Demographics of the current cohort of SARS-CoV-2 infected patients had similar characteristics as did the nationwide cohort study of the first consecutive 5,000 patients with COVID-19 in Qatar (4). While our study selected patients between 18 and 65 years of age, both cohorts consisted largely of males with a younger median age due to the relatively younger population in Qatar. Risk factors of ICU admission in our study and the national cohort study (4) included older age, male sex, higher BMI, and preexisting diabetes and hypertension. Other comorbidities such as chronic artery disease, liver disease or kidney disease were identified in both studies but did not reach significance in our smaller cohort. Of the first 5,000 consecutive cases in Qatar, 1424 patients (28.5%) required hospitalization, out of which 108 (7.6%) were admitted to ICU, and only 14 patients (0.28%) had died by 60 days after infection. In a relatively younger national cohort in Qatar, with a low comorbidity burden, COVID-19 was associated with low mortality (4), which was also reflected in our smaller cohort.
The Qatari population, with respect to COVID-19, is unique in the demographic characteristics (e.g. predominantly males with younger age) when compared to other populations such as that described in the ISARIC (1). However, our study of plasma profiling of SARS-CoV-2 infected patients and control subjects using the Olink Proteomics panels may be generalizable as the biological processes and pathways enriched in patients with severe complications in our subpopulation have also been reported in other studies. Additionally, the COVID-19 molecular severity score reported here was cross-validated in an independent, ethnically different, larger cohort from the Massachusetts General Hospital (MGH, USA). Interestingly, the analysis in the MGH cohort which included symptomatic SARS-CoV-2 negative patients showed that the molecular severity score could reflect a general biological response in symptomatic respiratory infection or disease irrespective of SARS-CoV-2 status although the score was higher and more predictive in SARS-CoV-2 positive patients.
Medical history of macular degeneration and of coagulation disorders (thrombocytopenia, thrombosis, and hemorrhage) were considered risk factors for higher morbidity and mortality in a recent study on 11,116 patients infected patients with SARS-CoV-2 (5). Moreover, RNA-Seq profiles from nasopharyngeal swabs in this study found several enriched immune-modulatory functions in SARS-CoV-2 infected patients versus controls, such as inflammatory response, interferon alpha response, and IL6-JAK-STAT3 signaling, which were also identified in our study. Activation of the complement and coagulation cascades was also among the most enriched gene sets (5), an observation corroborated by our plasma protein profiling results.
A multiplexed biomarker profiling of plasma from 49 SARS-CoV-2 infected patients (40 in ICU and 9 in non-ICU units) and 13 non-COVID-19, non-hospitalized controls identified multiple proteins in association with ICU admission and mortality, including HGF, RETN, LCN2, G-CSF, IL-6, IL-8, IL-6, IL-10, IL1RA and TNF-α (6), which were confirmed in our study. Importantly, the study also reported a unique neutrophil activation signature composed of neutrophil activators (G-CSF, IL-8) and effectors (RETN, LCN2 and HGF), with a strong predictive value to identify critically ill patients whereby the effector proteins strongly correlated with absolute neutrophil count (6). Our study not only identified those components of the neutrophil activation signature, but also found that the COVID-19 molecular severity score, a more comprehensive signature, also correlated with absolute neutrophil counts. Moreover, the neutrophil count was selected in the COVID-19 clinical risk score developed in our study. Another study deployed Olink Proteomics panels to measure 1,161 plasma proteins from 20 patients, 10 SARS-CoV-2 positive and 10 SARS-CoV-2 negative patients, admitted to ICU and 10 healthy controls (7). This study had a small sample size, therefore could not determine changes contributing to ICU admission and only reported mortality. Interestingly, it uncovered similar proteins and pathways as those identified in our study in association with COVID-19 severity, such as interleukins, CXCLs/chemokines, membrane receptors linked to lymphocyte-associated microparticles, cytoplasmic/cytoskeletal proteins, and nuclear proteins or transcription factors (7). Among their reported 20 top proteins differentiating patients with COVID-19 disease from healthy controls, 13 (65%) were also confirmed in our study, with two of them being components of our COVID-19 molecular severity score, namely IL6 and IL18R1. Of their reported 20 top proteins which differentiated ICU-admitted patients with COVID-19 versus non-COVID-19 disease, 12 (60%) were also found in our study with two being components of the molecular severity score, KRT19 and CCL7.
