Our study has identified a set of genes, the differential expression of which is associated with survival outcomes in severe COVID-19, particularly through their involvement in negative regulation of immune response, complement activation, and the defense response to the virus. To achieve this, we have conducted transcriptomic analysis of PBMCs derived from patients with severe SARS-CoV-2 Delta strain induced pneumonia and identified a number of genes differentially expressed between survivors and nonsurvivors in the period of 30 days after hospitalization in the ICU. To elucidate the role of identified DEGs in determination of survival outcome, we have conducted Gene Ontology Biological Processes (GO: BP) term enrichment. Most of the terms significantly enriched by DEGs were related to various immune and antiviral processes. A term–gene network has been constructed, based on the relation of terms and genes from GO. Significantly enriched terms can be split into five groups based on common genes—negative regulation of immune response (group 1), negative regulation of viral processes (group 2), humoral and complement-system activation (group 3), defense response to bacterium (group 4), and muscle contraction (group 5). Next, we will analyze potential role of significantly enriched gene ontology biological processes in COVID-19 and identify some key genes based on their position in the term–gene network and the current understanding of their role in the pathophysiology of COVID-19 and other diseases.

However, before this, it is worth mentioning that the small sample size impacts the generalizability of the study’s findings, especially in what concerns differential expression of particular genes. Since Gene Ontology enrichments impose additional levels of statistical testing on the data, we believe that identification of biological processes as a whole in the outcome of COVID-19, and less so individual DEGs, is the main takeaway of the study. Differential expression of identified key genes needs to be further validated on larger independent cohorts in order to confirm their role as important determinant in COVID-19 outcome and potential therapeutic targets.

The dual role of immune regulation in COVID-19

It has been previously established that the role of immune response in COVID-19 is twofold—on one hand, strong and timely activation of the immune system is necessary to prevent viral replication and stop the infection, but, on the other hand, the immune system hyperactivation that develops in the severe cases of COVID-19 contributes to its severity and lethality²⁸. Data obtained in our own study further contribute to this twofold picture. The significant enrichment of terms related to negative regulation of immune response and the fact that most of these genes are upregulated in survivors imply that the positive outcome of severe COVID-19 is partially based on downregulation of the immune system. However, significant enrichment of terms related to negative regulation of viral processes shows that strong antiviral immune response is still necessary for survival in severe COVID-19 cases. SARS-CoV-2’s ability to evade the immune system and its ability to cause reduced interferon response are well documented²⁹. The differential expression of genes related to these terms in the survivor group underlines the crucial role of effective immune response in counteracting COVID-19. The importance of interferon response is further stressed by group 2 terms being specifically related to three interferon induced DEGs—IFI6, IFITM3, and IFI44L—all of which are overexpressed in survivor group. It is further worth noting that only two DEGs are related to both groups 1 (negative regulation of immune response) and 2 (negative regulation of viral processes), with most being exclusively related to either group, suggesting that different sets of genes are involved in negative and positive components of immune system regulation. Taken together, these data points illustrate that the balance between immune activation and regulation appears critical for survival in severe COVID-19. While robust antiviral defense mechanisms are necessary, excessive immune responses, such as the dysregulated cytokine production observed in patients with severe and critically severe cases of COVID-19³⁰, can exacerbate the disease and cause additional harm.

Putting the terms “humoral response” (as a part of adaptive immunity) and “complement activation” (as a part of natural immunity) into a single group is most likely based on the involvement of the common genes, associated with both groups 1 and 3—SERPING1, VSIG4, C1QB, and C1QA. Before discussing these genes, however, it is worth examining the genes that are related to humoral response but not to complement activation. Both of them, DEFA3 and H2BC4, are also simultaneously related to the term “defense response to bacterium,” and these genes are most likely the reason behind enrichment of the “humoral response” term. Enrichment of the term “defense response to bacterium” appears at first glance to be a result of an error. However, upon further examination, this enrichment could have important implications for its role in the outcome of COVID-19. Genes LYZ³¹, DEFA3³², and H2BC4³³ encode proteins with well-documented antimicrobial properties. Among them, DEFA3 and H2BC4 are related to both terms humoral immune response and defense response to bacterium, while LYZ is related exclusively to defense response to bacterium. All three genes of these are upregulated in the group of nonsurvivors compared to survivors. Most of the genes used in GO enrichment are the opposite—upregulated in survivors compared to nonsurvivors (60 out of 77)—which makes it highly unlikely that this happened by chance. This may imply that downregulation/absence of upregulation of the antimicrobial and humoral response may play an important role in a positive outcome of COVID-19.

