Short- and mid-term temporal variability of the human urinary microbiota: a prospective observational cohort study

TL;DR

This study examines how the bacteria in human urine change over time, finding that while some details vary, most of the bacterial community remains consistent.

Contribution

The study provides new insights into the stability of the urinary microbiota over time, which is important for interpreting microbiome data in health and disease.

Findings

Overlapping bacterial operational taxonomic units (OTUs) accounted for more than 60% of the relative abundance in paired urine samples.

Alpha diversity (Chao1 index) showed significant differences between paired samples, but beta diversity did not.

Neither age nor sex of participants significantly influenced the variation in urinary microbiota composition.

Abstract

Understanding the temporal variability of the microbiome is critical for translating associations of the microbiome with health and disease into clinical practice. The aim of this study is to assess the extent of temporal variability of the human urinary microbiota. A pair of urine samples were collected from study participants at 3–40-month interval. DNA was extracted and the bacterial V4 hypervariable region of the 16S rRNA gene was sequenced on the Illumina MiSeq platform. The alpha diversity of paired samples was analyzed using Chao1 and Shannon indices and PERMANOVA was used to test the factors influencing beta diversity.

A total of 63 participants (43 men and 20 women with a mean age of 63.0 and 57.1 years, respectively) were included in the final analysis. An average of 152 ± 128 bacterial operational taxonomic units (OTUs) were identified in each urine sample from the entire cohort. There was an average of 41 ± 32 overlapping OTUs in each sample pair, accounting for 66.3 ± 29.4% of the relative abundance. There was a clear correlation between the number of overlapping OTUs and the relative abundance covered. The difference in Chao1 index between paired samples was statistically significant; the difference in Shannon index was not. Beta diversity did not differ significantly within the paired samples. Neither age nor sex of the participants influenced the variation in community composition. With a longer interval between the collections, the relative abundance covered by the overlapping OTUs changed significantly but not the number of OTUs.

Our findings demonstrated that, while the relative abundance of dominant bacteria varied, repeated collections generally shared more than 60% of the bacterial community. Furthermore, we observed little variation in the alpha and beta diversity of the microbial community in human urine. These results help to understand the dynamics of human urinary microbiota and enable interpretation of future studies.

The online version contains supplementary material available at 10.1186/s12866-025-03915-7.

Figures

The correlation between the number of overlapping OTUs and the relative abundance they are covering. The more OTUs are shared between two samples of urine taken apart, the higher relative abundance they account for (Pearson’s R² = 0.43, *p* < 0.001)

Microbial community indices of alpha diversity for two collection timepoints. (A) Chao1 and (B) Shannon index after 1,000× rarefaction. Boxplots show median, upper, and lower quartile, highest and lowest values. The first and second samples from the same subject are connected by orange lines

Overlap of the bacterial community from the two collection timepoints, grouped by the sex of the participants and cancer status. (A, B) number of overlapping OTUs (e.g. taxa detected in both collections). (C, D) averaged sum of relative abundances of overlapping taxa from collections at two timepoints ([abundance_1_ + abundance_2_] / 2). Boxplots show median, upper and lower quartile, highest and lowest values, Wilcoxon test p value is displayed

Relationship between taxa representation and age of subjects. (A) the number of OTUs which were detected in both collections as a function of age. (B) averaged sum of overlapping taxa relative abundance from collections at two timepoints as a function of age. The curve represents the fitted linear model with 0.95 confidence interval. Some urine pairs overlapping taxa represented nearly 100% of the community in both collections. In some cases, the relative abundance of the overlapping taxa was as low as 25%

Relationship between overlapping bacterial community and time between collections. (A) number of OTUs which were detected in both collections as a function of time between collections. (B) averaged sum of overlapping taxa relative abundance as a function of time between collections (*p* < 0.05). The curve represents the fitted linear model with 0.95 confidence interval

Full text

Background

The existence of the human urinary microbiota (UM) is now considered a fact [1, 2]. Both the lower and upper urinary tract harbor diverse microbial communities, some members of which have been cultured using expanded quantitative urine culture, while others have been detected by next-generation sequencing of the 16S rRNA gene. There is evidence that, in addition to the bacteriome, the UM also includes viruses (the virome) [3]and fungi (the mycobiome) [4], about which very little is known to date.

