1. Computational and Systems Biology
  2. Microbiology and Infectious Disease
Download icon

Microbial genetic and transcriptional contributions to oxalate degradation by the gut microbiota in health and disease

  1. Menghan Liu
  2. Joseph C Devlin
  3. Jiyuan Hu
  4. Angelina Volkova
  5. Thomas W Battaglia
  6. Melody Ho
  7. John R Asplin
  8. Allyson Byrd
  9. P'ng Loke
  10. Huilin Li
  11. Kelly V Ruggles
  12. Aristotelis Tsirigos
  13. Martin J Blaser  Is a corresponding author
  14. Lama Nazzal  Is a corresponding author
  1. NYU Langone Health, United States
  2. Vilcek Institute of Graduate Biomedical Sciences, United States
  3. Litholink Corporation, Laboratory Corporation of America Holdings, United States
  4. Department of Cancer Immunology, Genentech Inc, United States
  5. Center for Advanced Biotechnology and Medicine, Rutgers University, United States
Research Article
  • Cited 0
  • Views 1,255
  • Annotations
Cite this article as: eLife 2021;10:e63642 doi: 10.7554/eLife.63642

Abstract

Over-accumulation of oxalate in humans may lead to nephrolithiasis and nephrocalcinosis. Humans lack endogenous oxalate degradation pathways (ODP), but intestinal microbes can degrade oxalate using multiple ODPs and protect against its absorption. The exact oxalate-degrading taxa in the human microbiota and their ODP have not been described. We leverage multi-omics data (>3000 samples from >1000 subjects) to show that the human microbiota primarily uses the type II ODP, rather than type I. Furthermore, among the diverse ODP-encoding microbes, an oxalate autotroph, Oxalobacter formigenes, dominates this function transcriptionally. Patients with inflammatory bowel disease (IBD) frequently suffer from disrupted oxalate homeostasis and calcium oxalate nephrolithiasis. We show that the enteric oxalate level is elevated in IBD patients, with highest levels in Crohn’s disease (CD) patients with both ileal and colonic involvement consistent with known nephrolithiasis risk. We show that the microbiota ODP expression is reduced in IBD patients, which may contribute to the disrupted oxalate homeostasis. The specific changes in ODP expression by several important taxa suggest that they play distinct roles in IBD-induced nephrolithiasis risk. Lastly, we colonize mice that are maintained in the gnotobiotic facility with O. formigenes, using either a laboratory isolate or an isolate we cultured from human stools, and observed a significant reduction in host fecal and urine oxalate levels, supporting our in silico prediction of the importance of the microbiome, particularly O. formigenes in host oxalate homeostasis.

Introduction

Over-accumulation of oxalate in humans leads to toxicity (Asplin et al., 1998; Beck et al., 2013). The most common oxalate toxicity is calcium oxalate nephrolithiasis, which accounts for more than 70% of overall nephrolithiasis, affecting 9% of the US population with a 20% 5-year recurrence rate (Lieske et al., 2014; Rule et al., 2014; Saran et al., 2018; Scales et al., 2012; Stamatelou et al., 2003). Oxalate toxicity can also induce chronic kidney disease, an illness affecting more than 30 million Americans, via multiple mechanisms including the activation of the NALP3 inflammasome pathway (Knauf et al., 2013; Mulay et al., 2017; Mulay et al., 2013; Saran et al., 2018; Waikar et al., 2019), RIPK3-MLKL-mediated necroptosis (Mulay et al., 2016a), and oxidative stress-induced cell injury (Khan et al., 2006). In extreme cases, life-threatening systemic oxalosis occurs.

Humans lack endogenous oxalate-degrading enzymes. By contrast, the mammalian intestinal microbes can degrade oxalate, partially protecting their hosts against toxicity (Allison and Cook, 1981; Allison et al., 1986; Allison et al., 1977; Barber and Gallimore, 1940; Miller et al., 2014Azcarate-Peril et al., 2006Oxalobacter formigenes,Lactobacillus sp., Bifidobacterium sp., Enterobacteriaceae, and others can degrade oxalate in vitro (Mogna et al., 2014), and colonization with these taxa in rodent hyperoxaluria models showed reduction in urinary oxalate indicating oxalate degradation in vivo (Hatch et al., 2006; Klimesova et al., 2015; Kwak et al., 2006). Of these oxalate degraders, only O. formigenes is a specialist that uses oxalate as its sole energy source (Cornick and Allison, 1996). O. formigenes also is unique because it induces host oxalate secretion into the colonic lumen (Arvans et al., 2017).

The oxalate-degrading microbes in the human microbiota in vivo have not been characterized. Previous studies on oxalate-degrading microbes have been chiefly done in vitro or in animal models, and the relevance of those microbes to human health remains undefined (Fontenot et al., 2013; Klimesova et al., 2015; Kullin et al., 2014; Turroni et al., 2010; Turroni et al., 2007). This gap limits our understanding of the role of microbiota in diseases induced by oxalate toxicity. Here, we leveraged multi-omics data of the healthy human microbiome to characterize the oxalate-degrading microbes in vivo.

Inflammatory bowel disease (IBD) patients are at increased risk for oxalate toxicity, due to a condition called enteric hyperoxaluria (EH). In EH, enhanced bioavailability and hyperabsorption of intestinal oxalate result in oxalate nephrolithiasis (Corica and Romano, 2016; Liu and Nazzal, 2019). In the USA, >50,000 IBD patients suffer from EH and recurrent calcium oxalate kidney stones (Corica and Romano, 2016; McConnell et al., 2002). EH in IBD patients may reflect lipid malabsorption and increased gut permeability. However, an alternate hypothesis is that microbiota-based oxalate degradation is impaired in dysbiotic IBD patients, leading to increased oxalate absorption (Allison et al., 1986). We interrogated the microbiota of IBD patients to understand shifts in microbiota-based oxalate degradation functions and their metabolic consequences.

Results

Type I and type II microbial oxalate degradation pathways

To determine the oxalate degradation pathways (ODPs) used by human gut bacteria, we curated all experimentally validated microbial ODP from literature review and database searches (see Materials and methods) (Allison et al., 1985; Anand et al., 2002; Blackmore and Quayle, 1970; Daniel et al., 2004; Dumas et al., 1993; Foster et al., 2012; Pierce et al., 2010). We classified those ODPs into two types based on their enzymatic mechanisms and co-factor requirements. Type I ODPs cleave the oxalate carbon-carbon (C-C) bond in a single step (Figure 1A). The two recognized type I enzymes, oxalate oxidase and oxalate decarboxylase, are indistinguishable at the amino acid level; therefore, we refer to them jointly as oxalate oxidase/decarboxylase (OXDD) (Figure 1A). Type II ODPs consists of two enzymatic reactions requiring coenzyme A as co-factor (Figure 1A). First, a coenzyme A molecule is added to oxalate to form oxalyl-CoA via enzymes including formyl-CoA transferase (FRC) (Figure 1A). In the second step, oxalyl-CoA decarboxylase (OXC) metabolizes oxalyl-CoA into CO2 and formyl-CoA (Figure 1A).

Figure 1 with 1 supplement see all
Type I and type II microbial oxalate-degrading pathway (ODP).

(A) Schema of type I and type II ODP. Enzymes are annotated with corresponding KEGG IDs. OXDD, FRC, and OXC are the focus of the present study. (B). Cladogram of microbial genera that encode oxalate-degrading enzymes OXDD, FRC, and OXC. The three rings surrounding the cladogram indicate OXDD-, FRC-, or OXC-encoding genera, respectively.

Knowing the relevant oxalate degradation enzymes (ODEs), we then acquired all available protein homologs of the three ODE OXDD (n = 2836), FRC (n = 1947), and OXC (n = 1284), which enable homology search from UniProt Interpro (Mitchell et al., 2019; Mulder et al., 2005). By tracing the taxonomic origin of the genes encoding those homologs, we found that OXDD-coding taxa can be fungal or bacterial, whereas FRC- and OXC-coding taxa are strictly bacterial (Figure 1B). The frequent co-occurrence of FRC and OXC in individual genomes indicates encoding complete type II ODP (Figure 1B). As expected, OXDD, FRC, and OXC each are conserved within the same microbial class, but exhibit substantial divergence across classes (Figure 1—figure supplement 1). These data provide both a comprehensive inventory of ODPs and a reference set of ODP-encoding microbes, which enable analyses to elucidate those relevant to humans.

ODPs utilized by the gut microbiota of healthy humans

Next, we asked whether those ODPs are encoded or expressed by microbes within the intestinal tract of healthy humans. To do so, we computationally examined the presence of the ODE within the gut metagenome and metatranscriptome from publicly available samples of healthy humans. From five studies, we analyzed a total of 2359 metagenome and 1053 transcriptome samples from 660 and 165 healthy individuals, respectively (Figure 2—figure supplement 1, Supplementary file 1a). After quality-filtering (see Materials and methods), the sequencing reads were aligned to the unique OXDD, FRC, and OXC homologous proteins we had identified, using DIAMOND Blastx (Buchfink et al., 2015). Alignment pairs with >90% identity were retained for downstream analyses. The alignment cutoff was based on the protein identity of the inter- and intra-species ODEs and determined to be robust for distinguishing ODEs originating from differing microbial species (Figure 2—figure supplement 2).

We found that the majority of the healthy gut microbiomes include ODEs with at least one ODE detected in the metagenome of 607 (92%) of 660 subjects and the metatranscriptome of 132 (80%) of 165 subjects. In the metagenomes, the type II frc and oxc genes were more common (Figure 2A) and more abundant (Figure 2B) than the type I oxdd gene. Similarly, in metatranscriptomes, expressions of type II genes were more common and abundant (Figure 2C,D). Expression of oxdd was only detected in 10 (6%) of the 165 subjects, and the median RPKM was 2-log10 lower than those of frc or oxc (Figure 2C,D). Furthermore, frc and oxc – the coding genes of the two enzymes that catalyze the separate steps in type II PDP – were frequently co-expressed within the same microbiota, indicating expression of the complete type II ODPs (Figure 2—figure supplement 3).

Figure 2 with 3 supplements see all
Detection of type I and II ODE within the fecal metagenome and metatranscriptome of 660 and 165 healthy human subjects.

Prevalence (A) and abundance (B) of ODE in the fecal metagenome of five studies surveyed. Numbers written on the bottom bars indicate the numbers of subjects in whom the corresponding ODE is detected, and only those subjects were considered in (B). Prevalence (C) and abundance (D) of OXDD, FRC, and OXC in the fecal metatranscriptome of four studies surveyed. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, by proportion tests for (A) and (C), by multiple-adjusted Mann–Whitney tests for (B) and (C).

Figure 2—source data 1

Detection of OXDD, FRC, and OXC in the metagenome and metatranscriptome of healthy individuals.

https://cdn.elifesciences.org/articles/63642/elife-63642-fig2-data1-v2.xlsx

These data indicate that microbes utilizing type II rather than type I ODPs predominate in the human intestine. Such finding is consistent in all studies, despite the differences in source populations and sample preparation methods (Supplementary file 1a). For the remaining analyses, we focused on the type II ODPs.

Microbial species that transcribe the type II ODPs in vivo

Although multiple human commensal microbes are known to encode frc and oxc, whether they transcribe those genes in vivo has not been studied (Abratt and Reid, 2010; Cho et al., 2015; Fontenot et al., 2013; Mogna et al., 2014). We next characterized the microbial species transcribing these type II ODP genes in the microbiota of healthy humans. In the metagenomes of 660 individuals, oxc gene of multiple species, including Escherichia coli, O. formigenes, and several Muribaculaceae, Bifidobacterium, and Lactobacillus sp., was detected (Figure 3A, left); E. coli oxc was the most common (56% of subjects), followed by O. formigenes oxc (39% of subjects).

