key: cord-0335640-cuhmnir0
authors: Broderick, C.; Rivero Calle, I.; Gomez Carballa, A.; Gomez-Rial, J.; Li, H. K.; Mehta, R.; Jackson, H.; Salas, A.; Martinon-Torres, F.; Sriskandan, S.; Levin, M.; Kaforou, M.; BioAID Consortium,; Group, GEN-COVID Study
title: Pseudotemporal whole blood transcriptional profiling of COVID-19 patients stratified by clinical severity reveals differences in immune responses and possible role of monoamine oxidase B
date: 2022-05-27
journal: nan
DOI: 10.1101/2022.05.26.22274729
sha: b43dddbc929a57aecb728710aed057357b81de42
doc_id: 335640
cord_uid: cuhmnir0

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection is associated with highly variable clinical outcomes. Studying the temporal dynamics of host whole blood gene expression during SARS-CoV-2 infection can elucidate the biological processes that underlie these diverse clinical phenotypes. We employed a novel pseudotemporal approach using MaSigPro to model and compare the trajectories of whole blood transcriptomic responses in patients with mild, moderate and severe COVID-19 disease. We identified 5,267 genes significantly differentially expressed (SDE) over pseudotime and between severity groups and clustered these genes together based on pseudotemporal trends. Pathway analysis of these gene clusters revealed upregulation of multiple immune, coagulation, platelet and senescence pathways with increasing disease severity and downregulation of T cell, transcriptional and cellular metabolic pathways. The gene clusters exhibited differing pseudotemporal trends. Monoamine oxidase B was the top SDE gene, upregulated in severe>moderate>mild COVID-19 disease. This work provides new insights into the diversity of the host response to SARS-CoV-2 and disease severity and highlights the utility of pseudotemporal approaches in studying evolving immune responses to infectious diseases.

Outcomes from severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection vary 44

greatly from asymptomatic infection to severe, life-threatening multi-system disease. An 45 exaggerated hyperinflammatory immune response with neutrophil and macrophage activation and 46 "cytokine storm" is associated with severe clinical phenotypes [1-6], as well as impaired type 1 47 interferon signaling [7, 8] and T cell dysfunction [9] [10] [11] [12] , suggesting that severe disease is the 48 outcome of a complex interplay between the various components of the immune response and 49 SARS-CoV-2. 50

Studying the dynamics of the immune response to SARS-CoV-2 infection in differing severity 51 phenotypes can provide crucial insights into the pathogenesis of severe COVID-19 disease and 52

inform development of host-directed therapies and diagnostics. Time-course studies of anti-SARS-53

CoV-2 immune responses have thus far focused on proteomic and cellular approaches [13] [14] [15] [16] [17] [18] [19] . 54

Longitudinal transcriptomic studies, which can provide valuable insights into the temporal patterns 55

of immunological responses to infectious diseases [20] have been limited to date, and have focused 56 on biomarker discovery [21] and comparisons of acute disease with convalescence and recovery 57 [16, 22] . Longitudinal studies require the same patients to contribute sequential samples and can be 58 logistically challenging. Pseudotemporal approaches, where each patient may contribute just one 59

sample to the analysis, offer an alternative. 60

In this study we employed a novel pseudotemporal technique for exploring the trajectory of 61 transcriptomic responses in COVID-19. A pseudotemporal model was constructed from patients' 62 single timepoint samples, and the trajectory of gene expression over pseudotime was modelled for 63 mild, moderate and severe COVID-19 disease. Pseudotemporal differential analysis identified genes 64 that were significantly differentially expressed (SDE) over pseudotime and between the different 65 severity groups and pathway analysis was undertaken, aiming to elicit the temporal patterns of 66 biological mechanisms underlying the different severity phenotypes. Immunodeficiency Virus (HIV), as these conditions were thought likely to impact the transcriptome. 76

All patients had been scored on the World Health Organization 8-point ordinal scale for assessing 77 COVID-19 disease[23] Severity stratification was based on these scores, both for clinical status at 78 the time of research blood sampling ("Severity Sample") and for overall greatest severity during the 79 COVID-19 episode ("Worst Severity"). "Mild" was defined as those who scored 1-2, were 80 ambulatory and did not require hospitalisation for COVID-19; "Moderate" was those scored as WHO 81