As far as the niches of the human body are concerned that have been incorporated in the Human Microbiome Project [5], their microbial communities appear to be stable. In the gastrointestinal tract, the composition of the microbiome is generally consistent over time and even after antibiotic exposure or a temporary change in diet, the microbiota returns to its original state [6]. Not only the operational taxonomic units (OTUs) but also the metagenomic profile remain stable within a given individual [6]. Despite the exposure of the skin to various environmental factors, the stability of the community is maintained in the short and long term. A prolonged interval between sampling did not lead to an increase in the variation of the cutaneous microbiota within an individual [7]. In the oral cavity, the subgingival microbiota remains stable to a certain extent, although it varies from person to person and even between different sites of the oral cavity [8]. The vaginal microbiota undergoes major changes from birth to puberty to menopause, and these changes are driven by sex steroid hormones [9]. Vaginal microbial community composition is associated with menstruation and vaginal intercourse but, interestingly, not with oral probiotic use [9]. Pregnant women harbor a different and more stable vaginal microbiota than non-pregnant women [10].

Understanding the temporal variability/stability of the microbiota is crucial for translating associations of the microbiome with health and disease into clinical practice. However, until now it has been unclear how stable the UM is or how reliably a single sample reflects the microbial communities of a bladder over time. The present study aims to fill this knowledge gap and assess the degree of temporal variability of the human UM.

Methods

Population

The population of the present study is based on a cohort of subjects reported previously [11]. A total of 107 subjects were asked to provide a second urine sample for the assessment of their urinary microbiota stability. Those willing to re-attend provided a second sample of urine for the assessment of alpha and beta diversity. To expand the study cohort, members of the research staff (healthy volunteers) were asked to provide two samples of urine several months apart. Inclusion criteria for enrolment were as follows: age 18 years and older, no antibiotic treatment for any condition in the previous 6 weeks, no foreign body in the urinary tract (indwelling catheters or ureteric stents), no signs or symptoms of a urinary tract infection and no neurogenic lower urinary tract dysfunction. The study was conducted in accordance with the Declaration of Helsinki after previous approval by the Ethics Committee of the Institute for Clinical and Experimental Medicine and Thomayer Hospital with Multi-center Competence under the number G-19-01 and informed consent was obtained from all participants prior to enrolment. Participants´ enrollment into the study and sample collection was conducted between 27 May 2019 and 31 January 2023. The STORMS checklist for this study is attached as S1 Table [12].

Sample collection and handling

The first paired sample was collected either as a mid-stream, clean-catch voided specimen of urine (mid-stream urine, MSU) or obtained via transurethral catheterization under anesthesia before a surgical procedure. The second paired sample was uniformly collected as MSU. An aliquot of each sample was stored at 4 °C and processed within 24 h with standard urine culture; another aliquot was frozen at -20 °C on the same day for several months until DNA extraction and sequencing.

DNA extraction and sequencing

DNA extraction and 16S rRNA gene sequencing were performed as previously described [13] at the Institute of Microbiology of the Czech Academy of Sciences, Prague, Czech Republic. Briefly, DNA was extracted from urine samples using Eligene Urine Isolation Kit (Elisabeth Pharmacon, Ref. 90051-50) according to manufacturer’s instructions in a random order of subject groups to avoid batch effects. The whole amount of urine was vortexed for 15 s, 10 mL of urine was then centrifuged at 6,000× g for 20 min, the supernatant was discarded, and pellet resuspended in 200 µL of molecular grade water, 200 µL of MI3 solution, and 20 µL of Proteinase K was added. After 15 s vortexing, the mixture was incubated for 15 min at 65 °C. The lysate was centrifuged at 6,000× g for 5 min. The supernatant was transferred to microtube and 210 µL of MI4 solution added. The lysate was centrifuged for 1 min at 13,000× g. Due to a low microbial load of source samples, DNA extraction controls as well as negative controls for PCR reactions were included. Enrichment of microbial DNA was not performed. Yield and purity of extracted DNA were checked using NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific). Samples with even as low DNA concentration as 5 ng µL^− 1^ were included in the amplification step.

