Reprogramming of the intestinal epithelial-immune cell interactome during SARS-CoV-2 infection

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) represents an unprecedented worldwide health problem. Although the primary site of infection is the lung, growing evidence points towards a crucial role of the intestinal epithelium. Yet, the exact effects of viral infection and the role of intestinal epithelial-immune cell interactions in mediating the inflammatory response are not known. In this work, we apply network biology approaches to single-cell RNA-seq data from SARS-CoV-2 infected human ileal and colonic organoids to investigate how altered intracellular pathways upon infection in intestinal enterocytes leads to modified epithelial-immune crosstalk. We point out specific epithelial-immune interactions which could help SARS-CoV-2 evade the immune response. By integrating our data with existing experimental data, we provide a set of epithelial ligands likely to drive the inflammatory response upon infection. Our integrated analysis of intra- and inter-cellular molecular networks contribute to finding potential drug targets, and suggest using existing anti-inflammatory therapies in the gut as promising drug repurposing strategies against COVID-19.


Introduction
Since the first reported case in the province of Wuhan (China), severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has spread to almost every country in the world (Hui et al., 2020;Wu et al., 2020), posing an extraordinary threat to global public health (Deng and Peng, 2020;Han et al., 2020). Transmitted through respiratory droplets, aerosols and fomites, the virus can be detected in upper respiratory tract samples and is believed to primarily target airway and alveolar epithelial cells, vascular endothelial cells and lungresident macrophages (Tay et al., 2020). Once inside the host cell, SARS-CoV-2 releases viral RNAs which can be translated into proteins using host machinery (Merino et al., 2021).
SARS-CoV-2 infection is not limited to the lungs: other organs can be infected too, including the heart, kidney, brain, and the intestine (Gupta et al., 2020). In addition to directly infecting key organs, the main hurdle of SARS-CoV-2 infection is that, in some severe cases, it generates an excessive inflammatory response mediated by both the innate and adaptive immune systems . The overactivated inflammatory response, also known as cytokine release syndrome (CRS) or cytokine storm, is the result of high levels of circulating cytokines and chemokines, and it is thought to be responsible for the severe COVID symptoms some patients experience (Arunachalam et al., 2020). Yet, there is no clear understanding of which particular inflammatory pathways and cell types are driving these damaging inflammatory responses, and whether some organs are more important than others in initiating this (Stone et al., 2020). The role of gut microbes and previous infections were also raised as potential risk-increasing factors (Földvári-Nagy et al., 2021).
COVID-19 patients with severe symptoms show elevated expression of inflammatory cytokines (IL-2, IL-4, IL-6, IL-10 and IL-18; (Gu et al., 2020;Park et al., 2021) that are correlated with elevated levels of the gut inflammatory marker faecal calprotectin and an altered microbiome (Effenberger et al., 2020;Zuo et al., 2020). COVID-19 patients also often present various gastrointestinal (GI) symptoms such as vomiting, diarrhoea and abdominal pain Guo et al., 2020;Lin et al., 2020). Interestingly, patients with GI symptoms show decreased production of key proinflammatory cytokines and reduced disease severity and mortality following SARS-CoV-2 infection, indicating a potential role of the gut in the disease course (Livanos et al., 2021).
Recently, human intestinal organoids have been used as a tool to study SARS-CoV-2 infection in the gut and the inflammatory responses of specific intestinal epithelial cell types (Lamers et al., 2020;Stanifer et al., 2020;Triana et al., 2021a;Zang et al., 2020). These studies provided evidence that SARS-CoV-2 is able to infect and actively replicate in human intestinal cells (Lamers et al., 2020). From the organoid experiments, we learned that while most intestinal subpopulations are susceptible to SARS-CoV-2, enterocytes are the most affected (Lamers et al., 2020;Triana et al., 2021a). Studies using human intestinal organoids also revealed that, contrary to the limited type I and type III interferon (IFN) immune response observed in the lungs (Blanco-Melo et al., 2020;Hadjadj et al., 2020), the response to SARS-CoV-2 infection in the gut is characterised by the production of IFN and interferon stimulated genes (ISGs), and found that IFNs may actually provide protection to the intestinal epithelial cells against SARS-CoV-2 (Stanifer et al., 2020).
In inflammatory bowel disease (IBD), intestinal inflammation is known to drive dysregulated epithelial-immune cell interactions, often manifesting in extra-intestinal diseases (Weidinger et al., 2021). Hence, the question arises as to whether COVID-19 patients may share a similar dysregulated inflammatory response driven by gut epithelial-immune interactions as that observed in IBD patients. Indeed, examination of human intestinal samples has shown infiltration of lymphocytes and other inflammatory mediators in the lamina propria upon SARS-CoV-2 infection, suggesting that infection of gut epithelial cells results in the activation of local immune populations (Guo et al., 2021). Yet, the exact effects of viral infection in the gut and the role of epithelial cell-immune cell interaction in mediating the inflammatory response of the body are not known.
To our knowledge, no study has been carried out so far to analyze epithelial-immune crosstalk in the gastrointestinal tract upon SARS-CoV-2 infection. Hence, in this study we aim to model the effect of viral infection in host intestinal cells, the role of intestinal epithelial cell-immune cell crosstalk during infection, as well as their contribution to the inflammatory response. As miRNA-like sequences were recently found in the SARS-CoV-2 genome, and their potential role and targets predicted during infection (Mirzaei et al., 2021;Saçar Demirci and Adan, 2020), in a separate analysis, we assess the potential role of such viral regulators. Furthermore, we aim to assess potential tissue-specific differences between colon and ileum in these effects. To do this, we used previously generated single-cell RNA sequencing (scRNA-seq) data of SARS-CoV-2 infected ileum and colon-derived human organoids (Triana et al., 2021a) and of gut-resident immune cell populations (Martin et al., 2019;Smillie et al., 2019), as well as SARS-CoV-2-human miRNA/protein-protein interactions. We employ two independent tools, ViralLink and CARNIVAL, to reconstruct intracellular and intercellular networks, connecting intestinal epithelial cells and resident immune cells upon infection. With our integrated analysis, we provide a better understanding of the effect of viral infection on intestinal epithelial cells, and the role of intestinal epithelialimmune cell crosstalk during SARS-CoV-2 infection. Ultimately, our analyses may help to find key intercellular inflammatory pathways involved in these crosstalks, which could pave the way for potential successful strategies against the cytokine release syndrome associated-symptoms observed in severe cases of COVID-19.

