Diet posttranslationally modifies the mouse gut microbial proteome to modulate renal function

See allHide authors and affiliations

Science  18 Sep 2020:
Vol. 369, Issue 6510, pp. 1518-1524
DOI: 10.1126/science.abb3763

Microbiota protect the kidneys

Chronic kidney disease (CKD) afflicts millions of people globally. The first-line treatment for CKD is dietary intervention, so there may be a gut microbiota–associated component. Lobel et al. investigated the mechanistic links between the microbiota and protein intake, because the protein metabolites indole and indoxyl sulfate are known uremic toxins (see the Perspective by Pluznick). The authors used a mouse model of CKD precipitated by a paucity of the dietary sulfur–containing amino acids methionine and cysteine. Bacterial metabolism of sulfur-containing amino acids modulated indole production by sulfide inhibition of the enzyme tryptophanase, thus abrogating uremic toxicity by this metabolite in this model system.

Science, this issue p. 1518; see also p. 1426


Associations between chronic kidney disease (CKD) and the gut microbiota have been postulated, yet questions remain about the underlying mechanisms. In humans, dietary protein increases gut bacterial production of hydrogen sulfide (H2S), indole, and indoxyl sulfate. The latter are uremic toxins, and H2S has diverse physiological functions, some of which are mediated by posttranslational modification. In a mouse model of CKD, we found that a high sulfur amino acid–containing diet resulted in posttranslationally modified microbial tryptophanase activity. This reduced uremic toxin–producing activity and ameliorated progression to CKD in the mice. Thus, diet can tune microbiota function to support healthy host physiology through posttranslational modification without altering microbial community composition.

Chronic kidney disease (CKD) affects nearly 850 million people worldwide (1). Although dietary modification is a cornerstone of CKD treatment, the mechanistic roles of diet–microbiota interactions in CKD pathogenesis and treatment have been underexplored. Many diet–microbiome studies have focused on the effects of dietary fiber, fat, and carbohydrates (2). Less is known about the specific effects of dietary protein and amino acids, although 5 to 10% of dietary amino acids reach the colon, where most gut bacterial metabolism occurs (3). In humans, increasing dietary protein increases gut bacterial production of hydrogen sulfide (H2S), indole, and indoxyl sulfate (4, 5). Indole and indoxyl sulfate are uremic toxins, and H2S has diverse physiological functions, some of which are mediated by the posttranslational modification S-sulfhydration (6, 7). Although a vast number of studies have been performed in mammalian systems, the physiological roles of H2S in regulating gut bacterial function within a host are understudied. Additionally, whether there are bona fide opportunities to improve CKD by manipulating diet–microbiota interactions remains unclear.

Modifying patient dietary protein intake has been a clinical strategy used in the management of CKD for many decades, and more specifically, dietary sulfur–containing amino acid (Saa) intake has been reported to affect CKD progression in patients and disease models (8, 9). Given the gaps in understanding the relationships between dietary protein, gut microbial metabolism, and H2S and to address the role of gut microbial metabolism and diet in renal function, we used a mouse model of CKD that is driven by increased adenine (10) along with a Saa-based diet perturbation. We formulated isocaloric diets to represent extremes of mouse Saa consumption, relevant to human consumption, i.e., diets with low versus high amounts of methionine and cysteine (see table S1 for diet formulations and materials and methods) but with sufficient methionine to avoid methionine restriction (11). Conventionally reared, specific-pathogen-free (SPF) mice on a low–Saa plus adenine (Saa+Ade) diet had significantly increased serum creatinine levels compared with mice on a high-Saa+Ade (Fig. 1A), as well as more extensive and severe renal cortex histopathologic changes, including tubular dilatation and dropout, tubulitis with peritubular fibrosis, and cortical crystal deposition (Fig. 1, B to D). To determine the extent to which the Saa effects depend on the gut microbiota, we fed the Saa+Ade diets to gnotobiotically reared, germ-free (GF) mice. Serum creatinine and kidney damage were markedly reduced in the GF mice compared with SPF mice on the low-Saa+Ade diet, and there were similar phenotypes in the GF and SPF mice fed the high-Saa+Ade diet (Fig. 1, A to D). Given that the GF mice on the low-Saa + Ade diet still showed renal injury, although less than SPF mice, we examined the expression of a select panel of host genes implicated in CKD pathogenesis in both humans and the mouse adenine model (12). Spp1 (Osteopontin), Tgfb1, and Icam1 were increased in GF mice to greater levels than in SPF mice on the low-Saa+Ade diet (fig. S3D). By contrast, Ccl2 and Timp1 were increased to a greater extent in SPF mice on the low-Saa+Ade diet (fig. S3D). These data indicate that the microbiota may buffer expression of some host genes while stimulating expression of others. Overall, we found that a low-Saa diet exacerbated the CKD phenotypes observed and that the presence of a gut microbiota further magnified these effects.

