A Limited Effect of Chronic Renal Insufficiency on the Colon Microbiome

In this issue, Randall et al.¹ address whether experimental renal insufficiency reproducibly changes the colon microbiome. They report a meta-analysis of stool/colon microbiome data from studies of mice or rats. Data from ten studies reveal no common microbiome profile associated with renal insufficiency, and comparison of specific findings across studies shows only some weak qualitative similarities. Overall, their results weaken the hypothesis that a reproducibly defined set of microbiome alterations, or “uremic dysbiosis,” is a cause or consequence of renal insufficiency. Control animals in different studies are shown to have had widely different microbiomes, suggesting a basis for the variable findings among studies.

The findings of Randall et al. should be placed in the context of our rapidly expanding knowledge of the colon microbiota and its chemical products. The 1970s marked the beginning of genetic classification of bacteria on the basis of the sequence of the gene for the small subunit of ribosomal RNA (known as 16S rRNA).² This gene's highly conserved regions identify it as bacterial, while its interspersed variable regions act as a “molecular clock” and reveal the degree to which different microbes are genetically related. The application of PCR and automated DNA sequencing has revealed the existence of many thousands of new colon microbes, most of which cannot readily be grown in culture. The microbial composition of diverse types of samples (e.g., of stool, sea water, and soil) is therefore now most often established by relating the 16S rRNA gene sequences in DNA extracted from the samples to large databases of 16S rRNA gene sequences. Microbial species, or amplicon sequencing variants, are defined by their 16S rRNA sequences, while broader taxonomic groupings such as the phyla discussed by Randall et al. include microbes whose 16S rRNA sequences have a defined degree of similarity and are therefore more closely related to one another evolutionarily. Importantly, the functional properties of microbes cannot be perfectly inferred from their 16S rRNA sequences. Microbes with identical 16S rRNA gene sequences may have different metabolic capacities due to genetic changes that are independent of the 16S rRNA gene, such as acquiring a metabolic capacity through horizontal gene transfer.

Improved computational methods and reduced sequencing costs have spurred 16S rRNA analysis of the colon and other microbiomes in health and in disease. This work, as exemplified by the National Institutes of Health Human Microbiome Project launched in 2007, has opened a world of surprises.³ The 200 gram cell mass of the colon microbiome is seen to include hundreds of bacterial species in individual humans. The colon microbiome is somewhat limited in diversity, including only two to ten common phyla of the >100 microbial phyla identified to date. However the proportion of these phyla varies widely among healthy humans (Figure 1). Further studies have shown that the mix of colon microbes in individual adult humans remains relatively stable over years and tends to return toward its original composition after major perturbations such as those occasioned by antibiotic use. The health impact of the wide disparity among colon microbiomes in healthy humans in industrialized countries remains to be established. The causes of this disparity are also largely uncertain. Hypothesized influences include differences in the human host's gut motility and genetic makeup along with early exposure to different microbes. Diet is also an obvious potential influence. Of note, however, the variability among the colon microbiomes of healthy Americans was not reduced by exposure to a uniform modern American diet over 7 days.⁴ Exposure to different diets and for longer periods may produce greater changes, as observed with increased microbiome diversity when participants consumed high levels of fermented foods for 6 weeks.⁵

Figure 1.

Microbiomes that vary in composition are often functionally similar. The composition of 242 stool samples from healthy US adults (top) as assessed by 16S rRNA amplicon sequencing varies substantially more between individuals than functional capacity (bottom) as assessed by metagenomic sequencing. Each vertical bar represents one individual, and colors represent phyla (top) and broad functional categories predicted for gene sequences (bottom). The figure shows stool data abstracted from Figure 2 of the 2012 report of the Human Microbiome Project.³

