Abstract
Free full text
FUT2 genotype and secretory status are not associated with fecal microbial composition and inferred function in healthy subjects
ABSTRACT
Heritability analysis of the microbiota has demonstrated the importance of host genotype in defining the human microbiota. The alpha (1,2)-fucosyltransferase 2 encoded by FUT2 is involved in the formation of the H antigen and the SNP, rs601338 is associated with ABO histo-blood group antigen secretion in the intestinal mucosa. Previous studies have provided non replicated results for the association of this polymorphism with the composition and inferred function of intestinal microbiota. We aimed to assess this relationship in a large cohort of 1,190 healthy individuals. Genotyping was performed using the HumanCoreEXOME chip, microbial composition was addressed by 16S rRNA gene sequencing. Firmicutes, Bacteroidetes, and Actinobacteria were the dominant phyla in this cohort. Although we have sufficient power to detect significant associations of FUT2 genotype/ inferred phenotype with the microbiota, our data demonstrate that FUT2 genotype and secretor status is not associated with microbial alpha diversity, microbial composition or inferred microbial function after correction for multiple testing. Thus, FUT2 genotype and inferred phenotype are not associated with human fecal microbial composition and imputed function.
Introduction
The human gut microbiome is involved in a number of important metabolic functions.1 The major microbial phyla comprising the human gut microbiome are Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria2 At the genus taxonomy level, the composition varies markedly in healthy subjects, with some subjects having > 80% of their microbiota composition represented by a single taxa such as Bacteroides or Prevotella genera.3 Although, the function and composition of the human microbiome are subject to long-term temporal stability there is considerable short-term variability.4 The reason for high inter-individual variability of the composition of the microbiome in healthy individual is unknown; however, diet, environmental factors and host genetics likely contribute to the variability of the composition of the human gut microbiome.4-6
Heritability analysis of the microbiota has demonstrated the importance of host genotype in defining the human microbiota.6,7 Although our data in a large cohort failed to show any association with inflammatory bowel diseases (IBD) risk associated genetic polymorphisms.7 other studies suggest that specific genetic variants in key genes involved in mucus composition such as FUT2 are associated with microbial composition.8-11 Therefore, the association of IBD risk associated genetic polymorphisms with the diversity of the human microbiota remains controversial.12,13 The protein encoded by the alpha (1,2)-fucosyltransferase 2 gene (FUT2) is responsible for secretion of the ABO histo-blood group antigens in the mucosa. The FUT2 associated SNP rs601338 is associated with the secretor/non secretor status. The minor allele (A) confers a non-secretor phenotype and is associated with Crohn's disease (CD) susceptibility with odds ratio of 1.11 (1.071–1.143 95% CI, p < 10−15).13-15 Host-microbe interactions can involve FUT2 through modification of components of the mucus layer,16-19 and thus FUT2 is an ideal candidate to test for association with the gut microbiota variability. Indeed several studies have already investigated the role of FUT2 in the composition of the microbiota but large biological (mucosal versus stool sample or bile duct) and technical differences (DNA extraction and profiling method of the microbiota) exist between studies resulting in potentially different results.8,10,11,20 Indeed the FUT2 expression in tissue is the highest in salivary gland, then esophagus-mucosa, then terminal ileum, and then transverse colon21 and it was shown that the effects of the FUT2 genotype/secretor status on bacterial phyla decrease in distal region of the gastrointestinal tract.22 In addition detecting an association of FUT2 with microbial composition might also depend on other factor such as the stress applied to the host and/or diet of the host.19,23
We examined FUT2 associated SNPs with intestinal microbial composition and inferred function using a cohort of subjects collected as part of a prospective cohort study of healthy first degree relatives of CD subjects (The Genetic, Environmental Microbial (GEM) project). We analyzed the association of FUT2 with intestinal microbiota in 1,190 healthy first degree relatives of CD patients. Our results show that FUT2 genotype and secreting status are not significantly associated with fecal microbiota composition and inferred function.
Results
Demographics of the cohort
The cohort comprised a total of 1,190 healthy individuals of European descent. The majority of the subjects were recruited from several Canadian provinces with the highest number from Ontario, Alberta and Quebec (37.3%, 24.5% and 14.4%) and 2.0% where recruited from either Israel or USA (Supplementary Table 1). In the studied cohort, there were more females (55.5%) and the mean age was 19.8 ± 7.7 (mean ± SD).
