Alteration in Urease-producing Bacteria in the Gut Microbiomes of Patients with Inflammatory Bowel Diseases

Abstract

Background and Aims

Bacterial urease is a major virulence factor of human pathogens, and murine models have shown that it can contribute to the pathogenesis of inflammatory bowel diseases [IBD].

Methods

The distribution of urease-producing bacteria in IBD was assessed using public faecal metagenomic data from various cohorts, including non-IBD controls [n = 55], patients with Crohn’s disease [n = 291] or ulcerative colitis [n = 214], and patients with a pouch [n = 53]. The ureA gene and the taxonomic markers gyrA, rpoB, and recA were used to estimate the percentage of urease producers in each sample.

Results

Levels of urease producers in patients with IBD and non-IBD controls were comparable. In non-IBD controls and most IBD patients, urease producers were primarily acetate-producing genera such as Blautia and Ruminococcus. A shift in the type of the dominant urease producers towards Proteobacteria and Bacilli was observed in a subset of all IBD subtypes, which correlated with faecal calprotectin levels in one cohort. Some patients with IBD had no detectable urease producers. In patients with a pouch, the probiotic-associated species Streptococcus thermophilus was more common as a main urease producer than in other IBD phenotypes, and it generally did not co-occur with other Bacilli or with Proteobacteria.

Conclusions

Unlike all non-IBD controls, patients with IBD often showed a shift towards Bacilli or Proteobacteria or a complete loss of urease production. Probiotics containing the species S. thermophilus may have a protective effect against colonisation by undesirable urease-producing bacteria in a subset of patients with a pouch.

1. Introduction

Bacterial ureases catalyse the breakdown of urea to ammonia and carbonic acid, resulting in a pH increase.¹ Ureases protect the bacteria from acidity, either self-created [especially acidity caused by acetate- or lactate-producers] or host-produced [i.e., stomach derived],^2–4 and can also provide the capacity to use host urea as a nitrogen source.⁵

Although urease-producing bacteria are sometimes associated with positive health effects,^2,6,7 bacterial urease is primarily considered a major virulence factor found in multiple human pathogens, like Helicobacter pylori and Proteus mirabilis.⁸

Inflammatory bowel diseases [IBD], including Crohn’s disease [CD] and ulcerative colitis [UC], often display alterations of the gut microbiome^9,10 with unique microbial signatures.¹¹ Approximately a quarter of patients with UC may undergo total large bowel resection [proctocolectomy] with reconstruction of intestinal continuity by an ileo-anal pouch created from normal small bowel.¹² The majority of these patients develop inflammation of the pouch [pouchitis].^13–16

Urease-producing species of bacteria are a feature of IBD dysbiosis^17–19 and urease production specifically contributed to the pathogenesis of experimental colitis.²⁰ Furthermore, a recent study showed that P. mirabilis is significantly more prevalent and abundant in patients with CD than in controls and suggested that higher levels of urease-producing bacteria are a general feature of IBD.²¹ Assessing the role of urease in patients with IBD is therefore important in understanding disease pathogenesis and risk stratification.

Here, we critically examined the distribution of urease-producing bacteria in CD, UC, and pouchitis, using public shotgun metagenomic data from multiple cohorts. We tested the hypothesis that the distribution of urease-producers differs between patients with IBD and non-IBD controls, and analysed their diversity. We show that although the majority of patients with IBD have a urease-producing signature comparable to non-IBD controls, a substantial subset of patients have either no urease-producers at all or a faecal microbiome dominated by potentially harmful urease-producing bacteria. Such urease dysbiosis coincides with worse intestinal inflammation, reflected by faecal calprotectin levels. This may assist in patient stratification and precision medicine. We also show that in patients with a pouch, Streptococcus thermophilus is often a major urease producer, and suggest that probiotics containing S. thermophilus may have a protective effect against colonisation by undesirable urease-producing bacteria.

2. Methods

2.1. Data collection

We chose the largest publicly available IBD cohorts for which high quality [>2 m sequences per sample] shotgun metagenomic data were available.

