Investigation of a targeted panel of gut microbiome–derived toxins in children with chronic kidney disease

Abstract

Background

The gut–kidney axis is implicated in chronic kidney disease (CKD) morbidity. We describe how a panel of gut microbiome–derived toxins relates to kidney function and neurocognitive outcomes in children with CKD, consisting of indoleacetate, 3-indoxylsulfate, p-cresol glucuronide, p-cresol sulfate, and phenylacetylglutamine.

Methods

The Chronic Kidney Disease in Children (CKiD) cohort is a North American multicenter prospective cohort that enrolled children aged 6 months to 16 years with estimated glomerular filtration rate (eGFR) 30–89 ml/min/1.73 m². Data from the 2-year study visit were used for this analysis. Toxin quantification (Metabolon Inc., Durham, NC) was performed with ultra-high performance liquid chromatography/tandem mass spectrometry. Executive function and echocardiograms were assessed. Regression analysis examined the association of toxin levels with eGFR, CKD etiology, and neurocognitive and cardiac assessments (adjusted for age, sex, and urine protein:creatinine [UPCR]).

Results

There were 150 CKiD participants included in this study. All toxins levels were significantly inversely correlated with eGFR (Spearman’s rho − 0.45 to − 0.69). Children with non-glomerular CKD had significantly higher levels of 3-indoxylsulfate, phenylacetylglutamine, and p-cresol glucuronide. The toxin levels did not associate with neurocognitive outcomes. P-cresol glucuronide and phenylacetylglutamine negatively associated with left ventricular mass index z score, but did not associate with left ventricular hypertrophy.

Conclusions

Children with CKD have high levels of circulating gut microbiome–derived toxins. The levels of these toxins are strongly correlated with eGFR. There appear to be differences in toxin level based on glomerular versus non-glomerular etiology, even when accounting for the differences in eGFR between these two subgroups. In this sample, we did not detect any associations between these toxin levels and neurocognitive or cardiac outcomes.

Introduction

In children with chronic kidney disease (CKD), a dysregulated gut–kidney axis is hypothesized to mediate multisystem complications such as neurocognitive and cardiovascular dysfunction [1–11]. The gut–kidney axis refers to the biologic interactions between intestinal dysbiosis and reduced kidney function [5, 6, 12–14]. Reduced kidney function results in accumulation of fluid and circulating metabolic substrates (such as urea). These changes can cause alterations to intestinal microbiome composition (dysbiosis) and membrane permeability. In turn, these intestinal dysbiosis and membrane permeability alterations can lead to the production and accumulation of uremic toxins. These toxins are hypothesized to play a role in mediating the adverse outcomes of CKD.

Previous work in the Chronic Kidney Disease in Children (CKiD) cohort identified significant gut microbiome–derived metabolite associations with neurocognitive outcomes using global semi-quantitative metabolomics profiling [15]. In this study, we investigated the absolute quantification of a targeted panel of five gut microbiome–derived toxins in a subset of CKiD participants with available plasma samples from the 2-year study visit. The five toxins included in the panel were indoleacetate, 3-indoxylsulfate, p-cresol glucuronide, p-cresol sulfate, and phenylacetylglutamine. These five toxins were selected based on previous research showing associations with adverse outcomes in adults with CKD. These five toxins were also selected in part because Metabolon Inc. (Durham, NC) had validated absolute quantification protocols with commercially available analyte calibration standards. Table 1 reviews some existing knowledge regarding these gut microbiome–derived toxins. Among these, 3-indoxylsulfate, p-cresol sulfate, and phenylacetyl-glutamine are the most well-studied in the CKD population [16]. Phenylacetylglutamine has been associated with worse neurocognitive function in children, and cardiovascular and all-cause mortality in adults [15, 17]. P-cresol sulfate, p-cresol glucuronide, and 3-indoxylsulfate have been repeatedly shown to associate with adverse outcomes in adults with CKD [6, 18–24]. There are fewer studies linking adverse outcomes with indoleacetate, but there is a growing interest in tryptophan-indole metabolism alterations in CKD [25–27].

Table 1.