Figure 3 with 6 supplements see all
Expression of type II ODP of microbial species within the intestinal microbiota of healthy humans.

(A) Abundance and prevalence of OXC of microbial species in the metagenome (left) or metatranscriptome (right) of 660 and 165 subjects. Box plots indicate the abundance of microbial OXC (log10 RPKM) among subjects in whom OXC is detected, and are generated with ggplot2 with outliers excluded. Bar plots indicate the prevalence of microbial oxc, with percentage annotated. Microbial species are ordered by the corresponding metagenomic OXC prevalence. A parallel analysis for FRC is shown in Figure 3—figure supplement 5. (B) Detection of OXC and FRC of microbial species in the subject-matched metagenome and metatranscriptome, by study. For each microbial ODE, the subjects are divided into four groups (shown in different colors) based on the co-detection of ODE in the matched metagenome and metatranscriptome, with percent (%) of which reflected. The fran Study, from which E. coli ODP was detected in all subjects, used a sample extraction method known to induce E. coli, as noted in their publication (Franzosa et al., 2014). (C) Population-level contribution of individual species to metagenomic (left) or metatranscriptomic (right) OXC. The population-level contribution of each species was calculated at a relative scale (see Materials and methods) and plotted. Raw values can be found in Supplementary file 1a. The 10 species that have the highest metagenomic or metatranscriptomic contribution are shown. A parallel analysis for FRC is shown in Figure 3—figure supplement 3.

Figure 3—source data 1

Species contribution to FRC, and OXC in the metagenome and metatranscriptome of healthy individuals.

https://cdn.elifesciences.org/articles/63642/elife-63642-fig3-data1-v2.xlsx

In the metatranscriptomes of 167 individuals, oxc gene expression did not directly correlate with corresponding gene abundance or prevalence (Figure 3A, right). O. formigenes oxc expression was both most abundant and most prevalent (in 61% of subjects) in the metatranscriptomes (Figure 3A, right). Despite the detection of E. coli oxc in 56% of subjects, its transcript was present in only 12% of the subjects (Figure 3A). For Bifidobacterium and Lactobacillus species, for which oxalate degradation activity was reported in vitro and in animal models (Federici et al., 2004; Klimesova et al., 2015; Turroni et al., 2010), oxc expression was minimal (<5%) or not prevalent (Figure 3A right). The dichotomy between metagenomic and metatranscriptomic oxc was consistent across different studies (Figure 3—figure supplement 1, Figure 3—figure supplement 2), and also present for the other type II ODP gene frc (Figure 3—figure supplement 3).

To more rigorously examine ODP expression by individual taxa, we further co-analyzed the presence of ODP genes and transcripts matched by subject (Figure 3B). O. formigenes frc and oxc were transcribed in nearly all subjects in whom the genes were detected, as well as in others in whom the gene was not detected, indicating that their expression is common in vivo (Figure 3B). The under-detection of O. formigenes ODP genes in the metagenomes may reflect the highly variable abundance of the organism, often below the lower detection limit using gene-based methods (Barnett et al., 2016; Guo et al., 2017; Kelly et al., 2011; PeBenito et al., 2019). In contrast, E. coli frc and oxc were expressed in only a few subjects even when the corresponding genes were detected metagenomically (Figure 3B). These data indicate that in vivo E. coli rarely transcribe ODP, a pathway used for defense against oxalate-induced stress (Fontenot et al., 2013). In total, these findings demonstrate that ODP transcription varies widely in individual hosts, and by species.

The contributions of individual species to the global microbiota ODP

We then assessed the impact of individual species on global ODP by quantifying their population-level contributions (see Materials and methods). The contribution of O. formigenes to ODP increased from 17% to 63% from the metagenomic to the metatranscriptomic level, greater than the transcriptomic contributions of all other species combined (Figure 3C, Supplementary file 1b). Conversely, the E. coli contribution to ODP was markedly reduced from the metagenomic (36%) to the metatranscriptomic (7%) level (Figure 3C). Other species had low but varied contributions (Figure 3C). A parallel pattern was observed for frc (Figure 3—figure supplement 4, Supplementary file 1c). With the low activity of non-O. formigenes species, network analysis did not yield significant species-species interactions related to oxc transcription (data not shown). In summary, we found that the type II ODP genes, frc and oxc, are encoded by multiple gut microbes, but O. formigenes dominated this pathway at the transcriptional level. These data provide a baseline for examining disease-associated changes.

Validation of ODP detection using ShortBRED

ShortBRED (Kaminski et al., 2015) is a tool for microbiome functional profiling, which clusters protein homologs into clusters, and identifies marker peptide for each cluster, thus potentially achieving high specificity. We reanalyzed all samples using ShortBRED to validate our bioinformatics findings (see Materials and methods). Based on ShortBRED, FRC and OXC were significantly more abundant and more prevalent than OXDD; their abundances were significantly correlated with our previous results (Figure 3—figure supplement 5A). Consistently, O. formigenes was the species with the highest transcriptional activity for FRC (Figure 3—figure supplement 5B). Specifically, each of the three O. formigenes FRC homologs (C3 × 9Y2, C3 × 762, and C3 × 2D4) are distinct from other homologs and from each other; thus, each formed a singleton family (Supplementary file 2) with unique peptide markers (Supplementary file 3). The three O. formigenes FRCs are the most commonly transcribed among FRCs encoded by any taxon. They are present in the metatranscriptome of 50, 52, and 41% of the tested subjects, respectively (Figure 3—figure supplement 5B, right), and also are among the most abundant (ranks 1, 2, and 5) among all FRC proteins (Figure 3—figure supplement 5B, right).

The taxonomic contributions based on BURRITO (McNally et al., 2018) (see Materials and methods) (Supplementary file 1d) are highly concordant with findings using our prior method (Supplementary file 1b, metagenome, Healthy). E. coli and O. formigenes are consistently the two largest contributors, and several Bifidobacterium spp and Lactobacillus spp contributed to a lesser extent.

ShortBRED did not detect O. formigenes OXC in any sample. We reasoned that this under-detection is due to poorly-selected marker peptides. Indeed, for the protein cluster of O. formigenes OXC (consists of two O. formigenes OXC homologs: C3XBB9 and C3 × 545, with C3 × 545 as the centroid), the makers selected are as follows: (1)short (median length 16.2 amino acids-in comparison, the marker for O. formigenes FRC is 300 amino acids) and (2) improperly broken up from long continuous regions (Figure 3—figure supplement 6A, Marker #1–20). Therefore, we manually generated new markers (Markers #21–24) by combining markers that were one amino acid apart. Using these longer markers, ShortBRED successfully detected OXC using three of the four markers (except for the shortest one) with the expected patterns (Figure 3—figure supplement 6B,C).

In summary, all of our major conclusions examined were confirmed with ShortBRED. The diamond-mapping method utilizes reference full length and relies on best alignment scores to assign reads with higher sensitivity, while ShortBRED uses unique regions of reference with higher specificity. Because of their different advantages, the two methods are complementary and both were used for downstream analyses.

Increased enteric oxalate levels and reduced microbial ODP expression in IBD patients

IBD patients, particularly patients suffering from CD, frequently have EH, with oxalate hyperabsorption and calcium oxalate nephrolithiasis (Corica and Romano, 2016; Liu and Nazzal, 2019; McConnell et al., 2002). Ulcerative colitis (UC), regardless of severity and location, is associated with stone formation (Cury et al., 2013), but ileocolonic CD is associated with greater nephrolithiasis risk than either ileal or colonic involvement alone (Cury et al., 2013). We hypothesized that oxalate degradation by the intestinal microbiome may be impaired in IBD patients, leading to more luminal oxalate in the host available for passive absorption.

We tested this hypothesis, using the multi-omics data of IBD patients and healthy subjects from the iHMP-IBD study (Franzosa et al., 2019; Lloyd-Price et al., 2019). The patients were stratified by illness: UC (N = 30 subjects) and CD (N = 54), and the CD group was further divided into CD-L3 (N = 25), with ileocolonic phenotype at baseline and CD-nonL3 (N = 29) without, as defined by the L3 IBD Montreal classification (Satsangi et al., 2006). Consistent with the clinical nephrolithiasis risk, fecal oxalate relative abundances were elevated in both the UC (p=0.005) and CD (p=0.06) patients compared to healthy controls (Figure 4A). All of the CD risk was in the CD-L3 patients (p<0.001), and not in the non-CD-L3 patients, indicating that IBD location, particularly ileocolonic involvement, is key for EH risk (Figure 4A, Figure 4—figure supplement 1). The fecal oxalate levels were not clearly different in relation to disease activity (Damms and Bischoff, 2008; Jowett et al., 2001; Manz et al., 2012; Pathirana et al., 2018Figure 4—figure supplement 2). We observed increased fecal oxalate in association with higher inflammation levels (fecal calprotectin > 50 µg/g) (Figure 4—figure supplement 2C); since our number of study subjects was small, this relationship needs to be examined with larger patient cohorts in future studies.

Figure 4 with 5 supplements see all
Elevated fecal oxalate and reduced expression of microbiome ODP in IBD patients.

(A). Stool oxalate relative abundance (log10) in healthy, UC, CD, CD-L3, or CD-nonL3 subjects from HMP-IBD study. Fecal oxalate relative abundance was determined from untargeted metabolomics data from the iHMP-IBD; measurements related to oxalate were selected and normalized against total metabolites (percent abundance of all observed metabolites) for analysis. L3 refers to the ileocolonic phenotype, according to the Montreal Classification at baseline. Data derived from iHMP-IBD untargeted metabolomics measurements. Prevalence (B) and abundance (C) of OXDD, FRC, and OXC in metatranscriptomes of healthy, UC, CD, or CD-L3 subjects. The 165 healthy controls are combined from four studies (AMP, US_men, fran, HMP2). *: p<0.05, **: p<0.01, ***: p<0.001, ****: p<0.0001 by multiple-adjusted Mann-Whitney tests in (A) and (C), by proportion test in (B).

Figure 4—source data 1

Fecal oxalate and ODE expression in healthy and IBD individuals.

https://cdn.elifesciences.org/articles/63642/elife-63642-fig4-data1-v2.xlsx

The global transcripts of frc and oxc were reduced in the IBD patients compared to the controls in all four studies analyzed (Figure 4B, Figure 4—figure supplement 3). Expression of oxc was detected in 57% UC, 54% CD and 48% CD-L3 subjects, lower than the 71% in healthy individuals (Figure 4B). ODP expression was least impacted in the CD-nonL3 group (Figure 4BFigure 4—figure supplement 3), in which the fecal oxalate increase was not observed (Figure 4A). The total transcripts of frc and oxc were also significantly lower in IBD patients (Figure 4CFigure 4—figure supplement 3B) (p<0.001 for frc for all groups compared to healthy, p<0.01 for oxc in UC patients compared to healthy). We also observed that frc expression was significantly inversely correlated with oxalate relative abundance in the UC group (Figure 4—figure supplement 4A). The same inverse trend was marginally observed for oxc but was not statistically significant (Figure 4—figure supplement 4B); thus, this question will need to be examined in larger future cohorts. In contrast to reduced frc and oxc transcripts in IBD patients, the frc and oxc genes were significantly more abundant (Figure 4—figure supplement 5), indicating that the frc- and oxc-encoding taxa (e.g. Enterobascteriaceae) are enriched in the IBD gut but do not actively express ODP.

Collectively, these data showing the reduction of ODP-associated transcripts in the IBD patient microbiota suggest their role in the elevated intestinal oxalate levels, and possibly increased susceptibility to nephrolithiasis.

