Urinary Microbiome Dysbiosis in Children With Congenital Uropathies at Varying Risk for Urinary Tract Infections

Abstract

Introduction

Febrile urinary tract infections (UTIs) may occur in 30% to 50% of children with vesicoureteral reflux (VUR) or posterior urethral valves (PUVs), frequently leading to renal scarring despite chemoprophylaxis. Approximately 15% of children with uretero-pelvic junction obstruction (UPJO) may develop UTIs. However, investigations that can identify at-risk children before the first episode of UTI are lacking. In this exploratory study, we investigated the preinfection urinary microbiome in Indian children with congenital anomalies of the kidney and urinary tract (CAKUT) to determine whether microbiome alterations, metabolic potential, and antibiotic resistance profiles precede UTI.

Methods

In this prospective cohort study with follow-up, urine samples were collected from 80 children: 36 with newly diagnosed, antibiotic-naïve CAKUT (18 UPJO, 12 VUR, 6 PUV) and 44 controls. Patients were stratified a priori into low (n = 19) and high-risk (n = 17) groups using clinically defined UTI-susceptibility criteria. V3–V4 16S ribosomal RNA sequencing was used to define urinary microbial profiles. Alpha- and beta-diversity were compared using Shannon index and permutational multivariate analysis of variance (PERMANOVA), respectively. Sliding-window and network-based analyses were used to map dysbiosis gradients. Patients were followed-up longitudinally to assess UTI incidence. Identified dysbiosis-linked microbial markers at baseline were investigated using Kaplan-Meier and Cox-proportional hazard-based analyses as predictors of UTI-risk. Metabolic functions were inferred from taxonomic data. Antibiotic resistance patterns were characterized using the Comprehensive Antibiotic Resistance Database – Resistance Gene Identifier (CARD-RGI) and the World Health Organization Access, Watch, and Reserve classification.

Results

Urinary microbial alpha diversity declined significantly from controls to low-risk to high-risk groups (P = 0.002), accompanied by an increase in intragroup variability (P ≤ 0.005). PERMANOVA revealed distinct clustering by risk (R² = 0.11; P = 0.001). Dysbiosis scores inversely correlated with the first Kendall Principal Coordinates Analysis (PCoA) axis (ρ = −0.62; P < 0.001). With increasing risk of UTI, the commensal, control-associated genera declined along this axis while the facultative pathogens became dominant. Control-associated microbiomes favored short- and branched-chain fatty acid and spermidine production; high-risk microbiomes overproduced ammonia, putrescine, and cadaverine. Resistance to 18 of 22 routinely tested antibiotics was almost confined to the 31 risk-associated microbiomes (P = 0.001). During the median (interquartile range) follow-up of 564 (518–594) months, 14 of 36 children with CAKUT developed UTIs, and baseline depletion of health-associated microbial consortia correlated with reduced UTI-free survival. A panel of 10 species-level and 12 genus-level taxa were identified as health-associated markers negatively associated with future UTI-risk during follow-up investigation.

Conclusion

Children with CAKUT exhibit urinary microbiome dysbiosis before their first symptomatic UTI, characterized by loss of conserved health-associated taxa, metabolic imbalance, and broad-spectrum antibiotic resistance. These findings support the potential of microbiome-informed, noninvasive risk stratification and microbiome-tailored prophylaxis, while establishing the first Indian pediatric reference set for CAKUT-related UTI prevention.

Graphical abstract

UTI is one of the most common serious bacterial illnesses in childhood and a recognized driver of renal scarring that can culminate in chronic kidney disease when predisposing factors are present.¹ Children born with CAKUT, a spectrum that includes common anomalies such as PUVs, VUR, and UPJO carry a particularly heavy burden of UTI. Previous studies report that 30% to 50% of boys with VUR or PUVs, and ≤15% of infants with UPJO will experience ≥1 febrile UTI during childhood.² Breakthrough infections in VUR despite continuous chemoprophylaxis remain common; and circumcision, although beneficial, does not abolish the risk in these patients.³ Timely identification of those at greatest risk remains challenging: current prediction models rely largely on structural imaging and clinical history, but many children develop breakthrough infections despite optimal anatomic management and chemoprophylaxis.

The challenge is magnified by the rapid global increase in antimicrobial resistance in pediatric UTI.⁴ Recent studies on CAKUT-associated UTIs showed multidrug resistance in more than half of Escherichia coli, Klebsiella, and Pseudomonas isolates, compromising empirical regimens that rely on broad-spectrum antibiotics.⁵ Precision stewardship treating only the children in whom the next infection is genuinely imminent and tailoring the drug class to their predicted resistome has become a global priority. In this context, the role of the baseline urine microbiome becomes ever more important.

During the past decade, culture-independent methods have overturned the dogma of a sterile urinary tract, revealing a low biomass but metabolically active microbial community whose composition appears to influence bladder health.⁶ Adult studies link urinary “dysbiosis” to urgency incontinence, urolithiasis, and recurrent UTI, typically reporting a loss of alpha diversity, greater interindividual dispersion, and enrichment of facultative pathogens along a dysbiosis gradient.7, 8, 9 Pediatric data remain sparse, fragmented, and almost exclusively from a few countries.¹⁰ No study has yet characterized the urinary microbiome in Indian children with CAKUT, a population in which early-life microbial assembly and antibiotic exposure differ markedly from high-income settings.^11,12 Elucidating microbial signatures that precede febrile UTI in Indian children with CAKUT could therefore inform locally relevant risk-stratification tools and microbiome-directed interventions.

