM induced a rapid and transient increase in cytosolic calcium. These responses were not significantly reduced by the removal of extracellular Ca 2+ with EGTA (data not shown), suggesting that the rise in intracellular free Ca 2+ relates to a mobilization from calcium storing organelles rather then from an influx through ion channels in the plasma membrane. Treatment of the cells with PBP10 prior to stimulation did not alter the calcium responses, neither induced by fMLF nor by WKYMVM (Fig. 3 ). Hence, the PBP10-induced effect on FPRL1 signaling is not due to a specific inhibition of the calcium transients induced by the activated GPCRs. Figure 3 Effect of PBP10 on cytosolic calcium mobilization induced by peptide agonists. Neutrophil intracellular calcium mobilization was analyzed by monitoring Fura-2 fluorescence upon stimulation with the FPR agonist fMLF (100 nM; left) or the FPRL1 agonist WKYMVM (100 nM; right) in the absence (upper curves) or presence (lower curves) of PBP10 (1 μM). The curves are derived from a representative experiment. The calcium concentrations given in the right hand part of the figure are valid for the fMLF as well as the WKYMVM induced responses. PBP10 inhibits mobilization of CR3 The Ca 2+ elevation has been claimed to be required but not sufficient for the generation of an NADPH-oxidase activating signal from FPR [ 23 , 24 ]. Likewise, the mobilization through regulated exocytosis of neutrophil secretory storage organelles containing reserve pools of cell-surface receptors [ 25 ] has been suggested to be directly regulated by the cytosolic concentration of free Ca 2+ [ 26 , 27 ]. We found that despite the fact that PBP10 was without effect on the transient rise in intracellular Ca 2+ induced by WKYMVM, it blocked the secretory response (Fig. 4 ). In accordance with the inhibition of the NADPH-oxidase activity, PBP10 was without effect on the FPR induced granule mobilization (Fig. 4 ). This suggests that degranulation and NADPH-oxidase activity induced by FPRL1 both are on the same, PBP10-sensitive, signal transduction pathway in contrast to FPR, which induces both cellular functions by a PBP10-insensitive pathway. Figure 4 Effect of PBP10 on fMLF and WKYMVM induced mobilization of CR3 to the neutrophil cell surface. Neutrophils were incubated with or without PBP10 (1 μM) at 37°C for 5 minutes after which the cells were challenged with fMLF (100 nM) or WKYMVM (100 nM) for another 10 minutes. After fixation of the cells with paraformaldehyde and labeling with anti-CR3 antibody, the amount of CR3 on the cell surface was determined by FACS analysis. A representative histogram of PBP10 effect on WKYMVM mediated CR3 exposure is shown (inset), and the exposure of CR3 after cell stimulation with fMLF and WKYMVM is expressed as percent of control (without PBP10; mean ± SD of four independent experiments). *p < 0.05 compared to WKYMVM without PBP10. Regulation by the cytoskeleton Regulation of FPR signaling has been suggested to involve mechanisms that depend on direct receptor interaction with the membrane cytoskeleton [ 28 ]. It is well known that the neutrophil response to FPR and FPRL1 agonists is both augmented and prolonged in the presence of cytochalasin B [ 15 ], a fungal metabolite that inhibits re-organization of actin polymers and uncouples receptors from the cytoskeleton [ 13 ]. To investigate the involvement of cytoskeleton in the PBP10-sensitive signal transduction pathway, neutrophils were treated with cytochalasin B prior to activation by a receptor agonist (in the presence or absence of PBP10). Interestingly, the receptor selectivity of PBP10's inhibitory effect was lost when the receptors were first uncoupled from the cytoskeleton by cytochalasin B (Fig. 5 ). In accordance with the findings reported above, the NADPH-oxidase activity induced by WKYMVM was largely inhibited by PBP10 also in the presence of cytochalasin B. However, the drug introduced a PBP10 sensitivity also in the FPR-induced response. The inhibition of the FPR-mediated response required higher concentrations of PBP10, and the response was only partly inhibited suggesting that the PBP10-insensitive as well as the PBP10-sensitive signaling pathways were triggered simultaneously during activation of FPR uncoupled from the cytoskeleton. The transient rise in intracellular calcium induced by the active receptors was not affected by cytochalasin B (data not shown). In conclusion, both FPRL1 and FPR appear to possess the ability to activate neutrophils via a signaling route that is sensitive to PBP10, but that this signaling pathway normally is blocked for FPR, through association of the receptor with the cytoskeleton. Figure 5 Effect of PBP10 on oxidase activity induced by peptide agonists in the presence of cytochalasin B. Neutrophils were activated by fMLF (100 nM) or WKYMVM (100 nM) in the presence of cytochalasin B (2.5 μg/ml, final concentration) and the extracellular release of superoxide anion was recorded (expressed in Mcpm). (A) The figure shows the kinetics of the neutrophil response to fMLF in the absence (solid line) and presence (broken line) of PBP10 (5 μM). The results obtained are summarized in the inset, expressed as the integral values of oxidase activities from the controls (presence of cytochalasin B but not PBP10) and the activities in the presence of both cytocalasin B and PBP10. (B) The figure shows the kinetics of the neutrophil response to WKYMVM in the absence (solid line) and presence of PBP10 (broken line). The results obtained are summarized in the inset and expressed as the integral values of oxidase activities from the controls (presence of cytochalasin B but not PBP10) and the activities in the presence of both cytochalasin B and PBP10. Discussion The two formyl peptide receptor family members FPR and FPRL1 possess a high degree of amino acid identity in the signaling cytoplasmic domains [ 7 ], and the cell functions induced by the FPR and FPRL1 agonists are in most respects identical [ 10 - 15 ]. The indistinguishable responses of activated FPR and FPRL1 respectively, are expected, based on the fact that the two receptors are very similar in the regions suggested to be of importance for their interaction with the signaling, pertussis toxin-sensitive heterotrimeric G-protein. Dysfunctional variant FPR alleles (F110 replaced by an S and C126 replaced by a W) have been described that are associated with juvenile periodontitis and a deficiency in G-protein coupling [ 29 , 30 ], but FPRL1 contains the functional amino acids of FPR both in position 110 and in 126. Other G-protein coupling structures, identified through expression of different FPR mutants, suggest that the N-terminal part of the second transmembrane domain (S63 and D71) and the C-terminal interface of the third transmembrane domain (R123, C124 and C126) may be sites of interaction between the receptor and the G-protein [ 31 ]. In FPRL1 all but one of the amino acids in these suggested interaction sites are identical with those in FPR, and the signal-regulating NPXXY motif in the seven transmembrane domain (highly conserved among all GPCRs) is identical (sequence NPMLY [ 32 ]) in the two receptors. The exception is the serine in position 63 that in FPRL1 is replaced by a cysteine, but this difference seems to be of minor importance with respect to the effect of pertussis toxin, since both receptors are sensitive to the toxin. Despite the indistinguishable activation by FPR and FPRL1, functional differences between these two highly homologous receptors emerge when they are challenged by the membrane permeant polyphosphoinositide-binding peptide PBP10 (rhodamine B-QRLFQVKGRR) prior to activation. The FPRL1-mediated neutrophil activity was totally inhibited by the peptide PBP10 whereas there was no effect on FPR-mediated responses. As mentioned, the amino acids in the PBP 10 peptide correspond to the PIP 2 binding region segment 2 of the cytoskeletal protein gelsolin [ 16 ]. One of the enzymes competing with PBP10 for PIP 2 would be the phosphoinositide remodeling enzyme PI3K that converts PIP 2 into phosphatidylinositol 3,4,5-trisphosphate (PIP 3 ) which is of importance for cell locomotion and the associated dynamic reorganization of cytoskeletal components. The precise target for PIP 3 has however not yet been defined [ 33 ]. The inhibition of neutrophil function by Wortmannin, an inhibitor of PI3K, suggests that this signaling pathway is of importance for the NADPH-oxidase activity but the PI3K inhibitor lacked receptor specificity (i.e., both FPR and FPRL1 induced responses were inhibited) and we can thus rule out that the mechanism behind the PBP10 effect is a direct interference with the PI3K pathway. The basic properties of PBP 10 have been described earlier [ 16 , 21 , 34 ], and it has been shown that when coupled to rhodamine B the peptide possesses not only PIP 2 binding activity but also crosses the cell membrane of neutrophils and other cells. Interaction of rhodamine labeled peptides with PIP 2 depends on the peptide sequence, and truncation of the peptide reduces the PIP 2 binding affinity [ 21 ]. The truncated peptides were still membrane permeable and affected the oxidase activity only at higher concentrations. The unconjugated peptide QRLFQVKGRR exhibiting the same PIP 2 -binding characteristics as PBP10 but lacking the ability to cross the plasma membrane [ 16 ] had no effect on neutrophil oxidase activity. Taken together these data suggest that membrane permeation is required and PIP 2 binding is of importance for peptide-induced inhibition. Plasma membrane localized PIP 2 is utilized for generation of IP 3 which in turn is responsible for inducing transients in intracellular calcium [ 22 ]. Triggering of cells with fMLF or WKYMVM induced a rapid and transient increase in cytosolic calcium, but PBP10 did not have any inhibitory effect on the calcium transient. The Ca 2+ elevation has been claimed to be required but not sufficient for the generation of an NADPH-oxidase activating signal from FPR [ 23 , 24 ]. Likewise the mobilization of granule localized reserve pools of cell-surface receptors [ 25 ], has been suggested to be directly regulated by the cytosolic concentration of free Ca 2+ [ 26 , 27 ]. These experimental evidences rely, however, on methods that cannot distinguish a dependency on basal Ca 2+ levels from a requirement for a Ca 2+ transient, and we have earlier shown that receptor mobilization can occur and the oxidase can be activated without any transient rise in cytosolic Ca 2+ [ 13 , 35 , 36 ]. Despite the fact that PBP10 was without effect on the transient rise in intracellular Ca 2+ induced by WKYMVM, the secretory response was blocked and PBP10 selectively inhibited the FPRL1 induced granule mobilization. The GPCR family is very diverse and the transmission of signals by such receptors is a critical function in many cell/organ systems. Signaling through GPCRs is highly complex, evidently not only with respect to the wide variety of mechanisms that regulate different functional responses [ 2 ], but also with respect to the pathways used to regulate a defined cellular response through closely related receptors. The two FPR and FPRL1 genes, although originating from a common ancestral gene, appear to have undergone markedly different evolutionary events [ 37 ]. In contrast to FPR, which is characterized by a relatively high degree of single nucleotide polymorphism (five non-synonymous and two synonymous identified [ 37 ]), no FPRL1 polymorphism has been found. Only one of the FPR polymorphisms is located in the cytoplasmic regions; and the variant (containing the A346 → E exchange) has the same amino acid in that position in the cytoplasmic tail as FPRL1. A direct comparison of the amino acid sequences of FPR and FPRL1 reveal very small differences between the receptors in all intracellular domains except for the C-terminal tail. In the first intracellular loop the H57 in FPR is replaced by an R (H57 → R) in FPRL1, in the second loop differences are found in V125 → I and T133 → A, and in the third loop the Q231 → K and L233 → M exchanges are found. The only major differences between the two receptors are found in the cytoplasmic C-terminal tail in which 13 out of 45 amino acids differ, and it is worth noting that the amino acid exchange in nine of the positions that differ between FPR and FPRL1 are potential targets for phosphorylation. Five potential phosphorylation sites, suggested to be of importance for arrestin-binding and receptor desensitization in FPR [ 38 ], are missing in FPRL1 while two new sites have been added in this receptor. It seems reasonable to assume that the signaling route that is sensitive to PBP10 originates from this region of the FPRL1 receptor. Despite the fact that the differences between FPR and FPRL1 are limited, the identification of putative sites in FPRL1 of importance for the PBP10 sensitivity cannot be achieved through experiments performed with receptor chimeras, site-directed receptor mutants or deletions, since no cells equipped with the required effector functions are available for expression of the receptors. Regulation of FPR signaling has been suggested to involve mechanisms that depend on direct receptor interaction with the membrane cytoskeleton [ 28 ]. It is well known that the neutrophil response to FPR and FPRL1 agonists is both augmented and prolonged in the presence of cytochalasin B [ 15 ], a drug that inhibits re-organization of actin polymers and uncouples the receptors from the cytoskeleton [ 13 , 36 ]. The cytoskeleton is part of the signaling modulating machinery and we show that the receptor selectivity of the PBP10 inhibitory effect was lost when the receptors were uncoupled from the cytoskeleton. In accordance with the earlier described findings, the NADPH-oxidase activity induced by WKYMVM was largely inhibited by PBP10 also in the presence of cytochalasin B, but the drug introduced this sensitivity when fMLF was used as the triggering agent. The inhibition of the FPR-mediated response required higher concentrations of PBP10, and the response was only partly inhibited suggesting that the PBP10-insensitive as well as PBP10-sensitive signaling pathways were activated simultaneously by FPR's uncoupled from the cytoskeleton. It is interesting to note that when FPR was uncoupled from the cytoskeleton by cytochalasin B, PBP10 affected the sustained generation of superoxide but not the initial rate of production. This suggests that different signals are responsible for the triggering of the oxidase in the early and late phases of the response, respectively. It is reasonable to assume that the PBP10 sensitive signal generated by the uncoupled FPR is identical to that generated by FPRL1, however, we cannot at present exclude the possibility that also a novel FPR-triggered pathway is blocked by PBP10. The transient rise in intracellular calcium induced by the active receptors was not affected by cytochalasin B (data not shown), suggesting that this signaling route does not depend on interaction with the cytoskeleton. A possible explanation for the effects when PBP10 and cytochalasin B are combined is that signaling G-proteins compete with cytoskeletal proteins for the same site on FPR, and that this interaction involves a region of the receptor that differs between FPR and FPRL1. A 15 amino acid long sequence in FPR ( 322 FPR 336 ) has a fairly high (45–50%) identity with the actin-binding cytoskeletal proteins vinculin and coronin, and this region also participates in FPR interaction with the G-protein [ 28 ]. It is of interest to notice that the amino acid sequence in this region of FPRL1 differs from that of FPR in five positions and four of these are (in FPR) potential phosphorylation sites. Phagocytes express predominantly the G i α 2 complex of the pertussis toxin sensitive G-proteins and, to a lesser extent G i α 3 [ 39 ]. The molecular mechanism behind the difference in sensitivity to PBP10 between FPR and FPRL1 could possibly be that the receptors couple to different G-protein subtypes. FPR has, however, been shown to couple to both G-protein subtypes with similar efficiency [ 40 ]. It is however important to note that these experiments were performed in receptor-expressing cells in which nothing is known about the linkage between the receptors and the cytoskeleton and in which the proper cell function repertoire is missing. The relation of the biochemical/biophysical activities of the PBP10 peptide to its effect on the FPRL1-triggered cell function is not obvious and is likely to be complex. The sensitivity to PBP10 seems to be unique to FPRL1, but this receptor is expressed also in other cells such as astrocytes, neuroblastoma, and microglia cells [ 8 ]. FPR is also expressed in other cell types and whether the receptor selectivity/specificity of PBP10 is maintained in other cells remains to be determined. The biochemical/biophysical characterization of the ten strategically organized basic and hydrophobic amino acids of the gelsolin molecule included in PBP10 reveal that it may interact with a broad range of negatively charged phosphomonoesters and hydrophobic acyl chains of anionic phospholipids [ 41 , 42 ]. This suggests that in addition to blocking/competing activities which involve proteins that are regulated by cellular phosphoinositides, the peptide may function as a buffer of bioactive and signaling lipids. Although elucidation of the step in signal transduction that is disrupted by PBP10 requires much additional work, the receptor-specific and signal-selective effects of this peptide on neutrophil functions suggest that it has a potential as a tool to manipulate and help define how GPCRs produce and integrate the signals generated from activated receptors and to probe new signaling functions of polyphosphoinositides as well as defining the role as promotor/blocker of G-protein signaling of different cytoskeletal proteins. Conclusions The neutrophil formyl peptide receptor family members FPR and FPRL1 share 69% of amino acid identity and mediate almost indistinguishable cellular responses. Thus, the assumption that FPR and FPRL1 use the same signaling pathways has been generally accepted. However, in this study we clearly demonstrate a fundamental difference in intracellular signaling between these two very closely related neutrophil formyl peptide receptor members, one being PBP10 sensitive and the other not. This novel PBP10 sensitive signaling pathway utilized by FPRL1 is also used by FPR but only when the cytoskeleton network is disrupted. Methods Isolation of human neutrophils Neutrophil granulocytes were isolated from buffy coats obtained from healthy adults. After dextran sedimentation at 1 × g, hypotonic lysis of the remaining erythrocytes, and centrifugation in a Ficoll-Paque gradient [ 43 ], the neutrophils were washed twice and resuspended (1 × 10 7 /ml) in Krebs-Ringer phosphate buffer containing glucose (10 mM), Ca 2+ (1 mM), and Mg 2+ (1.5 mM) (KRG; pH 7.3). The cells were stored on melting ice and used within 120 min of preparation. Chemoattractants and stimuli The hexapeptide Trp-Lys-Tyr-Met-Val-Met-NH 2 (WKYMVM) and the annexin I peptide (annexin 9–25 Ac-QAWFIENEEQEYVQTVK) were synthesized and HPLC-purified by Alta Bioscience (University of Birmingham, Birmingham, United Kingdom), and Ross-Petersen ApS (Holte, Denmark). The formylated peptide N-formyl-Met-Leu-Phe (fMLF), C5a, and, phorbol myristate acetate (PMA) were from Sigma Chemical Co. (St. Louis, Missouri). IL-8 was from R&D systems (Minneapolis, MN) and dissolved in KRG containing 0.5% (w/v) bovine serum albumin. Serum amyloid A (SAA) was from Pepro Tech Inc. (UK). The C5a and SAA were dissolved in water while the other peptide agonists were dissolved in dimethyl sulfoxide to 10 -2 M and stored at -70°C until use. Further dilutions were made in KRG. PBP10 synthesis The peptide QRLFQVKGRR (gelsolin residues 160–169) and related peptides were prepared by solid phase peptide synthesis and coupled to rhodamine as described earlier [ 16 ]. Neutrophil NADPH-oxidase activity Neutrophil production and release of superoxide anions was measured by means of an isoluminol-enhanced chemiluminescence (CL) assay [ 44 ]. The CL activity was measured in a six-channel Biolumat LB 9505 apparatus (Berthold Co., Wildbad, Germany), using disposable 4-ml polypropylene tubes with a 900-μl reaction mixture containing 10 5 –10 6 neutrophils, horseradish peroxidase (4 U) and isoluminol (20 μM) with and without cytochalasin B or PBP10. The measuring tubes were equilibrated for 5 to 10 minutes at 37°C and the cells were activated by addition of a receptor specific peptide agonist, fMLF for FPR and WKYMVM for FPRL1. The light emission was recorded continuously. Details about the CL technique are given in [ 45 ]. Determination of changes in cytosolic calcium Neutrophils at a density of 2 × 10 7 /ml in Ca 2+ -free KRG supplemented with bovine serum albumin (BSA, 0.1%) were incubated with the acetoxymethylated derivative fura-2/AM (2 μM) at room temperature for 30 minutes. The cells were washed twice and resuspended in KRG, adjusted to 2 × 10 7 /ml and kept protected from light on ice until use. Cells with or without PBP10 were equilibrated for 5 minutes at 37°C, after which the peptide agonist was added. The fura-2 fluorescence was followed with a luminescence spectrometer (LS50B; Perkin Elmer Corp.) using excitation wavelengths of 340 nm and 380 nm, and an emission wavelength of 510 nm and the [Ca 2+ ] i was calculated as described earlier [ 46 ]. Determination of receptor exposure by FACS analysis The exposure of CR3 (CD11b/CD18) on the neutrophil cell surface was assessed by immunostaining and FACS-analysis. Cells challenged with peptide agonists with or without PBP10 were fixed with ice-cold paraformaldehyde and washed with FACSwash (PBS, 0.02% NaN 3 ), after which the cells were incubated at 4°C with phycoerythrin-conjugated anti-CR3 antibodies (CD11b; Becton Dickinson 10 μl/10 6 cells) before analysis using a FACScan (Becton Dickinson, Mountain View, CA). Reagents Horseradish peroxidase (HRP) was from Boehringer-Mannheim (Mannheim, Germany). Isoluminol, cytochalasin B, and pertussis toxin were purchased from Sigma. Dextran and Ficoll-Paque were from Pharmacia (Uppsala, Sweden), and Wortmannin was from Calbiochem (La Jolla, CA). Fura-2AM was from Molecular Probes Inc. (Eugene, OR). Statistic analysis Two-tailed, paired Students's t-tests were performed to determine statistical significance, and a P-value of < 0.05 was regarded as significant. Abbreviations GPCR – G-protein coupled receptor PBP – Phosphoinositide binding peptide FPR – formyl peptide receptor FPRL1 – formyl peptide receptor-like 1 PIP2 – phosphatidylinositol 4,5-bisphosphate PI3K – phosphatidylinositol 3-kinase fMLF – formyl-methionly-leucyl-phenylalanin Authors' contributions The scientific question raised in the paper was formulated during discussions between HF and CD, about the mechanisms behind receptor activation/deactivation/reactivation (see ref [ 38 ]). HF was responsible for most of the experiments but LB, JK and CM performed some of the experiments using techniques developed by CD and AK. PJ provided the PIP-binding peptides as well as suggestions for experiments. HF and CD wrote the first version of the paper, but contributions from all authors were important for the final outcome of the paper.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545074.xml
517726	Facilitating arrhythmia simulation: the method of quantitative cellular automata modeling and parallel running	Background Many arrhythmias are triggered by abnormal electrical activity at the ionic channel and cell level, and then evolve spatio-temporally within the heart. To understand arrhythmias better and to diagnose them more precisely by their ECG waveforms, a whole-heart model is required to explore the association between the massively parallel activities at the channel/cell level and the integrative electrophysiological phenomena at organ level. Methods We have developed a method to build large-scale electrophysiological models by using extended cellular automata, and to run such models on a cluster of shared memory machines. We describe here the method, including the extension of a language-based cellular automaton to implement quantitative computing, the building of a whole-heart model with Visible Human Project data, the parallelization of the model on a cluster of shared memory computers with OpenMP and MPI hybrid programming, and a simulation algorithm that links cellular activity with the ECG. Results We demonstrate that electrical activities at channel, cell, and organ levels can be traced and captured conveniently in our extended cellular automaton system. Examples of some ECG waveforms simulated with a 2-D slice are given to support the ECG simulation algorithm. A performance evaluation of the 3-D model on a four-node cluster is also given. Conclusions Quantitative multicellular modeling with extended cellular automata is a highly efficient and widely applicable method to weave experimental data at different levels into computational models. This process can be used to investigate complex and collective biological activities that can be described neither by their governing differentiation equations nor by discrete parallel computation. Transparent cluster computing is a convenient and effective method to make time-consuming simulation feasible. Arrhythmias, as a typical case, can be effectively simulated with the methods described.	Background Arrhythmias, a significant direct cause of death in heart diseases, are emergent and evolvable events that come with little prior warnings and allow limited response time [ 1 ]. Although ECG waveforms – the mapping of body surface potentials of cardiac cells – have routinely been used to diagnose arrhythmias, as integrated signals they tell us little about what happen at cell and ionic channel levels. They therefore are only marginally useful in guiding the clinical use of anti-arrhythmic drugs to treat disturbed cardioelectrical activity at cell and channel levels. Many of such drugs used today are ionic channel blocking agents [ 2 ]. To overcome the inherent limitations of clinical investigation, computational modeling and simulation has been widely recognized as a valuable alternative approach. Traditionally, cardiac modeling has centered on ECG simulation. Using the finite element method (FEM), the virtual heart and whole chest are partitioned into numerous elements representing a group of cells. The ECG is then simulated, based on computing the body surface potential of each cardiac element [ 3 - 5 ]. Basically, this method does not concentrate on cellular electrophysiological issues at the channel level, and thus fails to precisely associate macro level phenomena (ECG waveforms) with micro level activities and to make use of the considerable knowledge of cellular electrophysiology accumulated over the past decades. To improve the understanding of arrhythmias and to find effective perturbations, electrophysiological modeling using membrane equations is required so that mechanisms of arrhythmias at cell, channel, and even molecular levels can be investigated [ 6 - 8 ]. To study arrhythmias using a large-scale realistic electrophysiological model, two issues need to be effectively resolved: model building and operation. Though the widespread paradigm of modeling with C or C++ remains a workable choice, the huge number of cardiac cells in a realistic three-dimensional (3-D) whole-heart model and the numerous modifications of the model to simulate various pathological conditions, make more desirable efficient modeling based on transparent parallel computing. To build and run a model with parallel computing technologies, two strategies were separately developed in recent years. To provide transparent and parallel descriptions, cellular automata were used [ 9 - 11 , 50 ]; for efficient execution, distributed computing was adopted [ 12 , 13 ]. Yet, each strategy alone is not sufficient for the successful simulation of arrhythmias. On the one hand cardiac models built with traditional cellular automata are qualitative, and thus do not use the Hodgkin-Huxley (HH) action potential equations to describe channel electrical activity. Consequently, many arrhythmias, such as those triggered by early-after-depolarization (EAD) and delayed-after-depolarization (DAD), can not be simulated. On the other hand, physical parallelism on parallel computers is also not fully exploited in these cellular automata models. MPI, the programming protocol for distributed-memory multiprocessors (DMP), and OpenMP, the programming protocol for shared-memory multiprocessors (SMP), are not used [ 14 , 15 ]. Partly because of these two issues, arrhythmia simulations with a realistic whole-heart electrophysiological model have not been fruitfully conducted. There is compelling evidence that cardiomyocytes are not arranged in a uniformly connected continuum, as has often been assumed and simulated in the past. Along with nonlinear ionic channel electrical activity, discontinuous electrical propagation is another key feature of cardioelectrical activity [ 16 ]. The former demands a precise membrane equation based description; the latter requires a gap junction based discrete model. To efficiently build heterogeneous models containing different types of cardiomyocytes described by different membrane equation models and connected through different gap junctions remains a major challenge. In the present study, we propose a method using an extended, quantitative cellular automaton to build a discrete whole heart model with the data of the Visible Human Project (VHP) male cadaver [ 17 ]. An ECG simulation algorithm based on the membrane potential of each and every cell is designed and validated in a 2-D model built with the same method. Moreover, we combine cellular automata modeling with distributed parallel computing to realize efficient and affordable simulation. The parallel numerical solutions of the HH equations within a large number of cardiac cells are executed in parallel on a cluster with hybrid MPI and OpenMP programming. The parallel programming does not have to be manually coded in models built with the extended cellular automaton, because the modified compiler of the cellular automaton can automatically parallelize the codes, making parallelism fully transparent. The aim of this paper is to introduce the method and the whole-heart electrophysiological model. Results of performance evaluation on a four-node cluster are given. Based on this work, we conclude that quantitative modeling using extended and cluster computing enabled cellular automata is feasible and efficient, seamlessly binding conceptual and physical parallel computation. This method is suitable for a variety of computational intensive, tissue level modeling and simulation. Methods Cellular automata style quantitative computing Cellular automata were first introduced by John von Neumann and Stanislaw Ulam in the 1940s, and gradually used to solve a wide range of problems, including multicellular biological modeling in which a natural correspondence between each automaton cell and each biological cell is assumed [ 18 - 21 ]. Simple local interaction producing complex global behavior is common to a variety of natural phenomena, including cardioelectrical activity. Though traditionally cellular automata are regarded as discrete parallel systems, new functions can be obtained with non-standard implementations [ 22 , 23 ]. For language-based cellular automata, a program encoded in a language instead of a rule is shared by all cells and describes the behavior of each cell. A compiler translates the cell program into executable files. To enhance the portability of the language, a two-step compilation is usually adopted, with C/C++ files being the intermediate codes, allowing the extension of such cellular automata systems. Cellular is a cellular automata system based on the language Cellang [ 23 ], whose cell program comprises three parts: constant declarations, a cell declaration, and statements. The cell declaration defines a set of fields to store states of each cell between successive steps of computation. The cell array is described in a separate data file, whose output is piped to the cell program to provide the value of fields in each cell so as to enable computation. A predefined unchangeable variable time synchronizes the running of all cells. By conventions, in cellular automata computation quantitative computing is not fully supported and function call is not allowed. To overcome this inherent inadequacy of Cellang in quantitative computing and enable it to solve HH type membrane equations, we have added new language facilities, including the floating-point data type and the mathematical functions provided in the C libraries. Thus, Cellang is able to encode numerical solutions of the HH equations (Figure 1 ). A built-in function position() is also added to specify the global coordinates of the currently running cell. This function, in combination with the if-then statement, allows position-and time-dependent runtime perturbations to any cells, a function valuable for arrhythmia simulation. Viewing facilities are also extended to monitor and display simulation of electrical activitys at channel, cell, and organ level. Figure 1 Quantitative cellular automata with different neighborhoods. (a) Moor neighborhood. (b) A user-defined neighborhood. The radius in both cases is 1. Running a model may be no easier than building it. The large number of cells, the small time step Δt and the long running time for arrhythmia simulation make it impractical to run a whole heart model on any low-end computer. Since large-scale SMP machines are extremely expensive, the prevailing parallel computing platforms are clusters of small-scale SMP or DMP, and the protocols of programming in Fortran, C, and C++ on such computers are OpenMP and MPI. First, an endless while loop is used to realize iterative computation. Second, since for an n -dimensional model each field in the cell program is translated into an n -dimensional data array in the intermediate C program, within the while loop n successive for loops are employed to traverse the n dimensional cell space. Only the codes within the for loops, which are statements in the cell program, need to be parallelized. Instead of defining one data array with an offset for each field in the original Cellular system, we use two data arrays in a flip-flop manner to support parallel write operation. OpenMP provides a group of directives to be inserted into the C program to tell the compiler the region to be executed in parallel. Such directives appear before the for loops to dispatch the outermost for loop into a group of threads, whose number is set dynamically according to the available CPUs. The simulation of cellular activity is thus implicitly parallelized in a shared memory space. To parallelize a Cellang program on a platform with distributed memory, explicit cell space decomposition is inevitable. Every data array is equally divided into several subsets located in distributed memory bodies, and computed by autonomous computing nodes that communicate each other through MPI. Since a cell needs to access its neighboring cells over the maximal radius m to compute transjunctional currents, each subset should enclose m extra layers of cells on each side to ensure that cells at the boundary layer(s) can access correct neighbors that are located in the outermost m layers of the two neighboring subsets, respectively. Field values of cells in these extra layers should be swapped between neighboring subsets after each round of computation to ensure that cells access updated data. Such data exchange is the major factor for the excessive time needed by cellular automata models running on distributed memory platforms. If cells have many fields, the time spent on such operations may even offset the benefits of parallelism. To reach the maximal flexibility and portability for MPI-parallelization, a master-slave program structure is adopted (Figure 2 ). The number of slaves is set in the compilation command with an argument. The master is responsible for: Figure 2 The master-slave structure of distributed computing with MPI programming. • Reading data from the data file into the cell space and dispatching the cell space at time step 0; • collecting the value of a selected field in each cell from all slaves and displaying them at each time step; • updating the variable time and sending its new value to every slave to trigger the next round of computation; • doing some global, non-cellular automata style computing such as ECG simulation; Each slave process is responsible for: • receiving a subset of the cell space at time 0; • exchanging the boundary layers with neighboring slaves at each time step; • executing the cell program of cells in its subset at each time step; • sending the value of a selected field to the master at each time step for display; • receiving the new value of time at each time step. Finally, to run the model on a cluster of SMP nodes, a two-tier parallelism – a coarse-grained setting among nodes and a fine-grained setting within nodes – is applied. The hybrid programming is straightforward – insert OpenMP directives into the master and slaves , as in the case of pure OpenMP programming to spawn a group of threads within each node to implement cell level parallelism. Due to the stereotyped appearance of these codes, we let the compiler generate them each time we compile a model, leaving the number of slaves and threads within each slave as command line arguments, so as to make parallel computation entirely transparent to users. Construction of the anatomical model of the heart The anatomical model, which is a data file describing the distribution of the cell array and the initial value of fields in each cell, is independent of the cell program. The data file can be manually edited or generated by a program coded in C/C++. Due to the correspondence between each automaton cell and each biological cell, the building of the anatomical model is straightforward, even if a model has an irregular structure and a heterogeneous cell population. Usually, the Moore neighborhood is adopted, as is the case in our model. The data used to build the 3-D heart model are the axial images of the Visible Human Project (VHP) digital male cadaver [ 17 ], which contains about 125 thoracic slices at a 1 mm interval. Using an image processing program we developed, we enter a 128 × 128 coordinate system on each slice and use a computer mouse to mark the cells of different tissue with different colors (Figure 3 ). This process digitalizes each slice into a data file that reports the coordinates and type of each cell. The current heart model contains six kinds of cardiac tissues: sinoatrial node (SAN), atrioventricular node (AVN), atrium (AT), ventricle (VT), and trunk conduction bundle in the atrium (CBA) and in the ventricle (CBV). The distribution of trunk conduction bundles in each slice is determined by clinical experts, but the terminal distribution of the Purkinje fibers is generated by an algorithm at runtime. After processing all 125 slices, we use a program to merge the generated 2-D data files into a 3-D data file, which constitutes the anatomical model of the heart that occupies about 280,000 cells in the 128 × 128 × 128 cellular automata cell space. We also build an illustrative 2-D model containing the same kinds of cells and the same action potential models to evaluate the ECG simulation algorithms (Figure 6 ). The simulated normal and some abnormal ECG waveforms support the validity of the algorithm (Figure 7 ). Figure 3 Digitalizing slices of the digital male cadaver. Yellow color indicates ventricular tissue, and pink color indicates atrial tissue. Figure 6 The stratified heart walls. (a) The stratified 2-D model. (b) The epicardium-to-endocardium repolarization in the 2-D model based on stratified cardiac walls. Different color indicates different transmembrane potential. (c) Two ischemia areas based on stratified cardiac walls. (d) The stratified 3-D model (a section). Figure 7 The simulated ECGs with the 2-D model. (a) The normal ECG; two leads are at the middle of left and right chests (<190,50 > and <-80,50 >). (b) The normal ECG; two leads are in the cardiac cavities (<75,50 > and <32,50 >). (c) The ECG of endocardial ischemia (top line) and epicardial ischemia (bottom line); two leads are at the same positions as in (A). Two ischemia areas are shown in Figure 6(c). Heterogeneity, anisotropy and inhomogeneity The heterogeneity of a model is described in two ways. First, a specific field type , indicating cell type, is defined in the cell program, but whose value is initially stipulated in the anatomical model when we process tissue slices, such as: [x, y, z] = type, ...... x, y, and z are the coordinates. Values of other fields can be defined or not in the data file. With a nested if-then statement on the value of type, the cell program is divided into several parts, each being executed by cells of the specific type. Second, type can be modified at runtime to simulate pathological changes. After a cell type is changed, its program and therefore its behavior also change. In a tissue or organ, aside from heterogeneity, cells often show anisotropy and inhomogeneity that cannot be conveniently described while processing the raw data. To describe these two properties needs a bit of pattern formation programming [ 24 ], an interesting and challenging issue of biological modeling with cellular automata. In building the whole-heart model, the first relevant issue is cells at different layers in cardiac walls have different electrical properties [ 16 , 25 ]. To express this inhomogeneity, we developed a resolution-and dimension-independent algorithm to stratify cardiac walls into layers at the initial stage of runtime. The algorithm is comprised of three steps, and is run by all cells: Let the layer number of SAN cells (at least one SAN cell is on the epicardium) be 0 and the layer number of AVN cells (at least one AVN cell is on the endocardium) be 50. • If the current cell is neither a ventricular nor an atrial cell, then: if it connects to a cell whose layer is 0, its layer is 0; if it connects to a cell whose layer is 50, its layer is 50. • If the current cell is a ventricular or an atrial cell, then: visit all its neighboring cells and find out their minimal layer number, Min . The layer of this cell is Min+1 . A slice of the stratified 3-D model and of the stratified 2-D model is shown in Figure 6 . After stratification, the layer number of each ventricular cell is added to its HH equations as a special parameter to control and adjust its action potential duration (we do not use the layer number of atrial cells). This procedure ensures that epicardial cells have shorter action potential duration than endocardial cells, and the epicardium-to-endocardium repolarization is naturally and faithfully established. By assigning a large layer number to cells in the middle layers of the ventricular walls, an unusually long action potential duration is created, and the specific electrical property of M cells can be naturally simulated [ 25 ]. The slices of the digital cadaver do not provide any useful information on the distribution of terminal Purkinje fibers. Even if we know that most of them are located in the subendocardium, it is impractical to make a slice-by-slice manual description. Based on the stratified cardiac walls, this issue can be easily solved by using a lateral inhibition algorithm applied to cells on the ventricle subendocardium to generate a mesh-like Purkinje network. On completion of running the algorithm, some of the cell types are changed from ventricular cell to Purkinje cell. The more difficult engineering issue is the fine structure of cardiac cells and conduction fibers. Simulation with the 2-D model demonstrates that, even if the HH action potential model of Purkinje cells produces a very fast upstroke, which means a quick transjunctional conduction, yet a one-cell by one-cell conduction can never give normal propagation profiles as observed in Durrer's experiment, due to a too long transjunctional delay [ 26 ]. Nevertheless, our simulation shows that the rapid conduction in conduction fibers can be implemented by n -by- n cell communication among automaton cells of conduction fibers without physically building the fiber structure; n = 4 gives very satisfactory simulation results. Again, we propose that by using a pattern formation algorithm the physical construction of fiber structures is also feasible. In this scenario, a fiber is assembled by n successive automaton cells sharing the same, unique identity number, among which there is no conduction delay. Currently, with the 2-D model we find that even if letting each automaton cell stand for a discrete ventricular cell, the model works quite well in ECG simulation. Computation within and between cells Each automaton cell is a computing unit for action potential and ECG simulation. The electrical activity of each automaton cell of a specific cell type is described by the corresponding HH type action potential model. Five action potential models are employed to simulate activities of different cardiac cells (The cells of trunk conduction fiber in atriuma and ventricles use the action potential model of the Purkinje cell) [ 27 - 31 ]. Since the models are based on experimental data from different animal species, parameters of K + channels are slightly changed to produce action potential duration of human cardiac cells. Time constraints are the only reason for us to adopt early published, simpler action potential models. In the given simulations, all action potential models are solved using explicit Euler integration with a time step, Δt, of 0.01 ms [ 32 ]. An asynchronous adaptive time step method has also been developed to speed up simulation [ 49 ]. Electrical activities at channel (Figure 4 ), cell (Figure 5 ), and organ (Figures 6 , and 7 ) levels are monitored and captured at runtime. Figure 4 The channel level electrical activities of a ventricular cell. (a) The transjunctional currents from eight neighboring cells. (b) The state of gating variables. (c) The transmembrane ionic currents. Figure 5 The cell level electrical activities of a ventricular cell. The top line is the transmembrane potential; the middle line is the transmembrane current; the bottom line is the stimulating current received from neighboring cells. The cell is under progressive ischemia, which can be reflected in the change of action potential. Each automaton cell in the 3-D model has 26 neighbors, linked by gap junctions. When a cell depolarizes, driven by the potential difference between it and its neighboring cells, transjunctional currents generate and propagate to neighbors through gap junctions. Simulations with the 2-D model show that the simple static gap junction model, in which the resistance of the gap junction is a constant and the transjunctional current follows Ohm's law, works quite well. The dynamic gap junction models, in which the resistance of gap junction changes in a nonlinear manner with membrane potential, will significantly increase computational time [ 33 , 34 ] and create an extra burden for large-scale modeling. Values of gap junction resistance between different cells and in different directions are initially set according to available experimental data, and then tuned via simulation according to Durrer's experimental observations [ 26 ]. Excitation conduction between cells of the same layer follows an end-to-end propagation, and between cells of different layers is a side-to-side process. Parameters are also adjusted to fix the ratio of side-to-side conduction speed vs. end-to-end conduction speed to be one-third in ventricular cells and one-tenth in atrial cells [ 35 ]. Gap junction resistance can be modified at runtime to examine its effect on excitation propagation. For each cell in each round of computation, before solving the HH equations, the transmembrane potentials of all neighboring cells are checked and the transjunctional currents computed and summed to get the stimulating current, I stim , that the cell receives. The direction and speed of electrical propagation within the heart are jointly controlled by: (1) the stratification of the ventricular walls, (2) the speed of end-to-end and side-to-side conduction, (3) the ratio of end-to-end conduction speed to side-to-side conduction speed, and (4) the distribution of the trunk conduction bundle. Description of pathological activity In the cardiac model built with the extended cellular automata, abnormal electrical activities are grouped into four classes. The first class is based on anatomical defects in the heart that can be described by changing the type of some cells, either in the anatomical model or in the cell program. The Wolff-Parkinson-White syndrome caused by atrium-ventricle bypasses is a representative case. The second class is resulted from aberrant environments, especially abnormal ionic concentration (e.g. hyperkalemia) which significantly influences action potential. Abnormality of transjunctional conduction constitutes the third class, and the fourth class is is comprised of anomalous dynamics of cellular electrical activity itself. Changes in the HH equations can simulate these aberrations. In most cases, pathological changes affect more than one aspect. For example, in addition to providing an abnormal cell environment, ischemia also results in changed cellular electrical activity and altered gap junction resistance. To simulate the effect of ischemia, a special field blood, with normal value 1.0 is defined in the cell program, and introduced as an extra parameter in action potential and in gap junction models. To multiply the conductivity of the gap junction by an abnormal value of blood (e.g., blood 0.7) can simulate a lowered gap junction conduction speed. A small value for blood , when inserted into the equation computing Ca ++ current, affects the generation of the action potential. Dynamically modified blood can naturally simulate progressive ischemia (Figure 5 and Figure 7 ) [ 36 , 37 ]. Combining these factors, a variety of arrhythmias can be simulated. For example, by partially or completely blocking the activity of cells, or the conduction of gap junctions at specific locations, various conduction blocks such as AV (atrium-ventricle) block, LBBB (left bundle branch block) and RBBB (right bundle branch block) can be conveniently simulated. Many rapid arrhythmias are triggered by successive ectopic beats. Such abnormal beats can be produced in the model by providing cells at specific locations with extra stimulation or by changing them from atrial/ventricular to SAN cells. For abnormal propagations such as long QT syndrome, caused by the extra long refractory period of M cells, we can adjust, either dynamically or statically, the layer number of cells at the middle layers of the ventricular walls. Complex spatiotemporal patterns of arrhythmias are formed through discrete cell-cell communication. By these means, arrhythmic electrical activities can be simulated in flexible ways, different from those models created by cable equations or other partial differential equations (PDE). ECG simulation by computing field potential of every cell Linking ECG simulation directly with cellular and channel electrical activity is crucial for the understanding and treatment of arrhythmias. Since each automaton cell in our model is a computing unit, a cell level ECG simulation algorithm is developed to compute each cell's field point potential. The algorithm, implemented as a backend function coded in C language, shows strengths, because such cell level simulation builds a link between the action potentials of cells and the ECG waveforms. The limitation here is that this heart model does not include the chest, so that the contribution of thoracic tissues to body surface potential is much simplified. The model reads the membrane potential of every automaton cell and computes its field potentials at the standard lead locations. Several assumptions are made in the computations on account of established physical laws. (a) Each (automaton) cell has a spherical shape. Thus, the transmembrane potential, V m , is uniform on all parts of the cell membrane, except at gap junctions. (b) Gap junctions occupy the same area and are symmetrically distributed on each cell. (c) The distance from cell to field point is long enough so that the two solid angles subtended by the positive and negative sides of a cell membrane to a field point are equal. (d) Dielectric effects on field potential are neglected. σ media is the average conductivity of the 1-D tissues between a source cell and a field point; σ e the conductivity of the intercellular matrix; and σ i the conductivity of the intracellular cytoplasm. (e) For any neighboring cells, j and k , σ je = σ ke , σ ji = σ ki and φ je = φ ke , φ e is the potential in intercellular space. For convenience, we present the algorithm for the 2-D model here; the 3-D case is similar. The equation computing the field potential of an isolated cell at field point P is [ 38 ] σ i φ i - σ e φ e is the double layer strength of the cell membrane. When there is no propagation on the cell membrane, the field potential is zero. If the cell connects with eight adjacent cells, equation (1) becomes: Here, σ i φ i - σ ji φ ji is the double layer strength of the gap junction between the cell and its j th neighbor. The transjunctional potential difference between the two cells then becomes: φ i - φ ji = φ i - φ e + φ e - φ ji = φ i - φ e + φ je - φ ji = V m - V jm (3) S1 to S8 are areas of the eight gap junctions; Ω 1 to Ω 8 are solid angles subtended by them; and S0 denotes the remaining part of the cell membrane (Figure 8 ). Let the membrane area positive to P be S B and the area negative to P be S A . If: (a) S B = S A , (b) the same and even number gap junctions distribute symmetrically on S B and S A , and (c) all gap junctions have an equal area, then it can be proved that: Figure 8 Computing the field potential of single connected cells. This figure shows the distribution of gap junctions on the membrane of cells in a 2-D cell array and how gap junctions contribute to field potential. φ i is the potential within the current cell; φ 1 i is the potential within the first neighboring cell; is the normal direction of the transjunctional potential difference between the current cell and its first neighboring cell; ... Ω is the solid angle subtended by the current cell at field point P. Here, Σ K A is the sum of areas of gap junctions on S A , and Σ K B is the sum of areas on S B . In this circumstance, the first term in equation (2) is zero, and only the second term makes a contribution to field potential. The equation now becomes: Considering all Sj (j = 1...8) to be equal, and cos α 1 = cos α 5 = cos 90° = 0, we have: Here, is a constant that affects only the baseline, but not the shape of the ECG waveform. The field potential at point P, generated by the whole heart, is the superposition of all cells' contributions: Here, n is the number of automaton cells, and φ j (P) is computed with equation (5). Due to the time needed for the numerical solution of massive HH equations, the computing of σ media R 2 has to be much simplified, as described in equation (7), where σ media_i is the conductivity of tissue i , and R i is the length of tissue i on the line between the source cell and the field point P. As noted above, both the boundary effect and order of dielectric distribution are neglected. The 2-D (3-D) boundary between two tissues is treated as a 1-D boundary, and the ordered dielectric distribution is treated as an unordered distribution. Because of the huge number of cardiac cells and various possible field points on body surface, epicardium, endocardium and cardiac cavities, it is impossible to deal with numerous situations of boundary conditions between cells and field points. The method we adopt here is to make a runtime traverse from the source cell to the field point and meanwhile check the conductivity of each tissue and determine the 1-D space it occupies. The scaled distance is 1.0 between two perpendicular-connected cells, and 1.414 between two oblique-connected cells. Although the electrical property of non-cardiac tissues is much simplified, the electrical property of cardiac cells is intensively considered. Cardiac cells under different conditions have different conductivity [ 39 ]. It is σ Depo = 0.4 mhos/m in depolarization, σ Repo = 0.25 mhos/m in repolarization, σ Rest = 0.2 mhos/m in resting state, σ Infarct = 0.1 mhos/m in acute infarcted area, and σ Ischemia = 0.16 mhos/m in ischemic area. We let σ Blood = 0.6 and σ Chest = 0.1 be the average conductivity of tissues in chest [ 40 , 41 ]. We find ECG simulation is not clearly impaired by this simplified dielectric description, but benefits from the intensified cardiac tissue description. Results Factors that influence performance of the parallelized cellular automata model For the physically parallelized whole-heart electrophysiological model built with the extended cellular automata, several factors impair the simulation efficiency on a cluster of SMP. The first difficulty is the graphic display of the simulation. The main display window, as the critical resource in the parallelized program, can only be sequentially accessed by the cells. Thus, to display the updated state of all cells leads to a significant decrease of running speed. We mitigate this problem by displaying the cell state every 20 or even 50 steps, instead of every single step. The second problem is the overhead of communication among slaves located on different nodes. The time cost of communication, spent on the exchange of boundary sheets between neighboring slaves in each round, rises with the increase of slave number. An extreme case for a 128 × 128 × 128 model is that there are 128 nodes, and each slave deals with just one sheet, requiring every sheet to be exchanged in each round. The physical link between nodes also significantly affects performance. An additional limitation is that the overhead of fork/join operations in the OpenMP parallelized program degrades performance, although not very significantly. The OpenMP directive parallel can be inserted either before the endless while loop, or more simply, before the three for loops. If it is inserted before the for loops, the fork and join operations, which create and delete threads and allocate and collect memory for temporary variables in each thread, will be repeatedly executed in each round. The preferable way, which eliminates this unnecessary expenditure, is to insert the directive omp parallel before the while loop, and the directive omp for before the for loops. Between the omp parallel and the omp for , omp master is used to limit parallelism to only the cell program part. Finally, load balance becomes a limitation when a model has an irregular and/or heterogeneous structure. The heart, with four chambers and an uneven shape, is a typical case. Usually, on a cluster consisting of m identical nodes each containing n CPUs, a program is evenly dispatched to distribute the cell space across all nodes. We find that using this strategy, the 3-D heart model cannot reach the best load balance and performance because different nodes deal with different numbers of cardiac cells. Furthermore, even if each node contains the same number of cardiac cells, since different cells run different action potential models, the burden of computation remains unequal. Only a solid cubic model with homogeneous cells occupying the full cell space can ensure a best performance. Thus, if a model needs to run many simulations, an important issue is to find the best cell space partition. Evaluation of the 3-D cardiac model Pilot runs with the 3-D model have been made on a 4-node SUN computer cluster. The head node has eight UltraSparc CPUs, and the other nodes each have four identical CPUs. Five processes, 1 master and 4 slaves , are created at each run. The master and a slave are assigned to the head node, and each remaining slave is assigned to a node. Within each slave , four threads are created using the OpenMP parallel directive. Figure 9 gives the performance results for an even partition strategy. With this partition, due to the irregular structure of the heart and the heterogeneity of cardiac cells, the combination of 4 slaves -4 threads does not give the best performance. The 4 slaves- 2 threads setting behaves better because the overall cost is lower. . However, when the whole cell space is occupied by ventricular cells, the 4 slaves -4 threads version provides the best performance, as predicted. Figure 9 The performance of running the 3-D model on a cluster of SMP nodes. (a) When the 3-D cell space is sparsely occupied by the heart model (left), the evenly dispatch strategy does not produce the best performance. (b) When the cell space is fully occupied by ventricular cells (right), the best performance is guaranteed. Discussion Arrhythmias are a group of complex syndromes not well understood. Many difficult issues exist when investigating arrhythmias through computational modeling. We cover here only a few relating to model building and running. We adopt early published action potential models in the current 3-D model because they are simpler and more computationally affordable. In the first version of the Luo-Rudy model [ 30 ], there are six ionic channels, and the fluctuation of ionic concentration is not described. In the second version, there are many more channels, and the dynamic ion concentrations are described by another large group of ODEs [ 42 ]. It is straightforward to upgrade the cardiac model by replacing old action potential modelswith new ones. However, action potential models are not the whole story for a cardiac model. One has to make a compromise between the complexity of action potential models and the resolution of the cardiac model, because the computational burden also is due to the latter. Pollard et al . [ 43 ] in 1993 reported a cardiac model containing 400,000 computing nodes, but a much simplified action potential model was used [ 44 ]. This strategy was also adopted by other modelers [ 45 ]. We argue that full action potential models are important for modeling and understanding arrhythmias, especially those triggered by abnormal ionic electrical activities. However, to simulate more complex spatio-temporal cardioelectrical activities such as spiral waves in fibrillation, the current resolution of the model is undoubtedly insufficient. Technically, it is not difficult to adopt a higher resolution with more precise data, such as the digital female cadaver slices, which have an interval of 0.33 mm [ 17 ]. To double the resolution using a 256 × 256 × 256 coordinate system is another choice. By either method, the cell program, the pattern formation algorithms, and the ECG simulation algorithm remain the same – they are resolution-independent. This is a prominent and beneficial feature of such cellular automata modeling. Although experimental evidence show that fiber orientation takes a role in arrhythmia generation [ 46 , 47 ], to implement a highly realistic description of fiber orientation is expensive because of the difficulty in acquiring sufficient validated data [ 48 ]. Encoding these data into models is also quite complex, requiring the use of complex mathematical tools such as tensors. In our model, a new approach is proposed for stratifying heart walls. If fibers are simply built with cells of the same layer, they automatically acquire an arc shape (Figure 6d ) that can assume various orientations. Comparison of effectiveness between the two methods is currently not available due to the lack of sufficient simulations. Another issue relates to ECG simulation. Although simulation results of the 3-D model have not been acquired thus far, the simulated ECGs with the 2-D model in different lead locations and under physiological and pathological conditions are impressive and qualitatively agree with recorded ECGs, supporting the validity of the algorithm. Theoretically, the boundary effect of dielectrics is not negligible for field potential computing; yet in practice simplifications are often inevitable. So far we find that our simplified treatment of boundary effect on field potential does not visibly affect ECG simulation. There may be two reasons for this. First, the conductivity of the tissues (not including the heart) in the chest may not be significantly different. Second, the boundaries among different tissues are so irregular that the net boundary effect may be effectively neutralized. On the other hand, we find that to let cardiac cells under different conditions (depolarization, repolarization, resting state, and ischemia) have different conductivity can improve the ECG quality, indicating that the precise description of the source may be more important than the precise description of the dielectric. Finally, we point out that, as in single cell models where adaptive time steps can greatly improve the running performance, this same strategy can produce the same results in multicellular models built with this cellular automaton [ 49 ]. Simulation with the 2-D model shows that an overall speed improvement of 4.5 is reached. Conclusions Effective simulation of arrhythmias needs a whole-heart model, differential description of electrical properties of cardiac cells, membrane equation based computation, association between cellular activities and ECG generation, flexible description of pathological conditions, and long running time. To comprehensively address these issues, we develop a method based on cellular automata and parallel computing technologies to build large-scale electrophysiological models with extended cellular automata, and run such models on clusters of shared memory machines. The dynamically traced and captured electrical activities at channel, cell, and organ levels can substantially help us understand abnormal cardioelectrical activities through simulation. Simulation results with the 2-D model support the validity of the ECG simulation algorithm. Transparent cluster computing is a convenient and effective solution to the excessive time consumption of computational intensive simulation. In addition to reaching a mechanistic understanding of arrhythmias, an important goal of in silico research is to facilitate the discovery and evaluation of drugs. This helps to reduce the risk and cost of clinical trials, shorten the cycle of development, and remove randomness in candidate screening. A whole-heart electrophysiological model that links electrical activities at channel, cell, and organ levels can help achieve this result. The modeling method described in this paper shows the advantages of precisely linking cell and organ activities, exploiting the intrinsic parallelism in tissue/organ level biological activities. Besides modeling electrical activity, the method is also applicable to many other multicellular models in which quantitative description is required. Authors' contributions HZ develops the methods and built the initial 3-D model. AM provides data of conduction system distribution in ventricles. YS and GR help solve some technical issues. PD is responsible for the project.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517726.xml
529435	Oxytocin and cholecystokinin secretion in women with colectomy	Background Cholecystokinin (CCK) concentrations in plasma have been shown to be significantly higher in colectomised subjects compared to healthy controls. This has been ascribed to reduced inhibition of CCK release from colon. In an earlier study CCK in all but one woman who was colectomised, induced release of oxytocin, a peptide present throughout the gastrointestinal (GI) tract. The aim of this study was thus to examine if colectomised women had a different oxytocin response to CCK compared to healthy controls. Methods Eleven women, mean age 34.4 ± 2.3 years, who had undergone colectomy because of ulcerative colitis or constipation were studied. Eleven age-matched healthy women served as controls. All subjects were fasted overnight and given 0.2 μg/kg body weight of CCK-8 i.v. in the morning. Samples were taken ten minutes and immediately before the injection, and 10, 20, 30, 45, 60, 90 and 120 min afterwards. Plasma was collected for measurement of CCK and oxytocin concentrations. Results The basal oxytocin and CCK concentrations in plasma were similar in the two groups. Intravenous injection of CCK increased the release of oxytocin from 1.31 ± 0.12 and 1.64 ± 0.19 pmol/l to 2.82 ± 0.35 and 3.26 ± 0.50 pmol/l in controls and colectomised women, respectively (p < 0.001). Given the short half-life of CCK-8 in plasma, the increased concentration following injection could not be demonstrated in the controls. On the other hand, in colectomised women, an increase of CCK in plasma was observed for up to 20 minutes after the injection, concentrations increasing from 1.00 ± 0.21 to a maximum of 1.81 ± 0.26 pmol/l (p < 0.002). Conclusion CCK stimulates the release of oxytocin in women. There is no difference in plasma concentrations between colectomised and controls. However, colectomy seems to reduce the metabolic clearance of CCK. The hyperCCKemia in patients who had undergone colectomy is consequently not only dependent on CCK release, but may also depend on reduced clearance.	Background The gut hormone cholecystokinin (CCK) is synthesised in endocrine I cells in the mucosa of the upper small intestine [ 1 ] and is released into the blood after ingestion of fatty and protein-rich meals [ 2 ]. CCK has various effects on the gastrointestinal (GI) tract and acts on afferent vagal nerves [ 3 ], neurons of the myenteric plexus [ 4 ], and directly on muscle cells [ 5 ]. It is also synthesised in central neurons including hypothalamic, oxytocinergic neurons [ 6 ]. Circulating CCK is degraded in several sites, namely the kidney, liver and gut [ 7 , 8 ]. Oxytocin is synthesised in the supraoptic and paraventricular nuclei of the hypothalamus as part of a larger precursor polypeptide [ 9 ]. While the main effects are in the myoepithelial cells and uterine smooth muscle in the responses associated with the milk ejection reflex and parturition, the possibility has been raised that oxytocin also contributes to control of the GI motility [ 10 , 11 ]. Both exogenous and food-stimulated endogenous CCK stimulates the pituitary secretion of oxytocin in the rat through CCK-receptors on afferent vagal neurons [ 12 ]. In hypothalamus, both parvocellular neurons projecting to the dorsal vagal complex, and magnocellular neurons projecting to the pituitary, secrete oxytocin in response to CCK [ 13 ]. We have recently found that CCK also leads to oxytocin release in healthy women [ 14 ]. However, one of the women included was colectomised, and she was the only one who had no release of oxytocin in response to CCK [ 14 ], although colectomy leads to higher concentrations of CCK in plasma [ 15 - 17 ]. We have found mRNA for oxytocin and its receptor throughout the GI tract [ 18 ], as well as the fully expressed proteins (unpublished observation). We do not know if this has an autocrine and/or paracrine role in the gut, or if it also is released into the blood as a hormone. The aim of this study was therefore to examine if women who had performed a colectomy, had a different oxytocin response to CCK than otherwise healthy women with an intact GI tract. Methods Subjects Eleven women from the Departments of Medicine and Surgery at Malmö University Hospital (mean age 34.4 ± 2.3 years, range 22–42 years) were studied. They had all a history of colectomy. Two were colectomised because of slow transit constipation (STC) and had undergone a subtotal colectomy with the creation of ileo-rectal anastomosis. Proctocolectomy with ileal pouch-anal anastomosis had been performed in seven of them because of ulcerative colitis, and one because of familial multiple polyposis. The last patient has an ileostomy after subtotal colectomy, saving the rectum, because of ulcerative colitis. Thus, the subjects were cured from their original conditions. The time interval between the proctocolectomy/colectomy and this study was 10–149 months, with a mean of 48.5 ± 12.6 months. Eleven age-matched healthy women with preserved GI tract served as controls. Physical examination and laboratory routine screening were all within normal limits in both groups. The body weight was 68.7 ± 5.8 kg in the patients and 73.8 ± 6.1 kg in the controls. No drugs and no oral contraceptives or other hormonal treatments were allowed in either group. The experiments were performed at no specific stage of the menstrual cycle. None of the included subjects had participated in our former study [ 14 ]. Protocols The protocols were approved by the local Ethics committee at the University of Lund, and written informed consent was obtained from all subjects before the study was started. The possibility of pregnancy was excluded in all women. Experimental procedure All subjects were fasted overnight. In the morning they were given 0.2 μg/kg body weight of cholecystokinin octapeptide (CCK-8) (Clinalfa, Switzerland) as an intravenous injection. This bolus was chosen as it was the only dose giving raise to a weak, but not significant, oxytocin release in an earlier study [ 19 ]. Blood samples were taken through an intravenous catheter 10 min before and immediately before the injection, and 10, 20, 30, 45, 60, 90 and 120 min after the injection. Hormone analysis All blood samples consisted of 8.0 ml whole blood drawn into iced heparinised tubes. The plasma was separated and frozen at -20°C immediately after the experiment. Oxytocin was measured as described by Balment et al [ 20 ] using the Fourth International Standard for oxytocin (76/575). The lower limit of detection for this assay was 0.1 pmol/l with intra-assay and interassay variations of 5.4 and 11.8 %, respectively, at 2.5 pmol/l. The hormone was extracted from plasma using C 18 Sep Pak Columns (Waters Associates Ltd., Northwick, Middx., U.K.). The concentrations of CCK in plasma were measured using a highly specific and accurate radioimmunoassay as previously described [ 21 ]. The limit of detection for his assay is 0.1 pmol/l with intra-assay and interassay variations of less than 5 % and 15 %, respectively, at both 3.7 and 15 pmol/l concentrations. Statistical analysis The values are expressed as mean ± standard error of the mean (SEM). The basal value is the mean of the two fasting values. The peak value is the mean of the highest concentration in every subject after the injection. The total plasma CCK and oxytocin response was assessed by calculating the area under the plasma concentration time curve (AUC). The Kruskal-Wallis followed by Wilcoxon signed ranks test were used for assessment of the significance of the differences within and between the two groups. The Spearman rank test was used for calculating the correlation between CCK and oxytocin concentrations in plasma. Probabilities of less than 0.05 were considered significant. Results Plasma oxytocin concentrations The basal oxytocin concentration in plasma was similar in the two groups. The concentration was stable before the start of the experiments. Injection of CCK-8 led to an increase of the oxytocin secretion compared to basal values in both groups (p < 0.001) (Table 1 ). The increase in plasma concentration of oxytocin was observed after 10 min and persisted throughout the study. The highest concentration was found after 20 min (Fig 1 ). There was no difference of the AUC between the two groups, neither there was any difference between each time point studied (Fig 1 ). Table 1 Basal and peak plasma values of cholecystokinin (CCK) and oxytocin CCK (pmol/l) N = 11 Oxytocin (pmol/l) N = 11 Controls Basal 0.7 ± 0.1 1.3 ± 0.1 Peak 0.9 ± 0.1 2.8 ± 0.4* Patients Basal 1.0 ± 0.2 1.6 ± 0.2 Peak 1.8 ± 0.3+ 3.3 ± 0.5* Values are expressed as mean value ± standard error of the mean (SEM). Comparisons are made within groups; = p < 0.01, *** = p < 0.001, and between groups; + = p < 0.05. Wilcoxon signed rank test. The basal value is the mean of the two fasting values. The peak value is the mean of the highest concentration in every subject after the injection. Figure 1 The plasma concentration of oxytocin before and at different time points after an injection of 0.2 μg/kg body weight of cholecystokinin-8 (CCK-8). There were 11 subjects in each group. Values are given as mean and standard error of the mean (SEM). There was no difference between the groups neither when calculating values at different time points studied nor the area under the curve (AUC). Wilcoxon signed rank test. = control, = patient. Plasma cholecystokinin concentrations There was a tendency towards higher basal CCK concentration in patients, although not significant (Table 1 and Fig 2 ). CCK-8 has a half-life in plasma of about < 1 min (8). Therefore, no increase in plasma CCK could be detected in the control group after the intravenous injection of CCK-8 (Table 1 ). However, in colectomised women, an increase in plasma CCK concentrations was found after the injection compared to basal values (p < 0.002). The difference of peak value between the groups was significant (p < 0.04) (Table 1 ). The AUC differed significantly between the two groups (p < 0.04), but no difference was observed between values at each time point studied (Fig 2 ). Figure 2 The plasma concentration of cholecystokinin (CCK) before and at different time points after an injection of 0.2 μg/kg body weight of CCK-8. There were 11 subjects in each group. Values are given as mean and standard error of the mean (SEM). When calculating the area under the curve (AUC), there was a significantly increased AUC in the colectomised subjects compared to controls (p < 0.04). No difference was seen between the groups when comparing values at each time point. Wilcoxon signed rank test. = control, = patient. There was no correlation between CCK and oxytocin concentrations (data not shown). Neither was there any difference in CCK and oxytocin concentrations between patients with different diagnosis and those who had rectum saved or resected (data not shown). Discussion This study shows for the first time that CCK-8 increases the secretion of oxytocin in women. We have previously shown that exogenous CCK-33 and -39, and a fatty meal with endogenous CCK release, led to enhanced concentrations of oxytocin in plasma [ 14 ]. One patient in that study was colectomised, and she was the only one in whom no increase in oxytocin release was seen after CCK stimulation. This observation prompted the present study. In this study, there was no difference in plasma concentrations of oxytocin between colectomised and healthy controls. Thus, the oxytocin secreted into the blood after CCK stimulation seems not to origin from the colon. The oxytocin recently found in the colon may participate in autocrine and/or paracrine regulation of the gut while having no endocrine effects [ 18 ]. The patient group examined in this study was not homogenous, but it was not possible to include enough young women with colectomy after ulcerative colitis. We have earlier described the presence of oxytocin and its receptor throughout the gut, without any efforts to quantify the expression [ 18 ]. Only the effect of colon on plasma concentrations of oxytocin was measured in the present study. We do not know from this study if oxytocin from some other part of the gut is released into the plasma. It is difficult to conduct an experiment to examine the origin from the oxytocin release. CCK acts on receptors on afferent vagal nerves to stimulate the oxytocin release from the pituarity [ 12 ], and these receptors are present throughout the GI tract [ 3 - 5 ]. Therefore it is not possible to use CCK-receptor antagonists to distinguish between central or local CCK effects. CCK has been shown to stimulate oxytocin secretion in mammals in many studies [ 12 - 14 ]. However, in a previous study, intravenous injection of CCK-8, in the same dose as in our study, did not increase the concentration of oxytocin in plasma [ 19 ]. This may depend on methodological differences. Another possible explanation to the difference is the effect of gonadal hormones on the regulation of the oxytocin release from the posterior pituitary gland. In our study, only women were included, whereas Miaskiewicz et al [ 19 ] examined 13 men and one woman. Orally administered estrogen stimulates oxytocin secretion, and progesterone also affects release [ 22 ]. Lower plasma levels of these hormones in men may explain the absence of increased oxytocin secretion in men. In addition, one study has shown that testosterone inhibits the secretion of oxytocin from the pituitary gland [ 23 ]. Oxytocin is present in plasma in men, although at lower concentrations [ 24 , 25 ], and shows a circadian rythm [ 26 ]. Oxytocin may have similar effects on the GI tract in men and women, although the plasma concentrations differ. The effects of oxytocin have been only rudimentary examined. However, oxytocin has in one study been shown to enhance gastric emptying [ 10 ], and in a yet unpublished study, we have found that an oxytocin-receptor antagonist delayed the gastric emptying rate (unpublished observation). Further, we have demonstrated increased colonic peristaltis after oxytocin stimulation in healthy women [ 11 ]. Our finding of oxytocin release in response to a meal [ 14 ], and the presence of oxytocin receptors on the cells that regulate the gut motility (unpublished observation), suggest oxytocin to play a physiological role in the GI function. Several studies have reported that after colectomy in different species there are higher concentrations of CCK in plasma, both basal and postprandial, compared to healthy controls [ 15 - 17 ]. It has been suggested that this is due to depletion of an inhibitory factor of CCK secretion which is released from the colon. Peptide YY (PYY) is secreted from distal ileum and colon, and CCK is known to stimulate PYY secretion from the hindgut [ 27 - 29 ]. PYY then inhibits further CCK secretion [ 30 , 31 ]. As PYY is secreted from the hindgut, this peptide is substantially reduced after colectomy [ 26 ]. Thus, the reduced PYY concentration may explain the hyperCCKemia. In this study, the elevated CCK concentrations in the group of colectomised women were not due to increased secretion of CCK, as CCK was injected exogenously. Instead, the hyperCCKemia in the group of colectomised patients seems to be due to reduced degradation of the peptide injected. CCK is degraded in the kidney, liver and gut [ 7 , 8 ]. Our hypothesis is that PYY could contribute to the degradation as well as the secretion of CCK. Receptors for PYY have been found in the kidney and on hepatocytes, and PYY influences the renal and hepatocyte metabolism [ 32 - 34 ]. Alternatively, the different CCK concentrations could be due to reduced degradation in the colon in addition to the kidney and liver, as CCK-8 has been shown to be degraded in the gut in pigs [ 8 ]. It remains to be determined which mechanism contributes most to the hyperCCKemia observed after colectomy; increased CCK secretion, or decreased clearance. In the present study, the basal levels of CCK did not differ significantly in colectomised, as observed in earlier studies [ 15 - 17 ]. CCK has a wide range of effects on the GI tract. Three physiological effects on gut motility have been identified; contraction of the gallbladder [ 2 ], relaxation of the sphincter Oddi [ 35 ] and inhibition of gastric emptying [ 2 ]. Its role on colonic motility is controversial. While Barone et al [ 36 ] were able to demonstrate contractions, Niederau et al could find no effect of CCK [ 37 ]. Further, CCK increases pancreatic enzyme secretion [ 2 ]. It is not known if the hyperCCKemia observed in colectomised patients [ 15 - 17 ] has any impact on GI motility or health. Conclusions CCK stimulates the release of oxytocin in women, probably via an effect on the neurohypophysial system. There is no difference in plasma concentrations between colectomised women and women with intact GI tract. The hyperCCKemia observed in patients who have undergone colectomy is dependent not only on an increase in CCK release, but may also depend on a reduced degradation. It was beyond the aim of our study to determine the clearance of CCK. However, this should be evaluated further. Abbreviations AUC = area under the curve CCK = cholecystokinin GI = gastrointestinal PYY = peptide YY SEM = standard error of the mean Authors' contributions BO designed the study, included patients, paid for the most, performed the statistical analysis and drafted the manuscript JR carried out the radioimmunoassay for CCK and participated in the writing process MF carried out the radioimmunoassay for oxytocin and participated in the writing process All authors read and approved the final manuscript. Competing interests The authors declare that they have no competing interests. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529435.xml
521069	A method for detecting and correcting feature misidentification on expression microarrays	Background Much of the microarray data published at Stanford is based on mouse and human arrays produced under controlled and monitored conditions at the Brown and Botstein laboratories and at the Stanford Functional Genomics Facility (SFGF). Nevertheless, as large datasets based on the Stanford Human array began to accumulate, a small but significant number of discrepancies were detected that required a serious attempt to track down the original source of error. Due to a controlled process environment, sufficient data was available to accurately track the entire process leading to up to the final expression data. In this paper, we describe our statistical methods to detect the inconsistencies in microarray data that arise from process errors, and discuss our technique to locate and fix these errors. Results To date, the Brown and Botstein laboratories and the Stanford Functional Genomics Facility have together produced 40,000 large-scale (10–50,000 feature) cDNA microarrays. By applying the heuristic described here, we have been able to check most of these arrays for misidentified features, and have been able to confidently apply fixes to the data where needed. Out of the 265 million features checked in our database, problems were detected and corrected on 1.3 million of them. Conclusion Process errors in any genome scale high throughput production regime can lead to subsequent errors in data analysis. We show the value of tracking multi-step high throughput operations by using this knowledge to detect and correct misidentified data on gene expression microarrays.	Background Expression microarrays, with the capability to measure the mRNA expression level of tens of thousands of genes simultaneously, have found broad application in both clinical and basic research [ 1 - 7 ]. With the generation of large data sets from microarray experiments, statistical methods are needed to extract useful information. Many methods have had specific implementations written for the analysis of gene expression data, such as various forms of clustering, self ordered maps, singular value decomposition and significance analysis [ 8 - 13 ]. However, these methods all rely on the assumption that there are no gross process errors in the original data. Previous analyses of systematic errors in microarray data have focused on problems at the level of sample preparation, labelling, or hybridization. This report focuses on steps in the microarray production process prior to hybridization that may ultimately result in errors in the underlying data. Much of the microarray data published at Stanford is based on mouse and human arrays produced under controlled and monitored conditions at the Brown and Botstein laboratories and at the Stanford Functional Genomics Facility (SFGF). Nevertheless, as large datasets based on the Stanford Human array began to accumulate, a small but significant number of discrepancies were detected that required a serious attempt to track down the original source of error. Due to a controlled process environment, sufficient data was available to accurately track the entire process leading to up to the final expression data. In this paper, we describe our statistical methods to detect the inconsistencies in microarray data that arise from process errors, and discuss our technique to locate and fix these errors. We are able to fix those errors that originate at the level of any microtiter plate used during a multi-step microarray production process. The major process fail points in cDNA microarray production at Stanford are shown in Table 1 . It is at these points that misidentifications can occur. Other types of processes resulting in expression data will have their own possible fail points. Regardless of the particular process, for the sake of subsequent error checking it is important to track every instance where samples have moved from one microtiter plate to another, or from microtiter plate to microarray. Our process involves four such reallocations: From an archival 96-well block to a 96-well growth block, then to a 96-well PCR plate, then to four 384-well print plates and finally to 250 10–40 thousand element microarrays. Even in highly automated bar-coded and vision controlled systems the possibility exists that plates might become swapped, skipped, misordered, or rotated by 180 degrees during one of the process steps. Our own process is fairly well automated, but like most academic and commercial efforts our process involves hand loading of robots. For example in our case, during the transfer from 96 to 384-well format, it is possible to accidentally rotate a 96-well plate or misorder the 96 or 384-well plates. During printing a 384-well plate might be accidentally skipped, swapped or printed in the wrong orientation (rotated by 180 degrees). Even with the best engineering controls to prevent plate rotations, the potential exists for misidentified plates. Inconsistencies in our data were first detected both by visual inspection of microarray data as well as the appearance of anomalously large first components in singular value decomposition analyses of Ovarian Tumour data [ 6 ] traced to different production batches of arrays. Our algorithm was developed to detect and repair these types of errors, using the similarities in expression levels between sets of spots from different microarrays. The algorithm was used to check all of the microarray data in our database for which there was a sufficient process record. The general idea, illustrated in Figure 1 , is as follows: Partition the microarray expression data from a single microarray into sets based on the various microtiter plates the samples have been in throughout their process history. For example, at Stanford we keep our process fairly simple, with a minimum of plate changes, so our spotting material can be said to have existed in either 96 well or 384-well format during its entire process history. Thus, we partition the data into sets corresponding to the 96 or 384-well plate history. Next, an expression vector for each element of the partition (each plate) based on absolute (not relative) expression levels is formed and compared to every other expression vector from every other plate on many other array batches. A rank matrix of correlation coefficients is formed which should be close to unity on the diagonal and far from unity off the diagonal. Non-unitary diagonal elements indicate problems with that plate comparison. A rank comparison of the best correlations can be made to find swapped plates. The algorithm can be repeated assuming a rotated plate to check to see if the discrepancy can be attributed to a plate rotation. It should be noted that in cases where no process error is indicated, the method can still indicate the presence of problem plates, print batches or PCR rounds that should be flagged for particular attention in downstream analyses. Our algorithm (named MuFu for "MixupFixup") is for arrays produced in the Brown and Botstein laboratories and the Stanford Functional Genomics Facility. However, the ideas are general and can be applied to many other types of high throughput operations. In most of the research studies using our human microarray, a common reference [ 14 ] is compared against a tumour or tissue specimen. The common reference is usually labelled with the Cy3 dye, pseudo-colour green in most visualizations of the data, while the sample specimens are labelled with Cy5 dye, pseudo-colour red. Subtle corrections due to background subtraction issues and normalization are not important for this analysis. Because a common reference is used in a large number of hybridization experiments at Stanford, all of the Cy3 intensities (or Cy5 in some cases) across various kinds of experiments are comparable. We measure the similarities of two sets of spots by taking the correlations between the common reference intensities of these two sets of spots. For those experiments that do not use common reference, the comparison is made using experiments with similar samples in either the Cy3 or Cy5 channel. Figure 1 MuFu flowchart. Flow of the MuFu algorithm. Looping and re-partitions of the data are not shown. Results Finding misidentified data This example is from the analysis of experiments from an ovarian tumor dataset[ 6 ] that first led us to develop MuFu. Here, it was noticed that similar tissues were not clustering across batch boundaries as expected. Also, an anomalously large first component in a singular value decomposition analysis[ 9 ] pointed to problems at the array batch level. After some detective work, visual inspection uncovered anomalies in certain print plates as seen in Figure 2 . We were able to isolate the problem to distinct sets of 96-well plates that had been swapped during an upstream process step, probably during the transfer from 96-well plates to 384-well plates. Not wishing to go through this sort of process again and again, MuFu was developed to more succinctly recapitulate this finding and apply it to all data. Figure 2 Ovarian tumour data. Visual inspection of the anomalous spots from the Ovarian tumor data before and after applying MuFu. In each case the top row shows that the feature alignment is inconsistent with the contents of the plate. In the bottom row features group together as expected. With MuFu, a check using a 384-well plate partition of the data shows a discrepancy but no obvious plate rotation or misidentification. We then repartition and repeat our tests at the 96-well plate level. The results are shown in Table 2a where we show a 12 way pair wise comparison. The first four comparisons are for batches whose print plates are made from the first PCR round. The middle four compare the first PCR round to the second, and the last four compare batches from the second PCR round. Mismatches are evident in the middle set of comparisons. From Table 2b we see that the distance matrix identifies a match between plate h and plate i and plate n and plate o for all four comparisons of the first PCR round to the second. Table 2 Test for 96 well plate handling error. a) In this table we see that the middle bank of comparisons indicates a discrepancy between data from the first PCR run and the second, at the 96-well plate level. b) A check of the distance matrix results show that the swapped plates are h and i in one case, and n and o in the other. Bold indicates P-value = 0.3, italic indicates P-values between 1.0E-03 and 1.0E-04, while regular font indicates P-values < 1.0E-04. a) R_{m, m} PCR Round 1 vs 1 PCR Round 1 vs 2 PCR Round 2 vs 2 Plates A1 v A2 A2 v A3 A3 v A4 A4 v A5 A2 v B1 A3 v B2 A4 v B3 A5 v B4 B1 v B2 B5 v B3 B1 v B4 B5 v B4 {a, a} 1 1 1 1 1 1 1 2 1 1 2 1 {b, b} 1 1 1 1 1 1 1 1 7 1 3 1 {c, c} 1 1 1 1 1 1 1 1 1 1 1 1 {d, d} 2 1 1 1 1 1 5 9 1 1 1 1 {e, e} 1 1 1 1 1 1 1 2 1 1 2 1 {f, f} 1 1 1 1 1 1 1 1 1 1 1 1 {g, g} 3 1 1 1 1 1 1 1 1 2 1 1 {h, h} 1 1 1 1 343 322 408 294 1 1 1 1 {i, i} 2 1 1 1 421 290 402 359 1 1 1 1 {j, j} 1 1 1 1 2 1 1 1 2 1 4 1 {k, k} 1 1 1 1 2 1 1 1 1 1 1 1 {l, l} 1 1 1 1 1 2 1 1 1 1 1 2 {m, m} 1 1 1 1 1 1 1 1 1 1 1 1 {n, n} 1 1 1 1 255 141 20 167 1 1 1 1 {o, o} 1 1 1 1 330 248 357 277 3 1 5 6 {p, p} 1 1 1 1 1 3 1 1 3 2 1 1 {q, q} 1 1 1 1 4 3 3 2 1 1 1 1 {r, r} 1 1 1 1 1 1 1 5 1 1 5 1 {s, s} 3 1 1 1 1 1 1 1 1 1 1 1 {t, t} 1 1 1 1 1 1 1 1 1 1 1 1 {u, u} 1 1 1 1 1 1 1 1 1 1 1 1 {v, v} 1 1 1 1 1 1 7 1 1 1 1 1 {w, w} 2 1 1 1 1 16 1 2 1 1 1 1 b) A2 v B1 A3 v B2 A4 v B3 A5 v B4 D_{mm} D_{mn} D_{mm} D_{mn} D_{mm} D_{mn} D_{mm} D_{mn} MEAN 0.33 0.97 0.34 0.89 0.44 0.98 0.37 0.96 SD 0.2 0.11 0.2 0.13 0.21 0.11 0.21 0.12 {h, h} 0.99 1.02 1.07 1 {i, i} 1.09 1 1.07 1.06 {h, i} 0.22 0.09 0.33 0.31 {i, h} 0.28 0.07 0.11 0.14 {n, n} 0.98 0.91 0.85 0.92 {o, o} 1.01 1.02 1.07 1.02 {n, o} 0.1 0.51 0.45 0.52 {o, n} 0.04 0.17 0.11 0.05 In all four comparisons, the two distributions resolve themselves well, as can be seen in Figure 3 , leading to the conclusion that h and i are swapped, and n and o are swapped, most likely in the plates from the first PCR round. The ambiguity is broken by sequencing a small sampling of wells, which confirms that the misidentifications are in the print plates from the first round of PCR, and not the second. In Figure 4 we show the effect on the Ovarian data after the correction has been applied. Also, in Figure 2 we show how, via visual inspection of the actual spots on an array, one can verify that the fix has properly reorganized the data. The figure shows spots from the four different affected 96-well plates before and after the fix is applied. Figure 3 Match, mismatch distance distributions. Good separation between match and mismatch distance distributions at the 96-well plate level lends confidence to our ability to discriminate between chance matches and actual matches. The green bars refer to the distance distributions of matched plates and the red bars for mis-matched plates. Figure 4 Ovarian tumour clusters. In Ovarian tumor data it was noticed that similar experiments were not clustering as expected (upper cluster diagram). Using MuFu we were able to isolate the problem to a distinct set of 96-well plates that had been swapped during an upstream process step, probably during the transfer from 96-well plates to 384-well plates. The samples cluster together as expected after correction (lower cluster diagram). Finding a 384-well plate rotation We compare four arrays A1, A2, A3 and A4 from a particular print production batch A to four arrays B, C, D and E from four other distinct print production batches. The total number of 384 well microtiter plates in print batch A is 113. The results of the plate rotation test are shown in Table 3a . Here we see plate p , identified in both the rank and rotated-rank matrices across print batches is a clear candidate for a plate rotation. The data from the distance matrix comparisons is shown in Table 3b . In Figure 5 we show the distributions of the diagonal elements of the distance matrix and the rotated-distance matrix. These distributions are well resolved and allow us to easily distinguish data due to a plate rotation from noisy data. Table 3 Test for rotated plates. a) Plate rotation results from the rank matrices R and R' . The flagged comparisons indicate a plate rotation for plate p . b) Plate rotation distance matrix comparison. The data meet the criteria for a plate rotation. a) A1 v B A2 v C A3 v D A4 v E Plates R_{mm} R'_{mm} R_{mm} R'_{mm} R_{mm} R'_{mm} R_{mm} R'_{mm} {i, i} 1 102 1 95 1 82 1 93 {j, j} 1 18 1 102 1 16 1 64 {k, k} 1 52 1 99 1 102 1 41 {l, l} 6 108 3 35 9 76 1 22 {m, m} 1 51 1 79 1 85 1 68 {n, n} 1 14 1 35 1 41 1 93 {o, o} 1 92 1 92 1 102 1 76 {p, p} 78 1 48 1 46 1 65 1 {q, q} 1 117 1 96 1 87 1 111 {r, r} 1 118 1 110 1 115 1 112 {s, s} 1 116 1 108 1 109 1 113 {t, t} 1 69 1 93 3 20 1 103 {u, u} 1 116 1 103 1 72 1 84 {v, v} 1 83 1 104 1 85 1 64 b) A1 v B A2 v C A3 v D A4 v E D_{mm} D'_{mm} D_{mm} D'_{mm} D_{mm} D'_{mm} D_{mm} D'_{mm} MEAN 0.26 0.96 0.23 0.88 0.19 0.81 0.21 0.88 SD 0.14 0.1 0.14 0.14 0.1 0.17 0.13 0.14 {pp} 0.95 0.84 0.85 0.97 {pp} 0.21 0.14 0.23 0.24 P-value 4.10E-07 3.20E-14 6.60E-06 6.30E-08 2.10E-11 3.20E-04 2.50E-09 2.40E-06 Figure 5 Plate rotation distance comparison. Histogram of the distance comparisons for the plate rotated and not rotated cases. The distributions are well separated, where the green bars refer to the distance distributions for non-rotated plates and the red bars for rotated plates. Finding a misidentified 384-well plate In this example we compare four arrays A1, A2, A3 and A4 from a particular print batch A to four other arrays B, C, D and E from four other distinct print batches. As seen in Table 4a no plate is found to be a candidate for plate rotation, however we do find that plate q has a poor self-match comparison. Indeed, when the distance matrix is examined, we see the plate matches instead plate r across the four array comparisons. We check the three criteria for plate misidentification and summarize these data in Table 4b . In Figure 6 we show the distributions of the match and mismatch distributions. These distributions are quite distinct and give us confidence that we can distinguish the proper match from an accidental match. Table 4 Test for swapped plates. a) Plate r is seen here to have a problem, but from the table we see that it is most certainly not a plate rotation. b) A check of the distance matrix shows the {r, q} comparison to be quite good and satisfies the criteria for a match, indicating that plates r and q have been accidentally swapped. a) A1 v B A2 v C A3 v D A4 v E Plates R_{mm} R'_{mm} R_{mm} R'_{mm} R_{mm} R'_{mm} R_{mm} R'_{mm} {i, i} 1 82 1 48 1 107 1 76 {j, j} 1 45 1 87 1 87 1 89 {k, k} 1 18 1 45 1 63 1 15 {l, l} 1 85 1 106 1 103 1 93 {m, m} 1 30 1 49 1 53 1 37 {n, n} 1 115 1 113 1 37 1 102 {o, o} 1 24 1 64 1 82 1 100 {p, p} 1 110 1 102 1 104 1 93 {q, q} 1 118 1 111 1 110 1 91 {r, r} 77 7 48 12 51 22 16 23 {s, s} 1 55 1 44 1 90 1 88 {t, t} 1 115 1 101 1 110 1 99 {u, u} 1 118 1 93 1 74 1 95 {v, v} 1 15 1 13 1 29 1 71 b) A1 v B A2 v C A3 v D A4 v E D_{mm} D_{mn} D_{mm} D_{mn} D_{mm} D_{mn} D_{mm} D_{mn} MEAN 0.19 0.93 0.25 0.94 0.28 0.95 0.2 0.81 SD 0.14 0.07 0.19 0.07 0.17 0.07 0.16 0.1 {rq} 0.3 0.26 0.25 0.34 {rr} 0.97 0.95 0.98 0.72 P-value 1.30E-08 0 1.10E-04 0 1.90E-05 0 5.80E-04 1.30E-06 Figure 6 Misidentification distance comparison. Misidentified plate distance histogram showing good separation between the match and mismatch distance distributions. The green bars refer to the distance distributions of matched plates and the red bars for mis-matched plates. Discussion Out of the 265 million features checked in our database using MuFu, problems were detected and corrected on 1.3 million of them. That we were able to find and correct both previously known and unknown problems gives us confidence in the algorithm. That so few problems existed overall (0.5%) reassures us as to the robustness of our process in general. Microarray data, by its sheer volume, presents interesting laboratory information management challenges. The data arrive at the investigators desk after a significant number of steps. We have found that the better you track production, quality control, and experimental steps, the better chance you have of uncovering the reasons behind discrepancies that may appear in the data. Often, statistical analyses look only at the data presented as final expression values or ratios, without taking into account relevant quality control data. In our effort to understand our errors and the source of large systematic discrepancies in our data we have found the MuFu algorithm a useful test of certain classes of process errors, and as a check for general problems with specific process steps or microtiter plates. We use MuFu to find, verify and fix problems that can be attributed to an error in plate processing. We also flag plates for which we can find no specific problem but we see yield inconsistent results. These may be plates that, at some point in the process, had a problem (cross contamination, a PCR problem) that was not detected while the process step was being carried out. The fact that we can test the data, using our standard quality control hybridizations, for these types of quality issues is reassuring and has helped us gain confidence in our data. Obviously, there are many other classes of error that creep into microarray data. Aside from the gross process errors that are amenable to detection, as we have described here, there is also a large class of more subtle systematic errors that can contribute to the overall systematic error on the expression ratio measurement. Isolating the source of these individual errors is sometimes quite difficult. Properties of the microarray feature such as spot size, shape and uniformity can contribute, but the majority of systematic errors are introduced at the time the experiment is performed. Slide post processing, RNA quality, labelling, hybridization and washing all lend the possibility for introducing systematic errors. Improvements in protocols and hybridization apparatus have helped reduce these errors and should continue to do so in the future. As these sources of error are identified and eliminated, expression microarrays will continue to provide progressively more sensitive measures of gene expression. Conclusions Process errors in any genome scale high throughput production regime can lead to subsequent errors in data analysis. We have shown the value of tracking multi-step high throughput operations by using this knowledge to detect and correct misidentified data on gene expression microarrays. We generalized our procedure using a simple heuristic, which found and fixed several problems with the proper assignment of gene identifier with physical microarray feature. We found thirteen print runs (9K arrays) that had four plates mistracked, six print runs with single 384 plate rotations, and one instance of a plate rotation at the 96 well plate level. One skipped plate was detected, as well as one plate printed twice. Out of the 265 million features checked in our database using MuFu, problems were detected and corrected on 1.3 million of them. That we were able to find and correct both previously known and unknown problems gives us confidence in the algorithm. That so few problems existed overall (0.5%) reassures us as to the robustness of our process in general. A list of corrected arrays can be found at . Methods We follow the simple heuristic outlined here. The flowchart for the program is shown in Figure 1 . In the figure we do not include additional loops needed to repartition the data in different ways for different scenarios. Data partitioning 1. Begin by partitioning the gene expression data on an array into subsets according to plate. Other partitions can also be made but the plate level partition is the most useful for our purposes. Let A ij be the intensity measurements of array A , subset i and gene index j . The measurements are usually of the channel (Cy3 or Cy5) used as a common reference. For example, if we partition the data by the 384-well plate each feature once occupied at some point in its process history, A is the array id, i is the plate id, and j is the well index between 1–384. Let A i be the vector ( A i 1 ,..., A i 384 ). In our vernacular this is the 384-well plate expression vector for plate i . Let B ij and B i be the similar definitions for array B . We also reverse the data vectors from array A and label it A i ' . In our notation, A i ' = ( A i 384 ,..., A i 1 ) is the expression vector for a plate rotated by 180°. 2. Generate a distance matrix { D mn , 1 ≤ m, n ≤ N } in which each element D mn = 1 - corr ( A m , B n ). For the 384-well plate example, A m is the 384-well expression vector for plate m on array A , B n is the 384-well expression vector for plate n on array B , and D mn is the distance between the two vectors in this 384 dimensional space. We also generate the corresponding rotated-distance matrix { D' mn , 1 ≤ m, n ≤ N } for the plate rotation case, in which each element D' mn = 1 - corr ( A' m , B n ). We tried several correlation functions including Euclidean and Pearson but found the Pearson to be best suited to this task. 3. Generate a rank matrix { R mn , 1 ≤ m, n ≤ N } by converting the distances to ranks such that the row m in the rank matrix is the order statistic of the row m of the distance matrix. We also generate the corresponding rotated-rank matrix { R' mn } for { D' mn }. Ideally, for the case where we are comparing identical subsets from two different arrays we expect the diagonal elements of the rank matrix to all be equal to one, which means that each subset of genes from the first array matches its corresponding subset in the second array the best. In general, due to the statistical variation in array data, the diagonal elements are not all equal to one, even if there are no misidentification errors. The examples show that this does not hinder us from making a clear distinction between real problems in the data and statistical fluctuations. Identification of rotated plates A plate rotation may have an affect on a single microarray batch if it occurs during array printing or may persist across many print batches if it happens during a 96-well (PCR) process step. In any case, the mismatch will persist across many array comparisons. To check for plate rotations in a print batch, we compare an array sample ( A1, A2, A3, A4 in the example) from the print batch in question to a control sample of arrays ( B, C, D, E in the example) selected from several distinct print batches. By comparing the rank R , and rotated-rank R' , matrix assignments for comparisons across array batch boundaries we can quickly flag possible rotated plates. A visualization of this test is shown in Table 3a . In the table we have flagged the top 5% of all ranks in the rotated column and the bottom 95% in the unrotated column. If the flags agree across all comparisons, we have a strong indication that a plate rotation has occurred. If we see a flag raised in this test, but we cannot attribute it to a plate rotation, this may indicate a different class of process error. In particular, if the flag is raised for all pairwise comparisons of the batch being tested (in this case batch A ) against all of the control batches (in this case batches B, C, D and E ) in the non-rotated case, we conclude that the corresponding flagged plate or partition from batch A may be misidentified. Note that in the limit that the partitions of the array are all uniform in expression ratios there is a 5% probability of a spurious flag. For this reason it is better to use high quality, highly variegated control arrays for such tests. Next, to better resolve ambiguous cases and to check our rank matrix determination we use the distance rather than rank matrix. If, for example, a plate x is to be considered a rotated plate, the following three criteria must be met. 1. D' xx < D xx . The rotated-distance must be less than the non-rotated distance. 2. D' xx is close to the mean of the distribution, { D mm , 1 ≤ m ≤ N }, and D xx is close to the mean of the distribution { D' mm , 1 ≤ m ≤ N }. 3. D' xx is an outlier of the distribution { D' mm , 1 ≤ m ≤ N }, and D xx is an outlier of the distribution { D mm , 1 ≤ m ≤ N }. The second and third criteria above are valid if the distributions of { D' mm , 1 ≤ m ≤ N } and { D mm , 1 ≤ m ≤ N } separate well. In Figure 5 we showed that our data support this model. Finding misidentified plates If a plate has consistently poor self-match rankings and a plate rotation has been excluded as a possible source of error, this next step tests for a plate misidentification. From the rank matrix we can identify the best alternative match. We test a suspect plate x for a match with plate y with the following criteria: 1. D xy < D xx . The mismatch distance is shorter than the match distance. 2. D xy is close to the mean of the distribution { D mm , 1 ≤ m ≤ N }, and D xx is close to the mean of the distribution { D mn , 1 ≤ m ≠ n ≤ N }. 3. D xx is an outlier of the distribution { D mm , 1 ≤ m ≤ N }, and D xy is an outlier of the distribution { D mn , 1 ≤ m ≠ n ≤ N }. Again, the second and third criteria above are valid if the distributions of { D mm , 1 ≤ m ≤ N } and { D mn , 1 ≤ m ≠ n ≤ N } are well resolved. Authors' contributions IT, MF developed the heuristic and carried out analysis and implementation. MS brought this problem to our attention and provided test data. MD provided advice and checked our work with an alternative technique. All authors read and approved the final manuscript. Table 1 Process steps The process fail points for cDNA based microarray production. Steps shown in italics are accessible to the testing methods outlined here. Earlier steps may require sequencing to test. The process ID is used to identify steps where the problems, if any, are found. Process ID Problem Type Note -999 Unidentified 0.0 Source.General persists across all arrays at clone, 96 level 0.1 Source.Contamination 0.2 Source.MisID 1.0 Prep.General persists across all arrays at clone, 96 level 1.1 Prep.Contamination 1.2 Prep.MisID 1.2.1 Prep.MisID.OrderError 1.2.2 Prep.MisID.Rot 1.3 Prep.Fail 2.0 PCR.General persists for 1 PCR round at 96 level 2.1 PCR.Contamination 2.2 PCR.MisID 2.3 PCR.Fail 2.4 PCR.Cleanup 3.0 Replate.General persists for 1 PCR round at 96 level 3.1 Replate.Contamination 3.2 Replate.MisID 3.2.1 Replate.MisID.OrderError 3.2.2 Replate.MisID.96Rot 3.2.3 Replate.MisID.384Rot 4.0 Print.General persists for 1 print run batch at 384 level 4.1 Print.Contamination 4.2 Print.MisID 4.2.1 Print.MisID.OrderError 4.2.2 Print.MisID.Rot 4.3 Print.Fail dried out plate, too concentrated, too weak etc. 5.0 Scan.General anything after array production	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521069.xml