The paired end raw reads of the metagenomes of CD, UC, and non-IBD controls from the American PRISM cohort and the Dutch LLDeep and NLIBD cohorts were obtained from NCBI SRA with BioProject number PRJNA400072.²² The metadata for these cohorts contained data regarding antibiotic use 6 months preceding sampling, and the location of CD.

The paired end raw reads of the metagenomes of CD and UC patients from the Dutch 1000IBD project were obtained from EGA under accession number EGAS00001002702.²³ The metadata for these cohorts did not contain data regarding antibiotic use or the location of CD, but did contain clinical indices—Simple Clinical Colitis Activity Index [SCCAI] for UC and Harvey‐Bradshaw Index [HBI] for CD. In samples where the initial diagnosis did not match the provided index [UC samples with HBI values, etc.], disease phenotype was corrected so it matched the clinical index.

The paired end raw reads of the metagenomes of patients with a pouch were obtained from NCBI SRA under BioProject numbers PRJNA524170²⁴ and PRJNA637365.²⁵

Pouch phenotype was defined according to accepted clinical characteristics.¹⁵ A normal pouch was defined as having no clinical flares of pouchitis in the 2 years before the sample was taken, and no antibiotic or anti-inflammatory treatment [46 samples from 22 patients]. Recurrent acute pouchitis was diagnosed when episodes of acute pouchitis were followed by normal pouch function [20 samples from 15 patients]. Chronic pouchitis [CP] was defined as persistence of symptoms for more than 3 months or the need for chronic antibiotic therapy. Crohn’s-like disease of the pouch [CLDP] was diagnosed when having one or more of the following: pouch-related fistula, inflammation of the afferent limb or more proximal small bowel segment[s], and fibrostenotic disease of the pouch.²⁶ Data regarding antibiotic use in the month preceding sampling was provided for all the samples from patients with a pouch.

Samples from patients with CP [36 samples from 18 patients] and from patients with CLDP [19 samples from 10 patients] were combined into one group [CP/CLDP].

2.2. Bioinformatic analysis

2.2.1. Calculation of the percentage of urease-producers in the faecal microbiomes

UreA is a single copy gene; gyrA, rpoB, and recA are considered universal single copy genes, present in all bacteria, and can be used to calculate the percentage of urease producers in a sample.²⁷ This was done by dividing the total ureA coverage found in the sample [mean coverage of a nucleotide in the ureA gene, across all ureA genes found in the sample], by the mean coverage of the three marker genes in that sample.

2.2.2. Relative abundance calculation

The relative abundance of urease-producing bacteria from a given taxon in a sample was calculated by the coverage of ureA of this taxon in the sample divided by the total ureA coverage in the sample from all taxa.

2.2.3. Generating ureA and taxonomic marker gene databases

All nucleic acid and amino acid sequences of the genes ureA, gyrA, rpoB, and recA derived from genomes of bacterial isolates were downloaded from IMG JGI on March 10, 2021, based on their Pfam function, and integrated to local databases.^28,29 Each database was filtered for redundancy and incomplete gene sequences. The Pfam function for GyrA and its paralog ParC is the same, hence the gyrA database was additionally filtered by the gene name.

2.2.4. Sequence data quality control

Adapters and low-quality bases were removed using Trimmomatic with the default parameters.³⁰ FastQC was used to visually verify the quality before downstream analysis.³¹

2.2.5. Taxonomic assignment and quantification of ureA and the taxonomic marker genes

Since the metagenomic raw reads themselves are often too short for an exact identification of the gene by read mapping, we first constructed contigs of lengths ≥400 bp for each gene using an alignment-based gene centric assembler. Additionally, there are paralogous genes within a species with regions that may overlap with our genes of interest, and so not every read that maps to our custom database necessarily originates in a gene of interest.

We developed an algorithm to accurately identify and calculate the normalised abundance of the genes, which briefly, for ureA and for each marker gene:

1. constructs contigs of lengths ≥400 bp using protein alignment-guided gene centric assembly and identifies the possible genera each contig can originate from;

2. identifies the genera that may contribute the relevant gene to the sample [presence/absence] by a criterion such that at least 70% of the length of the gene from these genera can be covered by the contigs from step 1;

3. merges the genera that could not be differentiated from one another unequivocally [i.e., a group of genera that pass the criterion in step 2, yet have a contig that has a good similarity to all the genera in the group];

4. calculates for each taxonomic unit [a genus or a merged group of genera] the average coverage of a nucleotide in the gene from this taxonomic unit [by the original reads that were used to construct the matched contigs] as its abundance.