Dysbiosis in any microbiome is typically characterized by 2 interrelated phenomena. The first is the loss of features associated with a healthy, resilient state, often marked by high microbial diversity and the presence of key microbes that resist pathogen colonization.¹³ These keystone microbes often occupy central ecological roles¹⁴ and their loss is commonly followed by the second phenomenon: the emergence or overgrowth of atypical microbes, including potential pathogens. This pattern exemplifies the Anna Karenina principle (AKP) of microbiomes, which posits that under stress, microbiomes lose conserved core members and diverge along multiple unstable trajectories, resulting in multiple low-diversity configurations, typically pathogen-enriched. Whether the AKP applies to urinary microbiomes in children predisposed to UTIs remains currently unknown. Specifically, it remains unclear whether the pediatric urinary microbiome harbors keystone species whose loss leads to the emergence of pathogen-enriched microbiomes preceding the first febrile UTI. Likewise, the distribution of antibiotic-resistance profiles across health-associated versus risk-associated urinary microbes and the role of antibiotic prophylaxis in this UTI-preceding dysbiosis of the urine microbiome is poorly understood.

With this background, we conducted the first comprehensive urinary microbiome study in Indian children with CAKUT prone to UTIs and healthy controls. Using in-depth 16S ribosomal RNA sequencing of 80 pediatric urine samples, we compared microbial diversity, community structure, inferred metabolic capacity, and antibiotic resistance patterns among controls and CAKUT subgroups a priori stratified by a clinical risk criterion for risk of UTI. We hypothesized that pathogen-enriched, low-diversity microbiota would be detectable in high-risk children before symptomatic infection, and these communities would harbor broader resistomes than health-associated counterparts. We further anticipated that baseline urinary microbiome signatures would be associated with longitudinal UTI outcomes during follow-up. Specifically, depletion of health-associated microbial consortia at baseline was expected to correlate with reduced UTI-free survival.

Methods

Study Design and Setting

This single-center prospective cohort study was conducted between December 2023 and March 2025 in the Department of Pediatric Surgery at the All India Institute of Medical Sciences, New Delhi, India. Ethical approval was obtained from the Institutional Review Board (Ref. No: IEC-577/01.09.2023), and the study adhered to institutional guidelines for ethical research. Written informed consent was obtained from the parents of all children, with assent provided by participants as age appropriate.

Study Population

A total of 80 children, aged < 14 years, were consecutively included in this study. Of these, 36 children were newly diagnosed with one of the 3 most common congenital uropathies (PUVs, VUR, and UPJO), having a predisposition to UTI. The diagnosis of these anomalies was made as per the standard clinical and radiological criteria.¹⁵ None of them had a solitary kidney, or anatomic variations (duplex system, horseshoe kidney, crossed ectopic kidney, or crossed fused anomaly), or > 1 CAKUT phenotype. In addition, none of these patients were having active UTI or past history of UTI. In addition, they were neither on antibiotics at the time of recruitment (urine sample collection) nor were they having any prior history of antibiotic exposure. Following urine sampling, they were managed as per the standard treatment protocols of the Department of Pediatric Surgery.

The remaining 44 of 80 participants were controls diagnosed with a nongenitourinary disease during their outpatient department visit(s). None of the controls had any chronic illnesses. Exclusion criteria for this group were any known congenital or acquired kidney disease, prior history of UTI, presence of urinary symptoms, and abnormal findings on routine urine analysis.

Patient Groups

The distribution of 36 patients based on the disease phenotype was as follows: 18 with UPJO (2 bilateral), 12 with VUR, and 6 with PUVs. All patients with congenital uropathies were divided into 2 risk groups, that is, low and high-risk groups, based on their risk of developing UTI. Because CAKUT encompasses heterogeneous anatomical entities and there is no universally accepted cross-phenotype UTI-risk classification, we used an a priori clinical stratification aimed specifically at UTI susceptibility. Risk categories were assigned based on the presence of key infection-promoting mechanisms, that is, reflux facilitating bacterial ascent, urinary stasis from obstruction, and/or bladder dysfunction with incomplete emptying, together with age at presentation and presenting symptoms.⁵ Therefore, children with high-grade VUR (grade 4 or 5) were assigned a high risk of UTI, whereas those with low grades of VUR (grade 1–3) were assigned a low risk of UTI.¹⁶ In contrasts, the reported incidence of UTI in postnatally diagnosed PUVs is 50%.¹⁷ All children with PUV in our study were aged < 9 months and were uncircumcised; thereby putting them at high risk for UTIs.³ Among the patients with UPJO, the risk of UTI is generally low (between 0% and 15%).¹⁸ Therefore, all 16 children with unilateral UPJO were assigned a low risk of UTI. However, 2 patients with bilateral UPJO were assigned a high risk because of the presence of fever and urinary symptoms at the time of presentation in both. As a result, 19 of 36 children were grouped in low risk, whereas 17 of 36 were assigned high risk (Supplementary Table S1A).

Sample Collection, Processing, DNA Extraction, and Sequencing

The detailed description of the workflow adopted for the collection and processing of the urine samples, along with the strategy adopted for DNA extraction, transport and sequencing has been presented in Supplementary Text S1.

Preprocessing of Raw Reads and Taxonomic Processing of Microbiome Data

The processing of raw microbiome sequence data was performed using FastQC v0.12.1 and Trimmomatic v0.39.¹⁹ The taxonomic classification of the reads was performed using SPINGO v3.0.²⁰ The detailed protocol for these steps is described in Supplementary Text S2.

