Evidence for a distinct gut microbiome in kidney stone formers compared to non-stone formers

Abstract

The trillions of microbes that colonize our adult intestine are referred to as the gut microbiome (GMB). Functionally it behaves as a metabolic organ that communicates with, and complements, our own human metabolic apparatus. While the relationship between the GMB and kidney stone disease (KSD) has not been investigated, dysbiosis of the GMB has been associated with diabetes, obesity and cardiovascular disease. In this pilot study we sought to identify unique changes in the GMB of kidney stone patients compared to patients without KSD. With an IRB-approved protocol we enrolled 29 patients into our pilot study. 23 patients were kidney stone formers and six were non-stone forming controls. Specimens were collected after a 6h fast and were flash frozen in dry ice and then stored at −80 °C. Microbiome: determination of bacterial abundance was by analysis of 16 s rRNA marker gene sequences using next generation sequencing. Sequencing of the GMB identified 178 bacterial genera. The five most abundant enterotypes within each group made up to greater than 50 % of the bacterial abundance identified. Bacteroides was 3.4 times more abundant in the KSD group as compared to control (34.9 vs 10.2 %; p = 0.001). Prevotella was 2.8 times more abundant in the control group as compared to the KSD group (34.7 vs 12.3 %; p = 0.005). In a multivariate analysis including age, gender, BMI, and DM, kidney stone disease remained an increased risk for high prevalence for Bacteroides (OR = 3.26, p = 0.033), whereas there was an inverse association with Prevotella (OR = 0.37, p = 0.043). There were no statistically significant differences in bacterial abundance levels for Bacteroides or Prevotella when comparing patients with and without DM, obesity (BMI >30), HTN or HLD. 11 kidney stone patients completed 24 h urine analysis at the time of this writing. Looking at the bacterial genuses with at least 4 % abundance in the kidney stone group, Eubacterium was inversely correlated with oxalate levels (r = −0.60, p < 0.06) and Escherichia trended to an inverse correlation with citrate (r = −0.56, p < 0.08). We also compared bacterial abundance between uric acid (UA) stone formers (n = 5) and non UA stone formers (n = 18) and found no significant difference between them. We identified two genus of bacteria in the GMB that had significant association with KSD. Interestingly, components of the 24-h urine appear to be correlated to bacterial abundance. These preliminary studies for the first time associate differences in the GMB with kidney stone formation. Further studies are warranted to evaluate the potential causative role of preexisting dysbiosis in kidney stone disease.

Introduction

Kidney stone disease (KSD) causes significant morbidity and financial burden on public health. The lifetime prevalence of KSD is increasing, with data from the most recent National Health and Nutrition Examination Survey (NHANES; 2007–2010) reporting an 8.8 % population prevalence—a substantial increase from the 5.2 % prevalence reported in a prior NHANES cohort [1]. The rising prevalence of obesity and diabetes, two well established risk factors for KSD, together with population growth, is projected to increase its prevalence and contribute an additional $1.24 billion/year in US treatment costs by 2030 [2]. In addition to obesity and diabetes, a variety of systemic diseases such as the metabolic syndrome, hypertension, chronic kidney disease (CKD) and cardiovascular disease are associated with KSD [3-6].

Because pathological changes in metabolism are thought to play a key role, in the development of kidney stones, diet, as a primary determinant of an organisms metabolism, is generally believed to be an important contributing factor to KSD [7-10]. In the context of diet it is also increasingly recognized that the gut microbiome (GMB) is a modulating factor of diet-driven metabolism. Recent advances in sequencing of the human gut microbiome have led to important breakthroughs that describe the relationship of the GMB with diseases intimately linked to kidney stones like obesity and diabetes [11, 12]. Through the fermentation of fiber, carbohydrate, and protein by the colonic GMB small metabolites are generated that have wide and varying effects on an individual’s health [12-15]. If a unique GMB profile were to be identified in kidney stone patients, manipulation of that profile could change the metabolism of the individual thereby potentially altering ones kidney stone risk.

