Three highly pathogenic coronaviruses for human populations appeared in the past two decades. First, severe acute respiratory syndrome (SARS) was described in 2003 when the emergence of a human pathogen coronavirus (SARS-CoV) was noted (1–3). Second, in September 2012, Middle East Respiratory Syndrome Coronavirus (MERS-CoV) was reported to cause human infection (4). Third, in December, 2019, a novel coronavirus (SARS-CoV-2) causing coronavirus disease (COVID-19) was initially identified in Wuhan, China, and caused later a pandemic starting in early March, 2020 (5). The rapid, global spread of the virus is presently ongoing, by late January, 2021, over 4.0 million cases per week are being reported worldwide, with around 70.000 deaths per week (6). Organ dysfunction, particularly complex coagulopathy and respiratory failure often in the context of acute respiratory distress syndrome (ARDS), and multiple organ failure are associated with highest mortality, especially in the elderly (7).

The complement system is generally considered as part of the first line defense against pathogens, including viral infections, and this fact is marked by the multiple mechanisms by which viruses avoid complement attack (8). On the other hand, complement (with or without the contribution of antibodies) may also play an enhancing role in viral infections, as described for HIV (9–11). Furthermore, complement activation, especially generation of anaphylatoxin C5a, is involved in the development of acute lung injury caused by influenza A (12). Animal experiments also supported the involvement of complement activation in the development of SARS caused by CoV infection (13). Accordingly, complement inhibitors are considered as promising therapeutic options during the ongoing SARS-CoV-2 pandemic (14, 15), and the first COVID-19 cases treated by the C5 inhibitor monoclonal antibody eculizumab (16–18), and by the C3-targeting small peptide inhibitor AMY-101 (19) have recently been reported. Further in line with these considerations, Annane et al. reported promising survival results in a proof-of-concept, non-randomized study of 80 eculizumab treated intensive care unit patients with COVID-19 (20).

Complement system is a sensitive enzymatic cascade responsive through its lectin pathway to carbohydrate patterns of pathogens, or through its classical pathway to antibodies (21, 22). The alternative pathway (AP) represents an amplification loop on the level of the central component C3, with generation of powerful biological mediators including anaphylatoxins, opsonins, and lytic complexes (23). Patients with various infections may benefit from laboratory evaluation of the complex system, which can show the presence of complement deficiency, activation state, or consumption of the whole system, or its dedicated pathways (24, 25). Two recent studies have shown a clear association between respiratory failure, COVID-19 outcome, and systemic complement activation (26, 27) in COVID-19 patients. Furthermore, a large-scale multi-omic analysis of COVID-19 severity mapped several molecular features including the complement system, and identified strong associations between complement activation and severe COVID-19 phenotype (28). In a recent comprehensive review Perico and colleagues described multiple mechanisms of how endothelial cells, and complement system particularly, may contribute to severe thromboinflammation in COVID-19 (29). Finally, Eriksson et al. have identified complement activation through the MBL pathway as a novel amplification mechanism that contributes to pathological thrombosis in critically ill COVID-19 patients (30).

Based on the multiple potential interactions between viral pathogens and complement, and principally on the above described associations we hypothesized that complement activation may be an important factor in the pathogenesis of SARS-CoV-2 infection. Accordingly, the aim of our work was to describe activities of the alternative and classical complement pathways, levels of their components, regulators and activation products in COVID-19 patients, and to search for associations between severity and outcome of COVID-19 and complement activation, dysregulation, and consumption.

One hundred and two SARS-CoV-2 positive hospitalized patients were enrolled into this study ( Figure 1 ). Twenty-six patients who were all outpatients when infected with SARS-CoV-2, and were sampled in convalescent phase, formed the patient control group [sampled at mean 53 (SD 13), min: 26, max: 74 days after onset of symptoms].