Data Normalization, Diversity Investigations, and Computation of Other Microbiome-Level Summary Indices

The protocol for normalizing the data and for computing and comparing alpha diversity across groups as well as the computation of other microbiome-level summary indices (like Kendall-Uniqueness and Dysbiosis Score) has been summarized in Supplementary Text S3. In summary, various modules of R v4.3.1²¹ were used for normalization and for computation of the alpha diversity estimates. Beta-diversity was specifically assessed using PCoA,²² and PERMANOVA was performed to calculate the extent of separation between the study groups.²³

Sliding-Window–Based Approach to Understand the Shift in Microbial Composition Using Kendall Distance-Based PCoA

A sliding-window approach was employed to capture the gradual compositional variations of the urine microbiome along the PCoA1 axis of microbiome beta-diversity derived from the Kendall distance-based PCoA (Kendall-PCoA-1). This was the axis along which the microbiomes could be placed based on a gradual shift in phenotypes, ranging from controls (at one end) to low risk to high risk (at the other extreme end). The microbiomes were then divided into 9 equal, partially overlapping windows along the PCoA1 axis (Supplementary Figure S1A), and changes in the species abundance and prevalence were assessed across these windows. Each window had half of the overlap of the previous window.²⁴ A bubble plot was generated using the ggplot function from the ggplot2 v3.5.0²⁵ package in R. The species were classified into 3 different groups as follows: (i) species associated with the control group, (ii) species associated with the UTI risk group, and (iii) undefined species that are not well-defined in any of the 2 groups. This approach was employed at the species level and the genus level, and only microbes present in ≥ 5% of all the groups (controls, low risk, and high risk) were considered for this analysis (Supplementary Table S2A and B). Given the relatively small size of our study cohort, we additionally performed calculations using the pwr.r.test function pwr package version 1.3.0.²⁶

Network-Based Approach to Identify the Microbial Modules

Furthermore, for identifying the microbes-based modules and to represent their abundance change with respect to the UTI risk, a network using the Compositionality Corrected by Renormalization and Permutation (ccrepe) v1.36.0²⁷ package in R was employed and the relationships visualized using Cytoscape v3.10.2.²⁸

Computation of Inferred Metabolite Profiles

The metabolic profiles for each urinary microbiome were inferred from their corresponding species-level taxonomic compositions using methodologies described in previous studies,^22,29 using maps between microbes’ metabolic functionalities derived from experimentally validated metabolic-functional profiles (i.e., the production and consumption patterns) collated from multiple repositories and published studies^30,31 (Supplementary Text S4).

Follow-Up Clinical Assessment and Longitudinal Outcome Analysis

In addition to the microbiome characterization, longitudinal clinical follow-up investigation was performed to assess whether and if so, which baseline microbial features were associated with subsequent development of UTI. This investigation is described in detail in Supplementary Text S5. In brief, we first investigated the UTI-free survival patterns in the low-risk and high-risk patient groups using Kaplan-Meier analysis.³² Then, we investigated the specific microbiome features measured at the baseline for their associations with UTI incidences during the follow-up period using a combination of across-group Mann-Whitney Tests and Cox Proportional Hazards models.³³

Identifying the Resistance Gene in Microbial Species Using CARD-RGI

The Resistance Gene Identifier (RGI) software, along with the Comprehensive Antibiotic Resistance Database (CARD)³⁴ was used to identify the antibiotic-resistant genes (to 75 different antibiotics) in the species associated with different patient groups. The classification of antibiotics was based on the World Health Organization Access, Watch, and Reserve classification.³⁵ In addition, from the list of different antibiotics within the CARD-RGI program, we selected 22 antibiotics that are commonly employed for antibiotic susceptibility testing at our center. The details of this analysis are provided in Supplementary Text S6.

Results

The median (interquartile range) age of the patients and controls were 17 (8–60) months and 60 (36–109) months, respectively. At the time of recruitment, the patients were significantly younger than the controls (P = 0.0002). The sex distribution (male/female) was 28/8 and 26/18 among patients and controls, respectively. There was no significant difference in sex distribution among patients versus controls (P = 0.08). The baseline characteristics of the included participants are depicted in Table 1 (with the individual subject-specific data provided in Supplementary Table S1A–C). Among the patients, 33% (12/36) were having an antenatal diagnosis of CAKUT. Despite the observed difference in age between controls and patients, age was not significantly linked to the variation in microbiome composition (discussed later; Supplementary Table S1B).

Table 1.

Baseline characteristics of the cohort

Characteristics	Patients CAKUT (n = 36)	Controls (n = 44)
Median (IQR) age; in mos	17 (8–60)	60 (36–109)
Sex (M/F)	28 / 8	26 / 18
Antenatal diagnosis of CAKUT, n (%)	12 (33%)	-
CAKUT phenotype, n (%)		-
UPJO, n (%)	18 (50%)	-
VUR, n (%)	12 (33%)	-
PUV, n (%)	6 (17%)	-
UTI risk group		-
Low, n (%)	19 (53%)	-
High, n (%)	17 (47%)	-

CAKUT, congenital anomalies of the kidney and urinary tract; F, female; IQR, interquartile range; M, male; PUV, posterior urethral valve; UPJO, uretero-pelvic junction obstruction; UTI, urinary tract infection; VUR, vesicoureteral reflux.

UTI-Risk States Have Distinct Urine Microbiomes With Respect to Controls and are Characterized by Significantly Low Alpha Diversity and Increased Intragroup Heterogeneity

Initial investigation of the alpha diversity measures across the different microbiome groups revealed reduced Shannon diversity, especially for the high-risk patient group, compared with the controls at the species level (Supplementary Figure S2A and B) (control vs. high-risk: marginally significant Mann-Whitney test P = 0.076; low-risk vs. high-risk: significant Mann-Whitney test P = 0.028) and the genus level (Supplementary Figure S2C and D; Supplementary Text S7). Thus, higher UTI risk was associated with decreased diversity of the urinary microbiome.

Overall, PERMANOVA analysis showed no association between microbial community composition and age across either control or at-risk groups using both Bray–Curtis and Kendall distance metrics. Sex was associated with community composition only in the control group using Bray–Curtis distances, whereas no sex-related associations were observed in the at-risk group or with Kendall-based distances. Apart from this single observation, no other associations were detected, indicating that age and sex do not broadly explain microbiome variation in this cohort (Supplementary Table S1B). Profiling intermicrobiome community variations between the controls and the at-risk groups using 2 different distance measures (Bray-Curtis and Kendall) revealed significant differences in the species (Supplementary Figure S3A–D) (control vs. at-risk: PERMANOVA with Kendall distance R-squared: 0.11 and P = 0.001; PERMANOVA with Bray-Curtis distance R-squared: 0.05 and P = 0.02), (control vs. low-risk vs. high-risk: PERMANOVA with Kendall distance R-squared: 0.18 and P = 0.003; PERMANOVA with Bray-Curtis distance R-squared: 0.08 and P = 0.013); and genus level (Supplementary Figure S3E–H; Supplementary Text S7). We observed a clear clustering of urinary microbiomes corresponding to the control, low-risk, and high-risk groups, with maximum difference between the high-risk and the controls and low-risk microbiomes occupying intermediate position. This indicates a progressive “shift” in microbiome states from controls to those with increased UTI risk.