Intercellular analysis
Input data

Intestinal epithelial cells
Single cell transcriptomics data of colonoids and enteroids infected with SARS-CoV-2 was obtained from (Triana et al., 2021b).

Intestinal resident immune cells
Single cell expression data from ileal and colonic resident immune cells was obtained from (Martin et al., 2019) and (Smillie et al., 2019), respectively. For the analyses, data from healthy samples and uninflamed Crohn's disease samples was used for colonic and ileal immune cell populations, respectively. Following removal of all genes with count = 0, normalised log2 counts across all samples (separately for each cell type) were fitted to a gaussian kernel (Beal, 2017). All genes with expression values above mean minus three standard deviations were considered as expressed genes for the given cell type in the given intestinal location. For the intercellular ligand-receptor predictions, a representative collection of immune cells relevant in gut inflammation and SARS-CoV-2 infection based on previous literature was selected, which included all macrophages, T cells, B cells, plasma cells, ILCs, mast cells and a representative group of dendritic cells (Filbin et al., 2020;Martin et al., 2019;Schultze and Aschenbrenner, 2021;Sette and Crotty, 2021;Smillie et al., 2019). Cell type labels were maintained as originally published.
Ligand-receptor interactions (intercellular interactions; full list available at https://github.com/korcsmarosgroup/gut-COVID) were predicted between epithelial cells types and resident immune cells according to the following conditions: 1. The ligand is differentially expressed in the epithelial cell (upon SARS-CoV-2 infection -in directly infected or bystander cells) 2. The receptor is expressed in the immune cell in the relevant dataset (ie, ileal or colonic immune cells)