Description of the microbiota composition of the cohort
The three dominant bacterial phyla were Firmicutes (relative abundance of 64.2% ± 14.1), Bacteroidetes (26.9% ± 15.0), and Actinobacteria (5.0% ± 5.2) (Supplementary Figure 2). Of the 127 genera, we found that Blautia, Coprococcus, Ruminococcus, Bacteroides, Dorea, Roseburia, Faecalibacterium, Streptococcus and Oscillospira genera were found in all subjects. The remaining 118 genera were irregularly observed as present or not present across subjects. For example two individuals harbour 8,468 and 4,881 reads assigned to a single OTU (28.2% and 16.2% of their microbiota) assigned to Succinivibrio genus, an OTU usually found in less than 5% of the general population. This confirms that our cohort have the existence of a core microbiome24 coexisting with highly variable bacterial taxa of the human gut microbiome.4
Association of rs516246 with the gut microbiota
When examined against all OTUs, rs516246 genotype and the inferred FUT2 phenotype was not associated with microbial alpha diversity (Supplementary Tables 3, and Supplementary Figure 3). Because Bifidobacteriales taxon was previously associated with alpha diversity index we subsampled our microbiome data to perform an analysis restricted to this taxon.9-11 Our data showed that Bifidobacteriales alpha diversity estimated by Chao1, Simpson, Shannon and observed species indices were not associated with FUT2 genotype or secretor/non-secretor status (Supplementary Tables 4).
PCoA analysis of beta diversity performed on all OTUs as measured by unweighted UniFrac distances revealed that there was no clustering by genotype (R2 = 0.002, p-value = 0.31) or inferred phenotype (R2 = 0.001, p-value = 0.18) (Figure 1). Similar results was observed using Bray-Curtis distances across genotype (R2 = 0.002, p-value = 0.28) or inferred phenotype (R2 = 0.001, p-value = 0.13) (Figure 1). The difference in Bacteroidetes relative abundance reported previously among non-secretors was not replicated in our cohort (Figure 2).
PCoA plot using Unweighted Unifrac and Bray-Curtis distances for beta diversity measure. Individuals who are either homozygous major (GG), heterozygous (GA) or homozygous minor (AA), at rs516246 are colored in orange, blue, and red respectively. Individuals with a secreting status are coloured in green and individuals non-secretor are coloured in purple.
Beeswarm plot of the relative abundance of Bacteroidetes in individuals who are either homozygous major (GG), heterozygous (GA) or homozygous minor (AA), at rs516246 in 918 unrelated individuals of European descent. The lines represent the first (blue), second (bold red), and third quartiles (blue). Circles represent the relative abundance of the given family from an individual subject's sample.
We then applied the method published previously to analyze the microbiota at lower taxonomic levels.8 Briefly, OTUs were filtered based on an observed presence in at least 60% of the samples and with a minimum total count of 30 across samples. This strategy reduces the total number of OTUs from 12,863 down to 396, and consequently reduces the number of comparisons tested. Again, FUT2 genotype or inferred phenotype was not significantly associated with any of these filtered OTUs p>1.3 10−4 (Supplementary table 5).8
FUT2 genotype/ inferred phenotype are not associated with microbial composition even with less stringent filtering of OTU table
Filtering OTUs based on presence in at least 60% of the samples has important consequences on microbial composition characteristics (Supplementary Figure 1). A rarefaction at 30,000 reads, followed by filtering OTUs based on a presence in at least 60% of the samples decreased the number of reads to a mean of 23,952. For some individuals such filtering leads to an inaccurate estimation of the microbial composition (Supplementary Figure 1). By way of example, for five individuals, a single OTU assigned to Prevotella copri represents up to 40% of their microbiota. Filtering applied such that OTU's present in less than 60% of samples are removed would result in a finding that this bacteria represents <1% of the individual's entire fecal microbiota. In order to determine whether sensitivity to this filtering would alter the association between FUT2 and microbiota, we repeated the association analysis to include OTUs with an observed presence in at least 5% of the samples. This strategy reduces the total number of OTUs from 12,863 to 4,353 without altering the overall microbial composition (Procrustes of OTU table with no filtering against OTU table filtered based on 5% of OTUs prevalence, M2 = 0.0, p-value <10−6) and removed most singleton and/or spurious OTUs.25
Given this compositional analysis we then assessed the associations between bacteria at the level of Phylum down to individual OTUs. Using this filtering of OTUs we were able to detect several nominal associations but they did not survive correction for multiple testing (Supplementary table 6).