After investigating the intergroup variations, we probed the intragroup variations of the microbiome profiles. The within-group community variations (measured in terms of both Kendall and Bray-Curtis distances), reflecting microbial dissimilarity among individuals, were significantly higher for those with increased UTI risk (especially the high-risk group) (Supplementary Figure S3I–P). This was observed in reasonable reproducibility for both the species- and genus-level taxa. In general, high-risk urine microbiomes were associated with the highest intragroup heterogeneity, followed by the low-risk group, and then control microbiomes. The above results indicate that increased risk of UTIs is not only associated with a loss of microbiome diversity (putatively indicating a loss of diversity-associated major taxonomic members) but also with increased heterogeneity, which indicates a potential loss of putative core keystone members of the urine microbiome in this cohort.

Risk-Associated Microbiome Signatures are Characterized by an Increasing Degree of Dysbiosis and Gradually Separate From the Controls Along Specific PCoAs

During our investigation of microbiome beta diversity, we observed that along certain PCoAs, the placement of microbiomes correlated with their risk status. For example, along the first PCoA axis obtained based on the Kendall distance measure (Kendall-PCoA-1) (Supplementary Figure S3D), we observed that urine microbiomes from the high risk group were placed at the left most extremes (with the most negative Kendall-PCoA-1 values), followed by the low risk microbiomes, and finally nondiseased control microbiomes (at the right-most extreme with the most positive Kendall-PCoA-1 values).

We additionally quantified the extent of microbiome dysbiosis using 2 established measures, previously used for gut microbiomes, namely dysbiosis-score and Kendall-uniqueness^23,36 (Figure 1). As expected, the high-risk group microbiomes were characterized by significantly higher dysbiosis-scores. Both dysbiosis-scores and the Kendall-uniqueness of the microbiomes were significantly negatively correlated with their placement along Kendall-PCoA-1 values (Figure 1a and b, Supplementary Table S2C). These results indicate that specific principal coordinates such as Kendall-PCoA-1, could be used as a continuous measure capturing both UTI risk and microbiome dysbiosis. We next investigated the microbiome variations along these PCoA axes to identify the specific microbes associated (both negatively and positively) with dysbiosis and increased UTI-risk.

Figure 1.

Quantification of microbiome dysbiosis using dysbiosis score and Kendall uniqueness. (a) Dysbiosis scores were significantly higher in the high-risk group microbiomes and showed a significant negative correlation with Kendall-PCoA-1 values. (b) While Kendall’s uniqueness did not significantly differ across groups, Kendall-PCoA-1 was significantly correlated with Kendall’s uniqueness. These findings suggest that Kendall-PCoA-1 captures both UTI-risk and microbiome dysbiosis, providing a continuous measure of dysbiosis severity, and thus Kendall PCoA-1 was used to show microbiome shift using the window-based approach (Supplementary Figure S4). Kendall-PCoA-1, the first PCoA axis obtained based on the Kendall distance measure.

Sliding-Window–Based Investigation Along the Kendall-PCoA-1 Axis Captures the Microbiome Compositional Changes Associated With Dysbiosis and Increased UTI Risk

Consequently, a sliding-window–based investigation was performed as an attempt to “track” the microbiome shifts occurring from the nondiseased (control) state to the state of high risk of UTI. For this purpose, we arranged the microbiomes in ascending order of their species-level Kendall-PCoA-1 values. Subsequently, we divided this ordered list of 80 microbiomes into 9 overlapping groups of microbiomes (named from windows 1 to 9), such that window 1 consisted predominantly of high-risk group microbiomes and window 9 was dominated by control microbiomes (see Methods; Supplementary Figure S1). Finally, we profiled and compared the abundance as well as prevalence of different species-level microbes across the 9 windows (Supplementary Figures S1B and S4).

As the risk association with UTI increased from window 9 to window 1, specific microbial sets showed either a gradual decrease or an increase. We observed a specific set of 33 putatively health-associated microbial taxa that showed a decrease. We did power calculations to further validate the robustness of their associations with Kendall-PCoA-1. Twenty of these 33 taxa were identified with power ≥ 0.90 and P ≤ 0.05 (Figure 1). This included microbes such as Varibaculum cambriense, Corynebacterium massiliense, Arcanobacterium canis, multiple species of Prevotella (P timonensis/buccalis/bergensis), Porphyromonas (P asaccharolytica/bennonis/uenonis/somerae), Actinomyces (A radingae/coleocanis) and Mobiluncus (M curtisii/mulieris). Majority of these microbes showed a significant negative association with dysbiosis score, indicating their strong associations with nondiseased controls, but nonsignificant associations with Kendall uniqueness and Shannon index (diversity) (Supplementary Figure S4; Supplementary Table S2D–G).

The set of 41 microbes that increased from window 9 to window 1, were the UTI-risk–associated ones and included several previously reported UTI-pathogens such as E coli, Proteus mirabilis, K oxytoca, K pneumoniae, and P aeruginosa.^37,38 We observed additional previously unreported microbes such as Propionibacterium acnes, B cereus, Pre melaninogenica, C renale, Shigella flexneri, Enterobacter mori, Acinetobacter baumannii, C jeikeium, B gelatini, B funiculus, and B luciferensis, which showed an increase with UTI risk status. These members could be the hitherto unreported UTI-risk–associated microbes (or pathogens) and require further investigation of their role as etiological agents of UTIs in children. Interestingly, a majority of this group of microbes showed strong positive associations with both Kendall uniqueness and Shannon diversity. In addition, we observed a strong concordance between the shifts in the detection (prevalence) and the community representation (mean ranked abundance) of these microbes across the 9 windows, indicating that the alteration patterns are robust (Supplementary Figure S4A and B).