The role of ISG15 as a potential key gene in determining COVID-19 outcomes

In the gene–term network (Fig. 3), gene ISG15 occupies a key position, as it is the only gene that connects three different groups (1, 2, and 4) and is related to five different significantly enriched GO terms. ISG15 encodes interferon-inducible ubiquitin-like protein³⁴, which binds covalently to various substrates originating from both host cell and virus in a process called “ISGylation.” ISGylation plays an important role in inhibition of viral replication. The precise role and fate of ISGylated proteins are unknown; however, several possible mechanisms have been found. ISGylation can affect host protein localization, as in the case of filamin B³⁵, and it also could disrupt aggregation of viral protein complexes, as was observed in interaction of ISG15 with influenza B virus and human papilloma virus³⁶. ISG15 can also disrupt the egress of virus from host cells, as has been shown using the example of interaction of ISG15 with human immunodeficiency virus 1 (HIV-1), in which release of HIV-1 virions, but not production of HIV-1 proteins, were impeded by overexpression of ISG15³⁷. Besides interacting with various proteins through ISGylation, unconjugated ISG15 can act as a cytokine. Extracellular ISG15 acts as a ligand for integrin ITGAL³⁸, interacting with it and stimulating release of IFNγ and IL10 from NK cells through that interaction, with expression of IL10 being especially notable as an important factor in COVID-19 prognosis³⁰.

Previously, activation of ISG15 expression was identified in expressional profiling of nasopharyngeal swabs of patients infected with COVID-19^39,40,41. On the other hand, in vitro SARS-CoV-2 infection did not cause an increase in ISG15 mRNA production, which could be explained by specific properties of the used cell culture⁴². The important role of ISG15 can be also emphasized by the concentration and activity of ISG15 in a cell regulated by SARS-CoV-2. Papain-like proteinase PLpro plays an important role in this regulation. In SARS-CoV-2, PLpro is both necessary for normal processing of viral proteins and in regulation of posttranslational modification of host proteins, primarily ISG15, whereas in other coronaviruses it is primarily employed to prevent ubiquitination of cellular and viral proteins^43,44,45. The main targets of PLpro are complexes of ISG15 with IRF3 (interferon responsive factor 3, regulator of type 1 interferon response) and MDA5 (a protein from the family of RIG-I-like receptors, stimulating the synthesis of interferon and proinflammatory cytokines when binding to viral RNA)^46,47. This adaptation, allowing SARS-CoV-2 to target specifically ISG15 and ISGylated proteins, demonstrates the vital importance of ISG15 in antiviral immunity and suggests that increased expression and, consequently, activity of ISG15 play a crucial role in determining the outcomes of SARS-CoV-2 infection. Increased levels of ISG15 mRNA in the survivor group may lead to increased presence of ISG15 protein and, in turn, allow some of it to avoid inactivation by SARS-CoV-2. Based on our data and previous research, we consider overexpression of ISG15 to be a key event in determining a positive outcome of severe COVID-19 and believe it to be a valuable target for therapeutic intervention.

The complement system in COVID-19

Based on the raw number of related terms in the enrichment, eight key genes can be identified—C1QB, C1QA, ISG15, SERPING1, VSIG4, KLRD1, TRPM4, and HFE. Four of these genes are related to both groups 2 and 3 and to the terms “complement activation” and “negative regulation of immune response,” as well as the aforementioned “humoral response”—C1QA, C1QB, SERPING1, and VSIG4. Going further, we would like to focus our attention on these four genes and on their role and the wider role of the complement system in COVID-19. Also, it is worth noting that all four of these genes are upregulated in survivors compared to nonsurvivors.

