Crohn’s Disease–Associated Pathogenic Mutation in the Manganese Transporter ZIP8 Shifts the Ileal and Rectal Mucosal Microbiota Implicating Aberrant Bile Acid Metabolism

Abstract

Background

A pathogenic mutation in the manganese transporter ZIP8 (A391T; rs13107325) increases the risk of Crohn’s disease. ZIP8 regulates manganese homeostasis and given the shared need for metals between the host and resident microbes, there has been significant interest in alterations of the microbiome in carriers of ZIP8 A391T. Prior studies have not examined the ileal microbiome despite associations between ileal disease and ZIP8 A391T.

Methods

Here, we used the Pediatric Risk Stratification Study (RISK) cohort to perform a secondary analysis of 16S ribosomal RNA gene sequencing data obtained from ileal and rectal mucosa to study associations between ZIP8 A391T carrier status and microbiota composition.

Results

We found sequence variants mapping to Veillonella were decreased in the ileal mucosa of ZIP8 A391T carriers. Prior human studies have demonstrated the sensitivity of Veillonella to bile acid abundance. We therefore hypothesized that bile acid homeostasis is differentially regulated in carriers of ZIP8 A391T. Using a mouse model of ZIP8 A391T, we demonstrate an increase in total bile acids in the liver and stool and decreased fibroblast growth factor 15 (Fgf15) signaling, consistent with our hypothesis. We confirmed dysregulation of FGF19 in the 1000IBD cohort, finding that plasma FGF19 levels are lower in ZIP8 A391T carriers with ileocolonic Crohn’s disease.

Conclusions

In the search for genotype-specific therapeutic paradigms for patients with Crohn’s disease, these data suggest targeting the FGF19 pathway in ZIP8 A391T carriers. Aberrant bile acid metabolism may precede development of Crohn’s disease and prioritize study of the interactions between manganese homeostasis, bile acid metabolism and signaling, and complicated ileal Crohn’s disease.

Graphical Abstract

Open in new tabDownload slide

Introduction

ZIP8 is a key metal transporter that regulates manganese (Mn) homeostasis.¹ A pathologic mutation in ZIP8 (rs13107325, ZIP8 AlaA391Thr) is predicted to be one of the most deleterious in the human genome, and is associated with an increased risk of schizophrenia, obesity, dyslipidemia, and, important for this work, Crohn’s disease.^2-5 ZIP8 A391T is the coding variant associated with an increased risk of Crohn’s disease with the second highest prevalence in the population; the minor allele frequency (MAF) is approximately 0.07.⁶ In adults, the association between ZIP8 A391T and Crohn’s disease is most significant for the clinical phenotype of stricturing and penetrating ileal disease, which also remains the most difficult to effectively treat.²

The role of Mn homeostasis in Crohn’s disease remains to be fully elucidated.^2,7 Mn is a critical cofactor for key metalloenzymes that regulate redox homeostasis, glycosylation, and glucose metabolism, among other essential cellular activities, that can all have a potential role in intestinal homeostasis, inflammation, response to injury, and healing.¹ There are emerging reports that extremes of Mn dietary intake affects host fitness in mouse models of colitis and infection models.^7,8 Further, increased dietary Mn intake in human patients has been found to associate with transcriptional patterns closer to healthy individuals than those with active ileal Crohn’s disease.⁹ Given the association between ZIP8 A391T and complicated disease phenotypes, one possibility is that disruption of Mn homeostasis may drive fibrosing and penetrating disease.

Mn homeostasis in the gut is of particular interest because of the shared need between host and the resident microbes for essential metals.¹⁰ Therefore, it has been hypothesized that genetic variation in ZIP8 and effect on Mn homeostasis should also impact the microbiota. Thus far, 2 human studies have examined associations between A391T carrier status and the microbiome. In the initial description of the association between ZIP8 A391T and Crohn’s disease, genotype-specific associations with microbiota composition were studied using 16S ribosomal RNA sequencing data obtained from mucosal biopsies in the right and left colon. The cohort included healthy individuals (75 noncarriers, 22 carriers of ZIP8 A391T) and patients with Crohn’s disease in remission (58 noncarriers and 16 carriers of ZIP8 A391T). Greater than 500 individual operational taxonomic units (OTUs) were found to be significantly different in ZIP8 A391T carriers vs noncarriers, broadly characterized as a relative reduction in short-chain fatty acid–producing organisms.² In a subsequent study using a larger, population-based cohort of 446 individuals (416 noncarriers and 30 carriers of ZIP8 A391T) and 16S ribosomal RNA (rRNA) sequencing data derived from stool, no statistically significant difference was found in the microbiota composition in carriers vs noncarriers.¹¹