520820	An integrated 4249 marker FISH/RH map of the canine genome	Background The 156 breeds of dog recognized by the American Kennel Club offer a unique opportunity to map genes important in genetic variation. Each breed features a defining constellation of morphological and behavioral traits, often generated by deliberate crossing of closely related individuals, leading to a high rate of genetic disease in many breeds. Understanding the genetic basis of both phenotypic variation and disease susceptibility in the dog provides new ways in which to dissect the genetics of human health and biology. Results To facilitate both genetic mapping and cloning efforts, we have constructed an integrated canine genome map that is both dense and accurate. The resulting resource encompasses 4249 markers, and was constructed using the RHDF5000-2 whole genome radiation hybrid panel. The radiation hybrid (RH) map features a density of one marker every 900 Kb and contains 1760 bacterial artificial chromosome clones (BACs) localized to 1423 unique positions, 851 of which have also been mapped by fluorescence in situ hybridization (FISH). The two data sets show excellent concordance. Excluding the Y chromosome, the map features an RH/FISH mapped BAC every 3.5 Mb and an RH mapped BAC-end, on average, every 2 Mb. For 2233 markers, the orthologous human genes have been established, allowing the identification of 79 conserved segments (CS) between the dog and human genomes, dramatically extending the length of most previously described CS. Conclusions These results provide a necessary resource for the canine genome mapping community to undertake positional cloning experiments and provide new insights into the comparative canine-human genome maps.	Background Three major advances in the development of resources for mapping canine disease genes have been: 1) the development of a radiation hybrid (RH) map composed of large numbers of microsatellite markers and genes that link the canine and human genomes [ 1 ], 2) the development of canine specific whole chromosome paints that have allowed preliminary assignment of conserved segments between human and dog [ 3 - 5 ]; and 3) the publication of a 1.5x genome sequence of the dog [ 2 ]. The most recently published RH map of the dog comprises 3270 markers including 1596 microsatellite-based markers, 900 canine-specific cloned gene sequences and expressed sequence tags (ESTs), and an initial set of 668 canine-specific BAC-ends [ 1 ]. The map was constructed using the RHDF5000-2 whole genome radiation hybrid panel [ 6 ] and features markers mapped to 3009 unique positions, defining an average inter-marker distance of one megabase (Mb). The map also defines a minimal screening set of 325 highly informative well-spaced markers, to be used in the initiation of genome-wide scans [ 1 ]. A well-defined synteny between the dog and human genomes was established as a function of this work [ 1 ] and from extensive reciprocal chromosome painting studies [ 3 , 5 , 7 ]. The above mapping efforts are complemented by the recent release of a 1.5x sequence of a Standard Poodle genome [ 2 ]. The sequence includes 6.2 million sequence reads that span approximately 78% of the genome. More than 650 million base pairs (>25%) align uniquely to the human genome, and the resulting alignment includes fragments of putative orthologs for 18,473 of 24,567 annotated human genes. The current alignment supports most of the proposed comparative segments, but suggests that the final comparative dog-human map will be composed of at least 160 comparative blocks [ 2 ]. Using the above resources, canine researchers have undertaken genome-wide screens for linkage to a variety of disease loci as well as morphological traits defining differences between breeds. Many of these studies have met with success. Disease loci have been genetically mapped or otherwise localized in the dog for several disorders including: vision-associated disease such as progressive rod cone degeneration, early retinal degeneration, cone degeneration, and collie eye anomaly [ 8 - 11 ], kidney cancer [ 12 , 13 ], narcolepsy [ 14 ], rheumatoid arthritis [ 15 ], Severe Combined Immunodeficiency (SCID) [ 16 ], hip dysplasia [ 17 ], cystinuria [ 18 ], bleeding disorders [ 19 , 20 ], ceroid lipofuscinosis [ 21 ], and copper toxicosis [ 22 , 23 ]. In addition, quantitative trait loci (QTL) have been identified for principal components defining skeletal variation and hip dysplasia [ 17 , 24 ]. In general, investigators have been able to use the existing resources to localize traits of interest to an interval of 10–20 Mb, but in most cases the causative gene remains to be identified. To facilitate positional cloning efforts in the canine community, we have localized a large set of randomly selected canine-specific BACs onto the canine 5000-rad RH map, a subset of which we have also localized cytogenetically. The resulting map of 4249 markers includes 1760 mapped BAC-ends, 851 of which have been FISH mapped, to generate a dense, accurate and highly integrated map of the canine genome. Results General RH map characteristics This novel RH map of the dog genome contains 4249 markers of three different types: 900 genes, 1589 microsatellites and 1760 BACs. The map was generated by genotyping 1092 new BAC-end markers on the RHDF5000-2 panel, and recomputing the new vectors with those from the previous map [ 1 ] using MULTIMAP [ 25 ] and TSP/CONCORDE [ 26 ]. Table 1 summarizes the key features of the new map. The typing of 4249 markers resulted in an RH map containing 4106 markers that were eventually grouped and assigned to each of the canine chromosomes, leaving only 143 unlinked markers. Human orthologs were identified for 2233 mapped markers. The resolution limit of the RHDF5000 panel has been determined to be 4 cR5000 (600 Kb) [ 6 ]. Thus, markers falling within any 4 cR5000 or 600 Kb stretch cannot be ordered relative to one another with high confidence and, consequently, are reported as co-localized on the map. Table 1 Key features of the integrated canine RH/FISH map Number of Markers Unique BACs Unique BACs BACs Chromosome Size (Mb) (1) RH Mapped RH Positions (2) RH Mapped BAC RH Positions FISH Mapped RH/FISHed in Common Human Coordinates (3) Human/ Dog CS (4) CFA01 137 220 169 104 83 60 48 110 4 CFA02 99 140 99 55 43 37 27 78 4 CFA03 105 155 123 74 60 46 35 71 3 CFA04 100 147 106 60 51 37 29 80 3 CFA05 99 158 102 76 53 40 28 74 4 CFA06 87 122 69 53 35 29 19 69 3 CFA07 94 173 107 78 56 44 37 97 2 CFA08 86 135 98 64 50 30 25 77 1 CFA09 77 124 93 39 33 31 26 78 2 CFA10 80 121 91 53 41 28 22 68 3 CFA11 86 134 116 56 49 38 36 78 2 CFA12 85 181 127 71 55 32 29 94 1 CFA13 75 93 65 38 32 21 18 53 2 CFA14 72 114 85 54 42 31 28 62 2 CFA15 75 113 94 42 38 26 22 63 5 CFA16 73 93 79 42 36 23 23 44 3 CFA17 80 130 96 53 41 20 18 69 2 CFA18 66 118 99 50 45 30 27 71 2 CFA19 66 96 66 40 31 18 18 49 2 CFA20 66 139 95 52 41 20 16 74 2 CFA21 61 103 86 34 31 26 22 60 1 CFA22 61 113 76 52 33 17 15 70 1 CFA23 61 79 67 40 36 23 21 45 1 CFA24 73 84 66 41 33 18 16 42 1 CFA25 60 94 75 41 36 27 22 56 4 CFA26 48 85 66 39 33 19 13 44 3 CFA27 57 94 76 29 27 19 16 50 1 CFA28 55 84 70 39 35 26 24 47 1 CFA29 51 81 71 33 31 20 18 42 1 CFA30 47 63 50 26 21 18 16 38 1 CFA31 50 53 47 25 23 16 13 25 2 CFA32 51 56 46 33 28 14 15 27 1 CFA33 41 63 52 30 28 20 20 40 1 CFA34 50 65 49 23 13 15 15 47 2 CFA35 38 49 35 24 16 14 12 30 1 CFA36 41 61 52 26 26 13 13 40 1 CFA37 40 65 45 27 21 20 17 37 1 CFA38 38 32 28 16 14 9 9 16 1 CFAX 139 66 44 26 21 23 22 28 1 CFAY 27 10 10 2 2 2 1 1 1 unlinked 143 TOTAL 2797 4249 3090 1760 1423 1000 851 2233 79 Average Distance (Mb) Between Markers -- 0.66 0.91 1.59 1.97 2.80 3.48 1.25 Legend: Map statistics. (1) Chromosome size in Mb based on bivariate flow cytometry measurements (2) Unique RH positions with one or more markers (3) Human coordinates identified through BLAST analysis (see Methods) (4) Conserved segment identified by RH mapping with 2 or more loci Analysis of the entire dataset of 4249 RH vectors by MULTIMAP [ 25 ] at a Lod score of 8.0, and for a subset of regions a Lod of 9.0, resulted in 60 individual linkage groups that could be assigned to the 38 canine autosomes and two sex chromosomes. For chromosomes that were covered by multiple linkage groups, the data were merged and ordered as described in the Methods. FISH data combined with significant Lod scores between selected markers led to the unambiguous assignment and orientation of specific linkage groups to chromosomes and, ultimately, complete coverage of each chromosome. The resulting 4106 linked markers defined 3090 unique positions regularly spaced across all autosomes. The largest chromosome, canine chromosome 1 (CFA 1), contains 220 markers, while CFA 38, the smallest chromosome at 38 Mb, has only 32 markers (Figure 1 ). Excluding the sex chromosomes, the least dense chromosome, CFA 38, has a marker located every 1.2 Mb, while the most dense, CFA 12 and CFA 20, have markers positioned at an average of every 0.47 Mb. The increase in density is due solely to the addition of 1092 new BAC-ends, bringing the total to 1760 from the 668 reported in the previous version of the map. The mean interval size is now 1.59 Mb. The chromosome with the most BACs mapped to it is CFA 1, featuring 104 localized to 83 unique positions (Figure 1 ). See all chromosomes in the Supporting Online Material (SOM) at . Again, the chromosome with the fewest is CFA 38, with 16 BACs localized to 14 unique positions. This resource provides a dense scaffold on which to build regional physical maps and search for new genes. Of particular interest to disease gene mappers will be the 26 BACs that are RH mapped to the X chromosome, 13 of which are also localized by FISH. Only two BACs were localized to the Y chromosome, one of which is also FISH mapped. See Table 1 for details. Figure 1 Integrated FISH/RH map and dog/human comparative data for CFA1. BAC-ends also localized by FISH mapping are reported in the left panel, symbolized by a vertical bar along with BAC addresses. Asterisks () represent BAC clones that did not have a unique cytogenetic location (multiples hits are listed on SOM). RH mapped markers and their cumulative positions in centiRay 5000 units are reported to the right of the FISH map. Connecting lines between FISH and RH maps indicate integration points between maps. The right panel shows the human evolutionarily conserved segments, represented by colored boxes as determined by RH data. Human coordinates (in Mb) identified from dog/human sequence alignments (see Methods) are reported on the right most vertical bar. RH markers and their corresponding human sites are connected by lines and illustrate the rearrangement within synteny blocks. FISH mapped BACs A total of 1,000 BACs were assigned to a chromosome band, and then ordered along the length of the corresponding chromosome using a combination of metaphase and interphase multi-color FISH. CFA 1 is shown, for example, in Figure 2 . The panel of FISH mapped clones included clones representing both ends of each chromosome, with the exception of the centromeric end of CFA 6 (6q11-q12) and the centromeric 20% of CFA 9 (9q11-q12). Nine hundred and eighty-one out of 1,000 BACs had a unique cytogenetic location, and 851 were also ordered on the RH map. This yielded an average of one BAC-end that was both FISH and RH mapped every 3.48 Mb. See SOM material for data on all chromosomes and . Figure 2 Assignment of 60 canine BAC clones to CFA 1. BAC addresses alongside the ideogram of CFA 1 refer to clones from the RPCI-81 canine BAC library. A pair of colored spots spanning an interval of approximately 4 Mb represents the cytogenetic assignment of each clone. The color of the spots identifies the fluorochrome used to label the clone as follows: red = Spectrum Red; orange = Spectrum Orange; green = Spectrum Green; blue = DEAC; purple = Cy5. Multi-color FISH of neighboring clones, using both metaphase and interphase analyses, was used to establish the precise order of the clones along the length of the chromosome. Clones whose assignment is represented by a circular rainbow have been tentatively placed, but not yet co-localized with neighboring clones to establish one equivocal, linear order. BAC addresses followed by identify those clones that resulted in fluorescent signal at more than one location. Human orthologous regions (HSA) are reported on the left of the figure by vertical bars. Integration of FISH and RH data Analysis of co-linearity between the two maps was performed by connecting the BAC markers in common between the RH map and FISH data. This approach allowed the identification of markers that serve as anchors to each chromosome for both RH and FISH maps, thus validating their localization and marker order. Early in the assembly process, a small number of discrepancies between the two maps were easily identified through graphical drawings of all chromosomes, leading to a systematic re-examination of experimental data and, in most cases, consensus resolution. However, 19 BACs did not have one unique cytogenetic location. These are indicated with a double asterisk in Figure 1 and 2 , as well as the SOM and figures provided on all web sites. For each of these 19, one of their FISH localizations was in agreement with the corresponding RH map position. In addition, seven (7/19) had an identified human ortholog that was in agreement with the RH map localization. For an additional set of 24 BACs, conflict remained after rechecking both the FISH and RH data. These are indicated by "#" in all Figures. Interestingly, the DNA used to obtain FISH data for these clones PCR amplified successfully with the same primers used to obtain RH data, indicating that the DNA samples were the same and had not undergone sample mix-up. Finally, a human ortholog sequence was identified that was in agreement with the RH data for 15 (15/24) of these clones. Although we have no clear explanation for these discrepancies, it is of note that the two mapping methods used different DNA sequences to define the same marker. In the RH method, a pair of 25-mer oligonucleotides defining only one end of the dog insert DNA is used. By comparison, in the FISH analysis, the whole BAC encompassing approximately 150 Kb is hybridized. Since the DNA for the FISH localization was prepared from cultures initiated from single bacterial colonies, these anomalies suggest that at least a subset of the clones may be chimeric. An additional explanation is that the clones are hybridizing to genomic regions containing one member of a multi-gene family. Whatever the explanation for these discrepancies, it is noteworthy that the total number of discrepancies between the FISH and RH data represent 3% of the total data. This is in agreement with reports that the frequency of chimeric clones in this BAC library is very low [ 27 ]. Thus, in most cases, the co-linearity of the FISH and RH maps is perfect, i.e. the order of the BACs determined by multi-color FISH is identical to that derived from analysis of genotyping data using primers defining single BAC-ends. In addition to the conflicting results described above, in some cases, we note that the order for two closely positioned markers is inverted between the FISH and the RH data. However, for all of these cases the distance between the two markers is estimated to be less than 50–100 Kb, as these clones overlap in interphase nuclei. Thus, they are well within the 600 Kb resolution limit of the RH5000 panel. Such minor inaccuracies in marker order will not be detrimental to gene mapping, as the integrated map has fixed positions every 3.42 Mb with support from FISH data. The example of CFA 1 is provided (Figure 1 and 2 ). Please refer to SOM and websites for all other chromosomes. Synteny and conserved segments Two criteria were used to assign dog BAC clones to orthologous locations of the human genome. First, when compared to the complete human genome, each component of the paired end-sequences had greatest similarity to locations that are separated by 50–500 Kb. Although most BAC clones have inserts of 100–200 Kb, minor differences between the two genomes such as local duplication or loss of specific genes should be accommodated by the 50–500 Kb range. If one considers a single aligned end-sequence, the probability of the paired end-sequence aligning spuriously within 500 Kb in the human genome is approximately 0.03%. The second criterion was that each pair of end-sequences should align in a head-to-head orientation. With this additional condition, the probability of scoring pairs that contain spurious alignments falls to less than 0.01%. In the most recent version of the map [ 1 ], 75 conserved segments (CS) were detected within the 38 canine autosomes. In addition, two CS corresponding to the sex chromosomes and nine singletons were detected. These 77 CS were identified through the analysis of 820 canine markers having an unambiguous ortholog in the human sequence. In the present map, that number is more than doubled with 2233 markers having an ortholog in the human sequence. These newer data confirm all previously described CS, and incorporate two singletons into two novel CS for a new total of 79 CS. For the sake of clarity, the remaining singletons were not reported in this new version of the map. Interestingly, while the number of CS did not change significantly compared to the previous iteration of the RH map, their nature and composition were altered, as shown for CFA 1 (Figure 1 ). Four CS were identified for CFA1 that correspond, in order from centromere to telomere, to human chromosomes (HSA) 18, 6, 9 and 19. From the density of markers for which an ortholog can be identified, it is clear that the order of markers is consistent between CFA 1 and HSA 18. Thus, this CS would be considered a "conserved and ordered segment" (CSO). For the CS corresponding to HSA 6, however, two sub-segments are noted which are caused by an inversion. In each of these two sub-segments, the order of markers is comparable to the syntenic portion of HSA 6, and each canine segment would independently be considered a CSO. In addition to identification of some previously unknown CSO, the increased number of markers for which a human ortholog could be identified unambiguously allowed for a substantial increase in the size of many previously identified CS. At present, as shown in Figure 1 and SOM, CS in the dog are highly contiguous with the human genome. Indeed, only a few markers identified as human orthologs fall outside of a CS. This significantly refines our knowledge of evolutionary breakpoints between the dog and human genomes. Discussion The 4106 markers that constitute this new map occupy 3090 distinct positions, with an average of 1.3 markers per position. This reflects the fact that with 4106 genotyped markers, we are approaching the saturation level of 4500 markers predicted on the basis of size of the canine genome, and resolving power of the 5000 rad panel [ 6 ]. A detailed analysis of the co-localized markers indicates that in the majority of cases markers are co-positioned with other markers of different types, i.e. BACs and microsatellites, or microsatellite and genes, etc. Thus, all of the data will likely prove useful for mapping and cloning genes of interest, as the information provided at each location is non redundant. In addition, even when markers of the same type are co-localized, they are likely to be of value; any given microsatellite is not informative in every pedigree, and closely localized BACs may represent the beginning of an overlapping contig. A direct consequence of markers mapping to the same positions is that the mean distance between two adjacent positions is now 0.9 Mb, which is very similar to the 1 Mb calculated for the most recent published version of the canine map [ 1 ]. Again, this reflects the resolving power of the 5000 rad panel [ 6 ], as well as the method of map computation. TSP/CONCORDE considers markers co-localized if they map to a region of 0.9 Mb or less, regardless of RH panel mapping power [ 26 ]. The number and identity of conserved segments (CS) between the human and the dog genomes is 79. CS were identified by both FISH localization and RH mapping with four chromosomes showing four human/dog CS, and the remainder showing one to three. The number of markers and the resolving power of the 5000 rad panel allowed us to detect several instances in which CS appear to be split in sub-segments. In several such cases, the order of markers in the two adjacent sub-fragments is conserved, although the two sub-fragments are oriented in opposing directions. Following the previously agreed definitions, these sub-fragments are termed CSO for "conserved segment order" [ 28 ]. Such a situation is observed in CFA 1 (Figure 1 ). In other instances, one of the two adjacent sub-fragments appears to correspond to a CSO, while the other corresponds to a CS. Obviously a better distinction between CS and CSO segments would require genotyping of additional markers, ideally on a panel constructed with a higher dose of radiation, and with concomitantly higher resolution. Such experiments are currently underway using 10,000 genes derived from the available dog genome sequence [ 2 ] and a newly constructed 9000 rad panel (data not shown). As shown in Table 1 , and as indicated in the Results, 851 BACs were localized both by FISH and RH mapping, with a high level of concordance between the two methods. Thus, the map presented here provides a verified anchor point every 3.48 Mb. Unlike a FISH map that is not altered or modified by the addition of a new dataset, an RH map does change after adding information because different algorithms provide different solutions for local marker order. Each is statistically valid when ordering the same set of markers. However, discrepancies may arise as more markers are added to the map. Moreover, even one algorithm can generate maps with slightly different marker orders within localized regions, even when only minor adjustments are made to the dataset like removing seemingly redundant markers [ 29 ]. In the present study, the high degree of anchorage by a set of markers localized in different laboratories with differing technologies yields a high level of confidence for the integrated FISH/RH map. Of interest to many researchers will be the multiple applications that a cytogenetically defined and RH verified set of canine BAC clones can provide. For example, the ordered set of 981 FISH mapped canine BAC clones with a unique cytogenetic location will be welcomed by cancer geneticists as a means to characterize chromosome aberrations in canine tumors. The dense cytogenetic coverage available for most chromosomes allows the use of serial differential labeling of some or all clones in chromosome-specific panels designed to 'tile' structurally aberrant chromosomes. This approach will allow a much more accurate assessment of chromosome rearrangements than is possible with single color whole chromosome paint probes. In addition, the ordered clones will allow expansion of the current canine BAC microarrays, which in turn may be used for a variety of applications including array-based Comparative Genome Hybridization (CGH) analysis of canine tumors [ 30 ], array-painting of aberrant canine chromosomes, and investigation of cytogenetically cryptic copy number changes and gene dosage alterations in congenital abnormality syndromes. In addition, ordered arrays of BAC clones will be an important resource for refining the comparative cytogenetic data within the diverse karyotypes that comprise the Canidae . Finally, with the imminent release of the 7x canine genome assembly, the resources described in this paper will provide a long-awaited means to translate canine cytogenetics data into canine DNA sequence data, thus advancing our knowledge of canine and comparative genomics. Conclusions The presentation of a map containing 900 canine specific genes, 1589 microsatellites and 1760 BACs provides the canine genetics community with nearly all the resources it needs to undertake experiments aimed at both mapping and cloning traits of interest. A dense set of microsatellite markers (MSS-2) for undertaking genome wide scans was provided in the previous iteration of the map [ 1 ]. These same markers are integrated into the current map. Thus, within 0.7/0.8 Mb of any linked marker on the current map is now an array of anchored BACs for contig building, comparative mapping, and searching for new genes and splice variants. The current map defines some 79 conserved segments between human and dog. While we expect this number to approximately double when the 7x canine genome sequence is completed, we speculate that given the density of markers mapped and the distance between adjacent CS, new CS defined by the 7x sequencing effort will be short in size and harbor a limited number of genes. What remains in canine genome map building is the development of a very high resolution gene map that can assist in the assembly of the 7x canine genome sequence now underway, and provide a mechanism for moving easily between canine and human comparative segments. This is currently being undertaken. In the meantime, canine researchers can move forward with the continued development of the dog model for mapping and cloning genes of interest to both human and companion animal health. Methods Markers and primer selection Canine BACs were randomly selected from the RPCI-81 canine BAC library [ 27 ] and were end-sequenced as described previously [ 1 ]. DNA was prepared using standard automated approaches [ 31 ] and sequenced either at The Institute for Genomic Research (TIGR) or the University of Washington High Throughput Sequencing Center. Average read lengths were in excess of 700 bp. Trace files representing BAC-end sequences were imported from ABI sequencers and examined for sequence homology to cloning vectors, Escherichia coli ( E. coli ) and repetitive DNA sequences. BAC sequences were also compared using Cross_Match with the complete E. coli genome to remove contaminating sequences of bacterial host origin. Finally, sequences were examined for interspersed repeats and regions of low sequence complexity using RepeatMasker . Primers defining each BAC-end were designed to regions of high quality sequence data using Primer3 software . Primers of 25 bp lengths were preferentially designed in order to minimize problems associated with non-specific amplification, to generate amplicons of 200 to 500 bp, and to work under a single optimal set of PCR conditions. Genotyping Genotyping was performed using the RHDF5000-2 panel, which is comprised of 118 cell lines. The panel was constructed by fusing dog fibroblasts irradiated at 5000 rads with TK-HTK3 hamster cells and has an experimental retention frequency of 22% with a theoretical resolution limit of 600 Kb [ 32 ]. PCR reactions were carried out at the University of Rennes and the Fred Hutchinson Cancer Research Center (FHCRC) as described previously in 15 μl volumes [ 1 , 33 - 35 ] using the following touchdown program: 8 min 95°C, followed by 20 cycles of 30 sec 94°C, 30 sec 63°C decreasing by 0.5°C per cycle, 1 min 72°C and 15 cycles of 30 sec 94°C, 30 sec 53°C, 1 min 72°C and a final extension of 2 min 72°C. Primer pairs yielding either faint or spurious bands were rejected. PCR products were resolved and recorded as described [ 33 , 35 ] through a semi-automated data acquisition software. Quality control Duplicate genotypes were obtained for the 1092 new BACs added to the map. Data were considered consistent when the number of discrepancies between data sets was ≤ 16%, a value determined to correspond to a distance lower than the resolution limit of the RHDF5000-2 panel. In the rare cases where two independent typings yielded >16% discrepancies, a third typing was done and the resulting vector was either integrated into the map construction, or the marker was discarded if no agreement was observed between two of three genotypes. RH map construction RH vectors corresponding to the BAC-end markers and marker vectors of the previous map [ 1 ] were computed as a single data set using the MultiMap and TSP/Concorde algorithms [ 1 , 25 , 26 , 29 ]. Linkage groups were generated initially at a Lod score of 8.0, and where needed, at 9.0 to ensure strong statistical support. Individual linkage groups were analyzed using the multipoint approach of the rh_tsp_map version 2.0 of TSP/CONCORDE, as described previously [ 1 , 29 ]. Inter-marker distances are expressed in cR5000 units. Framework maps of well-spaced markers, supported by high quality data between adjacent markers, were initially generated for all chromosomes. Distance criteria were set to 4cR, corresponding to the resolution capacity of the RH panel used, and quality criteria were fixed to 4, corresponding to a maximum allowable number of ambiguous data within RH vectors, as determined from previous analyses [ 1 , 35 ]. Distance and ordering of markers within each group was then determined by the TSP/Concorde and rh_tsp_map-2 algorithms [ 26 , 29 ]. Markers that could not be ordered with a high confidence level were submitted for re-analysis by stepwise increases in the Lod score to > 9.0, forcing the linkage group to split into two or more groups, until satisfactory order with a high statistical confidence level was achieved for each resulting group. Groups were then merged and oriented into a unique dataset. The merging step utilized the cytogenetic data obtained as part of this study, as well as the 2-point Lod scores between the markers at the extremes of each linkage group. Lod scores were generated using the pairlod_dist software from the rh_tsp_map package [ 26 ]. FISH mapping A total of 851 BAC clones from the RH map were also localized by multi-color FISH analysis. DNA from each clone was prepared from 2.5 ml cultures using a BAC RealPrep (Qiagen, Valencia, CA) protocol. Two hundred nanograms from each sample were labeled using nick translation to incorporate one of five fluorochromes, Spectrum Red/Orange/Green dUTP (Vysis, Downers Grove, IL), diethylaminomethylcoumarin (DEAC)-5-dUTP (NEN/Perkin Elmer Life Sciences, Boston, MA), or Cy5-dUTP (Amersham Biosciences, Piscataway, NJ). Typically, 25 ng of each of five differentially labeled probes were pooled and precipitated in the presence of 15 μg of sonicated genomic dog DNA as competitor. Chromosome preparation, probe hybridization and post hybridization washes were performed as described previously [ 35 , 36 ]. Chromosomes were counterstained in 80 ng/ml 4', 6-diamidino-2-phenylindole (DAPI) and mounted in anti-fade solution (Vectashield, Vector Laboratories, Burlingame, CA). Images were acquired and processed using a multi-color FISH workstation comprising a fluorescence microscope (Axioplan 2ie, Zeiss) equipped with narrow pass filter sets and a cooled CCD camera (CoolSnapHQ, Photometrics, Tuscon, AZ) both driven by dedicated software (SmartCapture 2.3.1 Digital Scientific, Cambridge, U.K.). The digital image of each DAPI stained metaphase spread was processed using a high-pass spatial filter to reveal enhanced DAPI bands. Clones were assigned to a chromosome region according to the DAPI banded nomenclature of Breen et al. [ 35 , 36 ]. Refinement of probe order along the length of each chromosome was made by subsequent rehybridization to elongated canine chromosome preparations and/or by reference to interphase FISH analysis. Additional information may be found at . Alignment of dog BAC clones to orthologous regions of the human genome Nineteen 384-well plates of BAC clones from the RPCI-81 library [ 27 ] were selected at random, and end-sequence data were obtained from each clone using previously described methods [ 1 ]. Paired end-sequences for 1910 clones were masked for repetitive elements and searched against the human genome (NCBI build 31, November 2002, ). For 648 of the BACs (34%), each of the paired end-sequences gave a best hit to human genomic locations that are separated by 50–500 Kb, and aligned head-to-head (mean span of human genomic DNA = 191 Kb). The remaining 1262 BAC-end sequences were searched against scaffolds of the 1.5x assembly [ 2 ] using wu-blastn (matrix = identity, W = 40) to identify scaffold sequences that contained at least short overlaps (40 bases) of identical sequence. For 954 of the 1262, hits were detected for both of the paired end-sequences. The homologous scaffold sequences were trimmed to remove any sequence that extended beyond 5 Kb from the region of alignment. They were then masked for repetitive elements, and searched against the human genome using wu-blastn (E<0.1). Again, only the best hit was considered. For 604 of the BACs (32% of the original sample), the paired scaffolds gave the best hits to genomic locations that are separated by 50–500 Kb (mean = 202 Kb), and aligned with their component BAC-end sequences in a head-to-head orientation. Altogether a total of 1252 (648+ 604) canine paired BAC-end sequences demonstrated significant hits with the human sequence. Accession numbers, PCR conditions, primers and BAC-end sequences are available for all markers at: and . Abbreviations BAC-Bacterial Artificial Chromosome FISH-Fluorescence in situ Hybridization PCR-Polymerase Chain Reaction RH-Radiation Hybrid CGH-Comparative Genomic Hybridization cR-centiRays Mb-megabases Kb-kilobases Bp-base pair Min.-minutes E. Coli-Escherichia coli SOM-Supporting Online Material DAPI-Diamidino-2-phenylindole CS-Conserved Segments CSO-Conserved and Ordered Segments TIGR-The Institute for Genomics Research FHCRC-Fred Hutchinson Cancer Research Center Authors' contributions MB, RT, AS, and RH all contributed to the FISH mapping aspects of the manuscript, including isolation of BAC DNA samples, generation of canine chromosome preparations, probe labeling and purification, fluorescence in situ hybridization, and microscopy. MB performed all the cytogenetic analysis to generate detailed probe ordering. MB and RT completed the merging of RH and FISH data, and prepared the drawing of Figure 2 for both the manuscript and web sites. MB also wrote and edited relevant sections of the manuscript. CH did the statistical analysis for the localization of the 4249 markers on the radiation hybrid map and the integration of the FISH and RH maps. In addition, CH constructed Figure 1 (paper and web sites) and Table 1 and implemented the accompanying web site. GGM and EFK organized and oversaw the BAC-end sequencing, including data production and analysis, with CMF overseeing the ultimate effort at TIGR. Markers were RH-mapped in Rennes, France by EC and GE, and in Seattle, Washington, USA by HGP, GB, LS, and TDL. The Seattle data were checked, duplicated, data-entered and generally overseen by TDL. TDL also drafted portions of the methods section of the manuscript. RG computed the marker sequences using BLAST software against the human sequence to identify orthologoussequences and synteny conservation. RG also drew some of the chromosomes for Figures 1 (web sites). Efforts in Rennes, France were overseen by CA and FG, including analysis, quality control, supervision, and portions of manuscript production. Efforts in Seattle, Washington, USA were overseen by EAO including experimental design, quality control, supervision, and portions of manuscript production. Individual and joint grants funding this work were written by and awarded to EAO, MB, and FG. All authors read and approved the final manuscript. Links FISH Information : RH Map Information : and .	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC520820.xml
521733	Why Are So Many Bird Flowers Red?	Are bird-pollinated flowers red because bees - which might rob the flower of its nectar - cannot easily detect them, or might it be because of more subtle evolutionary trade-offs?	Most bird-pollinated flowers are both red and rich in nectar. The traditional explanation for this association is that, since red is inconspicuous to bees, it evolved to prevent bees from depleting the nectar of bird-pollinated flowers without effecting pollination. But bees can see, and they actually visit red flowers. So why are most bird-pollinated flowers red? To help answer this question, we need to consider how the outcomes of foraging decisions are affected by the community in which individuals live, and by the foraging options of other individuals. The Mystery Plants face a trade-off between attracting pollinators and remaining hidden from flower parasites (such as nectar robbers and seed predators). Consequently, there is often strong selection pressure for highly specific communication channels that can advertise the presence of their flowers to effective pollinators but not to other individuals. Many aspects of pollinator syndromes are best understood in these terms ( Proctor et al. 1996 ). For example, flowers that are pollinated by birds—bird flowers—produce nectar at much higher rates than those pollinated by bees ( Stiles 1981 ). If a bee is attracted to such a flower, it might sometimes remove nectar and pollen without providing an outcrossing service (i.e., bringing pollen from a different plant of the same species) to the flower. Therefore, bird-pollinated flowers should advertise their presence to birds, but not to bees. Following this line of reasoning, Peter Raven (1972) suggested more than thirty years ago that bird-pollinated flowers were predominantly red because ‘red is the only color of the spectrum that is at once inconspicuous to most insects and also an excellent “signal” of a high caloric reward for birds’. Raven's interpretation of inconspicuousness was soon transformed into invisibility; it was assumed that bees did not visit red flowers because they couldn't detect them ( Proctor et al. 1996 ; Vogel 1996). However, this interpretation no longer holds. Chittka and Waser (1997) have shown that red flowers are not actually invisible to bees. Indeed, typical bird flowers with no UV reflectance, such as the scarlet gilia ( Ipomopsis aggregata ) and the scarlet monkeyflower ( Mimulus cardinalis ) ( Figure 1 ), are routinely visited and exploited by different bee species (reviewed by Chittka and Waser 1997 ). Moreover, when bees are extremely abundant, they can drive birds away from red flowers. Echium wildpretii , an endemic of the Canary Islands, presents an entomophylous (‘insect-loving’) and an ornithophyllous (‘bird-loving’) subspecies ( Figure 2 ) that differ in flower colour: E. wildpretii trichosiphon , endemic to La Palma Island, has entomophylous, pink flowers, whereas E. wildpretii wildpretii , endemic to Tenerife Island, has ornithophyllous, red flowers, pollinated by generalist native birds and insects. E. wildpretii wildpretii is pollinated predominantly by birds early in the season until introduced honeybees ( Apis mellifera , which have increased enormously in number because of apiculture) deplete the nectar and displace nectar-feeding birds ( Valido et al. 2002 ). Figure 1 Typical Bird-Pollinated Flowers (A) Scarlet gilia Ipomopsis aggregata . Image courtesy of Clarence A. Rechenthin at US Department of Agriculture–National Resources Conservation Service PLANTS Database. (B) Scarlet monkeyflower, Mimulus cardinalis . Image by William & Wilma Follette at US Department of Agriculture–National Resources Conservation Service PLANTS Database ( USDA NRCS 1992 ). Figure 2 The Two Subspecies of Echium wildpretii (A) The pink flowers of E. wildpretii trichosiphon are pollinated by insects. (B) The red flowers of E. wildpretii wildpretii are pollinated by generalist native birds, unless birds are driven away by large densities of bees. Photos courtesy of Alfredo Valido. So if red flowers are not invisible to bees, why are most bird-pollinated flowers red? Perhaps birds are particularly apt at detecting red objects ( Chittka and Waser 1997 )? Again, this is not strictly true. Although all birds detect red objects and some birds do have their greatest spectral sensitivity and finest hue discrimination towards the long-wavelength (red) end of the spectrum ( Stiles 1981 ), they can also respond to ultraviolet light, and there is no evidence that, for example, hummingbirds have greater spectral sensitivity or greater spectral discrimination ability in the red part of the spectrum ( Goldsmith and Goldsmith 1979 ). Feeding experiments, where hummingbirds are given nectar in artificial flowers of different colours, show no inherited colour preferences; hummingbirds have temporary preferences that can be modified by conditioning ( Proctor et al. 1996 ). So are there other clues as to how this mystery might be solved? The Visual System of Bees One clue might come from the visual system of bees. Humans perceive light with a wavelength above approximately 600 nm as red ( Buser and Imbert 1968 ). Most bees have three types of colour receptors, with sensitivity peaks at 340, 430, and 540 nm ( Chittka 1996 ), although a very few bee species have sensitivity peaks at substantially longer wavelengths. For the majority, however, provided that the light source is sufficiently intense, red light (up to 650 nm) will stimulate the 540 nm receptor of bees ( Chittka and Waser 1997 ). Bees will therefore perceive red objects. To discriminate red flowers from their green background, bees must rely essentially on the difference between the intensity of the signal that flower and foliage generate on the bees' ‘green’ (540 nm) receptor ( Giurfa et al. 1996 ). Therefore, depending on the relative intensity of the green and red sources, bees may or may not be able to discriminate between red flowers and green foliage ( Chittka and Waser 1997 ). Because of the structure of their visual system, bees trained to feed at artificial red flowers take longer to find their goals than bees trained to feed at other-coloured flowers ( Spaethe et al. 2001 ). In a real environment, where red flowers would be more camouflaged against the different shades and intensities of the green foliage, the ability of bees to discriminate red flowers should be further reduced. Colour Vision and Niche Partition The fact that bees require more time to find red flowers than other-coloured flowers, together with some results from optimal-foraging theory, outlined here, could unlock the mystery and explain the association between red coloration and bird pollination in flowers. When different animals, either from the same or different species, are forced to share some resources, any degree of specialization tends to result in habitat selection ( Rosenzweig 1981 ). In 1992, Possingham, developed a ‘habitat selection’ model that showed how two nectar-feeding pollinator species, which differed in their foraging efficiency, would forage on two types of flowers. Although an abstract model, we can use it to illustrate how birds might interact with bees at different-coloured flowers. Consider a community that includes bees and birds, and red and blue flowers. Let us assume that the flowers differ only in their colour, that there are only two patches of flowers (one of blue, the other of red flowers), and that the density of flowers is the same in both patches. (For a general analysis, with the same qualitative results, see Possingham 1992 .) The question is: how many birds and bees should forage at the red and blue patches so that their intake of nectar is maximised? The expected intake rate is the average amount of nectar obtained per flower (or standing crop) divided by the time it takes to find and exploit a flower. If the flowers are the same distance apart and birds can detect red and blue flowers equally well, then travel time is independent of flower colour. Under these circumstances, an ecological equilibrium, with birds exploiting red and blue flowers equally, would indicate that the amount of nectar available from both flower colours was identical. Now add a few bees to this community of birds, sufficiently few that their intake of nectar is negligible. We know that the standing crop is the same at red and blue flowers. However, we also know that bees require more time to find red flowers than blue ones ( Spaethe et al. 2001 ), so their intake rate of nectar will be higher at blue flowers, and they will all go to the blue patch. If we continue to add bees one at a time to this community, then sooner or later, the number of bees will no longer be sufficiently low for us to ignore their depleting effect on the nectar available. What will happen at that point? Will bees now start visiting red flowers? Not yet. For a bee to visit the red patch, the difference in standing crop between red and blue flowers would have to be large enough to compensate for the difference in detection time. Before that happens, some birds will shift to the red patch. Indeed, since birds require the same time to detect red or blue flowers, some birds will move from the blue to the red patch as soon as bees start to noticeably reduce the nectar available from the blue flowers. What Possingham's model predicts, therefore, is that when the number of bees is large enough, all birds will forage at the red patch. Only when the difference in standing crop between red and blue flowers is so large that it compensates for the reduced detectability of red flowers, will bees start visiting the red patch. To conclude, there will be an association between red flowers and birds. Birds will exploit red flowers, and bees blue flowers. In addition, depending on the relative abundance of bees and birds (and of red and blue flowers), either birds or bees, but never both simultaneously, can also exploit the other flower type ( Figure 3 ). Figure 3 Possible Outcomes of Possingham's Model (1992) Each panel represents one possible outcome. If red flowers predominate (left), bees forage at blue and red flowers, while all birds forage at red flowers. If blue and red flowers are equally abundant (middle), there is complete resource partitioning, with bees foraging at blue flowers and birds at red flowers. If blue flowers become more abundant (right), all bees forage at blue flowers, while birds forage at red and blue flowers. Niche Partition and the Evolution of Red Flowers Possingham's model (1992) helps to explain the ecological association between flower colour and pollinator type, provided that both flower colours and pollinator types are present—but why did the red colouration of these flowers evolve in the first place? We believe that the model can also help explain the evolution of red coloration in bird-pollinated flowers. To understand the evolutionary process, consider a community where bees and birds are present, and where two flower species coexist. One flower type, the generalist flower, is blue and is efficiently pollinated by bees and birds alike. The blue Rocky Mountain penstemon, Penstemon strictus , provides a good example ( Castellanos et al. 2003 ). The other flower type, or bird flower, is yellow and is efficiently pollinated by birds, but not by bees—the red beardlip penstemon P. barbatus provides an example of this type ( Castellanos et al. 2003 ). If bee visits were costly for the ancestral bird flowers, they would experience a selective pressure to become red. Bees could impose several costs on the ancestral bird flowers; for example, the number of hummingbird visits may depend on the amount of nectar available in the flowers. Throughout evolutionary history, there will be variability and heritability in flower colour (as documented for Mimulus by Bradshaw et al. 1995 ). Since both bees and birds easily detect and efficiently pollinate generalist blue flowers, there is no particular reason to expect that their colour will evolve in one direction or another. Things are otherwise for bird flowers, which are more efficiently pollinated by birds. For simplicity, consider that, at any given time, this bird flower comes in only two shades of colour, one of them with a slightly longer wavelength (an orange morph). On an ecological timescale, yellow flowers will be visited mainly by bees and orange flowers mainly by birds. Orange flowers, being more efficiently pollinated by birds, will therefore have higher fitness than yellow flowers, and given enough time, there will be selection for bird flowers to become orange. In the absence of other costs, mutant flowers with higher wavelengths (i.e., becoming redder) can invade a population of yellow flowers so long as bird flowers continue to be visited by bees (unpublished data). So bird flowers will continue to shift their colour until bees are completely excluded from the bird flowers or until further shifts deteriorate detectability by birds. This explanation for the evolution of red coloration in bird-pollinated flowers differs from the one proposed by Raven (1972) in a key respect. In our view, the main point is not that bees fly over red flowers without seeing them; it is not even that they are unable to exploit red flowers efficiently in absolute terms. It is rather a question of relative efficiency that makes bees avoid red flowers when birds are depleting their nectar; it would work just as well if birds were colourblind and perceived red flowers as badly as flowers of other colours. Of course, Possingham's model (1992) is not incompatible with birds being more efficient than bees at exploiting red flowers, and the results would be strengthened if, as has been suggested ( Raven 1972 ; Chittka and Waser 1997 ), birds are better at detecting red flowers than blue ones. Toward a Solution Comparable problems can be found in other plant–pollinator systems. For example, when several species of bumblebees coexist, resource partitioning normally doesn't follow colour, but is dependent on different parameters: the corolla length of the plant and the proboscis length of the bee. Proboscis length affects the efficiency with which flowers of different depth are exploited ( Inouye 1980 ); bumblebees with long proboscises preferentially exploit flowers with deep corollas, while bumblebees with short proboscises exploit shallow flowers ( Heinrich 1976 ). But a bumblebee with a long proboscis can also exploit shallow flowers, and, to some extent, a bumblebee with a short proboscis can exploit deep flowers, if corollas are not too deep (although they will still leave some nectar behind). Indeed, when one bumblebee type is experimentally removed, the other one is seen to exploit both deep and shallow flowers ( Inouye 1978 ). The same, we believe, should happen with flower colour: the experimental removal of birds should lead to the systematic exploitation of red flowers by bees, at least when corolla tube morphology does not prevent bees from accessing the nectar. In fact, there is even no need to perform experimental bird removals, because plants provide us with a ready-made design: bees visit flowers searching for both nectar and pollen, while most birds exploit only the nectar. Hence, bees should readily collect pollen at red bird flowers. There are numerous examples of this, although in most cases they are indirectly documented. For example, solitary bees and syrphid and muscoid flies visit the red, hummingbird-pollinated flowers of Ipomopsis aggregata to collect pollen when hummingbirds visits are frequent, while bumblebee ( Bombus appositus ) visits to collect nectar are only common when hummingbird visits are rare ( Mayfield et al. 2001 ). Outside the native range of bird-pollinated plants, the same phenomenon can be observed: in Spanish gardens, the honeybee collects pollen from Aloe arborescens plants. Bees cannot access the nectar, concealed at the bottom of the corolla tube. This is opportunistically collected by birds such as the Sardinian warbler Sylvia melanocephala (unpublished data). Another comparison of interest concerns beetle-pollinated flowers, which in the Mediterranean region have open, bowl shapes and red coloration ( Dafni et al. 1990 ). Amphicoma beetles are more efficient pollinators of these flowers than commonly occurring bees ( Dafni et al. 1990 ), so the red coloration of these flowers might help to keep other visitors (possibly bees and flies) at bay. Indeed, other bowl-shaped flowers of different colours (such as yellow, white, and purple, e.g., in the genera Cistus and Helianthemum ) are commonly visited by pollen-collecting bees and bumblebees. A particularly interesting test case is provided by the corn poppy Papaver rhoeas ; in the eastern Mediterranean region, it is pollinated by beetles and does not reflect in the UV ( Dafni et al. 1990 ), while in central and western Europe it reflects in the UV ( Daumer 1958 ) and is pollinated by bees. Although refinements of Possingham's model, such as developing a prey-model version, or introducing stochasticity or several foraging constraints, might help us determine the extent to which we should expect resource partitioning along the colour dimension to take place, it is, in our view, far more pressing to determine the extent and conditions under which bees exploit red flowers (i.e., through comparisons of pollen vs. nectar exploitation, bird exclusion experiments, etc.), the detection time of red flowers against a natural background, and the effect of flower colour and size on flight mode in the field. Only then will we be able to fully unravel the factors that solve this fascinating mystery.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521733.xml
406401	The Human Sense of Smell: Are We Better Than We Think?	Gordon Shepherd challenges the notion - based on genetic evidence - that olfaction is less well developed in humans as compared to other mammals	“… a complete, comprehensive understanding of odor … may not seem a profound enough problem to dominate all the life sciences, but it contains, piece by piece, all the mysteries.” — Lewis Thomas One of the oldest beliefs about human perception is that we have a poor sense of smell. Not only is this a general belief among the public, but it appears to have a scientific basis. Recent genetic studies show a decline in the number of functional olfactory receptor genes through primate evolution to humans. Human evolution was characterized by the gradual ascendance of vision and reduction of smell, evidenced in the anthropological record by the progressive diminution of the snout as the eyes moved to the middle of the face to subserve depth vision ( Jones et al. 1992 ). Concurrently, the use of an arboreal habitat and the adoption of an erect posture moved the nose away from the ground, with its rich varieties of odors. However, some recent behavioral studies suggest that primates, including humans, have relatively good senses of smell. Resolution of this paradox may come from a larger perspective on the biology of smell. Here we begin by reassessing several overlooked factors: the structure of the nasal cavity, retronasal smell, olfactory brain areas, and language. In these arenas, humans may have advantages which outweigh their lower numbers of receptors. It appears that in the olfactory system, olfactory receptor genes do not map directly onto behavior; rather, behavior is the outcome of multiple factors. If human smell perception is better than we thought, it may have played a more important role in human evolution than is usually acknowledged. Gene Studies From rodents through the primate series to humans there is a progressive reduction in the proportion of functional olfactory receptor genes ( Rouquier et al. 2000 ; Gilad et al. 2004 ). Mice have approximately 1,300 olfactory receptor genes, of which some 1,100 are functional ( Young et al. 2002 ; Zhang and Firestein 2002 ), whereas humans have only some 350 functional genes of approximately 1,000 ( Glusman et al. 2001 ; Zozulya et al. 2001 ). The conclusion seems obvious: the low number of functional olfactory receptor genes in humans compared with rodents—and presumably most other mammals—is directly correlated with the evolutionary decline in the human sense of smell. Behavioral Studies Although these conclusions seem incontrovertible, they are challenged by some recent behavioral studies. One type of study shows that much of the olfactory system can be removed with no effect on smell perception. The olfactory receptor genes map topographically onto the first relay station, a sheet of modules called glomeruli in the olfactory bulb. Up to 80% of the glomerular layer in the rat can be removed without significant effect on olfactory detection and discrimination ( Bisulco and Slotnick 2003 ). If the remaining 20% of the glomeruli—and the olfactory receptor genes they represent—can subserve the functions of 1,100 genes, it implies that 350 genes in the human are more than enough to smell as well as a mouse. Another type of study has tested smell perception in primates, and has shown that, despite their reduced olfactory receptor gene repertoire, primates, including humans, have surprisingly good senses of smell ( Laska et al. 2000 ). Comparing the data on smell detection thresholds shows that humans not only perform as well or better than other primates, they also perform as well or better than other mammals. When tested for thresholds to the odors of a series of straight-chain (aliphatic) aldehydes, dogs do better on the short chain compounds, but humans perform as well or slightly better than dogs on the longer chain compounds, and humans perform significantly better than rats ( Laska et al. 2000 ). Similar results have been obtained with other types of odors. A third type of study demonstrating human olfactory abilities shows that in tests of odor detection, humans outperform the most sensitive measuring instruments such as the gas chromatograph. These results indicate that humans are not poor smellers (a condition technically called microsmats), but rather are relatively good, perhaps even excellent, smellers (macrosmats) ( Laska et al. 2000 ). This may come as a surprise to many people, though not to those who make their living by their noses, such as oenologists, perfumers, and food scientists. Anyone who has taken part in a wine tasting, or observed professional testing of food flavors or perfumes, knows that the human sense of smell has extraordinary capacities for discrimination. The Mystery Here, then, is the mystery: how can one reconcile a relatively high sensitivity to smell with a relatively low number of olfactory receptors in the nose? To answer this question, I think we need to look beyond the olfactory receptor genes and consider olfaction in its full behavioral context. This requires considering several overlooked aspects of the olfactory system: the nasal cavity, the oropharyngeal cavity, the olfactory brain, and the role of language. In this article I focus on behaviors related to conscious perception of ordinary smells. Pheromones, and the rich world of unconscious effects of odors and pheromones, are beyond the present scope (cf. Jacob et al. 2004 ), though they undoubtedly will add to the general conclusions. The Filtering Apparatus of the Nasal Cavity A marked difference between the noses of primates and other mammals is that in nearly all nonprimate mammals, the nasal cavities contain at the front a much-convoluted filtering apparatus (formed by the ethmo- and maxillo-turbinals) covered with respiratory membrane. This filtering apparatus is a biological air conditioner ( Negus 1958 ) with three key functions: cleaning, warming, and humidifying the inspired air. An important function of the filtering apparatus is presumably to protect the nasal cavity from infections. In many mammals, air drawn into the nose is often highly contaminated with bacteria from fecal material, decaying animal and plant material, and noxious fumes from the environment, all of which attack the olfactory epithelium. Rodents are susceptible to chronic rhinitis, which causes substantial loss of functioning olfactory receptor cells ( Hinds et al. 1984 ). This filtering, however, might have negative consequences for odor detection. Warming and humidification presumably enhance the odor-stimulating capacity of the inhaled air, but cleaning would remove odor molecules by absorbing them into the lining of the epithelium, an effect which could be large depending on the size of the filtering apparatus. If so, mammals with large snouts might have a large inventory of olfactory receptors at least in part to offset the loss of odor molecules absorbed by the filtering apparatus. How do these considerations relate to humans? The evolution of humans involved lifting the nose away from the noxious ground environment as they adopted a bipedal posture ( Aiello and Dean 1990 ). This would have reduced the need for the filtering apparatus and with it the losses of absorbed odor molecules. The large numbers of olfactory receptors and receptor cells would have come under reduced adaptive pressure and could accordingly be reduced in proportion. By this hypothesis, during human evolution the snout could be reduced in dimensions and complexity without compromising the ultimate amounts of odorized air reaching the olfactory epithelium. The reduced snout allowed the eyes to come forward and lie closer together to promote more effective stereoscopic vision. Thus, vision could become more dominant in humans without sacrificing unduly the sense of smell. Tests of this hypothesis are needed, including calculations of air flows and odor losses through the filtering apparatus in mammals with extensive filtering apparatuses compared with the simpler nasal cavities of primates. Humans Receive Richer Retronasal Smells Being carried in with inhaled air (the orthonasal route) is not the only way for odor molecules to reach the olfactory receptor cells. Odor molecules also reach the olfactory receptor cells via the retronasal route, from the back of the oral cavity through the nasopharynx into the back of the nasal cavity. Although the orthonasal route is the one usually used to test for smell perception, the retronasal route is the main source of the smells we perceive from foods and liquids within our mouths. These are the smells that primarily determine the hedonic (i.e., pleasurable or aversive) qualities of foods, and that, combined with taste and somatosensation, form the complex sensation of flavor. It is likely, for several reasons, that this is an important route for smell in humans. First, with the adoption of bipedalism, humans became increasingly wide ranging, with concomitant diversification of diet and retronasal smells. Second, the advent of fire, perhaps as early as 2 million years ago ( Wrangham and Conklin-Brittain 2003 ), made the human diet more odorous and tasty. From this time also one can begin to speak of human cuisines of prepared foods, with all their diversity of smells. Wrangham and Conklin-Brittain (2003) support the view that prepared cuisines based on cooked foods are one of the defining characteristics of humans. Third, added to the cooked cuisines were fermented foods and liquids, with their own strong flavors. These developments occurred among the early hunter-gatherer human cultures and continued through the last ice age. With the transition to agricultural and urban cultures 10,000 years ago, human cuisines changed by the advent of animal domestication, plant cultivation, use of spices, and of complex procedures, such as those for producing cheeses and wines, all of which produced foodstuffs that especially stimulate the smell receptors in the nose through the retronasal route and contribute to complex flavors. These considerations suggest the hypothesis that the retronasal route for smells has delivered a richer repertoire of smells in humans than in nonhuman primates and other mammals (see Figure 1 ). Research on retronasal olfaction is being actively pursued (reviewed in Deibler and Delwiche 2004 ). Studies are needed of the evolutionary pressures on this route in addition to the pressures on the evolution of the snout. Figure 1 Hypothetical “Odor Wheel” Representing and Comparing the Odor Worlds of Mouse and Human The inner part represents the different categories of odors for the mouse; the relative importance of each category for mouse smell-dependent behavior is indicated by the area of each wedge. The outer part represents the same categories for the human; the importance of each category for human smell-dependent behavior compared with the mouse is indicated by the area of each wedge. Note the greater importance of food odors for the human, reflecting the factors discussed in the text. Note also the retention of some sensitivity in humans to social odors and other odors prominent in rodents, though in many cases to still undetermined degrees. Based on numerous sources and the hypotheses discussed in the text. Humans Smell with Bigger and Better Brains Comparisons of the decreasing size of the olfactory system relative to expansion of the visual, auditory, and somatosensory systems usually focus on the olfactory bulb and lateral olfactory tract, which are relatively small. However, what matters more are the central olfactory brain regions that process the olfactory input as the basis for smell perception. These regions are more extensive in humans than is usually realized. The dedicated olfactory regions include the olfactory cortex, the olfactory tubercle, the entorhinal cortex, parts of the amygdala, parts of the hypothalamus, the mediodorsal thalamus, the medial and lateral orbitofrontal cortex, and parts of the insula ( Neville and Haberly 2004 ). These regions are involved in immediate processing of odor input and probably subserve the specific tasks of smell detection and simple smell discrimination. For more complex tasks, memory becomes important in comparing smells, thus involving the temporal and frontal lobes (e.g., Buchanan et al. 2003 ) and the specifically human higher association areas. It may be hypothesized that these regions enable humans to bring far more cognitive power to bear on odor discrimination than is possible in the rodent and other mammals. The reduced repertoire of olfactory receptor genes in the human is thus offset by the expanded repertoire of higher brain mechanisms. Rather than being restricted to a tiny part of the brain, olfactory processing of complex smells, such as those produced by human cuisines, draws on the enlarged processing capacity of the human brain. Language Is Necessary for Human Smell In the enlarged processing capacity for perceiving and discriminating odors, language plays a critical role. This seems paradoxical, for we have great difficulty describing a smell in words. Insight into this difficulty comes from the finding that different smells are represented in the olfactory bulb by different patterns of olfactory glomerular activity. These patterns function as virtual “odor images” ( Xu et al. 2003 ). It has been hypothesized that these odor images provide the basis for discrimination between odors, analogous to the way that retinal images are the basis for discrimination of visual pattern stimuli. The complex patterns constituting odor images may be considered as analogous to the complex patterns constituting visual images of faces. And just as we are very good at recognizing a human face, yet have difficulty describing it in words, we have a hard time describing and verbally comparing odor images. Because of this difficulty, describing a smell or a taste in words is very demanding. A professional wine tasting, for example, requires many steps: analysing both orthonasal and retronasal perception, comparing the two in memory with each other and with all other wines to be compared, identifying the constituent properties separate from the hedonic qualities, and finding the words to describe the process as it unfolds, leading to the final formulation to characterize the quality of the wine and identify it as distinct from all others. It may be characterized as hard cognitive work that only a human, among all the animals with olfactory organs, can do. It may be argued that this is what humans are adapted to do ( Wrangham and Conklin-Brittain 2003 ). This cognitive work is largely independent of the numbers of peripheral receptor cells and their genes. A good analogy is with language. There are some 17,000–20,000 auditory nerve fibers in the rat and cat and some 25,000–30,000 in the human (cf. Hall and Masengill 1997 ). This modest increase in the input from the peripheral auditory receptors provides little basis for the development of human speech and language, which had much more to do with the increase in the central brain mechanisms that elaborate the input. It may be hypothesized that a similar conclusion applies to human olfaction. Implications for Systems Biology A general result from these considerations is that there appears not to be a one-to-one relation between the number of olfactory receptor genes and the detection and discrimination of odors. This implies that we are dealing with a fundamental problem in relating genes to systems behavior: a given set of genes may not map directly onto a given behavior. In this respect the mystery being addressed here is a caution for the new era of “systems biology” and against any belief that behavior can be related directly to genomes, proteomes, or any other type of “-ome.” We are reminded instead that the functional ecology of the body is dependent on many factors. Conclusions Much about the sense of smell seems enigmatic and conflicting. This is partly because of the inherent difficulties in presenting smell stimuli, and partly because there is not yet a recognition of all the relevant mechanisms that are involved. It may be hoped that the hypotheses and mechanisms discussed here can help to address and resolve the mystery of the apparent noncorrelation of olfactory receptor gene numbers with smell acuity, and in doing so stimulate a major reassessment of human smell perception. Such an effort cuts across many academic disciplines. Molecular biologists need to continue their efforts to characterize the olfactory genomes of humans and nonhuman mammals more closely, to compare how different organisms sample odor space. Physiologists need to devise high-throughput systems to test these odor spaces. Behavioral neuroscientists need to develop increasingly accurate tests of olfactory function that enable comparisons across different species. Psychologists need to explore even more vigorously the subtle ways that smells can influence human behavior. Anthropologists and paleontologists need to study the olfactory parts of the cranium and face from this new perspective, to reassess the role that both orthonasal and retronasal smell may have played in primate and human evolution. The factors reviewed here suggest that the sense of smell is more important in humans than is generally realized, which in turn suggests that it may have played a bigger role in the evolution of human diet, habitat, and social behavior than has been appreciated. All of these considerations should stimulate a greater interest in this neglected sense.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC406401.xml
423130	Ecology Drives the Worldwide Distribution of Human Diseases	Identifying the factors underlying the origin and maintenance of the latitudinal diversity gradient is a central problem in ecology, but no consensus has emerged on which processes might generate this broad pattern. Interestingly, the vast majority of studies exploring the gradient have focused on free-living organisms, ignoring parasitic and infectious disease (PID) species. Here, we address the influence of environmental factors on the biological diversity of human pathogens and their global spatial organization. Using generalized linear multivariate models and Monte Carlo simulations, we conducted a series of comparative analyses to test the hypothesis that human PIDs exhibit the same global patterns of distribution as other taxonomic groups. We found a significant negative relationship between latitude and PID species richness, and a nested spatial organization, i.e., the accumulation of PID species with latitude, over large spatial scales. Additionally, our results show that climatic factors are of primary importance in explaining the link between latitude and the spatial pattern of human pathogens. Based on our findings, we propose that the global latitudinal species diversity gradient might be generated in large part by biotic interactions, providing strong support for the idea that current estimates of species diversity are substantially underestimated. When parasites and pathogens are included, estimates of total species diversity may increase by more than an order of magnitude.	Introduction Generally, the number of plant and animal species declines as one moves away from the equator ( Pianka 1966 ; Stevens 1989 , 1992 ; Rohde 1992 ; Brown 1995 ; Kaufman 1995 ; Rosenzweig 1995 ; Roy et al. 1998 ; Huston 1999 ; Chown and Gaston 2000 ; Hawkins and Porter 2001 ). This pattern, known as the latitudinal species diversity gradient, has been documented for many contemporary taxonomic groups (see Brown 1995 ; Rosenzweig 1995 ; Gaston and Blackburn 2000 ; Allen et al. 2002 ; Stevens et al. 2003 ). Over 30 hypotheses have been proposed to explain it ( Rohde 1992 ), and it is only over the past several years that the most credible candidates have been identified; these are hypotheses related to area, energy, and time ( Gaston and Blackburn 2000 ; Rahbek and Graves 2001 ) and to habitat heterogeneity and geometric constraints ( Rahbek and Graves 2001 ). The vast majority of studies exploring the latitudinal species diversity gradient have focused on free-living organisms, such as herbivores, mammals, and angiosperms, and with rare exception ( Hillebrand et al. 2001 ; Curtis et al. 2002 ; Nee 2003 ), none has examined large-scale latitudinal species diversity patterns of pathogenic microorganisms. Biotic interactions such as parasitism, predation, and symbiosis have been often invoked as a causal mechanism for the gradient (see Rohde 1992 ), but no serious attempts have been made to quantify its importance to biodiversity. Parasitic and infectious diseases (PIDs), in particular, could prove to be key in understanding large-scale patterns of species diversity on Earth since they comprise a major part of total biological diversity ( Combes 1995 ; Poulin 1998 ). Moreover, our understanding of human diseases and the existence of complete data sets provide an incomparable opportunity to explore the existence of a relationship between PID species richness and latitude, and to identify the determining factors of this latitudinal gradient. In recent years, research into nonrandom organization in parasite communities has turned, e.g., to the possible existence of nestedness. Nested structure is a hierarchical organization of species composition in which assemblages with successively lower species richness tend to be nonrandom subsets of richer assemblages ( Hanski 1982 ; Patterson and Atmar 1986 ; Patterson and Brown 1991 ; Poulin and Guégan 2000 ). Some species are widely distributed and occur in many communities, whereas other species have more restricted distributions and occur only in a subset of the richest samples ( Figure 1 ). When analysing the most important mechanisms responsible for generating nestedness, Wright et al. (1998) cited four candidate factors: random sampling, area, isolation, and habitat type. In the present study we seek an answer to the following question: To what extent is the global distribution of human pathogens specified by the properties of the physical environment or the organism itself, and to what extent does it depend on chance events? Figure 1 The Spatial Organization of Species Letters represent different PID species. Numbered rectangles represent different countries or areas. (A) Nested organization of species. Applying Diamond's theory, we here distinguish (1) “high-S” species, like species E, which are exclusively confined to the most species-rich communities; and (2) “tramps,” like species A, which occur mostly in richer communities but also in species-poor communities (e.g., measles, which is found in virtually every country). Thus, this nested pattern implies that some pathogens are restricted to the tropics, while others, more ubiquitous species, are widely and regularly distributed all over the world. (B) Random distribution of species, where no spatial organization occurs (see also Materials and Methods ). We examine the global spatial distribution of species richness for human PIDs, and test the hypothesis that human diseases follow a latitudinal species richness gradient, with low latitudes being the richest zones in pathogen species diversity. We then test two additional propositions: (i) PID assemblages show nested species patterns along latitudinal gradients, i.e., PIDs present at northern latitudes are also present in larger PID assemblages of equatorial zones, and (ii) PID assemblages may be strongly influenced by environmental climatic forces. Results The Latitudinal Gradient of Species Richness for Pathogens After correcting for cofactors (i.e., area and socio-demographic, physical, and environmental parameters) that could influence the relationships between latitude and PID species richness, we still found that species richness in human pathogens is strongly correlated with latitude ( Table 1 ). On average (seven times out of ten), tropical areas harbor higher pathogen species diversities compared to more temperate areas. Figure 2 A illustrates the change in PID species diversity with latitude across the two hemispheres. Figure 2 The Latitudinal Gradients of PID Species (A) Relationship between PID species richness and latitude across the two hemispheres. Linear relationships between PID species richness and latitude (dotted lines) are highly significant ( F = 12.29, df = 29, p = 0.0015 and F = 18.01, df = 130, p < 0.0001 for Southern and Northern hemispheres, respectively). No difference in disease species richness with latitude across the two hemispheres was observed (interaction: F = 2.68, df = 159 , p = 0.1036). Residuals of PID species richness on the y axis were extracted from minimal models controlling for the effects of confounding factors on PID species diversity estimates (see Materials and Methods ). Locally weighted regression (tension 0.5) did not change the general linear shape. Latitude is expressed in minute degrees. (B) Presence/absence matrix for the 229 distinct PID species across the hemispheres. The figure was generated by the Nestedness Temperature Calculator (see Atmar and Patterson 1995 ). The distribution is nonsymetrical because of the 224 studied countries, 172 countries are found in the Northern hemisphere versus only 52 in the Southern one. (B) indicates that PID species diversity decreases as one moves northwards or southwards from the equator. The black exponential curves are the occurrence boundary lines (see Materials and Methods ). The color scale indicates the nonuniform probability of state occupancy among all of the cells of the matrix, i.e., the probability of encountering a species as function of its position in the matrix. Black cells are highly predictable presences, whereas red cells are unexpected presences. (C) Monte Carlo–derived histogram after 1,000 permutations. The histogram represents the 1,000 values obtained after Monte Carlo permutations. The average theoretical value under the null hypothesis is compared to our real value, to assess the likelihood that the parent matrix was nonrandomly generated. The probability is highly significant ( p < 0.0001), confirming that the spatial organization of PID species richness on the largest scale matches the nested species subset hierarchy illustrated in Figure 1 A. The symmetrical Gaussian distribution indicates that 1,000 permutations are enough to obtain reliable variance estimates for probability calculations. Table 1 Minimal Models for Latitude Explaining PID Species Richness of Etiological Groups Of all factors included as potential predictors of PID species richness (see Materials and Methods ), Table 1 focuses on the emergence of latitude as a possible explanatory variable in minimal models. When significant, the probability value (p), the degrees of freedom (df), and the sign of slope (+/−) are given The Nested Organization of Pathogen Species over Large Scales Monte Carlo analyses confirmed an overall nested species pattern of global distribution in PID species richness ( N s = 2,481.4, R 0 and R 1 procedures, p < 0.0001) and showed diversity to be strongly nested, with some anecdotal differences across the different groups of etiological agents (all groups, p < 0.0001, except for vector-borne viruses, with the R 1 procedure [ N s = 1,787, p = 0.0015]). When considering the Northern and Southern hemispheres separately, both were highly nested (R 0 and R 1 procedures, N s = 6,602, p < 0.0001 and N s = 1,230, p < 0.0001, respectively). This was confirmed by the R 00 procedure used by the Nestedness Temperature Calculator program ( Atmar and Patterson 1995 ), which provides a useful graphic representation of the results ( Figure 2 B), showing that PID species diversity decreases as one moves northwards or southwards from the equator ( F = 28.2307, df = 161, p < 0.0001). The occurrence boundary lines (black exponential curves) were fitted by nonlinear regression ( y = 1.51 + 20.01e -0.29x and y = 1.65 + 35.87e -0.36x for Northern and Southern hemispheres, respectively). Results from Monte Carlo simulations confirmed that our nested matrix was nonrandomly generated ( p < 0.0001) ( Figure 2 C). The spatial organization of PID species richness on the largest scale matches the nested species subset hierarchy illustrated in Figure 1 A. Thus, pathogen species that compose a depauperate community in temperate conditions statistically constitute a proper subset of those occurring in warmer conditions, and evidence of pathogen species occurring in temperate areas but not in tropical ones was rare or anecdotal. It should be noted that, at this large spatial scale, our study demonstrates a nested pattern in PIDs , with a progression of species richness from polar regions to the equator, indicating that nestedness is strongly associated with latitude (see Figure 2 B). But this does not contradict the fact that some pathogens may be strict endemics of more temperate areas (e.g., Lyme disease). The Effect of Climatic Variables on Biodiversity Latitude is a proxy variable for a wide range of covarying bio-climatic factors and in itself has no meaning regarding factors potentially affecting species diversity. We therefore investigated the relationship between pathogen diversity and individual climatic variables reflected in the composite variable “latitude” ( Table 2 ). Results show significant positive correlations between pathogen species richness and the maximum range of precipitation after Bonferroni multiple corrections for all six of the PID taxa considered: bacteria (r = 0.3545, df = 213, p < 0.0001), viruses directly transmitted from person to person (r = 0.2350, df = 215, p < 0.0001), viruses indirectly transmitted via a vector (r = 0.3575, df = 215, p < 0.0001), fungi (r = 0.3554, df = 216, p < 0.0001), protozoa (r = 0.3744, df = 216, p < 0.0001), and helminths (r = 0.4270, df = 215, p < 0.0001). On the other hand, the relationship between PID species richness and monthly temperature range was only significant for three groups of pathogens: bacteria (r = 0.3016, df = 213, p < 0.0001), directly transmitted viruses (r = 0.2142, df = 214, p = 0.0015), and helminths (r = 0.2590, df = 213, p = 0.0001). In contrast to previous results ( Allen et al. 2002 ), we found no significant relationship between PID species richness and mean annual temperature. Finally, only the relationship between bacteria species richness and mean annual precipitation was significant ( r = 0.1987, df = 213, p = 0.0034). Very little difference was observed among hemispheres concerning these relationships (data not shown). Table 2 Relationship Between PID Species Richness by Etiological Group and Four Bio-Climatic Factors Pearson's correlation ( r ), sign of slope (+/−) and significance levels ( p ) are given. * indicates significance levels which become nonsignificant after the Bonferroni correction ( k = 6 multiple comparisons) Taken together, these findings indicate that the species richness of human pathogens, their spatial distribution and organization on a large scale, the maximum range of precipitation, and, to a lesser extent, monthly temperature might be intimately connected in generating the observed pattern of disease diversity. Discussion To our knowledge, this is the most comprehensive report of how PID species richness varies with latitude and the ecological factors behind observed trends. Our results support previous studies in showing that species diversity increases as one proceeds from the poles to the equator ( Pianka 1966 ; Stevens 1989 ; Rohde 1992 ; Brown 1995 ; Rosenzweig 1995 ; Chown and Gaston 2000 ). This similarity in the patterns of PID species and free-living organisms suggests that common mechanisms are at work. Regardless of whether PID richness simply tracks host diversity or, rather, is determined to a greater extent by exogenous factors, our analyses indicate that the most likely explanation for these patterns is the climatically-based energy hypothesis, i.e., that energy availability generates and maintains species richness gradients ( Rohde 1992 ; Gaston and Blackburn 2000 ; Allen et al. 2002 ; Hawkins et al. 2003 ). Many studies have identified correlations between gradients in species diversity and variation in climate ( Hill et al. 1999 ). Climate, in turn, largely determines the species of plants and animals that live in those areas. According to our results—and in contrast to the results of Allen and colleagues (2002) , who showed that environmental temperature was the best predictor of species diversity for terrestrial, freshwater, and marine ectotherm taxa—the maximum range of precipitation is highly correlated with the latitudinal gradient of pathogen species, with diversity significantly increasing with this climate-based factor. Interestingly, the annual variation of precipitation around the mean (and not the mean itself) was the best predictor overall of pathogen species distribution. This suggests that pathogen species, their vectors, or their hosts tend to be adapted to regions having more contrasted wetness and dryness conditions through the year (i.e., in tropical regions). Many parasites obviously require water or humid conditions to complete their life cycle, e.g., vector-borne diseases. So, the physical factor of precipitation variation may affect parasitic and infectious microorganism diversity, if the biological cyclicity of a variety of parasitic and infectious stages have adapted to the variability of precipitation. This might be why “latitude” does not appear in the minimal generalized linear models (GLIMs) for explaining the richnesses of bacteria, directly transmitted viruses, and fungi, these taxa being “internal” to the host, so less directly affected by environmental variability. Moreover, these taxa may more readily spread over longer distances via their hosts, and this should minimize the impact of environmental conditions. In contrast, taxa with “external” stages, like helminths or vector-transmitted pathogens, are more influenced by their environment. Nevertheless, other causes might explain why certain taxa do not conform to the general pattern, notably (1) the absence of possible explanatory variables in the GLIMs, (2) missing or imprecise information due to the large scale of our study, or (3) the real absence of correlations between the spatial distributions of certain taxonomic groups and the variables considered here. All three nestedness models (see Materials and Methods ) explained some of the variation in pathogen species across latitudes. Distance and isolation from pathogen species–rich regions in the tropics may sort PID species by their extinction–colonization dynamics ( Lomolino 1996 ). In addition, the availability of new hosts and reservoirs, passive sampling, and probabilistic filters screening species with particular characteristics (local habitat suitabilities, differential colonization capacities of species, and sustainability of viable populations within their environment) may further limit PID species ( Wright et al. 1998 ) and thus strongly affect the spatial organization of PID species. Nestedness might in fact be an inevitable second-order consequence of the same factors that cause variation in species richness and range size ( Gaston and Blackburn 2000 ). In addition, our results suggest that total species diversity on the planet might be substantially underestimated, especially because inventories generally focus attention on the most charismatic groups ( Shaw and Hochberg 2001 ), and little is known about the biodiversity of microorganisms associated with each considered group of organisms, i.e., hosts ( Ashford and Crewe 1998 ; Ashford 2000 ; Nee 2003 ). Based on a single host species, humans, we estimate that true tropical pathogen species diversity is greater than current estimates by a factor of about 22 in the Northern Hemisphere and about 37 in the Southern Hemisphere. If our work is representative of other (host) species, diversity may be currently underestimated by more than an order of magnitude, and based on our findings, this differential should increase as one goes from temperate to tropical latitudes. Our work quantitatively demonstrates that parasitic and pathogenic organisms, as representatives of biotic interactions, strongly amplify the general latitudinal gradient in species richness. The smallest organisms that have been neglected by science could very well be the biggest in generating the observed diversity pattern. The demonstration that parasitic and infectious organisms in humans do not constitute random assemblages at large spatial scales, but rather that many types of microorganisms show a predictable geographical distribution over the planet, could have important implications for public health policies. Our results show that climatic factors are of primary importance in explaining the occurrence and diversity of human pathogens, suggesting that global climate change might have cascading effects regarding the risks of PIDs. For instance, if specific temperate areas were to become more tropical, our results suggest that PID species and their associated vectors/reservoirs would be likely to colonize these changed areas. This would imply a progressive dissolution of the latitudinal effect and of the nested hierarchical structure as observed in the present study as pathogen species became more globally distributed. There is some recent evidence for this hypothesis (see Lindgren and Gustafson 2001 ). Other variables are indeed important in explaining global-scale patterns of human pathogens (e.g., modernization, urbanization, and pauperization, especially in developing countries). Thus, we do not mean to imply that latitude and surrogate variables are the only ones affecting PID species richness. Nevertheless, our results challenge the conventional wisdom that socio-economic conditions are of preponderant importance in controlling or eradicating diseases. These considerations indicate that a better understanding of PID species diversity and community dynamics in a changing world will be one of the major challenges in environmental epidemiology in the future. Materials and Methods Presence/absence matrix We compiled data on PID occurrence for a total of 332 different human pathogens, including bacteria, viruses, fungi, protozoa, and helminths distributed across 224 nations. Epidemiological data on PID species were extracted from the Global Infectious Diseases and Epidemiology Network database ( http://www.cyinfo.com ).The presence/absence matrix for the 229 distinct PIDs (after elimination of 103 unavailable values) across the Northern and Southern hemispheres was organised employing the Nestedness Temperature Calculator ( Atmar and Patterson 1995 ). One hundred seven ubiquitous pathogen species were eliminated from the database because the information they contained was entirely redundant with that of the most ubiquitous species already present in the matrix. The matrix of species presence/absence provides distributional information about which species occurs in which countries. GLIMs We employed GLIMs ( Crawley 1993 ; Venables and Ripley 1999 ) from the S-Plus statistical package ( Venables and Ripley 1999 ) to identify and characterize the effects of potential independent parameters and their interaction terms on PID species richness, which is the total number of human diseases known within the boundary limits of each country. It has been argued that species richness increases with increasing area sampled ( Hawkins and Porter 2001 ; but see Rohde 1997 ). Therefore, we included total surface area per country (in square kilometers) in our analyses, in order to control for its effect in the multivariate analysis. Similarly, we considered human population size and human population density per country (in persons per square kilometer), both highly colinear with surface area, as possible explanatory factors, since the number and density of human hosts may also influence parasite species richness ( Anderson and May 1991 ; Guégan et al. 2001 ). In addition, we considered a variety of environmental, demographic, and economical factors. Variables selected as environmental factors for each country were (1) continent, (2) hemisphere, (3) whether the country was insular or continental, (4) percentages of arable land, permanent pastures, permanent crops, irrigated lands, forest woodlands, and “other,” (5) mean latitude coordinate, centered at the country barycenter (in minute degrees), and (6) mean longitude (in minute degrees) from the Greenwich Meridian. Variables selected as demographic factors were (1) human population size, (2) human population density (persons per square kilometer), (3) human birth rate (births/1,000 people/year), (4) human death rate (deaths/1,000 people/year), and (5) annual population growth rate (average annual percent change in the population, resulting from a surplus or deficit of births over deaths and the balance of migrants entering and leaving a country). We employed the gross national product (per capita in United States dollars) as the economic factor, which is the value of all final goods and services produced within a nation in a given year, plus income earned by its citizens abroad, minus income earned by foreigners from domestic production. We also selected a few other variables linked to particular landscape practices (percentages of arable land, permanent pastures, permanent crops, irrigated lands, forest woodlands, and “other”), which were supposed to interact with the production of the nation. Data were collected from The World Factbook 2001 on the Internet ( http://www.cia.gov/cia/publications/factbook ) and from the appendix of Scott and Duncan (1998) . To relate richness to environmental factors, we employed a GLIM with a Poisson error and a log link function (see Wilson and Grenfell 1997 ). Factors and their interaction terms were selected by a backward stepwise elimination procedure from the general model according to the Akaike criterion ( Crawley 1993 ; Burnham and Anderson 2002 ). Deviances were compared using χ 2 statistics. Spatial autocorrelation analysis When data suggested nonlinear trends, explanatory variables were transformed and fitted again to improve their contribution to the models. Since close geographical neighbors (i.e., two countries sharing a boundary) probably also share common PID species, simple cross-country comparisons could include spatial autocorrelation artefacts ( Manly 1991 ). To test whether this influenced our regressions, we employed Monte Carlo simulations to calculate Moran's index (I) between the matrix of PID species richness and the matrix of distances across the 224 countries ( Manly 1991 ; Guégan and Hugueny 1994 ). The I value is bound between −1 and +1, with 0 indicating no spatial autocorrelation, and +1/−1 indicating a strong positive/negative autocorrelation, respectively. We first computed the correlation coefficient based on all pairs of neighboring countries, and we randomly estimated 99 coefficients each time, permuting the matching countries. The decision rule, ensuring significance at α = 0.01, consisted in rejecting the null hypothesis of the absence of spatial correlation if the correlation coefficient obtained for nonpermuted data was maximum among all 100 coefficients. The calculation of I using Monte Carlo simulations indicated no strong spatial autocorrelation ( I 0 = 0.08 equals I s = 0.11 at α = 0.01), suggesting that the close similarities between PID species richness and composition observed between neighboring countries conforms to the latitudinal diversity gradient. Nestedness analysis We also employed Monte Carlo simulations ( Manly 1991 ; Guégan and Hugueny 1994 ) to evaluate PID spatial organization at the largest scale. We used the data matrix of presence/absence values for 229 different pathogen species of the total dataset comprising 224 countries. We assessed the degree of nestedness of the system using two different, but complementary, analysis programs: (1) Nestedness ( Guégan and Hugueny 1994 ) and (2) Nestedness Temperature Calculator ( Atmar and Patterson 1995 ). Nested diversity patterns are identified when species found in depauperate communities represent nonrandom subsets of progressively richer communities ( Gaston and Blackburn 2000 ; Poulin and Guégan 2000 ). In procedure 1, pathogen species were either selected with uniform probability (null model R 0 ) or with a probability proportional to their incidence (R 1 ) ( Guégan and Hugueny 1994 ), whereas in procedure 2 we tested the null model R 00 ( Poulin and Guégan 2000 ; see also Cook and Quinn 1998 ; Wright et al. 1998 ; Gaston and Blackburn 2000 ). Nestedness Temperature Calculator generates simulated null matrices without either row or column constraints (hence “00”); only the total number of presences is fixed at the observed value. All three null hypotheses assume that sites are independent of one another ( Wright et al. 1998 ). According to the procedure adopted by the Nestedness Temperature Calculator (see Atmar and Patterson 1995 ), the matrix is first “packed” into a state of maximum nestedness, reordering rows and columns. By convention, the most species-rich country is placed along the top row, and the most widely distributed species is placed in the leftmost column, so as to concentrate presences in a corner of the matrix, and to minimize unexpected species absences and presences as in theoretical Figure 1 A. This will make differences in PID species distribution across countries readily perceivable. Moreover, not all unexpected species presences and absences are of equal informational value, and this must be taken into account. As we move away from the corner, where cells are most likely to be occupied, unexpected absences and presences begin to appear. The occurrence boundary lines (black exponential curves in Figure 2 B) are based on the distribution of unexpected species' presences and absences within the matrix. These curves determine the hypothetical boundary between the occupied area of the matrix and the unoccupied area. A color scale indicates the probability of a cell's occupancy. Nestedness Temperature Calculator also includes a Monte Carlo component to assess the statistical assurance that the parent data matrix was not randomly generated. To assess that probability, 1,000 randomized permutations were drawn to determine a baseline expectation. The result is a histogram representing the 1,000 “temperature” values obtained after permutations ( Figure 2 C). A black arrow indicates the “temperature” value observed with our master matrix. Lastly, the probability of obtaining this value by random is calculated.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC423130.xml
550653	SNP genotyping to screen for a common deletion in CHARGE Syndrome	Background CHARGE syndrome is a complex of birth defects including coloboma, choanal atresia, ear malformations and deafness, cardiac defects, and growth delay. We have previously hypothesized that CHARGE syndrome could be caused by unidentified genomic microdeletion, but no such deletion was detected using short tandem repeat (STR) markers spaced an average of 5 cM apart. Recently, microdeletion at 8q12 locus was reported in two patients with CHARGE, although point mutation in CHD7 on chromosome 8 was the underlying etiology in most of the affected patients. Methods We have extended our previous study by employing a much higher density of SNP markers (3258) with an average spacing of approximately 800 kb. These SNP markers are diallelic and, therefore, have much different properties for detection of deletions than STRs. Results A global error rate estimate was produced based on Mendelian inconsistency. One marker, rs431722 exceeded the expected frequency of inconsistencies, but no deletion could be demonstrated after retesting the 4 inconsistent pedigrees with local flanking markers or by FISH with the corresponding BAC clone. Expected deletion detection (EDD) was used to assess the coverage of specific intervals over the genome by deriving the probability of detecting a common loss of heterozygosity event over each genomic interval. This analysis estimated the fraction of unobserved deletions, taking into account the allele frequencies at the SNPs, the known marker spacing and sample size. Conclusions The results of our genotyping indicate that more than 35% of the genome is included in regions with very low probability of a deletion of at least 2 Mb.	Background CHARGE Association is characterized by ocular coloboma, cranial nerve abnormalities, common outflow tract heart defects, choanal atresia, cupped-shaped pinnae, Mondini dysplasia of the inner ear [ 1 , 2 ] and growth delay. The embryology and mechanisms of maldevelopment in CHARGE are not well understood. CHARGE Association may be genetically heterogeneous, a possibility supported by the rare and variable chromosomal aberrations observed in a few affected individuals. We have identified a de novo mutation in a semaphorin gene, SEMA3E , in an affected patient, identified upon mapping the translocation breakpoints in an unrelated individual with a de novo balanced translocation involving chromosomes 2 and 7: karyotype 46, XY, t(2;7)(p14;q21.11) [ 3 ]. Recently, Vissers et al . have reported mutations in CHD7 gene, a member of the chromodomain family in a substantial number of patients [ 4 ]. In a large number of children, however, the genetic mechanism of this complex birth defect remains unidentified. Graham has suggested that a subgroup of children with CHARGE Association have a recognizable syndrome [ 5 ] and may have a common pathologic and molecular basis. We have focused our study in this subset of patients and hypothesized that within this group, there is a common chromosomal region where recombination events lead to frequent microdeletion. In a previous study we used short tandem repeat (STR) markers spaced at an average of 5 cM to examine ten CHARGE case-parent trios for a large common deletion [ 6 ]. STR markers, because they are multi-allelic, generally are highly informative in each trio and there is little ambiguity regarding Mendelian transmission of parental alleles. The study did not identify any common deletion, but it was limited because of the relatively broad marker spacing. We have now extended the analysis by genotyping 3258 diallelic SNP markers with an average spacing of approximately 800 kb, in the same nuclear families. If microdeletion is the underlying basis of this disorder, then genotyping using this dense set of SNPs would be expected to uncover loci with loss of expected heterozygosity in the probands. We have taken into account the major and minor allele frequencies of each of the 3258 SNP markers and used a metric called the expected deletion detection, EDD [Belmont et al. , personal communication] to evaluate specific chromosomal intervals for the probability of detecting a deletion in the sample set. Although no deletion was detected in the CHARGE study sample, we conclude that this method could be generally useful in other studies in which small deletions occur as part of the allelic spectrum of disease. Methods Patients The diagnosis of CHARGE syndrome was established by examination by a participating clinical genetics specialist (CB, JWB, SRL). A medical history questionnaire was completed by the parents or by direct interview. For the core genotyping, 8 Caucasian and 1 Hispanic case/parent trios were selected based on the presence of 4 major criteria or three major and three minor criteria for the syndrome [ 7 ]. There were five affected males and four affected females, with ages ranging from five to twenty years. Blood was collected and transformed cell lines were established for these families. This research protocol was reviewed and approved by the Baylor College of Medicine Institutional Review Board. Genotyping Genotyping for this study was carried out on an Orchid BioSciences SNPstream UHT platform (Princeton, NJ) as previously described [ 8 ]. For this study, an initial set of ~4,200 T/C SNPs were identified and selected from the public databases for incorporation into a genome wide SNP panel. After selection, the complete set of SNPs was arranged into ~350 unique 12-plex reactions for the purpose of performing the assays on the UHT platform. The complete set of markers was then validated on a set of 5 CEPH pedigrees (40 individuals) and in 3 independent populations. A final set of 3258 markers was chosen from these combined datasets for analysis after eliminating SNPs that performed poorly through all populations, SNPs that failed both Hardy-Weinberg and Mendelian error calculations and any SNPs that were not polymorphic in the evaluated populations. An average genome wide spacing of ~800 kb between markers was achieved for this panel. This genotyping method utilized Orchid's single base primer extension chemistry (SBE) to identify which bases were present at the site of interrogation. Multiplexed reactions (12-plex) were performed in a single tube that incorporated labeled chain terminating nucleotides onto the ends of the SBE oligonucleotides. These reactions were then hybridized onto a microarray format, that facilitates the solid-phase sorting of the labeled extension-primers to a set of universal tagged primers arrayed on the surface of the plate. The universal tags were arranged on the surface of the microarray plate in a 384-well microwell layout. This microarray format created a generic design consisting of 384 4 × 4 arrays that contained 12 oligonucleotides that corresponded to 12 unique universal capture tags. The four additional oligonucleotides, plotted in each array, were used for positive and negative controls. Genotyping calls were determined by the presence or absence of incorporated dyes that appeared at each spot on the printed arrays. TaqMan polymerase chain reaction Two Assays-on-demand SNPs, rs422951 and ss1309424, flanking rs431722 were obtained from Applied Biosystems and genotyping was performed in 384 well-plates, using the TaqMan polymerase chain reaction-based method. The final volume reaction was 5 μl using 12 ng of genomic DNA, 2.5 μl of Taqman Master mix and 0.25 μl of 20X Assays-on-Demand SNP Genotyping Assay Mix. The plate was heated at 95° for 10 minutes, followed by forty cycles of denaturation at 92° for 15 seconds and annealing/extension at 60° for 1 minute. PCR plates were read on ABI PRISM 7900HT instrument with SDS v2.0 software. Individual genotypes that were ambiguous were excluded. Fish Bacterial Artificial chromosomes (BACs) were selected from the public database [ 9 ] and obtained from Children's Hospital Oakland Research Institute. DNA extraction was performed according to the standard protocol. Fluorescence in situ hybridization was performed as described elsewhere [ 6 ]. Detection of the digoxigenin labeled probe was performed with anti-digoxigenin conjugated to rhodamine, giving a red signal. Biotin labeled control probe was detected with FITC (fluorescein isothiocyanate), giving a green signal. The chromosomes were counterstained with DAPI and analyzed with a Zeiss Axioskop fluorescence microscope equipped with appropriate filter combinations. Approximately 10 metaphase preparations were scored for each hybridization. Data analysis The genotype error rate was estimated using the method of Gordon et al. [ 10 ] and as implemented in CUE [ 11 ]. Expected deletion detection (EDD) is a new method designed for this study, which uses the allele frequency, the marker spacing and the number of pedigrees sampled to estimate the probability that a common deletion would be missed because of ambiguous genotype outcomes. Qualitatively, the information available in a single SNP marker for the purposes of detecting a deletion by lack of expected heterozygosity in a case-parent trio is limited by the many genotype configurations that could appear consistent with Mendelian inheritance, but actually harbor a deletion. Inclusion of 2 or more SNP markers in a deletion interval decreases the likelihood that a common deletion goes undetected. Results SNP genotyping data were subjected to analysis for Hardy-Weinberg equilibrium. As an additional test of marker integrity, a transmission disequilibrium test was also performed for each of the SNP markers to examine for distortions in allele transmission in the trios. Unequal transmission of alleles from the heterozygote parents to the affected offspring was not determined by this analysis. Non-paternity was excluded in core pedigrees. Using the method of Gordon [ 12 ], we used the genotyping data to estimate the genotyping error rate at 0.02%. Analysis of the data showed transmission inconsistent with Mendelian inheritance for 22 markers on different chromosomes (20 with 1 inconsistency, 1 with 2 inconsistencies, and 1 with 4 inconsistencies). Given the underlying genotyping error rate, we could predict that markers with >2 inconsistencies would be highly unlikely to occur. One marker, rs431722 with overall call rate of 95%, showed Mendelian inconsistency in 4 trios. This SNP was found to lie within the intron 2 of the NOTCH4 gene on chromosome 6p21.32. Human DNA sequence from clone XXbac-300A18 (GenBank accession number AL662884) on chromosome 6p21 was used for FISH. This BAC clone encompasses the NOTCH4 gene, and was confirmed by PCR amplification of the clone sequence using NOTCH4 specific primers (data not shown). The analysis of the metaphase chromosomes after staining with DAPI showed two bright hybridization signals, indicating the presence of both alleles (Figure 1 ). Two additional SNPs, rs422951 and ss1309424, flanking rs431722 within 780 bp, were genotyped using TaqMan chemistry. The results showed inheritance of biparental alleles in all four pedigrees. The CHD7 locus on 8q12 was investigated additionally with FISH using RP11-33I11 (GenBank accession number AC113143) and RP11-414L17 (GenBank accession number AC023102). Using these BAC clones, microdeletion of this region was excluded in all the affected patients in this study sample. Because none of the pedigrees were consanguineous, we used parental data to estimate the allele frequencies for each of the 3258 SNP markers [ 13 ]. The EDD was then calculated for each chromosome. The percent coverage ranged from 51% on chromosome 20 to 20% on chromosome 15, with a mean of 36% for an autosomal deletion of 2 Mb (Figure 2 ). Discussion Despite various efforts to understand the molecular basis of CHARGE syndrome, with candidate genes sequencing [ 14 , 15 ], comparative genomic hybridization [ 16 ], and genome-wide scan for microdeletion(s) using microsatellite markers [ 6 ], the underlying molecular mechanism in many patients remained unknown until recently [ 4 ]. Based on the complex phenotype and clinical overlap with Velocardiofacial syndrome, it is a plausible hypothesis that in a subset of CHARGE patients with a homogenous phenotype, the underlying genetic mechanism is a cryptic submicroscopic deletion involving highly pleiotropic gene(s). To address this hypothesis, we had previously used microsatellite markers to ascertain loci with loss of expected heterozygosity in case-parent trios. SNPs are far more abundant than microsatellite markers but have not yet been used extensively in linkage and loss of heterozygosity (LOH) studies. The present study represents the application of SNPs to scan for potential submicroscopic deletions across the autosomes. Amos et al. have previously shown that SNP genotyping of child-parent trios provides valuable information about the presence of de novo microdeletion when sufficient families are studied [ 17 ]. However, this method is most appropriate when linkage disequilibrium is accounted for because of high SNP marker density. They have provided a general analytical framework and point out the effects of non-paternity, sample mix-up, and genotyping error in the interpretation of Mendelian inconsistency in case-parent trio data with biallelic markers. As expected, increases in rates of such phenomenon in the data decrease power to detect a microdeletion. In addition, they point out that heterogeneity in the position of a putative deletion also has a large impact on power. Their analysis is particularly apt given the probable future availability of extremely high density SNP marker maps and the technical capability to genotype hundreds of thousands of markers per research subject. However, they do not explore the effect of intermarker distance in the ability to detect deletions of various sizes. Assuming the genotyping error rate of 0.5%, the probability of observing >2 inconsistencies per marker is very low. The SNP marker rs431722 showed Mendelian inconsistency in four of nine pedigrees, with apparent loss of a parental allele in each case i.e. a frequency much higher than expected for the genotyping error rate. Interestingly, this SNP is located within the NOTCH4 gene on chromosome 6p21.32. The Notch gene family encodes highly conserved transmembrane receptors that are involved in intercellular signaling. The Notch signaling pathway plays an essential role in regulating embryonic vascular morphogenesis and remodeling [ 18 ]. Moreover, disruption of Notch signaling via mutation in the Notch ligand JAG1 is known to result in Alagille syndrome [ 19 ], nonsyndromic Tetralogy of Fallot (TOF) [ 20 ] and possibly nonsyndromic biliary extrahepatic atresia [ 21 ]. Since TOF is a heart defect commonly seen in CHARGE Syndrome, NOTCH4 was further studied with FISH as well as flanking SNP markers for microdeletion. The results, however conclusively excluded a discernible microdeletion at this locus. This screen is expected to detect deletions of about 1–2 Mb depending on the overlap of the SNP markers with the deletion interval. Variable coverage for each chromosome was determined for approximately 2 Mb microdeletion in this study. Almost 50% of chromosomes 7, 19 and 20 were excluded for any microdeletion greater than 2 Mb. The least coverage was observed for chromosome 15 and 18, with exclusion of 20% of the chromosome for the presence of a similar genetic aberration. Overall, we can estimate that approximately 36% of the genome had >80% chance for detecting a common 2 Mb deletion in at least 2 patients with CHARGE Syndrome. There are several limitations to this approach in studying the genetics of CHARGE syndrome. Although the marker density is high, the reduced amount of information per marker means that only some of the trios give the possibility of a conclusive result. Denser marker sets would be predicted to fill most of the gaps, but the regions around the centromeres are likely to be difficult with any currently available technique. The strategy out lined in this paper would work equally well for conventional Mendelian traits in which the mutant alleles included at least some deletions. Using a much denser SNP map of 1 marker every 5–10 kb, as is anticipated for whole genome association analyses, it would be possible to detect most deletions given a sufficient representation of deletions within the spectrum of gene mutations. Conclusions In this report we show that a SNP genotyping screen has excluded moderate length submicroscopic deletions in a subset of patients with CHARGE syndrome. Further analysis by microarray comparative genome hybridization methods or denser SNPs will allow a comprehensive assessment of the role, if any, of microdeletions in CHARGE syndrome. Competing interests The author(s) declare that they have no competing interests. Authors' contributions SRL carried out the data analysis, FISH and TaqMan assays and drafted the manuscript. AMS assisted in the data analysis. MP carried out the SNP genotyping, JWB conceived of the study, and participated in its design and coordination. He performed the statistical analysis and drafted the manuscript. SDF, CAB, and NLG participated in patient enrollment and LMM assisted in DNA extraction. WJC and JAT assisted in manuscript preparation. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC550653.xml
521082	Doublet method for very fast autocoding	Background Autocoding (or automatic concept indexing) occurs when a software program extracts terms contained within text and maps them to a standard list of concepts contained in a nomenclature. The purpose of autocoding is to provide a way of organizing large documents by the concepts represented in the text. Because textual data accumulates rapidly in biomedical institutions, the computational methods used to autocode text must be very fast. The purpose of this paper is to describe the doublet method, a new algorithm for very fast autocoding. Methods An autocoder was written that transforms plain-text into intercalated word doublets (e.g. "The ciliary body produces aqueous humor" becomes "The ciliary, ciliary body, body produces, produces aqueous, aqueous humor"). Each doublet is checked against an index of doublets extracted from a standard nomenclature. Matching doublets are assigned a numeric code specific for each doublet found in the nomenclature. Text doublets that do not match the index of doublets extracted from the nomenclature are not part of valid nomenclature terms. Runs of matching doublets from text are concatenated and matched against nomenclature terms (also represented as runs of doublets). Results The doublet autocoder was compared for speed and performance against a previously published phrase autocoder. Both autocoders are Perl scripts, and both autocoders used an identical text (a 170+ Megabyte collection of abstracts collected through a PubMed search) and the same nomenclature (neocl.xml, containing over 102,271 unique names of neoplasms). In side-by-side comparison on the same computer, the doublet method autocoder was 8.4 times faster than the phrase autocoder (211 seconds versus 1,776 seconds). The doublet method codes 0.8 Megabytes of text per second on a desktop computer with a 1.6 GHz processor. In addition, the doublet autocoder successfully matched terms that were missed by the phrase autocoder, while the phrase autocoder found no terms that were missed by the doublet autocoder. Conclusions The doublet method of autocoding is a novel algorithm for rapid text autocoding. The method will work with any nomenclature and will parse any ascii plain-text. An implementation of the algorithm in Perl is provided with this article. The algorithm, the Perl implementation, the neoplasm nomenclature, and Perl itself, are all open source materials.	Background Autocoding is a specialized form of machine translation. The general idea behind machine translation is that computers have the patience, stamina and speed to quickly parse through gigabytes of text, matching text terms with equivalent terms from an external vocabulary. Human translators often scoff at the output of machine translators, noting the high rate of comical errors. An often cited, perhaps apocryphal, example of poor machine translation is the English to Russian transformation of "out of sight, out of mind" to the Russian equivalent of "invisible idiot." Despite limitations, machine translation is the only way to transform gigabytes and terabytes of text. As long as clinicians, pathologists, radiologists, nurses, and scientists continue to type messages, reports, manuscripts and notes into electronic documents, we will need computers to parse and organize the resulting text. One of the many problems in the field of machine translation is that expressions (multi-word terms) convey ideas that transcend the meanings of the individual words in the expression. Consider the following sentence: "The ciliary body produces aqueous humor." The example sentence has unambiguous meaning to anatomists, but each word in the sentence can have many different meanings. "Ciliary" is a common medical word, and usually refers to the action of cilia. Cilia are found throughout the respiratory and GI tract and have an important role locomoting particulate matter. The word "body" almost always refers to the human body. The term "ciliary body" should (but does not) refer to the action of cilia that move human bodies from place to place. The word "aqueous" always refers to water. Humor relates to something being funny. The term "aqueous humor" should (but does not) relate to something that is funny by virtue of its use of water (as in squirting someone in the face with a trick flower). Actually, "ciliary body" and "aqueous humor" are each examples of medical doublets whose meanings are specific and contextually constant (i.e. always mean one thing). Furthermore, the meanings of the doublets cannot be reliably determined from the individual words that constitute the doublet, because the individual words have several different meanings. Basically, you either know the correct meaning of the doublet, or you don't. Any sentence can be examined by parsing it into an array of intercalated doublets: "The ciliary, ciliary body, body produces, produces aqueous, aqueous humor." The important concepts in the sentence are contained in two doublets (ciliary body and aqueous humor). A nomenclature containing these doublets would allow us to extract and index these two medical concepts. A nomenclature consisting of single words might miss the contextual meaning of the doublets. What if the term were larger than a doublet? Consider the tumor "orbital alveolar rhabdomyosarcoma." The individual words can be misleading. This orbital tumor is not from outer space, and the alveolar tumor is not from the lung. The 3-word term describes a sarcoma arising from the orbit of the eye that has a morphology characterized by tiny spaces of a size and shape as may occur in glands (alveoli). The term "orbital alveolar rhabdomyosarcoma" can be parsed as "orbital alveolar, alveolar rhabdomyosarcoma" Why is this any better than parsing the term into individual words, as in "orbital, alveolar, rhabdomyosarcoma"? The doublets, unlike the single words, are highly specific terms that are unlikely to occur in association with more than a few specific concepts. Very few medical terms are single words. In "The developmental lineage classification and taxonomy of neoplasms" there are 102,271 unique terms for neoplasms. All but 252 of these terms are multi-word terms [see Additional file 1 ] [see Additional file 2 ][ 1 ]. Of the 252 singletons, all but 34 are names of specific tumors ending in the suffix, "oma." "Oma" is short for "tumor." Single-word names of tumors ending in "oma" can be thought of as doublets with the first and second words fused together (i.e. osteoblastoma is "osteoblast" + "oma"). Some examples of "oma" terms are: acanthoma, adamantinoma, adenofibroma, adenomyoepithelioma, adenomyoma, adenosarcoma, ameloblastoma, etc.) In the entire taxonomy, there are only 34 singletons that do not end in "oma." These are, "acrochordon, carcinoid, cyst, dermoid, dip-nech, dipnech, erythroleukemia, fibroid, histiocytosis, leucaemia, leukaemia, leukemia, macroglobulinemia, mastocytosis, milia, milium, myelodysplasia, naevus, neuronevus, nevus, parapsoriasis, pre-leukaemia, pre-leukemia, precancer, preleukemia, premalignancy, preneoplasia, tylosis, verruca, verrucae, and wart". These singletons represent a mere 34/102,271 or 0.0003 of the neoplasm terminology. Medical autocoding can be considered a specialized form of machine translation. Medical autocoders transform text into an index of coded nomenclature terms (sometimes called a "concept index" or "concept signature"). Several innovative approaches to autocoding have used the higher information content of multiword terms (also called word n-grams) to match terms in text with terms in vocabularies or to enhance the content of vocabularies by identifying n-grams occurring in text that qualify as new nomenclature terms [ 2 - 4 ]. Unlike prior studies with n-grams, the method developed for this study does not use statistical inferencing or the information content of bi-grams to infer semantic meaning from natural language. The author has used the higher term specificity of doublets [bi-grams] to construct a simple and fast lexical parser. Lexical parsers are types of string-matching algorithms. In general, the overall speed of lexical parsers is determined by the speed with which the parser can prepare an array of all possible words and phrases contained in a block of text, coupled with the speed with which each of these phrases can be compared against all the terms in the nomenclature. The purpose of this paper is to describe a novel algorithm for autocoding based on finding runs of word-doublets that match a list of doublets extracted from a medical nomenclature. Using the same hardware and the same nomenclature, the speed and accuracy of the doublet method can be compared with another fast lexical parser. Implementation Nomenclature The nomenclature used is neocl.xml, previously described by the author and currently designated as "The developmental lineage classification and taxonomy of neoplasms"[ 1 ]. In the context of this manuscript, the purpose of the taxonomy is to provide a listing of all names of neoplasms, with synonyms grouped under a common code number. The current version of the neocl.xml file contains 102,271 unique names of neoplasms. In constructing the taxonomy, enormous effort was made to include every variant name for every known neoplasm of man. Variant names included different terms for the same concept and different ways of expressing an individual term (e.g. variations in word order). An example of the variety of synonyms encountered for a single tumor is shown for "adenocarcinoma of the colon." There are 44 synonyms listed in the taxonomy. These are: adenoca arising from colon, adenoca arising in colon, adenoca of colon, adenocarcinoma arising from colon, adenocarcinoma arising from large intestine, adenocarcinoma arising from the colon, adenocarcinoma arising from the large intestine, adenocarcinoma arising in colon, adenocarcinoma arising in large intestine, adenocarcinoma arising in the colon, adenocarcinoma arising in the large intestine, adenocarcinoma of colon, adenocarcinoma of large intestine, ca arising from colon, ca arising in colon, ca of colon, cancer arising from colon, cancer arising in colon, cancer of colon, carcinoma arising from colon, carcinoma arising in colon, carcinoma of colon, colon adenoca, colon adenocarcinoma, colon ca, colon cancer, colon cancers, colon carcinoma, colon carcinomas, colon with adenoca, colon with adenocarcinoma, colon with ca, colon with cancer, colon with carcinoma, colonic adenoca, colonic adenocarcinoma, colonic adenocarcinomas, colonic ca, colonic cancer, colonic cancers, colonic carcinoma, colonic carcinomas, large intestine adenocarcinoma, large intestine with adenocarcinoma. Input file The input file was created by a PubMed query on "pathology [ad] AND neoplasm [all]", at the U.S. government website [ 5 ]. The query gathered all abstracts from the pubmed database in which the term neoplasm occurs somewhere in the pubmed entry, and in which the affiliation of the author contains the word "pathology". The query yielded abstracts that are likely to contain names of neoplasms. The PubMed output file can serve as a good test for an autocoder that uses a neoplasm nomenclature. The PubMed search yielded 66,509 abstracts. All of the abstracts were downloaded into a single file from the PubMed site by setting the "Display" attribute to "Medline" and the "Send to" attribute to "file". This produced a 170,997,880 byte plain-text file. The file was given the filename tumor.txt, and this filename was used by the autocoders as a parsing input file. Although this file is not included with this manuscript, anyone in the world with internet access can obtain a near-identical file by repeating the same PubMed query. Doublet autocoder The doublet autocoder is supplied as an open source Perl script (doubcode.pl) with this manuscript [see Additional file 3 ]. Perl itself is an open source language that is bundled with the Unix and Linux operating systems. In the past decade, Perl has become very popular in the bioinformatics community. Perl interpreters are available at no cost and in versions suitable for virtually all operating systems. Perl can be obtained from the Comprehensive Perl Archives Network [ 6 ] or from Active State [ 7 ]. The algorithm used by the doublet autocoder is described: 1. Each phrase (term) in the nomenclature (neocl.xml) is converted into intercalated doublets, and each doublet is assigned a consecutive number. 2. Each nomenclature phrase is assigned the concatenated list of numbers that represent the ordered doublets composing the phrase. 3. Every text record (pubmed abstract in this case) is split into an array consisting of the consecutive words in the text record. 4. The text array is parsed as intercalated doublets. Intercalated doublets from the text that match doublets found anywhere in the nomenclature are assigned their numeric values (from the doublet index created for the nomenclature). Runs of consecutive doublets from the text that match doublets from the nomenclature are built into concatenated strings of doublet values. The occurrence of a text doublet that does not match any doublet in the nomenclature cannot possibly be part of a nomenclature term. Such text doublets serve as "stop" doublets between candidate runs of text doublets that match nomenclature doublets. 5. The runs of matching doublets are tested to see if they match any of the runs of doublets that compose nomenclature terms or if they contain any subsumed terms that match nomenclature terms. 6. The array of doublet runs extracted from the text that match nomenclature terms are cached in an external file. Phrase autocoder The phrase autocoding script, phrase.pl is included as supplementary file with this article [see Additional file 4 ]. The script was described in a prior publication [ 8 ]. It works by taking text records and parsing the text into all possible ordered phrases of all sizes up to a predetermined limit (5, in this case). The script then examines each resultant phrase to determine if it matches a term in the nomenclature. If there is a match, the term is added to the cache of matched terms. Results Speed The speed of both coders was compared using a 170+ Megabyte text. The doublet method coded the entire text in 211 seconds compared with 1,766 seconds required for the phrase method. Therefore the doublet method coded 8.4 times faster than the phrase method. The text coding rate for the doublet method is 0.8 Megabytes per second on a desktop computer with a modest 1.6 GHz Pentium processor. Both the phrase method and the doublet method coders used the same Programming Language, the same nomenclature, and the same data structure for the nomenclature (a simple associative array). Both coders grab sequentially occurring chunks of text, each chunk delimited by a double-newline delimiter (roughly the ascii equivalent of a paragraph delimiter) and place the grabbed text chunk into a temporary scalar variables. The difference between the two coders occurs when the chunks of text are parsed. The phrase coder takes the text and creates an array of every ordered combination of 1,2,3,4 and 5-word phrases contained in the record. This number of elements in the phrase array is about 5 times larger than the number of words in the record. The doublet coder parses the record into the set of all ordered doublet terms. This number of elements in the array of doublets is about the same as the number of words in the record. The phrase coder must try to match about 5 times as many terms as the doublet coder. Also, the phrase coder tries to match each phrase in the record array against the entire nomenclature. The doublet coder tries to match each doublet encountered in the input text against the collection of doublet terms extracted from the nomenclature. Because doublets of text, unlike individual words, tend to have unique meanings, only a small subset of the doublets encountered in the input text will match the set of doublets extracted from the nomenclature. Text doublets that do not match any doublets in the nomenclature are "skipped." Text doublets that match doublets from the nomenclature are concatenated to consecutive matching doublets until a non-matching doublet is encountered. The length of matching doublets sets the length of the candidate term. The algorithmic strength of the doublet method is that it eliminates the need to create and match [against a nomenclature] an array of all possible phrases of all possible lengths found in a textual record. Autocoding output The output of the doublet method and phrase method autocoders are provided as supplemental files with this manuscript [see Additional file 5 ] [see Additional file 6 ]. The analysis of the autocoding output was performed using another Perl script, doubcode.pl [see Additional file 7 ]. The analysis output file is also provided with this manuscript [see Additional file 8 ]. The two autocoder methods parsed a text exceeding 170 megabytes in length. The doublet method coded 4807 different [sometimes called unique] terms. The total number of matching term encountered (which includes replicate matches of terms within a record) by the doublet method is 467,391 terms. The phrase method coded 4557 different terms. There were 250 terms that the doublet method matched and the phrase method missed. There were zero terms that the phrase method matched and the doublet method missed. The improved performance of the doublet coder is accounted for by its ability to match terms of any length, while the phrase coder is limited to match terms that are no more than 5 words in length. Because the phrase coder parses chunks of input text into all allowed phrase size combinations, the performance of the phrase coder is slowed dramatically when the allowed phrase length is large. For instance, if the theoretical maximum phrase size were permitted for a text chunk 100 words in length, the phrase parser would need to create an array of a size equal to 100 × 100 or 10,000 phrases, matching each phrase against nomenclature. If the phrase size were restricted to 5 words, then the phrase parser would only need to create a phrase array of size 5 × 100 or 500. It is reasonable to restrict the allowed length of coded term phrases if it is known that the number of nomenclature terms with length exceeding the allowed limit is small. It is also reasonable to restrict the allowed length of coded term phrases is it is believed that terms of long length can be successfully represented by subsumed terms within the longer term. For example, "Refractory anemia with excess blasts in transformation" will be captured as a single term by the doublet method. The phrase method will miss the 7-word term but will code the subsumed terms, "refractory anemia" and "refractory anemia with excess blasts". In this specific instance, the doublet coder outperformed the phrase coder. However, few cancer terms actually exceed 5 words in length, and longer terms almost always contain modifying phrases (so-called term pre-coordinations) that add little to the fundamental meaning of the term. The following terms are examples of neoplasms extracted by the doublet method and missed by the phrase method. All of the listed terms exceed 5 words in length. "Acute myelogenous leukemia with minimal differentiation, acute myeloid leukemia with minimal differentiation, adenocarcinoma arising in the parotid gland, adenoid cystic carcinoma of submandibular gland, aids-related primary central nervous system lymphoma, atypical glandular cells of undetermined significance, atypical squamous cells of uncertain significance, carcinoid tumor arising from meckel diverticulum, extranodal marginal zone b-cell lymphoma of mucosa-associated lymphoid tissue, giant cell tumor of tendon sheath, giant cell tumour of tendon sheath, glandular malignant peripheral nerve sheath tumor, high-grade malignant peripheral nerve sheath tumor, hurthle cell carcinoma arising from thyroid, hyalinizing spindle cell tumor with giant rosettes, interdigitating dendritic cell sarcoma arising from the spleen, intermediate-grade b-cell non-hodgkin lymphoma, intraductal papillary mucinous neoplasm of pancreas, intraductal papillary-mucinous tumor with moderate dysplasia, kaposi sarcoma arising in the heart, large cell calcifying sertoli cell tumor, leiomyosarcoma arising from the pulmonary artery, low grade cervical glandular intraepithelial neoplasia, low-grade malignant peripheral nerve sheath tumor, lymphoma arising in the thyroid gland, malignant epithelioid peripheral nerve sheath tumor, malignant tumor of peripheral nervous system, malt lymphoma arising in the stomach, metanephric adenosarcoma in a young adult, monomorphic clear cell adenocarcinoma of salivary glands, ovarian serous borderline tumor with micropapillary and cribriform patterns, ovarian serous tumor of low malignant potential, paraganglioma arising in the cauda equina, perineurial malignant peripheral nerve sheath tumor, philadelphia chromosome negative chronic myeloid leukemia, phosphaturic mesenchymal tumor mixed connective tissue variant, pineal parenchymal tumor of intermediate differentiation, pleomorphic adenoma of minor salivary glands, polymorphous low-grade adenocarcinoma of minor salivary gland, polymorphous low-grade adenocarcinoma of minor salivary glands, polymorphous low-grade adenocarcinoma of salivary gland, poorly differentiated malignant peripheral nerve sheath tumor, porokeratotic eccrine ostial and dermal duct nevus, primary cutaneous anaplastic large cell lymphoma, primary gastric diffuse large b-cell lymphoma, primary signet-ring cell carcinoma of lung, primitive neuroectodermal tumor arising in the pancreas, pulmonary immunocytoma with massive crystal storing histiocytosis, rhabdoid transformation of tumor cells in meningioma, serous ovarian tumor of low malignant potential, sex cord tumor with annular tubules, signet ring cell adenocarcinoma of prostate, signet-ring cell mucin-producing adenocarcinoma of minor salivary glands, sinonasal desmoplastic small round cell tumor, smooth muscle tumor of uncertain malignant potential, smooth muscle tumor with an uncertain malignant potential, splenic lymphoma with circulating villous lymphocytes, squamous cell carcinoma of fallopian tube, submandibular gland carcinoma ex pleomorphic adenoma, tall cell variant of papillary carcinoma of thyroid, testicular large cell calcifying sertoli cell tumor, uterine tumor resembling ovarian sex cord tumor." The long terms found by the doublet coder and missed by the phrase coder represent a very small percentage of the different terms found by either coder in the long 170+ Megabyte text (250/4807 = 5%). A cursory review of these terms indicates that they represent rare lesions or rare variants of common lesions. Discussion Developers of medical autocoders seldom publish manuscripts. It is the author's perception, based on many years of activity in this field, that most autocoders are proprietary products produced for a very specific type of job. The author has never encountered any autocoder vendors who revealed the speed of their autocoders or shared any primary data that measures the performance of their autocoders. There really has never been any publication where one autocoder was compared against another. This is unfortunate because software developers may defer implementing brilliant ideas for autocoders, if they are uncertain whether better autocoders already exist. Even if competing software developers were to share performance data, there are no widely accepted standards for measuring the performance of medical autocoders. One of the purposes of this paper is to provide open access to algorithms, software, performance methodology and performance outcome data. Accuracy issues for lexical parsers The performance of autocoders is always a contentious issue. The reflexive approach to measuring autocoder performance usually involves selecting a small corpus of text and allowing a human expert in a language domain to carefully read through the lines of the text, creating a so-called gold standard against which the autocoder can be compared. The problems with this seemingly straightforward approach are many and have been explained by myself and others. The many limitations of "precision and recall" as measurements of indexing performance have been reviewed elsewhere [ 9 , 10 ]. It would be a useful exercise to re-examine what autocoding performance means to different people in contrast to what autocoding performance means within the context of lexical parsers [ 11 , 12 ]. First, every human coder is biased by their different perceptions of the knowledge domain. One coder may prefer "parsimonious" coding. In "parsimonious" coding, there is a "best" code that represents the ideas contained in a defined section of text. A review article on "liver cirrhosis" may contain many different terms, but the parsimonious coder may only preserve a single code for the entire article. Another coder, also a parsimonious coder, may not be so strict, but may want only the best term from among a group of subterms. So if "adenocarcinoma of endometrium" appears in text, she may want to preserve this term but omit the so-called atomic inclusive terms, "adenocarcinoma" and "endometrium." Another coder may want a complete listing of every matching term in a text, including terms that occur within larger terms. Still another informatician may want to include all ancestral terms for each term found in the text (i.e. terms not present in the original text but related to the textual concept). Finally, some coders seek to create "concept signatures" from text. A concept signature is the list of the concepts contained in a section of text. The relationship of one section of text to another section of text is determined by a quantitative representation of how closely their signatures match. Concept signatures are used to retrieve or organize related documents, not specific concepts [ 13 ]. Lexical coders are, in a sense, perfect autocoders because they don't use grammatic rules, they have no "exception" lists, and they never guess or interpret text. The only thing that lexical parsers do is to parse text, examining strings to see if they contain exact matches to terms from a nomenclature. When well-constructed, they do not make mistakes. If they miss a valid term that is present in the text, it is because the term was missing from the nomenclature. Developers of lexical autocoders place enormous demands on the curators of nomenclatures. The most common cause of missed terms arises when the text contains a modifying word that breaks the term into a phrase that no longer matches anything found in the nomenclature. For instance, "adenocarcinoma of the left lung," would be missed by a lexical parser if the nearest term in its terminology list is "adenocarcinoma of the lung." A rule-based parser or a semantic parser may have successfully teased out the "left" from the term and found the match. In this instance, the nomenclature failed the lexical parser because it did not list terms that indicated tumor laterality. Lexical parsers do not match terms that are absent from the text (i.e. no false positives). False-positive terms are possible in rule-based and semantic parsers if they create word patterns not present in the original text. Also, since lexical parsers strictly match phrases in text with phrases in nomenclatures, it is possible to achieve accurate results of dubious value. For instance, a lexical parser would parse "adenocarcinoma of the lung is not seen" and find a match against the neoplasm term, "adenocarcinoma of the lung." The concept is present in the text, even if though it is absent from the patient. It is a misleading but "true" positive. The doublet autocoder creates a string of words from a chunk of text. In the case of the example corpus (PubMed abstracts), all of the sentences from the abstract are squeezed into a single string of words, obliterating sentence boundaries. This means that if a sequence of words crossing a sentence boundary happens to match a term in the reference nomenclature, the coder will register a "pseudo-positive" term. Empirical evidence suggests that this theoretical error in the doublet autocoder simply does not occur. A 4 megabyte collection of abstracts chunked as whole abstracts or as chunked sentences (delimited by period-space-space) contained no instances of pseudo-terms created by the obliteration of sentence boundaries in the whole-abstract text chunks. For fastidious developers who wish to ensure that their parsers respect sentence boundaries, it is possible to pre-process text into sentences with a sentence parser [ 14 , 15 ]. Lexical parsers can be modified to provide an output that preserves the intended sense of the term as used in its context. For example, if the text includes the sentence, "Adenocarcinoma of the lung is not present." The "sensible" lexical parser may be modified to tag the "adenocarcinoma of the lung" with a negation modifier, preserving the intended sense of the term. The author has previously published an "in place" method of inserting codes directly into sentences, preserving modifier terms (including negations) [ 16 ]. Comparing parsers The two lexical parsers used the same programming language (Perl), the same nomenclature, the same data structure to hold the nomenclature, the same text corpus, and the same method for breaking the complete text into text chunks. Differences between the performance of the two autocoding methods were accounted for entirely by the algorithmic processes by which chunks of text are parsed into phrases that are matched against the common nomenclature. It is easy to compare speeds. The doublet parser required 211 seconds to parse the 170+ Megabyte file, while the phrase parser required 1,776 seconds. Three algorithmic properties contribute to the speed advantage of the doublet coder over the phrase coder: 1) The doublet coder parses text into a number of phrases that is 5-fold smaller than the number of phrases produced by the phrase coder; 2) Doublets that are not found in any of the nomenclature terms can be quickly excluded, 3) Successive text doublets that match doublets from the nomenclature can be quickly concatenated and tested for matches in the nomenclature. These are the algorithmic differences between the phrase coder and the doublet coder and presumably account for the improved speed of the doublet method. Comparing coding accuracy is somewhat trickier. As discussed, lexical parsers have no "false positive" matches. Lexical parsers, unlike natural language processors, do not construct context-based new word phrases from native text. Lexical parsers only match exact phrases found in text with exact nomenclature terms. Therefore, neither the doublet coder nor the phrase coder have false-positive term matches. Coding output was compared by examining the different terms matched by the two coders. The number of different terms found by the doublet method was 4,807, compared with 4,557 terms found by the phrase method. There were no terms found by the phrase method and missed by the doublet method. There were 250 terms missed by the phrase method and found by the doublet method (about 5% of the number of different terms found by the doublet method). For the most part, these terms represented rare neoplasms with pre-coordinated modifiers. The reason that such rare terms were encountered is due to the large size of the text (exceeding 170 Megabytes). When the doublet and phrase methods are judged by their ability to extract nomenclature from text, their output would be similar for small text sizes or when the primary purpose of the coding is to extract and count commonly occurring terms. Conclusions The doublet method is a novel approach to autocoding. It can autocode 0.8 Megabytes of text per second on desktop computer using a modest 1.6 GHz processor. For this manuscript, the doublet autocoder was tested on a corpus of medical text that exceeded 170 Megabytes in length and used a publicly available nomenclature of neoplasia containing 102,271 unique terms. In a side-by-side comparison with a publicly available fast lexical autocoder, the doublet method was 8.4-fold faster. The doublet algorithm is generalizable to any type or length of plain-text using any nomenclature. The algorithm and the Perl implementation script are available as open source documents. Availability and requirements The autocoding scripts are short programs written in Perl. Perl is a freely available open source programming language. Perl interpreters for virtually any operating system are available from several sites on the web [ 6 , 7 ]. These sites have links to rich sources of online information on the Perl language. The autocoding script will execute on any operating system hosting a Perl interpreter. It requires an external plain-text file (to be autocoded) and a nomenclature. With minor modification, the Perl scripts will use any parsable nomenclature that contains listed plain-text terms associated with alphnumeric concept codes. The Perl scripts (doublet.pl and phrase.pl) will work without modification using the current version of "The developmental lineage classification and taxonomy of neoplasms," which is provided with this manuscript [see Additional file 1 ] [see Additional file 2 ]. Competing interests None declared. Authors' contribution The work expresses the opinion of the author and does not represent policy of the U.S. government. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Neocl.xml is the developmental lineage classification and taxonomy of neoplasms, in XML format. Because neocl.xml exceeds 7 Megabytes when uncompressed, a gzipped version of the file is provided (neoclxml.gz). After downloading from the biomedcentral site, the filename should be provided with a .gz suffix (if absent from the filename as downloaded). After decompressing the file, the file shoud be renamed "neocl.xml". The file can be viewed on current web browsers, but experience has shown that many browsers lack sufficient memory to display the entire file. Otherwise, the file can be viewed on a wordprocessor or an ascii editor. Click here for file Additional File 2 Neoself.gz is the ascii flat-file version of the neocl.xml file. The file exceeds 16 Megabytes when expanded. Each line-record of the file consists of the numbered name of a different term, its code value, and the hierarchical list of its ancestral [class] terms from the developmental lineage classification and taxonomy of neoplasms. Click here for file Additional File 3 Doubcode.pl is a Perl script that implements the doublet method for text autocoding and expects an external file named tumor.txt as the source corpus and a file named neocl.xml as the reference nomenclature. Click here for file Additional File 4 Phrase.pl is a Perl script that implements the doublet method for text autocoding and expects an external file named tumor.txt as the source corpus and a file named neocl.xml as the reference nomenclature. Click here for file Additional File 5 Doub.out is the plain-text output of the doublet method for text autocoding (i.e. output of doubcode.pl). Click here for file Additional File 6 Quickcan.txt is the plain-text output file for phrase method for text autocoding (i.e. output of phrase.pl). Click here for file Additional File 7 Doubcomp.pl is the Perl script comparing the output files produced by doubcode.pl and phrase.pl. Click here for file Additional File 8 Doubcomp.txt is the plain-text output of doubcomp.pl. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521082.xml
