In this study, we compared leukocyte transcriptional profiling of COVID19+ to COVID19- ICU patients. The most striking result was the identification of unique and distinct transcriptional profiles in leukocytes obtained from COVID19+ ICU patients. The consistent differences between the transcriptional profiles of COVID19+ patients and COVID19- patients with non-COVID lung injury that we observed are similar to previous clinical reports indicating that COVID19+ patients do not exhibit “typical” symptoms of acute respiratory distress (3), and that COVID19+ patients have a unique inflammatory profile (5, 6). As indicated by both single-gene based and gene set (GSEA) approaches, the major disease-specific transcriptional responses of leukocytes in critically ill COVID19+ ICU patients were: (i) a robust overrepresentation of IFN related gene expression; (ii) a marked decrease in the transcriptional level of genes contributing to protein synthesis and bioenergy metabolism; and (iii) the dysregulated expression of genes associated with coagulation, platelet function, complement activation, and TNF/IL6 signalling.
Functional enrichment analysis also highlighted a marked shift in cellular population within the circulation with a significant enrichment for transcriptional profiles associated with CD14+ monocytes and CD33+ myeloid cells, underscoring the lymphopenia observed in COVID19+ ICU patients. Further, compared to COVID19- patients, COVID19+ patients exhibited an overall decrease in the expression of genes associated with these cell types, suggesting a potential shift in the innate immune response. Moreover, these data can be interpreted to indicate a decrease in CD14+ monocyte activation and/or a decrease in overall CD14+ monocyte number within COVID19+ ICU patients compared to COVID19- ICU patients, who exhibited increased white blood cells.
Disease-specific alterations in IFN signalling have been previously identified in patients with COVID19 (18), and our recent publication on a similar patient cohort identified significantly increased levels of IFNs in the plasma of COVID19+ ICU patients compared to both age- and sex-matched COVID19- ICU patients and healthy controls (5). Classical type I (α/β) and type II (γ) IFN signalling through IFN receptors and Janus kinase-signal transducer and activator of transcription (JAK-STAT) pathways activate IFN-stimulated genes (ISGs), in part through nuclear factor kappa B (NFkB) signalling, that mediate a complex web of responses that include, but are not limited to, pathogen sensing, sensitization, desensitization and inhibition of viral entry (19-21). Toll-like receptor (TLR) mediated cascades, independently identified in our network correlation analyses, are one of the major pathogen sensing pathways activated by type I IFN signalling. In parallel with the induction of ISG transcription, IFN signalling promotes a global shutdown in translation and cellular RNA degradation to induce the death of virus-infected cells. This correlates with the upregulation of E2F genes that regulate the core transcriptional machinery driving cell cycle progression, dictating the timing and fidelity of genome replication and ensuring genetic material is accurately passed on through each cell division cycle (21).
While transient activation of IFN signalling pathways often occurs during viral infections, sustained IFN signalling can paradoxically promote viral persistence and induce both immune suppression and chronic inflammation (22). For example, chronically increased IFN levels have been implicated in the inhibition of T cell proliferation and chemokine release, and the TRAIL-induced apoptosis of CD4+ T cells during both infectious and non-infections pathologies (23, 24). While TRAIL (TNFSF10) expression was not identified as significantly increased in our patient population, there was a trend towards enhanced TNFSF10 expression in COVID19+ compared to COVID19- patients at day 1 of ICU admission. In addition to TRAIL, sustained type I IFN levels can also induce the expression of pro-apoptotic molecules while enhancing the immunosuppressive actions of dendritic cells, monocytes, and macrophages through augmented expression of cytokines (22). Thus, sustained IFN signalling can effectively “derail” the normal innate immune response and promote viral persistence at both the cellular and tissue level.
Potentially relevant to COVID19+ ICU patients, chronically high levels of IFN have also been detected in, and postulated to be causal of, macrophage activation syndrome (MAS). Moreover, dysregulated and persistent macrophage activation has been previously described in COVID19+ patients (25). In this scenario, SARS-CoV-2 infection would lead to defects in lymphocyte cytolytic activity (26), where natural killer (NK) cells and cytolytic CD8+ T cells exhibit reduced capacity to lyse infected and otherwise activated antigen presenting cells, resulting in prolonged cell-cell interactions with persistent amplification of pro-inflammatory cytokine cascades (27). The subsequent cytokine response induces macrophage activation and haemophagocytic lymphohistiocytosis, ultimately contributing to multi-organ dysfunction and death (28). While this is a compelling working hypothesis, larger scale and more focussed analyses will be required to test this possibility.
In addition to IFN, our data also identified a significant enrichment in genes associated with TNF signalling in COVID19+ compared to COVID19- patients. TNF signalling is a critical pathway in the regulation of the proinflammatory response following infection, where it typically is observed in the very early stages of infection and returns to basal levels as the infection progresses (29, 30). With respect to COVID19, previous work by our group in a similar patient population identified a persistent elevation of circulating TNF in COVID19+ patients (5). Together, these studies suggest that augmented TNF signalling may be key in COVID19 and may provide a potential therapeutic target through the use of neutralizing antibodies or small molecule inhibitors as has been previously suggested (5, 31).