A more detailed explanation of all these steps can be found below.

2.2.6. Construction of ureA and taxonomic marker gene contigs and calculation of their average coverage

Since the metagenomic raw reads themselves are too short for an exact identification of the gene, contigs of lengths ≥400 bp were constructed as follows: both paired end raw reads of each sample were aligned against the UreA and the marker genes’ amino acid sequences using DIAMOND v0.9.24.125.³² The aligned reads were assembled using the MEGAN protein alignment-guided gene-centric assembler; the.daa files from both ends were converted to a single.rma file using MEGAN daa2rma, and the MEGAN gc-assembler was used to assemble the contigs, using the parameters -fun none -id ALL -len 400 -c false.³³ MEGAN gc-assembler works by looking at the aligned part in all the DIAMOND matches and at all the possible overlaps between these aligned parts, and also calculates for each contig it creates its average coverage—the average number of aligned parts that cover each nucleotide in the contig [by constructing a graph where the vertices are the aligned parts of the reads for each DIAMOND match and the weighted edges are the overlaps—and then sequentially removing paths of maximum weight from the graph as the output contigs].

—

2.2.7. Taxonomic assignment of ureA and taxonomic marker genes

The ureA and the taxonomic marker gene contigs were aligned against the nucleic acid sequences of the relevant genes using blastn v2.7.1 with the parameters -qcov_hsp_perc 90 and -perc_identity 94.³⁴ In case of multiple matches—all the matches with bit scores—≥99% of the best matches were considered as possible candidates for the taxonomic assignment.

The ureA contigs were long enough for an exact taxonomic assignment at the genus level in most cases [all blastn results were of the same genus]. For cases where a single species of a genus was presented in the dataset, the species designation was provided when identity was greater than 97%.

Since some of the genes have close paralogs [like ureG and some metallo-dependent hydrolases for ureA or parC for gyrA], we considered a gene assigned to a given genus to be present in a sample only if >70% of the nucleotides in the matching genes of this genus were covered by the aligned contigs in the sample. This condition was determined by taking the start and end positions of all the corresponding alignments by blastn to the appropriate genes of this genus, and finding the nucleotides covered in these genes by a sweep line algorithm.

After finding all the possible genera that can contribute the gene of interest to the sample, we scanned the BLASTn matches twice—once to correct the taxonomic assignment, i.e., to merge genera that could not be differentiated from one another unequivocally, the case where some contigs have good BLASTn alignments to several genera that passed the 70% threshold—and once again to calculate the abundance of the gene from each taxonomic unit [genus or a merged group of genera]. The calculation was done by multiplying the average coverage of each contig from the taxonomic unit by its length [to get the sum of coverages of the nucleotides in the contig], summing these values for all the contigs with the appropriate taxonomic assignment [to get the sum of coverages of the nucleotides in the complete gene], and then dividing by the complete gene length [to get the average coverage of a nucleotide in the gene again].

The only genera with ureA genes that could not be differentiated from one another unequivocally were Ruminococcus, Blautia, and Fusicatenibacter [Fusicatenibacter saccharivorans specifically], hence these genera were combined to a single operational taxonomic unit.

2.2.8. Quality filtering

Samples where the mean coverage of the taxonomic marker genes across all bacteria was less than 40X were removed from downstream analysis, since in those samples we are likely to be unable to detect urease-producers that constitute below 2.5% of the total community [1/40].