3-4, requiring hospitalisation with or without standard oxygen therapy; "Severe" was defined as 82 those scoring 5-8, who received at least one of high flow oxygen (>16 liters/ minute), non-invasive 83 ventilation, mechanical ventilation, inotropes, haemofiltration, and/ or who died. Given the 84 complexity of the planned analysis, it was important to minimise potential confounders. Thus, 85 samples were excluded if "Severity Sample" differed from "Worst Severity", or if patients had 86 received immunomodulating treatment for COVID-19 prior to research blood sampling, including 87 steroids, tocilizumab and interferon. "Symptom duration" was calculated as the number of days 88 between symptom-onset and research blood sampling and each sample was then assigned a 89

Pseudotime, based on symptom duration. 90

All statistical analyses were performed using the statistical software R ( the batch effect. 100

Pseudotemporal differential expression analysis, to identify genes significantly differentially 101 expressed (SDE) over pseudotime and between severity groups, was performed using MaSigPro 102

[27], with "Mild" as the comparator group, two degrees of freedom to capture linear and quadratic 103 trends and binomial distribution specified. MaSigPro follows a two-step regression strategy to find 104 genes with significant temporal expression changes and significant differences between groups.

Coefficients obtained in the second regression model were then used to cluster together significant 106 genes with similar expression patterns. Calculation of adjusted p-values was performed using the 107

Benjamini-Hochberg (BH) procedure [28] . Genes with an adjusted p-value < 0.05 were considered 108 SDE. 109

To complement and validate the MaSigPro analysis, time-course differential expression analysis was 110 performed in DESeq2, using the likelihood ratio test. SDE genes (BH-adjusted p-value < 0.05) 111

identified were identified for the factors Pseudotime, Severity and Severity: Pseudotime. 112

Pathway analysis of SDE genes, in which genes are mapped to biological processes and pathways, 113

was undertaken in gProfiler (https://biit.cs.ut.ee/gprofiler/gost). Reactome reference database was 114 used and significantly enriched terms defined as those with adjusted p-value <0.05. 115 3

Results 116

Whole blood RNA sequencing data were available for 99 COVID-19 patients and 10 healthy controls 118 (HC) from the UK, and 58 COVID-19 and 10 inflammatory bowel disease (IBD) patients from Spain, 119

with each patient contributing one sample. Of these, 29 COVID-19 samples (22 UK, 7 Spain) were 120 excluded due to presence of coinfection, haematological disorder or pre-existing 121

immunosuppression. The remaining 128 COVID-19, 10 HC and 10 IBD samples were taken forwards 122 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22274729 doi: medRxiv preprint for batch correction and normalisation. The HC and IBD samples, included to aid batch correction 123

and normalisation, were treated as independent groups and were removed after these steps. A 124 further 27 COVID-19 samples were excluded from the analysis due to prior treatment with 125 immunomodulators (steroids, tocilizumab, interferon) for COVID-19 (n=11), Sample Severity 126 category differing from Worst Severity category (n=15) or both (n=1). A further two samples from 127 individuals who remained asymptomatic were excluded as they could not be assigned to a 128

Pseudotime, as well as one sample which was an outlier with respect to symptom duration (42 days, 129 16 days longer than the next longest symptom duration) (Figure 1 ). 130 inputted as the MaSigPro variable "Time". 141

After batch correction and normalisation, principle component analysis confirmed that batch effects 143 had been removed (Supplementary figure 1). The pseudotemporal differential expression analysis in 144 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. phosphorylation of CD3 and TCR zeta chains and programmed cell death protein 1 (PD-1) signalling. 176

Cluster 6 was enriched for transcription, mRNA splicing and metabolism of RNA. (Supplementary file  177 3). 178

