key: cord-0844080-6xv0zse4
authors: Michels, James R.; Nazrul, Mohammad Shaheed; Adhikari, Sudeep; Wilkins, Dawn; Pavel, Ana B.
title: Th1, Th2 and Th17 inflammatory pathways synergistically correlate with cardiometabolic processes. A case study in COVID-19
date: 2021-12-04
journal: bioRxiv
DOI: 10.1101/2021.11.29.470414
sha: e89d8398a1e86adabb22593e3e7404ce5bb6720e
doc_id: 844080
cord_uid: 6xv0zse4

A predominant source of complication in SARS-CoV-2 patients arises from the cytokine storm, an elevated expression of inflammatory helper T-cell associated cytokines that can lead to tissue damage and organ failure. The high inflammatory burden of this viral infection often results in cardiovascular comorbidities. A better understanding of the interaction between the cytokine storm and cardiovascular proteins might inform medical decisions and therapeutic approaches. We hypothesized that all major helper T-cell inflammatory pathways (Th1, Th2 and Th17) synergistically contribute to cardiometabolic modifications in serum of COVID-19 patients. We proved our hypothesis by integrating Th1, Th2 and Th17 cytokines to predict expression of cardiometabolic proteins profiled by OLINK proteomics.

The respiratory virus SARS-CoV-2 has quickly spread around the world, resulting in a total of over 200 million reported cases and over 5 million reported deceased to date. 1 Despite incredible advancements in the development of vaccines aiding the prevention of cases, targeted treatments toward COVID-19 symptoms and effects are still needed. Complications such as the acute respiratory distress syndrome (ARDS), respiratory failure, hepatic and renal insufficiency are common in more severe cases. 2 Proteomic profiles are currently being investigated in order to study the effect of these cytokines in COVID-19 disease. 3 A better understanding of cytokines interactions with the cardiovascular system, might enhance development of novel immunomodulatory therapies and reduce mortality in COVID-19 vulnerable patients.

Previous studies have shown various helper T-cell responses in SARS-CoV-2 infections. One study in elderly SARS-CoV-2 patients found that the predominant response to infection was in the form of Th1-associated cytokines. However, the levels of Th17associated cytokines were also detected in COVID-19 severe disease. [4] [5] [6] , and Th2 secretions were associated to SARS-CoV-2 susceptibility. 7, 8 We have recently shown an association between a shift in Th1 network that severely affects cardiometabolic processes.

In this study we used the publicly available Massachusetts General Hospital COVID-19 registry 12 consisting of 383 patients (after removal of one patient with missing data), where 305 were COVID-19 positive and 78 were COVID-19 negative with other respiratory infections. Patients in this data set were classified by their severity as 42 COVID-19 positive and 7 COVID-19 negative deceased patients (who max score = 1), 67 COVID-19 positive and 16 COVID-19 negative intubated patients (who max score = 2) and 196 COVID-19 positive and 55 COVID-19 negative non-severe patients (who max score ≥ 3).

We included in our analysis serum proteins profiled by OLINK Explore platform, 12 such the Cardiometabolic panel consisting of 355 detected proteins, and all available Th1, Th2 and Th17 cytokines as previously described. 8 OLINK proteomics platform has been extensively used to profile targeted biomarkers and drug targets associated with various diseases and tissue types. 8, 17 The helper T-cell pathways were defined by 11 Th1 markers (CCL3, CCL4, CXCL11, CXCL10, CXCL9, IL2RA, IFNG, IFNGR1, IFNGR2, IL12B and IL1B), 14 Th2 markers (CCL11, CCL13, CCL17, CCL22, CCL24, CCL26, CCL7, IL10, IL13, IL33, IL4R, IL5, IL7R and TSLP) and 13 Th17 markers (CCL20, S100P, IL6, IL6R, LCN2, S100A12, CXCL1, PI3, IL17A, IL17F, CXCL3, IL12A and IL12B). 8

We first applied a linear regression model to associate the normalized expression (NPX) for each immune or cardiometabolic protein (dependent variable) with disease severity (independent variable) in both COVID-19 positive and negative patients. We then displayed heatmaps of the mean estimates of deceased, intubated and non-severe groups for all Th1, Th2 and Th17 cytokines, and all significantly differentially expressed car- COVID-19 positive patients   LTBP2  RNASE3  CHI3L1  CSTB  CA3  IGFBP1  SPP1  TNNI3  GPR37  CEACAM8  MB  GP2  RETN  GDF15  CXCL8  COL6A3  NPPB  FAM3C  NPDC1  TNC  DCN  IGFBP2  ACTA2  NTPROBNP  TFF3  REG3A  FABP4  REG1A  CCN3  CHIT1  REG1B  IL19  PLA2G2A  IL1RL1 (35 proteins) . Differentially expressed proteins were determined by | f old −change| ≥ 2 and FDR < 0.05 in any comparison (deceased versus non-severe, intubated versus non-severe or deceased versus intubated). All 35 proteins achieved statistical significance between deceased and non-severe groups. B. Heatmap of estimated mean expression of top significantly differentially expressed cardiometabolic proteins stratified by severity of COVID-19 negative patients with other respiratory conditions (8 proteins). Differentially expressed proteins were determined by | f old − change| ≥ 2 and FDR < 0.05 in any comparison (deceased versus nonsevere, intubated versus non-severe or deceased versus intubated). None of these 8 proteins achieved significance between deceased versus nonsevere patients. The heatmaps are z-score normalized with blue denoting decreased expression and red denoting increased expression. diometabolic markers between any comparison (deceased versus non-severe, intubated versus non-severe or deceased versus intubated) by False Discovery Rate (FDR) < 0.05.