Brief review of five gut microbiome–derived toxins

Toxin	Laboratory-model knowledge	General clinical knowledge	Pediatric CKD knowledge
3-Indoxylsulfate	In cultured rat cardiac myoblast cells, 3-indoxylsulfate induces hypertrophy [69] In endothelial cells, 3-indoxylsulfate promotes a prothrombotic and proinflammatory state, and leads to production of more reactive oxygen species, and induces arteriosclerosis [16, 70]	In adults with CKD, higher levels of 3-indoxylsulfate associated with increased aortic calcification, vascular stiffness, and cardiovascular mortality [18] Levels can be reduced in patients receiving chronic hemodialysis through probiotic supplementation [71]	In children receiving chronic hemodialysis or peritoneal dialysis, levels are elevated (median 5000 uMol/ mL) [29] In the 4C study, higher levels associated with increased risk for CKD progression and cardiovascular disease [51, 72]
Indoleacetate (also known as indole-3-acetic acid)	In cultured human endothelial cells, activated proinflammatory cyclooxygenase-2 pathway and increased production of reactive oxygen species [27] Indoleacetate’s downstream product, indolepropionate, has been shown to inhibit mitochondrial respiration [25]	Associated with higher rates of cardiovascular events and mortality in adults with CKD [27]	Limited
p-cresol glucuronide	A smaller proportion of gut microbiome-produced p-cresol enters circulation as p-cresol glucuronide [73]. In rat models, p-cresol affects both endothelial and renal tubular cells and has been shown to cause	In adults with CKD, a relative increase in p-cresol glucuronide compared to p-cresol sulfate associated with increased mortality and cardiovascular disease [19]	In children with autism spectrum disorder, higher levels associate with worse neurocognitive functioning [60]
p-cresol sulfate	proinflammatory signaling, endothelial dysfunction, and kidney fibrosis [24]	In adults receiving chronic hemodialysis, p-cresol sulfate levels predicted acute cardiac events, and also associated with biomarkers of endothelial dysfunction [22] It is protein-bound and not cleared with hemodialysis [24]	In children receiving chronic hemodialysis or peritoneal dialysis, levels are elevated (median 5000 uMol/ mL) [29] In the 4C study, levels did NOT associate with CKD progression or cardiovascular disease
Phenylacetylglutamine	In heart failure mouse models, phenylacetylglutamine promotes cardiac inflammation and fibrosis [74]	In adults with CKD, higher phenylacetylglutamine levels associated with higher risk of overall and cardiovascular mortality [17] In a general adult population, higher phenylacetylglutamine levels associated with heart failure severity and incident coronary artery disease [75,76] In adults receiving chronic dialysis, higher levels associated with worse neurocognitive outcomes [77]	Associated with worse neurocognitive outcomes in the CKiD study [15]

There is comparatively little pediatric-specific research investigating these gut microbiome–derived toxins [5, 28, 29]. These toxins are not assessed routinely as part of clinical practice, and pediatric CKD–specific reference ranges of their circulating levels have not been established. Previous work using a global semi-quantification approach is not easily translatable to clinical application. This study’s primary objective was to measure the absolute quantification of these five gut microbiome–derived toxins. This study’s secondary objectives were to assess the associations of these quantified levels with kidney function and neurocognitive and cardiovascular parameters.

Methods

Study participants

We leveraged data from the CKiD study, a large North American multicenter prospective cohort of pediatric CKD [30, 31]. CKiD participants included for this analysis were enrolled between January 2005 and December 2014 across 54 medical centers. Complete details of study design, methods, and cohort characteristics have been published previously [30]. Children between the ages of 6 months and 16 years were enrolled if they had an estimated glomerular filtration rate (eGFR) of 30–90 ml/min/1.73 m². Exclusion criteria included history of solid organ or bone marrow transplant, dialysis within 3 years, malignancy, HIV infection within 12 months, structural cardiac disease, and genetic syndromes with central nervous system involvement. The CKiD study was approved by the institutional review board of each participating center. The CKiD study adhered to the Declaration of Helsinki.

For this study, 150 participants were selected from the 2-year study visit based on frozen plasma sample availability. Participants with neurocognitive phenotyping from the 2-year study visit were prioritized.

eGFR was calculated using the CKiD U25 equations using both serum creatinine and cystatin C [32]. Hypertension was defined based on standardized CKiD study visit systolic or diastolic blood pressure readings greater than the 95^th percentile based on age, sex, and height [33]. Study participants reported if they were taking antibiotics, angiotensin-converting enzyme inhibitor, or angiotensin II receptor blocker (ACEi/ARB) medications. Anemia was defined based on hemoglobin less than the 5^th percentile based on age, sex, and self-reported race as defined by the Centers for Disease Control and Prevention [34]. Proteinuria was assessed with a random sample urine protein:creatinine (UPCR).

