Identification of a specific inflammatory protein biosignature in coronary and peripheral blood associated with increased risk of future cardiovascular events

1

and rationale 2 As an adjunct to coronary intervention, the Liquid Biopsy System (LBS, PlaqueTec, UK) 3 enables accurate intracoronary blood sampling at discrete sites simultaneously . We 4 investigated variation between local coronary and remote (peripheral) blood levels of a panel 5 of atherosclerosis-associated proteins and examined how this might relate to cardiovascular 6 risk assessment. 7

and Results 8 In a previous proof-of-concept trial, coronary blood samples were collected using the LBS in 9 28 patients. For 12 of these patients, sampling was conducted across the uninstrumented 10 lesion, prior to percutaneous coronary intervention (PCI). Peripheral blood samples were also 11 collected, at baseline and after PCI. Protein levels in coronary and peripheral plasma samples 12 were analysed by proximity extension assay (PEA, Olink). 13 Before PCI, in 10 out of 12 patients, coronary levels of hepatocyte growth factor (HGF), 14 pappalysin-1 (PAPPA) and spondin -1 (SPON1) were elevated compared with peripheral 15 levels, in some cases >10- fold. Following PCI, involving iatrogenic plaque rupture prior to 16 stenting, peripheral levels of these proteins were elevated to a similar degree as coronary 17 levels. In 2 patients, peripheral elevations of HGF, PAPPA and SPON1 (all >90th centile) were 18 observed at baseline, prior to PCI. The protein pattern that was identified, consisting of high 19 levels of a combination of HGF, PAPPA and SPON1 was absent in healthy control peripheral 20 blood, but when investigated in baseline peripheral blood samples from reference 21 cardiovascular and COVID-19 patient cohorts, was associated with the occurrence of major 22 adverse cardiovascular events (MACE) and mortality. 23

24 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 3 From investigation of coronary and peripheral blood samples, we identified a novel 1 inflammatory protein signature, which when present in peripheral blood appears to portend 2 worse outcomes. Measurement of these proteins could therefore aid identification of 3 individuals at high risk of cardiovascular events or death. 4 Translational Perspective 5 Through sampling of local coronary blood, we discovered a novel protein biosignature 6 consisting of a combination of elevated levels of HGF, PAPPA and SPON1. When this 7 biosignature was assessed in peripheral samples from reference cardiovascular and COVID-8 19 cohorts, it associated with the occurrence of MACE and mortality. The biosignature 9 protein levels correlated with markers of mast cell and neutrophil activity but not with CRP, 10 possibly indicating a specific inflammatory status. Early detection of this protein signal has 11 potential clinical utility to identify specific patients at increased risk of poor outcomes. 12 13

14 Despite current primary and secondary prevention strategies, coronary artery disease remains 15 the principal cause of death and disability worldwide, indicating much is still to be understood 16 about the mechanisms underlying residual cardiovascular risk (1, 2). Genetic insights have 17 suggested numerous proteins, lipids, and metabolites as identifiers of increased risk (3-6) and 18 several studies have identified circulating biomolecules associated with risk of major adverse 19 coronary events (MACE) . Principally, these studies analysed systemic venous biomolecule 20 levels and c orrelated with measures of disease severity ( 7) or retrospective analys es of 21 outcomes (8, 9). 22 Inflammation is known to contribute to atherosclerotic disease progression, destabilising 23 plaques and promot ing thrombosis; however, the specific causal components of these 24 inflammatory pathways and their links to other factors such as shear stress, wall strain and 25 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 4 flow disturbance are only beginning to be understood ( 10). Assessment of systemic 1 inflammation by measuring C -reactive protein (CRP) levels (an acute phase reactant) has 2 been linked with increased risk of myocardial infarction ( 11), and may be useful in stratifying 3 patients most likely to benefit from targeted anti-inflammatory treatments (12). Encouragingly, 4 large-scale randomised trials of both colchicine (13) and canakinumab (14), an inhibitor of the 5 inflammatory mediator IL-1β, have now demonstrated a reduction in cardiovascular events vs 6 placebo in patients with a history of myocardial infarction, demonstrating that modulating 7 inflammation can have manifest effects on risk of clinically important outcomes in 8 atherosclerotic vascular disease. 9 Systemic venous blood sampling reflects biomolecules derived from multiple organ systems 10 and vascular beds, making the identification of specific plaque-released factors problematic. 11

used to address this have included ex vivo sampling of atherosclerotic plaques (15, 12 16) and coronary thrombi (17, 18) or use of aspiration or guide catheters to sample blood 13 directly from the coronary artery (19-21). Samples collected from the coronary sinus, a venous 14 reservoir receiving drainage from the majority of the coronary circulation, have been compared 15 to blood sampled from the coronary ostium to identify transmyocardial gradients of certain 16 proteins, potentially reflecting coronary disease status more accurately than analysis solely of 17 systemic proteins sampled remotely (22). 18 The present study describes a novel protein pattern discovered in new analysis of coronary 19 and peripheral protein data from our previous study , where intracoronary samples were 20 collected using the PlaqueTec Liquid Biopsy System (LBS ) ( 23). We discovered not only 21 specific elevated signals in the coronary samples but also, in a minority of individuals, in 22 peripheral samples from a baseline blood draw. We then searched other published cohorts for 23 patients exhibiting this peripheral biosignature to assess prevalence an d possible links to 24 clinical outcomes. 25 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 5