Fig. 1 Dietary Saa and the gut microbiota modulate kidney injury severity in a mouse CKD model.

(A) Serum creatinine (Cre) levels of SPF and GF mice on low- versus high-Saa+Ade diet. (B and C) Representative hematoxylin and eosin (H&E) staining (B) and representative trichrome staining (C) of kidneys from mice in (A). (D) Histology-based renal injury scores. (E and F) SPF (E) and GF (F) mouse cecal sulfide levels detected in the lead acetate or methylene blue assay. (G) Normalized E. coli mean gene abundance in CKD patient samples compared with non-CKD controls (PTRI whole-genome shotgun sequencing dataset). (H) Serum Cre levels from ASF or ASFE. coli mice on low- versus high-Saa+Ade diets. (I and J) Representative H&E staining (I) and trichrome staining (J) of kidneys from mice in (H). (K) Histology-based renal injury scores. (L and M) ASF and ASFE. coli cecal sulfide levels detected by lead acetate (L) or methylene blue assay (M). Data represent two [(L) and (M)], three [(A), (D), (H), and (K)], or four [(E) and (F)] independent experiments. Symbols represent individual mice. Bars represent means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. Two-way analysis of variance (ANOVA) with Tukey’s post hoc test was used for (A), (D), (E), (F), (H), and (K); the Mann-Whitney test for was used for (L) and (M). Scale bar, 1 mm for the 40× magnification in (B) and 200 μm for the 200× magnification in (B) and (C).

We tested whether there is a link between dietary Saa and gut bacteria through microbial metabolism of cysteine to H2S. We measured cecal sulfide levels from GF and SPF mice fed low- versus high-Saa diets by using both the lead acetate and methylene blue sulfide detection assays (13). SPF mice on the high-Saa diet had higher cecal sulfide levels than those on the low-Saa diet (Fig. 1, E and F). GF mouse ceca had significantly less sulfide than SPF mice, regardless of Saa diet (Fig. 1, E and F). We did not observe any significant differences in the taxonomic abundances of the gut microbiota members between SPF mice on the low- versus high-Saa diets using 16S ribosomal RNA (rRNA) gene amplicon surveys (fig. S1), supporting the idea that the differences in cecal sulfide in healthy mice may be mediated by altered microbial function rather than changes in microbiota community structure.

