GEO (www.ncbi.nlm.nih.gov/geo/) is a huge public genomics data repository containing gene expression profiles of various diseases. We collected the datasets via using ulcerative colitis (UC) and chronic kidney disease (CKD) as keywords for search in the GEO database. The inclusion criteria were as follows: 1) The dataset contained at least fifteen samples in total; 2) The gene expression profiles were from adult human; 3) The included test samples should be from both patients and healthy individuals; 4) Raw data must be provided for further exploration. Finally, GSE66494 (CKD), GSE4183 (UC), GSE115857 (CKD), and GSE10616 (UC) were selected for further analysis (Table 1). For datasets GSE4183 and GSE10616, only samples from UC patients were included in the analysis. The GSM numbers of samples used in the analysis were provided in Supplementary Table 1.

Differentially expressed genes (DEGs) between diseased group and control group were identified using GEO2R (www.ncbi.nlm.nih.gov/geo/geo2r/) online analysis tool based on R packages (GEOquery and Limma). Genes with p-value < 0.05 and |log2 fold change (log2FC)|>1 were defined as DEGs. The R package “ggplot” was employed to visualize DEGs from datasets using volcano maps. The online Venn diagram tool was used to extract common DEGs both up-regulated or down-regulated between GSE66494 and GSE4183, or GSE115857 and GSE10616 respectively. Additionally, we chose GSE66494 and GSE4183 as our discovery datasets, and GSE115857 and GSE10616 as our validation datasets.

The above shared genes were submitted to the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (https://david.ncifcrf.gov/) online tool for further Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses. The enriched GO and KEGG pathways with p-value < 0.05 were selected and visualized using bar plots and bubble plots respectively.

Based on the overlapping DEGs, the Search Tool for the Retrieval of Interacting Genes (STRING) (http://string-db.org/) was employed for protein-protein interaction (PPI). The interaction sources of Textmining, Experiments, Databases, Co−expression, Neighborhood, Gene Fusion and Co−occurrence were adopted. The PPI pairs were extracted with an interaction score > 0.7 and then visualized by Cytoscape 3.9.0. The Cytoscape’s plug-in molecular complex detection (MCODE) was applied to the PPI network to explore key functional modules with selection criteria as follows: K-core = 2, degree cutoff = 2, max depth = 100, and node score cutoff = 0.2. And then, KEGG and GO analyses of the identified modular genes were performed using DAVID online tool and visualized by bubble plots and bar plots respectively.

The CytoHubba plug-in of Cytoscape identified hub genes from the PPI network, and then five algorithms (Radiality, MNC {Maximum Neighborhood Component}, MCC {Maximal Clique Centrality}, EPC {Edge Percolated Component}, and Degree) were used to confirm the final hub genes, which were illustrated in Venn diagram. Then, the hub genes were validated in GSE115857 and GSE10616. Ultimately, the above hub genes were submitted to GeneMANIA (http://genemania.org) to construct a co-expression network.

Receiver Operating Characteristic (ROC) curves were constructed using GraphPad Prism 9, and the area under the ROC curve (AUC) was calculated to assess the predictive value of the hub gene in four testing datasets, including GSE108112 (CKD), GSE200818 (CKD), GSE87466 (UC) and GSE47908 (UC).

To further reveal the immune cell landscape in renal and colonic biopsy, we employed the analytical platform CIBERSORT (https://cibersort.stanford.edu/) and LM22 signature to decode immune cell infiltration profiles of discovery and validation datasets. Pearson correlation analysis was used to identify the relationship between different immune cell phenotypes and hub genes, which was illustrated in lollipop plots.

Kidney biopsies in this study were obtained from patients with CKD-UC comorbidity (n=8), patients with CKD without reported comorbid bowel disease (including UC) (n=8) or living donors (n=8). Colon biopsies were derived from patients with CKD-UC (n=8), UC patients without kidney disease (n=8) or normal tissues from colonoscopy (n=8). Samples were obtained from the Department of Nephrology or Department of Pathology, The First Affiliated Hospital of Sun Yat-sen University. All participants provided informed consent and this study was approved by the Ethical Committee of the First Affiliated Hospital of Sun Yat-Sen University.