2.2.9. Taxonomic assignment of streptococci species

The species of Streptococcus that are urease-positive were determined according to the UreA sequences. These species are Streptococcus thermophilus, Streptococcus salivarius, Streptococcus vestibularis, and Streptococcus infantarius. To differentiate the probiotic S. thermophilus species from the other urease-positive ones, the total coverage of streptococcal ureases in each sample was split to S. thermophilus and other [‘non-thermophilus’] streptococci by calculating the ratio of S. thermophilus to the other urease-positive Streptococcus spp. found in the sample using MetaPhlAn2, as mentioned below.³⁵

2.2.10. Microbial dysbiosis index calculation

Microbial dysbiosis index [MDI] is based on two groups of taxa that were previously found to be over- or under-represented in patients with IBD compared with non-IBD controls.³⁶ Taxa that are increased in IBD belong to the families Enterobacteriaceae, Pasteurellaceae, Neisseriaceae, Fusobacteriaceae, Veillonellaceae, and the genus Gemella. Taxa that are decreased in IBD are members of Erysipelotrichaceae, Bifidobacteriaceae, Bacteroidales, and Clostridiales excluding Veillonellaceae. MDI is calculated using the following formula:

$logrelative abundance of taxa increased in IBDrelative abundance of taxa decreased in IBD.$⁠.

There were a few cases of samples with no taxa that are increased in IBD. In this case, we took the lowest relative abundance of a species in the sample, which estimates the minimal detectable threshold, as the abundance of these taxa.

2.2.11. Taxonomic profiling of the metagenomes

Taxonomic profiling was done using MetaPhlAn2 v2.6.0³⁵ which classifies metagenomic reads by mapping to a database of clade-specific marker genes. MetaPhlAn2 was run with the following parameters: --tax_lev ‘s’ [classify taxonomy to species level], --ignore_virus, and --ignore_eukaryotes to ignore viral and eukaryote reads, respectively.

2.3. Statistical analysis

All the statistical analyses were performed using R.

The difference in the percentage of urease-producers between different phenotypes [main IBD phenotypes or patients with a pouch] was tested using Kruskal Wallis test. The difference between each inflammatory phenotype and not inflamed phenotypes [non-IBD controls or patients with a normal pouch] was tested using the unpaired Wilcoxon test, with multiple tests correction by the FDR method.³⁷ Both tests were conducted using the stat_compare_means function from the ggpubr library.

The dependency between the disease phenotypes and the main urease producers in a sample [grouping by dominant taxonomic group] was tested using the Fisher’s exact test for count data, using the fisher.test function. Post hoc pairwise analyses were conducted with the Fisher’s exact test and the FDR method for multiple testing correction, using the fisher.multcomp function from the RVAideMemoire library.

To account for longitudinal sampling in patients with a pouch, a linear model was built using lme function from the nlme library, with the patient ID set as a random variable. The percentage of urease producers was transformed using the logit function from the car library and served as a fixed variable. The anova function was used on the linear model to obtain the p-value. Post-hoc pairwise analyses were performed using lsmeans function from the lsmeans library, with FDR correction for multiple testing. When performing Fisher’s exact test for count data in patients with a pouch, we randomly chose one sample from the main phenotype of each patient.

3. Results

3.1. The relative abundance of urease producers in the fecal microbiome in patients with IBD and non-IBD controls is comparable

First, we identified urease-producing bacteria in publicly available shotgun metagenomes of fecal samples from patients with CD, UC and non-IBD controls, from four cohorts [Table 1], by identifying genes encoding ureases and then quantifying their presence according to normalized read counts [see Methods].

We then tested whether there is a difference in the percentage of urease producers [among all bacteria present in the fecal community] between patients with CD, UC and non-IBD controls [Figure 1].

The median abundance of urease producers in patients with IBD and non-IBD controls was comparable. However, in 4.82% of metagenomes from patients with IBD no urease producers were present [0% abundance]. In those metagenomes having urease producers [>0% abundance], there was a trend of higher ratios of urease producers, i.e. urease producers tend to be fairly abundant members of those microbiomes.

3.2. IBD patients can have different urease-producing genera than those of non-IBD controls

Next, we analysed the taxonomic identity of urease producers in the metagenomes. Urease producers in non-IBD controls mostly belonged to the orders Clostridiales and Bacteroidales, across all the samples [Figure 2].