1 Coronary artery blood sample collection 2 As part of a proof of concept clinical trial of the LBS device (Clinical Study 1 – CS1, Clinical 3 Trials.gov: NCT02119767), intracoronary samples were obtained via sampling ports positioned 4 either side of an identified plaque in 28 patients (23). As described previously, all participants 5 provided informed written consent ; ethical approval for the study was granted by NRES 6 London, Chelsea 13/LO/0954 (23). This analysis focused on the 12 patients from the CS1 7 study that underwent blood sampling prior to PCI. Peripheral blood samples w ere taken at 8 baseline, prior to heparin administration, and also after PCI. LBS -derived and peripheral 9 samples were mixed with EDTA , and plasma isolated by centrifugation to generate platelet -10 poor plasma and stored at -80 ⁰C until analysis, as described previously (23). 11 Plasma samples were analysed by Proximity Extension Assay (PEA; Olink, Uppsala, Sweden), 12 initially using the CVD1 panel (discontinued) as described previously (23). Separately, plasma 13 samples were reanalysed using the Cardiovascular 3 (CVD3) and Inflammation panels, each 14 containing 92 proteins . The full list of proteins included in the analysis are detailed in the 15 Supplementary materials (File S1). Results are presented as normalised protein expression 16 (NPX) units, and log2 NPX values were converted into linear scale (using 2 NPX=linear NPX). 17 Assay characteristics with detection limits, validations and explanations of NPX units are 18 available from the manufacturer (http://www.olink.com). A ge - and sex-matched healthy control 19 systemic blood samples (BioIVT, Westbury, NY, USA) w ere analysed on the CVD3 and 20 Inflammation panels . CRP levels were analysed by high sensitivity CRP (hsCRP) ELISA 21 (Invitron, cat # IV3-105E; Eurofins, Abingdon, UK) and results provided in mg/L. 22 23 24 25 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 6

populations 1 To investigate prevalence and importance of the biosignature identified in CS1, other published 2 studies where the same proteins have been measured by PEA (Olink) were investigated. A 3 short summary of each cohort investigated is shown in Table 3, and details are provided below. 4 PACIFIC cohort 5 Plasma protein data f rom a study by Bom et al (7) was accessed from the journal website. This 6 study included 196 participants from the PACIFIC cohort with suspected cardiovascular 7 disease (CVD) who underwent coronary computed tomography angiography (CCTA). Proteins 8 were measured by PEA (Olink) using the CVD2, CVD3 and Inflammation panels. 9 Healthy control cohort (Olink) 10 Data on normal ranges of plasma pr oteins measured by PEA using Olink Explore 3072 for 300 11 healthy individuals was accessed at the manufacturer’s website (https://insight.olink.com/data-12 stories/normal-ranges). 13 IMPROVE cohort 14 The IMPROVE cohort r ecruited 3711 participants with at least 3 established CVD risk factors 15 (men, post-menopausal women, dyslipidaemia, hypertension, diabetes, smoking and family 16 history of CVD) (24, 25). Proteomic analysis of plasma was previously measured in IMPROVE 17 using the CVD1 panel (Olink Proteomics, Uppsala, Sweden) ( 26), and data on levels of 18 hepatocyte growth factor (HGF), Pappalysin- 1 (PAPPA) and Spondin- 1 (SPON1) were 19 extracted. Carotid-intima media thickness (c-IMT) was measured as previously described(25) 20 and carotid plaques, defined as c-IMT≥1.5mm; all subjects were followed to 3 years and major 21 adverse events recorded. Because of changing rules at some institutions 491 individuals from 22 one center in Finland have been excluded from the analysis. After quality control , 2901 23 participants had baseline protein and carotid measur ements included in the analysis and 24 followed up after 3 years. 171 cardiovascular events were recorded at 3-year follow-up, defined 25 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 7 as myocardial infarction, ischemic stroke, peripheral artery disease or revascularization 1 procedures. 2 Respiratory disease/COVID-1 9 cohort 3 384 patients enrolled in the emergency department during the COVID-19 peak between March 4 and April 2020 from Boston MGH had plasma samples analysed on the Olink Explore 1536 5 protein panel (27). Protein data and patient outcomes for this study were accessed via the 6 Olink website (Olink.com). 7 ESKD/COVID-19 c ohorts 8 An Imperial College study of end- stage kidney disease (ESKD) subcohort A where 55 patients 9 tested positive and 51 tested negative for COVID-19, and from subcohort B where 46 patients 10 tested positive and 11 tested negative for COVID -19 ( 28) h ad plasma or serum samples 11 analysed using PEA (Olink) assays on Inflammation, Immune Response, Cardiometabolic, 12 Cardiovascular 2, and Cardiovascular 3 panels. Protein data and patient outcomes for this 13 study were accessed via the journal website and personal communication with the authors. 14 Statistical analysis 15 In the CS1 cohort, to test for statistical significance between protein levels in plasma sampled 16 at different locations (peripheral v coronary, and pre- and post-PCI), a one-way ANOVA mixed-17 effects model, followed by a pairwise Holm-Sidak’s multiple comparison’s test was used (with 18 an alpha of 0.05). Pairwise c orrelation coefficients between proteins were calculated by 19 Spearman (rho coefficient) performed in GraphPad (version 9.0). In the respiratory disease 20 COVID-19 cohort , survival analysis , Kaplan-Meier curves and statistics were created in 21 GraphPad (version 9.0). 22 For the IMPROVE cohort, continuous variables are expressed as median and interquartile 23 range, categorical variables as number and percentages. HGF, PAPPA and SPON1 data were 24 standardized using the z -score. The correlation coefficient between the three biomarkers 25 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 8 (pairwise) and of each one of the three biomarkers with CRP (mg/L) wa s estimated by 1 Spearman’s (rho coefficient). Participants were categorized as two groups or ”scores”: “0” if 2 HGF, PAPPA and SPON1 were all <90th centile and “1” if HGF, PAPPA and SPON1 were all 3 ≥90th centile. Kruskall Wallis and chi square tests were used to analyze differences across 4 groups. A linear regression analysis was performed using the log transformed measures of c-5 IMT as dependent variable and the single biomarker or the biomarker score as independent 6 variables. A Cox regression model was used for event analysis. Data were adjusted by latitude 7 (main determinant of c -IMT in the IMPROVE cohort), age, gender and analytical batch. 8 Estimates were expressed as β coefficient and standard error (SE). All analyses were 9 performed in STATA v14. 10 11 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 9

1 Identification of protein bi osignature in coronary and peripheral blood samples 2 12 pa tients (45- 74 years, all male, Table 1) underwent blood sampling prior to any balloon 3 dilatation of an obstructive coronary plaque as part of a proof -of-concept clinical trial of the 4 LBS)(23); 11 of the 12 subsequently underwent PCI. 5 Using the CVD1 protein panel, levels of HGF and PAPPA were identified at much higher levels 6 (>10-fold) in the coronary samples compared with peripheral samples for 10/12 individuals 7 (Figure 1A, Supplement Figure S1) . SPON1 levels were also higher in coronary samples 8 compared with periphery in the same 10 patients but to a lesser extent than HGF and PAPPA. 9 The remaining 2 patients were found to have elevated levels of HGF, PAPPA and SPON1 in 10 peripheral as well as coronary samples (Figure 1 B, C and D), despite no differences in 11 demographics or sample location that may have explained these differences (Table 1). Other 12 coronary/peripheral protein peaks (Figure 1A) were not explored further as they did not have 13 the distinct patient pattern observed with absolute levels of HGF, PAPPA and SPON1. 