538289	Reconstruction of the pelvic ring using an autologous free non-vascularized fibula graft in a patient with benign fibrous histiocytoma	Background Benign fibrous histiocytomas (BFH) usually presents as a small benign lesion that predominantly occurs in the skin. Only few cases of BFH arising from bone have been reported, its occurrence in pelvic bones is even rarer. Case presentation A 34-year-old female presented with BFH at a rare anatomical location on both sides of the os ilium which was larger than earlier reported BFH of the bone. Surgical resection was performed successfully including resection of the inner pelvic ring and reconstruction of the linea terminalis using a non-vascularized fibular autograft. At 18 months after tumor resection and reconstruction of the pelvic ring, with interposition of a free vascularized fibula graft patient has an excellent clinical oncological and functional outcome. Conclusion Non vascularized fibular autograft is a useful reconstructive procedure in select patients.	Background Benign fibrous histiocytoma (BFH) is a tumor that occurs predominantly in the skin (also called dermatofibroma) and most commonly in younger individuals. The tumor typically presents as a painless nodule varying in size from a few millimeters to several centimeters. BFH of the bone has been a subject of increasing interest within the past few years. The term was initially introduced by Dominok in 1980 to describe a cystic lesion in the femur of a 66-year-old man [ 1 ]. Only a few cases of BFH of bone have been described in the literature since then, and even fewer located in the pelvis [ 2 - 6 ]. Histologically, BFH arising from soft tissues cannot be distinguished from those arising from bone [ 7 ]. Pain is most often the predominant presenting symptom [ 7 ], which helps to distinguish this tumor clinically from other fibrous lesions, such as non-ossifying fibroma. Peak incidence of BFH arising from bone is reported in the third decade [ 5 ]. On radiography the lesion is characterized by osteolytic lesions with a well defined sclerotic margin [ 6 ]. Magnetic resonance (MR) Imaging of BFH usually shows low signal intensity on T1- and high signal intensity on T2-weighted images. Peripheral contrast enhancement has also been described in BFH [ 4 , 8 , 9 ]. Computed tomographic (CT) scan typically show fibrous osteolytic lesions with cortical thinning [ 10 ]. Positive bone scans may be helpful in differentiate BFH from nonossifying fibroma [ 4 ]. Biopsy is mandatory to confirm the diagnosis: Histologically BFH represents a benign but diverse group of neoplasms that are characterized by both fibroblastic and histiocytic differentiation. Giant cells of the "Touton" type and a "storiform pattern" are typically seen [ 11 ]. BFH shows proliferation of benign oval or spindle cells resembling fibroblasts mingled with cells resembling histiocytes. Foam cells are a prominent component of the macroscopically yellow zones of the tumor. It may be difficult to differentiate BFH from low-grade malignant fibrous histiocytoma. Lack of marked pleomorphism or of atypical mitoses is suggestive of a benign diagnosis. On microscopic examination some forms of BFH show features of dysplasia, such as hypercellularity, mitotic activity or focal necrosis [ 2 ]. These entities are considered to be more aggressive types of BFH, that should be treated with wide surgical excision [ 2 ]; for non-aggressive types of BFH even intralesional resection can be effective [ 7 ]. A local recurrence rate of 5 to 25 % is reported in literature. The recurrence rate is typically related to the size of the tumor [ 2 , 12 ]. Consequently, careful clinical and radiological follow-up including regular MRI is recommended [ 2 ]. We present here a case of BFH treated with resection and reconstructed using non vascularized fibular bone graft. Case presentation A 34-year-old female presented with a 3 year history of a palpable swelling and mild pain in the left gluteal region. There was no history of preceding trauma or accident. Two years after initiation of symptoms without clinical progress, the patient became pregnant and noted a local growth of the mass. After giving birth to a healthy child, the patient consulted an orthopedic surgeon. A radiographic image of the pelvis revealed a large osteolytic lesion in the left iliac wing, sharply contoured with sclerotic margins, indicating a slow growth pattern (Figure 1 ). MR-imaging of the pelvis showed a mass originating from the os ilium with a large extra- and intrapelvic soft-tissue component (Figure 2 ). Open biopsy was performed that lead to the diagnosis of BFH. Figure 1 Radiograph of the pelvis showing a well-circumscribed osteolytic lesion in the left iliac bone; it is sharply marginated with a thin sclerotic rim and without any matrix calcifications. Figure 2 a-d: MRI shows a mass originating from the ileum and extending into the soft tissues both anteromedially and posterolaterally. In T 1 -weighted sequences the mass is isointense with muscle (a). It enhances after administration of gadolinium (b, c). In fat-suppressed T 2 -weighted sequences it has high signal intensity (d). After reference to the literature on BFH, surgical en bloc resection was considered to be the appropriate treatment modality [ 2 , 3 , 12 , 13 ]. In accordance to the imaging results, the iliac crest could be preserved, while the linea terminalis of the left side had to be resected (Figure 3 ). Figure 4 shows the resected tumor specimen. Histological study of the resected specimen confirmed the diagnosis of BFH (Figure 5 ). The surgical margins were found to be free of tumor confirming complete tumor resection. Figure 3 Defect after tumor resection, with disruption of the linea terminalis. Figure 4 Macroscopic view of the resection specimen in transverse orientation. Large yellow osteodestructive tumor originating from iliac bone and with extensive extraosseous parts. Invasion of massa lateralis of sacral bone. Sharp borders between tumor and bone are indicative of a slow-growing neoplasm. Figure 5 Histology of the resection specimen. Spindle tumor cells admixed with some multinucleated giant cells (left) and foamy macrophages (middle) and cholesterol clefts (right). (Hematoxyllin and Eosin x) The pelvic ring was then stabilized in a second operation using a non-vascularized fibular autograft that was impacted and stabilized with screws and a Kirschner-wire reconstructing the linea terminalis of the left pelvic ring (Figure 6 ). Post operatively she developed neurapraxia of the left lateral femoral cutaneous nerve, which resolved spontaneously within 6 months after surgery. Following two weeks of bed rest and limited restriction of movement up to 70° flexion of the hip joint, the patient was mobilized with partial weight bearing for three months. Figure 6 Status after interposition of the fibula transplant Radiographic and MRI controls at 3, 6 and 12 months postoperatively showed no evidence of local recurrence or secondary dislocation of the reconstruction (Figure 7 ). A diastasis of the symphysis was noted, which we ascribe to the pregnancy one year prior to surgery. After 18 months of surgery the patient is free of disease and ambulates with full weight bearing. Figure 7 Radiographic check 18 months after surgery, showing identical position of the fibular transplant Discussion A rare anatomical location of BFH accompanied by an extraordinary tumor size of 9.5 × 7 × 11 cm is presented in this report. To our knowledge, no BFH of the bone with a similar size has been published. Due to its size and location on both sides of the iliac wing, including a large osteolytic lesion of the os ilium , this tumor presented a challenge for surgical resection. There are only a few studies addressing the stability of the pelvic ring following partial hemipelvectomy [ 14 - 17 ]. A variety of problems, such as persistent instability and complications caused by postoperative infections and transplant dislocation, have been reported [ 14 - 17 ]. The majority of the reconstructions reported have followed extensive resection of malignant bone tumors. Allograft [ 18 ], non-vascularized and vascularized auto grafts (fibula, tibia, and femur) [ 15 , 19 , 20 ], endoprosthetic replacement and other osteosynthetic procedures (i.e. with Kuntscher rods and K-wires or transpedicular and iliac screw systems) [ 14 ] have been used with differing degrees of success and variable clinical outcome. After total sacrectomy adequate reconstruction was reported in one study to have been achieved with transpedicular and iliac screw systems [21]. However K-wires were found to be insufficient for reconstruction of the ilium [ 14 ]. Within recent years, the use of biological transplants for such reconstructions has been reported in a number of cases with functionally satisfactory results [ 18 , 15 , 20 ]. In our case, we decided to reconstruct the inferior pelvic ring with an impacted non-vascularized fibular autograft to provide stability to the pelvic ring. No data is available on the necessity for reconstructing the linea terminalis when the continuity of the pelvic ring remains intact (a bony bridge between sacrum and ilium was left intact in our case). We are of the opinion that stabilization of the inner pelvic ring is necessary to prevent stress fracture of the remaining bone bridge. After mobilization of the patient an increasing diastasis of the symphysis was noted (Figure 7 ). This might have been caused by moderate shortening of the implanted fibular graft, which can in turn result in rotation and tilting of the affected hemipelvis. In our case loosening of the symphysis following pregnancy and vaginal delivery prior to surgery too might have contributed to diastasis in our case. Oncologically the marginal tumor resection has proved to be an adequate treatment. Despite high recurrence rates, especially in the case of large BFH [ 2 ], no local recurrence has so far been observed in our case. With a follow-up of 18 months after tumor resection as well as reconstruction of the pelvic ring it seems to become a successfully performed treatment for this patient. Competing interests The author(s) declare that they have no competing interests. Authors' contributions PN prepared the manuscript and participated in the operations. KL carried out the radiographies and MR imaging and contributed the section on imaging MW performed histological analysis and wrote pathology part of manuscript LB and DP critically reviewed the manuscript for its scientific content and performed surgery.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538289.xml
340960	When Monkeys Learn Directional Tasks, Neurons Learn Too	null	If you've ever hit a patch of ice on the road that sent your car swerving left while you resolutely—and futilely—steered right to get back in your lane, you've experienced what neuroscientists call a “visuomotor rotation task.” On a dry road, your response would have been appropriate. But under icy conditions, the same sensory cue produces a decidedly negative result: a car fishtailing out of control. While you're figuring out what movements will straighten out the car, the neurons in your primary motor cortex—the region of the brain responsible for movement—are taking notes. Chances are, your next icy encounter was less dramatic. But how does your brain learn to produce a different movement in response to the same visual cue? Neuroscientists investigate such questions by recording and analyzing the electrical activity of neurons during learning and performance of new sensory-motor transformations. Such studies, for example, show that populations of neurons in different brain areas map sensory cues and desired arm motion by creating an internal representation of the corresponding sensory and motor coordinates in a way that allows flexible responses to changing conditions. In previous studies, Rony Paz and Eilon Vaadia, of The Hebrew University in Israel, found that neurons in the primary motor cortex that fire before monkeys move their arm in a particular direction have higher firing rates after the monkey learns to dissociate the arm direction from the cursor direction (an indicator of visual feedback). Interestingly, changes in activity preferentially occurred in a subset of neurons that were already tuned (that is, maximally activated during movement) to the direction experienced while learning. While many studies indicate that learning new tasks can generate specific changes in brain activity, it had not been clear how or if such changes improve the internal representation inside the brain. Specifically, is the neuronal code any “better” after learning? Now Paz and Vaadia show that while these neurons are firing at higher rates they are also transmitting more information about specific task parameters. Paz and Vaadia trained two rhesus monkeys to learn various visual-motor tasks—which involved operating a joystick to move a cursor on a screen—and then changed the relationship between the visual feedback (the cursor) and hand movement. Using information-theory analysis—which measures the amount of information that single neurons can tell about the movement—they were able to correlate neuron activity with direction of movement and, conversely, distinguish differences between directions based on neuron activity. Their analysis revealed that the neurons transmit more information about the direction of movement after the monkeys learn a task. To figure out what aspect of neuron activity conveys this improved information, Paz and Vaadia examined two features of neuron signaling—response variability and directional sensitivity—which they reasoned might plausibly accomplish this. Increased information content after learning a task, they found, corresponded to sensitivity to a single direction, and neurons attuned to that direction contributed to the increase. These findings suggest that subsets of directionally sensitive neurons increase their firing rates to more finely tune their sensitivity to that direction. By successfully reconstructing the movement direction from the neuron signals captured after learning a task, Paz and Vaadia also demonstrate that the observed learning improvement can be extracted to predict behavior. The authors argue that their results suggest a close association between properties of neurons—such as directional tuning of cells—and learning a skill that is focused on the same parameter—in this case, direction. Together with results from visual and auditory areas, they propose that similar mechanisms may control the interplay between neurons and learning throughout the central nervous system. Mutual information between neuronal activity and direction of movement	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC340960.xml
538262	Diversity and specificity in the interaction between Caenorhabditis elegans and the pathogen Serratia marcescens	Background Co-evolutionary arms races between parasites and hosts are considered to be of immense importance in the evolution of living organisms, potentially leading to highly dynamic life-history changes. The outcome of such arms races is in many cases thought to be determined by frequency dependent selection, which relies on genetic variation in host susceptibility and parasite virulence, and also genotype-specific interactions between host and parasite. Empirical evidence for these two prerequisites is scarce, however, especially for invertebrate hosts. We addressed this topic by analysing the interaction between natural isolates of the soil nematode Caenorhabditis elegans and the pathogenic soil bacterium Serratia marcescens . Results Our analysis reveals the presence of i) significant variation in host susceptibility, ii) significant variation in pathogen virulence, and iii) significant strain- and genotype-specific interactions between the two species. Conclusions The results obtained support the previous notion that highly specific interactions between parasites and animal hosts are generally widespread. At least for C. elegans , the high specificity is observed among isolates from the same population, such that it may provide a basis for and/or represent the outcome of co-evolutionary adaptations under natural conditions. Since both C. elegans and S. marcescens permit comprehensive molecular analyses, these two species provide a promising model system for inference of the molecular basis of such highly specific interactions, which are as yet unexplored in invertebrate hosts.	Background By definition, parasites have a negative effect on host fitness. Since parasites usually show a shorter generation time than their hosts, they are also able to adapt rapidly to newly arising host genotypes. Both characteristics together select for hosts with efficient counter-adaptations. Subsequently, parasites are favoured if they can circumvent these host countermeasures. Such interactions may result in a co-evolutionary arms race, consisting of repeated cycles of the emergence of new parasite offences and host countermeasures. Hence, parasite-host interactions can lead to extremely rapid evolutionary change [ 1 , 2 ]. As such, they are thought to be responsible for much of the complexity found in the immune system of animals [ 3 ]. They are also likely to account for the evolution of diverse genetic mechanisms, which aid in generating fast changes, including sexual reproduction and recombination [ 4 , 5 ]. They may also affect the evolution of other life-history traits, such as reproductive rate, longevity, or competitive ability, which compete for available resources with defence and virulence traits in host and parasite, respectively [ 6 , 7 ]. Co-evolutionary arms races between hosts and parasites (meaning here eukaryotic organisms, bacterial pathogens, and viruses that gain a fitness advantage from infecting and harming a host) are in many cases assumed to be determined by negative frequency dependent selection. In particular, rare parasite and host genotypes should be at an advantage because commonness facilitates evolution of host or parasite counter-adaptations, respectively [ 4 , 5 ]. Such frequency dependent dynamics rely on two important conditions: i) natural genetic variation in both host resistance and parasite virulence, and ii) natural genotype-specific interactions between hosts and parasites [ 1 , 4 ]. Empirical evidence for the presence of both of these prerequisites is still rare, especially for invertebrate hosts [ 8 , 9 ]. They include various associations between snails and trematodes (e.g., Potamopyrgus antipodarum versus Microphallus [ 10 ] or Bulinus globosus versus Schistosoma [ 11 ]), the association between the waterflea Daphnia magna and its microparasite Pasteuria ramosa [ 8 ], between Drosophila melanogaster and its parasitoid Asobara tabida [ 12 ], between the bumble bee Bombus terrestris and the trypanosome Crithidia bombi [ 13 ], or between the copepod Macrocyclops albidus and the cestode Schistocephalus solidus [ 14 ]. Clearly, more data is needed to determine the importance of parasite-mediated co-evolutionary arms races in nature. In this study, we evaluated differences in host resistance and parasite virulence, both defined in a broad sense and reciprocally as the effect of an infection on host condition (i.e. alive, morbid, or dead). In particular, we tested the presence of genetic variation and also the presence of strain- and genotype-specific interactions during the infection of the nematode Caenorhabditis elegans (Nematoda: Rhabditidae) by the Gram-negative bacterium Serratia marcescens (Enterobacteriaceae). C. elegans has recently been established as a model to study parasite-host interactions and in particular the genetics of host defence [ 15 - 17 ]. It is a soil inhabitant found in almost all temperate regions of the world. It seems to be common in decomposing material, where it feeds on diverse microorganisms [ 18 , 19 ]. About 50 natural strains are currently available. These strains are genetically very diverse, even when isolated from populations at a single location [ 18 , 19 ]. They also differ in many life-history traits, including their response towards the potential parasite Bacillus thuringiensis [ 20 ]. The parasites that C. elegans encounters under natural conditions have not yet been unambiguously identified. The ubiquitous soil-dwelling bacteria Pseudomonas aeruginosa , B. thuringiensis and S. marcescens are all likely candidates [ 17 ]. For our study, we chose S. marcescens , recently adopted as a model to study the genetic basis of virulence [ 15 , 21 ], as a pathogen as it is able to produce a persistent infection and it is likely to benefit from the infection [ 22 ], thus behaving as a true parasite of the nematode. Results In our main experiment, we compared the consequences of infection of eight different natural C. elegans strains with 5 different S. marcescens strains plus one control (heat-killed bacteria of S. marcescens Db11, a strain for which the genome sequence is now complete). The C. elegans strains were isolated from Münster, in Northwest Germany, and belong to four different microsatellite genotypes [ 18 ]. The S. marcescens strains originate from different locations around the world. The interaction between the two species was examined with the help of a survival assay, in which the survival of individual worms was monitored in the presence of a defined concentration of bacteria [ 20 ]. The survival assay was performed in 96-well plates on five occasions (runs). During each run, all possible bacterial and worm strain combinations were assayed in parallel, resulting in a total of 16 data points per factor combination per run and 80 data points per factor combination in total. 80 out of a total 3840 cases (2.08%) had to be excluded because of errors during automated worm-transfer (either no worm or more than one worm per well), resulting in between 75 and 80 usable data points per combination of worm and bacterial strains. The number of valid cases did not differ significantly among these factor combinations (likelihood ratio test [LRT], χ 2 = 0.904, d.f. = 35, P > 0.999). In the control (worms with dead S. marcescens Db11), only 12 out of 625 were not found in the category "alive" (1.92%). Of these, 9 were morbid and 3 were dead. The recorded number of live worms per strain did not differ significantly from 100% (LRT, χ 2 = 0.455, d.f. = 7, P > 0.999). It also did not differ significantly among the worm strains (LRT, χ 2 = 0.416, d.f. = 7, P > 0.999). These results show that the experimental set-up itself does not cause significant levels of dead or morbid worms and that it does not have a different effect on different worm strains. The different C. elegans strains show substantial differences as to their ability to survive in the presence of pathogenic S. marcescens (Fig. 1 ). In general, the strains MY6 and MY18 were most resistant, whereas MY14 and MY15 were most susceptible. Moreover, the strains with identical microsatellite genotypes generally produce similar but not identical levels of resistance. This suggests that these strains bear additional genetic differences, which were not resolved by microsatellite genotyping. At the same time, the different S. marcescens strains differ considerably in their effect on C. elegans (Fig. 1 ). Here, strain Sm2170 was most virulent, whereas strains Sma3 and Sma13 generally produced the fewest cases of mortality and morbidity. Since S. marcescens strains were grown under identical conditions and since some of them are already known to differ in phenotype (e.g. red pigmentation), the observed differences are most likely determined genetically. Most interestingly, the interaction between specific worm and bacterial strains seems to differ across the table. For instance, C. elegans strain MY10 is more susceptible to S. marcescens strain Sma13 than to ATCC274, whereas the opposite is true for C. elegans strain MY20 (Fig. 1 ). Similarly, host strain MY15 is more susceptible to pathogen strain ATCC274 than to strain Db11, whereas the pattern is reversed for almost all other host strains (Fig. 1 ). Figure 1 Treatment response for the different bacterial and worm strain combinations of the main experiment. The response is expressed as host condition (values for the whole experiment), such that the black area refers to the proportion of dead worms, grey to the proportion of morbid, and white to the proportion of live worms. For C. elegans , both strain (bottom line) and genotype (top line) designations are given. For S. marcescens , only strain names are listed. In general consistency with these observations, ordinal logistic regression (OLR) analysis indicates a significant effect of the factors bacterial strain, worm strain or genotype, the interaction between the two, and also experimental run on the treatment response (Table 1 ). The two respective models employed are significantly better than models without any predictors (model including worm strain as factor: LRT, χ 2 = 1285.63, d.f. = 199, P < 0.0001; model including worm microsatellite genotype as factor: LRT, χ 2 = 854.55, d.f. = 99, P < 0.0001). However, they are both significantly worse than the respective saturated models (model including worm strain: LRT, χ 2 = 491.32, d.f. = 199, P < 0.0001; model including worm genotype: LRT, χ 2 = 442.94, d.f. = 99, P < 0.0001). The latter test examines whether the model employed considers a sufficient number of factors or factor combinations to explain the variation found in the data. The results suggest that the model is not sufficiently complex. We decided against employing more complex models (e.g. consideration of host genotype nested in host strain in a single model), because the response variable is ordinal with only three categories (alive, morbid, dead), such that a larger number of predictor variables in the model would most likely lead to highly increased random error in the regression analysis. Thus, as an alternative, we analysed the data using association tests. Table 1 Ordinal logistic regression analysis of the importance of different factors in the main experiment. Source χ 2 d.f. P Consideration of worm strains as a factor Bacteria 272.78 4 < 0.0001 Worm 188.11 7 < 0.0001 BacteriaWorm 127.15 28 < 0.0001 Run [Bacteria, Worm] 835.27 160 < 0.0001 Consideration of worm genotypes as a factor Bacteria 193.77 4 < 0.0001 Worm 169.87 3 < 0.0001 BacteriaWorm 34.21 12 0.0006 Run [Bacteria, Worm] 477.14 80 < 0.0001 Ordinal logistic regression was based on a model, which contained bacterial strain, worm strain (alternatively worm genotype), the interaction between the two and run nested within both bacterial strain and worm strain/genotype as factors. The importance of different factors was assessed with the likelihood ratio test. Significant probabilities after Dunn-Sidák correction are given in bold. Two-way associations were analysed with the LRT. The results show a significant effect of either of the different factors on worm condition (Table 2 ). The relevance of these associations was further examined by taking into account a second predictor variable using the Cochran-Mantel-Haenszel (CMH) test of conditional independence. All previously identified associations remained significant, irrespective of the second predictor variable considered (Table 2 ). The only exception refers to the case where the factor worm strain was corrected by the factor worm genotype, suggesting that the observed variation among C. elegans strains is due to differences in genotypic composition. The remaining results indicate that the significant effect from one of the factors on the treatment response is independent of the significant effect from one of the other factors. This finding is consistent with the presence of an interaction effect from the factors bacterial strain and nematode strain/genotype, as above suggested by OLR. Table 2 Association analysis of the impact of different factors on worm condition in the main experiment . Factor Test χ 2 d.f. P Single factor effects Bacteria LRT 291.05 8 < 0.0001 Worm strain LRT 154.84 14 < 0.0001 Worm genotype LRT 136.29 6 < 0.0001 Run LRT 186.33 8 < 0.0001 Factor effects in consideration of one of the others (in brackets) Bacteria (Worm strain) CMH 196.74 4 < 0.0001 Bacteria (Worm genotype) CMH 196.44 4 < 0.0001 Bacteria (Run) CMH 192.08 4 < 0.0001 Worm strain (Bacteria) CMH 146.21 7 < 0.0001 Worm strain (Run) CMH 139.83 7 < 0.0001 Worm strain (Worm genotype) CMH 10.32 7 0.1713 Worm genotype (Bacteria) CMH 135.50 3 < 0.0001 Worm genotype (Run) CMH 129.34 3 < 0.0001 Run (Bacteria) CMH 52.56 4 < 0.0001 Run (Worm strain) CMH 51.09 4 < 0.0001 Run (Worm genotype) CMH 50.89 4 < 0.0001 The associations were assessed with the likelihood ratio test (LRT) or the Cochran-Mantel-Haenszel (CMH) test. Bold probabilities are significant after Dunn-Sidák correction. In the second experiment, we specifically addressed the presence of an interaction between two bacterial strains (Db11, ATCC274) and four host strains (MY8, MY10, MY14, MY15), the latter belonging to two different host genotypes. For this experiment, all factor combinations were included in each 96-well plate and in one experimental run. Only 7 out of 384 cases had to be excluded for the reasons given above (1.82%), resulting in 46 to 48 data points per factor combination. Again, the number of valid cases did not differ among factor combinations (LRT, χ 2 = 0.093, d.f. = 3, P = 0.996). In the control treatment of this experiment, all animals were alive. The second experiment confirmed the presence of variation in host resistance and pathogen virulence, although the overall level of virulence was lower than in the main experiment (Fig. 2 ). Subsequent OLR revealed a significant effect from the factor worm strain or worm genotype, and also the interaction between the bacterial strain and either worm strain or genotype. The effect of bacterial strains was significant before Dunn-Sidák adjustment of significance levels (due to multiple testing), but insignificant afterwards (Table 3 ). For these OLR analyses, the models employed were significantly better than a model without any predictors (model including worm strain as factor: LRT, χ 2 = 62.29, d.f. = 7, P < 0.0001; model including worm microsatellite genotype as factor: LRT, χ 2 = 59.12, d.f. = 3, P < 0.0001). Moreover, they were not significantly worse than the respective saturated models (model including worm strain: LRT, χ 2 = 4.62, d.f. = 7, P = 0.7060; model including worm genotype: LRT, χ 2 = 1.32, d.f. = 3, P = 0.7238), suggesting that they contained sufficient details to explain the observed variation. Figure 2 Treatment response for the different bacterial and worm strain combinations of the second experiment. The black area denotes the proportion of dead worms, grey the proportion of morbid, and white the proportion of live worms. Table 3 Ordinal logistic regression analysis of the importance of different factors in the second experiment. Source χ 2 d.f. P Consideration of worm strains as a factor Bacteria 4.89 3 0.0270 Worm 33.20 1 < 0.0001 BacteriaWorm 26.89 3 < 0.0001 Consideration of worm genotypes as a factor Bacteria 4.80 1 0.0284 Worm 31.97 1 < 0.0001 BacteriaWorm 24.50 1 < 0.0001 Ordinal logistic regression was based on a model, which contained bacterial strain, worm strain (alternatively worm genotype), and the interaction between the two as factors. The importance of different factors was assessed with the likelihood ratio test. Bold probabilities indicate significance after Dunn-Sidák correction. Subsequent performance of association tests generally corroborated the OLR analyses: The different predictor variables had a significant effect on the treatment response (LRT analysis in Table 4 ). With two exceptions, this was still true after correcting for one of the other predictors (CMH tests in Table 4 ). One of the exceptions refers to the factor bacterial strain, which no longer produced a significant effect if corrected by any of the other factors. This is in agreement with results from the OLR analysis. The other case shows that the factor worm strain becomes insignificant if corrected by worm genotype, which confirms the findings for the main experiment (see above). Consequently, the results clearly demonstrate that there are significant differences among host genotypes and, most importantly, that there are significant strain- or genotype-specific interactions between the two species. Table 4 Association analysis of the impact of different factors on worm condition in the second experiment. Factor Test χ 2 d.f. P Single factor effects Bacteria LRT 10.31 2 0.0058 Worm strain LRT 32.80 6 < 0.0001 Worm genotype LRT 30.46 2 < 0.0001 Factor effects in consideration of one of the others (in brackets) Bacteria (Worm strain) CMH 4.34 1 0.0372 Bacteria (Worm genotype) CMH 4.41 1 0.0358 Worm strain (Bacteria) CMH 29.21 3 < 0.0001 Worm strain (Worm genotype) CMH 0.93 3 0.8193 Worm genotype (Bacteria) CMH 28.37 1 < 0.0001 The associations were assessed with the likelihood ratio test (LRT) or the Cochran-Mantel-Haenszel (CMH) test. Bold probabilities are significant after Dunn-Sidák correction. Discussion We here provide evidence for the presence of i) genetic differences in resistance among natural C. elegans strains, ii) genetic differences in virulence among natural S. marcescens strains, and also iii) strain- or genotype-specific interactions between the two. The first of these points is generally consistent with our previous results on the presence of strain-specific differences in resistance of C. elegans towards Bacillus thuringiensis [ 20 ]. However, in the previous study, we compared C. elegans strains from different locations across the world, whereas in the present study all strains derive from the same place (the town of Münster in Northwest Germany) [ 18 ]. Previous microsatellite genotyping demonstrated that these strains are genetically extremely diverse [ 18 ]. Our present results highlight the fact that genetic diversity translates into phenotypic differences in resistance. Importantly, as these differences are present in one population, they could provide the basis for and/or represent the outcome of evolution under natural conditions. These conclusions are restricted to the host C. elegans , because the S. marcescens strains considered did not come from the same location. The observed strain- and genotype-specific interactions represent an important precondition for negative frequency dependent selection. As such, they may contribute to the emergence of co-evolutionary arms races [ 1 , 4 ]. The relevance of our results for the association between C. elegans and S. marcescens in the wild must currently be considered unclear. To date, it is unknown whether the two species indeed co-exist under natural conditions, even though it is strongly suggested by the fact that both – especially S. marcescens – are common soil inhabitants [ 19 , 23 ]. If they do co-exist, they clearly show the potential to engage in co-evolutionary interactions. In fact, in this case, the observed specificity may represent a signature of past counter-adaptations. Our results would then also suggest that such highly specific interactions are widespread among invertebrate hosts; they are currently only known in a few arthropods and molluscs (see the background section for examples). The situation is clearly different if the two species do not share the same natural habitat. In this case, the observed specificity must be the result of independent adaptations of parasite and host strains to other environmental conditions. Pleiotropy of such adaptations should then have produced the specific C. elegans - S. marcescens interactions as a side effect. For example, the C. elegans strains may have adapted differently towards environmental toxins. If the underlying detoxification mechanisms are also employed in the defence against pathogens, then this may result in the observed differences in resistance. Such mechanisms could indeed be of relevance in the interaction with S. marcescens , for which at least one toxin (hemolysin ShlA) was previously suggested to contribute to pathogenesis in C. elegans [ 22 ]. Moreover, such mechanisms may also account for highly specific interactions, even if the two species did co-exist in the wild, underlining the idea that past co-evolutionary events cannot be reliably deduced from the observation of specific interactions without further information (e.g. historical records of co-existence in nature or congruent phylogenies of host and parasite strains). Whatever its origin, the finding of high specificity in the interaction has further implications. The molecular basis of highly specific resistance is currently unexplored in invertebrate hosts. It could be due to the presence of different alleles of a certain cell surface protein targeted by specific parasite effector molecules. Such cell surface proteins have been suggested to be important for the interaction between C. elegans and Bt toxin, the main virulence factor of B. thuringiensis [ 24 , 25 ]. As a non-exclusive alternative, specificity may be a consequence of the inducible immune system as recently suggested for the specific interactions between the copepod M. albidus and the cestode S. solidus [ 14 ] or the waterflea D. magna and its microparasite P. ramosa [ 26 ]. The presence of an inducible system was recently demonstrated for C. elegans in response to S. marcescens [ 27 ], the fungus Drechmeria coniospora [ 28 ], and also the Bt toxin of B. thuringiensis [ 29 ]. Considering that diverse molecular tools are available for C. elegans , this nematode may in the future provide a valuable model system to dissect the molecular basis of specificity in invertebrate-pathogen interactions. Similarly, the observed highly specific virulence was previously unknown for S. marcescens . This bacterium is considered to be an opportunistic pathogen with a broad host range [ 23 ]. Hence, it should mainly possess unspecific virulence factors, which are effective against a large number of different taxa. Interestingly, some of the genes previously identified to contribute to pathogenesis in C. elegans also mediate virulence in other hosts ( Drosophila melanogaster ; mice), whereas other genes do not [ 22 ]. This already indicates some degree of specificity. Our results may now provide the basis for a molecular genetic characterisation of virulence factors that vary in their specific effects against different strains of a single host species. This information may potentially be of great value for understanding pathogenicity of S. marcescens in humans, where this bacterium has become a growing health problem, primarily in nosocomial infections [ 30 ]. Conclusions Based on the analysis of natural isolates of the nematode C. elegans and its potential microparasite S. marcescens , our study provides evidence for i) genetic variation in host susceptibility and parasite virulence, and also ii) strain- and genotype-specific interactions between the two. These two factors represent an important precondition for frequency dependent selection and as such for the emergence of co-evolutionary arms races. Such highly specific interactions were previously unknown for C. elegans or S. marcescens . Moreover, they have not as yet been reported for invertebrates other than molluscs and arthropods. At least for C. elegans , the observed variation was found among strains from the same population, such that it could indeed be of relevance for evolutionary changes under natural conditions. Taken together, these findings suggest that there is widespread potential for co-evolutionary interactions in animal hosts. Both C. elegans and S. marcescens represent important model organisms in biological research for which a diversity of manipulative techniques is available. Therefore, the association between these two species may in the future provide a valuable tool for the comprehensive analysis of such highly specific interactions. Methods We compared eight different natural C. elegans strains with 5 different natural S. marcescens strains plus one control. The C. elegans strains were isolated by HS and co-workers in 2002 from Münster, North-West Germany [ 18 ]. They are available from the Caenorhabditis Genetics Centre under strain numbers MY6, MY8, MY10, MY14, MY15, MY17, MY18, MY20 [ 31 ]. Some of these strains bear different genotypes: strains MY6 and MY18 have genotype EU4; MY8, MY10 and MY20 genotype EU3; MY14 and MY15 genotype EU2; and MY17 genotype EU5 [ 18 ]. Maintenance of worms, including feeding, worm transfer, synchronisation of cultures and cryo-preservation followed standard procedures [ 32 ]. These C. elegans strains had all been cryo-preserved within 5 generations after isolation [ 18 ]. They were then thawed only few generations before the start of the experiments to ascertain that they were subjected to selection towards laboratory conditions for the shortest possible time. One generation before the start of the experiment, the worm cultures were always synchronised using NaOH/NaOCl-treatment [ 32 ]. The S. marcescens strain Db11 was originally isolated by H. Boman [ 33 ]; the strain Sm2170 was obtained from T. Watanabe [ 34 ]; and strains ATCC274 [ 35 ], Sma3, Sma13 from G. Salmond (Cambridge, UK). These strains are known to differ in pigmentation, which is thought to correlate with virulence: Db11, Sma3, and Sma13 have no pigments, whereas ATCC274 and Sm2170 are pigmented [ 22 ]. They were also already shown to differ in virulence towards the main C. elegans strain N2 when tested on solid agar, whereby virulence varied in the following order (from high to low): Sm2170 > ATCC274 > Db11 > Sma3 = Sma13 [ 22 ]. Note, however, that the time-course of infection in liquid medium is much more rapid than on solid medium and that the underlying mechanisms of pathogenesis in the two cases are not identical, at least for Db11 (JJE and E. Pradel, unpublished observations). One day before the start of the experiment, the bacteria were grown in Luria Broth (LB) for about 18 h at 37°C. Their OD was then adjusted to a value of 0.1 by addition of LB. An OD 0.1 corresponds to a cell count of approximately 2 × 10 8 per ml. As a control, we used heat-killed bacteria of strain Db11 (incubation at 70°C for 15 min). These dead bacteria were previously shown to have lost their deleterious effects on C. elegans [ 22 ]. The interaction between C. elegans and S. marcescens was assessed using a simple survival assay [ 20 ]. For this, individual worms were confronted with a defined concentration of the pathogens in NGM solution and their survival checked after 24 h. The experiment was performed in 96-well plates. Each plate always contained the five different S. marcescens strains and the control, randomised across the plate. Only one C. elegans strain was examined per plate. Eight plates were analysed in parallel, each with one of the eight different C. elegans strains. Five runs of this set-up were performed, whereby the order in which strains were analysed during each run was randomised. This set-up resulted in a maximum of 80 data points per bacterial strain – worm strain combination. For a specific run, 50 μl NGM solution (without any bacteria) were first added to each well of the 96-well plates using a multi-channel pipette. Thereafter, individual worms were transferred to each well using the COPAS automated worm-sorter (Union Biometrica Inc.). For worm-sorting, we used synchronised L4 stage worms. Success of worm transfer was monitored. If no worm or more than one worm was transferred to a particular well, then it was excluded from further analysis. After worm transfer was completed, 50 μl NGM solution with pathogens were added to each well with a multi-channel pipette. These 50 μl contained 45 μl NGM solution and 5 μl of bacteria in LB with an OD of 0.1, resulting in a total of approximately 1 × 10 7 bacterial cells per well at the start of the survival assay. This concentration was found in a pilot study to permit detection of differences in survival among worm strains after 24 h. After this time period, the condition of the worms was recorded using the following three categories: i) alive (clearly visible body movements; in some cases only after being touched with a small pipette tip), ii) morbid (touching them with a small pipette tip resulted in retarded, very slow movements), iii) dead (no movements, even after being touched with a small pipette tip). After completion of the experiment, we re-assessed the interaction between two bacterial strains (Db11 and ATCC274) and four worm strains (MY8, MY10, MY14, MY15; the first two and the last two have identical microsatellite genotypes). The general set-up was the same as above. In this case, a total of four 96-well plates were studied at the same time. In contrast to the above experiment, each 96-well plate contained both the different worm and bacterial strains, randomised across the plate. This set-up results in a maximum of 48 data points per factor combination. In addition to the above, we included two 96-well plates as a control. These contained heat-killed Db11 bacteria and the four worm strains randomised across plates (48 data points per worm strain). After exposure to the pathogens, we confirmed that worms were indeed infected with the bacteria by analysis of some of the animals ( N = 40) using differential interference contrast microscopy and a fluorescent microscope (DMIRBE, Leica). The statistical analysis was performed with the program JMP version 5.0 (SAS Institute Inc.). Based on the hierarchical order of the categorical response variable (0, dead; 1, morbid; 2, alive), we used an ordinal logistic regression analysis (OLR) [ 36 , 37 ]. For the main experiment, we included bacterial strain, worm strain, the interaction between the two, and also run nested within both bacterial and worm strain as factors in the model. The whole analysis was repeated using worm microsatellite genotypes instead of worm strain in the model. For the second experiment, which was performed on a single occasion instead of separate runs, a full factorial model was employed, including bacterial strain, worm strain (alternatively, microsatellite genotype), and the interaction between the two as factors. For the main experiment, the lack of fit test was significant, indicating that the chosen model may not be sufficiently complex to explain the variation in the data (see results section). Therefore, we additionally employed association tests based on the inferred frequency counts for the different factor combinations. We specifically assessed the association between the treatment response (condition of the worms) and either of the following factors: bacterial strain, worm strain, worm microsatellite genotype, and run. The significance of the association was inferred using the likelihood ratio test (LRT) [ 36 , 38 ]. We further used the Cochran-Mantel-Haenszel test (CMH) to assess the conditional independence between the treatment response and one of the above factors in consideration of a second factor from the above list [ 36 ]. The response variable was treated as ordinal, while the predictors were treated as nominal (ordinal-nominal conditional association test) [ 36 ]. Multiple testing was accounted for by adjusting the significance level using the Dunn-Sidák procedure [ 38 ]. List of abbreviations used CMH, Cochran-Mantel-Haenszel test; LB, Luria broth; LRT, likelihood ratio test; NGM, nematode growth medium; OD, optical density; OLR, ordinal logistic regression. Authors' contributions HS designed the study, carried out the experiments, analysed the data and wrote the first draft of the manuscript. JJE participated in the design of the study, the interpretation of the data and revised the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538262.xml
523859	Age related differences in individual quality of life domains in youth with type 1 diabetes	Background Investigating individual, as opposed to predetermined, quality of life domains may yield important information about quality of life. This study investigated the individual quality of life domains nominated by youth with type 1 diabetes. Methods Eighty young people attending a diabetes summer camp completed the Schedule for the Evaluation of Individual Quality of Life-Direct Weighting interview, which allows respondents to nominate and evaluate their own quality of life domains. Results The most frequently nominated life domains were 'family', 'friends', 'diabetes', 'school', and 'health' respectively; ranked in terms of importance, domains were 'religion', 'family', 'diabetes', 'health', and 'the golden rule'; ranked in order of satisfaction, domains were 'camp', 'religion', 'pets', and 'family' and 'a special person' were tied for fifth. Respondent age was significantly positively associated with the importance of 'friends', and a significantly negatively associated with the importance of 'family'. Nearly all respondents nominated a quality of life domain relating to physical status, however, the specific physical status domain and the rationale for its nomination varied. Some respondents nominated 'diabetes' as a domain and emphasized diabetes 'self-care behaviors' in order to avoid negative health consequences such as hospitalization. Other respondents nominated 'health' and focused more generally on 'living well with diabetes'. In an ANOVA with physical status domain as the independent variable and age as the dependent variable, participants who nominated 'diabetes' were younger (M = 12.9 years) than those who nominated 'health' (M = 15.9 years). In a second ANOVA, with rationale for nomination the physical status domain as the independent variable, and age as the dependent variable, those who emphasized 'self care behaviors' were younger (M = 11.8 years) than those who emphasized 'living well with diabetes' (M = 14.6 years). These differences are discussed in terms of cognitive development and in relation to the decline in self-care and glycemic control often observed during adolescence. Conclusions Respondents nominated many non-diabetes life domains, underscoring that QOL is multidimensional. Subtle changes in conceptualization of diabetes and health with increasing age may reflect cognitive development or disease adjustment, and speak to the need for special attention to adolescents. Understanding individual quality of life domains can help clinicians motivate their young patients with diabetes for self-care. Future research should employ a larger, more diverse sample, and use longitudinal designs.	Background Quality of life (QOL) is now recognized as an important outcome for people with diabetes. In general, diabetes has been shown to negatively impact QOL [ 1 ]. Tighter glycemic control is associated with better QOL, despite the increased treatment demands it commonly requires [ 2 ]. As standards for optimal glycemic control get more rigorous, and as medical treatments for diabetes develop, a better understanding of the personal meaning of disease and related QOL would be beneficial. The measurement of QOL is evolving, and a state of the art review identified numerous different QOL measures [ 3 ]. Most QOL measures ask individuals to assess their QOL using predetermined items. This is true for both generic measures such as the Medical Outcome Study Short Form (SF-36) [ 4 ], and disease specific measures such as the Diabetes Quality of Life for Youth questionnaire (DQOLY) [ 5 , 6 ] and the Audit of Diabetes Dependent Quality of Life (ADDQoL) [ 7 ]. A more recent approach to QOL assessment is the development of individualized, or patient generated, measures that use an open ended question format. These measures allow respondents, from their own perspective, to identify the life domains that contribute most to their overall quality of life. This complimentary approach allows respondents to paint a fuller picture of their quality of life focusing on the domains that they consider important. The ADDQoL approaches this technique, in that it allows patients to indicate items on the measure that are 'not applicable' to their quality of life and weights remaining responses appropriately. Its paper and pencil format allow for its wider use than an interview format. However, it does not allow respondents to generate their own domains. By pre-selecting life domains, and/or limiting items to those that are diabetes relevant, we limit our breadth and depth of understanding of youth with diabetes. While diabetes impacts nearly all aspects of a young person's life, it may not be the central, organizing construct under which all other domains fall. If a child with diabetes was asked "What is important to you?" and "How is that important thing going for you right now?" diabetes may or may not be mentioned. For example, in a sample of adolescents with type 1 diabetes, fewer than 1/3 of participants ranked diabetes as the most important life domain, the remainder rated it much lower [ 8 ]. And when diabetes is considered important, the rationale for its importance may vary between respondents or by developmental stage. Furthermore, we know very little about the developmental aspect of QOL. Do QOL domains differ by age? Does the importance attributed to the domains differ? Does the rationale for their importance differ? These questions are not of solely theoretical interest. On an individual patient basis, such information may be clinically useful. Understanding the QOL domains that are important to an individual patient gives a provider information about how to motivate that individual for improved self-care. This is true for diabetes-related life domains and those domains that are not directly diabetes-related. For example, a provider treating a child who values athletics could use this information to motivate the child to improve diabetes self-care by demonstrating to him that improved glycemia could enhance his athletic performance. Using a patient's own value system to promote healthy behavior has been advocated by others [ 9 ] and is consistent with a patient centered approach to medical care. But, the tools to assess the life domains of importance for a given individual are required for such an approach. Walker and Bradley [ 8 ] found that diabetes nurse specialists who had considerable knowledge and experience of individual patients, were unable to predict accurately patients' ratings of their own QOL, nor the relative importance of the domains that constitute it, because of the subjective and complex nature of QOL. The Schedule for the Evaluation of Individual Quality of Life-Direct Weight (SEIQoL-DW) is a theory based [ 10 ] structured interview during which respondents nominate 5 life domains that are most important to their own QOL [ 11 ]. The SEIQoL-DW has been used with a variety of adult and geriatric medical populations including patients with HIV/AIDS [ 12 ], cancer [ 13 ], amyotrophic lateral sclerosis [ 14 ], psychiatric diagnoses [ 15 ], multiple sclerosis [ 16 ], motor neuron disease [ 17 ], Hodgkins lymphoma [ 18 ], and stem cell recipients [ 19 ]. In the only other study using the SEIQoL in youth with diabetes, Walker and Bradley [ 8 ] administered it to 15 adolescents with type 1 diabetes. What is reported here is a descriptive, exploratory study of individual quality of life in young people with diabetes. It is not the intention to advocate the use of the SEIQoL-DW in place of standard measures such as the DQOLY. Rather, the SEIQoL-DW interview was used to provide a window into the lives children and young people with diabetes beyond what is typically assessed with paper and pencil measures. Such an understanding helps us view the patient as a whole person, and approach them from their own unique perspective. Previously, we reported the appropriateness of using the SEIQoL-DW in the children from this sample under age 18 years [ 20 ]. This article reports age related differences in individual quality of life domains in an expanded sample that includes college-aged respondents. Methods Sample Participants were campers, and young counselors who had previously been campers, at an overnight summer camp for children with diabetes. The camp serves children with diabetes in northern New England. One hundred and twenty campers from 8–15 years old attend a 2-week session. The majority of the staff has diabetes. Measures Demographic and disease variables For campers, parents completed a survey of demographic and disease variables for themselves, their family, and their child with diabetes. Counselors who participated in the study completed the survey of demographic and disease variables themselves. Demographic variables included family structure, socioeconomic status, school performance, and race/ethnicity. Disease variables included disease duration, treatment regimen, HbA1c (gold standard measure of glycemic control), frequency of complications, and emergent use of health care services. Schedule for the Evaluation of Individual Quality of Life-Direct Weighting As described by Browne et al [ 19 ], there are three stages to administration of the SEIQoL-DW. In the first stage respondents nominate five life domains that they consider most important to their overall quality of life. Instructions were modified only slightly for youth, with an emphasis on making the language easy to understand and examples age appropriate. Participants were asked "For each of us, happiness and satisfaction in life depends on the areas of life which are important to us. When these important areas are going well, we are happy, but when they are going badly, we feel worried or unhappy. What is considered important varies from person to person. What is most important to you may not be so important to me or to your parents or friends and vice versa. I am interested in knowing what the most important areas of your life are at the moment. What are the five most important areas of your life at present – the things which make your life happy or sad at the moment?" If participants are unable to volunteer domains, examples are read from a standard list that is included in the SEIQoL-DW administration manual, and responses are noted as such. To assess a child's understanding of these directions, each participant was asked to "retell" the directions to the interviewer. If the child was unable to repeat the directions in basic terms and show understanding, they were excluded from the study. The appropriateness of the SEIQoL-DW in youth has been described elsewhere [ 20 ]. The authors took great care to ensure that those who were included yielded valid data, and those who did not were excluded. In the second stage, the respondents rate each domain on a 0–100 mm vertical visual analogue scale anchored at the two extremes by the terms 'best possible' and 'worst possible'. These anchors are designed to allow individuals to use their own criteria when assessing their status within each domain. The third stage involves a weighting procedure wherein the respondent judges the relative importance of each domain. In the original version of the SEIQoL a technique known as judgment analysis was used. Because of practical limitations of judgment analysis, a direct weighting procedure has been developed [ 21 ]. The direct weighting (DW) procedure of the SEIQoL-DW consists of asking participants to manipulate five stacked, centrally mounted, interlocking laminated disks. Each disk is a different color and is labeled with one of the five domains nominated by the individual. The disks can be rotated over each other to produce a dynamic pie chart where the relative size of each sector represents the weight the respondent attaches to a QOL domain. The proportion of the chart that each sector represents can be scored from a 100-point scale on the circumference. See figure 1 for example of the weighting instrument. Total quality of life is then calculated by multiplying each domain importance rating by the domain weighting and then summing the products. Examples of this scoring have been previously published [ 20 ]. Figure 1 A representative distribution of life domains on the weighting instrument. Studies have shown the SEIQoL to have good internal consistency, ranging from .6 to .9 [ 21 - 26 ]. It has also been shown to have adequate test-retest reliability, with Pearson's correlation >.70 [ 27 ]. The newer SEIQoL-DW has been shown to be sensitive to change, to have good construct validity [ 27 , 28 ], and to be psychometrically comparable to the SEIQoL [ 21 , 26 ]. A recent review concluded that the SEIQoL is superior to other patient generated QOL measures [ 29 ]. Procedures This project was carried out in accordance with the American Psychological Association guidelines for ethical conduct of research. It was approved by the University of Connecticut Health Center's institutional review board. One week prior to the two-week camp session, a letter was sent to the parents of campers, describing the study. Parents were sent a consent form for themselves, an assent form for their child, and a survey of disease and demographic data. Upon their arrival at camp, the materials were collected from parents and reviewed for completeness. Sixty one percent (n = 73) of campers and parents handed in completed questionnaires on the first day of camp. The most common reason given for not participating was lack of interest on the part of the child. During the 2-week camp session children were pulled one at a time from regular camp activities and administered the SEIQOL-DW. Participants were allowed to choose a sugar free treat (soda, gum, or mints) for their participation. During the one-week of pre-camp (when counselors are preparing for arrival of campers), counselors with diabetes who had previously been campers were approached and invited to participate in the study. Informed consent was obtained, and staff members were administered the SEIQoL-DW individually at their convenience. All eligible staff chose to participate (n = 17). Two interviewers completed all 90 interviews. They were the first author who is a clinical psychologist, and the third author who is a medical student with an M.P.H. in health behavior. Both interviewers studied the SEIQoL-DW manual, role-played giving the SEIQoL-DW, and conferred about validity, boredom, and fatigue scoring. Results Participants Consenters were compared to the total camp population for systematic differences. Results of a chi square test show no significant differences in gender. Results of an ANOVA show no significant differences in age, HbA1c, or duration of diabetes. Likewise, there were no significant differences between campers and counselors for quality of life total scores, gender, type of diabetes, age of diagnosis, number of daily injections, hypoglycemic episodes in the previous month, as well as number of sick days, hospitalizations, ketoacidosis episodes, and emergency room visits in the previous year. On average, participants were 13 years old, came from 2-adult homes (74%) with 1 or 2 siblings. All but one participant were European American. Most did "well" or "very well" in school (69%), and had parents with at least 2 years education beyond high school. Participants had diabetes for an average of about 6 years and had been attending diabetes camp for an average of 4 years. All were on multiple injection regimens or insulin pumps (n = 29), and their average glycemic control was fair (mean HbA1c = 8.02). Participants had missed an average of 3 days of school in the last year due to their diabetes, and had experienced about 6 episodes of hypoglycemia in the previous month. Table 1 Demographic and diabetes descriptive findings, n = 80 (numerical discrepancies reflect missing values) Mean (SD) Sex Male 47.5% (n = 38) Female 52.5% (n = 42) Age 14.0 (3.4) Age at diagnosis 7.1 (3.1) Years since diagnosis 7.0 (4.3) Most recent HbA1c 8.1 (1.8) # Injections/day 2.2 (2.1) # Children on CSII (n) 29 Diabetes sick days from school in last year 3.3 (5.4) Diabetes hospitalizations in last year 0 79.5 % (n = 61) 1–2 18.2% (n = 15) >2 2.2% (n = 2) DKA episodes in last year 0 72.9% (n = 53) 1–2 15.3% (n = 12) >2 11% (n = 10) Hypoglycemic episodes in last month 6.2 (6.5) Years at diabetes camp 4.6 (3.6) # of siblings 1.8 (1.4) Parent education (in years) 14.4 (2.6) Parent marital status Single/separated/divorced & living alone 16% (n = 12) Single/separated/divorced & cohabitating 17% (n = 14) Married 66% (n = 52) School performance Very poorly 1% (n = 1) Poorly 4% (n = 4) Ok 26% (n = 19) Well 33% (n = 26) Very well 36% (n = 30) Data from ten children were deemed invalid due to interviewers' judgment that the participant was unable to understand the SEIQoL-DW task. For example, one child did not understand the concept of 'importance', and instead only rated domains in terms of his happiness with them. All children whose data were deemed invalid were under 12 years old, with a mean age of 9.25. The data from these ten children (1/3 of those children under 12) are deleted from the following results. Table 1 presents these descriptive findings. Domain nomination Of the 400 total domains nominated by the 80 participants with valid data, only 21 domains (5%) were nominated with the assistance of the standard list. These 21 were for the third, fourth, and fifth domains. Thus, every respondent could nominate at least 2 domains without suggestion, and only a handful needed help with the additional 3 domains. Table 2 displays the domains and the frequency with which they were nominated. Table 2 SEIQoL-DW domains, with importance and satisfaction ratings Domain Number of times domain was nominated Mean importance rating out of 100 Mean satisfaction rating out of 100 Family 76 27.9 79.8 Friends 62 18.0 76.8 Diabetes 49 27.7 75.0 School 46 17.7 65.5 Health 30 22.9 80.1 Hobbies 26 15.3 66.2 Sports 17 14.2 67.6 Camp 14 15.5 89.9 Religion 12 30.6 83.8 Special person, such as a teacher 11 19.8 79.8 Approach to life, or mental attitude 10 20.0 69.6 Significant other (boyfriend/girlfriend) 9 19.0 62.5 Golden rule (treating others as would like to be treated) 9 20.1 64.7 The basics (housing, food, safety) 9 17.8 71.5 Career/future 7 15.5 62.8 Pet 7 18.3 82.1 Work 5 11.5 69.5 Nature 1 14.5 70.0 The most frequently nominated domain was 'family'. Family was nominated by 76/80 respondents. Answers were coded 'family' if the participant used the terms 'family' or 'parents'. Reasons given tended to involve either instrumental support (e.g., they provide me with clothes and a place to live) or emotional support (they love me and I can go to them with problems). Many respondents stated that their families were important because they helped with diabetes, e.g., purchased diabetes supplies, drove them to doctor's appointments, cooked nutritious food, and helped with treatment decisions. Results of a bivariate correlation indicate a significant negative correlation between age of respondent and the importance rating given to the 'family' domain, r = -.34, p < .01. There was no relationship between respondent age and 'family' satisfaction. The second most frequently nominated domain was 'friends'. Respondents tended to refer to a group of friends, rather than an individual friend. Friends were generally valued for their emotional support, their companionship, and the participant's ability to relax and have fun with them. The ability to "be myself" and still be accepted by friends was a common theme. Results of a bivariate correlation indicate a significant positive correlation between age of respondent and the importance rating given to the 'friends' domain, r = .35, p < .01. There was no relationship between respondent age and 'friends' satisfaction. School was nominated by many respondents as an important life domain. School was deemed important for a variety of reasons. Some respondents stated that school was important because learning is important in its own right. Some stated that school performance was important to assure acceptance to a good college and have a good job, or because their school performance was important to their parents. Age was not related to 'school' importance or satisfaction. The nomination of 'diabetes' as an important life domain was common. However, respondents' explanations for its nomination were varied. One type of response referred to taking proper care of diabetes in order to avoid negative consequences. These responses included things like eating well, self-monitoring blood glucose, keeping active, and taking injections on time to avoid medical complications, hospitalization, or death. Respondents who gave this sort of answer were fairly concrete and said things such as "I have to take my shots or I will end up in the hospital". Thirty four responses fell into this diabetes 'self-care behaviors' category. Another type of diabetes response referred to living well with diabetes. These responses included things like doing enjoyable activities despite diabetes, successfully negotiating diabetes treatment with parents, feeling proud of self when diabetes is controlled, receiving emotional support for diabetes, and keeping self, friends and family from worrying about diabetes. Respondents who gave this sort of answer provided more abstract explanations, and said things such as "I can't let diabetes stop me" and "It's important that people around me understand what it's like for me to have diabetes". Eleven responses were of this type. Results of an ANOVA with rationale for diabetes nomination (self care vs. living well with diabetes) as the independent variable, and age as the dependent variable revealed that those who provided 'self care behaviors' as a rationale were younger (M = 11.8 years) were than those who provided living well with diabetes as a rationale (M = 14.6 years) F (3, 40) = 2.88, p < .05. There were no group differences for age of diagnosis, number of daily injections, hypoglycemic episodes in the previous month, as well as number of sick days, hospitalizations, ketoacidosis episodes, and emergency room visits in the previous year. Among those who nominated 'diabetes' as a domain, age was not related to 'diabetes' satisfaction. Respondents who nominated 'diabetes' as a domain were asked whether or not they would do so if they were in a non-diabetic environment such as school, as opposed to a diabetes summer camp. All respondents stated yes, they would nominate diabetes. However, some went on to say that they might not use the word 'diabetes' per se, and instead might use the word 'health'. Twenty four respondents did in fact nominate 'health' as a domain (but not diabetes). Explanations for 'health' focused on what might be considered more general wellness. Individuals who nominated 'health' said things like "I have to be healthy in order to do the things I enjoy", "I like staying fit", "you can't be happy without good health", and "when I don't feel well I'm in a bad mood". Six other respondents nominated both 'diabetes' and 'health' as separate domains. Results of a one-way ANOVA reveal that respondents who nominated 'diabetes' only were significantly younger (M = 12.9 years) than respondents who nominated 'health' only (M = 15.9 years) F (3, 75) = 4.53, p < .01. There were no group differences for age of diagnosis, number of daily injections, hypoglycemic episodes in the previous month, as well as number of sick days, hospitalizations, ketoacidosis episodes, and emergency room visits in the previous year. Among those who nominated 'health' as a domain, age was not related to 'health' satisfaction. Domains nominated that are not on the standard list Many domains were nominated by this sample that are not included in the standard list. They include 'diabetes', 'school', and 'camp', all of which clearly reflect that this was a sample of school aged young people attending a camp for children with diabetes. Some respondents nominated a 'special person' in their lives, such as a teacher or a coach. Other novel domains were nominated the explanations for which are less obvious. One domain involved 'mental attitude' and referred to taking the right approach to life and maintaining a positive outlook. Another involved 'the golden rule' and referred to treating others as one wants to be treated, with fairness and respect. As stated above, the most frequently nominated domains were 'family', 'friends', 'diabetes', 'school', and 'health'. Domains ranked by importance were 'religion', 'family', 'diabetes', 'health', and 'the golden rule'. Domains ranked by satisfaction were 'camp', 'religion', 'pets', 'health', and 'family' and 'a special person' tied for fifth. Quality of Life Total SEIQoL-DW scores ranged from 34.9 – 97.7, M = 78.1, SD = 11.2. Normative data for children are not available. However, these values are similar to those of a small clinic sample of adolescents with type 1 diabetes that reported SEIQoL satisfaction data only [ 8 ]. This and other clinical samples (adults with cancer, with mental illness, transplant recipients) are shown in Table 3 . Table 3 Means and SDs for the SEIQoL-DW for the study and comparison samples Sample Mean SD Youth with diabetes 78.1 11.2 Adolescents with diabetes (satisfaction ratings only) [8] 75.3 N/A Adult HIV/AIDS patients [12] 58.4 21.59 Adult community serious mental illness [15] 69.04 24.58 Adult advanced cancer patients [26] 50.9 17.8 Adult stem cell transplant recipients [19] 63.21 17.55 Discussion The purpose of this study was to explore the individual QOL domains of youth with diabetes. Consistent with previous findings [ 8 ] results indicate that life domains nominated by individuals were thematic and shared many common characteristics, but varied substantially across respondents. Moreover, even when different respondents nominated identical life domains, their rationales for the importance of those domains varied. These findings underscore the personal nature of QOL, and highlight the benefit of allowing individuals to express their views of the life domains that determine it. The young people in this sample nominated nearly all the life domains that are offered on the SEIQoL-DW standard list (with the exception of finances). They also went on to nominate additional domains, notably what we have termed 'mental attitude' and 'the golden rule'. These domains are some of the more fundamental aspects of quality of life but are not typically nominated in adult samples. Their nomination raises interesting questions. It is possible that youth are more concerned with these life domains than adults. This may be particularly true for children at overnight summer camp where community living and group cooperation is fundamental and continually reinforced. It is also possible that people with diabetes are more concerned with these life domains than their non-diabetic counterparts. Perhaps these respondents, who have experienced a major negative life event with their diabetes diagnosis and may also have suffered teasing or rejection due to this diagnosis, are more sensitive to issues of positive thinking and treating others with respect. Because this was an uncontrolled study, these are, of course, hypotheses that can only be tested with further investigation of controls. Nearly all respondents nominated a domain that reflected physical status. Younger respondents (who were on average 12 years old) were more likely to focus specifically on diabetes, and emphasize the importance of diabetes self-care behaviors. These respondents stressed adherence to the diabetes treatment regimen in order to avoid medical complications. Older respondents (who were on average 15 years old) were more likely to focus on general health and emphasize the need to live well despite the difficulties of diabetes. These age related data need to be interpreted cautiously due to small sample size, homogeneity of the respondents, and the cross sectional design of the study. The literature would benefit from further use of individual QOL measures with children and with chronic illness populations such as diabetes, in longitudinal designs. Nonetheless, these data do suggest a change in how diabetes is conceptualized during adolescence. There are several possible explanations for these age differences. First, Piaget proposed that formal cognitive operations begin in adolescence. With formal operations, adolescents gain the ability to think about their own thinking, to imagine many possibilities, and to mentally generate possible outcomes and thus rely less on real objects and events. Abstract thought becomes possible. The observed shift in focus from the concrete to the abstract – from 'diabetes' to 'health' and from 'self-care behaviors' to 'living well' – may be a reflection of the cognitive development that occurs with formal operations. Studies have demonstrated a systematic progression of children's understanding of illness that corresponds to Piaget's framework [ 30 , 31 ]. Other research has shown that the growth of children's conceptualization of illness paralleled, but lagged behind, conceptual development of physical causality [ 32 ]. Indeed, cognitive development may explain some children under 12 were not able to comprehend the SEIQoL-DW instructions. Second, as they go through adolescence, youth with diabetes may view their physical well being differently due to more extensive and broadened life experience. Older youth may realize that strict adherence to the medical regimen comes at a cost of an inflexible lifestyle, and may value quality of life over strict medical management. They may become more willing to make compromises in self-care in order to participate more fully in normal activities. They may also, for the first time, be in a position to make these compromises as parental control of day-to-day diabetes management wanes in adolescence. Third, this shift may reflect adolescents desire to fit in with peers, normalize their disease experience, and assimilate their illness. Viewing physical status in terms of health, rather than diabetes, and diabetes in terms of living well, rather than disease management behaviors, serves these functions. This shift in diabetes conceptualization could serve to be adaptive, or conversely it may herald the poor self-care and decreased glycemic control that is often noted in adolescence and young adulthood. Wysocki, Hough, Ward and Green [ 33 ] found that adolescents with diabetes are at risk of various unfavorable behavioral and health outcomes and that adjustment to the disease during earlier adolescence may be a predictor of subsequent health-related behavior and health status. We did not find a relationship between glycemic control and specific life domains nominated; this could be a result of low statistical power or a true lack of relationship. Also, we did not measure psychological adjustment to diabetes, and it remains an empirical question whether the conceptual shift observed in the present study is related to adjustment, and if so, the direction of the relationship. Furthermore, this shift in itself may be less important than host variables such as health beliefs and social support. Perhaps for a well adjusted, well supported individual, the shift toward general health may indicate disease assimilation, where as in less adjusted, less supported adolescents, the shift away from diabetes may indicate a denial of the disease and a withdrawal from appropriate self-care. In addition to differences in domain nomination, we also observed age effects for the importance assigned to domains. As would be expected, age was significantly positively associated with the reported importance of friends, and a significantly negatively associated with the reported importance of family. As young people grow older, the value they place on family and friends changes, and peers become increasingly important. These data are consistent with previous findings that children with diabetes find friends more helpful in diabetes management than many adults [ 34 ]. Interventions that include peers, or a 'diabetes buddy' may be complimentary to those that target the family. A clinician who knows what a patient values can use that information to build rapport and to promote healthy behaviors. For example, while only 12 respondents nominated religion as a domain, it was given a higher importance rating than any other domain. Asking "how would God want you to care for your diabetes?" may be more fruitful with these children than a discussion of, say, the benefits of glycemic control for sports performance. These data should be interpreted cautiously for several reasons. First, they reflect a select sample of white, higher socioeconomic status, high academic achieving children, many of whom have had several years of diabetes camp experience. This group may nominate different life domains, and endorse a higher QOL, than children who come from more impoverished environments without disease specific psychosocial experiences. These findings should be viewed tentatively until findings are replicated with larger and more diverse samples, in longitudinal designs that employ controls. Conclusions This article reports age-related differences in health related quality of life domains in youth with type 1 diabetes. It was found that younger respondents nominated 'diabetes' as a domain and focused on 'self-care behaviors', whereas older respondents nominated 'health' and focused more on 'living well with diabetes'. Although limited by sample size and homogeneity, these findings point to age related differences that may help explain deteriorating glycemic control during adolescence. What is clear is that development into maturity is a difficult task made more difficult by diabetes, and that special attention must be given to this population. The Society for Adolescent Medicine [ 35 ] and the American Academy of Pediatrics [ 36 ] have both issued statements stressing the importance of transitional care from pediatrics to adult medicine for youth with special health care needs. Grey et al. [ 37 ] conclude that diabetes treatment teams need to pay attention to the psychosocial needs of all adolescent patients. As the medical management of youth with diabetes improves, so must the understanding of behavioral and psychosocial status. These data shed light on the quality of life domains valued by youth with diabetes. Authors' contributions JW conceived of the study, and was responsible for design, analysis, and manuscript preparation. GA assisted with data analysis, and manuscript preparation and revision. SL carried out many of the interviews, assisted with data recording, data entry, and study coordination.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC523859.xml
550647	Antiglucocorticoid RU38486 reduces net protein catabolism in experimental acute renal failure	Background In acute renal failure, a pronounced net protein catabolism occurs that has long been associated with corticoid action. By competitively blocking the glucocorticoid receptor with the potent antiglucocorticoid RU 38486, the present study addressed the question to what extent does corticoid action specific to uremia cause the observed muscle degradation, and does inhibition of glucocorticoid action reduce the protein wasting? Methods RU 38486 was administered in a dose of 50 mg/kg/24 h for 48 h after operation to fasted bilaterally nephrectomized (BNX) male adult Wistar rats and sham operated (SHAM) controls. Protein turnover was evaluated by high performance liquid chromatography (HPLC) of amino acid efflux in sera from isolated perfused hindquarters of animals treated with RU 38486 versus untreated controls. Results Administration of RU 38486 reduces the total amino acid efflux (TAAE) by 18.6% in SHAM and 15.6% in BNX and efflux of the indicator of net protein turnover, phenylalanine (Phe) by 33.3% in SHAM and 13% in BNX animals as compared to the equally operated, but untreated animals. However, the significantly higher protein degradation observed in BNX (0.6 ± 0.2 nmol/min/g muscle) versus SHAM (0.2 ± 0.1 nmol/min/g muscle) rats, as demonstrated by the marker of myofribrillar proteolytic rate, 3-Methylhistidine (3 MH) remains unaffected by administration of RU 38486 (0.5 ± 0.1 v. 0.2 ± 0.1 nmol/min/g muscle in BNX v. SHAM). Conclusion RU 38486 does not act on changes of muscular protein turnover specific to uremia but reduces the effect of stress- stimulated elevated corticosterone secretion arising from surgery and fasting. A potentially beneficial effect against stress- induced catabolism in severe illness can be postulated that merits further study.	Background As part of the complex uremic metabolic syndrome, pronounced disturbances of carbohydrate and lipid metabolism are commonly observed, as are pathologic changes of amino acid and protein turnover [ 1 ]. An increased net protein degradation in uremia was seen as early as 1949 by Persike and Addis [ 2 ], and in the same year, Bondy et al . [ 3 ] showed that adrenal hormones are involved therein. These early findings were validated in the late eighties by Schäfer et al . [ 4 , 5 ] who postulated a leading role for glucocorticoids as cause of the observed changes. Schäfer et. al used the experimental approach of inhibiting activation of the glucocorticoid receptor by enteral application of the potent glucocorticoid antagonist RU 38486 in acutely uremic rats [ 6 ] and found a decrease both in the accumulation of 3-methylhistidine, an amino acid that is produced in actomyosine catabolism and is not further metabolized, and in the activity of myofibrillar protease. However, while an effect of RU38486 on liver gluconeogenesis and urea synthesis in uremia could be demonstrated [ 7 , 8 ], so far there is no proof of a direct action of RU38486 on muscle metabolism in uremia. To address this problem, the present study made use of the classical experimental design of the isolated perfused hindquarter of the rat [ 9 ], in which roughly 40% of the rat body's total muscle mass can be evaluated under closely defined in vitro conditions. With this experimental design and by comparing sham-operated and bilaterally nephrectomized animals, the present study looked at the question to what extent does corticoid action specific to uremia cause the observed muscle degradation, and does inhibition of glucocorticoid action reduce the protein wasting? Methods Animal experimentation was carried out on male Wistar rats, aged 11–15 weeks, weighing 217–225 g, from the animal experimentation facilities of Heinrich Heine University, Düsseldorf. Permission to use animals for experimentation was given by Regierungspräsident Düsseldorf, file nr. 26.4203.1-217/87 according to German federal law. Surgery for nephrectomy, sham nephrectomy and preparation for perfusion was carried out under narcosis with hexobarbital (EVIPAN- Na: 15–20 mg/ 100 g BW). Bilateral nephrectomy was performed using a dorsal access, ligation of renal vessel string, and excision of the kidney, leaving the adrenal glands in place. Sham operated animals underwent the same manipulations except for the ligation and excision. After surgery, animals were fasted for 48 h until perfusion. Nephrectomized animals had access to drinking water on the day of surgery for 8 hours, then were deprived of liquid to avoid lung edema. Sham animals had free access to drinking water throughout. Animals were randomly assigned to one of four groups: bilaterally nephrectomized (BNX) and sham operated (SHAM) treated with RU38486 and untreated BNX and SHAM animals. For treatment, RU38486 was dissolved in phenylmethanol, then mixed with sesame oil to form a milky suspension which was injected into subcutaneously into the lateral abdomen in three subdoses within 24 h, adding up to a total dose of 5 mg/100 g/BW. 48 h after initial surgery, animals were narcotized and prepared for perfusion as previously described[ 10 ]. The hindquarter was linked to the recirculation system after full passage of 70 ml of pre- perfusion medium, as shown in illustration 1. The pre- perfusion medium was discarded and not used for the recirculation experiment. The perfusion was carried out with a half- synthetic medium on the basis of Krebs- Ringer- bicarbonate buffer (KRB), pH 7.38 [ 11 ]. Oxygen carriers were calf erythrocytes prepared from fresh calf blood sampled two days before experimentation and maintained with 300 mg/l Ampicillin and 220 ml/l citric acid/ glucose stabilizer. Bovine albumine maintained the physiologically correct oncotic pressure. 10 -6 mmol/l phentolamine were added to avoid vessel contractions. During perfusion, the arterial pH and perfusate oxygenation were monitored using a pH- meter and a total oxygen content analyzer (LEX- O2- CON, Lexington Instr., Mass., USA). At the beginning and end of perfusion, plasma samples were frozen for amino acid analysis. Amino acid analysis by HPLC was carried out using 25 μl of deproteinized perfusate sample, mixed with o-phthaldialdehyde (OPA)/3-mercaptopropionic acid to form OPA- adducts that were separated on a reversed phase column and measured by fluometry. Quantification was done by comparison with a standard amino acid mix including 3-methylhistidine. Of the 20 proteinogenic amino acids, cysteine, proline, and asparic acid were not included in HPLC analysis. Statistical analysis was done using the "Student" t- test for ungrouped, non- paired data with f = n1 + n2-2 and a significance level of p < 0.05. Results Loss of body weight (BW) During the 48 h fasting period between operation and perfusion, animals had a pronounced loss of BW (table 1 ). In SHAM, it was 35.5 ± 5.3 g; administration of RU 38486 reduced this to 27.6 ± 5.9 g (p < 0.05). Nephrectomized animals demonstrated a much less pronounced loss of BW due to a significant increase in tissue hydratation (Table 2 ). RU 38486 reduces the weight loss in nephrectomized animals, too; however, this effect is much less pronounced and lacks statistical significance (BNX 15.7 ± 4.8 g, BNX + RU 12.2 ± 4.8 g). Oxygen utilization and development of acidosis Oxygen utilization in the perfused muscle tissues is roughly the same in all four groups (data not shown) and equal to in vivo data previously reported [ 12 ] from rats after 24 hour fasting. As expected, pH dropped significantly lower during perfusion in the nephrectomized groups (SHAM: 7.378 ± 0.033, BNX 7.321 ± 0.018, p = <0.01; SHAM + RU 7.4 ± 0.031, BNX + RU 7.312 ± 0.013, p < 0.001). Parameters of amino acid and protein metabolism Total amount and spectrum of amino acids released during reperfusion: During perfusion, amino acids are released in varying amounts as shown in ill. 2. Nephrectomized animals (BNX) showed a general increase in amino acid release. This, however, is significant only in a few individual amino acids. The total amino acid efflux increases by 10,4 % (p = 0.05) without qualitative change. Notable exception is glycine, which is released to a lesser amount in nephrectomized animals. Amino acid release after treatment with RU 38486: Nephrectomy equally increases the amino acid efflux in animals treated with RU 38486 by 14.4% (p < 0.05) without change in spectrum. The total efflux of amino acids, however, is significantly reduced in the comparison SHAM/ SHAM+RU (-18.6%, p < 0.05) and BNX/ BNX+RU (-15.6%, p < 0.001). Release of 3-methylhistidine: 3-methylhistidine, a derivate of histidin mainly from actin and myosin in sceletal muscle and intestinal mucosa [ 13 ], is not reutilized after proteolysis, but excreted via the kidney as 3-methylhistidine or N- acetyl- 3-methylhistidine. During reperfusion of sham- nephrectomized animals, 3-methylhistidine is released from the hindquarter to a small amount that is incresed by roughly 300% in the nephrectomized animals. Administration of RU38486 has no effect on 3-methylhistidine efflux. (table 3 ). Discussion Background Acute renal failure is a catabolic state, and unfortunately the inherent acceleration of protein breakdown cannot be suppressed effectively by provision of exogenous nutritional substrates [ 14 ]. The situation is multicausal. Unspecific mechanisms induced by the process of acute disease, underlying illness and associated complications are just one side of the problem. On the other, one observes specific uremic effects, insulin resistance, hormonal derangements, metabolic acidosis, circulating proteases and other inflammatory mediators together with effects induced by the acute loss of renal function and the type and intensity of renal replacement therapy [ 1 , 15 - 17 ]. One factor that has for long been associated with the disturbances of protein metabolism is glucocorticoid action. The first observations date back to the nineteen-forties, when Persike and Addis [ 2 ] reported an increased urea- nitrogen production in experimental uremia, and Bondy and coworkers [ 3 ] demonstrated that adrenal hormones were involved in this dysregulation. Half a century later, it is still not fully understood to what extent steroid hormone action is responsible for the catabolic situation observed in renal insufficiency[ 18 ]. It has been shown that administration of high doses of glucocorticoids to adrenalectomized rats resulted in decreased protein synthesis, increased protein degradation, and a negative nitrogen balance [ 19 ]. In patients with chronic renal failure, a positive correlation between muscle proteolysis and the plasma cortisol level has been observed [ 20 ]. The in vivo influence of both glucocorticoids and metabolic acidosis on muscle proteolysis has been elucidated in whole- body leucine turnover studies in adrenalectomized rats [ 21 ]. These findings indicate that glucocorticoids play an important role in net protein degradation. Price formulated this so: "glucocorticoids are required but not directly responsible for the acidosis-induced increase in the mRNAs encoding proteins of this degradative pathway"[ 22 ]. Experimental approach In order to evaluate the relative importance of glucocorticoid action on protein metabolism in acute renal failure, an experimental setting was chosen that allowed to study glucocorticoid action indirectly by selective blockade of the glucocorticoid receptor with the potent antiglucocorticoid RU38486, a substance that binds to the receptor without activating the further process of transcription [ 23 ]. Parenteral administration of a total of 50 mg/kg BW/ d of RU38486 allowed to avoid additional irritation of the animals by a gastric catheter in the postoperative phase. Relevant action of RU38486 has been observed in enteral substitution at a dose of 20 mg/kg BW/d [ 6 ]. The degree to which RU 38486 blocks the glucocorticoid receptor depends very much on the mode of application, and the target tissue. While a recent study[ 24 ] shows that 80% of glucocorticoid receptors are blocked in rat muscle following oral application of mifepristone of 50 mg/ kg BW, Kim et al.[ 25 ] demonstrated effective blocking of glucocorticoid receptors in rat brain following subcutaneous application of 80 mg/ kg over two days. Schaefer et al.[ 6 ], on whose experimental set- up the present study was modeled, had reported significant effects of an oral dose of 20 mg/ kg on muscle. In view of this, the choice for the experimental procedure seems justified as the present study uses a substantially higher dose. The isolated perfusion procedure introduced by Ruderman[ 9 ] is well established for the representative study of muscle metabolism. In this setting, the perfused muscle mass is approximately 40% of the total muscle mass. Taking into consideration the different metabolic requirements of the perfused tissues, roughly 90% of oxydative metabolism occurs in the muscle[ 12 ], making this experimental setting truly a skeletal muscle preparation that permits the observation of even very discrete metabolic changes during reperfusion. In the given experimental setting, it is difficult to account for the in detail contribution of protein degradation, amino acid intermediate metabolism, and protein synthesis. Factors that modify the efflux are transport systems in the cell membrane [ 26 - 28 ], which can be concentration- dependent (system L) or acting against the concentration gradient (system A), and the intermediary metabolism within the muscle cells[ 29 ]. Numerous previous assessments of the metabolic situation in the isolated perfused hindlimb demonstrate that these factors are relatively minor contributors, while amino acid efflux is nearly exclusively characterized by the net balance of protein metabolism both in anabolic and catabolic situations[ 12 , 30 - 34 ]. It is mostly due to changes in skeletal muscle, with only minor contributions from other tissues in this preparation [ 35 - 37 ]. Results Sham- operated animals are catabolic at the time of perfusion, having lost about 36 g BW (see results), which is approximately 16% of initial BW. Rats of this age are still growing, with an increase of approx. 5 g/ day (2–3% BW)[ 38 ]. The weight loss is due to both lipolysis[ 39 ] and protein loss[ 40 ], which causes the typical increase of amino acid release in the hindquarter of fasting rats[ 12 ]. Compared to data[ 34 ] from non- operated rats fasted for 48 hours under otherwise identical conditions, the total amino acid release is increased by 30% in the sham- operated rats described here. While nephrectomy increases the amino acid release by approximately 15%, the relative decrease of amino acid release following administration of RU 38486 is similar in both nephrectomized and sham- operated animals. This indicates a stress- accentuated adaptation to fasting caused by corticosteron secretion[ 41 ] increased beyond the normal range, with a further effect of uremia. Increased amino acid serum concentrations during fasting are mostly due to an inhibition of protein synthesis[ 42 ], although proteolysis mainly of myofibrillary proteins does play a role[ 43 ]. However, corticosterone is only one of several effectors at play. RU 38486 affects neither acidosis nor lactate/ pyruvate ratio. Both factors may contribute to the continuously increased amino acid release. In the case of acidosis, this may be due to an action on acid inhibitable transporters such as system A which reduce the supply of nutrients to the cells[ 28 ]. Another possible mechanism is through inhibition of leptin by acidosis[ 44 ], which in neutral pH might counteract muscle wasting[ 45 ]. Balancing acidosis in chronically uremic rats with increased corticosterone secretion inhibited protein degradation, but had no effect on the defective protein synthesis[ 21 ]. More recently, RU 38486 was shown to be ineffective in blocking acid- mediated protein degradation as its action is only an indirect one, mediated via insuline- like growth factor I (IGF- I)[ 24 , 46 ]. These findings indicate that RU 38486 acts through an inhibition of the corticosterone- mediated decrease of protein- synthesis without affecting other factors that act predominantly on the level of protein degradation. While all these and more factors contribute to the muscle degradation seen in excess glucocorticoid situations, the mechanisms responsible in ultima causa remain still unclear[ 18 ]. Nephrectomy enhances the catabolic situation: the total amino acid efflux is increased by roughly 15% compared to sham- operated animals. The relative increase of amino acids that are not metabolized, such as phenylalanin and tyrosin, indicates that this effect is due to the acute and complex metabolic situation of uremia, without differenciating between inhibition of protein synthesis and stimulation of protein degradation. A multitude of effectors partake in this metabolic turmoil [ 15 ], of which glucocorticoids have been accused of playing a leading role [ 47 ]. At first sight, this opinion is supported by the finding that isolated hindquarters of animals treated with RU 38486 show a significant reduction of total amino acid efflux (16–19%, p < o.05- p <0.001) compared to untreated animals that underwent the same surgical procedure, indicating that RU38486 inhibits some common degrading influence on protein metabolism. By contrast, in the comparison of the two groups treated with RU38486, the amino acid release remains increased in nephrectomized animals, albeit to a lesser extent, stressing the very point that glucocorticoids are only one of several factors that contribute to the net protein wasting. Chronically uremic rats with increased corticosterone- secretion [ 48 ] showed a less pronounced increase in protein degradation when acidosis was balanced while the defective protein synthesis remained unchanged. In the present experimental setting, acidosis evolving during perfusion was not corrected for. Acidosis and glucocorticoid action are seen as concomitant factors in the activation of the ubiquitin- proteasome pathway of muscle proteolysis [ 49 ], and a pH- responsive element in the promoter region for the ubiquitin- proteasome pathway has been reported [ 47 ]. In the situation of uncorrected acidosis, it therefore seems likely RU 38486 may have had an inhibitory effect on the corticoid- induced decrease in protein synthesis without influencing the proteolytogenic effects of other putative agents. This presumption is supported by the finding that RU 38486 had no effect on the efflux of 3-methylhistidin. While this result is in contrast to Schäfer et al. [ 6 , 8 ], Lowell et al . found no reduction of the efflux of 3-methylhistidin after adrenalectomy in the perfused hindquarter of fasted animals [ 50 ], and in rats with chronic uremia, RU38486- resistant protein catabolism with unchanged release of 3-methylhistidin has been demonstrated in vivo by Teschner [ 51 ]. As responsiveness of protein synthesis and degradation to amino acid availability seem to be regulated differentially [ 52 ] and activation of glucocorticoid- mediated proteolysis occurs only at relatively elevated hormone levels [ 53 ] compared to the inhibition of protein synthesis [ 54 ], it seems possible to speculate that RU38486 may have a more pronounced effect on net protein catabolism at substantially higher doses. The presented data fail to show that RU38486 inhibits glucocorticoid action in the specific uremic setting while it clearly reduces the elevated net protein catabolism compared to non- operated animals. This suggests that glucocorticoid mediated protein wasting in acute uremia is rather a by- product of the overall stress, in the present experimental setting caused by surgery and fasting, than due to an independent action specific to uremia. While this finding abolishes hopes to counteract muscle wasting in uremia by administration of an anticorticoid drug and indirectly rather stresses the well described[ 55 ] clinical importance of a balanced acid- base status, it may open speculation about the usefulness of RU486 in post- traumatic states and severe illness. Conclusion Both sham- operated and nephrectomized animals show an increase in net protein catabolism. RU38486 clearly reduces the net protein wasting in both groups, but the increase in net protein catabolism observed over sham- operated animals remains unchanged in nephrectomized rats. The effect of antiglucocorticoid RU38486 may be attributed to an inhibition of fasting and operative stress- induced cortisol action which, even when within the physiological range, promotes increased protein turnover [ 56 ], and to a protective effect against the inhibition of protein synthesis. While RU38486 had no effect on net protein catabolism that could be specifically attributed to uremia, the demonstrated anticatabolic effectiveness in a stress accentuated metabolic situation should be studied more closely. Possible targets for therapeutic application under this aspect include post- traumatic states and severe illness. Competing interests The author(s) declare that they have no competing interests. Authors' contributions A.M. carried out animal experimentation, sample analysis, statistical analysis and wrote the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC550647.xml