Targeted metabolite quantification

Plasma samples were sent to Metabolon Inc. (Durham, NC) for untargeted ultra-high performance liquid chromatography tandem mass spectrometry–based (UHPLC-MS/MS) metabolomics quantification [35–37]. This is a different assay procedure from previous CKiD global metabolomics studies. Study samples, calibration samples, and quality control samples were included in the run. Calibration samples were prepared at eight different concentration levels. The quality control samples were Metabolon standardized, pooled healthy human adult donor plasma samples. The mass spectrometer was operated in negative mode using electrospray ionization. The peak area of the individual analyte product ion was measured against the peak area of the product ion of the corresponding internal standard. Quantitation was performed using weighted linear least squares regression analysis generated from fortified calibration standards prepared immediately prior to each run. Analyte concentrations that fell above the limit of quantitation were extrapolated, and reported values were given a comment of “above limit of quantification.” Quality control pooled donor sample mean analyte concentrations were also reported. We cannot make any direct statistical assessments to the pooled donor sample levels as we did not have access to Metabolon’s full internal data.

Neurocognitive phenotyping

Details of CKiD neurocognitive assessments have been published [38, 39]. For this study, we analyzed executive function assessments from the 2-year study visit. Executive function was assessed by the parent-completed Behavior Rating Inventory of Executive Function and the Delis–Kaplan Executive Function System Tower Subtest. These assessments have well-developed normative data and satisfactory psychometric properties. Assessments were standardized individually, and an average executive function z score was calculated. We focused on this composite assessment because a previous CKiD study using untargeted metabolomic profiling data identified gut microbiome–derived metabolite associations with executive function [15].

Cardiovascular phenotyping

Echocardiogram assessment data from the 2-year CKiD follow-up visit were analyzed for this study. Study participants underwent M-mode and Doppler echocardiography. All images were standardized at the Cincinnati Cardiovascular Imaging Research Core. Left ventricular mass was indexed to height^2.7 and normalized based on age and sex (LVMI z score). Left ventricular hypertrophy (LVH) was defined as an LVMI z score greater than the 95^th percentile [40–42]. Left ventricular diastolic dysfunction was characterized by an elevated filling pressure (E/e’ > 8.0), which has previously been shown to be the most prevalent marker of diastolic dysfunction in the CKiD study [4, 8, 43–45].

Statistical analyses

Toxin levels were described with median and interquartile range. Mean and standard deviation were also calculated using natural log-transformed metabolite levels, for which the results were back-transformed for reporting. In all subsequent regression analyses, toxin levels, eGFR and UPCR were natural log-transformed. All statistical analyses were run in RStudio. Representative R code including documentation of all packages utilized will be uploaded to a GitHub repository. The significance threshold was set at p < 0.05.

Two-sample t tests and chi-square tests assessed differences between participants with glomerular versus non-glomerular CKD etiology. Toxin associations with eGFR were described with Spearman correlation and multivariable linear regression (adjusting for age, sex, and UPCR). Differences in toxin levels based on glomerular versus non-glomerular diagnoses were analyzed using linear regression analysis adjusting for age, sex, eGFR, and UPCR.

We had previously published on global semi-quantification metabolomics data showing differences in gut microbiome–derived metabolites based on glomerular versus non-glomerular CKD etiology. We had hypothesized this was potentially related to antibiotic exposure among children with non-glomerular CKD. In this subsample with targeted absolute toxin quantification, we assessed potential toxin level differences based on antibiotic usage with linear regression analysis adjusting for age, sex, eGFR, and UPCR.

We had previously published on global semi-quantification metabolomic associations with executive function z score in the overall CKiD cohort [3]. In this subsample with targeted absolute toxin quantification, our replication analysis assessed toxin associations with executive function z score with multivariable linear regression, adjusting for age, sex, glomerular CKD etiology, eGFR, UPCR, hypertension, and CKD duration. These covariate adjustments are consistent with our previous analyses of neurocognitive outcomes. Toxin associations with cardiac assessments were assessed with multivariable regression adjusting for age, sex, glomerular CKD etiology, eGFR, UPCR, hypertension, ACEi/ ARB usage, CKD duration, and anemia (linear regression for LVMI z score and E/e’, logistic regression for LVH, and elevated filling pressure). ACEi/ARB usage and anemia were included as covariate adjustments in these analyses because these are clinical factors previously shown to associate with cardiac outcomes in pediatric CKD [46, 47].