The ligand-receptor interaction is present in OmniPath
Intercellular interactions were defined separately for directly infected epithelial cells and bystander epithelial cell populations in the ileum and in the colon. Enteroid epithelial data was paired with ileal immune cell data (Martin et al., 2019), while colonoid epithelial data was paired with colonic immune cell data (Smillie et al., 2019). Intercellular interactions were defined between every possible pair of epithelial cells and immune cells for each condition.

Scoring of ligands, receptors and immune cell types involved in ligandreceptor interactions
To assess the importance of specific ligands, receptors and immune cell types, additional parameters were computed using the ligand-receptor network. First, the number of interactions between each epithelial and immune cell type was computed by summing up all the possible interactions between each differentially expressed epithelial ligand and each of the receptors expressed by the specific immune cell type. Second, the number of immune cell types involved in each ligand-receptor pair was computed by counting the number of different immune cell types which were expressing the receiving receptor. Third, for each ligand, a "sum of receptor expression value" was computed for each interacting immune cell type, based on the number of interacting receptors and the mean expression level of the interacting receptors.

Data visualisation
Venn diagrams were generated using the 'gplots' R package. Heatmaps were generated using the 'ggplot2' and 'pheatmap' packages. Barplots were generated with the 'ggplot2' package. Network visualisations were done using Cytoscape (version 3.8.2) (Shannon et al. 2003;Su et al. 2014). All scripts used to generate these plots are available on the Github repository of the project (https://github.com/korcsmarosgroup/gut-COVID).

Intracellular analysis
Two previously established tools were employed to predict the effect of SARS-CoV-2 infection on intestinal epithelial cells: ViralLink and CARNIVAL (Liu et al., 2019;Treveil et al., 2021). Both tools, using related but distinct methods, infer causal molecular interaction networks. These networks link perturbed human proteins predicted to interact with SARS-CoV-2 viral proteins or miRNAs, to transcription factors known to regulate the observed differentially expressed ligands in infected epithelial cells.

Input data
To reconstruct the intracellular causal networks, three different a priori interactions datasets were used. Information on viral proteins and their interacting human binding partners was obtained from the SARS-CoV-2 collection of the IntAct database on 1st October 2020 (Hermjakob et al., 2004;Orchard et al., 2014). Predicted SARS-CoV-2 miRNAs and their putative human binding partners were obtained from (Saçar Demirci and Adan, 2020).
Intermediary signalling protein interactions known to occur in humans were obtained from the core protein-protein interaction (PPI) layer of the OmniPath collection using the 'OmnipathR' R package on 7th October 2020 (Türei et al., 2016). Normalised transcript counts and differentially expressed ligands were obtained from single cell transcriptomics data of colonoids and enteroids infected with SARS-CoV-2 obtained from (Triana et al., 2021a) as previously described.
ViralLink pipeline Intracellular causal networks were inferred using the ViralLink pipeline, as described in . Briefly, a list of expressed genes in infected immature enterocytes (originally known as "immature enterocytes 2" (MMP7+, MUC1+, CXCL1+)) from SARS-CoV-2-infected ileal and colonic organoids (Triana et al., 2021a) was generated from a normalised count table by fitting a gaussian kernel (Beal, 2017). The list of expressed genes in the infected immature enterocytes population was subsequently used to filter the a priori molecular interactions from OmniPath and DoRothEA, to obtain PPI and TF-TG subnetworks where both interacting molecules are expressed (described as "contextualised" networks). Transcription factors regulating the differentially expressed ligands were predicted using the contextualised DoRothEA TF-TG interactions and scored as described in . Human binding proteins of viral proteins and viral miRNAs obtained from the IntAct database (Hermjakob et al., 2004;Orchard et al., 2014) and (Saçar Demirci and Adan, 2020), respectively, were connected to the listed TFs through the contextualised PPIs using a network diffusion approach called Tied Diffusion Through Interacting Events (TieDIE) (Paull et al., 2013). In this model, all viral protein-human binding protein interactions were assumed to be inhibitory in action, based on likely biological function, and given a lack of clear literature evidence of proven action. All viral miRNA-human binding protein interactions were set as inhibitory based on biological action of miRNAs (Huang et al., 2011). The final reconstructed network contains "nodes", which refers to the interacting partners, and "edges", which refers to the interaction between nodes. Nodes include viral proteins and miRNAs, human binding proteins, intermediary signalling proteins, TFs and differentially expressed ligands. Edges include activatory or inhibitory interactions.
For both ileal and colonic data, separate networks were generated using the viral miRNA and viral protein as perturbations, and subsequently joined using the "Merge" function within Cytoscape to generate the final intracellular network. Nodes and edges were annotated according to their involvement in networks downstream of viral miRNAs or proteins. Further analyses were performed separately on the different layers of the network: miRNA specific, protein specific or shared nodes.