As an alternative method for evaluating genes differentially expressed over pseudotime and 180 severity, a time series analysis was conducted in DESeq2. In total, 6,042 were SDE with Severity and We have employed a novel pseudotemporal approach to explore trends in whole blood gene 199 expression in patients with varying COVID-19 severity phenotypes. We have identified genes 200 significantly differentially expressed over pseudotime and between severity groups, clustered these 201 genes together based on pseudotemporal trends and undertaken pathways analysis of these gene 202

clusters. Our analysis revealed that increasing severity was associated with upregulation of multiple 203

immunological, coagulation and platelet pathways and downregulation of RNA metabolism, 204 transcription and cellular metabolic pathways. We observed differing pseudotemporal trends: for 205 example, there was divergent expression of genes enriched for coagulation and platelet pathways 206 between severe and moderate groups from the earliest pseudo-timepoints, whereas expression of 207 genes enriched for humoral immunological pathways were similar between these groups at early 208 pseudo-timepoints before diverging after 14 days. 209

We observed upregulation of neutrophil activation pathways, highest in severe >moderate > mild 210 clinical phenotypes (Clusters 1 and 3) neutrophils play an early role in severe disease and also have a continuing role in those with 217 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Longitudinal studies, which provide valuable information about evolving immunological responses 290

to infectious diseases, are constrained by the need to collect multiple sequential samples from 291 patients, which can be logistically challenging. Our study demonstrates that pseudotemporal 292 approaches can be an informative alternative when sequential samples are unavailable. 293

To the best of our knowledge, this is the first reported analysis of pseudotemporal transcriptomic 294 trends in SARS-CoV-2 infection with comparisons of severity phenotypes, however it has some 295

limitations. There were differences in age and sex between the severity groups, with age and male 296 sex increasing with severity, which is in keeping with the epidemiology of COVID-19 disease [61, 62]. 297

Therefore it is possible that some of the SDE genes we have identified are driven by age or sex, 298 rather than COVID-19 severity. However, given COVID-19 severity, age and sex are so closely 299

intertwined, adjusting for these two variables could mask key drivers of severity, and thus our 300 unadjusted analysis may be a more sensitive approach. This study combines data from UK and 301

Spanish cohorts. Both cohorts were recruited and sampled during the first wave in early 2020, but 302 there may have been differences between the two countries, for example in government advice for 303 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022.

staying at home and clinical management. The complexity of this analysis required us to minimise 304 potential interference of the transcriptome by variables such as COVID-19 treatments and 305

coinfections. Therefore a strict set of pre-determined exclusion criteria were employed that 306 resulted in just two thirds of the samples being included in the analysis. Thus the sample size in 307 some of the later severity-pseudotime groups was modest. We only included samples for which 308 "Sample Severity" and "Worst Severity" classification were the same. Therefore, the results of this 309

analysis cannot inform predictions of severity or prognosis. 310

We have modelled and compared the pseudotemporal trajectories of whole blood transcriptomic 312 responses in patients with mild, moderate and severe COVID-19 disease. We have observed that 313

increasing disease severity is associated with upregulation of multiple immune pathways, including 314

neutrophil, complement and immunoglobulin-mediated responses, as well as platelet activation 315

and coagulation pathways, with varying pseudotemporal patterns identified over the clinical course.

A potential role of senescence is also suggested. This work highlights the complex interplay 317 between immunological, thrombotic and cellular factors that underlie the clinical spectrum of 318 COVID-19 disease. It demonstrates the utility of pseudotemporal approaches in studying host-319 pathogen responses. Proyectos GaIN Rescata-Covid_IN845D 2020/23 (GAIN, Xunta de Galicia). 341

The funders had no role in the design of the study; in the collection, analyses, or interpretation of 342 data; in the writing of the manuscript, or in the decision to publish the results. 343 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) ed. Vienna, Austria2020. 420 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22274729 doi: medRxiv preprint

Disease severity-specific neutrophil signatures in blood transcriptomes stratify COVID-19 358 patients

An 360 inflammatory cytokine signature predicts COVID-19 severity and survival

State Shift between Mild and Moderate COVID-19

Clinical and immunological features of 365 severe and moderate coronavirus disease 2019

Complex Immune Dysregulation in COVID-19 Patients with Severe Respiratory Failure

Proteins associated 370 with neutrophil degranulation are upregulated in nasopharyngeal swabs from SARS-CoV-2 patients. 371 PloS one

