Abnormalities of mucosal serotonin metabolism and 5‐HT3 receptor subunit 3C polymorphism in irritable bowel syndrome with diarrhoea predict responsiveness to ondansetron

Summary Background Irritable bowel syndrome with diarrhoea (IBS‐D) is a common condition, greatly reducing the quality of life with few effective treatment options available. Aim To report the beneficial response shown in our trial with the 5‐hydroyxtryptamine (5‐HT) receptor 3 antagonist, ondansetron in IBS‐D Methods A randomised, placebo‐controlled, cross‐over trial of 5 weeks of ondansetron versus placebo in 125 patients meeting modified Rome III criteria for IBS‐D as previously described. Patients were compared to 21 healthy controls. 5‐HT and 5‐HIAA were measured in rectal biopsies. Whole gut transit time was assessed using a radio‐opaque marker technique. Whole blood DNA was genotyped for an insertion polymorphism in the promoter region of the serotonin transporter gene SLC6A4, as well as single nucleotide polymorphisms (SNPs) of the tryptophan hydroxylase gene TPH1 and 5‐HT3 receptor genes HTR3A, C and E. Results Patients’ biopsies showed significantly higher 5‐HIAA levels (2.1 (1.2‐4.2) pmol/mg protein vs 1.1 (0.4‐1.5) in controls, P < .0001). 39 patients used < 4 mg/d (“super‐responders”) while 55 required ≥ 4 mg/d. 5‐HT concentrations in rectal biopsies were significantly lower in super‐responders (21.3 (17.0‐31.8) vs 37.7 (21.4‐61.4), P = .0357) and the increase in transit time on ondansetron was significantly greater (15.6 (1.8‐31) hours vs 3.9 (−5.1‐17.9) hours). Stool consistency responders were more likely to carry the CC genotype of the SNP p.N163K rs6766410 of the HTR3C gene (33% vs 14%, P = .0066). Conclusion IBS‐D patients have significant abnormalities in mucosal 5‐HT metabolism. Those with the lowest concentration of 5‐HT in rectal biopsies showed the greatest responsiveness to ondansetron.


| INTRODUC TI ON
When patients with irritable bowel syndrome with diarrhoea (IBS-D) are asked to report their most bothersome symptoms, erratic bowel habit and urgency are the most common. 1 Urgency and the associated occasional faecal incontinence cause significant distress and anxiety, which substantially impair their quality of life. Serotonin (5-hydroxytryptamine (5-HT)), a neurotransmitter in the gut has many effects which are relevant to this aspect of IBS, stimulating intestinal secretion and colonic motility. 2 5-HT 3 (5-HT 3 ) receptor antagonists including alosetron, cilansetron and ramosetron which block these effects and slow whole gut transit have been shown in meta-analysis to be effective treatments for IBS-D. 3 Alosetron was withdrawn because of adverse events including severe constipation and ischaemic colitis. 4 It has now been reintroduced under a FDA-managed risk evaluation and mitigation strategy but is not widely used. Ramosetron, a potent highly specific 5-HT 3 receptor antagonist has been shown to be similarly effective and recently another 5-HT 3 receptor antagonist, ondansetron, has also been shown to be effective without causing ischaemic colitis.
Our previous studies which focused on post-infectious IBS (PI-IBS), two-thirds of whom have IBS-D, had shown an increased number of 5-HT containing enteroendocrine cells. 5 We also showed in animal models of PI-IBS 6 that colonic mucosal 5-HT was elevated immediately after infection and 5-hydroxyindole acetic acid (5-HIAA)/5-HT ratios were increased while, serotonin transporter gene (SERT gene SLC6A4) mRNA expression was depressed up to 56 days post-infection, these two changes suggesting long-lasting accelerated mucosal 5-HT turnover. This was associated with enhanced firing of afferent neurons during colonic distension, a feature which could be blocked by ondansetron. This led us to hypothesise that increased 5-HT availability in the intestinal mucosa could be the driver of symptoms and the explanation of the benefit of 5-HT 3 receptor antagonists in IBS-D.
There are marked individual differences in responsiveness to 5-HT 3 receptor antagonists which have been correlated with common polymorphisms in key genes governing the synthesis and reuptake of 5-HT, as well as the structure of the 5-HT receptors. Responsiveness to alosetron was shown in one trial to be greater with the homozygous l/l variant of the 5-HTTLPR (serotonin-transporter-linked polymorphic region) of SLC6A4. 7 Furthermore, the gene TPH1 encoding tryptophan hydroxylase 1, the rate limiting enzyme for serotonin synthesis in enterochromaffin cells of the gut, contains several single nucleotide polymorphisms (SNPs), including rs4537731 and rs211105, which have been reported to predict responsiveness to ramosetron. 8 Finally, the SNP of HTR3C p.N163K rs6766410 predicts chemotherapy-induced vomiting, 9 which is known to be driven largely by serotonin. This led us to hypothesize that the sensitivity to ondansetron might in part be dependent on genetic variability due to polymorphisms in these genes.
We now present the biomarkers of mucosal 5-HT metabolism along with the genetic markers obtained from patients participating in our double-blind cross-over trial whose clinical results have been previously reported, 10 aiming to further characterise IBS-D patients who respond to ondansetron therapy.

