Semantic clustering analysis of E3-ubiquitin ligases in gastrointestinal tract defines genes ontology clusters with tissue expression patterns | BMC

Expression profiling of E3 Ub-ligases demonstrates tissue-specific gene combinations in the gastrointestinal tract

The expression profiling of E3 Ub-ligases and ubiquitination-related genes was performed on samples from stomach, small intestine and colon of WT mice (Additional file 1). We found that each organ of GIT has their specific set of up- and downregulated genes (Fig. 1), suggesting their organ-specific roles. For further analysis, the genes were divided into three groups, corresponding to defined Ub-ligases types. In colon samples, we detected 118 upregulated genes (sum of upregulated genes compared to intestine or stomach), in intestine 22 were upregulated (sum of upregulated genes compared to colon or stomach) and in stomach there were 78 upregulated (sum of upregulated genes compared to colon or intestine). No significant difference was found in the representation of individual Ub-ligases types with a p value 0.736 for upregulated genes in each organ. Genes from this cluster were expressed at the same level in stomach, small intestine and colon and might have the same functional activity for each organ.

Representative distribution of upregulated genes in stomach, small intestine and colon divided into main Ub-ligase classes

In the next step, we defined ontological terms for each gene and compared the distribution of ontology terms with theoretical distribution. We found 26 significantly enriched terms for genes, which were differentially expressed in specific parts of GIT (Table 1; Additional file 3: Table S2). These ontology terms displayed specific functions of given genes.

Ontology clusters of stomach represent genes that are involved in stress responses by regulating various intracellular signal transduction with association of the SCF ubiquitin ligase [27]. Ontology groups display that the small intestine is mostly represented by genes playing roles in immune and inflammatory responses. This ontology group is represented by the suppressor of cytokine signaling (SOCS) family of protein-encoded genes Socs1 and Socs3. These genes are responsible for negative regulation of cytokine signaling through the JAK/STAT3 pathway, and was mentioned as a probable substrate recognition component of a SCF-like ECS E3 ubiquitin-protein ligase complex [28, 29]. Next, there was a group of upregulated genes in the small intestine (14% out of all upregulated), which are responsible for negative regulation of the insulin receptor signaling pathway (Table 1; Additional file 2). The most representative gene for this group was Cish, which is also a member of SOCS family [30].

Ontology clustering the upregulated genes revealed enrichment of genes specifically involved in DNA repair, apoptosis, and catabolic processes. For instance, upregulation of the E3 ubiquitin-protein ligase Trim62 is a positive regulator of I-κB kinase/NF-κB signaling and DNA-binding of transcription factors (Table 1; Additional file 2) [31]. Mul1 and Trim13 (also known as Ret finger protein 2, RFP2), among the others were genes that take a role in positive regulation of cell death by modulating the innate immune response against viruses [32].

By applying semantic ontology analysis, we were able to find the groups of genes which belong to the same ontology cluster but which have a unique and specific GIT expression pattern. This kind of analysis allowed us to identify genes, which share similar functions in parallel regulatory networks. Additional file 3: Table S2 shows the ontology term combinations that were identified after applying semantic analysis on the GIT specific Ub-ligases and ubiquitination-related genes. Thus, ontology term combinations GO:0018193: peptidyl-amino acid modification and GO:0042326: negative regulation of phosphorylation unites five genes: Socs4, Socs5, Cbl, Socs1, Socs3 (Fig. 2A, B; Additional file 3: Table S2). Inside this ontology, genes Socs1, Socs3 are upregulated in the small intestine and downregulated in the colon, whereas genes Socs5 and Cbl exhibited the opposite expression. On the contrary, Socs4 does not show a significant difference in expression for colon and small intestine.

Selected examples of ontologically related genes in GIT. Heatmaps for organ specific (green columns, top) and general genes (orange columns, top) that are annotated simultaneously by ontology terms GO:0018193: peptidyl-amino acid modification and GO:0042326: negative regulation of phosphorylation (a), GO:0045309: protein phosphorylated amino acid binding and GO:0044267: cellular protein metabolic process (c), and GO:0045309: protein phosphorylated amino acid binding and PW:0000417: ubiquitin, ubiquitin like/proteasome degradation pathway (e). Schematic visualization of gene ontology and their more general terms for GO:0018193 and GO:0042326 (b), GO:0045309 and GO:0044267 (d), GO:0045309 and PW:0000417 (f). We note that these terms are highlighted by yellow color. Selected ontologies illustrate the ability of semantical clustering to group genes that carry the same biological function in different parts of the organ. Scheme shows the relationships among ontology terms related to biological processes that might help the analyst to orient in such extensive ontologies and then interpret the meaning of terms easier

