Re-analysis of the first public SARS-CoV-2 transcriptomics data

Last updated on November 29, 2024
Published on April 14th, 2020

Written by Axel Martinelli
⏱ 10 min read

Re-analysis Sar Cov2 datasets thumbnail image

Introduction

The first public SARS-CoV-2 transcriptome from Blanco-Melo et. al. is now available and we took the opportunity to upload it onto our Omics Playground platform for analysis. The datasets consist of transformed lung alveolar (A459) and primary human lung epithelium (NHBE) infected with different Multiplicity of Infections (MOI) of SARS-CoV-2 (collected after 24h), as well as transformed lung alveolar (A459) cells infected with either respiratory syncytial virus (RSV) (collected after 24h) or Influenza A virus (IAV) (collected after 9h).

Word Cloud

In order to replicate the main findings from this dataset, we first generated a wordcloud of the most significantly enriched gene sets in both SARS-CoV-2 infected cell types (Figure 1). In both cases, gene sets related to both viral infections and immune responses were particularly prevalent.

Figure 1. A) Word Cloud of enriched gene sets for A459 cells infected with SARS-CoV-2. B) Word Cloud of enriched gene sets for NHBE cells infected with SARS-CoV-2.
Figure 1. A) Word Cloud of enriched gene sets for A459 cells infected with SARS-CoV-2. B) Word Cloud of enriched gene sets for NHBE cells infected with SARS-CoV-2.

These enriched gene sets are also reflected in the similarity of SARS-CoV-2 with other viral expression profiles stored in our Omics Playground platform. A network of shared enriched datasets between various expression profiles (obtained using the “Signature” module in the platform) highlights the prevalence of core antiviral responses (Figure 2).

Figure 2. Network of shared enriched datasets between SARS-CoV-2 infected A549 cells and other gene expression profiles stored in the platform.
Figure 2. Network of shared enriched datasets between SARS-CoV-2 infected A549 cells and other gene expression profiles stored in the platform.

However, SARS-CoV-2 infections display unique features as well. Both IAV and RSV infections result in the differential expression of a significantly larger battery of genes compared to SARS-CoV-2 in A459 cells, with only four upregulated genes (SAT1, C3, IRF9 and OAS1) and no down regulated gene shared between all three species (Figure 3). 

It should be noted that the analysis performed here is based on the intersection of three different methods (EdgeR, DESEQ2 and limma), with a maximum FDR of 0.05 and a minimum log-fold change of 0.5, which may result in rather stringent criteria. 

SARS-CoV-2 infections shared more differentially expressed genes with RSV (47) than IAV (1) infections, which could highlight significant differences between coronavirus and influenza infections (Figure 3).

Figure 3. Venn diagram of the up- and down-regulated genes in the pairwise comparisons of IAV, RSV and SARS-CoV-2-infected A459 cells with their respective controls.
Figure 3. Venn diagram of the up- and down-regulated genes in the pairwise comparisons of IAV, RSV and SARS-CoV-2-infected A459 cells with their respective controls.

Evidence from the literature points to a distinctly muted initial immune response induced by coronavirus infections compared to other airways viral infections, particularly with reference to interferon induction, possibly a consequence of immunomodulation from the virus itself (you can read more about it in the paper by Geng Li et al.). 

Indeed, we can observe an almost completely absent expression of various type I IFN genes in SARS-CoV-2 infected cells compared to IAV- or RSV-infected cells (Figure 4). This could explain the drastic difference in differentially expressed gene counts between infections.

Figure 4. Log2CPM histograms of the read counts for A) Interferon alpha 1 (IFNA1), B) Interferon beta 1 (IFNB1) and C) Interferon lambda 1 (IFNL1). Expression is high in IAV- and RSV-infected cells, but very low or absent in SARS-CoV-2-infected cells and in control samples.
Figure 4. Log2CPM histograms of the read counts for A) Interferon alpha 1 (IFNA1), B) Interferon beta 1 (IFNB1) and C) Interferon lambda 1 (IFNL1). Expression is high in IAV- and RSV-infected cells, but very low or absent in SARS-CoV-2-infected cells and in control samples.

SARS-Cov-2 infections in NHBE cells induce a greater number of differentially expressed genes compared to infections in A459 (Figure 5). This difference could be explained both by the different host cell types, as well as the different MOI used to produce the infections. Only 19 differentially expressed genes were thus shared between SARS-CoV-2 infections, all of which were upregulated (Figure 5).

Figure 5. Venn diagram of the up- and down-regulated genes in the pairwise comparisons of SARS-CoV-2-infected A459 cells and NHBE cells with their respective controls.
Figure 5. Venn diagram of the up- and down-regulated genes in the pairwise comparisons of SARS-CoV-2-infected A459 cells and NHBE cells with their respective controls.

The list included several genes related to immune response, including Interleukin-8 (IL-8) (Figure 6). 

The latter is of particular interest, as IL-8 levels have been previously associated with viral bronchiolitis by Abu-Harb M et al., Bont L. et al. and McNamara P. S. et al.. Furthermore, BCG vaccination (which was recently shown to be protective against SARS-CoV-2 infections) induces an increase in IL-8 mRNA expression. 

It would be of interest to verify if the considerably increased levels of IL-8 in ventilated patients could have implications for the treatment of critical patients of the current pandemics.

Figure 6. List of differentially expressed genes shared between NHBE- and A459-SARS-CoV-2-infected cells with their cumulative fold-change.
Figure 6. List of differentially expressed genes shared between NHBE- and A459-SARS-CoV-2-infected cells with their cumulative fold-change.