Similar patterns of strong microbiome shift with UTI risk were observed at the genus level (Supplementary Text S7; Supplementary Figure S5; Supplementary Figure 3G and H; Supplementary Table S2H–K) as well as using a community-level network-based investigation of distinct coabundant taxonomic hubs in the urine microbial community (the ccrepe approach) (Supplementary Text S8; Supplementary Figure S6). Seventeen of the 23 genera that showed a decrease with increasing risk-status from window 9 to window 1 were identified with Kendall-PCoA-1 with P-value ≤ 0.05 and power ≥ 0.90 (Supplementary Figure S7), indicating robust association. Interestingly, for none of the taxa (both at the species- and genus-level), which increased as the risk-association increased from window 9 to window1, the correlation with Kendall-PCoA-1 crossed the statistical power threshold of 0.90. Thus, the risk-associated detrimental states are predominant with different UTI-pathogens that vary across the risk-associated states. In contrast, the normal or low-risk microbiome states are robustly associated with specific microbiome markers, indicating a strong health-associated signature. This underscores the shifts in microbial community composition across risk categories, with significant implications for understanding the microbiome’s role in UTI susceptibility and progression (Supplementary Figure S4B; Supplementary Table S2D–G). We next focused on identifying the putative functional or metabolic signatures of these microbial groups and validating whether (and if) any of these microbiome signatures at baseline correlates with future UTI-incidence using a follow-up investigation.

Longitudinal UTI Outcomes and Their Alignment With Baseline Microbiome Signatures

Building on the baseline microbiome patterns (Supplementary Figure S4; Supplementary Figure S5), we next examined whether these microbial signatures were associated with longitudinal UTI outcomes. Of the 36 children classified as low or high risk at baseline, 14 developed a clinically confirmed UTI during the median (interquartile range) follow-up (Supplementary Table S1C) of 564 (518–594) months. Kaplan–Meier analysis³² clearly showed significant separation in UTI-free survival between the 2 groups (log-rank test P-value = 0.0076), with high-risk children exhibiting an earlier and consistently higher incidence of UTI (Figure S2a), confirming that baseline clinical risk stratification predicted future UTI susceptibility.

To assess whether overall microbiome signatures at baseline related to these outcomes, we compared the baseline dysbiosis scores and Kendall-PCoA-1 between children who developed UTI and those who remained UTI-free. Although neither metric reached statistical significance (dysbiosis score: P = 0.20; Kendall-PCoA-1: P = 0.11), both trended toward greater deviation from control-like microbiome structure in the UTI group (Figure 2b and c). UTI episodes were disproportionately observed (48% vs. 11%) among children with dysbiosis score ≥ 0.95 (Figure 2d), most of whom belonged to the high-risk category (Supplementary Table S1C), suggesting an overlap between clinical risk, microbiome disruption, and subsequent infection.

Figure 2.

(a). Kaplan–Meier curves showing UTI incidence among low- and high-risk children (n = 36). High-risk children had significantly lower UTI-free probability (log-rank P = 0.0076). (b and d) Baseline dysbiosis scores were higher in children who later developed UTI (P = 0.20). UTI occurred more frequently in participants with dysbiosis ≥ 0.95, the majority of whom belonged to the high-risk group. (c) Kendall-PCoA-1 differed between UTI and UTI-free children, reflecting a shift toward higher-risk microbiome states, although not reaching statistical significance (P = 0.11). (e). Power analysis identified 20 high-confidence species (power ≥ 0.90), of which 10 were significantly or near-significantly enriched in controls and UTI-free children and depleted in those who developed UTI (P ≤ 0.05; false discovery rate ≤ 0.15). Corynebacterium massiliense showed the strongest negative association with UTI. (f) The cumulative rank abundance of these 10 species differed across groups, with lowest abundance in children who developed UTI (control vs UTI, P = 0.0006; UTI-free vs. UTI, P = 0.05). (g and h) Two species correlated with the number of UTI episodes: Corynebacterium massiliense (negative, P = 0.03) and Pseudomonas stutzeri (positive, P = 0.02). (i) Cox regression identified 14 species predictive of UTI during follow-up (P ≤ 0.05); 7 had infinite hazard ratios and 4 had hazard ratios > 1. Kendall-PCoA-1, the first PCoA axis obtained based on the Kendall distance measure; UTI, urinary tract infection.

We next investigated whether (and if so, which) microbiome taxa that varied across the entire risk landscape in our earlier analysis were significantly associated with UTI incidence and/or UTI-free survival during follow-up. We specifically focused on the subset of taxa (20 species-level taxa and 17 genus-level taxa) that showed association with Kendall-PCoA-1 (all of them negatively associated with risk states) with statistical power ≥ 0.90 (Supplementary Table S2l and m). Differential testing of these taxa revealed 10 out of the 20 species-level taxa and 12 out of the 17 genus-level taxa that were enriched at baseline in controls and at-risk subjects who did not develop UTIs as compared with those who developed UTIs (P ≤ 0.05; false discovery rate ≤ 0.15). Baseline C massiliense and the Alicyclobacillus exhibited the strongest negative association with future UTI-incidences at the species- and genus-level (Figure 2e and Supplementary Figure S7; Supplementary Table S2N and O), respectively. The cumulative ranked abundance of these 10 species-level taxa group produced a graded pattern across clinical groups, highest in controls, intermediate in at-risk UTI-free, and lowest in at-risk children who developed UTI with significant contrasts for control versus at-risk with UTI (P = 0.0006) and at-risk UTI-free versus at-risk with UTI (P = 0.05), and a borderline difference for control versus at-risk UTI-free (P = 0.07) (Figure 2f). A similar pattern was observed for the putatively health-associated genus-level 12 taxa consortia (significant contrasts for control vs. at-risk with UTI, P = 7.3 × 10⁻⁵; and at-risk UTI-free vs. at-risk with UTI, P = 5.6× 10⁻², and control vs. at-risk UTI-free, P = 0.03) (Supplementary Figure 7). These results clearly indicate that higher abundance of consortia of urine microbiome members is associated with UTI-free survival. These represent stable members, whose loss may predispose to infection. These results support a model in which loss of a set of potentially protective taxa is associated with increased infection vulnerability.