The primers 515F (5’- GTGCCAGCMGCCGCGGTAA) and 806R (5’- GGACTACHVGGGTWTCTAAT) [14] were used to amplify the hypervariable region V4 of the 16S rRNA gene. Each forward primer was barcoded by a custom sequence of nucleotides designed to multiplexing of different samples. PCR was performed in triplicates, and every reaction contained 5 µL of 5× Q5 Reaction Buffer for Q5 High-Fidelity DNA polymerase (New England Biolabs, Ref. B9027S); 0.25 µL Q5 High-Fidelity DNA polymerase (New England Biolabs, Ref. M0491L); 5 µL of 5× Q5 High GC Enhancer (New England Biolabs, Ref. M0491L); 1.5 µL of BSA (10 mg mL^− 1^, GeneOn, Ref. 209–005 W); 0.5 µL of PCR Nucleotide Mix (10 mM, Thermo Fisher Scientific, Ref. R0191); 1 µL of primer 515 F (10 µM); 1 µL of primer 806R (10 µM,); 1.0 µL of template DNA and sterile ddH_2_O up to 25 µL. Conditions for amplification started at 94 °C for 4 min followed by 25 cycles of 94 °C for 45 s, 50 °C for 60 s, 72 °C for 75 s and finished with a final setting of 72 °C for 10 min.

The successful PCR was confirmed using agarose gel electrophoresis. Three PCR reactions were pooled together in order to randomize PCR bias and to get higher DNA yield. Pooled samples were purified by MinElute PCR Purification Kit (Qiagen, Ref. 28004) and mixed in an equimolar amount according to the concentration measured on the Qubit 2.0 Fluorometer (Thermo Fisher Scientific). Sequencing libraries were prepared using the TruSeq DNA PCR-Free Kit (Illumina, Ref. 20015962) according to manufacturer’s instructions while following official ligation protocol for the kit used. Sequencing was performed on Illumina MiSeq in a 2 × 250 bases sequencing run. Raw fastq files were retrieved and used as input to the bioinformatic pipeline described in the following section.

Statistical analyses

Demographic and clinical data were analyzed as continuous or categorical variables and reported as mean and standard deviation (SD)/interquartile range (IQR) or counts with percentages as appropriate.

The sequencing data were processed using SEED 2.1.2 [15]. Pair-end reads were merged using fastq-join [16]. Sequences with ambiguous bases were omitted as well as sequences with average quality PHRED score < 30. The chimeric sequences were detected and removed using USEARCH 8.1.1861, and clustered into OTUs using the UPARSE algorithm [17] at a 97% similarity level. This step yielded an average of 15,107 ± 8,700 sequences per sample (SD, min = 129). To filter out non-bacterial sequences, the most abundant sequence from each OTU was assigned to the closest hit from the GenBank database by NCBI BLAST 2.10.1. Taxonomy was assigned using DECIPHER 2.30.0 [18] and threshold 40 with IDTAXA algorithm trained on SILVA SSU database r138 [19]. This step yielded an average of 11,524 ± 8,669 sequences per sample (SD, min = 19).

If one of the paired samples showed Goods sequence coverage < 85, the pair was excluded from analysis. Global singleton OTUs were removed before analysis of OTUs occurrence and relative abundance of individual OTUs was calculated by normalizing the abundance of each OTU within a sample so that the total abundance of all OTUs in that sample summed to 1. Alpha and beta diversity analyses were performed using the packages tidyverse 2.0.0 [20], vegan 2.6-4 [21], QsRutils 0.1.5 [22], GUniFrac 1.8 [23] and tidylog 1.0.2 [24] in R 4.3.1 [25]. For alpha diversity analyses, rarefaction was done following [26]: each sample was randomly subsampled to 1,000 sequences while excluding subjects with at least one sample below this threshold (Table S2). Subsampling was iterated 1,000 times and per-sample average summary statistics was calculated. The following indices were calculated [27]: Chao1 (reflecting species richness) and Shannon diversity index (reflecting species evenness). Normality was tested by Shapiro–Wilk test, differences among groups were tested by Student’s t-test, Kruskal-Wallis or Wilcoxon paired test as appropriate and by PERMANOVA based on the Bray–Curtis dissimilarity matrix. Temporal trends in the number of shared OTUs and their relative abundance were modelled by fitting a linear mixed model (estimated using REML and the nlminb optimizer) using the package nlme v3.1-164 [28] and results were reported using the package report v0.5.8 [29]. Standardized parameters of the model were obtained by fitting the model on a standardized version of the dataset. 95% Confidence Intervals (CIs) and p-values were computed using a Wald t-distribution approximation. The results were considered statistically significant at the level p < 0.05.

Results

After exclusion of samples with insufficient Good´s coverage (6 subjects), a total of 63 subjects and their paired samples of urine were included in the analysis (Table 1 and extended Table S2). These samples were provided by 43 men (mean age 63.0 ± standard deviation (SD) 15.2 years) and 20 women (mean age 57.1 ± 13.6 years). The time gap between the two collections ranged from three to 40 months (median 24.8, IQR 14.9 months).