The complement system plays a role in antiviral infections by opsonizing viruses and virus-infected cells, directly neutralizing viruses outside of cells, lysing virus-infected cells, and inducing virus-specific immune and inflammatory responses⁴⁸. The complement system can be activated through three possible pathways—classical, nonclassical, and lectin. Generally, it is considered that all the pathways of complement system activation lead to the same protein cascade and activation of the same effector mechanisms. However, based on our own and previously obtained data, it appears that the pathway of complement system activation could be important for determining outcomes of COVID-19. As previously mentioned, significantly increased expression of genes C1QA and C1QB, encoding both subunits of the first complement system component (related specifically to the classical pathway of complement activation), was observed in our comparison. Currently, it is not completely clear how interaction with SARS-CoV-2 activates the complement system. It has been shown previously that all of the major SARS-CoV-2 proteins (S, N, M, E) interact with the first component of the complement system, engaging the classic pathway of complement system activation⁴⁹. However, additional data exist that contradict these finding and implies that SARS-CoV-2 proteins do not directly bind to C1Q, instead activating the complement through the lectin pathway⁵⁰. Complement system proteins deposits are found in various tissues of COVID-19 patients, e.g., alveolar, kidney, and liver capillaries. However, most of those deposits consist of complement system proteins related to a nonclassical complement-system activation pathway⁵¹. A retrospective examination of concentration of several complement system components (C1Q, C3, C4, and C5) in a cohort of 74 patients with COVID-19 has demonstrated that a high level of expression of nonclassical and lectin complement system activation pathway proteins are linked to negative outcomes of COVID-19, while no analogous connection was found in the case of classical pathway⁵². It has also been shown that in a cohort of 71 patients, levels of C1q in serum were decreased in severe cases of COVID-19 when compared with patients with mild COVID-19⁵³. While not all changes in expression that can be observed in comparisons of mild and severe cases must necessarily also be present in comparisons of survival and nonsurvival outcomes, the assumption can be made that transcriptomic profiles of patients with mild cases of COVID-19 will be closer to those of surviving patients and, correspondingly, the transcriptomic profiles of patients with severe cases will be closer those of patients with terminal outcomes. Under this assumption, we can observe that our results are in accordance with previously obtained data, since we have observed an increase in the expression of both subunits of first complement component in the survivor group. Taken together with previously published data, increased expression of C1QA and C1QB in the survivor group implies that activation of the complement system via the classical pathway may contribute to a positive outcome of the disease. However, whether this relationship is casual in nature or is an association should be interpreted carefully, since it has also been shown that the concentration of С1q, unlike that of IL6 and IL10, is not an independent predictor of clinical outcomes, being instead dependent on the concentrations of other components of the “cytokine storm”³⁰.

The twofold role of the complement system in determining COVID-19 outcomes

Previously, the twofold role that immune system regulation plays in COVID-19 pathogenesis has been mentioned. The role of the complement system in SARS-CoV-2 pathogenesis also fits a twofold paradigm. On one hand, hyperactivation of the complement system has long been implicated in development of respiratory failure in both COVID-19⁵⁴ and other diseases⁵⁵, and targeting the complement system as an intervention appears a possible strategy for improving outcomes in severe COVID-19^56,57. On the other hand, the complement system plays an indispensable role in combating infectious diseases⁵⁸, and inhibiting it as an intervention opens the door to a significant increase in the risk of treatment-induced infection-related side effects^56,57 due to disruption of its role in immune response. Several putative reasons for the twofold role of the complement system in severe SARS-CoV-2 pathogenesis can be posited. The first is timing, with early activation being beneficial and the later role of the complement becoming detrimental⁴⁸; the second being the “Goldilocks zone”⁵⁸, with both too much activation and too little being detrimental; and the third one being that the particular pathway of activation determines whether complement activation will be detrimental or beneficial. Our own data could be considered as supporting the third hypothesis, seeing as we have identified that an increase in the expression of components of the complement system specifically involved in only the classic complement activation pathway is connected to survival outcomes. Previous research has also shown that outcomes of COVID-19 are associated with changes in concentration of components related to particular pathways of complement-system activation, but not just increase or decrease in activation of system overall^51,52,53. Based on our data and previous research, increased expression of C1QA and C1QB could to be involved in pathogenesis and lethality of COVID-19, and could be considered a molecular marker of milder cases, and stimulation of their expression could be a possible option for therapeutic intervention. However, role of the complement system in SARS-CoV-2 is complex, and therapeutic interventions involving complement system are not without risk. Further research is required in order to better understand how complement system affects outcomes in SARS-CoV-2 infections.