Next, we integrated all Th1, Th2 and Th17 immune mediators by elastic net regularized generalized linear models with 10-fold cross-validation, using cv.glmnet function from glmnet R package, to predict expression levels of each cardiometabolic protein on the OLINK panel based on helper T-cell immune profiles.

We then ranked the best predictions among all 355 cardiometabolic markers by Pearson correlation coefficient calculated between the real and predicted value, and considered as best fit those correlations with r ≥ 0.7 and FDR < 0.05.

To visualize the interaction between immune and cardiometabolic markers with r ≥ 0.7 we used igraph R package. The edges of the network represent non-zero coefficients of the elastic net regression model with absolute value ≥ 0.05. We used this threshold to exclude weak interactions and filter noise.

We performed all statistical analyses using R programming language.

We first analyzed all cytokines of T-helper cell types pathways (Th1, Th2 and Th17) in both COVID-19 positive and COVID-19 negative patients and displayed them as heatmaps of mean expression estimates stratified by disease severity (Figure 1 ). We observed an increasing trend in protein expression from non-severe to deceased COVID-19 patients in most of Th1 and Th17 mediators and in more than 50% of Th2 mediators. We found significant increases (p − value < 0.05) in Th1 (IFNGR1, CCL3, CCL4, CCL5, CXCL9, CXCL10, IL1B, IL2RA), Th2 (CCL11, CCL13, CCL7, CCL24, IL4R) and Th17 (S100A12, S100P, CCL20, LCN2, PI3, CXCL1, IL17A, IL6) mediators in deceased as compared to nonsevere COVID-19 patients (Figure 1 A, Supplementary Table 1 ). In contrast, COVID-19 negative patients with other respiratory infections did not show significant increasing trends in deceased as compared to non-severe patients in the Th1, Th2 and Th17 pathways (Figure 1 B, Supplementary Table 2 ).

Next, we evaluated 355 cardiometabolic proteins detected by OLINK Explore Cardiometabolic panel. We found that 35 of these proteins were strongly associated with COVID-19 severity (| f old − change| ≥ 2, FDR < 0.05) as shown in Figure 2 A (Supplementary Table 3 ). All these 35 differentially expressed proteins were significantly increased in deceased compared to non-severe COVID-19 patients, and several of them (i.e. LTBP2, RNASE3, CHI3L1, CSTB, RETN, GDF15, CXCL8, PLA2G2A, IL1RL1 and NADK) also showed significantly elevated expression in intubated compared to the non-severe patients.

In contrast to this strong cardiometabolic signal associated with COVID-19 severity, COVID-19 negative patients with other respiratory infections showed no statistically significant differences between deceased and non-severe patients, and only 8 proteins were increased in intubated as compared to non-severe patients (Figure 2 B, Supplementary Table 4 ).

Next, we sought to test our hypothesis that all Th1, Th2 and Th17 cytokines contribute to cardiometabolic modifications associated with SARS-CoV-2 infections. We integrated all Th1, Th2 and Th17 immune markers by elastic net regularized linear regression to predict the abundance of each cardiometabolic protein in COVID-19 positive patients. By this approach we were able to rank all the cytokines and immune mediators by their predictive potential.

We evaluated the best fit of our predictions by Pearson correlation coefficient computed between the measured and predicted expression values for each of the 355 cardiometabolic proteins (Supplementary Table 5 ). We identified 186 significant predictions (r ≥ 0.7 and FDR < 0.05), and further represented them as a graph structure, where the edges (links) represent the nonzero coefficients (abs ≥ 0.05) corresponding to each cytokine. We found that all Th1, Th2 and Th17 immune pathways synergistically describe the expression of these 186 cardiometabolic proteins. Most of the links in our networks were positive connections (Figure 3) , suggesting that an increased production in cytokines stimulates the overall production of cardiometabolic proteins. Figure 3 A shows all positive links between Th1 (yellowgreen), Th2 (blue) and Th17 (green) cytokines and the significantly predicted cardiometabolic proteins (gray), highlighting the extent of the immune system's impact on cardiometabolic processes.

