Pache.co Information Technologies

Classification and identification of CD genes

The PRISMA flow chart for the selection of studies and genes is shown in Fig. 1. The PRISMA checklist is provided as supplementary information (File S1). The PubMed search imported into PubTerm identified 2496 articles, which referred to 1172 genes. The genes were reduced to 1055 after filtering for human genes (Fig. 1, Table S1). Then, each gene was carefully curated, annotated, and categorized as described in methods. The curation revealed that 400 genes were somehow potentially related to CD while 655 genes were not related due to annotations errors, the gene mention was casual, no association was found, or no evidence of variation was shown (mainly in subsequent gene expression changes). From the 400 genes, 81 were finally categorized as associated with a related disease such as IBD in general, UC, familial diarrhea syndrome, colorectal cancer, or chronic lymphocytic leukemia. Ninety-three genes were classified to other genetic alterations because its gene was uncertain, which included genes identified through haplotypes or intergenic regions in GWAS. Thus 226 genes were confirmed as associated with CD from this curation (Fig. 1).

Summary of categorized genes for Crohn’s disease. The numbers at left in arrows at the bottom represent the genes from search [1], while the numbers at right correspond to search [2]. *SNX20 was added manually

Full size image

Summary of categorized genes for Crohn’s disease. The numbers at left in arrows at the bottom represent the genes from search [1], while the numbers at right correspond to search [2]. *SNX20 was added manually

We also used GWAS Catalog [30] as a source of gene information. From 22 publications for CD risk, we obtained the 525 comprising variants. Variants were further filtered by removing those whose tagged gene was not reported, were not significant, or the gene or variant was duplicated, leaving only 133 genes (Table S2). From these, 77 were already categorized in the PubMed curation described above. Thus 56 genes were added to our list of genes, 27 intergenic variants were assigned to other genetic alterations, and 29 to GWAS evidence within gene. A list of the considered PubMed abstracts is shown in Supplementary Information (files S2 and S3). In summary, we identified 256 genes associated with diverse aspects of CD (Fig. 1).

A total of 126 genes were found to have experimental evidence of variants in CD. The top 26 genes of this category mentioned in more than 15 abstracts are shown in Table 1, while the genes with less than 15 abstracts are summarized in Table 2 and detailed in the Supplementary Information (Table S4). The topmost frequent genes for this category are well-known for their association with CD [11, 12], such as NOD2, TNF, IL23R, ATG16L1, TLR4, IL10, SLC22A4, SLC22A5, and IRGM (Table 1).

Besides the above 126 genes (Tables 1 and 2), we also found 71 genes associated with CD that were categorized as GWAS evidence within gene where an SNP is located within genomic coordinates, either an intron or exon (Table 3). Additionally, 18 genes were found to be specifically associated with treatment response in CD and 41 related to disease complications (Table 3).

Location of the functional variants

Of the 126 genes corresponding to the categories of experimental evidence of variants plus 71 genes with GWAS evidence within gene, only 17 genes (< 12%) were found to be annotated in ClinVar [32]. In this context, to support our systematic categorization, a supplementary file is provided with the information referring to the location of the variants that were not found in ClinVar (Tables S4 and S5). This information was revised and obtained either by the original paper or by the information related to the SNP reported at dbSNP [43].

In-silico functional analysis

To provide an overview of the 126 genes with experimental evidence to CD, a functional bioinformatics analysis was performed [33, 44]. For this, we assessed whether the genes prioritized in our study are indeed statistical and biologically relevant. We performed gene set enrichment analysis in multiple databases containing different biological terms, including pathways (KEGG), diseases (GAD), and gene ontology (GO) terms. Detailed results of all enriched gene sets are present in the Supplementary Information. Because the number of significant terms was high, repetitive, and difficult to interpret, we grouped the terms by biological and genetic similarity (see Methods).

Regarding diseases, as expected, the most similar term to CD is IBD and UC, validating our strategy (Fig. 2). We observed a dense group of genes and diseases where CD is located close to other groups of autoimmune and inflammatory diseases, certain types of cancer, hypersensitivity disorders caused by allergies and intolerance, infections by virus and bacteria, pregnancy complications, and metabolic complications.

Functional and expression analysis of CD-associated Genes. AThe y-axis comprises the GO, KEGG, and disease terms. The x-axis comprises the 126 genes used for the analysis. The heat map show values from 0 to 1, corresponding to the average presence of a gene within all terms merged in each group. Diseases include 10 groups, autoimmune (Lupus Erythematosus Systemic (LES), Psoriasis (PS), Diabetes type 1 (DT1), Vitiligo and Arthritis Rheumatoid (RA), Chronic/Inflammatory diseases (Psoriasis, Endometriosis, Sarcoidosis, and Cystic fibrosis), infections (Leprosy, Tuberculosis, Sepsis, Dengue, Hepatitis, and HIV), cancer (Meningioma, cervical, lung, esophageal, liver, ovarian, stomach and prostate cancer), hypersensitivity (Asthma, Atopy, Celiac disease, and dermatitis), pregnancy complications (Abortion, preeclampsia and premature birth), vascular diseases (Atherosclerosis, restenosis, and thromboembolism), brain and mental diseases (Depression, Migraine, Parkinson and Schizophrenia) and metabolic complications (Hypercholesterolemia, Obesity, Diabetes type 2 (DT2) and metabolic syndrome). B Differential Gene Expression of Genes represented by fold changes of all the CD genes, which show to be significantly different in at least one comparison. * denotes significance at q < 0.1 (holm p-value adjusted). Figures were mainly rendered in R software (https://cran.r-project.org/)

