A Perspective on How Fibrinaloid Microclots and Platelet Pathology May be Applied in Clinical Investigations

shows the broad swathe of phenomena on which we focus, as they occur in response to an external stimulussuch as an infection. It recognizes that a chief set of pathways necessarily involves an external event—commonly an infection—that leads to inflammation (assessed via increases in the levels of inflammatory cytokines) and directly or indirectly to the formation of microclots and the hyperactivation of platelets. These phenomena can then feedback in a kind of autocatalytic cycle (in very unfavorable cases leading to a "cytokine storm"). Fig. 1 ^{44 45 46}

The presence and concentration of molecules,including pro- (and anti-) inflammatory molecules vary considerably between individuals, even in "health", not least because of variations in ageand in the gut microbiome.Considerable variation similarly exists in those with the "same" disease,such as an infection(including by SARS-CoV-2), the patterns often correlating with severity and/or outcome. The same is true in myalgic encephalitis/chronic fatigue syndromeand long COVID.Consequently, the presence of a soup of inflammatory molecules in circulation that individually might bind to the numerous receptors on platelets, and have direct protein–protein interactions, may therefore be a crucial predictor of the detailed effects of any systemic inflammation. These interactions should be seen as central to the cause and effect of disease development and presentation of both diverse and overlapping symptoms. Clotting and platelet pathologies should therefore not simply be seen as a predictor of clinical thrombosis. Such assumption would be oversimplifying the pathophysiological value of both identifying and studying the underlying causes of the presence of the inflammatory molecules in circulation. ^{47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62}

There are many reasons why we might anticipate that platelet activation in chronic inflammatory diseases might have many origins, not least the plethora of receptors on platelets that can be activated individually by numerous circulating inflammatory molecules (seeandfor a basic list of the types of membrane receptors and their functions). Platelets are mostly seen as participating "only" in clot formation during wound healing, but they are actually most important as signaling entities in immuno-thrombosis. Signaling through different receptors can lead to different degrees of platelet activation; however, when a platelet is activated, it cannot return to its original state. Different diseases and even individuals with the same disease have a unique combination and concentration of inflammatory molecules in circulation. It would only need a few of these molecules to bind to any one or more of the various receptors on platelets, to cause platelet hyperactivation. The exact nature of the inflammatory molecule trigger may be different, but the end results will be an "outside-in" signaling activation that, in turn, results in "inside-out" platelet signaling and platelet hyperactivation. This is therefore a case where platelet activation and the concept of "cause and result" should be discussed with care. Nonetheless, microscopy imaging might be particularly useful in studying platelet hyperactivation, as the preparation methods we employ in our research allows for minimal handling of samples and short timeframes between obtaining the sample and doing the analysis.shows examples of microscopy imaging of platelets from control and long COVID samples. Fig. 2 Table 1 Fig. 3

We think it will be helpful to rehearse some of the known facts about fibrinaloid microclot formation in the form of a "ten things"style (); also please seefor examples of fluorescence microscopy imaging of microclots. Many proteins can adopt a more thermodynamically stable microstate with no change in the primary structure (sequence), in which the more stable contains an ordered β-sheet "amyloid" structure(see). Normally, however, it is present in a less stable state that is kinetically more accessible during and following its synthesis. The more stable (labeled PrPSc) is separated from the initial state (PrPC) via a large energy barrier. This is true for amyloid proteins generally and is illustrated here for classical prion proteins. ⁶³ Table 2 Fig. 4 ¹⁵ Fig. 5 ^{17 64}

offers a comparative examination ofreanalyzing fluorescence microscopy data, previously published,extending our prior analysis to include healthy participants, individuals with T2DM, and those currently experiencing acute COVID-19. Reexamining some of our previously collected micrograph data from previously published papers, we also provide evidence that microclots are visible in scanning electron microscopy (SEM) of whole blood samples.showcases a selection of SEM micrographs of healthy whole blood and conditions where clotting pathologies are well-known. In, a few plasma deposits (microclots) are discernible among or on the erythrocytes, whileshow whole blood SEM micrographs of samples from an individual diagnosed with systemic lupus erythematosus (), acute COVID-19 during the initial wave in 2020 (), rheumatoid arthritis (), and a sample from a patient diagnosed with Alzheimer's-type dementia (). We only realized that these deposits were of significant importance during the last few years. Fig. 6 ²⁶ Fig. 7 Fig. 7A Fig. 7B to F Fig. 7B Figs. 7C and D Fig. 7E Fig. 7F microclot areas

We remain relatively ignorant of the details of precisely what governs fibrin self-assembly following fibrinopeptide release (e.g., why it stops at a certain and nonconstant fiber diameter, even in health,and why it can vary with disease). One thing we do know, however, is that microclots covary with (and in our view are thus largely responsible for) the severity of disease in both acute and long COVIDand in fact contribute to widespread thrombotic endothelialitis in long COVID. ^{65 66 67 62}

Since the early discoveries that molecules such as oestrogenscould induce the formation during clotting of "dense matted deposits" (that we now refer to as fibrinaloid microclots, because of their amyloid character), we have also found that low concentrations of molecules such as bacterial lipopolysaccharide,lipoteichoic acid,serum amyloid A,and SARS-CoV-2 S1 spike proteincan also do so. These dense matted deposits could be induced using both whole blood and platelet-poor plasma. It is moderately to highly unlikely that they are binding in the same locations as each other (the binding site of spike is known), and thus, the morphology of the clots they induce will be the same. In a similar vein, nonamyloid substances can also bind to and induce the aggregation of established amyloid proteins such as bacterial DNA promoting tauor β-amyloidaggregation. ^{68 69 14 21 70 34 71 72 73}