In the metagenomes of patients with IBD most samples had dominant urease producers belonging to the same orders as non-IBD controls [mostly members of Clostridiales belonging to the genera Blautia and Ruminococcus], yet Bacilli or Proteobacteria [e.g. Klebsiella, Haemophilus and Morganella] were more abundant [present in 38.43% of patients with IBD vs 25.45% in non-IBD controls, p-value = 0.077, Fisher exact test, Figure 3], and in some metagenomes, bacteria belonging to these taxonomic groups were the dominant urease producers. The main urease producers among the Bacilli were oral streptococci and the probiotic/food-derived species Streptococcus thermophilus. Notably, some metagenomes had no urease producers at all [Figure 3].

Among CD subtypes in the American cohort, which had disease location data, the prevalence of samples with no urease producers was highest when the ileum was involved [ileal or ileocolonic CD], and lowest in colonic CD. Notably, the identity of urease producers was meaningful, as urease producing Proteobacteria were markedly higher in patients with an ileocolonic disease [Supplementary Figure 1], available as Supplementary data at ECCO-JCC online, and generally much more prevalent in American patients with CD compared with Dutch ones 8 out of the 66 American samples had Proteobacteria as the main urease producers, compared with 3 of the Dutch ones [p-value = 0.0004, Fisher’s exact test, Supplementary Figure 2], available as Supplementary data at ECCO-JCC online]. This further corresponded to a trend, in which Proteobacteria in general were more abundant in American patients with IBD, constituting an average of 7.9% of the gut community in samples from American, and 1.5% in Dutch patients, respectively [see supplementary Figures 3, 5], available as Supplementary data at ECCO-JCC online]. Interestingly, no urease-encoding E. coli strains could be detected in any of the samples, even though 107 of the samples had E. coli strains in >1% relative abundance according to the taxonomic profiling [see Methods]. In subsequent analysis we divided the metagenomes to groups according to their dominant urease producers [i.e. the urease producers that constitute more than 50% of all the urease producers in the sample] – Clostridiales and Bacteroidales [which were the main urease producers in non-IBD controls], Bacilli and Proteobacteria [the atypical urease producers that were more prevalent in patients with IBD], probiotics-associated species [S. thermophilus and Bifidobacterium longum] or no urease producers at all. Only 9 out of the 560 samples could not be classified to any of the groups [Table 2].

Next, we excluded from further analyses the probiotic-dominated samples, since the presence of these species is probably due to dietary changes or supplement ingestion rather than stable colonization, as well as the ‘no dominant group’ due to the extremely small number of samples having this profile.

An association between the group of main urease producers and IBD compared with non-IBD controls was observed [Table 2, Fisher’s exact test, p = 0.0142], where microbiomes of all non-IBD controls were dominated by Bacteroidales and Clostridiales, whereas microbiomes of patients with IBD could also be dominated by other groups. Post hoc pair-wise analyses did not detect statistically significant differences between phenotypes [Supplementary Table 1], available as Supplementary data at ECCO-JCC online]. CD location was not significantly associated with any group of urease producers in the American cohort, though this may be due to the small sample size.

The above-mentioned trends differed between the American and the Dutch cohorts, and the association between IBD and Bacilli and Proteobacteria-dominated urease producers was prominent in the American cohort [Fisher’s exact test, p-value = 0.0045, see also Supplementary Table 2], available as Supplementary data at ECCO-JCC online], but did not reach statistical significance in the Dutch cohorts [Fisher’s exact test, p-value = 0.7047,]. This may be explained by the lower overall IBD-associated dysbiosis, reflected in the microbial dysbiosis index [MDI^36] in Dutch compared with American patients with IBD [median MDI of -3.2 vs -2.66; Wilcoxon test, p = 0.0001, Supplementary Figures 4, 5], available as Supplementary data at ECCO-JCC online].

To test for a possible bias caused by antibiotic use, we performed the analysis on the American cohort after excluding samples with antibiotic use in the 6 month prior to sampling or with no data regarding antibiotic use from the analysis [21 out of the 148 samples]. Reassuringly, the association remained statistically significant when excluding those samples [Fisher’s exact test, p-value = 0.018].