In the next post, we will look into potential inhibitory drugs against SARS-CoV-2 infections, compare these results with those obtained from previous posts on SARS and MERS coronavirus, and finally look for potential biomarkers of infection unique to SARS-CoV-2 and/or other coronavirus infections.

Drug Connectivity Analysis

Using the “SystemsBio” tab on the Omics Playground platform, we consulted the Drug Connectivity Map database to identify potentially inhibitory drug profiles against both SARS-CoV-2 infected cell lines from the Blanco et al dataset. Analysis revealed that in the profiles obtained from both cell lines, several Drug Modes of Actions were shared (Figure 1A and 1B). In particular the following inhibitors for the following pathways were in common between the two infected samples: PI3K, MEK and mTOR. This is consistent with previous results based on SARS- and MERS-coronavirus infections (here and here).

Figure 1.​ ​A​) Modes of action of potential inhibitors against A549 cells infected with SARS-CoV-2. ​B​) Modes of action of potential inhibitors against NHBE cells infected with SARS-CoV-2.
Figure 1.​ ​A​) Modes of action of potential inhibitors against A549 cells infected with SARS-CoV-2. ​B​) Modes of action of potential inhibitors against NHBE cells infected with SARS-CoV-2.

On an individual drugs level, it was interesting to notice the presence of the JAK-inhibitor TG-101348 (​Figure 2A​) based on the NHBE-infected cells profile analysis. JAK-inhibitors act as anti-inflammatory drugs and ​may be beneficial ​against the cytokine storm occurring in severe SARS-CoV-2 infections. Bariticinib, a JAK-inhibitor has been recently suggested as a ​potential treatment​ of SARS-CoV-2 and a ​trial study​ on the efficacy of Jakafi (another JAK-inhibitor) to treat SARS-CoV-2 infections is ongoing.

In our previous posts we had identified the kinase inhibitor trametinib (with tested activity against MERS coronavirus infections in vitro) as a potential inhibitor. Trametinib was also suggested as an inhibitor in both NHBE- and A549-SARS-CoV-2-infected cell lines (​Figure 2B-2C​) from the Blanco et al study. Furthermore, the related drug selumetinib (​Figure 2D​) was also highlighted in the analysis of the profile from infected NHBE cell line (but not in the profile from infected A459 cells).

An intriguing compound prominently suggested by the NHBE-infected cells profile is fostamatinib (​Figure 2E​). Although the inhibitory potential against A459-infected cells was not statistically significant, it also showed a negative correlation there. ​Fostamatinib​ is a spleen tyrosine kinase (SYK) inhibitor used in the treatment of immune thrombocytopenia, an autoimmune disease resulting in low platelet counts. SYK may be involved in the ​recognition​ of viral infection, so it may be involved in inducing cytokine storms during coronavirus infections or may be directly modulated by the virus to evade immune responses. In any case, there is no evidence in the literature of the use of SYK-inhibitors in a viral infection context, so this could potentially represent a novel target pathway for drug repurposing.

GSEA plot for target genesets against the expression profile of SARS-CoV-2 infected NHBE cells.
Figure 2. A​) GSEA plot for the TG-101348 geneset against the expression profile of SARS-CoV-2 infected NHBE cells. ​B​) GSEA plot for the trametinib geneset against the expression profile of SARS-CoV-2 infected NHBE cells. ​C​) GSEA plot for the trametinib geneset against the expression profile of SARS-CoV-2 infected A549 cells. ​D​) GSEA plot for the selumetinib geneset against the expression profile of SARS-CoV-2 infected NHBE cells. ​E​) GSEA plot for the fostamatinib geneset against the expression profile of SARS-CoV-2 infected NHBE cells.

We have previously identified the expression levels of the SSX2 gene in infected cells as a potential biomarker for coronavirus infections. Unfortunately, no reads in any sample were detected for the SSX2 gene. 

However, the presence of cells infected from two other pathogens (IAV and RSV) meant that a biomarker analysis by disease could be performed. 

The decision tree produced by the “Find biomarkers” module (Figure 3A) indicated that the MX1 gene was highly expressed during SARS-CoV-2 and RSV infections (Figure 3B), but not IAV infections (albeit the latter was collected after 9h, while the former after 24h post-infections). 

MX1 is a gene associated with defense against viral infections, so it’s elevated expression is expected and the difference with IAV may be due to the different timepoints. RSV and SARS-CoV-2 infection profiles could be distinguished by TP63 (a transcription factor) gene expression levels, which were strongly repressed upon RSV infection (as also observed in the literature) compared to control samples, but were unaffected by coronavirus infections in either cell types.

Biomarker analysis foir SARS dataset using Omics Playground
Figure 3.​ ​A​) Decision tree of potential biomarkers distinguishing Sars-CoV-2, IAV, RSV and uninfected cells. ​B​) Boxplot of the log2-fold read counts for gene MX-1. ​C)​ Boxplot of the log2-fold read counts for gene TP63.

From this brief analysis, it looks like there are several similarities between SARS-CoV-2 and RSV infections, which could help in the repurposing of drugs. Potential inhibitors shared mode of actions with other coronavirus infections and also included a JAK-inhibitor, consistent with ongoing drug trials.

Unlock the full potential of your RNA-seq and proteomics data.

Try Omics Playground

About the Author

Axel Martinelli

Axel Martinelli’s academic background is in molecular biology and parasitology. He earned a Ph.D. on the genetics of strain-specific immunity against malaria infections and a master’s degree in bioinformatics with specialization in the analysis of omics data. During his postdoctoral career, he worked on genomics and transcriptomics studies and is currently the head of biology at Bigomics Analytics.