538276	Gapped alignment of protein sequence motifs through Monte Carlo optimization of a hidden Markov model	Background Certain protein families are highly conserved across distantly related organisms and belong to large and functionally diverse superfamilies. The patterns of conservation present in these protein sequences presumably are due to selective constraints maintaining important but unknown structural mechanisms with some constraints specific to each family and others shared by a larger subset or by the entire superfamily. To exploit these patterns as a source of functional information, we recently devised a statistically based approach called c ontrast h ierarchical a lignment and i nteraction n etwork (CHAIN) analysis, which infers the strengths of various categories of selective constraints from co-conserved patterns in a multiple alignment. The power of this approach strongly depends on the quality of the multiple alignments, which thus motivated development of theoretical concepts and strategies to improve alignment of conserved motifs within large sets of distantly related sequences. Results Here we describe a hidden Markov model (HMM), an algebraic system, and Markov chain Monte Carlo (MCMC) sampling strategies for alignment of multiple sequence motifs. The MCMC sampling strategies are useful both for alignment optimization and for adjusting position specific background amino acid frequencies for alignment uncertainties. Associated statistical formulations provide an objective measure of alignment quality as well as automatic gap penalty optimization. Improved alignments obtained in this way are compared with PSI-BLAST based alignments within the context of CHAIN analysis of three protein families: G i α subunits, prolyl oligopeptidases, and transitional endoplasmic reticulum (p97) AAA+ ATPases. Conclusion While not entirely replacing PSI-BLAST based alignments, which likewise may be optimized for CHAIN analysis using this approach, these motif-based methods often more accurately align very distantly related sequences and thus can provide a better measure of selective constraints. In some instances, these new approaches also provide a better understanding of family-specific constraints, as we illustrate for p97 ATPases. Programs implementing these procedures and supplementary information are available from the authors.	Background As the genome projects continue to generate sequence data, it is increasingly common to find protein superfamilies with thousands of members in the protein database. Given sufficient numbers of sequences, sensitive iterative search and alignment procedures, such as PSI-BLAST [ 1 ] and SAM [ 2 ], often reveal that protein families previously thought to be distinct are, in fact, distantly related. Protein structural analysis likewise reveals subtle evolutionary relationships between protein families sharing very little sequence similarity. Since our ability to make protein structure and function predictions depends in large part on alignment accuracy, it is thus important to develop alignment methods able to handle these increasingly large and diverse sets of distantly related sequences. Certain protein families within these large superfamilies are often very highly conserved across distantly related organisms. Such proteins include, for example, certain metabolic enzymes, DNA replication and repair factors, certain structural proteins, such as actin, the motor protein dynein, and regulatory and signalling factors, such as protein kinases and Ras-like GTPases. While many of these proteins seem relatively well characterized, we still cannot account for the strong selective constraints preserving their observed high degree of sequence conservation across major taxonomic groups. Presumably these patterns of conservation contain implicit information regarding still unknown functional mechanisms. To access this information, we recently developed a statistically based approach, called c ontrast h ierarchical a lignment and i nteraction n etwork (CHAIN) analysis [ 3 ], that identifies, categorizes, and statistically characterizes co-conserved patterns in multiple alignments. The power of this approach strongly depends on the quality of the alignment, which thus motivated the development of the theoretical concepts and strategies described here. Aligning distantly related sequences presents unique algorithmic and statistical challenges because such proteins often only share a minimal structural core with sizable insertions occurring between, and even within, core elements. Classical dynamic programming-based multiple alignment procedures typically have considerable difficulty spanning across these insert regions because the log-odds scores associated with weakly conserved core elements are often too low to offset the substantial gap penalties that such insert regions incur. This problem is further exacerbated when core elements contain short insertions or deletions within them. To address this problems, we previously devised motif (or block) based multiple alignment procedures [ 4 - 6 ] that can easily jump over non-homologous insert regions. This approach seems easier to justify than attempting to align regions for which there is no statistical evidence of relatedness. A block based alignment strategy thus seeks to detect islands of subtle sequence similarity within otherwise dissimilar sequences. Fortunately, even when the conserved motifs are very subtle, such a procedure can take advantage of large numbers of available sequences to detect weak, yet statistically significant similarities. Altschul at the National Center for Biotechnology Information (NCBI) likewise sought to address this problem through generalized affine gap costs [ 7 ], but the utility of this approach is unclear, as the NCBI currently does not support any public programs based upon it. The programs MUSCLE [ 8 , 9 ] and MAFFT [ 10 ] also are designed to avoid alignment of non-homologous regions and in other respects are generally superior to more widely used multiple alignment programs, such as Clustalw [ 11 ] and T-coffee [ 12 ]. Because MUSCLE and MAFFT can handle large data sets, we explored the use of these programs for CHAIN analysis (Neuwald, unpublished). Somewhat surprisingly, these failed to achieve the degree of accuracy needed to detect subtle, co-conserved patterns, such as those recently identified and structurally confirmed within P loop GTPases [ 3 ]. We found that, although these programs align regions globally conserved in the sequences well, for several large test sets they fail to accurately align regions conserved only within more closely related subsets. This is, of course, a major drawback to their general application for CHAIN analysis. By contrast, PSI-BLAST [ 1 ], which seems less likely to produce high quality global alignments given its simple alignment procedure nevertheless in many cases does a better job of aligning database sequences relative to the query. Thus PSI-BLAST (albeit with some modifications to improve alignment accuracy [ 3 ]) has turned out to be more generally useful than these other methods for CHAIN analysis, which like PSI-BLAST is query centric. Note, however, that a systematic comparison of various methods within the context of CHAIN analysis has not yet been done. More relevant to our purpose here, another drawback to the use of MUSCLE, MAFFT, and similar programs for CHAIN analysis is that these will align randomly generated sequences – a characteristic incompatible with the statistical basis of CHAIN analysis. MUSCLE and MAFFT perform well on small sets of relatively diverse representative sequences, such as the BALIBASE benchmark sets [ 13 ], because they incorporate heuristics that unfortunately also can compromise statistical rigor and, as a result, confuse random noise with biologically valid homology. Statistically the best alignment for random sequences is the 'null alignment', that is the procedure should leave such sequences unaligned – a property of PSI-BLAST that played a key role in choosing it for CHAIN analysis. To maintain statistical rigor in our formulations here, we will 'let the data speak' by modelling only those characteristics of the sequences that can be justified by the input data. Such an approach cannot be applied, however, to small benchmark alignment sets, because these lack sufficient sequences – less than the number of amino acids whose parameters are being estimated. Thus, while a rigorous statistical approach has severe limitations when applied to small datasets, it works very well when applied to large, diverse sets of distantly related sequences, as demonstrated, for example, by some of our earlier analyses [ 14 - 16 ]. Two other theoretical issues, which are important to the multiple alignment problem, are devising an objective measure of alignment quality and an efficient strategy for finding the best alignments based on this measure. Our previous methods [ 4 - 6 ] addressed these issues using a Bayesian statistical approach for modelling an arbitrary number of multiply aligned ungapped blocks, each of arbitrary length, in conjunction with a Gibbs sampling procedure for exploring the 'space' of all such alignments. Gibbs sampling is a Markov chain Monte Carlo (MCMC) method that iteratively realigns the sequences with probability proportional to how much the model is thereby improved. Theoretically, beginning from an arbitrary starting alignment, this process will ultimately sample alignments according to the posterior distribution defined by our Bayesian model. Exploring the alignment space in this way is more efficient than taking a greedy approach (one that always chooses the transition to the best alignment) because an element of chance allows the sampler to maneuver around locally optimal traps. Within this MCMC sampler we implemented specific operations on the alignments, including those allowing for realignment of a sequence against the alignment model, shortening or lengthening of blocks, and creation of recombinant alignments. Such operations function like catalysts to help the sampler avoid or more quickly escape from local optima. Here we expand on the number of these operations and modify our Bayesian model to allow for short insertions or deletions within blocks. In theory, such an approach could be used to sample representative multiple alignments from the posterior distribution, which is relevant to CHAIN analysis because this could be used to adjust position-specific amino acid frequencies for alignment uncertainty. Doing so for the model and operations described here, however, is non-trivial and thus is a topic for a future publication built upon this one. Our primary objective here is merely to obtain the optimal alignment. Thus we also introduce various annealing-like strategies for luring the sampler toward optimum alignments. These include simulated annealing, which is applied within sampling routines, and other intervention strategies. Our primary motivation for developing and implementing these concepts and strategies is to improve CHAIN analysis, as is illustrated here for G-protein α subunits, which belong to the P loop GTPase class [ 17 ], prolyl endopeptidases, which belong to the α , β -hydrolase fold class [ 18 , 19 ], and transitional endoplasmic reticulum (p97) ATPases [ 20 ], which belongs to the AAA family [ 21 - 23 ] within the AAA+ class [ 14 , 24 , 25 ]. Problem definition The fundamental problem addressed here is to identify the essential features – the common structural core – characteristic of a large set of distantly related proteins. Given an input sequence set, we build a Bayesian statistical model with adjustable parameters to reflect the relationships among the proteins. We also design a stochastic search algorithm, with an MCMC sampler as its backbone, to explore possible alignments and corresponding model parameters in order to find alignment models that best 'explain' the input data. The model parameters specify, for example, the number and lengths of the motifs, their locations within each sequence, the residue frequencies observed at each position in each motif, and other properties (described below). We may thus envision our sampler as searching through a discrete space where each point, corresponding to a particular alignment, has a probability associated with it. The probability function appears fairly smooth inasmuch as nearby points (similar alignments) have roughly comparable probabilities. As the sampler traverses from one point to another, it favors moves toward the better alignments, that is, toward that part of the alignment space with greater posterior probability. Since it is computationally prohibitive for the sampler to consider many transitions at one time, a key design issue is the selection of allowed transitions between points. Results and discussion The block-motif model We first define the alignment model in precise mathematical terms, which provides a scoring scheme that allows us to judge which alignment is better than another. Here, for the sake of conciseness and readability, we will keep the discussion on a conceptual level whenever possible. Interested readers can consult our earlier publications for further details [ 4 , 6 ]. As illustrated in Fig. 1 , this previously described block-based motif model assumes that the aligned core of each protein sequence consists of m co-linear ungapped motifs, of width w 1 ,... , w m , respectively. Each motif is modelled by a position specific frequency matrix Θ i , whereas residues outside the motif blocks follow a common frequency distribution. Independent prior Dirichelet distributions are employed for these frequency parameters. Since both m and the w i 's are unknown, we assume that they are uniformly distributed in a certain range a priori (see [ 6 ] for details). We also employ a "fragmentation model," which allows non-informative aligned columns to be ignored by the motif model. Although we use no explicit gap penalties between motifs, our prior imposes a large penalty on alignments with large m . Let S denote the sequence data and let A denote the motif alignments (which also includes m and the w i 's). Then the posterior alignment distribution is: P ( A \| S ) ∝ P ( A ) ∫ P ( S \| A , Θ) P (Θ) d Θ. Based on this distribution, our algorithm (as implemented in the PROBE program [ 5 ]) attempts to maximize P ( A \| S ), the so-called " maximum a posteriori (MAP)" score. Hidden Markov models for gapped motifs A major drawback of the previous block-based alignment approach is that it disallows insertions or deletions within motif blocks. Here we describe hidden Markov model (HMM) [ 26 , 27 ] structures for insertions and deletions, which will be used by our current algorithm via the operation GAPALIGN (see below). The general architecture for these HMMs is given in Fig. 2 , and detailed descriptions, including the definition for our scoring function g ( A , Λ), are given in Methods. For an intuitive notion of how within-motif penalties influence the total MAP score, consider a gap-opening penalty of say 20 bits (i.e., p = 1/2 20 ) and an extension penalty of 2.5 bits. Then, for example, the overall MAP would need to improve by 25 bits in order to justify a 'surgical operation' on a sequence involving an insertion of three dummy residue (i.e., to 'correct' a deletion in a sequence) or a deletion of three residues (i.e., to 'correct' an insertion in a sequence). The statistical problem is thus that of finding the right penalty so that the sampler only adds insertions or deletions when the data provides sufficient justification. In a Bayesian context, this justification is based on the posterior inference of the overall number of insertions and deletions from what it finds in the aligned sequences. Markov chain Monte Carlo methods The Bayesian analysis described in Methods provides us the posterior distribution of the alignment up to a normalizing constant. Although this distribution defines the answer to our problem, namely inferring the optimal alignment, it is difficult to make sense out of it because of the huge size of the alignment space. Fortunately, recent progress in using MCMC methods for statistical analysis has made it possible to study this function. MCMC methods, of which the Gibbs sampler is a special case, refer to a set of techniques developed by physicists since the 1950s to simulate variables from a given probability distribution up to a normalizing constant. The central idea of these techniques is to evolve a Markov chain, each step of which perturbs the current state (alignment) slightly, with the equilibrium distribution of the chain being the target distribution. A MCMC scheme is usually constructed in two steps: (i) propose a new state according to a certain reversible transition rule, and (ii) accept or reject the proposal according to the probability ratio between the proposed and the current states [ 28 ]. The broad utility and general applicability of these techniques are exemplified and popularized by recent developments in statistics: if one can sample from g ( A , Λ) one obtains a set of "typical" alignments according to the posterior distribution, which provides information regarding the most likely alignment(s) supported by the data and its variability. In practice, however, one may wish to find the optimum of this function and explore only around this optimum considering the difficulty of summarizing a set of distinct alignments in a meaningful way. MCMC is also an important ingredient of an optimization technique termed "simulated annealing" [ 29 ], of which we will develop a variation. A good MCMC scheme should have the following property: (a) its transition rules should collectively allow the sampler to access every point in the space; (b) these transitions should also allow for global changes, such as, for example, recombination between two alignments; and (c) the acceptance rate of these proposals should be reasonable (10~50%). The sections below will focus on designing such transitions for multiple alignment. An algebraic system for touring the alignment space The elementary mathematical operations of addition and subtraction define a means of transitioning between points in the discrete space of natural numbers. "Global" operations, such as multiplication and integer division, allow transitions between more distant points in this space. Likewise, we define both elementary and global operations on multiple alignments as a means of transitioning between points in alignment space. In this case a set of unaligned sequences (termed the null alignment) serves the same role as the natural number zero. Formal mathematical descriptions of the alignment and of certain simple operations are provided in our earlier papers [ 4 , 6 ]. Since the new operations described here involve various combinations of these simple operations, it is straightforward to derive these new operations from the previously published descriptions. There are two issues to consider in the design of multiple alignment operations. First, the reversibility of MCMC algorithms requires that every operation have an "inverse" so that the sampler can readily transit in either direction. Second, to help find the optimal alignment according to our Bayesian model, which is our main objective, annealing techniques and less restrictive acceptance rules should be considered for certain complex operations. By doing so the target alignment distribution has to be distorted to some degree, though the global optimum of the distribution remains the same. All alignments described here are collinear multiple alignments (CMAs), which are defined to contain zero or more motif blocks arranged collinearly in each sequence. Partial or complete deletion of any motif from a particular sequence is modelled by aligning that motif against null residues ('-'), which the sampler may insert anywhere in the sequence. Sequences may also contain more than one repeat of the entire protein domain, each of which is modelled by the full set of motifs. (The identification of repeat domains will be described elsewhere; Spouge and Neuwald, unpublished.) For clarity, we describe operations deterministically, though it should be kept in mind that our sampler applies these stochastically. Elementary operations The HideInsert operation (inverse ShowInsert) is applied to 'surgically' remove a region of the sequence that appears to correspond to a typically short insertion within a conserved motif. This operation thus changes the real sequence into an idealized sequence that, presumably, more closely resembles the canonical characteristics of the protein class. As a result, the sampler needs to maintain both a real and an idealized version of each protein's sequence and to store the operational derivation used to obtain the ideal sequence from the real. Algorithmically it is convenient to deal with insert regions in this way because otherwise the sampler would need to look up the locations of insertions and deletion within each sequence when applying other operations. The FillDeletion operation (inverse UnfillDeletion) likewise converts a sequence that contains a deletion of either part of or all of a motif into an idealized sequence in which the deletion has been filled in with null or 'dummy' ('-') residues. Note that HideInsert and FillDeletion merely define data structure interconversions that allow basic operations, which were initially defined for ungapped motifs, to be efficiently applied to gapped motifs. The Align operation assigns motif positions within a sequence and thereby adds that sequence to the alignment, UnAlign removes the sequence from the alignment. Note that these operations disallow gaps within motif blocks. The AddColumn and DeleteColumn operations add and remove aligned columns, respectively. Note that these operations may add or remove columns internal to a motif as well as at the edges. Moreover, AddColumn may also insert a column an arbitrary number of residues beyond the current edge of a motif. This is important for motif 'fragmentation' [ 4 , 30 ], a procedure that allows certain nonconserved positions inside of a motif to be ignored by the alignment statistical model. Compound operations Elementary operations can be combined in a coordinated manner in various ways to produce compound operations that better facilitate escape from local traps. For example, GapAlign (inverse UngapAlign) combines the row operation Align with the sequence operations HideInsert and FillDeletion in order to add a sequence to an alignment with insertions and deletions. The GapAlign operation is performed using dynamic programming to obtain a gapped alignment of a sequence against a statistical model of the current alignment. The trace back procedure determines how to apply the HideInsert and FillDeletion operations to the true sequence and how the Align operation is then applied to the resultant idealized sequence. We define several compound operations on a motif block: AddBlock, ShiftRight, and TrimRight (with inverses: DeleteBlock, ShiftLeft, and TrimLeft, respectively). Another compound operation, MoveColumn, which transfers a column from one position to another within a block, is its own inverse. Conceptually, AddBlock and DeleteBlock simply iteratively apply the AddColumn and DeleteColumn operations, respectively. Because our motif alignments are collinear, the position of an added block within each idealized sequence must be specified in a manner consistent with this collinear arrangement and, in order to add a new block in this way, the sampler may need to insert null residues at certain positions within some of the idealized sequences. This is an example of operational flexibility. Similar operational flexibility is required for the ShiftRight and ShiftLeft operations, which remove one or more columns from one end and append it to the other end of a motif. TrimRight and TrimLeft allow poorly conserved residues to be trimmed from a motif block based on their relative entropy. These operations thus provide a means to manually edit motif-based alignments as discussed below. Three compound operations involving two motif blocks are: TransferColumn, Splitblock and FuseBlocks. TransferColumn deletes a column from one block and adds it to another block. Splitblock splits a single block into two leaving two contiguous motif blocks in each of the idealized sequences. During future realignment operations the sampler typically induces these abutted blocks to drift apart. Splitblock's inverse operation, FuseBlocks, merges two blocks into one, which typically requires forced realignment of motif positions in each sequence in order to join the blocks together. All such forced realignments are followed by additional optimization via sampling prior to deciding whether to reject or accept this new configuration. We thus typically have to violate the MCMC's acceptance-rejection rule to enable such a move, which distorts the target distribution. The awkwardness of this procedure may be advantageous, however, inasmuch as it forces the sampler out of local traps in alignment space. Fig. 3 illustrates the effect of applying compound operations during Gibbs sampling. Recombinational operations As an aid to locating the optimum alignment, we define recombination operations that combine the best features of two distinct, fairly well refined alignments. These operations require that the sampler first generate a population of fairly well refined alignments starting from distinct, randomly selected points in alignment space. All of these input alignments must, of course, contain the same set of sequences. The Recombine operation must be applied to two alignments that are fairly similar because the sampler needs to locate at least one crossover point between them. A crossover point is a set of positions, one position in each aligned sequence, such that the same set of blocks in the first alignment lie to the left of each of those points, while the same set of blocks in the second alignment lie to the right of each point. Because this requirement often proves difficult to satisfy for every sequence, we define the Recombine operation flexibly by allowing a certain number of sequences to violate this rule. In this case, violating sequences are removed prior to recombination and sampled back in afterwards (using the GAPALIGN operation). The Intersect operation takes as input two distinct alignments and produces a new alignment containing only those aligned columns common to corresponding motifs in both input alignments. More precisely, we first find the common blocks shared by the two alignments, where a common block is defined as two aligned motif blocks (one in each alignment) that overlap within corresponding sequences. To allow for some flexibility, these are defined as blocks for which at least some minimum fraction (say 50%) of the sequences are consistently aligned in both input alignments. (Inconsistently aligned sequences are removed from the alignment prior to performing this operation.) Then, for each pair of common blocks, we find the sub-block shared by both blocks. Next, we create a new alignment containing only these Intersecting sub-blocks. Finally, sequences that were inconsistently aligned between the two starting alignments are sampled back into the resulting alignment. The Intersect operation allows the sampler to be reinitialized starting with a consensus alignment that aligns only those regions with high likelihood scores and eliminates those regions about which the sampler is less certain. Subsequent sampling will then extend these sub-blocks, add new blocks, and explore more extensively the alignment space. Parameter settings for operations There are no absolute rules on how to choose parameter settings for these algebraic operations, such as, for example, the maximum increase in motif length allowed during the MoveColumn operation or the number of disordered blocks to tolerate for the Recombine operation. We find, in fact, that it often matters little which settings are used and the slight degree to which it does matter depends on the particular protein class being analyzed. As a result, any biologically reasonable parameter settings work well. For example, since weakly conserved motifs are never a hundred residues long, motif blocks typically should be limited to no more than, say, fifty residues in length. Nevertheless our algorithm tolerates unreasonable parameter settings, because then it either simply rejects the corresponding alignment space transitions (though with some degradation in performance) and/or learns to avoid applying useless operations through its memory module, as described below. High level sampling strategies Having specified various operations on the alignment space, we now need to specify when and how often to apply them, as well as how to escape from local traps and thus to most rapidly converge on an optimum or nearly optimum alignment. Providing the sampler with a memory Since some of the alignment operations are computationally expensive, it would be helpful to avoid applying them over and over again when this proves to be unfruitful. For example, if the sampler has already converged on the correct number of motifs, applying the AddBlock operation may be a waste of time. On the other hand, we don't want to eliminate any operation entirely, as at some point it may be useful. To do this we define both short-term and long-term sampling memories. The short-term memory allows a rapid response to sudden changes while the long-term memory adds stability so that the sampler does not over respond to short term trends. Details are given in Methods. Simulated annealing with a thermostat Let the target alignment distribution be denoted generically as π ( X ). As the sampler converges on near optimum alignments, typically it has difficulty 'dropping' into the global optimum of π ( X ) because the chance of selecting the highest probability alignment is still very small due to the sheer number of near optimum alignments. This is true for the same reason that the most likely outcome of obtaining exactly 5,000 heads and 5,000 tails in 10,000 flips of a fair coin is extremely unlikely. A standard way around this problem is to take power of π ( X ) to some exponent, renormalizing it and using the "powered-up" distribution, denoted as π T ( X ) ∝ π 1/ T ( X ) with the "temperature" parameter varying from a very large value to near-zero, for sampling. This procedure is a key component of simulated annealing [ 29 ], which has the same effect on sampling as lowering the temperature has on annealing of single stranded DNA into double stranded DNA in solution. By 'cooling' the system (i.e., letting T → 0), we raise the probability of high-density points and lower the probability of low-density points, so as to allow the best alignment to win out over alignments that are nearly as good. If the temperature is lower too abruptly, however, the sampler may get trapped in a sub-optimum alignment, so that the annealing strategy needs to be devised carefully. We have built a 'thermostat' into the sampler that keeps track of variations in the ( T = 1) probability densities of the sampled alignments. If the variance of log π ( X ) in a given number K of consecutive iterations at a given temperature is below a certain threshold (so that the posterior probabilities barely change), the sampler may be stuck in a (presumably local) optimum, and the thermostat raises the temperature a bit. On the other hand, if the log π ( X ) are varying wildly and, in particular, if they are greatly diverging from the best (i.e., highest probability) alignment found thus far, then the sampler may be wandering away from near optimum alignments and the thermostat lowers the temperature. This approach thus attempts to keep the sampler just above its 'glass transition temperature' [ 31 ], designated T g . Details are given in Methods. Since there are no absolute criteria for determining whether the sampler has actually found the optimum alignment, it is necessary to devise heuristics for terminating the computation. We retain the same criterion used in earlier Gibbs samplers, such that if the alignment fails to improve after a specified number of sampling cycles, then the program stops and returns the best alignment found. Since picking the right number of cycles depends heavily on the number and nature of the input sequences (as well as the user's patience), the user can modify this parameter. As an alternative strategy, two or more programs may also be run in parallel until they both converge on the same alignment. Progressive refinement strategy When painting a picture, it is helpful to first draw a rough sketch so that details will end up in the right place relative to each other. Similarly the sampler uses the following progressive refinement strategy to avoid being too "shortsighted." There are five stages to this strategy. In the first stage, the sampler applies the Align operation, which aligns the sequences against contiguous ungapped blocks; it also applies compound ungapped motif operations. The initial numbers of block motifs and columns in each block are sampled from binomial distributions with means between roughly 5~10 blocks and 10~30 columns each, respectively. In the second stage, which is introduced after the sampler begins to converge on a local optimum under the ungapped block-motif model, elementary and compound column operations are introduced, which allow these ungapped blocks to 'fragment', thereby permitting nonconserved columns to be ignored by the alignment model (mathematical details are found in [ 6 ]). Recombination operations are also applied during and after this stage. In the third stage, the GapAlign operation based on a simple gapped sampling procedure [ 14 ] with very conservative gap penalties is introduced, which allows the sampler to add short gaps within motif blocks and to delete part or all of a block. In the fourth stage, the number of blocks is fixed (although other operations are retained) and recombination and simulated annealing procedures are used to help guide the sampler into a (hopefully) global optimum. These first four stages are implemented in the program GISMO (see below). A fifth stage, which is implemented in the program GARMA (see below), recombines a set of alignments independently found by GISMO and optimizes the recombinants using a GapAlign procedure based on the HMM model described above. (Here we apply another annealing strategy, termed prior annealing, where early on low HMM gap penalty priors are used to introduce gaps more liberally, and later high HMM gap penalty priors are used to eliminate less convincing gaps.) GapAlign sampling is performed by Viterbi alignment of the sequence against the HMM where the HMM emission and transition probably parameters are sampled from the posterior distribution. Afterwards the resultant alignment is either rejected or accepted based on our new scoring function g ( A , Λ). Manual application of alignment operations Despite attempts to codify and fully automate optimization of a multiple sequence alignment, the algorithm may still create an alignment model that lacks certain properties observed to be biologically important for a particular class of proteins. Take the situation, for example, where a motif, which occurs as a single block in most of the proteins, is split in two by a sizable insertion in other proteins and where the sampler, due to the a priori parameter settings chosen before the analysis, fails to split this motif into two blocks. In this case, a biologically more meaningful alignment may be achieved by manually intervening to split this ungapped region (followed, ideally, by additional optimization via MCMC sampling perhaps using adjusted prior probabilities). To accommodate such tweaking, we thus allow manual application of various operations. We find that splitting and trimming of aligned blocks are particularly helpful in this regard. Such manually modified alignments then may be reintroduced into a population of similar alignments for recombination and selection via our genetic algorithm [ 5 ] followed by further optimization. Implementation and examples The theoretical concepts and strategies just described were implemented in the programs GISMO ( Gi bbs-like s ampling with m ultiple o perations), GARMA ( g enetic a lgorithm for r ecombinant m ultiple a lignment) and GAMBIT ( g apped a lignment with M CMC- b ased i ndel t empering). GARMA recombines the output alignments provided by GISMO and then applies simulated annealing strategies on the recombinants. GAMBIT performs on a single alignment the same optimization procedures that GARMA performs on recombinants. Manual application of alignment operations may be performed using another program, TweakAln. These programs along with sample alignments are available from the authors. Multiple alignment of thousands of sequences in this way may take substantial time (e.g., overnight on a 10-processor Linux cluster), but this is not critical because, once performed for a particular protein class, such an alignment can be updated readily by seeding the sampler with a previously optimized alignment. Here we apply these programs to several large protein classes within the context of CHAIN analysis, which is our primary reason for generating such alignments. Application to CHAIN analysis CHAIN analysis both decomposes into distinct categories and quantifies the sequence constraints associated with conserved patterns in a multiple alignment. This yields evolutionary clues regarding the underlying structural mechanisms presumably preserving these patterns. Aspects of these mechanisms can be inferred by comparing category-specific selective constraints with known structures of members of the protein class being investigated, as illustrated in three recent publications [ 3 , 32 , 33 ]. 'Contrast hierarchical alignments', such as are shown in Figs 4 , 5 , 6 , are the primary output from CHAIN analysis. In constructing such an alignment, three sets of related sequences are multiply aligned: (i) a 'displayed set', (ii) a 'foreground set', which is a superset of the displayed set, and (iii) a 'background set'. The displayed set corresponds to the aligned sequences of interest within the foreground set (i.e., only the alignment for these sequences is actually shown). The foreground set corresponds to the sequences whose selective constraints are being measured. These are not shown explicitly, but rather are merely represented by conserved patterns and residue frequencies shown below the displayed alignment (as in Fig. 4A ). The original CHAIN analysis procedure uses a modified version of the PSI-BLAST algorithm to align these sequences. Here these PSI-BLAST alignments are compared with motif-based foreground alignments created using GISMO, GARMA, GAMBIT, and TweakAln. CHAIN analysis measures selective constraints in terms of the difficulty of randomly drawing the amino acids observed at a particular position in the foreground alignment from the distribution at that position in the background alignment. In the examples here, unless specified otherwise, the overall frequency of amino acids generally observed in proteins serves as an implicit background set at each position. Foreground positions with compositions closely resembling the background presumably are subject to little or no selective constraints, while positions with compositions strikingly different from (i.e., that contrast with) the background are subject to strong constraints. In Figs 4 , 5 , 6 these constraints are displayed in the histograms above the alignments. G α and P loop GTPases We first examine in this way G protein α subunits. G proteins [ 17 ] are heterotrimers, consisting of an α , a β and a γ subunit, that mediate transduction of extracellular signals to the cellular interior. As do many members of the P loop GTPase class, the G α subunit functions as a binary switch that is turned on by binding GTP in response to the signal and thereby relays this information to downstream components of the pathway. This switch is turned off by hydrolysis of GTP to GDP, an event mediated by GTPase activating proteins (GAPs). G α subunits are unique among such GTPase switches inasmuch as their GAP domain is contained within the G α polypeptide chain itself rather than existing as a distinct protein. This unique arrangement presents particular difficulties for CHAIN analysis because, during subsequent iterations, the PSI-BLAST algorithm tends to slightly overextend the alignment beyond G α 's region of homology to other P loop GTPases and into the C-terminal region of the GAP domain. As a result, the foreground patterns for the Walker A motif are mistakenly aligned against the C-terminal end of the GAP domain (Fig. 4B ). By contrast, the Gibbs sampler avoids this misalignment problem because it can readily jump over the internal GAP domain (Fig. 4A ). This thus illustrates how our motif-based approach avoids a serious problem encountered by PSI-BLAST. α , β -hydrolase fold enzymes Similar misalignment problems may be encountered between motif regions even when the aligned proteins lack large inserts. This is seen, for example, when aligning α , β -hydrolase fold proteins [ 18 , 19 ], which correspond to a large class of enzymes possessing a catalytic triad (typically consisting of a serine, an aspartate and a histidine) at their active sites. These three residues are involved in an electron transfer mechanism and thus are generally very highly conserved, despite the often very weak pairwise similarity between many members of this class. CHAIN analyses of prolyl oligopeptidases reveals that our motif-based alignment assigns very strong selective constraints to all three of these catalytic residues, the aspartate and histidine of which are shown in Fig. 5A . This is as expected, because conservation of one member of the catalytic triad is highly correlated with conservation of the other two, as the α , β -hydrolase electron transfer mechanism requires all three residues. In contrast, the PSI-BLAST alignment assigns a strong selective constraint to the catalytic serine (not shown in Fig. 5 ) but much weaker constraints to these other two catalytic residues (Fig. 5B ). This is because the PSI-BLAST algorithm finds it much easier to correctly align the catalytic serine but, due to weak sequence similarity, often either misaligns or fails to extend the alignment into the C-terminal region of this domain. (The fraction of sequences that fail to align with this region is indicated near the bottom of Fig. 5B ). Thus our motif-based approach again provides a better measure of the selective constraints acting on these residues. P97 an AAA+ ATPase Improved identification of a short insertion within a motif by our approach is illustrated through CHAIN analysis of p97, a transitional endoplasmic reticulum AAA+ ATPase (recently reviewed in [ 20 ]). AAA+ ATPases are a large and diverse class of chaperone and chaperone-like proteins [ 14 , 24 , 25 ]. They are characterized by the presence of one or more AAA+ modules, each of which consists of an α , β -fold domain, which it shares with other P loop NTPases, followed by a helical bundle domain. P97 contains two AAA+ modules, designated D1 and D2; our analysis was performed on the D1 module, whose structure is known [ 34 ]. These AAA+ modules often associate to form homohexameric complexes such that a prominently conserved arginine (R362A in Fig. 6 and 7 ) and a conserved acidic residue (D333 in Figs 6 and 7 ) in one module are positioned near a Walker B conserved acidic residue (E305 in Fig. 7 ) and a bound ATP-Mg 2+ in an adjacent AAA+ module. When our motif-based approach was applied (with prior annealing) to AAA+ ATPases (Fig. 6A ), it introduced within the Box VII motif of the p97 D1 module a two-residue insertion (most often a phe-gly; F360-G361 in Figs 6 and 7 ) immediately before a prominently conserved arginine (R362). By contrast, the PSI-BLAST alignment tends to misalign this region and, consequently, obscures both the two-residue insertion and the prominence of the conserved arginine (as indicated by the histogram height over this position; see Fig. 6B ). The phenylalanine within this insert forms a CH- π interaction with an alanine (A409 in Figs 6 and 7 ) within the adjacent AAA+ module's three-helix bundle domain. Notably, an arginine often occurs at this alanine position in related AAA+ modules and is believed to sense bound ATP in the adjacent AAA+ module. (The region containing this arginine thus is termed the 'sensor II region'.) PSI-BLAST again does a poorer job aligning this sensor II arginine against A409 of p97 compared with our motif-based method. The improved motif-based alignment thus better reveals how the p97 AAA+ D1 module presumably utilizes an alternative configuration for sensing and responding to bound nucleotide relative to typical AAA+ modules (Fig. 7 ). In particular, two highly conserved p97 family-specific features – namely the phe-gly insertion, which is highly conserved in eukaryotes though replaced by a pro-gly in eubacteria and archaea, along with a third well conserved arginine directly preceding this insert (R359 in Figs 6 and 7) – are likely to perform an important role associated with p97's unique cellular function. Conclusions With a view to improving alignments for CHAIN analysis, we have enhanced our earlier motif-based methods by developing (i) a HMM for insertions and deletions within motifs, (ii) an expanded algebraic system of operations on multiple alignments and (iii) various annealing and sampling strategies that facilitate rapid convergence on optimum or near optimum alignments. Furthermore, our approach, due to its rigorous statistical basis, fills a gap left by current multiple alignment methods inasmuch as it aligns only those characteristics of the input sequences that may be justified statistically. Thus it is useful for statistical analysis of conserved patterns in multiple alignments. Our statistical model likewise provides objective criteria for evaluating curated alignments, thereby guiding manual application of various operations. In the future, our MCMC sampling methods could be used to estimate alignment uncertainties, which will be useful for estimating background amino acid frequencies for CHAIN analysis. These approaches also serve as a starting point for further enhancements that integrate MCMC sampling, HMM and PSI-BLAST methods, which, based on our earlier analyses [ 16 ], seem likely to improve both alignment accuracy and search sensitivity. When this motif-based approach was applied to CHAIN analysis of families belonging to large and diverse protein classes, we found numerous examples, three of which are described here, where this does a better job of revealing subtle, biologically important sequence features than does PSI-BLAST. This is in large part due to the ability of our statistical model and sampling strategies to find weakly conserved islands of homology within a sea of essentially nonconserved regions. While this motif based approach will not become the default method for CHAIN analysis – especially considering that PSI-BLAST alignments also may be optimized using these approaches – it, nevertheless, often more accurately aligns very distantly related sequences and thus can provide a better measure of selective constraints in this situation. Methods HMM architecture We model gaps within motif blocks through the HMM shown in Fig. 2 . The corresponding probability matrix for transitions between HMM states internal to the i th motif is: where 1 ≤ i ≤ m and 1 ≤ x < w i and where M, I, and D denote match, insertion and deletion states, respectively. The probability matrix for transitions between motifs is: where 1 < i < m and where these transitions each emit a string of zero or more residues. Note that the contribution to the log-posterior probability of the lengths of these strings and of their emission probabilities (as well as those of M and I states) are specified by our ungapped statistical model [ 6 ], upon which this HMM is based and thus are unspecified by the HMM. Note also that the treatment we provide here easily can be generalized to cases where transitions I → D and D → I are allowed or where gap penalties are motif-specific. Statistical inference of indel penalties For a given alignment A , let f ( A ) be its log-posterior probability as in [ 6 ]. If we allow insertions and deletions within motifs, then each motif i within each sequence S k is associated with a "path" through the HMM indicating its alignment against motif model Θ i . Let the collection of these paths be Λ. Next, we denote the total number of transitions of type M → M, M → I, ..., by N mm , N mi , N md , N im , N ii , N dm , N dd . It then follows that the likelihood of the gap parameters is with independent prior distributions ( α o , β o , 1 - α o - β o ) ~ Dirichlet( a o , b o , n m - a o - b o ), α e ~ Beta( a e , n i - a e ), and β e ~ Beta( b e , n d - b e ), where a o , b o , n m , a e , n i , b e , n d are prior pseudo counts given by the user. The corresponding maximum likelihood estimates (MLEs) are The joint posterior distribution for the alignment and gap parameters is g ( A , Λ, α , β ) ∝ P ( S \| A , Λ) × P ( A ) Λ h (Λ \| α , β ) P ( α , β ), where P ( S \| A , Λ) × P ( A ) is computed the same way as in the original block-motif model [ 6 ], and P ( α , β ) = Dirichlet( a o , b o , n m - a o - b o ) × Beta( a e , n i - a e) × Beta( b e , n d - b e ). Given the alignment Λ, we have the conditional posterior distribution Sampling on this distribution can be performed by drawing the following random variables: Parameter collapsing For computational efficiency, we integrate out the α and β to get This gives rise to a new posterior g ( A , Λ) with h (Λ) replacing h (Λ \| α , β ) P ( α , β ) in our previous formula [ 6 ] and frees us from having to fix or update the gap parameters. This also allows us to determine the optimum posterior gap penalties based on the sequence data. Prior specifications Suppose that we expect to see one insertion in every K 1 residues and one deletion in every K 2 residues. Then we set and set n o to reflect the strength of this conviction. We suggest using priors reflecting conservative gapping where, for example, K 1 = K 2 = 1000 and n o = n M N , where is the total number of match positions in all of the motifs and N is the total number of aligned sequences. For gap extension prior probabilities, if one expects to see an average insertion length of L 1 , and deletion length of L 2 , then we let We set the prior pseudo counts n 1 to be equal to the total number of expected insertions within motifs n M / K 1 . Likewise, n 2 is set equal to the expected number of deletions n M / K 2 . In order to have different gap parameters for each motif, one need only keep specific counts of insertions and deletions for each motif, as the formula h (Λ) then applies to each motif individually, and we only need to multiply these h ( ) functions together when computing the total 'penalty'. The sampler's memory For long-term memory we monitor among the sampler's previous iterations the number of times N o (where typically, N o = 25) that a type " o " operation has been applied and the number of times n o that it was "successful" (i.e., resulted in an increase of the posterior probability). The same is done for short-term memory except that in this case we monitor the number of short-term successes m o over M o previous applications (where typically M o = 5). At the next iteration, we then assign a probability of applying this operation, where w s ≥ 0 and w l ≥ 0 are the weights given to the short and long-term memories, respectively, and where w p ≥ 0 specifies the minimum frequency at which this operation is applied. Typically, we set w s = w l = 1 and 0.2 ≤ w p ≤ 0.66, so that operations that previously proved to be unfruitful will only be performed about one-tenth to one-third as often as those that always yield improvements in the alignment. The sampler's thermostat We define an intuitive sampling temperature T ' = 300/ T and, thus, π T ' ( X ) ∝ π 300/ T ( X ). On this 'pseudo-degrees-Kelvin' scale sampling from the true distribution π ( X ) (i.e., 300°) corresponds to sampling at 'room temperature'. After a period of sampling at room temperature until 'convergence', which is defined by the sampler's failure to improve the MAP after a specified number of iterations, simulated annealing is initiated. During this stage, whenever the probability densities of the sampled alignments averaged over say 20 iterations fluctuate by more than some maximal value, say Δlog( p) ≥ 50 nats, the temperature is lowered by 1–5°. If, on the other hand, the probability densities of the sampled alignments fluctuate on average less than some minimal value, say Δlog( p) ≤ 5 nats, the temperature is raised by say 1°. (The precise parameters used are not critical and may depend somewhat on the input sequence set.) This period of thermostatic sampling is again applied until convergence. Authors' contributions AFN developed the algorithmic strategies and early ad hoc approaches conceptually similar to the statistically rigorous procedures described in Methods, which were designed by JSL. AFN implemented the procedures and performed the sequence analyses. Both authors wrote and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538276.xml
544949	Time and frequency domain methods for quantifying common modulation of motor unit firing patterns	Background In investigations of the human motor system, two approaches are generally employed toward the identification of common modulating drives from motor unit recordings. One is a frequency domain method and uses the coherence function to determine the degree of linear correlation between each frequency component of the signals. The other is a time domain method that has been developed to determine the strength of low frequency common modulations between motor unit spike trains, often referred to in the literature as 'common drive'. Methods The relationships between these methods are systematically explored using both mathematical and experimental procedures. A mathematical derivation is presented that shows the theoretical relationship between both time and frequency domain techniques. Multiple recordings from concurrent activities of pairs of motor units are studied and linear regressions are performed between time and frequency domain estimates (for different time domain window sizes) to assess their equivalence. Results Analytically, it may be demonstrated that under the theoretical condition of a narrowband point frequency, the two relations are equivalent. However practical situations deviate from this ideal condition. The correlation between the two techniques varies with time domain moving average window length and for window lengths of 200 ms, 400 ms and 800 ms, the r 2 regression statistics ( p < 0.05) are 0.56, 0.81 and 0.80 respectively. Conclusions Although theoretically equivalent and experimentally well correlated there are a number of minor discrepancies between the two techniques that are explored. The time domain technique is preferred for short data segments and is better able to quantify the strength of a broad band drive into a single index. The frequency domain measures are more encompassing, providing a complete description of all oscillatory inputs and are better suited to quantifying narrow ranges of descending input into a single index. In general the physiological question at hand should dictate which technique is best suited.	Introduction Common oscillations in neurophysiological activity in the human motor system have been well documented. During voluntary muscle contraction, the human central nervous system drives motor neurons at a range of frequencies which cause common modulations in the firings of these neurons. These drives are reviewed in [ 1 ] and [ 2 ] where they are summarized into four broad frequency ranges: (1) A low frequency drive at around 1–3 Hz (2) A neurogenic component of physiological tremor that occurs between 5–12 Hz and is likely to have both spinal and supraspinal components. (3) A corticospinal drive in the beta (15–30 Hz) range (4) A corticospinal drive in the low gamma (30–60 Hz) range, that increases in importance with stronger contractions and is called the Piper rhythm. There are two distinct approaches toward the identification of these drives. The majority of the literature has examined common modulation to motor units using frequency domain methods. This methodology was first introduced by Rosenberg and colleagues [ 3 ] and applied by Farmer and colleagues [ 4 ] who used coherence analysis to identify both a significant low frequency and beta-band association between motor unit firings in the 1–12 Hz and 15–30 Hz frequency ranges respectively. Subsequently, coherence analysis has become an established technique to study bivariate motor system measurements and a number of works have used this to investigate corticomuscular interactions [ 5 - 8 , 1 ]; tremor [ 9 ]; aging [ 10 ]; oscillatory drive in Parkinson's disease [ 11 , 12 ]; dystonia [ 13 ]; stroke [ 14 ]; and cortical myoclonus [ 15 ]. A separate body of literature has focused specifically on the low frequency common drive. This drive was first identified by De Luca and colleagues [ 16 ] who demonstrated that the firing rates of concurrently active motor units (MUs) were modulated in a highly interdependent manner. They low-pass filtered the impulse trains corresponding to MU firing times to obtain the time-varying mean firing rates which they high-pass filtered at 0.75 Hz. They then performed a time domain cross-correlation analysis between pairs of zero-mean signals representing the fluctuations in mean firing rates. Peaks occurring near the zero lag location in the normalized cross correlations implied that those firings rates were essentially simultaneously modulated with virtually no time delay. This phenomenon was termed 'common drive' to indicate a common excitation to the motor neuron pool that results in concurrent fluctuations in the firing rates of motor units from the same pool. A number of subsequent studies have utilized this technique to investigate the relationship of this drive to handedness [ 17 - 19 ]; different proprioceptive conditions [ 20 ]; exercise [ 21 ]; task and disease [ 22 ]; and aging [ 23 ]. These works have established this time domain method as an accepted means of quantifying the common low frequency modulation of MU firings. In a recent review [ 1 ], it was suggested that the low frequency common drive first identified by De Luca and colleagues [ 16 ] using time domain methods is essentially the same low frequency drive as detected by Farmer and colleagues [ 4 ] using frequency domain methods. In this paper we explore the relationship between the two techniques using mathematical and experimental approaches. Analytic methods Frequency domain methods: Coherence The coherence between two zero-mean stationary random processes x 1 ( t ) and x 2 ( t ), at frequency f , is defined as: where ( f ) is the cross spectral density and ( f ) and ( f ) are the auto spectral density functions of x 1 ( t ) and x 2 ( t ) respectively. The coherence function is a complex quantity and its squared magnitude provides a bounded measure of linear association between the two series, taking on a value of 1 for a perfect linear relationship and a value of 0 to indicate that the series are uncorrelated. In practice, we are often limited to a single time-limited realization of each random process and hence it is necessary to estimate the magnitude squared coherence, , by windowing the time series to obtain multiple sections as follows: where * denotes complex conjugation, N is the number of data segments employed and X 1 n ( f ) and X 2 n ( f ) are the discrete Fourier transforms of the n th data segments of x 1 ( t ) and x 2 ( t ). This estimate is biased and its probability density function for non-weighted and non-overlapping windows has been analytically determined [ 24 ]. This may be used to derive the value of the estimated coherence, with a particular probability of occurrence, α , that would be obtained when the true value is zero. Any value exceeding this level is considered to be unlikely to be a false indication of coherence with ( α × 100) % confidence. This confidence level is given by [ 24 , 3 ] E α = 1 - (1 - α ) 1/( N -1) (3) The resolution of the coherence estimate is determined from the inverse of the length of the windowed sections, i.e., for a 2 s window, the coherence resolution will be 0.5 Hz. Figure 1 depicts a typical coherence plot computed for the spike trains of two MU's and the associated 95% confidence level. The coherence plots reveal the bandwidth and values of significant coherence for the given resolution. Results from coherence analyses are usually quantified in terms of either the peak value and its frequency or the frequency range of significant coherence. In Figure 1 there is significant coherence between 0.5 and 3.5 Hz and the peak value of coherence is 0.46 and occurs at 1.5 Hz. Figure 1 Example of a magnitude squared coherence plot. Magnitude coherence between two motor unit spike trains recorded from the FDI muscle. The dashed horizontal line indicates significance at the 95% confidence level. Significant coherence occurs between 0.5 and 3.5 Hz with the peak coherence of 0.46 occurring at 1.5 Hz. Time domain methods: Cross correlation The cross correlation between two zero-mean stationary random processes x 1 ( t ) and x 2 ( t ) is defined as: where E [·] is the estimation operator. Assuming ergodicity, for single time-limited realizations of each random process, this is determined using the integral: where * denotes complex conjugation and τ is the time lag between the signals. The Fourier transform of the cross correlation function, defines the cross-spectrum, ( f ). Cross correlation functions are unbounded measures and are typically normalized by the values of the autocorrelations at zero lag to bound the estimate between -1 and 1. The autocorrelation functions are the time domain equivalent of the auto power spectra and their value at zero lag represents the total energy in the signal. The normalized and bounded measure is known as the cross correlation coefficient, , which provides a measure of the linear association between the two signals at a given time lag and is given by: The original method employed by De Luca and colleagues [ 16 ] represents the time series as a binary pulse train with ones corresponding to the firing times of the MUs and zeros comprising the remainder of the signal. A moving average window is then used to smooth these binary signals, which is analogous to filtering the time-series with a low-pass filter. These smoothed signals are depicted in figure 2a . A high-pass filter is then used to remove the mean bias of the signal as 'shown in figure 2b . The filtered signals are subsequently cross correlated and an index obtained from the peak value of the normalized cross correlation function within a specified lag window. Here we term this index the time domain common drive coefficient. Figure 2 Example of the construction of a low frequency common drive plot. A low-pass, moving average Hanning window filter of length 400 ms was applied to two motor unit spike trains recorded from the FDI muscle. (a) A 5 s epoch of the time-varying smoothed firing rates; (b) the high-pass filtered version of the smoothed firing rates shown in (a); and (c) the low frequency common drive coefficient function between two motor unit spike trains. This results in an effective pass band of 0–5 Hz. The peak of the signal is 0.75 and occurs at a lag of 3.5 ms. Figure 2c depicts the normalized cross correlation function for the same two motor unit spike trains used in Figure 1 obtained using a moving average Hanning window of 400 ms duration and high pass filtering at 0.75 Hz. The function is displayed for lags up to ± 400 ms and the peak of the signal indicated at a lag of 3.5 ms. The time domain common drive coefficient is measured as 0.75. Relationship between cross correlation and coherence The cross correlation coefficient is related to coherence using a similar analysis to Gardner [ 25 ] as follows: We begin with the real and stationary signals x 1 ( t ) and x 2 ( t ), where x 2 ( t ) = α x 1 ( t - τ 0 ) + n ( t ) is a scaled and time-delayed version of x 1 ( t ), with additive uncorrelated, zero mean noise, n ( t ). The cross-correlation function is given by since the cross correlation with the noise is zero everywhere. The cross spectrum is given by Assume that the signals y 1 ( t ) and y 2 ( t ) result from passing the signals x 1 ( t ) and x 2 ( t ) respectively through a tunable narrowband bandpass filter with transfer function denoted by H ( f ): where and Δ are the center frequency and bandwidth of the ideal bandpass filter. The cross-correlation function between filtered signals y 1 ( t ) and y 2 ( t ) is given by: where ( f ) is the cross power spectral density function of y 1 ( t ) and y 2 ( t ). The cross power spectrum may be written in terms of its magnitude and phase, . Since for stationary, real signals, the autocorrelation is real and even and hence, ( f ) is real, the phase of the cross spectrum is given by (equation 8): Thus replacing ( f ) with \| H ( f )\| 2 ( f ) in equation 10 we get since is real for real x 1 ( t ) and x 2 ( t ). Similarly for the autocorrelation functions we get and Thus as Δ → 0 we obtain the expression for the normalized cross correlation function as: The peak of the cross-correlation function occurs at the time delay, τ = τ 0 . Thus Thus we see that the peak of the normalized cross-correlation function between two signals after ideally bandpass filtering to contain a single frequency, is identical to the magnitude of the coherence function of the original signals at the frequency of the filter. The phase of the coherence function is the same as the phase of the cross-spectrum and provides the time delay. For a less ideal filter that spans several frequencies the relationship is less precise and may be derived as follows: Let W ( f ) be the new filter transfer function and thus the normalized cross-correlation function is: where f 1 and f 2 are the cut-off frequencies of the filter. Thus when multiple frequencies are present, this may be thought of as taking the weighted summation of the cross-correlation functions at each frequency present and normalizing this by the product of the weighted summations of the autocorrelations across all frequencies present. The more narrow band the filter used, the more similar the time domain correlation and frequency domain magnitude coherence measures. As the filter encompasses a greater range of frequencies, measures from the two methods will increasingly deviate. The low frequency time domain method employed by De Luca and colleagues [ 16 ] utilized a moving Hanning window as a low pass filter. The cut-off frequency of the filter is dependent on the time constant of the filter which is typically 400 ms [ 16 , 21 ] but values up to 0.95 s have also been used [ 26 ]. However different window lengths will modify the relationship between this time domain measure and the coherence function. The effect of varying window length may be illustrated by obtaining an expression for the filter transfer function. The equation for the Hanning window is given as: where τ is the length of the window. The discrete Fourier transform of this is given as (Kay, 1988): where Figure 3 depicts the transfer function power spectrum (\| W ( f )\| 2 ) for τ = 200, 400 and 800 ms. The figure clearly demonstrates that as the length of the analysis window decreases, the bandwidth of the filter increases. Therefore the only information that can be ascertained with shorter windows is that the frequency of the common modulating input lies somewhere within the frequency range specified by the window. Longer windows result in a better correlation with coherence values at lower single frequencies (close to zero), while shorter windows lump into a single value a weighted expression of the coherence values in the frequency range which they span. Figure 3 Magnitude squared spectra of Hanning window filters. Magnitude spectra of the transfer functions of Hanning window filters for three different time constants, τ = 200 ms (dotted line), τ = 400 ms (dashed line) and τ = 800 ms (solid line). As the time constant increases the bandwidth of the filter decreases and its magnitude increases. Experimental methods In this section we demonstrate the relationship between time and frequency domain based methods to estimate the common modulating drive using empirical data. The methods are applied to data collected during isometric contractions of the First Dorsal Interosseous muscle at 20% of maximal effort. Two contractions where the activities of 4 and 5 MUs were identified yielded a total of 16 pairs of coactive MUs. The periods of concurrent activity of these MU pairs ranged between 30 s to 1 minute and were further divided into pairs of non-overlapping 10 s intervals resulting in a total of 50 pairs of 10 s long spike trains. Each method was applied to these spike train pairs and the correlation between the results yielded by the two methods were investigated as discussed below. The time domain method was used to estimate low frequency common drive according to the method described by De Luca and colleagues [ 16 ]. Three different time domain estimates were formed by smoothing the spike trains using Hanning windows of length 200, 400 and 800 ms respectively. These smoothed firing rate signals were then digitally high pass filtered with a low frequency cut-off of 0.75 Hz using a third order Butterworth filter to remove the mean bias discharge rates. The cross correlation coefficient function of these high pass filtered records was then obtained and the peak value of this function within ± 50 ms of the zero time lag was recorded and termed the time domain common drive coefficient. The coherence analyses were performed in a similar manner to the procedure of Rosenberg and colleagues [ 3 ] for point process data. The spike trains were represented as binary pulse trains with ones corresponding to the firing times of the MU's and zeros comprising the remainder of the signal. Fourier transforms of these trains were obtained for each appropriately windowed section and then averaged according to equation (2). However where Rosenberg and colleagues [ 3 ] do not use overlapping or tapered data windows, we used overlapping, tapered Hanning windows of 2048 ms to optimize the variance and bias of the estimate. With any non-parametric spectral estimation technique, there is a trade-off between the variance and both the bias and resolution of the estimation. A window size of 2048 ms, gives a frequency resolution of 0.49 Hz, which is adequate to discriminate frequencies for our purposes. However, when analyzing 10 s of data using 2048 ms non-overlapping windows, only 5 different records are available and this small number of records will increase the variance of the estimate. Furthermore rectangular windows introduce an estimation bias due to the effect of their sidelobes. These concerns may be reduced by using the Welch periodogram method which uses tapered windows (to reduce spectral leakage and therefore the estimation bias) and overlapping windows (to increase the total number of windows and hence reduce the variance). The minimum variance for this method is obtained using an overlap of 62.5% [ 27 ]. The frequency corresponding to the first zero-crossing of the Hanning filter was obtained according to equation (20) and the peak value of the coherence in the range between 0.75 Hz and this frequency was recorded. A linear regression between the time domain common drive coefficients and corresponding frequency domain peak coherence values was performed to determine whether a linear relationship between the two indices existed. The regression r 2 values are reported at a significance level of p < 0.05. Results and discussion Figure 4a,4b,4c displays the regression between the low frequency time domain common drive coefficients for Hanning windows of length 200, 400 and 800 ms and peak low frequency coherence. All regressions are significant at p < 0.05 and the r 2 statistics are 0.56, 0.81 and 0.80 respectively. A unitary slope line is displayed in the figure and this describes the theoretical relationship between the two indices. These results indicate that for the larger 400 ms and 800 ms windows, the time domain method is more closely correlated with the coherence estimate, with the 400 ms window yielding a marginally better fit. The data for the smaller 200 ms window exhibits a consistent bias, with the coherence estimate larger than the time domain common drive estimate, whilst the 400 ms and 800 ms windowed data are more evenly distributed around the unitary slope line, indicating less bias. There are a number of possible factors that could contribute to the observed mismatches between the two methods. Figure 4 Regression plots for low frequency common drive time versus frequency domain techniques. Regression plots for low frequency common drive time versus frequency domain techniques. Three different moving average Hanning windows were used to low pass filter the time series for the time domain method. The time constants for the filters are as follows: (a) τ = 200 ms, (b) τ = 400 ms, (c) τ = 800 ms. All regressions are significant at p < 0.05 and the r 2 statistics are (a) 0.56, (b) 0.81 and (c) 0.80. The unitary slop line is indicated in the figures as a dashed line and represents the ideal mathematical relationship. As demonstrated in Figure 2 , the cross-correlation peak can occur at lags slightly different than zero. A time delay or misalignment has been shown to introduce a bias into the coherence estimate that is proportional to the delay and coherence magnitude and inversely proportional to the FFT epoch duration [ 24 ]. However for delays of the order of magnitude of ± 50 ms and FFT lengths of approximately 2 s, this type of bias is very small and unlikely to account for the observed differences between the time and frequency domain estimates. The use of a short duration window in the time domain method results in the inclusion of multiple frequencies in the time domain correlation estimation according to equation (18). The bandwidths of descending oscillatory drives may be variable. Thus when the descending drive occupies a narrow bandwidth and the time domain window includes a greater range of frequencies than this bandwidth, this will bias the time domain estimate to be lower than the peak coherence value as is the case in figure 4a . Alternatively should the drive span a broader bandwidth, the time domain measure would encompass all the correlated frequencies into a single value and would thus be different than the value obtained from any single peak coherence frequency. This idea is illustrated in Figure 5 where a typical coherence plot is displayed. Superimposed on this are vertical lines representing various moving average filter cut-off frequencies. The 0.75 Hz high pass cut-off frequency is also displayed. Thus from the figure we see that in this case the coherence occupies a fairly broad bandwidth from 1–5 Hz, peaking at 1.5 Hz. The cut-off frequency of the 200 ms filter is approximately 10 Hz and thus the time domain estimate will include coherence values at all these frequencies which would make it significantly different from the peak coherence. The 400 ms and 800 ms windows would better correlate with the peak coherence frequencies and the 400 ms window would provide a better overall index encompassing the full bandwidth of the drive. However, if the middle peak at 8 Hz were stronger and actually the main peak, the wider time windows would miss it altogether. This emphasizes the importance of a priori knowledge in choosing the appropriate time windows in the time-domain based method. Therefore in summary, the time domain measure is more effective in quantifying a range of frequencies into a single index and the peak coherence estimate is better at representing the coherence at any single frequency. Figure 5 Magnitude squared coherence between two motor unit spike trains recorded from the FDI muscle. Magnitude coherence between two motor unit spike trains recorded from the FDI muscle. The vertical dotted lines from left to right represent the cut-off frequencies of the 0.75 high-pass filter, and the 800 ms, 400 ms and 200 ms moving average low-pass filters. The peak coherence occurs at 1.5 Hz. Coherence estimates are typically formed from data records of around 1–5 minutes in length [ 4 , 28 , 29 ] or from pooled coherence measures of repeated trial measurements [ 30 ]. This increases the number of non-overlapping windows in the calculation, thereby reducing the variance of the coherence estimate. Non-overlapping, rectangular windows are traditionally preferred due to the clear relationship with significance levels. Overlapping, tapered windows will allow coherence to be estimated from shorter data segments and parametric techniques, in particular multivariate autoregressive (MAR) methods are suggested for the analysis of very short duration data segments [ 31 ]. When using short records of data (<5 s), the coherence estimates are likely to be significantly biased. However, the time domain method is more robust for such short data lengths and would therefore be preferred in these situations. The time domain method uses a high pass filter to remove the mean bias from the smoothed signals, whereas the frequency domain coherence method simply subtracts the mean component of the signal prior to forming the estimate. Although similar, these two methods are not identical and may further explain some of the variation between the time and frequency domain techniques. A further possibility is to employ a low order polynomial detrending technique instead of high pass filtering or subtracting the mean. In general, a visual examination of the smoothed firing rate signals would indicate whether this would be necessary. It is straight forward to quantify any time delay using the time domain technique. Although this is also possible with the frequency domain technique, this delay information is incorporated in the phase of the estimate and is therefore 2 π periodic and would thus yield the same result for integer multiples of delay. For significant coherence present over a band of frequencies, Mima and colleagues [ 32 ] suggest a constant phase shift plus constant time delay regression model to compute time delays from coherence estimates. However for narrow band descending drives, the time domain technique provides a clearer estimate of any time delay. It should be noted that the time domain technique may be generalized to cover any arbitrary frequency. This would necessitate bandpass filtering the signals to the frequency range of interest, removing the mean trends and then evaluating the cross-correlation function. Although this requires a priori knowledge of the drive bandwidth, this method would then be able to provide a single index to quantify a particular bandwidth. Conclusions Two separate bodies of literature offer techniques to estimate band limited common oscillatory input to motor neurons. These techniques are based in either the time or in the frequency domain. Indices derived from both measurement techniques are well correlated with each other and in the theoretical limit, the techniques are shown to be mathematically equivalent. However, for practical purposes there are a number of minor discrepancies which may favour the use of one particular method for a given application. The time domain method offers greater resolution in time (the latency of the correlations are easily revealed) at the expense of the requirement of a priori knowledge of the bandwidth of the common modulating drive. On the other hand the frequency domain technique reveals information regarding the frequency distribution of the common modulating drives but it is more difficult to obtain accurate estimates of the coherence with short signal lengths as well as of the exact delay. Time domain methods of estimation are preferred for short data segments and are well suited to quantifying the strength of a broad band drive into a single index. This proves useful in quantitative, comparitive analyses of the common behavior of MUs such as statistical tests that investigate the effects of aging or disease. Frequency domain measures tend to be more encompassing as they provide a complete description of all common oscillatory inputs. This facilitates qualitative analysis of distribution of coherence across frequencies and hence leads to a better understanding of the nature of the common inputs. They are well suited for estimation from large data segments, that may be assumed to be stationary, and are better able to quantify narrow ranges of descending inputs into a single index. Thus the selection of one technique over another should be dictated by the nature of the physiological question to be addressed.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544949.xml
544961	Australian primary care policy in 2004: two tiers or one for Medicare?	The recent primary care policy debate in Australia has centred on access to primary medical (general practice) services. In Australia, access is heavily influenced by Commonwealth Government patient rebates that provide incentives for general practitioners not to charge copayments to patients (bulk billing). A steady decline in key access indicators (bulk billing) has led the Howard Government to introduce a set of changes that move Medicare from a universal scheme, to one increasingly targeted at providing services to more disadvantaged Australians. In doing so, another scene in the story of the contest between universal health care and selective provision in Australia has been written. This paper explores the immediate antecedents and consequences of the changes and sets them in the broader context of policy development for primary care in Australia.	Introduction Primary health care and community care can be thought of as a set of health programs and services. Most discussions of the primary health and community care services sector suggest that it has the following characteristics: (1) It is the first point of contact with the health system. This may occur through general practice, community health services, and pharmacies. There is also some overlap between primary care and hospital emergency departments, particularly for less complex and intensive presentations. (2) Services are provided in community and ambulatory settings and at home. (3) There is an emphasis on continuing relationships between service providers and consumers over extended periods of time. (4) Services have a more comprehensive and holistic approach. (5) There is an emphasis on early detection and illness prevention services such as maternal and child health programs and population health programs including health promotion. Primary health and community care is the most visible and commonly used part of the health system. In 1999–00 the Commonwealth provided approximately $6 billion through the Commonwealth Medical Benefits and Pharmaceutical Benefits Schemes. States and Local Government provided approximately $1.8 billion for 'community and public health' which includes allied health, counselling, nursing and a range of primary and secondary prevention and health promotion programs. The Commonwealth, through direct outlays ($6 million) and private health insurance premium rebates ($97 million), also provided $103 million for dental services, with the States and Territories contributing $305 million. This does not include the substantial funds committed to the various forms of community support for people with disabilities, chronic illness and mental illness [ 1 ]. Primary health and community care services face unique challenges. Over the past three decades primary health services have come under significant pressure to address a more complex and diverse range of community needs. Deinstitutionalisation, the introduction of new health and information technologies, the increasing prevalence of chronic disease and more general social and economic trends have had a significant impact on primary health and community support services. This has resulted in concerns about the equity, quality and efficiency of services and programs. Arguably, there is a need for a national primary health care policy in Australia. One that would address system integration, care pathways and team practice, work force development, payment arrangements, governance, performance management and accountability. However, current Commonwealth reforms are focused on important, but relatively, narrowly focused solutions to the decreasing affordability and access for general medical services. This article focuses primarily on the recent debate that surrounds this issue. Recent Policy Medicare is a Commonwealth Government, tax funded, social insurance scheme that provides rebates for general (primary) and specialist medical services and optometry. In Australia, it is the principal national program for ensuring equitable access to primary medical services. Over the last two years there has been a fiercely contested debate about the future of Medicare. Medicare was introduced to provide universal access to affordable medical care. Up until recently Medicare simply provided a rebate of 85% of the Commonwealth Government determined schedule fee for medical and diagnostic services. Practitioners were free to charge patients a copayment as well. Where they did not apply a copayment, they could bill the Commonwealth for the rebate and receive bulk payments direct from the Commonwealth for these services, thereby avoiding administrative costs and delay. This payment method, which became known as bulk billing, ensured that services were effectively free to the patient at the point of service. Medicare has been very successful, particularly for general practice services. It is strongly supported by the Australia community because it provides affordable access to medical services. Despite historical resistance by the Australian Medical Association, it has been widely supported by practitioners because it provides them with a universal, simple and predictable revenue stream. Bulk billing increased steadily from the introduction of Medicare in 1984/85 to approximately 70% in the mid-1990s. Bulk-billing rates for GP services have generally been about 10% higher than the overall bulk-billing rates over the last decade, reaching a plateau of about 80% in the mid-1990s. They declined significantly after 2000. Average GP bulk billing fell to 68% by September 2003. GP bulk billing rates are now similar to the overall CMBS bulk billing rates for all services [ 2 ]. As bulk billing rates declined, disquiet and concern about access to medical services rose amongst stakeholder interests. More generally, the overall decline in bulk billing came on top of considerable disparity in equity of access between rural and urban settings. Bulk billing rates in inner city areas with high per capita GP ratios were 30% higher than those in rural settings with low per capita ratios. A number of remote rural areas had difficulty attracting any GPs at all. Analysis of the reasons for the decline in bulk billing and the disparities between rural and metropolitan settings suggest a strong relationship between the supply of GPs and the capacity to charge copayments and the importance of the steady decline in the relative value of Commonwealth rebates for GP services over time. There is considerable evidence that GPs manage demand for their services to maintain their income [ 3 ]. As the number of GPs increased with introduction of Medicare, particularly in inner city areas, per capita utilisation of GP services increased sharply. Average out of pocket costs for patients fell as bulk billing increased. The Commonwealth effectively provided the 'floor price' for services in areas of high supply and high competition leaving patient throughput rates as the primary means for increasing revenue. Urban areas with greater levels of disadvantage had higher bulk billing rates and shorter consultation times. In higher socio-economic status areas, where patients have a greater capacity to pay, there were lower bulk billing rates and longer consultation times. In the decade to 2003, changes to GP training, migration and demographic ageing lead to a stabilization and decline in the supply of GP services. Over the same period, the relative value of Medicare rebate income for GP services fell by about 10 percent compared with average weekly ordinary time earnings. A decline that was probably even greater when compared to specialist incomes. In response, GPs began to experiment with price increases (co-payments) to improve their relative incomes [ 4 ]. Interestingly, bulk billing rates in relatively under supplied rural settings remained relatively stable at about 50% of consultations. This is about the level of consultations one would expect people on low incomes, who are eligible for concessional welfare benefits, to use. A finding that suggests that patient capacity to pay sets a 'floor' bulk billing rate at about this level [ 4 ]. It is worth noting that many of these effects were predictable from the reforms to General Practice introduced in the early 1990s. In particular, tight management of GP supply through changes to training programs and restrictions on overseas trained medical practitioners were introduced in order to reduce growth in aggregate Medicare expenditure. However, the reforms recognised that a move away from fee for service payment would also be required in the longer term. To this end a Better Practice Program to pay GPs on a per capita basis was introduced. Over time, it was intended that a significant proportion of Medicare payments would be made by practice based, per capita payments. Progressively this would have allowed a shift toward more comprehensive, integrated practice and a greater focus on quality and preventive services. However, while the supply of GPs was successfully constrained, per capita payments remained a marginal component of the payment system. In response to concerns about the fall in the bulk billing rate, the Commonwealth Government proposed a "Fairer Medicare" package in April 2003. The package introduced a participating practice scheme. GP practices that agreed to charge a no gap fee to concessional patients were to be eligible for increased Medicare rebates for these patients. The level of the proposed increase for the rebate was $1 in metropolitan city practices, $2.95 in non-metropolitan city practices, $5.30 in rural centre practices, and $6.30 in outer rural and remote areas. Participating practices were to continue to have the capacity to determine fees for non-concession cardholders, including the option of bulk billing. However, if they chose not to bulk bill these patients, they were to be able to charge the patient the co-payment and claim the Medicare rebate direct from the Health Insurance Commission through HIC online billing facilities. Effectively, non-concession cardholders were to be charged a gap payment thereby avoiding the transaction costs involved in claiming a rebate through the Medicare scheme themselves. A new MBS safety net was to be available for those covered by concession cards with out-of-pocket costs greater than $500 in a calendar year. Charges in excess of the scheduled fee were to be included, as were the costs of specialist and diagnostic services. Eighty per cent of out-of-pocket costs above the $500 threshold were to be met through this safety net. Private health insurers were to be able to offer insurance coverage for the cumulative cost of out-of-hospital medical services over $1,000 for a family in a calendar year. This included costs above the scheduled fee across a range of out-of-hospital services, including GP and specialist consultations and diagnostic tests. The Commonwealth estimated that insurance products for this coverage were likely to cost around $50 per year for families, and the 30% private health insurance rebate was to apply to these products. The Government's package also included proposals to introduce additional medical school places, additional GP training places, additional nurses and allied health professionals in general practice, and measures for veterans. The Fairer Medicare package resulted in considerable debate and criticism, much of which was considered by the Senate Select Committee on Medicare [ 5 ]. In part the Committee concluded that: • Equitable access to general practice services regardless of income or geography is fundamental to good health care. • GP income from bulk billing had not kept pace with increases in average weekly ordinary time earnings and this had contributed to declining bulk billing rates and increased out of pocket charges. • Shortages in the supply of GPs are emerging as result of compositional changes in the workforce, changes in practice patterns and population ageing. • The Commonwealth's 'Fairer Medicare' proposals were inconsistent with the principles of Medicare. • The differential rebate payments for concessional patients were unnecessary because these patients were already largely receiving bulk billed services • The introduction of the new safety net arrangements creates a two tier system of access to GP services It became apparent that the Senate would not pass the legislation required to enact the Commonwealth's package. Consequently, the Commonwealth presented its 'MedicarePlus' extensions and revisions to the original proposal in November 2003 [ 6 ]. This package was passed by the Australian Senate with the support of four independent Senators. The MedicarePlus proposals dropped the participating practice scheme. Instead the Commonwealth proposed to increase the rebate for all concessional patients by $5 in metropolitan areas and $7.50 in remote, rural and regional areas (including the State of Tasmania). The increased rebate was also extended to children under 16. The safety net provisions were modified to provide an 80% rebate for out of hospital medical costs for concessional patients and those whose income fell below specified tax thresholds after $300 of out of pocket expenses and after $700 for the remainder of the population. Other aspects of the original proposal were largely retained and extended. These included training places for GPs, medical graduates and nurses. Additionally, it was proposed to introduce a Medicare Benefits Schedule item for nursing support in general practice and improved internet access and online billing for GPs. MedicarePlus also provides rebates for up to five allied health consultations delivered to patients with a chronic condition or complex care needs, for and on behalf of a GP. Similarly, dental treatment care plans will be funded for these patients where they have significant dental problems that exacerbate their condition. The total estimated cost for MedicarePlus to 2006/07 was estimated at $2.85 billion. Policy Analysis Initial reactions to the Commonwealth's proposals were mixed. A number of patient and provider groups have criticized the new arrangements as undermining the principle of universality that underpins Medicare. Criticisms have also focused on the narrow focus of MedicarePlus on fees for general practitioners. More specifically, MedicarePlus is likely to have differential effects on affordability and access to GP services in rural and metropolitan settings. In metropolitan settings, the introduction of a $5 differential rebate for bulk billing concessional payments is sufficient to increase net GP incomes to about the AWOTE relativities that applied prior to the decline in bulk billing. However, with current levels of bulk billing still at over 65% in metropolitan areas, virtually all concessional patients are already bulk billed. The proposal is therefore subject to substantial dead weight loss. No incentives to bulk bill non concessional patients (other than non concessional children aged less than 16) are included. In metropolitan areas, the gap between the average Medicare rebate and the average patient billed service is around $13 for patients not covered by the differential rebate, compared to $8 for concessional patients and children under 16. Within system constraints, GP incomes are optimized by bulk billing concessional patients and children under 16 and charging copayments for other patients. Doing so also largely addresses patient capacity to pay issues. In the absence of major changes to supply or GP costs, it is therefore likely that bulk billing rates in metropolitan areas will continue to decline over time, until they stabilise at around the level of services for concessional patients plus non concessional children under 16 (around 60% in metropolitan areas, assuming children under 16 have average population consultation rates). The proportion of children is highest in outer metropolitan regions and lowest in the inner city. Bulk billing rates have declined most in outer metropolitan areas. It is therefore plausible that this measure will have the greatest differential impact in outer metropolitan regions. In rural settings, where there are GP shortages, the differential rebate (which is higher than in metropolitan settings) could substantially increase GP incomes. However, GP supply factors ensure GPs have considerable capacity to increase copayments within the limits of patient capacity to pay. In general, bulk billing rates in rural settings are now at or below the consultation rate for concessional patients. The new arrangements are therefore likely to protect bulk billing rate for concessional patients and are likely to see the rate increase to the level of concessional consultations (around 55 – 60%). The effect of differential rebates for non concessional children under 16 on overall bulk billing rates is less clear in rural settings. With the increased rebate, there remains a gap of approximately $5.50 between the new rebate and the average patient billed service. As for metropolitan settings, there are no additional incentives to bulk bill other non concessional patients. Given the greater capacity to charge copayments, this measure may be less successful in encouraging bulk billing than in metropolitan areas. The safety net provisions in MedicarePlus have significant inflationary potential for out of hospital medical service fees. Concessional patients and those who qualify for Family Tax Benefit A are eligible for an 80% rebate on out of hospital costs once they incur $300 of out of pocket costs. There is no cap on the rebate under the safety net. Average out of pocket costs for patient billed GP services are currently about $13. The safety net is therefore reached in 20–25 consultations. The safety net provisions will be invoked more quickly when specialist medical practitioners, diagnostic imaging services and pathology are required. Average copayments are two or three times higher for specialist medical practitioners than for GPs. Effectively, the introduction of the safety net removes constraints on medical practitioners associated with concerns about patient capacity to pay. This introduces moral hazard for practitioners and consumers. Practitioners have incentives to increase their fees and provide more services than necessary knowing the safety net will protect patients. Patients have incentives to consume more services than are necessary because they are effectively insured by the safety net. However, the initial threshold and value of copayments act as balancing disincentives for utilization. Clearly if the initial threshold and the copayments were less, the hazard would be greater and vice versa. This trade off is likely to impact differently depending on need, capacity to pay and supply factors. For example, there may be paradoxical adverse effects for patients with significant ongoing health costs who are currently bulk billed because GPs and specialists have concerns about their capacity to meet aggregate out of pocket costs over time. This is particularly true for aged pension recipients with chronic illness. With the introduction of the safety net, the potential for incurring unmanageable costs is significantly reduced and therefore bulk billing rates for this group may decline. Whether effects like these are experienced in practice will depend on factors such as the real value of GP rebates, patient need, capacity to pay, GP supply and regulatory constraints. Overall, the design features of the Howard Government recent changes to Medicare are intended to, and will produce a two tier system. Access to primary medical services for people on low incomes will be relatively well protected, but those above the income threshold will see a steady decline in bulk billing and an increase in out of pocket costs for these services. Additionally, the poorly designed safety net will have inflationary consequences. Future Directions The policy and political contest around Medicare has an extended pedigree. The conservative Liberal/National Party Coalition has long held the position that government should primarily provide health services for those who are unable to provide for themselves and that those who are able to make their own way should do so, particularly by taking out private health insurance. From this perspective the role of government is to provide an appropriate regulatory environment, incentives and sanctions to take up private insurance and a targeted safety net for the disadvantaged. Their preferred model was developed and refined in the 1950s and 1960s during the period of the Menzies Government and reintroduced in stages during the late 1970s and early 1980s by the Fraser Government [ 7 ]. On the other hand, the Australian Labor Party has advocated tax funded, universal access to publicly funded health care provided on the basis of need, rather than capacity to pay. The Whitlam Government settled the basic architecture of Labor's approach in the early 1970s. Setting aside the brief flirtation with a national community health program for primary care, it established a universal system of public hospital access through the States and a tax funded, social insurance scheme to underwrite equitable access to medical and related services. Notwithstanding Howard Government claims of strong support for Medicare, the pendulum has now swung a considerable distance back toward the traditional Liberal/National Party preferred model. If history is a guide, now that the incentives to take out private health insurance and the safety net is in place, the next steps are regulatory mechanisms to exclude higher income earners from accessing publicly funded health services. While debates about access and equity are critical, they are only part of the overall picture. Recontesting the basic access and equity principles of the health system every decade or so misses a number of important emerging issues. There is now emerging evidence that closer integration of clinical decision-making and purchasing for enrolled populations in primary care settings through funds pooling and local agreements and contracts has the potential to increase innovation, reduce costs and improve outcomes. These principles are being explored or actively implemented in a number of countries comparable to Australia, including the United Kingdom and New Zealand [ 8 ]. There is clearly a need to reconsider the development of a national policy for primary health and community support services. Such a policy might include the following elements to address the issues which have been discussed above: • National primary health and community care goals and objectives. For example, these might broadly set out equity, efficiency and quality criteria for the Australian primary health and community support system. • National performance indicators. For example, these indicators could be used to report on and benchmark the quality, access, efficiency and utilisation of the primary health and community support system and its impact on acute, sub acute and residential care. • Population based planning, allocation and monitoring. For example, funding allocation models and system governance arrangements based on the health care needs of geographically defined residential populations (e.g. Divisions of General Practice, Area Health Authorities, Districts, Primary Care Partnerships) that promote continuity of care and service integration could be considered. • Coordinated service pathways for health issues and conditions. For example, consistent best practice models linking prevention, early intervention, primary care, acute care, rehabilitation and community support should be developed for all major chronic diseases, mental illness and alcohol and drug problems. • Payment systems. For example a program to develop integrated payment models and systems for primary and community support services could be established and linked to Commonwealth/State agreements (e.g. AHCAS, HACC) and own purpose funding streams. This might include consideration of capitated, case based, and contract funding to replace or compliment existing arrangements for primary care services. • National workforce planning and analysis for primary health and community support services. • A national evaluation, research and development program in primary health and community support services. • National planning and priority setting processes for primary health and community care to ensure greater alignment of Commonwealth and State priorities.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544961.xml
439008	A New Role for a Synaptotagmin Protein in Calcium-Dependent Exocytosis	null	The hardest working molecules in cell biology, proteins abound in a dazzling variety of shapes and sizes to carry out an equally impressive array of tasks. Many proteins function within the cell, while others get shipped out to locations both near and far. Specialized organelles within the cell package the traveling proteins into cargo containers called vesicles. Vesicles move through a highly regulated transportation system until they reach their ultimate destination, either inside or outside the cell, and release their cargo. When vesicles fuse to the plasma membrane and release their cargo outside the cell the process is called exocytosis. Exocytosis allows macromolecules to leave the cell without compromising the structural integrity of its membrane. “Professional” secretory cells specialize in producing copious quantities of their protein product and sending vesicles packed with their customized issue to the plasma membrane. Once there, the vesicles wait for a signal to fuse with the membrane. The signal most often comes in the form of a transient and localized increase in calcium ion levels. Over a decade ago, researchers discovered that calcium-triggered exocytosis also occurs in “nonprofessional” secretory cells. In these cells, the process was thought to be important for healing ruptured plasma membrane, though the identity of the vesicles responsible remained unknown. It has been suggested that these vesicles are lysosomes, enzyme-filled organelles that break down waste and extracellular debris ingested by the cells. Sanford Simon, with his colleagues Jyoti Jaiswal and Norma Andrews, previously confirmed that calcium does specifically trigger exocytosis of lysosomes in the “nonprofessional” secretory cells. Calcium-triggered exocytosis is thought to require the services of a family of proteins called synaptotagmins. But the fifteen members of the synaptotagmin family diverge from this job description in various ways, calling its role into question. Synaptotagmin VII (Syt VII)—the synaptotagmin member expressed on the lysosomes—is present in most tissues in organisms ranging from worms to humans. This protein functions in processes requiring lysosomal exocytosis and during invasion by trypanosome parasites, such as the one that causes Chagas disease. In this issue of PLoS Biology , Jaiswal, Simon, and colleagues investigate the molecular mechanisms underlying calcium-triggered lysosomal exocytosis, focusing on the role of Syt VII. The researchers took cells from two lines of mice: one lacked the functional Syt VII protein and the other produced normal levels of Syt VII. First they labeled the surface and interior cavity of lysosomes in these cells with fluorescent tags and triggered an increase in cells' calcium level; then they watched the behavior of single lysosomes releasing their contents in real time. Most of the lysosomes from normal cells released only a portion of their contents and had very small fusion pores that remained open to the outside of the cell for only a short time. Interestingly, while proteins in secretory vesicle membranes typically diffuse into the plasma membrane during exocytosis, proteins in the lysosomal membrane stayed near the site of fusion. In the cells from Syt VII–deficient mice, Simon and colleagues discovered, to their surprise, that this protein isn't necessary for calcium-triggering of lysosomal exocytosis. Calcium-triggered exocytosis not only occurred in these cells, it happened more rapidly than in normal cells. Plus most ofthese deficient lysosomes fused completely and their membrane proteins fully diffused into the plasma membrane. Simon and colleagues argue that these results show that Syt VII restricts rather than facilitates lysosomal exocytosis. It does so by limiting the formation and size of fusion pores and by preventing lysosomal membrane proteins from integrating into the plasma membrane.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC439008.xml
516799	Hormonal Regulation of Plant Growth and Development	Besides environmental factors, plant growth depends upon endogenous signals. Bill Gray examines what these hormonal signals are and how they act to regulate many aspects of growth and development.	Phytohormones: What Are they? Plant growth and development involves the integration of many environmental and endogenous signals that, together with the intrinsic genetic program, determine plant form. Fundamental to this process are several growth regulators collectively called the plant hormones or phytohormones. This group includes auxin, cytokinin, the gibberellins (GAs), abscisic acid (ABA), ethylene, the brassinosteroids (BRs), and jasmonic acid (JA), each of which acts at low concentrations to regulate many aspects of plant growth and development. With the notable exception of the steroidal hormones of the BR group, plant hormones bear little resemblance to their animal counterparts ( Figure 1 ). Rather, they are relatively simple, small molecules such as ethylene gas and indole-3-acetic acid (IAA), the primary auxin in the majority of plant species. The concept of plant hormones originates from a classical experiment on phototropism, the bending of plants toward light, carried out by Charles Darwin and his son Francis in 1880. The Darwins were able to demonstrate that when oat seedlings were exposed to a lateral light source, a transported signal originating from the plant apex promoted differential cell elongation in the lower parts of the seedling that resulted in it bending toward the light source. This signal was subsequently shown to be IAA, the first known plant hormone. Figure 1 Chemical Structures of the Plant Hormones A partial list of the responses elicited by each hormone is provided below. Ethylene gas promotes fruit ripening, senescence, and responses to pathogens and abiotic stresses. IAA (an auxin) regulates cell division and expansion, vascular differentiation, lateral root development, and apical dominance. Cytokinins are adenine derivatives first identified by their ability to promote cytokinesis. JA is a volatile signal that modulates pollen development and responses to pathogen infection. The BRs regulate cell expansion and photomorphogenesis (light-regulated development). GAs are diterpenoid compounds that promote germination, stem elongation, and the induction of flowering. ABA promotes seed dormancy and is involved in several stress signaling pathways. What Do They Do? Virtually every aspect of plant growth and development is under hormonal control to some degree. A single hormone can regulate an amazingly diverse array of cellular and developmental processes, while at the same time multiple hormones often influence a single process. Well-studied examples include the promotion of fruit ripening by ethylene, regulation of the cell cycle by auxin and cytokinin, induction of seed germination and stem elongation by GA, and the maintenance of seed dormancy by ABA. Historically, the effects of each hormone have been defined largely by the application of exogenous hormone. More recently, the isolation of hormone biosynthetic and response mutants has provided powerful new tools for painting a clearer picture of the roles of the various phytohormones in plant growth and development. How Do They Work? Plant biologists have been fascinated by the regulatory capacity of phytohormones since the time of their discovery, and the notion that hormone levels or responses could be manipulated to improve desired plant traits has long been an area of intense interest. Perhaps the best-known example of this is the isolation of dwarf varieties of wheat and rice that led to the “green revolution” in the second half of the 20th century, which is credited with saving millions of people around the globe from starvation. These dwarf varieties have shorter stems than wild-type, making these plants less susceptible to damage by wind and rain. The molecular isolation of these “dwarfing genes” has revealed that they encode components of the GA biosynthesis and response pathways ( Peng et al. 1999 ; Sasaki et al. 2002 ). To elucidate the molecular mechanisms underlying phytohormone action, several researchers have utilized the genetically facile model plant Arabidopsis thaliana to isolate mutations that confer altered response to applied hormone. Molecular and biochemical analysis of the gene products defined by these mutations, coupled with expression studies aimed at identifying the downstream target genes that mediate hormonal changes in growth and development, has begun to unlock some of the mysteries behind phytohormone action. While no hormone transduction pathway is completely understood, we now have a rudimentary understanding of many of the molecular events underlying hormone action. Several reviews covering the individual hormone pathways in greater detail have recently been published ( Turner et al. 2002 ; Gomi and Matsuoka 2003 ; Himmelbach et al. 2003 ; Kakimoto 2003 ; Dharmasiri and Estelle 2004 ; Guo and Ecker 2004 ; Wang and He 2004 ). Common Themes Regulation by proteolysis has emerged as a resounding theme in plant hormone signaling. The ubiquitin-mediated degradation of key regulatory proteins has been demonstrated, or is at least likely, for all of the phytohormone response pathways ( Smalle and Vierstra 2004 ). In the case of auxin, the response pathway is normally subject to repression by a large family of transcriptional regulators called the Aux/IAA proteins ( Figure 2 ). These proteins dimerize with members of the auxin response factor (ARF) family of transcription factors, thus preventing ARFs from activating auxin-responsive genes ( Tiwari et al. 2004 ). Upon an auxin stimulus, an SCF (SKP1/Cullin/F-box protein) ubiquitin ligase ( Deshaies 1999 ) containing the TIR1 F-box protein ubiquitinates the Aux/IAA proteins, marking them for degradation by the 26S proteasome thereby de-repressing the response pathway ( Gray et al. 2001 ). The hormone promotes the Aux/IAA–TIR1 interaction; however, the molecular mechanisms behind this regulation are unclear. Most yeast and animal SCF substrates must be post-translationally modified, usually by phosphorylation, before they are recognized by their cognate F-box protein. Despite numerous efforts to identify auxin-induced modification of Aux/IAA proteins, no such signal has been discovered, raising the distinct possibility that auxin uses a novel mechanism to regulate SCF–substrate interactions. Figure 2 The Ubiquitin-Mediated Proteolysis of Aux/IAA Proteins Regulates Auxin Response (A) Wild-type Arabidopsis thaliana and the axr2-1 mutant. axr2-1 is a dominant gain-of-function mutation in an Aux/IAA gene that confers reduced auxin response. The mutant axr2-1 protein constitutively represses auxin response because it cannot be targeted for proteolysis by the SCF TIR1 ubiquitin ligase. The effect of the mutation on AXR2 stability is shown in a pulse-chase experiment (inset). Wild-type and axr2-1 seedlings were labeled with 35 S-methionine and AXR2/axr2-1 protein was immunoprecipitated either immediately after the labeling period (t = 0) or following a 15-minute chase with unlabeled methionine (t = 15). (B) A simplified model for auxin response. In the absence of an auxin stimulus, Aux/ IAA proteins inhibit ARF transcriptional activity by forming heterodimers. Auxin perception (by an unknown receptor) targets the Aux/IAA proteins to the SCF TIR1 complex, resulting in their ubiquitination and degradation, thereby de-repressing the ARF transcription factors. Among the ARF targets are the Aux/IAA genes themselves, which produce nascent Aux/IAA proteins that restore repression upon the pathway in a negative feedback loop. Ethylene and cytokinin are both perceived by receptors sharing similarity to bacterial two-component regulators. Common in prokaryotes, but apparently restricted to plants and fungi in eukaryotes, these modular signaling systems involve a membrane-bound receptor containing an intracellular histidine kinase (HK) domain ( Wolanin et al. 2002 ). Ligand binding activates the kinase, resulting in autophosphorylation and initiation of a series of phosphotransfer reactions that culminates with the activation of a response regulator protein that functions as the effector component of the pathway. Cytokinin signaling appears to largely follow this paradigm ( Kakimoto 2003 ). Ethylene response, however, appears more complex ( Guo and Ecker 2004 ). Ethylene is perceived by a family of five receptors. ETR1 and ERS1 contain a consensus HK domain, however, the HK domains of ETR2, ERS2, and EIN4 are degenerate and lack elements necessary for catalytic activity. This fact, together with studies of “kinase-dead” mutants of ETR1 , suggests that HK activity is not required for ethylene response. Mutations that abolish ethylene binding in any of the five receptor genes are dominant and confer ethylene insensitivity, indicating that the receptors function as negative regulators of the ethylene pathway. Genetic and molecular studies have positioned these receptors upstream of the Raf-like MAP kinase kinase kinase, CTR1, which interacts with the receptors and also acts as a negative regulator ( Figure 3 ). The integral membrane protein, EIN2, and the transcription factors EIN3 and EIL1 are positive regulators of ethylene signaling downstream of CTR1. Current models propose that hormone binding inactivates the receptors, thus resulting in down-regulation of CTR1 activity. Since the identification of CTR1, biologists have speculated that a MAP kinase cascade may be involved. Only recently, however, have putative MAP kinase kinase and MAP kinase components of the ethylene pathway been identified ( Chang 2003 ). Interestingly, these kinases appear to positively regulate ethylene response, suggesting that CTR1 must inhibit their function. If so, this would represent a novel twist on the traditional MAP kinase signaling paradigm. Precisely how the ethylene signal is transduced to the EIN3 and EIL1 transcription factors remains unclear. However, the recent finding that ethylene stabilizes these transcription factors, which are targeted for degradation by an SCF complex in the absence of ethylene, clearly indicates a role for the ubiquitin pathway ( Guo and Ecker 2003 ; Potuschak et al. 2003 ). One of the known targets for EIN3 is the ERF1 transcription factor, which activates several genes involved in a subset of ethylene responses. Figure 3 A Model for the Arabidopsis Ethylene Response Pathway Ethylene is perceived by a family of two-component receptors containing a consensus (unshaded) or degenerate (shaded) HK domain (H). Three of the receptors also contain a C-terminal receiver domain (R). The receptors negatively regulate ethylene response together with CTR1 in a complex on the endoplasmic reticulum membrane. Perception results in reduced receptor and CTR1 activities and activation of a MAP kinase kinase, which transmits the signal through the EIN2 membrane protein, ultimately resulting in the activation of a transcriptional cascade in the nucleus. The EIN3 and EIL1 transcription factors regulate primary response genes including ERF1 , which activates a subset of secondary ethylene-induced genes involved in defense responses. EIN3/EIL1 abundance is regulated in an ethylene-dependent manner by SCF complexes containing F-box proteins encoded by the ethylene-induced genes EBF1 and EBF2 . Positive- and negative-acting components of the pathway are indicated in green and red, respectively. Solid lines indicate regulation that is likely to be through direct interactions. Dotted lines indicate speculative interactions based on genetic studies. Signal Integration and Combinatorial Control Long ago, plant physiologists noted the apparent antagonistic interactions between some of the phytohormones, such as between auxin and cytokinin in the regulation of root–shoot differentiation and between GA and ABA in germination. Other processes are synergistically regulated by multiple hormones. While it has long been obvious that hormones do not function in discrete pathways, but rather exhibit extensive cross-talk and signal integration with each other and with environmental and developmental signaling pathways, the molecular basis for such coordinated regulation has been unclear. Several recent findings have begun to elucidate the molecular details of some of these events. One example of such signal integration was recently described for the ethylene and JA pathways ( Lorenzo et al. 2003 ). Genetic studies had previously implicated both hormones as important regulators of pathogen defense responses, as well as of the wounding response and other stress-related pathways. Additionally, microarray analysis has identified a large number of genes that are responsive to both hormones. The ERF1 transcription factor was recently found to be an intersection point for these two signaling pathways ( Lorenzo et al. 2003 ). Like ethylene, JA rapidly induces ERF1 expression, and treatment with both hormones synergistically activates ERF1 . Induction of ERF1 by both hormones alone or in combination is dependent upon both signaling pathways, and constitutive overexpression of ERF1 rescues the defense-response defects of both ethylene- and JA-insensitive mutants. These findings suggest that ERF1 represents one of the first signaling nodes identified in the complex web of hormonal cross-talk. The auxin and BR pathways also appear to converge and mutually regulate some developmental processes. Both hormones promote cell expansion, and microarray studies have revealed that as many as 40% of all BR-induced genes are also up-regulated by auxin ( Goda et al. 2004 ; Nemhauser et al. 2004 ). BR is perceived by the cell surface receptor kinase BRI1 ( Wang and He 2004 ). The SHAGGY/GSK3-type kinase BIN2 acts as a negative regulator of the pathway downstream of the receptor. In the absence of a BR signal, BIN2 phosphorylates the transcription factors BES1 and BZR1, targeting them for proteolysis by the 26S proteasome. Upon a BR stimulus, BIN2 is inactivated, allowing BES1 and BZR1to accumulate in the nucleus, where they are presumably involved in regulating BR-responsive genes. Using combined genetic, physiological, and genomic approaches, Nemhauser and colleagues (2004) were able to demonstrate that auxin and BR regulate Arabidopsis hypocotyl (embryonic stem) elongation in a synergistic and interdependent fashion. Elevating endogenous auxin levels rendered plants more sensitive to BR application in hypocotyl elongation assays, and this response was dependent upon both the auxin and BR signaling pathways. Genetic studies suggest that the convergence of these two pathways occurs at a late point in hormone signaling, perhaps at the promoters of the many genes responsive to both hormones. In support of this notion, bioinformatic analysis identified distinct sequence elements that were enriched specifically in the promoters of auxin-induced, BR-induced, and auxin/BR-induced genes. Many Unanswered Questions While great strides have been made in recent years in understanding the molecular basis of phytohormone action, many fundamental questions remain. Receptors and other upstream signaling components remain to be identified for the majority of the phytohormones. Equally important are the elucidation of hormonal networks and the integration of these networks with the morphogenetic program, such that our understanding of hormone action can be placed in a developmental context.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516799.xml