Yet, importantly, given the strongest association between ZIP8 A391T and ileal Crohn’s disease, the ZIP8 genotypic effects on the ileal mucosa–associated microbiome have not been studied.² Therefore, the purpose of this work is to determine if there is an association between ZIP8 A391T carrier status and the ileal mucosal microbiota composition in patients with Crohn’s disease. We hypothesized that the ileal mucosal microbiota composition of patients with ZIP8 A391T would be distinct from noncarriers and inform mechanistic hypotheses underlying why the genetic association between ZIP8 A391T and Crohn’s disease is strongest for ileal disease location with complex features.

The RISK cohort provides a one-of-a-kind resource of clinical, genetic, and mucosal microbiome profiling in newly diagnosed, treatment-naïve, pediatric patients with ileal Crohn’s disease. In this work, we demonstrate that ZIP8 A391T carriers exhibit trends toward sex-specific associations with clinical characteristics. We performed a secondary analysis of 16S rRNA gene sequencing data obtained from ileal and rectal mucosa to study associations between ZIP8 A391T carrier status and microbiota composition. We found a restricted number of zero-radius OTUs (zOTUs) associated with ZIP8 A391T. When the data were collapsed and analyzed at the genus level, we found a significant decrease in Veillonella in the ileal mucosa of ZIP8 A391T carriers. In human studies, Veillonella abundance has been reported to interact with bile acids and increased fibroblast growth factor 19 (FGF19) signaling.¹² We therefore hypothesized that ZIP8 A391T may interact with bile acid homeostasis. Using a mouse model of ZIP8 A391T and consistent with our hypothesis, we demonstrate a Zip8 genotype effect on the abundance of liver and fecal bile acids and decreased Fgf19 signaling. We further confirm dysregulation of FGF19 in an independent cohort of adult patients with Crohn’s disease; plasma FGF19 levels are lower in ZIP8 A391T carriers compared with noncarriers with ileocolonic Crohn’s disease. These data prioritize further evaluation of the interactions between Mn homeostasis and bile acid metabolism and signaling and if perturbations in these pathways predispose to and possibly drive complicated ileal Crohn’s disease.

Methods

IBD Plexus cohort

For these analyses, data from IBD Plexus was provided via an institutional agreement with the Crohn’s and Colitis Foundation. The ZIP8 genotype at rs13107325 was available for 927 patients. De-identified clinical metadata were available for 927 patients. The MAF of ZIP8 A391T was 0.079. We considered an additive model in which all ZIP8 A391T carriers (heterozygous and homozygous) were analyzed as a single group. The clinical characteristics of this cohort are described in the Results. From this main cohort, we analyzed 16S rRNA sequencing from ileal and rectal mucosal biopsies obtained at diagnosis for 240 patients for whom ZIP8 genotype data were also available (microbiome subcohort). Within the microbiome subcohort, the ZIP8 A391T MAF was 0.077 (36 patients). The age at diagnosis ranged from 1.75 to 16.92 years with a mean age 12.15 years. The study population was 61% male. The mean body mass index (BMI) at diagnosis and mucosal sampling was 17.38 kg/m². Montreal classification L3 disease location was the most common (n = 163 [68%]), followed by L1 (n = 57 [23.8%]), L2 (n = 12 [5%]), and L0 (2.1%). Over 96% of the patients were classified as B1 (nonstricturing, nonpenetrating) disease phenotype at time of diagnosis and mucosal sampling. Covariates included in this analysis included A391T carrier status, age at diagnosis, sex, BMI, disease location, and disease phenotype. Data on disease location were missing for 3 patients and B status was missing for 1 patient; these patients were excluded from the analysis, bringing the total microbiome subcohort to 236 patients.

Analysis of ZIP8 Genotype–Clinical Phenotype Associations

The cohort was divided by ZIP8 A391T carrier status and associates with key clinical phenotypes were performed. Specifically, we studied the interaction between ZIP8 genotype (with and without sex) and age and BMI at diagnosis, disease location, need for surgery, and need for anti-tumor necrosis factor medications with P < .05 considered significant. Fisher’s exact testing was used to determine enrichment of ZIP8 A391T carriers in patients who progressed to penetrating disease compared with stricturing disease. Mean time to progression from B1 (inflammatory) to complicated phenotypes (Montreal classification B2, B2 + B3, or B3) was compared using t test. Kaplan-Meier survival analysis was performed using the survfit function in R (R Foundation for Statistical Computing version 4.2.1) for all patients, males, and females stratified by carrier and noncarrier statuses. The log-rank test was performed to compare survival curves between carriers and noncarriers.