Paired test in family members with FUT2 inferred phenotype discrepancy
Finally, we applied a different strategy to assess the relationship between secretor/non-secretor and microbial composition. Within the original cohort we had identified 102 individuals from 46 families who did not share the same secretor status (Secretor and non-secretor individuals from the same family). A paired t-test for related subjects failed to identify significant associations with taxa after correction for multiple testing (with the lowest p-value being 0.001 for Tissierellaceae taxa) (Supplementary table 7). The significant difference in phylotype abundances reported previously, namely, the increase in Bacteroidetes among non-secretors8,9 was not confirmed in this analysis (Supplementary Figure 4).
Imputed function failed to detect any association between imputed microbial function and FUT2 genotype and secretor/non-secretor status
Microbial functions were imputed and tested for association with the FUT2 genotype/inferred phenotype (See Material and Methods, Supplementary table 8, Supplementary table 9). It was previously shown that microbiome from secretor individuals separate from non-secretor.8 Our PCoA analysis of beta diversity performed on all OTUs as measured by Bray-Curtis distances revealed that there was no clustering by inferred phenotype (R2 = 0.0004, p-value = 0.77).We did not identify any association of FUT2 genotype or inferred phenotype with predicted metagenome composition by KEGG after correction of the p value for multiple testing (Supplementary table 9). Previous associations8 in amino acid biosynthesis, in carbohydrate and lipid metabolism, in cofactors and vitamins metabolism and in glycan biosynthesis and metabolism in non-secretor (AA) individuals as compared to secretor (GG) were not replicated in our dataset. However some subcategories have nominal association with FUT2 genotype such as K05988 (p-value <0.004) and K01799 (p-value <0.03) involved in carbohydrate metabolism, or K03918 (p-value <0.02) and K13831 (p-value <0.04) involved in amino acid metabolism (Supplementary table 9).
COG categories related to carbohydrate transport and metabolism (class [G]) were not significantly associated in pair comparison. Examining subcategories of COG specifically related to fucose metabolism, i.e. COG2407 (L-fucose isomerase and related proteins), COG3594 (Fucose 4-O-acetylase and related acetyltransferases), COG0738 (Fucose permease), and COG3669 (Alpha-L-fucosidase) failed to identify a significant difference across genotype groups. However subcategory COG4154 (Fucose dissimilation pathway protein FucU) had a nominal association with FUT2 genotype (p-value <0.03) but this did not survive correction for multiple testing. Other COG function were nominally significant but were not closely link to fucose metabolism of the mucus (COG3465, an uncharacterized protein YwgA COG3676, a transposase and inactivated derivatives, COG3728, a phage terminase, and COG3774, a mannosyltransferase).
Discussion
We examined a cohort of healthy first degree relatives of CD patients to assess if FUT2 polymorphism and inferred phenotype was associated with microbial composition and inferred function using 16S rRNA gene sequencing and PICRUSt software analysis to infer bacterial community function. The size of the cohort we examined allowed for the study to be sufficiently powered to detect the previously reported effect size for the association of FUT2 with microbial composition.8-11 Indeed, standard procedure for genetic-traits association was carefully applied that include, control of population stratification by restricting analysis to individuals of European descent, control of type I error by excluding related subjects (n = 272 individuals). With a total number of 918 unrelated individual, the power calculations in our study showed that we have >80% power to detect the association of FUT2 rs516246 associations with 10 out of 13 bacterial phyla (Supplementary Tables 10) based on prior studies effect size.8,9 Despite the fact that the size of our cohort is larger than that of previous studies,8-11 our results indicated that FUT2 is not associated with fecal microbial diversity, composition or inferred function. However, our results are in agreement with another recent study using a large cohort of 1,503 twins from the United Kingdom,26 as well in 3 different microbiome GWAS studies that all failed to identify any association of FUT2 and fecal microbiome composition, even after adjusting for age, sex covariates and taking into account family structure.7,27,28
Alpha diversity was not significantly different across genotype using Chao1 and number of species index. These two indexes were applied and previously reported to be decreased in AA and GA as compared to GG, with a reported p values of 0.012 and 0.085 respectively.8 However, another study using PCR Denaturing Gradient Gel Elecrophoresis (DGGE) and pyrosequencing to characterise microbial diversity11 also failed to find any differences in alpha diversity with respect to FUT2 genotype. In addition, the same group found different results between DGGE, pyrosequencing and HITChip methods applied to the exact same cohort.10,11 After subsampling our dataset to this taxon, we failed to replicate any association with Bifidobacteriales alpha diversity. Our sequencing protocol was directed at V4 region of the 16S which a region known to have adequate amplification and assignment of members of Bifidobacteriaceae family.29-31 Thus, we believe that our results concerning Bifidobacteriaceaceare are robust.