Given these findings and with the goal of more effectively modeling gut microbial activity shifts that could occur in CKD patients, we sought out publicly available CKD patient gut microbiota profiling studies to identify taxa enriched in CKD patients compared with healthy individuals. We re-analyzed fecal 16S rRNA gene amplicon datasets from Xu et al. (14) and from Southern Medical University (NCBI accession PRJEB5761), a fecal PhyloChip study by Vaziri et al. (15), and a fecal whole-genome shotgun sequencing dataset from Promegene Translational Research Institute (PTRI) (NCBI accession PRJNA449784). Enforcing stringent statistical cutoffs [linear discriminant analysis (LDA) score of >4 for LDA effect size and fold change of >2 for the PhyloChip analysis] revealed a clear and robust signal of Enterobacteriaceae enrichment in CKD patients (fig. S2, A to C). Although the 16S rRNA gene amplicon analyses did not afford strain-level Escherichia coli identification, the PhyloChip analysis showed a significant increase in the combined mean abundance of seven E. coli strains measured in fecal samples of CKD patients with end-stage renal disease compared with control subjects (fig. S2D). Further analysis of the PTRI whole-genome shotgun sequencing dataset strengthened this finding, as we found a higher normalized E. coli mean gene abundance in CKD patient samples than in non-CKD controls (Fig. 1G). Given these findings from reanalysis of human CKD gut microbiota and the genetic tractability and relatively well-characterized proteome of E. coli, we focused on the effects of E. coli in the adenine-driven CKD model. As the mice we obtain from Jackson Laboratory do not harbor any Enterobacteriaceae members (fig. S1) and to carry out a carefully controlled study of gut microbial activity in a reproducible model system, we used gnotobiotically reared mice colonized with the altered Schaedler flora (ASF) to which we added E. coli K-12. The ASF is a simplified microbial community consisting of eight bacterial species, none of which is related to Enterobacteriaceae (16). We used ASF mice, rather than monocolonized mice, because ASF mice are more physiologically similar to SPF mice (16). E. coli colonization was similar on the low- and high-Saa diets with and without adenine (fig. S3, A and B), and we did not observe changes in the relative abundance of ASF members (fig. S3C). On the low-Saa+Ade diet, ASF mice colonized with E. coli (ASFE. coli) had higher serum creatinine and more extensive tubulitis, tubular atrophy and dropout, peritubular fibrosis, and cortical crystals than ASF mice (Fig. 1, H to K). By contrast, ASFE. coli and ASF mice on the high-Saa+Ade diet had similar serum creatinine levels and milder renal parenchymal pathology compared with their littermates on the low-Saa+Ade diet (Fig. 1, H to K). As with SPF mice, we found higher cecal sulfide levels in ASFE. coli mice on the high- versus low-Saa+Ade diet (Fig. 1, L and M). To determine if changes in renal function would occur in these models in the absence of the adenine insult, we examined serum creatinine levels in SPF and ASFE. coli mice on the low- versus high-Saa diet. Notably, the low-Saa diet and E. coli were sufficient to increase serum creatinine levels in mice, and no overt histologic abnormalities were present (fig. S3E). Similar results were obtained with SPF mice on the Saa diets (fig. S3F). Overall, these results support that E. coli interacts with dietary Saa to modulate kidney function.

Given our observations regarding cecal H2S in SPF and ASFE. coli mice on the Saa diets and the literature on how H2S can posttranslationally modify mammalian proteins leading to a range of physiologic effects, we investigated the effects of H2S on E. coli. In lead acetate sulfide detection assays, E. coli, grown either aerobically or anaerobically, produced sulfide from cysteine in a dose-dependent manner without any effects on growth (Fig. 2A and fig. S4, A to C). To serve as a control for the effects of endogenous H2S production on E. coli physiology, we generated an isogenic strain harboring a deletion of decR, which encodes a transcriptional activator of the cysteine desulfhydrase genes yhaO and yhaM, which drive cysteine-derived sulfide production in E. coli (17). Deletion of decR resulted in significant reduction of sulfide production, with no effect on growth kinetics (Fig. 2, A and B; fig. S4, A to D). Of note, SseA (3-mercaptopyruvate sulfurtransferase), which has been reported as a major source of cysteine-derived sulfide in E. coli (18), did not affect sulfide production in our system (fig. S4E). Sulfide exerts its effects through generation of polysulfides that modify cysteine residues, resulting in S-sulfhydration (19). To identify E. coli proteins that are S-sulfhydrated (R-S-S), we adapted a pulldown method that specifically enriches for S-sulfhydrated proteins (20). This technique leverages maleimide binding to free thiols and persulfides, resulting in thioether bonds, and the ability of dithiothreitol (DTT) to break disulfide bonds but not thioether bonds (Fig. 2C). We observed a robust enrichment of S-sulfhydrated proteins in DTT-eluted samples of wild-type (WT) E. coli lysates grown in medium supplemented with cysteine (Fig. 2D). We validated the pulldown assay’s specificity and found that treating bacterial lysates with H2O2, and hence oxidizing free thiols, reduced the detection of S-sulfhydrated proteins (fig. S4D). By contrast, treatment with sodium hydrosulfide (NaHS), a fast-reacting sulfide donor, induced higher S-sulfhydration levels in bacterial lysates (fig. S4D). We detected a higher level of S-sulfhydration in E. coli lysates grown in medium supplemented with cysteine than in E. coli grown in Luria broth (LB) alone (fig. S4, E and F). By contrast, lysates of ΔdecR bacteria, which produce less H2S, grown in cysteine-supplemented LB broth showed less S-sulfhydration than WT E. coli (Fig. 2E). Next, we sought to characterize the E. coli sulfhydrome using quantitative tandem mass tag (TMT) liquid chromatography–multistaged mass spectometry (LC-MS3) analysis. This analysis revealed that most identified proteins enriched in E. coli lysates that were eluted with DTT were indeed S-sulfhydrated, compared with the same samples not treated with DTT (Fig. 2F). Furthermore, most detected S-sulfhydrated proteins were enriched in WT versus ΔdecR E. coli, as expected from the strains’ different abilities to produce sulfide from cysteine. Ranking of the S-sulfhydrated proteins by their q values (DTT versus non-DTT) revealed the top 10 most abundant S-sulfhydrated proteins (Fig. 2F and fig. S4G). Although most of these proteins are highly expressed during logarithmic bacterial growth and are expected to be highly abundant, tryptophanase (TnaA) was overrepresented. Overall, our quantitative proteomics analysis identified 212 proteins as S-sulfhydrated with high confidence (table S2), and hypergeometric distribution analysis revealed 13 cellular pathways enriched with S-sulfhydrated proteins, several of which are related to protein translation (fig. S4H).