To detect p-Akt, ICAM1 and myeloperoxidase (MPO) in human specimens, slides were processed to remove paraffin and then hydrated in alcohol and phosphate-buffered saline. Then the slides were blocked with 10% normal donkey serum (Sigma, D9663) and incubated with primary antibodies, respectively, at 4°C overnight, including ICAM1 (Santa Cruz, sc-8439, 1:50), p-Akt (CST, 4060S, 1:50), and MPO (abcam, ab208670, 1:150). Later, slides were stained with corresponding fluorescence-labeled secondary antibodies for 1h at room temperature, including anti-rabbit Alexa Fluor 488 (Thermo Fisher, A21206, 1:1000) and anti-mouse Alexa Fluor 546 (Thermo Fisher, A11030, 1:1000). Nuclei were stained with DAPI. Slides were then mounted with Prolong Gold Antifade reagent (Invitrogen, P36934), and examined on a ZEISS LSM880 confocal microscope.

Unpaired t-test was used to detect the difference between two groups. And comparison among more than two groups of immunofluorescence results was assessed by the analysis of variance (ANOVA) using the statistical software GraphPad Prism and P values < 0.05 were considered statistically significant. Statistical significance is defined as * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, or ns (not significant).

As shown in the workflow chart summarized in Figure 1, CKD microarray dataset GSE66494 and UC microarray dataset GSE4183 were downloaded from the GEO database. Details of selected datasets are provided in Table 1. Subsequently, DEGs were identified (6604 in GSE66494 and 2473 in GSE4183) after screening with the criteria of p-value < 0.05 and |log2FC| > 1, which were illustrated by the volcano plots in Figures 2A, B, respectively. In addition, a total of 784 shared differentially expressed genes were found after taking the intersection of CKD DEGs and UC DEGs based on Venn diagram analysis (Figure 2C). Finally, 462 co-up- or down-regulated DEGs in GSE66494 and GSE4183 were obtained after excluding genes showing opposite expression trends.

To decipher the biological functions of the common DEGs, GO and KEGG pathway enrichment analyses were performed on 462 shared genes between CKD and UC using the DAVID online tool. For the biological processes of GO enrichment analysis, these genes were primarily enriched in immune-related biological processes, such as inflammatory response, immune response, neutrophil chemotaxis, and chemokine-mediated signaling pathway (Figure 3A). For cellular components and molecular functions, these DEGs were enriched in extracellular space, extracellular region, chemokine activity, cytokine activity, heparin binding, and receptor binding. In terms of KEGG pathway enrichment analysis, significantly enriched pathways included cytokine-cytokine receptor interaction, pathways in cancer, PI3K-Akt signaling pathway, chemokine signaling pathway, TNF signaling pathway, IL-17 signaling pathway and NF-kappa B (NFκB) signaling pathway (Figure 3B). Taken together, these results indicate that inflammatory response plays a vital role in both CKD and UC, and that extracellular immune mediators such as chemokines and cytokines and PI3K-Akt signaling pathway largely participate in the process.

To confirm our findings, we chose GSE115857 for CKD and GSE10616 for UC as validation datasets, the details of which are presented in Table 1. Initially, DEGs with p-value < 0.05 and |log2FC| > 1 were extracted from these two datasets (2001 in GSE115857 and 1450 in GSE10616). Then, Venn diagram analysis was performed on DEGs of the two groups, and a total of 155 common DEGs were spotted (Supplementary Figure 1A). Afterward, genes with contrary expression trends were removed, and the remaining 117 DEGs were introduced to GO and KEGG enrichment analyses. For GO enrichment analysis, inflammatory response was enriched, and DEGs were mostly presented in extracellular space as well (Supplementary Figure 1B). In terms of KEGG pathway enrichment analysis, PI3K-Akt signaling also ranked top as displayed in Supplementary Figure 1C.