Spearman correlation with the number of UTI episodes identified baseline abundance of C massiliense to be negatively correlated with UTI episode counts (P = 0.03) (Figure 2h). Notably, these results align with the network analysis, where several Corynebacterium species clustered within the control-associated microbial hub (Supplementary Figures S4 and S6). This suggests that specific Corynebacterium species may act as stabilizing community members whose depletion compromises urinary tract resilience.

In contrast, the known UTI-pathogen Ps stutzeri was positively correlated with UTI episode counts (P = 0.02) (Figure 2g). Further, species-wise Cox proportional hazards models identified 14 species whose baseline abundance significantly predicted UTI risk during follow-up (P ≤ 0.05); 7 species had infinite hazard ratios, and 4 showed HR > 1 (Figure 2i). Notably, 4 Cox-associated taxa, C imitans, E coli, C amycolatum, and S flexneri were also flagged by the sliding-window compositional analysis (Figure 1), reinforcing their candidacy as robust microbiome-derived predictors of UTI susceptibility (Supplementary Table S2P). Together, the strong convergence of baseline microbiome signatures and the follow-up findings indicates that the urinary microbiome exhibits reproducible signatures of both health and UTI susceptibility. These signatures are not merely correlational markers of baseline risk but predict future infection trajectories, highlighting their potential utility as early microbial biomarkers for UTI surveillance in children.

Inferred Functional Profiling Identifies Putative Microbial Metabolic Functionalities With Disease Risk

We investigated the putative metabolic functionalities of the UTI-risk–associated and control-associated microbiome states. Given that the current dataset included 16S amplicon data, we computed the metabolic functional profile of the different urine microbiomes using a predictive metabolic approach^31,39 and subsequently correlated the metabolic functionality abundance profiles with the species-level Kendall-PCoA-1 values to identify specific functionalities associated with control and UTI-risk states (Supplementary Figure S8A).

A clear variation in the inferred metabolic functionalities that emerged from these investigations pertained to the metabolism of short-chain and branched-chain fatty acids (SCFAs and BCFAs). Although the production of multiple SCFAs or BCFAs (propanoate, butyrate, valerate, isovalerate, and isobutyrate) showed a significant positive association with Kendall-PCoA-1, the microbial consumption of these metabolites showed a negative association (indicating an increased abundance of microbes associated with SCFAs or BCFAs with increased UTI risk). Thus, the landscape of microbiome variations from control-associated states to high-risk states indicated a shift from SCFA- or BCFA-producing microbiome (thereby indicating their enrichment) to a SCFA- or BCFA-consuming microbiome (indicating depletion or reduced levels). In addition to the multiple beneficial roles of SCFAs, BCFAs such as valerate have been previously shown to prevent the growth of pathogens such as Clostridium difficile.³⁹

In a similar manner, we observed a positive association between the control state and the microbial production of phenylacetic acid, spermidine, and succinate (some of which have been associated with immune activation and antibacterial activities).⁴⁰ In contrast, the high-risk associated states were associated with the production of multiple detrimental and toxic metabolites such as ammonia, ethanol, and acetone as well as the production of putrescine and cadaverine (Supplementary Figure S8A). In addition, risk-associated microbiomes were associated with the production of putrescine and cadaverine, which have been shown to be elevated during pathogenic UTI infections.⁴¹ The risk-associated states were associated with increased consumption of multiple sugars and amino acids, whereas the control microbiomes were predicted to have higher levels of butanol or propanol consumption.

Thus, the microbiome-composition-predicted metabolic functional profiles captured many previously reported metabolite links associated with UTI-risk and identified likely metabolites to be involved in immune-system activation and antibacterial activities.

Antibiotic Resistance Profiles in UTI and Control-Associated Microbial Species

Because the current dataset was amplicon-based, we investigated the presence of resistance genes to different antibiotics in the previously reported genomes of these species using the CARD-RGI pipeline (See Methods; Supplementary Figure S8B; the World Health Organization Access, Watch, and Reserve – Supplementary Table S3). We considered 55 species that showed progressive shift in the sliding-window–based approach (Supplementary Figure S4B) and whose genome sequences were available in our reference genome collection (See Methods).

The CARD-RGI analysis revealed noticeable differences between UTI risk-associated and control-associated species. Among the 55 species, 31 were associated with UTI risk patients (identified based on the increase in species abundance and prevalence in the UTI risk patients; Supplementary Figure S4A and B), whereas 24 were linked to control patients. Four species associated with controls, Po uenonis, P asaccharolytica, A canis, and A haemolyticum, showed no presence of any antibiotic resistance gene. We observed marginally significant enrichment in the distribution of antibiotic resistance sites between UTI-risk–associated and control-associated microbes (P = 0.053; Supplementary Figure S9A).

Furthermore, we classified the antibiotics according to the World Health Organization Access, Watch, and Reserve classification (Supplementary Table S3). Of the 75 antibiotics, 51 were included in the AWaRe classification. The pathogenic microbial species isolated from the at-risk group have resistant genes to all 51 different antibiotics (access-15, watch-30, reserve-6). In contrast, the microbes isolated from the control group lacked resistant genes to many commonly prescribed antibiotics, namely penicillin, linezolid, nitrofurantoin, cefoxitin, tigecycline, trimethoprim, norfloxacin, piperacillin, amoxicillin, ceftazidime, ciprofloxacin, ampicillin, amikacin, meropenem, imipenem, etc. (Supplementary Table S3).