517493	Automatic annotation of protein motif function with Gene Ontology terms	Background Conserved protein sequence motifs are short stretches of amino acid sequence patterns that potentially encode the function of proteins. Several sequence pattern searching algorithms and programs exist foridentifying candidate protein motifs at the whole genome level. However, amuch needed and importanttask is to determine the functions of the newly identified protein motifs. The Gene Ontology (GO) project is an endeavor to annotate the function of genes or protein sequences with terms from a dynamic, controlled vocabulary and these annotations serve well as a knowledge base. Results This paperpresents methods to mine the GO knowledge base and use the association between the GO terms assigned to a sequence and the motifs matched by the same sequence as evidence for predicting the functions of novel protein motifs automatically. The task of assigning GO terms to protein motifsis viewed as both a binary classification and information retrieval problem, where PROSITE motifs are used as samples for mode training and functional prediction. The mutual information of a motif and aGO term association isfound to be a very useful feature. We take advantageof the known motifs to train a logistic regression classifier, which allows us to combine mutual information with other frequency-based features and obtain a probability of correctassociation. The trained logistic regression model has intuitively meaningful and logically plausible parameter values, and performs very well empirically according to our evaluation criteria. Conclusions In this research, different methods for automatic annotation of protein motifs have been investigated. Empirical result demonstrated that the methods have a great potential for detecting and augmenting information about thefunctions of newly discovered candidate protein motifs.	Background With the completion of many genome sequencing projects and advances in the methods of automatic discovery of sequence patterns (see Brazma [ 1 ] and Brejova et al [ 2 ] for reviews), it is now possible to search or discover protein sequence motifs at the genome level. If one regards protein sequences as "sentences" of the biological language with amino acids as the alphabet, then protein motifs can be considered as words or phrases of that language and determining the function of a motif is equivalent to determining the sense of a word. Identifying biological sequence motifs has been a fundamental task of bioinformatics, which has led to the development of several motif (pattern) databases, such as PROSITE, BLOCKS, SMART and Pfam [ 3 - 6 ]. These databases are usually constructed by studying the set of protein sequences that are known to have certain functions and extracting the conserved sequence motifs that are believed to be responsible for their functions. However, the number of motifs that can be extracted in this way is quite limited, and it has been a major challenge to discover new motifs. With the advent of algorithms and programs that can automatically discover sequence motifs from any given set of sequences [ 1 , 2 , 7 - 9 ], it is possible to mine a large number of sequences to find novel motifs without necessarily knowing their functions and to compile a dictionary of biological language accordingly. An essential task involved in the compilation of such a dictionary is to determine the function (the meaning) of newly identified protein motifs. Here, we report development of general methods that can be used to predict the function of protein motifs by mining the knowledge in the Gene Ontology. The Gene Ontology™ (GO) project [ 10 ] is a concerted effort by the bioinformatics community to develop a controlled vocabulary (GO terms) and to annotate biological sequences with the vocabulary. A biological sequence is described in three different aspects, namely, biological process, cellular component, and molecular function. The standardized annotation with a controlled vocabulary is the main advantage of Gene Ontology, which facilitates both communications among scientists and information management. Both the number of annotated sequences and the number of GO terms associated with individual sequences in the Gene Ontology database are increasing very rapidly. Moreover, natural language processing techniques are also being used to automatically annotate gene products with GO terms [ 11 , 12 ]. Thus, it can be foreseen that the annotations of protein sequences in the Gene Ontology database will become more and more detailed, and have a great potential to be used as an enriched knowledge base of proteins. The basic approach for determining the function of a motif is to study all the sequences that contain the motif (pattern). Intuitively, if all the functional aspects of the sequences matching a motif are known, we should be able to learn which function is most likely encoded by the motif, based on the assumption that every protein function is encoded by an underlying motif. This means that we would need a knowledge base of protein sequences, in which the functions of a sequence are annotated as detailed as possible. In addition, we would also need prediction methods that can work on a given set of protein sequences and their functional descriptions to reliably attribute one of the functions to the motif that matches these sequences. To determine the function of any novel motif, we would first search the protein knowledge base to retrieve all the functional descriptions of the proteins containing the motif, and then use such prediction methods to decide which function is encoded by the motif. In this research, we use the Gene Ontology database as our protein knowledge base and explore statistical methods that can learn to automatically assign biological functions (in the form of GO terms) to a protein motif. Our approach is based on the observation that the Gene Ontology database contains protein sequences and the GO terms associated with the sequences. In addition, the database also contains information of known protein motifs, e.g. the PROSITE patterns that match the sequences. Thus, the protein sequences in the database provide a sample of potential associations of GO term with motifs, among which some are correct (i.e., the GO term definition matches the functional description of the motif) and some are not. This provides us an opportunity to perform supervised learning to identify discriminative features and use these features to predict whether a new association is correct or not. Current Gene Ontology database is implemented with relational database system, which allows one to perform queries like "retrieve all GO terms associated with the sequences that matches a given motif" and vice versa . However, the database usually returns more than one GO terms that may or may not describe the function of the motif in the query. Thus, we need methods to disambiguate which GO term describe the function of the motif (assign a GO term to a motif) and determine how confident we are as the assignment is concerned. We use statistical approaches to learn from known examples and cast disambiguation task into a classification problem. Furthermore, the probability output by the classifier can be used to represent its confidence for the assignment. Recently, Schug et al [ 13 ] published their result of automatically associating GO terms with protein domains from two motif databases – ProDom and CDD [ 14 , 15 ]. Their approach is to use protein domains to BLAST [ 16 ] search against GO database and assign the molecular functional GO term from the sequence matching the domains with most significant p -value. They found that, in the database they worked with, most sequences only had one functional GO term. Therefore, they could assign the GO term of a sequence to the motif that matched with highest score with fairly good accuracy. However, due to restrictive assumption that each sequence has only one GO term, their approach can not address the potential problem that a sequence matching a motif has multiple associated GO terms, which is common case now, and how to resolve such ambiguity. Results The data set We use the May 2002 release of the Gene Ontology sequence database (available online [ 17 ]), which contains 37,331 protein sequences. For each sequence, a set of GO terms assigned to the sequence is identified, and a set of PROSITE patterns that match the same sequence is also retrieved. If both sets are nonempty, all the possible pattern-term combinations formed by the two sets are produced. Table 1 shows an example association of GO terms with PROSITE motifs. The protein MGI\|MGI:97380 from the database is assigned seven GO terms and the sequence also matches two PROSITE patterns. Thus, as cross product of two sets, 14 distinct associations are produced. Note that the same pattern-term association may be observed multiple times within the database. A total of 4,135 GO terms, 1,282 PROSITE motifs, and 2,249 distinct PROSITE-GO associations have been obtained from this database. Using the information stored in the Gene Ontology and PROSITE, we manually judged a set of 1,602 cases of distinct PROSITE-GO associations to determine whether the association is correct or not. The PROSITE-GO association set has been judged in two different ways. One way is to label an association as correct if and only if the definition of the GO term and the PROSITE motif match perfectly according to the annotator. Gene Ontology has the structure of a directed acyclic graph (DAG) to reflect the relations among the terms. Most terms (nodes in the graph) have parent, sibling and child terms to reflect the relation of "belonging to" or "subfamily". The second way of judging GO-PROSITE association is to label an association as correct if the GO term and the PROSITE motif are either exact match or the definitions of GO term and PROSITE motif are within one level difference in the tree, i.e., the definition of GO term and the PROSITE motif have either a parent-child relation or a sibling relation according to the GO structure. Thus we have two sets of labeled PROSITE-GO associations, the perfect match set and the relaxed match set (with neighbors). Both sets are further randomly divided into training (1128 distinct associations) and test (474 distinct associations) sets. Since the test sample size is fairly large, the variance of the prediction accuracy can be expected to be small. Thus we have not considered any alternative split of training and test sets. Measuring term-motif associations Intuitively, we may think of the GO terms assigned to a protein as one description of the function of a protein in one language (human understandable) while the motifs contained in the protein sequence as another description of the same function in a different language (biological). We would like to discover the "translation rules" between these two languages. Looking at a large number of annotated sequences, we hope to find which terms tend to co-occur with a given motif pattern. Imagine that, if the sequences that match a motif are all assigned a term T , and none of the sequences that do not match the motif is assigned the term T , then it is very likely that the motif pattern is encoding the function described by term T . Of course, this is only an ideal situation; in reality, we may see that most of, but not all of the proteins matching a motif pattern would be assigned the same pattern, and also some proteins that do not match the motif may also have the same term. Thus, we want to have a quantitative measure of such correlation between GO terms and motif patterns. A commonly used association measure is mutual information (M.I.), which measures the correlation between two discrete random variables X and Y [ 18 ]. It basically compares the observed joint distribution p ( X = x , Y = y ) with the expected joint distribution under the hypothesis that X and Y are independent, which is given by p ( X = x ) p ( Y = y ). A larger mutual information indicates a stronger association between X and Y , and I ( X;Y ) = 0 if and only if X and Y are independent. For our purpose, we regard the assignment of a term T to a sequence and the matching of a sequence with a motif M as two binary random variables. The involved probabilities can then be empirically estimated based on the number of sequences matching motif M ( NM ), the number of sequences assigned term T ( NT ), the number of sequences both matching M and assigned T ( NT - M ), and the total number of sequences in the database. Table 2 shows the top five terms that have the highest mutual information with PROSITE motif PS00109, which is the specific active-site signature of protein tyrosine kinases, along with the related counts. We set out to test whether we can use mutual information as a criterion to assign a GO term to a PROSITE motif. One approach is to use a mutual information cutoff value c to define a simple decision rule: assign term T to motif M , if and only if I ( T;M ) ≥ c . For a given cutoff c , the precision of term assignment is defined as the ratio of the number of correct assignments to that of the total assignments according to the cutoff c . In Figure 1 , we plot the precision at different mutual information cutoff values. It is easy to see that, in general, using a higher (i.e., stricter) cutoff, the precision is higher; indeed, the Pearson correlation coefficient between the precision and the cutoff is 0.837. This suggests that mutual information is indeed a good indicator of the correlation However, a drawback of such an approach is that, given a motif, sometimes, many observed motif-term associations can have mutual information above the cutoff value, making it difficult to decide which pair is correct. While in other cases, the mutual information of the observed motif-term pairs may all be below the cutoff value, but we still would like to predict what terms are most likely to be appropriate for the motif. To address this problem, we can use a different cutoff strategy, and adopt a decision rule that assigns a GO term to a motif based on the ranking of mutual information, which is a common technique used in information retrieval text categorization [ 19 ]. More specifically, for each PROSITE motif M in the annotated data set, all observed motif-term associations containing M are retrieved and ranked according to mutual information, then the term that has highest mutual information is assigned to M . Alternatively, if we use this approach to facilitate human annotation, we can relax the rule to include GO terms that have lower ranks, thus allowing multiple potential GO terms to be assigned to a motif, assuming that a human annotator would be able to further decide which is correct. In this method, the key in making a decision is to select a cutoff rank that covers as many correct associations as possible (high sensitivity) while also retrieves as fewer incorrect associations as possible (high specificity). The optimal cutoff can be determined by the desired utility function. Figure 2 shows the Receiver Operating Characteristic (ROC) curve [ 20 ] of assigning GO terms to PROSITE motifs in our data set according to the rank of motif-term associations. The two curves are for the two different labeled association sets (i.e., perfect match and relaxed match) respectively. The areas under the two curves are 0.782 and 0.735 respectively, which can be considered as fairly good. We also plot the precision, also referred to as positive predictive value, in panel B. The precision is calculated as the percent of predicted assignments that are truly correct. As shown in panel B, if we assign the GO terms at the top rank for all PROSITE motifs, 50–70% of the cases will be predicted correctly. As we loosen the threshold to include lower ranked terms, we would assign more terms to a motif, and as expected, precision would decline. But even at rank 5, we still have a precision of about 50%. Also shown in Table 2 , with respect to the PROSITE pattern of tyrosine kinase (PS00109), most of the top five associated GO terms are related to kinase activity and the term with the highest rank is the most specific. Predicting motif functions using logistic regression While the mutual information measure appears to give reasonable results, there are three motivations for exploring more sophisticated methods. First, the mutual information value is only meaningful when we compare two candidate terms for a given motif pattern; it is hard to interpret the absolute value. While a user can empirically tune the cutoff based on some utility preferences, it would be highly desirable to attach some kind of confidence value or probability of correctness to all the potential candidate motif-term associations. Second, there may be other features that can also help predict the function (term) for a motif. We hope that the additional features may help a classifier to further separate correct motif-term assignment from wrong ones. Third, there exist many motifs with known functions (e.g., those in the PROSITE database), and it is desirable to take advantage of such information to help predict the functions of unknown motifs. This means that we need methods that can learn from such information. In this section, we show that the use of logistic regression can help achieve all three goals. Specifically, we use logistic regression to combine the mutual information with other features, and produce a probability of correct assignment. The motifs with known functions serve as training examples that are needed for estimating the parameters of the regression function. Feature extraction and parameter estimation We now discuss the features to be used in logistic regression, in addition to the mutual information discussed in the previous section. The goal is to identify a set of features that is helpful to determine whether association of any pair of a GO term and a motif is correct or not, without requiring specific information regarding the function of GO term and motif. For a distinct motif-term pair, we collect following frequency-based features: (1) The number of sequences in which the GO term ( T ) and PROSITE motif ( M ) co-occur ( NT-M ). (2) The number of sequences in which T occurs ( NT ). (3) The number of sequences in which M occurs ( NM ). (4) The number of distinct GO terms ( G ) seen associated with M ( NG\|M ). (5) The number of distinct PROSITE patterns ( P ) seen associated with T ( NP\|T ). In addition, we also consider, as a feature, the similarity of the sequences that support a motif-term pair. Intuitively, if a motif is conserved among a set of diverse sequences, it is more likely that the motif is used as a building block in proteins with different functions. Thus, the average pair-wise sequence similarity of the sequence set can potentially be used as a heuristic feature in the logistic regression classifier. Given a set of sequences, we use a BLAST search engine to perform pair-wise sequence comparisons. We devised a metric AvgS to measure the averaged pair-wise sequence similarity per 100 amino acids (see methods) and use it as an input feature for classifier. To cast the prediction problem as a binary classification problem, we augment our data set of motif-term pairs with a class label variable Y , so that Y = 1 means correct assignment and 0 means incorrect. We represent a motif-term pair by a vector of features X = ( X 1 ,..., X k ), where k is the number of features. The seven features/variables used in our experiments are NT-M, NT, NM, NG\|M, NP\|T, AvgS , and M.I. . Suppose we have observed n motif-term pairs, then we have n samples of ( y i , x i ), i = 1, 2, ..., n , where, y i is the correctness label and x i is the feature vector for the corresponding motif-term pair. Our goal is to train a classifier which, when given a motif-term pair and feature vector X , would output a label Y with value 1 or 0. Alternatively, we can also consider building a classifier which outputs a probability that Y = 1 instead of a deterministic label. Thus, our task is now precisely a typical supervised learning problem; many supervised learning techniques can potentially be applied. Here, we choose to use logistic regression as our classification model because it has a sound statistical foundation, gives us a probability of correct assignment, and can combine our features naturally without any further transformation. In order to build a model only with the truly discriminative features, it is a common practice to perform feature selection for logistic regression. We use a combined forward and backward feature selection algorithm. Starting from the intercept, we sequentially add features into the model and test if the log-likelihood increases significantly; we keep the current feature if it does. After the forward selection, we sequentially drop features from the model, to see if dropping a feature would significantly reduce the log-likelihood of the model; if it does, we exclude the feature from the model, otherwise continue. When testing the significance, we use the likelihood ratio statistic G , given by 2 l ( D\|β f )/ l ( D\|β - f ), where, l ( D\|β f ) and l ( D\|β - f ) are the log-likelihood of the model with feature f and the model without feature f , respectively. Since we add or drop one feature at a time, G follows χ 2 distribution with degree of freedom of 1 [ 21 ]. We use the p -value of 0.1 as a significant threshold. Figure 3 illustrates the procedure of feature selection. We found that the average pair-wise similarity of supporting sequence set does not contribute to the model significantly and so excluded it; all other variables contribute to the model significantly. The results of parameters estimation are show in the Table 3 . Logistic regression classification After fitting the model using the training set, we tested the model on the test set, i.e., we used the model to compute an output p ( Y i = 1\| X i ) for each test case. Table 4 shows an example of computed conditional probability of correct assignment for the GO terms associated with the protein motif possible the motif "PS00383", which is the "tyrosine specific protein phosphatases signature and profiles". The table 4 lists top 5 GO terms, which are observed to be associated with the motif and ranked according to the conditional probability returned by logistic regression. As the results from the logistic regression are the conditional probability that an association of a GO term with a given motif is correct, we need to decide the cut off threshold for making decision. We calculate the sensitivity and specificity for a different threshold from 0.1 to 0.9 with a step of 0.1 and plotted the ROC curves as shown in Figure 4 . The areas under the logistic regression ROC curves are 0.875 and 0.871 for perfect match and relaxed match test set respectively. The precision of the rules is plotted in panel B, where we see that, as the rule becomes more stringent (using a higher threshold), predictions generally become more accurate. We noticed that the precision on the perfect match test set is more variable. This is probably due to the fact that this data set has fewer cases with Y = 1, thus, a small change in the number of cases introduces a large change in percentage. For example, when the threshold is set at 0.9, only three cases are covered by the rule and two of them are correct, thus percent correct drop to 66%. To see whether the additional features are useful, we also performed ROC analysis using different mutual information cutoff threshold on the perfect match test set. The result is shown in Figure 4 panels C and D . We see that using mutual information alone performs almost as well as logistic regression with additional features. However, the area under the curve (0.816) is smaller than that of logistic regression (0.875), indicating that logistic regression does take advantage of other features and has more discriminative power than mutual information alone. The coefficients β 1 , β 2 and β 3 for the three features NT-M , NT and NM , which are also involved in the calculation of mutual information, have a very interesting interpretation – they indicate that the roles of these three variables in the logistic regression model actually are to compromise the effect of mutual information! Indeed, according to the formula of the mutual information, a strong correlation corresponds to a high NT-M , low NT , and low NM , but the coefficients shown in Table 3 clearly suggest the opposite. We believe that this actually corrects one drawback of mutual information – over-emphasizing the correlation but ignoring the support or the strength of evidence. For example, if a term is rare, say occurs only once in the data set, then it would have a very high mutual information value (due to an extremely low NT ) with respect to any pattern matched by the sequence to which the term is assigned. But, intuitively, one occurrence is very weak evidence, and at least should be regarded as weaker than when we have a term occurring 10 times in total and co-occurring 9 times with the same motif. The key issue here is that mutual information only reflects the correlation between variables, but does not take into account the strength of evidence, therefore, tends to over-favor the situation where there is a perfect correlation but very little evidence. However, the number of sequences in which the co-occurrence happens, which is called the "support" for the association, is also very important. The coefficients for the other two parameters, NG\|M and NP\|T , are also meaningful. Their negative signs indicate that the more terms a motif co-occurs with or the more motifs a term co-occurs with, the less likely a particular association is correct. This also makes sense intuitively, since all those co-occurring terms can be regarded as "competing" for a candidate description of the motif's function, so the more terms a motif is associated with, the competition is stronger, and thus the chance that any particular term is a correct description of function should be smaller. Thus, the logistic regression model not only performs well in terms of prediction accuracy but also gives meaningful and logically plausible coefficient values. Discussion In this paper, we explore the use of the Gene Ontology knowledge base to predict the functions of protein motifs. We find that the mutual information can be used as an important feature to capture the association between a motif and a GO term. Evaluation indicates that, even used alone, the mutual information could be useful for ranking terms for any given motif. We further use logistic regression to combine mutual information with several other statistical features and to learn a probabilistic classifier from a set of motifs with known functions. Our evaluation shows that, with the addition of new features and with the extra information provided by the motifs with known functions, logistic regression can perform better than using the mutual information alone. This is encouraging, as it shows that we can potentially learn from the motifs with known functions to better predict the functions of unknown motifs. This means that our prediction algorithm can be expected to further improve, as we accumulate more and more known motifs. Although we have so far only tested our methods on the known motifs, which is necessary for the purpose of evaluation, the method is most useful for predicting the functions of new and unknown motifs. For the future work, we can build a motif function prediction system and apply our algorithm to many candidate new motifs e.g., those discovered using TEIRESIAS, SPLASH or other programs. This would further enable us to perform data mining from the Gene Ontology database in several ways. For example, we can hypothesize the functions of a large number of novel motifs probabilistically, then we will be able to answer a query, such as "finding the five patterns that are most likely associated with the GO term tyrosine kinase". This is potentially very useful because it is not uncommon that substantial knowledge about the functions and sub-cellular location of a given protein is available even though a structural explanation for the functions remains obscure. On the other hand, we believe that our methods will facilitate identifying potentially biological meaningful patterns among the millions of patterns returned by pattern searching programs. A sequence pattern that associates with certain GO term with high M.I. or probability is more like to be a meaningful pattern that that with low scores. Furthermore, our methods can also be used in automatic annotation of novel protein sequences as suggested in Schug et al and Rigoutsos et al [ 9 , 13 , 22 ]. Our methods provide different approaches to associate sequence patterns with functional descriptions. After associating functional descriptions (in the form of GO term) to motifs, we can determine what motifs a novel protein sequence matches and correspondingly transfer the functional descriptions associated with motifs to the sequence. One key advantage of our methods is that the probability of correctness for a GO-motif association can be considered as confidence or uncertainty. This enables one to optimize the automatic annotation according to Bayesian decision theory and minimize the risk of incorrect annotation. Having stated the potential uses of our approaches, we also realize that there exist some limitations for our methods. For example, in order to predict the function of a newly identified sequence pattern correctly, we would require functional annotations of the sequences of GO database be complete and accurate, which may not always be the case. In this paper, we mainly used the motifs with known function to evaluate the capability of the methods developed in this research. Our result shows that the methods work well with known sequences patterns. Currently, the annotation of motif function with GO term is carried out manually at the European Bioinformatics Institute (the GOA project). Such approach is warranted because human annotation is more accurate than automatic ones. However, as the amount of information regarding protein functions accumulates and a large number of new potential motifs are discovered, it will be very labor intensive to annotate the potential association of protein function and protein patterns. By then, the methods studied in this research will potentially prove to be useful to discover the underlying protein motifs that are responsible for the newly annotated function. For example, the methods can be used as prescreening to narrow down to the most possible associations of protein function and motifs, thus facilitate human annotation. Conclusions In summary, we have developed methods that disambiguate the associations between of Gene Ontology terms and protein motifs. These methods can be used to mine the knowledge contained in the Gene Ontology database to predict the function of novel motifs, discover the basis of a molecular function at primary sequence level and automatically annotated the function of novel proteins. Methods Mutual information Mutual information is defined as follows In which the probabilities p ( X = x , Y = y ), p ( X = x ) and p ( Y = y ) can be empirically estimated from the data by counting occurrence/co-occurrence followed by normalization. Sensitivity and specificity The sensitivity and specificity of the rules are calculated as where TP (True Positive) is the number of associations labeled as correct among the retrieved motif-term pairs meeting the ranking cutoff criteria, FN (False Negative) is the number of associations labeled as correct but not retrieved, TN (True Negative) is the number of associations labeled as incorrect and not retrieved, and FP (False Positive) is the number of associations labeled incorrect but are retrieved. Averaged sequence similarity Calculation of the average pair-wise sequence similarity per 100 amino acids ( AvgS ) of a sequence set is as follows Where S ij is raw BLAST pair-wise similarity scores between the sequence i and sequence j ; L i and L j are the lengths of sequences i and j , respectively; n is the number of sequences in the set; and δ ( i , j ) is a delta function which equals 1 if i = j and 0 otherwise. Logistic regression The logistic regression model is a conditional model that assumes the following linear relationship between p ( Y = 1\| X ) and X 1 , ..., X k : where, β = ( β 0 , β 1 , ..., β k ) is the parameter vector. We can fit the logistic regression model (i.e., estimate the parameters) using the Maximum Likelihood method – essentially setting the parameters to values at which the likelihood of the observed data is maximized (Hosmer and Lemeshow 1989, Hastie et al 2001). In our experiments, we use iteratively reweighted least squares (IRLS) algorithm [ 23 ] to fit the logistic regression model. All features are normalized to zero mean and unit variance before training.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517493.xml
528844	Are the effects of nicotinic acid on insulin resistance precipitated by abnormal phosphorous metabolism?	Nicotinic acid is a unique cholesterol modifying agent that exerts favorable effects on all cholesterol parameters. It holds promise as one of the main pharmacological agents to treat mixed dyslipidemia in metabolic syndrome and diabetic patients. The use of nicotinic acid has always been haunted with concerns that it might worsen insulin resistance and complicate diabetes management. We will discuss the interaction between phosphorous metabolism and carbohydrate metabolism and the possibility that worsening of insulin resistance could be related to adrug induced alteration in phosphorous metabolism, and the implications of that in medical management of diabetes and metabolic syndrome patients with mixed dyslipidemia.	Background Nicotinic acid functions in the body after conversion to nicotinamide adenine dinucleotide (NAD) in the NAD coenzyme system. Niacin reduces total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), triglycerides (TG), apolipoprotein B-100 (Apo B) and Lipoprotein (a) (Lp (a)). Niacin increases high-density lipoprotein cholesterol (HDL-C), apolipoprotein A-I (Apo A-I), lipoprotein A-I, and HDL2: HDL3 ratio. The severity and type of underlying lipid abnormality determines the extent of response to niacin treatment [ 1 ]. There is evidence suggesting that niacin reduces the risk of a coronary artery or cerebrovascular disease in the setting of secondary prevention, and when combined with a bile acid-sequestrant, promotes regression and decreases progression of atherosclerotic cardiovascular disease [ 2 , 3 ]. Niacin is known to cause transient, small but statistically significant dose-related reductions in phosphorous levels. These effects were studied for extended release niacin in placebo-controlled trials, mean percentage changes from baseline (± standard error) in phosphorous were (-4.0 ± 1.3) for the 500 mg once daily dose, (-7.2 ± 1.1) for the 1000 mg once daily dose, (-9.1 ± 1.1) for the 1500 mg once daily dose, and (-12.6 ± 1.6) for the 2000 mg once daily dose [ 4 ]. Presentation of the hypothesis It has been long known that niacin induces impaired fasting glucose. I was also suggested that this effect is most likely transient. The mechanism behind this phenomenon has not been fully elucidated [ 5 ]. Multiple studies were conducted to study effects of nicotinic acid on exacerbating insulin resistance. One retrospective study suggested that the use of moderate doses of extended release nicotinic acid (average, 1000 mg/d) in reasonably controlled diabetics was associated with improved glycemic control with HbA1c levels decreased by 0.5% ± 0.3% due to aggressive hypoglycemic therapy; as most of these patients had their insulin or oral diabetes regimen intensified [ 6 ]. Another retrospective study using unmodified niacin concluded similar results [ 7 ]. The (ADMIT) trial, was one of the first studies to demonstrate the safety of nicotinic acid in patients with diabetes. Diabetic patients (468 participants, 125 with diabetes and peripheral arterial disease) were randomized to treatment with either placebo or niacin (2500 mg/d) for 18 weeks. Patients randomized to niacin therapy had modest increases in fasting glucose (+8 mg/dL). There was no significant change in HbA1c levels or in diabetes treatment regimen in the niacin-treated diabetic patients group [ 8 ]. Another prospective Trial (ADVENT), was a 16-week, double-blind, placebo-controlled trial, 148 patients were randomized to placebo (n = 49) or 1000 (n = 45) or 1500 mg/d (n = 52) of extended release niacin. Fasting plasma glucose levels in both treatment groups were unchanged from placebo at the study's end. Meanwhile, the mean HbA1c levels were only slightly elevated in the 1500 mg arm, compared with placebo. The 1500 mg nicotinic acid arm required a small increase in anti-diabetic treatment regimen [ 9 ]. It appears from the clinical data presented above, that nicotinic acid at higher doses could results in alterations in glycemic control of patients with an already abnormal glucose metabolism. Phosphate is primarily an intracellular anion. Insulin is implicated in the transport of glucose and phosphate from the intra-cellular to the extra-cellular space. Studies have suggested a link between decreased serum phosphorous and disturbed carbohydrate metabolism, eventually leading to hyperglycemia [ 10 , 11 ]. Chronic hypophosphatemia inhibits glucose transport. Further more, reduced serum phosphorous levels significantly decrease the phosphorylation of carbohydrate intermediates in glycolysis and glycogenesis. It seems intuitive to suspect that the transient decrease in plasma phosphorous level might explain the transient disturbance in carbohydrate metabolism and hyperglycemia induced by nicotinic acid. Also, studies suggest the existence of a dose response relationship in both instances. One way to test this hypothesis will be to administer nicotinic acid to both healthy subjects, and to those with impaired fasting glucose or subjects with diabetes, then measure simultaneously at different doses serum phosphorous, and correlate the changes to changes in serum glucose and serum insulin either individually or combined as a measure of insulin resistance. It will be of interest to study if such a relationship exists in patients with the metabolic syndrome diagnosis, since literature suggests the existence of an abnormal phosphorous metabolism in these patients [ 10 , 12 ]. Implications of the hypothesis This hypothesis has several important insights and implications: 1. If valid, it suggests the possibility of a simple clinical way to prevent nicotinic acid induced worsening of glycemic control by dietary phosphorous supplementation, eliminating the need to intensify expensive diabetic regimens in diabetic patients that need nicotinic acid treatment. 2. If valid, it will aid in lessening the theoretical anxiety of the possibility of precipitating diabetes in glucose intolerant or metabolic syndrome patients in need of nicotinic acid treatment for the management of dyslipidemia. 3. Metabolic syndrome and diabetics stand to benefit greatly from the new evidence implicating the importance of increasing HDL-cholesterol in cardiovascular event reduction. 4. This hypothesis offers new insights into the mechanism of development of glucose intolerance.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC528844.xml
503394	Complementary and alternative medical therapies for chronic low back pain: What treatments are patients willing to try?	Background Although back pain is the most common reason patients use complementary and alternative medical (CAM) therapies, little is known about the willingness of primary care back pain patients to try these therapies. As part of an effort to refine recruitment strategies for clinical trials, we sought to determine if back pain patients are willing to try acupuncture, chiropractic, massage, meditation, and t'ai chi and to learn about their knowledge of, experience with, and perceptions about each of these therapies. Methods We identified English-speaking patients with diagnoses consistent with chronic low back pain using automated visit data from one health care organization in Boston and another in Seattle. We were able to confirm the eligibility status (i.e., current low back pain that had lasted at least 3 months) of 70% of the patients with such diagnoses and all eligible respondents were interviewed. Results Except for chiropractic, knowledge about these therapies was low. Chiropractic and massage had been used by the largest fractions of respondents (54% and 38%, respectively), mostly for back pain (45% and 24%, respectively). Among prior users of specific CAM therapies for back pain, massage was rated most helpful. Users of chiropractic reported treatment-related "significant discomfort, pain or harm" more often (23%) than users of other therapies (5–16%). Respondents expected massage would be most helpful (median of 7 on a 0 to 10 scale) and meditation least helpful (median of 3) in relieving their current pain. Most respondents indicated they would be "very likely" to try acupuncture, massage, or chiropractic for their back pain if they did not have to pay out of pocket and their physician thought it was a reasonable treatment option. Conclusions Most patients with chronic back pain in our sample were interested in trying therapeutic options that lie outside the conventional medical spectrum. This highlights the need for additional studies evaluating their effectiveness and suggests that researchers conducting clinical trials of these therapies may not have difficulties recruiting patients.	Background Back pain is one of the most common and costly health problems in developed countries, where more than half of adults suffer from this condition each year [ 1 ] and 70% to 80% suffer from it at some time in their lives [ 2 ]. Patients with back pain are often dissatisfied with standard medical care [ 3 ], especially in comparison to care provided by alternative providers [ 4 - 6 ]. In fact, back/neck pain is the number one condition for which Americans seek complementary or alternative medical (CAM) care. During 1997, 30% of Americans with back problems visited CAM practitioners, especially chiropractors and massage therapists, for this condition and another 18% used CAM self-care [ 7 ]. Yet, to date, few of these therapies have been adequately evaluated for effectiveness, in part because of methodological challenges including recruiting sufficient numbers of participants, designing reasonable interventions and selecting appropriate control groups. Prior to designing and conducting two pilot clinical trials evaluating the effectiveness of five different CAM therapies for chronic low back pain (LBP) in older (65+ years) and younger (20 through 64 years) adults, we sought to refine recruitment strategies. As part of this effort, we surveyed chronic low back pain patients about their interest in trying each of these five CAM therapies if included in their health plan benefits and as part of two separate clinical trials. In this descriptive and exploratory study, we also collected information about their knowledge of, experience with, and perceptions of each of these therapies. Methods Study design From April to October 2001, we conducted telephone interviews with 249 patients who were currently suffering from non-specific low back pain that had persisted at least three months. The patients were members of a non-profit managed health care system (Group Health Cooperative in the Puget Sound region of Washington State) and a large multi-specialty group practice (Harvard Vanguard Medical Associates, Boston, Massachusetts). Our goal was to interview 150 patients from Group Health and 100 patients from Harvard Vanguard who were otherwise healthy and spoke English, with 50% of the interviews from adults 65 years or older. The Institutional Review Boards of Group Health Cooperative, Seattle WA and Harvard Pilgrim Health Care, Boston, MA approved the study. Sample Using automated visit data, we identified and mailed letters to 787 patients with working phone numbers who visited a Group Health (n = 422) or a Harvard Vanguard (n = 365) primary care provider and had a diagnosis consistent with non-specific low back pain between 12 and 52 weeks previously. Because we were planning to use the results of the survey to help us refine recruitment strategies for two pilot randomized trials, our exclusion criteria for this study were those we planned to use in the subsequent trials. We therefore used the automated visit data to exclude the following individuals from our sample: • those whose back pain may have been due to a specific disease or condition (i.e., sciatica, herniated disc, spondylolisthesis, spine fracture, vertebral fracture, cancer); • those who had other pain conditions that could complicate the interpretation of trial results (rheumatoid arthritis, ankylosing spondylitis, fibromyalgia); • those who were inappropriate candidates for one or more of the clinical trial treatments (aneurism, coagulation disorders, osteoporosis) or who were unlikely to be able to give informed consent or to participate in the baseline and follow-up assessments of the trial (Alzheimers disease, dementia, major psychoses, blindness, deafness). Before administering the in-person survey, we asked eight screening questions and then excluded individuals who did not have back pain at the time of the interview or who had not had back pain for at least 12 weeks, who had previously had low back surgery, who reported having fractured a vertebrae, who were pregnant, who were involved in back-pain related litigation, who had serious health problems, or who could not speak English. We tried to contact all mailees, but could not reach 28 (7%) persons from Group Health and 73 (20%) from Harvard Vanguard despite at least seven phone calls. Among the 394 Group Health and 292 Harvard Vanguard patients who were contacted, 57 from Group Health and 81 from Harvard Vanguard refused the interview and we could not determine their eligibility status, 195 from GHC and 104 from Harvard Vanguard were ineligible upon screening, and 142 from Group Health and 107 from Harvard Vanguard were eligible and interviewed. Thus, we were able to screen 70% of all mailees for eligibility (80% from Group Health and 58% from Harvard Vanguard). All screened and eligible mailees were interviewed. In both areas, individuals aged 65 and older were more likely to refuse screening (in Seattle: 19% of 210 older adults vs. 8% of 212 younger; in Boston: 31% of 184 older adults vs. 13% of 181 younger) and those under 65 were less likely to be contacted by phone (in Seattle: 12% younger vs. 1% older could not be contacted; in Boston: 31% younger vs. 9% older could not be contacted). Most patients were ineligible because they were not experiencing back pain at the time of the interview (n = 82 from Group Health and n = 68 from Harvard Vanguard) or their pain had not persisted for three months (n = 39 from Group Health and n = 11 from Harvard Vanguard). Survey questionnaire We conducted a phone interview that lasted an average of 17.7 minutes (SD = 6.1; range = 8 to 50 minutes). It included questions about demographic characteristics (age, race/ethnicity, education); back pain characteristics (years since first episode of back pain lasting more than two weeks, number of days of pain in the last six months, bothersomeness of pain on a 0 to 10 scale, expectations of pain level one year from the time of interview, and use of medications in the past week); self-reported knowledge (measured on a five point scale) of five CAM treatments or self-care methods (acupuncture, chiropractic, massage, meditation, t'ai chi); previous use of these therapies for any reason and for back pain specifically (and helpfulness of the therapy for back pain relief); perceived harm from previous use of these therapies; expectations of helpfulness of these therapies for current back pain; willingness to try these therapies if offered by the health plan for no additional cost and for a $10 per visit co-pay; willingness to participate in two hypothetical clinical trials, one evaluating acupuncture, chiropractic, and massage and another involving massage, meditation, and t'ai chi. (Respondents were told that the control group in both trials would receive a book about self-management of back pain.) Finally, respondents were asked about which treatment they most preferred among those offered in each trial. Gender and geographic location were obtained for respondents from the enrollment files of each healthplan. General definitions of each therapy were provided only for respondents who informed the interviewers they did not know what a particular therapy was. Acupuncture was described as a system of healing that involved inserting hair thin needles into acupuncture points just beneath the skin or using other methods, such as heat, to stimulate these points, whereas chiropractic was defined as a system of therapy that uses manipulation to adjusts the spine and other body parts to "promote normal nerve functions". Massage was described as the systematic rubbing and manipulation of muscle and other tissues to relieve bodily infirmities, while meditation was defined as a "self-directed practice for relaxing the body and calming the mind". Finally, tai chi was described as a Chinese martial art that uses slow and smooth body movements and is often practiced for its purported health benefits. The survey was pre-tested on a convenience sample of 15 people (both older and younger) in Seattle and 5 people in Boston. Statistical analyses We analyzed the data using the SAS statistical software version 6.12 (SAS Institute, Cary, NC). Descriptive data were characterized using percentages or medians. For each of the five CAM therapies, we performed separate exploratory logistic regressions to identify specific characteristics associated with 1) high degree of knowledge (4 or 5 on a 5-point scale) (five separate models), 2) prior use (for any reason and for back pain) (10 separate models), 3) high expectations of helpfulness for current back pain (7 to 10 on a 0 to 10 scale) (five separate models), 4) greatest likelihood of trying therapy for no additional cost (all five therapies) and for a $10 per visit co-pay (acupuncture, chiropractic, massage only) (eight separate models), and 6) greatest likelihood of participating in each of the two hypothetical clinical trials (two separate models). Thus, a total of 30 separate logistic models were created. For each of the 30 dependent variables, we identified potential predictor variables in advance and evaluated them in preliminary models. In Table 1 , the potential predictor variables evaluated in the preliminary models for each of the 28 therapy-specific dependent variables are indicated by an X. In addition, we modeled the likelihood of being "definitely willing" to participate in a hypothetical clinical trial of acupuncture, chiropractic, and massage and of being "definitely willing" to participate in a hypothetical clinical trial of massage, meditation, and t'ai chi. In both models, we evaluated the following 22 variables as potential predictor variables of being "definitely willing" to participate in the hypothetical clinical trial: demographic characteristics (age, gender, race, education, geographic location), prior use of each of the therapies included in the trial (i.e., acupuncture, chiropractic, and massage for one trial and massage, meditation, and t'ai chi for the other trial) for any reason (and for back pain), knowledge of these three included therapies, prior perceived harm from these three included therapies, years since first back pain, symptom bothersomeness, high expectations of each included CAM therapy for current back pain, number of days of back pain in last six months, and medication usage in the week prior to the interview. Table 1 Potential Predictor Variables Evaluated in 28 Therapy-Specific Logistic Regression Models Dependent Variables for Logistic Regressions Potential Predictor Variable High Knowledge of Therapy* Prior Use of Therapy* Prior Use of Therapy for Back Pain* High Expectations of Success of Therapy* Likelihood of Trying Therapy at No Cost* Likelihood of Trying Therapy for $10 Co-pay** Geographic location (Boston vs. Seattle) X ¶ X X X X X Age (65+ vs. < 65) X X X X X X Gender (female vs. male) X X X X X X Race (white, non-white) X X X X X X Education (no college vs. some college) X X X X X X ≥ 5 years since first back pain X X X ≥ 90 days of LBP in last 6 mo. X X X High symptom bothersomeness (7 – 10) on a 0 – 10 scale X X X High knowledge of therapy (4 or 5) on a 1 – 5 scale X X X Prior use of therapy X X X Prior use of therapy for back pain X X X High expectations of therapy (7 – 10) on a 0 – 10 scale X X Medication usage in past week X X Prior harm from therapy X X * Separate models were done for each of the five therapies (acupuncture, chiropractic, massage, meditation, t'ai chi) ** Separate models were done for acupuncture, chiropractic, and massage. ¶ An X indicates that a particular potential predictor variable was evaluated in a model with the specific dependent variable. Initially, we evaluated potential predictor variables in preliminary models containing five or fewer independent variables. Any independent variable associated with the dependent variable at a p value of 0.15 or less in a preliminary model was a candidate for the appropriate final model. We used a backwards elimination procedure to evaluate candidate predictor variables and to determine the final models [ 8 ]. All variables with a p value of 0.01 or less were retained in the final model. Odds ratios (OR) are presented along with 95% confidence intervals (95% CI). Table 4 presents the odds ratios that describe the significant associations (p < 0.01) for each of the 28 therapy – specific dependent variables. Table 4 Predictors of Knowledge of, Experience with, Expectations about, and Willingness to Try Five Complementary and Alternative Medical (CAM) Therapies Odds ratios* (95% CI) for the independent predictor variables used in the final models for each CAM therapy Dependent Variable Acupuncture Chiropractic Massage Meditation T'ai chi High Knowledge of specific therapy (4–5) Tried acupuncture: 43.6 (16.7–113.6) Tried chiropractic: 12.8 (6.2–26.7) Tried massage: 7.6 (4.0 – 14.7) Tried meditation: 11.6 (4.6–29.3) LOGISTIC NOT VALID Bostonian: 4.8 (1.9–12.3) Previously tried specific therapy No associations Bostonian: 0.5 (0.3 – 0.8) 65+ yrs: 0.4 (0.2 – 0.7) Female: 2.5 (1.3 – 4.8) No associations Previously tried specific therapy for back pain No associations None 65+ yrs: 0.3 (0.2 – 0.6) No associations LOGISTIC NOT VALID High expectations of specific therapy (7 – 10) 65+ yrs: 0.4(0.2 – 0.8) Knowledge: 2.9 (1.6 – 5.2) 65+ yrs: 0.3 (0.2 – 0.5) No associations No associations Tried acupuncture: 4.3 (2.1 – 9.0) Very likely to try specific therapy for free High expectations: 15.4 (3.6 – 66.1) High expectations: 27.4 (9.5 – 79.3) High expectations: 16.4 (7.4 – 36.5) High expectations: 3.6 (1.7 – 7.7) High expectations: 14.3 (5.4 – 38.3) Tried meditation: 2.4 (1.3 – 4.5) Very likely to try specific therapy for $10 /visit co-pay High expectations: 6.8 (3.0 – 15.5) High expectations: 8.1 (4.2 – 15.7) High expectations: 6.4 (3.5 – 11.4) NOT QUERIED NOT QUERIED Bostonian: 2.3 (1.4 – 4.0) Bostonian: 1.8 (1.001 – 3.2) These odds ratios describe the significant associations (p < 0.01) for each of the 28 therapy – specific dependent variables. For example, we found that those who had tried acupuncture were 43.6 times more likely to have high knowledge of acupuncture. No other variables were related to high knowledge of acupuncture. * These logistic regression models did not converge. Categorization for independent variables: Age (<65; 65+) Knowledge of therapy (1–3; 4–5) Gender (M; F) Expectations of therapy (missing through 6; 7+) Geography (Seattle; Boston) Prior Use of therapy (no; yes) Results Characteristics of respondents Most study participants were white, women, and had attended college (Table 2 ). Most had had back problems for at least five years, had experienced back pain at least 90 days in last six months, had used medications in the prior week, and expected little change in their pain in a year. Table 2 Demographic and Back Pain Characteristics of 249 Survey Respondents Characteristic Percent Location (Boston) 43 Age (< 65) 52 Women 60 White 80 Attended some college 57 At least 5 years since first back pain lasting longer than 2 weeks 60 90+ days of LBP in last 6 mo. 61 High symptom bothersomeness in the past week (≥ 7) on 0 – 10 scale 42 Used medication for LBP in the past week 56 Expect pain to be similar in a year 72 Missing data – last variable has 10 missing values (4% of all observations), 1 variable has 5 (2%), all others have 3 or fewer. Knowledge of, experience with, perceptions of, and willingness to try CAM therapies Except for chiropractic, most participants reported little or no knowledge of these therapies (Table 3 ). In logistic regressions, prior use of a therapy consistently predicted high knowledge of that therapy (Table 4 ). Table 3 Knowledge of, Experience with, Expectations about, and Willingness to Try Five CAM Therapies* Acupuncture (N = 249) Chiropractic (N = 249) Massage (N = 249) Meditation (N = 249) T'ai Chi (N = 249) Knowledge about Therapy (%) 1 – 2 (1="no knowledge") 69 44 52 72 91 3 17 22 24 15 6 4 – 5 (5="a lot of knowledge") 14 34 24 13 3 Ever tried therapy (%) 18 54 38 27 8 Ever tried therapy for LBP (%) 11 45 24 7 0.4 Median helpfulness for LBP among prior users (0 to 10 scale) 5 6 7 5 ** Pain or harm reported by prior users (%) 13 23 13 5 16 Median expectation of helpfulness for current LBP (0 to 10 scale) 5 5 7 3 5 Did not provide expectation rating (%) 25 10 9 12 24 High expectations of helpfulness for current LBP (7 to 10 on 0 to 10 scale) (%) 19 28 48 15 16 Very likely to try therapy if primary care provider thought reasonable and no extra cost (%) 64 51 69 27 41 Very likely to try therapy if primary care provider thought reasonable and $10 co-pay (%) 51 42 56 NA NA NA = Not Asked. * Each column refers to a specific therapy and the specific question about the therapy is shown in the first column. ** Only 1 person had tried t'ai chi for low back pain previously. All variables, except expectations of helpfulness of current LBP (where % are given in the table) have missing values for < 5% of respondents. More than half of the participants had tried chiropractic compared with 38% who had tried massage and substantially fewer who had tried the other therapies (Table 3 ). No demographic characteristics were related consistently to use of these therapies (Table 4 ). Chiropractic and massage were also the most commonly used of the therapies specifically for low back pain. Users of massage rated treatment helpfulness higher than did users of other therapies (Table 3 ). Reports of harm or increased pain were highest for chiropractic (23%) and lowest for meditation (5%). Respondents believed that massage would be most helpful for their current back pain (median rating of 7) and that meditation would be least helpful (median rating of 3) (Table 3 ). One quarter of all respondents were unable to rate their expectation of acupuncture or t'ai chi, compared to about 10% for the other therapies. Respondents 65 years of age or older were less optimistic than younger respondents about the helpfulness of acupuncture and massage (Table 4 ). High expectations of helpfulness of chiropractic were more common in those with high knowledge of this therapy and high expectations of helpfulness of acupuncture were more common among those who had tried it (Table 4 ). More than half of the respondents said they would be "very likely" to try acupuncture, chiropractic, or massage if provided by their health plan for no additional cost and their physician felt it was reasonable. Fewer respondents said they would be "very likely" to try meditation training (27%) or t'ai chi (41%) under those circumstances (Table 3 ). In logistic regression models, the strongest predictors of being very likely to try a particular therapy were high expectations of a therapy and, for meditation, prior use of the therapy (Table 4 ). About 80% of those very likely to try acupuncture, chiropractic, or massage for no additional cost were also very likely to try it for a $10 per visit co-pay (Table 3 ). Paralleling the finding for free care, the strongest predictor of willingness to try a therapy for a $10 per visit co-pay was high expectations of success for that therapy. Respondents from Boston were more willing to try acupuncture. Those reporting harm or pain from chiropractic were less willing to try this therapy again. Willingness to participate in a clinical trial More than half of respondents were "definitely willing" to participate in each of two hypothetical clinical trials about which they were asked and less than 5% were definitely unwilling to participate (Table 5 ). When asked which of the treatments in each trial they would most prefer, respondents preferred massage and acupuncture in the trial of acupuncture, chiropractic, and massage and, in the second trial, strongly preferred massage to meditation. However, a significant fraction (24%) expressed a preference for t'ai chi. We found no demographic, back pain, or CAM characteristics associated with being "definitely willing" to participate in the hypothetical trial of acupuncture, chiropractic, and massage. People who were "definitely willing" to participate in the hypothetical trial of massage, meditation, and t'ai chi were more likely to have high expectations of meditation (OR = 3.1, 95% CI = 1.4 – 7.0). Table 5 Willingness to Participate in Clinical Trials of CAM Therapies for Low Back Pain and Preference for Therapies Percent (N = 249) Definitely willing to participate in clinical trial of acupuncture, chiropractic, massage, and a self-help back pain book (%)* 62 Preferred treatment among above: Massage 43 Acupuncture 35 Chiropractic 18 None or Other 3 Book 1 Definitely willing to participate in clinical trial of massage, meditation, t'ai chi, and a self-help back pain book (%)** 53 Preferred treatment among above: Massage 63 T'ai Chi training 24 Book 5 Meditation training 4 None or Other 4 Missing values – < 4% of responses for each variable are missing. Your healthplan is thinking about conducting a study evaluating several treatments for people with chronic low back pain. In this study, participants would have a one in four chance of being assigned to one of the following treatments: acupuncture, chiropractic, massage, or a book designed to help patients better understand their low back pain. Participants would be expected to try the treatment they were assigned to at least once. Participants would still retain access to their usual care and participation in this study would be free. If you were asked to take part in a study like this would you be willing to participate? * Same question was asked, but the treatments were massage, meditation training, and t'ai chi training. Discussion Our findings suggest that many patients would be willing to try specific CAM therapies for back pain, especially if they had high expectations for their helpfulness. Interestingly, we found no consistent relationships between high expectations for a particular therapy and either previous use of that therapy or high self-perceived knowledge of that therapy. Our findings regarding knowledge, previous use, and expectations for these therapies were largely similar for Seattle and Boston and for older and younger adults. However, those over 65 years old were less likely to have high expectations of acupuncture and massage and to have tried massage previously. Since we conducted the study in two metropolitan areas where CAM use is fairly common, our results might not represent the CAM views of patients with back pain in more rural areas or in other regions of the country. Another limitation of our study was that 30% of people we attempted to contact could not be assessed for eligibility, leading to the possibility of a high non-response rate. Because we have almost no information on the characteristics of the individuals with unknown eligibility, we do not know if they differ from those included in the study, and cannot adequately estimate the magnitude and direction of potential biases regarding interest in CAM that might exist in our sample. However, the fractions of individuals who were unable to be assessed for eligibility were similar among those less than 65 years of age and those 65 and older in each metropolitan area (45% vs. 40% in Boston, respectively; 20% vs. 20% in Seattle). Respondents showed a clear preference for receiving hands-on treatments delivered by a practitioner compared to attending classes that teach self-care techniques. Whether this reflects a preference for provider-oriented, more passive, therapies or the belief that classes teaching these specific self-care therapies would be less effective is not clear. Unfortunately, our interview did not include questions about yoga, which has recently received more popular press than meditation or t'ai chi as a self-care therapy for back pain [ 9 , 10 ]. Survey respondents were not enthusiastic about "meditation training" as a treatment for back pain. Relatively few of those who indicated prior use of meditation for physical or mental health problems had used the forms of meditation most commonly taught in a medical setting (e.g., mindfulness meditation). Consequently, studies recruiting patients to participate in interventions including meditation training may need to carefully describe the treatment in terms of a concrete goal (e.g., stress reduction). There is still relatively little knowledge about and experience with acupuncture and t'ai chi even in Boston and Seattle where use of CAM therapies is generally high. In fact, about one – quarter of respondents were unable to provide an expectation of the helpfulness of acupuncture or t'ai chi. Nevertheless, substantial fractions of participants were willing to try acupuncture and t'ai chi as a treatment if their primary care provider thought it reasonable, and in the case of acupuncture, even if they had to pay a $10 co-pay each visit. Our finding that people in our sample reported being almost as willing to try acupuncture as massage, despite less knowledge of, expectations about and experience with it, is intriguing and requires further inquiry. Although participants in this study reported more knowledge of and experience with chiropractic, they were more enthusiastic about massage. A recent survey [ 11 ]of 46,000 Consumer Reports subscribers found that among those who had experienced back pain, the relatively few who had tried deep tissue massage rated it more favorably than those who had tried medications or physical therapy. The use of massage in this country has been growing steadily since the 1960's, with the largest increases in the 1990's [ 12 ]. In fact, in surveys of CAM use in the US population conducted in 1990 and 1997, Eisenberg et al. [ 7 ] found that massage as a treatment for various medical conditions had increased 61% over the seven-year period, while chiropractic remained fairly stable. By 1997, the estimated percentage of US adults who had used chiropractic was similar to that who had used massage, 11%. The relative popularity of massage may result from the more positive experiences of those who have tried it compared with chiropractic or acupuncture, and higher expectations that massage would be helpful for their current pain. Moreover, chiropractic users were more likely to report treatment related "harm" or "pain" than were users of massage. Implications for clinical trials Most survey respondents indicated they were "very willing" to participate in our two hypothetical clinical trials evaluating different CAM treatments for chronic back pain. Massage was the preferred treatment in both trials, but more than one in five survey respondents stated a preference for acupuncture and t'ai chi. In view of the long-standing popularity of chiropractic, surprisingly few respondents reported chiropractic as their top choice. Nonetheless the finding that massage was substantially more popular than chiropractic mirrors the results among acute low back pain patients in a clinical trial who were randomized to a choice of acupuncture, chiropractic, massage, or usual care or to usual care alone [ 13 ]. In that study, 52% of the participants said they would choose massage if given a choice, compared with only 24% who said they would choose chiropractic if given a choice. This finding could reflect the fact that many people have access to chiropractic as part of their current health care coverage [ 14 ]. Despite low levels of knowledge about t'ai chi and acupuncture, the finding that over 40% of respondents indicated they were very likely to try these therapies suggests that recruiting enough subjects for clinical trials involving these therapies may be feasible if moderate to large patient populations are available. Recruiting patients for meditation trials, however, is likely to be difficult. Consequently, when we recruited patients for a pilot trial that included a stress reduction intervention based on the principles of mindfulness meditation, we chose to describe it as "Mindfulness Based Stress Reduction" rather than mindfulness meditation. We believe that clinical trials evaluating obviously different treatments for chronic low back pain, such as massage and meditation, may have problems retaining subjects who do not receive the treatment (e.g., massage) that attracted them to the study. This problem may be exacerbated if patients have an exceptionally strong preference (or dislike) for one treatment. Inclusion of multiple CAM modalities in a single study risks tempting potential participants to sign up for the study in the hope of receiving a desired treatment, and then dropping out if they receive a different treatment. In addition, if one treatment is vastly more popular than another, it could be difficult to disentangle the effects of patient expectations and treatment efficacy per se, leading to difficulties in interpreting positive study outcomes. This problem is compounded by concerns about the subjective nature of back pain outcomes, the difficulty in masking participants to study treatment, and the strong skepticism of some researchers that CAM treatments can be effective, even when results are impressive. Masking patients to treatment is quite difficult in studies of many types of conventional as well as CAM treatments if the treatments involve a physical modality, such as massage, or active participation of the patient in the treatment, as in t'ai chi. In such circumstances, using masked outcomes assessors is important to minimize bias. We also suggest that patient (and provider) expectations for treatment and prior experience with each treatment, be measured and, if appropriate, controlled for in the analyses. Finally, if a particular therapy is shown effective in clinical trials in different populations, mechanistic studies will be important for determining how these therapies achieve their effects. Such studies are especially important to convince skeptics that CAM therapies actually have specific effects. In the meantime, the high and rising public interest in CAM therapies, especially for musculoskeletal conditions [ 12 ], highlights the importance of evaluating the effectiveness of various CAM treatments for back pain and our findings suggest that recruiting for these efforts may not be difficult. Conclusions Most patients in our sample were interested in trying options for treating chronic back pain that lie outside the conventional medical spectrum, even within the context of a clinical trial. This was true even among patients who had relatively little knowledge of or experience with the therapy. Given our limited knowledge about the effectiveness of most CAM therapies, there is a clear need for additional studies evaluating their effectiveness. Fortunately, our results suggest that researchers will not find it difficult to recruit patients interested in participating in clinical trials of many of the CAM therapies. Competing interests None declared. Authors' contributions KJS participated in the development of the questionnaire and took primary responsibility for the analysis of the data and for drafting the manuscript. DCC was the PI on one of the grants funding the study, participated in the development of the questionnaire and the analysis of the data and played a major role in drafting the manuscript. MTC participated in the development of the questionnaire and the analysis of the data. JE and JBS coordinated the project and oversaw the data collection. RBD participated in the analysis of the data and provided statistical oversight. DME was the Principal Investigator on one of the grants funding the study, participated in the development of the questionnaire and the analysis of the data. All authors read and approved the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC503394.xml
516772	Sequence periodicity of Escherichia coli is concentrated in intergenic regions	Background Sequence periodicity with a period close to the DNA helical repeat is a very basic genomic property. This genomic feature was demonstrated for many prokaryotic genomes. The Escherichia coli sequences display the period close to 11 base pairs. Results Here we demonstrate that practically only ApA/TpT dinucleotides contribute to overall dinucleotide periodicity in Escherichia coli . The noncoding sequences reveal this periodicity much more prominently compared to protein-coding sequences. The sequence periodicity of ApC/GpT, ApT and GpC dinucleotides along the Escherichia coli K-12 is found to be located as well mainly within the intergenic regions. Conclusions The observed concentration of the dinucleotide sequence periodicity in the intergenic regions of E. coli suggests that the periodicity is a typical property of prokaryotic intergenic regions. We suppose that this preferential distribution of dinucleotide periodicity serves many biological functions; first of all, the regulation of transcription.	Background DNA sequence periodicity with the period about 10–11 base pairs (bp) has been long known in eukaryotic DNA sequences. It was discovered recently in prokaryotic sequences as well [ 1 - 6 ]. The periodicity in Eubacteria sequences usually shows the period close to 11 bp [ 1 ]. This period is clearly different from the structural helical period of 10.5–10.6 bp/turn [ 7 , 8 ]. The difference was interpreted [ 1 , 2 ] as a possible reflection of the sequence dependent writhe of prokaryotic DNA. In the work [ 9 ] it was demonstrated that the periodicity in the bacterial genomes, in E. coli as well, is distributed in a non-uniform way, in scattered segments of the size 100–150 bases. It was also known for a long time that quite a few DNA promoter regions of E. coli possess the sequence periodicity of AA and TT dinucleotides [ 10 ]. The sequence periodicity of AA/TT dinucleotides is frequently associated with sequence-dependent DNA curvature, which is known to play an important role in the initiation of transcription of many genes (for reviews, see [ 11 - 15 ]). Using different models and approaches for prediction of intrinsic DNA curvature it was shown that many E. coli promoters have upstream curved sequences [ 16 , 17 ]. Pedersen et al. [ 18 ] showed that promoter area frequently has an unusual sequence structure. This region possesses higher DNA curvature, more rigid and less stable. Moreover, in our study of prokaryotic terminators of transcription (Hosid and Bolshoy, submitted) we have found that in E. coli DNA curvature peaks are frequently located downstream of the CDS. Since the dinucleotide periodicity with the period close to the helical repeat is associated with DNA intrinsic curvature [ 19 - 23 ], the curvature distribution along DNA would suggest similar distribution of DNA sequence periodicity. In this work, the sequence dinucleotide periodicity in E. coli and its distribution along the genome are systematically analyzed. A strong preference of intergenic regions to express the sequence periodicity of AA, AC, GC, and TT dinucleotides is discovered. Results and Discussion Positional autocorrelation analysis of the nucleotide sequences is an appropriate tool to detect all major characteristic distances in the sequences, the periodicities in particular. The complete genome of E. coli, as well as its coding and noncoding regions, was subjected to this procedure. Resulting autocorrelation profiles for all 16 dinucleotides (data not shown) were further analyzed by Fourier transform. In Fig. 1 the corresponding spectra are shown. The analysis demonstrates presence of the sequence periodicity of AA and TT dinucleotides with a period close to 11 bp mostly in intergenic regions, and weaker periodicity of AC and GC notably exclusively in intergenic regions. All 16 dinucleotides show periodicity of 3 bp, a well-known characteristics of the coding sequences, e.g. [ 24 , 25 ]. Weak 2 bp periodicity of AT and TA is also observed in intergenic regions. It indicates, perhaps, presence of tandem ApT repeats. A weak 10 bp periodicity of GC in intergenic regions, probably, corresponds to terminator regions (work in progress). The amplitudes of the 11 bp periodicity of AA and TT are the highest, even comparable with 3 bp coding periodicity. We, thus, focused on AA and TT distributions. Figure 1 Periodograms of the distance distributions of 16 dinucleotides in E. coli genome. The complete nucleotide sequence of E. coli, as well as subsets of its coding and noncoding regions, was subjected to the positional autocorrelation analysis for all 16 dinucleotides separately. Resulting autocorrelation profiles were after that analyzed by Fourier transform. The black lines correspond to the whole genome, the blue curves – to the coding sequences, and the red curves – to the noncoding sequences. To screen the genome of E. coli and find out where the periodical regions are located, we chose the period 11.2 bp [ 1 , 2 , 5 ] and this study (Fig. 1 ); and the window of 150 bp [ 9 , 26 ]. We used periodical AA and TT probes with the above periodicity to correlate with the E. coli genome sequence and to detect the periodical sites. This calculation shows that the periodicity is not evenly distributed along the E. coli genome. In Fig. 2 , the typical maps for several large segments of the E. coli genome are shown. The periodicity is distinctly located in certain regions. Many of the peaks observed are found to correspond to the intergenic regions (indicated by the black bars at the top). For example, two such peaks of periodicity in Fig. 2a correspond to the intergenic regions. Three such maxima are observed in Fig. 2b , three in Fig. 2c , and two in Fig. 2d . For the genome sections in Fig. 2 about 2/3 of the intergenic regions are associated with the local periodicity. Figure 2 Four examples of periodicity maps for fragments of E. coli genome. The maps were smoothed by running average with window 51 bp. The black bars on the top of the plot correspond to positions of intergenic regions. To verify the apparent strong correlation between the intergenic regions and AA/TT periodicity, we split intergenic regions in several families by size and analyzed the subsets separately by aligning (centering) the regions and summing up the respective local periodicity distributions. The combined maps for intergenic regions with a size from 50 to 150 bp, from 150 to 250 bp, from 250 to 350 bp, from 350 to 450 bp, and from 450 to 550 bp are shown in Fig. 3 . This figure demonstrates, indeed, that intergenic regions are typically periodic, irrespective of the size. The average amplitudes of the observed periodicities – 0.1–0.25 units – are comparable with the amplitudes in Fig. 2 , which indicates that, indeed a large proportion of the intergenic regions are periodical. Figure 3 The averaged maps of periodicity are synchronized at the centers of intergenic regions and smoothed by a running average of 51 bp. Five families of the intergenic regions with different lengths are presented: a) 100 ± 50, bp 1073 sequences, b) 200 ± 50 bp, 602 sequences, c) 300 ± 50 bp, 319 sequences, d) 400 ± 50 bp, 160 sequences, and e) 500 ± 50 bp, 78 sequences. The black bars at the bottom of the each figure correspond to the average intergenic region. The gray bands around black dashed lines correspond to standard deviations around randomized background. To verify the choice of the period 11.2 bases, we calculated the periodicity maps for highly populated group of the regions of the size 200 ± 50 bp, by assuming different periods in the range 10.5–12.5 bases. The resonance 3D plot in Fig. 4 indicates that the best-fit period is 11.3 ± 0.4 bp, which confirms earlier estimates of the E. coli DNA sequence periodicity. Figure 4 The 3D resonance plot for the intergenic regions of length 200 ± 50 bp. The maximum of resonance the plot corresponds to period 11.3 ± 0.4 bp. The contour around the maximum is also shown as a projection at the base line level. The spectral analysis (Fig. 1 ) and examples of the periodicity distribution maps (Fig. 2 ) show that apart from described correlation among the intergenic regions and AA/TT periodicity, there are numerous sites of periodicity located within coding sequences. Work is in progress to find out the functional relevance, if any, of these sites. Conclusions The observed concentration of the sequence periodicity in the intergenic regions corroborates earlier results and suggests that the periodicity is a typical property of the intergenic regions. Methods Genome data The sequence of the whole genome of Escherichia coli K-12 MG1655, locus U00096, 4639221 base pairs, was taken from the National Center of Biotechnology Information . Intergenic regions were identified in accordance with the annotation to this genome of E. coli and gathered in a separate dataset. Fourier transform of positional autocorrelation function Autocorrelation profile X was calculated for each dinucleotide separately. For the calculation of ApA autocorrelation, for example, we calculated the number of occurrences of pairs ApA – ApA in a distance k , and designated it by X k . Spectral analysis of autocorrelation profile X was obtained using the following formulae: where f p is normalized wave-function amplitude of period p , X is an autocorrelation profile for one chosen dinucleotide, X i is its value in position i , is its average value, and W is a maximal considered autocorrelation distance (in our case 100 bp). Sequence periodicity As a probe of periodicity the sine waves with period T were taken to describe idealized periodical distribution of AA and TT dinucleotides within window W . The probes were correlated with E. coli sequences by moving the probes along the sequences and calculating the value C for every position. where i is an index of a dinucleotide position in the window W and The value C max is introduced for the normalization purposes. It is calculated as follows: where i is a position in the window W and Ideally periodical sequence segments would be, therefore, described by C = 1, while segments with no periodicity would correspond to C = 0. The results of these calculations are presented as maps of the sequence periodicity. The four sample maps are shown in Fig. 2a,2b,2c,2d . Synchronization of the maps The maps around intergenic regions were combined (summed) separately for the groups of similar sizes of the intergenic regions. Five such groups were analyzed: 100 ± 50 bp, 200 ± 50 bp, 300 ± 50 bp, 400 ± 50 bp, and 500 ± 50 bp. For each group the maps were synchronized at the respective intergenic centers and the sums of the maps were calculated and smoothed by a running average within 51 bp. The standard deviations for the combined plots were estimated by generating random sequences of the same size and dinucleotides composition for each group separately and averaging the respective periodicity maps. The resonance plot The resonance 3D plot for the intergenic regions of length 200 ± 50 bp was built from calculations with different periods T in the interval 10–12.5 bp. One-third (202) of the most periodic maps of this group was taken for the calculation. The maps for different periods T were smoothed five times by a running average over 51 bp. The baselines were set to 0. The surface of 3D plot was smoothed 3 times by a running average over 9 point square elements, on the grid with separations 0.1 bp for T , and 20 bp for sequence position. Competing interests None declared. Authors' contributions SH carried out all graphics. ENT and AB participated in the design of the study and analysis of results. All authors drafted the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516772.xml
546192	Rapid, single-tube method for quantitative preparation and analysis of RNA and DNA in samples as small as one cell	Background Current methods for accurate quantification of nucleic acids typically begin with a template preparation step in which DNA and/or RNA are freed of bound proteins and are then purified. Isolation of RNA is particularly challenging because this molecule is sensitive to elevated temperatures and is degraded by RNases, which therefore have to be immediately inactivated upon cell lysis. Many protocols for nucleic acids purification, reverse transcription of RNA and/or amplification of DNA require repeated transfers from tube to tube and other manipulations during which materials may be lost. Results This paper introduces a novel and highly reliable single-tube method for rapid cell lysis, followed by quantitative preparation and analysis of both RNA and/or DNA molecules in small samples. In contrast to previous approaches, this procedure allows all steps to be carried out by sequential dilution in a single tube, without chemical extraction or binding to a matrix. We demonstrate the utility of this method by quantification of four genes, Xist , Sry and the two heat-inducible hsp70i ( hsp70.1 and hsp70.3 ), as well as their RNA transcripts in single mouse embryos and in isolated blastomeres. Conclusion This method virtually eliminates losses of nucleic acids and is sensitive and accurate down to single molecules.	Background Real-time polymerase chain reaction (PCR) in combination with reverse transcription (RT) provides a powerful tool for accurate quantification of DNA and RNA copy numbers and has opened the way to the study of subtle modulations of gene expression in small numbers of cells, as well as small-scale genetic analyses aimed at establishing chromosome numbers, the presence of mutations, or allele dropout. The reliability of these measurements, however, depends on the accuracy of each step, including preparation and recovery of RNA and/or DNA, reverse transcription of RNA into cDNA, and quantifiable and specific amplification of all desired sequences. The importance of optimizing each of these steps is well recognized [ 1 ], as is the need to minimize the number of tube-to-tube transfers in order to avoid the loss of templates and decrease the risk of contamination. This risk is posed by environmental RNases, material carried over from sample to sample, as well as previously generated amplicons present on laboratory equipment. Sequential performance of several steps in a single-tube is therefore highly desirable, especially when starting with small numbers of target molecules, such as chromosomes of individual cells or a few virus particles [ 2 - 5 ]. Our laboratory has already demonstrated that bound proteins prevent reliable PCR amplification of genomic DNA and that a thorough proteolytic digestion followed by heat inactivation solves this problem [ 6 ]. For accurate gene expression studies, RNA molecules also need to be released intact and free of proteins from all subcellular compartments, but proteases cannot be used both because they are not fast enough to inhibit the RNases (particularly endogenous RNases, released in the sample upon cell disruption) and because RNA is sensitive to the high temperatures required for protease inactivation [ 7 ]. Commercial kits for RNA purification therefore commonly employ either chaotropic agents or lysis buffers containing strong detergents, or a combination of the two, in order to achieve rapid denaturation of proteins. Nucleic acids are then extracted to remove these chemicals, because their presence interferes with subsequent enzymatic reactions. Alternatively, some RT-PCR kits bypass nucleic acid purification in favor of a simple dilution step, but in this case only a small aliquot of the lysed sample can be added to the RT mixture, due to volume restrictions. This approach introduces imprecision of its own and makes single cell analysis impossible. On the other hand, gentler lysis conditions that are compatible with single-tube analysis of a whole small sample do not remove proteins completely, resulting in substandard template preparations. For instance, protocols involving simple freeze-thaw cycles to produce cell lysis do not generate protein-free RNA or DNA. Similarly, mild detergents that do not lyse the nuclear membrane preclude quantification of DNA or RNA located in the nucleus, and are unlikely to completely remove proteins bound to cytoplasmic RNA. The chaotropic agent guanidine isothiocyanate (GITC) has long been the chemical of choice for nucleic acid preparation. It is particularly useful for RNA studies [ 8 , 9 ], because it rapidly denatures all cellular proteins, as well as serum proteins, including RNases, added to culture media. GITC has also proven superior to all other tested methods for the recovery of either DNA or RNA extracted from mummified tissue [ 10 ]. Due to its strong chemical action, GITC at high concentrations offers the further advantage of allowing safe storage of the samples until they are processed for quantification. For the same reason, however, all traditional protocols require removal of GITC prior to RT and PCR to avoid inactivation of the enzymes involved. Typically GITC is removed by extraction with phenol-chloroform and purification of the nucleic acids through alcohol precipitation cycles [ 9 ], or by absorption of the freed RNA to a matrix such as glass fiber filters, silica-gel membranes, magnetic beads or proprietary compositions, usually followed by elution in a relatively large volume. Both these approaches are time-consuming and involve a number of steps that can lead to incomplete RNA recovery. In view of these limitations we devised an alternative strategy in which the sample is collected and denatured in a minimal volume of a GITC solution, briefly heated to allow dry storage, and, when needed, is directly analyzed in the same tube by performing quantitative RT-PCR in a volume large enough to lower GITC concentration to negligible levels. This new procedure, hereafter referred to as PurAmp (patent pending) and described here in full for the first time, is the only available method that allows a whole sample, such as a single embryo or cell, to be processed from lysis to RT-PCR in the same tube, under conditions that permit precise quantification of both RNA and genomic DNA copy numbers. This fully optimized method is sensitive enough to detect specific sequences within a single chromosome in one cell, yet robust enough to measure the presence of thousands of RNA molecules released from hundreds of cells. PurAmp has made it possible for us to conveniently investigate expression levels of two gene types essential for early mouse embryo development: Xist , responsible for X-inactivation and dosage compensation in female cells [ 11 - 13 ], and the heat-shock inducible hsp70.1 and hsp70.3 , jointly called hsp70i [ 14 , 15 ]. Xist RNA is a noncoding transcript that exerts its particular function of gene-silencer by coating the inactive X-chromosome. Unlike hsp70i RNA and most other mRNAs, it is therefore localized in the cell nucleus and particularly challenging to extract. Besides their biological relevance, both Xist and the hsp70i genes offer the advantage of naturally-occurring unambiguous controls for the specificity of transcripts amplified with RT-PCR. Xist RNA is, in fact, virtually absent from male cleavage stage embryos [ 16 - 19 ], while hsp70i RNA is predominantly synthesized in response to stress although minimal levels of hsp70i transcripts are normally present in embryonic cells. A careful quantitative analysis of hsp70i heat and culture stress-response in preimplantation embryos at different developmental stages, and its implications for development, will be presented elsewhere (C. Hartshorn, A. Anshelevich and L. J. Wangh, in preparation). In addition, we have been able to detect and quantify the genomic sequences of Xist , the hsp70i and the male sex-determining gene Sry [ 20 ]. Because the number of these sequences is known and very low in samples comprised by an identifiable number of cells, such as early embryos that have undergone few cleavages, their precise quantification provided an optimal internal control to demonstrate the strength of this novel technique. The quick and reliable detection of DNA (or RNA) in very low copy number is, however, not limited to the role of internal control, but holds much wider utility for a variety of applications such as genetic studies and detection of viral sequences in a sample. Further, this method can be successfully employed for the study of individual cells, as shown by the present report, and is easily adaptable to analysis of subcellular fractions or aliquots from bodily fluids; it also minimizes the use of toxic chemicals and the possibility of contamination, while allowing dry storage of the collected samples. All these features concur to render PurAmp ideally suitable for fast but highly sensitive gene and gene expression screening of multiple samples, including small whole specimens or fractions of larger ones. Results Single-tube Xist RNA, Xist DNA and Sry DNA quantification in individual male and female blastocysts The PurAmp method presented in this study is performed in a single tube from cell lysis to cDNA or genomic DNA amplification, thus eliminating possible loss of template molecules due to procedures such as phase separation and recovery, repeated washing and re-suspension of nucleic acid pellets, elution from binding matrices and vessel-to-vessel transfer. This strategy offers an immediate improvement in the precision of gene expression analyses, at the same time shortening considerably the experimental protocol compared to traditional methods. In order to validate our method, we initially measured the Xist RNA content of a group of female mouse embryos at the blastocysts stage and quantified Xist and Sry DNA copy numbers in their male counterparts. Our previous analysis of these parameters in single female and male mouse embryos at different developmental stages provided us with an ample pool of data obtained with commercially available nucleic acids preparation methods [ 17 - 19 ], which we used as a reference for comparison with our new results. Figure 1 shows the real-time PCR plots obtained from six PurAmp-treated single embryos. Following RT, the accumulation of multiplexed Xist/Sry amplicons was detected using two molecular beacons conjugated to different fluorescent dyes. Both amplicons span intronless sequences of the genes [ 18 ], thus allowing in either case the simultaneous measurement of cDNA (when present) and genomic DNA copies. Each color in the plots of Fig. 1 identifies a specific embryo and is used for both its Sry and Xist signal. Three embryos were identified as male based on the presence of the Sry amplicon (Fig. 1 , upper panel, lines in blue hues) and on the fact that in each case the Xist fluorescent signal (Fig. 1 , lower panel, lines in blue hues) arose at the same "threshold cycle" (C T ) as the Sry signal (see Methods for a definition of C T and details on signal quantification). This indicates the presence of the same number of copies of Sry and Xist templates, as expected for male blastomeres that contain one copy of the Sry gene on the Y-chromosome and one copy of the Xist gene on the X-chromosome. Neither gene is expressed in male blastocysts [ 16 , 17 ], and therefore these three samples contained only genomic Xist and Sry DNA. The three remaining embryos did not generate any Sry signal (Fig. 1 , upper panel, lines in red hues), while their Xist signals arose significantly earlier than the others (Fig. 1 , lower panel, compare lines in red and blue hues). Based on these data they could be identified as female embryos which contained thousands of copies of Xist RNA in addition to two copies of genomic Xist DNA per cell. These results are fully in agreement with those of our earlier analyses using traditional methods of nucleic acid preparation. Quantification of the real-time PCR data obtained from the three male embryos in Fig. 1 indicated that, on average, each embryo contained 125 ± 83 (mean ± s.d.) copies of Xist genomic DNA, consistent with the previous estimate of 165 ± 101 genomes per male blastocyst [ 17 ]. Both values are higher than the expected cell number per embryo at this stage (60–100, depending on culture conditions), due to endoreduplication in trophoblast cells [ 21 ], a phenomenon also responsible for some variability between samples. In contrast, a total of five female blastocysts analyzed via the PurAmp method yielded an average of 12,600 ± 5079 copies of Xist cDNA + genomic DNA per embryo. Since each female blastocyst contains about 250 copies of the Xist genomic sequence (twice the number of a male embryo), the accumulation of Xist RNA per female blastocyst averages above 12,000 copies, considerably higher than our previous measurement of 6797 ± 2894 copies obtained using a multistep nucleic acids isolation procedure [ 17 ]. The data in Figure 2 demonstrate that the amplification efficiency of both the Xist and Sry sequences is neither decreased nor increased by the presence of diluted denaturing solution (0.4 mM GITC, as in the PurAmp protocol) during real-time PCR. In fact, additional experiments revealed that Xist/Sry real-time PCR was unaffected by a GITC concentration as high as 20 mM (not shown). Taken together these results suggest that the higher levels of Xist cDNA measured using the PurAmp method are due to an improvement in RNA recovery at the initial step of cell lysis. Quantification of low-to-high Xist RNA and DNA copy numbers in single embryos and blastomeres As shown above, blastocysts are comprised by many cells and contain hundreds of copies of the Xist and Sry genes and thousands of copies of Xist transcripts, with rather wide sample-to-sample fluctuations. In order to more carefully determine the quantitative capability of the new assay, we next analyzed embryos at earlier developmental stages containing lower and, in some cases, precisely known numbers of template copies. Figure 3 illustrates the real-time PCR plots of the Xist amplicons generated in the course of two separate experiments by, right-to-left, i) a 3-cell male embryo (yellow); ii) a 4-cell male embryo (green); iii) a single blastomere isolated from a 4-cell female embryo (light purple); iiii) a 4-cell female embryo (red); iiiii) a female blastocyst (blue). The gender of each embryo was confirmed by the detection of an Sry -specific fluorescent signal in male samples (Fig. 3 , inset). The quantitative analysis of these results confirmed that the 3-cell male embryo contained 3 copies of the Xist gene, while the 4-cell male embryo contained 6 copies of the Xist gene, indicating that DNA duplication had occurred in two of the blastomeres. It has long been known that two of the blastomeres of a 4-cell embryo divide ahead of the other two [ 22 ], an observation in agreement with our finding. The numbers of Xist templates measured in these male embryos also confirmed the expectation that these samples did not contain Xist RNA because Xist is not expressed in male cells [ 13 , 17 ]. Conversely, the Xist signal of the female 4-cell embryo arose about five cycles earlier than the Xist signal of the 4-cell male embryo (compare red and green curves), denoting the presence of Xist transcripts (157 copies of Xist RNA assuming diploidy of all cells and calculated from a total of 165 cDNA + genomic DNA templates), albeit at considerably lower levels than those measured in the female blastocyst (9750 copies of Xist RNA assuming the aforementioned average number of genomes per blastocyst of 125, and based on 10,000 copies of total Xist cDNA + genomic DNA templates). These measurements are consistent with other studies demonstrating that Xist transcripts are accumulated in the developing embryo beginning at the late 2-cell stage [ 23 ] and with our previously published Xist developmental profile [ 17 ]. The quantitative accuracy of the PurAmp method was further confirmed by the fact that the Xist signal generated by a single blastomere isolated from a 4-cell female embryo arose 2.2 cycles later than the Xist signal of the whole 4-cell female embryo (compare light purple and red curves, Fig. 3 ). A left-to-right shift of 2 cycles is exactly what is expected for a fourfold decrease in template numbers quantified by real-time PCR amplification. Figure 4 illustrates this point by showing Xist RNA + genomic DNA levels in two individual blastomeres isolated from a 4-cell female embryo, as compared to Xist template levels measured in intact 4-cell embryos of different sex and then calculated on a per cell basis. Even at this early developmental stage, the presence of Xist RNA is clearly detectable in the female samples, absent from the male, and not affected by the blastomere isolation procedure [ 18 ]. Hsp70i RNA and DNA measurements in heat shocked and non-heat shocked single embryos and blastomeres In order to more extensively test the validity of the PurAmp approach to template quantification in single cells, we measured transcript levels of the heat shock-inducible genes hsp70.1 and hsp70.3 in blastomeres isolated from embryos at the pre-compaction 8-cell stage, when cells can be easily counted and separated. The sequences of these two genes are almost entirely identical, they are located on the same chromosome and they encode the same protein [ 14 , 15 ]. For this reason, there has been some confusion in their identification and nomenclature in past studies. Heat-inducible hsp70 transcription, previously indicated as hsp70.1 expression, is now more precisely designated as the sum of hsp70.1 and hsp70.3 ( hsp70i ) RNAs. A preliminary set of experiments was carried out on embryos at the blastocyst stage, when heat shock response is fully established [ 14 ], with the goal of evaluating the effect of hyperthermia on hsp70i expression. Like Sry , the hsp70i are naturally intronless genes and therefore once again our pair of PCR primers simultaneously amplified both genomic DNA and cDNA sequences. The data in Table 1 clearly indicated that heat shock (see Methods) produced a sharp rise in hsp70i template numbers due to the presence of thousands of copies of hsp70i RNA, although these numbers were considerably lower when samples were prepared with a multistep/multitube phenol-chloroform extraction [ 17 , 18 ] rather than with the PurAmp method. During these initial experiments, embryos were allowed to recover for 30–40 minutes after heat shock. Under these conditions, however, only five out of seven blastocysts exposed to hyperthermia showed an increase in hsp70i RNA levels. Based on these results, the duration of the recovery period was increased to at least two hours in all following experiments, eliminating the finding of "non responsive" embryos. Copy numbers of hsp70i RNA were then quantified in whole 8-cell embryos that had or had not been exposed to a temperature increase, as summarized in Table 2 . The number of hsp70i genomic DNA copies measured in the absence of RT was consistent with the presence of four copies of the genes per cell, one hsp70.1 and one hsp70.3 on each chromosome 17, and with the fact that some of the cells analyzed had already duplicated their DNA. Only a minimal amount of hsp70i RNA, calculated as the difference of template copy numbers obtained with and without reverse transcription ( hsp70i cDNA + genomic DNA less hsp70i genomic DNA), was present in non-heated embryos, indicating that the embryos were not stressed by culture conditions [ 24 ]. Expression of hsp70i RNA increased sharply after a 30-minute heat treatment, followed by a recovery period of either 2 or 3 hours necessary for transcripts synthesis and accumulation. Some of the heat-shocked embryos were harvested intact, while others were dissociated in single blastomeres. The average numbers of hsp70i template copies per blastomere were calculated from the isolated cells (not all cells of dissected 8-cell embryos could be recovered) and compared to the average per blastomere values obtained from whole embryos. The results show that the amount of hsp70i RNA + genomic DNA per cell calculated by these two approaches is very similar. A post-heating recovery period of 3 hours rather than 2 hours increased the hsp70i RNA levels only slightly, indicating that the onset of transcription and the major build-up in RNA occur quickly. Single cells derived from non-heated embryos contained only trace amounts of hsp70i RNA, consistent with whole embryo measurements, and PCR efficiency was, again, unaffected by the PurAmp components (not shown). Efficiency of DNase treatment within the PurAmp protocol The PurAmp method described above automatically results in the quantitative recovery of genomic DNA, which can then be used as a quality control and a convenient internal standard for the simultaneous recovery and measurement of mRNA [[ 17 - 19 ], see Discussion]. Applications such as microarray analysis, however, are based on RNA-only amplification. For this reason, we introduced a DNase digestion step preceding RT in the Xist/Sry PurAmp protocol and analyzed the efficacy with which the genomic DNA was degraded. Genome numbers in embryos at the morula stage were calculated by counting Xist and Sry copies in five male samples (as detailed for blastocysts, see above) and averaged at 21.3 ± 8.9, consistent with the fact that embryos at this stage are normally comprised of 16-to-32 cells. After treatment with DNase I, only 0.8 ± 1.5 genomes per embryo were still present in a group of 15 male samples, demonstrating that the enzyme had successfully degraded 96.3% of the DNA. Figure 5 illustrates the effects of DNase digestion on Xist (upper panel) and Sry (lower panel) DNA in a group of eight male and eight female single embryos. As expected, control male samples (blue lines) contained equal numbers of Xist and Sry copies, as shown by the equal C T values, corresponding to the genomic DNA copy number. Both amplicons were absent from DNase-treated embryos that could be identified as male because they were devoid of Xist RNA (yellow lines). In contrast, control female samples (red lines) contained Xist RNA and DNA but lacked Sry . DNase treatment caused a delay in the Xist signals arising from female embryos (green lines), consistent with the elimination of all 42 copies of genomic Xist DNA in each embryo plus some decrease in the number of Xist transcripts. The amount of RNA recovered in these samples averaged at 75% of control levels. The fact that some RNA is lost during the DNase step is not surprising as it is well known that some RNA hydrolysis is unavoidable (see Discussion), and is not linked to the single-tube procedure. We anticipate that further optimization of the available DNase protocols and reagents will minimize this problem. Thus, the data in Fig. 5 demonstrate overall that a DNase digestion step can be successfully inserted within the PurAmp procedure, without disruption of the DNase enzymatic activity. Discussion It is increasingly clear that individual cells in a population do not exhibit identical patterns of gene expression and, hence, that expression profiling is more informative if it is quantitative and carried out at the single cell level [ 25 - 27 ]. This consideration is particularly relevant to current efforts aimed at understanding early mammalian embryogenesis in which totipotent cells generated during the first few cell divisions gradually become committed to particular lines of development. The mechanisms of this process are under intensive scrutiny and appear to be rooted in differential gene expression resulting from epigenetic modifications. It is in this context that we have been measuring RNA levels in single cells of cleaving embryos [ 18 , 19 ] and it is to increase the reliability of these measurements that we have now developed the PurAmp method. This completely single-tube approach is easy to use, eliminates loss of material, and improves the quantitative accuracy of gene and gene expression studies. First, cell lysis and protein denaturation occur very rapidly upon delivery of the sample to crystalline GITC, thus ensuring both protein removal from DNA and RNA and inactivation of cellular nuclease that would otherwise quickly degrade RNA [ 8 ]. Transcripts localized in the nucleus, such as Xist RNA, are freed and made available to reverse transcription as well as cytoplasmic mRNA molecules, a result unattainable by mild detergent treatment that leaves nuclei intact [ 25 ]. Second, the brief heating period after cell lysis enhances complete denaturation of proteins and also reduces the volume of the sample, thereby further increasing the guanidine concentration. The semi-dry sample can then be safely stored without risk of nuclease activity. Third, carrying out cell lysis in nanoliter volumes allows a manifold dilution of the chaotropic agent after addition of the RT cocktail, so that RT can be performed on the whole sample and in the same vessel in which it was collected without inhibition of the enzymatic activity. Finally, RT and PCR can be carried out immediately after cell lysis, rather than after cumbersome and lengthy nucleic acid preparation procedures, thereby further reducing the time required to process many samples, as well as the risk of contamination. Our quantitative measurements of Xist RNA levels in developing mouse embryos highlight one of several merits of the PurAmp method over the traditional, multistep approach to nucleic acids purification [ 17 , 18 ]. In fact, while genome numbers obtained with the two methods are similar as expected, Xist RNA levels are higher with the single-tube protocol. The same culture conditions and procedures were used in the two groups of experiments, making it unlikely that differences in embryo quality were the cause of the increase in Xis t RNA. We, therefore, conclude that the higher levels of Xist RNA observed using the new procedure reflect improved template preparation with efficient inactivation of RNases and reduced loss of RNA molecules. Xist RNA is known to trigger X-chromosome silencing through interactions with numerous proteins and possibly with the nuclear matrix scaffold [ 28 , 29 ]. The results presented in this study clearly show that the very high initial concentration of GITC thoroughly breaks up protein-RNA interactions, but the denaturant does not inhibit subsequent RT once is diluted. Similarly, our quantification of hsp70i templates in heat-shocked blastocysts supports the view that larger pools of RNA are detected in PurAmp-treated samples than when using phase separation-based nucleic acid extraction. The later method, in fact, presents several steps that require extreme care to avoid loss of material, including complete recovery of the upper phase, thorough precipitation of all nucleic acids molecules, and repeated re-suspension and washing of barely visible pellets. All these manipulations render the results obtained with this technique particularly operator-dependent, while, in contrast, PurAmp simply requires sequential addition of reagents into the same tube. Once the sample is delivered to the LysoDot in the reaction vessel (see Figure 6 and Methods section), therefore, this technique is much less dependent on the operator's specific skill. Individual blastomeres of pre-compaction mouse embryos are easily harvested due to their size, and laser zona drilling efficiently preserves RNA pools allowing dependable single-cell analysis [ 18 ]. We thus used measurements of Xist and hsp70i RNA levels in single blastomeres to further validate the quantitative accuracy and reliability of the PurAmp method, as shown by the fact that transcript levels in individual cells are comparable to average RNA levels per cell calculated from whole embryos. Based on these results we anticipate that PurAmp will prove useful for quantification of RNA levels in small pieces of tissue from many sources, as well as single cells and even fractions of cells such as neuronal dendrites and axons [ 30 ]. The small volume in which denaturation is carried out is also amenable to analysis of biological material isolated by laser capture microdissection or laser pressure catapulting [ 31 ]. Genomic DNA has recently been proposed as the optimal standard for gene expression studies [ 32 ] and it is the required internal standard when cDNA is quantified with the strategy of amplification competition [ 33 ]. In this case, DNA and RNA are purified together, as in our experiments, and one set of primers is designed to co-amplify a genomic sequence that spans an intron as well as the corresponding intronless cDNA. The alternative strategy that we developed makes use of primer sets that do not span introns and therefore amplify genomic DNA sequences that have the same length and composition as their corresponding cDNA's, eliminating any possible difference in PCR efficiency for the two types of templates [[ 17 - 19 ]; C. Hartshorn, A. Anshelevich and L. J. Wangh, in preparation]. We have found it very informative to measure genomic DNA copy numbers in addition to RNA levels of the genes under study, because this strategy provides a reliable internal control for primer specificity and for nucleic acids recovery, particularly when performing single-cell analyses. In the case of early mouse embryos, detection of the Sr y gene, which is not expressed at those stages, has also allowed us to identify the sex of each embryo. The recovery and quantification of genomic DNA together with RNA has previously enabled us to establish genome number averages for developing embryos [ 17 , 19 ]. While these numbers are very similar to the number of cells in early embryos, measurements of DNA copies are more accurate because they indicate whether the cells have completed S phase. Genome quantification becomes even more critical after the late 8-cell stage, because the embryos compact making it very difficult to count individual cells. Moreover, endoreduplication takes place in trophoblasts at the blastocyst stage, greatly increasing the number of genome copies present in those cells [ 21 ]. All these factors render the counting of DNA copy numbers important if gene expression data are to be calculated on a per-genome basis, independently from a cell's ploidy. A further reason to preserve DNA molecules in preparations for RT-PCR is that all DNase digestion protocols currently available lead to partial hydrolysis of RNA when the enzyme is heat-inactivated in the presence of divalent cations at the end of the reaction [ 34 ]. We have consistently found a decrease in amplified cDNA in DNase-treated samples, particularly when performing RNA isolation with traditional methods (unpublished results), even when a chelating agent was added prior to the heating step. Incomplete RNA recovery after DNase inactivation in the presence of EDTA was not evident in past reports, due to the use of non-quantitative methods of nucleic acids analysis [ 35 ]. Our real-time PCR results, however, agree with numerous more recent findings [see ref. [ 36 ] for an overview of DNase-related problems]. Efforts have been made, therefore, to devise alternative ways to eliminate the DNase once digestion has occurred. These methods, however, depend on removal of the enzyme which, in turn, implies manipulations such as phenol extraction that may still generate nucleic acids loss. For all of the above reasons as well as the fact that we are working with very small amounts of material, we prefer single-tube preparation-to-amplification of both RNA and DNA templates, an approach made possible for the first time by the procedure described in this paper. Previously reported single-tube template preparation protocols, in fact, have been aimed at measuring only specific RNAs and employ lysis buffers containing low concentrations of the mild detergent NP-40 [ 25 , 37 ], or they bypass the lysis step altogether and are limited to neuron studies [ 38 ]. While these methods are valuable for detection of protein-free RNA molecules, they utilize non-denaturing conditions, as clearly demonstrated by the addition of proteic RNase inhibitors to the extraction buffers, and therefore preclude quantitative analysis of DNA [ 6 ] as well as of protein-bound RNA pools. Conclusions Due to its ability to thoroughly remove proteins from both RNA and DNA molecules in a rapid and simple way, PurAmp is suitable to a wide variety of applications, including gene expression quantification, studies on genetic mutations, and viral detection. Because the presence of DNA is undesirable for certain applications, such as microarray-based expression profiling, we have also shown that a DNase digestion step can be easily included in the single-tube format. We anticipate that treatment with other enzymes, such as cellulase in the case of plant cells, can similarly be inserted into the PurAmp protocol to digest other "undesired" components of particular cells prior to amplification. Thus, PurAmp is a very flexible technique that affords the investigator a variety of ways of processing the contents of a lysed sample with a heightened level of precision (Fig. 6 ). Methods Embryo culture and single blastomere isolation For most experiments, frozen late 2-cell stage embryos (B6C3F1 females bred with B6D2F1 males) were obtained from Embryotech Laboratories, Inc. (Wilmington, MA), and were cultured as previously described [ 17 ] until the desired stage of development. For the DNase experiment, frozen 8-cell embryos obtained from the same source were grown to the morula stage. Blastocyst stage embryos used for hsp70i measurements were also grown from frozen 8-cell embryos. Single blastomeres were isolated from either 4-cell embryos (for Xist measurements) or pre-compaction 8-cell embryos (for hsp70i measurements) after drilling the zona using a ZILOS-tk™ zona infrared laser optical system (beam = 1480 nm) (Hamilton Thorne Biosciences, Inc., Beverly, MA), according to a procedure developed in our laboratory and described elsewhere [ 18 , 19 ]. PurAmp multiplex measurements of Xist/Sry RNA + DNA in individual embryos or blastomeres All experimental procedures were carried out using rigorous precautions aimed at avoiding or destroying environmental RNases contamination [ 17 - 19 ]. Dried droplets of denaturing solution, hereafter called "LysoDots", were prepared prior to embryo collection by delivering 20-nl aliquots of the denaturing solution (see below) to the inside surface of the lids of PCR-grade reaction tubes (Applied Biosystems, Foster City, CA). Precise measurement of the droplets size was obtained following the method previously described by Wangh [ 39 ]. The denaturing solution composition was: 0.25% sarcosyl, 2 M GITC, 100 mM β-mercapto-ethanol, 0.01 M sodium citrate, pH 7.0 (all reagents from Stratagene, La Jolla, CA), 1% (vol/vol) dimethylsulfoxide (Sigma Chemical Company, St. Louis, MO). LysoDots were prepared in advance, allowed to dry under sterile conditions, and then stored at room temperature in closed PCR tubes. Immediately before harvesting, individual embryos were placed in 3 ml of Dulbecco's PBS devoid of calcium and magnesium chloride [ 17 ]. Dulbecco's PBS containing 0.4% polyvinyl pyrrolidone (both products from Sigma) was used when isolating single blastomeres [ 18 ]. After one wash in the same buffer, each embryo or cell was aspirated into a glass capillary having an internal diameter of 0.2 mm [ 39 ] and tapered at the end so that the inner volume of the tapered tip would contain about 20 nl. Tapering was obtained by pulling the glass capillaries in a Micro-Pipette Puller (Industrial Science Associates, Inc., Ridgewood, NY). The embryo (or cell) was expelled directly onto the LysoDot in a volume of PBS as close as possible to 20 nl. Microscope observation revealed that the GITC crystals dissolved instantly upon addition of the sample-containing PBS and, thus, that cell lysis occurred immediately. Tubes were closed upside down and heated at 75–77°C for 5 minutes, after which their content was once again dry or semi-dry. The samples were then stored at -20°C until the next step. In order to perform reverse transcription, each sample was carefully re-solubilized in the lid by addition of 6 μl of Random Hexamers mixture (4.2 ng/μl) in DEPC-treated water (all RT reagents were from a ThermoScript™ RT-PCR System kit, Invitrogen, Life Technologies, Carlsbad, CA). Tubes were closed, inverted, briefly centrifuged and incubated for 5 minutes at 65°C in order to allow primer/RNA hybridization. The remaining reagents needed for RT were then added to the tube in a volume of 4 μl, and the reaction was carried out according to the protocol suggested by the manufacturer. As previously described [ 17 - 19 ], all RT reagents were used at the suggested concentrations except for the absence of DTT, but volumes were halved so that each assay was performed in just 10 μl, which increased to 10.5 μl after RNase H digestion. The full volume of each sample was then mixed with 89.5 μl of complete PCR amplification cocktail containing sequence-specific molecular beacons as detection probes [ 40 ]. Multiplex real-time PCR of Xist and Sry genomic DNA + cDNA templates was thus performed in a final volume of 100 μl, as detailed elsewhere, in the presence of 4 units of Taq DNA polymerase (Promega, Madison, WI) [ 18 , 19 ]. Real-time PCR was carried out in an ABI Prism ® 7700 Sequence Detector (Applied Biosystems, Foster City, CA) and fluorescence readings were taken at the annealing temperature. PurAmp assay for hsp70i RNA and DNA measurements in individual embryos or blastomeres Embryos were heat-shocked at 43°C for 30 minutes, followed by a recovery period of 30–40 minutes (blastocysts), or 2–3 hours (8-cell embryos, as indicated) at 37°C. The hsp70i assay was carried out similarly to the one for the Xist/Sry multiplex, by sequential dilutions of denaturant, RT and PCR reagents. The procedure for collection and lysis of the samples was the same. Dry samples were re-solubilized with 6 μl of random decamer primers (8.3 μM) in nuclease-free water (all RT reagents were from a Cells-to-cDNA™ II kit, Ambion, Inc., Austin, TX). After a 3 minute incubation at 75°C to optimize primer binding to RNA, all other reagents needed for RT were added to the sample and the reaction was carried out according to the manufacturer's instructions. As for the Xist/Sry assay, all RT mixture components were used at the suggested concentrations, but volumes were halved so that RT was performed in a final volume of 10 μl. An RNase H digestion step was included at the end of RT, as in the case of the Xist/Sry assay. Real-time PCR was carried out in a final volume of 100 μl, by adding the PCR reagents to the sample after completion of RNase H digestion. The chosen hsp70i primers were localized at positions 1245/1305 of the hsp70.1 GenBank sequence with accession number M35021 (5' CCGCCTACTTCAACGAC 3', upstream primer; 5' ATCCGCAGCACGTTTA 3', downstream primer) and were identical to sequences within the hsp70.3 gene, previously known as hsp70A1 (GenBank sequence with accession number M76613) [ 14 ]. Because the hsp70i was the only amplicon generated in this assay, it was not necessary to design a sequence-specific detection probe. In this case, SYBR ® Green, a fluorescent dye that binds to double-stranded DNA, was used as fluorescent probe for real-time PCR. The specificity and purity of the amplicon was confirmed by both gel electrophoresis and analysis of the melt profile, as previously described [ 17 ]. The composition of the cocktail for hsp70i PCR was the following: 50 mM Tris, pH 8.3, 3 mM MgCl 2 , 0.3 μM each primer, 0.25 mM each dNTP, 1:62,500 SYBR Green (from a "10,000X concentrate in DMSO" purchased from FMC BioProducts, Rockland, ME), and 4 units of Taq DNA polymerase (Promega, Madison, WI). The polymerase was incubated at a 1:1 (v/v) ratio with Platinum ® Taq antibody (Invitrogen) for 5 minutes before addition to the reaction mixture (hotstart PCR). The cycling profile was: 95°C for 5 minutes; 10 cycles consisting of the following four steps: 95°C (20 sec), 64°C (30 sec), 72°C (30 sec), 84°C (15 sec); 35 cycles with the following four steps: 95°C (20 sec), 59°C (30 sec), 72°C (30 sec), 84°C (15 sec). Fluorescence readings were acquired at 84°C, in order to exclude fluorescent signals due to the possible formation of primer dimers late in the reaction. A number of embryos were also processed as "No RT" controls, with the same protocol used for the other samples but without inclusion of reverse transcriptase in the RT mixture. These controls were used to quantify hsp70i genomic DNA copy numbers in the absence of cDNA [ 17 ]. Multistep/multitube nucleic acid extraction For the preliminary studies on hsp70i expression in blastocysts, some of the embryos were processed using a commercially available multistep/multitube kit, as previously described [ 18 ]. Briefly, nucleic acids (DNA and RNA) from each sample were purified using phenol:chloroform:isoamyl alcohol phase separation (Micro RNA Isolation Kit, Stratagene, La Jolla, CA) with a ratio of 100 μl of phenol and 45 μl of chloroform/isoamyl alcohol solution per assay. Transfer RNA (10 μg/assay; Sigma Chemical Company, St. Louis, MO) was added as a co-precipitant. Pellets were washed twice, once in isopropanol followed by overnight precipitation at -20°C and once in 75% ethanol, and then nucleic acids were reverse transcribed and analyzed by real-time PCR exactly as detailed above for the PurAmp-treated samples. Quantification of Xist , Sry and hsp70i amplicons Calculation of template copy numbers was based on the "threshold cycle" (C T ) at which each fluorescent signal was first detected above background. C T values were compared to standard scales obtained from analysis of male mouse genomes at known copy numbers, as detailed previously [ 17 - 19 ] and further exemplified in the Result section. Briefly, a two-fold difference in the number of templates amplified results in a shift of one cycle between two C T determinations. A lower C T value indicates an earlier detection of the fluorescent signal and therefore more templates present at the start of the reaction [reviewed in ref. [ 41 ]]. One male mouse genome contains one copy of Sry (Y-chromosome), one copy of Xist (X-chromosome) and four copies of hsp70i (two of hsp70.1 and two of hsp70.3 , both on chromosome 17). Insertion of a DNase I digestion step within the Xist/Sry PurAmp protocol Embryos were grown to the morula stage and were individually harvested and lysed as detailed above. Controls were processed for multiplex detection of Xist/Sry with the described PurAmp protocol. RNA-only samples were prepared by inserting a DNase digestion step prior to RT, as follows. Lysed, dried samples were re-suspended with 4 μl of DNase mixture containing: 20 mM Tris-HCl, pH 8.4; 2 mM MgCl 2 , 50 mM KCl (Invitrogen's DNase I Reaction Buffer) and 1 unit of DNase I (Ambion) in nuclease-free water. After an incubation of 20 minutes at room temperature, the reaction was terminated by adding 1 μl of a 10 mM EDTA solution, pH 8.0 (Invitrogen). The nuclease was inactivated by heating the samples at 65°C for 10 minutes, according to the protocol recommended by Invitrogen. One μl of a 25 ng/μl RT primer solution was then added to the sample, so that the final primer concentration was now the same as in the assay without DNase (see above). RT and PCR were then carried out as detailed for the "No DNase" assay. Authors' contributions CH devised the finalized form of the PurAmp method, carried out the Xist/Sry measurements and drafted the manuscript. AA established heat shock conditions and performed the hsp70i quantification experiments. LJW coordinated the study and contributed to all aspects of its design.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546192.xml