Full size image

Functional and expression analysis of CD-associated Genes. AThe y-axis comprises the GO, KEGG, and disease terms. The x-axis comprises the 126 genes used for the analysis. The heat map show values from 0 to 1, corresponding to the average presence of a gene within all terms merged in each group. Diseases include 10 groups, autoimmune (Lupus Erythematosus Systemic (LES), Psoriasis (PS), Diabetes type 1 (DT1), Vitiligo and Arthritis Rheumatoid (RA), Chronic/Inflammatory diseases (Psoriasis, Endometriosis, Sarcoidosis, and Cystic fibrosis), infections (Leprosy, Tuberculosis, Sepsis, Dengue, Hepatitis, and HIV), cancer (Meningioma, cervical, lung, esophageal, liver, ovarian, stomach and prostate cancer), hypersensitivity (Asthma, Atopy, Celiac disease, and dermatitis), pregnancy complications (Abortion, preeclampsia and premature birth), vascular diseases (Atherosclerosis, restenosis, and thromboembolism), brain and mental diseases (Depression, Migraine, Parkinson and Schizophrenia) and metabolic complications (Hypercholesterolemia, Obesity, Diabetes type 2 (DT2) and metabolic syndrome). B Differential Gene Expression of Genes represented by fold changes of all the CD genes, which show to be significantly different in at least one comparison. * denotes significance at q < 0.1 (holm p-value adjusted). Figures were mainly rendered in R software (https://cran.r-project.org/)

Gene sets from GO are divided into cellular components, biological processes, and molecular functions. Within biological processes terms, we observed significant enrichment in processes related to response to bacteria, positive regulation of nitric oxide (NO), ERK, and NFκβ, apoptosis, cell proliferation, response to lipopolysaccharide, inflammatory and immune response. For molecular function, cytokine activity, interleukine-1 receptor binding, receptor activity, and protein homodimerization activity were significantly enriched by the CD-risk genes analyzed. For cellular components, only membrane, plasma membrane, and extracellular region were significant. Overall the significantly enriched GO terms point to known CD terms such as the immune response, cytokine activity, and signaling receptors as the primary source of functional causes (in terms as Immune Response, Infection, Interleukin-1 receptor binding, Cytokine and defense response, Cytokine activity, Autoimmune Disease).

Additionally, among the enriched pathways identified were JAK-STAT signaling pathway, cytokine receptor interaction, NOD-like receptor signaling pathway, NF-kappa, TNF signaling, Toll-like receptor (TLR) signaling, T cell receptor signaling, and Osteoclast differentiation. These signaling pathways converge on the activation of NF-κB, a protein complex that controls the transcription of DNA cytokine production and cell survival [45]. In addition, the KEGG comorbidities identified are infectious diseases caused by bacteria, protozoa and virus, and IBD, which are reliable associations due to the relationship between microbes and CD [46]. The mapping is, therefore, an excellent guide to connect genes and important biological aspects of CD (Fig. 2).

Among the genes present in the enriched biological terms, we noted two distinct groups depending on the frequency of their presence in the gene sets and their number of abstracts found by PubTerm, designated as common and sporadic (Fig. 2). Briefly, the 34 common genes are highly related to diseases and biological terms and well-studied. In comparison, the 92 sporadic genes are associated with particular diseases or biological terms and not as studied in CD as the common genes. Among the common group, the most shared genes across concepts are NOD2, TNF, ICAM1 NFKBIA, NFKB1, TNFRSF1A, CD14, ACE TLR4, and TLR9, which are involved in both TNF and of NF-κβ signaling pathways [47,48,49]. Also, IFNG, IL1B, IL6, IL23R, IL10, IL4, IL12B, IL1RN, IL18, and IL4R, which are all well-known cytokines or related genes, and contribute to the inflammatory response and cytokine interaction process [50,51,52]. The sporadic group comprised 92 genes that were much less frequent among enriched terms, where ATG16L1, SLC22A4, IRGM, SLC22A5, TNFSF15, NOD1, PTPN2, PTPN22, and DLC5 being more frequently mentioned.

DNAH12, ERAP2, FUT2, ORMDL3 and, TRAIP were more specifically enriched in the CD disease term and less common for the remaining terms.