We found that FUT2 genotype and inferred phenotype was not associated with microbial composition. This is different from previous studies which have generally been small in sample size.8-11 Using PCR-DGGE fingerprinting technology to characterise the microbiome, it has been shown that non-secretors have lower relative abundance and diversity of the bifidobacterial taxa.10,11 Wacklin et al. reported that this association was weak and pyrosequencing and HITChip characterization of the microbiota did not replicate all these findings in the same cohort.10,11 Another study has shown that FUT2 secretor status was associated with gut microbiota compositio;n9 however, only 47 individuals were studied including 29 with CD, which could indicate a confounding effect of disease phenotype rather than genotype.9 The association of FUT2 with intestinal microbial the composition and function in the cecum and sigmoid colon from 39 healthy subjects has been reported. However these authors only included OTUs observed in at least 60% of the cohort reducing the number of OTUs from 4074 to 419. Here, a q-value threshold of 0.25 for a part of their analysis was used rather than the conventional p-value threshold of 0.05, thus increasing type I error.8 For comparison, we applied the same strategy, but we could still not replicate previous results.8 Finally a small cohort of 35 individuals comprising 8 non-secretor individuals found no difference in alpha diversity and beta diversity but differences in Prevotellaceae, Paraprevotellaceae were observed.32 Finally, a more recent large cohort study also found that secretor statuses are not associated with stool microbiota composition in 1,500 twins.26 Our results are thus in accordance with this larger cohort study and suggest that studies with small sample size should be replicated in well powered studies.
FUT2 genotype/ inferred phenotype was not associated with imputed microbial function assessed using PICRUSt. A report from Tong et al. was the only previous study to evaluate the function of the human microbiota in the context of FUT2.8 FUT2 was shown to be associated with several KEGG pathways including carbohydrate and lipid metabolism, cofactors and vitamins metabolism and glycan biosynthesis and metabolism. In our dataset none of these pathways associations were replicated. The discordance between our report and that of Tong et al could be due to numerous technical differences: DNA extraction (PowerSoil DNA Isolation Kit versus QIAamp DNA Stool Mini Kit), sequencing technology (HiSeq versus MiSeq). We used the same primers to sequence the 16S gene; however, HiSeq allows the generation of a higher numbers of reads per run compared to MiSeq.30 Due to lower number of sequences generated by MiSeq we chose a rarefaction of 30,000 sequences per sample while Tong et al. used a rarefaction of 300,000 sequences per sample. This difference in rarefaction depth might affect the presence/absence of particular OTUs with more rare OTUs and related inferred functions detected. As PICRUSt software uses OTU composition to infer bacterial function, it might be sensitive to OTUs richness and prevalence resulting in different conclusion.33 A report by Langille et al. has demonstrated that only 105 16S sequences are required to accurately impute bacterial function.34 Thus we believe that our conclusions are robust even with a lower number of sequences per sample.
The major difference between our study and previous reports is the bio-specimen used to investigate intestinal microbial composition. Tong et al. used mucosal lavage collected from the cecum and sigmoid colon.8 Rash et al. used colonic sigmoid biopsies,9 while in our study faecal samples were used. The microbial composition is known to differ across luminal and fecal samples.35 Also, there is a decreasing gradient of fucose and ABH blood group expression from ileum to rectum.36,37 Thus the alpha (1,2) fucosylated components are likely decreased in stool as compared to sigmoid or cecal samples. In addition, age is an important factor that is associated with differences in microbial composition.38 The mean age of the GEM cohort (19.8 ± 7.7) was lower than previous studies8,10,11 and thus the microbiota composition and/or inferred function in our population might be different from these studies. Nevertheless, the overall microbial composition in our cohort was consistent with that described in microbiome studies2,6 (Supplementary Figure 2). Finally technical differences might explain some discrepancy such as DNA extraction, microbial identification method, primers, and regions of 16Sr RNA.39,40 Indeed, DNA extraction was different across all these study and include the FastDNAH SPIN Kit for Soil (MP Biomedicals),10,11 prep DNA/RNA Mini Kit (Qiagen),9 and PowerSoil DNA Isolation Kit (MO BIO Laboratories)8 while the QIAamp DNA Stool Mini Kit (Qiagen) was used in this study. These DNA extractions are known to result in differences in microbiome profiling even if the same sample was used for each extraction.39 In addition, Waclkin et al. use DGGE, pyrosequencing and HITChip methods, while Rausch et al. used pyrosequencing and Tong et al. used Illumina HiSeq 2000 while Miseq was used in the GEM project.41 All these methods make it difficult to directly compare the biological output of each study. To summarise, stool sampling, DNA extraction and the age of the cohort are the three most important factors which differ between this and previous studies which could explain the absence of a FUT2 association with microbial composition and inferred function observed in this study.