Relatively little is known about the role of the GMB in KSD, and there is scant literature on the subject. The role of gut microbiota in diseases linked to the metabolic syndrome provides motivation to investigate the relationship between the microbiota and kidney stone disease. Therefore, in the proof-of-principal pilot study reported here, we determined if there were differences between the GMB of kidney stone patients and that of non-stone formers. We also sought to correlate 24-h urine parameters with bacterial abundance. This approach does not address whether the microbiota are causal in this relationship to kidney stone disease, but it provides incentive for more definitive studies, and allows us a first examination of the entire gut microbiota in kidney stone disease.

Methods

The study protocol and recruitment of the 29 human subjects were approved by the Albert Einstein internal review board (IRB). After a 6 h fasting period, participants in the study were asked to provide urine (approximately 10 ml) and a fecal sample (using Q-tip like applicators). These samples were flash frozen in dry ice and then stored at −80 °C until analysis. All collections were obtained prior to antibiotic use.

Kidney stone patients (N = 23)

For this pilot study we did not discriminate between the first time kidney stone formers and recurrent stone formers. All stone types were included. Stone composition was defined by the constituent with the largest contribution to the stone make up. We excluded patients actively receiving chemotherapy or having received chemotherapy in the last 1 year. All patients reporting antibiotic use within 2 weeks of enrollment were excluded. All patients having undergone bariatric surgery were excluded. 24-h urine specimens were analyzed through Litholink.

Control patients (N = 6)

No self-reported history of kidney stones. We excluded those actively receiving chemotherapy or who received chemotherapy in the last 1 year. All patients having undergone bariatric surgery were excluded. All patients reporting antibiotic use within 2 weeks of enrollment were excluded.

Gut microbiome analysis

100 μl fecal sample collected in Specimen Transport Medium (STM) (Qiagen, USA) was extracted for microbiome DNA using two methods, bead-beating based PowerSoil DNA isolation kit (MO BIO laboratories, USA) and chemical-lysis based QIAamp DNA mini kit (Qiagen, USA), following the manufacturer’s protocol. Upon finishing, the purified DNA was rinsed in 100 μl elution buffer (pH 8.0) and measured for concentration using Qubit 2.0 (Thermo Fisher Scientific, USA). The microbiome DNA was PCR amplified using barcoded primers spanning the V4 variable region of the 16S rRNA gene [16]. Barcoded PCR products from all samples were pooled at approximately equal molar DNA concentrations [17] and sequenced on an Illumina MiSeq (Illumina Inc., USA) by the Albert Einstein Epigenomics Shared Facility using paired-end reads.

The short Illumina reads were processed using our previously published bioinformatics pipeline [18] and 3rd-party taxonomy software, QIIME [19], usearch [20] and pplacer [21], based on the latest microbiome 16S rRNA sequence database (Greengenes Database, version May, 2013) for an operational taxonomic unit (OTU) classification. Compositions of microbiome communities were summarized by a proportion at the genus level. If unavailable, higher levels were assigned.

Statistical and comparative analyses

All plotting and statistical comparisons of microbiome community were performed in R v3.1.3 using scripts developed in-house. The characterized reads were displayed using a heat map and squash hierarchical clustering, which is based on a weighted UniFrac distance [22] of sample microbiome using the Jackknifed beta diversity in QIIME and normalized with respect to the diameter of the taxa reference tree. Principal component analysis was performed in R using the prcomp function where different comparisons were evaluated for association with one or more principal components of the microbiome community. A two-sided p value of <0.05 was considered statistically significant.

We compared the prevalence of the major enterotypes within the GMB between KSD cases and controls using Chi-square tests. After identifying two major enterotypes within each group, we fitted two separate logistic regression models for associations of KSD with Bacteroides and Prevotella respectively, adjusting for age, gender, DM, and BMI. We compared the average bacterial abundance of Bacteroides and Prevotella between patients with and without DM, HTN, HLD or BMI < or >30 using Student’s T tests. Among kidney stone patients with completed 24 h urine measurements at the time of this data collection (n = 11) we assessed correlations between 24 h urine parameters and bacterial abundance using Spearman correlation coefficients. All tests were two-sided (p < 0.05). Data analysis was carried out in Stata version 13.