Inborn errors of type I 373 IFN immunity in patients with life-threatening COVID-19

Autoantibodies 375 against type I IFNs in patients with life-threatening COVID-19

Functional exhaustion of antiviral 377 lymphocytes in COVID-19 patients

Clinical and immunological features of 379 severe and moderate coronavirus disease 2019

ComBat-seq: batch effect adjustment for RNA-seq count 421 data

Moderated estimation of fold change and dispersion for RNA-423 seq data with DESeq2

maSigPro: a method to identify significantly 425 differential expression profiles in time-course microarray experiments

Controlling the False Discovery Rate: A Practical and Powerful 428 Approach to Multiple Testing

Dysregulation of Immune Response in 431 Patients With Coronavirus 2019 (COVID-19) in Wuhan, China

434 Characterisation of the blood RNA host response underpinning severity in COVID-19 patients

The predicting roles of 437 carcinoembryonic antigen and its underlying mechanism in the progression of coronavirus disease 438 2019

Neutrophil extracellular 440 traps in COVID-19

Targeting potential drivers of COVID-19: Neutrophil extracellular traps. The Journal of experimental 443 medicine

Abnormal coagulation parameters are associated with poor 445 prognosis in patients with novel coronavirus pneumonia

Thromboembolism risk of 448 COVID-19 is high and associated with a higher risk of mortality: A systematic review and meta-449 analysis

Platelet 451 activation and platelet-monocyte aggregate formation trigger tissue factor expression in patients 452 with severe COVID-19

SARS-CoV-2 interacts with platelets and 454 megakaryocytes via ACE2-independent mechanism

Platelet gene 457 expression and function in patients with COVID-19

Platelets Can Associate with 459 SARS-Cov-2 RNA and Are Hyperactivated in COVID-19

A comprehensive, longitudinal analysis of 461 humoral responses specific to four recombinant antigens of SARS-CoV-2 in severe and non-severe 462 COVID-19 patients

Delayed production of 464 neutralizing antibodies correlates with fatal COVID-19

Single-cell landscape of 466 immunological responses in patients with COVID-19

Cell fusion induced by 468 ERVWE1 or measles virus causes cellular senescence

470 Induction of DNA double-strand breaks and cellular senescence by human respiratory syncytial 471 virus

SILAC Analysis Reveals 473 Increased Secretion of Hemostasis-Related Factors by Senescent Cells

The senescence-associated secretory phenotype: 476 the dark side of tumor suppression

Virus-induced 478 senescence is a driver and therapeutic target in COVID-19

Splicing, Translation, and Protein Trafficking to Suppress Host Defenses

Structural 483 basis for translational shutdown and immune evasion by the Nsp1 protein of SARS-CoV-2

ZAP-70 in Signaling, Biology, and Disease. Annual 486 review of immunology

Downregulation of 488 CD45 Signaling in COVID-19 Patients Is Reversed by C24D, a Novel CD45 Targeting Peptide. Frontiers 489 in medicine

The Role of PD-1 in Acute and 491 Chronic Infection

Monoamine oxidase is a 493 source of oxidative stress in obese patients with chronic inflammation (1)

Monoamine oxidase expression during 496 development and aging

Metformin alleviates 498 monoamine oxidase-related vascular oxidative stress and endothelial dysfunction in rats with diet-499 induced obesity. Molecular and cellular biochemistry

Monoamine oxidase 501 inhibition improves vascular function in mammary arteries from nondiabetic and diabetic patients 502 with coronary heart disease

Monoamine 505 oxidases are mediators of endothelial dysfunction in the mouse aorta

Oxidase-Related Vascular Oxidative Stress in Diseases Associated with Inflammatory Burden. 509 Oxidative medicine and cellular longevity

Metabolomics and computational analysis of the role of monoamine oxidase activity in delirium and 512 SARS-COV-2 infection

Relationship between COVID-19 infection and 514 neurodegeneration: Computational insight into interactions between the SARS-CoV-2 spike protein 515 and the monoamine oxidase enzymes

Risk factors of critical & mortal COVID-19 517 cases: A systematic literature review and meta-analysis

Risk factors for Covid-19 severity and fatality: a 520 structured literature review