Abstract:Cardiovascular diseases, the leading cause of death in Western nations, carry enormous socioeconomic costs and have been a focus of clinical and basic research throughout the past century. Genetic analyses have become a staple of such research, and several single nucleotide polymorphisms (SNPs) are now known that correlate well with an increased risk for thrombotic events. Most of these SNPs, including the Leiden variant of factor V and the 2021 A allele of prothrombin, are associated with enhanced fibrin deposition. Individuals who carry these alleles are at increased risk of venous thrombosis, because fibrin is the dominant component of thrombi in the veins. In arteries, however, platelets play a more prominent role in thrombus formation, and these SNPs have little prognostic value. Until 1996, none of the SNPs present in genes determining platelet function had been suggested to influence the risk of arterial thrombosis. The seminal but still controversial publication of Weiss et al. (1), identified the PlA2 allele as a significant risk factor for acute myocardial infarction and unstable angina, particularly in younger Caucasian males. As discussed below, a massive amount of clinical data regarding this association has accumulated since that time, but the literature is equivocal in the extreme. Skeptics demand a mechanism by which this SNP could give rise to important differences in platelet function. The article by Vijayan et al. (2) in this issue of the JCI begins to address their concern. In essence, these authors have established that the PlA1 and PlA2 alleles create multiple differences in the function of the platelet integrin αIIbβ3, strengthening the case for an association between the polymorphism and cardiovascular disease. The PlA1/PlA2 polymorphism lies within the gene for the β3 (GPIIIa) subunit of αIIbβ3 (GPIIb-IIIa). This integrin serves as a receptor for adhesive proteins fibrinogen and von Willebrand factor and is critical for the cohesive properties of platelets. Recognition of these proteins as soluble ligands requires αIIbβ3 to be activated by exposing platelets to one of several agonists. The transition of the receptor from a resting to an activated state depends on inside-out signals, which are transmitted across the cell membrane to render the extracellular face of αIIbβ3 competent to bind soluble ligands. Occupancy of αIIbβ3 also induces a series of outside-in signals culminating in the activation of numerous kinases and phosphatases and in rearrangement of the platelet cytoskeleton, allowing platelets to spread on an adhesive surface. Recognition of fibrin or fibrinogen as an immobilized substrate does not require inside-out signaling, but still induces outside-in signaling. PlA1/PlA2, one of several allelic systems on αIIbβ3, was originally recognized in transfusion medicine as a cause of neonatal alloimmune thrombocytopenia and posttransfusion purpura. The structural basis of the PlA1/PlA2 polymorphism arises from a single base difference, a thymidine or a cytosine, which leads to an amino acid difference at position 33 in the mature β3 subunit: a leucine in PlA1 or a proline in PlA2 (3). Because this simple substitution introduces multiple neoepitopes within β3 (4), it may exert substantial and long-range effects on the conformation and, consequently, on the function of αIIbβ3. Since the initial report that the PlA2 polymorphism is a cardiovascular risk factor, no fewer than 17 studies have presented findings on this topic. Of these, 9 support a role for the PlA2 allele as a thrombotic risk factor, and 8 do not. By far the largest of these studies was the analysis by Ridker et al. (5) who found no evidence of an association between the PlA2 allele and myocardial infarction, stroke, or venous thrombosis among 14,916 men. However, several reports within the last 2 years (6, 7) identified a significant association of PlA2 with the severity of coronary artery disease or stenosis. Some of these studies (7) suggested that clinical consequences could even be seen with expression of the heterozygous PlA1/2 genotype. The search for a mechanism by which the PlA1/PlA2 polymorphism might affect platelet function has only added to this confusion. For example, although Goodall et al. (8) noted a difference in binding of soluble fibrinogen to platelets with PlA1 or PlA2, other investigators (9, 10) found no such evidence. In addition, Feng et al. (11) reported that PlA2 platelets activated at lower doses of one platelet agonist (adenosine diphosphate), whereas Lasne et al. (12) reported that higher doses of another platelet agonist (thrombin receptor activating peptide) were required to activate PlA2 platelets. As summarized in Figure Figure1,1, other functional differences between PlA1 and PlA2 platelets have been reported (Figure (Figure1),1), but no consistent pattern has emerged (13, 14). Figure 1 Functional consequences of the PlA2 allele within αIIbβ3. Previous reports have used platelets and have yielded conflicting results, whereas the study by Vijayan et al. (2) identifies differences in outside-in signaling using a cell line ... A fundamental problem confronting investigators in such studies is the notorious variability of platelet function assays. To circumvent this problem, Vijayan et al. (2) established cell lines expressing either the PlA1 or PlA2 forms of αIIbβ3. Working with these cells, the authors found no differences in soluble fibrinogen binding to the PlA1 and PlA2 cells. However, significantly more PlA2 than PlA1 cells adhered to immobilized fibrinogen. The greater adhesion of the PlA2 cells was associated with greater spreading and more extensive actin polymerization on the fibrinogen substrate. Such changes are likely to reflect differences between PlA2 and PlA1, in the nature or the efficiency of outside-in signals that are generated when αIIbβ3 encounters a thrombogenic substrate. Consistent with this interpretation, tyrosine phosphorylation of pp125FAK — a kinase that is frequently activated during integrin signaling — was enhanced in the PlA2 cells. Also, clot retraction, another response that depends on outside-in signaling, was enhanced in the PlA2 cells. The latter observation is particularly notable because retracted clots are more resistant to thrombolysis, so clots containing PlA2 platelets would be expected to persist longer and pose greater danger than clots that form with only PlA1. To be sure, the differences in αIIbβ3 function between the PlA1 and PlA2 cells were quite subtle, but marked differences would not be expected: the PlA2 allele is proposed as a risk factor, and not a cause of thrombosis. The challenge now lies in verifying whether the differences identified by Vijayan et al. are indeed responsible for the thrombotic risks associated with the PlA2 polymorphism. Additional epidemiological studies and data analyses are needed to determine whether the PlA2 genotype is a risk of cardiovascular disease. At a more general level, the possibility remains that still other platelet SNPs of equal or even greater prognostic value can be identified.

Deep neural networks architectures from the perspective of manifold learning

Deep Manifold Transformation for Dimension Reduction

Deep Manifold Transformation for Nonlinear Dimensionality Reduction

Deep Manifold Computing and Visualization

Deep Manifold Part 1: Anatomy of Neural Network Manifold

Exploring the Manifold of Neural Networks Using Diffusion Geometry

Deep Manifold Computing and Visualization Using Elastic Locally Isometric Smoothness

Markov-Lipschitz Deep Learning

ManifoldNet: A Deep Network Framework for Manifold-valued Data

A Comprehensive Review of Deep Neural Network Interpretation Using Topological Data Analysis

A geometric viewpoint of manifold learning

Experimental Observations of the Topology of Convolutional Neural Network Activations

The Training Process of Many Deep Networks Explores the Same Low-Dimensional Manifold

Deep Extrinsic Manifold Representation for Vision Tasks

A Theoretical Study of Neural Network Expressive Power via Manifold Topology

How deep learning works --The geometry of deep learning

Matrix Manifold Neural Networks++

Intrinsic Dimension, Persistent Homology and Generalization in Neural Networks

The Pl(A2) allele and cardiovascular disease: the pro(33) and con.

Probing Biological and Artificial Neural Networks with Task-dependent Neural Manifolds