340948	Learning-Induced Improvement in Encoding and Decoding of Specific Movement Directions by Neurons in the Primary Motor Cortex	Many recent studies describe learning-related changes in sensory and motor areas, but few have directly probed for improvement in neuronal coding after learning. We used information theory to analyze single-cell activity from the primary motor cortex of monkeys, before and after learning a local rotational visuomotor task. We show that after learning, neurons in the primary motor cortex conveyed more information about the direction of movement and did so with relation to their directional sensitivity. Similar to recent findings in sensory systems, this specific improvement in encoding is correlated with an increase in the slope of the neurons' tuning curve. We further demonstrate that the improved information after learning enables a more accurate reconstruction of movement direction from neuronal populations. Our results suggest that similar mechanisms govern learning in sensory and motor areas and provide further evidence for a tight relationship between the locality of learning and the properties of neurons; namely, cells only show plasticity if their preferred direction is near the training one. The results also suggest that simple learning tasks can enhance the performance of brain–machine interfaces.	Introduction Practice can induce behavioral improvement that is often specific to the situation experienced during the practice sessions (or “training”). Such findings suggest that changes occur in neurons with fine selectivity (or “tuning”) for the stimuli experienced or the movements made during training. In the visual system, for example, behavioral improvement is specific to the trained stimulus, such as the orientation of a light bar ( Fiorentini and Berardi 1980 ; Crist et al. 1997 ), and is paralleled by specific changes in neurons that are tuned to the orientation of a light bar ( Schoups et al. 2001 ) or, in other experiments, the direction of visual motion ( Zohary et al. 1994 ). In the auditory system, changes in response properties of single neurons and cochleotopic maps are specific to the parameters characterizing the sound ( Suga et al. 2002 ). In the motor system, skill acquisition induces expansion in the cortical representation of the used forelimb ( Nudo et al. 1996 ) and enhance synaptic connections in the trained contralateral hemisphere ( Rioult-Pedotti et al. 2000 ). A line of studies found that when monkeys perform reaching movements and adapt to directional errors induced by force fields, primary motor cortex (M1) cells shift their preferred direction (PD) in about the same way as for the muscle activity needed to perform the task ( Gandolfo et al. 2000 ; Li et al. 2001 ; Padoa-Schioppa et al. 2002 ). We have recently shown that learning a local rotational visuomotor task can induce an elevation in the activity of single neurons in M1 ( Paz et al. 2003 ) and that these changes are observed only in a specific subpopulation of neurons, those with a PD close to the movement direction used during the learning. Whereas many studies indicate that learning can induce specific changes in brain activity, this finding does not necessarily imply that newly learned skills are “better” represented in the brain. The crucial question is this: Do neurons encode task parameters, such as movement direction, any better after learning? In the motor system, such improved encoding ( Chen and Wise 1997 ) can be used for decoding by downstream areas and as an efference copy for further computation ( Wolpert and Ghahramani 2000 ; Sommer and Wurtz 2002 ). It can also be used by an external observer to allow for more accurate prediction of behavior ( Laubach et al. 2000 ). In this paper, we examine two questions. First, do learning-induced changes in firing rates provide more information on the task? And, second, what aspect of the cells' activity contributes mostly to this improvement? To address the first question, we employed an information-theory analysis ( Cover and Thomas 1991 ; Rieke et al. 1997 ) to calculate the mutual information (MuI) (see Figure 2 ) between cells' activity and direction of movement. Informational measures have two relevant advantages. First, they use the full distribution (estimated from the data) of neuronal activity and do not assume any specific shape of the tuning curve or noise distribution. This allows for a more fine-tuned examination of learning-related changes. Second, they provide a measure as to how well different directions can be differentiated, based on neuronal activity. To address the second question, we examined two features of the neuronal response that could contribute to the increase in information: response variability and the slope of the tuning curve. Finally, to demonstrate that the observed increase in information can be extracted, we use the neuronal activity to decode the actual movement direction. Figure 2 MuI between Neuronal Activity and Direction of Movement The example shows a simulation of the activity of one cell during 64 movements to evenly spaced eight directions, presented in a random order (eight trials per direction). Each dot in the raster plots a and b describes the spike count of the cell in a specific trial. Without prior knowledge about the direction of movement (A), a large uncertainty exists about the responses of the neuron. However, ordering the trials according to the movement direction (B) reveals a large reduction in the uncertainty about the cell responses. The probability p( r,d ) of observing a trial with direction d and spike count r is shown in (C); along with a specific conditional distribution p ( r \| d 0 = 90) . The entropy is a measure of the uncertainty about movement direction: H(D) = log(8) = 3 bits, in the case that all eight directions have equal probability to occur. The conditional entropy is defined as and describes the mean uncertainty about direction given the cell response. The MuI I ( R ; D ) = H ( D ) − H ( D \| R ) measures the reduction in uncertainty about movement direction given the response of the cell. The MuI is symmetric, in the sense that it also measures the reduction in uncertainty about cell response given the direction of movement I ( R ; D ) = H ( D ) − H ( D \| R ) . This relation is graphically depicted in (D). Results Monkeys adapted to visuomotor rotations on a daily basis by altering the relationship between the visual feedback (cursor) and the hand movement ( Figure 1 ). Learning was confined to only one target in space, i.e., learning that is local in direction. We tested neuronal sensitivity to direction by comparing the information content conveyed in the firing rate of single cells during the pre- and post-learning epochs (identical task of standard movements to eight directions spanning the two-dimensional working surface, only differentiated by a learning epoch). We specifically looked for a change in representation that was related selectively to the learned direction, i.e., the hand direction that was used to bring the cursor to the target during the transformation. Figure 1 Behavioral Paradigm and Movement Kinematics (A) Session flow (left to right). Every session (day) consisted of pre-learning, learning, post-learning, and relearning epochs. Pre- and post-learning epochs were standard eight-target tasks with a default (one-to-one) mapping between cursor movement and the movement of the hand. In the learning epoch, only one target (upwards) appeared, and a visuomotor rotational transformation was imposed on the relationship between movement of the hand and cursor movement. The example shown is for a transform of −90° (see Materials and Methods for a full description). (B–D) Similar kinematics pre- and post-learning. (B) Example of 1-day trajectories from the two epochs; the transform in this session was of –45°. (C) Velocity profiles. Peak velocity was slightly lower in the post-learning epoch ( t -test, p = 0.05), but no difference was found between the learned direction and other directions ( t -test, p = 0.3). (D) Improvement in directional deviation was calculated as the deviation of the instantaneous hand direction from the required target direction, calculated every 10 ms starting from the go-signal. All four movement types (learned and nonlearned, pre- and post-learning) exhibited the same temporal pattern. Here and for analysis of neuronal activity, we excluded the first trials in the post-learning epoch—those exhibiting significant aftereffects due to learning. Activity was measured from the hold period that immediately follows the target appearance, but before the go-signal, and was therefore termed preparatory activity (PA). There were three reasons for this choice. First, such PA has been reported in many motor cortices and is thought to participate in movement planning and in computing visuomotor transformations ( Kurata and Wise 1988 ; Alexander and Crutcher 1990 ; Kalaska et al. 1997 ; Shen and Alexander 1997 ; Zhang et al. 1997 ; Kakei et al. 2001 ). Second, as previously found in this experimental paradigm, learning-related changes have only been reported for this period ( Paz et al. 2003 ). Third, as a means of eliminating any kinematic-related changes ( Wise et al. 1998 ), we further verified that movements shared similar kinematics before and after learning (see Materials and Methods ; Figure 1 ). Mutual Information The MuI between one-cell activity and direction of movement is exemplified in Figure 2 . We compared the MuI between pre- and post-learning ( Figure 3 A). The figure depicts the distributions of MuI between direction and spike count for all cells ( Figure 3 A, corrected for bias) for pre-learning (dashed line) and for post-learning (solid line). No difference was found between the MuI on the population level, either by comparing the distributions (Kolmogorov–Smirnoff, p = 0.3) or by comparing their means (paired t -test, p = 0.53). We further tested the average information about direction conveyed by each spike by normalizing each cell's information by its firing rate and again found no significant difference (inset in Figure 3 A; Kolmogorov–Smirnoff, p = 0.25, paired t -test, p = 0.7). Figure 3 Comparing MuI of Single Cells Pre- and Post-Learning (A) Distributions of single-cell information about direction of movement pre-learning (dashed) and post-learning (solid). No significant difference was found between the distributions (Kolmogorov–Smirnoff, p = 0.3). The inset shows the MuI per spike, calculated by dividing the information per cell by the cell's firing rate (Kolmogorov–Smirnoff, p = 0.25). (B) Improvement in information of individual cells. Histogram of p -values for all cells; a significant ( p < 0.01, χ 2 ) number of cells ( n = 37) had a p -value greater than 0.95, representing cells that significantly increase their information content about direction after learning; 18 cells had a p -value lower than 0.05, representing cells that decreased their information content, but this was found to be only marginally significant ( p = 0.06, χ 2 ). (C) Histograms of difference in information, post- minus pre-learning, for all cells (upper) and only for cells that increase ( p > 0.95) or decrease ( p < 0.05) their information content significantly (lower). (D) Circular histogram for PD of cells that significantly increased their information. The cells' PDs were normalized to the learned direction in each cell recording session, revealing a unimodal distribution (Rayleigh test, p < 0.05). The upper inset shows the circular histogram for all cells and lower inset shows the circular histogram for cells that decreased their information; in both cases, the distributions seem homogenous (Rayleigh test, p > 0.1). Although the population as a whole did not change significantly, single neurons could still increase or decrease their information about direction. To explore this, we probed each neuron individually for changes in MuI. Using a bootstrap method, we shuffled trials from pre- and post-learning and randomly reselected two different groups of trials, we then calculated the MuI for each group and the difference between the two MuIs. The procedure was repeated 1,000 times to produce a distribution of MuI differences. The actual MuI difference (between the pre- and post-learning) was compared to this distribution to obtain a p -value. A high p -value means that the MuI in the post-learning epoch was significantly higher than the MuI in the pre-learning epoch. Figure 3 B plots the histogram of the p -values for all cells. A significant number of cells showed an increase in MuI with a p -value larger than 0.95 (black in Figure 3 B; n = 37 out of 177, p < 0.01, χ 2 ), a nearly significant number of cells showed a decrease in MuI with a p -value lower than 0.05 (white/transparent in Figure 3 B; n = 18, p = 0.06), while all the rest did not (gray in Figure 3 B). We also examined the actual change in information content for all cells ( Figure 3 C, upper) and specifically for the cells that had a significant change ( Figure 3 C, lower). Following the rationale explained in the Introduction, the association between the learned parameter (direction) in local rotational transformations and the sensitivity of many cells to direction, we probed for a relation between cells' PD and the learned direction. Figure 3 D plots a circular histogram of PDs of cells that exhibited a significant increase in their MuI. For the plot, we normalized each cell's PD to the learned direction in its recording session, and this revealed a unimodal distribution (Rayleigh test, p < 0.05) with its center on the learned direction. In contrast, the PD distributions of the whole population ( Figure 3 D, upper inset) and of cells that significantly decrease their information content ( Figure 3 D, lower inset) did not exhibit this trend and seemed homogenous. To test that this change in information is indeed owing to the learning of visuomotor transformations and not owing to the mere repetition of a single movement during the learning epoch, we conducted the same analysis for control, repetition sessions. Only a nonsignificant ( p > 0.1, χ 2 ) number of cells (eight out of 126) had a p -value greater than 0.95 ( Figure 4 A). Further, this population did not exhibit any specific distribution of PDs ( Figure 4 B; Rayleigh test, p > 0.1). Figure 4 Changes Were Not Observed after Mere Repetition of Movement to One Direction Same as in Figure 3B and 3D, but for control sessions that included the mere repetition of standard, nontransformed movement to one target during the learning epoch. The number of cells that exhibited an increase in their information content was not significant ([A] right bar, eight out of 126), and their distribution was homogenous and showed no specific relation to the direction of the repeated movement (B). Individual Information per Direction The MuI represents the information that a cell's spike count conveys about all the eight tested directions. We further investigated how much information a cell conveys about one direction in particular, which is termed the individual information per direction (DI) ( Rolls et al. 1997 ; Buracas et al. 1998 ) and is measured as the reduction in uncertainty about the spike counts, given a specific direction. We calculated the DI of each cell for each of the eight possible directions, pre- and post-learning. The distribution of the differences between the post-learning DI and pre-learning DI for the learned direction was significantly above zero ( Figure 5 A, “‘Learned”'). This indicates that after learning, cells' firing rates conveyed more information about the learned direction. Figure 5 A also shows that information about other nonlearned directions did not change. As with the MuI, to probe for the directional tuning of the cells, we plotted the circular histogram of PDs of cells that increased their information about the learned direction (a positive post-learning minus pre-learning). Again, a unimodal distribution (Rayleigh test, p = 0.01) was found with its peak on the learned direction ( Figure 5 B). Figure 5 Comparing Individual DI (A) Mean (with 95% confidence intervals, by fitting a Gaussian distribution) of post-learning information minus pre-learning information for one direction. Abscissa represents the distance from the learned-movement direction; all directions were normalized according to the learned direction in the cell's session. An increase is evident only for the learned-movement direction, with mean at 0.1 and 95% confidence intervals at 0.036 and 0.164. (B) Circular histogram of PDs for cells with a positive difference of post-learning minus pre-learning information about the learned direction (Rayleigh test, p = 0.01). Possible Origins for Improvement in Information Information theory makes use of the complete (estimated from data) stimulus–response distribution and thus does not tell us what feature in cell activity primarily contributed to the increase in information content. However, we found that the increase in information is specific to a single-learned direction and that cells with a PD close to the learned direction mainly contributed to this increase. We have previously reported that cells with PD close to the learned direction increased their firing rate after learning when movement was to the learned direction ( Paz et al. 2003 ). We therefore explored more closely this elevation in firing rates and its relationship to the increase in information content. Figure 6 A histograms the net changes in activity (post- minus pre-learning) at the cells' PDs for the whole population. Figure 6 B shows the same net changes for cells that significantly increased their information about direction, where a significant positive trend was found (by fitting a normal distribution; see legend to Figure 6 B). We further aligned each cell tuning curve on the cell's PD and calculated the average tuning curve. This revealed that this group of cells indeed elevated their activity mainly around their PD ( Figure 6 C). Figure 6 Learning-Induced Elevation of Information and Activity (A and B) Histograms of changes in firing rate in the PD (post-learning minus pre-learning) for all the cells (A) and for cells that significantly increased their information (B). The horizontal line below the histogram represents its mean and the 95% confidence intervals, by fitting a Gaussian distribution. (C) Average tuning curves (baseline subtracted, ± SEM) of cells that significantly increased their information ( n = 37). Comparing pre-learning (gray) and post-learning (black). Cell tuning curves were first aligned to each cell's PD. Two natural features of a cell's tuning curve can contribute to the improvement in information content. First, a cell can increase the slope of the tuning curve just near the learned direction, and thus small changes in direction can lead to a larger difference in the cell's response, providing a better differentiation of direction based on the neuronal activity (illustrated in Figure 7 A). Second, cells can reduce the variability of their response near the learned direction. This is also termed “reliability,” because when variability is low, each single report made by the cell is more reliable (illustrated in Figure 7 B). A standard method for characterizing this is the Fano factor ( Berry et al. 1997 ), calculated as the variance of the response divided by its mean. We correlated the net change in information content (post-learning minus pre-learning) to these two factors: change in slope near the learned direction ( Figure 7C1 – 7C3 ) and change in the Fano factor ( Figure 7D1 – 7D3 ). Figure 7 shows that whereas no systematic change in the corresponding factor was found for the whole population ( Figure 7 C1 for slope and Figure 7 D1 for FF), a significant positive trend was observed for the population of neurons that significantly increased their information after learning. This trend was obvious for the slope factor ( Figure 7 C2) and also, but to a much lesser extent, for the Fano factor ( Figure 7 D2). Figure 7C3 and 7D3 extends this relation and shows the correlation between the corresponding factor and the increase in information. A significant positive correlation was only found for the slope factor and only for cells that significantly increased their information ( Figure 7 C3, black asterisks and line). No correlation was observed between the change in slope and the change in information for the rest of the cells ( Figure 7 C3, gray dots) or between the change in Fano factor and the change in information, either for the whole population ( Figure 7 D3, gray dots) or for those that significantly increased their information ( Figure 7 D3, black asterisks). Further, the increase in the slope of the tuning curve near the learned direction was specific to this direction only and to cells that significantly increased their information content ( Figure 8 ). Figure 7 Increased Slope of Tuning Curve Is Correlated with the Increase in Information Possible mechanisms for improving the information content of single cells. (A and C) The slope of the tuning curve at the learned direction indicates the magnitude of change in activity in response to small changes in direction. The higher slope suggests that nearby directions can be discriminated better. (B and D) Reliability of coding. The variability at each direction indicates how well different directions can be differentiated based on single trials. (C1 and D1) Histograms of the difference between pre- and post-learning for the corresponding mechanism for the whole population of cells. The horizontal line below the histogram represents its mean and the 95% confidence intervals. (C2 and D2) Histograms for cells that significantly increased their information about direction. (C3 and D3) Correlation between the difference in information (post- minus pre-learning) and the corresponding mechanism. Gray dots are all the cells, and black asterisks are cells that significantly increased their information content. The black line represents the linear regression fit. The corresponding Pearson correlation (C) and its significance ( p -value) are designated. The histogram in (C2) is shifted to the right, indicating that cells that increased their information content also increased the slope of the tuning curve in the learned direction. In these cells only, a significant ( p = 0.002) correlation coefficient ( c = 0.492) was found. Figure 8 Slope Increase Is Specific to the Learned Direction Mean change (± SEM) in the slope of the tuning curve surrounding each direction, for cells that significantly increased their information content (black) and for the rest of the cells (gray). These results suggest that cells increased the slope of their tuning curve near the learned direction and improve the information content in their activity. Cells can use several strategies to do so and we considered three possibilities: first, by shifting their tuning curve and positioning the learned direction at a better “slope-wise” location on the tuning curve (illustrated in Figure 9 A); second, by narrowing the tuning curve ( Figure 9 B); and, third, by local changes increasing or decreasing specific points near the desired (learned) location ( Figure 9 C). Although the three possibilities are not mutually exclusive and might be interrelated, we attempted to distinguish among them by correlating the change in information to each one. Figure 9 D– 9 F shows that the increase in information was correlated with the increased firing rate at the learned direction ( Figure 9 F1– 9 F3), but not with shifts in PD ( Figure 9 D1– 9 D3) or with the narrowing of tuning curves ( Figure 9 E1– 9 E3). We therefore suggest that cells locally increased their firing rate to increase the slope of their tuning curve at the learned direction. Figure 9 Increased Information after Learning Is Correlated with Elevation of Firing Rate in the Learned Direction Possible mechanisms for increased slope of the tuning curve in the learned direction. (A and D) Shift of PD, i.e., shifting the whole tuning curve, may position the learned direction at a higher slope location. (B and E) Narrowing of the tuning curve, as measured by the width at half-height. (C and F) Local changes (Increase) in activity in the learned direction can increase the slope. This is similar to the observed learning-induced changes in our data (see Figure 6C). In (A)–(C), an illustration of the measured difference is indicated. (D1–D3, E1–E3, and F1–F3) Same format as in Figure 4 for the three possible mechanisms. The histogram in (F2) is shifted to the right, indicating that cells that increased their information content also elevated their firing rate in the learned direction. In these cells only a significant ( p < 0.001) correlation coefficient ( c = 0.566) was found (F3, asterisks and line). Decoding Movement Direction We hypothesized that the improved information regarding movements in the learned-movement direction would lead to an improved ability to reconstruct movements from population activity. To test this assumption, we applied two reconstruction methods: the population vector (PV) approach, a widely used decoding scheme for M1 activity ( Georgopoulos et al. 1988 ; Moran and Schwartz 1999 ), and a maximum a posteriori (MAP) estimator ( Sanger 1996 ). For the PV analysis, we selected 129 of the 177 cells, only including cells that exhibited directional tuning by a cosine fit. Neurons were pooled according to the learned-movement direction in their recording session, and we computed the PV from the pre-learning and post-learning activity. Figure 10 A shows the deviation of the PV direction, i.e., the difference between the PV prediction and the actual movement direction for the four possible learned-movement directions. A marked and statistically significant improvement was observed in the predicted direction ( p < 0.05 for all four learned directions, bootstrap and t -test). We verified that this improvement was due to learning in two ways: first, by shuffling trials from the pre-learning and the post-learning epochs, and second, by shuffling cells from days with different transformations. In both cases, the mean of the distribution of improvements was not significantly different from zero. Furthermore, the improvement in the PV prediction was specific to the learned-movement direction. Figure 10 B shows the signal-to-noise ratio (mean/SD) of improvements in PV accuracy (the difference between the accuracy of the pre-learning prediction and the post-learning prediction). We normalized each session directions to the learned direction in the session. A statistically significant improvement was found only for the learned-movement direction (χ 2 , p < 0.01). This improvement in the PV prediction can be accounted for by the enhanced firing of cells with a PD near the learned-movement direction, as shown above (see Figure 6 ). Cells with their PD close to the learned-movement direction made a larger contribution to the PV, but mostly when the movement was in that direction. Because each cell contributes a weighted vector in the direction of its own PD, only the learned-movement directions showed improvement in PV accuracy. This improvement in prediction due to altered directional tuning is reminiscence of studies that examined learning of visuomotor associations in frontal eye fields ( Chen and Wise 1996 , 1997 ) and of studies showing evolvement of directional tuning in M1 when monkeys received real-time visual feedback of brain-controlled trajectories ( Taylor et al. 2002 ). Figure 10 Improved Decoding of Movement Direction Only for the Learned Direction (A and B) Using PV. (A) PV errors given as the distance in degrees between the predicted and the actual direction for the four learned-movement directions (± SEM, bootstrap test). (B) Signal-to-noise ratio (mean/SD) of PV improvement (pre-learning deviation minus post-learning deviation) for all directions (four learned directions are pooled together and all other directions are normalized to them). A significant improvement was observed only for the learned direction ( p < 0.005, Bonferoni correction for multiple tests, i.e., the eight directions). (C and D) Using a MAP estimator, we predicted 100 times the actual hand direction using neuronal activity. Shown is the fraction of correct predictions for pre-learning (C) and post-learning (D). A significant increase was observed only for the learned direction ( p < 0.005, Bonferoni correction for multiple tests, i.e., the eight directions). The dashed line is the chance level (0.125). The PV method includes several assumptions about the coding and the decoding of the M1 population activity and is not guaranteed to be optimal ( Sanger 1994 ; Snippe 1996 ; Pouget et al. 2000 ; Scott et al. 2001 ). Therefore, we also tested the performance of a probabilistic approach. Using a MAP estimator, we predicted the movement direction for all possible directions, including the learned-movement direction pooled and normalized from all sessions. Figure 10 C depicts the success rate for 100 repetitions (by cross-validation) for each direction. Figure 10 D shows the same, but in the post-learning epoch. A higher success rate of correctly predicting the movement direction can be observed for learned direction only in the post-learning epoch (χ 2 , p < 0.01, chance level is at 0.125;dashed line in Figure 10 D). This indicates that after learning and by using this decoding method, we could better predict the actual movement direction from neuronal activity. Discussion This report describes improved encoding and decoding of specific directions by neurons in M1 of monkeys after learning a visuomotor skill that requires learning only for one direction in space. Our results suggest a close link between properties of neurons, such as directional tuning of cells, and learning a skill that is local in the same parameter, in this case direction, a finding that is concordant with ideas and findings in sensory systems ( Zohary et al. 1994 ; Suga et al. 2002 ; Sharma et al. 2003 ). The fact that the increased information we found was associated with an increased slope of the tuning curve, as also reported in a recent visual study ( Schoups et al. 2001 ), further suggests that similar mechanisms may govern neuronal interactions and learning throughout the central nervous system. The fact that improved information in neuronal activity was evident mainly for the learned direction is in accordance with studies showing confined generalization of learning a sensorimotor skill, one that requires adaptation to directional errors. The width of the behavioral generalization function (i.e., the angular distance from the learned direction where aftereffects could still be observed) was similar for our monkeys ( Paz et al. 2003 ) and in human studies, ranging from 45° ( Gandolfo et al. 1996 ; Krakauer et al. 2000 ) to 90° ( Imamizu et al. 1995 ; Thoroughman and Shadmehr 2000 ). The neuronal changes we previously observed occurred mainly for cells with PD within 30° of the learned direction, and the change in slope observed in this study was sharply focused and not seen for directions 45° away from the learned direction (see Figure 8 ; note, however, that changes in firing rate were wider [see Figure 6 ]). While narrower primitives reasonably lead to narrower generalization function ( Donchin et al. 2003 ), the exact generalization width depends not only on the primitives' width, but also on the connectivity and the specific model used. These are still largely unknown. An intriguing result in this study is that learning-related changes were observed and persistent in the post-learning epoch, when performing a standard task that required no transformation. Further, measured kinematics was the same as in the pre-learning epoch. If the improved information can be used, why isn't it? First, our monkeys were trained on a task that did not require improved performance in the standard task after learning, but did encourage them to reserve learning for future use of the same visuomotor task. This is in agreement with our previous report, showing that the monkeys retained the task until the performance of the relearning epoch (i.e., they exhibited immediate recall rather than learning in this second learning epoch), and suggests that the neuronal change should persist but somehow gated. Indeed, everyday behavior shows that we can learn new tasks without interfering with the performance of existing ones. An alternative possibility is that we did not measure the appropriate kinematic variable that was altered and improved due to the neuronal changes. For example, a task that would demand finer directional sensitivity (i.e., angular distance of less than 45°) might show a change in performance after learning. It is also worth noting that our experiment was performed in a local region in space and was not constrained to a specific posture ( Scott and Kalaska 1997 ) or joint or muscle combination ( Scott et al. 2001 ). Therefore, we cannot conclude that locality and specificity of change in information content are related to external direction of movement. Our results may be consistent with other reference frames as well ( Mussa-Ivaldi 1988 ; Todorov 2000 ). One important question is what kind of learning can induce such an increase in information content. Although psychophysics studies have shown that adapting to new kinematics and/or dynamics environments results in the formation of internal representations in the brain (reviewed by Kawato 1999 ; Wolpert and Ghahramani 2000 ), changes were also observed after extensive training and mere repetition ( Nudo et al. 1996 ). Moreover, many sensory systems exhibit stimulus-related adaptations ( Dragoi et al. 2000 ; Suga et al. 2002 ), where repeated presentations of a stimulus induce a change in activity of neurons. To control for this possibility, we conducted sessions with a repetition condition, which entailed a one-target task without angular transformations. Cells recorded in these sessions did not exhibit a change in their information content, and PV analysis produced similar results before and after repetition. An alternative explanation could be attention-related modulations ( Spitzer et al. 1988 ). We discuss elsewhere why this is an unlikely source for the changes we observed ( Paz et al. 2003 ), yet we cannot rule out the possibility that increased attention might lead to similar improvement in information. MuI measures are more often used in sensory research, describing the information that neurons convey about a presented stimulus, and only few papers have applied such measures to the motor system (e.g., Hatsopoulos et al. 1998 ). We believe this stems from the fact that in sensory systems, neurons respond to the stimulus, whereas in the motor system, neurons “cause” the movement. In this study, we treated direction of movement as a stimulus to which the neuron responds. This can be justified because MuI is a symmetric measure and the point of view can be reversed; e.g., we can interpret the results as neuronal activity → movement, rather than movement → neuronal activity. More importantly, frontal motor fields, M1 included, are only part of the brain's learning system and project to many brain areas that take part in processing an upcoming movement, such as the basal ganglia and cerebellum ( Middleton and Strick 2000 ). Therefore, M1 activity may be decoded by those areas involved in coplanning of the movement. Moreover, an efference copy of the planned motor command is probably used for error estimation and correction ( Wolpert and Ghahramani 2000 ; Sommer and Wurtz 2002 ). Indeed, we are aware of our movements before they have actually started ( Haggard and Magno 1999 ). This suggests that when learning new sensorimotor tasks, activity in M1 should not only produce the correct behavior, but also change in a way that enables other brain areas a better readout of the motor command. This will allow more efficient computation and better control of the forthcoming movement. Although higher information content implies better encoding by neurons, it does not entail better decoding; this is highly dependent on the algorithm used and on the error function introduced. Since our task involved manipulation of movement direction and since real-time prediction of movement trajectory has taken on major interest in recent years ( Wessberg et al. 2000 ; Taylor et al. 2002 ), we used the discrepancy between the actual movement direction and the predicted one from neuronal activity as the error signal (either categorical, for the MAP, or continuous, for the PV). The MAP method ( Sanger 1996 ; Zhang et al. 1998 ) is theoretically optimal ( Seung and Sompolinsky 1993 ) and requires fewer assumptions on the tuning-curve shapes and distribution of PDs ( Sanger 1994 ), but requires larger amounts of data to estimate the true distributions ( Pouget et al. 2000 ). The PV method has been shown to be robust in many scenarios ( Georgopoulos et al. 1988 ; Moran and Schwartz 1999 ) and very useful in predicting hand movement in real time ( Taylor et al. 2002 ). In our experiment, both methods yielded a better prediction of the learned-movement direction during its planning stages and long before its initiation (see also Laubach et al. 2000 ). Although we cannot determine whether neurons further downstream use this improved information to decode a previous layer of neurons, we believe it is possible. Further, our findings could lead to improved strategies for recovering trajectory information from populations of M1 cells ( Wessberg et al. 2000 ; Serruya et al. 2002 ; Taylor et al. 2002 ). The specificity of the learning is of high importance here. The large number of degrees of freedom, the complexity of movements, and the technical difficulties of recording many neurons simultaneously are only starting to be addressed, and a plausible strategy might require learning and practicing specific and essential movements. Our results suggest that this would modify brain activity in a way that would enable earlier and better readout of brain activity from fewer neurons. Materials and Methods The experimental setup and data acquisition procedures are described in detail in Paz et al. (2003 ). The sampled cells were taken from the same database. Physiological procedures. Two female rhesus ( Macaca mulatta ) monkeys (approximately 4.5 kg) were implanted with recording chambers (27 × 27 mm) above both the right and left hemispheres. Animal care and surgical procedures complied with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (rev. 1996) and with the Hebrew University guidelines supervised by the Institutional Committee for Animal Care and Use. The monkeys were seated in a dark chamber, and eight microelectrodes were introduced into each hemisphere. The electrode signals were amplified, filtered, and sorted (MCP-PLUS, Alpha-Omega, Nazareth, Israel), and all spike shapes were sampled at 24 KHz. We used a template-based method for real-time isolation of spike shapes (MSD, Alpha-Omega). Penetration locations were verified by MRI (Biospec Bruker 4.7 Tesla, Bruker BioScences, Billerica, Massachusetts, United States) before recordings. At the end of each session, we examined the activity of neurons evoked by passive manipulation of the limbs and applied intracortical microstimulation (50 ms of 200-μs cathodal pulses at 300 Hz) to evoke movements. Only penetration sites that evoked single-joint shoulder or elbow movement at thresholds of lesser than or equal to 40 μA were used in this study. In one monkey, we also made anatomical observations, to verify the accurate penetration sites relative to the central sulcus. Behavioral paradigm. Monkeys moved a manipulandum to control the movement of a cursor on a video screen located 50 cm from their torso and eyes with the goal of moving the cursor from a starting point at the center of the screen (origin) to a visual target in a delayed go-signal paradigm; this required the monkey to hold (as verified by hand velocity and EMG) the cursor in the origin circle for a random 750–1,500 ms after the target onset. The disappearance of the origin indicated the go-signal. In each session (day), four consecutive epochs were introduced: (1) pre-learning epoch (more than 100 trials), a standard, eight-target task in which the target direction was randomly chosen from eight possible directions uniformly distributed over the circle; (2) learning epoch (more than 30 trials), a transformed, one-target task in which only one target (upwards, 90°) was presented and a rotational transformation was introduced between the cursor on the screen and the manipulandum; (3) post-learning epoch (more than 100 trials), where the default eight-target task was presented again; and (4) relearning epoch, same as the learning epoch. Rotations were 90°, 45°, –45°, or –90° and were chosen randomly for each session, but fixed for the duration of the learning epoch in a session. Note that learning here is local in direction since only one target direction was introduced during the learning epoch. The term learned-movement direction refers to the direction of hand movement needed to bring the cursor to the target for these rotations (thus, there were four possible learned-movement directions in this study: 0°, 45°, 135°, and 180°, associated with the –90°, –45°, +45°, and +90° transforms, respectively). Monkeys were trained for several months with the standard eight-target, task but did not see the transformations before the recordings. To achieve learning on a daily basis during the whole recording period (rather than switching between pre-learned behaviors), a different rotational transformation was randomly chosen for each day from the set of four possible transformations. To observe systematic change in the activity of neurons, the same transformation was repeated (greater than or equal to four repetitions for each transformation and each monkey, on different days). Note that in this paradigm, the monkeys learn the visuomotor rotation by repeated performance of a single movement (to the learned direction). To test whether the repetition could be responsible for the neuronal changes observed, we conducted control sessions. These sessions (termed “‘repetition”' sessions) consisted of a one-target task without any visuomotor transformation (namely, a standard task to one direction only). We performed 16 such sessions (ten with repeated movements to 90° and six with movements to 180°). Data analysis. Psychophysics studies have shown that immediately after learning, humans exhibit aftereffects, which is evidence for the formation of an internal representation of the newly acquired skill ( Lackner and DiZio 1994 ; Shadmehr and Mussa-Ivaldi 1994 ; Kawato 1999 ). This has been observed in monkeys as well ( Paz et al. 2003 ). To compare neuronal activities for movements with same kinematics, we excluded the first trials (three to five) in the post-learning epoch that exhibited significant aftereffects (measured as the directional deviation at peak velocity from a straight movement and compared to the distribution of deviations before learning). For the remaining trials, we compared velocity profiles, initial direction as a function of time, and actual trajectories to verify that there were similar to the trajectories in the pre-learning epoch (see Figure 1 B–1D). We also compared reaction times and perpendicular deviations at peak velocity and endpoint locations. No significant difference was found between the pre- and post-learning in all three groups ( t -test, p > 0.1). We also verified that learning was the same during the whole recording period. We divided the recording period into two to three consecutive segments and compared (1) learning rates in the learning epochs and (2) aftereffect magnitudes and washout rates in the post-learning epoch ( Paz et al. 2003 ). To further avoid changes in activity that result from any kinematic or dynamic differences, and since learning-related changes were only observed in activity taken from preparation for movement (before the go-signal), here we only report neuronal activity from this period, i.e., activity during the 600 ms following the target appearance but before the go-signal. We isolated 177 cells (113 from monkey W and 64 from monkey X) based on (1) the lack of significant change in activity during the first-hold period (during which no information was available about the upcoming trial) for the pre-learning epoch versus the post-learning epoch (by Mann-Whitney U -test); (2) the results of a one-way ANOVA showing a significant effect for direction; (3) more than five trials per direction both pre- and post-learning. We calculated spike counts in the 600-ms range following the target onset, referred to as the PA. Examining the neurons for changes in PD did not reveal any systematic or significant changes (bootstrap test, three of 177 showed a significant change) and PDs were uniformly distributed (Rayleigh test). MuI between the direction of the movement and each cell response was calculated by standard methods ( Cover and Thomas 1991 ) using the formula where d is the direction of movement and r is the number of spikes (see Figure 2 ). We used either the direct method for calculating P(r) or by assuming a Poisson distribution with the mean taken from all trials. We compensated for the limited number of trials (bias correction) by applying either analytical correction ( Panzeri and Treves 1996 ) or by shuffling trials between directions to obtain mean baseline and confidence intervals for the MuI; since both methods produced similar qualitative results, we report here the direct method, corrected analytically. For calculating the individual DI that each neuron conveys, we used the following formula ( Rolls et al. 1997 ; Buracas et al. 1998 ) (for alternative definitions, see DeWeese and Meister 1999 ): calculated separately for each direction d . To predict the direction of hand movement based on neuronal activity, we used two standard decoding methods: (1) MAP estimator. This was carried out by assuming Poisson distribution of rates and independency between neurons. We sampled (with repetition) 100 cells and then cross-validated by selecting randomly one trial from each cell and direction, calculating the cell's mean firing rate from the rest of the trials and used the following formula ( Sanger 1996 ) to obtain the most likely direction: where σ i ( d ) denotes the mean firing rate of cell i in direction d and r i is the rate in the randomly drawn trial. For ease of computation, we took the log of the probability and did not calculate N , the normalization factor. The process was repeated 100 times and performed separately for the pre-learning and post-learning. (2) PV analysis ( Georgopoulos et al. 1988 ; Schwartz 1993 ). One hundred twenty-nine cells (91 from monkey X and 38 from monkey W) were characterized as directionally tuned by fitting a cosine function ( r 2 > 0.5). The cells' PDs were homogenously distributed both pre- and post-learning (Rao test, pre-learning, p = 0.4, post-learning, p = 0.5). We performed two bootstrap tests for significance. First, we shuffled trials from pre- and post-learning and calculated the difference between the deviations of the PV prediction pre-learning to that of the post-learning. The process was repeated 1,000 times to obtain confidence intervals. Second, we shuffled cells from days in which different transformations were learned and again obtained confidence limits. This process tests whether the improvement in prediction was indeed related to the specific direction learned in the session.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC340948.xml
535815	Application of methods of identifying receptor binding models and analysis of parameters	Background Possible methods for distinguishing receptor binding models and analysing their parameters are considered. Results and Discussion The conjugate gradients method is shown to be optimal for solving problems of the kind considered. Convergence with experimental data is rapidly achieved with the appropriate model but not with alternative models. Conclusion Lack of convergence using the conjugate gradients method can be taken to indicate inconsistency between the receptor binding model and the experimental data. Thus, the conjugate gradients method can be used to distinguish among receptor binding models.	Background Most medicinal preparations and biologically active substances do not penetrate into cells and must therefore exert their influence on intracellular processes by interaction with specific protein molecules at the cell surface [ 1 - 3 ], for which the name "receptors" is in common use. Hormones and drugs that interact with receptors are known as "ligands". Data from research in molecular biology, and also results from indirect studies, have established the following schemes of ligand-receptor interaction [see [ 4 - 6 ] represented by the general models: Non-cooperative interaction between ligand and receptor : where R is the receptor molecule, L is the ligand molecule, RL is the ligand-receptor complex, and k +1 and k -1 are respectively the kinetic constants of formation and dissociation of the complex. Cooperative interaction between ligand and receptor Interaction of one ligand with N types of binding sites Let us note that the ligand-receptor interaction can also involve a combination of all three of these schemes. The most frequently used method for studying ligand-receptor interactions is the radioreceptor method [ 7 ], based on measuring the amount of radioactively labelled ligand bound in some defined manner to the appropriate receptor. Thus, experimentally, direct measurements of ligand-receptor complex concentration, [ RL ] are determined. The investigator has to solve two basic interrelated problems [ 6 ]: 1. discrimination among the ligand-receptor binding models (1–3 or modifications thereof); 2. determination of parameters that adequately relate the model to the experimental data. From a pharmacological point of view, the most important parameters are the following: [ R 0 ] (initial receptor concentration), and K d = k -1 / k +1 (dissociation constant) [ 7 ] The concentration of receptors and the dissociation constant can be changed. Modification of these parameter values can occur in many physiological and pathophysiological situations. For instance, the receptor concentration can reflect functional receptor modifications, and the dissociation constant can reflect genetic alterations of the receptor [ 6 ]. To solve the two interrelated problems a series of graphic methods can be deployed, of which the most frequently used is the Scatchard method [ 7 , 8 ]. However, the application of graphic methods in many cases is limited because of experimental errors and/or receptor binding complexity [ 9 , 10 ]. In particular, graphic methods are inapplicable for definition of the cooperative binding parameters and for analysis of non-equilibrium binding. Regression methods can be found for the measurement of ligand-receptor interaction constants [ 11 ]. As a matter of fact, these procedures computerize the graphic methods. Therefore, both regression methods and graphic methods are of limited applicability. The present paper argues that it is very difficult or impossible to discriminate reliably among receptor binding models or to analyse the parameters by traditional analytical methods. Materials and methods Let us write the law of mass action for each ligand-receptor interaction scheme as: For the scheme (1) But [ R ] = [ R 0 ] - [ RL ], [ L ] = [ L 0 ] - [ RL ]. So equation (4) can be rewritten: This differential equation relates to the class of Rikkatty equations. It can be solved analytically with the help of a special substitution [ 12 ], but in all other cases the substitutions [ R ] = [ R 0 ] - [ RL ], [ L ] = [ L 0 ] - [ RL ] do not generate analytically soluble equations. Therefore, all equations of this form were solved numerically using the Runge-Kutta method [ 13 , 14 ]. The differential equations are as follows: For scheme (2): For scheme (3): Numerical solution of equations (5–7) was carried out to determine [ RL ] u . Random error assuming the normal distribution law was superimposed on the magnitude of [ RL ] u , and was calculated at 5, 10, 20 or 100 points. The magnitude [ RT ] m was calculated using parameters other than [ RL ] u from models (1–3). These parameters were applied to the determination of [ RL ] u by the following functional minimization: Φ = ([ RL ] u - [ RL ] m ) 2 . (8) For functional minimization as per equation (8), Newton's method and its variants (the conjugate gradients method and coordinate descent method in various modifications) were used [ 15 - 17 ]. The iteration procedure stopped, when Φ/[ RL ] u was constant on the next iteration step. It is clear from the literature [ 6 ] that [ R 0 ] and K d cannot be <10 -15 M or >10 -5 M. Hence the iteration procedure could be improved by re-scaling these parameters logarithmically, making 10 -15 M equivalent to -1 on the new scale and 10 -5 M equivalent to 1. Results and discussion The functional (8) contour plots are shown in fig. 1 . From this figure, the degree of correlation between the parameters [ R 0 ] K d can be seen. Therefore the magnification of the random error in evaluating the magnitude of [ RL ] u displaces the functional (8) global maximum from its true values. In a sufficiently large neighbourhood of the global maximum, the functional magnitude (8) is practically invariant. However, this modification becomes more essential for evaluating the ratio of the functional (8) to basis vector of values [ RL ] u . Therefore this ratio was used with the inhibiting criterion choice. Figure 1 The functional (8) contour plot. The various methods of functional minimization are illustrated: a. The second derivative Newton method b. The conjugate gradients method c. The coordinate descent method The Newton method converges only in the close neighbourhood of the global maximum. However, modifications of the Newton method using second derivatives allow convergence to the global maximum after 1–2 iterations (fig. 1 , line 1). The conjugate gradients method converged after 2–3 iterations (fig. 1 , line 2). When magnification of the random error in the evaluation of [ RL ] u was taken into account, the convergence of the conjugate gradients method varied less than that of the Newton method. The coordinate descent method required an indeterminately large number of iterations before satisfactory convergence was reached. Use of the exhausting coordinate descent method accelerated the convergence procedure, but the number of iterative steps remained large (fig. 1 , line 3). It can be shown that 5 points suffice to identify the parameters of model (1) using the conjugate gradients method, whereas this method required >10 points for identifying the parameters in a more complicated model. The Newton methods required >7 and 12 points respectively, and the coordinate descent method required >10 and 18 points. Functional (8) behaviour was analysed with respect to the evaluation of [ RL ] m using an incorrect binding model. In particular (see fig. 2 ), the functional (8) contour plot for model (1) with the attempt to approximate the given model by scheme (2). It follows from the figure that a discordant receptor binding model results in functional (8) contour plot modification. Figure 2 The functional (8) contour plot with an inadequate choice of receptor-binding model. Thus, the modification of the functional (8) contour plot from the type in fig. 1 to the type in fig. 2 can be used as the criterion for choosing a receptor binding model. With the right choice, the contour plot is similar to that represented in fig. 1 . With the incorrect choice, the contour plot is similar to that shown in fig. 2 . It appears that when an incorrect choice of the receptor binding model has been made, the conjugate gradients method does not lead to convergence, whereas in some cases the Newton method converges to one of the local minima. Therefore, lack of convergence using the conjugate gradients method suggests an incorrect choice of receptor binding model. Conclusion Possible methods have been explored for discriminating among models for receptor binding model and for defining the relevant parameters. The procedure devised allows one to determine the receptor binding model and its parameters, even when the application of graphical methods is difficult or impossible . As seen here, lack of convergence in the conjugate gradients method indicates that an incorrect choice of model has been made. It is also shown that for the defining the parameters of the correct model, 5–10 data points are sufficient.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535815.xml
515284	Site-specific mutagenesis of Drosophila proliferating cell nuclear antigen enhances its effects on calf thymus DNA polymerase δ	Background We and others have shown four distinct and presumably related effects of mammalian proliferating cell nuclear antigen (PCNA) on DNA synthesis catalyzed by mammalian DNA polymerase δ(pol δ). In the presence of homologous PCNA, pol δ exhibits 1) increased absolute activity; 2) increased processivity of DNA synthesis; 3) stable binding of synthetic oligonucleotide template-primers (t 1/2 of the pol δ•PCNA•template-primer complex ≥2.5 h); and 4) enhanced synthesis of DNA opposite and beyond template base lesions. This last effect is potentially mutagenic in vivo . Biochemical studies performed in parallel with in vivo genetic analyses, would represent an extremely powerful approach to investigate further, both DNA replication and repair in eukaryotes. Results Drosophila PCNA, although highly similar in structure to mammalian PCNA (e.g., it is >70% identical to human PCNA in amino acid sequence), can only substitute poorly for either calf thymus or human PCNA (~10% as well) in affecting calf thymus pol δ. However, by mutating one or only a few amino acids in the region of Drosophila PCNA thought to interact with pol δ, all four effects can be enhanced dramatically. Conclusions Our results therefore suggest that all four above effects depend at least in part on the PCNA-pol δ interaction. Moreover unlike mammals, Drosophila offers the potential for immediate in vivo genetic analyses. Although it has proven difficult to obtain sufficient amounts of homologous pol δ for parallel in vitro biochemical studies, by altering Drosophila PCNA using site-directed mutagenesis as suggested by our results, in vitro biochemical studies may now be performed using human and/or calf thymus pol δ preparations.	Background Many Drosophila melanogaster homologs of the proteins required for both DNA replication and repair have been identified and in several cases purified to apparent homogeneity. These include DNA polymerase α holoenzyme [ 1 , 2 ], DNA polymerase δ(pol δ) [ 2 - 4 ], replication protein A (RP-A; [ 5 ]), replication factor C (RF-C; e.g., see [ 6 - 9 ]) and various origin recognition complex (ORC) subunits (see e.g., [ 10 , 11 ]). Moreover, complete replication of DNA containing the SV40 origin of replication has been reconstituted in vitro using purified SV40 T-antigen and Drosophila cell-free extracts [ 7 ]. A protein about which much information has been obtained is proliferating cell nuclear antigen (PCNA). Drosophila PCNA was first identified both as a highly purified protein able to substitute, albeit poorly, for human PCNA in a cell-free SV40 DNA replication system reconstituted from purified proteins [ 12 ] and by Yamaguchi et al. [ 13 ] who used an oligonucleotide probe to detect the Drosophila PCNA cDNA and gene, express the protein in E. coli and deduce its complete amino acid sequence. Further results indicated that in flies, PCNA was encoded by a single gene located at position 56F5-15 on the right arm of chromosome 2. This was subsequently identified as the Drosophila mus 209 locus [ 14 ]. Recently, a second Drosophila PCNA gene of limited homology to the original and of unknown biological function has also been found [ 15 ]. Protocols have been established for purification of wild-type human PCNA from tissue culture cells [ 16 , 17 ], unmodified wild-type human PCNA after regulated expression in E. coli [ 18 ] and NH 2 -terminally his-tagged but otherwise wild-type human PCNA, also engineered for bacterial expression [ 19 ]. All were comparably effective at stimulating mammalian pol δ. Similar protocols have been developed for Drosophila PCNA and strategies for site-directed mutagenesis have been devised and implemented [ 20 ]. Recently, Zhang et al. [ 21 ] (see also [ 22 ]) as well as others (e.g., see [ 23 ]) identified the interdomain connector loop of PCNA (amino acids 119-133 of human PCNA) as crucial for binding pol δ. Of note, relative to wild-type PCNA, mutations of the molecule within this region such as glutamine at position 125 changed to glutamic acid (Q125E) promoted increased pol δ-processivity [ 21 ]. In human PCNA, residues 123, 126, 127 and 128 were defined as being essential for interaction with pol δ [ 21 ]. Comparison of human with Drosophila PCNA sequences in this region indicated that of these four amino acids, three (residues 126, 127 and 128) are identical. The fourth, residue 123, is glutamine (Q123) in wild-type Drosophila PCNA. The corresponding residue in human PCNA is valine (V). To investigate the role of the interdomain connector loop of PCNA on the effects of PCNA on pol δ, we mutagenized residues within this region of Drosophila PCNA so that they more nearly resembled human amino acids. After bacterial expression and purification, we tested the effects of these site-specifically modified ("humanized") Drosophila PCNA molecules on purified calf thymus pol δ (two-subunit form; see [ 17 , 24 ]). Calf thymus and human pol δ are highly similar in amino acid sequence [ 25 - 27 ] and can, for our purposes, be used interchangeably. "Humanization" of a single Drosophila PCNA residue, conversion of Q123 to V (Q123V), conferred upon it, enhanced ability to affect several properties of calf thymus pol δ. More extensive mutagenesis, in which the entire interdomain connector loop of Drosophila PCNA (amino acids 119-133) was replaced by the corresponding human residues, was still more effective at stimulation of calf thymus pol δ, than either wild-type or Q123V Drosophila PCNA. However, it was considerably less effective than wild-type human PCNA at altering the properties of calf thymus pol δ. These results therefore suggest that in addition to the interdomain connnector loop, other regions of PCNA are also important effectors of pol δ activity. They also provide a means to couple operationally, the considerable power of in vivo genetic analyses performed in Drosophila with the sophistication of mammalian biochemistry. Results To study the role of the interdomain connector loop of PCNA (amino acids 119-133), we compared human and Drosophila homologs. Of the 15 interdomain connector loop residues, nine are identical between the two; identical residues are shaded (Fig. 1A ). Overall, Drosophila PCNA is >70% identical to that from mammals (e.g., humans; see [ 13 ]). Others showed that PCNA residues 123, 126, 127 and 128 were essential for interaction with pol δ [ 28 ]. Of these four, only one (residue 123) differs between flies and humans. Also shown is a model constructed from the X-ray crystallographically determined structure of PCNA indicating the locations of the sites to be mutated in Drosophila PCNA (Fig. 1B ). Shown (Fig. 1B ) is the X-ray crystal structure of human PCNA. The Drosophila homolog is assumed to be similar. Figure 1 Structure and structural rationale for mutating Drosophila PCNA. A: amino acid sequences of the interdomain loops of Drosophila (designated D.m.) and human (designated H.s.) PCNA. Gray boxes indicate amino acids identical for both organisms; arrows show amino acids thought essential for interaction of human PCNA with human pol δ. Amino acid 123 is the only one which is both essential and different in Drosophila versus human PCNA. B: the "front" side of the human PCNA trimer. Amino acids 119-133 of the interdomain loops are highlighted by showing their α-carbon atoms as black spheres. The α-carbon atom of Val123 is shown as a larger gray sphere. Purification of wild-type and site-specifically mutated PCNA Four NH 2 -terminally his-tagged PCNA variants were highly purified; purity for each is shown (Fig. 2 ). First constructs were prepared encoding 1) NH 2 -terminally his-tagged wild-type human PCNA; 2) NH 2 -terminally his-tagged wild-type Drosophila PCNA (dPCNA) and two dPCNA derivatives; 3) one in which amino acid 123 was mutated from glutamine to valine (Q123V dPCNA); and 4) the other, in which Drosophila amino acids 119-133 were replaced by the corresponding human sequence (dr119-133h dPCNA). Then all four were transformed separately into E. coli (strain M15 [pREP4]) and respective proteins were expressed. Finally bacteria were lysed and his-tagged proteins were purified using various procedures including Ni 2+ -IDA Sepharose chromatography. The purity of each was determined by SDS-PAGE and is shown as indicated (Fig. 2 ). The identity of wild-type human PCNA was confirmed using mouse monoclonal anti-mammalian PCNA antibody PC10; the identity of wild-type Drosophila PCNA was confirmed using affinity purified polyclonal anti- Drosophila PCNA antibodies prepared in rabbits [ 12 ] (not shown). Figure 2 SDS-PAGE analysis of his-tagged PCNA purified from E. coli extracts after regulated bacterial expression. Purification and SDS-PAGE were as described (Experimental Procedures). Lane 1, 0.4 μg wild-type human PCNA was subjected to electrophoresis. Lane 2, 0.8 μg wild-type Drosophila PCNA was subjected to electrophoresis. Lane 3, 0.8 μg Drosophila PCNA containing valine substituted for glutamine at position 123 was subjected to electrophoresis. Lane 4, 0.45 μg Drosophila PCNA containing amino acids 119-133 substituted with the corresponding human PCNA amino acids was subjected to electrophoresis. Migration positions of molecular mass standards are indicated to the right of the figure. Stimulation of calf thymus pol δ activity by highly purified wild-type versus selected mutant PCNA fractions Calf thymus pol δ (apparently homogeneous two-subunit form; see [ 24 ]) was purified and assayed for polymerase activity in the presence of varying concentrations of both highly purified wild-type and specific mutant PCNA molecules. We showed previously that either calf thymus or human PCNA could be used interchangeably as stimulatory co-factors for calf thymus pol δ [ 29 ] (see also [ 12 , 18 , 19 ]). Assays were performed using poly(dA)-oligo(dT) as described (Experimental Procedures). As can be seen, human PCNA resulted in robust stimulation of calf thymus pol δ; much less stimulation was observed for wild-type Drosophila PCNA (Fig. 3 ). Mutation of Drosophila PCNA resulted in substantially increased stimulation of calf thymus pol δ; both substitution of a single amino acid (Q123V dPCNA) and replacement of the entire fly interdomain connector loop with corresponding human amino acids (dr119-133h dPCNA) had demonstrable effects. Of note, at relatively high concentrations, Drosophila PCNA but with the entire fly interdomain connector loop replaced by corresponding human amino acids (dr119-133h dPCNA) was similarly effective to wild-type human PCNA at stimulating the activity of calf thymus pol δ; however, it was considerably less effective at lower concentrations (Fig. 3 ). This suggests an effect on binding of PCNA to pol δ and/or on mutant PCNA multimerization. Figure 3 Effect of various purified PCNA fractions on the DNA polymerase activity of calf thymus pol δ. Calf thymus pol δ was incubated in a reaction mixture as described (see Materials and Methods) for 5 min at room temperature. Each incubation contained 10 ng of pol δ. DNA product synthesized was determined after placing 5-μl aliquots on Whatman DE-81 filters and subsequently washing with a 5% (w/v) solution of Na 2 HPO 4 •12H 2 O. Radioactivity retained on filters was then determined by liquid scintillation counter. Reaction mixtures contained increasing amounts, as indicated on the abscissa, of various PCNA samples, also as indicated. The effects of highly purified wild-type versus selected mutant PCNA fractions on the processivity of incorporation by calf thymus pol δ To examine further, the stimulation of calf thymus pol δ by both wild-type and specific mutant PCNA molecules, we examined effects on processivity of nucleotide incorporation. Processivity is defined as the number of deoxyribonucleotides incorporated each time a DNA polymerase binds its template-primer. As can be seen, without PCNA (Fig. 4 lane 1), pol δ is essentially a distributive enzyme incorporating only a few nucleotides as a result of each binding event. With increasing concentrations of wild-type human PCNA (concentrations increasing from right to left as indicated), processivity of incorporation increases dramatically (Fig. 4 lanes 2–4). This correlates quite closely with the PCNA-mediated activity increase (see Fig. 3 ). Wild-type Drosophila PCNA had relatively much less effect on the processivity of calf thymus pol δ (Fig. 4 lanes 5–7; concentrations again increasing from right to left as indicated). This is also consistent with activity data presented herein (Fig. 3 ) as well as with results reported previously [ 12 ]. When mutants of Drosophila PCNA were tested, both Q123V dPCNA (Fig. 4 lanes lanes 8–10; concentrations again increasing from right to left as indicated) and dr119-133h dPCNA (Fig. 4 lanes lanes 11–13; concentrations again increasing from right to left as indicated), promoted increased pol δ processivities, again consistent with increased activities (Fig. 3 ). Increases were concentration-dependent, also as expected. Figure 4 Effect of various purified PCNA fractions on the processivity of nucleotide incorporation by calf thymus pol δ. Incorporation of [α 32 P]dTMP by calf thymus pol δ was monitored by standard denaturing PAGE. The substrates used were (dA) ~500 -(dT) 12–18 as template-primer and [α- 32 P]dTTP. Concentrations of PCNA, both wild-type and mutant proteins, are as indicated. h, human; dr, Drosophila melanogaster. NH 2 -terminally his-tagged-PCNA fractions are as indicated; wt, wild-type; Q123V, recombinant Drosophila PCNA containing a single amino acid, glutamine at position 123, changed to valine; dr119-133h, recombinant Drosophila PCNA containing the entire interdomain connector loop (amino acids 119-133) replaced with the corresponding human PCNA amino acids. Stable complex formation among pol δ, 32 P-labeled oligonucleotide template-primer and highly purified wild-type versus selected mutant PCNA fractions PAGE band mobility shift assays were used to evaluate, in an essentially qualitative manner, the stability of complex formation among calf thymus pol δ, labeled template-primer and highly purified wild-type versus selected mutant PCNA molecules. As can be seen, wild-type Drosophila PCNA promoted almost no pol δ•PCNA•template-primer complex formation (Fig. 5 ). In contrast, complex-formation with both Drosophila PCNA mutants (Q123V dPCNA and dr119-133h dPCNA) was readily detectable but neither gave results as robust as those seen with wild-type human PCNA (Fig. 5 ). Figure 5 Effect of various purified PCNA fractions on calf thymus pol δ•PCNA• 32 P-labeled oligonucleotide template-primer complex formation. Complex formation among pol δ, various purified PCNA fractions and 32 P-labeled synthetic oligonucleotide template-primers (30-21-mers) was monitored by standard non-denaturing PAGE-band-mobility-shift assays [32]. Each incubation contained 10 ng of pol δ, 70 ng of PCNA and 0.1 pmol/reaction (useable 3'-OH) of annealed template-primer. NH 2 -terminally his-tagged-PCNA fractions are as indicated; wt, wild-type; Q123V, recombinant Drosophila PCNA containing a single amino acid, glutamine at position 123, changed to valine; dr119-133h, recombinant Drosophila PCNA containing the entire interdomain connector loop (amino acids 119-133) replaced with the corresponding human PCNA amino acids. DNA synthesis beyond chemically defined template base lesions promoted by highly purified wild-type versus selected mutant PCNA fractions As a final test, we examined the abilities of various PCNA fractions to promote pol δ-dependent DNA synthesis beyond template base lesions (TLS). PCNA-dependent TLS by pol δ was first reported by O'Day et al. [ 30 ] and subsequently analyzed in detail biochemically [ 29 ]. The structure of the synthetic oligonucleotide used for evaluation is shown in Fig. 6A . For the data shown (Fig. 6B ), X represents the model abasic site (hereafter termed the abasic site [ 31 ]) used previously for many of our studies (e.g., see [ 29 ]). The mobility of the labeled 21-mer primer, PAGE-purified but without any subsequent enzymatic incubation is shown (Fig. 6B lane 1). When calf thymus pol δ alone was added, primer extension opposite the template abasic site was detected but there was no discernible elongation of the resulting 22-mer primer and no full-length product (30-mer) was observed; some degradation of the 21-mer primer, presumably resulting from the activity of the intrinsic pol δ 3'-5' exonuclease, was seen (Fig. 6B lane 2). Addition to incubations of wild-type Drosophila PCNA resulted in slight but readily detectable DNA synthesis beyond the template abasic site; this included some full-length 30-mer (Fig. 6B lane 3). Relatively more full-length 30-mer was seen when Q123V mutant Drosophila PCNA was included in addition to calf thymus pol δ (Fig. 6B lane 4) and still more full-length 30-mer was seen when dr119-133h Drosophila PCNA was added (Fig. 6B lane 5). Clearly, the greatest amount of full-length 30-mer product was seen when wild-type human PCNA was incubated with calf thymus pol δ (Fig. 6B lane 6). Of note, wild-type human PCNA also promotes the tightest complex formation between calf thymus pol δ and 32 P-labeled template-primer DNA (see Fig. 5 ). Figure 6 Effect of various purified PCNA fractions to promote nucleotide incorporation by calf thymus pol δ beyond chemically defined template base lesions. A: the structure of the 5'- 32 P-labeled 30-21-mer template-primer; only the primer (21-mer) was radiolabeled and X indicates the position of a modified tetrahydrofuran moiety (model abasic site) on the 30-mer template. B: lane 1, gel-purified primer alone was subjected to electrophoresis; lanes 2–6, incubations were formulated as indicated with the template-primer shown in A followed by standard denaturing PAGE. h, human; dr, Drosophila melanogaster. For lanes 2–6, each incubation contained 0.5 pmol of labeled primer (3'-OH ends) annealed to 0.5 pmol of template (3'-OH ends), 10 ng pol δ and 70 ng PCNA as indicated. NH 2 -terminally his-tagged-PCNA fractions are as indicated; wt, wild-type; Q123V, recombinant Drosophila PCNA containing a single amino acid, glutamine at position 123, changed to valine; dr119-133h, recombinant Drosophila PCNA containing the entire interdomain connector loop (amino acids 119-133) replaced with the corresponding human PCNA amino acids. Discussion Although human PCNA and Drosophila PCNA are more than 70% identical at the level of primary amino acid sequence, wild-type Drosophila PCNA is only a very poor substitute for human PCNA in cell-free reactions with calf thymus pol δ. This is documented both in this report and previously [ 12 , 32 ]. However, mutating only a single Drosophila PCNA amino acid, glutamine at position 123 (Q123) to valine (V), leads to a dramatic enhancement in the abilities of Drosophila PCNA to stimulate calf thymus pol δ. Effects were shown on total activity (Fig. 3 ), processivity (Fig. 4 ), pol δ•PCNA•template-primer complex formation (Fig. 5 ) and extended DNA synthesis beyond a template abasic site (Fig. 6 ). Replacing the entire interdomain connector loop of Drosophila PCNA (amino acids 119-133) with the corresponding residues from human PCNA resulted in additional enhancement (Figs. 3 , 4 , 5 , 6 ), but in neither case were the mutants of Drosophila PCNA (Q123V dPCNA or dr119-133h dPCNA) equivalent to wild-type human PCNA in the stimulation of calf thymus pol δ. Our data indicate that although a single Drosophila PCNA amino acid at position 123 (in addition to conserved residues 126–128) is very important for pol δ-stimulation, the further enhancement of stimulation seen when the entire interdomain connector loop of Drosophila PCNA (amino acids 119-133) was replaced with the corresponding residues from human PCNA suggests that other residues in this loop are also involved directly in binding pol δ. Alternatively, it is possible that loop residues other than 123 and 126–128 play a secondary or indirect (e.g., conformational) role in positioning crucial amino acids so as to optimize their direct binding to pol δ. In this context, we would like to call attention to the fact that at relatively low concentrations, dr119-133h dPCNA is considerably less effective than wild-type human PCNA in stimulating the activity of calf thymus pol δ; at higher concentrations, dr119-133h dPCNA and wild-type human PCNA stimulate calf thymus pol δ similarly. This implies complex protein-protein interactions between PCNA and pol δ such that biochemical properties recorded in dilute solutions in vitro may not accurately predict properties manifest at much different and generally much higher intranuclear concentrations present in vivo . Alternatively, PCNA must be present as a trimer (three-subunit ring) in order to function. Since the equilibrium among monomer, dimer and trimer was shown to depend on PCNA protein concentration [ 33 ], it is certainly possible that the difference observed between dr119-133h dPCNA and wild-type human PCNA actually reflects differences in the K eq for PCNA multimerization. These two possibilities, concerning both complicated pol δ•PCNA interactions and PCNA multimerization, are not mutually exclusive. Similarly, the fact that replacement of the entire interdomain connector loop of Drosophila PCNA (amino acids 119-133) with the corresponding residues from human PCNA did not result in a molecule as effective in stimulating calf thymus pol δ as human PCNA suggests that regions other than the interdomain connector loop are important for pol δ-stimulation. Our data do not address the question of whether these putative "other regions" affect pol δ directly (e.g., like the interdomain loop) or indirectly (e.g., through conformational effects on other regions of the molecule that do bind pol δ directly). Additional mutagenesis studies may shed light on this issue. For example, based on experiments of others, it seems likely that the extreme C-terminus of PCNA also interacts directly with pol δ (see [ 23 , 34 - 36 ]). Hence it may be of interest to perform similar mutagenesis experiments to those reported here, focusing instead on the C-terminal region of Drosophila PCNA, rather than the interdomain connector loop. We think it should also be noted that both Oku et al. [ 35 ] and Ola et al. [ 36 ] prepared hybrid proteins between human and S. cerevisiae PCNA. As in our studies, Ola et al. [ 36 ] found that regions other than the interdomain connector loop of PCNA were important for interaction with pol δ. These authors suggested that additional interacting regions were likely to exist both in the PCNA C-terminus and N-terminus. It may also be of interest to prepare double-mutants, first in the interdomain connector loop of Drosophila PCNA, thereby allowing efficient in vitro function with purified calf thymus pol δ, and then elsewhere in the PCNA molecule corresponding to interesting sites defined phenotypically by in vivo genetic studies of others. For example, it might be possible to determine if particular mus 209 mutations leading to enhanced mutagen sensitivity among affected organisms (see [ 37 ] and references therein) alter any functional interactions between PCNA and pol δ in vitro . Results of such studies could lead to novel biochemical insights regarding the mechanism(s) by which point mutations in the Drosophila PCNA gene lead to enhanced mutagen sensitivity among animals bearing these mutations. The strategy taken here will presumably allow study of interactions between PCNA and other proteins with which it interacts. In this context, we think it important to note that partial effects on pol δ-stimulation have been recorded. This suggests that our methodology will also allow detection of partial rather than complete effects on the binding of other proteins. Interactions between PCNA and many of the molecules with which it interacts have recently been mapped [ 23 ] and for example, one might immediately compare interactions between several mammalian proteins (e.g., human RF-C, DNA ligase I, FEN I and/or p21) and both various wild-type and mutant PCNA molecules described in this paper. Functional (e.g., effects on pol δ activity) as well as direct binding measurements may be made. As with PCNA•pol δ interactions, it may ultimately be feasible to correlate interesting PCNA molecules defined phenotypically using genetic analyses performed in living animals and biochemical studies of specific PCNA•protein binding. For example, do mutagen sensitive mus 209 animals bear mutations in a region of PCNA responsible for MSH binding? Both MSH3 and MSH6 were reported to possess a consensus motif for binding to the interdomain connector loop of PCNA [ 38 ]. Finally, we think it important to note that pol δ has most recently been reported to contain at least four subunits (see e.g., [ 39 , 40 ]) yet all experiments performed here were with the two-subunit form of the enzyme purified from calf thymus. We and others have shown that the larger subunit, p125, is catalytic while the smaller, p50, does not seem to contact the DNA closely (see e.g, [ 41 ]), but instead, is required for processivity-stimulation by PCNA (e.g., see [ 42 ]) to which it apparently binds. It is also clear that PCNA binds to what has been termed, the third pol δ subunit, p68 or p66 in mammalian systems [ 39 , 43 , 44 ], Cdc27p in S. pombe [ 40 ] and Pol32p in S. cerevisiae [ 45 , 46 ]. Clearly the physiologically important interaction between PCNA (either mutant or wild-type) and this third pol δ subunit was omitted from our analyses, but could markedly affect any or all of the responses of polymerase to PCNA that we reported here. Conclusions Through our experiments, we showed that Drosophila PCNA could be "humanized" and that "humanization" (mutation of key Drosophila residues to human ones) increased effects on mammalian pol δ. The highly purified two-subunit form of pol δ was used for all of our studies. It is possible, though we think it unlikely, that different conclusions would be reached if a different form of pol δ (three-or four-subunit) was used. Nevertheless two of the effects we observed could be considered beneficial. They were enhancement of polymerase activity and processivity. A third effect seems likely to be detrimental, at least over the long term, that is increased synthesis opposite and beyond a chemically defined template base lesion (TLS). Our data suggest that all three of these effects result from enhancement of PCNA-dependent stability of the pol δ•PCNA•template-primer complex. In other words, in the range that we have studied, the more tightly pol δ binds to DNA, the greater its activity, the greater its processivity, but also the more likely it is to catalyze TLS. Our results provide an explicit approach to correlate in vivo genetic studies with rigorous in vitro biochemistry. Methods Materials Unlabeled deoxyribonucleoside triphosphates (dNTPs) were from Boehringer-Mannheim; [α- 32 P]ATP and [α- 32 P]dTTP were from Amersham Corp. E. coli DNA polymerase I Klenow fragment without 3'-5' exonuclease activity (exo-), was expressed and purified according to standard protocols [ 47 ]. Terminal deoxynucleotidyl transferase (TdT) was from Sigma. Micrococcal nuclease was from Boehringer-Mannheim. Pfu DNA polymerase was from Stratagene. Ni 2+ -IDA Sepharose was from Pharmacia (Piscataway, NJ). Acrylamide and methylene bis-acrylamide were from Eastman Organic Chemicals and for protein SDS-PAGE, were further purified by adsorption of impurities to activated charcoal. For PAGE of nucleic acids, they were purified by adsorption to an ion exchange resin. All other materials were of reagent grade and were used without additional purification. Proteins PCNA was purified to apparent homogeneity from calf thymus [ 17 ] as was pol δ [ 24 , 48 ]. Human PCNA cDNA was cloned into a bacterial expression vector and human PCNA was purified from an E. coli extract, also to apparent homogeneity [ 18 ]. D. melanogaster PCNA was purified to apparent homogeneity identically after bacterial expression [ 13 ]. A his-tag was added to the NH 2 -termini of both human and Drosophila PCNA by cDNA insertion into pQE30 (Qiagen, Valencia, CA) using Bam H1 and Hind III restriction endonuclease sites. Nucleic acids Templates and primers, all of defined sequence, were synthesized conventionally by Dr. F. Johnson and colleagues (Stony Brook). Before use, they were purified by standard denaturing PAGE [ 49 ]. All other DNA manipulations were performed according to standard techniques [ 49 ]. Methods Much of the methodology was described in detail previously [ 12 , 19 , 20 , 24 , 29 , 32 , 41 , 50 , 51 ]. SDS-PAGE was according to Laemmli [ 52 ] as modified [ 53 ] on minigels or as reported previously [ 54 ]. For immunoblots, proteins were transferred electrophoretically to nitrocellulose [ 55 ] and resulting replicas were probed with antibodies. Reactivity was visualized colorimetrically [ 56 ] with alkaline phosphatase-conjugated goat anti-IgG antibodies [ 57 , 58 ] and a one-solution phosphatase substrate (Kirkegaard and Perry, Gaithersburg, MD). Immunologic detection of human PCNA was with mouse monoclonal antibody (mAb) PC10 (Oncogene Sciences, Uniondale, NY). Detection of Drosophila PCNA was with affinity purified polyclonal rabbit anti- Drosophila PCNA antibodies [ 12 ]. Restriction endonucleases were from Boehringer (Indianapolis, IN) and were used according to the vendor's instructions. DNA sequencing performed in both directions was according to Sanger et al. [ 59 ] using a fluorescence-based method and an ABI 373 (Applied Biosystems, Foster City, CA) automated DNA sequencer. Site-directed mutagenesis of Drosophila PCNA Site-directed mutagenesis of NH 2 -terminally his-tagged Drosophila PCNA was performed exactly as described [ 20 ] to generate either the Q123V protein or chimeric molecules containing the entire Drosophila PCNA sequence except for amino acids 119-133 which were replaced by the corresponding residues from human PCNA. Purification of his-tagged PCNA Purification of his-tagged PCNA to apparent homogeneity was performed exactly as previously described [ 20 ]. Characterization was by SDS-PAGE (Fig. 2 ) and immunoblot analysis. DNA polymerase δ incubations Assays of pol δ on synthetic oligonucleotide template-primers were performed essentially as previously described [ 24 ]. Primers were 5' end-labeled with T4 polynucleotide kinase in the presence of [γ- 32 P]ATP. Afterward, labeled primer was annealed to an unlabeled template. The standard reaction mixture for pol δ contained 40 mM Bis-Tris, pH 6.7, 6 mM MgCl 2 , 1 mM dithiothreitol, 10% glycerol and 40 μg/ml bovine serum albumin. Additional details are provided in the figure legends. Incubations were terminated by addition of standard stop solution and aliquots were subjected to 12% PAGE in the presence of 7 M urea and 15% formamide. After electrophoresis, gels were subjected to autoradiography and/or Molecular Dynamics 445 SI PhosphorImager analyses. Pol δ processivity Processivity was evaluated qualitatively using (dA) ~500 annealed to (dT) 12–18 (both from Pharmacia) in a final volume of 5 μl containing 6 nmol poly(dA) (nucleotide), 0.2 nmol (dT) 12–18 (nucleotide), 10 μM dTTP, 100 μCi [α- 32 P]dTTP, 40 mM Bis-Tris, pH 6.7, 6 mM MgCl, 1 mM dithiothreitol, 10% glycerol, 40 μg/ml bovine serum albumin, 10 ng of highly purified pol δ and various quantities of different PCNA samples as indicated. Assays were for 5 min at room temperature and were stopped by addition of standard PAGE stop solution and PAGE in the presence of 7 M urea. After electrophoresis, gels were subjected to autoradiography and/or Molecular Dynamics 445 SI PhosphorImager analyses. Nondenaturing PAGE band mobility shift assays Nondenaturing PAGE band mobility shift assays were performed essentially as previously described [ 32 ] but without MgCl 2 and otherwise as detailed in the figure legend. EDTA was included in each incubation and in the gel electrophoresis buffer at a final concentration of 3 mM. Authors' contributions DJuM performed all enzymologic and mobility shift assays with DNA polymerase δ in combination with both wild-type and various mutant PCNA molecules. He also designed, engineered and characterized all recombinant PCNA molecules. DJuM expressed several recombinant proteins in bacteria and purified them. Finally, he participated in DNA polymerase purification and drafted the original manuscript. MM expressed some recombinant proteins in bacteria and purified them. She also purified and characterized most DNA polymerase substrates. HM participated in DNA polymerase purification and manuscript preparation. PAF advised DJuM on execution and interpretation of experiments and assisted both in figure design and all other aspects of manuscript preparation. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC515284.xml
314470	A New Gene That Shapes Mouse Pigmentation Patterning	null	Scientists have long known that variation in animal color patterns carry far more than cosmetic significance. Darwin first connected pigmentation with adaptive advantage, noting that male finches with bright red plumage enjoyed greater reproductive success than their drab competitors. Explaining why coloration confers such advantages, however, has proved somewhat easier than showing how it arises. Biologists studying how neighboring regions of the vertebrate body plan develop differences in appearance and form have identified a small number of signaling pathways common to all animals. How and whether these pathways also control the developmental expression and variation of surface attributes like hair color, hair density, and hair length are unclear. By studying an old mouse mutant called droopy ear , Gregory Barsh and colleagues show that a member of the well-known family of T-box genes is required for a key pigmentation pattern in mice. Many vertebrate species—be they fish, bird, or mammal—have a much lighter belly than back. Studies in mice indicate these dorsoventral pigment differences arise from differential expression of the Agouti gene in the ventral and dorsal regions of the developing mouse; Agouti produces a pale yellow color and thus mice with light bellies have Agouti expressed in their ventral but not dorsal region. Droopy ear was discovered more than 50 years ago by virtue of its effects on head and ear shape, but it also affects pigmentation patterns; mutant mice have expanded ventral-specific Agouti expression into the dorsal region. First, Sophie Candille, a graduate student in Barsh's laboratory, searched for the gene that underlies the defect in droopy ear . When the researchers homed in on the chromosomal region known to harbor droopy ear , they found Tbx15 —a member of the T-box gene family. T-box genes are found in a wide range of species and play diverse roles during embryonic development. In the droopy ear mouse, Tbx15 carries a mutation that makes the protein nonfunctional. The researchers made certain that Tbx15 really is the droopy ear gene by deleting most of the gene's coding region and showing that this “knocked-out” gene produces the typical droopy ear mouse. The pattern of embryonic Tbx15 expression—determined by observing messenger RNA transcripts in developing tissues of the head, trunk, and limbs—suggests that early expression of Tbx15 in the dorsal flank sets coordinates for dorsoventral differences in hair length and pigmentation. Candille et al. demonstrate that the regional pigment differences characteristic of adults is indeed established soon after embryonic Tbx15 expression. So this boundary in pigmentation is set up very early during development. Interestingly, the early coordinates of the future pigment boundary do not correspond to any other known developmental boundary. The Tbx15 pigmentation effects seen in these lab mice, the researchers note, resembles coat variations in other mammals, including German shepherds and an endangered mouse whose lighter dorsal markings once gave it an adaptive advantage on the white sand reefs where it lives (sadly, such markings offer no protection against loss of habitat). T-box genes are also found in humans; mutations in Tbx1 , Tbx4 , Tbx5 , and Tbx22 can cause developmental abnormalities of the heart, limbs, or of the head and neck. Mutations of human Tbx15 have not yet been identified, but could contribute to regional differences of pigmentation (in dorsal and ventral surfaces of the limbs, for example) or to development of the head and neck. The identification of Tbx15 adds a new player to the genes that help pattern the developing embryo—attention now turns to the controls that regulate Tbx15 and the Tbx15 targets, which set up the pattern. Dorsoventral pigment boundaries in mouse and human	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC314470.xml
514603	Short-term cytotoxic effects and long-term instability of RNAi delivered using lentiviral vectors	Background RNA interference (RNAi) can potently reduce target gene expression in mammalian cells and is in wide use for loss-of-function studies. Several recent reports have demonstrated that short double-stranded RNAs (dsRNAs), used to mediate RNAi, can also induce an interferon-based response resulting in changes in the expression of many interferon-responsive genes. Off-target gene silencing has also been described, bringing into question the validity of certain RNAi-based approaches for studying gene function. We have targeted the plasminogen activator inhibitor-2 (PAI-2 or SERPINB2) mRNA using lentiviral vectors for delivery of U6 promoter-driven PAI-2-targeted short hairpin RNA (shRNA) expression. PAI-2 is reported to have anti-apoptotic activity, thus reduction of endogenous expression may be expected to make cells more sensitive to programmed cell death. Results As expected, we encountered a cytotoxic phenotype when targeting the PAI-2 mRNA with vector-derived shRNA. However, this predicted phenotype was a potent non-specific effect of shRNA expression, as functional overexpression of the target protein failed to rescue the phenotype. By decreasing the shRNA length or modifying its sequence we maintained PAI-2 silencing and reduced, but did not eliminate, cytotoxicity. ShRNA of 21 complementary nucleotides (21 mers) or more increased expression of the oligoadenylate synthase-1 (OAS1) interferon-responsive gene. 19 mer shRNA had no effect on OAS1 expression but long-term selective pressure on cell growth was observed. By lowering lentiviral vector titre we were able to reduce both expression of shRNA and induction of OAS1, without a major impact on the efficacy of gene silencing. Conclusions Our data demonstrate a rapid cytotoxic effect of shRNAs expressed in human tumor cell lines. There appears to be a cut-off of 21 complementary nucleotides below which there is no interferon response while target gene silencing is maintained. Cytotoxicity or OAS1 induction could be reduced by changing shRNA sequence or vector titre, but stable gene silencing could not be maintained in extended cell culture despite persistent marker gene expression from the RNAi-inducing transgene cassette. These results underscore the necessity of careful controls for immediate and long-term RNAi use in mammalian cell systems.	Background Gene silencing is a powerful tool with which to study protein function. Gene inactivations in mice have revolutionised the way we study both basic biology and a plethora of disease types [ 1 , 2 ]. Gene silencing in human cells has, until recently, proven difficult to achieve [ 3 ]. Research with plants, flies and worms recently uncovered a mechanism by which eukaryotic cells target mRNAs, and perhaps even genetic loci, for specific gene silencing. This process is termed RNA interference (RNAi). RNAi can also be induced in mammalian cells using double-stranded RNAs (dsRNAs), and has become the method of choice for targeted knock-down of gene expression in mammalian cells [ 4 ]. The apparent specificity of RNAi also enables allele-specific gene targeting [ 5 ]. Initial studies using RNAi in mammalian cells centred around transient knock-down of target gene expression, either using direct transfection of synthetic short interfering RNA (siRNA) [ 6 ], transfection of in vitro transcribed siRNA [ 7 ] or transient expression of short dsRNA via transfection of plasmid DNA bearing RNA Polymerase III promoter-driven expression cassettes [ 8 , 9 ]. Short dsRNAs of 19 to 29 base-paired nucleotides, complementary to the target mRNA, were expressed as 2 complementary RNAs or as a hairpin structure (shRNA), and resulted in knock-down of the target message. While these initial RNAi methods gave a rapid phenotypic read-out in vitro, stable knock-down of gene expression is required for monitoring long-term effects on cell function, for example, in developing tumors in vivo or in cell-based gene therapy approaches. Efficient delivery of RNAi-inducing dsRNA or expression cassettes is required for effective transient and long-term studies. Transfer of functional shRNAs using lentiviral vectors appears to be a valid approach for effective, stable construct delivery to both cell lines [ 10 ] and primary cells [ 11 - 13 ]. More recently, using several different expression systems and target cells, reports have highlighted the utility and specificity of the RNAi approach [ 14 - 17 ]. Maintaining RNAi-inducing dsRNA below 30 nucleotides in length was thought to avoid activation of the interferon-induced anti-viral response. PKR is a key anti-viral regulator and its expression can be induced by the interferon response [ 18 ]. PKR is activated when bound to dsRNA longer than 30 nucleotides, despite interacting with shorter dsRNA molecules [ 19 ]. Four recent reports have pointed towards limitations to using RNAi as a tool in mammalian cells. The first demonstrated off-target gene silencing [ 20 ], highlighting the redundancy of short nucleotide sequences in the human transcriptome. The second characterised the expression profile of genes as a result of lentiviral vector-mediated RNAi. Interferon regulated gene expression was increased even with dsRNAs as short as 19 nucleotides [ 21 ]. The third report demonstrated similar interferon response gene up-regulation, after transfection of cell lines with synthetic siRNAs as short as 21 nucleotides [ 22 ]. Finally, Scacheri et al documented significant siRNA sequence-dependent changes in the expression of non-targeted proteins [ 23 ]. In this work we used a simple approach for gene silencing in human tumor cell lines, using lentiviral vectors for stable delivery of shRNAs. We aimed to study the effects of targeting the plasminogen activator inhibitor-2 (PAI-2 or SERPINB2) mRNA on cell survival in the presence of pro-apoptotic stimuli. In addition to its inhibitory activity on the urokinase plasminogen activator, PAI-2 is thought to have anti-apoptotic properties [ 24 , 25 ]. Its molecular targets in this respect are unknown. A recent report demonstrated a functional interaction between PAI-2 and the retinoblastoma protein cell cycle regulator [ 26 ]. Using lentiviral vectors for delivery of RNAi-inducing expression cassettes we achieved potent PAI-2 gene silencing, accompanied by a rapid cytotoxic effect. The degree of cytotoxicity was proportional to shRNA length and induction of an interferon response gene could be detected when shRNA of 21 complementary base pairs or more was expressed. The phenotype was not target gene specific, as PAI-2 overexpression failed to rescue cytotoxicity and control hairpins were also cytotoxic. Using lower vector titre, reduced shRNA expression and interferon response induction was measured without compromising gene silencing. Using a 19 complementary base pair shRNA expression vector, which reduced PAI-2 expression and induced no initial cytotoxicity or interferon response, transduced cell marker gene expression was maintained but gene silencing lost in long-term cell culture. Our results highlight the need for careful controls to monitor specificity and maintenance of gene silencing when using RNAi for stable loss-of-function studies in mammalian cells. Results Efficient transfer of RNAi-inducing cassettes using lentiviral vectors Lentiviral vectors were generated which deliver an expression cassette for human U6 promoter-driven expression of short hairpin RNA (shRNA), with exact homology to the human PAI-2 mRNA. The vector expression cassette also bears the enhanced green fluorescent protein (GFP) gene under the control of the EF-1α promoter, and an internal ribosome entry site (IRES) sequence (see Figure 1A ). This cassette allows permanent expression of GFP in transduced cells, and the possibility of concomitant overexpression of a further cDNA, between the EF-1α promoter and IRES sequences, not used here. The shRNA sequences were chosen from the PAI-2 mRNA to include a 5' guanosine at the U6 promoter transcriptional start site, to exclude the 5' and 3' 100 nucleotides of the PAI-2 open reading frame, and to be between 30 and 70 % guanosine/cytidine rich. As controls, we have used a vector leading to expression of GFP alone (EGFP), a vector with the U6 promoter and transcriptional termination signals but lacking a hairpin encoding sequence (U6PT), and vectors leading to expression of scrambled sequences of certain hairpins. Figure 1 demonstrates the efficient transduction of Isreco-1 (IS-1) human colorectal carcinoma cells with one such PAI-2 targeting vector (sh325). Sh325 is designed for expression of a shRNA with a 25 nucleotide double-stranded stretch (a 25 mer) to target the PAI-2 mRNA. As controls, we used U6PT and a scrambled sequence (sh325scr) vector. Four days after transduction, each cell population expressed high levels of GFP, as a marker for transduction (Figure 1B ). Compared to non-transduced cells, or cells transduced with the U6PT control vector, we measured a clear knock-down of endogenous PAI-2 protein and mRNA in cells transduced with the sh325 vector (see figure 1C and 1D ). Figure 1 Effective gene silencing using lentiviral vectors for RNAi. The gene transfer cassette common to each vector for RNAi is shown in A. Each construction for RNAi was designed for expression of a shRNA, homologous to the target mRNA or with a scrambled sequence, driven by the RNA polymerase III-controlled human U6 promoter and ending with a terminator (T) sequence. The shRNA is represented by two arrows which encode 19 to 25 nucleotide complementary sequences and are joined by an eight nucleotide loop (L). EGFP expression is via the EF-1α promoter, oriented in the opposite direction, driving an IRES sequence and the EGFP gene. Each cassette is flanked by the HIV long terminal repeats (LTR), of which the 3' LTR is modified to ensure that the vectors are self-inactivating upon integration (SIN). B shows flow cytometry analysis of non-transduced IS-1 cells and cells four days after transduction with the U6PT control vector, a vector for expression of shRNA complementary to a region of the PAI-2 mRNA (sh325) and a vector for expression of a shRNA with a scrambled sh325 sequence (sh325scr). C shows an immunoblot for detection of PAI-2 in the cell lysate of these cells. NS highlights a single non-specific band which is consistently detected in PAI-2 immunoblots using IS-1 cell lysates. In D, PAI-2 mRNA levels from the same samples are measured by QRT-PCR of cDNA, using the ΔCT method and hypoxanthine phosphoribosyl transferase (HPRT) as the control gene. Each target gene was detected in duplicate, error bars represent the standard deviation of mean values. However, the control vector with a scrambled sequence (sh325scr) also reduced PAI-2 mRNA and protein levels. Equal sample loading for immunoblots was confirmed by Ponceau S staining of nitrocellulose membranes (data not shown) and the presence of equal amounts of a PAI-2 monoclonal antibody-reactive non-specific band in each blot (NS in Figure 1C ). Rapid cytotoxic effect of RNAi vectors Many of the initial loss-of-function studies using RNAi have measured the phenotypic effect of gene silencing in the immediate time frame after introduction of the siRNA or RNAi-inducing expression vector. As seen in Figure 1 , four days after transduction with our vectors appears to be sufficient for efficient target gene silencing. Transduction with lentiviral vectors leads to stable long-term integration of the desired transgene cassette, a key advantage in their use compared to other transient or less stable expression systems. Thus we reasoned that transduction with RNAi-inducing cassettes, using lentiviral vectors, would also be stable unless the reduction in target gene expression gave transduced cells a significant growth disadvantage or cytotoxic phenotype. Four to five days after transduction, IS-1 cells bearing the sh325 construct or cells transduced with a scrambled sh325 sequence rapidly changed morphology, compared to U6PT-transduced control cells. Sh325-transduced cells began to disintegrate into small particles and detach from cell culture dishes. After 10 days most of the sh325-transduced cells were dead while the U6PT-transduced cells were growing like the parent cell line. The scrambled hairpin vector-transduced cells gave a weaker cytotoxic phenotype, with deteriorating cell morphology and some detachment of transduced cells. Figure 2A shows the morphology of IS-1 cells 6 days after transduction. To understand further this cytotoxic effect, we performed quantitative RT-PCR (QRT-PCR) on RNA isolated from IS-1 cells, 4 days after transduction with the same vectors, in order to measure the levels of the 2'5'-oligoadenylate synthetase-1 (OAS1) mRNA after transduction with each vector. The OAS1 gene is recognised as an interferon response gene and has also been monitored elsewhere when using RNAi [ 21 ]. We measured increases in OAS1 expression in both sh325 and scrambled sh325 vector-transduced cells, whereas control-transduced cells (U6PT) had equal OAS1 levels to non-transduced cells (see Figure 2B ). Figure 2 Cytotoxicity and OAS1 induction with RNAi vectors. In A, morphology was observed using phase contrast microscopy of non-transduced IS-1 cells, or cells six days after transduction with vectors leading to expression of no shRNA (U6PT), a 25 mer shRNA targeting PAI-2 (sh325) and a scrambled 25 mer control shRNA (sh325scr). B shows comparison of OAS1 expression in non-transduced cells or cells four days after transduction with U6PT, sh325 and sh325scr vectors, by QRT-PCR. Each target gene was detected in duplicate, error bars represent the standard deviation of mean values. 19 mer shRNAs induce less cytotoxicity than longer hairpins and do not increase OAS1 expression As both the target gene-specific and the scrambled sequence 25 mer shRNAs, sh325 and sh325scr, induced the OAS1 interferon response gene, we generated further vectors for delivery of shRNAs with reduced hairpin length. We reduced the length of sh325, from 25 to 23, 21 and 19 nucleotides and named the novel vectors sh323, sh321 and sh319, respectively. The truncations were made at the 3' end of the 25 nucleotide sense strand and therefore the 5' of its complementary anti-sense sequence (see Table 1 ). Each vector was used to transduce IS-1 cells and the growth of GFP positive cells monitored at 4 and 10 days after transduction, compared to U6PT control-transduced cells (see Figure 3A ). Targeting of the PAI-2 mRNA and protein was monitored, four days after transduction, by QRT-PCR and immunoblotting of cell lysates (Figure 3B and 3C ). While each PAI-2 targeted vector successfully reduced PAI-2 mRNA and protein four days after transduction, a strong negative selection was seen for shRNA-expressing cells after a further six days of culture. This selective pressure on transduced cells was stronger with the 21 mer, 23 mer and 25 mer shRNAs than with the shorter sh319-derived 19 mer (Figure 3A ). OAS1 mRNA levels were measured by QRT-PCR of transduced cell cDNA four days after transduction (Figure 3B ). The cells transduced with the 21 mer, 23 mer and 25 mer shRNAs showed induction of OAS1 mRNA, however, contrary to our expectations, highest OAS1 levels were obtained with the 21 mer shRNA. Loss of GFP positive cells over time was comparable for 21 mer, 23 mer and 25 mer hairpin constructs. To determine whether the lack of OAS1 induction was specific to sh319 or common to other 19 mer shRNAs, further transductions and QRT-PCR analysis were performed on mRNA from IS-1 cells transduced with sh319, sh321 and sh319scr vectors. sh319scr encodes a shRNA with a scrambled sh319 sequence. This analysis also confirms the specificity of the PAI-2 silencing, by comparing sh319 to sh319scr. Figure 3D shows that no induction of OAS1 was measured using sh319 or sh319scr vectors and sh319scr had no effect on the PAI-2 mRNA level. Table 1 Construct details and shRNA sequences. Vector names and the shRNA sequences they encode. In comments, numbers are coding human PAI-2 mRNA nucleotides (adapted from accession number M18082). Hairpin sequence Name sense loop antisense Comments sh319 GCGCACACCUGUACAGAUG CAAGCUUC CAUCUGUACAGGUGUGCGC PAI-2 684–702 sh321 GCGCACACCUGUACAGAUGAU CAAGCUUC AUCAUCUGUACAGGUGUGCGC PAI-2 684–704 sh323 GCGCACACCUGUACAGAUGAUGU CAAGCUUC ACAUCAUCUGUACAGGUGUGCGC PAI-2 684–706 sh325 GCGCACACCUGUACAGAUGAUGUAC CAAGCUUC GUACAUCAUCUGUACAGGUGUGCGC PAI-2 684–708 sh319scr GUCAUACCGGCAAGGAUCC CAAGCUUC GGAUCCUUGCCGGUAUGAC scrambled sh319 sh325scr GGCCGGAGAUAAGUUCACUCAACUC CAAGCUUC GAGUUGAGUGAACUUAUCUCCGGCC scrambled sh325 sh119 GAAGACCAGAUGGCCAAGG CAAGCUUC CCUUGGCCAUCUGGUCUUC PAI-2 151–169 EGFP No hairpin, empty vector. none U6PT Human U6 promoter/terminator, no hairpin. none Figure 3 Shorter shRNA length reduced, but did not eliminate, cytotoxicity. A represents flow cytometry analysis of IS-1 cells 4 and 10 days after transduction with U6PT, sh319, sh321, sh323 and sh325 vectors. GFP expression is detected, and the percentage of GFP expressing cells was determined using the M1 gating shown (percentage GFP positive cells is shown in each histogram). B shows a comparative analysis of PAI-2 and OAS1 mRNA, in the samples described in A, 4 days after transduction. Data were generated by QRT-PCR and error bars are as described in previous figures. In C, cell lysates from samples of transduced cells described in A and B were subjected to immunoblotting with anti-PAI-2 monoclonal antibodies. Ponceau S staining served as a gel loading control, as did comparison of a single non-specific band (NS) in the immunoblot. D shows QRT-PCR analysis, as in B, for IS-1 cell mRNAs after transduction with or without U6PT, sh319, sh319scr and sh321 vectors. Cytotoxicity is not target gene specific To determine if all or part of the cytotoxic effect seen with our shRNAs was due to down-regulation of PAI-2, we generated an IS-1 cell line which overexpresses functional PAI-2. A lentiviral vector was produced which delivers the wild type PAI-2 cDNA, and used to transduce IS-1 cells. This resulted in a homogeneous population of cells which overexpress PAI-2 (see Figure 4A , IS-1 PAI-2 cells). Using immunoblotting of PAI-2/u-PA complexes, formed by mixing IS-1 PAI-2 cell lysates with low molecular weight u-PA, we demonstrated that this overexpressed protein was functional (see Figure 4B ). We transduced these cells with the series of shRNA-delivering vectors described in Figure 3 (sh325, sh323, sh321 and sh319), to test whether functional PAI-2 overexpression could reverse the cytotoxic phenotype. As even endogenous PAI-2 is not completely silenced using these vectors we reasoned that the RNAi they deliver would not be capable of functionally silencing overexpressed PAI-2. As predicted, our PAI-2 targeting shRNAs were unable to completely reduce the overexpressed PAI-2 protein levels (see Figure 4C , compared to the relative non-specific band intensity in Figure 3C ). However, the cytotoxic effect seen with the parent IS-1 cell line was also clearly apparent in the PAI-2 overexpressing cells. We monitored the loss of GFP positive cells in the transduced PAI-2 overexpressing cell populations and saw almost identical kinetics, compared to the parent cell line (compare Figure 3A and Figure 4D ). These data show that the cytotoxic effect is not target gene specific. We also measured the level of PAI-2 mRNA in this experiment, by QRT-PCR. Despite a several-fold decrease in overexpressed PAI-2 protein level (see immunoblot in Figure 4C ) we were unable to detect knock-down of the overexpressed mRNA (Figure 4E ). In a similar manner to the IS-1 parent cell line, transduction of PAI-2 overexpressing cells with sh325, sh323, sh321, but not the sh319 vector, induced the OAS1 interferon response gene (Figure 4F ). To exclude IS-1 cell-specific effects of our vectors we transduced IS-1 cells and HeLa cells with a GFP control, the sh319 and the sh319 scrambled sequence (sh319scr) vectors. We monitored the percentage of GFP positive cells at day 4, 8 and 11 after transduction and observed similar selective loss of GFP positive cells for the sh319 and sh319scr vectors, in both cell lines (Table 2 ). Figure 4 PAI-2-targeted RNAi with overexpression of functional PAI-2 in IS-1 cells. IS-1 cells were transduced with a lentiviral vector for delivery of the human PAI-2 cDNA under control of the CMV promoter. A shows a flow cytometry analysis for detection of PAI-2 in transduced IS-1 PAI-2, and non-transduced IS-1 cells. Both cell types were fixed, permeabilised and labelled with anti-PAI-2 monoclonal antibodies, then incubated with PE-labelled secondary antibodies. In B, the functional activity of overexpressed PAI-2 was assessed by immunoblotting of cell lysates, for PAI-2 expression, with or without the addition of 10U of u-PA. u-PA alone is included in lane 1. In C, PAI-2 protein levels were assessed in cell lysates from IS-1 PAI-2 cells, after transduction with U6PT, sh319, sh321, sh323 and sh325 vectors. NS in B and C highlights a single non-specific band which is consistently detected in PAI-2 immunoblots of IS-1 cell lysates. D represents flow cytometry analysis of IS-1 PAI-2 cells 4 and 10 days after transduction with U6PT, sh319, sh321, sh323 and sh325 vectors. GFP expression is detected, and the percentage of GFP expressing cells was determined using the M1 gating shown (percentage GFP positive cells are given in each histogram). E and F show comparative analyses of PAI-2 and OAS1 mRNA expression, respectively, in samples described in D. Samples were analysed 4 days after transduction. Data were generated by QRT-PCR, each target gene was detected in duplicate, error bars represent the standard deviation of mean values. Table 2 Percentage GFP positive cells over time in shRNA-expressing IS-1 and HeLa cells. Transduced cells were assessed for GFP expression by flow cytometry. GFP positive cells were gated equally for each cell type, 4, 8 and 11 days after transduction with EGFP, sh319 and sh319scr vectors. % GFP positive cells IS-1 HeLa Vector Day 4 Day 8 Day 11 Day 4 Day 8 Day 11 EGFP 95 95 93 92 94 93 sh319 86 70 56 79 68 50 sh319scr 90 76 59 82 64 45 Reducing lentiviral vector titre can reduce shRNA expression level and OAS1 induction, while maintaining gene silencing To understand whether lentiviral vector titre and resulting shRNA expression levels influence the non-specific effects described, IS-1 cells were transduced with the U6PT, sh319 and sh321 vectors as described above and also using a 10-fold reduction in vector titre (vector titre details are given in the figure 5 legend). Both sh319 and sh321 vectors effectively down-regulate PAI-2, but only the sh321 vector induces OAS1 expression (see Figure 3B ). Using 10-fold lower vector titre had little impact on PAI-2 mRNA silencing, as seen using QRT-PCR (Figure 5A ), and resulted in a small decrease in OAS1 mRNA induction (Figure 5B ) in sh321-transduced cells. To assess whether lower vector titre resulted in lower shRNA expression, total RNA from cells transduced with different titres of sh321 vector was subjected to RNase digestion after hybridization with a 32 -P labelled RNA probe, designed to protect the first 19 nucleotides of the sh325 shRNA. Upon hybridization, this "sh3" probe should protect short RNA expressed in sh321 vector-transduced cells from RNase digestion. As negative controls, RNA from U6PT-transduced cells was subjected to the RNase protection procedure, and the sh3 probe was treated with RNase without target RNA. As a positive control a probe for the Mir-16 miRNA was constructed and used to detect endogenous Mir-16 miRNA in U6PT-transduced cells using the same protocol (Figure 5C ). Sh3 probe-protected RNA was detected in sh321-transduced cells. Reducing the viral titre clearly reduced the expression of shRNA and this correlated with reduced OAS1 induction. Sh3 probe-protected RNA was not detected in control-transduced cells and the sh3 probe was completely digested in the absence of target RNA (Figure 5C ). This data demonstrates that vector-derived shRNA expression can be reduced without impacting gene silencing and that lower expression correlates with a reduced interferon response. Figure 5 Efficient target gene silencing with reduced OAS1 induction and lower shRNA expression. Total RNA was isolated from IS-1 cells (CTRL) or IS-1 cells 4 days after transduction with U6PT, sh319 or sh321 vectors. After reverse transcription, cDNA was analysed for expression of PAI-2 (A) and OAS1 (B). Data were generated by QRT-PCR, each target gene was detected in triplicate, error bars represent the standard deviation of mean values. 1 ml or 0.1 ml of each lentiviral vector stock was used, hence the designations 1 and 0.1. Vector titres were approximately 10 6 transducing units per ml resulting in a multiplicity of transduction of approximately 10 for 1 ml used or 1 for 0.1 ml used. In C, shRNA expression was detected in total cell RNAs using a modified RNase protection protocol. Total RNA was mixed with radiolabelled probes for hybridization and RNase protection. Samples were resolved on a 15 % Acrylamide/8M Urea/TBE gel and RNase protected probes detected by autoradiography. Lane 1 shows the Mir-16 probe without RNase digestion or target RNA, lane 2 is as lane 1 with RNase digestion and lane 3 as lane 2 with U6PT-transduced cell RNA as hybridization target. Lane 4 shows the sh3 probe without RNase digestion or target RNA, lane 5 as lane 4 with RNase digestion and lane 6 as lane 5 with U6PT-transduced cell RNA as hybridization target. In lane 7 sh321-transduced cell RNA was used as the hybridization target for the sh3 probe with RNase digestion, and lane 8 is as lane 7 except that cells were transduced with 10-fold less vector titre (sh321 1 and sh321 0.1). Known nucleotide lengths (Ntds.) for the probes and the protected Mir-16 endogenous RNA are marked. Loss of long-term gene silencing despite persistent transduction marker gene expression In an attempt to generate a cell line with stable PAI-2 mRNA silencing without interferon response induction, we generated additional vectors which deliver 19 to 25 nucleotide shRNAs targeting different regions of the PAI-2 mRNA. Of these, the sh119 construct reduced PAI-2 expression, did not induce OAS1 and had no effect on cell morphology one week after transduction (data not shown). In parallel, we transduced IS-1 cells with the sh119 vector or a GFP control vector. We achieved high percentage transduction rates which were monitored for over two weeks (Figure 6A ). GFP positive cells, from control- and sh119-transduced populations were sorted twice, using flow cytometry, to further enrich the GFP positive population of each (EGFPs and sh119s). Sh119-transduced cells showed PAI-2 gene silencing 10 days after transduction but not after one month of cell culture (Figure 6B and 6C ). GFP marker gene expression was maintained in both sorted cell populations. During the prolonged culture period we noticed that, while the percentage of GFP positive cells remained stable, the intensity of GFP detected was reduced in the sh119-transduced cells, compared to GFP alone controls (Figure 6A ). This was accompanied by a significant reduction in integrated vector copies (data not shown). Thus, although the marker protein was maintained at a reduced expression level, the prolonged culture period selected against cells with effective gene silencing. This suggests the presence of subtle cytotoxic effects of short dsRNA expression which are not apparent in the initial post transduction period and are in the absence of interferon response gene induction. Overexpression of PAI-2 in the IS-1 cells did not reduce the long-term selective effect on transduced cells or GFP expression levels in sh119-transduced cells (see Table 3 ). Figure 6 Long-term gene silencing is not stable, despite persistent marker gene expression. In A, IS-1 cells transduced with EGFP control or sh119 vectors were analysed by flow cytometry 3 and 17 days after transduction, and compared to non-transduced cells. EGFP expressing cells, from both transduced cell populations, were selected by cell sorting and named EGFPs and sh119s. 39 days after transduction, these cells were analysed for EGFP expression. Percentage EGFP positive cells, assessed by the M1 gating shown, are given in each histogram. In B, PAI-2 expression in EGFP and sh119 vector-transduced cell lysates were analysed by immunoblotting, 10 days after transduction. C is the same PAI-2 immunoblot as B, performed using EGFPs and sh119s cell lysates, 33 days after transduction. Table 3 Percentage GFP positive cells and mean GFP fluorescence in sh119-transduced IS-1 and IS-1 PAI-2 cells. Transduced cells were assessed for GFP expression by flow cytometry. GFP positive cells were gated equally for each cell type, 5 and 31 days after transduction with the sh119 vector. Mean F is the mean fluorescence of gated GFP positive cells. % GFP positive cells Mean F of GFP positive cells Day5 Day31 Day5 Day31 IS-1 cells 85 40 146 83 IS-1 PAI-2 cells 82 46 180 96 Discussion Here we report the use of lentiviral vectors for the delivery of expression cassettes designed for RNAi-induced stable knock-down of gene expression. We undertook this approach because of the promise of RNAi to easily create cell lines that are specifically deficient in one protein component. By using lentiviral vectors for shRNA expression, high transduction efficiencies can be achieved, avoiding effects due to clonal selection of phenotypically different cells. We successfully targeted the PAI-2 mRNA with the aim of studying the effects of reducing PAI-2 activity on cell sensitivity to apoptosis-inducing stimuli. PAI-2 activity has previously been implicated in protection from apoptosis [ 24 ]. The cytotoxic effect we observed, in cells transduced with RNAi-inducing vectors, appeared to correlate well with the reduction in PAI-2 protein levels. However, rapid selective growth pressure on cells bearing shRNA constructs with scrambled sequences, having no complementarity to the PAI-2 mRNA, suggested non-specific effects rather than a PAI-2-related phenotype. Using GFP as a marker gene, delivered by all vectors, enabled very sensitive detection of selective effects on transduced cells even when initial cell culture suggested stable transduction and cell growth. We were able to detect increased expression of an interferon response gene, OAS1, in cells transduced with all hairpins of 21 or more base-paired nucleotides. A 21 mer hairpin induced the most potent OAS1 induction. These results, and those of others [ 21 , 22 ], suggest that dsRNAs of less than 30 nucleotides can induce an interferon response, even if they cannot directly activate protein kinase R [ 19 ]. The absence of OAS1 induction in cells transduced with 19 mer shRNAs implies that the search for appropriate hairpin sequences should be limited to stretches of this length or less. Overexpression of functional PAI-2 did not rescue the cytotoxic effects or OAS1 induction observed in cells transduced with a series of vectors for expression of different length shRNAs. This result, the cytotoxicity associated with scrambled sequence hairpin-encoding constructs, and the same selective pressure seen on the growth of transduced HeLa cells, which do not express detectable PAI-2 mRNA or protein (Table 2 and data not shown), lead us to conclude that the phenotype we have seen in IS-1 cells is not PAI-2-specific. Without the ability to track transduced cells, via GFP expression, this conclusion would have been more difficult to obtain. In cells engineered to overexpress functional PAI-2, our RNAi-inducing vectors clearly reduced PAI-2 protein levels but did not significantly reduce the overexpressed PAI-2 mRNA. This suggests that in the presence of high concentrations of targeted mRNA, the machinery necessary for RNAi-induced mRNA cleavage is saturated and mRNA down-regulation undetectable. As protein levels are nevertheless reduced, the shRNA may be functioning post-transcriptionally, perhaps in a similar manner to natural miRNA. This phenomenon has been described elsewhere for siRNAs [ 27 ]. IS-1 cell transduction, using one PAI-2 targeting vector (sh119), initially appeared stable, compared to control-transduced cells. PAI-2 protein levels were clearly reduced 10 days after transduction and selective pressure on cell growth appeared to be minimal, as the percentage of sh119 GFP positive cells was apparently stable at about 80 %, 17 days after transduction. However, after cell sorting of GFP positive cells and prolonged cell culture of over one month after transduction, the PAI-2 antigen measured in sh119s (s for selected) cells was restored to control-transduced cell levels. Despite maintenance of transduction marker expression, gene silencing was absent. The GFP expression levels in selected sh119s cells was reduced after one month of growth, compared to cells monitored three days after transduction. It is possible that we selected cells with a greatly reduced, non-cytotoxic shRNA expression level, as we have detected a reduced number of integrated vector copies. The negative selective effect on cells transduced with the sh119 vector was not due to the suppression of PAI-2 expression, as PAI-2 overexpressing cells showed the same negative selection. In all experiments in which a selective pressure on growth was apparent on shRNA-expressing cells, reduced percentage GFP positive cells and reduced GFP expression in transduced cells was measured over time. We hypothesised that very high expression levels of the various shRNAs is cytotoxic. This could occur via high numbers of transcriptionally active vector integration events or integration at transcriptionally active chromosomal regions. Both might be controlled using a tightly regulatable expression system, which has been described [ 15 ], but may require careful dosage in a gene- and cell-specific manner. Our data demonstrate the importance of appropriate controls for using RNAi, as proposed in a recent editorial [ 28 ]. These include suppression of the RNAi phenotype by target gene overexpression, use of scrambled dsRNAs, and monitoring of non-specific gene expression in particular of interferon-responsive genes such as OAS1. Also, if considering the use of stable RNAi, it is imperative that stable knock-down is demonstrated as well as stable marker gene expression. In the cell culture system we used, conclusions drawn from experiments several days after RNAi delivery mask effects which are only apparent days later by monitoring the percentage of transduced cells. Such effects are likely be present in experiments using regulatable RNAi systems or using exogenously added dsRNA, where the experimental data linked to gene targeting may be collected before other effects are seen. The molecular events which lead to the long-term effect we have documented may well be underway during this experimental period. In the light of our own data and other recent reports [ 20 - 22 ], solutions for the induced cytotoxic effects we describe here include testing a series of target sequences, using dsRNA of no more than 19 nucleotides at low effective vector doses, and careful monitoring of transduced cell phenotype with and without functional target gene overexpression. Long-term monitoring of gene silencing appears to be necessary in stable systems, even in the presence of marker gene expression. Conclusions Our study demonstrates vector-derived RNAi in tumor cell lines and points towards the necessity of careful, but clearly feasible, controls when using RNAi for stable gene suppression in short- and long-term experiments. Methods Cell lines Isreco-1 (IS-1) cells were a gift from Dr. B. Sordat (ISREC, Lausanne). 293T cells were a gift from Dr. D. Trono (Geneva University Medical Centre). HeLa cells were purchased from the European collection of cell cultures, ECACC number: 93021013. All cells were maintained in DMEM supplemented with 10 % Fetal Bovine Serum and 10 mM HEPES pH 7.4 (IS-1 medium) (purchased from Invitrogen). Plasmid constructions Gene transfer plasmids, for RNA interference using lentiviral vectors, were constructed using the backbone of ploxEWiresGFP, a gift from Dr. P. Salmon, Geneva University Medical Centre. The human U6 gene was amplified by PCR using HeLa cell genomic DNA as template and the oligonucleotides U6-ClaI-F 5' GATC ATCGATAAGGTCGGGCAGGAAGAGGGCCTATTTCCC 3' and U6-ClaI-R 5' GATCATCGATTGGTAAACCGTGCACCGGCGATAAACG 3'. The 483 base pair PCR product was digested with ClaI, inserted into the ClaI site of pTRE2 hyg (Clontech), and its sequence verified by DNA sequencing. The U6 promoter and gene sequence corresponded to nucleotides 65 to 527 of Genbank accession number M14486. This plasmid was used as template for PCR reactions to amplify U6 promoter-driven expression cassettes. Each PCR product included the U6 promoter with shRNA-encoding sequences beginning at the U6 +1 site, and a run of 6 or 7 thymidine bases for an RNA polymerase III transcription termination signal. Each PCR introduced flanking ClaI restriction sites. PCR products were directly cloned into the pGEM Teasy (Promega) plasmid and sequenced. ClaI fragments of positive clones were excised and ligated into the unique ClaI site of ploxEWiresGFP. The U6 promoter cassettes in the resulting lentiviral vector plasmids were verified by DNA sequencing. Each construct used for vector production had the same orientation of the U6 expression cassette with respect to the P EF-1α-iresGFP region of the plasmid. A schematic of the gene transfer cassette is given in Figure 1A . Details of shRNA sequences used for each construct are given in Table 1 . The gene transfer plasmid for PAI-2 overexpression was constructed by replacing the EGFP gene from ploxCW-GFP (a gift from Dr. P. Salmon, Geneva University Medical Centre) with the type B human PAI-2 (SERPINB2) [ 29 ] open reading frame. Vector production and transduction Lentiviral vectors were produced by three plasmid co-transfection of 293T cells, essentially as described previously [ 30 ]. Vectors were harvested 48 hours after transfection, passed through 0.45 μm filters and used directly on target cells in a 1:1 ratio with IS-1 medium in a total volume of 2 ml. Transductions were performed in 3 cm diameter 6-well plates, on cells seeded the previous day at 5 × 10 4 or 1 × 10 5 cells/well. Upon addition of vectors, plates were centrifuged for one hour at 1000 g in the presence of 8 μg/ml polybrene (hexadimethrine bromide, Sigma). After 24 hours, target cells were washed twice with PBS and cultured in IS-1 medium until analysis. Where indicated, viral titre was determined by transducing cells with 10 μl of lentiviral vector conditioned medium from 293T cell producer cells and measuring the % of GFP positive cells by flow cytometry. We routinely achieve 10 6 effective transducing units per ml of producer cell conditioned medium, which results in a typical multiplicity of transduction of 10. Comparisons between cell populations transduced with different vectors is made by flow cytometry analysis of GFP positive cells. Antibodies Anti-human PAI-2 monoclonal antibody 3750 was purchased from American Diagnostica. PE-labelled goat anti-mouse antibodies were purchased from Pharmingen. HRP-conjugated goat anti-mouse antibodies were purchased from Bio-Rad. Flow cytometry Flow cytometry was performed using Becton Dickinson FACScan, FACStrack or FACScalibur instruments at the Geneva University Medical Centre flow cytometry facility. GFP was detected in cells detached and resuspended in FACS buffer comprising 1 % BSA in PBS supplemented with 0.05 % sodium azide. For the detection of intracellular PAI-2, a method adopted from Dr. M. Ranson (University of Wollongong, Australia) was used. Cells were detached and fixed in 0.25 % paraformaldehyde (PAF)/PBS for one hour on ice. Cells were permeabilised in 0.1 % saponin/PBS for 30 minutes at room temperature. Fixed, permeabilised cells were incubated for 30 minutes in PBS/0.5 % BSA/0.1 % saponin containing 2 μg/ml 3750 anti-PAI-2 monoclonal antibody, washed twice in PBS/0.1 % saponin and incubated for 30 minutes in 0.1 % saponin/0.5 % BSA/PBS/goat anti-mouse-PE antibodies. Finally, cells were washed twice in 0.1% saponin/PBS, twice in PBS and resuspended in 2.5 % PAF for flow cytometric analysis. Sorting of live GFP positive cells was performed using a FACStar+ instrument (Becton Dickinson). Western blotting of cell lysates Cells in suspension were lysed in 10 mM Tris-HCl (pH 7.4), 10 mM NaCl, 0.5 % NP-40, 3 mM MgCl 2 , 5 mM DTT and 1 mM PMSF for one hour on ice. Lysates were centrifuged at 16000 g for 5–10 minutes to remove nuclei and precipitates. Supernatant protein concentrations were measured using the Bio-Rad protein assay with BSA in lysis buffer as a standard. Cell lysates were separated by reducing SDS-PAGE and transferred to nitrocellulose membranes. Equal total protein lysate was used for each blot, between 2.5 and 10 μg depending on the assay. Membranes were blocked for 1 hour at room temperature in TBS-0.1 % Tween 20/5 % non-fat milk, and probed using antibodies in TBS-0.1 % Tween 20/5 % non-fat milk. The 3750 anti-PAI-2 antibody was used at a concentration of 1 μg/ml. Microscopy Phase contrast microscopy was performed using a Zeiss Axiovert 100 M instrument on live cells. Images were collected using a Hamamatsu CCD camera (ORCA-100). PAI-2 functional activity To measure PAI-2 functional activity, IS-1 and IS-1 PAI-2 cells were lysed in 1 % NP-40, 150 mM NaCl and 50 mM Tris pH 8.0. Thereafter, 6 μg of cleared total cell lysate was incubated with and without 10 U of low molecular weight urokinase (u-PA). 3 μg total cell lysate samples were then subjected to reducing SDS-PAGE and immunoblotting as for other PAI-2 immunoblots. Quantitative RT-PCR analysis of mRNA Total RNA was isolated from cells using RNA preparation kits from Qiagen or TRIZOL ® reagent (Invitrogen). cDNA was generated using ImpromII ® reverse transcriptase (Promega) and random hexamers, according to the manufacturers instructions, typically using 1 μg of total RNA per reaction. Quantitative PCR was performed using an Applied Biosystems Prism 7000 instrument using Applied Biosystems SYBR ® green master mix reagent and oligonucleotide pairs to detect hypoxanthine phosphoribosyl transferase (HPRT), PAI-2 and oligoadenylate synthase-1 (OAS1) cDNA. 5' to 3' primer sequences were as follows: HPRT forward TATTGTAATGACCAGTCAACAG, HPRT reverse GGTCCTTTTCACCA GCAAG, PAI-2 forward GGGTCAAGACTCAAACCAAAG, PAI-2 reverse CCTTTGAAGTAGACAGCATTC, OAS1 forward AGGTGGTAAAGGGT GGCTCC and OAS1 reverse ACAACCAGGTCAGCGTCAGAT. Data were analysed using Applied Biosystems Prism software and the ΔC T method. Briefly, target gene expression was normalised to the HPRT endogenous reference gene for each sample. The difference between mean threshold PCR cycle values for target and control genes gave the ΔC T value. This was then calibrated to the control sample in each experiment to give the ΔΔC T value, where the control had a ΔΔC T value of 0. The fold target gene expression, compared to the calibrator value, is given by the formula 2 -ΔΔCT . Error bars represent the standard deviation of each target gene value, after evaluating the expression 2 -ΔΔCT+s and 2 -ΔΔCT-s , where s is the standard deviation of the ΔΔC T value. All reactions were performed in duplicates or triplicates. ShRNA detection Expression of shRNAs was detected using the mir Vana™ Probe Construction and miRNA Detection kits from Ambion (Austin), according to the manufacturers instructions. These kits employ in vitro transcription for radiolabelled probe generation and an RNase protection protocol for detection of small RNA expression, respectively. Briefly, radiolabelled RNA probes incorporating α- 32 P-UTP (Amersham) were constructed using T7 polymerase-driven transcription templates. Templates were designed to generate RNAs which hybridize to the endogenous Mir-16 miRNA, as a control, or the first 19 nucleotides of the sh325 shRNA (see table 1 ). The sh3 probe should therefore hybridize to any of the sh319/sh321/sh323 or sh325-derived shRNAs. Before hybridization, radiolabelled probes were purified by migration of the transcription reaction on a 12 or 15% Acrylamide/8 M Urea denaturing gel and elution of radioactive probe bands after detection by autoradiography. 3 μg of total cellular RNA was used for all hybridizations and RNase protections, supplemented with 2 μg yeast RNA as a carrier. Control reactions without target RNA but with or without RNase digestion included 5 μg yeast RNA. After hybridization and RNase digestion, protected probes were detected by autoradiography after migration in a 15% acrylamide/8 M Urea denaturing gel. Authors' contributions RF carried out all of the experiments in this study and contributed to its conception, design and description. EKOK conceived the study and participated in its design and description. Both authors approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514603.xml
