Targeting Gut Microbiome With Prebiotic in Patients With CKD: The TarGut-CKD Study

Abstract

Introduction

Disruption of gut microbiota underpins some of the metabolic alterations observed in chronic kidney disease (CKD).

Methods

In a nonrandomized, open-label, 3-phase pilot trial, with repeated measures within each phase, we examined the efficacy of oligofructose-enriched inulin (p-inulin) in changing the gut microbiome and their metabolic products in 15 patients with CKD. The stability of microbiome and metabolome was studied during the pretreatment phase (8 weeks), a p-inulin treatment phase (12 weeks), and a post treatment phase (8 weeks) of the study.

Results

Study participants completed 373 of the 420 expected study visits (88.8%). Adherence to p-inulin was 83.4%. 16S rRNA sequencing was performed in 368 stool samples. A total of 1085 stool, urine, and plasma samples were subjected to untargeted metabolomic studies. p-inulin administration altered the composition of the gut microbiota significantly, with an increase in abundance of Bifidobacterium and Anaerostipes. Intersubject variations in microbiome and metabolome were larger than intrasubject variation, indicating the stability of the gut microbiome within each phase of the study. Overall metabolite compositions assessed by beta diversity in urine and stool metabolic profiles were significantly different across study phases. Several specific metabolites in stool, urine, and plasma were significant at false discovery rate (FDR) ≤ 0.1 over phase. Specifically, there was significant enrichment in microbial metabolites derived from saccharolysis.

Conclusion

Results from our study highlight the stability of the gut microbiome and the expansive effect of p-inulin on microbiome and host cometabolism in patients with CKD. Findings from this study will enable rigorous design of microbiome-based intervention trials.

Graphical abstract

The gut microbiota is a complex, dynamic consortium with coordinated networks of metabolic interactions among themselves, with the host and the ecosystem. The symbiotic microbiota is involved in numerous physiological processes.¹ Gut dysbiosis is characterized frequently by decreased diversity, and relative alterations in the abundance of selected microbial taxa.² Patients with CKD have dysbiosis and dysmetabolism, with accumulation of several gut-microbiota derived retention solutes in the blood.^1,3, 4, 5 It is possible that early alterations in microbiome in CKD are adaptive but become maladaptive with progressive loss of kidney function contributing to uremic toxicity.⁶

The scientific community has sought to capitalize on the metabolic potential of microbiome through probiotics, prebiotics, xenobiotics, nutritional modifications, and genetically engineered bacteria.7, 8, 9 Despite successful completion of the Human Microbiome Project¹⁰ and the Metagenomics of the Human Intestinal Tract project,¹¹ the lack of integrative metagenomic-metabolomics data has greatly impeded investigating gut microbiome as a therapeutic target in CKD. Prior to launching a full-scale prebiotic trial, it is important to understand the composition, function, and stability of the gut microbiota in CKD.

Prebiotics are microbial feed supplements that facilitate the growth of beneficial gut bacteria. The purpose of the Targeting Gut Microbiome and p-inulin in CKD (TarGut-CKD) pilot trial was to test the feasibility of conducting a rigorous clinical trial that would determine the safety of the prebiotic, p-inulin and to explore changes in gut microbiome and their metabolic products in patients with CKD. The repeated sampling strategy used in our study allowed us to assess the intraperson and interperson variability in the microbial composition and their metabolic products.

Methods

The Targeting Gut Microbiome and p-inulin in CKD study was conducted by the Pilot Studies in CKD Consortium, which was established by the National Institute of Diabetes and Digestive and Kidney Diseases to conduct early-phase studies of interventions for patients with CKD. The study was designed by the steering committee with input from an external advisory board. The study execution and compliance with protocol was rigorously monitored by the steering committee. The Cleveland Clinic served as the data coordinating center. Study participants for this study were recruited from George Washington University. The schedule for study procedures is shown in Figure 1a.

Figure 1.

Study design (a) Schedule of study procedures and (b) CONSORT flow diagram of this study.