Results

A total of 150 participants were included in this analysis. Participant characteristics are reported in Table 2. Participants with glomerular disease were older (p < 0.001), and had higher eGFR (p < 0.001), higher UPCR (p < 0.001), shorter CKD duration (p < 0.001), and executive function z scores (p < 0.001). Participants with glomerular disease were more likely to be using ACEi/ARBs (p < 0.001). Participants with non-glomerular disease had higher calcium, phosphorous, and intact parathyroid hormone levels (all p < 0.001).

Table 2.

Participant characteristics

	Overall sample	Glomerular	Non-glomerular
Participant characteristics
N	150	53	97
Age (years)	14 (11, 17)	15 (12, 17)	13 (10, 16)
Male sex	92 (61%)	28 (53%)	64 (66%)
BMI z score	0.5 (− 0.3, 1.3)	0.6 (− 0.1, 1.8)	0.4 (− 0.3, 1.0)
CKD characteristics
eGFR (ml/min/1.73 m2)	53 (39, 72)	67 (49, 80)	47 (36, 65)
Urine protein:creatinine (mg/mg)	0.2 (0.1, 0.7)	0.3 (0.1, 0.8)	0.2 (0.1, 0.7)
Anemia	43 (29%)	19 (36%)	24 (25%)
Hypertension	20 (13%)	8 (15%)	12 (12%)
ACEi/ARB usage	91 (61%)	43 (81%)	48 (49%)
Antibiotic usage	30 (20%)	7 (13%)	23 (24%)
CKD duration (years)	9 (5, 13)	4 (2, 8)	11 (9, 14)
Additional lab values
Calcium (mg/dL)	9.4 (9.1, 9.7)	9.3 (9.0, 9.5)	9.6 (9.2, 9.8)
Phosphorus (mg/dL)	4.3 (3.8, 4.8)	4.2 (3.8, 4.8)	4.4 (3.8, 4.9)
Intact parathyroid hormone (pg/mL)	51 (32, 75)	40 (31, 63)	57 (35, 86)
Neurocognitive outcomes (overall n = 116, glomerular n = 38, nonglomerular n = 82)
Executive function z score	− 0.1 (− 0.9, 0.5)	0.1 (− 0.6, 0.4)	− 0.1 (− 0.9, 0.5)
Cardiovascular outcomes (overall n = 108, glomerular n = 36, nonglomerular n = 72)
Left ventricular hypertrophy	8 (7%)	3 (8%)	5 (7%)
Left ventricular mass index z score	− 0.4 (− 1.2, 0.6)	− 0.3 (− 1.0, 0.6)	− 0.5 (− 1.2, 0.5)
Abnormal filling pressure (E/e’> 8.0)	17 (16%)	4 (11%)	13 (18%)

Continuous variables are presented as median (interquartile range). Discrete variables are presented as n (%). There were characteristic differences between the glomerular and non-glomerular subgroups based on two sample t tests and chi-square tests. Participants with glomerular disease were older (p < 0.001), had higher eGFR (p < 0.001), had higher UPCR (p < 0.001), shorter CKD duration (p < 0.001), and higher parental assessment of executive function z scores (p < 0.001). Participants with glomerular disease were more likely to be using ACEi/ARBs (p < 0.001)

Toxin levels and distribution are reported in Table 3. In conjunction, we report the sample means of the Metabolon-provided pooled healthy adult donor samples. The mean and median levels of all five toxins in children with CKD were above the pooled healthy donor mean levels. We cannot make direct statistical comparisons as we did not have the standard deviations of the pooled healthy donor means. We report the number of participants with toxin levels detected above Metabolon’s quantification threshold. In this table, we also report the toxins’ association with eGFR. All toxins were inversely correlated with eGFR (p < 0.001 for Spearman correlation and multivariable linear regression for all metabolites), with the magnitude of these correlations being within the moderate to strong range. Figure 1 visualizes the relationship between the toxin levels and eGFR. Table 4 also reports toxin distribution based on CKD stage.

Table 3.

