Proteomic Analysis of Cardioembolic and Large Artery Atherosclerotic Clots Using Reverse Phase Protein Array Technology Reveals Key Cellular Interactions Within Clot Microenvironments

Thrombus characteristics are dependent on clot composition, but identification of the etiology based on histological analysis has proved inconclusive. Identification of proteomic signatures may help to differentiate between clots of different etiologies such as cardioembolic, large artery atherosclerotic, and other known etiologies, information that could enhance an individualized medicine approach to secondary stroke prevention. In this study, total protein extracts from cardioembolic (n=25) and large artery atherosclerotic (n=23) thrombus specimens were arrayed in quadruplicate on nitrocellulose slides and immunostained for 31 proteins using a Dako Autostainer (Agilent Technologies, Inc., Santa Clara, USA). We quantified 31 proteins involved in platelet and/or endothelial function, inflammation, oxidative stress, and metabolism. Pathway analysis showed more heterogeneity and protein network interactions in the cardioembolic clots but no specific correlations with clot etiology. Reverse-phase protein arrays are a powerful tool for assessing cellular interactions within the clot microenvironment and may enhance understanding of clot formation and origination. This tool could be further explored to help in identifying stroke etiology in large vessel occlusion patients with embolic stroke of an undetermined source.


Introduction
Ischemic stroke is caused by a thromboembolic occlusion of cerebral arteries. Treatment is focused on fast and efficient removal of the occluding clot with either intravenous thrombolysis or with endovascular thrombectomy. Although the occluding clot is the primary target of treatment, little is known about its internal organization and inflammatory or cytokines that are involved in the formation or evolution of clots. Also, cellular pathways and cells governing the structure of clots has become a hot topic in the medical literature. A recent subgroup analysis from the Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands (MR CLEAN) Registry has confirmed that thrombus computed tomography characteristics of cardioembolic (CE) stroke are distinct from those of non-CE stroke. Non-CE strokes were associated with the presence of hyperdense artery sign, a higher clot burden, a shift towards a more proximal thrombus location, and longer thrombi compared with a CE cause [1]. The differences in imaging characteristics between stroke etiologies are likely a result of differences in the composition of the occlusive thrombus resulting from dissimilarities in the clotting mechanisms. Histological analysis can reveal microscopic clot composition, but studies relating etiology to clot composition have thus far proved inconclusive [2]. Identification of proteomic signatures of thrombi from different stroke etiologies may prove invaluable in identifying the true etiology of the roughly 40% of ischemic stroke patients that have no known cause, which complicates effective secondary stroke prevention in these patients [3]. This study sought to investigate the proteomic signature of CE and large artery atherosclerotic (LAA) clots.

Patients
All patients underwent comprehensive neurovascular imaging for evaluation of the presence of carotid atherosclerotic disease, including CT angiography from the aorta to the intracranial arterial vasculature. All patients also underwent a standard cardiac evaluation, including a transthoracic echocardiogram (TTE)/transesophageal echocardiogram (TEE), electrocardiogram (EKG). In cases in which no abnormality was found on the initial cardiac workup, prolonged cardiac monitoring was performed using either a Holter monitor or implantable cardiac monitoring device. Patients were categorized according to Toast criteria [4]. The study was approved by the Mayo Clinic Institutional Review Board, and a waiver of consent was granted.

Specimens
Thrombus specimens were collected from patients undergoing mechanical thrombectomy at Mayo Clinic between October 2016 and July 2019 ( Table 1). The CE (n=25) and LAA (n=23) thrombus specimens were fixed in 10% neutral buffered formalin for up to 24 hours and processed into paraffin blocks (FFPE). Two representative slides from each clot (3µm-thick sections, two-section on each slide) were stained with Martius Scarlett Blue stain. After regular histological staining, three 50μm sections were cut from each FFPE block, placed into a microcentrifuge tube, and stored at 4°C. Two tissue sections from each specimen were deparaffinized in the tube using two exchanges of xylene (three minutes at 50°C). The tissue sections were rehydrated in graded ethanol (100%, 95%, 70%, water with protease inhibitors (Roche cOmplete™ Mini), and lysed in protein lysis buffer (450μL tissue protein extraction reagent (Pierce T-PER™), 450μL tris-glycine 2X sodium dodecyl sulfate (SDS) sample buffer (Invitrogen™), and 100μL tris(2-carboxyethyl)phosphine (Pierce™ TCEP Solution, neutral pH, Bond-Breaker™ ). Tissue lysates were heated at 95°C for eight minutes and stored at -80°C.
Arrays were scanned at 600dpi on a flatbed scanner (PowerLook, UMAX®, Dallas, USA). Spot (pixel) intensity was analyzed using ImageQuant™ ver. 5.2 (GE Healthcare, Chicago, USA), with mean local area background subtraction. Signal intensity for each spot was calculated using the freely available data reduction algorithm (RAS ver. 16) [5]. Spot intensities were normalized to total protein/spot.

Histological composition of LAA and cardioembolic clots
The Martius scarlet blue (MSB) histological analysis revealed that there were slight differences in the histological clot composition between the LAA and cardioembolic clots ( Figure 2)

Proteomic heterogeneity exists within classes of thrombus specimens.
Data for two specimens in the LAA group were not used for hierarchical clustering and network analyses due to missing data (poor precision between technical replicates) for CD63, CD45, and collagen IV. Unsupervised two-way hierarchical clustering revealed two primary clusters of specimens; however, each cluster showed heterogeneity in relation to the etiology of the clot (Figure 3). The protein levels for the 31 proteins in our study could not clearly discriminate CE and LAA thrombus types. relationships between the expressions of all variables (rows) for all thrombus specimens. The 'variables' concentrations are depicted using a color scale from red (high), to white, to blue (low).