Results

Kidney stone disease and the GMB

Table 1 describes patient characteristics. We successfully collected stool samples from six of our control patients, and 23 of our kidney stone patients. We identified 178 genera of which the five most abundant enterotypes within each group made up greater than 50 % of the bacterial abundance identified. The most weighted clustering was seen from Prevotella and Bacteroides which together made up approximately 42 % of the case abundance and 45 % of the control abundance. Other less well-represented genera did not cluster as strongly (Fig. 1a, b). In the control group Prevotella genus was the most abundant, while in the KSD group Bacteroides genus was the most abundant. Each was found to be differentially expressed between control and kidney stone patients. As illustrated in Fig. 2a, b, Bacteroides was 3.4 times more abundant in the KSD group as compared to control (34.9 vs 10.2 %; p = 0.001). Prevotella was 2.8 times more abundant in the control group as compared to the KSD group (34.7 vs 12.3 %; p = 0.005). We also compared bacterial abundance between uric acid (UA) stone formers (n = 5) and nonUA stone formers (n = 18) and found no significant differences between them. In a multivariate analysis including age, gender, BMI, and DM, kidney stone disease remained an increased risk for high prevalence for Bacteroides (OR = 3.26, p = 0.033), whereas, there was an inverse association with Prevotella (OR = 0.37, p = 0.043). There were no statistically significant differences in bacterial abundance levels for Bacteroides or Prevotella when comparing patients with and without DM, obesity (BMI >30), HTN or HLD.

Table 1.

Key characteristics of patients and controls

	All patients	Stone formers	Controls	P
Age: mean (SD)	53.7 (15.2)	53.7 (15.4)	53.5 (16.0)	0.978
Sex: n (%)
Female	20 (69)	18 (78)	2 (33)	0.034
Male	9 (31)	5 (22)	4 (67)
Ethnicity: n (%)
African American	6 (20)	3 (13)	3 (50)	0.073
Hispanic	8 (28)	8 (35)	0
White	15 (52)	12 (52)	3 (50)
Stone type: n (%)
Calcium based	16	16 (55)	N/A	N/A
Uric acid	5	5 (17)	N/A
BMI: mean (SD)	28.9 (6.7)	29.3 (6.4)	27.0 (7.9)	0.454
Comorbidities: n (%)
CKD	2 (7)	2 (9)	0	0.454
DM	12 (41)	10 (43)	2 (33)	0.653
HLD	14 (48)	12 (52)	2 (33)	0.411
HTN	19 (66)	15 (65)	4 (67)	0.947

Comparisons made using Chi-square and Student’s T. Two patients had uncommon stone compositions

Bold value indicates statistically significant (p value < 0.05)

Fig. 1.

a Heat map comparing the microbiome between patients with kidney stone disease (N = 23, red) and control (N = 6, green). The subjects are grouped by phylogeny and thereby not numerically. Note the clustering of Bacteroides and the Prevotella genera with the two groups of patients. b Highlighted heat map demonstrating the 5 most abundant genera in the kidney stone group. Note the clustering of Bacteroides and the Prevotella genera among the two groups of patients

Fig. 2.

a Gut microbiome abundance plot: Bacteroides was 3.4 times more abundant in the KSD group as compared to control (34.9 vs 10.2 %; p = 0.001). Prevotella was 2.8 times more abundant in the control group as compared to the KSD group (34.7 vs 12.3 %; p = 0.005). b The composition of the microbiota between kidney stone formers and non-stone forming controls. Percentage of each bacteria genus between cases (n = 23) and controls (n = 6) was plotted in the bar chart. The Wilcoxon Mann–Whitney tests with p < 0.05 were highlighted in bold. Genera with less than 500 of total reads were not presented

11 kidney stone patients completed 24 h urine analysis at the time of this writing (Table 2). Looking at the bacterial genera with at least 4 % abundance in the kidney stone group, Eubacterium was inversely correlated with oxalate levels (r = −0.60, p < 0.06) and Escherichia trended to an inverse correlation with citrate (r = −0.56, p < 0.08)

Table 2.