Loss of O. formigenes and its ODP-associated transcripts in IBD patients

We next sought to identify the microbial species accounting for the reduced ODP transcripts in the IBD patients. E. coli and O. formigenes with the largest ODP contributions at the genomic and transcriptional level, respectively, were notable. Using gene and transcript jointly as markers, ODP expression by O. formigenes was detected in ~25% of UC and CD patients (Figure 5A), significantly less than in healthy persons (~70%), either when studies were combined (Figure 5A) or separate (Figure 3B). In contrast, E. coli ODP was detected in nearly all IBD subjects and was transcribed more frequently compared to healthy subjects (Figure 5A). Consistent with the low overall prevalence, transcripts for O. formigenes ODP expression were less abundant (lower RPKM values) in UC, CD, and in CD-L3 patients compared with controls (p<0.01 for all groups) (Figure 5B). In the IBD subjects, the observed O. formigenes ODP genes were always actively expressed (Figure 5A). Significantly elevated fecal calprotectin levels were observed when O. formigenes was absent in healthy individuals and in CD-L3 patients (Figure 4—figure supplement 2D).

Differential ODP expression by human gut microbes in healthy and disease states.

(A) Detection of microbial OXC and FRC in the subject-matched metagenome and metatranscriptome from healthy subjects, UC, CD, or CD-L3 patients. For each species shown, the subjects are divided into one of four categories based on the co-detection of ODE in the matched metagenome and metatranscriptome. (B) Expression of microbial FRC and OXC in the metatranscriptomes of healthy subjects, UC, CD, or CD-L3 patients. Boxplot reflects the subjects, in whose metatranscriptome the corresponding enzyme is detected. *p<0.01, **<0.0001 by multiple-adjusted Mann–Whitney tests.

Figure 5—source data 1

Species contribution to FRC, and OXC in IBD individuals.xlsx.

https://cdn.elifesciences.org/articles/63642/elife-63642-fig5-data1-v2.xlsx

In total, these data indicate that the absence of O. formigenes colonization or colonization below the level of detection is responsible for the reduction in global ODP transcripts. E. coli, and Lactobacillus and Bifidobacterium spp. use ODPs to defend against oxalate-induced acid stress; their upregulation in IBD appears secondary to the elevated oxalate levels present.

Effect on O. formigenes colonization on the host urinary and fecal oxalate levels

To validate our bioinformatics prediction that O. formigenes is an important oxalate-degrading organism that can influence oxalate homeostasis in vivo, we examined whether the colonization of O. formigenes results in significant reduction in urinary oxalate in mice (Figure 6A). Mice from our gnotobiotic facility which had a total microbiota 2–3 log10 lower than conventional mice were fed with diet supplemented with 1% sodium oxalate and 0.5% calcium. The 1% dietary oxalate is lower than previous studies (Hatch et al., 2006; Hatch et al., 2011), in order to reduce chronic kidney damage described previously (Mulay et al., 2016b). This diet resulted in significant increase of oxalate (23.3% and 174.8% increase in fecal and urinary oxalate, respectively) compared to normal chow (data not shown). We then colonized these mice with either of two O. formigenes strains, including a widely studied lab strain (OXCC13) and a human O. formigenes isolate from the stool sample of a patient with primary hyperoxaluria type 1 (PH1) isolated as described (Allison et al., 1985). Continued colonization with each was established by qPCR (Figure 6—figure supplement 1). Using targeted oxalate assays, the O. formigenes-colonized mice had significantly lower urinary and fecal oxalate levels, compared to the non-colonized mice (Figure 6B–C); the values from the two measurements tended to correlate in individual mice (Figure 6D).

Figure 6 with 2 supplements see all
Effect of O. formigenes colonization on fecal and urinary oxalate.

(A) Study design of the mouse experiment. C57Bl/6 mice from our gnotobiotic facility were assigned to three groups. At days 0, 2, and 7, mice were gavaged (blue arrowheads) with O. formigenes strain OXCC13 (n = 5), O. formigenes freshly isolated from a primary hyperoxaluria type 1 (PH1) subject (n = 6), or O. formigenes culture medium alone (Media). Mice were fed with normal chow supplemented with 1% sodium oxalate and 0.5% calcium from day −5 until sacrifice. Urine was obtained from a 48 hr collection (one to two mice per pool) prior to sacrifice, and feces were collected at sacrifice (blue arrowhead). (B, C) Urinary and fecal oxalate in three mouse groups. Urinary oxalate normalized by creatinine in the 48 hr urine samples and fecal oxalate levels per gram of stool samples in the three experimental groups. *p<0.05, by Tukey’s multiple comparisons test. (D) Relationship between fecal and urinary oxalate. Correlation coefficient was computed using Spearman’s r.

Figure 6—source data 1

Fecal and urinary oxalate in relation to O. formigenes colonization in mice.

https://cdn.elifesciences.org/articles/63642/elife-63642-fig6-data1-v2.xlsx

In our model, fecal O. formigenes averaged ~5×106 by qPCR (Figure 6—figure supplement 1). In contrast, O. formigenes was below the detection limit (<102) in the jejunum and ileum of most mice (Figure 6—figure supplement 2). These data suggest that O. formigenes primarily colonizes and degrades oxalate in the host colon, which is known to be an important site for oxalate absorption.

Taken together, these findings suggest that O. formigenes is sufficient to reduce host urinary and fecal oxalate levels, and support our in silico prediction of the importance of O. formigenes to host oxalate homeostasis, with protection against oxalate-induced toxicity.

Discussion

Oxalate degradation by the human microbiota has been known since the 1940s (Allison and Cook, 1981; Allison et al., 1986; Allison et al., 1985; Allison et al., 1977; Barber and Gallimore, 1940), but the taxa involved in vivo has not been systematically described. We present the first comprehensive study of human oxalate-degrading microbes and define their individual contributions. We successfully distinguished the taxa that are actively transcribing ODP from those that encode the pathway, but with low expression, by co-analyzing metagenome and metatranscriptome data.

Our finding that multiple human gut microbes encode ODPs is consistent with prior studies (Abratt and Reid, 2010; Klimesova et al., 2015; Mogna et al., 2014; Stern et al., 2016; Ticinesi et al., 2018). But surprisingly, at the transcriptional level, O. formigenes dominates the global ODP, which is consistent with O. formigenes being an oxalate autotroph, as well as FRC and OXC being the most abundant proteins during both exponential and lag stages (Ellis et al., 2016). These findings were consistent using our method and ShortBRED, which have their unique analytic strengths. The diamond-mapping method utilizes protein full-length information and is completely agnostic. By setting a stringent identity cutoff of 90%, we retain only the high-confidence alignment pairs. The finding generated using this 90% cutoff is consistent with the observations generated through the second method ShortBRED, which uses ODE-specific marker peptides, which provide evidence that the cutoff is highly effective at preventing spurious alignments. By contrast, ShortBRED generated highly specific markers but is limited by several parameter choices (CD-Hit, centroid protein selection, length and identity of short identical regions and final marker). Because of their specific advantages, the two methods are complementary.

The contrasting genetic and transcriptional ODP differences we observed highlight the importance of analysis beyond the gene level for microbiome studies. There are two critical advantages of an approach that co-analyzes metagenome and metatranscriptome: (1) Detection of genes in the metagenome does not ensure that they are being actively utilized by the indicated taxon. Thus, analyzing metatranscriptomic data allows distinguishing the taxa that are actively contributing to a biological process by generating the relevant proteins. In contrast, for those taxa that are not, the analysis provides a new view of their metabolism, in that although they have the potential for using a certain pathway or metabolite, it is not active under the particular condition tested. (2) Having metatranscriptomic data or other functional readout is particularly useful for cross-comparisons of the microbiome of different cohorts (i.e., diseased versus healthy subjects, in humans or experimental animals). For example, in our study, ODP genes were increased in IBD patients, whereas transcripts were significantly reduced. Specifically, the increased ODP gene abundances were due to the over-representation of E. coli strains in IBD patients, which uncommonly transcribe ODP in vivo. In contrast, the decrease of oxalate-degrading gene expression is caused by the loss of O. formigenes, which is the dominant microbe that transcribes this pathway. Having observed the differential abundance for a gene does not necessarily indicate functional shifts, as gene abundance is driven by the most abundant taxa, which could be transcriptionally silent. As such, the transcriptional evidence is a better indicator for evaluation of microbiome functional differences. The integrative multi-omics analysis framework built for this study, (now deposited on Github), can be extended to a broad range of microbiome functions.

EH is frequent in IBD patients, particularly in CD, and in those who underwent Roux-en-Y gastric bypass; the latter population is optimal for study since pre- and post-treatment samples can be easily collected. In the present study, we examine IBD patients, a risk population for EH and nephrolithiasis, and show that the impaired metabolic activity of the microbiota is correlated with enteric oxalate levels, thus nephrolithiasis risk. That fecal oxalate relative abundances were elevated in the iHMP-IBD UC patients, and in the CD patients with ileocolonic involvement, is consistent with their high nephrolithiasis risk (Cury et al., 2013; Hylander et al., 1979; Hylander et al., 1978). In this population at risk for EH, the ODP gene abundances are increased, but the extent of gene expression is decreased. This contrast between gene abundance and expression is due to differential ODP transcription in particular microbes.

IBD patients are known to have low levels of O. formigenes colonization (Kumar et al., 2004). The oxalate-degrading specialist O. formigenes is a gram-negative anaerobe susceptible to multiple antibiotics including macrolides, tetracyclines, metronidazole, and clindamycin (Kharlamb et al., 2011; Mittal et al., 2005). Thus, the reduced colonization of O. formigenes in IBD patients might reflect the frequent antibiotic treatments they receive. Human data showed that a single course of antibiotics to eradicate Helicobacter pylori results in the long-term suppression of O. formigenes colonization (Kharlamb et al., 2011). This supports the notion that antibiotics exacerbate microbiome dysbiosis and may lead to secondary conditions such as EH. These findings raise a further potential cost of antibiotic treatment in IBD patients, which should be weighed against potential benefits, especially when infection is not clearly demonstrated. Other changes in the intestinal milieu of IBD patients affecting pH, oxygen, and bile acid levels have been shown to impact O. formigenes in vitro (Duncan et al., 2002Allison et al., 1985).

Our findings that O. formigenes is the main contributor to oxalate degradation in the healthy state, but is diminished in the IBD population, provides a strong rationale for O. formigenes-based restoration therapy. Restoring O. formigenes has been tested in primary hyperoxaluria (PH) patients, but yielded mixed results (Hoppe et al., 2011; Hoppe et al., 2017). However, PH patients might not be the best subjects for restoration because their oxalate toxicity is caused by hepatic overproduction of oxalate, and microbial degradation is restricted to the oxalate secreted into the gut lumen. In EH, the oxalate overload is in the intestinal lumen, which provides a nutrient-rich environment for microbes such as O. formigenes to (re)colonize and degrade oxalate prior to absorption into the host circulation. A recent study (Canales and Hatch, 2017) of a surgery-induced EH rat model via Roux-en-Y gastric bypass (RYGB), indicated the potential of O. formigenes for treating EH in the post-RYGB rat, O. formigenes strain OXWR achieved 100% colonization and decreased urinary oxalate by 74% compared to 39% in the sham-operated group. However, levels of O. formigenes colonization in EH patients need to be established in future studies. Although PH is a rare disease, EH in IBD patients is common and growing more so, with currently limited treatment options.

