Considering the anticipated exacerbation of the global prevalence of IBD, the imperative for the discovery of non-invasive biomarkers and new therapeutic targets has intensified. To this end, a meta-analysis of blood specimens was employed to discover potential biomarkers of IBD. Datasets GSE119600, GSE94648, GSE86434, and GSE717304 were merged since they had a higher number of genes among the retrieved datasets, allowing for a more comprehensive analysis of the transcript variations present in the samples. Differential expression analysis of the integrated data, following batch-effect correction, led to the identification of 357 genes whose expression alterations had an identical direction both in CD and UC, which were subsequently defined as DEGs in IBD.
By reflecting the pathways activated following the immune response, functional enrichment analysis implied the biological relevance of the DEGs in the pathogenesis of IBD. It is worth noting that the CC results of GO analysis markedly underscored granules and vesicles. The terms including 'specific granules,' 'secretory granules,' 'tertiary granules,' and 'ficolin-1-rich granules' explicitly refer to neutrophil function. The spectrum of terms reflects various aspects of their ability to recognize pathogens and release immune mediators, such as lactoferrin, lysozyme, and elastase, in response to pathogen invasion. Recent investigations have unveiled the significance of neutrophils in the pathogenesis of IBD. Multiple susceptibility genes associated with IBD are involved in orchestrating neutrophil responses against pathogens. Furthermore, genetic defects that disrupt neutrophil functions lead to the formation of IBD-like intestinal conditions manifested by inflammation and changes in gut microbiota [14]. Lactoferrin, which is primarily secreted by mature neutrophils and highlighted by the specific granule (GO:0042581), represents intestinal inflammation and has been suggested as a biomarker in stool specimens [15]. Considering lactoferrin's anti-inflammatory, immunomodulatory, and barrier-protective properties, its administration has been suggested as a therapeutic strategy [16].
The results from KEGG pathway analysis also highlighted the enrichment of neutrophil extracellular trap (NET) formation. Neutrophils, as pioneer immune cells involved in orchestrating immune responses against pathogens, release NETs and play a dual role in gut inflammation. On the one hand, they release proteases such as elastase and matrix metalloproteases (MMP), which directly damage the epithelial barrier in the gut, leading to amplification of inflammation. On the other hand, NETs contribute to the clearance of damage-associated molecular patterns (DAMPs) in the peripheral blood, resulting in an attenuated inflammatory response. Multiple investigations have shown that NETs exhibit heightened abundance in the inflamed mucosal tissue, fecal matter, and peripheral blood of IBD patients, particularly during periods of active disease [17]. To conclude, the results from the functional analysis, in line with recent advancements in the field, underscored the contribution of neutrophils to orchestrating immune responses in IBD.
Of the 357 DEGs in IBD, a PPI network encompassing known interactions of 277 nodes was extracted using a confidence threshold of 0.4. By interacting with 87 nodes, TNF was the most prominent hub gene in the PPI network of IBD patients’ blood. The results of the present meta-analysis, in contrast to previous studies [18], demonstrated a significant downregulation of the TNF gene. Separate differential expression analysis on each retrieved dataset indicated significant downregulation of TNF in the GSE119600, GSE71730, GSE169568, and GSE112057 datasets, while the alterations of this gene were not statistically significant in the remaining datasets (Supplementary Fig. S1). This discrepancy may be attributed to a variety of factors, such as potential biases in gene expression resulting either from small sample sizes or unaccounted pretreatment conditions. Nevertheless, TNF is a central cytokine involved in orchestrating gut inflammation in IBD and has long been considered a major therapeutic target [19]. Although inhibition of TNF-α has been widely reported to be a beneficial treatment, recent paradoxical reports of its drawbacks make its systemic inhibition challenging [20, 21]. Recent cohort studies have revealed that IBD patients receiving anti-TNF therapy have an elevated risk of developing psoriasis, rheumatoid arthritis (RA), and hidradenitis suppurativa [22]. Given that the TNF level is increased in the intestine of patients with IBD and correlated with the severity of the disease, recent investigations have suggested localized inhibition of TNF in the intestine rather than systemic treatment to avoid the adverse effects of systemic administration [23].