With 1,190 individuals, this study represents one of the largest human cohorts to assess the association of FUT2 genotype and inferred phenotype with the human gut microbiota. We found that FUT2 was nominally associated with numerous inferred microbial functions and with microbial composition but none of these associations survive correction of the p-value for multiple testing. In addition no clustering based on FUT2 genotype could be observed and the alpha diversity was similar across FUT2 genotype and secreting status. However dietary habits, variation in stool sampling, and DNA extraction might explain discrepancy with other studies. In conclusion we found that FUT2 was not associated with human microbiota. Genetic factors that are associated with microbial composition and function remain to be defined and replicated.7,27,28
Methods
Patient recruitment
Subjects were recruited as part of a prospective cohort study of healthy first degree relatives of CD subjects between 6 – 35 years of age from 2008 to 2013 (The Genetic, Environmental Microbial (GEM) project) (http://www.gemproject.ca/, Table 1, Supplementary Table 1). After consent, each subject provided stool and blood samples and completed a standardized questionnaire to exclude any history or symptoms of IBD or gastro-intestinal disease as defined by the clinical sub-committee of the GEM Project (Supplementary Note 1). Demographic information and environmental risk data were recorded (Supplementary Note 1, Supplementary Note 2, Table 1, and Supplementary Table 1). All subjects and/or their guardians gave written informed consent to participate in the study. Subjects were excluded if they had received antibiotic treatment within three months prior to their recruitment into the study. The study cohort comprised 1,190 asymptomatic individuals of European descent. The study was approved by the Mount Sinai Hospital Research Ethics Board (Toronto – Managing center) and each participating recruitment center.
Table 1.
Demographic and genetic data in unrelated individuals of European descent.
| | Total number of unrelated Subject | FUT2 genotype rs516246 | ||
|---|---|---|---|---|
| (918) | GG (244) | AG (436) | AA (238) | |
| Gender | F (501) | 119 | 246 | 136 |
| M (417) | 125 | 190 | 102 | |
| Age | Mean | 19.4 | 19.7 | 20.8 |
| St Dev | 8.0 | 7.4 | 7.8 | |
Genotyping of healthy first degree relatives by HumanCoreEXOME chip
Blood was collected from subjects and genomic deoxyribonucleic acid (DNA) extracted using the Gentra Puregene Blood Kit (Quiagen, CA, USA, catalog #158389). DNA samples were quantified by Nanodrop dilutions. DNA was then prepared at a final concentrations of 20 ng/µl and aliquoted into 96-well reaction plates. SNP genotyping was performed using the HumanCoreExome-24 v1.0 chip (Illumina, Inc. San Diego, CA). We restricted the analysis to subjects with self-declared Caucasian ethnicity due to the fact that rs601338 (W143X, G428A) is the most common cause of secretor status in European ancestry, while other polymorphisms may be responsible in other ethnicity as reviewed previously.42 Since, rs601338 is not genotyped in the HumanCoreEXOME chip we used rs516246, which is in strong linkage disequilibrium with rs601338 to infer secretor phenotype. According to the International HapMap Project (HapMap3 release 2, Northern and Western Ancestry population) these two SNPs are in perfect linkage disequilibrium with one another (r2 = 1.0).43 The minor allele (A) frequency was 49.9% and the genotype distribution were in Hardy-Weinberg equilibrium (p-value = 0.27). Individuals with AA genotype were defined as non-secretor while AG and GG were defined as secretor. Genotyping data and corresponding stool identifiers is available in Supplementary Table 11.