14 In post-PCI samples, the majority of patients had higher levels of HGF, PAPPA and SPON1 15 peripherally (Figure 1E) compared with pre-PCI samples, suggesting that the conditions of the 16 PCI, including iatrogenic plaque rupture and resultant response to injury - inflammation and 17 release of plaque constituents – caused a systemic elevation of these proteins. 18 Confirmation of protein biosignature in coronary and per ipheral blood samples 19 Initially considered outliers, s amples from the 2 patients with high peripheral levels of HGF, 20 PAPPA and SPON1 prior to PCI were re-tested using the ‘Inflammation’ and ‘CVD3’ panels 21 (Olink), which include HGF and SPON1 from the now-discontinued CVD1 panel but do not 22 include PAPPA. Plasma samples from the 10 individuals with low baseline but elevated post-23 PCI biosignature and from 12 healthy individuals (age- and sex-matched) as a control series 24 were also included in these assays for comparison. 25 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 10 Repeat testing of both HGF and SPON1 demonstrated consistent results (Figure 2A and B). 1 The 10 patients with low HGF and SPON1 in the peripheral samples prior to PCI had similar 2 levels to healthy controls, indicating that the elevated levels seen in 2 patients were well above 3 ‘normal’. Of note, some differences between absolute NPX values measured on CVD1 and 4 CVD3/Inflammation panels were seen, most likely attributed to the different dilutions and 5 dynamic ranges of the newer panel assays. Analysis of coronary/peripheral protein ratios using 6 the two new panels revealed 5 further proteins with distinctly higher levels in the coronary 7 plasma compared with peripheral plasma; C-X-C motif chemokine 9 (CXCL9) , C-C motif 8 chemokine ligand 28 (CCL28), TNF-related weak inducer of apoptosis (TNFSF12 or TWEAK) 9 (Figure 2C), tissue factor pathway inhibitor (TFPI) and azurocidin-1 (AZU1) (Figure 2D). 10 When comparing absolute coronary and peripheral protein values a distinct pattern was seen 11 for HGF, which to varying extent was mirrored by the 7 other proteins. The 2 patients with the 12 biosignature had high levels of AZU1 and CXCL9, however, one other patient in each case 13 had high peripheral levels (Figure 3A) . Some healthy control participants had relatively high 14 levels of AZU1 and CXCL9. The peripheral biosignature pattern appeared to be most 15 consistent for 6 proteins: HGF, PAPPA, SPON1, CCL28, TFPI and TWEAK. 16 Analysis of C-R eactive Protein (CRP) levels 17 To investigate whether the peripheral bi osignature pattern seen in 2 patients could be 18 explained by acute- phase systemic inflammation, CRP levels were measured using a high-19 sensitivity assay (hsCRP). Levels were uniformly low (<5mg/L), and CRP did not correlate with 20 HGF levels (Figure 3B), suggesting systemic inflammation was not a factor. 21 Investigating the peripheral bi osignature in healthy individuals 22 To further investigate the biosignature in healthy cohorts, data available from Olink on 23 variability (IQR) of 2943 proteins in a healthy cohort of 300 individuals, the 6 biosignature 24 proteins had relatively low variability, compared with in the PACIFIC and CS1 cohorts 25 (Supplement Figure S2). If a minority of participants had unusually high levels of the 6 proteins, 26 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 11 this would raise IQR values relative to the other proteins measured. Observing relatively low 1 IQRs for the 6 proteins in this cohort suggests it was unlikely that any of the healthy participants 2 had the biosignature. 3 Investigating the peripheral bi osignature pattern in patients with suspected coronary artery 4 disease: PACIFIC cohort 5 A study by B om et al (7) represents a 196-patient subset of the PACIFIC cohort with suspected 6 coronary artery disease. Two of the 196 patients displayed clearly raised levels of all six 7 biosignature proteins (Figure 4). To allow comparisons across different cohorts, a biosignature 8 definition of >90 th centile for HGF and also PAPPA and SPON1 was used, which identified 9 three participants with the biosignature, all with low CRP levels (hsCRP<2.5mg/L). None of the 10 demographic variables provided clues to differentiate the biosignature-displaying subjects from 11 the rest of the cohort. 12 Investigating the per ipheral biosignature pattern in patients with multiple CVD risk factors: 13 IMPROVE cohort 14 In the IM PROVE population, three of the biosignature proteins were measured: HGF, PAPPA 15 and SPON1. 39 subjects were identified with the biosignature defined as >90th centile levels 16 for HGF, and also PAPPA and SPON1. This was a small minority (1. 3%) of the cohort that 17 comprised 26 men and 13 women (Table 2A). The signature proteins associated with a thicker 18 c-IMTmax (P=0.033) (Table 2B). The biosignature proteins displayed a positive and significant 19 correlation with each other ( Table 2C). No correlation was observed between biosignature 20 proteins and CRP levels , except for HGF. However, the biosignature group had higher CRP 21 levels compared with the remaining cohort (Table 2A). During the 3-year follow-up, 20.5% (8 22 of 39) of the patients with the biosignature suffered a cardiovascular event, compared with 23 5.7% (163 of 2862) in the remaining cohort. Using a Cox regression model, participants with 24 the biosignature had a significantly increased risk of future cardiovascular events; HR 3.16 25 (95%CI:1.55-6.45), p=0.002. Additionally, when applying a 90 th percentile definition for 26 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 12 identification of participants with the highest levels of individual proteins and analysing event 1 rates for HGF (15 of 171, 8.8%), PAPPA (17 of 171, 9.9%) or SPON1 alone (20 of 171, 11.7%), 2 suggested that the biosignature combination of proteins performs better at predicting MACE 3 compared with each protein analysed independently. Furthermore, for participants with CRP 4 above the 90th percentile, 28/171 had an event (16.4%) which suggests that the biosignature 5 is also better than CRP at detecting the occurrence of MACE. 6 Investigating the peripheral bi osignature pattern in patients with r espiratory disease: COVID-7 19 cohort 8 The biosignature was al so investigated in raw data available from the MGH Boston COVID-19 9 cohort (27) using the same biosignature definition to identify individuals with >90th percentile 10 values for HGF, and also PAPPA and SPON1. Analysis of protein levels at day 0 (first sample 11 collected from 358 individuals) identified 11/358 patients (3.1%) with the biosignature, a slightly 12 higher prevalence compared with IMPROVE (Supplement Figure S3 and Summary results in 13 Table 3). For individuals identified with the biosignature, 7 of the 11 (63.6%) died within 28 14 days of admission to hospital, while 42 of the 373 (11.3%) remaining patients died. For the 4 15 surviving subjects with the biosignature, 2 were ‘intubated, ventilated and survived to 28 days’ 16 and 2 were ‘hospitalized, supplementary O2 required’. Survival analysis indicated a significant 17 statistical difference in probability of survival between those with and without the biosignature 18 (P<0.001, Supplement Figure S4). Individuals with the biosignature had higher rates of cardiac 19 events within the first 72 hours of presentation (5/11, 45% ), compared with those without the 20 biosignature (29/373, 8% ). Participants with the biosignature also had higher rates of pre-21 existing kidney disease (5/11, 45%) compared with participants without the biosignature 22 (56/373, 15%) and had similar CRP ranges to those without the biosignature (Supplement 23 Figure S5 and Summary results in Table 3 and Table 4). 