Distribution of metabolite levels and how they associate with eGFR

	Metabolon-pooled healthy human donor mean (ng/mL)	CKiD sample mean (± 2SD), (ng/mL)	CKiD sample median (IQR), (ng/mL)	Samples with level > validated upper limit of detection, n (%)	Spearman’s rho	Estimate (95% CI) (for every 10% higher eGFR, x% difference in toxin level)
3-indoxylsulfate	567	1331 (205, 8624)	1453 (811, 2455)	6 (4%)	− 0.69	– 10.8 (− 13.1, − 8.4)
Indoleacetate	297	382 (133, 1098)	348 (263, 573)	0	− 0.53	– 5.6 (− 7.2, − 4.0)
p-cresol glucuronide	17.9	25 (2, 317)	27 (11, 65)	1 (1%)	− 0.45	– 10.3 (− 14.0, − 6.5)
p-cresol sulfate	4088	4889 (440, 54,368)	5861 (2845, 10,148)	14 (9%)	− 0.60	– 9.4 (− 12.8, − 5.8)
Phenylacetylglutamine	494	745 (123, 4541)	750 (435, 1297)	2 (1%)	− 0.59	– 10.0 (− 12.3, − 7.7)

Toxin level distributions and their associations with kidney function. Means of the Metabolon-pooled healthy adult donor plasma samples also provided from the Metabolon data curation report. Additional descriptive statistics of these Metabolon samples were not available in this report

Fig. 1.

Visualization of toxins’ association with eGFR. Association of circulating metabolite levels with kidney function, both as ml/min/1.73 m² versus ng/mL and on natural log-transformed axes. All five were inversely associated with kidney function. There was still a degree of variability despite this strong linear relationship

Table 4.

Toxin levels based on CKD stage

Toxin levels in ng/mL	Stages 1 and 2 (≥ 60)	Stage 3 (30–59)	Stages 4 and 5 (< 30)
n	60	75	15
3-indoxylsulfate	750 (449, 1252)	1833 (1296, 2716)	4419 (3690, 7649)
Indoleacetate	289 (233, 385)	396 (289, 652)	772 (515, 1089)
p-cresol glucuronide	14 (6, 29)	33 (17, 64)	100 (63, 215)
p-cresol sulfate	3103 (1643, 5681)	7662 (4774, 11,320)	17,092 (10,452, 20,370)
Phenylacetylglutamine	502 (267, 782)	870 (595, 1332)	2458 (1790, 3430)

There were differences in toxin levels based on glomerular versus non-glomerular CKD etiology. Participants with glomerular CKD had lower levels of 3-indoxylsul-fate (difference estimate = − 37% [95% confidence interval: − 63, − 11], p < 0.01), p-cresol glucuronide (− 43% [− 83, − 2], p < 0.05), and phenylacetylglutamine (− 41% [− 65, − 16], p < 0.01). The differences in these toxins based on CKD etiology are illustrated in Fig. 2.

Fig. 2.

Toxin level differences based on CKD etiology. Density plots on a natural log-transformed scale of circulating toxin levels by glomerular versus non-glomerular CKD etiology. The Metabolon-pooled healthy adult donor sample means are also indicated. Children with non-glomerular CKD had higher levels of 3-indoxylsulfate, phenylacetylglutamine, and p-cresol glucuronide

Thirty participants (20%) in the overall sample reported antibiotic usage. Among the entire sample, there were no differences in toxin level based on antibiotic usage. Twenty-three participants (24%) with non-glomerular CKD reported antibiotic usage. Among participants with non-glomerular CKD, p-cresol sulfate levels were lower in those reporting antibiotic usage (− 56% [− 102, − 9], p = 0.02).

There were no toxin associations with executive function z score or elevated filling pressure (Table 5). Phenylacetyl-glutamine and p-cresol glucuronide had significant negative associations with LVMI z score, but there were no associations with LVH. Full results of these analyses are reported in Table 5.

Table 5.