Research on PI3K-Akt signaling pathway mainly focused on Akt, which is the most important downstream effector in the PI3K-Akt pathway, and phosphorylation is indispensable for Akt activation (21, 22). Therefore, to validate the PI3K-Akt signaling pathway across disease states, we performed immunofluorescence staining to detect phosphorylated Akt (p-Akt) levels in colon and kidney biopsies from healthy individuals, CKD patients, UC patients, and those suffering from both diseases (CKD-UC). As shown in Figure 3C, p-Akt was sparsely expressed in healthy kidney tissues and significantly increased in tubular epithelial cells (TECs) of CKD renal biopsies. Moreover, the expression level of p-Akt was further elevated in CKD patients who also developed UC. An interesting parallel occurred in the colon, where p-Akt, nearly unexpressed in normal human colon, was highly upregulated in colon biopsies from UC patients. A more abundant expression of p-Akt was detected in colon tissues of patients diagnosed with both UC and CKD (Figure 3D). To sum all, these findings vigorously indicate that immune regulation and PI3K-Akt signaling pathway might play crucial roles in the occurrence and progression of these two diseases.

In order to further identify the potential relationships between proteins encoded by the shared DEGs of CKD and UC, the PPI network with an interaction score > 0.7 was performed by STRING and visualized with Cytoscape, including 438 nodes and 317 edges (Figure 4A). Next, the MCODE plug-in was used to detect significant gene clustering modules. Eight modules consisting of 52 common DEGs and 118 interaction pairs were extracted, and the top three significant modules were shown in Figures 4B-D. GO analysis of these modular genes revealed that these modules were mainly enriched in inflammatory response, chemokine-mediated signaling pathway and chemotaxis (Figure 4E). KEGG enrichment analysis showed that these modules were highly associated with cytokine-cytokine receptor interaction, chemokine signaling pathway, TNF signaling pathway, and IL-17 signaling pathway (Figure 4F). Altogether, these data indicate that inflammation is important in both CKD and UC.

To investigate the potential genes playing critical roles both in the occurrence of CKD and UC, hub genes of common DEGs were determined by the CytoHubba plug-in. We calculated the scores of common DEGs by five algorithms (Radiality, MNC, MCC, EPC, and Degree) and then took the intersection of the top 20 hub genes obtained from different algorithms by Venn diagram analysis (Table 2). Finally, nine candidate hub genes were identified, including CXCL8, CCL2, CD44, ICAM1, IL1A, CXCR2, PTPRC, ITGAX, and CSF3 (Figure 5A). The full names and related functions of these candidate hub genes are shown in Table 3. Interestingly, the majority of these genes were involved in immune cell migration and adhesion, suggesting that the infiltration of inflammatory cells is crucial in the progression of both CKD and UC.

We then validated our findings in GSE115857 for CKD and GSE10616 for UC (Supplementary Figure 2A). After intersecting the candidate hub genes from discovery cohorts and validation cohorts, we obtained only one shared hub gene, ICAM1 (Figure 5B), which was shown to be significantly upregulated in both CKD and UC of validation cohorts (Supplementary Figures 2B, C), suggesting that ICAM1 may be of great importance in both diseases.