VSIG4 and SERPING1—Potential key regulators of the complement system in COVID-19

Another key upregulated gene related to negative regulation of the immune system, humoral system, and complement activation is VSIG4, a gene encoding V-set and Ig domain-containing four proteins, also known as the “complement receptor of the immunoglobulin superfamily” (CRIg). This gene is expressed by macrophages and dendritic cells, participating in ingestion of complement-opsonized particles. VSIG4 also inhibits activation of a nonclassical complement pathway by binding to C3b^59,60, leading additional credence to the idea that the particular pathway of complement activation is one of the factors determining survival outcomes in COVID-19 patients, since its overexpression is linked to survival in our data. Besides participating in phagocytosis of complement-opsonized particles and complement system regulation, engagement of VSIG4with its ligands in macrophages also causes immunosuppressive and anti-inflammatory effects⁶¹.

The role of VSIG4in COVID-19 pathogenesis is not fully understood. However, data points exist that point to its increased expression being both possibly detrimental and possibly beneficial. Reduced expression of VSIG4 leading to inhibition of anti-inflammatory activity was previously described in СOPD (chronic obstructive pulmonary disease) as an important factor in its pathogenesis⁶². Since, based on our GO: BP enrichment, we consider negative regulation of the immune system being important in survival for severe COVID-19, its previously demonstrated immunosuppressive and anti-inflammatory activity could be beneficial. On the other hand, overexpression of this protein is considered an unfavorable factor in ARDS (acute respiratory distress syndrome)⁶³ and lung cancer^64,65, which could also mean its overexpression could be a detrimental factor in COVID-19, seeing how respiratory failure is one of the more dangerous symptoms in severe cases of COVID-19. Therefore, increased VSIG4 can be both a detrimental and beneficial factor in severe COVID-19. Its overexpression in our data suggests that it could play a protective role in severe COVID-19 through modulation of the complement system response, inhibition of alternative pathway of complement activation and anti-inflammatory and immunosuppressive activity as a receptor. Based on its position in gene–term network and its role in several important processes in COVID-19 pathogenesis, we consider VSIG4 to be one of the key genes and a potential biomarker and target for therapeutic intervention.

The final complement-system related differentially expressed key gene is SERPING1. SERPING1encodes C1-INH, a serine protease inhibitor (serpin), which plays a role in inhibition of the innate immune responses and complement system activation⁶⁶. C1-INH is capable of inhibiting all three pathways of complement activation and has shown the capability of limiting the complement and inflammation-related damage in multiple injury models^67,68. C1-INH inhibits those pathways via forming covalent bonds with C1r and C1 s (classic pathway), MASP-1 and MASP-1 (lectin pathway), and C3b (alternative pathway)⁶⁶, and it is worth noting that, since this is a posttranslational mechanism, an observed increase of C1QA and C1QB expression on the transcriptional level may be not connected with increased activity of C1q. Being an interferon-induced gene, SERPING1could be posited to play a role of negative feedback mechanism, limiting immune system hyperactivation and, therefore, improving outcomes⁶⁶, which is especially important, considering the propensity of severe COVID-19 to cause immune-related damage to various tissues. SERPING1is implied to play a role in COVID-19 pathogenesis with its expression being shown to be increased compared to healthy individuals in reanalysis of three cases of from 2020⁶⁹. It has also been shown that the T allele of variant rs78958998 is associated with increased risk of viremia in COVID-19 cases⁷⁰. Overall, its beneficial role as a negative feedback mechanism against runaway immune processes fits very well with current understanding of the possible role of immune-mediated damage in severe COVID-19.

Currently, to our knowledge, no therapies for COVID-19 or other infectious diseases have been developed that specifically target ISG15, SERPING1, or VSIG4 genes. As an IFN-induced gene, ISG15 can be upregulated by introduction of a wide variety of compounds, such as type I IFN, double-stranded RNA, lipopolysaccharide, or all-trans-retinoic acid⁷¹. However, the effect of such compounds is not specific to inducing ISG15 expression, and they have a wide variety of effects on many different biological processes and induce a variety of different responses. IFN treatment appears to be the most promising among those, however evidence for efficacy and safety of IFN treatments in SARS-CoV-2 is mixed⁷², leading to outcomes that are better than control in some and worse than control in other reports. Currently, treatment of hereditary angioedema with recombinant SERPING1 (Ruconest, Salix Pharmaceuticals) is approved in some jurisdictions⁷³, and it could be interesting to investigate it for efficacy in SARS-CoV-2 cases, although it is important to be mindful of how angioedema pathophysiology is different from that of COVID-19. As for VSIG4, anti-VSIG4 antibodies have been demonstrated to be able to induce immune response in vitro, in vivo, and ex vivo assays, but the relevance of these findings for SARS-CoV-2 remains to be shown, especially considering the fact that, in the original paper presenting those findings, the goal was to suppress activity of VSIG4, not activate it⁷⁴.