Human 16S rRNA Sequencing Data Acquisition and Analysis

Publicly available 16S rRNA gene sequence files for eligible RISK patients were retrieved from the National Center for Biotechnology Information Sequence Read Archive using the SRAToolKit (version 2.9.3). The UNOISE3 algorithm provided in USEARCH (version 11.0.667) was used to process paired-end FASTQ files into zOTUs.¹³ The fastq_mergepairs function was used to merge the paired-end reads allowing for a maximum of 10 differences in the overlapping region and a merged read length of 240 to 270 bp. Sequences with no more than 1 expected error were retained as the seeds to form the zOTUs. Default settings as implemented via the unoise3 function were used for error correction and chimera removal and mapped to zOTUs using the USEARCH otutab command with default settings. The naive Bayesian classifier as implemented in QIIME2 (version 2019.7) was used to perform taxonomic classification against the SILVA 13.8 reference database trimmed to the v4 region.¹⁴ Corrected sequences were inserted into the SILVA 13.8 phylogenic tree using the fragment insertion via the SATé-enabled phylogenetic placement technique.¹⁵ Samples with at least 3000 reads after error correction and zOTUs observed at >0.0001% relative abundance were retained for analysis. Observed zOTUs (eg, richness), Shannon diversity, and phylogenic diversity were calculated after subsampling to the lowest observed read depth (3111 reads), and multiple linear regression was used to test for differences in alpha diversity estimates according to ZIP8 A391T genotype adjusted for age at diagnosis, sex, BMI, disease location (Montreal L classification) at diagnosis, and Montreal B status. Beta diversity ordinations of the first 2 principal component analysis axes were performed on the zOTU count matrix after a variance stabilizing transformation–centered log-ratio transformation.¹⁶ Ordinations of the first 2 principal coordinates analysis axes performed on the subsampled zOTU count matrix were also generated using the Bray-Curtis dissimilarity, Jaccard distance (binary), and UniFrac (weighted and unweighted) distances. Differential zOTU relative abundance according to ZIP8 A391T status was assessed using zero-inflated negative binomial regression as implemented in the glmmTMB package (version 1.0.2). Models were adjusted for age at diagnosis, sex, BMI, disease location at diagnosis, and Montreal B status. The geometric mean of pairwise ratios method was used to compute a scaling factor to account for differences in sequencing depth and included as a model offset.¹⁷ Outlying counts were truncated at the 97% percentile as previously recommended for count-based models of microbiome data.¹⁸ zOTUs seen in fewer than 20% of samples were filtered prior to testing, and the Benjamini-Hochberg false discovery rate (FDR) correction was applied to control the proportion of false positive results. Log2 fold changes were estimated from the model predicted values. Similar analyses were conducted after aggregating zOTUs to genus-level classifications and for rectal mucosa samples.

Animal Studies

Wild-type (WT) C57/BL6 mice and Zip8 393T knock-in (KI) mice were used in these studies.¹⁹ Male mice (21-22 weeks old) were obtained from our colony and were housed in a single facility with a 12-hour light/dark cycle. Mice were fed conventional chow (Harlan/Envigo Teklad Global 18% Protein Extruded Diet 2018SX). For stool collection for bile acids, mice were housed in metabolic cages to permit 24-hour stool collection. At time of sacrifice, mice were euthanized using 5% isoflurane. Stools and organs were harvested as previously reported.¹⁹

Quantitative polymerase chain reaction

RNA was extracted using the Direct-zol RNA Miniprep Kit (Zymo Research; #R2052) as per the manufacturer’s protocols. Purified RNA concentrations were measured using wavelengths of 260/280 nm on the Beckman Coulter DU 800 spectrophotometer. From equal amounts of RNA, complementary DNA was generated using iScript (Bio-Rad). Real-time quantitative polymerase chain reaction was performed and analyzed using Power SYBR Green Master Mix (Thermo Fisher Scientific) reagents on the QuantStudio 6 flex real-time polymerase chain reaction system. The sequences of the sense and antisense primers were as follows: Cyp7a1 Forward 5ʹ-TCC CTC CTT TGA AAA ACG TG-3ʹ, Cyp7a1 Reverse 5ʹ-GAG GTT CTG AGG CTG TGC TC-3ʹ; Fgf15 Forward 5ʹ-TGA GCC ATC CAG TTG TGT CC-3ʹ, Fgf15 Reverse 5ʹ-CCA CTG GAG AAT TTG GGG CT-3ʹ; GAPDH Forward 5ʹ-CAT CAC TGC CAC CCA GAA GAC TG-3ʹ; GAPDH Reverse 5ʹ-ATG CCA GTG AGC TTC CCG TTC AG-3ʹ. GAPDH was included as an internal control, and relative expression was calculated using the 2(–ΔΔCt) method.