Mean values of 24 h urine (n = 11) constituents and Spearman correlation coefficients to major bacterial genera by fraction of gut flora

Constituent	Mean	SD	Range	Bacteroides		Escherichia		Eubacterium		Faecalibacterium		Prevotella
				R	p	R	p	R	p	R	p	R	p
Stone risk factors
SS CaOx (unitless)	4.4	2.67	0.89–9.6	−0.216	0.548	0.122	0.737	−0.286	0.423	0.049	0.894	0.065	0.858
Ca (mg/day)	128.5	61.4	40–221	−0.223	0.536	0.054	0.882	0.004	0.991	0.176	0.627	−0.082	0.821
Ox (mg/day)	26.4	6.66	20–39	0.241	0.503	0.125	0.731	−0.601	0.066	−0.256	0.476	−0.207	0.567
Cit (mg/day)	535	116.00	396–782	−0.313	0.379	−0.567	0.088	0.221	0.540	0.151	0.677	0.464	0.176
SS CaP (unitless)	0.52	0.51	0.04–1.54	−0.489	0.152	0.061	0.868	0.233	0.517	0.039	0.915	0.123	0.735
pH (unitless)	5.7	0.33	5.1–6.1	0.067	0.854	0.071	0.846	0.071	0.845	0.206	0.568	−0.455	0.187
SS UA (unitless)	1.21	0.62	0.33–2.01	0.009	0.981	0.023	0.950	0.072	0.843	−0.025	0.945	0.210	0.561
UA (g/day)	0.57	0.15	0.33–0.81	0.046	0.900	0.235	0.513	0.103	0.777	0.046	0.899	−0.296	0.407
Dietary factors
Na (mEq/day)	150	39.6	112–230	−0.260	0.468	0.351	0.319	0.244	0.497	−0.459	0.182	−0.043	0.906
K (mEq/day)	51.5	11.6	34–68	−0.017	0.963	−0.414	0.234	0.253	0.480	0.087	0.812	−0.168	0.642
Mg (mEq/day)	92.6	24.1	56–146	0.145	0.689	0.178	0.624	−0.025	0.946	−0.355	0.314	−0.123	0.735
P (mEq/day)	0.70	0.16	0.49–0.98	−0.470	0.170	0.138	0.703	0.340	0.337	−0.532	0.113	0.337	0.342
NH4 (mEq/day)	27.1	5.72	16–34	−0.493	0.148	−0.361	0.306	0.296	0.407	0.003	0.993	0.554	0.096
Cl (mEq/day)	152	37.7	106–223	−0.202	0.576	0.339	0.337	0.196	0.588	−0.469	0.171	−0.102	0.780
Sul (mEq/day)	35.8	14.5	9–56	−0.252	0.482	−0.388	0.268	0.646	0.044	0.379	0.281	0.204	0.573
UUN (g/dL)	10.3	2.02	7.3–13.5	−0.481	0.160	−0.391	0.265	0.264	0.461	−0.200	0.580	0.502	0.140
PCR (g/kg)	1.03	0.2	0.8–1.5	−0.130	0.739	0.006	0.987	0.390	0.300	−0.078	0.842	0.401	0.285
Normalized values
Cr (mg/day)	1390	259.00	888–1750	−0.109	0.764	−0.556	0.095	0.219	0.543	0.312	0.380	−0.007	0.985
Cr/kg (mg/kg × day)	18.6	4.18	13.6–25.3	0.261	0.497	−0.194	0.618	0.506	0.165	0.423	0.257	−0.179	0.645
Ca/kg (mg/kg × day)	1.76	0.94	0.5–2.7	0.016	0.968	0.189	0.626	−0.028	0.943	0.222	0.566	−0.105	0.789
Ca/Cr (unitless)	99	51.9	24–165	−0.087	0.811	0.356	0.313	−0.146	0.687	−0.016	0.965	−0.085	0.816

Genera representing <4 % abundance of the total bacterial count were not presented

Bold values indicates statistically significant (p value < 0.05)

Discussion

Although our pilot studies were performed in a relatively small study group the comparative analysis of the GMB from stone formers and non-stone formers would provide the first published evidence that kidney stone patients may have a distinct gut microbial profile as compared to controls. Other studies have shown that diabetes and obesity, two known risk factors for KSD, are modulated by the GMB. A relationship between GMB and T2D has been clearly suggested by two independent studies that compared the fecal microbiome from healthy and T2D subjects [23, 24]. Manipulation of the GMB has also been shown to alter disease states like obesity and DM. When the GMB of a lean nondiabetic individual is transplanted into an obese diabetic individual, insulin sensitivity improved within 6 weeks and the metabolite profile of the obese patient starts to look more like the lean patient [25]. If a GMB signature could be identified in kidney stone patients, GMB manipulation might represent a novel preventative treatment for KSD.