CARNIVAL pipeline
Intracellular causal networks were inferred using CARNIVAL and associated tools for analyses of expression data as described in (Liu et al., 2019). For simplicity, we refer to the pipeline as described in (Liu et al., 2019) as the CARNIVAL pipeline. Briefly, PROGENy is used to infer pathway activity from the log2 FC of the infected immature enterocytes 2 gene expression data (Schubert et al., 2018). Next, using the TF-TGs interactions (from DoRothEA) and the differential expression data from infected organoids, VIPER was used to score TF activity based on enriched regulon analysis (Alvarez et al., 2016). Here, only the top 25 TFs regulating at least 15 target genes were taken forward, and a correction for pleiotropic regulation was included. Finally, CARNIVAL applied an integer linear programming approach to identify the most likely paths between human interaction partners of viral proteins or miRNAs and the selected TFs, through PPIs from OmniPath, considering the activity scores from PROGENy and VIPER. Viral protein-human binding protein interactions signs were specified to CARNIVAL as 'inhibitory', based on likely biological function, and given a lack of clear literature evidence of proven action. All viral miRNAhuman binding protein interactions were also set as inhibitory based on biological action of miRNAs (Huang et al., 2011).

Network functional analysis
Functional overrepresentation analysis was performed on the networks constructed as above-mentioned using the R packages 'ClusterProfiler' and 'ReactomePA', for Gene Ontology (GO) (Ashburner et al., 2000)) and for Reactome (Fabregat et al., 2018;Yu and He, 2016;Yu et al., 2012)

Selection of ligands involved in the inflammatory process
To assess the importance of specific ligands in driving the inflammatory process upon SARS-CoV-2 infection, the list of differentially expressed ligands in infected immature enterocytes in both colon and ileum was validated using independent data from three previously published studies. To identify ligands whose expression was induced by cytokines, ligands were compared to DEGs in human colonic organoids exposed to cytokines from (Pavlidis et al., 2021). To identify ligands already known to influence immune cell population, ligands were compared to two databases: ImmunoGlobe, a manually curated intercellular immune interaction network (Atallah et al., 2020), and ImmunoeXpresso, a collection of cell-cytokine interactions generated through text mining (Kveler et al., 2018).
Finally, to identify ligands that could directly explain blood cytokine level changes in COVID-19 patients via direct immune cell regulation, ligands were compared to the data from a large dataset we recently manually compiled using COVID-19 patient publications ).

Data availability
The workflow (and necessary input data) and the full ligand-receptor interaction tables are available in the GitHub repository of the project (https://github.com/korcsmarosgroup/gut-COVID). All other relevant data is in the main text and in supplementary files.