Total Bile Acid Quantification in Mice

Either 10 mg of liver tissues or feces were homogenized with 1 mL of cold isopropanol and heated at 50 °C for 2 hours for extraction of bile acids. The extracts were centrifuged at 2000 g for 15 minutes and fecal or hepatic bile acid concentrations in the collected supernatant were measured using a commercial colorimetric total bile acid assay kit (Cell Biolabs; #STA-631) as per the manufacturer’s instructions. Each bile acid standard and sample were assayed in duplicate. The endpoint calorimetric results were read at a primary wavelength at 405 nm and at a secondary wavelength at 630 nm using a microplate spectrophotometer.

1000IBD Validation Cohort

To confirm the observed dysregulation of FGF19 signaling in mice carrying ZIP8 A391T, we leveraged proteomic data from an adult genotyped cohort of patients with Crohn’s disease from the Netherlands.^20,21. This cohort is part of a large, multiomics IBD project in which patients with an established diagnosis of IBD were included at the University Medical Center Groningen. Patients were enrolled from 2009-2019, and after written informed consent was provided, detailed phenotypic information and multi-omics data were collected. The 1000IBD study was approved by the Institutional Review Board of the University Medical Center Groningen (no. 2008/338) and conducted in accordance with the principles of the Declaration of Helsinki (2013). Proteomic data, acquired through proximity extension assay technology (Olink Proteomics), from 567 adult patients with Crohn’s disease were used to compare plasma FGF19 levels between ZIP8 A391T carriers and noncarriers across clinical phenotypes. The median age at time of sampling was 38 years (interquartile range, 27-53 years), and 64% of patients were female. Median BMI was 24 kg/m² (interquartile range, 21.6-28.0 kg/m²). Ileal (Montreal L1) and ileocolonic (Montreal L3) disease locations were the most common (both 35%), followed by L2 (20%), and L4 either in combination (7%) or in isolation (2%). Regarding disease behavior, 50% of patients were classified as B1 (nonstricturing, nonpenetrating) at time of sampling, followed by B2 (32%) and B3 (18%). The majority of patients in this cohort were in clinical disease remission (~70%) at time of sampling.

Statistics

For the human cohort studies, the statistical tests used are described previously or within the figure legends. For the animal experiments, we performed statistical analysis using Prism GraphPad Prism 9 (GraphPad Software; RRID:SCR_000306 https://scicrunch.org/resources/data/record/nlx_144509-1/RRID:SCR_000306/resolver?q=SCR_000306&i=rrid:scr_000306). The individual data points are plotted and the error bars represent the mean and SEM. When 2 groups were compared, t test was used with P < .05 considered statistically significant.

Animal Study Approval

Animal experiments were approved by the Johns Hopkins University Animal Care and Use Committee in accordance with the Association for Assessment and Accreditation of Laboratory Animal Care International.

Results

Crohn’s Disease Phenotypes and Behaviors Associated With ZIP8 A391T in Newly Diagnosed Pediatric Patients

The RISK cohort provides the unique opportunity to study effects of disease-associated variants such as ZIP8 A391T on clinical disease phenotypes, progression, and treatment response. The RISK cohort is a one-of-a-kind resource with clinical and mucosal microbiome analysis from ileum and rectum. Patients were followed from time of diagnosis with Crohn’s disease for up to 5 years. We sought associations between ZIP8 A391T carrier status and important clinical phenotypes. The MAF of ZIP8 A391T in this cohort was 0.079. We studied if ZIP8 A391T carrier status influenced age at diagnosis, extent of disease at diagnosis, BMI at diagnosis, or over follow-up along with progression of disease phenotype from inflammatory to stricturing or penetrating disease (change in Montreal B status), treatment with anti-tumor necrosis factor α therapies, or surgery over the period of follow-up.²² We tested for both an interaction with ZIP8 genotype and interaction with sex and made 3 key observations: (1) there was a trend toward earlier age at diagnosis in male ZIP8 A391T carriers (P = .10); (2) there was an interaction between ZIP8 genotype and BMI at diagnosis, with lower BMI in male ZIP8 A391T carriers (P = .07); and (3) female ZIP8 A391T carriers were modestly protected from surgery (P = .08) (Table 1). We tested time to progression from B1 (inflammatory disease) to stricturing and/or penetrating disease (B2, B3, or B2 + B3 disease) using Kaplan-Meier survival analysis. We found no statistically significant difference between ZIP8 A391T carriers and noncarriers in aggregate or when stratified by sex (Supplementary Figure 1). These data provide clinical observations of modest significance and highlight the possible sex-specific effects of ZIP8 A391T. We do not find a direct relationship between ZIP8 A391T carrier status, rapidity of disease progression, or development of complicated disease phenotypes, although the data from the larger adult cohorts demonstrate the association with complex disease.²