Another ontology term combination GO:0045309: protein phosphorylated amino acid binding and GO:0044267: cellular protein metabolic process includes Fbxw7, Nedd4, Btrc, Cblb and Socs3 genes (Fig. 2C, D; Additional file 3: Table S2). In this group, Socs3 is upregulated in the small intestine and downregulated in the colon, while genes Nedd4, Btrc, Cblb are characterized by the inversed expression pattern. Expression of Fbxw7 did not significantly differ between small intestine and colon. Interestingly, the same gene set (with additional Cbl gene, which is a paralog of Cblb) belongs to another ontology term combination GO:0045309: protein phosphorylated amino acid binding and PW:0000417: ubiquitin, ubiquitin-like/proteasome degradation pathway. For this ontology group, these genes display a similar tissue expression pattern (Fig. 2E, F; Additional file 3: Table S2) representing the theory that the same genes might be involved in multiple regulatory pathways.

Moreover, Socs genes represented the most illustrative expression pattern in the GIT, particularly Socs1, Socs3, Socs4, and Socs5. These genes appeared in 9 out of 10 ontology combinations. Thus, Socs5 was always downregulated in the small intestine and upregulated in the colon and stomach. Socs1 and Socs3 showed upregulation in the small intestine and downregulation in the colon and Socs1 was downregulated in the stomach tissue. Socs4 did not show any difference in expression between SI and colon which indicates its equal contribution for homeostasis of these tissues.

Genes from the same ontology cluster alter their expression pattern after induced epithelial damage in colon

In order to reveal possible parallel networks, we used a model of epithelial regeneration. We hypothesized that the genes involved in tissue regeneration might be masked by steady state homeostasis, thus their function might become apparent after tissue-challenged conditions, such as epithelial inflammatory damage. For this purpose, we induced epithelial damage by treating mice with DSS, which is widely used for mouse colitis models [23]. Following expression profiling detected 35 Ub-ligases and ubiquitination-related genes with altered expression in the distal colon after DSS treatment (Additional file 4; Additional file 5: Fig. S2). Then the localization of expression of 35 Ub-ligases and ubiquitination-related genes was monitored in the DSS treated and untreated distal colon tissue. It was observed that most of genes changed their expression pattern from the crypt base to apical part along the colon crypt, which is related to the disruption of homeostatic cell balance (Fig. 3A) in the crypt after damage. In the untreated colon, 22 out of 35 genes were detected on the luminal (apical) side of the crypt, whereas after DSS treatment only 12 genes remained, and others translocated either to the crypt base or spread over the entire crypt (Fig. 3B–D). This alteration in the expression pattern could be associated with the damaged and missing apical cells due to the inflammation. Similarly, a shift in expression localization was also observed in the case of genes that were originally expressed in the crypt base.

Spatial changes in expression of Ub-ligases of the same ontology group inside colonic crypts after induced epithelial damage. a Schematic illustration of the cell type residence in colonic crypt during homeostasis (based on (22, 38, 39, 41)). b–d Epithelial damage causes a shift in the expression pattern of ubiquitin ligases and ubiquitination-related genes in distal colon along the crypt axis. e In situ hybridization images of the DSS treated and untreated colon demonstrate spatial changes in expression of genes obtained from the ontology combination GO:0045309 and GO:0044267, GO:0045309 and PW:0000417. Red arrow depicts the gene Cbl expression upregulated after the DSS-induced damage and blue arrows display the other specific genes downregulated after the damage. Scale = 50 µm

As for expression in the crypt body, we observed more genes that displayed expression in the affected area, some of which showed strong signal (Cbl, Fbxl5, Rnf19b, Apc2) (Fig. 3B–E; Additional file 5: Fig. S3). This observation might be a result of inflammation, local immune responses and/or robust epithelial regeneration that occurs after the DSS treatment, suggesting an essential role of those Ub-ligases in the mentioned processes. For instance, high presence of Apc2 (component of the anaphase-promoting complex/cyclosome (APC/C)) could be associated with dynamic cell cycle regulation for the maintaining of tissue homeostasis [33]. On the other hand, there were genes with significantly decreased expression after the treatment. Some E3s kept their original intramucosal location of expression (for example, Trim25, Smurf2, Trim11), whereas others were no longer detected (such as Bmi1, Asb11, March7, Btrc) (Additional file 5: Fig. S3) which indicates that these Ub-ligases are not necessary in the tissue damage response. Notably, together with Ub-ligases, Brcc3 (a part of multisubunit BRCC complex with Lys 63-linked deubiquitinating activity [34]) retained its expression pattern after the treatment, but expression levels were significantly decreased in the damaged tissue (Additional file 5: Fig. S3).