Fig. 2 Characterization of E. coli S-sulfhydrome reveals that TnaA is a highly S-sulfhydrated protein.

(A) E. coli sulfide production, determined with lead acetate. (B) E. coli sulfide production, determined with methylene blue. OD670, optical density at 670 nm. (C) Schematic of S-sulfhydrated protein pulldown method. The S-sulfhydrated protein is highlighted in blue, to distinguish it from the native protein in red. (D) Silver staining of E. coli lysates subjected to S-sulfhydration pulldown and eluted either with or without DTT. m.w., molecular weight. (E) Silver staining of lysates from WT or ΔdecR E. coli lysates subjected to S-sulfhydration pulldown. (F) Heat map of the relative quantity of the 212 S-sulfhydrated proteins by TMT LC-MS3 analysis from S-sulfhydration pulldown fractions from WT E. coli samples eluted with or without DTT and ΔdecR mutant samples eluted with DTT. Proteins ordered based on q score for enrichment in the DTT-eluted versus non-DTT–eluted samples. Data represent two (E), three [(D) and (F)], four (A), or six (B) independent experiments. Bars represent means ± SEM. **P < 0.01. Statistical significance was determined with a linear model test (A), two-way Kruskal-Wallis test with Dunn’s post hoc test (B), or two-way ANOVA with Tukey’s post hoc test (F).