The exploration of regulatory miRNAs targeting hub genes of the PPI network underscored four families of miRNA, namely, miR-548, miR-181, miR-92, and miR-513 families. Among these, the four members of the miR-181 family, including miR-181a-5p, -181b-5p, -181c-5p, and − 181d-5p, as well as miR-150-5p, were found to target TNF. In both mice and humans, the miR-181 family demonstrates an abnormal expression pattern following intestinal inflammation. By conducting qPCR, a significant downregulation was reported for the levels of miR-181a and miR-181b in the colitis-induced mouse model, whereas in UC patients, the downregulation was observed only for MIR-181A. The therapeutic potential of miR-181 was reflected by the observation that miR-181 is required for the restoration of the intestinal epithelial barrier, which was attributed to its regulatory effect on the Wnt signaling pathway. According to the miRNA-mRNA network in this study, it can also be concluded that the beneficial effect of miR-181 could partially stem from its inhibitory effect on TNF, which is abundant in the inflamed intestinal mucosal layer [24]. A study on serum levels of miRNAs in patients with UC detected miR-150 as the only downregulated miRNA, which was reported by qPCR [25]. In addition to targeting TNF-a, miR-150-5p is involved in the immunopathology network of diseases by targeting IL-6 and IL-1b. Furthermore, by targeting c-Myb, it has been reported to be positively correlated with the induction of apoptosis in intestinal epithelial cells. Although there is inconsistency in the previous reports about whether miR-150-5p is upregulated or downregulated in the colonic specimens, its significant dysregulation in IBD has been documented by multiple studies [26].
There have also been reports of an association of the three remaining families with IBD. Among the members of the miR-92 family, qPCR analysis demonstrated that miR-92a has an elevated expression level in the blood of UC patients compared to healthy individuals. Additionally, ROC analysis revealed an AUC above 0.9 for this miRNA, reflecting its potential as a biomarker [27]. In the miR-513 family, miR-513a-3p was detected as a target of circRNA_102610, an upregulated circRNA in PBMCs of patients with CD [28], according to the miRanda and TargetScan databases [29], which reflect its potential involvement in the pathogenesis of IBD.
According to the extant corpus of literature, there is an absence of reports regarding the IBD-associated dysregulation of the miR-548 family members identified in this investigation [26, 30, 31]. This particular family of microRNAs was highlighted by the miRNA-mRNA regulatory network in this study due to its regulatory effect on CD163 and KLRD1. Examination of the expression of CD163, a marker of monocytes and macrophages, in biopsies obtained from the colon by qPCR showed its upregulation in treatment-naive patients compared to healthy controls [32]. The association of KLRD1, which is primarily expressed in NK cells [33], with IBD has yet to be directly reported.
Seven modules were singled out from the results of WGCNA, meeting the criterion of p-value < 0.01. Among these, modules 10 and 11 were substantially involved in viral and bacterial infections. Previous studies have reported an elevated susceptibility for developing Epstein-Barr virus (EBV) [34], influenza [35], Salmonella [36], and tuberculosis [37] in patients with IBD, reflecting shared potential pathogenic mechanisms. Reports, however, demonstrated no increased risk of hepatitis C virus (HCV) [38] or Covid-19 [39] in IBD patients.
Module 12, which exhibited enrichment and a negative correlation with IBD, was further analyzed to identify the key genes within this module. While 11 genes, namely CD8A, IFNG, PRF1, KLRD1, CD247, CD2, EOMES, GZMA, IL2RB, GNLY, and KLRB1, were detected as highly connected nodes in the PPI network of this module, only PRF1 was selected as the hub regarding module membership and gene significance with IBD. Of the 11 genes identified as hub genes in the PPI network of module 12 in this study, 5 genes, including CD8A, CD2, IL2RB, PRF1, and GNLY, were previously reported as hub genes in the shared PPI network of UC and non-alcoholic fatty liver disease (NASH) [40]. In a study on gene expression patterns of circulating CD4 + T cells of patients with CD, PRF1 was identified as a hub gene in a gene module particularly activated in the penetrating subtype of CD, a severe form of the disease [41]. Moreover, differential methylation analysis detected altered methylation of the PRF1 in the blood of CD patients compared to healthy individuals [42].
Of the 357 DEGs in IBD, 298 genes with consistent correlation and expression direction were analyzed using RF and LASSO machine learning methods. FEZ1 and NLRC5 were identified by both analyses and demonstrated acceptable diagnostic efficacy in integrated data. However, the diagnostic potential of NLRC5 failed to meet the AUC > 0.7 threshold in two of the three validation datasets that included data on the gene. Meanwhile, FEZ1 met the threshold in six of the eight analyzed datasets.