Reconstructing an epithelial-immune interactome
Our previously published data on ileal and colonic human organoids infected with SARS-CoV-2 suggested that immature enterocytes were the main epithelial population affected by infection (Triana et al., 2021a). In this study, we wanted to further investigate the effects of epithelial infection on epithelial-immune cell crosstalk in the gut by integrating single cell data and network biology approaches (Figure 1).
To do this, we integrated epithelial cell (Triana et al., 2021a) and immune cell ( These numbers are lower than those predicted to be differentially expressed upon infection by (Triana et al., 2021a), indicating that some ligands are not affected by direct upstream signalling changes but by more complex mechanisms, or the original knowledge network used as input for the analysis did not contain information about such ligands (Menche et  Interestingly, viral miRNAs that were predicted to regulate upstream the altered intracellular signalling were mostly different between colon and ileum (miR_10,11,16,18 in the colon and miR_4,5,6,18 in the ileum).
Additionally, by analysing these networks, we observed that NOTCH1 and SMAD4, seem to be central to the intracellular signalling cascade in the colon, by receiving several signals driven by viral miRNAs and viral proteins, respectively (Supplementary Figure 3).
Interestingly, both the Notch and TGF-β SMAD-dependant signaling pathways are involved in intestinal epithelial cell homeostasis, including stem cell maintenance, progenitor cell proliferation (Carulli et al., 2015) and maintenance of cell differentiation (Yamada et al., 2013), suggesting a modulation of these pathways upon infection. In the ileal network, JAK2 and CREB1, as well as SMAD2, SMAD 3 and ERK2 (MAPK1) seem to play a central role in the intracellular PPI signalling driven by viral miRNAs and viral proteins, respectively, and JAK2 and both SMAD2 and SMAD3 were also upregulated upon infection (Supplementary involved in these intercellular interactions, assessing any potential similarities or differences between the colon and ileum (Figure 1 and Methods).
Upregulated ligands in infected immature enterocytes were largely shared between colon and ileum, with one ligand (FAS) uniquely upregulated in the colon ( Figure 2B). Shared upregulated ligands included mainly cytokines and chemokines (CXCL2/3/10 and tumor necrosis factor (TNF-a)) and the adhesion factor ICAM1. Interestingly, several additional chemokines (CSF1, CXCLs, TNFSFs) and adhesion factors (PLAU, EFNA) were upregulated in the ileum upon infection, which we did not find in the colon ( Figure 2B).
Additionally, 38 receptors on immune cells targeted by upregulated ligands in the colon were all shared with the ileum, and were mainly represented by chemokine receptors (CXCRs, CCRs) (Figure 4). Epithelial-immune interactions driven by upregulated ligands were also mostly shared in the colon and ileum (1 unique to colon, 219 unique to ileum, 66 shared) ( Figure 6 & Supplementary Figure 8A, 8B).
Next, to understand which ligands were driving the most interactions with specific immune cell types, we scored them based on the number of ligand-receptor interactions they had with the different immune cell types analysed (Figure 1 and Methods). Chemokines  Figure 9A, 9B), and RIPK1 in the colon only ( Figure 9A).
To understand the role of these epithelial-immune interactions, we performed a functional overrepresentation analysis of the participating upregulated epithelial ligands and receiving receptors on immune cells (Figure 1, Supplementary Figure 8, and Methods). In line with the extensive overlap in upregulated intercellular interactions (Figure 6), most functions were shared between colon and ileum, and included chemotaxis (GPCR signalling, chemokine signalling), immunity (interleukin signalling), apoptosis (caspase activation) and angiogenesis (VEGFA-VEGFR2 pathway) (Supplementary Figure 9A, 9B). One colonicspecific function was related to pro-inflammatory responses (TNF signalling) and one ilealspecific function was related to stem cell renewal (BMP signalling) (Supplementary Figure   9A, 9B).