24 Investigating the per ipheral biosignature pattern in patients with multiple CVD risk factors: 25 ESKD COVID-19 cohort 26 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 13 The biosignature was then investigated in an Imperial College study of 2 end-stage kidney 1 disease (ESKD) cohorts: subcohort A and B ( 28). Of the patients that tested positive for 2 COVID-19, 2/55 patients (3.6%) in subcohort A and 2/46 patients (4.3%) in subcohort B were 3 identified with the biosignature in first samples collected ( Supplement Figure S 6 and 4 Summarised in Table 3 ). In subcohort A, both patients with the biosignature died during the 5 study, while in subcohort B both patients survived and were classed as having ‘severe’ or 6 ‘moderate’ disease. In subcohort A, CRP correlated with HGF levels but not with the other 7 biosignature proteins ( 28). Participants had wide ranges of CRP levels and no statistical 8 differences were found between groups (Supplement Figure S7 and Table 4). 9 10 Biosignature association with Tryptase, MPO, and Syndecan-1 l evels 11 The biosignature pattern in plasma could possibly be induced via endogenous heparin release 12 from mast cells, and/or neutrophil activity and glycocalyx damage. To test if this notion is 13 consistent with proteins known to be linked with these processes, correlation analysis was 14 performed between the biosignature proteins and tryptase, myeloperoxidase (MPO), and 15 syndecan-1 levels in peripheral blood. Significant positive correlations were found between 16 some or all of the biosignature proteins with tryptase, MPO, and syndecan-1 levels in cohorts 17 where this data was available (Table 4, and Supplement File S1). 18 19 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 14

1 Analysis of coronary and peripheral proteins from patients with coronary artery disease 2 undergoing PCI revealed a novel biosignature, which prompted investigation of the peripheral 3 protein pattern in other cohorts. Using the definition of >90 th centile levels for HGF, and also 4 PAPPA and SPON1, the biosignature was found in a minority of patients from a cohort of 5 individuals with suspected coronary disease (PACIFIC cohort), and in a cohort with known 6 CVD risk factors (IMPROVE). The biosignature was also detected in a minority of individuals 7 in two COVID-19 cohorts but appeared absent in healthy cohorts and in healthy control plasma, 8 which raised the question as to whether this biosignature might reflect an unstable, higher-risk 9 cardiovascular state. 10 The known functions of the biosignature proteins are summarised in Table 5. All have direct or 11 indirect roles in inflammation and endothelial function and some have been associated with 12 risk of adverse events. For example, elevated HGF levels have been linked with worse 13 outcomes in dialysis patients (29) and also more recently in COVID-19 cohorts, with neutrophil 14 activation suggested as a possible source of the high HGF levels preceding critical illness (30). 15 In a genome-wide meta-analysis of 85 proteins in over 30,000 individuals (SCALLOP study) 16 several proteins were identified as causative in disease and included HGF ( positively 17 associated with high triglyceride levels), SPON1 (positively associated with atrial fibrillation) 18 and PAPPA ( positively associated with type 2 diabetes) ( 31). Interestingly, in the Imperial 19 College study of ESKD and COVID -19 patients, SPON1 and CCL28 elevated levels were 20 linked with increased risk of death, while increased levels of TWEAK were associated with 21 reduced risk of death(28). TWEAK plays a role in inflammation and repair in several diseases 22 and is thought to promote the development of atherosclerosis, but lower circulating levels have 23 been linked to higher cardiovascular risk ( 32). Furthermore, 18 proteins found to be 24 independently associated with cardiovascular death included HGF and SPON1 (9). However, 25 the combination of specific proteins at raised levels described here has not been reported 26 previously. 27 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 15 The raised levels of biosignature proteins in the periphery may be due to local changes causing 1 release of proteins from the endothelial glycocalyx, activated/damaged cells or activated 2 neutrophils or mast cells, or inflammation- driven increased de novo synthesis. Of particular 3 relevance, mast cells contain heparin ( 33) and mast cell activation leading to the release of 4 endogenous heparin could increase blood levels of the biosignature proteins via release from 5 the endothelial glycocalyx. Increased mast cell activity, as measured by serum tryptase levels, 6 was reported as an indicator of poor CVD outcomes and tryptase levels did not correlate with 7 CRP levels in that study, suggesting CRP is not a surrogate marker for mast cell activity (34). 8 In the present study, individuals with the biosignature in the CVD cohorts had low -mid CRP 9 levels (<5mg/L) which indicates a low to moderate mortality risk (35). The biosignature protein 10 levels did not correlate with CRP levels, except in the IMPROVE and ESKD/COVID-19 cohorts 11 where CRP correlated positively and significantly with one biosignature protein, HGF (28). 12 However, it should be noted that all patients in the CS1 study were taking statins, which may 13 have impacted CRP levels. In cohorts where tryptase data was available, positive correlations 14 between biosignature proteins and tryptase were detected, suggesting there may be a link with 15 mast cell activity. Additionally, the biosignature also appeared to be linked with neutrophil 16 activity (MPO levels) and glycocalyx damage (Syndecan-1 levels). 17 To understand if an elevation in biosignature proteins is relevant to patient outcomes , we 18 explored its presence in the IMPROVE study population of individuals at high cardiovascular 19 risk. The biosignature group had a significant association with the risk of cardiovascular events 20 and the biosignature protein combination appeared to be better at predicting MACE than the 21 individual proteins or CRP. In the MGH study of patients hospitalised due to COVID-19-related 22 symptoms (27), patients that exhibited the biosignature had higher death rates compared to 23 patients without the biosignature. A similar observation was made in a cohort of patients with 24 ESKD and COVID-19 (subcohort A) (28), where higher mortality rates occurred in those with 25 the biosignature, compared with patients without the biosignature. However, in a second small 26 ESKD COVID-19 cohort (subcohort B) (28) those with the biosignature had severe or moderate 27 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 16 disease and survived. Some differences between cohort A and B might help to explain 1 dissimilar results related to the biosignature, a lthough equally the results could indicate that 2 the biosignature may not be a reliable risk indicator for more severely ill ESKD patients. 3 The biosignature appears to be present in a minority of individuals , which perhaps explains 4 why it is has not previously been described. It might have been overlooked in other studies if 5 unusually high protein levels were considered as outliers. Furthermore, since the biosignature 6 was discovered by comparing coronary and peripheral protein levels, the LBS sampling 7 technology and study design may have provided a unique opportunity to make accurate 8 coronary measurements for comparison with the periphery. 9