| Design
The design was as previously described. 10 In brief, 125 patients with IBS-D meeting Rome III criteria 11 were recruited from gastroenterology clinics in Nottingham and Manchester and randomised to receive either 5 weeks of ondansetron, then 5 weeks of placebo or placebo followed by ondansetron, with at least 2 weeks washout between the two treatment periods. Doses were titrated for the first 3 weeks of each treatment period, maintaining a constant dose for the final 2 weeks. Twenty-one healthy volunteers were recruited as controls for comparison. They completed the same questionnaires as IBS-D patients, underwent rectal mucosal biopsy and provided normal values for colonic transit studies.

| Data collection
The following personal baseline data were collected: age, sex, Hospital Anxiety and Depression Scale (HADS) 12

| Biopsy analysis
All 70 patients recruited from Nottingham were invited to undergo unprepared flexible sigmoidoscopy to obtain high rectal/sigmoid biopsies and 57 consented to this additional procedure. Two of the samples were immediately frozen in liquid nitrogen for analysis of serotonin and 5-HIAA content by high-performance liquid chromatography (HPLC), two in RNALater for gene expression studies, one for incubation at 25°C in 95% O 2 for assay of 5-HT release from a 0-30 minute period as previously described 20 (Data S1). Finally, one biopsy was embedded in paraffin for routine histology to exclude microscopic colitis. Real-time polymerase chain reaction (PCR) was used to assess the relative expression of TPH1 with respect to the housekeeping gene hypoxanthine phosphoribosyltransferase 1 (HPRT1). RNA was extracted and analysed according to the previously reported method. 21

| Plasma 5-HIAA
We assessed fasting plasma 5-HIAA using HPLC as previously described. 22  About 10% of the samples were repeated to ensure genotyping accuracy.

| Responder definition
Our original study was a pilot study, so we used stool form rather than pain as our primary endpoint since it was known to have had a much lower placebo response rate. Therefore, in this paper, we used the US Food and Drug Administration (FDA) definition of a "stool consistency responder" as a "patient who experiences a 50 percent or greater reduction in the number of days per week with at least one stool that has a consistency of Type 6 or 7 compared with baseline". 24 Our pain assessments on a 0-3 scale did not allow us to calculate a pain responder rate according to FDA guidelines which were published after our study was initiated.

| Data analysis
Analysis was performed using Graphpad Prism version 7.0c (GraphPad Software, La Jolla California, USA). Efficacy parameters were calculated for each patient as the difference in the endpoints measured in the last 2 weeks of the ondansetron and placebo periods.
Baseline and rectal biopsy data were analysed for differences between patients and healthy volunteers using unpaired t-tests and Mann-Whitney tests on parametric and nonparametric data, respectively. Subsequent analysis based on their stool form responder status and mean ondansetron dose was done as below.
Genotype data was correlated with stool form responder status, baseline clinical features, biopsy results, final ondansetron dose, whole gut transit time and TPH1 mRNA expression. Chi-squared tests were performed for stool form responder status, and one-way ANOVA and Kruskal-Wallis tests were performed on parametric and nonparametric data respectively.

| Participants
Healthy volunteers were age and sex matched to patients. Mean patient and control ages were 41 and 43 years and percentage female were 71% and 76% respectively. The patients taking part in the trial were, as expected, more anxious, more stressed and showed greater somatic symptoms with significantly higher HAD, perceived stress scale and PHQ-12 scores compared to controls (Table 1). There were no differences between patients consented for biopsy and those Note: Data are mean (SD), unless stated. P values, were significant, demonstrate differences between the patient population and healthy volunteers with no difference between the two patient groups. P values are obtained from one-way ANOVA and Kruskal-Wallis tests for parametric and nonparametric data respectively. who did not. As required for entry into the trial 10 IBS-D patients reported significantly higher bowel frequency.

| Rectal biopsies
Rectal biopsies were obtained from 57 of the 125 patients completing the trial and 21 controls. Biopsy 5-HIAA levels were significantly higher in IBS-D patients ( Figure 1) as were the 5-HIAA /5-HT ratios (Table 1). There were, however, no significant differences between patients and controls for biopsy 5-HT nor 5-HT release in the first 30 minutes (see Table 2).