Downregulated epithelial ligands upon infection impact antigen presentation and focal adhesion pathways
Downregulated ligands in infected immature enterocytes were partially shared between colon and ileum, but were tissue-specific to a large extent ( Figure 2B). Additionally, receptors on immune cells targeted by downregulated ligands were partially shared between colon and ileum (66), but several of them were tissue-specific (63 unique to colon, 38 unique to ileum) (Figure 4). In line with this, while some downregulated interactions in infected immature enterocytes were shared (104)

Implication of epithelial ligands in the inflammatory process
With our experimental data-based analysis, we pointed out several differentially expressed ligands in the epithelial-immune cell network relative to infected immature enterocytes which could play a role in driving the inflammatory process upon SARS-CoV-2 infection. To validate their importance during immune reactions, we exploited independent data from three previously published studies (Figure 1).

First, by comparing the differentially expressed ligands upon SARS-CoV-2 infection to DEGs
in human colonic organoids exposed to inflammatory cytokines (Pavlidis et al., 2021), we identified 24 ligands whose expression change is regulated by cytokines during intestinal inflammation (Table 1, and Methods). These ligands are more probable to contribute to the inflammatory responses upon infection. Next, by comparing ileal and colonic ligands to data from ImmunoGlobe, a manually curated intercellular immune interaction network (Atallah et al., 2020) and ImmunoeXpresso, a collection of cell-cytokine interactions generated through text mining (Kveler et al., 2018), we identified 12 ligands previously known to influence immune cell populations (Table 1, and Methods). The full list of affected immune cell types for each epithelial ligand is available in Table 2B. Finally, to understand which ileal and colonic ligands could explain blood cytokine level changes of COVID-19 patients via direct immune cell regulation, we used data from , and identified 6 ligands capable to create the detected blood cytokine levels during infection ( Table 1, and   Methods).
Using this assessment, we were able to rank the differentially expressed ligands for their importance in the inflammatory process, and subsequently listed the 18 highest ranked ligands, for which there is strong evidence of their role in epithelial-immune cell interactions during the inflammatory SARS-CoV-2 disease response ( Table 1) (Figure 3A, 3B). Furthermore, pathways related to cell cycle (negative regulation of G2/M transition) and cell proliferation were also altered upon infection (Figure 3A, 3B), in line with a previous phosphoproteomics study finding a correlation with cell cycle arrest upon SARS-CoV-2 infection (Bouhaddou et al., 2020). Interestingly, several pathways involved in cell differentiation, cell migration and epithelial polarization, were also modulated upon infection in our study (Figure 3A, 3B).
By using available ligand-receptor interaction data, we aimed to elucidate mechanisms by which infected epithelial cells in the gut recruit innate and adaptive immune cell populations to find key interactions driving the immune response in the gut. In the colon, the number of downregulated ligands (29) was higher compared to upregulated ligands (6) Figure 9A, 9B).
Notably, four chemokine receptors identified by our study (CXCR6 in the ileum, CCR1/2 and CCR9 in both ileum and colon) are coded in a genomic region that has been found associated as a COVID-19 risk locus on chromosome 3 (Schultze and Aschenbrenner, 2021). Importantly, recruitment of neutrophils by CXCL8 in the lung, which presented the epithelial ligand driving most interactions in the ileum in our study (Figure 2D), has been associated with disease severity in COVID-19 patients, supporting the role of chemokinedriven immune cell recruitment in disease manifestation (Park and Lee, 2020).
In both colon and ileum, we found strong downregulated interactions driven by epithelial HLAs (HLA-A, B, C) and B2M, a subcomponent of the major histocompatibility complex I (MHC I) (Figure 2C, 2D). According to our analysis, these ligands were mainly binding to KLR receptors, which are mainly presented on NK cells (Supplementary Figure 8A, 8B).