FEZ1, a kinesin adaptor with multiple biological functions, has been widely studied in the context of its role in neurons [43]. For instance, FEZ1 is a well-established susceptibility gene for schizophrenia (SCZ) [44], and its downregulation in the peripheral blood of patients with the disease has been reported by qPCR analysis [45]. Given that the results of the present study highlighted FEZ1 as a key gene in IBD, and in light of its contribution to brain abnormalities, including attention deficit hyperactivity disorder (ADHD), autism spectrum disorders (ASD), and SCZ [46], it can be hypothesized that FEZ1 serves as a pathogenic link in the gut-brain axis. The correlation between IBD and the risk of developing SCZ remains controversial, as some studies have reported an increased risk of SCZ in patients with IBD, whereas others found no significant correlation [47]. However, a population-based cohort study observed an elevated incidence rate of IBD in SCZ patients [48]. Similar to the downregulation of FEZ1 observed in this study, FEZ1 has also been documented as a downregulated gene in the blood of patients SCZ [45, 49] and bipolar disorder (BD) [50]. To conclude, although the pathogenic role of FEZ1 in IBD has yet to be elucidated, its identification as an IBD-specific gene coupled with the consistent records of its downregulation patterns in SCZ and BD suggests that FEZ1 is involved in the gut-brain axis.
Emerging investigations have indicated that the function of FEZ1 is not limited to neurons. FEZ1 is involved in the pathways of innate immunity by modulating the translocation of heat shock protein 8 (HSPA8) to the nucleus and the expression of interferon-stimulated genes [43]. Furthermore, through direct interactions with NBR1, SMC3, IMMT, RIF1, and RAB3GAP2, FEZ1 is involved in autophagy networks implicated in CD [51]. Dysregulated autophagy is acknowledged as a pathogenic mechanism in various inflammatory autoimmune disorders, including RA [52], psoriasis [53], multiple sclerosis [54], and IBD [55]. Particularly in IBD, autophagy contributes to the clearance of pathogens, release of granules, inflammasome functioning, expression of pro-inflammatory cytokines, and endoplasmic reticulum stress [56]. FEZ1 was shown to be upregulated in the blood of CD risk allele carriers (rs13361189) compared to non-carriers, suggesting a potential role for FEZ1 in the pathophysiology associated with this genetic variant [57].
Previous studies have revealed significant blood alterations of FEZ1 in patients diagnosed with other autoimmune disorders, such as systemic lupus erythematosus (SLE) and type 1 diabetes (T1D). In a study on circulating T cells of patients with SLE, the analysis of RNA-seq data by both generalized linear model (GLM) and support vector machine (SVM) methods identified FEZ1 as a prominent gene, with an AUC exceeding 0.9 in ROC analysis [58]. Microarray analysis of PBMCs of patients with T1D revealed that FEZ1 is downregulated, with a log fold change (logFC) of -1.13 and an adjusted p-value of below 0.001 [59].
NLRC5 is a unique member of the NLRC subfamily of the NOD-like receptor (NLR) family, characterized by a caspase recruitment domain (CARD) without an acidic domain [60]. NLRC5 regulates the transcription of genes in T cells and contributes to the T cell response, primarily through complexes that lead to PANoptosis, involving mediators of pyroptosis, apoptosis, and necroptosis. This process occurs following heme exposure during inflammatory conditions [61]. Given that intestinal bleeding increases the heme levels in the mucosal layer [62], it can be assumed that the NLRC5-mediated response in T cells could be further amplified [61], underscoring the role of NLRC5 in IBD pathogenesis. In a study on intestinal mucosa, differential methylation analysis identified NLRC5 as a hypomethylated gene in patients with IBD compared to healthy controls [63].
In summary, this study aimed to identify blood RNA markers with diagnostic potential in patients with IBD. FEZ1 exhibited acceptable diagnostic capability through the integrated transcriptomics approach applied. However, the findings of this study are based on computational analyses, and their clinical applicability has yet to be established. This study was limited by the lack of comprehensive biological information from patients, which hindered the accurate matching of case and control groups. Confounding factors, such as smoking history, gender, or age, may have implicitly influenced the results. Nevertheless, according to the phenotypic data retrieved by the GEOquery package, an initial phase of filtration aimed at minimizing disparities between groups was executed to ensure that blood constituted the sample origin and that no prior treatments were documented.