| Stool form responders
107/125 IBS-D patients completed daily stool diaries to allow assessment of response, of which 82 patients met FDA criteria for stool form responders. As shown in Table 3, they had significantly lower baseline average abdominal pain scores and pain occurred on fewer days per week. They also reported lower average urgency scores and urgency on fewer days per week. Responders tended to be younger, with lower HAD scores and to have slower colonic transit; however, these differences failed to reach conventional statistical significance. There were no significant differences in either baseline stool form or frequency (Data S1).

| Features of patients showing increased responsiveness to ondansetron (super-responders)
About

| Effect of genotype on treatment response
Stool form responders were more likely to carry the homozygous CC genotype p.N163K rs6766410 of the HTR3C gene which was found in 33% of responders compared to 14% of nonresponders, (P = .0066, Table 5). The polymorphism in HTR3E c.*76G > A rs561098476 showed a trend towards association with more responders carrying the GG genotype compared to nonresponders, although this was  (Table S1). Surprisingly no significant effect was seen for the two TPH1 SNPs on TPH1 mRNA levels (Data S1).
When the 5-HT released from the biopsies over 30 minutes was expressed as a percentage of the biopsy 5-HT concentration, there was a trend for this to be elevated but this was not statistically significant owing to wide variability in this measure. Comparing Tables 2 and 4 it can be seen that healthy volunteers' mucosal 5-HT values were similar to the group requiring ≥ 4 mg. This suggests that it is the super-responders who have abnormally low mucosal 5-HT though the other 5-HT parameters were similar to both healthy controls and the other patients.
The super-responders showed a four-fold greater change in transit even although they were taking a dose which was a quarter of that observed in those requiring larger doses, again objectively supporting the idea that super-responders are very much more sensitive to the drug. The striking slowing of transit seen in super-responders has important implications. We have previously reported from this same patient group that the faecal protease levels correlated negatively with transit time and positively with average urgency in IBS-D patients. 31 As others have confirmed, urgency is strongly related to fast transit in IBS-D 32 but just exactly why slowing transit helps is uncertain. We hypothesise that, by allowing more time for the colonic microbiota to deconjugate bile acids and degrade endogenous proteases, 31 the slowing of transit may prevent the sensitisation of the rectum that these endogenous irritants can cause. However, further intervention trials using other nonserotonergic agents to slow transit, such as loperamide, will be needed to decide what is unique about the 5-HT 3 receptor antagonist's action.
Our finding that stool form responders were more likely to carry the CC genotype p.N163K rs6766410 of HTR3C fits with the data of Fasching et al 9 which suggested CC was more sensitive to chemotherapy-induced vomiting, known to be driven largely by serotonin. This substitution has a predicted possible functional effect on the receptor (polyphen2 33 score 0.798). If those with CC genotype are more sensitive to the effects of 5-HT this might help explain the benefit of a drug which blocks its action.
Like many mechanistic studies which make substantial demands on patients we were probably underpowered for many of our secondary endpoints. Since our study was carried out on patients referred to secondary care they can only be generalised to this population which may differ from that seen in primary care. We only requested biopsies for patients recruited in Nottingham where the laboratory facilities and staff needed were available and we allowed patients to choose whether to have the extra biopsies. We felt that this was important to facilitate recruitment to the main trial and as Table 1 shows, we did not find any differences between patients who opted to have biopsies compared to those who did not, so we do not think this introduced any bias.
Although we chose to study variants in the HTR3C gene with previous evidence for potential influence on response to 5-HT, these findings need to be interpreted with caution as none of the genetic analyses were corrected for multiple testing and being post hoc need to be replicated before they can be accepted. We were unable to link known SNPs in the genes encoding the SERT gene SLC6A4 or TPH1 with a clinically significant pattern of symptom features nor responsiveness to ondansetron, as has been suggested by others for the related drug ramosetron, 34 but our numbers were probably too low for such an analysis. Also relevant is the fact that we did not find a correlation between TPH1 SNPs examined and mRNA levels for TPH1.
These data are compatible with data in Gtex 35 showing that rs21105 has no significant eQTL signatures. rs4537731 has significant eQTL signatures in skin and thyroid but not in nerve or gut tissue.
Our pilot study asked subjects for pain scores on a 0-3 scale, which is insufficient to define a decrease in pain by at least 30% as recommended by the FDA. Consequently, we defined response by the change in stool consistency which is suboptimal in an IBS study.
Future studies should take this into account by using an 11-point pain scale allowing an assessment using both pain and stool consistency response according to FDA guidance.
While psychological factors are thought to be important in IBS we found no differences in anxiety nor somatisation between superresponders and non-super-responders, in keeping with ondansetron's known lack of central effects.
Overall, our study is important for highlighting the heterogeneity of IBS-D and indicates that personalised medicine approach is both necessary and possible if we know the mode of action and can easily assess the key factors determining response to particular drugs.

DISCL AIMER
This is a summary of independent research funded by the National