One of the key clinical features of COVID19 is upregulation of the pro-thrombotic phenotype and microvascular complications, leading to COVID19–associated coagulopathy (7). This is characterized by venous, arterial, and microvascular thrombosis despite the use of anti-coagulant therapies (32-38). Our study provides compelling evidence of COVID19 associated changes in coagulation-related gene expression levels that may exacerbate thrombosis caused by endothelial cell injury and platelet activation in COVID19+ ICU patients (7). In particular, disease-specific upregulation of SERPINE1 (encoding plasminogen activator inhibitor-1; PAI-1), VWF (von Willebrand factor), and GZMB (Granzyme B) gene expression levels in COVID19+ ICU patients may contribute to thrombotic disorders in COVID19 patients (5, 39). Granzyme B is a serine protease that may be involved in thrombus formation through induction of endothelial damage and endothelial cell apoptosis (40), which we recently found to be the most discriminatory inflammatory analyte for identifying COVID19 status in ICU patients (5). Assessed in combination, the COVID19-associated changes in the expression of genes involved in platelet activation, hemostasis, fibrin clot formation, platelet cytosolic calcium levels, and Thromboxane A2 expression, in parallel with transcriptional changes in genes regulating chemokine levels and granulocyte activation, identified in this study may provide both a unique transcriptional profile that identifies and reflects the pathophysiological mechanism(s) involved, and potential therapeutic targets for the treatment of patients with life-threatening COVID19 infections.
We have also identified a decrease in the expression of genes required for heme metabolism in the immune cells from COVID19+ ICU patients. These findings may be relevant to viral sequestration of iron from haemoglobin. Many viruses require iron for replication, and utilize a variety of mechanisms to decrease cellular iron metabolism to increase iron availability. Cell free iron can bind to damage recognition receptors, and in addition to driving oxidative stress, cell-free hemoglobin has been shown to be injurious in the lungs of patients with ARDS (41). Although little is known about the importance of iron and heme metabolism in SARS-CoV-2 or other coronavirus infections, depriving the virus of iron may represent a promising adjuvant therapeutic strategy.
As shown in Figure 5, COVID19+ ICU patients also exhibited changes in the expression of genes regulating the addition of ubiquitin moieties and/or small ubiquitin-like moieties (SUMO) to proteins. Viruses rely heavily on the host’s cellular replication machinery, including the ability to by-pass and/or exploit cellular ubiquitin and SUMOylation conjugating systems, to successfully proliferate and achieve infection. Many viruses are able to target essentially every step of ubiquitination and SUMOylation processes, including the activation of genes that encode ubiquitin ligases or other molecules that alter the intracellular pools of free ubiquitin and SUMOs available for conjugation to proteins that modulate replication processes (42).
Another intriguing network correlation identified was neurotrophin signalling. Neurotrophins are structurally related neuropeptides and include nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF). These peptides play key roles in the survival and development of peripheral and central nervous systems. In the context of COVID19, neurotrophins are likely to be components of a neuroimmune response to infection, as shown previously in the early stages of the H1N1 influenza pandemic (43). Neurotrophins have been previously implicated in the pathogenesis of other lung conditions associated with airway inflammation and hyperreactivity, such as asthma (44). Moreover, it is plausible that the dysregulation of neurotrophin-encoding genes in COVID19+ ICU patients may be linked to neural damage resulting from SARS-CoV-2 infection of the brainstem and/or other central nervous system (CNS) sites (45). Brain stem infections with SARS-CoV-2 have been postulated to dysregulate CNS responses to changes in the concentrations of CO2 and O2, further compounding the ARDS induced by infection of the lungs (46).
Collectively, the data from our exploratory study identify a unique leukocyte transcriptional profile in COVID19+ ICU patients vs. COVID19- ICU patients. However, it is important to note that, while novel, these findings are derived from a study with some design limitations. First, all the data in this study were derived from critically ill patients at a single admission time point. As such, we were unable to assess any transcriptional changes associated with ICU admission or track these changes over the course of the patients time in the ICU. Second, the study included only 7 COVID19+ and 7 COVID19- patients. However, even with what could be considered to be a small sample size, we were still able to identify a unique transcriptional profile associated with SARS-CoV-2 infection. Importantly, these studies, which are supported by previous findings by our group (5-7), will be hypothesis-generating for future studies of disease severity and/or outcome. Third, the COVID19-specific transcriptional profiles reported here have not been independently verified. However, our findings were consistent with previous studies comparing COVID19 patients with healthy controls, and with the COVID19-induced changes in plasma proteins identified in our recently published studies from a similar patient cohort (5-7). For example, increased expression of the GZMB and SDC1 genes, as well as the increased expression of multiple genes encoding IFN signalling molecules, are supported by findings of increased abundance of granzyme B, syndecan-1, and IFNs respectively in the plasma of COVID19+ vs. COVID19- ICU patients (5-7).
We also utilized multiple approaches to analyze the transcriptome data, including both single-gene as well as gene set enrichment analysis (i.e., GSEA). GSEA applies a threshold-free overrepresentation analysis strategy to evaluate genome wide expression profiles and to determine whether a priori defined sets of genes show statistically significant, cumulative changes in gene expression that correlate with phenotype (COVID19+ vs. COVID19-). While single-gene methods can identify individual gene effects, this approach may be undermined by the variance across individuals seen in complex disease states. GSEA is complementary to single-gene analysis in providing a framework with which to examine changes in higher levels of biological organization, such that alterations in gene expression associated with a disease can manifest at the level of biological pathways or co-regulated gene sets, rather than individual genes. Thus, while we did not independently assess the changes in expression of individual genes identified in these RNAseq analyses, the collective support provided by previous publications in combination with the use of multiple analysis techniques validates our overall findings.