3.3. The identity of urease producers in patients with IBD is associated with inflammation

To test whether changes in urease producers [either Bacilli / Proteobacteria dominance, or complete loss of urease producing bacteria] correlated with intestinal inflammation, we analysed metagenomes [CD and UC] for which we also had corresponding data regarding the fecal inflammatory marker calprotectin, and in the Dutch 1000IBD cohort we analysed metagenomes for which we also had a clinical index - SCCAI for UC samples or HBI for CD samples.

Calprotectin levels were classified into low <50 mcg/g, intermediate-50–200, and high >200.³⁸ Overall, no correlation was observed [Fisher’s exact test, p = 0.17]. When separately analysing data from the American cohort, which had higher dysbiosis [see above], calprotectin levels were associated with the group of the main urease producers [Table 3; Fisher’s exact test, p = 0.0063]. Notably, all urease-negative metagenomes in the American cohort came from patients with high calprotectin levels [Table 3]. The difference between urease-negative metagenomes and Bacteroidales / Clostridiales dominated urease producers reached a borderline statistical significance in the American Cohort when comparing low vs high calprotectin groups in the pairwise post-hoc analysis after correction by FDR [Fisher’s exact test, p = 0.0901, see Supplementary Table 3], available as Supplementary data at ECCO-JCC online]. The association between calprotectin levels and the main urease producers was also significant when including only samples that originated from patients who were not treated with antibiotics [Fisher’s exact test, p = 0.0366].

When analysing data from patients with UC from the Dutch 1000IBD cohort, SCCAI index was classified into 2 bins - low ≤4 and high ≥5.^39,40 Overall, no correlation was observed between the dominating urease producers and the SCCAI index [Fisher’s exact test, p = 0.3], probably due to the small number of samples in the high SCCAI bin [n = 15].

When analysing data from patients with CD from the Dutch 1000IBD cohort, the HBI was classified into 2 bins - low ≤4 and high ≥5.^41,42 A significant association was observed between the dominant urease producers and the HBI [Fisher’s exact test, p = 0.0179, Table 4]. The difference between metagenomes with Bacilli / Proteobacteria-dominated urease producers and Bacteroidales / Clostridiales-dominated urease producers across the groups [of high vs low HBI] was statistically significant in the pairwise post-hoc analysis after correction by FDR [Fisher’s exact test, p = 0.0438, see Supplementary Table 4], available as Supplementary data at ECCO-JCC online].

When combining both CD and UC [each with its respective clinical index bin] the association was even stronger [Fisher’s exact test, p = 0.0086], and the difference between metagenomes with Bacilli / Proteobacteria dominated urease producers and Bacteroidales / Clostridiales-dominated urease producers across the groups [of high vs low index] was also statistically significant in the pairwise post-hoc analysis after correction by FDR [Fisher’s exact test, p = 0.0305].

3.4. Urease dysbiosis corresponds to low bacterial diversity

Given that lack of the typical urease producers corresponded to higher calprotectin levels and disease scores, we hypothesized that such urease dysbiosis may reflect a more general imbalance in those patients. We therefore tested whether different urease production patterns correspond to different bacterial diversity. Samples that had typical urease producers had the highest average diversity, while the probiotics-dominated, Proteobacteria/Bacilli-dominated and no urease producers all had much lower diversity [Figure 4]. This confirms that urease dysbiosis tends represents one fact of the disturbed microbiota of those patients.

3.5. Low levels of urease producers in patients with chronic pouchitis

Our next step was to focus on patients with UC who underwent pouch surgery. Those patients represent different features of UC and CD, and when pouch inflammation develops [pouchitis] it is the most antibiotic –responsive IBD. We analysed the metagenomes of fecal samples of different pouch clinical phenotypes: normal [non-inflamed] pouches, recurrent acute pouchitis, chronic pouchitis and Crohn’s-like disease of the pouch – CLDP [henceforth grouped as CP/CLDP]. In this dataset, most patients were longitudinally sampled [Table 5]. Some patients changed their phenotype during follow up – one normal patient progressed to CP [p67], four patients progressed from normal to recurrent acute pouchitis [p63, p64, p65 and p78], and five patients progressed from recurrent acute pouchitis to CP [p29, p50, p68, p70 and p72].