Participants

The Targeting Gut Microbiome and p-inulin in CKD trial was a nonrandomized trial that consisted of 3 sequential phases consisting of a pretreatment phase (weeks 1–8), a p-inulin treatment phase (weeks 9–20), and a posttreatment phase (weeks 21–28). Participants with an estimated glomerular filtration rate of 15.0 to 50.0 ml/min per 1.73 m² were recruited from the outpatient clinics at the George Washington University School of Medicine. Other major inclusion criteria were age ≥18 years and self-reported stool frequency of at least every other day. The major exclusion criteria were the following: (i) use of prebiotics, probiotics, or antibiotics during the past 8 weeks; (ii) consumption of probiotic yogurt during the past 2 weeks; (iii) current infection, inflammatory bowel disease, chronic diarrhea, or Clostridioides difficile infection; and (iv) hemoglobin <9.0 g/dl within the past 4 weeks.

Biosample Collection

Study participants were provided a commercial “toilet hat” stool specimen collection kit (specimen container, shipping box, Styrofoam cooler, and cold packs, Fisherbrand Commode Specimen Collection System [Thermo Fisher Scientific, Waltham, MA]), sample aliquot tubes, and gloves. Participants generated 10 aliquots from each stool sample at home, stored the samples at 4 °C, and transported the samples in Styrofoam coolers with ice packs to the clinical trial unit within 24 hours. Urine and stool samples were collected weekly coinciding with the day of stool sample collection in the morning and stored at −20 °C. Blood samples were stored as serum and plasma aliquots at −80 °C. All blood and urine samples were collected in a fasting state.

Intervention

p-inulin (Prebiotin, provided by Jackson GI Medical, Harrisburg, PA) was administered at a dose of 8 g twice daily. Study participants added the content to approximately 200 ml of liquid and consumed. A dose reduction to 4 g twice daily was permitted for gastrointestinal (GI) side effects. The dose and duration of treatment for the present study were based on previous dose finding studies.12, 13, 14 During the dosing phase, participants were instructed to keep all unused study agent packets and bring them for assessing adherence by packet count.

Dietary Intake and GI Symptom Assessments

The Block Food Frequency Questionnaire was administered at baseline and weeks 8, 20, and 28.¹⁵ The GI Symptom Rating Scale was administered at baseline and every 4 weeks.¹⁶ The GI Symptom Rating Scale is a disease-specific instrument consisting of 15 items combined into 5 symptom clusters depicting reflux, abdominal pain, indigestion, diarrhea, and constipation.¹⁶ At each study visit, participants were asked about any significant clinical events and antibiotic use.

16S RNA Sequencing

The microbiome profile of weekly collected stool samples was determined using 16S rRNA sequencing performed at the Baylor College of Medicine Alkek Center for Metagenomics and Microbiome. Briefly, fecal bacterial genomic DNA was extracted using the DNeasy PowerSoil Kit (Qiagen). The 16S rRNA genes were amplified using degenerate primers that target the V3-V4 hypervariable region. Primers contained molecular barcodes and adapters for polymerase chain reaction products to be pooled and sequenced directly on the Illumina MiSeq platform (2x250bp protocol). Pooling depth targeted at least 20,000 merged reads per sample on average. Rarefaction and collector’s curves of microbial community data were constructed using sequence data for each sample to ensure sampling the majority of diversity present.

Metabolomic Profiling

Metabolomic studies were performed at the same time points as the microbiome studies. Untargeted metabolomics profiling was performed at the West Coast Metabolomics Center at the University of California Davis using a gas chromatography time-of-flight mass spectrometry platform. ChromaTOF version 4.50 and Binbase version 5.0.3 were used for gas chromatography time-of-flight mass spectrometry data processing.¹⁷ In order to account for the systematic error associated with large sample sets, we employed random forest (SERRF) for eliminating the unwanted systematic variations.¹⁸

Outcomes

The primary microbiome and metabolome associated outcomes were as follows: (i) interparticipant and intraparticipant variability in stool, urine, and plasma metabolites during each of the pretreatment, treatment, and posttreatment phases; (ii) interparticipant and intraparticipant variability in the gut microbial composition across the pretreatment, treatment, and posttreatment phases; and (iii) interparticipant and intraparticipant change in plasma metabolomic profiles and bacterial composition during the pretreatment, treatment, and posttreatment phases.

