Various studies implicated the serotonin system and serotonin type 3 receptor genes, in particular, in the pathophysiology of IBS. We aimed to validate these findings in a large multi-center endeavor. However, the only SNP holding true to being associated with IBS was the HTR3E SNP c.*76G > A (rs56109847 = rs62625044) in females with IBS-D.

The lack of replication of the SNPs in HTR3A, HTR3B, and HTR3C with IBS may be attributed to an initial association with comorbid conditions like psychiatric disorders including depression and anxiety as well as pain disorders including migraine and fibromyalgia. Only recently, IBS was termed as gut-brain disorder and its comorbid conditions increasingly recognized. Accumulating evidence suggested an involvement of disturbed brain-gut axis communication. In fact, HTR3A rs1062613 had initially been associated with conditions seen in IBS-like depression and anxiety [27, 28], and had also been correlated with the severity of IBS symptoms and anxiety. It has also been linked to amygdaloid activity [29, 30], early life trauma and altered emotional networks in the human brain as well as with the onset of depression [31]. More recently, this SNP - in interaction with childhood trauma - was shown to differentially modulate central serotonin activity [32]. Lately, this SNP has been reported to contribute to pain intensity in temporomandibular disorder myalgia [33]. Furthermore, the HTR3B variant p.Tyr129Ser (rs1176744) has also been associated with bipolar affective disorder and female major depression as well as pain catastrophizing, a coping style characterized by excessively negative thoughts and emotions related to pain [22, 34,35,36]. Moreover, it has been associated with IBS, increased anxiety and alexithymia (a personality trait characterized by the inability to identify and describe emotions) [24]. All HTR3 SNPs have been proven 317 to be functional in earlier studies [12].

Interestingly, rs1062613 in HTR3A has also been associated with dyspepsia [37] and hypersensitivity in gastroesophageal reflux disease, possibly due to reduced 5-HT3 receptor activity in the descending serotonergic pathway [38]. Furthermore, in a functional imaging study of responses to rectal balloon distension, risk allele carriers presented with significantly more activation in the right amygdala, left insula, and left orbitofrontal cortex, suggesting that individuals carrying HTR3 polymorphisms may respond differently to gut-derived signals in brain regions of negative emotion, body recognition, and that discrimination of the stimulus that might be enhanced in 5-HT3 receptor signaling [24].

In summary, disturbed peripheral and central 5-HT-mediated signaling seems to shape the phenotypes of these complex conditions. In line with this, functional MRI studies have confirmed that 5-HT3 receptors are important for the neural brain networks involved in emotional processing, learning, cognition, visceral perception, and pain processing. HTR3 variants have been linked to functional and structural differences in brain regions relevant to these traits such as the amygdala, frontal cortex, temporal lobe, and insula [25, 30]. Therefore, an individual make-up of 5-HT3 receptors may specifically modulate relevant neural circuits, thereby making individuals prone to develop one or another disorder. The association findings of HTR3 variants with various complex disorders may additionally be attributed to a common genetic basis of neurodevelopmental disorders. 5-HT3 receptors are involved in shaping neuronal structure and plasticity in distinct brain regions during development and maturation [39]. HTR3-associated disorders may share neurodevelopmental and functional CNS and/or ENS disturbances, which may be altered by HTR3 variants. Modified responses may be distorted by disturbed 5-HT3 receptor signalling.

To understand the underlying molecular pathomechanisms and to allow the dissection of the respective disturbances that predispose to particular traits, deep phenotyping of patients and control individuals is mandatory.

In our qPCR analyses, none of the HTR3 genes except HTR3E was found to be robustly expressed in any of the tested GI tissues from the small and large intestine. In contrast to our data, HTR3A was found to be robustly expressed in various brain and GI tissues in the GTex data set (Supplementary SD Fig. 4, access 16.12. 2020). This discrepancy can currently not yet be explained; however, but might be biased by inflammation in a given tissue. According to single cell data in the protein atlas, the cellular expression within the colon differs remarkably [41]. Whereas HTR3E was mainly found to be expressed in enteroendocrine and Paneth cells as well as in two different types of enterocytes (https://www.proteinatlas.org/ENSG00000186038-HTR3E/celltype/colon), the largest extent of HTR3A expression was mainly found in B cells of the colon, but only marginally in T cells as well as in another type of enterocytes as compared to HTR3E (https://www.proteinatlas.org/ENSG00000166736-HTR3A/celltype/colon).

The subordinate role of HTR3B and HTR3C within the GI tract is furthermore underlined by expression data from the GTEx portal (https://gtexportal.org/home/). This data shows the highest expression in the small and large intestines (terminal ileum data from n = 187 donors, median TPM (transcripts per million) = 0.4364, colon transverse data from n = 406 donors, median TPM (transcripts per million) = 0.4273). This is in line with earlier studies indicating that HTR3E is restricted to the GI tract, colon and ileum, respectively [42] (Supplementary SD Fig. 5, GTex access 16.12. 2020).

The expression in other tissues was even lower (e.g., brain cortex, n = 255 donors, median TPM = 0.019, all others: median TPM below 0.000).