First, we tested whether there is a difference in the percentage of urease producers between the different pouch phenotypes, based on all samples [regardless of the patient identity of each sample]. Samples from CP/CLDP had a lower percentage of urease-producing bacteria compared with those with a normal pouch and recurrent acute pouchitis [Figure 5].

To account for longitudinal sampling, we constructed a linear model setting patient identity as a random effect. The overall difference between the pouch phenotypes remained significant in this model [p-value = 0.0282], and the difference between CP/CLDP and recurrent acute pouchitis was significant in the post-hoc analyses [least square means, FDR adjusted p-value = 0.0236]. Different pouch phenotypes differed in their urease producers even when excluding the samples taken from patients who took antibiotics within the month prior to sampling [p-value = 0.0485], and the difference between CP/CLDP and recurrent acute had a borderline significance in the post-hoc analyses [least square means, FDR adjusted p-value = 0.0576].

3.6. Diversity of urease producers in patients with a pouch

Next, we analysed the taxonomic identity of urease producers in the metagenomes of patients with a pouch. In patients with a normal pouch, the main urease producers were Blautia, Ruminococcus, S. thermophilus or other streptococci. Two samples had Megamonas as a main urease producer, and some samples had no urease producers at all, similar to what we observed for patients with CD and UC, above [Figure 6].

In general, in patients with pouchitis, the main urease producers were similar to those observed in the normal pouches. Nevertheless, in patients with the more severe phenotypes [CP/CLDP], the prevalence of samples with no urease producers was higher. Curiously, one patient in this group had Bifidobacterium scardovii as a main urease producer [Figure 7], which is an oral species sometimes observed in urinary tract infections^43,44 which was absent from all other metagenomes in our cohorts. Another patient in this group had a urease-encoding E. coli as a main urease producer [>80% relative abundance among the urease producers, and 6.5% of the overall bacterial community, Figure 7]. However, no statistically significant association was found between clinical pouch phenotype and main urease producer.

The probiotic/food-associated species S. thermophilus was more common as a main urease producer in patients with a pouch than in other IBD patients. Interestingly, S. thermophilus generally did not co-exist with Proteobacteria, and tended to be the main urease producer in the samples in which it was present [Figures 5 and 6]; this trend was more prominent in patients with a pouch than in CD or UC; 20.66% of the samples from patients with a pouch had S. thermophilus as the dominant urease producers, vs 2.38% of the samples from CD and UC patients.

To test whether there is a connection between pouch inflammation and the changes in urease producers, we also separated the faecal samples to two subgroups based on their calprotectin level—≤200 mcg/g or >200 mcg/g—and observed no significant difference between these groups.

4. Discussion

Whereas urease production is an important trait of several human pathogens, in the gut the urease producers tend to be primarily acetate-producing members of the Clostridiales order such as Blautia and Ruminococcus which are considered to be beneficial to human health,⁴⁵ but not the putatively IBD-associated R. gnavus.⁴⁶ This is not surprising, since 20–25% of all the urea produced by the human body enters the intestinal tract,⁴⁷ providing urease producers a competitive advantage due to a better supply of nitrogen as well as resistance to acidity. The latter may be of specific importance as pH in the caecum is acidic—5.7. Similar to the colon, the healthy pouch is an acidic environment⁴⁸ and hence is expected to select for urease-producing bacteria. In contrast, the ileum has a pH of around 7.4, making the benefit of urease production smaller for bacteria that colonise that organ.

Urease production is often a strain-specific function of a species,^49,50 making it difficult to infer it based on 16S rRNA gene-based data; shotgun metagenomics is thus required for its accurate quantification in microbiomes. Indeed, although urease-encoding E. coli strains exist, we observed no such strains in UC or CD, and only a single occurrence of such a strain in a patient with pouchitis. Most IBD samples had the same commensal urease-producing bacteria as non-IBD controls. Analysis of shotgun faecal metagenomes allowed us to observe: in which IBD phenotypes urease producers have attained unusual dominance and constituted the majority of the bacterial population; and in which patients urease-negative metagenomes were observed. Thus, it is a shift in the type of the dominant urease producers rather than an increase in their proportion that constitutes a disease-associated state that one would label ‘dysbiosis’ in IBD.