Whilst, there are presently no published reports associating GMB and KSD, recent data have established an interaction between the GMB and the kidney. Established data have shown that modification of microbiota composition could affect the outcome of glomerulopathies, and recent data indicate that renal inflammation can be modulated with the production of metabolites generated through intestinal fermentation [26]. Short-chain fatty acids (SCFAs) are such metabolites, produced from the fermentation of complex carbohydrates by the intestinal microbiota, and have been shown to ameliorate acute kidney injury. Notably, treatment with short chain fatty acid producing bacteria was found to reduce ischemic damage in a mouse model of renal ischemia reperfusion injury [13]. Khan et al. have suggested that an over-abundance of reactive oxygen species in kidney stone formers may be part of the catalyst for plaque formation and secondary inflammation seen in KSD [27]. Interestingly, data from our lab show that the SCFA butyrate, considered as an important mediator of inflammation [28], was found to be 2.3 times less abundant in kidney stone patients as compared to controls (p = 0.04). Secretion of metabolites is one function of the GMB and it has been shown that metabolite levels often cluster based on enterotype function [29].

It is thought provoking that manipulation of the gut microbiome could influence metabolite levels within the kidney and thereby possibly modulate kidney stone pathology. Our data suggested that bacterial levels do correlate with lithogenic parameters in that Eubacterium was inversely correlated with oxalate levels (r = −0.60, p < 0.06) and Escherichia trended to an inverse correlation with citrate (r = −0.56, p < 0.08).

We, like others, believe that the role of diet represents a major factor in the pathogenesis of KSD [7-10], however we also postulate that a greater understanding of its role may benefit from a new focus on the metabolic implications of diet as influenced by the gut microbiome (GMB). It is presently not known how the GMB changes in response to a kidney stone event or the dietary modification that can follow. It may be that dietary recommendations counseled in a kidney stone clinic are insufficient to significantly alter a patient’s core gut microbiota. Although diet has been shown to rapidly alter an individual’s gut microbial composition [30], data also indicates that long-term baseline diet is strongly associated with a core enterotype partitioning within an individual that may be resistant to small dietary changes. To study the stability of the GMB in response to dietary change, one controlled diet study randomized patients over a 10 day period according to a fixed diet they were asked to maintain. Compositional changes were significant and rapid, but the magnitude of the change was modest and not sufficient to switch individuals between enterotype clusters [31]. We identified 178 bacterial types at the genus level and over 50 % of the total bacterial count was made up of just 5 core enterotypes. With Prevotella and Bacteroides so heavily weighted, and likely to represent part of the groups core enterotype, the other less dominant genera did not cluster well. Perhaps, these less abundant genera are more easily susceptible to environmental and dietary changes. Additionally, clustering subpopulations within the enterotype might only be detectable with substantially more samples. Whether dietary modification has the capability to “switch” the core of one’s GMB in a kidney stone patient is not known, but opens a wide and not previously investigated avenue for continued urolithiasis research. The increasing prevalence of kidney stones and the high stone recurrence rates that are reported despite dietary counsel provide motivation to identify a GMB profile in kidney stone patients as a first step in understanding the role the GMB plays in kidney stone patients.

Limitations

As this was a pilot study aimed in part at determining the feasibility of microbial profiling in this population, several caveats must be noted. Our study was not sufficiently powered to examine the microbiota between first time and recurrent stone formers. Our study did not control for diet. Our study is not a matched cohort study, although multivariate analysis was completed to eliminate confounders. Expanded studies are needed to substantiate these preliminary results; however, although this pilot study involved only a small number of patients significant changes in the GMB were identified. Antibiotics were not given in the near term before specimen collection. However, patient reported antibiotic use has inherent flaws. Nonetheless, data has suggested that the GMB can be substantially altered by the use of antibiotics and can maintain an altered state, that while broadly similar to its native state, community changes after a prolonged period can be maintained [32]. Only 11 out of our 23 kidney stone patients have returned a 24 h urine analysis at the time of this writing.

Conclusion

We identified two genus of bacteria not previously considered to be associated with kidney stone disease. Interestingly, components of the 24-h urine appear to be correlated to bacterial abundance. This is a preliminary data that requires further study. Ongoing work to accrue more patients and to further understand these associations is underway.