Taxonomic profiling of the gut microbiota
Stool sampling from 1,190 subjects (comprising 918 unrelated subjects) was performed as described previously.41 Briefly, stool samples were collected in FB Commode Specimen Collector (Fisher Scientific, Waltham, MA, catalog #23-038032), put into Polypropylene vials (Starplex Scientific Inc, Etobicoke, ON, catalog #V302-F) and kept in freezers, then shipped from study sites and stored at -80°C. Fecal bacterial DNA was extracted using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany, catalog #51504) with slight modifications which included physical disruption of the bacterial cell wall using ceramic beads. The V4 hypervariable region of bacterial 16S ribosomal RNA (16S) was sequenced in paired-end mode (2 × 150 base pair) on MiSeq platform (Illumina Inc., San Diego, CA, USA).30 The resulting paired reads were assembled using PANDAseq v 2.7 to generate an amplicon size of 250 base pairs.44 To test for inter and intra-sequencing run variation, a subset of 45 samples belonging to 12 different subjects were replicated (2–8 replicates depending on subjects) for quality control (Supplementary Table 2). Sequencing across multiple MiSeq runs was consistent (Supplementary Table 2). Assembled reads were demultiplexed and processed by the quantitative insights into microbial ecology (QIIME v1.8.0) pipeline using the default parameters.45 Chimeric sequences were identified de novo and reference based and then removed using usearch61.46 The non-chimeric sequences were then clustered into operational taxonomic units (OTUs) at 97.0% sequence similarity using a closed reference-based picking approach with UCLUST software against Greengenes database 13_8 of bacterial 16S sequences.47 All samples had a minimum of 30,000 reads after quality filtering. Alpha diversity was determined after rarefication at a depth of 30,000 reads per sample using Chao1, Simpson, Shannon and observed species indices.48
Power calculation
Power calculation was performed using QUANTO v1.2.4.49 Power calculation was defined as the power at two-sided alpha of 5% to detect difference between genotypes using a recessive model, the observed coefficients of determination (R2), and observed minor allele frequency in 918 unrelated individuals.
Association of gut microbiota with FUT2 genotype and inferred phenotype
Since including related subjects in genetic association analysis inflates type 1 error,50 we choose to restrict the first part of this study to 918 unrelated individuals of European descent. To assess if FUT2 genotype/ inferred phenotype is associated with the microbial composition we applied the same methods as Tong et al.8 which showed that rs516246 genotype (AG and GG vs AA) is associated with alpha diversity measured by Chao1 index and beta diversity estimated by using Bray-Curtis and unweighted UniFrac distances between samples.8,51 In our cohort, we found that rs516246 genotype (AG and GG vs AA) have a similar alpha diversity. In addition, because Bifidobacteriales taxon was also previously associated with alpha diversity as determined by DGGE,10 we restricted our microbiome dataset and investigated this taxon specifically. Bacterial community grouping was assessed using the Adonis test for 10,000 permutations on the first five dimensions of the principal coordinates analysis (PCoA).52 Pair-wise comparisons between FUT2 secretor versus non secretor was assessed using the non-parametric Kruskal-Wallis test to identify differentially abundant phylum taxa and OTU levels of taxonomy.
Analysis of bacterial taxa and OTUs were performed using two different filtering. The first analysis applied the filtering used in Tong et al.8 Briefly, to analyze at lower taxonomic levels, we filtered out low-abundant OTUs based on the criteria of minimum total observation count of 30 across all samples and being observed in at least 60% of the samples, reducing the total number of OTUs from 12,863 to 396.8 A second approach was then applied because filtering OTU that are observed in at least 60% of the samples had consequences on characterization of each individual's microbiota (Procrustes of OTU table with no filtering against OTU table filtered based on 60% of OTUs prevalence, M2 = 0.01, p-value <10−6) with some individuals having a biased microbiota profiles after applying this filtering (Supplementary Figure 1–2). Such approaches also decrease the number of comparison made. For such reason, we decided to apply less stringent criteria to filter OTUs that are observed in at least 5% of the samples and a minimum read count of 30 across all samples, leaving 4,353 OTUs (R software v2.14.1; http://CRAN.R-project.org). We then analysed bacterial composition based on OTUs grouped within the same taxonomic assignment at each level of bacterial taxonomy. A Bonferroni corrected p-value threshold of 0.05 was considered as significant (while 396 comparisons were performed at OTU level of bacterial taxonomy and 166 bacterial taxa). Raw p-values are reported. All genotype classes were compared in addition to secretor status and the p-value were corrected for intra-group compared.