The finding that patients with IBD, who have no urease producers in their intestinal metagenomes, have high calprotectin levels is intriguing and may be highly relevant to disease pathogenesis, as lack of urea-using bacteria when there is a constant supply of this nutrient is likely to be temporary. That lack of urease producers could be due to the high oxidative stress brought about by intestinal inflammation, which suppresses the growth of obligatory anaerobic urease-producing bacteria, especially Clostridia such as Ruminococcus or Blautia. Such urease-negative conditions may be especially susceptible to subsequent invasion and blooms of urease-producing Proteobacteria such as Klebsiella or Proteus [although the latter genus was not observed in the tested cohorts] which have a relatively high tolerance to reactive oxygen species while also being able to benefit from having urea as a nitrogen source.^5,51,52 Such ‘urease dysbiosis’, in which the normal urease producers are replaced by bacteria that are considered potentially harmful in the context of IBD, was observed in about 6.07% and 6.87% of UC and CD patients, respectively in the IBD cohorts analysed, but in none of the non-IBD controls.

Similar to what was observed in patients with CD, a substantial fraction of patients with a pouch had no urease producers, especially those with the more severe disease phenotypes [CP or CLDP]. Interestingly, in patients with a pouch where the dominant urease producer was S. thermophilus, a major constituent of both probiotic yoghurt and commercial higher dosage probiotics and often prescribed for patients with IBD [specifically those with a pouch],⁵³ no dominance of Proteobacteria or oral streptococcal species⁵⁴ was observed. This may suggest that when such probiotics are consumed in substantial amounts, they could potentially provide some colonisation resistance against harmful urease-producing bacteria.^21,55 This conjecture certainly requires further validation, but it could represent a means for identifying patients more likely to benefit from the use of probiotic products with high S. thermophilus contents, namely those patients who have low or no indigenous urease producers in their microbiomes. Moreover, the enzymatic activity of urease in faeces can be easily quantified using a sensitive colorimetric assay,⁵⁶ and commercial kits are available from multiple suppliers, making such a precision medicine approach relatively easy to implement in the clinic.

Acknowledgements

The authors thank Leah Reshef for helpful suggestions.

Metagenomic sequences from the American PRISM cohort and the Dutch LLDeep and NLIBD cohorts are available via SRA [https://www.ncbi.nlm.nih.gov/sra] with BioProject number PRJNA400072.²²

Metagenomic sequences from the Dutch 1000IBD project are available on the European Genome-Phenome Archive [https://www.ega-archine.org], accession no: EGAS00001002702.²³

Metagenomic sequences of patients with a pouch from our study are available via SRA [https://www.ncbi.nlm.nih.gov/sra] with BioProject numbers PRJNA524170²⁴ and PRJNA637365.²⁵

The scripts used for analysing the data in this article are available at [https://github.com/GophnaUrease/scripts].

Funding

This work was supported by a generous grant from the Leona M. and Harry B. Helmsley Charitable Trust, and by the Israeli Ministry of Science and Technology. VD was partially supported by a fellowship from the Edmond J. Safra Center for Bioinformatics at Tel Aviv University.

Conflict of Interest

ID: consultation/advisory boards for Pfizer, Janssen, Abbvie, Takeda, Roche/Genentech, Celltrion, Celgene/BMS, Rafa Laboratories, Neopharm, Sublimity, Sangamo, Arena, Gilead, Galapgos, DSM, Wild bio, Food Industries Association, Cambridge Healthcare, Integra Holdings; speaking/teaching for Pfizer, Janssen, Abbvie, Takeda, Roche/Genentech, Celltrion, Celgene/BMS, Nestle, Falk Pharma, Ferring. Grant support: Pfizer, Altman Research.

Author Contributions

RR and UG designed the study; ID, KR, NW provided clinical insights; RR designed and established bioinformatics pipelines; RR, VD analysed data; RR, UG, and ID wrote the manuscript draft.

References