The Impact of NOD2 Genetic Variants on the Gut Mycobiota in Crohn’s Disease Patients in Remission and in Individuals Without Gastrointestinal Inflammation

Abstract Background and Aims Historical and emerging data implicate fungi in Crohn’s disease [CD] pathogenesis. However, a causal link between mycobiota, dysregulated immunity, and any impact of NOD2 variants remains elusive. This study aims to evaluate associations between NOD2 variants and faecal mycobiota in CD patients and non-CD subjects. Methods Faecal samples were obtained from 34 CD patients [18 NOD2 mutant, 16 NOD2 wild-type] identified from the UK IBD Genetics Consortium. To avoid confounding influence of mucosal inflammation, CD patients were in clinical remission and had a faecal calprotectin <250 μg/g; 47 non-CD subjects were included as comparator groups, including 22 matched household [four NOD2 mutant] and 25 non-household subjects with known NOD2 genotype [14 NOD2 mutant] identified by the NIHR BioResource Cambridge. Faecal mycobiota composition was determined using internal transcribed spacer 1 [ITS1] sequencing and was compared with 16S rRNA gene sequences and volatile organic compounds. Results CD was associated with higher numbers of fungal observed taxonomic units [OTUs] [p = 0.033]. Principal coordinates analysis using Jaccard index [p = 0.018] and weighted Bray‐Curtis dissimilarities [p = 0.01] showed Candida spp. clustered closer to CD patients whereas Cryptococcus spp. clustered closer to non-CD. In CD, we found higher relative abundance of Ascomycota [p = 0.001] and lower relative abundance Basidiomycota [p = 0.019] phyla. An inverse relationship was found between bacterial and fungal Shannon diversity in NOD2 wild-type which was independent of CD [r = -0.349; p = 0.029]. Conclusions This study confirms compositional changes in the gut mycobiota in CD and provides evidence that fungi may play a role in CD pathogenesis. No NOD2 genotype-specific differences were observed in the faecal mycobiota.


Introduction
Crohn's disease [CD] is a chronic relapsing inflammatory disease of the gastrointestinal [GI] tract, which affects the quality of life of over 1.5 million individuals in North America and Europe. 1,2 Although inflammation may occur at any point along the GI tract, CD commonly affects the terminal ileum and colon 3 and can lead to significant tissue damage, often necessitating the introduction of biologic therapy, and in many patients surgical intervention. [4][5][6] Whereas the aetiology and pathogenesis of CD remain unresolved, the likely mechanisms involve complex interactions between predisposing genes and environmental exposures, which lead to aberrant immune responses against the gut microbiota, resulting in an imbalanced microbial community. [7][8][9][10] CD patients have altered gut bacterial communities that include imbalances in Bacteroidetes, Firmicutes, and Proteobacteria phyla. [11][12][13] The gut microbiota of the healthy adult GI tract includes the fungal mycobiota [14][15][16] and a study by Sokol and colleagues, which examined the faecal mycobiota in adult CD patients, reported fungal dysbiosis. 17 Furthermore, a number of studies evaluating the faecal mycobiota in CD report increased prevalence of Candida spp., [17][18][19] although the link between increased abundance and disease pathogenesis remains obscure.
The strongest genetic association for CD susceptibility is in the gene encoding the nucleotide-binding oligomerisation domaincontaining protein 2 [NOD2; also known as CARD15]. 20,21 The three most common CD-associated NOD2 mutations, R702W and G908R, result in amino acid substitutions and L1007fsinsC results in a premature stop codon and dysfunctional NOD2. In Caucasians, up to 50% of CD patients carry at least one NOD2 mutation and individuals who carry two mutated NOD2 alleles have a 20-to 40-fold increased risk of developing CD. 20,21 NOD2 is a cytosolic pattern recognition receptor that is highly expressed in dendritic cells and macrophages 22 and Paneth cells, 23,24 and variably expressed in intestinal epithelial cells. 25 NOD2 is activated by muramyl dipeptide [MDP], a bacterial cell wall component, 26,27 and upregulates expression of cytokines, chemokines, and defensins, and triggers adaptive immune responses. 28 The three major CD-associated mutations are located within the ligand recognition domain of the NOD2 protein and are therefore defective in their ability to sense MDP and trigger autophagy, which impairs gut bacterial handling. 29,30 A number of studies have investigated the impact of NOD2bacterial interactions in CD, [31][32][33][34][35] including a retrospective study by Frank et al., which found NOD2-specific compositional shifts in the intestine-associated bacterial community. 34 Recently and in contrast, Kennedy and colleagues evaluated the impact of NOD2 variants on the faecal bacterial community in well-phenotyped CD patients compared with matched controls, and found no NOD2-specific alterations in the bacterial communities. 35 An intriguing study by Wagener and coworkers demonstrated that NOD2 is also activated by the fungal wall component chitin and induces interleukin 10 secretion, so NOD2-dependent recognition of chitin particles dampens inflammatory responses. 36 Currently, it is unknown what impact NOD2 genetic variants may have on the gut mycobiota in healthy individuals or patients with CD.
There were three aims of this study. First, we aimed to investigate the faecal mycobiota of CD patients in remission and of non-CD individuals, to determine whether there are unique fungal signatures that distinguish between the two groups, and if NOD2 genotype impacts upon the composition of the faecal mycobiota. Second, we compared the relative abundance of fungal and bacterial communities in order to identify potential correlations that may be implicated in CD and/ or NOD2 genotype. Third, we aimed to determine whether there are specific correlations between volatile organic compounds [VOC] and bacteria or fungi detectable in the faecal contents.