Safety and tolerability outcomes included the following: (i) GI symptoms, (ii) early discontinuation or reduction in p-inulin dose, and (iii) adverse events. Feasibility outcomes were as follows: (i) the proportion of protocol-specified blood and stool sample collections; (ii) adherence to p-inulin assessed by returned packets at weeks 12, 16, and 20; and (iii) participant withdrawal during each phase of the study.

Sample Size Determination

The study aimed to enroll 10 individuals who completed phases 1 and 2 and provided at least 2 stool samples during weeks 1 to 4, 2 stool samples during weeks 5 to 8, 3 stool samples during weeks 9 to 15, and 3 stool samples during weeks 15 to 20. Based on the distance-based multivariate analysis of variance using generalized UniFrac distances simulated from a 2-dimensional circular space,¹⁹ the target sample size of 10 participants with repeated measurements was anticipated to allow the detection of moderate-to-large changes in bacterial community membership, evenness, richness, and lineages. For the effect of p-inulin on metabolites, a sample size of 10 participants was expected to detect an effect size (Cohen’s d) of 1.7 based on a paired t-test, assuming that 25 metabolites were tested using the Benjamini-Hochberg procedure to control the FDR.²⁰

Statistical Analyses

Descriptive statistics were used to summarize the characteristics of the study population. Means and SDs were used for symmetrically distributed continuous data, medians, and interquartile ranges for skewed continuous data, and frequencies and percentages for categorical data. For alpha diversity, the Shannon index was used to capture the richness (number of types of organisms) and evenness (uniformity across organisms) for each participant at each time point. Alpha diversity across the treatment phase was assessed using a mixed-effects model that treated the Shannon indices in each treatment phase as repeated measures.²¹ For beta diversity, the weighted UniFrac distance was used for the microbiome data and the Euclidean distance was used for the metabolome data. Their results were visualized using principal coordinate analysis and assessed using permutational multivariate analysis of variance with permutations constrained within time to investigate changes in microbial compositions or metabolomic expressions across the treatment phase.²² The difference between the intraparticipant and interparticipant variability was compared using an approximate randomization method.²³ For changes in the composition of the microbiome (or metabolome) across the treatment phase, changes in the distances from baseline (i.e., week 2) were assessed using linear mixed-effects models by treating the distances in each treatment phase as repeated measures. The intraclass correlation coefficient for each taxon or metabolite was estimated using a linear mixed-effects model. To identify differentially abundant taxa or metabolites, the abundance levels of taxa or the expression levels of metabolites in each treatment phase were treated as repeated measurements, and their differences across the treatment phase were assessed using linear mixed-effects models. Univariate analysis between each taxon and each metabolite was performed using mixed effects models. In all analyses, the following covariates were included: age, sex, race, diabetes, estimated glomerular filtration rate, and urine albumin-to-creatinine ratio. The Benjamini-Hochberg procedure was used to adjust for multiple testing in order to control FDR. All analyses were performed in R.²⁴

Study Approval

The study abided by guidelines laid out by the Declaration of Helsinki and the Declaration of Istanbul. The trial was registered at Clinicaltrials.gov (NCT03348592). The study was approved by the Institutional Review Boards at George Washington University and the Data Coordinating Center. All participants provided written informed consent before initiating study procedures and were compensated for time and effort.

Results

Patients

This is a nonrandomized, open-label, 3-phase crossover trial with repeated measures within each phase. Among 17 consenting eligible individuals, 13 completed the treatment phase, as shown in the CONSORT (CONsolidated Standards Of Reporting Trials) flow diagram (Figure 1b). Serious adverse events during the study period are described in Supplementary Table S1. Characteristics of the participants involved in this study are described in Table 1. Study participants completed 373 of the 420 expected study visits (88.8%), providing 372 of the expected 373 urine samples (99.7%), 345 of the expected 373 blood samples (92.5%), and 3659 of the expected 3730 stool aliquots (98.1%) (Table 2). Adherence to p-inulin was 83.4%, with some patients experiencing GI symptoms (Table 3, Supplementary Tables S2 and S3). There was a significant decrease in carbohydrate intake during the treatment phase compared to pretreatment phase by a 2-sided, paired t-test (n = 13, 142.95 ± 100.30 vs. 98.41 ± 55.71, P = 0.037; Supplementary Table S4).

Table 1.