In addition, investigating the resistance to 22 antibiotics, frequently employed for antibiotic-susceptibility testing at our center (Supplementary Figure S8B, highlighted in red), we observed that, except for vancomycin, teicoplanin, erythromycin, and tetracycline, the resistance genes to the 18 other antibiotics were noticeably absent in the control-associated species but reasonably detected in the risk-associated species. This difference was statistically significant (P = 0.001) (Supplementary Figure S9B).

Discussion

The definition of a healthy or normal microbiome varies by body site, host age, and biological context. Although reduced microbial diversity is often interpreted as dysbiosis in the gut, this concept does not uniformly apply to the urinary tract. Several studies have shown that the urinary microbiome (urobiome) in adult women is a low-diversity protective urinary microbiome dominated by specific Lactobacilli, particularly L crispatus, associated lower risk of UTI.^6,36 However, this adult-derived paradigm may not be directly applicable to children.

These observations may not translate to pediatric urobiomes. Unlike adults, the pediatric urobiome is not sterile but dynamic, influenced by factors such as age, sex, toilet-training status, antibiotic exposure, and urologic history, suggesting that it represents a developing ecosystem rather than a stable, adult-like community.42, 43, 44, 45, 46 Notably, multiple studies on the pediatric urobiome have linked lower alpha diversity with UTI risk and increased abundance of uropathogens.^47,48 These observations highlight the importance of initiatives in studying the urinary pediatric urobiome.

In this regard, this study depicts a progressive loss of microbial diversity from controls to low-risk to high-risk groups and a concomitant rise in interindividual heterogeneity, suggesting that the more vulnerable urinary tracts harbor increasingly unstable communities. Overall, the microbiome alteration associated with UTI disease risk observed in this study align with the AKP as applied to microbiomes.¹³ The AKP refers to scenarios wherein all healthy microbiomes tend to be similar, with dysbiotic microbiomes disrupted in distinct ways, thus exhibiting greater variability among each other and in comparison with healthy microbiomes. AKP in microbiomes has been observed in approximately 50% of human diseases, including aging-associated dysbiosis.^23,49 We observed a similar pattern in this study. In addition, the follow-up analyses revealed that participants with greater baseline microbiome dysbiosis experienced disproportionately higher incidences of UTIs during the follow-up period.

The disrupted states are often characterized by the loss of key symbiotic organisms, conferring crucial benefits to the host, including resistance to pathogen colonization. Therefore, to understand microbiome-related diseases and develop microbiome-informed diagnostics or therapeutics, identification of these sets of health-associated core symbiotic microbes within different human-associated microbial communities is essential. Although several studies have focused on identifying such core microbes in the gut,¹⁴ our knowledge of such health-associated core microbes for other human symbiotic microbiomes, especially the urinary microbiome, remains limited.

Herein, by characterizing microbiome alterations along a normal–to–high-risk gradient and validating the identified signatures through follow-up investigations, we identified consortia of putatively health-associated microbes comprising 10 species-level taxa (led by C massiliense) and 12 genera (dominated by Alicyclobacillus genus). Not only were these consortia enriched in health-associated states, but higher baseline abundances were inversely associated with subsequent UTI incidence, indicating their diagnostic potentials for measuring urine microbiome health.

Although this list needs to be validated and refined using data from additional studies, many of these microbes have been reported in previous studies to be present in the urine microbiome of healthy individuals.^6,50, 51, 52, 53, 54, 55, 56 Preliminary investigations of the inferred metabolic profiles identified putative metabolic capabilities associated with the nondiseased microbiomes that were previously linked to immune activation and protection against pathogenic infections.39, 40, 41

In contrast, microbiome states linked to progressively increased UTI risk were characterized by gradually increased abundance of a number of UTI-associated pathogenic species (e.g., E coli, K spp., Ps aeruginosa, Pro mirabilis, Enterobacter spp., Ps stutzeri, A baumannii, Kocuria rhizophila).^37,38,57, 58, 59 Many of these microbiome markers were validated in our follow-up investigation, where higher baseline levels of these taxa were associated with increased future UTI risk. Overall, the convergence of microbiome signatures and longitudinal follow-up data suggests that urinary dysbiosis represents a measurable continuum of risk that can be detected before symptoms arise and may help refine microbiome-informed surveillance and prophylactic strategies.

Our urinary microbiome data mirror the findings of Hong et al., who demonstrated that preterm infants develop a sharp, pathogen-specific bloom of Enterobacteriaceae or Enterococcaceae in stool samples preceding culture-positive UTI, with relative abundances that distinguished future cases from controls with 79% to 92% sensitivity and 67% to 100% specificity. We extend this trajectory even, that is, before the onset of symptoms, children with CAKUT exhibit a parallel collapse of commensal hubs of taxa such as C massiliense, multiple species of Prevotella, Varibaculum, Porphyromonas, Alicyclibacillus, etc, and an enrichment of the pathogens, including P stutzeri, E coli, Klebsiella, Proteus, and Enterococcus, directly within the urine. Together, these 2 studies delineate a contiguous “gut–bladder axis” of risk in which pathogen-specific alterations first arise in the gut and can be subsequently detected in the urine before the first UTI episode. This convergence strengthens the biological plausibility of using microbial fingerprints, rather than anatomical markers alone, to anticipate breakthrough infections. Moreover, the shared early expansion of antimicrobial-resistant Enterobacteriaceae across gut and urine underscores the need for prophylactic strategies that preserve keystone anaerobes while avoiding broad-spectrum antibiotics that may further tip the balance toward dysbiosis.