10 Although the a priori sample size for this investigation was small, the confirmation of elements 11 of the inflammatory biosignature in much larger and well-characterised populations, particularly 12 when correlated with increased risk of future events, is compelling. Initially, timing of sample 13 collection and the presence of heparin had potential to explain the pre- and post-PCI disparity, 14 however sequential comparison with other cohorts confirmed this was not the case. Although 15 its observation in reference cohorts strengthens the validity of the biosignature, comparison of 16 unequal groups with and without the biosignature results in wide confidence intervals. The 17 choice of threshold of >90 th centile for HGF, PAPPA and SPON1 was arbitrary but aided the 18 identification of participants with the highest NPX values for all of these proteins within a cohort 19 and allowed comparisons across cohorts. F or future applications, it would be desirable to 20 optimise threshold selection and to measure proteins using standard concentration units 21 (pg/ml). Nevertheless, our findings should be considered hypothesis -generating, and 22 prospective studies on larger at-risk populations are warranted, with a focus on mechanisms 23 underpinning the abnormally -elevated biomolecules and their role in the aetiology of 24 cardiovascular risk. 25 26 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 17

1 A biosignature was detected in individuals with known CVD, suspected CVD or with CVD risk 2 factors. The biosignature associated with a higher rate of MACE, and in COVID-19 cohorts it 3 appeared to be associated with increased risk of death. The biosignature has the potential to 4 be developed as a non-invasive blood test indicative of a specific type of vascular inflammation 5 and we can speculate its use as a warning signal, especially in cases where CRP levels raise 6 no concern, and where alternative anti-inflammatory therapies may be required. An upcoming 7 300 patient study using the same intracoronary sampling device will enable this finding and 8 the underlying mechanisms to be further investigated on a larger scale. 9 10 11 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 18 Acknowledgments 1 We would like to thank participants in the clinical trials detailed in this manuscript, and to the 2 hospital staff involved in blood sample collection and processing. 3 We are grateful for the open access to raw data from the PACIFIC, MGH Boston and Imperial 4 cohorts and to James Peters and Jack Gisby (Imperial College London) for additional data on 5 subcohort B and personal communication. 6 This study was funded by PlaqueTec Ltd. Collaborative work on IMPROVE with BG was 7 funded by Stiftelsen Sigurd & Elsa Goljes minne (LA2022-0133), Stiftelsen Professor Nanna 8 Swartz fond (2022-00472) and Hjärt-Lungfonden (20210472). IMPROVE was funded by the 9 Vth European Union (EU) program. 10 11 Disclosures 12 DP and SW are employees of PlaqueTec. NEJW was employed by PlaqueTec at the time that 13 this work was performed but is now an employee of Shockwave Medical. SPH is a consultant 14 to PlaqueTec. Part of the work is published in “International Patent Application No. 15 PCT/GB2022/051281, published as WO 2022/243703 A1”. 16 17 18 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 19

1 1. Baigent C, Keech A, Kearney PM, Blackwell L, Buck G, Pollicino C, et al. Efficacy and safety of 2 cholesterol-lowering treatment: prospective meta-analysis of data from 90,056 participants in 14 3 randomised trials of statins. Lancet. 2005;366(9493):1267-78. 4 2. Virani SS, Alonso A, Aparicio HJ, Benjamin EJ, Bittencourt MS, Callaway CW, et al. Heart 5 Disease and Stroke Statistics-2021 Update: A Report From the American Heart Association. 6 Circulation. 2021;143(8):e254-e743. 7 3. Yao C, Chen G, Song C, Keefe J, Mendelson M, Huan T, et al. Genome-wide mapping of 8 plasma protein QTLs identifies putatively causal genes and pathways for cardiovascular disease. Nat 9 Commun. 2018;9(1):3268. 10 4. Klarin D, Damrauer SM, Cho K, Sun YV, Teslovich TM, Honerlaw J, et al. Genetics of blood 11 lipids among ~300,000 multi-ethnic participants of the Million Veteran Program. Nat Genet. 12 2018;50(11):1514-23. 13 5. Kraus WE, Muoio DM, Stevens R, Craig D, Bain JR, Grass E, et al. Metabolomic Quantitative 14 Trait Loci (mQTL) Mapping Implicates the Ubiquitin Proteasome System in Cardiovascular Disease 15 Pathogenesis. PLoS Genet. 2015;11(11):e1005553. 16 6. Libby P, Ebert BL. CHIP (Clonal Hematopoiesis of Indeterminate Potential): Potent and Newly 17 Recognized Contributor to Cardiovascular Risk. Circulation. 2018;138(7):666-8. 18 7. Bom MJ, Levin E, Driessen RS, Danad I, Van Kuijk CC, van Rossum AC, et al. Predictive value 19 of targeted proteomics for coronary plaque morphology in patients with suspected coronary artery 20 disease. EBioMedicine. 2019;39:109-17. 21 8. Lepojarvi ES, Piira OP, Kiviniemi AM, Miettinen JA, Kentta T, Ukkola O, et al. Usefulness of 22 Highly Sensitive Troponin as a Predictor of Short-Term Outcome in Patients With Diabetes Mellitus 23 and Stable Coronary Artery Disease (from the ARTEMIS Study). Am J Cardiol. 2016;117(4):515-21. 24 9. Wallentin L, Eriksson N, Olszowka M, Grammer TB, Hagstrom E, Held C, et al. Plasma 25 proteins associated with cardiovascular death in patients with chronic coronary heart disease: A 26 retrospective study. PLoS Med. 2021;18(1):e1003513. 27 10. Engelen SE, Robinson AJB, Zurke Y-X, Monaco C. Therapeutic strategies targeting 28 inflammation and immunity in atherosclerosis: how to proceed? Nature Reviews Cardiology. 29 2022;19(8):522-42. 30 11. Aday AW, Ridker PM. Targeting Residual Inflammatory Risk: A Shifting Paradigm for 31 Atherosclerotic Disease. Front Cardiovasc Med. 2019;6:16. 32 12. Ridker PM, MacFadyen JG, Everett BM, Libby P, Thuren T, Glynn RJ, et al. Relationship of C-33 reactive protein reduction to cardiovascular event reduction following treatment with canakinumab: 34 a secondary analysis from the CANTOS randomised controlled trial. Lancet. 2018;391(10118):319-28. 