Although our study has significant findings with the potential for translational and mechanistic studies, we acknowledge study limitations. The fact that we only used sequences of proteins instead of the whole genomes during read mapping could lead to false taxa assignments due to possible horizontal gene transfer events. Our analysis was limited to the currently characterized ODPs; therefore, we cannot rule out the existence of other enzymes in the human microbiota to degrade oxalate. Our observations largely rely on metatranscriptomics data, and therefore could be limited by technical biases. It is more difficult to acquire high-quality metatranscriptome, as RNA is less stable and subject to degradation during sample preparation. However, the high consistency across the separate studies (Figure 2, Figure 2—figure supplement 3, Figure 3—figure supplement 1) done using different sample collection methods, library preparation, and sequencing methods (Supplementary file 1a), suggests that our findings are robust with reference to technical variation.

Also, we did not have access to urinary oxalate or ascertain kidney stone history in our IBD cohort, both critical variables to extend our conclusions. However, previous studies demonstrated intestinal oxalate is predominantly absorbed paracellularly (Binder, 1974; Knauf et al., 2011; Saunders et al., 1975), and oxalate absorption is determined by concentration gradient, gut permeability, and oxalate bioavailability. Thus, colonic oxalate levels should correlate with both renal oxalate and with risk for CaOx stones. In addition to the untargeted metabolomics data used in this study, targeted measurement of fecal and urinary oxalate is desired in future controlled human studies. Oxalate transport in the human intestine has not been completely elucidated, nor has the relative importance of the small and large intestine to oxalate absorption been affirmatively determined. Therefore, colonization site and biogeography may be important factors for oxalate-degrading microbes to reduce host oxalate absorption. The ODP-transcribing microbes identified in the present study are based on fecal samples, which may bias towards organisms colonizing the colon, but not the small intestine. Furthermore, we showed in mice that O. formigenes colonization significantly affects oxalate homeostasis in vivo but further colonization with synthetic communities of different oxalate degraders is needed to identify which oxalate-degrading species have the largest impact on oxalate degradation and overall oxalate homeostasis.

In prior human studies, O. formigenes alone can effectively decrease host fecal or urinary oxalate levels, when host are exposed to high oxalate levels (Canales and Hatch, 2017; Jiang et al., 2011; Li et al., 2016; Li et al., 2015). In our study, O. formigenes colonization was below the level of detection in the jejunum and ileum; therefore, it is likely that O. formigenes predominately colonizes and performs oxalate degradation in the colon, reducing dietary oxalate being absorbed into the host. A study with germ-free mice (Li et al., 2016) observed that colonization of O. formigenes strain OXCC13 decreased the mouse fecal oxalate, which is consistent with our findings, but not urinary oxalate. However, several differences were noted between their study and ours, most importantly the use of different mouse strain and their use of a lower oxalate and higher calcium diet (0.1% oxalate and 1% calcium). In our study, both O. formigenes strains we tested reduced host urinary and fecal oxalate, although PH1 that was isolated from human showed a greater effect. This difference suggests that establishing the metabolic activity and host adaptability is critical for evaluating the therapeutic potential of individual O. formigenes strains.

Materials and methods

Meta-omics data of the human microbiome

Request a detailed protocol

Metagenomic and metatranscriptomic data of healthy human subjects were collected from five and four studies, respectively (Abu-Ali et al., 2018; Ehrlich, 2011Le Chatelier et al., 2013 Franzosa et al., 2014; Lloyd-Price et al., 2019; Petersen et al., 2017; Schirmer et al., 2018). Metagenomic and metatranscriptomic data of healthy humans and IBD subjects were collected from the iHMP-IBD study (Lloyd-Price et al., 2019). Each sample was cleaned by KneaData to remove low-quality reads and host-associated reads. The metabolic profiles of each sample were surveyed using HUMAnN2 v0.11.1 (Franzosa et al., 2018) under parameters --prescreen-threshold 0.01, --pathways-database metacyc_reactions_level4, metacyc_pathways_structured, and --protein-database uniref50, for the comparison in Figure 2—figure supplement 1. Fecal oxalate relative abundance was determined from untargeted metabolomics data from iHMP-IBD; measurements related to oxalate were selected and normalized against total metabolite (percent abundance of all observed metabolites) for analysis.

Homologous proteins of ODE

Request a detailed protocol

The protein homolog families of OXDD, FRC, and OXC were characterized by UniProt Interpro (Mitchell et al., 2019; Mulder et al., 2005) (V70) in protein families IPR017774, IPR017659, and IPR017660, respectively. We acquired the taxonomic origin and amino acid sequences of 2699 OXDD, 1947 FRC, and 1284 OXC homologs. Protein homologs that are 100% identical were then removed, leaving 2519 OXDD, 1556 FRC, and 1037 OXC unique homologs, which were used as a reference database of ODEs for subsequent query against the meta’omics data. Oxalate oxidoreductase (Figure 1A), a recently discovered enzyme for which there only is limited information (Anand et al., 2002; Dumas et al., 1993Grąz et al., 2016; Kumar et al., 2011; Svedruzić et al., 2007; Tanner et al., 2001), was not considered in this present study.

Pairwise identity between ODE protein homologs

Request a detailed protocol

Multiple sequence alignments were performed among the unique protein homologs separately, by muscle (Edgar, 2004) in seaview v4.7 (Gouy et al., 2010), and alignments were trimmed and imported into R. The pairwise alignment distance d was calculated using function dist.alignment in the seqinR package (Charif et al., 2005) based on identity or Fitch matrix (Fitch, 1966). The alignment distance d was subsequently converted to percent protein identity 100*(1d2), following the documentation of dist.alignment.

Detection of ODE in the meta-omics data

Request a detailed protocol

We used two different approaches with complementary methodologies to ensure the conclusions are robust.

  1. The first method we used is an agnostic approach that uses full protein sequences with high sensitivity. The quality-filtered meta’omics data were aligned against the reference protein databases consisting of the unique ODE homologs, by diamond blastx (Buchfink et al., 2015), with best hit returned (--max-target-seqs 1). Alignments with identity <90% were arbitrarily filtered out to prevent non-ODE reads from aligning to the ODE-specific reference proteins due to local similarity (Figure 2—figure supplement 2) By setting a stringent identity cutoff of 90%, we retain only the high-confidence alignment pairs. The abundance of each ODE protein homolog was calculated as reads per kilobase per million (RPKM) in each sample. When multiple timepoints were available, each subject was represented by the mean measurements across all samples provided.

  2. We also used ShortBRED, which compares ODEs with all other known proteins to identify highly specific marker peptides, and thus can achieve high specificity. Using ShortBRED-identify, FRC, OXC, and OXDD were clustered into 202, 190, and 846 families respectively, with a centroid/representative protein selected for each family (Supplementary file 2). Then those centroids were compared against the uniref90 reference protein database (the ODE homologs were excluded), to remove the common region for identification of short peptide markers (Supplementary file 3). Lastly, filtered meta-omics reads were mapped against those peptide markers using ShortBRED-quantify with parameters --pctlength 0.5 and --id 0.9, to calculate the RPKM for each protein family. Default parameters were used unless noted otherwise.

Population-level contribution to ODE

Request a detailed protocol

The population-level contribution of a species to ODE was designed as a measurement to take both prevalence and abundance information into consideration. It is calculated for each ODE separately, based on their abundances (RPKM values). Using oxc as the example, suppose there are Moxc-coding species and N samples. In any given sample j, the contribution of species i to OXC, cij, is represented by its relative oxc abundance, calculated from

cij= zij/i=1Mzij 

where the zij denotes the RPKMoxc of species i in sample j. In this way, we normalize across samples with the total contribution in any OXC-positive samples fixed to 1, and to 0 in any OXC-negative samples.

The population-level contribution of species i: Ci, can be subsequently calculated from summating contribution of species i in N samples, as follows

Ci = j=1Ncij 

Note that population-level contribution of species monotonically increases with sample size N. Therefore, it has been transformed to relative scale when being compared across different populations or different sample types (metagenome vs. metatranscriptome), such as in Figure 3C, Figure 3—figure supplement 4 and 12.

BURRITO for linking function to taxa

Request a detailed protocol

BURRITO (McNally et al., 2018) was used to deconvolve ODE genes into taxa at the species level. Taxonomic profiles of healthy metagenome (n = 2539 samples) were generated by Metaphlan2 (Truong et al., 2015) under default parameters. In each sample, RPKM of all FRC or OXC homologs were summed as total abundance of frc or oxc genes and then supplied to BURRITO. The genomic content file was derived from taxonomic annotation of protein homologs from UniProt Interpro. We assumed one copy of frc and oxc for each genome to minimize bias, as such information is not available for all species.

Network analysis

Request a detailed protocol

A network analysis of oxc or frc expression from microbial species used SpiecEasi (Kurtz et al., 2015). The raw RPKM values were used, and networks were constructed under default parameters method='mb', sel.criterion='bstars', lambda.min.ratio = 2e-2nlambda = 100, and pulsar.params=list(rep.num = 20, ncores = 2).

Code availability

Request a detailed protocol

Source code of the pipeline can be found on Github via https://github.com/ml3958/FindTaxaCtrbt (Liu, 2021; copy archived at swh:1:rev:13bbc4662f458bff327e348162bf51d875ed34d3). Downstream analysis scripts are available per request.

O. formigenes isolation and culturing

Request a detailed protocol

O. formigenes strain PH1 was isolated from the stool sample of a primary hyperoxaluria patient using methods described in Allison et al., 1985. Both strains were cultured in defined oxalate broth as described (Allison et al., 1985), except that the concentration of oxalate is 50 mM. O. formigenes was cultured at 37°C in an anaerobic chamber before each mouse gavage.

O. formigenes colonization of mice

Request a detailed protocol

A total of 17 C57Bl/6 male mice maintained in our gnotobiotic facility were used. The median of mouse baseline microbiome 16S qPCR measurements was 2.6 × 105, which is significantly lower than that of SPF mice (usually 108 or 109 in our experiments, representing <0.1% of the absolute abundance). The taxa detected in the baseline samples (Supplementary file 1e) did not lead to colonization resistance to the introduced O. formigenes strains.

At days 0, 2, and 7, mice were gavaged with 100 µl from a 24 hr growing culture of O. formigenes strain OXCC13 (n = 5), O. formigenes freshly isolated from a PH1 subject (n = 6), and O. formigenes culture medium alone (n = 6). Mice were fed normal chow supplemented with 1% sodium oxalate from study commencement (day −5) until mouse sacrifice at day 35. We did not perform an explicit power analysis since there is no previous data on our animal model and experimental conditions. We thought that five to six mice per group is a reasonable number of mice per group to show a significant change in urinary oxalate. Biological replicates are parallel measurements of biologically distinct samples that capture random biological variation, which may itself be a subject of study or a noise source. In each mouse group, we performed five to six biological replicates by gavaging material (culture media, growing culture of strain OXCC13 or PH1) into five to six mice.

DNA extraction and O. formigenes quantitative PCR

Request a detailed protocol

Murine fecal pellets or intestinal contents were collected and stored at −80°C until DNA was extracted, using the MoBio 96-well extraction kit, following the manufacturer’s instructions. We confirmed that mice were colonized with O. formigenes at sacrifice using qPCR of fecal samples (Figure S16). qPCR was used to quantitate the number of copies of the oxc mRNA using the LightCycler 480 SYBR Green I Master Mix and run using the LightCycler 480 system. Paired primers (forward 5′-TGT-TTG-TCG-GCA-TTC-CTA-TC-3′, reverse 5′-TTG-GGA-AGC-AGT-TGG-TGG-3′) were used under the conditions: 95°C for 10 min, followed by 40 cycles of: 95°C for 23 s, 63°C for 20 s, 70°C for 40 s, and a final 30 s at 40°C as described (Pebenito et al., 2019).