Study participants
This study used clinical data, stool samples, 16S rRNA gene sequences [accessible from the European Nucleotide Archive, accession number PRJEB21593], and faecal volatile organic compounds [VOC] collected previously for a study to investigate the impact of NOD2 genotype on faecal bacterial community profiles in CD and non-CD individuals. 35  . NOD2 mutant CD patients were matched for age, gender, and geographical location to a wild-type NOD2 CD patient. To avoid the confounding influence of inflammation, CD patients were confirmed to be in clinical remission as defined by physician assessment and a faecal calprotectin <250 μg/g. 6 All non-CD subjects had a faecal calprotectin <100 μg/g. 6 Healthy controls were stratified by the same NOD2 genotypes, ie NOD2 wild-type and NOD2 mutant [compound heterozygotes, other homozygotes, and single heterozygotes]. Two non-CD comparator groups were recruited for this study. 35 First an environmental control group, where possible a household member [usually an unrelated spouse] of CD participants, was recruited for stool sample collection and saliva for NOD2 genotyping.

Bioinformatic analysis
Sequences were processed using Mothur [v1.39.5], 37 and paired end reads were merged using make.contigs with trimoverlap set to true. Any sequence with an ambiguous base was removed from the dataset. Reads were assigned taxonomy using the UNITE database [v8] 38 and only fungal reads were retained for further analysis. A rarefied observed taxonomic units [OTU]

Volatile organic compounds
The VOC data were generated from a previous study 35 using a previously described method. 40

Statistical analysis
Statistical analyses were performed using Prism v. 8  Significance of parametric variables was determined using the two-sample t test for two-category comparisons or paired t tests for matched pair comparisons. Correlation between continuous variables were assessed using non-parametric Spearman rank correlation tests, or parametric variables were assessed using Pearson's correlation test. Basidiomycota/Ascomycota ratios were calculated for each sample by dividing the percentage relative abundance of Basidiomycota by the percentage relative abundance of Ascomycota. Principal coordinate analysis [PCoA], using unweighted Jaccard or weighted Bray-Curtis distance, was performed to investigate relationships in β-diversity, and between-variable distances were analysed by PERMANOVA. Where applicable, p-values were corrected using the false-discovery rate algorithm [FDR]. 41 To determine the associations between the relative abundance of either fungal genera or bacterial genera and the intensity of stool VOC, we performed sparse partial least squares regression analysis in canonical mode, using MixOmics. 42

Characteristics of the study population
The initial cohort comprised 113 study participants of Caucasian ethnicity [Supplementary Table 1, available as Supplementary data at ECCO-JCC online]; 32 samples were excluded from the study and the exclusion criteria are summarised in Figure 1A. A total of 81 participants remained and the characteristics of this cohort are summarised in Table 1A. Three household control subjects could not be NOD2 genotyped, leaving 78 participants for genotype-stratified analyses [ Figure 1B]. There were 34