Participant characteristics (n = 15)

Variable	n (%) or mean ± SD
Age (years)	64.5 ± 11.4
Male	11 (73.3%)
Race
Black or African American	12 (80.0%)
White	3 (20.0%)
Diabetes mellitus	7 (46.7%)
eGFR (ml/min per 1.73 m2)	30.0 ± 9.87
UACR (mg/g)	860 ± 882
Hemoglobin (g/dl)	11.3 ± 1.53
WBC (1000/mcl)	6.69 ± 1.85
Platelet count (1000/mcl)	235 ± 54.8
BUN (mg/dl)	39.1 ± 11.9
Creatinine (mg/dl)	3.02 ± 1.35
Cystatin C (mg/l)	2.34 ± 0.84
Calcium (mg/dl)	9.53 ± 0.51
Phosphorus (mg/dl)	4.29 ± 0.84
FGF-23 (RU/ml)	295 ± 184
Glucose (mg/dl)	125 ± 74.9
CRP (ug/ml)	7.41 ± 12.2
Vitamin D	2 (14.3%)
Lipid lowering agents	7 (50.0%)
Insulin	0 (0.0%)

eGFR, estimated glomerular filtration rate; UACR, urine albumin-to-creatinine ratio.

Table 2.

Study visits and sample collection completed

Participant ID	Study Visits Completed	Performed/Expected
		Urine Collections	Stool Collections	Stool Aliquots	Blood Collections
1	27	26/27	26/27	260/270	27/27
2	27	27/27	27/27	268/270	27/27
3	14	14/14	14/14	131/140	11/14
4	28	28/28	28/28	279/280	28/28
5	28	28/28	28/28	280/280	28/28
6	26	26/26	26/26	252/260	26/26
7	25	25/25	25/25	250/250	25/25
8	28	28/28	28/28	280/280	28/28
9	28	28/28	28/28	280/280	28/28
10	23	23/23	19/23	189/230	7/23
11	28	28/28	28/28	280/280	28/28
12	7	7/7	7/7	70/70	6/7
13	28	28/28	28/28	280/280	28/28
14	28	28/28	28/28	280/280	20/28
15	28	28/28	28/28	280/280	28/28
Total	373/420	372/373	368/373	3659/3730	345/373
Percentage	(88.8%)	(99.7%)	98.7%	(98.1%)	(92.5%)

Table 3.

Adherence to p-inulin

Participant ID	Mean Packet Count (%)			Mean
	Week 12	Week 16	Week 20
1	80.4	100.9a	100.9a	94.1
2	93.5	77.1	85.7	85.4
3	.	5.6	.	5.6
4	76.3	103.4a	111.1a	96.9
5	96.3	106.7a	96.6	99.9
6	99.1	100	100	99.7
7	98.8	100	72.2	90.3
8	96.4	109.7a	90.9	99.0
9	100	88.4	95.8	94.7
10	29.9	87.1	65.6	60.9
11	100	102.7a	55.4	86.0
12	.	.	.	.
13	100	100	100	100
14	58	100	75	77.7
15	80.4	77.2	74.3	77.3
Mean	85.3 ± 21.0	89.9 ± 26.3	86.4 ± 16.5	83.4 ± 25.1

^aConsumed more than the recommended dose.

p-inulin Altered the Composition of Gut Microbiota

16S rRNA sequencing performed in 368 stool samples detected 216 microbial genera. The top 30 most abundant genera are shown in Figure 2a. A total of 30 genera were significantly different over the study phases (FDR ≤ 0.1; Supplementary Table S5). The differentially abundant genera with FDR ≤ 0.05 and the median abundance greater than zero are shown in Figure 2b. We found that the abundance of Bifidobacterium and Anaerostipes were increased, whereas Lachnospira, Moryella, Negativibacillus, Ruminococcaceae, and Erysipelotrichaceae were reduced during the treatment phase.

Figure 2.