Besides Olink technology, mass spectroscopy was used in two studies to identify deregulated proteins in SARS-CoV-2 infected patients. The first study used liquid chromatography-mass spectrometry (LC-MS) to profile 31 patients with SARS-CoV-2 infection, where the disease severity was graded according to the WHO outcome scale. The study identified 27 potential biomarkers that were differentially expressed (8). Although none of these biomarkers were identified in our study, the biological functions reported in their study were also captured in our analysis, including complement factors and the coagulation system, inflammation modulators, and pro-inflammatory factors upstream and downstream of interleukin 6. The second study used mass spectroscopy for proteomic and metabolomic analysis of sera from 46 COVID-19 and 53 control individuals and identified 93 proteins which were differentially expressed in sera from severe COVID-19 patients (9). Of these 93 differentially expressed serum proteins, 17 were included in our panel profiling of plasma with 11 (65%) of these proteins also being significantly deregulated in our cohort. More specifically, 8 of the 11 proteins were upregulated (VWF, PVR, GRN, NID1, VCAM1, SAA4, CD59, CDH1), whereas the remaining three proteins were downregulated (FETUB, APOM, and IGFBP3) in the severe cases in our study. Using mass spectroscopy, the authors developed a model of 22 proteins and seven metabolites (29 sera factors) to stratify patients according to severity (9). None of the protein biomarkers reported in this study were identified in our study; however, there is a strong overlap in biological functions, including the release of IL-6 and TNF-α, inflammatory responses, activation of the complement system and protein phosphorylation.
Interestingly, both MS-based studies agreed on ten protein biomarkers in their classifiers, 10/27 (37%) for the first study which included serum and plasma (8), and 10/22 (45%) in the second study which only used serum (9). Although our study agrees on the biological functions identified in the two MS-based studies, the lack of agreement with the named biomarkers may be due to the use of plasma in our study compared to serum in the MS studies. We cannot exclude that the MS-based studies are more comprehensive and less biased than the panel profiling used in our study. However, it should be noted that there was a small overlap between all the proteins detected by mass spectroscopy in sera (prior to statistical analysis) from the Shen et al. study (9) and the proteins profiled in plasma in our study; 134 common proteins out of the 791 (17%) proteins detected by mass spectroscopy and the 894 (15%) proteins profiled in our study.
Altogether, our study identified several biological pathways described in previous proteomic studies of sera or plasma of patients with severe COVID-19 complications. In addition to their potential biomarker value, the protein profiles can also be used to predict potential drugs for intervention. Our drug-protein interaction analyses shortlisted HGF, MPO and CXCL10 as targets that could influence most of the interactions between the plasma proteins upregulated in severe COVID-19 cases. Notable examples of possible drugs include flutamide which can target MPO, ACE2 and IL2RA and has been proposed as a possible drug for COVID-19 treatment based on ACE2 interaction network analysis (10). Methylene Blue can modulate MPO, VWF and CPA1, which were upregulated in our severe cases and had been tested in combination with other drugs in a clinical trial with critically ill COVID-19 patients in Iran (11), whilst a wider clinical trial (NCT04370288) has been designed. Thalidomide is another example that targets one of the shortlisted proteins (HGF) in addition three other upregulated proteins in severe cases (IL6R, VWF, and F2R). Its use for COVID-19 was reported for a single case in China (12) and led to recovery, and two clinical trials (NCT04273581 and NCT04273529) have been registered. However, thalidomide’s side effects and its previous dark past has been raised as serious concerns for its use to treat COVID-19 patients, and may have to be strictly limited to use in men and post-menopausal women (13).
Furthermore, methotrexate inhibits HGF and two other upregulated proteins in severe cases in our study, S100A12 and SULT2A1. This drug has been reported to inhibit SARS-CoV-2 virus replication in vitro via purine biosynthesis, thereby potently inhibiting viral RNA replication, viral protein synthesis, and virus release. As such, methotrexate was proposed as an effective measure to prevent possible COVID-19 complications (14). The use of methotrexate to treat COVID-19 patients or prevent complications has not been tested; however, a large comparative cohort study suggested that patients with recent TNF inhibitors and/or methotrexate exposure do not have increased COVID-19 related hospitalization or mortality (15).
In addition to the potential targeting of HGF, MPO and CXCL10 as highly interconnected proteins, our analysis identified ribavirin as a treatment option based on the upregulation of VWF and CST3 (Cystatin C) in patients with severe COVID-19 complications. Ribavirin, an oral nucleoside analogue, has been tested in combination with injectable interferon beta-1b and the oral protease inhibitor (lopinavir-ritonavir) in a randomized phase 2 trial to treat COVID-19 patients. Compared to lopinavir-ritonavir alone, the triple combination was safe and effectively shortened the duration of virus shedding, decreased cytokine responses, alleviated symptoms, and facilitated the discharge of patients with mild to moderate COVID-19 disease (16). A follow-up trial has been registered (NCT04494399) to test the combination of ribavirin with interferon beta-1b, without lopinavir-ritonavir, to treat patients with COVID-19.
In conclusion, our study identified deregulated proteins in the plasma of patients with severe COVID-19 complications that may inform therapeutic interventions. The 46-protein signature identified in our study was developed as the COVID-19 molecular severity score and used to stratify patients according to COVID-19 severity in an independent cohort. The COVID-19 molecular severity score could predict outcomes up to 28 days post-admission and from as early as three days of admission. We used the molecular severity score to select clinical parameters available at the time of admission and generated a scoring system to develop the molecularly trained clinical risk score. The molecular severity and the clinical risk scores developed here have the potential to stratify SARS-CoV-2 infected patients at early stages according to their risk of developing complications to prospectively inform healthcare management and clinical decision-making to prevent complications and mortality.