521684	Strengths and weaknesses of EST-based prediction of tissue-specific alternative splicing	Background Alternative splicing contributes significantly to the complexity of the human transcriptome and proteome. Computational prediction of alternative splice isoforms are usually based on EST sequences that also allow to approximate the expression pattern of the related transcripts. However, the limited number of tissues represented in the EST data as well as the different cDNA construction protocols may influence the predictive capacity of ESTs to unravel tissue-specifically expressed transcripts. Methods We predict tissue and tumor specific splice isoforms based on the genomic mapping (SpliceNest) of the EST consensus sequences and library annotation provided in the GeneNest database. We further ascertain the potentially rare tissue specific transcripts as the ones represented only by ESTs derived from normalized libraries. A subset of the predicted tissue and tumor specific isoforms are then validated via RT-PCR experiments over a spectrum of 40 tissue types. Results Our strategy revealed 427 genes with at least one tissue specific transcript as well as 1120 genes showing tumor specific isoforms. While our experimental evaluation of computationally predicted tissue-specific isoforms revealed a high success rate in confirming the expression of these isoforms in the respective tissue, the strategy frequently failed to detect the expected restricted expression pattern. The analysis of putative lowly expressed transcripts using normalized cDNA libraries suggests that our ability to detect tissue-specific isoforms strongly depends on the expression level of the respective transcript as well as on the sensitivity of the experimental methods. Especially splice isoforms predicted to be disease-specific tend to represent transcripts that are expressed in a set of healthy tissues rather than novel isoforms. Conclusions We propose to combine the computational prediction of alternative splice isoforms with experimental validation for efficient delineation of an accurate set of tissue-specific transcripts.	Background The large difference between cells from different tissues is the consequence of a complex regulatory machinery guiding the tissue specific expression of genes and their transcripts. Several genes have been described to exhibit differential splicing patterns for different tissues (E.g. PDE1C [ 1 ]; IRF-3 [ 2 ]) that result either in alternative proteins or affect the regulation of the respective gene product [ 3 ]. Due to the large number of genes generating alternative transcripts as well as by the complicated splicing machinery involving a large variety of different proteins, mis-splicing events are also frequently observed. Some of these artificial splice isoforms are already linked to specific diseases like Hemophilia A, Marfan syndrome etc. [ 4 , 5 ]. The resource mainly used to predict tissue-specific expression is the rapidly expanding repertoire of expressed sequence tags (ESTs) in the public databases representing a wide spectrum of tissue types. Unlike serial analysis of gene expression (SAGE) which mainly yields the tissue specific expression of genes [ 6 ], the EST data additionally allow the identification of alternatively spliced transcripts [ 7 - 11 ]. Besides the detection of the existence of alternative splice isoforms the tissue annotation of ESTs can be used for the computational prediction of the expression pattern of these transcripts where the tissue-wise count of transcript-specific ESTs with respect to a random background distribution defines an expression level [ 12 - 14 ]. Transcripts that are significantly over-represented by ESTs derived from a single tissue are usually defined as being tissue-specifically expressed. However, different cDNA construction protocols like normalization [ 15 ] include subtractive hybridization and PCR amplification steps introducing an artificial enrichment of ESTs from lowly abundant transcripts. The level of enrichment depends on the number of normalization/amplification steps performed, measured as Cot or Rot [ 16 ]. This inconsistency in the correlation of the number of ESTs observed for a transcript and its real expression level may affect the reliability of computational predictions of tissue-specifically expressed transcript. Since the EST-based prediction of expression patterns might already be error-prone because of the lack of sufficient numbers of EST sequences for each tissue this might be further complicated by different cDNA library protocols. Consequently, EST data related to normalized cDNA libraries are excluded from analysis in several computational approaches that aim at predicting tissue-specific expression [ 13 , 17 ]. Because of these uncertainties we combined our computational prediction of alternative splice variants and their expression pattern with experimental validation of these iso-forms via RT-PCR on 40 different tissue samples in order to evaluate the predictive potential of ESTs. Results The EST-based prediction of alternative splice iso-forms revealed 427 genes each contributing at least one potential tissue-specifically expressed variant. These variants show specificity for 28 different tissue types, where brain, testis and placenta account for approximately half of these transcripts (see additional file 1 ). Many of these genes (n = 210) exhibit isoforms that were exclusively detected due to ESTs derived from normalized libraries. These form a significant fraction (p-value: 8e-19) of the total genes that show tissue specific transcripts, since the number of ESTs derived from normalized libraries (896,645) is only 30% the total EST count (3,084,576) in tissues for which tissue specific isoforms exist. Out of the 20 isoforms tested experimentally (see additional file 3 for details of experiments), 15 isoforms could be successfully verified in some tissue (Table 1 ). The remaining five variants are either likely to resemble rare transcripts according to the respective library construction protocol, or as in case of a disease-specific isoform (Hs.272688), the appropriate tissue sample was not available for experimental testing. Only four of the isoforms predicted based on the basis of normalized libraries could be validated using the standard RT-PCR conditions. For five additional isoforms a more refined protocol had to be applied in order to detect bands of significant strength. More sensitive PCR conditions frequently revealed expression in more tissues indicating low expression of the isoforms in these tissues. These results show the tendency of normalized libraries to be enriched for low-abundant transcripts. Table 1 RT-PCR validation results for tissue and disease-specific splice isoforms. The experiments are categorized into three groups viz. tissue specific isoforms predicted via ESTs related to non-normalized libraries (1 to 4), tissue specific isoforms predicted only via ESTs derived from normalized libraries (5 to 16) and disease-specific isoforms (17 to 20). For some of the variants represented by normalized libraries, standard PCR did not reveal the isoforms. However, five of these isoforms were detected using refined PCR conditions. The experiments frequently validated the isoforms and the tissue type, but the predicted specificity was rarely verified. Isoform Gene Chr. Unigene EST Evidence ESTs Cycles Isoform Specificity Comment (Most sensitive PCR) Norm. Level 1 Unknown 11 Hs.112250 testis 3 39 + + 2 ISGF3G 14 Hs.1706 stomach 10 39 + - Ubiquitous 3 MRPL42 12 Hs.112110 stomach-lymph 5 39 + - Ubiquitous 4 SGN3 17 Hs.6076 testis 3 39 - ? 5 PC326 13 Hs.279882 testis 9 39 + + testis [36] Rot-5 6 LMO7 13 Hs.5978 brain 5 39 + - brain, testis, eye(?) [18] 7 HRD1 8 Hs.274122 brain 3 39 + - brain, eye, thymus, salivary gland, kidney 8 Unknown 1 Hs.24119 pancreas 4 39 + - approx. 10 tissues Cot-20 9 BCLG 12 Hs.11962 testis 4 39,78 ?,+ +,+ Cot-5 10 RBPMS 8 Hs.80248 placenta 4 39,78 ?,+ -,- Ubiquitous 11 SCML1 X Hs.109655 testis 12 39,78 ?,+ +,- approx. 6 tissues Rot-5 12 WNK1 12 Hs.432900 kidney 3 39,78 ?,+ +,- Digestive system [28] Cot-25 13 NY-CO-10 5 Hs.23557 testis 3 39,78 -,+ ?,+ Cot-5 14 Unknown 11 Hs.169100 testis 3 39,78 -,- ?,? Rot-5 15 Unknown 16 Hs.48396 breast 4 39,78 -,- ?,? Cot-230 16 CIDE-A 18 Hs.249129 breast 4 39,78 -,- ?,? Cot-230 17 KCNAB2 1 Hs.298184 tumor 29 39 + - Ubiquitous 18 SNRP70 19 Hs.174051 stomach ascites 25/26 39 + - Ubiquitous 19 RAB1 14 Hs.227327 tumor 39/95 39 + - fetal(kidney, thymus, liver, spleen), ovary [19] 20 Unknown 7 Hs.272688 tumor 12 39,78 -,- ?,? relevant tumor sample not in set The predicted expression of the isoforms in a single tissue could not be confirmed for half of the variants analyzed (standard conditions). However, the isoforms were always detected to be expressed in the tissue that was originally predicted by our software. The observed expression pattern of the 'unspecific' isoforms ranges from expression in only a few, sometimes related tissues (LMO7 [ 18 ]: brain, eye, testis, Fig. 2 ; HRD1: brain, eye, thymus, salivary gland, kidney) to ubiquitous expression (MRPL42, ISGF3G). Those variants that were validated to be specifically expressed frequently originate from testis. Increasing the sensitivity of the RT-PCR revealed another testis-specific variant. At the same time the variants of the genes WNK1 and SCML1 were no longer defined as being tissue-specifically expressed since they were now also detected in a few additional tissues (Table 1 : isoform 11 & 12). The number of genes with transcripts exclusively expressed in tumors is relatively large (1120) as compared to the number of genes revealing tissue specific isoforms. Interestingly, 2 out of 4 such disease-related transcripts (Table 1 : isoform 17–20) were ubiquitously expressed although the large number of ESTs covering these variants was suggesting a high significance of the prediction. The tumor associated isoform described by Wang et al. [ 19 ] was observed to be expressed in several fetal tissues along with ovary. Figure 2 RT-PCR validation experiment of a putative brain-specific isoform ( LMO7 ) . (A) The additional exon is detected in all tissues (primers F1, R1). (B) The primer pair F1-R2 located on exons flanking the extra exon results in two products where the shorter one is observed in brain, testis and eye (weak band). The predicted brain-specific expression pattern is, in fact, not specific. The entire dataset for human as well as the gel images from the RT-PCR experiments is available at . Discussion Consistent with previous work [ 11 ] our approach of combining computational and experimental validation yields a high success rate in predicting the existence of splice variants. In line with the expected general enrichment of clones derived from lowly expressed transcripts in normalized cDNA libraries our experimental results confirm the expression of the predicted low abundance transcripts. Consequently, those isoforms that could not be validated experimentally may also reflect real biological signatures of extremely rare transcripts since they are often represented just by heavily normalized libraries (Cot 230, CIDE-A + Hs.48396). While the methods used in the construction of normalized libraries (PCR amplification, subtraction, size selection) increase the sensitivity of the detection of transcripts they unfortunately disturb the rough correlation between the expression level of a transcript and the observed number of related clones that is usually maintained in non-normalized libraries. Therefore, in these cases, the larger number of ESTs found for a specific transcript will profess to deal with a higher expressed transcript, also implying a higher confidence in the prediction although the sequences may be derived from the same although amplified clone. From the computational point of view the artificially increased number of ESTs affects the likelihood to predict tissue-specifically expressed transcripts since the prediction mainly relies on the count of ESTs [ 12 , 13 ]. Nevertheless, our experimental results show that especially isoforms predicted to be expressed exclusively in testis could be successfully validated, while other isoforms frequently appear to be expressed in a set of additional tissues that were not suggested by the ESTs. Surprisingly, the absence of supporting EST evidence for the variants LMO7 and ISGF3G is not caused by the lack of the respective cDNA libraries but may rather reflect differences in the tissue samples (e.g. enrichment of different cell types from the same organ, developmental differences) used for library construction. In the context of tumors, our data shows that the predicted tumor-specific expression of isoforms derived from ESTs usually tends not to reflect the experimentally validated expression pattern. Rather it suggests expression in a collection of different tissues although the large number of related ESTs derived from tumor would imply a high confidence in the EST based prediction. Since tumor cells often show an up-regulation of a larger number of transcripts involved in various pathways [ 20 , 21 ] the tumor-specific transcripts predicted based on the EST data may just reflect this general deregulation of gene expression. The large number of predicted tumor-related isoforms further supports this hypothesis. Nevertheless, some transcripts detected via EST data may still serve as potential tumor markers like in case of the gene PRAME [ 22 ] where the EST data as well as the experimental data suggests specific expression in testis and in a variety of different tumors (see additional file 2 ). Overall, ESTs are an extremely powerful tool to reliably unravel alternative transcripts independent of the level of expression. The functional relevance of the low abundant transcripts is not yet clear, especially if the isoforms do not affect the coding sequence. These isoforms may either be related to processes like nonsense-mediated decay (NMD: [ 23 , 24 ]) or they might be some kind of non-functional leakage of the splicing machinery. Nevertheless, since many lowly expressed genes are already known to have important regulatory functions [ 25 - 27 ] this may also hold true for a not yet defined fraction of the alternative isoforms we detected via normalized libraries. In contrast to the prediction of the existence of isoforms, the task of predicting their expression pattern is much more error-prone since EST data always covers only a subset of potential tissues with variable sensitivity. The fuzzy terminology of tissue-specific expression that is frequently used to describe significant expression in a discrete tissue or a set of tissues, is therefore strongly biased by the sensitivity of computational and experimental methods (SCML1; WNK1 [ 28 ]). Beside these technical difficulties, the definition of specificity may also depend on the regulatory network that mediates tissue-specificity. While isoforms expressed in testis are specifically expressed in a more strict sense, other isoforms are expressed in a small set of (not necessarily related) tissues eventually pointing to alternative regulatory mechanisms acting with different stringency, e.g. involving transcription factors [ 29 ], [ 30 ] and/or DNA methylation [ 31 , 32 ]. Conclusions The separate evaluation of EST data from non-normalized as well as from normalized cDNA libraries will help to categorize transcripts into highly and lowly abundant ones thus facilitating the integration of EST-based predictions with expression data from microarray experiments. We suggest that large-scale analysis of tissue-specific transcripts should be ideally based on a computational prediction of isoforms that ranks candidate transcripts, tightly coupled with experimental validation via RT-PCR or DNA microarray experiments [ 33 ]. Such an approach will lead to a comprehensive set of verified isoforms suitable for a wide range of applications in the functional analysis of the regulation of tissue-specific expression. This will also improve the detection of tumor-related isoforms that do not just reflect a general up-regulation of gene expression. Methods The basis of our work is the tissue/tumor annotation of ESTs is GeneNest database [ 34 ] and the quality prediction of alternative splicing [ 11 ], visualized in SpliceNest database [ 10 ]. Library classification The cDNA libraries of the GeneNest database were semi-automatically categorized into non-normalized, normalized/subtracted and PCR-based libraries by screening for the appropriate keywords in the original annotation of the respective EMBL database entries. All libraries for which none of the keywords were found were defined as being non-normalized. PCR-based libraries like those derived by ORESTES PCR were not used for the current analysis. Additionally, to avoid miscounting caused by PCR amplification, ESTs of the same library and with identical start/end positions in the alignment were treated as a single sequence. Since the level of normalization of different libraries may differ depending on the number of rounds of subtractive hybridizations performed, we also extracted the normalization level (measured as Cot or Rot: [ 16 ]) as far as it was noted in the respective entries. Increasing Cot-values hereby reflect the enrichment of clones derived from low abundant transcripts in the respective cDNA library. Besides the categorization of cDNA libraries according to the construction methods used we further split these groups into libraries derived from healthy or disease tissue. Finally, ESTs of the four groups of cDNA libraries (healthy/non-normalized, healthy/normalized, disease/non-normalized, disease/normalized) were either analyzed separately or data of normalized and non-normalied libraries were combined. Prediction of tissue specific alternative splicing Alternative splice isoforms in the SpliceNest database are revealed by aligning EST consensus sequences (putative transcripts) related to one gene to the appropriate genomic sequence. Significant differences in the boundaries of the putative exons are interpreted as alternative splicing events. For all exon-exon-boundaries that define a certain splice iso-form the annotation of ESTs covering the respective boundary is evaluated. Isoforms overrepresented by ESTs from particular tissue are tagged as putative tissue/tumor specific splice isoforms. Several parameters (e.g. number of ESTs from a particular tissue, number of ESTs from other tissues, number of associated mRNA sequences etc.) are computed for these isoforms and finally stored in a relational database system. The refined set of tissue and tumor specific variants is then generated by setting the requirement of at least 3 ESTs in both alternative forms. Fig. 1 describes such a prediction using GeneNest and SpliceNest visualizations. Since the counts of ESTs per tissue-specific splice event were frequently below 5, we considered it inappropriate to apply statistical methods as were used by Xu et. al. ([ 12 ]). Figure 1 Detection of brain specific splicing in gene LMO7 . The top part of the figure is a visualization of gene LMO7 in SpliceNest, showing parts of three transcripts with exons displayed as red blocks, connected by lines representing introns. The middle exon of the top transcript (Hs5978.1) is missing in the second transcript (Hs5978.2) and is therefore highlighted as an alternative splice event (green bar). The boundaries corresponding to this exon as well as the corresponding intron are visualized as vertical lines in the GeneNest database (left and right box respectively). Both regions are covered by several ESTs depicted by horizontal arrows with corresponding tissues encoded in coloured rectangles towards the left of each EST. Upon comparing the tissue distribution of these alternative regions it is evident that the middle exon of transcript Hs5978.1 is covered by ESTs derived from several tissues, while the corresponding exon junction that lacks this middle exon, in transcript Hs5978.2, is represented by ESTs derived from brain only, thereby revealing this as a brain specific splice event. Experimental verification A set of putative tissue specific (n = 16) and disease-related (n = 4) alternative splice events was arbitrarily selected for RT-PCR experiments. PCR primers were generated on the alternatively spliced exon as well as on either side of the event (Fig. 2 ) using the primer design software GenomePRIDE ([ 35 ]). For the subsequent RT-PCR experiment, total RNA was prepared using the single-step guanidinum method according to the manufacturer's instructions (TRIZOL, Gibco BRL). First strand cDNA synthesis was carried out in 20 μl reaction using the Omniscript Reverse transcriptase (Qiagen) and the oligo(dT) primers with 2 μg of total RNA. RT-PCR was carried out in a 20 μl reaction in 1 × buffer [1.5 mM Mg2+, 0.2 mM dNTPs, 0,4 μM primers each, 1 Unit of Taq polymerase (Roche)] and 1 μl of cDNA. Amplification steps were as follows: 95°C for 90 sec; 9 cycles of 94°C for 20 sec, 64°C for 10 sec decreasing the annealing temp for 1°C with each cycle (touchdown), 72°C for 20 sec; followed by 30 cycles of 94°C for 20 sec, 55°C for 10 sec, 72°C for 20 sec, followed by an extension at 72°C for 10 min. For the refined PCR, the amplification step was repeated with identical PCR conditions but with 2 μl of PCR product instead of 1 μl of cDNA. All PCR products were resolved on 2% agarose gels run at 90 V/20 cm for 1.5 h in TAE buffer. Gels were then manually examined for exact size, genomic contamination and the tissues in which the bands are observed. As a control, a fraction of variants were sequenced using the ABI Prism BigDye Terminators and the ABI Prism 3100 sequencer (Applied Biosystems). Authors' contributions SG wrote the prediction software as well as designed PCR primers for experimental analysis. SH and MV provided guidance for the computational work. DZ performed the RT-PCR experiments with the guidance of BK. Supplementary Material Additional File 1 List of tissues for which tissue specific transcripts are predicted. This is a text file with a listing of all tissues for which specific trascripts exist along with the number of ESTs related to individual tissues. Furthermore, the ESTs derived from normalized libraries and the specific variants predicted via such ESTs are also listed. Click here for file Additional File 3 Detailed description of RT-PCR experiments This is an excel file containing the primer sequences used for RT-PCR experiments along with detailed comments on the gel pictures subsequently obtained. Click here for file Additional File 2 RT-PCR picture (jpeg file) showing the expression pattern of gene PRAME This gene shows specific expression for several tumor types, along with testis as the only normal tissue. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521684.xml
314464	Cell-Passage Activity Is Required for the Malarial Parasite to Cross the Liver Sinusoidal Cell Layer	Liver infection is an obligatory step in malarial transmission, but it remains unclear how the sporozoites gain access to the hepatocytes, which are separated from the circulatory system by the liver sinusoidal cell layer. We found that a novel microneme protein, named sporozoite microneme protein essential for cell traversal (SPECT), is produced by the liver-infective sporozoite of the rodent malaria parasite, Plasmodium berghei . Targeted disruption of the spect gene greatly reduced sporozoite infectivity to the liver. In vitro cell invasion assays revealed that these disruptants can infect hepatocytes normally but completely lack their cell passage ability. Their apparent liver infectivity was, however, restored by depletion of Kupffer cells, hepatic macrophages included in the sinusoidal cell layer. These results show that malarial sporozoites access hepatocytes through the liver sinusoidal cell layer by cell traversal motility mediated by SPECT and strongly suggest that Kupffer cells are main routes for this passage. Our findings may open the way for novel malaria transmission-blocking strategies that target molecules involved in sporozoite migration to the hepatocyte.	Introduction Malaria is one of the most devastating infectious diseases in the world, killing more than 1 million people per year. Malaria is transmitted by bites of infected mosquitoes that inject sporozoites under the skin. The first obligatory step for these parasites to establish infection in humans is migration to hepatocytes, where they proliferate and develop into the erythrocyte-invasive form ( Sinnis 1996 ). This liver-invasive stage has been demonstrated as a promising target for antimalarial strategies that aim to establish sterile immunity against the malarial parasite ( Nussenzweig et al. 1967 ; Hoffman et al. 1996 ). However, the mechanisms underlying the parasite's liver infection are largely unknown. In particular, it has been controversial how sporozoites reach the hepatocytes that are separated from blood circulation by the liver sinusoidal layer. The routes the sporozoites use to cross this layer, the modes of motility on which their migration is based, and the molecules of the parasite involved in this process are poorly understood. Malarial parasites develop into sporozoites in the mosquito midgut and then invade the salivary gland, where they wait to be transferred to the mammalian host ( Menard 2000 ). Once injected by mosquito bites under the skin, sporozoites enter the blood circulation and are carried to the liver by the bloodstream ( Sinnis and Nussenzweig 1996 ; Menard 2000 ; Mota and Rodriguez 2002 ). In the liver, they are thought to be arrested on the inner surface of the liver sinusoidal vein and then leave the vein and infect the hepatocytes by crossing the sinusoidal wall ( Sinnis and Nussenzweig 1996 ). This wall is a single-cell layer mainly composed of sinusoidal endothelial cells and Kupffer cells, which are hepatic macrophages. Several models have been proposed to explain how sporozoites cross this layer. Some authors proposed that sporozoites infect hepatocytes after crossing the liver endothelial cell through fenestrations in this cell ( Vanderberg and Stewart 1990 ), but these openings are too small for sporozoites to freely pass through ( Mota and Rodriguez 2002 ). Other authors have suggested that Kupffer cells are gates for sporozoites to access hepatocytes, based on the ultrastructural observation that sporozoites were found inside Kupffer cells shortly after intravenous inoculation ( Mota and Rodriguez 2002 ). This Kupffer cell hypothesis, however, has not been convincingly demonstrated, because other tools for probing into this event were lacking. Furthermore, the observation that the sporozoites in Kupffer cells sometimes have a vacuole around them makes the conclusion uncertain. Some authors have proposed that sporozoites are passively engulfed by Kupffer cells and then carried to the hepatocyte ( Meis et al. 1983 ), and some have proposed that this migration is based on active motility accompanied by vacuole formation ( Pradel and Frevert 2001 ). The malarial parasite has no locomotory organelles such as flagella or cilia. Motility of the host-invasive stages of the malarial parasite, including the sporozoite, is dependent on secretion of micronemes that are organelles occupying the cytoplasm of the parasite ( Sultan 1999 ; Menard 2001 ). Micronemal components, which may include several attachment proteins, are secreted from the apical pore during parasite movement and are translocated backwards along the parasite cell surface by actomyosin motors of the parasite. This surface movement is believed to generate traction for parasite-invasive motility. Salivary gland sporozoites display three modes of motility in vitro dependent on secretion of micronemes ( Mota and Rodriguez 2002 ). One mode is gliding motility on a solid surface, which can be observed as circular movement on a glass slide, probably representing gliding motility on the cell surface. The other two are cell-invasive motilities: cell-infection and cell-traversal motility ( Mota et al. 2001 ; Kappe et al. 2003 ). Cell-infection motility is accompanied by vacuole formation and is followed by parasite development into exoerythrocytic forms (EEFs). Cell-traversal motility, on the other hand, involves plasma-membrane disruption and is followed by migration through the cytoplasm and eventual escape from the cell. Recently, Mota et al. (2002 ) revealed that this type of cell-invasion motility can be identified by conventional cell-wound assay. According to the observation that passage through some hepatocytes by this motility precedes hepatocyte infection, they proposed the hypothesis that this motility is necessary for sporozoites to be activated for hepatocyte infection ( Mota et al. 2002 ). However, the role of this motility in liver infection remains unclear. Aiming at identification of molecules involved in sporozoite infection, we screened an expressed sequence tag (EST) database of the salivary gland sporozoite of a rodent malarial parasite, Plasmodium berghei . In this paper, we report a novel microneme protein, named SPECT (sporozoite microneme protein essential for cell traversal), which is specifically produced by the liver-infective sporozoite and is essential for the sporozoite's cell-passage ability. By using spect -disrupted parasites, we show that cell-passage ability of the sporozoite plays a critical role in malarial transmission to the vertebrate host and is required for sporozoites to access hepatocytes by traversal of the liver sinusoidal cell layer. In addition, we provide a model of sporozoite liver infection, which suggests an answer to the question of how sporozoites reach the hepatocytes. Results Identification of cDNA Encoding SPECT from P. berghei Salivary Gland Sporozoite EST Database Sporozoites acquire the ability to infect the mammalian liver after infection of the mosquito salivary glands ( Sultan et al. 1997 ), indicating that novel protein synthesis for liver infection begins in this stage ( Matuschewski et al. 2002 ). To search for malarial genes involved in liver infection, we screened an EST database of P. berghei salivary gland sporozoites. We assembled 3,825 ESTs, obtained 502 contigs, and screened them for genes encoding secretory proteins or membrane-associated proteins, which may participate in host–parasite interactions. This screening was started from the contigs containing a high number of ESTs, since the number of ESTs may correlate with the expression levels of the respective genes. In this process, we identified a contig composed of ten ESTs, encoding a putative secretory protein of 241 amino acids ( Figure 1 A). The expected molecular mass for the N-terminal signal sequence-processed form of this protein was 25 kDa. We named this protein SPECT (sporozoite microneme protein essential for cell traversal), since it is essential for sporozoite passage through a host cell, as described later. Figure 1 Sequence Analysis of spect cDNA (A) Nucleotide sequence of spect cDNA (top lane) and the deduced amino acid sequence (bottom lane) are shown. The predicted N-terminal signal sequence is underlined. The numbers indicate positions of the nucleotides starting from the 5′ end. The asterisks indicate the termination codon. (B) A comparison of deduced amino acid sequences of P. berghei spect (top) and P. falciparum spect (bottom). Gaps are introduced to obtain optical matching by using GENETIX-MAC software. Asterisks or dots show conserved or similar residues, respectively. The amino acid numbers from the first Met residue are shown on the left of each line. Southern blot analysis showed that the spect gene is a single-copy gene (data not shown). Sequence analysis of the spect gene identified four introns (data not shown). A computer search of Plasmodium genome databases ( Carlton et al. 2002 ; Gardner et al. 2002 ) revealed that this gene is conserved through several Plasmodium species. The orthologous protein in P. falciparum, the clinically most important human malaria parasite, shared 45.6% sequence identity with P. berghei SPECT ( Figure 1 B). SPECT Is Produced Specifically by Salivary Gland Sporozoites and Localized in Micronemes The expression profile of this gene in the malarial life cycle was investigated. Immunofluorescent analysis in all host-invasive stages showed that SPECT production was restricted to sporozoites in the salivary gland ( Figure 2 A). It is noteworthy that SPECT is not detected in sporozoites in the midgut, because this expression profile strongly suggests that SPECT is specifically involved in liver infection. Western blot analysis revealed SPECT as a 22 kDa protein in salivary gland sporozoites, but not in midgut sporozoites ( Figure 2 B), confirming that SPECT is produced after invasion into the salivary gland. Immunoelectron microscopy showed that SPECT is localized in the sporozoite to micronemes that are secretory organelles occupying the cytoplasm ( Figure 2 C). Micronemes are common to motile stages of Plasmodium parasites and play a central role in host-invasive motility ( Sultan 1999 ; Menard 2001 ). Taken together, these results indicate that SPECT plays a role in the liver-invasive motility of the sporozoite. Figure 2 SPECT Is a Microneme Protein Specifically Produced in the Liver-Infective Sporozoite Stage (A) Indirect immunofluorescence microscopy of all four invasive forms of the malarial parasite (indicated over the panel). Parasites were stained with primary antibodies against SPECT, followed by FITC-conjugated secondary antibodies. SPECT was detected only in the salivary gland sporozoite, the liver-infective stage. The corresponding phase-contrast or DAPI-stained image (Phase or DAPI) is shown under each image. Scale bars, 5 μm (B) Western blot analysis of SPECT production in the midgut sporozoite (M) and the salivary gland sporozoite (S). Lysate of 500,000 sporozoites was loaded onto each lane and detected with the same antibody used in (A). SPECT was detected as a single band of 22 kDa (arrowhead) only in the salivary gland sporozoite. (C) Immunoelectron microscopy of sporozoites in the salivary gland. Ultrathin sections of a mosquito salivary gland infected with sporozoites were incubated with the same antibody used in (A) followed by secondary antibodies conjugated with gold particles (15 nm). Particles were localized to micronemes (Mn) but not to rhoptories (Rh). Axial (inset) and vertical images are shown. Scale bars, 0.5 μm. SPECT Plays an Important Role in Sporozoite Infection of the Host Liver To investigate the function of SPECT protein, we generated spect -disrupted parasites by homologous recombination ( Figure 3 A). The spect disruptants were selected by the antimalarial drug pyrimethamine and were separated from wild-type parasites by limiting dilution. Disruption of the spect locus was confirmed by Southern blot analysis ( Figure 3 B). To exclude the possibility that the spect -disrupted populations obtained were derived from a single clone, two independently obtained spect -disrupted populations ( spect (−)1 and spect (−)2) were used in the following experiments. Figure 3 Targeted Disruption of the spect Gene (A) Schematic representation of targeted disruption of the spect gene. The targeting vector (top) containing a selectable marker gene is integrated into the spect gene locus (middle) by double crossover. This recombination event resulted in the disruption of the spect gene and confers pyrimethamine resistance to disruptants (bottom). (B) Genomic Southern blot hybridization of wild-type (WT) and spect (−) populations. Genomic DNA isolated from the respective parasite populations was digested with EcoT22I and hybridized with the probe indicated in (A) by a solid bar. By integration of the targeting construct, the size of detected fragments was decreased from 1.9 kbp to 1.2 kbp. The result is shown for two independently prepared populations, spect (−)1 and spect (−)2. (C) Immunofluorescence microscopy of the wild-type (WT) and spect (−) parasite. Sporozoites were collected from the salivary gland and stained with primary antibody against SPECT followed by FITC-conjugated secondary antibodies. The apical end of each sporozoite is indicated by an arrowhead. In the intra-erythrocytic stage, SPECT gene disruption did not affect parasite proliferation, as the growth rates in rat blood were almost the same in the spect -disrupted and wild-type populations (data not shown). Furthermore, disruption of the gene did not affect parasite development in the mosquito vector, as numbers of sporozoites residing in the midgut and in the salivary glands were similar in the spect -disrupted and wild-type populations ( Table 1 ). Table 1 SPECT Disrupted Parasites Develop Normally into Sporozoites and Invade the Salivary Gland in the Mosquito Vector Mosquitoes were fed on mice infected with spect (−) parasite populations or wild-type polyclonal populations. Sporozoites were collected separately from the midgut and the salivary glands of mosquitoes 24–28 d after feeding and then counted. Each value is the mean of the number with its standard error from three independent experiments Next, the liver infectivity of the spect -disrupted sporozoites was examined. Rats were intravenously inoculated with sporozoites, and the progress of parasitemia, the percentage of infected erythrocytes, was measured in the exponential growth period (from 3.5 d to 5 d after inoculation). It is thought that the parasitemias reflect the liver infectivity of the respective parasite populations, since the growth rates of their intraerythrocytic stages are similar (shown by the parallel slopes of the increase in parasitemia in Figure 4 ). Based on the average parasitemia at 3.5 d after inoculation of 30,000 sporozoites, the liver infectivities of the two disruptant strains were estimated to be 15- and 28-fold lower, respectively, than that of the wild-type. These results are consistent with the observation that the parasitemias after injection of 30,000 disruptant sporozoites were lower than that from 3,000 wild-type sporozoites. Figure 4 Targeted Disruption of spect Results in Reduction of Sporozoite Infectivity to the Liver (A) The salivary gland sporozoites of each parasite population were injected intravenously into five rats. The parasitemia of each rat was checked by a Giemsa-stained blood smear after inoculation on the days indicated. The average parasitemia after inoculation of 30,000 sporozoites was significantly low in disruptant populations, whereas their growth rates in the blood were essentially the same as the wild-type. The numbers of parasites inoculated were as follows: 30,000 spect (−)1 (open circles), 30,000 spect (−)2 (open triangles), 30,000 wild-type (filled circles), and 3,000 wild-type (filled squares). Values shown represent the mean parasitemia (± SEM) of five rats. (B) The salivary gland sporozoites (500,000) of wild-type or spect -disrupted parasites were inoculated intravenously into 3-wk-old rats. After 24 h, the livers were fixed with paraformaldehyde and frozen. The number of EEFs on each cryostat sections was estimated by indirect immunofluorescence analysis using anti-CS antiserum. Values shown represent the mean number of EEFs per square millimeter (± SEM) of at least three rats. The liver infectivity was also evaluated by the number of early EEFs. Frozen sections of the rat liver was prepared 24 h after sporozoite injection and EEFs were counted by immunofluorescence microscopy. As shown in Figure 4 B, EEFs were approximately 30-fold decreased by spect gene disruption. This reduction rate agrees well with that estimated by parasitemia. These results indicate that SPECT plays a role in the process of sporozoite invasion into the liver. SPECT Is Essential for Sporozoite Cell-Passage Ability Localization of SPECT in micronemes indicates its involvement in the invasive motility of the sporozoite. The motility of spect -disrupted sporozoites was investigated by three in vitro assays corresponding to three modes of motility of the sporozoite. First, we checked gliding motility on a solid surface, which is essential for sporozoite infectivity. Most disruptants displayed a typical circular movement, and the proportion of motile sporozoites was almost identical in disruptant and wild-type parasites (63.6% and 67.5%, respectively), showing that their gliding motility is not affected by SPECT gene disruption. Second, we examined the ability of the sporozoites to infect hepatocytes. This was assayed by formation of EEFs in a human hepatoma cell line, HepG2 ( Hollingdale et al. 1981 ). As shown in Figure 5 A, the disruptants formed EEFs in similar numbers to the wild-type, indicating that they retain normal infectivity to the hepatocyte. Third, we examined cell-traversal ability that takes place prior to hepatocyte infection. This was estimated by the number of membrane-wounded cultured cells that were labeled by uptake of fluorescein isothiocyanate (FITC)-conjugated dextran from the medium ( Mota et al. 2001 ). As shown in Figure 5 B, the cell-wound assay using HeLa cells showed that the disruptants lost their cell-passage activity completely. The same results were obtained in HepG2 cells (data not shown). These results revealed that SPECT is specifically involved in cell-traversal ability and suggest that lack of this ability reduced liver infectivity of the disruptants. Figure 5 spect Disruption Results in Loss of Cell-Passage Activity of the Sporozoite (A) spect disruption does not affect sporozoite ability to infect hepatocytes. (Top panel) Comparison of EEF numbers between disruptants ( spect (−)) and wild-type (WT) parasites. Salivary gland sporozoites were added to HepG2 cells and cultured for 48 h. EEFs formed were counted after immunofluorescence staining with an antiserum against CS protein. (Bottom panels) Representative fluorescence stained images. (B) Sporozoites lacking SPECT cannot traverse HeLa cells. (Top) Comparison of cell-passage activity between disruptants and wild-type parasites. Salivary gland sporozoites were added to HeLa cells and incubated for 1 h with FITC-conjugated dextran (1 mg/ml). Cell-passage activity was estimated by the number of cells wounded by sporozoite passage, which were identified by cytosolic labeling with FITC-conjugated dextran. (Bottom) Representative fluorescence stained images. All data are mean numbers of EEFs or FITC-positive cells in a one-fifth area of an 8-well chamber slide with standard errors for at least three independent experiments. Cell Passage Ability Is Necessary for Sporozoites to Traverse the Sinusoidal Layer Cells and to Access Hepatocytes To access the hepatocytes, sporozoites must cross the sinusoidal layer, which separates them from the circulation. We assumed that SPECT was necessary for this process. Since Kupffer cells are major components of this layer and have been reported as the main gates for sporozoite access to the hepatocyte, we prepared Kupffer cell-depleted rats by intravenous injection of liposome-encapsulated dichloromethylene diphosphonate (Cl 2 MDP) ( Vreden et al. 1993 ; van Rooijen and Sanders 1994 ) and tested them for infection by disruptant and wild-type sporozoites. As shown in Figure 6 A, infectivities of spect -disruptants assessed by parasitemia were increased by 22- and 53-fold by Kupffer cell depletion and, as a result, became equal to that of the wild-type. The numbers of early EEFs detected in the liver sections were also almost identical in wild-type and spect -disrupted parasites ( Figure 6 B). These results show that the disruptants' loss of infectivity is localized at the sinusoidal cell layer and that the cell-passage ability of the sporozoite is necessary to cross this layer and, specifically, the Kupffer cells. Figure 6 Restoration of spect (−) Sporozoite Infectivity in Kupffer Cell-Depleted Rats (A) Liposome-encapsulated Cl 2 MDP (filled points) or PBS (open) was injected intravenously into rats. After 48 h, 30,000 sporozoites of spect (−)1 (circles), spect (−)2 (triangles), or wild-type (squares) populations were inoculated intravenously. Parasitemia of each rat was checked by Giemsa-stained blood smears after inoculation on the days indicated. Values shown represent the mean parasitemia (± SEM) of five rats. (B) Salivary gland sporozoites (500,000) of each parasite population were inoculated intravenously into Kupffer cell-depleted rats. After 24 h, the livers were fixed with paraformaldehyde and frozen. The number of EEFs on each cryostat section was estimated by indirect immunofluorescence analysis using anti-CS antiserum. Values shown represent the mean number of EEFs per square millimeter (± SEM) of at least three rats. Discussion It has been reported that the Plasmodium sporozoite has the ability to traverse cultured cells rapidly ( Mota et al. 2001 ), but the role of this process in liver infection has remained unclear. On the other hand, it is poorly understood how the sporozoite migrates from the circulatory system to the hepatocyte. In this paper, we address these issues using a gene-targeting technique. We have shown that the cell-traversal activity of the sporozoite is necessary for it to leave the circulatory system by crossing the liver sinusoidal cell layer. These results are the first to reveal the role of cell-traversal activity in malarial transmission. In vitro cell invasion assays showed that spect -disrupted sporozoites completely lose cell passage activity, but preserve normal infectivity to the hepatocyte (see Figure 5 ). These results clearly demonstrated that these two cell-invasion activities are independent of each other. This conclusion contradicts the hypothesis proposed by Mota et al. (2002 ) that cell passage activates the sporozoite for hepatocyte infection. They assumed that sporozoites traverse some hepatocytes before infecting a hepatocyte and that this passage alters their mode of cell invasion from passage to infection ( Mota et al. 2002 ). Our results, however, demonstrated that lack of previous cell passage has no influence on the infectivity to hepatocytes. This independence was confirmed in vivo by the result that disruptants and wild-type showed the same liver infectivities in Kupffer cell-depleted rats (see Figure 6 ). Therefore, sporozoites may change their mode of invasive motility according to other factors, which remain to be elucidated. We suppose that secretion of the micronemal contents during gliding on the cell surface might be one such factor, since this motility may precede hepatocyte infection as discussed below. Our results indicate that the liver sinusoidal barrier is not perfect, since a small proportion of the spect -disrupted sporozoites can infect the liver (see Figure 4 ). It is supposed that this layer may have a few openings and the disruptants can migrate through them by gliding along the epithelial cell surface. In Kupffer cell-depleted rats, on the other hand, both disruptants and wild-type may migrate through the numerous gaps created among the endothelial cells, resulting in elimination of the phenotypic difference. Since Kupffer cells constitute approximately 30% of the sinusoidal cells ( Bouwens et al. 1986 ), their depletion from this layer may leave many gaps that cannot readily be repaired. Supposedly, sporozoites cross these gaps in the same way as they migrate through the few gaps in normal rats. Experiments using Kupffer cell-depleted rats indicate that Kupffer cells are not involved in sporozoites targeting the liver, because the depletion did not reduce the susceptibility of rats to sporozoite infection. Thus, sporozoites seem to be first arrested on the endothelial cell surface or on the glycosaminoglycans extending through endothelial fenestrations and then migrate to Kupffer cells ( Cerami et al. 1992 ; Pradel et al. 2002 ). If so, gliding motility on the cell surface would be necessary for the sporozoite to migrate from initial attachment sites to Kupffer cells (or to gaps) along the inner surface of the sinusoidal layer as well as for the sporozoite to migrate through gaps. These assumptions imply that after Kupffer cell depletion, sporozoites can arrive at the hepatocyte by gliding motility alone, in accord with the observation that the disruptants can infect Kupffer cell-depleted rats with the same infectivity as the wild-type. Our results strongly suggest that Kupffer cells are main gates for sporozoites to access hepatocytes. Previous electron microscopic studies have reported that sporozoites are observed in Kupffer cells after intravenous inoculation, and some of them are found within vacuoles ( Meis et al. 1983 ; Pradel and Frevert 2001 ). Based on this observation, it has been speculated that sporozoites invade the Kupffer cell by a motility distinct from passage that does not involve parasitophorous vacuole formation. Our results, on the contrary, indicate that sporozoites cross the layer by the same cell-passage motility as observed in vitro. We think this discrepancy indicates the following two possibilities. One is that the vacuole formed in the Kupffer cell after rupture of its cell membrane is different from the parasitophorous vacuole formed in the hepatocyte, although their differences cannot be distinguished by electron microscopy. Another possibility is that the parasites seen in vacuoles were phagocytosed ones and not in the process of invasion. In fact, if their invasion mode is cell-traversal motility, as we believe, this event may be rapidly completed and difficult to catch by electron microscopy. Therefore, many phagocytosed parasites could be included among those seen. Taking the evidence together, we propose that the sporozoites access the hepatocyte through Kupffer cells by the same cell-traversal motility that has been identified in vitro, and we propose a model for sporozoite liver infection in Figure 7 . Figure 7 Schematic Representation of Sporozoite Migration to and Infection of Hepatocytes (Left) Sporozoites migrate to the space of Disse through the Kupffer cells. [1] The sporozoite (Sp) in the circulatory system is sequestered to the sinusoidal endothelial cell (EC) by specific recognition of the cell surface or glycosaminoglycans extending from the hepatocytes (He) through fenestration. [2] The sporozoite begins to glide on the epithelial cell surface. [3] Encountering a Kupffer cell (KC), the sporozoite ruptures the plasma membrane, passes through the cell, and enters into the space of Disse. Thus, the sporozoite gains access to hepatocytes. This step requires SPECT. [4] The sporozoite infects a hepatocyte with formation of a vacuole and develops into EEF in the hepatocyte. (Right) An alternative route to the hepatocyte. A small number of sporozoites, which find gaps in the sinusoidal layer while gliding, migrate to hepatocytes directly through the openings without need for cell passage and infect the hepatocytes. Likewise, in Kupffer cell-depleted rats, both wild-type and spect (−) sporozoites can enter hepatocytes through numerous gaps present between the sinusoidal endothelial cells. In this study we have established the significance of cell-passage ability of the sporozoite in malaria transmission and have demonstrated that this ability is necessary for breaking through the liver sinusoidal barrier. Cell-traversal activity plays an important role in other invasive stages of the malarial parasite, including the ookinete, which migrates through the epithelial cells of the mosquito midgut, and the sporozoite in the oocyst, which is released from the mature oocyst and then migrates through the salivary gland cell. Our study revealed that another cellular barrier is present in the malarial life cycle and sporozoites must break through this barrier by cell-traversal activity. Our recent work has identified two other genes that are involved in the cell passage activity of the sporozoite. Like SPECT, the products of these genes have a secretory protein-like structure and are localized in the micronemes. Furthermore, sporozoites disrupted for these genes have similar phenotypic character to spect -disrupted ones, including impaired cell-passage ability, decreased liver infectivity with similar reduction rate, complete restoration of the infectivity in Kupffer cell-depleted rats, normal gliding motility, and normal hepatocyte infectivity (unpublished data). This suggests that the cell-traversal ability of the sporozoite is realized by cooperation of several microneme proteins. We suggest that these molecules could be targets for antimalarial strategies, since success in crossing this layer is critical for the malarial parasite to establish infection in humans. Elucidation of the molecular mechanisms of passage may lead to novel malaria transmission-blocking strategies that prevent sporozoites from gaining access to the hepatocyte. Materials and Methods Parasite preparations Female 6–10-wk-old BALB/c mice (Japan SLC, Inc., Hamamatsu, Japan) infected with the P. berghei ANKA strain were prepared by peritoneal injection of infected blood that was stored at −70°C. For the purification of sporozoites, infected mosquitoes were dissected 24–28 d after the infective blood meal. The salivary glands and midgut were separately collected in medium 199 on ice and then gently ground to release the sporozoites. Ookinetes and erythrocytic-stage parasites were prepared as described previously ( Yuda et al. 1999 ; Kariu et al. 2002 ). Genomic Southern blot hybridization Genomic DNA of P. berghei parasites (2 μg) was digested with ClaI, EcoRI, EcoT22I, HindIII, or XbaI, separated on 1.2% agarose gel and then transferred to a nylon membrane. DNA fragments were amplified by PCR using genomic DNA as template with the following primers: 5′-TGGGCAATTTTGCCTTTAAAAACG-3′ and 5′-TTTCATTGTGTTAAACGATAAGTG-3′. They were labeled with [ 32 P]dCTP and used as probes. Antibody preparation and Western blot analysis Recombinant SPECT without signal sequence was expressed as a glutathione S-transferase (GST)–fusion protein using the pGEX 6p-1 system (Amersham Bioscience, Uppsala, Sweden). The recombinant protein was purified with a GST column and used for immunization of rabbits. Specific antibodies were affinity purified using a N-hydroxysuccinimide-activated column (Amersham Bioscience) coupled with recombinant SPECT protein. For CS antiserum production, the peptide DPPPPNANDPAPPNAN, corresponding to the repeat region, was conjugated to keyhole limpet hemocyanin as a carrier and used for the immunization of rabbits. Western blot analysis was performed as described previously ( Kariu et al. 2002 ). Immunofluorescence microscopy and immunoelectron microscopy Immunofluorescence microscopy was performed as described previously ( Kariu et al. 2002 ). Purified parasites were fixed in acetone for 2 min. The slides were incubated with anti-SPECT rabbit antibodies and then with FITC-conjugated secondary antibody (Zymed. South San Francisco, California, United States). For nuclear staining, 4′,6-diamidino-2-phenylindole (DAPI) (0.02 μg/ml final concentration) was added to the secondary antibody solution. Immunoelectron microscopy was performed as described previously ( Yuda et al. 2001 ). In brief, purified parasites were fixed in 1% paraformaldehyde–0.1% glutaraldehyde for 15 min on ice. After embedding in LR Gold resin (London Resin Company Ltd., London, United Kingdom), ultrathin sections were incubated with anti-SPECT antibodies and then with secondary antibody conjugated to gold particles (15 nm diameter) (AuroProbe, Amersham Pharmacia Biotech, Uppsala, Sweden). The samples were examined with a Hitachi H-800 transmission electron microscope (Hitachi, Tokyo, Japan) at an acceleration voltage of 100 kV. Targeted disruption of the spect gene For construction of the targeting vector, two fragments of the spect gene were amplified by PCR using genomic DNA as template with the primer pairs 5′-CGCGAGCTCGCAATATGGTATTAAATTTTGGGCTAGCCA-3′ and 5′-CGCGGATCCGGTATTTTCATTGTGTTAAACGATATGTGA-3′ and 5′-CCGCTCGAGGTCCTATTTATCATTTTAAAATGTGTTTTATC-3′ and 5′-CGGGGTACCAATCGTCATAAATAGGAGTTATGAAGT-3′. These fragments were cloned into either side of the selectable marker gene in pBluescript (Strategene, La Jolla, California, United States). The gene targeting experiment was performed as described previously ( Yuda et al. 1999 ). Evaluation of sporozoite infectivity to rats Sporozoites collected from mosquito salivary glands were suspended in medium 199 and then injected intravenously into 3-wk-old female Wistar rats (Japan SLC, Inc., Hamamatsu, Japan) ( n = 5). Before each inoculation, sporozoites were checked for their ability to glide in vitro to confirm that they contained over 60% motile sporozoites. Parasitemia was checked every 12 h by a Giemsa-stained blood smear. Measurement of the number of EEFs in the infected liver Sporozoites (5.0 × 10 5 ) were intravenously inoculated into a 3-wk-old female Wistar rat. After 24 h, the liver was perfused with PBS followed by 4% paraformaldehyde. The liver was further fixed in 4% paraformaldehyde for 6 h and frozen in liquid nitrogen. Cryostat sections (20 μm) were prepared from the left lobe and fixed in acetone for 2 min on a glass slide. The EEFs were detected by immunofluorescence staining using rabbit anti-CS antiserum and FITC-conjugated secondary antibody. At least 12 sections were examined under an Olympus (Tokyo, Japan) BX60 fluorescence microscope (200×) and the number of EEFs per square millimeter was calculated. EEF development assay in vitro The EEF formation assay was performed as described previously ( Hollingdale et al. 1981 ) with minor modifications. HepG2 cells (5.0 × 10 5 ) were plated in 8-well chamber slides. Sporozoites (5.0 × 10 3 or 5.0 × 10 4 ) were suspended in 100 μl of complete medium and added to this culture. After 2 h, the media were replaced with 400 μl of fresh complete medium supplemented with 3 μg/ml glucose. The slides were incubated for 2 d with medium changed twice a day and were fixed in acetone for 2 min. The EEFs were detected by immunofluorescence staining as described above. The number of EEFs in one-fifth of the area of each well was counted under an Olympus BX60 fluorescence microscope (200×). Cell-traversing activity assay The traversing activity of the sporozoite was examined using a standard cell-wounding and membrane repair assay ( Mota et al. 2001 ). HepG2 cells (2.5 × 10 5 ) or HeLa cells (5.0 × 10 4 ) were inoculated into 8-well chamber slides (Nunc Inc., Napierville, Illinois, United States). Sporozoites were added 2 d later to cells for 1 h in the presence of 1 mg/ml FITC-labeled dextran (10,000 MW, lysine-fixable; Molecular Probes, Inc., Eugene, Oregon, United States). The cells were incubated for an additional 3 h in complete culture medium and fixed with 4% paraformaldehyde in PBS. The number of FITC-positive cells was counted under a fluorescence microscope. Depletion of rat Kupffer cells For depletion of Kupffer cells, 3-wk-old female Wistar rats were injected intravenously with 120 μl of liposome-encapsulated Cl 2 MDP or an equal volume of PBS as control. After 48 h, sporozoites were injected into a tail vein and the parasitemia was checked by Giemsa-stained blood smears. Cl 2 MDP liposomes were prepared as described elsewhere ( van Rooijen and Sanders 1994 ). Elimination of Kupffer cells was confirmed by immunoperoxidase staining after liver perfusion with PBS followed by fixation with 4% paraformaldehyde in PBS. Cl 2 MDP was a gift from Roche Diagnostics (Mannheim, Germany).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC314464.xml
516028	Cross-species global and subset gene expression profiling identifies genes involved in prostate cancer response to selenium	Background Gene expression technologies have the ability to generate vast amounts of data, yet there often resides only limited resources for subsequent validation studies. This necessitates the ability to perform sorting and prioritization of the output data. Previously described methodologies have used functional pathways or transcriptional regulatory grouping to sort genes for further study. In this paper we demonstrate a comparative genomics based method to leverage data from animal models to prioritize genes for validation. This approach allows one to develop a disease-based focus for the prioritization of gene data, a process that is essential for systems that lack significant functional pathway data yet have defined animal models. This method is made possible through the use of highly controlled spotted cDNA slide production and the use of comparative bioinformatics databases without the use of cross-species slide hybridizations. Results Using gene expression profiling we have demonstrated a similar whole transcriptome gene expression patterns in prostate cancer cells from human and rat prostate cancer cell lines both at baseline expression levels and after treatment with physiologic concentrations of the proposed chemopreventive agent Selenium. Using both the human PC3 and rat PAII prostate cancer cell lines have gone on to identify a subset of one hundred and fifty-four genes that demonstrate a similar level of differential expression to Selenium treatment in both species. Further analysis and data mining for two genes, the Insulin like Growth Factor Binding protein 3, and Retinoic X Receptor alpha, demonstrates an association with prostate cancer, functional pathway links, and protein-protein interactions that make these genes prime candidates for explaining the mechanism of Selenium's chemopreventive effect in prostate cancer. These genes are subsequently validated by western blots showing Selenium based induction and using tissue microarrays to demonstrate a significant association between downregulated protein expression and tumorigenesis, a process that is the reverse of what is seen in the presence of Selenium. Conclusions Thus the outlined process demonstrates similar baseline and selenium induced gene expression profiles between rat and human prostate cancers, and provides a method for identifying testable functional pathways for the action of Selenium's chemopreventive properties in prostate cancer.	Background Gene expression profiling, along with other methods to evaluate the global changes in genomes, provides the opportunity to understand whole scale changes present in human biology. Yet the sheer mass of data presented by these techniques often makes subsequent analysis difficult. Techniques such as gene expression profiling may result in the identification of hundreds if not thousands of differentially expressed genes that may be associated with the biological process, but may also represent noise related to the biological and technical variation. In an economic environment where limited resources are available for the follow-up and validation of potential target genes methods must be provided for the prioritization and sorting of data. Previous methods have relied heavily on the mapping of metabolic pathways or transcription factor binding sites [ 1 - 5 ]. These processes rely on the premise that the metabolic pathways associated with a given disease are well delineated, or that groups of proteins with very similar structural or functional design are involved in the disease process. In situations where these assumptions may not be true, alternative methods for the sorting of the data are needed. Here we demonstrate an alternative approach using comparative genomics and animal models of human prostate cancer to sort and identify genes involved in the response of prostate cancer cells to the proposed chemopreventive agent Selenium [ 6 , 7 ]. This process takes advantage of the continued sequencing of multiple animal genomes and the ability to produce gene expression profiles in multiple species. Through the use of these techniques one can leverage established animal models to identify genes associated with human disease processes, as is demonstrated here with the identification of Insulin-like growth factor-2 Binding protein 3 (IGFBP3) and retinoid-X-receptor alpha (RXRalpha). Results Generation of common genes and homologs Sequence validated gene libraries for both the rat and human DNAs were obtained from Research Genetics (Huntsville, AL), and were supplemented with additional DNA samples obtained from the University of Iowa rat clone sequencing program [ 8 ]. The majority of the rat DNAs, and a subset of the human DNAs were resequenced by Dr. J. Quackenbush at TIGR through a joint Program in Genomic Applications consortium. The GeneBank accession numbers for the 19,200 individual human or rat clones present in the recent slide printings were used to query the NCBI Unigene database to return the associated Unigene IDs. Unigene IDs were returned for virtually all identified clones, and were placed in an Oracle database where they were compared with the downloaded NCBI Homologene dataset (build 106) of rat, mouse, and human homologues. Of the 19,200 clones, 5740 genes were identified with homologues present on both the rat and human slides. This homologue set was used for the subsequent comparisons across species. Similar global and prostate gene expression profiles between rat and human prostate cancer cell lines We have sought to compare the rat and human prostate cancer transcriptomes in an effort to judge the degree of similarity between the two cell types. Because the use of differentially expressed genes would bias the comparison by eliminating the majority of genes that do not show any difference, we used the absolute level of expression for each gene and compared the rat and human genes for significant differences in absolute expression levels. In order to derive the absolute level of expression for individual genes in human or rat prostate cancer cells we used expression values derived from the associated self-self hybridizations performed for each cell line. The experiments were facilitated by the use of slides that have been quality controlled for the quantity of spotted target DNA through the use of a FITC label third dye [ 9 ]. These slides were subsequently imaged for FITC fluorescence and sorted based on the similar amounts of target DNA present on each slide [ 10 ]. Using the third dye quality control correlation coefficients of greater than 0.80 are routinely achieved between slide replicates [ 9 ]. In this manner comparisons of bound hybridized probe can be made across slides with a degree of confidence. RNA samples from cells were harvested, labeled, and homotypically hybridized to establish the baseline level of consistency within the hybridizations. Performing slice analysis on the normalized homotypic gene expression data across all the self-self hybridization slides within a species and retaining genes that demonstrated consistent expression patterns within two standard deviations of the mean expression value was performed to remove a degree of error from the technical replicates. Using the third dye as a baseline for comparison, these common expressed genes were then broken down into their component Cy3 or Cy5 expression vectors and used to build the transcriptomes for each gene using their absolute expression values. These transcriptomes were then used to compare expression values between the rat and human cell lines. These genes were annotated and gene homologues identified from the NCBI Homologene[ 11 ] dataset of rat-human homologues. Thus from a dataset of 5740 homologues, 2883 genes were found that were present within this experimental dataset and expressed in both the rat and human prostate cancer cell lines, and thus could be used for comparative genomics. These samples were processed using the Multiexperiment Viewer mircoarray statistical analysis and visualization program developed by TIGR [ 12 ]. Files were loaded and visualized for comparison across the 2883 common expressed genes in a self-organizing tree algorhythm [ 13 ] (figure 1 ) and analyzed for similarities in global expression patterns. The hierarchical clustering in self-organizing trees failed to demonstrate a pattern of clustering between species. T-test analysis [ 12 , 14 ] between the human and rat cell lines identified 58 genes (2%) which demonstrated significantly different expression patterns between species (p < 0.01 with Bonferroni correction). Thus in these comparisons, 2826 genes, or 98% of the genes examined, failed to demonstrate a statistically significant difference in expression between the human and rat prostate cancer cell lines. Using principle components analysis (figure 2 , [ 12 ]) these studies can be visualized, and demonstrate while there is some clustering of the rat and human prostate cancer cell lines, the differences are not significant. Thus when comparing gene expression patterns in rat and human cell lines one will detect significant species-specific differences in expression in 1 out of every 50 genes, with the majority of the genes demonstrating similar expression patterns. Figure 1 Gene expression profiles for human and rat prostate cancer cells. Clustering of the expressed genes in the human (LNCAP, DU145, PRO4, LN4, and PC3 derivatives) and rat (AT3, MatLyLu, and PAIII) prostate cancer cell lines based on the common homologs as defined within to NCBI Homologene database. Raw data files are available for review from the corresponding author. Figure 2 Principal Components Analysis of Rat and Human Prostate Cancer Cell Lines. There is a clustering of the human (Pro4-purple, LN4-dark-blue, PC3S-light blue, PC3US-yellow) and rat (MatLyLu-red, AT3-magenta, PAIII-green) prostate cancer cell lines in the same quadrant. The degree of separation within the quadrant was not significant by T-testing. Each sample is presented in duplicate based on independent Cy3 and Cy5 vector profiles. The presence of a large quantity of genes whose expression may be related to general cellular functions, as opposed to prostate specific metabolism, could infuse a significant amount of homogeneity to the data. In the presence of such homogeneity it may be impossible to identify the true differences that are related to prostate cellular function, and thus the perceived similarities may be artifactual. To address this issue we sought to repeat the analysis using only prostate related genes. To generate a list of such genes we used cDNAs in eight normal human prostate cDNA libraries present in the NCI Cancer Genome Anatomy Project [ 15 ]. Generation of a list of common genes proved impossible, as the combination of more than four of the cDNA lists resulted in the number of common genes being reduced to zero. A similar result was obtained when one attempted to generate a list of commonly expressed genes across multiple different cancer cDNA libraries. As an alternative approach we developed a list of 12,008 expressed genes were identified based on their presence in at least one of the eight normal human prostate cDNA libraries. The human Unigene IDs for each of the expressed genes were then used to identify the associated rat homologues from Homologene [ 11 ] and yielded 2,269 homologous rat genes (18.9%), of which 1,319 (58.1%) had associated prostate cancer gene expression data. These 1,319 prostate expressed genes were then used to repeat the comparative genomics. Similar visual and clustering results were identified for the prostate transcriptomes. T-test analysis [ 12 , 14 ] between the human and rat cell lines identified 30 prostate expressed genes (2%) which demonstrated significant differential expression between species (p < 0.01 with Bonferroni correction, while 1,289 genes (98%) failed to demonstrate a significant difference in expression across species. Thus even when only prostate expressed genes are considered, similar results were obtained. Between the rat and human prostate cancer cell lines the patterns of expression are similar for 49 of 50 genes examined. Comparison of global and prostate specific differential gene expression profiles between rat and human prostate cancer cell lines treated with selenium While global gene expression profiles appear to be similar between rat and human prostate cancer cell lines one wonders whether the response to specific physiologic stimuli may elicit similar transcriptional changes. If so, one may be able to infer a degree of homology in their biological response to the stimuli. This has already been observed on a physiological level for the rat models of prostate cancer. For example, rat and human prostate cancers respond very similarly to chemotheraputic and environmental agents including hormonal agents (both respond), cyclophosphamide (neither respond), high fat diets (increased incidence), and soy isoflavones (decreased incidence) [ 16 - 22 ]. In an effort to evaluate these similar biological responses we have compared the transcriptomes between rat and human prostate cancer cell lines treated with the proposed prostate cancer chemopreventive agent Selenium. Samples from the human PC3 and rat PA-III cell lines were treated with Selenium and examined for differential gene expression profiling. These two cell lines were chosen based on their similar biologic characteristics, as both cell lines were derived from androgen independent metastatic tumors, and thus represent tumors with similar biologic potential [ 23 , 24 ]. The cells were treated with twenty-five micromolar Selenium for either 6 hours or 5 days, to identify both immediate changes in gene transcription or changes related to the long term exposure to Selenium. Due to our interest in prostate cancer we have attempted to choose a form and concentration of Selenium that would be reflected in the ongoing prevention trials such as the SELECT prostate cancer prevention trial [ 25 , 26 ]. In this trial patients receive Selenium in the form of Selenized baker's yeast. Previous HPLC and electrospray mass spectroscopy studies have demonstrated that 85% of the Selenium in yeast is present as selenomethionine [ 27 ]. Selenomethionine has previously been used in in-vitro studies of prostate cancer cells[ 28 , 29 ]. These studies demonstrated an inhibition of prostate cancer cell proliferation over a broad range of concentrations, while an IC50 and/or decreased expression was seen at concentrations above 70 micromolar selenomethionine. To avoid the general effects of cell inhibition or cell death while focusing on the effect of Selenium we chose a lower concentration of 25 micromolar selenomethionine. These changes, while not resulting in increased cell death, did cause decreased cell division and increased doubling time in both species (data not shown). Common rat and human homologous genes demonstrating differential expression by greater than two standard deviations were identified and included 1123 genes after 6 hours and 1053 genes after 5 days of exposure to Selenium. When the expression patterns of these genes were compared across species by T-test and principle component analysis as outlined above 713 genes (25%) were found to have statistically significant differences in expression between species (p < 0.01 with Bonferroni correction). Thus when comparing rat and human samples, while the majority of the gene expression changes are similar, in at least one in four genes (p = 0.75) one can detect significant species specific differences in expression alteration when cells are treated with Selenium. Yet similar physiologic changes (decreased cellular proliferation, increased cell death) were observed in both species. These changes represent the desired physiologic changes one would expect for the chemopreventive effects of Selenium, and could be dissected by examining the common transcriptional changes seen in both species with respect to Selenium. Combined differential expression patterns for selenium responsive genes identify common gene pathways Because some of the differences in the rat and human prostate cancer cell line transcriptomes may be related to confounding variables such as culture methods, cell passage number, or time in culture, an effort was made to focus on genes that are common, and as such may define the similar Selenium based cell proliferative changes. The subsets of 1123 and 1053 differentially expressed genes (6 hours and 5 days respectively) were analyzed for genes that demonstrate similar changes in expression with respect to Selenium across species. Of these differentially expressed genes, 291 and 309 demonstrated up-regulation in rat and human cells at 6 hours and 5 days respectively. Likewise, 261 (6 hours) and 216 (5 days) demonstrated down-regulation in the presence of Selenium. When these subsets were further analyzed to identify genes with similar levels of up or down-regulation (defined as ratio differences within 0.2 units of each other) 81 genes were identified at 6 hours and 73 at 5 days (table 1-see additional file 1 ). These genes included 40 ESTs or genes with limited associated data, and 90 defined genes with associated gene data. Twenty-four of the genes were common to Selenium treatment at both 6 hours and 5 days. Additional information related to these genes was obtained using the GeneInfo data mining tool. This tool was developed by the authors (MWD, XW, HL, GZ) to allow for the rapid identification of supplemental data from the biomedical literature related to genes of interest. In brief, the tool allows one to cut and paste a list of genes based on either Unigene or Genebank IDs and search PubMed for associated references based on annotations of the associated gene names. Additional search terms can be stipulated by the user based on their knowledge of the biological process or in response to results received from the previous search. Results are returned in a table that lists the number of references that met the search criteria and provides a hyperlink to the associated references for either downloading or viewing. In this way the user is allowed to direct queries in an open manner based on their own experience or unpublished data. In this manner searches were conducted using the list of genes and the search terms "prostate cancer", "Selenium", and "apoptosis" (table 1-see additional file 1 ). IGFBP3 and RXR-alpha are expressed in the prostate, induced by selenium, and downregulated in prostate cancer Of the 154 genes identified with similar cross-species differential expression changes with respect to Selenium, two genes were identified that had unique features based on their associated references and interrelated functions. These genes, IGFBP3 and RXR-alpha were both up-regulated with respect to Selenium and could be used to suggest a model for Selenium action in prostate cancer. PXR-alpha is upregulated in both rat and human prostate cancer cells at 5 days in response to Selenium. Likewise, IGFBP3 is upregulated after six hours of Selenium treatment in both species. These two genes both contained Medline references with respect to prostate cancer, but had not yet been implicated in Selenium action. Western blotting performed on the human prostate cancer cell line PC3 with respect to Selenium validated the bioinformatically identified expression data (figure 3 ). To confirm the role of these two proteins in the prostate immunohistochemical studies on prostate cancer tissue microarrays were performed to identify IGFBP3 and RXR-alpha in both normal, nodular hyperplasia (benign prostatic hypertrophy), high grade prostatic intraepithelial neoplasia (HGPIN), invasive carcinoma, and metastatic prostatic carcinoma (table 2 ). These studies demonstrate that both IGFBP3 and RXR-alpha are expressed in the normal human prostatic epithelium (figure 4 , table 2 ). IGFBP3 is also expressed in the prostatic basal cells. Patterns of expression were predominantly nuclear, a finding that has been described for both proteins [ 30 ]. In addition, staining for IGFBP3 was also noted in the prostatic stroma, consistent with IGFBP3's associated function as a secreted protein. Decreased levels of IGFBP3 was noted in prostatic cancers when compared to normal prostate epithelium (p = 0.0044). Along with this decreased expression there was a distinct shift in the protein localization nuclear to cytoplasmic was observed (p < 0.00001), and in cases where expression was still present, there were decreased numbers and intensity of cell staining. IGFBP3 expression was similar in HGPIN, invasive carcinoma, and metastatic carcinoma. The level and pattern of IGFBP3 expression in nodular hyperplasia was similar to that seen in normal prostate tissues, and significantly different from the expression seen in cancer samples (p = 0.0036 and p < 0.00001 respectively). RXR-alpha expression was also significantly downregulated in prostate cancer when compared to normal prostate epithelium or nodular hyperplasia (p < 0.0001), and was similar to that seen in HGPIN and metastatic carcinoma. RXR-alpha expression was consistently nuclear in the samples studied, and while the intensity of staining was similar, in the remaining positive cancer cases there were decreased numbers of cells staining (8.6 +/- 12.6% in malignant epithelium vs 20.0 +/- 25.5% in normal epithelium). Figure 3 Expression of IGFBP3 and RXR-alpha with respect to Selenium. Western blotting reveals an induction of RXR-alpha or IGFBP-3 protein after Selenium treatment of human PC3 prostate cancer cells (arrows, upper row). Western blotting of immunoprecipitations from rat PAIII cells (bottom row) reveal RXR-alpha in immunoprecipitated IGFBP3 extracts (right panel) and IGFBP-3 in immunoprecipitated RXR-alpha extracts confirming and extending the reported interactions between the human proteins[40]. Table 2 Expression of IGFBP3 and RXRalpha in Prostatic Epithelium Normal Prostate Nodular Hyperplasia HGPIN Prostate Cancer Metastatic Cancer IGFBP3 Positive cases 105 62 49 202 25 Negative cases 5 1 9 36 8 Statistics (comparison) p = 0.0036 (cancer) N.S. (cancer) p = 0.0044 (normal) N.S. (cancer) IGFBP3 Intensity (avg+/-std) 2.47 +/- 0.70 2.49 +/- 0.65 2.57 +/- 0.82 2.74 +/- 0.56 2.79 +/- 0.49 Percentage cells (avg+/- std) 8.3 +/- 13.5 7.5 +/- 12.5 8.8 +/- 15.2 4.4 +/- 6.6 8.5 +/- 12.6 Nuclear cases 92 59 40 94 8 Cytoplasmic cases 22 6 18 152 11 Statistics (comparison) p < 0.00001 (cancer) p = 0.065 (cancer) p < 0.00001 (normal) N.S. (cancer) RXRalpha Positive cases 92 58 35 112 16 Negative cases 10 3 31 125 19 Statistics (comparison) p < 0.00001 (cancer) N.S. (cancer) p < 0.00001 (normal) N.S. (cancer) RXRalpha Intensity (avg+/-std) 2.73 +/- 0.51 2.78 +/- 0.50 2.83 +/- 0.38 2.76 +/- 0.49 3 +/- 0 Percentage cells (avg+/- std) 20.0 +/- 25.5 23.2 +/- 25.7 8.4 +/- 12.5 8.6 +/- 12.6 4.2 +/- 4.6 Nuclear cases 92 58 35 107 16 Cytoplasmic cases 2 0 6 9 0 Statistics (comparison) N.S. (cancer) N.S. (cancer) N.S. (normal) N.S. (cancer) Figure 4 Expression of IGFBP3 and RXRalpha in human prostate tissues. Immunohistochemical staining for IGFBP3 is present as brown staining in normal prostate (A) and prostate cancer (C). Similarly RXRalpha expression is present in normal prostate (B) and lost in prostate cancer (D). All images recorded at 100× magnification. Discussion Leveraging cross-species bioinformatics in the prioritization of gene data Through the use of cross-species comparisons of the number of differentially expressed genes to be examined after 6 hours and 5 days of Selenium treatment was dropped from 9453 and 7768 to 1123 and 1053 respectively, an 87–89 percent reduction of the sample size. Even with the use of multiple timepoints, the number of differentially expressed genes was only reduced in a single species study to 5934, less than half. By using comparative genomics the final dataset was reduced to 154 genes, providing a greater than 100 fold enrichment of the data. Thus by leveraging the additional biological species the ability to reduce the final analysis pool was substantial. This process only works if the species used have biological relevance to the disease in question. The choice of rat prostate cancer cell lines was made based on their use as an animal model for the study of prostate cancer [ 31 ]. The animal systems have been extensively used in the study of hormonal carcinogenesis, and in particular have been of value as a model of environmental and dietary effects on prostate cancer [ 18 - 20 ]. Previous studies have identified similar effects of rat animal models and prostate cancer cell lines to soy based diets [ 17 - 19 ], high fat diets [ 20 - 22 ], hormonal chemotherapeutics (Pollard, personal communication) and standard chemotherapy [ 32 , 33 ]. While comparative gene expression profiling has been performed, this has usually been through cross-species hybridizations to leverage RNA studies in species where sufficient expressed transcripts in a given species have not been identified for the production of species-specific gene expression slides, in particular for microbial genomes [ 34 - 37 ]. Thus the approach taken here leverages the production of species-specific gene expression profiles along with the increasing amount of gene homolog data generated by the sequencing of additional animal genomes. It is expected that with future genome efforts additional cross-species studies will be possible that leverage the knowledge of additional animal models in the study of disease. Similarities in prostate cancer transcriptomes across species For both overall and prostate expressed genes, we have failed to identify a significant difference in the transcriptomes between rat and human prostate cancer cell lines. This general similarity in transcriptomes may be due to the inherent biological similarities of the cell lines and/or their underlying biological origin. While the studies sought to utilize prostate cancer cell lines with similar biological potentials (established cell lines all derived from metastases) the degree of diversity present within the samples may account for some of the residual differences still identified. In addition, the extended period of time that these cell lines have been used has allowed for the continued in-vitro evolution of the cells, and could possibly extend those genomic differences. Yet the common clustering of the rat and human cell lines together suggests there are still significant similarities in their biological potential. This is also demonstrated by the similar biological potential of the cell lines when treated with a given stimulus, in this example Selenium. This parallels the similar physiological properties observed in the rat models of human prostate cancer. Based on these features we demonstrate that it is possible to identify functionally significant genes related to Selenium response by using comparative genomics. These findings also support the use of animal models in the study of human prostate cancer by suggesting that there is enough inherent genomic similarity that valuable insights may be gained from animal systems. Comparative genomics identifies functionally significant genes with respect to selenium chemoprevention A true test of the profiling method is the identification of genes that have a functional significance to the experimental system. In this case we have identified a series of genes, which when examined with additional data mining techniques, identifies genes with associated roles related to apoptosis (IGFBP3, RXRalpha, dynamin-2), antioxidant protection (selenoprotein N, peroxiredoxin I, zinc metalloprotease, glutathione S transferase), cell cycle (CDC26-anaphase promoting complex, kinetochore associated protein), and protein balance (proteasome subunit beta-4, ubiquitin conjugating enzyme). In addition, the ability to sort the identified genes by their associated biomedical literature allowed the focus to shift to IGFBP3 and RXRalpha. Retinoids, through the retinoid X receptor, have been shown to induce the expression of IGFBP3 [ 38 ]. In concert these two proteins act to induce apoptosis in cancer cell lines [ 39 ]. In particular, recent data has shown that these proteins work in synergy to enhance apoptosis in prostate cancer, and that there is a physical interaction between these two proteins in prostate cancer cells[ 40 ]. Further validation and confirmatory data is presented here that demonstrates the selenium induced expression and interaction between both RXRalpha and IGFBP3 in prostate cancer cells, along with their expression in normal prostate epithelium and subsequent down-regulation in malignant prostatic epithelium. This allows one to pose a model by which the restoration of IGFBP3 and RXRalpha levels by Selenium treatment may lead to the disruption of prostate tumorigenesis. This model is testable, and if validated, would present not only a mechanism by which Selenium may exert its effect, but provide a biomarker for assaying the effect of Selenium supplementation in the ongoing prostate cancer prevention clinical trials. Conclusions Using gene profiling on highly controlled spotted cDNA arrays we have demonstrated that similar baseline and selenium induced gene expression profiles can be identified between rat and human prostate cancer cells. This has allowed us to filter our gene expression data to identify genes whose transcriptional response to Selenium is similar across species, and by so doing focus our discovery process on specific common physiologic pathways. Two such proteins, RXR-alpha and IGFBP-3, which may be located in a common pathway, have been identified as dysregulated in human prostate cancers. This provides further support that the cross-species methods employed here can identify genes with roles in human prostate cancer. Methods Cell culture and selenium treatment Cell lines were received from ATCC, Rockford, MD, (LNCap, DU-145, MatLyLu, AT3), from Drs. Paul Lindholm and Andre Kadjacsy-Balla (LN4, Pro4, PC3, PC3-NI(PC3US), PC3-I(PC3-S)), or Dr. Morris Pollard and Mark Suckow (PA-III). These cells were cultured in RPMI (DU-145) or DME medium supplemented with 10% fetal calf serum, 10 mM glutamine, and 10 mM sodium pyruvate, and passaged 1:8 or 1:10 when the cells reached 70–80% confluence with trypsin-EDTA. For the Selenium studies PC3 or PAIII cells from a single cell stock were seeded at 1 × 10EE4 cells per ml and grown to 50% confluence at which time the culture medium was changed to either standard growth medium (above) or medium supplemented with twenty-five micromolar Selenium (Seleno-DL-methionine, Sigma cat# S3875, St. Louis MO). The cells were then cultured for an additional 6 hours or 5 days. Cells that reached 80% confluence prior to the five day timepoint were split using trypsin-EDTA and replated in either control or selenium-containing medium for the duration of the experiment. Cells were monitored for viability and cell growth with parallel growth curves conducted in triplicate, this data demonstrated the previously described [ 41 , 42 ] decrease in cellular proliferation (data not shown) observed in the presence of Selenium. RNA isolation and quantitation RNA was isolated from cells using Trizol (Invitrogen cat # 15596018, Carlsbad, CA) and subsequently examined for quality using agarose gel electrophoresis and Gelstar nucleic acid stain against known RNA standards and failed to demonstrate significant degradation based on the presence of high molecular weight RNA species, and intact 28s and 18s ribosomal RNA bands. DNA library preparation and amplification Sequence-verified rat and human libraries (Research Genetics, Huntsville, AL, and University of Iowa cDNA clone set, IA), consisting of 41,472 human clones and 36,000 rat clones were used as a source of probe DNA. A subset of 200 randomly selected clones were chosen from these libraries, resequenced locally, and demonstrated clone accuracy of 92%. We have opted to reformat libraries from 96 to 384-format for culture growth/archiving, PCR, purification, and printing. This has reduced the number of plates of our 41,472 human clone library from 432 to a more manageable 108, and the rat clone library from 375 to 94. The library was reformatted and subsequently manipulated using slot pin replicator tools (VP Scientific, San Diego, CA). Cultures were grown in 150 ul Terrific Broth (Sigma, St. Louis, MO) supplemented with 100 mg/ml ampicillin in 384 deep-well plates (Matrix Technologies, Hudson, NH) sealed with air pore tape sheets (Qiagen, Valencia, CA) and incubated with shaking for 14–16 hr. Clone inserts were amplified in duplicate in 384-well format from 0.5 μl bacterial culture diluted 1:8 in sterile distilled water or from 0.5 μl purified plasmid (controls only) using 0.26 μM of each vector primer {SK865 5'-fluorescein-GTC CGT ATG TTG TGT GGA A-3' and SK536: 5'-fluorescein-GCG AAA GGG GGA TGT GCT G-3'} (Integrated DNA Technologies, Coralville, IA) in a 20 μl reaction consisting of 10 mM Tris-HCl pH8.3, 3.0 mM MgCl 2 , 50 mM KCl, 0.2 mM each dNTP (Amersham, Piscataway, NJ), 1 M betaine, and 0.50 U Taq polymerase (Roche, Indianapolis IN). Reactions were amplified with a touchdown thermal profile consisting of 94°C for 5 min; 20 cycles of 94°C for 1 min, 60°C for 1 min (minus 0.5° per cycle), 72°C for 1 min; and 15 cycles of 94°C for 5 min; 20 cycles 94°C for 1 min, 55°C for 1 min, 72°C for 1 min; terminated with a 7 min hold at 72°. PCR reactions analyzed for single products by 1% agarose gel electrophoresis analysis. Products from replicate plates were pooled and then purified by size exclusion filtration using the Multiscreen 384 PCR filter plates (Millipore, Bedford, MA) to remove unincorporated primer and PCR reaction components. Forty wells of each 384-well probe plate were quantified by the PicoGreen assay (Molecular Probes, Eugene, OR) according to the manufacturers instructions. After quantification, all plates were dried down, and reconstituted at 150 ng/μl in 3% DMSO/1.5 M betaine. Array slide fabrication A single printing array containing 19,200 elements (human) or 2 arrays of 9,600 (rat), were printed on poly-L-lysine coated slides prepared in-house (1–2 arrays/slide) as previously described [ 9 ]. Printing was conducted with a GeneMachines Omni Grid printer (San Carlos, CA) with 16 or 32 Telechem International SMP3 pins (Sunnyvale, CA) at 40% humidity and 22°C. To control pin contact force and duration, the instrument was set with the following Z motion parameters, velocity: 7 cm/sec, acceleration: 100 cm/sec 2 , deceleration: 100 cm/sec 2 . All slides were post-processed using the previously described nonaqueous protocol[ 9 ]. Slide coating was performed as described previously [ 43 ]. Image files on all arrays were collected after blocking (fluorescein), and again after hybridization (Cy3 and Cy5) with a ScanArray 5000 (GSI Lumonics, Billerica, MA). Experimental design and bioinformatics based data analysis The experimental design utilized two biological replicates for each comparison with each replicate incorporating a Cy3/Cy5 dye flip. In addition, self-self hybridizations were performed for each sample to ensure experimental accuracy and evaluate expression bias. Comparisons were organized in a loop design for either human or rat prostate cancer cell lines, or were run as two-sample comparisons of baseline untreated control and Selenium treated cells. Array image TIFF files were analyzed with Gleams software (Nutec Sciences, Atlanta, GA). Additional TIFF file analysis, data normalization, clustering, and principle components analysis was performed using the Spotfinder, MIDAS and MultiExperiment Viewer Software from The Institute for Genomic Research (TIGR, Rockville, MD, [ 44 ], [ 12 ]) and used default values set in the MCW Practical Guide to TIGR Software Use (M. Datta, unpublished). In brief, image expression data was used as channel intensity minus background and intensity thresholds were set at a value of 300. Images were analyzed as dye flip pairs normalized using MIDAS with LocFit based LOWESS normalization and slice analysis set at two standard deviation cutoffs and a sample data population of 500 [ 45 ]. Samples were then averaged across two dye flip replicate pairs with removal of zero/dropped values using locally developed averaging software from the BEAR microarray suite (M. Datta, submitted). These final averaged values were subsequently annotated using the BEAR suite annotator and used for pattern identification and correlation with gene homologs. Homologous genes were identified from the NCBI homologene database ftp files and parsed using local scripts and databases present in the Bioinformatics Program,[ 46 ]. Additional data mining to identify references in the biomedical literature associated with specific genes and user chosen search terms was performed using the locally developed GeneInfo data tool (M. Datta, submitted). Raw data files, along with analyzed data subsets are available for use and study and can be obtained via a secure ftp site after contacting the corresponding author mdatta@mcw.edu . Protein purification, western blotting, and immunoprecipitation Protein extracts were prepared and immunoprecipitations and/or western blots made from five day twenty-five micromolar Selenium treated or control PC3 or PAIII prostate cancer cell lines as described previously[ 47 ]. In brief, ten micrograms of total protein were run on pre-cast 12% reducing SDS PAGE gels (Bio-Rad Labs, Hurcules, CA) and transferred to PVDF membranes. After blocking with caseine blocking buffer (Bio-Rad Labs, Hurcules, CA) the PVDF membranes were incubated with either anti-RXR-alpha or anti-IGFBP-3 antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) at 200 μg/ml dilutions, washed, and incubated with anti-rabbit secondary antibody (2 μg/ml) and developed with ECL Chemiluminescence (cat. RPN2108, Amersham Biosciences, Piscataway, New Jersey). Immunoprecipitations were carried out using 200 microgram samples of total cellular protein, which after preclearing with protein A agarose beads was sequentially incubated with either anti-RXR-alpha (1 μg/ml) or anti-IGFBP-3 (1 μg/ml) antibodies, washed, incubated with anti-rabbit protein A agarose beads, washed, and the protein pellet western blotted with the complimentary antibody (anti-IGFBP-3 or anti-RXR-alpha, respectively), and developed with ECL Chemiluminescence. Tissue microarray production, immunohistochemistry, and analysis After expedited institutional review board approval normal prostate tissues and prostate cancer samples were obtained from de-identified discarded patient specimens. The formalin-fixed paraffin embedded specimens were prepared as 5 micron sections. Tissue microarrays were prepared from donor tissue blocks as 0.6 mm cores in 12 (4 × 4) or (5 × 5) grids with between 192 to 300 samples and used in the preparation of 5 micron sections. Immunohistochemistry was performed using primary rabbit polyclonal antibodies to the insulin-like growth factor binding protein 3 (IGFBP3, 1:300, Santa Cruz Biotechnology, Santa Cruz, CA), or retinoic-X-receptor alpha (RXR-alpha, 1:800, Santa Cruz Biotechnology, Santa Cruz, CA) using methods previously described [ 48 , 49 ]. In brief, endogenous peroxidase from deparaffinized sections were blocked with Methanol/Acetic acid, and after treatment with blocking serum (ABC kit, Pierce Biotechnology, Rockford, IL) samples were incubated for 30 minutes with either anti-IGFBP3 (1:300) or anti-RXRalpha (1:600). Sections were subsequently washed, and incubated with anti-rabbitt secondary antibody conjugated to horseradish peroxidase and counterstained with Mayers hematoxalin. Antigen retrieval (90 C waterbath for 10 minutes) was used for RXRalpha. Positive controls for each antibody included nuclear staining in Sertoli cells [ 50 ] and lymphocytes[ 51 ]. Positive staining was recorded and scored on a 0–2 scale (0 = no staining, 1 = staining that does not obscure the hematoxalyn counterstain, 2 = staining that obscures the hematoxalyn counterstain). Evidence of positive staining was recorded as presence of staining (yes/no) or percent of epithelial or basal cells staining (number of cells staining over total number of cells). Patterns of staining (nuclear, cytoplasmic, membranous, diffuse extracellular) were also recorded. All samples were analyzed and recorded by two separate personnel, including a trained urologic pathologist (MWD, BM). Statistical analysis was performed using Chi-squared probability analysis. Abbreviations None declared. Authors contributions M.W.D. was responsible for the conception and implementation of this project in association with P.J.T., M.S., and M.P. H.L., X.W., and G.Z. were actively involved in the programming, database construction, and testing of the software. M.S. and M.H. were responsible for spotted cDNA construction, hybridization, and experimental analysis along with M.W.D. Cell culture, western blots, immunoprecipitations, and selenium treatments were performed by M.S. with assistance by B.M. Tissue microarray staining and analysis was performed by M.W.D., R.D., T.B., and B.M. All the authors reviewed and accepted the final version of the paper. Supplementary Material Additional File 1 Table 1, Word document, Table of the genes identified in the selenium gene expression studies. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516028.xml
549538	Changes in lipids over twelve months after initiating protease inhibitor therapy among persons treated for HIV/AIDS	Background Protease inhibitors are known to alter the lipid profiles in subjects treated for HIV/AIDS. However, the magnitude of this effect on plasma lipoproteins and lipids has not been adequately quantified. Objective To estimate the changes in plasma lipoproteins and triglycerides occurring within 12 months of initiating PI-based antiretroviral therapy among HIV/AIDS afflicted subjects. Methods We included all antiretroviral naïve HIV-infected persons treated at St-Paul's Hospital, British Columbia, Canada, who initiated therapy with protease inhibitor antiretroviral (ARV) drugs between August 1996 and January 2002 and who had at least one plasma lipid measurement. Longitudinal associations between medication use and plasma lipids were estimated using mixed effects models that accounted for repeated measures on the same subjects and were adjusted for age, sex, time dependent CD4+ T-cell count, and time dependent cumulative use of non-nucleoside reverse transcriptase inhibitors and adherence. The cumulative number of prescriptions filled for PIs was considered time dependent. We estimated the changes in the 12 months following any initiation of a PI based regimen. Results A total of 679 eligible subjects were dispensed nucleoside analogues and PI at the initiation of therapy. Over a median 47 months of follow-up (interquartile range (IQR): 29–62), subjects had a median of 3 (IQR: 1–6) blood lipid measurements. Twelve months after treatment initiation of PI use, there was an estimated 20% (95% confidence interval: 17% – 24%) increase in total cholesterol and 22% (12% – 33%) increase in triglycerides. Conclusions Twelve months after treatment initiation with PIs, statistically significant increases in total cholesterol and triglycerides levels were observed in HIV-infected patients under conditions of standard treatment. Our results contribute to the growing body of evidence implicating PIs in the development of blood lipid abnormalities. In conjunction with the predominance or men, high rates of smoking, and aging of the treated HIV-positive population, elevated lipoproteins and triglycerides may mean that patients such as these are at elevated risk for cardiovascular events in the future.	Introduction Abnormalities in the lipid metabolism of persons infected with human immunodeficiency virus (HIV), potentially induced by the disease itself and the medications used for treatment, were first reported in the early 1990s[ 1 ]. Reductions in high- (HDL) and low-density lipoprotein (LDL) cholesterol were observed amongst persons infected with HIV and increases in triglycerides were observed among persons with AIDS[ 1 ]. Following the introduction of protease inhibitors (PI), morphological changes in fat distribution were reported. This was followed by numerous reports of metabolic disturbances, including glucose and lipid abnormalities presenting as insulin resistance, impaired glucose tolerance, hyperglycemia, type 2 diabetes mellitus [ 2 - 5 ], elevated serum triglycerides, LDL and very low density lipoprotein cholesterol, apolipoprotein B, E, and lipoprotein(a) [ 2 - 4 , 6 ]. The combination of metabolic disturbances and morphological changes are now described as the HIV-related "lipodystrophy syndrome"[ 2 , 3 ]. Up to one-half of subjects treated with PIs have shown elevated levels of triglycerides, total cholesterol (TChol), LDL, insulin, and fasting glucose [ 7 - 10 ]. The onset of metabolic changes appears to occur soon after initiation of treatment, often as quickly as within several weeks[ 11 ]. The high prevalence of disturbances of lipoproteins and triglycerides, the rapidity of their onset, and large changes that have been observed in randomized trials have led to concern regarding the potential impact of PIs on the cardiovascular health of persons with HIV/AIDS. However, the direction and magnitude of changes in plasma lipoproteins and triglycerides induced by PIs has yet to be quantified in an observational setting. The aim of this study was to quantify the magnitude of change in lipoprotein and triglyceride levels over twelve months following any initiation of PI-based treatment in a cohort of subjects treated for HIV/AIDS in a large tertiary care institution. Methods We included all antiretroviral naïve HIV-infected persons treated at St-Paul's Hospital, British Columbia (BC), Canada, who initiated use of PIs between August 1996 and January 2002. Plasma lipoprotein and triglyceride levels were obtained from measurements taken over the course of regular monitoring. Longitudinal effects of the impact of PIs on lipoproteins and triglycerides were estimated using statistical models that accounted for correlation due to repeated measurements on the same individuals. Study Setting and Population In BC, antiretroviral drugs have been centrally distributed at no cost to eligible HIV-infected individuals since 1986[ 12 , 13 ]. In October 1992, the HIV/AIDS Drug Treatment Program became the responsibility of the BC Centre for Excellence in HIV/AIDS. Since December 1996, the mainstay of treatment for HIV/AIDS has been highly active antiretroviral therapy (HAART) including two nucleosides and either a PI or a non-nucleoside reverse transcriptase inhibitor (NNRTI). Typically, HIV-infected subjects receiving antiretroviral therapy are monitored by physicians at intervals no longer than three months at which time prescriptions are renewed or modified based on clinical and laboratory parameters, and necessary laboratory tests are conducted. This research received ethical approval from the Institutional Review Board of Providence Health Care in BC. Exposure to antiretroviral therapy for HIV/AIDS At the BC Centre for Excellence in HIV/AIDS, records of CD4+ T-cell counts and a profile of dispensed antiretroviral therapy are routinely maintained, including the: prescription fill dates, medications prescribed, and amount dispensed. Records of dispensed antiretroviral medications, (including nucleoside analogs, PI and NNRTI) were used to determine the drugs dispensed and available to each subject during each month of follow-up. Subjects were considered unexposed until the first month that a PI was dispensed. The exposure at the time of the lipid measurement was calculated as the cumulative number of consecutive months of drug exposure up until and including the month in which a measurement was taken. As we were interested in estimating changes that occur in plasma lipoproteins and triglycerides within the first year after treatment initiation, we included measurements with a cumulative PI exposure up to a maximum of twelve months. Subjects were considered unexposed thirty days after the last PI was dispensed and, at that time, the cumulative exposure was set to zero. Plasma lipoproteins and triglycerides Approximately one-half of HIV-infected patients in BC are followed at one treatment centre located at St. Paul's Hospital, a large teaching hospital in Vancouver. For these subjects, in addition to virological testing, the hospital laboratory records the results of plasma lipoproteins and triglycerides. Subjects were routinely instructed to fast for 12 hours prior to the sample being drawn. Blood samples were collected in 10-1 EDTA-coated vacutainer tubes. Plasma was separated by centrifugation for 10 minutes at 2,000 revolutions per minute. TChol in the plasma was determined using an enzymatic method[ 14 ] and plasma triglyceride was determined as previously described[ 15 ]. HDL cholesterol was determined using a heparin manganese precipitation of apo B-containing lipoproteins[ 16 ] and LDL cholesterol was calculated using the Friedewald formula[ 17 , 18 ]. For some subjects, the first lipid measurement was taken prior to the dispensation of any antiretroviral drug ("treatment naïve" baseline lipid measurements). Other subjects had their first recorded plasma lipoproteins and triglycerides after HAART was initiated (initiation lipid levels). Statistical analysis We used linear mixed effects models[ 19 ] to estimate the effect of PI-based HAART on changes in plasma lipoproteins and triglycerides over twelve months. As blood lipid levels were always greater than 0, the responses were log transformed prior to model fitting. Effect estimates were obtained via restricted maximum likelihood estimation using the R © function lme(.). An analysis of variance indicated that simple parallel-line models, where the exposure effect was assumed to be constant over subjects, could adequately explain variation in the data compared to more complex models with random slopes. To account for correlated responses due to repeated measurements on the same individual, the correlation structure of the model was specified to include only a random intercept for each subject. The following explanatory variables were included in the models: age, sex, and CD4+ T- cell count at PI initiation, concomitant use of NNRTI, and a measure of adherence to antiretroviral therapy. Exposure to NNRTI was considered time-dependent and quantified using the same algorithm that was used to calculate time dependent PI exposure. Adherence was calculated by dividing the number of months of antiretroviral medications dispensed by the number of months of follow-up in the first twelve months after treatment initiation. Incomplete adherence represents the gap between the time that the previous medication supply ran out until the next refill date, and/or until the last contact date with the program. This method is reliably associated with both clinical outcomes and un-timed drug level monitoring[ 20 , 21 ]. For each outcome and associated regression coefficient β, the quantity exp(β)-1 represented the adjusted monthly percent change in that plasma lipid fraction. These values were annualized using the formula exp(12 × β)-1. Ninety five percent confidence intervals (95% CI) for the effect estimates were obtained using the standard error for each regression coefficient. Two sensitivity analyses were conducted. The first analysis involved restricting the analysis to lipid measurements which were taken on subjects prior to any change from baseline antiretroviral therapy. The second sensitivity analysis included only subjects who had lipid measurements taken prior to initiation of therapy. Results There were 679 subjects who were eligible for analysis, 91% of whom were male (Table 1 ). The median age was 38 years at initiation of therapy and the median baseline CD4+ T-CELL COUNT was 210 cells/mm 3 . All subjects initiated therapy that included a nucleoside analog and a PI. Table 1 Demographic and clinical characteristics at treatment program enrollement among 679 subjects treated with PI-based HAART initiation for HIV/AIDS at St Paul's Hospital, Vancouver, BC, August 1996 – January 2002 Characteristic* Age (y) Mean (SD) Median 39 (8.9) 38 Sex (% male) 91 Status (% alive) at end of follow-up 93 CD4 (cells/mm 3 ) Mean (SD) Median 255 (222) 210 Abbreviations: SD = Standard deviation; PI protease inhibitor. Subjects were followed for a median of 47 months (Table 2 ). During that follow-up time, subjects had a mean of 73% of months in which a PI had been dispensed, with a mean of 31 prescription refills. The mean adherence in the first year was 89%, and the median adherence was 100% indicating that more than one-half of subjects were dispensed all their medications. A total of 3,010 lipid measurements were used in the analysis, with 400 subjects having had two or more lipid measurements during the course of follow-up. There were a median of 3 (IQR 1–6) measurements per subject. The median time between lipid measurements was 3 (IQR 2–5) months for all subjects. Table 2 Use of PIs among 679 subjects initiating PI-based HAART for HIV/AIDS at St Paul's Hospital, Vancouver, BC, August 1996 – January 2002 Characteristic* Number of months between initiation of PI and end of observation, median (IQR) 47 (29–62) % of follow-up time with dispensed therapy mean (SD), median for PI 73 (30) 84 Mean (SD) number prescriptions for* PI 31 (19) % adherence in first year mean (SD) 89 (21) median (IQR) 100 (92 – 100) Abbreviations: SD = Standard deviation; IQR = Interquartile range; NRTI = nucleoside reverse transcriptase inhibitors; PI protease inhibitor; NNRTI = non-nucleoside reverse transcriptase inhibitor * PIs available during follow-up included: indinavir, nelfinavir, saquinavir, and ritonavir. All data taken from time of first therapy initiation (treatment program enrollement). Time between first dispensation of HAART (enrollment) and last date of follow-up. * All prescriptions are of 30 day duration Table 3 shows the percentage change in plasma lipoprotein and triglyceride levels after 12 months of PI use, after adjustment for age, sex, and CD4+ T- cell count at initiation, concomitant use of NNRTI, and adherence. Statistically significant increases of about 20% were observed in TChol, the lipoprotein fractions, and triglycerides. Table 3 Baseline plasma cholesterol and triglycerides and estimated percent changes after 12 months when using PI among 679 subjects treated for HIV/AIDS at St Paul's Hospital, Vancouver, BC, August 1996 – January 2002 Number of measurements (subjects) Mean lipid baseline measurements (SD) % change* (95% CI) TChol 1620 (529) 4.2 (1.0) 20 (17, 24) HDL 1250 (419) 1.0 (0.3) 22 (15, 29) LDL 677 (295) 2.5 (0.8) 12 (5, 20) N-HDL 1247 (419) 3.1 (1.2) 20 (15, 26) TRG 1743 (556) 1.9 (1.2) 22 (12, 33) Abbreviations: PI protease inhibitor; TChol = total cholesterol; HDL = high-density lipoprotein cholesterol; LDL = low- density lipoprotein cholesterol; N-HDL = Non-HDL Cholesterol; TRG = triglycerides * Adjusted for age, sex, and CD4 + cell count at first prescription for HAART, concomitant use of non-nucleoside reverse transcriptase inhibitor and adherence to antiretroviral therapy in the first year of treatment The sensitivity analyses indicated that the models were robust to changes in the sub-population studied and the exposures that were included. In the analysis that included subjects who had a baseline measurement prior to start of HAART, a plasma profile prior to antiretroviral therapy initiation showed a similar pattern but with wider confidence intervals. In the analysis which was restricted to lipid measurements taken on subjects prior to switching from baseline therapy, the effect estimates were similar to those in Table 3 . Discussion In patients treated for HIV infection with HAART in a naturalistic setting, we observed that treatment with PIs was estimated to result in 20% increases in the levels of total cholesterol, HDL- and LDL-cholesterol and triglycerides twelve months after treatment initiation. Our results contribute to the growing body of evidence implicating PIs in the development of blood lipid abnormalities[ 22 ] and are in keeping with findings of randomized trials where a change from PI-based HAART to NRTI- or NNRTI-based HAART was associated with improvements in the lipid profile[ 23 ]. The observed increases in LDL-cholesterol and triglycerides were expected based on results reported from randomized trials. The finding of an increase in HDL-cholesterol has been reported only in some[ 24 ] but not all randomized trials. Low HDL levels and other forms of dyslipidemia, and disturbances in glucose metabolism, as well as central obesity, have been shown to be strong independent risk factors for cardiovascular morbidity and mortality in the general population[ 25 , 26 ]. Furthermore, the clustering of risk factors leads to greater cardiovascular morbidity and mortality than would be expected to occur in relation to each component alone[ 25 ]. These factors, when considered together, provide grounds to suspect that persons being treated for HIV infection with PI are at an increased risk of cardiovascular disease[ 8 ]. The increases in triglycerides caused by use of PI is of concern because there is growing evidence that these lipids are an independent risk factor for cardiovascular disease[ 27 , 28 ]. Two large observational cohort studies, published in 2003, showed discrepant results regarding the risks of cardiovascular disease among persons treated with HAART[ 29 , 30 ]. A retrospective administrative claims database study from the US Veteran's Affairs indicated that there was no relation between the use of antiretroviral therapy and the risk of cardiovascular or cerebrovascular hospitalizations [ 29 ]. A prospective multi-country collaborative study, with clinical events validation from eleven cohorts of HIV-infected persons showed that the incidence of myocardial infarction increased with longer exposure to combination antiretroviral therapy[ 30 ]. This study has a number of features that add credence to the results. First, while only based at one hospital, the study population comprised about one-half of the treated population in BC. The study sample had similar demographic composition (age and sex) and CD4+ T-cell count at initiation of therapy as other treated subjects in the province. We believe that these results are therefore generalizable to PI-regimen treated HIV-infected patients in BC, as well as to other target populations with similar demographic characteristics. Second, as all dispensed PIs and other antiretroviral medications are paid for and recorded centrally, there was complete information on all HAART medications available to all study subjects. We adjusted for adherence using each subjects' refill compliance in the first year of treatment. Third, data on the plasma lipid profile, including both lipoproteins and triglycerides, were of high quality. The measurements, while lacking the feature of being taken at uniformly spaced intervals as in closely monitored system such as randomized trial, represent the actual experience in an observational setting. Fourth, the enrollment period lasted over five years, allowing a large sample size. We employed repeated measures analyses to account for the correlated structure of the data. This study was also subject to several limitations. First, lipid measurements were not collected in a predetermined manner as is typical in a prospective research study. Consequently, the timing of measurements was irregular and not all lipid fractions were measured when blood was drawn. Second, despite instructions some subjects may not have fasted for the full 12 hour period prior to blood draws. Non-fasting triglyceride measurements are still useful for predicting cardiovascular disease and death[ 31 ]. Third, other potent risk factors for changes in lipid profiles such as family history and diet were not considered. It is possible but unlikely that the type of triple therapy prescribed by physicians could have been influenced by the belief that PIs may negatively impact lipid levels, likely leading to an attenuation of the estimates. Fourth, a number of factors contributed to a bias towards underestimating the true changes: for some subjects baseline values were not obtained prior to treatment initiation; it was not possible to adjust for concomitant use of lipid modifying agents such as statins because these data were not available; subjects with large changes in their lipid profiles may have been switched to another class; and treatment interruptions would also have affected exposure estimates. As a result of these limitations, the actual changes in plasma lipoprotein and triglycerides may be greater than reported in this study. Fifth, we assumed that patients who stop therapy or skip a month of therapy can be modelled as treatment naïve with respect to any subsequent therapy. More sophisticated statistical techniques are available[ 32 ] for handling this limitation, but appropriate software and resolution of technical issues are still in the developmental stages. Lastly, it is possible that the PI effect on lipids and triglyceride levels is not a class effect. The recently introduced PI, atazanavir sulfate, has been shown in several randomized clinical trials not to adversely affect lipid levels [ 33 - 38 ]. Due to changes in the HAART regimen that occurred because of resistance, non-compliance, or other reasons, it was not possible to isolate the effect of individual PIs. Over the period under study, most subjects initiating PI-based HAART were prescribed either indinavir or nelfinavir. Treating PI-induced dyslipidemia As PI-based HAART is very effective in increasing survival in HIV-infected patients[ 39 ], discontinuing PI is undesirable, even in patients with dyslipidemia. However, as PI-induced dyslipidemia is often asymptomatic and typically occurs in younger patients whose baseline risk of cardiovascular disease is low, the need for primary prevention is often under-appreciated in clinical practice[ 40 ]. When it is recognized, rather than discontinuing or switching PI therapy, one response is to initiate pharmacologic therapy with statins and/or fibrates[ 41 ]. Statins have been shown to be effective drugs to prevent cardiovascular disease in non-HIV patients[ 26 ]. While the role of statins in preventing cardiovascular disease in HIV patients remains to be demonstrated definitively, several smaller randomized controlled trials that have shown that these medications have beneficial effects on PI-induced dyslipidemia [ 42 ]. Until recently, pravastatin was the preferred statin for treating PI-induced dyslipidemia because it does not require metabolism by the cyp 3A4 system. However, recent studies have shown that moderate lipid lowering with pravastatin was less effective than intensive lipid lowering with atorvastatin in reducing: progression of atherosclerotic lesions [ 43 ] in patients requiring coronary angiography; and death or major cardiovascular events in patients with an acute coronary syndrome[ 44 ]. Atorvastatin is a powerful and effective statin, though there are still only limited published studies with this agent in HIV patients[ 45 ]. Because of metabolism by cyp 3A4, the dose of atorvastatin should be reduced one-half in patients receiving PIs. Lovastatin and simvastatin require metabolic activation by cyp 3A4 and should be avoided in patients receiving PIs. Rosusvastatin, a recently introduced statin has features that make it attractive for treating patients with PI-induced dyslipidemias: it does not require metabolism by the cyp 3A4 system and it effectively reduces LDL cholesterol. The safety and effectiveness of rosuvastatin in reducing clinical outcomes in HIV patients remain to be demonstrated. Fibrates are effective in lowering triglycerides and increasing HDL and have been shown to be effective in HIV patients[ 42 ]. Despite the impact on plasma lipids and triglycerides, only one trial has shown fibrates to reduce clinical outcomes[ 46 , 47 ]. Therefore, the role of fibrates in reducing cardiovascular disease outcomes remains to be fully elucidated. The main role of fibrates is to reduce the risk of pancreatitis associated with hypertriglyceridemia. In HIV patients, gemfibrozil should be limited to monotherapy because of the risk for myopathy. Fenofibrate is the preferred drug if combination therapy with a statin is contemplated. The efficacy of other lipid lowering drugs such as niacin and cholesterol transport blockers (ezetimibe) remains to be demonstrated. Bile acid sequestrants are contraindicated in patients receiving HAART. We conclude that there was an estimated 20% increase in lipoproteins and triglycerides in the first year after initiating PI-based antiretroviral therapy in HIV-infected patients under conditions of standard treatment. Whether these increases continue beyond one year is of considerable interest because lipoproteins and triglycerides are known risk factors for cardiovascular disease. In conjunction with the predominance or men, high rates of smoking, and aging of the treated HIV-positive population, elevated lipoproteins and triglycerides may mean that patients such as these are at elevated risk for cardiovascular events in the future.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC549538.xml