The fibrinaloid microclots that we discovered and that can be induced my miniscule amounts of molecules such as bacterial LPS or the SARS-CoV-2 spike protein are resistant to normal fibrinolysis for a least two reasons: (1) amyloid-type proteins are inherently more resistant to proteolysisbecause of their crossed-β structure,and (2) because they also entrap inhibitors of normal proteolysis such as α2-antiplasmin and plasminogen activator inhibitor 1 (PAI-1).Consequently, they are able to block microcapillaries, leading to tissue ischemia and hypoxia, from which the great majority of symptoms can be seen to follow.Importantly, our analysis provides a mechanistic link between external trigger events and the pathologies of present interest, as well as a candidate set of targets for treatment. ^{35 74 15 75 76 74 17 77}

To date, our microscopy imaging and analyses have mostly been semiquantitative, showing the existence of different microclots in various diseases. As with much of modern, postgenomic biology,especially that influenced by "deep learning", it is time to move from a "discriminative" strategy. We suggest a "generative" strategy, in which we seek to solve the "inverse problem" (as in) by using the observables (microclot properties) to infer their "cause" (i.e., the diseases with which they are associated). The observables include the distribution of clot sizes and morphologies,the diameter of individual fibers,their ability to cross-seed other amyloidogenic proteins,the spectral properties of the different stains when attached to fibers,the susceptibility of the clots to proteolysis by different proteases,and the extent to which they are naturally present in platelet-poor plasma versus being induced by the in vitro addition of thrombin. ^{78 79 80 26 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 35 97 17 34 98}

Each property of an individual example in a system of interest can be seen as an element of a vector describing that system or as a dimension in multidimensional space.When we have a series of properties of both a sample "as a whole" and indeed of individual objects therein, it becomes increasingly easy to discriminate them.Care is needed, however,since even comparatively small random variations in normal distributions can appear significant in individual dimensions when there are many to choose from, even in "unsupervised" methods (in which class membership, such as a particular disease, is not known). In favorable cases (e.g.,), though, these are entirely sufficient. Supervised methods, in which a model is "trained" to predict an output set of properties (such as a particular disease) from "input" variables such as omics data, are even more prone to overtraining and other kinds of bias.The solution here (e.g.,) is to assess any predictions taken from such a mathematical model using separate "validation" examples whose class membership or other output properties are known but which are not used in the construction of the model. ^{99 100 101 100 102 103 104 105 102 106}

Another modern trend is to recognize that the amount of unlabeled data available normally vastly exceeds that of labeled data and that such data can be used in the training of a supervised model; these methods are known as "semisupervised." They have become preeminent in deep learning modelsbased on variational autoencodersand transformers,especially in natural languageand image processing. ^{79 107 108 109 110 111 112 113 114 115 116}

In variational autoencoders, one essentially clusters input examples into a (much) lower dimensional space. This still allows considerable discrimination; however, even a normalized vector of just 20 in which an individual may be in the upper or lower half admits 2(approximately 1 million) possibilities.In the case of microclots, we consider that (distributions in) the many 1,000s of individual metabolitesand proteinsin serum or plasma can potentially each affect the size and shape of the microclots. This "harvesting" of all the molecules to which the fibrinogen is exposed, and which then determines how it polymerizes, effectively concentrates the vast numbers of metabolites, proteins, and even transcripts into a smaller number of dimensions; the microclots essentially act as a surrogate for the metabolome, proteome, and transcriptome present in the plasma at the time of clotting. Consequently, we are optimistic that an analysis of the detailed morphological and spectral properties of microclots will indeed serve to discriminate, possibly quite finely, individuals with different diseases. 20 ^{109 117 118 119 120 121}

In our previous work, we have focused more or less qualitatively on the presence of fibrinaloid microclots that we discovered using microscopy imaging while recognizing that some of their measurable properties differ in various conditions or diseases. Within the terminology of the deep learning agenda,this is to be seen as a "discriminative" approach. The opposite strategy, amounting to the solution of the inverse problem, is referred to in general as a "generative" strategy. Here the aim, now clearly worthwhile, is to develop and exploit a more quantitative analysis in the prediction of diseased states, and the success of any treatment both in modifying the microclot properties and in curing or ameliorating the diseases. Various laboratories in the United States, United Kingdom, and Germany have successfully implemented microclot imaging and will also be publishing their results shortly. We have been using and described here, fluorescence microscopy imaging (and also briefly SEM methods) to visualize microclot presence and platelet hyperactivation. In addition, we are also developing imaging flow cytometry methods that may, in future, direct a meaningful translational outcome that could guide treatment options. We also suggest that automation of our microscopy imaging using software modalities like the MetaSystems platform (), might be particularly useful for unbiased quantitative analysis of microscopy imaging of both platelets and microclots. Other research teams are developing innovative strategies that integrate real-time one-dimensional-imaging fluorescence with deformability cytometry within a single instrument, denoted as RT-FDC.RT-DC techniques are also currently being adapted to study and characterize the mechanical classification of microclots in whole blood, at rates of several hundred cells per second. We believe that the research agenda, using various novel methods, as set out here, will make the enterprise of unbiased quantitative analysis of microclots, worthwhile. We suggest that researchers direct increased attention toward a previously disregarded fraction of the blood sample. Microclots found in platelet-poor plasma could hold a significant diagnostic value for clotting-related issues in inflammatory diseases and other similar conditions, including postviral syndromes. https://metasystems-international.com/↗ ^{79 122}