Alpha and Beta Diversity Are Unchanged in Carriers of ZIP8 A391T at the Ileal and Rectal Mucosa

The next step of the analysis was to examine the ileal mucosal microbiome in the subcohort of patients with 16S rRNA gene sequencing data available (236 patients, ZIP8 A391T MAF = 0.077). The cohort was segregated by ZIP8 A391T carrier status, independent of disease activity. Here, we found no material differences in alpha diversity as measured by the number of observed zOTUs (eg, richness; beta = −1.54, SE = 13.39, P = .91), Shannon diversity (beta = 0.16, SE = 0.13, P = .21), or phylogenetic diversity (beta = −0.11, SE = 0.72, P = .88) according to ZIP8 A391T carrier status (Supplementary Figure 2). Further, there was no clustering of samples according to ZIP8 A391T for any beta diversity ordination examined (Supplementary Figure 3).

Six OTUs and a Single Genus, Veillonella, Exhibit Differential Abundance in ZIP8 A391T Carriers at the Ileal Mucosa

Six zOTUs were flagged as differentially abundant in carriers of ZIP8 A391T. Four of these met an FDR threshold of <0.05 (2 others with FDR < 0.10). All differentially abundant zOTUs showed reduced relative abundance in the ileum in ZIP8 A391T carriers. The largest fold change difference was for zOTU141 mapping to the genus Dorea but was unclassified at the species level (log2 fold change = −5.75). Two other differentially abundant zOTUs mapped to Veillonella dispar (zOTU49, zOTU70). The final 3 zOTUs were classified to Faecalibacterium prausnitzii (zOTU3311), Bifidobacterium unclassified (zOTU47), and Bacteriodes ovatus (zOTU32) (Figure 1A-1C).

We then aggregated reads to study genus-level phylotypes. Again, there were no differences in alpha or beta diversity based on ZIP8 A391T carrier status. Veillonella was the single genus that was significantly different by ZIP8 A391T carrier status with a composite log2 fold change of –1.83 (FDR P = .006) (Figure 1D).

Erysipelotrichaceae Is Increased and Bifidobacterium Is Decreased in the Rectal Mucosa of ZIP8 A391T Carriers

In the same patients, rectal biopsies were also collected to study the mucosa-associated bacterial composition. Here, again, there were no differences in alpha or beta diversity by ZIP8 genotype. Analysis of differentially abundant taxa revealed only 2 zOTUs associated with ZIP8 A391T: Erysipelotrichaceae unclassified was enriched in ZIP8 A391T carriers (log2 fold change 2.34, FDR P = .082) and Bifidobacterium unclassified (zOTU47) was reduced in A391T carriers (log2 fold change –4.06, FDR P = .082) (Figure 1E).

Bile Acid Signaling Is Perturbed in the Mouse Model of ZIP8 A391T

In human studies, the abundance of Veillonella increased in patients treated in a clinical trial of the FGF19 analog aldafermin as a treatment for nonalcoholic steatohepatitis.¹² FGF19 is a gut hormone that is induced in ileal enterocytes by bile acid stimulation, prompting downregulation of hepatic bile acid synthesis.²³ These data position Veillonella as sensitive to bile acid composition and abundance.¹² Further, the abundance of Erysipelotrichaceae has been reported to correlate with cholesterol metabolism in humans and dysregulation of bile acid receptor signaling in mice.²⁴ Bile acids have been reported to have a growth-inhibitory effect on Bifidobacterium species, with the majority of tested species exhibiting weak-to-moderate tolerance to bile.^25,26 Given that ZIP8 A391T is associated with disease of the ileum—the key site of bile acid enterohepatic circulation—and the connections to bile acid metabolism in the differentially abundant mucosal bacteria found in our microbiome analysis, we hypothesized that ZIP8 A391T alters bile acid metabolism.

We recently established a mouse model of ZIP8 A391T, sequence equivalent to mouse Zip8 393-Thr.¹⁹ The Zip8 393-Thr KI mice recapitulate features of the genetic associations with ZIP8 A391T, specifically reduced blood Mn and susceptibility to colonic injury using the dextran sulfate sodium colitis model—a finding that has been replicated in an independently generated Zip8 393T-KI mouse.²⁷ Therefore, we used the Zip8 393T-KI mice to specifically test an interaction between Zip8 393T-KI and bile acid metabolism.