Toxin associations with clinical outcomes

Executive function z score
Metabolite	Estimate	p value	95% CI lower	5% CI upper
3-indoxylsulfate	− 0.01	> 0.05	− 0.03	0.02
Indoleacetate	− 0.01	> 0.05	− 0.04	0.03
Phenylacetylglutamine	− 0.02	> 0.05	− 0.05	0.00
p-cresol sulfate	− 0.02	> 0.05	− 0.04	0.01
p-cresol glucuronide	− 0.01	> 0.05	− 0.03	0.00
Left ventricular mass index z score					Left ventricular hypertrophy, by LVMI z score > 95th-percentile for age and sex
Metabolite	Estimate	p value	95% CI lower	95% CI upper	Odds ratio	p value	95% CI lower	95% CI upper
3-indoxylsulfate	− 0.02	> 0.05	− 0.05	0.01	1.17	> 0.05	0.92	1.46
Indoleacetate	− 0.03	> 0.05	− 0.08	0.01	0.77	> 0.05	0.59	1.02
Phenylacetylglutamine	− 0.05	0.003	− 0.08	− 0.02	0.85	> 0.05	0.66	1.10
p-cresol sulfate	− 0.02	> 0.05	− 0.04	0.00	1.02	> 0.05	0.88	1.19
p-cresol glucuronide	− 0.03	0.004	− 0.05	− 0.01	0.95	> 0.05	0.82	1.10
E/e’					Abnormal filling pressure, by E/e’ > 8.0
Metabolite	Estimate	p value	95% CI lower	95% CI upper	Odds ratio	p value	95% CI lower	95% CI upper
3-indoxylsulfate	− 0.01	> 0.05	− 0.07	0.05	1.02	> 0.05	0.94	1.12
Indoleacetate	0.01	> 0.05	− 0.08	0.10	1.06	> 0.05	0.94	1.21
Phenylacetylglutamine	− 0.03	> 0.05	− 0.10	0.03	0.95	> 0.05	0.87	1.05
p-cresol sulfate	− 0.03	> 0.05	− 0.09	0.02	0.94	> 0.05	0.87	1.02
p-cresol glucuronide	− 0.02	> 0.05	− 0.06	0.01	0.96	> 0.05	0.91	1.02

There were no significant associations between toxin levels with clinical outcomes. For executive function, we report estimated percent-change in z score per 10% increase in metabolite level. For LVMI z score and E/e’, we report change per 10% increase in metabolite level. For LVH and elevated filling pressure, we report odds ratios per 10% increase in metabolite level

Discussion

We examined a targeted panel of five gut microbiome–derived toxins and their associations with kidney function and clinical outcomes in children and adolescents with CKD. An important aspect of this study was confirming relationships that have been observed in adults with CKD and in smaller samples of children with CKD [5, 17–20, 24, 29, 48, 49]. We reported levels of these gut microbiome–derived toxins measured against calibrated standards, and how these levels correlated with eGFR. We also showed that children with non-glomerular CKD had higher levels of several of these substances compared to children with glomerular CKD.

In current clinical practice, we do not routinely assess the levels of these gut microbiome–derived toxins in children with CKD. There is an emerging understanding of how the gut microbiome is perturbed in CKD [6, 28, 50]. This intestinal dysbiosis can result in the circulating accumulation of gut microbiome–derived toxins, which in turn may mediate the adverse outcomes of CKD [9, 13, 14]. As this research area advances, it is important to support efforts in translating this new knowledge toward clinical application. Specific reference ranges for children with mild–moderate CKD for these five gut microbiome–derived toxins have not been established. Our study leveraged a large sample from the CKiD cohort to report plasma levels assayed against standardized concentrations. Our results were similar to observed median levels of 3-indoxylsulfate (1128 ng/mL) and p-cresol sulfate (3237 ng/mL) in a large European population of children with CKD stages 3–5, providing some degree of external replication [51].

One interesting observation of this study is that for all five assayed substances, the mean and median levels among CKiD participants were higher than the pooled healthy adult sample means. This observation should be interpreted with caution, as we are not able to make direct biostatistical inferences using this information provided by Metabolon Inc. However, it does raise important questions that should drive continued investigation, such as in a non-CKD population, are there age-related differences in plasma concentrations of these substances, and what factors may be driving these differences?

Prior research and this study have shown that although there is a strong correlation between circulating toxin levels and kidney function, there is a substantial variability in toxin level distribution [51]. We echo these previous studies in stating the need for identifying what other CKD-related factors may drive the observed variability in toxin levels.