TnaA (Fig. 2F), a secreted enzyme that catalyzes the degradation of tryptophan to indole, pyruvate, and ammonia, offered a potential connection between our S-sulfhydrome analysis and the phenotypes we observed in the CKD mouse model. Indoles are a class of bacteria-produced molecules that not only regulate bacterial physiology (21) but also participate in bacteria–host interactions (22). Indoles can be transported through the portal vein to the liver, where they are oxidized, yielding the uremic toxin indoxyl sulfate (23). For these reasons, TnaA emerged as an attractive target for investigating host–microbe interactions in our CKD mouse model. We replaced the E. coli TnaA chromosomal copy with a cloned tnaA-his under its native promoter. We then validated our S-sulfhydrome results by analyzing TnaA S-sulfhydration in WT versus ΔdecR E. coli lysates using Western blot analysis and found reduced TnaA S-sulfhydration in ΔdecR lysates (Fig. 3A). E. coli lysates treated with H2O2 and NaHS showed reduced and increased TnaA S-sulfhydration, respectively (Fig. 3B). Since the S-sulfhydration pulldown method reduces the S-sulfhydrated cysteine residue (i.e., removes the S-sulfhydration), we could not pinpoint the exact cysteine residues being S-sulfhydrated, as TnaA has seven cysteines. Therefore, we purified natively expressed TnaA-His from E. coli grown in LB supplemented with cysteine and performed LC–tandem MS (LC-MS/MS) analysis to detect and map the S-sulfhydration. We detected several TnaA-His peptides that had a +32-Da addition, matching the molecular weight of S-sulfhydration on a cysteine residue (fig. S5). As oxidation of a cysteine residue to sulfinic acid (R-S-O2) results in same mass shift and given the potential for oxidation during our analysis, we could not rule out that such oxidation occurs. However, an S-sulfhydrated cysteine can be oxidized to sulfinic acid (R-S-S-O2), resulting in a +64-Da increase, a shift that results from oxidation of an S-sulfhydrated cysteine or a second S-sulfhydration (R-S-S-SH). We were able to detect a +64-Da shift in several cysteine residues of TnaA (fig. S5 and table S3). Although we found evidence that six of the seven TnaA cysteine (C) residues were S-sulfhydrated (C148, C281, C294, C298, C352, and C383), we could not rule out that cysteine residue C413 is also S-sulfhydrated, as our coverage of TnaA (~78%) did not include peptides with high confidence within this region. TnaA cysteine residues are involved in its enzymatic activity (24), as mutation of C298 results in altered TnaA activity (25). To study the effect of S-sulfhydration on TnaA activity, we measured indole concentrations by both Kovac’s reagent method and LC-MS/MS analysis of bacterial cultures. We found that supplementing LB broth with cysteine or NaHS reduced indole concentrations in the supernatants (Fig. 3, C and D). Also, supporting sulfide’s role in TnaA inhibition, ΔdecR E. coli had higher indole levels than WT E. coli when grown in LB supplemented with cysteine (fig. S6A), and TnaA expression was similar under these conditions (fig. S6B). To show that indole production was dependent on TnaA, we ablated TnaA activity by using an isogenic tnaA739::kan mutant (tnaA mut) and did not detect indole in the culture supernatant (fig. S6C). To test that S-sulfhydration inhibits TnaA activity in a direct manner, we used a reductionist approach using purified E. coli TnaA. We observed that incubation with disodium tetrasulfide (Na2S4), a polysulfide donor, led to TnaA S-sulfhydration (fig. S6D) and reduced enzymatic activity by 60% in vitro (Fig. 3E). As an assay control, we added DTT, which should reduce S-sulfhydrated TnaA to its functional native form, and we observed TnaA activity more than tripled (Fig. 3E). To provide a more physiological context for TnaA inhibition by cysteine-derived sulfide, we measured the activity of TnaA purified from WT and ΔdecR E. coli cultures grown with cysteine and found that TnaAΔdecR had higher activity (fig. S6E). Collectively, these results support that S-sulfhydration of E. coli TnaA reduces its activity as measured by indole production from tryptophan, both in vitro and in bacterial cultures.

Fig. 3 S-Sulfhydration inhibits the indole-producing enzymatic activity of E. coli TnaA.

(A) Representative Western blot analysis of TnaA-His from WT and ΔdecR E. coli lysates subjected to S-sulfhydration pulldown. Loading controls show RpoD in the flowthrough. (B) Same method as in (A) but with E. coli lysates treated with NaCl, H2O2, or NaHS. (C) LC-MS/MS analysis of indoles in WT E. coli cultures with cysteine or NaHS. (D and E) Kovac’s assay for indole production in WT E. coli cultures with cysteine or NaHS (D) or with purified TnaA enzyme supplemented with NaCl, Na2S4, or DTT (E). Data represent three [(A) and (E)], four [(B) and (D)], or five (C) independent experiments. Bars represent means ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. Data were analyzed with the Mann-Whitney test (A), two-way ANOVA with Tukey’s post hoc test [(B) and (E)], or two-way Kruskal-Wallis test with Dunn’s post hoc test [(C) and (D)].