Table 1. Table with the summary of demographic information of the study population and character of urine collections. The table depicts basic demographic and clinical data, interval between collections and type of collectionCollection interval (months)21.2 ± 10.7Sexn_females_ = 20n_males_ = 43Age61.1 ± 14.8BMI27.6 ± 4.6Height (cm)176.1 ± 8.5Weight (kg)86.4 ± 19.1Bladder cancer (n)28Kidney stones (n)10Ureteric stones (n)12Bladder stones (n)0First sample obtained via transurethral catheterization (n)26

Standard urine culture was negative at 10^5^ colony forming units/mL for all samples except one (no. 307) at the first sampling. The second sample was culture-negative in 61 subjects. Among the study participants, 28 were patients with non-muscle invasive bladder cancer (stage Ta-T1) presenting for a transurethral resection of the bladder tumor, 18 were patients admitted for elective surgery for a non-malignant condition (benign prostate hyperplasia or upper urinary tract stone disease) and 17 were volunteering members of staff who provided two urine samples at least three months apart.

In each urine sample, a mean of 152 ± 128 OTUs were identified in total bacterial community. In each pair of samples from an individual subject, 41 ± 32 (27% ± 21%) of OTUs were overlapping, i.e. detected in both collections. These OTUs present in both collections accounted for an average of 66.3% ± 29.4% (median 72.7%, IQR 49.7%) of the microbial community in urine samples in terms of relative abundance. There was a clear correlation between the number of overlapping OTUs and the relative abundance they were accounting for (Pearson’s R² = 0.43, p < 0.001) (Fig. 1).

Fig. 1. The correlation between the number of overlapping OTUs and the relative abundance they are covering. The more OTUs are shared between two samples of urine taken apart, the higher relative abundance they account for (Pearson’s R² = 0.43, p < 0.001)

Median value of Chao1 index was lower by 15.4 (IQR = 129.6) when comparing the second collection to the first one (Wilcoxon test, p < 0.05, Fig. 2A). Shannon index showed no significant difference in community evenness between individual collections (Fig. 2B). Comparison of groups with and without bladder cancer and groups with and without urinary stones did not show differences in alpha diversity between these health conditions. Similarly alpha diversity values did not differ when comparing MSU and catheterized samples during the first collection. When comparing males and females the only difference was shown for Chao1 estimate during the second collection (Table S3).

Fig. 2. Microbial community indices of alpha diversity for two collection timepoints. (A) Chao1 and (B) Shannon index after 1,000× rarefaction. Boxplots show median, upper, and lower quartile, highest and lowest values. The first and second samples from the same subject are connected by orange lines

In addition to alpha diversity metrics, the sum of bacterial relative abundance represented by overlapping taxa as well as their identity was evaluated in relation to biological sex. There was no difference in the number of overlapping OTUs and the relative abundance of overlapping community between males and females (Fig. 3A and C). Subjects with bladder cancer showed lower count of OTUs which overlapped in both collections than subjects without cancer (35 ± 34 and 45 ± 31, respectively, Wilcoxon test p < 0.05, Fig. 3B). The average relative abundance of overlapping taxa was not different between subjects with bladder cancer and patients without cancer (Wilcoxon test p > 0.05, Fig. 3D). Based on the design of our previous project [11], to which the present study is an extension, 26 of 63 (41%) urine pairs contained catheterized urine from the first collection (all second samples were MSU). When MSU was compared to MSU in both samples, there were 47 ± 32 overlapping OTUs; when catheterized urine from the first collection was compared to MSU from the second sampling, the proportion of overlapping OTUs decreased to 32 ± 31 OTUs (Wilcoxon test p = 0.01).

Fig. 3. Overlap of the bacterial community from the two collection timepoints, grouped by the sex of the participants and cancer status. (A, B) number of overlapping OTUs (e.g. taxa detected in both collections). (C, D) averaged sum of relative abundances of overlapping taxa from collections at two timepoints ([abundance_1_ + abundance_2_] / 2). Boxplots show median, upper and lower quartile, highest and lowest values, Wilcoxon test p value is displayed