Imputed function of the gut microbiota
The function of the fecal microbial communities was imputed using PICRUSt V0.1.34 Briefly, The OTU table was used as the input file for metagenome imputation after a rarefaction step to 30,000 sequences per sample. The pre-calculated table of gene counts was used to identify the gene counts in the organisms present in the microbiome.53 The Kyoto Encyclopedia of Genes and Genomes (KEGG)54 and clusters of orthologous groups (COG)55 databases were used to identify the gene families. The metagenomic prediction had a relatively low weighted nearest sequenced taxon index (0.07 ±. 0.017). Pair-wise comparisons between secretor/non-secretor were assessed using the non-parametric Kruskal-Wallis test to identify differentially abundant inferred functions from KEGG and COG pathways. A Bonferroni correction for multiple testing was applied and significant association was considered below a corrected p-value threshold of 5%. All Procrust analyses were based on Bray-Curtis distance.56 Monte Carlo simulation was performed using 10,000 permutations on the first five dimensions of the PCoA.
Funding Statement
This study was supported by grants from Crohn's and Colitis Canada, Canadian Institutes of Health Research (CIHR) Grant #CMF108031 and the Helmsley Charitable Trust. Williams Turpin is a recipient of a Postdoctoral Fellowship Research Award from the CIHR Fellowship/ Canadian Association of Gastroenterology (CAG)/ Ferring Pharmaceuticals Inc. and a fellowship from the Department of Medicine, Mount Sinai Hospital, Toronto. Mark Silverberg is supported in part by the Gale and Graham Wright Chair in Digestive Diseases. David Kevans is a recipient of a CIHR /CAG/Abbvie IBD Fellowship Award. Osvaldo Espin-Garcia is a recipient of a CIHR STAGE fellowship. Konstantin Shestopaloff is a recipient of an Ontario Graduate Scholarship.
Abbreviations
- CD
- Crohn's disease
- COG
- clusters of orthologous groups
- DGGE
- Denaturing Gradient Gel Elecrophoresis
- IBD
- inflammatory bowel disease
- FUT2
- alpha (1,2)-fucosyltransferase 2
- GEM
- The Genetic, Environmental Microbial project
- KEGG
- The Kyoto Encyclopedia of Genes and Genomes
- OTU
- operational taxonomic units (OTUs)
- PANDAseq
- PAired-eND Assembler for DNA sequences
- PICRUSt
- Phylogenetic Investigation of Communities by Reconstruction of Unobserved States
- PCoA
- Principal Coordinates Analysis
- QIIME
- Quantitative Insights Into Microbial Ecology
- SNP
- single nucleotide polymorphism
Declarations
Ethics approval and consent to participate
All subjects and/or their guardians gave written informed consent to participate in the study. The study was approved by the Mount Sinai Hospital Research Ethics Board (Toronto – Managing center) and local centers (07-0322-E).
Consent for publication
Not applicable.
Competing interests
All authors disclose no potential conflicts (financial, professional, or personal) that are relevant to the manuscript.
Authors in the CCC IBD GEM project research team
The GEM Project Research Consortium is composed of: Maria Abreu, Paul Beck, Charles Bernstein, Kenneth Croitoru, Leo Dieleman, Brian Feagan, Anne Griffiths, David Guttman, Kevan Jacobson, Gilaad Kaplan, Denis O. Krause*, Karen Madsen, John Marshall, Paul Moayyedi, Mark Ropeleski, Ernest Seidman, Mark Silverberg, Scott Snapper, Andy Stadnyk, Hilary Steinhart, Michael Surette, Dan Turner, Tom Walters, Bruce Vallance, Guy Aumais, Alain Bitton, Maria Cino, Jeff Critch, Lee Denson, Colette Deslandres, Wael El-Matary, Hans Herfarth, Peter Higgins, Hien Huynh, Jeff Hyams, David Mack, Jerry McGrath, Anthony Otley, and Remo Panancionne. (* deceased)
Availability of data and materials
16S sequences are available the study accession number PRJEB14839 (NCBI BioProject database).
Acknowledgments
We thank the members of the GEM Global Project Office: Cristina Bravi, Deema Couchman, Nisha Ganeswaren, Alexandra Keludjian, Kevin Ow, Rachel Caplan, Melissa Greaves, Amy Craig-Neil, Ana Olteanu, Nellie Allam, Amanda Garrioch, Denis Ng, Venus Onay and Isabelle Yeadon for administrative support. We thank Dennis Cvitkovitch for his helpful scientific discussion of the project.