Although we detected TnaA S-sulfhydration in vitro for both purified protein and TnaA from bacterial cell lysates and demonstrated that this modification inhibited its activity, we had yet to determine if this posttranslational modification occurred within the gut in response to dietary Saa and resulted in measurable physiological consequences for the host. To investigate this, we began by providing ASFE.coli mice with the high- and low-Saa diets. Although mice on the diets harbored similar levels of E. coli (fig. S7A), we detected higher TnaA S-sulfhydration in the cecal contents of mice on the high-Saa diet than in those on the low-Saa diet (Fig. 4A). None of the eight ASF bacterial genomes encode a tnaA gene, and we could not detect indoles in the ASF mouse cecal contents using LC-MS/MS, implying that E. coli is the sole producer of indoles in this model (fig. S7B). Taking advantage of that distinction, we measured indole in the cecal contents of ASFE. coli mice on the two diets. We found that mice on the high-Saa diet had significantly lower indole levels, demonstrating that high dietary Saa not only increased TnaA S-sulfhydration but also that this modification was sufficient to affect TnaA activity in vivo (Fig. 4, B and C). To strengthen the links between diet, microbial metabolism, and kidney damage, we leveraged the CKD mouse model by using the low-Saa+Ade diet, with which we observed the most renal injury, and the gnotobiotic ASF mice we used previously (Fig. 1). We colonized these mice with either WT E. coli (ASFE.coli) or with one of two isogenic mutants, tnaA mut or ΔdecR (ASFtnaA mut or ASFΔdecR, respectively). Unlike the ASFE.coli mice, no indoxyl sulfate was detected in the serum of ASFtnaA mut mice, as there was no tryptophanase present within the gut microbiota. As E. coli ΔdecR is deficient in producing sulfide from cysteine, TnaA remains less S-sulfhydrated and more highly active (fig. S6A). Consistent with that observation, ASFΔdecR mice showed reduced cecal sulfide (fig. S7C), increased cecal indole (fig. S7D), and increased serum indoxyl sulfate relative to ASFE. coli mice (Fig. 4D). Mice colonized with WT E. coli had higher serum creatinine levels than mice colonized with the tnaA mut strain (Fig. 4E). Concomitant with the serum indoxyl sulfate levels, mice colonized with the ΔdecR strain had the highest serum creatinine levels (Fig. 4E). Colonization of the three different E. coli strains was similar (fig. S7E). Histological findings of more severe tubulointerstitial damage, fibrosis, cortical crystal deposition, and more extensive parenchymal involvement mirrored the trends observed for indoxyl sulfate and creatinine for E. coli ΔdecR versus WT E. coli (Fig. 4, F and G). We also examined these mice on the high-Saa+Ade diet. Consistent with prior observations (Fig. 1), ASFE.coli mice on this diet showed milder phenotypes, although ASFΔdecR mice had slightly increased serum creatinine compared with the parental and tnaA mut strains (fig. S7, F to K). Collectively, these data support that a dietary component can generate a posttranslational modification of a gut microbial protein that affects extraintestinal host function, and our findings also furnish mechanistic insight into how host–diet–microbiota interactions can contribute to disease states such as CKD.

Fig. 4 Dietary Saa modulate cecal indole levels, serum indoxyl sulfate levels, and kidney function in a mouse CKD model.

(A and B) Western blot analysis of TnaA of S-sulfhydration pulldown and flowthrough samples from cecal contents (A) and Kovac’s assay measurement of indole levels in cecal contents (B) from ASFE. coli mice on Saa diets. (C) LC-MS/MS analysis of indole levels in cecal contents from ASFE. coli mice on Saa diets. Left, spectra representative of an experiment with three mice per group and indole standard. (D) LC-MS measurements of serum indoxyl sulfate in ASF mice on the low-Saa+Ade diet and colonized with E. coli strains. (E) Serum creatinine (Cre) levels in ASF mice colonized with E. coli strains and on low-Saa+Ade diets. (F and G) Representative H&E staining (F) and representative trichrome staining (F) of kidneys from mice in (E). (H) Histology-based renal injury scores. Data in (A), (B), (C), (D), (E), and (H) represent three independent experiments. Symbols represent individual mice. Bars represent means ± SEM. *P < 0.05; **P < 0.01. Data were analyzed with the Mann-Whitney test (A to D) or two-way ANOVA with Tukey’s post hoc test [(E) and (H)]. Scale bar, 1 mm [40× magnifications in (F)] or 200 μm [200× magnification in (F) and (G)].