One potential clinical factor driving toxin level variability is antibiotic usage. One of the most notable findings in this study is the difference in toxin levels between participants with non-glomerular versus glomerular CKD. Although the non-glomerular subgroup within this study had lower eGFR compared to the glomerular subgroup, the toxin level differences remained significant after adjusting for eGFR in multivariable regression analyses. We hypothesized that this may be related to antibiotic exposure differences between the two subgroups. Children with non-glomerular CKD largely have congenital anomalies of the kidney and urinary tract (CAKUT) and may have increased antibiotic exposure related to prophylaxis and treatment of urinary tract infections [52–56]. This may translate to an increased risk for intestinal dysbiosis [54, 57–59]. In this sample, a higher proportion of participants with non-glomerular CKD reported antibiotic usage. However, only p-cresol sulfate levels were significantly lower among participants with non-glomerular CKD and antibiotic usage. How the gut microbiome and associated toxin production is affected by antibiotic usage is incompletely described in pediatric CKD. In addressing this study, we are limited in that only 30 of 150 participants reported antibiotic usage. We are also not able to assess potential differences related to urinary tract infection prophylaxis or treatment antibiotic dosing.

An alternative hypothesis to the observed glomerular versus non-glomerular differences is that children with CAKUT are at risk for bowel-bladder dysfunction, which in turn can mediate gut microbiome disturbances [53]. A fascinating intersecting research area is in children with autism spectrum disorder with normal kidney function, among whom bowel-bladder dysfunction is prevalent, changes in gut microbiome composition and elevated levels of these toxins have been demonstrated [60, 61]. Some of these gut microbiome–derived toxins have been associated with worse neurocognitive function, similar to what has been previously reported in CKiD [15, 60, 62].

It is important to acknowledge that these toxin levels are still theoretical biomarkers. More studies are needed to potentially validate these substances as biomarkers of either gut microbiome disturbances or predictors of adverse CKD outcomes. This will be of clinical importance as there are evolving questions regarding the indications, benefits, and risks of antibiotic urinary tract infection prophylaxis in the CAKUT population [63–67]. As more data emerge regarding the potential negative health impact of intestinal dysbiosis in CKD, the implication of these findings needs to be weighed against the benefits of urinary tract infection prophylaxis.

An important negative finding of our study is that we did not detect any association between these five metabolites with selected clinical outcomes in this CKiD sample. In linear regression analyses, phenylacetylglutamine and p-cresol glucuronide negatively associated with LVMI z score, which is contrary to what has been observed in previous studies in adult CKD [9, 10, 18, 21–23, 25]. This association did not translate into significant associations with LVH. Phenylacetylglutamine was negatively associated with executive function in a much larger CKiD study using untargeted metabolomic profiling [15].

There are several possible reasons why this study could have failed to detect potential biologic relations between these substances and our selected clinical outcomes. In addition to unmeasured confounding, bias may have been introduced based on the sampling of the CKiD cohort. Notably, there was a lower prevalence of LVH compared to the overall CKiD cohort (7% in this subsample compared to 15–20% in the overall CKiD cohort) [41]. CKiD participants with more longitudinal follow-up have also previously been shown to be a healthier subsample of the full cohort [15, 68].

Another intrinsic bias of this study was the selection of five of the more commonly studied gut microbiome–derived metabolites based primarily on adult data. We do not know if these gut microbiome–derived toxins affect children and adults with CKD in the same way.

An important future investigation would be to assess potential relationships between antibiotic exposure and differences in circulating gut microbiome–derived metabolite levels in children with CKD. Another important direction would be correlating these circulating metabolite levels with measured gut microbiome perturbations through stool metagenomic sequencing. How dietary intake affects gut microbiome–derived uremic toxin generation could also be further studied. Additional analyses could also investigate potential differences related to free versus protein-bound toxin levels.

In summary, our study reported plasma concentrations of 3-indoxylsulfate, indoleacetate, phenylacetylglutamine, p-cresol sulfate, and p-cresol glucuronide in a large sample of children with mild–moderate CKD. A key contributing strength of this study is that these measurements were assayed against standardized concentrations, as reference ranges for clinical or research use have not yet been established for the pediatric CKD population. Another interesting finding was that children with non-glomerular CKD had higher levels of 3-indoxylsulfate, phenylacetylglutamine, and p-cresol glucuronide, raising questions as to what CKD-related factors may be driving differences in these gut microbiome–derived substances. As the interest in the interactions between intestinal dysbiosis and CKD continues to grow, continued research dedicated to study these substances and their potential use as biomarkers in pediatric CKD is essential.

Supplementary Material

Supplementary Information The online version contains supplementary material available at 10.1007/s00467-024-06580-6.

Data availability

Parties interested in accessing deidentified CKiD data may contact Judith Jerry (jjerry@jhu.edu) at Johns Hopkins University.