Author contributions
Contributed equally – ADP and KC; Jointly supervised research- ADP and KC; Conceived and designed the experiments- WT, MSS, ADP, KC and The GEM Project Steering Committee; Performed the experiments- WT, WX, LX, KC, Performed statistical analysis; – WT, OEG, WX, LX, ADP; Analyzed the data – WT, LB, MSS, MIS, WX, GMH, DK, KS, OEG, DSG, LX, ADP, KC; Contributed reagents/materials/analysis tools and significant subject recruitment – WX, LX, ADP, AG, RP, AO, and the GEM Project Consortium; Wrote the paper- WT, ADP, KC.
References
Articles from Gut Microbes are provided here courtesy of Taylor & Francis
Full text links
Read article at publisher's site: https://doi.org/10.1080/19490976.2018.1445956
Read article for free, from open access legal sources, via Unpaywall:
https://www.tandfonline.com/doi/pdf/10.1080/19490976.2018.1445956?needAccess=true
Citations & impact
Impact metrics
Citations of article over time
Alternative metrics
Smart citations by scite.ai
Explore citation contexts and check if this article has been
supported or disputed.
https://scite.ai/reports/10.1080/19490976.2018.1445956
Article citations
Patterns of Human Milk Oligosaccharides in Mature Milk Are Associated with Certain Gut Microbiota in Infants.
Nutrients, 16(9):1287, 25 Apr 2024
Cited by: 0 articles | PMID: 38732534 | PMCID: PMC11085179
Relationship between ABO Blood Group Alleles and Pancreatic Cancer Is Modulated by Secretor (FUT2) Genotype, but Not Lewis Antigen (FUT3) Genotype.
Cancer Epidemiol Biomarkers Prev, 32(9):1242-1248, 01 Sep 2023
Cited by: 0 articles | PMID: 37342060 | PMCID: PMC10527950
Microbiota-induced regulatory T cells associate with FUT2-dependent susceptibility to rotavirus gastroenteritis.
Front Microbiol, 14:1123803, 27 Feb 2023
Cited by: 0 articles | PMID: 36922975 | PMCID: PMC10008897
Colon mucus in colorectal neoplasia and beyond.
World J Gastroenterol, 28(32):4475-4492, 01 Aug 2022
Cited by: 4 articles | PMID: 36157924 | PMCID: PMC9476883
Review
Free full text in Europe PMCMicrobiome-associated human genetic variants impact phenome-wide disease risk.
Proc Natl Acad Sci U S A, 119(26):e2200551119, 24 Jun 2022
Cited by: 12 articles | PMID: 35749358 | PMCID: PMC9245617
Go to all (26) article citations
Data
Data behind the article
This data has been text mined from the article, or deposited into data resources.
BioStudies: supplemental material and supporting data
SNPs (2)
- (3 citations) dbSNP - rs601338
- (3 citations) dbSNP - rs516246
Similar Articles
To arrive at the top five similar articles we use a word-weighted algorithm to compare words from the Title and Abstract of each citation.
Faecal microbiota composition in adults is associated with the FUT2 gene determining the secretor status.
PLoS One, 9(4):e94863, 14 Apr 2014
Cited by: 96 articles | PMID: 24733310 | PMCID: PMC3986271
Gut Microbiome Composition and Metabolic Capacity Differ by FUT2 Secretor Status in Exclusively Breastfed Infants.
Nutrients, 15(2):471, 16 Jan 2023
Cited by: 10 articles | PMID: 36678342 | PMCID: PMC9866411
Relevance of secretor status genotype and microbiota composition in susceptibility to rotavirus and norovirus infections in humans.
Sci Rep, 7:45559, 30 Mar 2017
Cited by: 52 articles | PMID: 28358023 | PMCID: PMC5372083
Fucosyltransferase 2: a genetic risk factor for primary sclerosing cholangitis and Crohn's disease--a comprehensive review.
Clin Rev Allergy Immunol, 48(2-3):182-191, 01 Jun 2015
Cited by: 31 articles | PMID: 24828903
Review
Alterations in fecal microbiota composition by probiotic supplementation in healthy adults: a systematic review of randomized controlled trials.
Genome Med, 8(1):52, 10 May 2016
Cited by: 233 articles | PMID: 27159972 | PMCID: PMC4862129
Review
Free full text in Europe PMC
Funding
Funders who supported this work.
CIHR (1)
Grant ID: CMF108031
Government of Canada | Canadian Institutes of Health Research (CIHR) (1)
Grant ID: #CMF108031