We found that 31 of the 35 cardiometabolic proteins associated with COVID-19 severity (Figure 2 A) were also significantly predicted by the cytokine storm, except for CA3, TNNI3, NPPB, CHIT1 which did not reach our prediction threshold for the "goodness of fit" (r ≥ 0.7 and FDR < 0.05). Most of the cytokines and immune mediators positively contributed to these predictions. In Figure 4 we highlighted 20 of these cardiometabolic proteins that were both significantly associated with COVID-19 severity and significantly predicted by the cytokine storm, and in addition were found to be associated with cardiovascular inflammation and hypertension in other studies. [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] Across these 20 cardiovascular markers, the most common association with Th1 pathway was represented by IFNGR1 (85%), while the most common Th2 associations included CCL11 (50%) and CCL7 (40%), and the most common Th17 associations included PI3 (50%), LCN2 (50%) and IL6 (45%).

In this paper we explored the relationships between key immune mediators of COVID-19 cytokine storm and cardiometabolic proteins measured by OLINK Explore platform. 12 Using predictive modeling, we integrated Th1, Th2 and Th17 cytokines and found 186/355 significant cardiometabolic predictions. Of these 186 cardiometabolic proteins, 31 were also significantly associated with COVID-19 severity. We highlighted 20 of these 31 proteins that were also associated with cardiovascular inflammation and hypertension in previous studies. For example, LTBP2 has been previously reported as a marker of human heart failure 18 . CHI3L1 levels have been correlated with severity of coronary disease 19 and carotid atherosclerotic plaque, 20 as well as stroke 21 . Elevated GDF15 levels have been associated with higher risks in multiple cardiovascular diseases such as stable coronary artery disease, acute coronary syndrome, and heart failure 22 . Elevated CXCL8 expression levels have been identified in cases of atherosclerotic plaque 23 . SPP1 expression has been found to be higher in response to ischemia associated with stroke 24 , myocardial infarction 25 , and peripheral artery disease 26 . FABP4 has been found to contribute to the development of atherosclerosis, and studies had shown that lower levels of FABP4 protect against atherosclerosis to a degree 38 . REG1A has been shown to have high levels of expression in hearts of patients who died of myocardial infarction 27 . TFF3 levels in sera have been linked to the prediction of major adverse cardiovascular events 28 as well as being identified as a possible biomarker for myocardial infarction 29 . TNC has been previously linked to many cardiovascular diseases in humans such as pulmonary thromboembolism 30 and hypertension 31 . IL1RL1 has been studied as a marker of cardiac disease, and found to have increased expression in the lung of patients with heart failure 32 . CSTB has been identified as a relevant biomarker associated with chronic heart failure patients 33 . NT-proBNP has been shown to be a reliable biomarker in diagnostic evaluation and outcome prediction in cases of acute heart failure, especially in dyspnoeic patients 34, 35 . One study suggested that NTproBNP been seen as a signal of heart failure, valvular heart diseases, pulmonary hypertension 36 . Higher IGFBP1 expression has been previously related to lessened cardiovascular risk factors and decreased presence of atherosclerosis in elderly patients 37 .

Our study highlights the potential role of helper T-cells in the production of cardiometabolic proteins in SARS-CoV-2 infection, suggesting the association of the severe cytokine storm with cardiovascular-associated complications. While Type 1 and Type 17 helper T-cells have been extensively associated with the immune system's response to SARS-CoV-2 infection, the role of Type 2 helper T-cells is still poorly understood. Type 2 inflammation characterizes allergic and autoimmune reactions and has been previously associated with an increased risk of vascular inflammation. 10, 11 In addition, Th2 inhibition has been suggested to offer protection against COVID-19 symptoms 17, 39 , and hence Th2 inhibitors are currently being tested in cases of patients with atopic conditions 9 . Our study suggests that in synergy with Th1, both Th2 and Th17 pathways play an important role in the overproduction of cardiovascular-associated proteins. While Th1 cytokines are important mediators in fighting against any viral infection, Th2 and Th17 pathways are signals triggered by the severe autoimmune response to the unknown pathogen. Hence inhibiting Th2 and Th17 cytokines specifically with immunomodulators may help reduce cardiovascular inflammation without reducing the body's immune capabilities to fight the infection.

We acknowledge the use of a limited OLINK assay analysis rather than a whole genome analysis, and the potential misdiagnosis of COVID-19 cases during the first wave of the pandemic as limitations of our study.

In summary, while additional research to explore the link between helper T-cells and cardiovascular inflammation is needed, our data suggests that major immune axes (Th1, Th2 and Th17 pathways) are synergistically linked to a myriad of cardiometabolic proteins, which may potentially explain the cardiovascular complications associated with cytokine storm in COVID-19 patients. Future work will include further investigation of the relationships between immune and cardiovascular pathways and a detailed comparisons with other viral infections and autoimmune conditions to create a more complete map of immunecardiovascular interaction of human body.

JRM contributed to statistical analysis, data visualization and writing the manuscript. MSN and SA contributed to data visualization. DW provided valuable insight into the methodology. ABP designed the study, performed statistical analysis and data interpretation, and wrote the manuscript.

There are no conflicts to declare.

The Lancet

Jornal de Pediatria

Vascular disease

The authors would like to a acknowledge MGH for making the COVID-19 registry available and the Department of Biomedical Engineering at UM for in kind support.