Fecal and urinary collections and oxalate measurements

Request a detailed protocol

Mice were housed (n = 1 or 2) in metabolic cages, and 48 hr collections (1–2 mice per urine pool) were made under mineral oil into vessels containing crystal thymol as a preservative. Urinary oxalate (mg/dl) and creatinine (mg/dl) concentrations were determined in acidified (HCl) samples collected from all mice over a 48 hr period by Litholink Corp (Chicago, IL). Fecal pellets were collected at the end of the urine collections. Fecal pellets were acidified using 2M HCl, vortexed for 20 min, and then centrifuged at 21,000 g at room temperature, and supernatant fecal water collected using described methods (Jiang et al., 2011). Fecal water oxalate was measured using an oxalate calorimetric assay (Abcam, ab196990, Cambridge, UK) per the manufacturer’s instructions. Technical replicates are repeated measurements of the same sample that represent independent measures of the random noise associated with protocols or equipment. Fecal oxalate and qPCR were measured in duplicates.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for Figures 2–5.

The following previously published data sets were used

References

    1. Dumas B
    2. Sailland A
    3. Cheviet JP
    4. Freyssinet G
    5. Pallett K
    (1993)
    Identification of barley oxalate oxidase as a germin-like protein
    Comptes Rendus De l'Academie Des Sciences. Serie III, Sciences De La Vie 316:793–798.
    1. Hylander E
    2. Jarnum S
    3. Frandsen I
    (1979)
    Urolithiasis and hyperoxaluria in chronic inflammatory bowel disease
    Scandinavian Journal of Gastroenterology 14:475–479.
    1. Pathirana WGW
    2. Chubb SP
    3. Gillett MJ
    4. Vasikaran SD
    (2018)
    Faecal calprotectin
    The Clinical Biochemist. Reviews 39:77–90.

Decision letter

  1. Wendy S Garrett
    Senior Editor; Harvard T.H. Chan School of Public Health, United States
  2. Peter Turnbaugh
    Reviewing Editor; University of California, San Francisco, United States
  3. Denise Dearing
    Reviewer; University of Utah, United States
  4. Eric Brown
    Reviewer; Broad Institute, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Oxalate is critical for kidney stones yet the bacteria responsible for its metabolism in humans remain poorly understood. Herein, the authors use a multi-disciplinary approach to study the abundance and expression of genes for human gut bacterial oxalate metabolism in healthy subjects and patients with inflammatory bowel disease. They go on to show that Oxalobacter formigenes significantly alters oxalate levels in mice. These analyses provide a critical step towards a more comprehensive view of oxalate metabolism and its role in health and disease.

Decision letter after peer review:

Thank you for submitting your article "Microbial contributions to oxalate metabolism in health and disease" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Wendy Garrett as the Senior Editor. The following individual involved in review of your submission have agreed to reveal their identity: Eric Brown (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

Liu and colleagues present a series of finding related to gut bacterial oxalate metabolism. First, they curate a set of previously described proteins and generate a reference database of homologs based upon Interpro annotations. Then, they re-analyze previously published meta-genomes and transcriptomes to find hits to these reference genes in the gut microbiomes of healthy and IBD subjects. Surprisingly, they find an inverse association between oxalate levels and the total transcripts of bacterial oxalate degradation genes. They also include data showing that Oxalobacter formigenes (a model oxalate degrading bacterium) impacts oxalate in mice. These analyses are a good example of how the microbiome field as a whole can utilize complex multi-omic datasets to help answer specific questions of clinical importance. However, there are multiple limitations and points which need to be addressed prior to publication.

Essential revisions

1. The claim that O. formigenes is the dominant oxalate degrading species is not well supported by any of the current data. The sequence analysis is based on a presumably partial knowledge of the full scope of enzymes capable of this activity, so it remains unclear if alternative species or pathways are important to consider. The mouse experiment is used as a "validation" but only shows that this species is sufficient to impact oxalate not that it is necessary in humans. A valuable first step would be to colonize germ-free mice with O. formigenes along with multiple other oxalate degraders, then perform leave-one-out experiments to test which species have a marked impact on oxalate levels when removed.

2. The approaches used to assign genes and species are not state-of-the-art and may not be entirely reliable. I'd suggest trying ShortBRED (Huttenhower lab) or a related tool to quantify the protein families of interest. FishTaco and BURRITO (Borenstein lab) could be used to help link taxonomy to function. This is an important point since it relates to the claim that O. formigenes is the source of most transcripts. Furthermore, it's unclear if these genes are horizontally transferred (which could be assessed by comparing gene and species trees). If so, simple read mapping could assign genes to the wrong genomes. Ignoring these other tools, the validation shown in Figure S3 doesn't make much sense to me. The threshold of 90% misses many of the intraspecies comparisons. I'm also concerned that Oxalobacter, the focus of this work, only has a handful of representative genes, which will make it difficult to reliably assign reads to this genus let alone to O. formigenes specifically.

3. Some attempt needs to be made to experimentally address the counter-intuitive observation that higher substrate (oxalate) is associated with lower expression, which runs counter to how most bacterial genes are regulated. What accounts for the downregulation? Is this related at all to the environment within the IBD gut?

4. The way the oxalate levels in the feces are presented is problematic. In the manuscript, the authors make multiple mentions the observed abundance of oxalate is a "fecal concentration of oxalate" when in fact it is the relative abundance of oxalate as measured by LC-MS. These data are not measuring concentration but relative abundance, which can be influenced by other non-biological factors such as how well the metabolite is ionized in each sample by LC-MS. Authors should not these are relative abundance calculations and not concentrations (for example Line 13 describing Figure 4A). Furthermore, the authors should indicate whether the samples were normalized between cohorts and how they were, and whether the relative abundance measurement is correct for within sample differences (% abundance of all observed metabolites) or a raw abundance? For example, the large difference between IBD and healthy stool could lead to less total metabolites being extracted and ionized thus data normalization between samples may actually increase the correlations you are seeing between oxalate and oxalate-degrading enzyme expression by meta-transcriptomics.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Microbial genetic and transcriptional contributions to oxalate degradation by the gut microbiota in health and disease" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, including Peter Turnbaugh as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Wendy Garrett as the Senior Editor.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1. In Figure 4 the raw values in the y-axis when reporting oxalate relative abundance are unclear which units? Are they actually negative. The figure legends in general should include more specific information for the metabolomic data so it is easier to interpret.

2. The authors still refer to "oxalate concentrations" in the figure legend for figure 4.

3. It is unclear why the calprotectin cut-off was 50ug/mL please cite literature or a reason for using this value. It seems like there is in fact a trend with inflammation potentially.

4. Along that note Pearson analysis could show whether calprotectin levels correlate with Oxalobacter, something worth mentioning for the differences in abundance across IBD vs healthy controls and will be useful for the field if in fact this were to ever be utilized as a probiotic by others in the future.

5. More clarification is needed in the manuscript text on how comparing metagenomic analysis with metatranscriptomics can successfully pinpoint which taxa contribute to a disease pathway (in this case oxalate degradation). I still find this comparison confusing as a reader and potential pitfalls to this approach should be more clearly stated (sequencing depth as mentioned).

https://doi.org/10.7554/eLife.63642.sa1

Author response

Essential revisions

1. The claim that O. formigenes is the dominant oxalate degrading species is not well supported by any of the current data. The sequence analysis is based on a presumably partial knowledge of the full scope of enzymes capable of this activity, so it remains unclear if alternative species or pathways are important to consider. The mouse experiment is used as a "validation" but only shows that this species is sufficient to impact oxalate not that it is necessary in humans. A valuable first step would be to colonize germ-free mice with O. formigenes along with multiple other oxalate degraders, then perform leave-one-out experiments to test which species have a marked impact on oxalate levels when removed.

We agree with the reviewer that we do not know the full scope of enzymes capable of this activity, however we based our analysis and conclusions on the published literature, which indicates that oxalate degradation is through one of the 2 pathways (Figure 1). We now acknowledge this limitation in the Discussion.

We changed the description of the animal experiment to “To validate our bioinformatics prediction that Oxalobacter formigenes is an important oxalate degrading organism that can influence oxalate homeostasis in vivo, we examined whether the colonization of O. formigenes results in significant reduction in urinary oxalate in mice.”

We also thank the reviewer for the suggestion for a ‘leave-one-out’ experiment, which is in our future plans to test the hypothesis suggested by the data. We agree with the reviewer that an experiment in which we could give GF mice a synthetic community of potential oxalate-degrading microbes, and monitor their in vivo colonization, ODP transcription, and host oxalate metabolic responses will be valuable. However, such an experiment would require careful design and multiple pilot experiments. We would first need to test different diets of varying oxalate levels to establish a mouse model with intra-colonic oxalate concentrations faithfully mimicking the human intra-colonic milieu and avoiding toxicity to the oxalate-degrading community with high concentrations of dietary oxalate, as has been described (Miller, Dale et al. 2017). We also need to test multiple strains from the same species of oxalate degraders to ensure rigor. Thus, in combination with the current COVID situation limiting our capabilities to perform further animal experiments, we would like to argue that this current experiment is beyond the scope of this manuscript and propose to perform it when possible in the future, as part of a future manuscript. We have now added this limitation to the Discussion and the need for validating our results using animal models colonized with these different microbiota. We believe that the present manuscript has a number of important observations, not requiring this question to be answered, for it to still substantially advance the field.

2. The approaches used to assign genes and species are not state-of-the-art and may not be entirely reliable. The approaches used to assign genes and species are not state-of-the-art and may not be entirely reliable. I'd suggest trying ShortBRED (Huttenhower lab) or a related tool to quantify the protein families of interest.

As the reviewer suggested, we reanalyzed abundances of oxalate degrading enzymes (ODEs) using ShortBRED. First, using ShortBRED-identify default parameters, we clustered FRC, OXC, and OXDD homologs included in this study into 202, 190, and 846 families, respectively. Marker peptides were identified for each of the ODE protein families (see Methods). Next, using ShortBRED-quantify, we mapped the multi-omics reads of 2359 metagenomic and 1053 transcriptomic samples to those marker peptides and quantified abundances of each ODE using ShortBRED-quantify.

The results generated by these new ShortBRED analyses validated all of our prior major findings. We found that:

1. FRC and OXC, but not OXDD, are frequently detected in the human gut metagenome and metatranscriptome (New Figure 3—figure supplement 5A), which is consistent with our previous observation (Figure 3).

2. O. formigenes is the species with the highest transcriptional activity for FRC (New Figure 3—figure supplement 5B). Specifically, each of the three O. formigenes FRC homologs (C3X9Y2, C3X762, and C3X2D4) are distinct from other homologs and from each other. Thus each formed a singleton family (New Supplementary file 2) with unique peptide markers (New Supplementary file 3). Based on the ShortBRED analysis, the three O. formigenes FRCs are the most commonly transcribed among the FRCs encoded by any taxon. They are present in the metatranscriptome of 50, 52, and 41 percent of the tested subjects, respectively (New Figure 3—figure supplement 5B, right), and also are among the most abundant ones (rank 1,2 and 5) among all FRC proteins (New Figure 3—figure supplement 5B, right). Therefore, O. formigenes is the dominant microbial source for FRC in the human microbiome (See below for OXC data).

3. Based on ShorBRED-quantify results, the global transcripts for frc and oxc are reduced in all IBD groups compared to healthy individuals, particularly in UC and CD-L3 IBD patients (New Figure 3—figure supplement 6), which is consistent with our previous observation (Figure 4).