First, we measured total bile acid abundance in stool and found that total bile acids were increased in Zip8 393T-KI mice compared with WT animals (Figure 2A). This suggested either increased production or decreased absorption of bile acids. Therefore, we measured total bile acid abundance in the liver, which was also increased (Figure 2B).

Again, human studies of treatment with an FGF19 analog showed increased Veillonella.¹² Due to the increased bile acids in the liver of Zip8 393T-KI mice and decreased Veillonella in the ileal mucosa of Zip8 A391T carriers, we hypothesized that ileal Fgf15 (murine analog of FGF19) would be decreased. Decreased Fgf15 would result in the loss of negative regulation of de novo bile acid synthesis in the liver. Fgf15/FGF19 is regulated by Farnesoid X receptor activation in ileal epithelial cells after bile acid resorption. Enterohepatic circulation of Fgf15 regulates bile acid synthesis via Cyp7a1; therefore, it is a critical metabolite that determines bile acid metabolism. We used quantitative polymerase chain reaction to quantify Fgf15 messenger RNA (mRNA) levels in the mouse ileum. Consistent with our hypothesis, Fgf15 mRNA was decreased in Zip8 393T-KI mice with a reciprocal loss of negative regulation of hepatic Cyp7a1 mRNA showing increased Cyp7a1 mRNA in the livers of the Zip8 393T-KI mice compared with WT animals (Figure 2D, E). These data confirmed the hypothesis that there is an interaction between ZIP8 genotype and bile acid metabolism.

Plasma FGF19 Levels are Decreased in ZIP8 A391T Carriers Consistent With Aberrant Bile Acid Signaling

To further establish this interaction, we leveraged data from an adult cohort of patients with Crohn’s disease (1000IBD) to test the hypothesis that FGF19 levels are decreased in ZIP8 A391T carriers compared with noncarriers. FGF19 in Crohn’s disease has been studied by multiple groups, finding that FGF19 levels are decreased in patients with Crohn’s disease compared with individuals without Crohn’s disease, and FGF19 levels are decreased in inflamed mucosa compared with uninflamed mucosa.²⁸ FGF19 levels are lower in patients with ileal involvement and lower in patients with stricturing and penetrating disease.²⁰ It is known that FGF19 levels are affected by ileal resection and dietary patterns.^20,29 Here, we asked if FGF19 levels are modified by ZIP8 genotype using the 1000IBD cohort.²¹ ZIP8 A391T was not significantly associated with FGF19 levels when analyzed across all patients with Crohn’s disease. However, when the cohort was restricted to patients who did not undergo Crohn’s disease–related surgery and stratified based on Montreal classification, ZIP8 A391T was associated with decreased FGF19 levels in patients with ileocolonic disease (L3) compared with noncarriers (P = .01) (Figure 2F).²² In contrast, this was not observed in patients with colonic disease (L2), while a trend toward lower circulating FGF19 levels was observed in patients with solely ileal disease (L1) involvement (P = .26). These data provide further evidence that decreased FGF19 signaling is a feature of disease in carriers of ZIP8 A391T with ileal involvement.

Discussion

In this work, we examine clinical phenotypic associations of ZIP8 A391T in a cohort of newly diagnosed pediatric patients with Crohn’s disease. We identify associations between ZIP8 A391T carrier status and the ileal and rectal mucosal microbiome that implicate aberrant bile acid metabolism. We demonstrate that bile acid signaling is perturbed with ZIP8 A391T in a dedicated animal model, which we validated in an independent cohort of patients with Crohn’s disease. The unique feature of our study is that the data from the animal model suggest perturbation of bile acid signaling may be present in individuals who carry ZIP8 A391T prior to any Crohn’s disease–related inflammation.