Downregulation of HLAs represents a common immune evasion mechanism of viruses (Koutsakos et al., 2019), and has recently been discovered as a mechanism that SARS-CoV-2 protein ORF8 may use to escape host immune surveillance (Park, 2020).
Uniquely in the colon, we found strong downregulated interactions driven by epithelial laminins (LAMB3 and LAMC2) and integrins, with T cells and macrophages as the main immune cell types targeted upon infection (Figure 2C, 6A). Laminin-integrin binding contributes to focal adhesion of immune cells to the inflamed tissue (Simon and Bromberg, 2017) the overrepresentation of focal adhesion pathways and RHO GTPase signalling (Supplementary Figure 9A), which is involved in the migration of leukocytes to the site of infection (Biro et al., 2014). Overall, downregulation of laminins could represent a strategy for immune evasion following viral infection uniquely in the colon. Furthermore, laminins are known to play a role in shaping the architecture of intestinal mucosa, and an altered expression has been observed in Crohn's disease, a type of IBD, driven by pro-inflammatory cytokines TNF-α and IFN-γ (Bouatrouss et al., 2000;Francoeur et al., 2004;Mahoney et al., 2008).
Finally, calmodulin genes (CALM1, CALM2, CALM2) were predicted to drive several downregulated ligand-receptor interactions (Figure 2C, 3D), mainly binding to cyclic AMPspecific phosphodiesterases (PDEs) (PDE1A, PDE1B, PDE1C) on immune cells in both tissues upon infection (Supplementary Figure 8). PDEs, whose activation is calcium/calmodulin dependent, are responsible for cyclic AMP (cAMP) degradation in T cells, which is a potent inhibitor of T-cell activation (Bjørgo et al., 2011). Hence, the  (Figure 2A). Notably, previous reports suggests that IgA is the main type of immunoglobulin induced by mucosal infection of SARS-CoV-2, stressing the importance of the crucial role played by IgA-mediated mucosal immunity in anti-SARS-CoV-2 infection (Sterlin et al., 2021). Interestingly, in the colon most of these cell-cell interactions were driven by downregulated ligands, including laminins, HLAs and calmodulins, possibly suggesting a decreased antigen presentation and calciumdependent activation of these cell types (Figure 2A, 2C). Conversely, in the ileum these interactions were driven by upregulated ligands, mainly cytokines/chemokines (TNF-a, CXCLs, CSF1) and adhesion factors (ICAM1, PLAU), possibly suggesting increased recruitment of these cell types to the epithelium (Figure 2A, 2D). The methodology we used for our analysis has some limitations. When constructing the intracellular causal network, the effect of SARS-CoV-2 proteins towards human binding partners was always considered as inhibitory. However, this is not always the case. In the future, with increasingly available data, a more refined model could be generated.
Furthermore, two different single cell transcriptomics datasets were used for colonic and ileal immune cell populations, due to the unavailability of both datasets from the same experiment. Similarly, IBD uninflamed data and healthy data were used for the ileum and colon respectively, as healthy control scRNAseq immune cell data for both tissues was not available at the time of the analysis. Finally, the a priori resources used to infer the intracellular and intercellular interaction networks may have some intrinsic limitations associated with them (Dimitrov et al., 2021)  JSR received funding from GSK and Sanofi and consultant fees from Travere Therapeutics. Table 1. Key differentially expressed ligands produced by infected immature enterocytes drive the inflammatory process upon SARS-CoV-2 infection. Table   showing a  whether ligand expression was found to be regulated by cytokines during inflammation based on results from (Pavlidis et al., 2021). Ileal data was not available (n.d.) in this study, so no conclusions could be drawn for ileal ligands. 'Known to affect immune cells' indicates whether the ligand was found to be regulated by immune cells using data from ImmunoGlobe (1) and ImmunoeXpresso (2) databases. 'Directly explain patient blood cytokine levels' indicates whether the ligand was found to directly regulate blood cytokine levels in COVID-19 patients from .          Within the "Node Table", rows indicate whether the node or edge belong to intracellular signals stemming from viral miRNA only, viral protein only or both ("shared").