Fungal diversity of patients with CD and non-CD individuals
Using the UNITE ITS database, we identified 523 fungal OTUs that were detectable in the stool samples of the cohort, and this identified 184 genera.
Comparing the within-sample diversity [alpha-diversity] between CD patients and non-CD subjects, we found a significant difference in observed OTUs [ Figure 2A; p = 0.033] but there was no difference in Shannon diversity [ Figure 2A; p = 0.19]. We examined betadiversity, which considers between-sample variation, of the fungal community using Jaccard indices [considers presence or absence] and Bray-Curtis dissimilarity [considers relative abundance]. A significant difference was observed in the PCoA using Jaccard indices [ Figure 2B; R 2 = 0.0184, p = 0.018] and Bray-Curtis dissimilarities [ Figure 2B; R 2 = 0.0246, p = 0.01; PERMANOVA]. The genus Candida and Cryptococcus appeared most discriminatory where Candida was most associated with CD and Cryptococcus was most associated with non-CD. The dominant phyla in the stool mycobiota were Ascomycota and Basidiomycota [Supplementary Table S2, available as Supplementary data at ECCO-JCC online]. Ascomycota were significantly more abundant in CD ( Figure 2C; false-discovery rate adjusted [FDR-Adj] p = 0.001), whereas Basidiomycota were significantly less abundant in CD compared with non-CD [ Figure 2C; FDR-Adj p = 0.019]. We sought to determine whether there were significant differences in the relative abundance of fungal genera and found no significant difference in genus abundance [ Figure 2D]. The Basidiomycota/Ascomycota abundance ratio was found to be significantly lower in CD patients [ Figure 2E; p = 0.0051]. We investigated whether there were correlations in alpha-diversity between the bacterial and fungal communities, but no significant correlations in the number of observed OTUs [ Figure 2F] or Shannon diversity [ Figure 2G] by disease status were found.

CD-specific changes in the mycobiota compared with household matched individuals
We sought to compare the fungal mycobiota composition between CD and non-CD shared household contacts. Analysis of the alpha-diversity revealed CD patients had a significantly higher number of OTUs [ Figure 3A; p = 0.0045, paired t test] and Shannon diversity [ Figure 3A; p = 0.048, paired t test]. PCoA using Jaccard and Bray-Curtis indices showed no clustering in the fungal community composition [ Figure 3B]. There was no significant difference in the relative abundance of any fungal phyla or genera between CD and household contacts [ Figure 3C and D]. The Basidiomycota/ Ascomycota abundance ratio was also not significantly different between groups [ Figure 3E; p = 0.083, Wilcoxon matched pair test]. We investigated whether there were correlations in alpha-diversity metrics between bacterial and fungal communities in CD and household matched subjects, and found no significant correlations in either the number of observed OTUs [ Figure 3F] or Shannon diversity [ Figure 3G].

Faecal mycobiota-NOD2 genotype association
To examine the impact of NOD2 genotype on the stool mycobiota, we compared the relative abundance of fungal phyla and genera between 42 NOD2 wild-type vs 36 NOD2 mutant subjects [19 compound heterozygotes, three frameshift homozygotes, 10 other homozygotes, and four single heterozygotes]. These analyses were independent of Crohn's disease. Comparing alpha-diversity, we found no significant differences in observed OTUs [ Figure 4A; FDR-Adj p = 0.43] or Shannon diversity [ Figure 4A; FDR-Adj p = 0.57]. The fungal profiles were comparable between NOD2 wild-type and NOD2 mutant subjects using Jaccard indices [ Figure 4B; R 2 = 0.0131, p = 0.494] and Bray-Curtis [ Figure 4B;  relative abundance of Ascomycota [ Figure 4C; FDR-Adj p = 0.28] or Basidiomycota [ Figure 4C; FDR-Adj p = 0.23] or genera [ Figure 4D]. The Basidiomycota/Ascomycota abundance ratio was not found to be significantly different [ Figure 4E; p = 0.182, Mann-Whitney test]. We investigated whether there were correlations in alpha-diversity between bacterial and fungal communities and found no significant       Finally, we sought to determine whether NOD2 would exert the strongest effect on the alpha-diversity of the stool mycobiota in CD patients and non-CD subjects matched for age, gender, and geography. We compared nine CD NOD2 wild-type subjects vs nine CD NOD2 mutant subjects and found no significant differences in observed OTUs [Supplementary Figure S1B