This first known study of urinary microbiome evaluation from an Indian pediatric setting demonstrates that pathogen-specific shifts in microbial diversity and antibiotic-resistance profiles arise well before symptomatic UTI, opening a noninvasive window for early risk stratification, precision antimicrobial stewardship, and microbiome-targeted interventions in children with CAKUT. Our window and network analysis showed that the shift from control to high-risk group is driven by progressive loss of health-associated microbes and gain of high antimicrobial resistance–associated facultative pathogens. This finding opposes the practice of blanket, broad-spectrum antibiotic prophylaxis in all children with CAKUT prone to UTI. Only those children whose profiles already harbor multidrug-resistant uro-pathogens may warrant early metagenomics-guided therapy, whereas those dominated by commensals might be managed conservatively.

In addition, the loss of these control-associated hallmark microbes can be quantitatively used to devise dysbiosis scores and risk-indices that could identify children whose urinary communities are already shifting toward a high-risk configuration well before a febrile UTIs occurs. These variables could be incorporated into outpatient surveillance visits for early risk stratification, allowing pediatricians to intensify monitoring or initiate prophylaxis only in a very specific subset of patients. These findings could eventually aid in refining clinical decision-making, especially in children with VUR, where breakthrough UTIs often trigger an escalation to surgery. Furthermore, such risk indices could serve as an adjunct criterion for the identification of children who might benefit from earlier surgical procedures, potentially preventing renal scarring.

Moreover, the consistent depletion of anaerobic, SCFA- or BCFA-producing commensals across risk windows suggests a potential therapeutic target. Pilot microbiome restoration trials of bladder-instilled probiotics or orally administered prebiotics aimed at restoring specific taxa or their metabolites can be safely designed and timed using the risk-indices as an early efficacy measure, to pinpoint where along the spectrum each child sits. This can be used to develop personalized follow-up schedules: low-risk windows could justify longer intervals between visits, whereas high-risk windows would prompt closer urine monitoring or imaging.

The findings of this study must be interpreted within the context of certain limitations. First, this study includes only 1-time evaluation of urinary microbiome, and it challenges causal inference. Prospective longitudinal sampling is needed to validate and establish cause-and-effect relationships. Second, children were enrolled from 1 tertiary hospital, which limits generalizability and underpowers the subgroup analyses, for example, comparison of different congenital anomalies. Third, though we have used SPINGO classification that is typically customized for species-level classification of 16S amplicon datasets and has been used in multiple previous studies,^22,23,60,61 16S ribosomal RNA sequencing itself has limitations with respect to taxonomic resolution. Further, the sequencing of the V3-V4 region of the 16S ribosomal RNA genes has been associated with lower species-level taxonomic resolution in certain studies,^61,62 although this specific region has also been used in other urobiome studies.^47,63 Nevertheless, the sample-specific taxonomic profiles (especially at the species-level) need to be further validated using shotgun metagenomic and associated multi-omic studies. Fourth, the study did not account for important confounders, including diet, hygiene, catheter use, etc. Thus, though the identified microbial signatures are intriguing, they are currently at the stage of hypothesis generation. To exploit and use these findings clinically, future work should focus on larger cohorts with standardized risk stratification, longitudinal sampling, longer follow-up and correlation with long-term clinical outcomes. To address these, we are currently expanding our investigations to validate our initial findings on larger stratified cohorts. In the future, the predictive value of the dysbiosis score needs to be tested prospectively before any definite conclusions are drawn in this regard.

Conclusion

The findings of this study reveal that the composition of the urinary microbiome displays a gradient that strongly correlates with the UTI risk status of children with CAKUT. Controls harbor a microbiome typically enriched with a conserved health-associated microbial consortium, the depletion of which in at-risk children is accompanied by reduced diversity and enrichment of heterogeneous pathogen-dominated states, with these changes being more pronounced in high-risk than in low-risk children. We also observed distinct variations in metabolic capabilities and antibiotic resistance profiles across the microbes associated with different risk states. Importantly, the observation that microbial alterations preceded symptomatic infections in the children with CAKUT supports the concept that urinary microbiome monitoring can detect subtle risk transitions before the onset of clinical UTI. Collectively, these findings provide an initial framework for early UTI risk stratification, inform precision antimicrobial stewardship, and lay the groundwork for microbiome-targeted prevention in children with CAKUT prone to UTI.

Disclosure

All the authors declared no competing interests.

Supplementary Material

Supplementary File (PDF and xlsx)

Text S1. Detailed protocol for sample collection, processing, DNA extraction and transport (PDF). Text S2. Preprocessing and taxonomic classification of microbiome data (PDF). Text S3. Data normalization and diversity investigations (PDF). Text S4. Computation of inferred metabolite profiles (PDF). Text S5. Longitudinal follow-up investigation of the association of different risk-groups and baseline microbiome characteristics on future UTI incidences (PDF). Text S6. Identifying the Resistance gene in microbial species using CARD-RGI (PDF). Text S7. Investigation of microbiome variations with UTI risk at genus level (PDF). Text S8. Distinct coabundant microbial hubs predominate the urine microbiome and differentially associate with UTI-risk (PDF). Figure S1. Method adopted to divide the PCoA-1 into different overlapping windows and the composition of microbiome of different study groups in each of the window (PDF). Figure S2. Alpha diversity analysis with different study cohorts (PDF). Figure S3. Beta-diversity analysis between different study cohorts (PDF). Figure S4. Window-based approach to show microbiome shift across different risk groups at species level (PDF). Figure S5. Window-based approach to show microbiome shift across different risk groups at Genus level (PDF). Figure S6. Coabundance networks using the ccrepe approach to show the modules associated with different study groups (PDF). Figure S7. Abundances of the 12 genera (out of 17 with statistical power ≥ 0.90) (PDF). Figure S8. (A) Inferred metabolic functional profiling of urine microbiomes and their association with UTI-risk state; (B) Antibiotic resistance profiles in UTI and control-associated microbial species (PDF). Figure S9. Comparison of antibiotic resistance sites across control and UTI-risk–associated groups for different antibiotics (PDF). Table S1. Baseline and follow-up metadata of participants (xlsx). Table S2. Microbiome shifts and correlation analyses (xlsx). Table S3. The World Health Organization Access, Watch, and Reserve classification of Antibiotics (xlsx).