However, ShortBRED relies on many arbitrary decisions (including CD-Hit, centroid protein selection, length and identity of short identical regions and final marker). It is not possible to empirically determine the best parameter combinations. For example, ShortBRED did not detect O. formigenes OXC in any sample, which contradicts all of our other analyses using ShortBRED and our method (Figure 3 and Figure 3—figure supplement 3), and contradicts the well-established fact that O. formigenes colonizes >30% of humans (Kelly, Curhan et al. 2011, PeBenito, Nazzal et al. 2019). We reasoned that this under-detection is due to poorly-selected marker peptides. Indeed, the makers selected for the protein family of O. formigenes OXC (New Figure 3—figure supplement 6A, markers 1-20) are (1), short (median length 16.2 amino acids, as a comparison: the marker for O. formigenes FRC is 300 amino acids), and (2), improperly broken up from long continuous regions (New Figure 3—figure supplement 6A). Therefore, we manually generated new markers (Markers #21-24) by combining markers that were one amino acid apart. Using these longer markers, ShortBRED now successfully detected OXC using 3 of the 4 markers (except for the shortest marker) with the expected patterns (New Figure 3—figure supplement 6B, C).

ShortBRED and our previous method are different and both have unique strengths. Our previous method uses protein full length information and is completely agnostic, relying on best alignment hit to assign read to reference. Our method successfully detected O. formigenes FRC and OXC (Figure 3 and Figure 3—figure supplement 3). Thus, the two methods have different advantages and are complementary.

In summary, using a different methodology, several of our major conclusions were confirmed. We now report the results in the revised manuscript in the Results section Validation of ODP detection using ShortBRED and in New Figure 3—figure supplement 5, Figure 3—figure supplement 6. We also discuss the pros and cons of the two methods to help readers interpret the data.

FishTaco and BURRITO (Borenstein lab) could be used to help link taxonomy to function.

Thank you for this excellent suggestion. We now used BURRITO (McNally, Eng et al. 2018) to link function to taxa based on the taxonomic composition generated by Metaphlan2 (Truong, Franzosa et al. 2015) (See Methods). Only metagenomes (n=2359 samples) were included for this analysis because the method was designed to be used with metagenomic data in which the taxonomic and gene abundance are linearly linked.

The taxonomic contributions based on BURRITO (New Supplementary file 1d) are highly concordant with the findings using our original method (Supplementary file 1b, metagenome, Healthy). E. coli and O. formigenes are consistently the top two contributors, and several Bifidobacterium spp and Lactobacillus spp contributed to a lesser extent. We now added those new results into the revised manuscript.

Furthermore, it's unclear if these genes are horizontally transferred (which could be assessed by comparing gene and species trees). If so, simple read mapping could assign genes to the wrong genomes.

The reviewers’ concern regarding horizontal gene transfer (HGT) is well-taken. FRC and OXC homologs encoded by the same Class of bacteria generally cluster together (Author response image 1), suggesting this pathway is evolutionary conserved at high taxonomic levels. As microbiomes are highly individual-specific, to address HGT, one needs to assemble reads from each metagenome to highquality contigs to provide fine resolution to assess for HGT. Currently such a task is technically and computationally challenging. As it is not the focus of current study, we acknowledge these limitations and now include the following sentence in the Discussion:

“The fact that we only used sequences of proteins instead of whole genomes during read mapping could lead to false taxonomic assignments due to possible horizontal gene transfer events”

Author response image 1
Phylogenetic analysis of OXC (A), FRC (B) uniref100 proteins.

Each tip represents a protein homolog, which is color-coded by the microbial Class associated with the encoding. The tip size is proportional to the prevalence of the corresponding protein in the metatranscriptomes of 165 healthy individuals. Homologs with prevalence >1% are annotatedwith text.

I'm also concerned that Oxalobacter, the focus of this work, only has a handful of representative genes, which will make it difficult to reliably assign reads to this genus let alone to O. formigenes specifically.

We thank the reviewer for raising this concern. First, Oxalobacter is currently a small genus consisting of only two known species (The other species Oxalobacter vibrioformis was isolated from anoxic freshwater sediment and only described once in 1989 (Dehning and Schink 1989). Thus, we are confident that both Oxalobacter genus and O. formigenes species are highly-identifiable taxa. Second, the frc and oxc genes of O. formigenes species are distant from other species, as evidenced by the fact that ShortBRED CD Hit clustered them as individual families at the amino acid level) (New Supplementary file 2). Especially for FRC, all sequenced O. formigenes genomes encode three copies of the frc gene with conserved operon structures (Author response image 2A). The three frc genes are further grouped into two forms (Author response image 2B), both of which were identified in the current study. Collectively, the O. formigenes frc and oxc genes in our study are highly representative, and the read assignments for them should be highly specific as well.

Author response image 2
Operon structure and phylogenetic relationship of FRCs of four O.formigenes strains OXCC13, HC1, HOxBLS, and OXK, for which whole genome sequence is available.

(A). Schematic representation of the operon structures for the three FRC genes. The structures are conserved across all four strains. We named the three FRCs as α-, β-, and γ-FRC (from top to bottom). (B). Maximum likelihood phylogenetic tree for the α-, β-, and γ-FRCs from four O. formigenes strains, based on their amino acid sequence. The sequence alignment and tree were generated using phylogeny.fr using “simple click” mode.

The threshold of 90% misses many of the intraspecies comparisons.

We thank the reviewer for this comment. The 90% cutoff on an alignment hit is to prevent non-ODE reads being aligned to our ODE reference proteins at a low identity score. We now clarify this point in the text, as follows:

“By setting a stringent identity cutoff of 90%, we retain only the high confidence alignment pairs. The finding generated using this 90% cutoff is consistent with the observations generated through a second method ShortBRED, which uses ODE-specific marker peptides, which provide evidence that the cutoff is highly effective at preventing spurious alignment.”

3. Some attempt needs to be made to experimentally address the counter-intuitive observation that higher substrate (oxalate) is associated with lower expression, which runs counter to how most bacterial genes are regulated. What accounts for the downregulation?

The Reviewer is correct, and we thank him/her for the comment. The ‘downregulation’ of oxc or frc expression by microbiome referred to the reduction of total transcripts associated with oxc and frc. The suppression or absence of O. formigenes cells is largely responsible for the decrease in ODP transcripts in the IBD gut (Figure 5A). To avoid confusion, we have revised our manuscript, replacing “downregulation of ODP” with “global reduction of ODP-related transcripts” throughout.

Is this related at all to the environment within the IBD gut?

IBD patients regularly receive antibiotics, and O. formigenes is susceptible to commonly used antibiotics (Mittal, Kumar et al. 2005, Kharlamb, Schelker et al. 2011, Lange, Wood et al. 2012, Liu, Koh et al. 2017). Our unpublished human data in healthy adults (in revision, 2021) demonstrated that a single course of antibiotics (clarithromycin and metronidazole) to eradicate Helicobacter pylori results in the persistent suppression of O. formigenes colonization for at least 6 months. Other changes in the intestinal milieu of IBD patients affecting pH, oxygen levels, and increasing bile acids, have been shown to impact O. formigenes in vitro (Allison, Dawson et al. 1985, Duncan, Richardson et al. 2002). We have now added these potential explanations into the Discussion.

4. The way the oxalate levels in the feces are presented is problematic. In the manuscript, the authors make multiple mentions the observed abundance of oxalate is a "fecal concentration of oxalate" when in fact it is the relative abundance of oxalate as measured by LC-MS. These data are not measuring concentration but relative abundance, which can be influenced by other non-biological factors such as how well the metabolite is ionized in each sample by LC-MS. Authors should not these are relative abundance calculations and not concentrations (for example Line 13 describing Figure 4A). Furthermore, the authors should indicate whether the samples were normalized between cohorts and how they were, and whether the relative abundance measurement is correct for within sample differences (% abundance of all observed metabolites) or a raw abundance? For example, the large difference between IBD and healthy stool could lead to less total metabolites being extracted and ionized thus data normalization between samples may actually increase the correlations you are seeing between oxalate and oxalate-degrading enzyme expression by meta-transcriptomics.

We thank the Reviewer for these helpful comments. We address the Reviewer’s comment point-by-point below:

Authors should note these are relative abundance calculations and not concentrations (for example Line 13 describing Figure 4A). Furthermore, the authors should indicate whether the samples were normalized between cohorts and how they were, and whether the relative abundance measurement is correct for within sample differences (% abundance of all observed metabolites) or a raw abundance?

We thank the reviewer for this comment. As suggested, we have repeated all analyses based on the relative abundance of oxalate (% abundance of all observed metabolites in each sample), to account for sample differences (New Figure 4, which replaces the prior Figure 4). This point now is indicated in the Methods under section Meta-omics data of the human microbiome.

Using relative abundance, we reached the same observation that fecal oxalate is elevated in all IBD cohorts compared to healthy individuals (New Figure 4). The correlation between oxc expression and oxalate relative abundance showed negative, but not statistically significant, trends in all IBD groups (New Figure 4—figure supplement 4A), but frc was significantly inversely correlated with oxalate relative abundance in the UC group (New Figure 4—figure supplement 4A). In the revised manuscript, we now report the results based on both absolute and relative abundances.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Essential revisions:

1. In Figure 4 the raw values in the y-axis when reporting oxalate relative abundance are unclear which units? Are they actually negative. The figure legends in general should include more specific information for the metabolomic data so it is easier to interpret.

As requested by a prior reviewer (Prior Review point 4), Figure 4A reports log10 relative abundance of oxalate as a percent of the total metabolites measured (See Methods). The negative values are due to log10 transformations as indicated in the figure. To improve clarity, we now include that information in the figure legend as follows:

“(A). Stool oxalate relative abundance (log10) in healthy, UC, CD, CD-L3 or CD-nonL3 subjects from the HMP-IBD study. Fecal oxalate relative abundance was determined from untargeted metabolomics data from iHMP-IBD; measurements related to oxalate were normalized against total metabolites (percent abundance of all observed metabolites) for analysis.”

2. The authors still refer to "oxalate concentrations" in the figure legend for figure 4.

We thank the reviewer for catching this error. We now have updated the figure axis and legend of Figure 4 and the related Supplemental Figures 1, 2, 4, and 5 to indicate “Fecal oxalate log10 relative abundance”.

3. It is unclear why the calprotectin cut-off was 50ug/mL please cite literature or a reason for using this value. It seems like there is in fact a trend with inflammation potentially.

The fecal calprotectin cutoff was determined based on the pertinent literature (Damms and Bischoff, 2008; Manz et al., 2012; Pathirana, Chubb, Gillett, and Vasikaran, 2018) and to be consistent with current test kits (Pathirana et al., 2018). We have now included citations to these references in the main text and in the Figure Legend.

As the Editor/reviewer points out, increased fecal oxalate with higher inflammation levels (fecal calprotectin >50µg/g) is indeed present. A sub-analysis with samples with fecal calprotectin > 50ug/ml showed a positive correlation between fecal calprotectin and oxalate (new Figure 4—figure supplement 2C), which needs to be examined with larger patient cohorts in future studies. This point now is included in the text, as “We observed increased fecal oxalate in association with higher inflammation levels (fecal calprotectin >50µg/ml) (Figure 4—figure supplement 2C); since our number of study subjects was small, this relationshp needs to be examined with larger patient cohorts in future studies.”

4. Along that note Pearson analysis could show whether calprotectin levels correlate with Oxalobacter, something worth mentioning for the differences in abundance across IBD vs healthy controls and will be useful for the field if in fact this were to ever be utilized as a probiotic by others in the future.