Gut microbial composition altered by p-inulin. (a) Top 30 abundant genera (b) Differentially abundant genera over phases (FDR < 0.05). FDR, false discovery rate.

p-inulin Altered the Diversity of Gut Microbiota

The change in Shannon diversity of each patient over weeks is shown in Figure 3. Alpha diversity assessed by Shannon index was significantly different between pretreatment and treatment phases. The overall microbial composition (i.e., beta diversity) assessed by permutational multivariate analysis of variance with the weighted UniFrac distance was significant (P = 0.002) across treatment phases (Figure 4a). Interpatient variability in microbial composition was significantly larger than intrapatient variability (Figure 4b). In the comparison with changes in microbial composition from baseline (week 2) using the weighted UniFrac distance, during-treatment phase was significantly larger than pretreatment phase (Figure 4c). The intersubject variability in microbial composition is relatively stable at each time point and each treatment phase as shown in Figure 4d and e, respectively. Distributions of the intraclass correlation coefficient for the genera suggested an increase in intrasubject variability in the gut microbiome during the treatment and posttreatment phase compared to the pretreatment phase (Figure 4f). The area under the intraclass correlation coefficient distribution curve greater than 0.5 was significantly decreased at the posttreatment phase when comparing with pretreatment phase (Figure 4g).

Figure 3.

Effects of p-inulin on gut microbial alpha diversity. (a) Alpha diversity within subject, and (b) alpha diversity over weeks.

Figure 4.

Effects of p-inulin on gut microbial beta diversity. (a) Weighted UniFrac distance, PERMANOVA for treatment phase: P < 0.001. (b) intravariability and intervariability (16S). (c) Change in distance across treatment phases. (d) Intersubject variability. Levene’s test of homogeneity: P = 0.002. Post hoc: Bonferroni correction. (e) Intersubject distance (weighted UniFrac), P = 0.381. (f) Density of ICCs for bacterial genera. (g) Area under the curve for bacteria genera with ICC > 0.5. ICC, intraclass correlation coefficient; PERMANOVA, permutational multivariate analysis of variance.

p-inulin Altered Stool Metabolites

Untargeted metabolomics was performed in 368 stool samples. Weekly observations showed changes in the intersubject variability between weeks (Figure 5a). Variations between subjects were larger than variations within subjects (Figure 5b). Changes in the stool metabolic profile measured by the Euclidean distance from the baseline were not significantly different over study phases (Figure 5c). The overall composition of the stool metabolites exhibited significant difference over phases of the study (Figure 5d), and 6 stool metabolites were significant at FDR ≤ 0.1 over phase (Figure 5e). Notably, the level of a secondary bile acid, deoxycholic acid in stool samples was significantly different (P = 0.048), suggesting that the changes in gut microbial composition induced by p-inulin treatment further affected the production of microbial metabolites.

Figure 5.

Effects of p-inulin on stool metabolites. (a) Intersubject variability of stool metabolites. Levene’s test of homogeneity: P < 0.001; Post hoc: Bonferroni correction. (b) Between vs. within variation for stool metabolites. (c) Euclidean distance from week 2 for stool metabolites. (d) Principal component analysis of stool metabolites. PERMANOVA for treatment phase: P = 0.037. (e) Significantly different stool metabolites (FDR < 0.1). FDR, false discovery rate; PERMANOVA, permutational multivariate analysis of variance.

p-inulin Altered Plasma Metabolites

Untargeted metabolomics was performed in 345 plasma samples. Intersubject variability in plasma metabolites was noted in weekly assessments (Figure 6a). Variations between subjects were larger than variations within subjects (Figure 6b). Changes in the plasma metabolic profile from baseline over study phases were not significant (Figure 6c). Although the overall composition of plasma metabolites did not differ significantly over study phases (Figure 6d), 2 plasma metabolites, raffinose and N-acetylornithine, were significant at FDR ≤ 0.1 (Figure 6e and f). Plasma levels of p-cresol sulfate and indoxyl sulfate did not change significantly during the treatment phase (Figure 6g and h).

Figure 6.