We further examined the response of genes that were grouped into the same ontology term combination GO:0045309 and GO:0044267, GO:0045309 and PW:0000417. In this analysis, Socs1 and Socs3 genes were downregulated in the colon while Btrc, Bmi1 and Cbl were upregulated. With the help of in situ hybridization we identified the specific expression regions in colon for each of these genes. Also, we saw that each expression region can differ under pathological conditions (Fig. 3E). In homeostasis, Btrc was highly expressed by the cells of the crypt apex, but its expression decreased remarkably after injury. A similar situation was observed for Bmi1, which was originally present in the apical part of the crypt and crypt base, respectively. Socs1 and Socs3 were localized to the crypt apex. In the damaged or regenerated tissue, they maintained this expression pattern, but their expression level decreased because of either missing or re-structured epithelia (Fig. 3E).

On the contrary, Cbl showed high expression at both the apical part of the crypt and crypt base, but DSS-induced damage significantly dispersed its expression through the entire crypt body. This could be explained by the potential communication of Cbl with signaling pathways maintaining stem/progenitor/mature cell balance during tissue regeneration (for example, through the protein tyrosine kinase-mediated signaling as mentioned in [35]. In conclusion, the results of ontological clustering and the subsequent in situ hybridization for the accordingly picked Ub-ligase genes displayed in Fig. 3 clearly suggest which cluster of Ub-ligases might be involved in colon mucosa inflammation and regeneration. This thorough analysis also exhibits the temporo-spatial dynamics of this Ub-ligase cluster function under these specific conditions.

Ub-ligases from the same ontology cluster expressed by several cell types in colon indicating their compensatory potential

To determine cell-type specific distribution of Ub-ligases in the colon, we used published single-cell RNA sequencing data of the murine colon as a reference [22]. We hypothesized that Ub-ligases from the same ontology cluster might be involved in the parallel signaling pathways in distinct cell types, but be active under the certain homeostatic condition or differentiation stage. This way, only Ub-ligase related genes were selected from the entire scRNA-seq dataset (n = 367) and were processed by the Seurat package [36]. For a cell subtype visualization, we performed principal component analysis (PCA), then the 10 most significant principal components were projected to two dimensions with Uniform Manifold Approximation and Projection technique (UMAP), the cells were then colored by their classification label [37]. We used established cell markers to determine cell types in proximal and distal colon, including enterocytes (Krt20+, Slc26a3+) [38], goblet cells (Atoh1+, Spdef+) [39, 40], tuft cells (Dclk1+) [41], chromaffin (also known as enteroendocrine) cells (Chga+, Chgb+) [42], proliferating (Lgr5-, Mki67+) and non-proliferating (Lgr5+ , Mki67−) stem cells (SCs) [17]. With the help of UMAP visualization, Ub-ligases were grouped into several clusters that correspond to specific cell type (Fig. 4A). However, there was no strict tissue association between distal and proximal parts of colon, and clusters there demonstrate just partial overlapping.

Distribution of Ub-ligases in colon. a UMAP analysis demonstrates that the colon Ub-ligases may be grouped into 13 cell specific clusters (labeled by colors). b–f Ub-ligases from the same ontology combination group displaying distributional expression among several cell types

To further analyze the Ub-ligase distribution throughout the colon, we focused on genes that were clustered into the same ontology combination groups GO:0018193 and GO:0042326, GO:0045309 and GO:0044267, GO:0045309 and PW:0000417 (Additional file 3: Table S2). These genes showed a shift in expression after DSS-induced inflammation (Fig. 3E). Thus, Socs1 and Socs3 were mostly expressed by stem cells that clustered as Lgr5+ undifferentiated, Lgr5+ amplifying undifferentiated SCs and goblet cells both in proximal and distal colon, together with enterocyte cells of proximal colon (Fig. 4B, C). Besides this, Socs3 was also typical for Lgr5-undifferentiated SCs cluster (Fig. 4C). However, Socs1 and Socs3 genes were not typical either for chromaffin, tuft cells or goblet cell at the crypt apex. Btrc was abundantly expressed in the clusters of goblet cells, Lgr5+ amplifying undifferentiated and Lgr5+ undifferentiated SCs (Fig. 4D), and it was rather scarce in chromaffin, enterocyte, and tuft cell clusters. As for Cbl, Btrc displayed comparable distribution in all cell clusters with the higher concentration in the enterocyte and Lgr5+ undifferentiated SCs clusters (Fig. 4E). Finally, Bmi1 showed the lowest cell cluster specificity with equal distribution through all clusters with relatively higher specificity for Lgr5+ undifferentiated and Lgr5+ amplifying undifferentiated SCs clusters (Fig. 4F). These results support the evidence that Ub-ligases are expressed by various cell types along the tissue, but their functions slightly differ depending on the cell type, developmental stage and homeostatic condition, as it was observed also after the colon injury. This might bring a novel proof of semantically defined similarity of those Ub-ligases and compensatory potential in relation to each other.