We agree, thank you. Using detection of O. formigenes ODP genes and transcripts jointly as a marker for its presence, we observed significantly elevated fecal calprotectin levels when O. formigenes is absent in healthy individuals and in CD-L3 patients. This analysis suggests that colonization by O. formigenes and gut inflammation could potentially be linked. As suggested by the Editor, we now include new Supplemental Figure 2D, which shows the relationship of O. formigenes status and fecal calprotectin level, without using an arbitrary cut-off. Although we show this Figure, we indicate that the underlying mechanism is unknown and needs to be examined in future studies. The new text that we have included is as follows: “Significantly elevated fecal calprotectin levels were observed when O. formigenes was absent in healthy individuals and in CD-L3 patients (Figure 4—figure supplement 2D) “

5. More clarification is needed in the manuscript text on how comparing metagenomic analysis with metatranscriptomics can successfully pinpoint which taxa contribute to a disease pathway (in this case oxalate degradation). I still find this comparison confusing as a reader and potential pitfalls to this approach should be more clearly stated (sequencing depth as mentioned).

We appreciate the reviewer’s suggestion. We now more clearly describe the two critical advantages of an approach that co-analyzes metagenome and metatranscriptome in the discussion, as follows:

“(1) Detection of genes in the metagenome does not ensure that they are being actively utilized by the indicated taxon. Thus, analyzing metatranscriptomic data allows distinguishing the taxa that are actively contributing to a biological process by generating the relevant proteins. In contrast, for those taxa that are not, the analysis provides a new view of their metabolism, in that although they have the potential for using a certain pathway or metabolite, it is not active under the particular condition tested. (2) Having metatranscriptomic data or other functional readout is particularly useful for cross-comparisons of the microbiome of different cohorts (i.e., diseased versus healthy subjects, in humans or experimental animals). For example, in our study, ODP genes were increased in IBD patients, whereas transcripts were significantly reduced. Specifically, the increased ODP gene abundances were due to the over-representation in IBD patients of E. coli strains that uncommonly transcribe ODP in vivo. In contrast, the decrease of oxalate-degrading gene expression is caused by the loss of O. formigenes, which is the dominant microbe that transcribes this pathway. Having observed the differential abundance for a gene does not necessarily indicate functional shifts, as gene abundance is driven by the most abundant taxa, which could be transcriptionally silent. As such, the transcriptional evidence is a better indicator for evaluation of microbiome functional differences.”

In addition to the previous description of the limitations of our methods, we now have added another limitation that “it is more difficult to acquire high-quality metatranscriptome, as RNA is less stable and subject to degradation during sample preparation”.

[Editors' note: we include below the reviews that the authors received from another journal, along with the authors’ responses.]

Reviewer 1

The authors have taken advantage of accrued data associated with the Microbiome Project to identify gut microbes in humans that are involved in the breakdown of oxalate. They importantly and for the first time use both metagenomics and metatranscriptomics to better understand microbial contributions to oxalate degradation in the human gut. They also performed analyses in various types of IBD to determine if these patients have perturbations of the oxalate degrading microbiome. Some novel aspects of the study were (1) classifying the abundance of oxalatedegrading organisms based on the enzymes involved and their co-factor requirements (2) revealing that the contribution of the oxalate auxotroph O. formigenes to the oxalate degrading pathway is greater than the transcriptomic contributions of all other species combined (3) demonstrating that oxalate degradation pathways were reduced and fecal oxalate elevated in patients with IBD compared with healthy controls. This study opens the doorway to the development of strategies that may off-set the increased oxalate absorption associated with IBD.

We appreciate the reviewer’s comment on the multiple novel aspects of the study. We agree that as the first systemic study on oxalate-degradation by human gut microbes, this work will open new avenues of therapeutic strategies for enteric hyperoxaluria (EH). This study for the first time describes the relative importance of different microbial taxa in human oxalate degradation, and identifies the most relevant clinical indication (Crohn’s disease with the ileocolonic phenotype).

Points that should be addressed:

1. Figure 4- Indicate if this is a log scale.

Thank you for this suggestion. The fecal oxalate level in panels A and D was on a log10 scale, but this was not reflected on the axis title. We now have updated the axis title as “Log10 fecal oxalate by LC-MS” (please see new Figure 4 on page 37).

2. Due to the importance of fecal oxalate measurements, the authors should add to their limitations that a more refined and targeted direct measure of fecal oxalate is warranted to substantiate these claims.

Thank you for the suggestion. We have now added a sentence in the Discussion section that states “In addition to the untargeted metabolomics data used in this study, targeted measurement of fecal and urinary oxalate is desired in future controlled human studies.”.

In addition, in a mouse experiment that we now present, both fecal and urine oxalate levels were directly measured using targeted assays (please see the new Figure 6B and Figure 6C on page 39).

3. Figure S10 (B). Legend indicates OXC abundance, but axis title is frc.

Thank you for pointing out this error. We now have updated that figure (see page 54).

Reviewer 2

The analysis is overall well done and convincing.

We thank the reviewer for the positive feedback.

1. Inflammasome pathways have been implicated but may not be the only mechanism of renal damage in hyperoxaluria.

Thank you for the comment. We agree with the reviewer that there are other mechanisms of calcium oxalate nephrotoxicity. Due to the word limits, we did not list all relevant studies. Now we have added text to the Introduction about other potential mechanisms of calcium oxalate nephrotoxicity.

2. There is no direct evidence that elevated intestinal oxalate is key to EH, although there us indirect evidence, especially in older literature from the 1980s

We thank the reviewer for this comment. We agree that there is no direct evidence yet linking fecal oxalate to EH. However, in prior EH human studies and animal models, fecal oxalate was not measured. In our study, we observed increased fecal oxalate in ulcerative colitis patients and Crohn’s disease patients with the ileocolonic phenotype, which is the same subgroups as have clinical nephrolithiasis risk (Cury, Moss, and Schor, 2013), suggesting that intestinal oxalate could be a marker for EH. In our mouse model, fecal and urine oxalate tended to correlate (please see new Figure 6D on page 39).

3. Lack of Urine oxalate data in these patients is a major weakness, and should be stressed more in the discussion1

We thank the reviewer for this comment. As above, we now provide mouse data on oxalate levels (new Figure 6 on page 39), which addresses the weakness pointed out by the reviewer.

Also, this limitation is stated in the Discussion as: “we did not have access to urinary oxalate or ascertain kidney stone history in our human IBD cohort, both critical variables to extend our conclusions”. However, previous studies have shown that intestinal oxalate is predominantly absorbed paracellularly (Binder, 1974; Knauf et al., 2011; Saunders, Sillery, and McDonald, 1975) along its concentration gradient; as such, colonic oxalate levels should correlate with both renal oxalate and with risk for CaOx stones.

4. It is interesting that the oxalate degrading taxa and genes are reduced in patients at risk for EH, despite the metabolomics data that fecal oxalate is increased. This merits some discussion

We thank the reviewer for this comment. In patients at risk for EH, the abundance of ODP genes is increased (Figure S10), but the extent of gene expression is decreased (Figure 4). This contrast between gene abundance and expression is due to differential ODP transcription in particular microbes. Specifically, the increase of ODP gene abundances was due to the overrepresentation of E. coli that rarely transcribe ODP in vivo. In contrast, the decrease of ODP-gene expression is caused by loss of O. formigenes, which is the dominant microbe that transcribes this pathway. This point is now clarified in the Discussion.

5. There is no good evidence that oral Oxalobacter could be used to recolonize EH patients, or would effectively reduce urinary oxalate. This part of the discussion should be more guarded

We thank the reviewer for this comment. We agree that O. formigenes colonization in EH patients needs to be established in future studies. However, in EH, the intestinal lumen is an oxalate-rich environment, which supports O. formigenes colonization and oxalate degradation prior to absorption into the host circulation. Furthermore, a recent study (Canales and Hatch, 2017) of a surgery-induced EH rat model via Roux-en-Y gastric bypass (RYGB), indicated the potential of O. formigenes for treating EH. in the post-RYGB rat, O. formigenes strain OXWR achieved 100% colonization and decreased urinary oxalate by 74% compared to 39% in the sham-operated group.

As the reviewer suggested, we now have added both points into the Discussion.

Minor comment

1. Be consistent in use of calcium-oxalate vs calcium oxalate. I would favor not using the hyphen.

Thank you. We have deleted the hyphen in “calcium-oxalate” throughout the manuscript, and changes are tracked in the document.

Reviewer 3

In the manuscript entitled “Microbial contributions to oxalate metabolism in health and disease”, Liu et al. have performed a meta-analysis of human fecal multi’omics datasets focusing on describing the gene abundance and expression of oxalate degradation pathway (OPD) genes across these data sets, including IBD cohorts.

The authors introduce the background and significance of this study highlighting that oxalate (systemic) toxicity is of significant importance as a risk factor of kidney stones and CKD. The authors then introduce that humans lack ODP but that gut microbes do, but that there are “gaps in our understanding of the role of the microbiota in diseases induced by oxalate toxicity”. The goals of the manuscript were to characterize ODP in human gut metagenomes and metatranscriptomes using pre-existing data sets.

In Ffigure 1, the authors search gut metagenomes for genes in type 1 or type 2 ODPs and assign to microbial taxa. The main conclusion is that type I genes occur in both bacteria and fungi, type I only in bacteria. In Figure 2, the authors then show that most human gut samples have ODP genes and that type II genes are much more prevalent that type I. Focusing on type 2 ODP, in Figure 3 the authors then assign type 2 ODP genes to various bacterial taxa, mostly belonging to E. coli (Ec) and Oxalobacter formigenes (Of). Comparing metagenomics (MGX) and metatranscriptomics (MTX), the authors show that while for Of, gene content and transcript correlated, for Ec transcript as not often detected and that transcript were dominated in MTX by

Of. The authors then in Figure 4 analyze data from IBD case and control cohorts (iHMP-IBD study), show that oxalate concentrations are higher in feces from people with IBD than control. Finally, in IBD compared to controls, the authors find decreased expression of type II ODP in IBD compared to controls. The reduced expression of ODP genes in IBD was assigned to Of (Figure 5).

General Commentary

The goals of this study characterizing ODP in human gut metagenomes and metatranscriptomes were met. This referee cannot formally critique the multi-omics methods or approach as not my area of expertise. I will critique this at the level of the novelty of the conclusions from this study, rather than a technical critique. There have been a number of microbiome compositional analysis associating oxalate nephrocalcinosis known of which were cited and this lack of citation misrepresents the field. The assignment of ODP genes to various taxa is a first comprehensive bioinformatics analysis, but not entirely novel, i.e. that organisms known to have this pathway were identified. The finding that ODP transcripts are dominated by Of compared to Ec is again “new” information but not necessarily unexpected as Of has an obligate requirement for oxalate and thus must express ODP genes for survival while Ec does not.

We appreciate the reviewer’s comments. Although we agree that some oxalatedegrading microbes have been individually characterized in vitro before, we believe our study provided novel insights with more depth and breadth, based on the following ideas:

1. We analyzed six ODPs that belong to two distinct types, and reached the novel conclusion that type I ODPs are encoded by both fungi and bacteria, but type II is exclusively encoded by bacteria, and that the human microbiome is associated with type II ODP.

2. Our study focuses on transcription of multiple oxalate-degrading enzymes in vivo using metatranscriptomics data. Such transcriptional evidence directly informs the functional potential of microbiota oxalate degradation. In contrast, the previous “microbiome compositional analysis” mentioned by the reviewer focused on microbiome population differences based on genetic-level changes (16S rRNA, and metagenomic sequencing). The lack of correlation between ODP genes and transcripts demonstrates that transcriptional data are critical when assessing the activity of these genes. Collectively,

this evidence illustrates the robustness of our analyses compared to prior studies that used gene-based measurements.

3. Although multiple oxalate-degrading microbes have been identified, as the reviewer pointed out, the relative importance of each taxon has been in debate for decades(Miller and Dearing, 2013). Therefore, there have been both academic and commercial efforts using several other bacteria to alleviate oxalate toxicity (Cho, Gebhart, Furrow, and Lulich, 2015; Ellis, Shaw, Jackson, Daniel, and Knight, 2015; Lieske, 2017; Miller, Choy, Penniston, and Lange, 2019; Ticinesi, Nouvenne, and Meschi, 2019). We now provide clear evidence that demonstrates the dominant role of O. formigenes in oxalate degradation within the human microbiota. Moreover, we demonstrate potential clinical indications (e.g. Crohn’s disease of the ileocolonic phenotype, and Ulcerative colitis) when O. formigenes is lost.