Effects of p-inulin on plasma metabolites. (a) Intersubject variability of plasma metabolites. Levene’s test of homogeneity: P < 0.001; Post hoc: Bonferroni correction (b) Between vs. within variation for plasma metabolites. (c) Euclidean distance from W2 for plasma metabolites. (d) Principal component analysis of stool metabolites. PERMANOVA for treatment phase: P = 0.285. (e) Significant plasma metabolite raffinose (FDR < 0.1). (f) Significant plasma metabolite N-acetylornithine (FDR < 0.1). (g and h) Plasma levels of indoxyl sulfate and p-cresol sulfate (P > 0.05). FDR, false discovery rate; PERMANOVA, permutational multivariate analysis of variance.

p-inulin Altered the Urine Metabolites

Untargeted metabolomics was performed in 372 urine samples. Changes in the intersubject variability were noted between different weeks (Figure 7a). Variations between subjects were larger than variations within subjects for urine metabolites (Figure 7b). Changes in the urine metabolic profile from the baseline were not significantly different over phases of the study (Figure 7c). The overall composition of the urine metabolic profile revealed significant difference over phases of the study (Figure 7d). Univariate analysis showed that 8 urine metabolites were found to be significant at an FDR ≤ 0.1 (Figure 7e).

Figure 7.

Effects of p-inulin on urine metabolites. (a) Intersubject variability for urine metabolites. Levene’s Test of Homogeneity: P < 0.001. Post hoc: Bonferroni correction. (b) Between vs. within variation for urine metabolites. (c) Euclidean distance from W2 for urine metabolites. (d) Principal component analysis of urine metabolites. PERMANOVA for Treatment Phase: P = 0.042. (e) Significant urine metabolites (FDR < 0.1). FDR, false discovery rate; PERMANOVA, permutational multivariate analysis of variance.

Discussion

Imbalance in microbial composition and altered bacterial metabolites are described in CKD. The gut microbiome remains a prime target for restoring metabolic synergy in CKD.²⁵ End products of protein fermentation have detrimental effects on human health.²⁶ Consumption of nondigestible fermentable fibers attenuates proteolysis by the gut microbiota through functional reconfiguration of metabolic pathways.^27,28 However, lack of definitive data regarding the intraindividual variability of gut microbiome and metabolome has impeded the design and interpretation of results from the microbiome-based therapeutics in patients with CKD. In this 3-phase pilot trial with repeated sampling within each phase, we examined the shift in microbiome and metabolomic profile in response to p-inulin treatment. Adherence to study protocol was excellent and p-inulin treatment was well-tolerated. Within each individual, gut microbiome and metabolome tend to be stable over each study phase. p-inulin supplementation altered the abundance of several microbial genera and enrichment of specific microbiota-derived metabolites in stool, plasma, and urine.

The gut microbiota exists in a state of stable equilibrium and is resilient and resistant to changes.²⁹ Faith et al.³⁰ studied 37 healthy individuals and showed that the gut microbiota is remarkably stable over a course of 5 years. Loss of “keystone taxa” in chronic diseases such as CKD leads to gut dysbiosis, characterized by decreased diversity, and also alterations in the abundance of selected microbial taxa.^2,31 In our study, alpha diversity, a measure of richness and evenness and beta diversity, an index of variability in community composition among samples, changed significantly during the treatment phase. Diet has a significant impact on gut microbiome profile. During the treatment phase, carbohydrate intake decreased significantly. Reduction in alpha diversity in response to prebiotic consumption has been reported, which has been attributed to increase in beneficial bacteria including Bifidobacteria.³² The documented stability of the gut microbiota in patients with CKD has major implications for designing microbiome-centric clinical trials.

CKD is characterized by an increase in abundance of Proteobacteria and Fusobacteria, with depletion of short-chain fatty acid producing bacteria.^3,6 Prebiotics are believed to increase the abundance of Bifidobacterium and also Lactobacillus, Anaerostipes, and Faecalibacterium in individuals without CKD.^33,34 However, improved understanding of the microbial ecosystem redundancy and complementary function has challenged the traditional concept of selectivity and specificity of prebiotics.³⁵ We noted significant changes in abundance in Bifidobacterium, Anaerostipes, Lachnospira, Moryella, Negativibacillus, Ruminococcaceae, and Erysipelotrichaceae during the intervention phase. Notably, Bifidobacteria, which are endowed with the enzymes capable of fermenting complex carbohydrates were significantly increased.³⁶ However, recent studies have suggested functional redundancy in substrate utilization is unrestricted by taxonomic boundaries.^2,31 Irrespective, increased availability of readily fermentable fiber substrate could increase the saccharolytic-to-proteolytic fermentation ratio, establishing a metabolic synergy in CKD.^25,37