Fungal and bacterial interactions
To determine whether there were correlations in the relative abundance of fungal and bacterial genera, we performed sparse partial least squared [sPLS] regression. We identified pairwise positive correlations in the faecal mycobiota in samples between genera Wallemia and Anaerostipes, Debaryomyces and Pseudobutyrivibrio, Cryptococcus and Ruminococcus, Saccharomyces and Escherichia/ Shigella, and Sordariomyces and Clostridium-sensu-stricto-1 [ Figure 5A]. Previous work on these samples profiled the VOC in the faecal contents of the CD patients and non-CD subjects and found the concentrations of pentanoic acid, 2-butanone, acetone, and 2-hexanone, 5-methyl were significantly lower in CD, and the concentrations of 2-piperidinone and butanoic acid, 3-methyl-, ethyl ester were significantly higher in CD. 35 We sought to look at the relationship of these CD-discriminating VOCs with the relative abundance of fungal genera, employing a canonical correlation analysis approach [ Figure 5B], and identified a strong positive correlation between CD-associated Candida spp. and the CD-associated VOC 2-piperidinone, and between Debaryomyces and the CD-associated VOC 2-hexanone, 5-methyl-.
We also sought to determine the relationship of the VOCs with the bacterial genera [ Figure 5C] and identified strong inverse relationships between Prevotella with 2-piperidinone; the bacterial genera Dorea, Blautia, and Alistipes showed a negative correlation with acetone and 2-butanone. The genus Faecalibacterium was negatively associated with pentanoic acid and 2-hexanone, 5-methyl-, and Pseudobutyrivibrio inversely correlated with the concentration of pentanoic acid.