35 13. Tardif JC, Kouz S, Waters DD, Bertrand OF, Diaz R, Maggioni AP, et al. Efficacy and Safety of 36 Low-Dose Colchicine after Myocardial Infarction. N Engl J Med. 2019;381(26):2497-505. 37 14. Ridker PM, Everett BM, Thuren T, MacFadyen JG, Chang WH, Ballantyne C, et al. 38 Antiinflammatory Therapy with Canakinumab for Atherosclerotic Disease. New England Journal of 39 Medicine. 2017;377(12):1119-31. 40 15. de Kleijn DP, Moll FL, Hellings WE, Ozsarlak-Sozer G, de Bruin P, Doevendans PA, et al. Local 41 atherosclerotic plaques are a source of prognostic biomarkers for adverse cardiovascular events. 42 Arterioscler Thromb Vasc Biol. 2010;30(3):612-9. 43 16. Lee R, Fischer R, Charles PD, Adlam D, Valli A, Di Gleria K, et al. A novel workflow combining 44 plaque imaging, plaque and plasma proteomics identifies biomarkers of human coronary 45 atherosclerotic plaque disruption. Clin Proteomics. 2017;14:22. 46 17. Alonso-Orgaz S, Moreno-Luna R, Lopez JA, Gil-Dones F, Padial LR, Moreu J, et al. Proteomic 47 characterization of human coronary thrombus in patients with ST-segment elevation acute 48 myocardial infarction. Journal of proteomics. 2014;109:368-81. 49 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 20 18. Ramaiola I, Padro T, Pena E, Juan-Babot O, Cubedo J, Martin-Yuste V, et al. Changes in 1 thrombus composition and profilin-1 release in acute myocardial infarction. Eur Heart J. 2 2015;36(16):965-75. 3 19. Funayama H, Ishikawa SE, Kubo N, Katayama T, Yasu T, Saito M, et al. Increases in 4 interleukin-6 and matrix metalloproteinase-9 in the infarct-related coronary artery of acute 5 myocardial infarction. Circ J. 2004;68(5):451-4. 6 20. Yong AS, Pennings GJ, Chang M, Hamzah A, Chung T, Qi M, et al. Intracoronary shear-related 7 up-regulation of platelet P-selectin and platelet-monocyte aggregation despite the use of aspirin and 8 clopidogrel. Blood. 2011;117(1):11-20. 9 21. Chandran S, Watkins J, Abdul-Aziz A, Shafat M, Calvert PA, Bowles KM, et al. Inflammatory 10 Differences in Plaque Erosion and Rupture in Patients With ST-Segment Elevation Myocardial 11 Infarction. J Am Heart Assoc. 2017;6(5). 12 22. Bambrough P, Peverelli M, Brown AJ, Giblett JP, Bennett MR, West NEJ, et al. Trans-13 Myocardial Blood Interleukin-6 Levels Relate to Intracoronary Imaging-Defined Features of Plaque 14 Vulnerability and Predict Procedure-Induced Myocardial Infarction. Cardiovasc Revasc Med. 15 2022;39:6-11. 16 23. West NEJ, Corrigan JP, Owen RHG, Hoole SP, Brown AJ, Blatcher S, et al. Percutaneous 17 Sampling of Local Biomolecule Gradients Across Coronary Artery Atherosclerotic Plaques. JACC Basic 18 to translational science. 2017;2(6):646-54. 19 24. Baldassarre D, Nyyssönen K, Rauramaa R, de Faire U, Hamsten A, Smit AJ, et al. Cross-20 sectional analysis of baseline data to identify the major determinants of carotid intima-media 21 thickness in a European population: the IMPROVE study. Eur Heart J. 2010;31(5):614-22. 22 25. Baldassarre D, Hamsten A, Veglia F, de Faire U, Humphries SE, Smit AJ, et al. Measurements 23 of carotid intima-media thickness and of interadventitia common carotid diameter improve 24 prediction of cardiovascular events: results of the IMPROVE (Carotid Intima Media Thickness [IMT] 25 and IMT-Progression as Predictors of Vascular Events in a High Risk European Population) study. J 26 Am Coll Cardiol. 2012;60(16):1489-99. 27 26. Folkersen L, Fauman E, Sabater-Lleal M, Strawbridge RJ, Frånberg M, Sennblad B, et al. 28 Mapping of 79 loci for 83 plasma protein biomarkers in cardiovascular disease. PLoS Genet. 29 2017;13(4):e1006706. 30 27. Filbin MR, Mehta A, Schneider AM, Kays KR, Guess JR, Gentili M, et al. Longitudinal 31 proteomic analysis of severe COVID-19 reveals survival-associated signatures, tissue-specific cell 32 death, and cell-cell interactions. Cell Rep Med. 2021;2(5):100287. 33 28. Gisby J, Clarke CL, Medjeral-Thomas N, Malik TH, Papadaki A, Mortimer PM, et al. 34 Longitudinal proteomic profiling of dialysis patients with COVID-19 reveals markers of severity and 35 predictors of death. Elife. 2021;10. 36 29. Malatino LS, Mallamaci F, Benedetto FA, Bellanuova I, Cataliotti A, Tripepi G, et al. 37 Hepatocyte growth factor predicts survival and relates to inflammation and intima media thickness 38 in end-stage renal disease. Am J Kidney Dis. 2000;36(5):945-52. 39 30. Meizlish ML, Pine AB, Bishai JD, Goshua G, Nadelmann ER, Simonov M, et al. A neutrophil 40 activation signature predicts critical illness and mortality in COVID-19. Blood Advances. 41 2021;5(5):1164-77. 42 31. Folkersen L, Gustafsson S, Wang Q, Hansen DH, Hedman ÅK, Schork A, et al. Genomic and 43 drug target evaluation of 90 cardiovascular proteins in 30,931 individuals. Nature Metabolism. 44 2020;2(10):1135-48. 45 32. Diaz-Lopez A, Bullo M, Chacon MR, Estruch R, Vendrell J, Diez-Espino J, et al. Reduced 46 circulating sTWEAK levels are associated with metabolic syndrome in elderly individuals at high 47 cardiovascular risk. Cardiovasc Diabetol. 2014;13:51. 48 33. da Silva EZM, Jamur MC, Oliver C. Mast Cell Function:A New Vision of an Old Cell. Journal of 49 Histochemistry & Cytochemistry. 2014;62(10):698-738. 50 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 21 34. Morici N, Farioli L, Losappio LM, Colombo G, Nichelatti M, Preziosi D, et al. Mast cells and 1 acute coronary syndromes: relationship between serum tryptase, clinical outcome and severity of 2 coronary artery disease. Open Heart. 2016;3(2):e000472. 3 35. Kaura A, Hartley A, Panoulas V, Glampson B, Shah ASV, Davies J, et al. Mortality risk 4 prediction of high-sensitivity C-reactive protein in suspected acute coronary syndrome: A cohort 5 study. PLoS Med. 2022;19(2):e1003911. 6 36. Decker PA, Larson NB, Bell EJ, Pankow JS, Hanson NQ, Wassel CL, et al. Increased hepatocyte 7 growth factor levels over 2 years are associated with coronary heart disease: the Multi-Ethnic Study 8 of Atherosclerosis (MESA). American heart journal. 2019;213:30-4. 9 37. Haljasmägi L, Salumets A, Rumm AP, Jürgenson M, Krassohhina E, Remm A, et al. 10 Longitudinal proteomic profiling reveals increased early inflammation and sustained apoptosis 11 proteins in severe COVID-19. Scientific Reports. 2020;10(1):20533. 12 38. Nilsson E, Kastrup J, Sajadieh A, Boje Jensen G, Kjøller E, Kolmos HJ, et al. Pregnancy 13 Associated Plasma Protein-A as a Cardiovascular Risk Marker in Patients with Stable Coronary Heart 14 Disease During 10 Years Follow-Up-A CLARICOR Trial Sub-Study. Journal of clinical medicine. 15 2020;9(1). 16 39. Klimczak-Tomaniak D, Bouwens E, Schuurman A-S, Akkerhuis KM, Constantinescu A, Brugts J, 17 et al. Temporal patterns of macrophage- and neutrophil-related markers are associated with clinical 18 outcome in heart failure patients. ESC Heart Failure. 2020;7(3):1190-200. 19 40. Bernardi S, Voltan R, Rimondi E, Melloni E, Milani D, Cervellati C, et al. TRAIL, OPG, and 20 TWEAK in kidney disease: biomarkers or therapeutic targets? Clin Sci (Lond). 2019;133(10):1145-66. 21 41. McMahon M, Skaggs BJ, Grossman JM, Sahakian L, Fitzgerald J, Wong WK, et al. A panel of 22 biomarkers is associated with increased risk of the presence and progression of atherosclerosis in 23 women with systemic lupus erythematosus. Arthritis Rheumatol. 