Entirely Of the most novel findings of the study are the inverse correlation between Of ODP gene expression and oxalate levels in IBD. This raises interesting questions whether dysfunction in Of ODP gene expression may lead to increased gut oxalate in IBD but not healthy guts. Furthermore, why would Of not express these genes in these situations?

We thank the reviewer for this comment. Indeed, our findings indicate for the first time that microbiota-based oxalate degradation is significantly associated with the elevated fecal oxalate levels in IBD patients. The ODP of O. formigenes, which was the dominating ODPtranscribing microbe in health, was detected at significantly lower levels in IBD patients. The biological nature of O. formigenes as an oxalate autotroph (Cornick and Allison, 1996), requires that it continually transcribe ODP for both carbon source and for energy, leading to ATP production (Cornick and Allison, 1996). Therefore, the lack of O. formigenes ODP transcription most likely indicates its absence in IBD patients. Previous studies have demonstrated its loss of colonization is associated with antibiotic use, which is common in IBD patients (Liu et al., 2017). Our study now raises testable hypotheses that common treatments including antibiotics and other medications with antibacterial activities have caused the loss of these bacteria in IBD patients.

In general, I am under impressed by this manuscript. The description of ODP pathways in metagenomes and the expression by Oxalobacter formigens and E. coli is not particularly unexpected. The inverse relationship between oxalate levels in IBD and Oxalobacter expression is interesting but again I can't equate this finding with the level of findings of this journal.

We appreciate the reviewer’s time and comments. We have provided evidence (above) of how our study has provided novel insights in this area in both health and disease, and have included new data from a mouse model supporting the important in vivo role of O. formigenes in oxalate homeostasis (new Figure 6 in on page 39).

References:

Binder, H. J. (1974). Intestinal oxalate absorption. Gastroenterology, 67(3), 441-446.

Canales, B. K., and Hatch, M. (2017). Oxalobacter formigenes colonization normalizes oxalate excretion in a gastric bypass model of hyperoxaluria. Surg Obes Relat Dis, 13(7), 1152-1157. doi:10.1016/j.soard.2017.03.014

Cho, J. G., Gebhart, C. J., Furrow, E., and Lulich, J. P. (2015). Assessment of in vitro oxalate degradation by Lactobacillus species cultured from veterinary probiotics. Am J Vet Res, 76(9), 801-806. doi:10.2460/ajvr.76.9.801

Cornick, N. A., and Allison, M. J. (1996). Assimilation of oxalate, acetate, and CO2 by Oxalobacter formigenes. Can J Microbiol, 42(11), 1081-1086. doi:10.1139/m96-138

Cury, D. B., Moss, A. C., and Schor, N. (2013). Nephrolithiasis in patients with inflammatory bowel disease in the community. Int J Nephrol Renovasc Dis, 6, 139-142. doi:10.2147/IJNRD.S45466

Ellis, M. L., Shaw, K. J., Jackson, S. B., Daniel, S. L., and Knight, J. (2015). Analysis of Commercial Kidney Stone Probiotic Supplements. Urology, 85(3), 517-521. doi:https://doi.org/10.1016/j.urology.2014.11.013

Knauf, F., Ko, N., Jiang, Z., Robertson, W. G., Van Itallie, C. M., Anderson, J. M., and Aronson, P. S. (2011). Net intestinal transport of oxalate reflects passive absorption and SLC26A6-mediated secretion. J Am Soc Nephrol, 22(12), 2247-2255. doi:10.1681/ASN.2011040433

Lieske, J. C. (2017). Probiotics for prevention of urinary stones. Annals of translational medicine, 5(2), 29-29. doi:10.21037/atm.2016.11.86

Liu, M., Koh, H., Kurtz, Z. D., Battaglia, T., PeBenito, A., Li, H.,... Blaser, M. J. (2017). Oxalobacter formigenes-associated host features and microbial community structures examined using the American Gut Project. Microbiome, 5(1), 108. doi:10.1186/s40168-017-0316-0

Miller, A. W., Choy, D., Penniston, K. L., and Lange, D. (2019). Inhibition of urinary stone disease by a multi-species bacterial network ensures healthy oxalate homeostasis. Kidney Int, 96(1), 180-188. doi:10.1016/j.kint.2019.02.012

Miller, A. W., and Dearing, D. (2013). The metabolic and ecological interactions of oxalate-degrading bacteria in the Mammalian gut. Pathogens, 2(4), 636-652. doi:10.3390/pathogens2040636

Saunders, D. R., Sillery, J., and McDonald, G. B. (1975). Regional differences in oxalate absorption by rat intestine: evidence for excessive absorption by the colon in steatorrhoea. Gut, 16(7), 543-548. doi:10.1136/gut.16.7.543

Ticinesi, A., Nouvenne, A., and Meschi, T. (2019). Gut microbiome and kidney stone disease: not just an Oxalobacter story. Kidney Int, 96(1), 25-27. doi:https://doi.org/10.1016/j.kint.2019.03.020

https://doi.org/10.7554/eLife.63642.sa2

Article and author information

Author details

  1. Menghan Liu

    1. NYU Langone Health, New York, United States
    2. Vilcek Institute of Graduate Biomedical Sciences, New York, United States
    Present address
    Department of Biological Sciences, Columbia University, New York, United States
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9390-8194
  2. Joseph C Devlin

    1. NYU Langone Health, New York, United States
    2. Vilcek Institute of Graduate Biomedical Sciences, New York, United States
    Contribution
    Data curation
    Competing interests
    No competing interests declared
  3. Jiyuan Hu

    NYU Langone Health, New York, United States
    Contribution
    Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Angelina Volkova

    1. NYU Langone Health, New York, United States
    2. Vilcek Institute of Graduate Biomedical Sciences, New York, United States
    Contribution
    Data curation
    Competing interests
    No competing interests declared
  5. Thomas W Battaglia

    NYU Langone Health, New York, United States
    Contribution
    Data curation
    Competing interests
    No competing interests declared
  6. Melody Ho

    NYU Langone Health, New York, United States
    Contribution
    Data curation
    Competing interests
    No competing interests declared
  7. John R Asplin

    Litholink Corporation, Laboratory Corporation of America Holdings, Chicago, United States
    Contribution
    Data curation, Writing - review and editing
    Competing interests
    is an employee of Litholink
  8. Allyson Byrd

    Department of Cancer Immunology, Genentech Inc, South San Francisco, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Investigation, Writing - review and editing
    Competing interests
    is an employee of Genentech
  9. P'ng Loke

    NYU Langone Health, New York, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6211-3292
  10. Huilin Li

    NYU Langone Health, New York, United States
    Contribution
    Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  11. Kelly V Ruggles

    NYU Langone Health, New York, United States
    Contribution
    Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0152-0863
  12. Aristotelis Tsirigos

    NYU Langone Health, New York, United States
    Contribution
    Conceptualization, Resources, Data curation, Supervision, Funding acquisition, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
  13. Martin J Blaser

    Center for Advanced Biotechnology and Medicine, Rutgers University, New York, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Investigation, Writing - review and editing
    For correspondence
    martin.blaser@cabm.rutgers.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2447-2443
  14. Lama Nazzal

    NYU Langone Health, New York, United States
    Contribution
    Conceptualization, Resources, Data curation, Supervision, Funding acquisition, Investigation, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    Lama.Nazzal@nyulangone.org
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0106-5060

Funding

National Institute of Allergy and Infectious Diseases (NIAID) (U01AI22285)

  • Martin J Blaser

National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) (R01DK110014)

  • Huilin Li

Rare Kidney Stone Consortium (U54 DK083908)

  • Lama Nazzal

The C & D and Zlinkoff Funds

  • Martin J Blaser

Oxalosis and Hyperoxaluria Foundation (career development grant)

  • Lama Nazzal

TransAtlantic Partnership of the Fondation LeDucq

  • Martin J Blaser

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dr. David Goldfarb and Xuhui Zheng for their helpful comments. We thank Dr. Tim Borbet for the helpful discussion and input. Funding: This study was supported in part by U01AI22285, R01DK110014, and the Rare Kidney Stone Consortium (U54 DK083908) from the National Institutes of Health, by the C and D and Zlinkoff Funds, Oxalosis and Hyperoxaluria Foundation-American Society of Nephrology career development grant, and the TransAtlantic Partnership of the Fondation LeDucq.

Ethics

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee (IACUC) protocols (#IA16-00822) of the New York University Langone Medical Center.

Senior Editor

  1. Wendy S Garrett, Harvard T.H. Chan School of Public Health, United States

Reviewing Editor

  1. Peter Turnbaugh, University of California, San Francisco, United States

Reviewers

  1. Denise Dearing, University of Utah, United States
  2. Eric Brown, Broad Institute, United States

Publication history

  1. Received: October 1, 2020
  2. Accepted: March 23, 2021
  3. Accepted Manuscript published: March 26, 2021 (version 1)
  4. Version of Record published: April 22, 2021 (version 2)

Copyright

© 2021, Liu et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,255
    Page views
  • 153
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Computational and Systems Biology
    2. Epidemiology and Global Health
    Ewan Colman et al.
    Research Article

    Humans and other group-living animals tend to distribute their social effort disproportionately. Individuals predominantly interact with a small number of close companions while maintaining weaker social bonds with less familiar group members. By incorporating this behavior into a mathematical model, we find that a single parameter, which we refer to as social fluidity, controls the rate of social mixing within the group. Large values of social fluidity correspond to gregarious behavior, whereas small values signify the existence of persistent bonds between individuals. We compare the social fluidity of 13 species by applying the model to empirical human and animal social interaction data. To investigate how social behavior influences the likelihood of an epidemic outbreak, we derive an analytical expression of the relationship between social fluidity and the basic reproductive number of an infectious disease. For species that form more stable social bonds, the model describes frequency-dependent transmission that is sensitive to changes in social fluidity. As social fluidity increases, animal-disease systems become increasingly density-dependent. Finally, we demonstrate that social fluidity is a stronger predictor of disease outcomes than both group size and connectivity, and it provides an integrated framework for both density-dependent and frequency-dependent transmission.

    1. Computational and Systems Biology
    2. Neuroscience
    Sean R Bittner et al.
    Research Article

    A cornerstone of theoretical neuroscience is the circuit model: a system of equations that captures a hypothesized neural mechanism. Such models are valuable when they give rise to an experimentally observed phenomenon -- whether behavioral or a pattern of neural activity -- and thus can offer insights into neural computation. The operation of these circuits, like all models, critically depends on the choice of model parameters. A key step is then to identify the model parameters consistent with observed phenomena: to solve the inverse problem. In this work, we present a novel technique, emergent property inference (EPI), that brings the modern probabilistic modeling toolkit to theoretical neuroscience. When theorizing circuit models, theoreticians predominantly focus on reproducing computational properties rather than a particular dataset. Our method uses deep neural networks to learn parameter distributions with these computational properties. This methodology is introduced through a motivational example of parameter inference in the stomatogastric ganglion. EPI is then shown to allow precise control over the behavior of inferred parameters and to scale in parameter dimension better than alternative techniques. In the remainder of this work, we present novel theoretical findings in models of primary visual cortex and superior colliculus, which were gained through the examination of complex parametric structure captured by EPI. Beyond its scientific contribution, this work illustrates the variety of analyses possible once deep learning is harnessed towards solving theoretical inverse problems.