Our work adds to the limited studies that examine how this pathogenic mutation may define a subgroup of patients with Crohn’s disease. We take particular note of the studies that have reported on features of healthy individuals who carry ZIP8 A391T. In the study of the colonic mucosa, ZIP8 A391T carriers exhibited a relative depletion of genera associated with short-chain fatty acid production, while the study of the stool microbiota in healthy carriers did not show a strong association between ZIP8 A391T and the microbiome.^2,11 Although these data could be interpreted as conflicting, we argue that they represent analysis of different microbial compartments within the gut and that the sensitivity to detect changes driven by mutation in a single gene may differ across the compartments. The relative decrease in Veillonella that we found in the ileal mucosa of ZIP8 A391T carriers is further notable, as Veillonella was otherwise reported to be increased in patients with Crohn’s disease in the RISK cohort at diagnosis compared with healthy control individuals.³⁰ Like our previous report that carriers of ZIP8 A391T have a unique plasma N-glycome compared with the population of Crohn’s disease patients analyzed as a whole, the association with decreased Veillonella with ZIP8 A391T continues to lend evidence to subtle but unique features associated with ZIP8 A391T.¹⁹ The KI mouse model of ZIP8 A391T provided a reductionist model to study bile acid homeostasis and demonstrate the interaction with ZIP8 genotype.

Aberrant bile acid signaling in Crohn’s disease has long been recognized, although the specific pathomechanisms have not been fully dissected. The paradigm has been that with Crohn’s disease–related inflammation, the critical physiologic functions of the ileum, like enterohepatic circulation, are impaired. In multiple studies, patients with Crohn’s disease have an increased abundance of primary bile acids.^31,32 Dysbiosis in patients with Crohn’s disease has been linked with decreased conversion of primary bile acids to secondary bile acids, and change in the bile acid composition has been linked to downstream immune effects.³³ Because bile acid receptors are expressed on a number of immune cell types including T cells, B cells, dendritic cells, monocytes, natural killer cells, and granulocytes, a number of recent studies have shown that bile acids affect immune responses in IBD like regulatory T cell populations and differentiation of effector T cell populations.³⁴

In addition to changes in bile acid composition, bile acid uptake is changed in patients with Crohn’s disease. Inflammation in the ileum downregulates the primary bile acid transporter, apical sodium dependent bile acid transporter (ASBT).³⁵ Additionally, Crohn’s disease medications can regulate ASBT, including the ileal-targeting steroid budesonide that induces expression of ASBT.³⁶ Downstream from ASBT, patients with Crohn’s disease have lower levels of FGF19 in plasma compared with healthy individuals. FGF19 levels are the most reduced in patients with ileal involvement, prior ileal resection, or active inflammation, and associate with dietary patterns of patients with Crohn’s disease.²⁸ FGF19/Farnesoid X receptor agonists are in clinical trials for the treatment of steatohepatitis and have also been proposed for the treatment of IBD.¹²Veillonella abundance seen with FGF19 agonism in patients with steatohepatitis provided an important mechanistic link between bile acid perturbations and the FGF19 pathway in our model.¹² It is further notable that V dispar is the taxa most associated with the alterations of the fecal microbiome in patients with autoimmune hepatitis.³⁷ Consistent with our model, autoimmune hepatitis is associated with increased FGF19 levels and decreased bile acid synthesis.³⁸ In the Zip8 393T-KI mice, we confirm decreased Fgf15 mRNA levels with a reciprocal increase in Cyp7A1 mRNA in the liver and increased total bile acid levels in the liver. There are important animal studies that have shown efficacy of Fgf15 agonism in reducing colitis severity in the chemical-induced colitis model of dextran sulfate sodium, and we would prioritize preclinical studies in the Zip8 393T-KI mice.²⁸

To complement the animal model, the relationship between ZIP8 A391T and altered FGF19 signaling was validated in a cohort of adult patients with Crohn’s disease. Again, FGF19 levels in patients with Crohn’s disease can be affected by multiple factors, including active ileal inflammation, prior surgeries, and dietary patterns.^20,29 Despite these multiple determinants of FGF19 signaling, we were able to show that ZIP8 A391T was associated with decreased FGF19 levels in patients with ileocolonic disease (L3) compared with noncarriers, while excluding patients with a known history of ileocecal resections. These data suggest that FGF19 agonism may have particular efficacy in ZIP8 A391T carriers with ileal disease involvement, and that impaired release of FGF19 into the circulation in these patients is likely driven by intestinal inflammation and impaired gut barrier integrity, which both may result in disrupted bile acid homeostasis.^28,39,40 It is provocative to consider perturbation of FGF19 signaling may underlie the pleiotropic associations between ZIP8 A391T and human disease, including schizophrenia and obesity, in which bile acid metabolism has is known to be changed.^41,42 Further work is needed to determine the specific underlying drivers of ZIP8-associated disease and how they all may interact.