2014;66(1):130-9. 24 42. Winckers K, Siegerink B, Duckers C, Maurissen LF, Tans G, Castoldi E, et al. Increased tissue 25 factor pathway inhibitor activity is associated with myocardial infarction in young women: results 26 from the RATIO study. J Thromb Haemost. 2011;9(11):2243-50. 27 28 29 30 31 32 33 34 35 36 37 38 39 40 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 22 Tables 1 2 3 Patient ID Sex BMI Smoking Diabetes Cardiac status Statin Peripheral Sample location 07 Male 41 NA No Stable angina Atorvastatin Sheath 09 Male 32 Former No Stable angina Simvastatin Sheath 10 Male 38 Former Non-IDDM Stable angina Simvastatin Sheath 22 Male 27 Former No Stable angina Simvastatin Guide catheter 27 Male 44 Current No Stable angina Simvastatin Cubital vein 43 Male 26 Current No Stable angina Atorvastatin Cubital vein 44 Male 31 Never No Stable angina Atorvastatin Sheath 45 Male 42 Former Non-IDDM Stable angina Simvastatin Cubital vein 53 Male 26 Former No Unstable angina Atorvastatin Sheath 54 Male 28 Former No Stable angina Atorvastatin Sheath 58 Male 28 Former Non-IDDM Stable angina Simvastatin Sheath 65 Male 20 Current No NSTEMI Simvastatin Sheath 4 Table 1. Demographic characteristics for the PlaqueTec CS1 trial. 5 All patients with blood samples collected before PCI are listed. Locations of the blood sample taken 6 peripherally, prior to heparin infusion. Note details for the 2 patients with the biosignature at baseline, 7 27 and 58, are displayed in bold. Abbreviations: insulin-dependent diabetes, IDDM; Non-ST -elevation 8 myocardial infarction, NSTEMI. 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 23 1 A 2 Biomarker score 0 (n=2862) Biomarker score 1 (n=39) Age (y) 65 (60-67) 65 (62-70) Gender (M/F) 1314/1548 26/13 Cardiovascular risk factors n (%) Family History CHD 1660 (58) 25 (66) Smoking 401 (14) 5 (13) Diabetes 789 (28) 13 (33) Hypertension 2216 (77) 35 (90) BMI (kg/m2) 27 (24-29) 28 (25-31) Biochemistry LDL-cholesterol (mmol/L) 3.5 (2.9-4.3) 3.7 (2.8-4.3) HDL-cholesterol (mmol/L) 1.20 (1.01-1.46) 1.13 (0.94-1.29) Triglycerides (mmol/L) 1.34 (0.96-1.95) 1.87 (1.28-2.67) CRP (mg/L) 1.92 (0.82-3.67) 3.2 (2.49-4.63) Drugs Statin 1348 (40) 10 (24) Antiplatelet 566 (17) 0 Ultrasonographic Measures C-IMT mean (mm) 0.83 (0.73-0.98) 0.91 (0.82-1.06) C-IMT max (mm) 1.84 (1.39-2.41) 2.22 (1.64 -2.71) C-IMT mean-max (mm) 1.17 (1.02-1.37) 1.29 (1.15-1.55) Presence of plaque (n) (%) 1898 (66) 31 (79) 3 B 4 N b (SE) p Biomarker score c-IMTmean 2900 .039 (.29) 0.178 c-IMTmax 2900 .026 ( .012) 0.033 c-IMTmean-max 2901 .0106 ( .042) 0.158 5 C 6 HGF PAPPA SPON1 HGF 1 PAPPA 0.37* 1 SPON1 0.55* 0.38* 1 CRP 0.2275* -0.1030 0.0391 7 Table 2. IMPROVE cohort and biosignature 8 A. Demographic characteristics for IMPROVE according to the plasma levels for HGF, and also PAPPA 9 and SPON1 included in the biomarker score categorized as below (0) and over (1) the 90th percentile 10 for all three proteins. Continuous variables are reported as median (interquartile range) and categorial 11 variables as number (percentages); CHD: coronary heart disease; Presence of plaque: number of 12 study participants where a plaque has been recorded. 13 B. Association of the biomarker score 1 (> 90 th percentile) with measures of c-IMT mean or maximum 14 (max) after adjustment by age, gender and latitude by linear regression. (Note that latitude is highly 15 correlated with c-IMT and plaque). 16 C. Pairwise correlation between the biomarkers in analysis expressed by Spearman rho coefficient, 17 * p<0.0001 18 19 20 21 22 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 24 Cohort Biosignature (n) No biosignature (n) Biosignature prevalence (%) Biosignature CV event (%) No biosignature CV event (%) Biosignature mortality (%) No biosignature mortality (%) PlaqueTec CS1 - CVD patients undergoing PCI procedure 2 10 16.7 NA NA NA NA PACIFIC - participants with suspected CVD undergoing CCTA 7 3 193 1.5 NA NA NA NA IMPROVE - participants with >3 risk factors for CVD, 3 year follow up for CVD events 39 2862 1.3 20.5 6 NA NA MGH Boston cohort - majority COVID-19 positive, n=358 day 0 samples 27 11 347 3.1 45.5 7.7 63.6 11.3 Imperial College ESKD subcohort A COVID-19 positive, first samples 28 2 53 3.6 NA NA 100 13.2 Imperial College ESKD subcohort B COVID-19 positive, serum 28 2 44 4.3 NA NA 0 21.7 Imperial College ESKD subcohort A COVID-19 negative 28 1 50 2.0 NA NA NA NA Imperial College ESKD subcohort B COVID-19 negative, serum 28 0 11 0 NA NA NA NA 1 Table 3. Summary of biosignature prevalence and outcomes in different cohorts 2 The table displays comparisons of biosignature prevalence and outcomes between cardiovascular disease 3 cohorts, a respiratory disease/COVID-19 cohort and an ESKD/COVID-19 cohort. To identify participants with 4 the biosignature a definition of >90th centile for HGF, and also >90th centile for PAPPA and SPON1 levels 5 was used. Individuals identified with the biosignature were in the minority, but a had higher incidence of 6 cardiovascular events in IMPROVE and the respiratory disease/COVID-19 cohort, and a higher mortality in the 7 respiratory disease/COVID-19 cohort and the ESKD/COVID-19 subcohort A compared with individuals without 8 the biosignature. Abbreviations: CV, cardiovascular; CVD, cardiovascular disease; CCTA, coronary computed 9 tomography angiography; ESKD, end-stage kidney disease; NA, not available. 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 25 1 Cohort Biosignature CRP (mg/L) No biosignature CRP (mg/L) CRP higher in Biosignature group? Biosignature proteins correlate with CRP Biosignature proteins correlating positively with MPO Biosignature proteins correlating positively with tryptase Biosignature proteins correlating positively with syndecan-1 CVD PlaqueTec CS1 3.4 (2.1-4.7) 12.8 (0.7-15) no no HGF, TFPI, CCL28 NA NA CVD PACIFIC 7 <2.5 2.5 n=33 no NA HGF, PAPPA, SPON1, TWEAK, TFPI NA NA CVD IMPROVE 3.2 (2.49-4.63) 1.92 (0.82- 3.67) Yes HGF HGF, PAPPA, SPON1 NA NA COVID-19 27 0-180+ (Category 1-5) 0-180+ (Category 1-5) no NA NA HGF, PAPPA, SPON1, TWEAK, TFPI, CCL28 HGF, PAPPA, SPON1, TFPI, CCL28 ESKD subcohort A 28 220.7 (163.5- 278) 79.9 (0.6- 386.1) no HGF HGF, PAPPA, SPON1, TWEAK, TFPI, CCL28 HGF NA ESKD subcohort B 28 197.6 (34.4- 360.7) 155.6 (18.3- 405.6) no NA none TWEAK, CCL28 NA 2 Table 4. Comparison of biosignature CRP levels and correlations with MPO, tryptase and 3 syndecan-1 in different cohorts 4 For CRP levels, medians (range) are shown. For correlations, proteins with a positive and significant (p<0.05) 5 Spearman pairwise correlation are shown. Results of correlation analysis are detailed in the Supplement File 6 S1. NA, not available. 