In-hospital patients (n = 102) were divided into three severity subgroups (without or with nasal oxygen support, or necessitating intensive care unit treatment) when sampled for this study. Sampling was done when the patient was first hospitalized in Central Hospital of Southern Pest National Institute of Haematology and Infectious Diseases or in Semmelweis University, or when the patient was transferred from a different hospital to our centers. Therefore, the elapsed time between onset of symptoms and sampling varied between 1 day and 63 days, as shown on Figure 2 . Typically, sampling was done on day 8 (median) after onset of symptoms [this is day 6 (median) after first PCR positivity]. Sixty-seven percent of patients were sampled on the first 2 weeks after onset, and the remaining one-third on weeks 3–9 ( Figure 2 ). Extra-corporeal membrane oxygenator treatment, non-invasive ventilation, or high-flow oxygen therapy was not used for patients in this study, therefore, no patients fall into WHO category 5.

Severity of COVID-19 changed over time during the hospital stay (as presented on Figure 1 ), therefore, severity was assessed at a second time when “peak” was reached. Thirty-three in-hospital patients required nasal oxygen support during the hospital stay. During the observation period 25 patients died, 17 survivor patients were discharged from ICU. Baseline characteristics revealed significant differences across various severity categories in comorbidities, complications, and inflammatory markers of COVID-19 ( Table 1 ). Severe COVID-19 disease was associated with frequent occurrence of diabetes mellitus and malignant diseases (p < 0.0001). Patients, who died later, had four comorbidities (median), whereas survivors had only two ( Table 1 ). Increased numbers of in-hospital complications, such as pneumonia, respiratory failure, sepsis, thromboembolic events, and acute kidney injury, were more common among those who required ICU treatment or died ( Table 1 , p < 0.0001). Lymphopenia, increased neutrophil count, and gradually increasing IL-6, C-reactive protein (CRP), troponin, and ferritin levels were all present in patients with more severe forms of COVID-19 ( Table 1 ).

Our study provides the first observational proof that biomarkers of complement overactivation and consumption are useful tools for the prediction of in-hospital mortality of COVID-19 patients. We observed a continuous increase of complement activation markers C3a and sC5b-9 across WHO categories of COVID-19 severity ( Figure 4 and Table 2 ), with highest levels in patients presenting with critical illness. C3 concentrations were remarkably decreased in the group of non-survivors ( Figure 4 ), with significantly elevated C3a/C3 ratios, reflecting overactivation and consumption ( Figure 5 ). This difference translated to significant prediction of mortality in Cox proportional-hazards survival models, with hazard ratios of 6.1 (2.1–17.8, 95% CI) for those having increased ratio of C3a/C3. The model remained significant after adjustment for important covariates of critical illness, including comorbidities, in-hospital complications, or CRP.

The demonstration of complement C3 overactivation and consumption, and its strong association with COVID-19 severity and outcome, is intriguing and confirms recent observations. First, the RCI-COVID-19 group demonstrated activation of complement with elevated C3a, sC5b-9, and C3c levels in severe COVID-19 (27). High levels of complement activation markers were associated with mortality and thromboembolic complications, however, prediction of mortality was not formally evaluated in survival models. Recently, the activation through the MBL pathway and pathological thrombosis in critically ill COVID-19 patients was described (30), highlighting the critical role of complement activation in important complications of SARS-CoV2 infection. In addition, Fang et al. (19) reported low serum C3 levels in association with COVID-19 related mortality. Specifically, their study identified approximately seven times higher risk of death, per 1 g/L decrease in serum C3 level. In the same study complement activation has not been analyzed, hence whether decreased C3 was due to overactivation and consumption, or protein loss/decreased production, remained elusive. Of note, complement activation was shown to predict COVID-19 progression in chronic hemodialysis patients (36), and C5a was considered as an earlier marker, than C3a. Furthermore, we also observed decreased C4 levels in the group of non-survivors, indicating a potential involvement of the lectin or classical pathways behind complement activation and consumption. It is of note, that levels of C1q antigen [a protein produced mainly by monocytes and macrophages (37)] were the highest in deceased COVID-19 patients, suggesting the potential role macrophage overactivation in the fatal outcome of the disease (38). These observations require detailed molecular analysis in future studies.