Because the gut microbiota-generated metabolites may translocate to the systemic circulation from the intestine and excreted in the urine, we performed metabolomic studies in stool, plasma, and urine. In general, the individual metabolome profile was stable in our study during the study phases. Previous studies have also indicated that serum metabolome is stable over time.³⁸ p-inulin treatment effect on a global metabolomic profile measured by the Euclidian distance was not significantly different over study phases; however, several specific stool, plasma, and urine metabolites were altered. Many products of carbohydrate metabolism were increased in the urine, including raffinose, 1-kestose, and beta-gentiobiose. Raffinose is a trisaccharide that cannot be digested by humans because the human digestive system does not produce α-1,6-galactosidase enzyme.³⁹ Raffinose is metabolized by the gut microbes, generating hydrogen, carbon dioxide, and methane, which causes flatulence, but promotes the growth of beneficial bacteria such as Bifidobactera.⁴⁰ 1-kestose is a bifidogenic fructo-oligosaccharides that enables the growth of butyrate producing bacteria such as Faecalibacterium prausnitzii.⁴¹ Beta-sitosterol, a bioactive phytosterol with immunomodulatory, antiinflammatory, and lipid-lowering effects42, 43, 44 was enriched in the urine during the treatment phase. 4-methylcatechol was increased during the postintervention phase in the urine. This flavonoid microbial metabolite derived from rutin attenuates oxidative stress and has antihypertensive effect.^45,46 Interestingly, an inverse correlation between 4-methylcatechol and p-cresol was observed because of a potential competition between rutin degradation and p-cresol.^45,46 The observed changes in metabolites are possibly mediated by the metabolism of prebiotic, change in microbial composition, alterations their metabolic machinery driven by substrate availability during the treatment phase, and legacy effect during the posttreatment phase.⁴⁷

The biosynthesis of p-cresol relies on the metabolism of tyrosine by gut bacteria such as Clostridioides and Coriobacteriaceae.⁴⁸ In a pilot study involving 22 patients on maintenance hemodialysis, Meijers et al. showed that treatment with 20 g of p-inulin for 4 weeks reduced the generation and serum concentration of p-cresol.¹⁴ Other studies have also reported potential benefits of a symbiotic therapy in reducing uremic toxins.^49,50 Indoxyl sulfate is derived from tryptophan metabolism, which has been implicated in cardiovascular morbidity and mortality in CKD.⁵¹ In this study, we did not observe a significant change in plasma levels of p-cresol sulfate or indoxyl sulfate.

This study has several strengths, including the intensive study design with 3 sequential phases with multiple sampling within each phase; simultaneous metabolomic studies in stool, plasma, and urine; meticulous patient selection; and adequate treatment duration. Inclusion of a postintervention phase permits us to assess the persistence of treatment effects. The weaknesses of this study include the relatively small sample size with participants recruited from a single study site. These findings need to be validated in a larger independent patient cohort. We acknowledge that urine metabolomics could be influenced by variation in water intake which might obfuscate interpretation.

In conclusion, the microbiome and metabolome are relatively stable within each phase of the study. Supplementation with p-inulin alters microbiome profiles with an increase in abundance of microbial metabolites derived from carbohydrate metabolism. This pilot study shows feasibility data for launching a full-scale clinical trial testing the utility of prebiotics in patients with CKD. Based on our findings, we opine that in a steady state, assessment of the gut microbiome and metabolome at one time point should be sufficient in patients with CKD to draw valid conclusions.

Appendix

Members of the Pilot Studies in CKD Consortium

Cynthia Kendrick, Tamara Isakova, Jennifer J. Gassman, Linda F. Fried, Myles Wolf, Alfred K. Cheung, Kalani L. Raphael, Joe Ix, John P. Middleton, Susan Mendley, Michael F. Flessner, and Dominic S. Raj, for the Pilot Studies in CKD Consortium.

Disclosure

All the authors declared no competing interests.

Supplementary Material

Table S1. Serious adverse events during the study period.

Table S2. P-inulin associated gastrointestinal symptoms.

Table S3. Adherence to p-inulin among those who completed the intervention phase.

Table S4. Dietary intake assessed by food frequency questionnaire.

Table S5. Differentially abundant taxa over phase (FDR< 0.1). These P-values are based on the likelihood ratio test, that is, we tested if a taxon was associated with the study phases.