Determinants of outcome in the Delta vs. Alpha variant of COVID-19

Next, we would like to compare our data obtained in this study to the results of our own previous investigations of COVID-19. Previously, in a different paper, we have demonstrated that a lethal outcome in COVID-19 caused by SARS-CoV-2 Delta variant infection was associated with increased levels of pro-inflammatory cytokines⁷⁵. However, in another paper, we have shown a downregulation of LDL particle-receptor pathway activity in ICU patients who survived severe COVID-19 infection¹⁴, which was in agreement with results obtained in independent studies, including a meta-analysis of 22 published works^76,77,78. It is worth considering why those findings have not been replicated in the present study. The methodology of the present study and the 2021 study are very similar, from patient selection to the analysis of differential expression. One of the significant differences is in the variant of SARS-CoV-2 causing the disease. The Alpha variant was the most widespread variant in St. Petersburg during the patient cohort collection of the 2021 study¹⁴. The Alpha variant is characterized by seven missense mutations (N501Y, A570D, D614G, P681H, T716I, S982 A, and D1118H) and three deletions (69/70 del and 144 del) in spike protein S. The Delta variant is characterized by a different set of mutations (T19R, G142D, Δ157–158, L452R, T478 K, D614G, P681R, and D950 N), and the identified differences could affect the interaction of S protein with ACE2 and immune-system proteins^10,79,80. This could lead to both differences in virus infectivity and activation of different metabolic pathways in pathogenesis of COVID-19. The role of particular missense variants in interaction with cellular proteins is not clearly understood. For example, there have been no mutations identified in the gene encoding PLpro between Delta and Alpha variants of SARS-CoV-2^81,82. However, no differences in the expression of ISG15 between survivor and nonsurvivor groups have been identified in our data in the case of severe COVID-19 caused by infection by the Alpha variant of SARS-CoV-2. This suggests that the nature of our findings may be variant-specific. However, despite our best effort, several confounding variables are present, and could not be fully accounted for, as will be discussed further in the Study Limitations section. Therefore, interpretation of strain-specificity in our findings needs to be done cautiously. Nevertheless, the differences between the results of this and the previous study are evident. Significantly enriched biological processes in this study are almost exclusively related to the immune system, whereas there was a high number of metabolism-related significantly enriched processes in our previous study¹⁴. One possible explanation of such differences would be the role of evolving immune evasion of SARS-CoV-2. With SARS-CoV-2 constantly getting better at avoiding the immune system⁸³, the role of subtle differences of immune system regulation and activity in determining outcomes of COVID-19 should also increase.

Further study of patients infected by different variants of SARS-CoV-2 may elucidate the role of the variant- specific aspects of pathogenesis of COVID-19. We believe that continuous research of COVID-19 cases caused by different variants of constantly evolving SARS-CoV-2 is important. To accomplish this, our suggestion would be to attempt to conduct periodic studies of SARS-CoV-2 infection using various methods, such as RNA-seq, while keeping the design of each study as constant as possible. This would allow keeping as many of the confounding variables the same as possible and to investigate the variant-specific aspects of pathogenesis of COVID-19. In our work, taking together the present paper and a previous publication of a similarly designed investigation of COVID-19 infection caused by the Alpha variant¹⁴, we have attempted to do exactly that. Fortunately, thanks to the spread of vaccinations, improving standard of care, and reduced load on health-care systems, number of lethal cases of COVID-19 has significantly gone down. However, this impairs our ability to adhere to the same study design in the future (comparison of survivors and nonsurvivors), which suggests that, going forward, in a series of experiments on different strains and variants of SARS-CoV-2, we will need to employ different comparisons, such as, e.g., severe versus mild cases.