Importantly, our results provide observational support to the milestone paper of Yu et al. published in the journal Blood. The group at Johns Hopkins University reported that SARS-CoV-2 spike proteins bind heparan sulfate on cell surface and overactivate mainly the alternative pathway by interfering with its main soluble regulator, Factor H. Moreover, ACH145951, a small molecule Factor D inhibitor, blocked SARS-CoV-2 spike protein induced AP activation (39). Our clinical observations on the strong prediction of in-hospital mortality by the complement C3 overactivation and consumption provide independent observational support to the conclusion of the authors, who suggested that SARS-CoV-2 spike protein induced AP activation may have profound implications in the multiorgan dysfunction, coagulopathy, and endothelial injury, all characteristic of COVID-19.

Finally, in their elegant retrospective study Ramlall et al. (40) reported that history of age-related macular degeneration (AMD)—a condition known to be linked to complement dysregulation—is a risk factor for COVID-19 morbidity and mortality. In addition, with a candidate-driven genetic association study authors identified multiple putative risk loci in genes of complement regulators and components that were associated with clinical outcome of SARS-CoV-2 infection (40). Authors concluded that history of AMD predisposes patients to poor clinical outcome following SARS-CoV-2 infection, and variants in critical regulators of complement are associated with this inferior outcome. By showing the strong association between complement dysregulation (overactivation and consumption, as expressed by the C3a/C3 ratio, Figure 5 ) and severity and mortality of COVID-19, our current results corroborate the findings of Ramlall et al. Lastly, the C3 S/F polymorphism has been reported as a potential confounder of COVID-19 related mortality (41). C3 S/F polymorphism is functionally active (42), the F allele, being reported as a confounder of increased mortality, has also been associated with increased tendency to activate the alternative pathway. Therefore, inherited factors determining the activity and regulation of complement activation might indeed play important roles in the pathogenesis of COVID-19.

Inflammatory markers were closely related to COVID-19 severity ( Table 1 ), which is an observation already reported at the beginning of the pandemic (43). Among the complement factors, Factor B, the enzymatic component of alternative pathway C3-convertase, was most closely related to inflammatory markers ( Table 3 ) and showed the highest elevation across severity categories. Complement activation markers C3a and sC5b-9 showed significant, constant increase across severity groups of COVID-19 patients ( Table 2 , Figure 4 ). It is tempting to speculate that the particular tendency for complement overactivation and the “cytokine storm” after SARS-CoV-2 infection is based on the strong induction of hepatic acute phase reaction (including for example Factor B). Due to the high level and availability of alternative pathway components C3 and Factor B, complement activation with overproduction of inflammatory and cell damaging activation products is constant and increasing across all severity categories of COVID-19, and it is most probably the exhausting regulatory capacity that makes individuals prone to complement overactivation and consumption, and finally to death. Results of genetic association studies seem to support this assumption (40).

In COVID-19 viral pneumonia, acute respiratory distress syndrome and respiratory failure are frequent, life-threatening complications often linked to thromboinflammation and vascular damage. Analyzing pulmonary biopsy and autopsy samples of SARS-CoV-2 infected patients with immunochemistry, Magro and coworkers were the first to describe septal capillary injury accompanied by extensive deposits of MASP-2, C4d, and terminal complement complex (44). Authors concluded that COVID-19 associated pauci-immune, complement mediated lung injury is distinct from the typical ARDS. Supporting the contribution of systemic complement activation to the development of COVID-19 associated respiratory failure are the recent observations of Holter et al. (26). Authors observed that increased admission C4d and sC5b-9 levels were associated with the development of respiratory failure and higher need for oxygen therapy, and that complement activation correlated to anti-viral antibody and ferritin levels, but not to viral load. Our observations are in line with these results, since there were significant increases in C3a and sC5b-9 levels in correlation with COVID-19 severity, with highest levels in critical patients often also having respiratory failure ( Figure 4 and Table 2 ). Furthermore, complement factor and activation product levels were associated with CRP and ferritin concentrations ( Table 3 ), reflecting the relationship between the extent of inflammation and complement dysregulation.