We found that sulfide derived from bacterial metabolism of dietary Saa regulates E. coli indole production through inhibition of tryptophanase by S-sulfhydration. Our work shows that a dietary component can be metabolized by the microbiota to generate a posttranslational modification of microbial proteins that affects host physiology and offers a framework for how host–diet–microbiota interactions can contribute to or stave off disease states such as CKD (fig. S8). Although metatranscriptomics have provided a window into functional changes within a community, our results show that a single modification on one specific protein can mediate such effects. Thus, in our mouse CKD model a subtle dietary change, which does not result in microbial composition changes, reveals that production of indoles by E. coli is differentially affected by levels of sulfide endogenously produced by gut bacteria. Our work indicates that bacterial metabolism not only may have direct effects on host physiology but also may influence microbe–microbe interactions driven by bacterial posttranslational modifications mediated by host diet.

However, our findings need to be interpreted cautiously and their limitations noted. Our observations that GF mice exhibit a renal failure phenotype in the adenine model, though milder than in SPF mice, indicate that host factors, as well as redox status (e.g., altered glutathione levels) (8), also play a role in dietary Saa modulation of kidney function. Indole-independent dietary Saa effects on renal function are also indicated by the phenotype of ASF mice colonized with the E.coli tnaA mut strain; these mice lack indole and indoxyl sulfate yet still exhibit a CKD phenotype on the Saa+Ade diet. Thus, the contribution of dietary Saa and bacterial TnaA indole production to renal failure in human CKD warrants further investigation. The 2-fold increase in cecal indole culminating in a 10-fold serum indoxyl sulfate increase in the mouse CKD model is notable. The increase in indoxyl sulfate could be driven, in part, by indole-independent mechanisms in the adenine CKD mouse model that lead to a reduction in its excretion and accumulation in the serum. Alternatively, the kinetics of indole production in the cecum could be augmented by the rate of indole oxidation and accumulation in the liver and its release into the serum.

In summary, diet is a crucial aspect in managing CKD (26, 27), and we hypothesize that administration of TnaA inhibitors, such as sulfide donors, may help reduce gut indole levels and thus mitigate kidney damage. In support of this concept and its broad application, other gut bacteria, especially members of the Bacteriodetes phylum, also encode TnaA homologs (28), and a high degree of homology exists between bacterial TnaA alleles (fig. S9). Our study elucidates an interaction between diet, microbial metabolism, and kidney function that is mediated by posttranslational protein regulation. These findings might shed light on managing CKD and provide clinical approaches to improve human health that target the microbiota and the enzymatic activities of its proteome.

Supplementary Materials

Materials and Methods

Fig. S1 to S10

Tables S1 to S5

References (2963)

MDAR Reproducibility Checklist

References and Notes

Acknowledgments: We thank J. X. Wang (Small Molecule Core Mass-Spectrometry, Harvard University), B. Budnik and R. Robinson (Harvard Center for Mass Spectrometry Proteomics), and R. Kuntz and S. Thakurta (Thermo Fisher Scientific Center for Multiplexed Proteomics, Harvard University). We also thank L. Ricci for graphic design (fig. S8), J. K. Lang and Kate Rosinski (Harvard T.H. Chan Gnotobiotic Center for Mechanistic Microbiome Studies) for technical support, and members of the Garrett lab for their discussion of this work. Funding: This work was supported by a CRI Irvington postdoctoral fellowship to L.L., NIH T32 AI118692 and F31 DK121375 to Y.G.C., and R01CA202704 and R24DK110499 to W.S.G. Author contributions: Conceptualization: L.L. and W.S.G. Formal analysis: L.L. Funding acquisition: W.S.G. Investigation: L.L., Y.G.C., and J.N.G. Software: L.L. Supervision: W.S.G. Visualization: L.L., Y.G.C., and W.S.G. Writing – original draft: L.L. and W.S.G. Writing – review and editing: L.L., Y.G.C., J.N.G., and W.S.G. Competing Interests: The authors declare no competing financial interests. W.S.G. is on the science advisory boards of Kintai Therapeutics, Leap Therapeutics, Evelo Biosciences, Tenza Inc., and SanaRx. On 21 February 2020, related to this work, a patent application was filed, U.S. Application no. 62/979,638 . Data and materials availability: All data are available in the manuscript or the supplementary materials. Raw 16S rRNA gene amplicon sequences were deposited in NCBI SRA databank under the bioproject accession PRJNA603373. Bacterial strains generated for this study are available from the Garrett Laboratory under an MTA with Harvard University.

Stay Connected to Science

Navigate This Article