521690	The pattern of methacholine responsiveness in mice is dependent on antigen challenge dose	Background Considerable variation exists in the protocols used to induce hyperresponsiveness in murine models of allergic sensitisation. We examined the effect of varying the number of antigen exposures at challenge on the development of methacholine responsiveness in systemically sensitised mice. Methods BALB/c mice were sensitised with ovalbumin (OVA), challenged with 1, 3 or 6 OVA aerosols. Lung function was measured using low frequency forced oscillations and partitioned into components representing the airways (R aw ) and lung parenchyma (tissue damping (G) and tissue elastance (H)). Responsiveness to inhaled methacholine (MCh), inflammatory cell profile and circulating IgE were assessed 24 and 48 hours after challenge. The threshold dose of MCh required to elicit a detectable response (sensitivity) and response to 30 mg.mL -1 (maximal response) were determined for each compartment. Results Sensitivity ; All three OVA protocols resulted in an increased sensitivity to MCh in R aw but not in G or H. These responses where present at 24 and 48 hrs, except 1 OVA aerosol in which changes had resolved by 48 hrs. Maximal response ; 1 OVA aerosol increased maximal responses in R aw , G and H at 24 hrs, which was gone by 48 hrs. Three OVA aerosols increased responses in H at 48 hrs only. Six OVA challenges caused increases in R aw , G and H at both 24 and 48 hrs. Eosinophils increased with increasing antigen challenges. IgE was elevated by OVA sensitisation but not boosted by OVA aerosol challenge. Conclusions The pattern of eosinophilia, IgE and MCh responsiveness in mice was determined by antigen dose at challenge. In this study, increased sensitivity to MCh was confined to the airways whereas increases in maximal responses occurred in both the airway and parenchymal compartments. The presence of eosinophilia and IgE did not always coincide with increased responsiveness to inhaled MCh. These findings require further systematic study to determine whether different mechanisms underlie airway and parenchymal hyperresponsiveness post antigen challenge.	Background Persistent asthma is an allergic disease characterised by airway inflammation ([ 1 - 5 ]) and hyperresponsiveness to external stimuli ([ 1 ]). Mouse models of allergic airway sensitisation are often used to elucidate the pathobiology of this disease ([ 6 - 8 ]). To date, a number of techniques have been used to measure changes in lung function in response to bronchoconstricting agents in murine models of allergic bronchopulmonary inflammation (see [ 6 , 8 , 9 ] for reviews). One method that has gained recent popularity is unrestrained barometric plethysmography, which uses a 'pseudo-flow' measurement to derive a dimensionless parameter known as enhanced pause (Penh). There are now several publications in the literature which claim to have documented airway hyperresponsiveness in allergen-driven murine models based on methacholine induced changes in Penh. However, it has also been well documented that Penh does not correlate with changes in the physiology of the lung ([ 10 - 14 ]), especially in chronic disease states ([ 15 ]). In contrast, the low frequency forced oscillation technique (LFOT) is able to provide sensitive measurements of respiratory system input impedance (Zrs) in the mouse, that are partitioned into components representing airway and parenchymal compartments by fitting the constant-phase model ([ 16 - 18 ]). Using LFOT, Tomioka et al . ([ 17 ]) found that systemic sensitisation followed by three antigen challenges, one of the most common allergen models utilised in studies using Penh, resulted in hyperresponsiveness that was confined primarily to the tissue compartment of the lung. This has important implications for the interpretation of results obtained with Penh that have demonstrated mechanisms underlying allergic inflammation in mice given that a significant portion of the respiratory system hyperreactivity to MCh in human asthmatics is a result of the response of the conducting airways ([ 19 ]). One of the most common methods for inducing allergic bronchopulmonary inflammation in mice involves systemic sensitisation with a specific antigen and Th-2 skewing adjuvant, usually ovalbumin (OVA) adsorbed onto aluminium hydroxide (Alum), followed by airway challenge with the same antigen ([ 20 - 22 ]). However, considerable variations exist between studies in terms of the dose of antigen used during airway challenge. To date, a number of studies have found that airway hyperresponsiveness is increased by increasing the dose of antigen at challenge ([ 23 - 25 ]). However, these studies, which used different doses of antigen at challenge as part of a broader intervention protocol, have used Penh ([ 23 , 24 ]) or a measure of total lung resistance ([ 25 ]) to examine the resulting changes in lung physiology. As yet, no studies have systematically examined the effect of the dose of antigen at challenge on the subsequent development of hyperresponsiveness using a technique that is able to partition the reactivity of the lungs into airways and tissue compartments. Hyperresponsiveness of the respiratory system to bronchoconstricting agents, and other outcome parameters such as those that reflect inflammation and allergic sensitisation, are often measured at different times post challenge. In an examination of the kinetics of hyperresponsiveness in an OVA model of allergic sensitisation in mice using a single dose of antigen at challenge, Tomkinson et al . ([ 26 ]) found that responsiveness to methacholine (MCh) is maximal 24 hours post challenge, has begun to resolve by 48 hours, and has returned to baseline levels beyond that time. The kinetics of responsiveness to MCh in other studies, however, are often overlooked and it is yet be determined if altering the dose of antigen at challenge has an influence on the timing of peak responsiveness to bronchoconstricting agents. The aim of this study was to systemically investigate the effect of antigen dose at challenge on the pattern of hyperresponsiveness to inhaled MCh in a murine model of allergic bronchopulmonary inflammation. Methods Animals 8 week old specific pathogen free female BALB/c mice were purchased from the Animal Resources Centre, Murdoch, Western Australia. Mice were housed in a controlled environment with a 12 hr light:dark cycle and provided with an OVA free diet and acidified water ad libitum . All experiments were approved by the Institutional Animal Ethics and Experimentation Committee. Sensitisation protocols Mice were sensitised by intraperitoneal (i.p.) injection with 20 μg of OVA (Sigma, St Louis, USA) suspended in 200 μL of Alum (Alu-gel-S, Serva, Heidelberg, Germany) on days 0 and 14. Mice were then challenged with either 1, 3 or 6 OVA (1% w/v in PBS) aerosols delivered with an ultrasonic nebuliser (UltraNeb ® , DeVilbiss, Somerset, Pennsylvania) for 30 minutes on consecutive days starting at day 21 (Fig 1 ). Two additional groups of mice served as controls; a naïve group and a group sensitised with i.p. OVA and challenged with a single PBS aerosol using the protocol described above. Figure 1 Timeline for sensitisation and data collection. Timeline for the protocols used to induce allergic bronchopulmonary inflammation and timing for bronchoalveolar lavage (BAL), serum IgE measurement and assessment of hyperresponsiveness to inhaled methacholine (MCh). Mice were systemically sensitised with two intraperitoneal injections of OVA/Alum on day 0 and 14, challenged with either 1 (A), 3 (B) or 6 (C) OVA aerosols (1%) for 30 minutes starting at day 21. Respiratory mechanics Changes in Zrs were measured using a modification of the low frequency forced-oscillation technique (LFOT) as described previously ([ 27 ]). Briefly, mice were anaesthetised with an i.p. injection of a solution containing xylazine (2 mg.mL -1 , Troy Laboratories, NSW, Australia) and ketamine (40 mg.mL -1 , Troy Laboratories, NSW, Australia) at a dose of 0.01 mL.g -1 . Mice were tracheostomised with a 10 mm section of polyethylene tubing (1.27 mm OD: 0.86 mm ID) and ventilated ( flexiVent , Scireq, Montreal, Canada) at 450 b.min -1 with a tidal volume of 8 mL.kg -1 and a positive end expiratory pressure (PEEP) of 2 cmH 2 O. The lung volume history of the mice was standardised prior to measurement of lung mechanics using two deep inflations and three P-V curves. The respiratory system input impedance (Zrs) was measured during periods of apnea using a 16 s signal containing 19 mutually prime sinusoidal frequencies ranging from 0.25 to 19.625 Hz. The constant phase model ([ 16 ]) was then fit to the real and imaginary parts of the Zrs spectrum allowing the calculation of airway resistance (R aw ), tissue damping (G), tissue elastance (H) and hysteresivity (η) ([ 28 ]). Methacholine responsiveness Changes in respiratory mechanics following inhaled MCh were measured either 24 or 48 hrs after the last OVA aerosol. Following measurement of baseline Zrs, mice were exposed to a 90 s saline aerosol delivered with an ultrasonic nebuliser (UltraNeb ® , Devilbiss, Somerset, Pennsylvania). Zrs was then measured every minute for the next 5 minutes. This aerosol procedure was repeated with 1/2 log 10 incremental doses of MCh from 0.1 to 30 mg.mL -1 with Zrs measured every minute for at least 5 minutes after the aerosol until the parameters calculated from the constant phase model had peaked. Inflammatory cell counts Separate groups of mice, sensitised using the same protocol described above, were anaesthetised and tracheostomised 24 or 48 hrs after their last aerosol. BAL fluid was collected by slowly infusing and withdrawing a 1 mL aliquot of PBS containing BSA (bovine serum albumin, 20 mg.mL -1 , CSL, Victoria, Australia) and lidocaine (35 mg.mL -1 , Sigma, St Louis, USA) from the lungs three times. The BAL was then centrifuged at 2000 rpm for 4 mins. The supernatant was removed and the pellet resuspended in PBS. The cells were stained with trypan blue to determine viability and the total cell count (TCC) obtained by counting the cells with a haemocytometer. Differential counts were obtained from the cytospin sample, stained with Leishman's stain and examined using light microscopy. Three hundred cells were counted from each sample to determine the relative proportions of each cell type. Serum IgE In a separate group of mice, serum samples were periodically collected for analysis of total IgE. An additional control group was included in the analysis of serum IgE consisting of mice sensitised with PBS/Alum. Sera were diluted 1:7.5 in Delfia Assay buffer (Wallac Oy, Turku, Finland). The diluted sera were analysed for the presence of total IgE by time-resolved fluorescence (TRF) assays. Briefly, 96-well plates (Nunc Maxisorp, Denmark) were coated overnight at 4° C with anti-mouse IgE (R35-72; BD PharMingen, San Diego, USA). Plates were blocked with 200 μl of 0.5% BSA in TRIS-HCl pH 7.4 for 1 hour at room temperature on a plate shaker. For all subsequent steps a volume of 50 μl per well was used and incubations were performed for 1 hour at room temperature unless otherwise indicated. Between steps, plates were washed five times with wash buffer (TRIS-HCl pH 7.8 Tween20). Mouse anti-TNP IgE (BD PharMingen, San Diego, USA) was used as an interassay standard. Biotinylated anti-mouse IgE (R35-118; BD PharMingen, San Diego, USA) was added to the wells at 2 μg.mL -1 . Straptavidin-conjugated Europium (Wallac Oy, Turku, Finland) was incubated at 1:500 for 30 minutes and plates washed eight times thereafter. Delfia enhancement solution (Wallac Oy, Turku, Finland) was added and the plates were agitated on a shaker for 10 minutes prior to reading the fluorescence on a Wallac Victor 2 counter (Wallac Oy, Turku, Finland). The detection limit of this assay is approximately 100 ng.mL -1 . Statistical analysis Log 10 transformed inflammatory cell and immunoglobulin data were compared using ANOVA and Tukey's post-hoc test. Responses in R aw and G to inhaled MCh at the maximum dose used (30 mg.mL -1 ) were expressed as a percentage of the response to the saline aerosol and compared using non-parametric ANOVA on ranks and Dunn's post-hoc test. Responses in H were expressed as a percentage of the response to saline, log 10 transformed and compared using ANOVA and Tukey's post-hoc test. The threshold dose of MCh where there was a detectable change in R aw , G or H (termed sensitivity hereafter) was interpolated from the raw dose response curve as the upper limit of the 99% CI of the 5 measurements taken following the saline aerosol (Fig. 2 ). The sensitivity data were compared using ANOVA and Tukey's post-hoc test. All data were analysed using SigmaStat 2.03 and p values < 0.05 were deemed to be significant. Figure 2 Technique for sensitivity calculation. Schematic representation of the technique used for calculation of the threshold dose of MCh (sensitivity) required to induce a detectable increase in R aw , G and H. Results Methacholine responsiveness The degree and time of observed maximum MCh induced responses in R aw , G and H varied substantially between treatments (Fig. 3 ). A summary of statistical comparisons of sensitivity to MCh and percentage response to the maximum dose (30 mg.mL -1 ) between treatment groups and naïve mice is presented in Table 1 . Sensitisation followed by challenge with a single PBS aerosol did not cause an increase in sensitivity or maximum responsiveness to MCh compared to naïve mice. Figure 3 Dose response curves to inhaled methacholine. Dose response curves (expressed as a % of the response to saline aerosol) for mice systemically sensitised with OVA/Alum and challenged via the airways with 1 (left), 3 (centre) or 6 (right) OVA aerosols. Mice were challenged 24 (●) or 48 (▲) hours after the last OVA aerosol. Dose response curves from naïve mice (○) are also shown. All data are expressed as mean ± SEM (n = 7–8). * indicates significance (p < 0.05 vs naïve mice; ANOVA on Ranks, Dunn's post-hoc for R aw and G; ANOVA, Tukey's post-hoc for H). Table 1 Summary of sensitivity and maximum responses to methacholine in airway and parenchymal lung compartments. Summary of the threshold dose (sensitivity) required to elicit a detectable increase in airway resistance (R aw ), tissue damping (G) and tissue elastance (H) for naïve mice, mice systemically sensitised with OVA/Alum and challenged with PBS and mice systemically sensitised and challenged with OVA. Also shown is the percentage change in R aw , G and H in response to the maximum does of methacholine used (30 mg.mL -1 ). Data are presented as the mean (SEM). Challenge Assessed after last aerosol (hr) Sensitivity - Threshold dose of MCh (mg.mL -1 ) Response at 30 mg.mL -1 MCh Raw G H Raw § G § H § p* p* p* p* p* p* Naïve - 0.54(0.14) - 0.51(0.25) - 0.10(0.02) - 235.5(21.5) - 138.8(6.1) - 145.4(5.1) - 1 PBS aerosol 24 and 48 pooled 0.55(0.30) ns 0.25(0.10) ns 0.06(0.01) ns 242.9(19.1) ns 141.5(4.7) ns 143.4(2.7) ns 1 OVA aerosol 24 0.09(0.03) 0.012 0.15(0.04) ns 0.05(0.01) ns 514.0(82.7) <0.05 275.5(30.6) <0.001 250.2(24.5) <0.001 48 0.46(0.19) ns 0.18(0.05) ns 0.17(0.07) ns 326.4(32.6) ns 213.0(14.6) ns 179.3(12.6) ns 3 OVA aerosols 24 0.12(0.03) 0.012 0.35(0.20) ns 0.06(0.01) ns 271.2(43.1) ns 180.3(41.7) ns 178.9(39.9) ns 48 0.20(0.07) 0.034 0.16(0.04) ns 0.08(0.01) ns 348.8(46.0) ns 243.6(42.7) ns 233.0(31.6) 0.045 6 OVA aerosols 24 0.17(0.05) 0.019 0.12(0.05) ns 0.06(0.01) ns 456.7(43.4) <0.05 358.2(88.4) <0.05 298.5(52.2) 0.02 48 0.16(0.05) 0.034 0.15(0.06) ns 0.06(0.01) ns 396.0(23.3) <0.05 304.8(28.5) <0.05 295.6(30.6) 0.018 § expressed as a % of saline response * vs naïve values One OVA aerosol A single OVA aerosol was sufficient to induce a significant increase in MCh responsiveness in the airways, seen as both a lower threshold dose of MCh required to induce a response (increased sensitivity) and increased response at the 24 hour time point (Table 1 ). In the parenchymal compartment, no increase in sensitivity was seen but a significant increase in maximal response was seen for both G and H. This heightened sensitivity and response had diminished, back to the level seen in naive mice, 48 hours after the OVA aerosol. Three OVA aerosols Three OVA aerosols resulted in significantly increased airway (but not parenchymal) sensitivity to MCh at both the 24 and 48 hour time points (Table 1 ). However, there was no increase in maximum response at 24 hours in R aw , G or H and only an increased response in H after 48 hours but not R aw and G. Six OVA aerosols Six OVA aerosols resulted in both significantly increased airway sensitivity and maximal responses to MCh at 24 and 48 hours post-challenge. Increased maximal responses, but not increased sensitivity, were also seen in the parenchymal compartment at both the 24 and 48 hour time points. Inflammatory cell counts Challenge with a single PBS aerosol following systemic sensitisation with OVA did not cause a significant increase in TCC in the BAL (p = 0.552) compared to naïve mice (Fig. 4 ). There was, however, a significant increase in TCC in mice challenged with a single OVA aerosol (p = 0.032) and a further increase in TCC following 3 OVA challenges (p < 0.001). Exposure to 6 OVA aerosols did not cause any further increase in TCC above levels observed in mice exposed to 3 OVA aerosols (p = 0.805) but remained significantly higher than mice challenged with 1 OVA aerosol (p < 0.001). Time of sampling after the last aerosol with any of the protocols did not have a significant impact on TCC (p = 0.357). Figure 4 Total cell counts from bronchoalveolar lavage. Total cell counts (TCC) from the bronchoalveolar lavage (BAL) of naïve BALB/c mice, mice systemically sensitised and challenge with OVA aerosols and mice systemically sensitised with OVA and challenged with PBS. Samples were collected 24 (grey) and 48 (black) hours after the last aerosol. Data are expressed as mean ± SEM (n = 5–6). Exposure to a PBS aerosol following antigen sensitisation did not cause an increase in TCC (p = 0.552). In contrast, a single OVA aerosol was sufficient to cause a significant increase in TCC (p = 0.032). Exposure to 3 OVA aerosols caused a further increase in TCC (p < 0.001) but 6 OVA aerosols did not cause an increase in TCC beyond those observed in mice exposed to 3 OVA aerosols (p = 0.805). The number of aeroallergen challenges also had a significant impact on the number of eosinophils (p < 0.001) and macrophages (p < 0.001) in the BAL. There were significant increases in the number of eosinophils in sensitised mice challenged with 1 (p = 0.032), 3 (p < 0.001) and 6 (p < 0.001) OVA aerosols (Fig. 5 ) compared to naïve mice. The numbers of eosinophils in the BAL of mice exposed to 3 and 6 aerosols were significantly higher than those exposed to a single OVA aerosol (p < 0.001 and p < 0.001 respectively) but were not significantly different from each other (p = 0.805). The number of macrophages in the BAL were also higher in mice exposed to 3 (p < 0.001) and 6 (p < 0.001) OVA aerosols compared to naïve mice. As with TCC, time of sampling after the last aerosol did not have a significant impact on the number of eosinophils (p = 0.357) or macrophages (p = 0.079) in the BAL. Low levels of neutrophils were observed in BALs from OVA challenged mice sampled at 24 hours but not in mice sampled 48 hours after the last OVA aerosol (Fig. 5 ). Lymphocyte numbers were not significantly elevated in the BALs from any of the treatment groups ( data not shown ). Figure 5 Differential cell counts from bronchoalveolar lavage. Differential cell counts from the bronchoalveolar lavage (BAL) of naïve BALB/c mice, mice systemically sensitised and challenge with 1,3 or 6 OVA aerosols and mice systemically sensitized with OVA and challenged with a single PBS aerosol. BALs were collected 24 and 48 hours after the last aerosol. Data are expressed as mean ± SEM (n = 5–6). There was a significant increase in the number of eosinophils (p = 0.032) in the BAL following a single OVA aerosol. Exposure to 3 or more OVA aerosols caused a further increase in the number eosinophils (p < 0.001), compared to 1 OVA aerosol, and an increase in the number of macrophages (p < 0.001) compared to naïve mice. There were neutrophils present in the BALs of some mice but only in those groups sensitised and challenged with OVA and only in BALs sampled 24 hours after the last aerosol. Serum IgE Total serum IgE was significantly increased at day 21 (p < 0.001), 7 days after the second injection of OVA/Alum, compared to naïve mice (Fig. 6 ). In contrast, serum IgE levels at day 14, after a single injection, were not significantly elevated (p = 0.438) compared to naïve mice. The total serum IgE response to systemic sensitisation, in the absence of subsequent antigen aerosol challenge, peaked at day 22 and partially declined by day 27. However, this decrease was not statistically significant (p = 0.511). There was no further increase in the total serum IgE in mice that were sensitised and subsequently challenged with OVA aerosols compared to those that were only systemically sensitised (p = 0.842). Total serum IgE levels were not significantly greater in mice sensitised with PBS/Alum and challenged with OVA ( data not shown ). Figure 6 Total serum IgE obtained from time resolved fluorescence. Total IgE obtained from time resolved fluorescence assay of serum collected from systemically sensitised (i.p. OVA/Alum on day 0 and day 14) but not challenged with aerosolised antigen (white bars). The vertical bars represent total serum IgE from mice sensitised and challenged with either 1, 3 or 6 OVA aerosols. Serum samples from these mice were collected 24 (grey bars) and 48 (black bars) hours after the last aerosol. Data are expressed as mean ± SEM (n = 10). Two intraperitoneal injections of OVA/Alum were sufficient to induce increased levels total IgE by day 21 (p < 0.001) compared to naïve mice. Exposure to OVA aerosol challenges did not cause a further increase in total IgE (p = 0.842). Discussion Varying the number of aeroallergen challenges in a systemically sensitised murine model of allergic bronchopulmonary inflammation altered the degree and timing of hyperresponsiveness to inhaled MCh. A single OVA challenge increased airway sensitivity to inhaled MCh 24 hours after the challenge, while sensitivity remained elevated for 48 hours after three and six challenges. OVA challenge did not increase parenchymal sensitivity at any level. In contrast to sensitivity measurements, the maximum response to 30 mg.mL -1 MCh showed a variable pattern. A transient response was observed in both airway and parenchymal compartments after a single OVA aerosol. After 3 OVA aerosols significant increases were seen in the tissue compartment at 48 hours, while after 6 OVA aerosols an elevated response was seen in the airway and parenchymal compartments that persisted beyond 48 hours. There was a significant influx of inflammatory cells in the BAL in response to OVA aerosols, however, the presence of this inflammation did not always result in hyperessponsiveness to inhaled MCh. Murine models using 2 systemic allergen sensitisations followed by 3 aeroallergen challenges are prevalent in the literature ([ 20 , 29 - 31 ]) and have been reported to demonstrate airway hyperresponsiveness to MCh. However, these studies have used enhanced pause (Penh), which is derived from unrestrained barometric plethysmography, to measure changes in lung physiology. As Penh cannot differentiate between constriction in the airways and changes in the tissue compartment of the lungs, it is impossible to tell where the responses to MCh are localised, if indeed they are true physiological responses ([ 10 - 14 ]). In contrast, our study, using 2 systemic sensitisations and 1,3 or 6 challenges, has demonstrated clear airway, tissue or mixed compartment responses to methacholine which is dependent on the number of aerosol challenges delivered. In our hands, the more common model of 2 systemic sensitisations followed by 3 OVA challenges resulted in increased responsiveness to the maximum dose of MCh that was confined to the tissue compartment of the lung. This finding is consistent with a previous study by Tomioka et al . ([ 17 ]), which also used a forced oscillation technique to measure changes in lung mechanics in OVA sensitised and challenged mice. The fact that the response was confined to the tissues is of interest as the aim of these models is to mimic the human asthmatic condition, in which a significant portion of reactivity of the lungs is localised in the conducting airways ([ 19 ]). This work emphasises the importance of measuring bronchoconstriction with physiological techniques capable of compartmentalising responses within the lungs. By varying the antigen dose at challenge we have revealed a system with the potential to allow investigation of transient or prolonged responsiveness to MCh that is localised in the airways, tissues, or both. Further investigation is needed in order to understand the mechanisms that are influencing the site of responsiveness. Typically, most human studies measure MCh responsiveness in terms of sensitivity as they report the concentration of MCh required to produce a 20% fall in FEV 1 . We have shown that it is possible to determine sensitivity to inhaled MCh in mice and that only the airway compartment shows heightened sensitivity following allergic sensitisation and challenge. While increased maximal responses can be seen in both airway and parenchymal compartments, depending on which model is used, no increase in parenchymal sensitivity is seen with any of the models we used. As such, these findings reinforce the value of using lung function techniques that are capable of assessing airway and parenchymal mechanics separately. Total serum IgE was significantly elevated following systemic sensitisation but was not increased by aerosol challenge. There was, however, a tendency for total serum IgE to decline by day 27 in mice that were systemically sensitised but not challenged with OVA aerosols, compared to mice additionally exposed to 6 OVA aerosols. It is possible that if the study had been extended to include further exposure to antigen over subsequent days, a difference would have been detected between mice that were only sensitised and mice that were sensitised and challenged. Given that antigen specific IgE and other immunoglobulin subtypes were not measured in this study, further work is required to characterise the effect of dose of antigen at challenge on the development of antibody responses to OVA in mice. The protocol used in the present study induced significant eosinophilia after a single airway challenge. The degree of eosinophilia increased with increasing number of airway challenges. This finding is consistent with several previous studies using similar protocols to induce allergic inflammation in the lungs of mice ([ 20 , 29 - 31 ]). While the level of activation of the eosinophils was not measured in the present study, the 61% eosinophilia found after 6 OVA aerosols was much higher than those that are typically found in human asthmatics ([ 32 ]). Given the significant and prolonged parenchymal response to inhaled methacholine following 6 OVA aerosols and the level of eosinophilia present, it is likely that this model more closely parallels an allergic alveolitis ([ 33 ]) than the airway inflammation commonly seen in humans. In recent studies there has been some focus on the association, or lack thereof, between indicators of systemic sensitisation, such as the levels of serum antibodies, airway inflammation and AHR ([ 34 ]). In a review of the role of IgE in the induction of eosinophilic airway inflammation and AHR, Hamelmann et al . ([ 35 ]) concluded that systemic methods of sensitisation resulted in high levels of IgE and eosinophilic airway inflammation in BALB/c mice. In these models, AHR was determined to be dependent on eosinophils but not IgE. However, the results of our study, which uses a similar protocol to those reviewed by Hamelmann et al . ([ 35 ]), show that the presence of eosinophils did not always coincide with an increase in responsiveness to MCh. Three OVA aerosols resulted in a significant eosinophilia after 24 hours but an increase in the response to the maximum dose of MCh was not evident until 48 hours post challenge. In contrast, a single OVA challenge resulted in hyperresponsiveness to MCh that had resolved by 48 hours while the levels of eosinophils remained significantly elevated. The levels of total serum IgE were equivalent across all challenge doses suggesting that, while the presence of IgE may be necessary to initiate the allergic response, its presence at a particular measurement time point does not necessarily relate to the presence of hyperresponsivenss. Conclusions The findings of the present study demonstrate the significant impact of changing antigen challenge dose in a murine model of allergic bronchopulmonary inflammation. Given the variability of the inflammatory profile and characteristic responses observed in this study, it is clear that investigators must carefully characterise their allergen-driven murine models to ensure the model used contains the characteristic of interest. Future studies need to be directed at understanding the mechanisms that underlie airway and parenchymal hyperresponsiveness post antigen challenge. Authors' contributions GRZ carried out the animal studies and drafted the manuscript. CvG carried out the IgE analysis and assisted in the interpretation of results and editing the manuscript. PAS assisted in the interpretation of results and editing the manuscript. PGH assisted in the conceptualisation of the study and interpretation of the results. PDS and DJT assisted in the conceptualisation of the study, interpretation of the results and editing the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521690.xml
546186	Differences in codon bias cannot explain differences in translational power among microbes	Background Translational power is the cellular rate of protein synthesis normalized to the biomass invested in translational machinery. Published data suggest a previously unrecognized pattern: translational power is higher among rapidly growing microbes, and lower among slowly growing microbes. One factor known to affect translational power is biased use of synonymous codons. The correlation within an organism between expression level and degree of codon bias among genes of Escherichia coli and other bacteria capable of rapid growth is commonly attributed to selection for high translational power. Conversely, the absence of such a correlation in some slowly growing microbes has been interpreted as the absence of selection for translational power. Because codon bias caused by translational selection varies between rapidly growing and slowly growing microbes, we investigated whether observed differences in translational power among microbes could be explained entirely by differences in the degree of codon bias. Although the data are not available to estimate the effect of codon bias in other species, we developed an empirically-based mathematical model to compare the translation rate of E. coli to the translation rate of a hypothetical strain which differs from E. coli only by lacking codon bias. Results Our reanalysis of data from the scientific literature suggests that translational power can differ by a factor of 5 or more between E. coli and slowly growing microbial species. Using empirical codon-specific in vivo translation rates for 29 codons, and several scenarios for extrapolating from these data to estimates over all codons, we find that codon bias cannot account for more than a doubling of the translation rate in E. coli , even with unrealistic simplifying assumptions that exaggerate the effect of codon bias. With more realistic assumptions, our best estimate is that codon bias accelerates translation in E. coli by no more than 60% in comparison to microbes with very little codon bias. Conclusions While codon bias confers a substantial benefit of faster translation and hence greater translational power, the magnitude of this effect is insufficient to explain observed differences in translational power among bacterial and archaeal species, particularly the differences between slowly growing and rapidly growing species. Hence, large differences in translational power suggest that the translational apparatus itself differs among microbes in ways that influence translational performance.	Background Translational power is the rate of protein synthesis of a cell or culture, normalized to the amount of biomass invested in the protein synthesis machinery. We are introducing the term 'translational power' to describe precisely the same concept (and the same quantitative parameter, see Methods) that was originally defined as 'ribosome efficiency' [ 1 - 3 ]. In recent years, this concept has more commonly been called 'translational efficiency' [ 4 , 5 ], particularly in discussions of codon usage bias [ 6 - 8 ]. Although we are reluctant to depart from established terminology, we do so to avoid an inconsistency with the meaning of 'efficiency' as it is used in many other areas of science and in common parlance. In the physical sciences and in many areas of biology, the efficiency of a process refers to a comparison of output to input, in particular to the fluxes of useful energy and/or mass (e.g., the efficiency of a heat engine [ 9 ], trophic transfer efficiency [ 10 ]). These scientific meanings of 'efficiency' are consistent with the common notion that a process obtaining the desired output with little waste is highly efficient. According to these conventions, calculations of efficiency make no direct reference to the rate at which a process occurs. Physicists and engineers use a distinct term, 'power,' to refer to the rate of energy consumption or the rate at which work is performed [ 11 ]. The semantic distinction between power (or rate) and efficiency is important, because in many real and idealized physical systems, the laws of thermodynamic result in an inherent tradeoff between power and efficiency [ 9 ]. In biology, several attempts to argue for the universality of power-efficiency tradeoffs [ 12 , 13 ] have justifiably been criticized for the misapplication of thermodynamic arguments [ 14 - 16 ]. Nonetheless, many specific tradeoffs have been demonstrated in a wide range of organisms that can be described as evolutionary choices between power (increased rates of biological processes such as resource acquisition, metabolism or organismal growth) and efficiency (increased biological output measured as probability of survival, production of biomass, number of progeny, etc. per unit resource) [ 17 - 24 ]. Among bacteria, comparisons of coexisting species or strains have also provided evidence for power-efficiency tradeoffs [ 25 - 28 ], as have comparisons of engineered mutant strains [ 29 , 30 ]. However, the absence of apparent tradeoffs in some carefully designed studies of bacteria demonstrates that such tradeoffs are not inevitable [ 31 - 33 ]. Even if power-efficiency tradeoffs occur only in some biological contexts, it is valuable to maintain a semantic distinction between power (implying rapid rate) and efficiency (implying low waste). However, the terms 'ribosome efficiency' and 'translational efficiency' blur this distinction, because they refer to a rate – the quantitative measure of ribosome efficiency [ 1 ] is expressed in units of (time -1 ). We prefer the term 'translational power', which refers to the rate of protein synthesis of a cell or culture, normalized to the mass of the translational apparatus, in a manner that is more consistent with the connotations of 'power' and 'efficiency' derived from other areas of science and from colloquial usage. Translation rate (a synonym of 'protein chain growth rate' [ 3 , 34 ], meaning the rate of amino acid polymerization per active ribosome) is one component of translational power, but translational power reflects other properties of the protein synthesis system as well, most notably the fraction of ribosomes that are active (see Methods, also chapter 6 of reference [ 34 ]). Intuitively, translational power measures the capacity of the protein synthesis subsystem to drive replication of the cell, the protein-dominated autocatalytic system to which it belongs. The concept and a quantitative metric of translational power were first introduced to facilitate comparisons of translational performance between different growth rates within a single bacterial strain [ 1 ]. The initial belief that translational power is nearly constant in a strain across a wide range of growth rates, based both on empirical data and theoretical arguments [ 2 , 34 ], has gradually given way to the current understanding that translational power increases with growth rate, at least in E. coli [ 3 , 4 , 35 , 36 ]. The question of whether translational power varies between microbial species has been investigated only rarely, in four studies that each compared a single slowly-growing microbial species to E. coli [ 37 - 40 ]. In each case, translational power was found to be higher in E. coli than in the slowly growing comparison strain. Although each of these studies discusses this unexpected result, only one of them references the same result from another study. In previous work, the consistent association of low translational power with slowly growing microbes appears to have escaped notice; however, our reanalysis of the data from these four studies as well as additional published data (presented in Results) suggests that the association is robust. One factor capable of affecting translational power is the biased usage of synonymous alternative codons. In the standard translational code, 18 of the 20 amino acids are encoded by more than a single codon, but in many microorganisms, synonymous codons are not used with equal frequency. The pattern first found in E. coli and Bacillus subtilis turns out to be common: the majority of genes within an organism show a preference for the same subset of codons, but the degree of bias towards the preferred subset is correlated with the expression level of the gene [ 41 , 42 ]. For some time, the consensus has been that such a pattern reflects selection for translational power [ 7 , 8 ]. Codon bias increases translational power because preferred codons tend to be translated more rapidly than synonymous alternatives [ 43 - 45 ]. This effect can be attributed to the high abundance of tRNAs cognate to the preferred codons, to a canonical base pair interaction at the codon wobble position between preferred codons and their cognate tRNAs, or to both these factors [ 7 , 8 ]. Codon bias resulting from selection for translational power (or for any other translation-dependent benefit) is correlated with gene expression level because the benefit accrues during each instance of translation, so the selective pressure for preferred codons is stronger in more highly expressed genes [ 7 , 8 ]. In contrast to the codon bias caused by translational selection, codon bias that is consistent in both magnitude and direction in genes that vary widely in expression level is explained most easily by mutational bias acting on DNA [ 8 , 46 ]. While the effects of both translational selection and mutational bias are evident in some microbial genomes with moderately biased G+C content [ 47 , 48 ], organisms with strong mutational bias (very high or low G+C content) have been reported to show very little [ 49 ] or no [ 50 - 52 ] evidence of translational selection. Theoretical calculations indicate that if the strength of mutational bias exceeds a certain critical threshold, any pre-existing codon preferences that conflict with the mutational bias will be reversed [ 53 ]. In this case, codon use is almost entirely determined by the mutational bias, which influences genes equally regardless of expression level. Note that while the degree of codon bias and the gene expression level would not be correlated among genes from such a genome, this does not necessarily imply that deviations from the average (biased) codon usage would be selectively neutral, nor that the fitness effects of any such deviations would be independent of gene expression level. The absence of a correlation between codon usage and gene expression level has also been reported in some organisms with moderate G+C content, in particular the spirochete Treponema pallidum [ 54 ] and the proteobacteria Helicobacter pylori [ 55 ]. The lack of evidence for translational selection in these organisms requires an explanation, since they lack a strong mutational bias that could obscure such evidence. It has been suggested that rapid exponential growth confers little or no fitness benefit in these strains [ 8 , 55 ], consistent with their slow growth rate and other characteristics of their ecological niche. If so, these organisms would not experience selection for translational power. If variation in the strength of selection for translational power leads to differences in the degree of codon bias between microbes (superimposed on any differences in codon bias that can be attributed to variation in mutational bias), we wondered whether differences in codon bias could in turn explain the observed differences in translational power between microbes. An estimate of the effect of biased codon use on the overall rate of translation would depend on knowledge of absolute or relative translation rates in vivo for each codon. Unfortunately, these data are incomplete even for E. coli , and are not available for other microbes. Therefore, we approach the issue by framing the following question: How much faster is the translation rate of E. coli than the expected translation rate of a hypothetical organism that has the same proteome composition and the same investment in translational machinery as E. coli , but which lacks codon bias? Here we report results from a simple mathematical model developed to address this question. For convenience, we will refer to the hypothetical E. coli -like organism with uniform use of synonymous alternative codons as 'Uni'. By 'same proteome composition', we mean that over a cell generation, each amino acid is incorporated into protein the same number of times in Uni and in E. coli , although for the 18 amino acids specified by multiple codons, the individual codons will differ in frequency. By 'same investment in translational machinery', we mean that the total biomass of the translational apparatus is the same in Uni and in E. coli , although ideally the allocation of that biomass among various components of the apparatus in Uni would be optimized for unbiased codon usage. However, in order to apply empirical codon-specific translation rate data, we will impose a more stringent requirement on Uni, that the abundance of each individual component of the translational apparatus will be unchanged in comparison to E. coli . Due to this restriction, and due to the incomplete data for codon-specific translation rates, we make no claim to be able to answer our question precisely. However, our approximations are adequate to conclude that differences in codon bias alone are unlikely to account for differences in translational power of the magnitude inferred from macromolecular analysis of slowly growing and rapidly growing microbes. Results Comparisons of translational power among microbes We know of 4 studies that have made explicit comparisons of translational power between different microbial species; in each case, the comparison was made between E. coli and a single slowly growing strain [ 37 - 40 ]. One of these studies relied on original measurements of E. coli [ 38 ]; the remaining studies made comparisons using the E. coli data of Bremer and Dennis [ 3 ]. Although growth rates and translation rates vary with temperature [ 56 ], at least 2 of the 4 studies [ 39 , 40 ] compared data from strains grown at different temperatures without compensating for temperature effects. One of 2 studies that made comparisons based on the number of ribosomes per cell volume appears to have assumed that E. coli cell volume is constant over a range of growth rates [ 39 ], which is unlikely. We have reanalyzed the data from these studies (as described in Methods) to provide consistent comparisons of translational power between E. coli and other strains. In addition, we applied the same comparative methodology to every microbial species for which we could find the requisite data in the literature. The list of species that could be included is surprisingly short; most studies reporting both the protein and RNA content of microbes growing at known rates have involved E. coli or closely related enteric bacteria. Table 1 summarizes the comparisons of translational power between E. coli and all other species. The comparisons of translational power in Table 1 are based on the fastest growth rate for which data are available for each of the comparison organisms, because at submaximal growth rates, there may be a reduction in the average translation rate [ 4 , 57 ], in the active fraction of ribosomes [ 35 , 36 ], or both. Either of these phenomena would reduce translational power. However, the comparisons to E. coli are not always based the fastest E. coli growth rate, but rather on the growth rate at which E. coli makes a comparable investment in the translational apparatus as the comparison organism. A comparison at similar investment levels reflects the expectation that the selective pressure to maximize translational power increases with the biomass invested in the apparatus [ 4 , 58 ]. If the comparisons had always been made to the fastest E. coli growth rate (i.e., where E. coli translational power is highest), the disparity in translational power would be greater for most of the comparisons shown. Even with the conservative comparisons displayed in Table 1 , the published data suggest that translational power varies considerably between strains, particularly for comparisons between microbes adapted to different ranges of growth rates. While translational power is higher in E. coli and other rapidly growing organisms, it is lower in slowly growing organisms, ranging from less than 17% to 42% of the value for E. coli . Hence, if differences in the degree of codon bias are to explain these differences in translational power, we would expect codon bias to be capable of accelerating the rate of translation by 2.5-fold to 6-fold. In summarizing the comparisons of Table 1 as a contrast between slowly growing and rapidly growing microbes, we are not relying on the actual growth rates shown in the third column, especially since chemostat growth rates are necessarily constrained below the maximal growth rate for a strain. Instead, we have relied both on well-recognized growth characteristics for some species (e.g., Sphingopyxis alaskensis and Rickettsia prowazekii are slow growers, Salmonella enterica and Enterobacter aerogenes are rapid growers), and on the number of copies of the ribosomal RNA ( rrn ) operon per genome. High rrn copy number is an adaptation permitting rapid growth [ 59 , 60 ], while low rrn copy number is characteristic of microbes adapted for slow growth [ 39 , 61 ]. Estimates of the translation rate benefit of codon bias We define the translation rate benefit of codon bias in E. coli as s bias , the fractional increase in the time required to replicate the E. coli proteome if the actual codon bias of E. coli were to be replaced with uniform use of synonymous codons (Equation (10) in Methods). Our estimates of s bias depend on the relative translation rates of individual codons in vivo , and on the frequency with which each codon is used in synthesizing the proteome. The sources we have used for these data, and the details of several adjustments made to the source data, are described in the Methods section. All data used in our estimates of s bias are presented in Table 2 . Because the codon-specific translation rate data are incomplete even for E. coli , we have explored 4 different scenarios (described in Methods) for extrapolating from the empirical rate data to obtain an estimate of s bias over all codons. Scenarios 1–4 are increasingly complex, and represent deliberate attempts to assign translation rates to the unmeasured codons in a way that increases s bias while remaining consistent with patterns found in the empirical data. Furthermore, in Scenario 5, we apply a theoretical approach [ 62 ] for predicting optimal codon-specific translation rates that does not rely on empirical translation rate measurements at all, but only on codon frequency and tRNA abundance data. Estimates of s bias for all scenarios are presented in Figure 1 . The empirical translation rate data used in Scenarios 1–4 reflect ternary complex selection at the ribosomal A-site, but not translocation of the newly-formed peptidyl-tRNA from the A-site to the P-site [ 45 ]. Thus, for these scenarios we show two estimates of s bias that are based on different assumptions regarding the relative duration of translocation and ternary complex selection. The white bars of Figure 1 are based on the assumption that the duration of translocation is negligible for all codons in comparison to the duration of ternary complex selection. The cross-hatched bars of Figure 1 are based on the assumption that translocation requires a finite amount of time that is constant for all codons, but short in comparison to the time required for ternary complex selection [ 63 ]. In Scenario 5 the duration of translocation is not treated explicitly, but the theoretical rate predictions refer to the entire cycle of translational elongation. Hence, we have grouped the estimate from Scenario 5 with other estimates that account for the duration of translocation. Our estimates of the benefit of codon bias in E. coli relative to the complete absence of codon bias range from 0.6 – 1.4 if translocation time is neglected, or from 0.4 – 1.1 with the more realistic assumption that translocation requires a short amount of time. We have also estimated the benefit of codon bias in E. coli relative to the limited degree of codon bias that might be found in an actual low-bias organism, rather than making a comparison to the biologically unrealistic standard of strictly uniform synonymous codon use. We took T. pallidum as our example of a microbe with limited codon bias, since it is a slowly growing bacterium with little mutational bias (52.7% G+C) that has also been reported to lack translational selection [ 54 ]. The T. pallidum genome has the second-most uniform codon use over all predicted genes (assessed as Wright's effective number of codons [ 64 ]) among 108 bacterial and archaeal species for which complete genome sequences were available in June, 2003 (data not shown). Our method for generating a set of low bias codon frequencies from T. pallidum genome codon frequencies is described in Methods. Estimates of the translation rate benefit of codon bias for E. coli relative to low bias codon frequencies are shown by the black bars of Figure 1 , again assuming a short, invariant duration of translocation. The estimated benefits range from 0.2 – 0.6; as expected, these estimates are smaller than estimates derived from a comparison to strictly uniform codon usage. Because the theoretical estimates of Scenario 5 fall in the middle of the corresponding ranges of empirical estimates from Scenarios 1–4, we are confident that our results are not merely an artifact of unrecognized errors in the empirical rate measurements. The benefit of codon bias calculated for individual amino acids Our definition of s bias can be applied over any subset of codons, in particular, it can be applied to the codons of each amino acid separately. While all amino acids with multiple codons except proline contribute positively to s bias in all scenarios, the magnitude of that contribution is highly variable between amino acids (Figure 2 ). Codon bias accelerates the translation of most amino acids only slightly in E. coli , because most non-preferred codons are not particularly rare in the E. coli proteome, compared to the preferred synonym. For example, among the 9 amino acids encoded by 2 codons, on average the preferred codon is 2.9-fold more abundant than the non-preferred codon. Of these amino acids, asparagine shows the greatest difference between preferred and non-preferred codon frequencies, with GAC being 5.2-fold more abundant than GAU. Even if the disparity in codon-specific translation rates is unrealistically large, the ratio of the frequencies of preferred to non-preferred codons in E. coli constrains the maximum possible value of s bias . For asparagine, even if the preferred codon were translated instantaneously (i.e., infinitely faster than the non-preferred codon), the difference between using the non-preferred codon at 16% of asparagine residues in E. coli instead of at 50% of asparagine residues in Uni corresponds to only about a 3-fold acceleration of translation ( s bias ≈ 2) for this amino acid. With more realistic disparities between the translation rates of preferred and non-preferred codons, the largest estimate of s bias for asparagine in any of our scenarios is less than 0.2. In other words, we estimate that codon bias in E. coli leads to no more than a 20% decrease in the time required to translate all asparagine codons in the proteome (Figure 2 ). The amino acids with the largest values of s bias are leucine, isoleucine, and arginine (Figure 2 ). Although these amino acids are not rare, they possess between them the six rarest codons in E. coli , each encoding less than 0.1% of the proteome. (An average codon encodes 1.6% of the proteome.) The frequencies of the most and the least abundant synonyms for leucine, isoleucine and arginine differ by 74-fold, 83-fold, and 1460-fold, respectively. (The higher ratio for arginine reflects the extreme rarity of AGG, which is 17-fold less abundant than the second rarest E. coli codon, AUA encoding isoleucine.) Since the translation rates measured or assumed for the 6 rarest codons are quite slow, their increased abundance in Uni accounts for the much of the additional time required for replicating the Uni proteome. If these six codons remained as rare in Uni as they are in E. coli , while all other synonymous codons were used without bias in Uni, the translation rate benefit estimated under Scenario 4 (the scenario producing the largest estimate of s bias ) would be reduced by almost half (data not shown). The influence of these 6 codons is such that the estimate of s bias is quite sensitive to the translation rates assigned to them, in contrast to the relative insensitivity of s bias to the exact translation rates assigned to most codons. Discussion We want to know whether reduced codon bias could account for the lower translational power measured in at least some slowly growing bacteria, in comparison to E. coli . We approach this issue by its converse, calculating how much faster the proteome is replicated in E. coli than it would be in the complete absence of codon bias. If we take our estimates at face value, we would conclude that even during rapid growth when the proteome is most biased and translation is fastest, s bias is unlikely to be much larger than 1 (cross-hatched bars of Figure 1 ), which corresponds to a 2-fold increase in the average translation rate. An effect of this magnitude approaches the smaller disparities in the comparisons of translational power between E. coli and slowly growing strains shown in Table 1 , but could not explain the roughly 5-fold difference in translational power between E. coli and S. alaskensis , R. prowazekii , Halobacterium cutirubrum , or sulfate-reducing strain PT2. However, there are two reasons to think that the benefit of codon bias for E. coli , in comparison to most actual slow-growing organisms, is even less than this estimate. The first reason is that we have prevented our hypothetical Uni from adapting to the codon frequencies we have assigned to it, by keeping the abundance of each component of the translational apparatus fixed. The data do not suggest that maximizing translational power has been the only selective pressure influencing codon use in E. coli [ 45 , 65 ]. If it had been, the codon with the highest rate constant for ternary complex selection among synonymous alternatives would always be the preferred codon, since it would permit faster translation with a lower biomass investment in cognate tRNA. Of 10 amino acids with multiple codons for which codon-specific translation rate measurements exist [ 44 , 45 ], leucine, serine and proline are not consistent with this prediction. On the other hand, it seems clear that selection for rapid translation has exerted some, and perhaps the major influence on the coevolution of codon frequencies and tRNA abundance in E. coli . The codon with the highest rate constant is the preferred codon for 7 of the 10 amino acids for which data are available. Other considerations (possibly including error avoidance [ 66 ], interactions between adjacent tRNA anticodons [ 67 ], or factors unrelated to translation [ 68 ]) may have been more influential than the inherent characteristics of the codon-anticodon interactions for determining the preferred codons encoding leucine, serine and proline. However, the importance of rapid translation remains evident in that E. coli still translates the preferred codons quickly for 2 of these 3 amino acids, albeit with a larger investment in tRNA than would be necessary if the interaction between the preferred codon and its cognate tRNA occurred more readily. At a larger scale, the correlation across all codons between frequency and cognate tRNA abundance [ 69 , 70 ] is best explained as a response to selection for rapid translation, as is the pattern of increased bias towards rapidly translated codons with increased levels of gene expression [ 45 ]. Without asserting that the distribution of tRNA abundance in E. coli necessarily produces the fastest possible translation rate for the E. coli codon frequency distribution, it is clear that selection for translational power has been a significant factor in the co evolution of codon frequencies and cognate tRNA abundances in E. coli . Thus, it is very unlikely that we have attained the maximum possible translation rate for Uni by matching the E. coli distribution of tRNA abundance values (in the form of a particular distribution of codon-specific translation rates) to the very different codon frequency distribution of Uni. For this reason, our estimates confound the translation rate benefit of codon bias in E. coli with the penalty of a suboptimal allocation of translational resources in Uni. The second reason that our approach overstates the relative benefit of codon bias for E. coli in comparison to actual slow-growing organisms is that actual microbes are not completely devoid of codon bias. Assessing s bias in E. coli in comparison to a biologically plausible standard for low codon bias, instead of in comparison to the implausible standard of no codon bias whatsoever, reduces the estimated benefit in E. coli by about half (black bars of Figure 1 ). Only a slight bias in codon use is sufficient to obtain a substantial benefit of faster translation because only a few codons in E. coli are translated much more slowly than the median rate (Table 2 ). Moderate avoidance of only these few codons can provide a considerable acceleration of the average translation rate without generating a dramatic bias in overall codon use. Our estimate of a biologically plausible standard for low bias codon frequencies is deliberately conservative, underestimating the degree of bias expected in most slowly growing microbes, for two reasons. First, our low bias codon frequencies are based on the genome codon frequencies of T. pallidum , as if all predicted genes in the genome were expressed equally. Correspondence analysis performed at the level of individual genes failed to uncover evidence that codon use varies with expression level in T. pallidum [ 54 ]. If this were true, the proteome codon frequencies would indeed be similar to genome codon frequencies, regardless of variability in gene expression levels. However, a more sensitive analysis using codon frequencies summed over a set of putative high expression genes indicates that codon use in such genes is more biased than codon use in the genome as a whole. This conclusion is based on a comparison of Wright's effective number of codons [ 64 ] calculated for codon frequencies summed over all predicted genes annotated as ribosomal proteins or translation elongation factors (Nc = 52.7) or calculated for codon frequencies summed over all predicted genes in the genome (Nc = 55.2) [ 71 ]. The failure to observe this low level of codon bias in the previous analysis based on individual gene sequences [ 54 ] can probably be attributed to high gene-to-gene variability in codon frequency estimates based on the small samples of codons represented by individual genes. Thus, even for T. pallidum , the proteome codon frequencies appropriate for estimating the benefit of codon bias will be more biased than the genome-derived low bias codon frequencies shown in Table 2 . The second reason our low bias codon frequencies underestimate the degree of codon bias in most slowly growing microbes is that T. pallidum is essentially free of the influence of mutational bias, with a genome G+C content of 52.7%. In contrast, many slow-growing microbes have more extensive codon bias that can be attributed mostly or entirely to the biased nucleotide composition of the genome (e.g., R. prowazekii [ 52 ], H. pylori [ 55 ], Borrelia burgdorferi [ 54 ], Buchnera aphidicola [ 72 ], Mycoplasma genitalium [ 73 ], and Chlamydia species [ 74 ]). If codon bias derived from mutational bias, like codon bias derived from translational selection, permits more rapid translation for the same investment in translational machinery, the use of low bias codon frequencies derived from T. pallidum will underestimate the translation rate of many slow growing strains. We believe that codon bias derived from mutational bias does, indeed, have the potential to accelerate translation. The translation rate benefit of codon bias depends on matching preferred codons with cognate tRNAs that are abundant and/or that form 3 canonical base pairs [ 7 , 8 ]. Even when codon use is determined by mutational bias in the DNA replication and repair systems [ 46 ], not by selection acting simultaneously on codons and their cognate tRNAs via translation-associated effects, selection for translational power can influence the relative abundance and anticodon sequence of tRNA species. Relatively few mutations are sufficient to influence the identity and abundance of tRNA molecules in an organism, in comparison to the number of mutations required to influence proteome codon frequencies. (Consider that 45 mutations could allow a single mutation in the anticodon wobble position or in the regulatory region of many or even all tRNA genes, depending on the organism, while 45 mutations could alter the identity of less than 0.5% of the >9,000 codons in genes encoding ribosomal proteins and translational elongation factors.) Hence, the mutation-selection balance argument invoked to explain diminished codon bias in genes expressed at low levels in many strains [ 8 , 75 ] also suggests that the distribution of tRNAs can be influenced by translational selection that may be too weak to create a dramatic effect on codon usage. In fact, if codon use is biased in the same direction in all genes (as expected if the source of codon bias is mutational bias), instead of being biased only in highly expressed genes, it would increase the selective pressure for adaptation of the tRNA pool. Hence, it would be very surprising if the anticodons and the relative abundances of tRNA molecules in organisms with high or low G+C content did not reflect their biased use of codons. This prediction is confirmed by the only two studies we have found of tRNA abundance in microbes with extreme G+C content, involving Mycoplasma capricolum (25% G+C) [ 76 ] and Micrococcus luteus (74% G+C) [ 77 ]. M. capricolum , but not M. luteus , can be considered a constitutively slow-growing strain. As expected, cognate tRNA abundance in both organisms is correlated with codon frequency, both across all codons and within synonymous codon families [ 76 , 77 ]. For M. capricolum , this is accomplished largely without the tRNA gene dosage effects that are important for E. coli [ 70 ] and B. subtilis [ 78 ], since 28 of the 29 M. capricolum tRNA genes are present in only a single copy [ 76 ]. These examples indicate that selection for translational power is operative even for organisms in which the codon bias is determined by mutational bias instead of translational selection, and even for slowly growing organisms. Because codon bias from any source can be exploited to obtain higher translational power, the estimates of s bias for E. coli compared to codon frequencies derived from T. pallidum will overstate the benefit that exists for E. coli relative to most other slowly growing microbes that have greater mutational bias. In summary, we believe the translation rate benefit of codon bias in E. coli is likely to be less than 0.6 (see black bars of Figure 1 ) relative to an actual slow-growing organism that shows limited codon bias, such as T. pallidum , and substantially less than 0.6 relative to a slow-growing organism with more extensive codon bias. We do not mean to suggest that the advantage of translating as much as 60% faster than a competitor is unimportant. Clearly, the benefit of codon bias for E. coli must be substantial, considering that it arises from the aggregate effect of many thousands of preferred codons that are stably maintained in the E. coli genome, despite the randomizing influence of mutation acting at each individual codon. On the other hand, the influence of codon bias on the average translation rate is far smaller than the differences in translational power observed between microbes adapted to different ranges of growth rates. For differences in codon bias to explain the difference in translational power between E. coli and S. alaskensis , s bias would have to be about 5; to explain the difference between E. coli and R. prowazekii , s bias would have to be about 3. Is it possible that the comparisons of translational power presented in Table 1 are flawed? The colorimetric assays used for RNA and protein measurement in these studies are indeed dependent on procedural details, such that comparisons between laboratories and between studies are less reliable than comparisons within a study. Nonetheless, variation between species in the estimates of translational power presented in Table 1 do not appear to result simply from large random errors around a common mean. Estimates of translational power for slowly growing species with few rrn operons cluster around low values; the reverse is true for species capable of rapid growth with higher numbers of rrn operons. In addition, our own measurements of 10 bacterial species (including E. coli , S. alaskensis and 8 recent soil bacterial isolates) reproduce the same pattern; we have found differences in translational power that are comparable in magnitude to those shown in Table 1 [ 79 ]. Hence, we believe the comparisons in Table 1 are an adequate representation of the differences in translational power between rapidly growing and slowly growing microbes. Conclusions Because codon bias influences translational power, and because the degree of codon bias due to translational selection may differ systematically between rapidly growing and slowly growing strains, we investigated the parsimonious hypothesis that observed differences in translational power between microbial species could be explained by differences in the degree of codon bias. However, based on the analysis reported here, such an explanation is not plausible. Instead, differences in translational power between rapidly growing and slowly growing species suggest that the translational apparatus itself has different performance characteristics in rapidly growing and slowly growing microbes. Methods Translational power, translation rate and the active fraction of ribosomes Conceptually, we define translational power as the rate of protein synthesis in a cell or culture, normalized to the biomass invested in the protein synthesis system. We intend the term to be synonymous with 'translational efficiency' [ 4 , 5 , 8 ]; our rationale for departing from established terminology is provided in the Introduction. The protein synthesis system is comprised of ribosomes, elongation factors, tRNAs, tRNA synthetases, mRNAs, and numerous other components. Measuring the mass of the entire system is not trivial, because it includes a variable fraction of the cell's protein. However, since the protein synthesis system includes essentially all the cell's RNA, we follow Kjeldgaard and Kurland [ 1 ] in using RNA mass ( R ) as an index of the biomass invested in the entire system. For a culture in balanced, exponential growth, the instantaneous rate of increase of any culture component is d X /dt = μ X , where μ is the specific growth rate and X is the mass of the component present in the culture at that moment. Hence, μ P is the rate of protein synthesis in a culture containing mass P of protein. Thus, our quantitative measure of translational power is: This quantitative measure of translational power will be consistent with the conceptual definition as long as RNA is a nearly constant fraction of the mass of the entire protein synthesis system. Translational power reflects both the average translation rate and the fraction of active ribosomes in a cell or culture, which we demonstrate as follows, using the approach of chapter 6 of reference [ 34 ]. 'Translation rate' refers to the rate of amino acid polymerization of an active ribosome. The average translation rate of a cell or culture is the rate of amino acid polymerization in the entire culture divided by the total number of active ribosomes: We know that the mass rate of protein synthesis in a culture in balanced growth is μ P . Units of protein mass can be converted to a number of amino acids by dividing the protein mass by the average mass of an amino acid: number of amino acids polymerized per unit time = μ P /(average mass of amino acid) (3) The number of ribosomes in a culture containing a mass R of RNA can be found by multiplying R by the fraction of RNA that is ribosomal, and then dividing by the mass of RNA in a ribosome. However, only a fraction of these ribosomes are active at any given time. Thus: Substituting Equations (3) and (4) into Equation (2) yields: After rearranging terms in Equation (5), we have: where The quantity μ P/R in Equation (6) is the quantitative measure of translational power from Equation (1) [ 1 , 3 ]. From Equation (6), it is clear that translational power reflects both the average translation rate and the active fraction of ribosomes in a cell or culture. What of the term we have labeled C , implying a constant? The two quantities in the numerator, the mass of RNA in a ribosome and the average mass of an amino acid, are indeed constant or nearly constant, both within a strain at different growth rates, and across strains. However, despite the constant ribosomal fraction of RNA reported in reference [ 3 ], other data indicate that the rRNA fraction decreases from about 85% to about 75% as growth rate declines in E. coli from 1.7 hr -1 to 0.28 hr -1 [ 70 ], a result which is expected on theoretical grounds [ 4 , 65 ]. This variation is not dramatic; it would reduce translational power by only 12%, if the average translation rate and active fraction of ribosomes were unchanged. Data are also available from 2 of the 4 studies that have compared translational power between E. coli and a slowly growing strain. The rRNA fraction is reported as 84% for H. cutirubrum at specific growth rates of both 0.10 hr -1 and 0.05 hr -1 , after the authors made the deliberately generous assumption that messenger RNA comprises 5% of the total RNA [ 37 ]. The rRNA fraction is about 85% for R. prowazekii at a specific growth rate of ~0.07 hr -1 , after a correction is made for 2–3% messenger RNA [ 38 ]. These data suggest that variation between microbial species in the ribosomal fraction of RNA is limited, even when comparing species that grow at very different rates. Comparisons of translational power based on published data Table 1 summarizes comparisons of translational power between E. coli and all other bacterial and archaeal species for which we could find both the protein content and the RNA content of cultures growing at known rates. Throughout this work, E. coli is represented by the Bremer and Dennis data [ 3 ], which are typical of the data reported for E. coli in many other studies. Similarly, comparisons between E. coli and 2 closely related species of enteric bacteria, S. enterica and E. aerogenes , are made using only a single representative study for the latter strains, chosen from among several published reports. For the remaining species, only a single published study was available for comparison, except for one species represented by two studies, both of which are included. For strains not grown at 37°C, we assume that the growth rate, but not the macromolecular content, would be altered by growth in the same medium at a different temperature [ 80 ]. The growth rates reported for these strains were adjusted to the growth rates expected at 37°C using the linear range of the relationship reported in reference [ 56 ]. (Although this temperature-growth rate relationship was generated with E. coli , the comparison is mathematically identical whether the temperature correction is applied to E. coli or to the comparison strain.) The comparisons in Table 1 use the fastest growth rate for which data are available for the comparison organisms, and use data for E. coli growing at a rate such that it matches the comparison organism for investment in the translational apparatus. (For two of the comparison strains, translational power differed considerably between the fastest growth rates obtained in different culture conditions; both values are reported.) One of three measures was used to gauge the level of investment in the translational apparatus, depending on the quantity measured in the original study. The possible measures were the number of ribosomes per cell volume, the ratio of protein to ribosomal RNA, or the ratio of protein to total RNA. Values of these quantities for E. coli were interpolated between adjacent data points to estimate the growth rate at which E. coli made the same investment in the translational apparatus as the comparison organism. The translational power of the comparison organism at the fastest available growth rate was then expressed as a percentage of the translational power of E. coli at the 'same investment' growth rate. A comparison at similar investment levels reflects the expectation that the selective pressure to maximize translational power increases with the biomass invested in the apparatus [ 4 , 58 ]. If the comparisons had always been made to the fastest E. coli growth rate (i.e., where E. coli translational power is highest), the disparities in translational power would be greater for most of the comparisons shown. Calculation of the translation rate benefit of codon bias Consider a cell in which a total of C i codons of type i are translated during a single cell generation, so that the sum over all sense codons C = Σ C i is the total number of codons translated during a cell generation. (Hereafter we refer to the translational output over a cell generation as the proteome.) If we define c i = C i / C as the proportion of all codons of type i in the proteome and r i as the average translation rate of codons of type i , the total time required for replication of the proteome (i.e., the proteome generation time) will be where R # is the average number of ribosomes active in translation over the cell cycle and the sum is over all sense codons. Codon bias in favor of rapidly translated codons will reduce g p in comparison to uniform codon use. If a mutation changes the fitness of an organism from w to w ', the benefit of the mutation is typically described as s , where w '/ w = 1 + s . By analogy, and considering g p to be inversely related to fitness, we can express the translation rate benefit of codon bias as The protein content (and thus C ) is the same in Uni as in E. coli by hypothesis. With the restrictive condition that the abundance of each individual component of the translational apparatus is unchanged in Uni, ribosome content ( R # ) will be the same also. Hence, the C / R # term of g p in Equation (8) cancels from both the numerator and denominator of Equation (9) for s bias , leading to Since amino acid frequencies are identical in E. coli and Uni, the disparities in translation rates between synonymous codons largely determine the magnitude of the translation rate benefit of codon bias. We will use the same codon-specific translation rates (the r i 's) for both Uni and E. coli , again invoking the restrictive stipulation that the abundance of each individual tRNA species is unchanged. If rate constants for the interaction of each codon with each of its cognate tRNA species were known, we could calculate the optimal tRNA abundance distribution for the codon frequencies of Uni, and infer the resulting codon-specific translation rates [ 62 , 65 ]. However, in vivo codon-specific translation rate data are available only as codon averages, including translation from all tRNA species cognate to each codon. Hence, rate constants specific to each codon-cognate tRNA pair cannot be calculated from the available data for the codons translated by multiple tRNA species, and thus we cannot calculate an optimal tRNA abundance distribution for Uni. Instead, we have constrained Uni to maintain the same tRNA distribution and codon-specific translation rates as E. coli . Insofar as the E. coli rates reflect an allocation of tRNA abundance that would be sub-optimal for Uni (as we argue in the Discussion section), our approach will tend to overestimate of the benefit of codon bias in E. coli , a conservative error for our purposes. Data sources All data used in our estimates of s bias are reported in Table 2 . For the codon frequencies used in synthesizing the proteome of E. coli , we rely on the data of Dong et al . at a specific growth rate of 1.73 hr -1 [ 70 ], compiled from public gene sequence databases and protein abundance data derived from 2D gel electrophoresis studies [ 81 , 82 ]. The absolute codon frequencies shown in Table 2 have been recalculated from [ 70 ] with initiation and stop (including selenocysteine) codons removed. As expected, the translation rate benefit of codon bias was found to increase monotonically with growth rate, when calculated by any of the scenarios described below, using the proteome codon frequencies and tRNA abundance data from the range of growth rates reported in reference [ 70 ] (data not shown). This increase in s bias reflects simply the increasing bias in both proteome codon usage and relative tRNA abundance with increasing growth rate. Since we are interested in the maximum effect of codon bias, we report results from only the highest growth rate for which data are available. To investigate the importance of low levels of codon bias, we applied Equation (10) either with Uni having strictly uniform use of synonymous codons, or with Uni assigned a set of low bias codon frequencies (Table 2 ). The low bias frequencies were generated from relative codon frequencies over all predicted genes in the complete genome sequence of T. pallidum [ 71 ]. By relative codon frequencies, we mean the absolute frequency of a codon divided by absolute frequency of the amino acid it encodes. The set of T. pallidum relative codon frequencies for a particular amino acid were multiplied by the absolute frequency of that amino acid in the E. coli proteome; the resulting set of absolute codon frequency values were assigned to the codons of that amino acid in the low bias set so as to retain the same rank order of codon frequency among synonyms as exists in the E. coli proteome. For example, the absolute frequency of isoleucine and the identity of the 1st, 2nd and 3rd most common isoleucine codons are the same in the low bias set as in the E. coli proteome. However, the relative frequencies of the 1st, 2nd and 3rd most common isoleucine codons in the low bias set are the same as the relative frequencies of the 1st, 2nd and 3rd most common isoleucine codons in the T. pallidum genome. To represent codon-specific translation rates, we use the relative rate data (the quantity R tRNA /R shift ) of Curran and Yarus [ 45 ] for the 29 sense codons beginning with U or C (YNN codons, Y = pyrimidine). Although incomplete, this is by far the largest data set available for in vivo translational kinetics. The original publication transposed values reported for two arginine codons, CGC and CGA [ 83 ]; we have corrected this error. We also revised the rate measured for CGA downward, to account for interference from the bulky wobble position inosine-adenine base pair in the P site that results from translation of a CGA codon. Such interference is strongly suggested to slow selection of a ternary complex at the codon subsequent to CGA [ 83 ]; such an effect would not have been measured with the experimental system of reference [ 45 ], but is appropriate to include as a codon-specific effect of CGA on translation rate. In the absence of more precise data, we reduced the translation rate measured for CGA by a factor of 3, the factor by which CGA reduces read-through of a following stop codon by a suppressor tRNA in comparison to CGC [ 83 ]. This adjustment to the CGA rate brings these results into rough agreement with those of Sorensen and Pedersen [ 84 ], who used an experimental approach that would have detected a consistent effect of CGA on the translation rate of the subsequent codon, attributing it to slow translation of CGA itself. The relative rates of reference [ 45 ], modified as described above, are listed in Table 2 . The relative rates reported by Curran and Yarus [ 45 ] do not reflect the entire translational cycle, but rather the time required for selecting a cognate ternary complex at an empty, codon-programmed ribosomal A site, which is believed to occupy the majority of the elongational cycle [ 63 ]. Although peptide bond formation may be very rapid, the time required for the EF-G-catalyzed translocation of the ribosome to the subsequent codon (and the associated movement of P- and A-site tRNAs) may not be much shorter than the time needed for EF-Tu-catalyzed ternary complex selection [ 63 ]. Hence, in addition to calculations made using rates of ternary complex selection to represent an entire cycle of translational elongation (assuming, in effect, that the duration of translocation is negligible), we also made calculations after modifying the reported rates by adding an invariant 'translocation time' to the variable 'ternary complex selection time' for all codons. The duration of translocation per codon was set at 40% of the average time required to select a ternary complex containing tRNA phe at a UUU codon, consistent with the only quantitative measure of translocation rate that has been made in conditions approximating those in vivo [ 63 ]. Results from both sets of calculations (white and cross-hatched bars of Figure 1 ) are presented for each scenario (described below) that is based on these ternary complex selection rates. For convenience, elsewhere in this report we refer to the relative rates of reference [ 45 ] as translation rates, rather than using the more accurate but cumbersome expression 'ternary complex selection rates'. To calculate the total abundance of cognate tRNA for each codon, we assign cognate specificity largely according to Björk [ 85 ], and use the tRNA abundance data from Dong et al . [ 70 ]. We differ from Björk only in assuming that the leucine and glycine tRNAs with uridine in the anticodon wobble position (for which nucleotide modifications have not been characterized) will read codons ending in U, A and G, instead of A and G only. This would be the case if the wobble position U is modified to cmO 5 U, as is done for each of the other 6 amino acids encoded by a full box of the translational code (i.e., amino acids for which the four XXN codons are synonyms). Following Björk, we assume that 40% of the tRNAs for glutamate, glutamine and lysine with uridine in the anticodon wobble position are modified to mnm 5 Se 2 U and thus read codons ending in A or G; the balance of these tRNA species are assumed to have mnm 5 S 2 U in the wobble position and read A-ending codons only [ 85 ]. The abundance of two pairs of isoaccepting tRNA species (Gln1 + Gln2 and Ile1 + Ile2) were reported as summed values by Dong et al . [ 70 ], since these individual species were not separated under the experimental conditions applied. We have resolved the summed values to the abundance of individual species using the ratios of the individual abundance values reported by Ikemura [ 69 ]. We show cognate tRNA abundance data in Table 2 as a percentage of total tRNA, omitting initiator and selenocysteine tRNAs; the sum of all values is greater than 100%, reflecting the partially overlapping specificity of many tRNA species. Scenarios for extrapolating from incomplete empirical translation rate data We address the incompleteness of codon-specific translation rate data in several ways. In Scenario 1, we assume that the effects of biased use of YNN codons on translation rate can be used to represent the effects of bias over all codons, without assigning particular translation rates to the unmeasured codons. However, since the YNN codons are almost half of all sense codons but only account for about a third of all expression (Table 2 ), they must be less highly expressed, on average, than the RNN codons (R = purine). Consequently, selection for translational power may have been weaker among YNN codons than RNN codons. Scenarios 2–4 address this potential deficiency by applying various strategies of assigning translation rates to the unmeasured codons that are consistent with observed patterns, but that could allow the effect of codon bias on translation rate to be greater among RNN codon than YNN codons. Scenario 5 abandons empirical codon-specific translation rate measurements completely, assigning translation rates to all codons on the basis of the proteome codon frequency and cognate tRNA abundance of E. coli , assuming optimality (i.e., maximal translation rate) according to theory developed by Solomovici et al . [ 62 ]. Scenario 1 The 29 YNN codons encode 10 amino acids, 9 of which have multiple codons. For 7 of these 9 amino acids, the most common synonym is the codon with the fastest translation rate. One of the remaining amino acids is serine, for which the two fastest-translated codons are the two most abundant, although in reverse order, with relatively small differences between the two in both rate and abundance. Only proline appears to be anomalous; the 2 most abundant codons encode over 90% of all proline residues in the proteome [ 70 ], but support ternary complex selection about 3.5-fold more slowly than the 2 least abundant codons [ 45 ]. It has been suggested [ 45 ] that this anomaly could be adaptive; if proline, because of its unique structure, is found preferentially between protein domains [ 86 ] where slow translation may be important to permit cotranslational folding [ 87 , 88 ]. If proline is the only amino acid for which such contrarian selection pressure is more important than selection for translational power, including proline codons in a sample intended to represent all codons will lead to an underestimate of s bias . Hence, in Scenario 1 we apply Equation (10) over YNN codons, with the calculated translation time for non-proline YNN codons weighted by a factor of 3.2, which scales the expression level of these codons to the expression level of all non-proline codons. In other words, we assume the effects of codon bias on translation rate among the 25 non-proline YNN sense codons are representative of the effects of codon bias among all 57 non-proline sense codons, whereas the translation rates measured for proline codons are applied only to themselves. Scenario 2 Curran and Yarus noted that among highly expressed genes, there is a significant tendency for rapidly-translated codons to be used frequently, although the relationship appears to be nonlinear [ 45 ]. We observe the same pattern comparing their relative rate data to the proteome codon frequency data of Dong et al . [ 70 ] at the highest growth rate. For non-proline YNN codons, the best fit (R 2 = 0.56) of a quadratic relationship passing through the origin between the codon frequency and translation rate data of Table 2 is c i = 0.205 r i - 0.522 r i 2 . We use this equation to predict translation rates from codon frequency for all RNN codons, as shown in Table 2 . Since our objective is to obtain a reasonable estimate the codon-specific translation rate for codons which have not been measured, not to defend a particular model of the relationship between codon frequency and translation rate, we make no attempt to justify a quadratic fit in comparison to other possible functional relationships. The predicted rates for RNN codons and the measured rates for YNN codons (Table 2 ) are used with Equation (10) to estimate the translation rate benefit of codon bias under Scenario 2. Scenario 3 The preceding scenario applied to the YNN codons tends to predict translation rates among synonymous alternatives that are not as disparate as those actually observed. Furthermore, the fit of a functional relationship between codon frequency and translation rate among YNN codons is better when only preferred codons are considered, instead of all codons. Hence, we fit a quadratic relationship passing through the origin to data from 10 preferred non-proline YNN codons, obtaining c i = 0.352 r i - 1.611 r i 2 (R 2 = 0.81). Among the 10 preferred codons, we include UGG, the sole tryptophan codon, and UUG, the preferred leucine codon within the UUR split box although not the preferred leucine codon overall. We then apply this equation to predict translation rates from codon frequencies for 12 preferred RNN codons, including AUG, the sole methionine codon, and AGG and AGC, the preferred arginine and serine codons within their respective split boxes, although not the preferred codons overall. For non-preferred RNN codons, translation rate is predicted by multiplying the predicted rate for the preferred synonym (within the full or split box) by the ratio of the square roots of the codon frequencies for the non-preferred and preferred codons: This relationship was chosen both because a dependence on the square root of codon frequency has been suggested repeatedly in theoretical investigations of optimal translation rates [ 62 , 65 , 89 , 90 ], and because for all non-preferred RNN codons, this relationship leads to a greater disparity of predicted translation rates compared to the preferred synonym than the regression of Scenario 2. (It also predicts a greater translation rate disparity than is observed for the majority of non-preferred YNN codons.) When both the quadratic regression for preferred codons and Equation (11) for non-preferred codons are applied to predict the translation rate of non-proline YNN codons, the correlation of predicted with measured translation rates is comparable to that attained with Scenario 2 (R 2 = 0.57). The predicted rates for RNN codons and the measured rates for YNN codons (Table 2 ) are used with Equation (10) to estimate the translation rate benefit of codon bias under Scenario 3. Scenario 4 This scenario is generated in three steps, with the goal of generating an estimate of the translation rate benefit of codon bias that is consistent with the most extreme empirical observations. First, three rare RNN codons (AGG and AGA for arginine and AUA for isoleucine, all with c i < 0.1%) are assigned the slowest relative translation rate observed among YNN codons ( r i = 0.6 for the rare leucine codon CUA). Second, the translation rates for preferred RNN codons within full or split boxes (except AGG) are estimated according to the regression equation described for Scenario 3. Finally, the translation rates for non-preferred codons (except AGA and AUA) are predicted from the preferred synonym using the ratios of the most disparate translation rates observed empirically among synonymous alternatives, treating split boxes and full boxes of the translational code separately. The most extreme ratio observed among translation rates in a split box is 3.375, for glutamate codons in the study of Sorensen and Pedersen [ 84 ]. The most extreme ratios observed for translation rates of codons in a full box is 1:1.3:1.6:24 for the CUN leucine codons in the study of Curran and Yarus [ 45 ]. (Exploring other rate values 1 ≤ x ≤ y ≤ 24 in ratios of the form 1: x : y :24 failed to find any that greatly increased the estimated benefit beyond that using the leucine ratios, data not shown.) Although this scenario is based on extreme observations, applying these 3 rules to the non-proline YNN codons leads to a correlation of predicted and measured translation rates (R 2 = 0.67) somewhat better than that obtained under Scenario 2 or Scenario 3. The predicted rates for RNN codons and the measured rates for YNN codons (Table 2 ) are used with Equation (10) to estimate the translation rate benefit of codon bias under Scenario 4. Scenario 5 In contrast to the preceding scenarios that extend codon-specific translation rate measurements of YNN codons in various ways to make estimates of the effect of codon bias over all codons, Scenario 5 incorporates a theoretical prediction of the optimal translation rates for all codons based only on codon frequency and cognate tRNA abundance data. While this approach necessarily involves additional assumptions, it has the advantage of drawing on data that is more complete and less likely to be influenced by unrecognized experimental errors. Solomovici et al . [ 62 ] assume that selection on synonymous codon frequencies reflects intrinsic differences in rate constants for a cognate tRNA interacting with preferred and non-preferred codons, while the total tRNA abundance and amino acid composition are fixed. They demonstrate that the fastest overall translation rate is obtained when the square roots of synonymous codon frequencies are proportional to the rate constants for cognate tRNA interacting with the codons. They assume further that the rate constants for the interaction of all non-degenerate or preferred codons with their preferred cognate tRNA are identical, so the translation rate for these codons is proportional to cognate tRNA abundance. We modified the approach of reference [ 62 ] to reflect greater degeneracy in translation than assumed by the original authors ([ 85 ], also the comments earlier in this section), and applied it using the codon frequency and tRNA abundance data of Dong et al . [ 70 ], modified as shown in Table 2 . The predicted relative translation rates for YNN codons (i.e., the recalculated quantities d ij and d im, j of reference [ 62 ] for codons with single or multiple cognate tRNAs, respectively) are not in good agreement with observed relative rates of Curran and Yarus [ 45 ] (R 2 = 0.30). However, the empirical codon frequencies of Dong et al . [ 70 ] are correlated more closely with predicted relative rates of Scenario 5 (R 2 = 0.70) than with the empirical relative rates of Curran and Yarus [ 45 ] (R 2 = 0.31). A good correlation between the predicted translation rates and the empirical codon frequencies is expected, since the codon frequencies were used to generate the predictions. However, the poor correlation between predicted and empirical translation rates could reflect the inadequacies in any of 3 areas: 1) the assumptions of Solomovici et al . [ 62 ], 2) the rate measurements of Curran and Yarus [ 45 ], and/or 3) the codon and tRNA data of Dong et al . [ 70 ]. Alternatively, the discrepancy between predicted optimal translation rates and empirical rates may indicate that the phenotype of E. coli is not perfectly optimized for maximal translation rates (as suggested in reference [ 65 ]), either because of genetic drift or because of conflicting selection pressures. Nonetheless, the disparity between the relative rates of synonymous preferred and non-preferred codons for most amino acids are greater with the predicted rates of Scenario 5 than with the observed rates. Hence, Scenario 5 will generate a higher estimate of the translation rate benefit of codon bias than would a strict application of the empirical codon-specific translation rates. (In fact, none of our scenarios are strict applications of the empirical rates; Scenarios 1–4 also deliberately extrapolate from the empirical rates in ways that will increase the estimated benefit of codon bias.) The predicted translation rates for all codons (Table 2 ) are used with Equation (10) to estimate the translation rate benefit of codon bias under Scenario 5. Authors' contributions LD conceived of the project, collected and analyzed the data, developed the mathematical model, and drafted the manuscript. TMS helped plan the project, critiqued the work as it progressed, and edited the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546186.xml
524515	A step ahead: combining protein purification and correct folding selection	The success of recombinant protein expression seems unpredictable and even good yields of soluble proteins do not guarantee the correct folding. The search for soluble constructs can be performed by exploiting libraries and speeded up by automation, but these approaches are money and time consuming and the tags used for affinity purification can mask the real stability of the target proteins. The ideal purification protocol would include the structure quality control. A recent paper commented in this article describes a phage-display method to screen for antibodies that are able to re-fold after heat-denaturation and can be selectively affinity-purified only if monodispersed. It turned out that the proteins with high recovery performance after heat-shock were also suitable for efficient recombinant expression.	Introduction The possibility to produce recombinant proteins instead of recovering the native molecules offers the double advantage of higher yields and of a simplified purification protocol using affinity chromatography. At least half a dozen of the purification tags that have been proposed so far are routinely fused to the target proteins and used to perform affinity purification. E. coli is the most popular host for the expression of heterologous proteins but its simplified cell organization can be limiting for the expression of correctly folded recombinant proteins. No bioinformatic tools can predict if a construct will be expressed soluble in bacteria and, therefore, time-consuming cloning steps and expression optimization tests must be considered. In most of the cases the affinity purification protocols are effective. However, the costs of the resins and proteases necessary to remove the tags can become a limiting factor. Furthermore, the different requirements for the chromatography steps and proteolysis make difficult to conceive automatic systems for obtaining purified homogeneous protein. As an alternative, we showed that the fusion of a target protein with a thermostable partner can be purified to homogeneity by heating [ 1 ]. Because the recovered target proteins resulted correctly folded only in some cases the method seems rather suitable for the preparation of antigens than for functionally active molecules. We do not expect that heated proteins recover the native structural features but it is a common simplification to assume that a soluble protein is correctly folded and companies commercialize vectors with tags that "improve the solubility". Completely underestimated is the fact that a very soluble fusion partner can keep in solution unfolded target proteins. The work of the group around Travé showed the false results generated using fusion proteins and suggested a method for the evaluation of the aggregation state [ 2 , 3 ] and to consider the monodispersity as required parameter. Therefore, methods would be envisaged that combine the purification to the selection for the correct folding. An important contribution in this direction is the recent paper published by Greg Winter's group [ 4 ]. Discussion The authors [ 4 ] describe a method for selecting antibody heavy chain variable regions resistant to the heat-induced aggregation. The antibody domains were displayed at the tip of filamentous bacteriophage and recovered by affinity binding to Protein A or to the specific antigen after the heating step. The purification was dependent on the correct folding of the antibodies since aggregates did not bind to the ligand. Once expressed in bacteria the selected antibodies showed a high yield and the property of reversible unfolding. In conclusion, the selection for the feature "re-folding from denatured/aggregation state" enabled the isolation of constructs adapted to recombinant expression. The results suggest that for the selected proteins the mechanisms leading to the re-folding into the native state are common to those that organize the folding of linear amino acid chains. Functional genomic relies on the possibility of screening fast and efficiently large number of clones for their expression and correct folding. Several approaches have been suggested over the past years. An indirect method considers marker genes activated by misfolding [ 5 ] to discriminate aggregation-prone constructs. Otherwise, the solubility of reporter fusion partners has been considered [ 6 , 7 ]. Nevertheless, as well as for fusions with MBP or GST the correct folding of the fusion partner does not automatically mean that the target protein reached its native structure. The elegance of the method described by Winter and co-workers relies on the use of a ligand that recognizes only the folded state of the protein to purify: the quality control is inclusive in the affinity purification. Furthermore, the phage-display format allows for the identification of the corresponding clone. As pointed out by the authors, such an approach is limited to those cases for which a conformation-dependent bait is available or, at least, a reliable method exists to discriminate between native and aggregated proteins. The logic of the experiment reminds me to the protocol used to select in vivo , directly and exclusively, for the class of conventional antibodies able to fold in the cytoplasm [ 8 ]. It is difficult to envisage a method applicable to all protein classes for selecting constructs that will express correctly folded proteins. Nevertheless, it is still possible to improve the yield of recombinant proteins that tend to aggregate. At least part of the unfolded proteins is not definitely trapped in aggregates. Re-solubilisation from bacterial inclusion bodies happens in vivo [ 9 ], the involvement of specific chaperones in the disaggregation process has been illustrated [ 10 , 11 ] and we used their co-expression to boost the bacteria re-folding machinery and increase the recombinant target protein yield [ 12 ]. Our unpublished data show that the chaperone-dependent solubilised protein is correctly folded, namely the recombinant chaperones are integrated into the in vivo protein quality control. Conclusions The recombinant expression of proteins often induces the formation of soluble aggregates and several of such aggregates conserve sufficient features for being recovered by affinity chromatography. The simplification of having considered a purified soluble protein as a protein in native state has generated false results [ 2 ]. Therefore, methods that enable to selectively purify only correctly folded proteins [ 4 ] are welcome because couple purification and quality control. Unfortunately, their application is limited to few single protein classes for which a suitable binder exists. One important information of the Winter's group article is that the protein recovery after heat-shock, namely the possibility to re-fold correctly, correlates with correct folding in recombinant expression. Since the measurements of the aggregation index needs very small amounts of proteins [ 3 ] it would maybe worthy to screen using the aggregation index of heated domains to check if the Winter's group observation is a general rule and its application useful to select potentially soluble constructs in absence of specific binders. Finally, the heat-selection can provide useful insights about the molecular features involved in the (re)-folding/disaggregation mechanisms.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524515.xml
524501	The pivotal role of cholesterol absorption inhibitors in the management of dyslipidemia	Elevated low-density lipoprotein (LDL)-cholesterol is associated with a significantly increased risk of coronary heart disease. Ezetimibe is the first member of a new class of selective cholesterol absorption inhibitors. It impairs the intestinal reabsorption of both dietary and hepatically excreted biliary cholesterol. Ezetimibe is an effective and safe agent for lowering LDL-C and non HDL-C. Short term clinical trials have established the role of ezetimibe monotherapy and its use in combination with statins. Furthermore, ezetimibe and statin combination therapy increased the percentage of patients who achieved their LDL-C treatment goal. Studies using surrogate markers of atherosclerosis have suggested a possible role of ezetimibe in combating atherosclerosis. Ezetimibe provides an effective therapeutic strategy for the management of homozygous familial hypercholesterolemia (HoFH) and sitosterolemia. The lack of outcomes and long term safety data is attributed to the relatively recent introduction of this medication.	Background Over 60 million Americans suffer from cardiovascular disease (CHD). The incidence of CHD and stroke has been on the rise partly because of the increase in life expectancy and the explosive epidemic of diabetes and the metabolic syndrome [ 1 ]. CHD is responsible for about 38% of the overall mortality in the United States making it the number one killer of Americans [ 2 ]. Animal and human studies have established the role of cholesterol in the development and progression of atherosclerosis. LDL-cholesterol (LDL-C) constitutes approximately 60–70 % of total serum cholesterol. Epidemiological studies directly implicated LDL-C to the development of atherosclerosis and CHD. Furthermore, LDL-C level appears to be directly related to the development and recurrence of CHD [ 3 ]. Animal studies suggested a protective effect of low LDL-C against atherosclerosis [ 2 ]. Multiple human trials examining the relationship of LDL-C lowering in primary and secondary prevention of CHD have demonstrated the impact of reducing LDL-C levels on decreasing CHD and CHD related mortality [ 4 - 8 ]. Most of the landmark CHD prevention trials involved the use of statin medications. LDL-C remains the primary target of treatment in most instances, and statins are the mainstay of LDL-C lowering treatment [ 9 ]. The National Cholesterol Education Program/ Adult Treatment Panel III (NCEP/ATP III) updated guidelines (table 1 ) for detection and treatment of dyslipidemia envisioned LDL-C below 100 mg/dL to be optimal for all patient risk categories. These more aggressive guidelines resulted in doubling of the number of patients that are not at target LDL-C levels as compared to previous guidelines [ 2 ]. Recent NCEP/ATP III update data suggested even lower LDL-C levels than previously advocated, making it harder to achieve the treatment in many instances and recommended the use of combination therapy if needed to help achieve the treatment targets. The NCEP/ATP III update emphasized "the lower, the better" hypothesis [ 10 ]. Table 1 Synopsis of the updated ATP III LDL-C Goals and Cut-points for TLC and Drug Therapy in Different Risk Categories and Proposed Modifications Based on Recent Clinical Trial Evidence Risk Category Goal TLC Drug Therapy High risk : CHD or CHD risk equivalents (10-year risk >20%) < 100 mg/dL (optional goal: <70 mg/dL) ≥ 100 mg/dL ≥ 100 (<100 mg/dL: consider drug options) Moderately high risk : 2+ risk factors (10-year risk 10% to 20%) < 130 mg/dL ≥ 130 mg/dL ≥ 130 mg/dL(100–129 mg/dL; consider drug options) Moderate risk : 2+ risk factors (10-year risk <10%) < 130 mg/dL ≥ 130 mg/dL ≥ 160 mg/dL Lower risk : 0–1 risk factor < 160 mg/dL ≥ 160 mg/dL ≥ 190 mg/dL (160–189 mg/dL: LDL-lowering drug optional) Cholesterol Absorption inhibitors Ezetimibe, a cholesterol absorption inhibitor, is the first agent of a new class of lipid-lowering compounds that selectively inhibits the intestinal absorption of cholesterol and related phytosterols. Ezetimibe undergoes extensive glucuronidation to an active metabolite in the intestinal mucosa [ 11 ]. Ezetimibe acts on brush border of the small intestine and decreases biliary and dietary cholesterol from the small intestine uptake into the enterocytes. Ezetimibe is primarily metabolized in the small intestine and liver via glucuronide conjugation with subsequent biliary and renal excretion [ 12 ]. Ezetimibe does not affect the absorption of fat-soluble vitamins, triglycerides, or bile acids [ 13 ]. After oral administration, ezetimibe is absorbed and extensively conjugated to a pharmacologically active phenolic glucuronide (ezetimibe-glucuronide) [ 14 ], the drug and its metabolite have a half-life of approximately 22 hours [ 8 ]. Concomitant food administration (high fat or non-fat meals) had no effect on the extent of absorption of ezetimibe when administered in the 10-mg clinical dose [ 15 ]. Ezetimibe and ezetimibe-glucuronide are highly bound (>90%) to human plasma proteins [ 16 ]. Plasma concentrations for total ezetimibe were about 2-fold higher in older individuals (>65 years), levels were similar in adolescents to healthy adults and may be higher in women than in men [ 8 ]. In patients with severe renal disease, ezetimibe level was increased approximately 1.5-fold, compared to healthy controls [ 17 ]. Ezetimibe had no significant effects on the bioavailability of warfarin, fenofibrate, HMG CoA reductase inhibitors, or digoxin [ 16 , 18 - 20 ]. Adverse experiences were reported in approximately 2% of patients treated with ezetimibe and included fatigue, arthralgia, diarrhea, abdominal pain and back pain. Angioedema and rash were reported after general clinical use of this medication [ 16 ]. With co-administration of ezetimibe and statins the adverse event profile was similar to that for statins alone. In a recently published case report, the authors described two patients whose creatinine kinase (CK) increased after the addition of ezetimibe to statin therapy causing one of the patients to experience myalgia and tendinopathy. This finding raises the question of whether ezetimibe can be implicated in precipitating increased risk of stain-associated myopathy [ 21 ]. The Role of Ezetimibe in Clinical Practice Indications for use 1. Primary Hypercholesterolemia (heterozygous familial and non-familial), Ezetimibe is indicated in this case for use as both mono and combination therapy. 2. The reduction of elevated total-C and LDL-C levels in patients with homozygous familial hypercholesterolemia (HoFH) either as primary or as an adjunct to other lipid-lowering treatments. 3. In patients suffering from Homozygous Sitosterolemia, as adjunctive therapy to diet for the reduction of elevated sitosterol and campesterol levels in patients with homozygous familial sitosterolemia. Monotherapy Multiple studies conducted to examine the effects of ezetimibe monotherapy have concluded that this drug was effective in lowering LDL-C versus placebo. Analysis of multicenter, double-blind, placebo-controlled trials demonstrated that ezetimibe at the 10 mg once daily clinically approved dose significantly modified cholesterol and cholesterol subtypes in patients with hypercholesterolemia when compared to placebo. Ezetimibe significantly lowered total-Cholesterol (TC) (12 %), LDL-C (18 %), apolipoprotein B (Apo B) (15 %), and triglycerides (TG) (7%) and increased high density lipoprotein (HDL-C) (3.5%) [ 22 - 24 ]. Lipoprotein (a) [Lp (a)] was not significantly affected by Ezetimibe 10 mg once a day treatment [ 25 ]. In a case series report, we analyzed the effects of Ezetimibe on cholesterol particle size and number using NMR technology (Lipo science, Raleigh, NC). We found that Ezetimibe lowered cholesterol particle number by a mean 26 % and had no significant effect on cholesterol particle size [ 26 ]. The effects of ezetimibe on cholesterol and its subtypes were not influenced by risk-factor status, gender, age, race, time of administration, or baseline lipid profile [ 22 ]. The overall incidence of adverse effects with ezetimibe monotherapy was similar to placebo. Combination therapy The struggle to achieve the NCEP/ATP III guidelines LDL-C goals through primary utilization of statins is often frustrating for the clinician. In a study to examine the efficacy of statin titration on attainment of LDL-C goal, the authors concluded that for high risk patients, approximately half were able to achieve their LDL-C goal at the appropriate statin starting dose, and only one third of the titration group were able to achieve the NCEP/ATP III cholesterol goal [ 27 ]. Now with the very recent publication of the update to the NCEP/ATP III, clinicians are faced with even lower goals of LDL-C, Making combination therapy a must in more cases than previously advocated by the NCEP/ATP III [ 10 ]. HMG-CoA reductase inhibitors (statins) act on the rate-limiting step to inhibit HMG-CoA conversion to mevalonate, effectively decreasing LDL-C synthesis. They result in a decrease of LDL-C ranging between 30–60 %, depending on the individual statin and the dose administered. Statin induced LDL-C lowering appears to be effective in reducing CHD and CHD related mortality and morbidity. The extent of CHD and CHD related events reduction is proportionate to the extent of LDL-C reduction [ 10 , 28 ]. LDL-C reduction trials have demonstrated a reduction in CHD related events by approximately 20–40% [ 4 - 8 ]. Reasons to initiate combination therapy to treat hyperlipidemia include: further LDL-C lowering, reducing side effects related to higher doses of statins, modifying other risk factors besides LDL-C such as HDL-C and TG. Increasing the dose of statins has a limited effect on reducing LDL-C, as it is well established that doubling the dose of a statins leads to a 5 % more reduction in total TC and 7 % more reduction in LDL-C with each doubling [ 29 ]. Although statins have demonstrated similarity in CHD related events, they are heterogeneous not only in LDL-C lowering efficacy but also in their safety profiles. The bulk of the statins effect on LDL-C occurs at the initial recommended dose and they are safer when used at doses below the maximal recommended dose. Statins are the most effective drugs known to modify LDL-C, but in terms of HDL-C and TG modifying capacity, other classes of lipid lowering medications used alone or in combination with statins offer higher efficacy. The metabolism of cholesterol is an intricate process that involves both produced and ingested cholesterol. The mechanism of action of HMG-CoA reductase inhibitors affects the production of cholesterol, whereas that of cholesterol absorption inhibitors affects absorbed cholesterol, thereby offering potential synergism of action when the medication are used in combination. Trials examining the efficacy have demonstrated synergism and consistency in LDL-C lowering in the absence pharmacokinetic interaction between the statins and ezetimibe. In a relatively large, multicenter study, involving patients with primary hypercholesterolemia already receiving statin monotherapy (but who had not met their NCEP ATP II target LDL-C goal), patients were randomized to receive either ezetimibe or placebo in addition to their current statin therapy. At the conclusion of this 8 week study, the ezetimibe and statin groups were found to have a significantly lower total-C, LDL-C, Apo B, and TG, and increased HDL-C when compared to the statin only and placebo groups. Furthermore, LDL-C reductions induced by ezetimibe were generally consistent across all statins groups [ 30 ]. Another multicenter, double-blind, randomized trial examined the effects of ezetimibe on patients suffering from (HoFH). At the initiation of the trial patients were receiving either atorvastatin or simvastatin. The addition of ezetimibe reduced LDL-C by an additional 20.5 % in contrast to only 6.7 % reduction that resulted from doubling the statin dose [ 31 ]. Similar results were demonstrated in high-risk patients with familial heterozygous hypercholesterolemia (HeFH) [ 32 ]. The addition of ezetimibe to statins is superior to treatment with statins alone in lowering non-HDL-C, ezetimibe co-administered with simvastatin lowered non-HDL-C by 47.1% whereas, simvastatin monotherapy lowered non-HDL-C by 33.6% when results were pooled across different doses [ 33 ]. In terms of modifying risk factors other than LDL-C, the co-administration of ezetimibe with statins had a more favorable effect on HDL-C and TG when compared to statins therapy alone [ 33 ]. In conclusion, the addition of ezetimibe to statins produced further lowering of LDL-C of approximately 15–20 % with no apparent increase in side effects. This effect was superior to that observed by doubling the dose of the statins. Furthermore, the lowering produced was consistent. Ezetimibe co-administration with fibric acid derivatives was examined in a randomized, evaluator-blind, placebo-controlled, parallel-group study of 32 healthy hypereholesterolemics. Ezetimibe co-administration with fenofibrate was found to produce clinically significant reductions in LDL-C (36.3%) compared to the fenofibrate group (22.3%) with a more favorable TG and HDL-C profile [ 34 ]. Sitosterolemia Sitosterolemia is a rare inherited disorder caused by mutation in either the ABCG5 or ABCG8 genes located on chromosome (2p21) [ 35 ]. First described by Bhattacharyya and Connor in 1974 in two sisters of German and German-Swiss ancestry with normal mental development. The patients presented with tendinous and tuberous xanthoma and elevation of beta-sitosterol, campesterol and stigmasterol (plant sterols) in the blood [ 36 ]. Affected individuals have increased intestinal absorption of plant sterols (mainly sitosterol) that are usually absorbed in minute amounts in normal individuals. Additionally, these patients have diminished clearance of plant sterols, leading to very high levels of plant sterols in the plasma. Patients suffering from sitosterolemia have severely depressed hepatic cholesterol biosynthesis, and decreased levels of HMG-CoA reductase enzyme [ 37 ]. Clinical manifestations include: tendon and tuberous xanthomas, episodes of hemolysis, accelerated atherosclerosis, and premature coronary artery disease. It is important to note that close to 50 % of these patients have normal cholesterol levels [ 38 - 41 ]. A recently reported trial demonstrated that treatment with ezetimibe reduces plant sterol levels in patients with sitosterolemia. The authors reported a decrease in sitosterol concentrations by 21% and campesterol by 24 % [ 42 ]. Ezetimibe and atherosclerosis It is generally accepted that atherosclerosis is an inflammatory disorder. It is believed that the atherosclerotic process begins with endothelial cell activation, which is triggered by multiple factors such as oxidized lipoproteins. Cholesterol lowering agents as a group have demonstrated great efficacy in prevention and cessation of the progression of atherosclerosis. The efficacy of ezetimibe in monotherapy or in combination on CHD morbidity and mortality has not been well established. One of the unique features of cholesterol absorption inhibitors is their ability to modify post-prandial hyperlipidemia. There is increasing evidence that post-prandial lipoproteins (particularly cholesterol-rich chylomicron remnant) are atherogenic. Ezetimibe has the potential to reduce the cholesterol content of chylomicrons by up to 60% [ 44 ], which may lead to a lower atherogenic potential of chylomicron remnants [ 43 ]. High-sensitivity C-reactive protein (hs-CRP) is an inflammatory mediator whose levels correlate with increased coronary risk. Ezetimibe co-administered with simvastatin resulted in significant incremental decreases in hs-CRP in patients with primary hypercholesterolemia. Changes in individual lipid parameters did not explain the observed decreases in hs-CRP and were possibly consistent with an additional anti-inflammatory effect compared with simvastatin monotherapy [ 45 ]. In a prospective trial to study effects of ezetimibe co-administered with atorvastatin in patients with primary hypercholesterolemia, ezetimibe plus atorvastatin significantly provided an additional (10%) lowering of hs-CRP versus atorvastatin alone [ 46 ]. The unanswered questions Ezetimibe and its class of cholesterol absorption inhibitors are new, and there is a lack of outcomes data to explore whether its cholesterol modifying effects will translate to lower CHD mortality and morbidity. The safety of this medication has not yet been established with long term trials data as most of the studies conducted were short term. With the advent and increased utilization of combination therapy in the management of dyslipidemia, further trials are needed to explore the efficacy, indications and safety profile of ezetimibe use in combination with Peroxisome proliferator-activated receptors (PPARs), niacin and bile acid resins. The increased popularity of special weight loss diets such as the high protein diet, poses questions of whether such diets will alter the efficacy or safety of cholesterol absorption inhibitors. Finally, the efficacy of statins in reducing CHD related events has lead to the controversial hypothesis regarding whether or not statins poses a pleiotropic (non lipoprotein) effect. If a pleiotropic effect exists, one might argue that a statin at a higher dose might be more beneficial than combination therapy producing the same effect. Conclusions Ezetimibe is the first clinically approved cholesterol absorption inhibitor. It is effective in lowering LDL-C as monotherapy or in combination with statins. The use of combination LDL-C lowering medication is expected to become a much more common modality of treatment, especially after the recent NCEP/ATP III update. Ezetimibe offers further lowering of LDL-C and non HDL-C that is consistent and probably safer than increasing the dose of the individual statin. It also provides another effective treatment option for HoFH and sitosterolemia patients. Because of its recent introduction, we still lack both outcomes and long term safety data.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524501.xml

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.