An important novel observation of the current study is the association of mortality in COVID-19 patients with markers of complement overactivation and consumption, independent of comorbidities and in-hospital complications. A prospective cohort study from New York City identified higher age, decreased oxygen saturation at presentation, comorbidities (chronic heart and renal failure and malignant disease), and increased C-reactive protein, D-dimer, procalcitonin, and troponin levels as strong predictors of in-hospital mortality (45). Since in-hospital complications, including development of respiratory failure, sepsis, thromboembolic complications, and renal failure, are strong determinants of progression to critical illness and intensive care unit treatment, we adjusted the Cox survival models for the total number of in-hospital complications. Prediction of mortality by markers of complement overactivation and consumption remained still significant after adjustment for in-hospital complications ( Table 4 ). Adjustment (in separate models) for age and C-reactive protein or delay between disease onset and sampling, did not change the results, either ( Table 4 ).

Current evidences support that SARS-CoV-2 infection leads to immune dysfunction, widespread endothelial injury, complement associated coagulopathy, and systemic microangiopathy (29). Our data reinforce the importance of clinical trials with complement inhibitors to limit complement mediated tissue damage and to decrease mortality in COVID-19. Multiple available inhibitors of complement cascade, including anti-C5 monoclonals eculizumab (17, 18, 46) and ravulizumab (19, 47), narsoplimab, a monoclonal antibody against MASP-2 (48), the compstatin-based complement C3 inhibitor AMY-101 (49), C1-inhibitor (50), and the anaphylatoxin C5a blocking antibody IFX-1 (vilobelimab) (51) are currently being evaluated for COVID-19. Preliminary observations on safety and efficacy results are promising (20), but it needs more time to make firm conclusions on the utility of such drugs for COVID-19.

This study was limited by its relatively small size enrolling 102 in-hospital patients (25 non-survivors) with confirmed COVID-19 in two tertiary care centers. However, the severity groups were nearly equally sized and follow-up was complete for all of the patients, allowing for reliable statistical analysis. Post-hoc power analysis yield P > 0.8 for both C3a and sC5b-9, when comparing the 25 non-survivors and the 27 in-hospital patients without need for oxygen support ( Figure 4 ). The low number of non-survivors in the study allowed adjustment for only one confounder, but since our study collected all of the relevant clinical data about comorbidities and in-hospital complications, we were able to adjust for the most important confounders ( Table 4 ). One of the strengths of our study was the detailed characterization of large panel complement profile by methods regularly evaluated in external proficiency testing (25). In addition, our strategy to measure complement activity and factor levels together with activation markers allowed us to identify the clinical relevance of complement overactivation together with signs of consumption, a phenomenon indicating the presence of complement dysregulation. Results on the relationship between activity of lectin pathway and disease severity are heavily influenced by genetic factors strongly regulating mannose-binding lectin levels, but due to space limitations, the results of detailed analysis will be presented separately (manuscript in preparation).

The raw data supporting the conclusions of this article will be made available by the corresponding author, without undue reservation.

The studies involving human participants were reviewed and approved by Hungarian Ethical Review Agency (ETT-TUKEB). Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

BM, ZF, DC, LH, EK, LC, and PK designed and performed laboratory determinations, interpreted data, and drafted the manuscript. GS, ZZP, and ZP conceptualized research, collected and analyzed clinical information and laboratory data, conducted statistical analysis, interpreted data, and wrote the manuscript. MR, VM, ZI, JG, LG, PR, BS, BL, JS, IB, TM, and IV-N conceptualization, collected and analyzed clinical information, interpreted and supervised data, and drafted the manuscript. All authors contributed to the article and approved the submitted version.

The research was financed by the Higher Education Institutional Excellence Programme of the Ministry of Human Capacities in Hungary, within the framework of the molecular biology thematic program of the Semmelweis University, by the National Office for Innovation and Research [KH130355, and “Befektetés a jövőbe” (2020-1-1-6-JÖVŐ-2021-00013) to ZP]. The study was performed in frame of the Premium Postdoctoral Fellowship Program of the Hungarian Academy of Sciences (PPD2018-016/2018 to DC). ZP and LH are supported by funds of the EU MSCA project CORVOS 860044.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.