To further verify the reliability and clinical significance of the hub gene discovered by bioinformatics analysis, we obtained the colon and kidney biopsies from healthy individuals, CKD patients, UC patients, and CKD-UC comorbidity patients with the same grouping settings as above, and detected ICAM1 by immunofluorescence. As shown in Figures 5D, E, ICAM1 was rarely expressed in healthy kidney and colon tissues, while significantly upregulated in renal TECs and vascular endothelium of CKD patients and colonic mucosae of UC patients, consistent with previous reports (23, 24). Strikingly, the expression levels of ICAM1 were further dramatically elevated in both kidney and colon biopsies from CKD-UC patients, thus confirming the important role of ICAM1 in comorbid features and pathogenic mechanisms. On this basis, we analyzed the co-expression network and related functions of ICAM1 based on the GeneMANIA database. ICAM1 possessed a complex interaction network with physical interaction of 77.64%, co-expression of 8.01%, predicted of 5.37%, co-localization of 3.63%, genetic interaction of 2.87%, pathway of 1.83%, shared protein domains of 0.60% (Figure 5C). Twenty genes associated with ICAM1 were identified, mainly related to leukocyte migration, regulation of leukocyte migration, T cell activation involved in immune response, positive regulation of leukocyte migration, lymphocyte activation involved in immune response, and cellular extravasation. These results re-emphasize the importance of immune cell-related inflammation in these two diseases, and ICAM1 might be the key regulatory gene.

To further determine which types of immune cells in the kidney and colon are related to both CKD and UC, the proportions of 22 types of immune cells within kidney and colon samples were quantified by CIBERSORT (Figure 6A). Compared to healthy controls, both CKD and UC were commonly characterized by a remarkable enhancement of neutrophils, M0 macrophages (Mφ0), M1 macrophages (Mφ1), and CD4⁺ T memory cells (Figures 6A-D).

Furthermore, the relationships between these immune cells and ICAM1 expression levels were analyzed by Pearson correlation analysis. Intriguingly, the expression level of ICAM1 was positively correlated with neutrophils, Mφ0, Mφ1, CD4⁺ T memory cells, activated mast cells, gamma delta (γδ) T cells, and eosinophils, while negatively linked with regulatory T cells, naive B cells, plasma cells, resting NK cells and resting mast cells in both CKD and UC (Figures 7A, B). These matching trends of immune infiltration indicate that the upregulation of ICAM1 resulted in a similar immune cell landscape in the two diseases. In particular, neutrophils, Mφ0, Mφ1, and activated CD4⁺ T memory cells were the top four significant immune cells positively associated with ICAM1, suggesting that ICAM1 may specifically regulate these four types of immune cells in CKD and UC. The correlation patterns were highly consistent in validated cohorts (Supplementary Figure 3).

ICAM1 is a well-known adhesion receptor regulating leukocyte recruitment. In particular, existing evidence suggested that ICAM1 primarily mediates the infiltration of neutrophils in CKD and UC (23, 25). Therefore, we validated the relationship between ICAM1 and myeloperoxidase (MPO) positive neutrophil infiltration in colon and kidney biopsies from healthy controls, CKD patients, UC patients and CKD-UC patients. As expected, ICAM1 expression levels and neutrophil infiltration were both markedly increased in the colon and kidney tissues of UC patients and CKD patients compared with healthy controls, and were further upregulated in CKD-UC patients (Figures 7C, D). Most importantly, there was a significant positive correlation between the expression levels of ICAM1 and neutrophil infiltration, which was more pronounced in comorbid patients (Figure 7E, F). These results suggest that ICAM1 and subsequent neutrophil infiltration mediated by it might play an important role in the co-occurrence of CKD and UC.

Next, we employed ROC curve analysis to explore whether hub gene ICAM1 possesses diagnostic efficiency in the four testing datasets GSE108112 (CKD), GSE200818 (CKD), GSE87466 (UC) and GSE47908 (UC). Details of these datasets are provided in Table 1. The AUC values in these four datasets were greater than 0.7, demonstrating that the predictive capabilities of ICAM1 are excellent. Specifically, in GSE108112 (AUC: 0.8106) and GSE200818 (AUC: 0.7754), ICAM1 displayed an excellent performance in distinguishing CKD patients from healthy controls (Figures 8A, B). Likewise, ICAM1 worked well to separate UC patients from healthy individuals in GSE87466 (AUC:0.9573) and GSE47908 (AUC:0.8274) (Figures 8C, D).