Variants in HTR3 genes might also alter the drug response. For example, in a recent pharmacogenetic study, stool consistency response to treatment with the 5-HT3 receptor antagonist ondansetron correlated with the CC genotype of the SNP p.N163K rs6766410 of the HTR3C gene [43] in IBS-D patients. To what extent HTR3 SNPs might further influence drug response and therefore be relevant to novel pharmacogenetic treatments, will be the subject to future studies.

We are well aware that our study has some limitations. Methodological variations exist since we combined data from 14 centers from different countries with some samples of rather small size. However, to address the impact of the study size on the outcome of the meta-analysis, funnel plots were generated. Funnel plots of estimated log-ORs against their precision (1 divided by the standard error of the log-ORs) for each study were fairly symmetric around the overall effect estimate, i.e., no publication bias was evident (see Supplementary SD Fig. 6).

The number of tested SNPs was limited, and therefore we could not correct our data for population stratification using genetic principal component analysis. To mitigate potential bias, we restricted our analysis to individuals of European ancestry only. Of note, the replicated association of the HTR3E SNP seems to also hold true for the Asian population since four earlier studies confirmed this finding in case control studies from China [18,19,20,21].

How far our association finding may apply to other populations remains to be evaluated in future studies. Remarkably, some SNP genotype frequencies for HTR3B and HTR3C significantly deviated from the HWE expectation in patient subgroups from the German, Greek, US, and UK cohorts—a phenomenon which we cannot explain at present (see Supplementary SD Table 6). In addition, the HTR3A SNP data deviated from HWE in controls from Greece. HTR3E SNP data were in accordance with HWE for all cohorts but one US patient sample (IBS overall) and one German control sample (female controls). However, the estimated effect size and direction for the HTR3E SNP are in line with the overall results.

Another limitation relates to the definition of the IBS phenotype. We used Rome II and/or Rome III criteria and therefore could not refer to a uniform symptom classification. We are aware that, for example, the Rome III criteria [44] allow a wider diagnostic range of IBS, especially for IBS subtypes. To explore differences between ROME II- and ROME III-based diagnoses, a subgroup analysis was performed. Stratified analyses according to Rome II or Rome III criteria generally revealed stronger associations for Rome III diagnoses: IBS overall OR = 1.10, 95%CI (0.86–1.40) for Rome II compared to OR = 2.45, 95%CI (1.79–3.34) for Rome III; IBS-D OR = 1.31, 95%CI (0.96–1.80) for Rome II compared to OR = 2.23, 95%CI (1.51–3.27) for Rome III; IBS-D females OR = 1.57, 95%CI (1.07–2.29) for Rome II compared to OR = 1.94, 95%CI (1.16–3.22) for Rome III.

We are also facing a sex bias, since more females than males were included. Yet, this phenomenon might just reflect the fact that the prevalence of IBS is higher in women.

Moreover, we cannot exclude that IBS patients are included in the control samples since GI symptomatology that might qualify as IBS was not rigorously excluded in the control population. For genetic studies, a purely symptom-based IBS classification is only of limited specificity when it comes the identification of mechanistically diverse phenotypes of IBS or its subgroups [11]. Consequently, the assessment of additional symptoms as well as intermediate or quantitative traits, are all warranted to dissect the genetics underlying IBS subtypes and to correlate these to symptoms/biomarkers, applying a standardized and robust deep phenotypic tool [40]. Accounting for variations in the GI phenotype and also factoring the impact of common comorbidities and psychiatric phenotypes, personality traits, and somatization, in particular, should be a mandatory feature of future studies [40]. Of course, control individuals should be appropriately characterized to exclude IBS sufferers. Most of our samples came from tertiary referral centers, and our findings may therefore not hold true for all IBS patients.

The small number of samples in the differential expression analysis was another limitation. However, a strength of our study is the fact that we did not limit our study to only one GI region but instead, we included samples from the small and large intestines, whereas previous studies were limited to only one or two regions of the GI tract [45]. In addition, combined genotyping and expression analysis from the same individual was not possible due to the current design and should thus be a target in future studies.

In conclusion, meta-analysis confirmed the role of the HTR3E SNP rs56109847 = rs62625044 in females with IBS-D. Expression analysis revealed reduced HTR3E levels in the sigmoid colon of IBS-D patients. This underlines the relevance of HTR3E in the pathogenesis of IBS-D. To what extent this and the other non-replicated SNPs shape the GI phenotype in IBS and how they interfere with other behavior- and pain-related genetic variants and impact on communications via the gut-brain axis, remains currently unknown. Future studies are warranted to assess how these variants correlate with behavior, pain perception, and bowel habits and how far gene–gene and gene–environment interactions affect the individual susceptibility to chronic GI disorders. Our currently ongoing studies within the international H2020 consortium DISCOvERIE (Development, dIagnosis and prevention of gender-related Somatic and mental COmorbiditiEs in iRritable Bowel Syndrome In Europe, www.DISCOvERIE.eu) that implemented deep phenotyping guidelines [40] will allow us to get assess how HTR3 variants relate to central and peripheral phenotypes.

A better understanding of the contribution of HTR3 SNPs in shaping different phenotypes relevant to IBS and its comorbidities is of clinical importance and may allow insights into IBS pathophysiology and refine the targeting of 5-HT3 receptors in therapeutic interventions.