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 26 1 Protein (Uniprot ID) Function Plasma levels and CVD outcomes HGF (P14210) Protective role in atherosclerosis: anti -apoptotic, anti-fibrotic, pro-angiogenic, anti- inflammatory. Linked with unstable carotid atherosclerosis and elevated triglyceride levels. High levels linked with CVD risk and mortality(9, 36) Causal role in raised triglyceride levels(31) Increased risk of death in COVID- 19 cohort(37) PAPPA (Q13219) Metalloprotease that cleaves IGF binding proteins, allowing IGF to bind to its receptor, which can have several effects on vascular cells including release of inflammatory cytokines. High levels linked unstable plaques and type 2 diabetes High levels linked with CVD risk(38) Causal role in Type 2 diabetes(31) SPON1 (Q9HCB6) Inhibitor of calcification in bone. Regulated by fibroblast growth factor 2 (FGF2), retinoic acid, IGF and Toll-like Receptor-4 agonists. Upregulated in osteoarthritis. Inhibitor of vascular smooth muscle cell proliferation and migration. Role in atrial fibrillation (AF). Higher levels linked with adverse events in CHD cohort (39) and mortality(9) Causal role in AF(31) Increased risk of death in COVID- 19 cohort(28) TWEAK (O43508) Binds to TWEAKR inducing apoptosis, proliferation and migration of endothelial cells. Stimulates cytokine secretion and chemokine IL -8, via NFkB activation. Linked with endothelial dysfunction. Low levels linked with higher CVD risk(40) Higher levels with higher CVD risk in SLE and controls (41). Higher levels with lower death rates in COVID-19 cohort(28) TFPI (P10646) Binds factor X and interacts with lipoproteins. Anti- thrombotic and associated with inflammation and endothelial dysfunction. High levels linked with cardiovascular damage and myocardial infarction(42) CCL28 (Q9NRJ3) Chemokine that binds receptors CCR10 and CCR3 and exerts chemotaxis in T, B cells and eosinophils. Induced by pro -inflammatory cytokines and bacteria. CVD not directly tested Predictor of death in a COVID-19 cohort(28) 2 Table 5. Summary of functions of the biosignature proteins 3 4 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 27 Figure Legends 1 2 Figure 1. Comparison of coronary and peripheral protein levels – CVD1 protein panel 3 A. Protein levels in plasma samples from 12 patients from the CS1 cohort, expressed as ratio of 4 coronary/peripheral levels (CVD1 panel proteins on x-axis). Each datapoint represents an individual 5 patient, with each patient displayed by a different colour. Data from proximal coronary plasma 6 samples is shown (see Supplement Figure 1 for distal coronary data). Note 3 proteins indicated 7 have high coronary/peripheral protein ratios for 10 of the 12 patients. 8 B. HGF absolute protein levels (NPX units) in coronary and peripheral plasma, pre- (n=12) and post-9 PCI procedure (n=11). Each datapoint represents an individual patient. Note that 2 patients had 10 much higher levels of HGF (shown in red) in peripheral samples prior to PCI compared with the 11 other 10 patients. 12 C. As for B with absolute levels of PAPPA (NPX units). Note that 2 patients had much higher levels of 13 PAPPA (shown in red) in peripheral samples prior to PCI compared with the other 10 patients. 14 D. As for C with absolute levels of SPON1 (NPX units). Note that 2 patients had much higher levels 15 of SPON1 (shown in red) in peripheral samples prior to PCI compared with the other 10 patients. 16 E. Post-PCI/pre-PCI ratio of systemic samples (CVD1 panel). Growth Hormone (GH). Note similar 17 patterns to A, indicating that for 10 of the 12 patients, conditions of the PCI procedure raised the 18 peripheral levels of these specific proteins. For 2 patients, these proteins were already at high levels 19 peripherally at baseline, before the procedure. 20 21 Figure 2. Comparison of coronary and peripheral protein levels repeat analysis of CS1 plasma 22 in CVD3 and Inflammation protein panels 23 A. Repeatability measurements of CS1 coronary and peripheral samples for HGF in Inflammation 24 panel and comparison with healthy controls (age- and sex-matched for the CS1 cohort). Note the 25 2 patients with much higher peripheral levels of HGF and SPON1 are shown in red. 26 B. As for A with but with SPON1 using the CVD3 panel. (Note that patterns for HGF and SPON1 27 measured on the CVD1 panels in Fig 1B and D were similar, but NPX values differed between the 28 2 panels due to different dilution factors and dynamic ranges of the assays. 29 C. Coronary:peripheral protein ratios, Inflammation panel. Comparing with Figure 1A, a similar HGF 30 pattern was observed. Further proteins with similar patterns were identified: CXCL9, CCL28 and 31 TWEAK. 32 D. Coronary:peripheral protein level ratios, CVD3 panel. Comparing with Figure 1A, a similar but more 33 distinct SPON1 pattern was observed. Further proteins with similar patterns were identified: TFPI 34 and AZU1. 35 36 Figure 3. CS1 patient and healthy control peripheral proteins levels 37 A. Individual patient absolute levels (NPX) for selected proteins, coronary (red) and peripheral (black), 38 and comparing with 12 healthy, age- and sex-matched control plasma samples, data from the 39 CVD3 and Inflammation panels. These raw data plots revealed that levels of these proteins in 40 coronary arteries of patients are clearly higher than healthy, age- and sex -matched control 41 peripheral levels (n=12). This effect was less pronounced for AZU1 and CXCL9. PAPPA control 42 levels were not determined as PAPPA is not included on the CVD3 and Inflammation panels. Thus 43 2 patients had a clear systemic pattern of elevated HGF, SPON1, TFPI, CCL28, TWEAK and 44 PAPPA. 45 B. No correlation of hsCRP with HGF levels Spearman R = -0.01 (P = 0.99). 46 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint 28 Figure 4. Comparison of CS1 with PACIFIC cohort 1 Absolute levels for peripheral proteins are displayed for CS1 cohort (left) and data was downloaded 2 from Bom et al 7, PACIFIC cohort (right). Age-matched control levels can be compared with CS1 3 cohort as they were analysed together (PAPPA was not included). NPX levels between CS1 and 4 PACIFIC cohorts cannot be compared directly, but patterns can be identified, displaying similarities 5 in the expression levels of the 6 proteins. 2 participants in the PACIFIC cohort displayed distinctly 6 higher levels of the 6 proteins compared with the majority of participants (indicated in red). Using a 7 definition of >90th centile for HGF and also PAPPA and SPON1 identified 1 additional participant 8 (indicated in red). 9 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint Graphical Abstract . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint A B C E D coronary/peripheral peripheral post-PCI/pre-PCI GH Pre-PCI Post-PCI Pre-PCI Post-PCI Pre-PCI Post-PCI Figure 1 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint Inflammation panel D C A B CVD3 panel Figure 2 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint A B Figure 3 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint CS1 cohort PACIFIC cohort Protein levels (NPX) Patient ID Participant ID CS1 age-matched healthy Individual participants Figure 4 . CC-BY-NC-ND 4.0 International licenseIt is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint The copyright holder for thisthis version posted December 15, 2023. ; https://doi.org/10.1101/2023.04.06.23288168doi: medRxiv preprint