Discussion
This study reports that CD patients in clinical remission have higher fungal diversity compared with a non-inflammatory, non-CD comparator group. We found significantly higher observed OTUs in CD, and the fungal profiles of CD patients clustered distinctly from that of non-CD subjects, with Candida spp. and Cryptococcus spp. found to cluster with CD and non-CD subjects, respectively. We identified shifts in the fungal phyla composition in CD patients, notably lower Basidiomycota that include Cryptococcus spp. and higher Ascomycota that include Candida spp. However, following correction for multiple comparisons, there were no significant differences found in the relative abundance of fungal genera in CD patients and non-CD individuals. We also report significantly higher alpha-diversity in CD patients compared with their respective household matched control. A strength of this study was that it included well-characterised CD and non-CD participants of known NOD2 genotype. We compared the faecal mycobiota of subjects stratified by NOD2 genotype and in the presence or absence of CD, and found that no differences were evident.
We found CD-specific shifts in the abundance of Ascomycota and Basidiomycota. The Basidiomycota/Ascomycota ratio in non-CD subjects was approximately 1:2, and this increased to 1:8 in CD patients. Ascomycota dominated the fungal community in CD patients, and this finding was independent of intestinal inflammation. Our findings confirm previously identified shifts in fungal phyla in the stool mycobiota of CD patients compared with non-CD controls. [17][18][19]43 The study by Sokol and colleagues reported that, during intestinal inflammatory flare, the abundance of Basidiomycota increased markedly and the abundance of Ascomycota decreased, and that the Basidiomycota/Ascomycota ratio increased with increasing inflammation. 17 In contrast, our study excluded any participants with evidence of disease activity, and found that Ascomycota was the dominant phylum in CD patients and that the Basidiomycota/ Ascomycota ratio was decreased.
A number of faecal mycobiota studies in CD patients have found an association with Candida albicans, although the functional importance of this association remains unknown. 17,19,43 Our data found that C. albicans was the most abundant species in CD patients and also in individuals with CD-associated NOD2 mutations. A study by Hoarau and colleagues found that Candida tropicalis was significantly higher in CD patients and correlated positively with serum anti-Saccharomyces cerevisiae antibodies [ASCA]. 18 In our study, C. tropicalis was the eighth most abundant species identified in CD patients; however, there were no significant differences found between CD and non-CD subjects. It was also interesting to note that one CD patient was detectable for nearly 80% of all C. tropicalis ITS1 reads and this patient had CD-associated NOD2 mutations.
Recently, Limon and co-authors found that Malassezia restricta was more abundant in CD patients with the single nucleotide polymorphism CARD9 S12N , which normally expresses a key signalling adaptor that plays an important role in anti-fungal immunity. 44 Curiously, we did not detect M. restricta in our cohort. This discrepancy is potentially due to sampling differences. We used stool samples, whereas Limon and colleagues used intestinal wash samples, which have different microbial communities compared with stool. 45 The observation may also relate to demographic differences of the respective cohorts; all our study participants were recruited from four locations in the UK, whereas the subjects in the Limon study were recruited from California, USA.  Our study found that fungal diversity is higher in CD and this was independent of inflammation. This contrasts with the wellestablished lower bacterial diversity seen in CD, [11][12][13]35 and may indicate that fungi fill this niche in CD. The inclusion of matched VOC data from a previous study 35 enabled us to look at associations between metabolites and relative bacterial and fungal abundance. One of the primary routes with which the gut microbiota may interact and cross-communicate with each other and the host is through the production of volatile metabolites. Previous studies report that the faecal concentration of the ketone 2-piperidinone is higher in CD during active disease and remission. 35,46 The present study sought to evaluate correlations between fungal and bacterial communities with faecal VOC, and identified a positive correlation between Candida spp. and 2-piperidinone and a negative correlation between 2-piperidinone and Prevotella spp., which is intriguing. Future studies to assess if this compound has anti-bacterial properties that confer a competitive advantage on yeast, and enables Candida spp. to outcompete commensal bacteria such as Prevotella, would be helpful in interrogating the functional relationship between these microbes and VOC.
A study by Wagener and colleagues found that there were chitindependent pathways of NOD2 activation that lead to interleukin-10 secretion and promote intestinal homeostasis, 36 leading to the hypothesis that the mycobiota of CD patients with NOD2 mutations would differ from that of wild-type individuals. Our findings suggest that there is no difference in the faecal mycobiota between NOD2 wild-type or mutant individuals with or without CD. However, CD patients with NOD2 mutations often have ileal involvement, a more aggressive fistulising and fibrostenotic disease phenotype, and increased risk of postsurgical disease recurrence after ileal resection. [47][48][49] Evidence from animal models 50 and human studies 45 show that there is a modest correlation between the microbial contents of the stool and ileal wash samples, which suggests that future studies are warranted which should include paired samples to determine the impact of NOD2 genotype on the ileal-associated microbiome.
The present study has a number of strengths. The study included a cohort of well-characterised CD patients and non-CD subjects with known NOD2 genotypes. Importantly, the study excluded participants with raised faecal calprotectin levels, thereby reducing the confounding influence of inflammation, which is a key stressor to the gut mycobiota. 17 The study design enabled comparisons between CD vs non-CD controls, CD NOD2 wild-type vs CD NOD2 mutant, non-CD NOD2 wild-type vs non-CD NOD2 mutant, and CD vs matched household. The latter grouping was important, as unaffected individuals sharing the same household environment with a CD patient may also show signs of bacterial dysbiosis. 51 Our findings show that the fungal community of CD patients are richer and more diverse than the mycobiota of their respective household controls.
Some study limitations should be noted. Reflective of the complexity of undertaking a genotype-stratified study, despite recruitment from the UK IBD Genetics cohort via several IBD centres, the absolute number of participants and the numbers of participants in each group [CD patients, non-CD subjects, matched household subjects, wild-type NOD2, and mutant NOD2] and of samples in this multi-omic analysis, are relatively small. Furthermore, this study evaluated the impact of the most common NOD2 genetic variants on the mycobiota and stratified participants into two groups [wildtype NOD2 and mutant NOD2], as we recognized that it was underpowered to examine the effects of individual genotypes including the effect of rarer NOD2 variants. 52 To remove the confounding influence of inflammation, our study recruited CD patients in remission. However, it is possible that the greatest impact of NOD2 genotype on the gut mycobiota will manifest during active disease. We also did not undertake functional analysis regarding NOD2 activity in the respective CD-associated NOD2 mutations, and thus did not determine whether there was aberrant fungal sensing or handling in NOD2 mutant subjects compared with NOD2 wild-type. Finally, this study characterised fungal taxonomy by employing an ITS1 amplicon DNA sequencing approach, and we recognise that there may be variation in the taxonomic resolution of the fungal communities by targeting alternative genetic markers such as ITS2, 18S, and 28S rRNA. 53 In conclusion, this study shows that fungal diversity is higher and bacterial diversity is lower in CD. We show CD-specific changes in fungal diversity, which importantly is independent of inflammatory flare. Using stool samples, we could not identify specific associations of NOD2 genotype with the mycobiota in CD patients or non-CD subjects. Candida spp. clustered more closely with CD and highlights the need for further studies to characterise whether Candida is a gut commensal or a pathogen in CD patients. Future investigations should also explore the impact of NOD2 genotype on the ileal-associated mycobiota in larger cohorts stratified by NOD2 genotype, in order to resolve these complex mucosal interactions.