Wide differences among healthy subjects have made it difficult to define a “healthy” microbiome. Extensive efforts to identify changes in the microbiome characteristic of disease states other than renal insufficiency have also had limited success. Perhaps not surprisingly, the most dramatic changes have been associated with active inflammatory bowel disease.⁶ Here is important to distinguish between the microbiome's composition as defined by the proportion of different phylogenetic groups and the microbiome's metabolic function(s). Continued developments in sequencing and computational methods now allow the genomes of different microbial species to be reconstructed from samples of mixed microbial DNA. This “metagenomic” analysis allows characterization of the microbiome according to the prevalence of all genes within the sample, including those involved in metabolic processes. A remarkable finding of the Human Microbiome Project was that microbiomes of widely differing phylogenetic composition (determined by 16S rRNA sequencing, Figure 1, top) encoded similar metabolic functions (determined by metagenomics, Figure 1 bottom). With the functional capacity of different microbes in mind, we should consider what might cause colon microbial alterations in CKD. Metagenomic analysis might confirm the suggestion that high urea and uric acid concentrations promote the growth of colon microbes possessing urease and uricase.⁷ However, even an analysis of the microbiome's genetic capacity does not provide a reliable measure of its metabolic activity. Assessment of colon microbial activity on the basis of analysis of mRNAs (metatranscriptomics), proteins (metaproteomics), or metabolites (metabolomics) has not been as widely used for reasons including the cost and technological challenges of these approaches. Indeed, 16S sequencing and metagenomics only partially avoid these challenges because variable recovery of microbial DNA from stool samples likely introduces error in the description of the microbiome's genetic profile.

The experimental data analyzed by Randall et al. were obtained largely in animals with about one-third normal renal function. As they describe, human studies have also so far not identified a reproducible change in the microbiome characteristic of more advanced renal failure. As they note, references to “uremic dysbiosis” thus seem premature. The colon microbiome may however contribute to the ill effects of renal failure without any change in its composition. It produces numerous compounds foreign to mammalian metabolism which, often after conjugation by the liver, are normally excreted by the kidneys.⁸ Well-known examples are the amino acid breakdown products indoxyl sulfate and p-cresol sulfate. As cited by Randall et al., human studies suggest that end-stage renal failure does not greatly affect the production of the best known of these compounds and that differences in microbial metabolite production in humans with and without renal function are due more to differences in diet than to renal failure.^9,10 The accumulation of such compounds has been considered to promote illness in two ways. First, their accumulation could cause “uremic” symptoms in patients with advanced CKD. Second, their accumulation could accelerate the progression of CKD. Other studies have suggested that increased entry into the circulation of endotoxin and other large microbially derived substances causes inflammation in advanced CKD.

Ultimately, we can prove that the colon microbiome causes illness only if we can modify illness by manipulating it. Such proof, which may be considered to represent fulfillment of a modified version of Koch postulates, has not been obtained in human renal disease. Human health benefit has been obtained from fecal transplantation for recurrent C. difficile infection. In renal disease, feeding dietary fiber has been found to reduce the production of selected colon-derived solutes in patients and to limit the progression of CKD in animals. Overall, the microbiome has proven hard to manipulate. Selected microbes introduced as probiotics in animals and humans with established microbiomes generally do not persist. Efforts are now directed toward the development of “designer microbiomes” with specific metabolic capacities which could eventually be installed in diseased humans to reprogram metabolic output and alter aspects of host biology in beneficial ways.¹¹ More detailed knowledge of the microbiome and its products will hopefully direct future testing of whether such manipulation of the microbiome can alleviate illness in patients with renal failure.

Disclosures

Dr. Fischbach is a co-founder and director of Federation Bio and Viralogic, a co-founder of Revolution Medicines, a member of the scientific advisory. M. Fischbach also reports Consultancy: NGM Bio; Ownership Interest: Kelonia, NGM Bio; Patents or Royalties: Federation Bio; and Advisory or Leadership Role: Federation Bio, Kelonia, Board of NGM Biopharmaceuticals, and an innovation partner at The Column Group. Dr. Sonnenberg is a co-founder of Novome Biotechnologies, January AI, and Interface Biosciences; he serves on the scientific advisory board of BCD Biosciences. J. Sonnenburg also reports Ownership Interest: BCD Biosciences, Second Genome; Research Funding: Abbott Nutrition, Clorox; Honoraria: Biocodex Microbiota Foundation (spouse, SAB), The Cranberry Institute (spouse, SAB); and Patents or Royalties: Novome Biotechnologies. Dr. Meyer has served as a consultant for Baxter. T. Meyer also reports Research Funding: Outset Medical; Honoraria: Renal Research Institute; and Advisory or Leadership Role: ASN editorial board, KI editorial board. The remaining author has nothing to disclose.

Funding

This work was supported by a National Institutes of Health award (R01 DK118426). Dr. Guthrie received further support from a Howard Hughes Medical Institute Hanna H. Gray fellowship.

Author Contributions

M.A. Fischbach, L. Guthrie, T. Meyer, and J.L. Sonnenburg conceptualized the study, wrote the original draft, and reviewed and edited the manuscript.