Why ZIP8 genotype modifies bile acid production and FGF19 signaling remains to be elucidated. The intersectionality with Mn homeostasis and bile acid metabolism has been recently suggested: expression of SLC30A10, a key Mn exporter, increased in mouse primary ileal enteroids stimulated with bile acid pools enriched for 12α-hydroxylated bile acids.⁴³ 12α-hydroxylated bile acids are increased in individuals with insulin resistance, influence intestinal fat absorption, and contribute to hepatosteatosis. But how and why Mn homeostasis in the liver or ileum is regulated by bile acids and if this regulation would be perturbed by ileal inflammation is uninvestigated.⁴⁴ Further, a primary epithelial disorder with either ZIP8 A391T or changes in Mn homeostasis that affects bile acid uptake will require further study.

We acknowledge that the conclusions from the clinical phenotype-genotype analysis are limited by sample size and statistical power, particularly when stratifying by sex. It is interesting that there was a trend toward earlier diagnosis in male ZIP8 A391T carriers. We have shown in the Zip8 393T-KI mice that the differential effect of ZIP8 A391T on Mn homeostasis is greater in male mice.¹⁹ These data are congruent with the longstanding appreciation of the influence of sex on Mn homeostasis.⁴⁵ Overall, although we do not definitively prove an association with a more aggressive clinical course, these data support the study of the role of sex-specific effects of ZIP8 A391T and aberrant Mn homeostasis in Crohn’s disease.

Conclusions

As we move toward precision medicine in IBD, the goal remains to identify subgroups of patients with discrete pathogenic drivers that can be specifically targeted. ZIP8 A391T is a foundational pathologic variant in the human genome given its pleiotropic disease associations. ZIP8 A391T has challenged us to reconsider Mn homeostasis in human health. The associations between ZIP8 A391T and metabolic phenotypes (obesity and dyslipidemia) and hepatic fibrosis push the hypotheses that ZIP8 A391T–driven pathology may critically link enterohepatic physiology and provide novel insights into the gut-liver interactions in Crohn’s disease, which may be regulated by bile acid metabolism.^46-48 Future studies will focus on the mechanism driving decreased FGF19 signaling and investigate targeting of FGF19 in the animal model of Zip8 393T-KI with the goal of developing a genotype-specific therapeutic approach.

Acknowledgments

The authors acknowledge the many patients who have participated to build this research resource. They thank Dr Subra Kugathasan and David Hercules at Emory University for assistance providing de-identified data from the RISK cohort. Statistical support was provided by the Biostatistics, Epidemiology and Data Management Core at Johns Hopkins. IBD Plexus de-identified patient data were provided through an agreement with the Crohn’s and Colitis Foundation facilitated by Orlando Green and Angela Dobes with local sponsorship by Dr Maria Oliva-Hemker, Johns Hopkins University School of Medicine.

Author Contribution

K.B.: analysis, visualization, writing, reviewing, editing. V.T.: validation, analysis, visualization, writing, reviewing, editing. N.O.: methodology, validation, analysis, visualization, writing, reviewing, editing. Y.H.: methodology, validation, analysis, visualization, writing, reviewing, editing. A.R.B.: resources, analysis, visualization, writing, reviewing, editing. S.H.: resources, analysis, visualization, writing, reviewing, editing. L.C.: analysis, visualization, writing. L.S.: validation, analysis. R.K.W.: methodology, resources. L.A.D: methodology, resources, reviewing, editing, funding acquisition. J.M.P.M.: conceptualization, methodology, analysis, resources, writing the first draft of the manuscript, reviewing, editing, visualization, supervision, and funding acquisition.

Funding

Funding was provided by the National Institutes of Health (DK114478 to J.M.P.M.), the Johns Hopkins Specialized Center for Research Excellence in Sex Differences (U54AG062333 – pilot funding to J.M.P.M.), the American Gastroenterological Association (J.M.P.M.), the Doris Duke Early Clinician Investigator Award (#2020147 to J.M.P.M.), and the Crohn’s and Colitis Foundation (N.O., Y.H., L.A.D.). The 1000IBD proteomic study was supported by the collaborative T cell–driven Immune Mediated Inflammatory Diseases project (LSHM18057-SGF) financed by the PPP allowance made available by Top Sector Life Sciences and Health to Samenwerkende Gezondheidsfondsen to stimulate public-private partnerships and cofinancing by health foundations that are part of the Samenwerkende Gezondheidsfondsen.

Conflicts of Interest

There are no direct conflict of interests to disclose related to the work presented. J.M.P.M. has received funding to support an IBD fellowship program through Pfizer and Johnson and Johnson and investigator initiated research funding from Pfizer. J.M.P.M. has been a site clinical investigator for a clinical trial funded by Seres Pharmaceutical.

Data Availability

IBD Plexus data are available through application to the Crohn’s and Colitis Foundation. 1000IBD data are available through collaboration with the Weersma research group.

References

Author notes