Cardioembolic clots possess diverse protein signaling interactions
A pairwise protein linkage analysis was conducted using Spearman's rho non-parametric correlation to map inter-related signaling kinases. Fifty strong protein linkages were noted for the CE (rho≥0.85, p<0.001), while only 18 strong linkages were noted for the LAA group (  The protein linkages (correlations) in CE clot specimens were diverse and more abundant compared to LAA clot specimens (Figure 4). Predominant linkages between PPAR-gamma and arginase-1, CD63, CD234, PKCαβ Thr 638/641, and von Willebrand factor (vWF) were found in CE clots compared to LAA clots ( Table  3). Multiple protein linkages within inflammatory/immune cell proteins arginase-1, neutrophil elastase, CD234, and CD63 were found in CE clot specimens. In contrast, the LAA clots did not have any unique protein-protein correlations compared to the CE clots and lacked significant vWF protein linkages.

FIGURE 4: Protein interconnections and network signaling differ between thrombus etiology
A correlogram of CE thrombus specimens reveals numerous, strong protein-protein interconnections, shown by the dark blue dots (A). Diverse protein network signaling predominates in CE thrombus specimens (B). In contrast, LAA clots show very few protein-protein interactions (C), with sparse network interconnections (D).

Discussion
Our study advances the knowledge of the human cerebrovascular thrombi composition by focusing on the proteomic analysis of cardioembolic and atherothrombotic thrombi. Our results support the general notion that direct analysis of the clots could help in the unveiling, in the future, biomarkers which potentially could assist in stroke diagnosis and differentiating stroke etiologies. Our study comparing the proteomic signatures of CE versus LAA clots demonstrated a number of interesting findings. We found predominant linkages between PPAR-gamma and arginase-1, CD63, CD234, PKCαβ Thr 638/641, and von Willebrand Factor (vWF) in CE clots suggesting that platelet signaling dominates in CE clots compared to LAA clots. The multiple protein linkages within inflammatory/immune cell proteins lend additional support to the observation of platelet-immune cell cross-talk in CE clot specimens, supported by the white blood cell enrichment seen in the histological analysis of CE emboli in this and previous studies [9]. Proteomic signaling does not necessarily correspond with cellular proportions on histological staining, but it does reflect activated pathways and intra-and inter-cellular communications present in the occlusive clot, influenced by the etiology. Detailed analysis including individual patient histories and medications will likely demonstrate strong correlations with certain protein networks, a study we plan to perform with the larger sample currently under investigation.
The widespread use of mechanical thrombectomy devices has resulted in the availability of clot material for histopathological and proteomic analysis. Studies relating histological clot composition to etiology have thus far proved inconclusive [2]. Identifying stroke etiology in this population is very important due to the fact that up to 30% of large vessel occlusions (LVO) patients do not have a stroke source identified, which complicates secondary stroke prevention strategies. Secondary stroke prevention in this population is essential in preventing recurrent large vessel occlusion, an occurrence that can have a devastating effect on the patient [9]. RPPAs are a powerful tool for assessing proteomic interactions within the cellular microenvironment and may enhance our understanding of the clot formation process in various etiologies, which is why we sought to investigate the proteomic signatures of thrombi from LAA and cardiac etiologies.
Previous proteomics studies have provided some insight, but much more needs to be explored to identify pathways that could aid in differentiating stroke etiology. A previous proteomic study found that two inflammation proteins (integrin alpha-M and mitochondrial superoxide dismutase) are associated with high blood low-density lipoprotein (LDL) [10]. In another study, these two proteins failed to differentiate between cardioembolic vs. atherosclerotic etiology [11]. Also, another study analyzed four thrombi, with 1,600 proteins identified [12]. To identify proteomic signatures that can help differentiate between clots of different etiologies such as cardiac and LAA, 31 proteins involved in platelet and/or endothelial function, inflammation, oxidative stress, and metabolism were quantified using reverse phase protein arrays (RPPA) [13][14][15]. Unsupervised two-way hierarchical clustering revealed two primary clusters of specimens, but despite the difference in etiology between the two thrombus types, the resulting proteomic clusters were quite heterogeneous within each group. However, we did find that platelet signaling dominates in CE clots compared to LAA clots and hope to explore this hypothesis further in larger cohort studies.
Protein signaling networks comprise a series of interconnected proteins and their post-translationally modified forms [6,16], which are generally organized within specific biological processes, such as inflammation or endothelial integrity. Crosstalk between pathways provides redundancy and resiliency within the cellular microenvironment. Mapping these signaling networks provides functional insights into metabolic processes [13,16,17]. The diverse protein signaling interactions found in the clots were determined primarily by analysis of protein phosphorylation cascades. Phosphorylation events are transient functional indicators of active kinase signaling pathways. Specific phosphorylation sites, at any point in time, define the functional character of protein-protein interactions. A set of interacting proteins within an active pathway will tend to be simultaneously phosphorylated in the signaling cascade. Thus, for a given set of patient observations over a disease cohort, correlations among individual protein pairs imply active interactions in a linked pathway [6,18].

Conclusions
While we demonstrated common protein interactions in thromboembolic disease, consisting of immune, endocannabinoid, and metabolism proteins, CE protein signaling networks are comprised of complex interactions across biological pathways and differ from those seen in LAA patients. Our findings suggest that RPPA of retrieved thrombi in large vessel occlusion could be a promising tool in determining stroke etiology. However, larger studies are needed.

Additional Information Disclosures
Human subjects: Consent was obtained or waived by all participants in this study. Animal subjects: All authors have confirmed that this study did not involve animal subjects or tissue. Conflicts of interest: In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: This work was funded by the National Institutes of Health grant number R01 NS105853. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work.