Once we noted two clear sets of common and sporadic genes that were associated with specific terms, we considered whether the genes might also be grouped by other phenotypes that could explain CD symptoms. Thus, we used the Gene Network tool, which clusters genes with similar Human Phenotype Ontology. Five sub-networks were identified from the 126 genes categorized as experimental evidence. We then merged three highly interconnected sub-networks and subsequently analyzed the three resultant modules (73, 26, and 26 genes respectively for Module 1 in green, Module 2 in purple, and Module 3 in blue, as shown in Fig. 3). Next, for each module, the genes were also functionally analyzed to identify their potential phenotypic consequences. For this, we also used the Gene Network tool. Only exclusive terms for each module that were significant after Bonferroni correction were analyzed (see Methods).

Network analysis for the three main groups identified. Group 1 (green): 73 genes, Group 2 (purple): 26 genes and Group 3 (blue): 27 genes. Figure adapted from Gene Network tool

Full size image

Network analysis for the three main groups identified. Group 1 (green): 73 genes, Group 2 (purple): 26 genes and Group 3 (blue): 27 genes. Figure adapted from Gene Network tool

For the first module (green), 60 phenotypes were retrieved; most of them related to severe symptoms such as Ocular complications, Altered immune system, Sepsis, Bowel incontinence, Heart complications, Endocrine abnormalities, Dysphagia and constipation, and muscle problems. For the second module (purple), 11 terms were identified mainly related to Neoplasm of the gastrointestinal tract, Hyperhidrosis, Respiratory complications, and Visual impairment. These symptoms are among common abnormalities detected in IBD patients [53,54,55]. Finally, for the third module (blue), only four consequences were identified, abnormality of glutamine metabolism, abnormality of the small intestine, and two remaining terms related to the facial skeleton. The information related to the module assigned to each gene is provided in Table S3.

Gene expression analysis

A recent gene expression analysis of CD, UC, and controls in the colon and ileum showed that 1008 genes were differentially expressed [56]. We, therefore, explored whether the genes found genetically associated with CD in our systematic review were related to those differentially expressed (DE) between CD or UC relative to their normal ileum and colon gene expression. From the 126 genes, 67 genes were found to be DE for both UC and CD (Fig. 2B). The overlap between the 67 genes with those 1008 is highly significant (p = 1^− 290, hypergeometric test), suggesting that our 126 genes are particularly enriched in DE genes. Except for SLC39A8, FAS, IRF5, HSPA1L, and PTEN, the vast majority were indeed more significantly associated with UC than with CD. Moreover, the majority of the genes were less expressed in CD relative to controls. We noted two solute carriers more expressed in CD than in normal colon or ileum; SLC22A4 (ergothioneine, carnitine, tetraethylammonium) probably for detoxification, and SLC22A5 (carnitine), whose variants are reported to affect the function of carnitine and organic cation transporters [57]. There are also two less expressed solute carriers, SLC11A1, with a role in the susceptibility of humans and animals to several infections [58] and SLC39A8, associated with gut microbiome composition [59].

Drug-gene interaction analysis

The main therapeutic drugs for CD are azathioprine [60], infliximab, adalimumab, certolizumab pegol, ustekinumab, vedolizumab [61,62,63], prednisone, hydrocortisone, and hydrocortisone acetate [61]. To explore drug-gene interactions, an analysis was performed in DGIdb [41] using these drugs. There is a total of 78 genes that had a reported interaction with these CD therapeutic drugs, 10 of which were identified in our systematic review (Table 4). We reasoned that focusing on drugs that target the gene variants associated with IBD could be a strategy for CD drug repurposing [64]. Thus, to search for alternative drugs for CD, we used 13 other drugs commonly employed in the treatment of chronic autoimmune and inflammatory diseases [64, 65]. We identified 13 CD genes (IL1B, IL1RN, IL6, ABCB1, XIAP, IFNG, ICAM1, NLRP3, JAK2, PPARG, PTGS2, APOE, and SMAD3) that show some interaction and therefore could be further explored as possible treatments for CD (Table 4).

To compare the generated list of genes with genetic panels already in use, we benchmarked within those panels in the GTR [42]. We found 21 panels, of which 19 were specific to Crohn’s disease containing only 2 genes, NOD2 and IL6. The two remaining tests were not specific for Crohn’s, IBD, and related diseases. These tests considered 70 genes, of which 22 were identified as functional variants for CD in our curation (including NOD2 and IL6). Thus, from the 256 genes we found (Tables 3 and S3), 225 genes were not included in any panels for CD or IBD-related disorders.

We also verified the identifications at the Open Targets (OT) platform, whose pipeline includes a fine mapping of variants [31]. Filtering 3093 genes for CD for genetic association score > 0.8, 178 genes were identified. Of these, 3 genes (SNN, SH2B3, and SKAP2) were not identified in the set of 1092 curated genes. From the 126 genes identified here having experimental evidence, 39 showed a low genetic association score for CD (0.5 to 0.79), 8 genes showed a good score (0.8 to 1.0) for IBD, and 52 genes do not have genetic data information in OT for CD nor IBD. Some of the 52 genes show variations that have not been reported in GWAS, explaining their absence in OT. Examples include NOD1, ABCB1, IL1RN, MEFV, and IL18 having variants that could not be easily identified in GWAS because there are triallelic changes [66], deletions [67], and VNTR [68]. This comparison shows that even GWAS information can leave aside some information of other variants detected through other technologies or methodologies.