Linear modelling of taxa counts and overlap of their relative abundance (response variables) revealed a non-significant trend towards lower overlap with increasing subject age (fixed effect, Fig. 4A and B). The number of overlapping OTUs remained stable regardless of the time interval between the first and second urine collections (Fig. 5A). Having said that, the mean relative abundance of overlapping taxa significantly decreased over time, as demonstrated by a linear model using the average relative abundance of overlapping OTUs from both collection points [(abundance_1_ + abundance_2_) / 2] as the response variable and collection interval as the fixed effect (p < 0.05; Fig. 5B). To account for potential confounding variables, a linear mixed model was fitted to predict mean relative abundance of the overlapping taxa with collection interval, age of subjects, sex and their cancer status as predictors. The model included subjects’ ID as a random effect (formula: mean abundance ~ months + age + sex + cancer, random = ~ 1 | id). The explanatory power of the model related to the fixed effects alone (marginal R²) was found to be 0.64. The model’s intercept, corresponding to months = 0, age = 0, sex = F, and karc = 0, was determined to be 99.78 (95% CI [71.22, 128.33], t(58) = 6.99, p < 0.001). Within this model, the effect of time between collections was found to be statistically significant and negative (beta = -0.66, 95% CI [-1.26, -0.05], t(58) = -2.16, p = 0.035; Std. beta = -0.31, 95% CI [-0.59, -0.02]). It means that the relative abundance of the overlapping taxa is decreasing by 0.66% for each additional month between collections. The indirect fixed effects of age, sex and cancer status were determined to be statistically non-significant for relative abundances of shared taxa. Similarly, when including collection method as another fixed effect in the linear mixed model, it was determined as non-significant.

To determine, if there is a strict clustering between groups of subjects with shorter and longer collection intervals (Fig. 5B), the average relative abundance was calculated for the group with 3–15 months collection interval and for the group with 15–40 months collection interval. The values were overlapping to some extent (81.7 ± 20.9% and 61.1 ± 21.4%, respectively). The linear model showed the presence of negative correlation which was not significant in contrast to the models which included subjects with full range of collection intervals (3–40 months).

Fig. 4. Relationship between taxa representation and age of subjects. (A) the number of OTUs which were detected in both collections as a function of age. (B) averaged sum of overlapping taxa relative abundance from collections at two timepoints as a function of age. The curve represents the fitted linear model with 0.95 confidence interval. Some urine pairs overlapping taxa represented nearly 100% of the community in both collections. In some cases, the relative abundance of the overlapping taxa was as low as 25%

Fig. 5. Relationship between overlapping bacterial community and time between collections. (A) number of OTUs which were detected in both collections as a function of time between collections. (B) averaged sum of overlapping taxa relative abundance as a function of time between collections (p < 0.05). The curve represents the fitted linear model with 0.95 confidence interval

In terms of beta diversity, the subjects’ identity was a stronger driver for community composition than the time between collections (R^2^ = 0.04 and 0.02, respectively, p < 0.05). When differences in beta diversity were tested for the influence of sex, subjects’ cancer status and collection method, each variable explained 2% of variability (R^2^ = 0.02 each, p < 0.05). There were certain OTUs frequently observed in both paired samples and in more than 50% of male and female subjects (Table 2).

Table 2. Taxa detected in both the first and second urine sample by frequency. Taxa detected in both the first and second urine sample with more than 50% prevalence in either sexGenus (Class)OTU IDpresence in males (%)presence in females (%) Staphylococcus (Bacilli) CL0001140 (93%)17 (85%)unclass. *Enterobacteriaceae (Gammaproteobacteria)*CL0000731 (72%)16 (80%) Cutibacterium (Actinobacteria) CL0002434 (79%)14 (70%) Finegoldia (Clostridia) CL0002129 (67%)14 (70%) Peptoniphilus (Clostridia) CL0003527 (63%)14 (70%) Corynebacterium (Actinobacteria) CL0004131 (72%)12 (60%)unclass. *Enterobacteriaceae (Gammaproteobacteria)*CL0000427 (63%)13 (65%) Streptococcus (Bacilli) CL0002028 (65%)0 (0%) Fusobacterium (Fusobacteriia) CL000760 (0%)13 (65%) Enhydrobacter (Gammaproteobacteria) CL0005026 (60%)0 (0%) Campylobacter (Campylobacteria) CL000130 (0%)12 (60%) Dialister (Negativicutes) CL000300 (0%)12 (60%) Gardnerella (Actinobacteria) CL000030 (0%)11 (55%) Bacteroidales (Bacteroidia) CL000060 (0%)11 (55%) Anaerococcus (Clostridia) CL0001523 (53%)0 (0%) Corynebacterium (Actinobacteria) CL0005422 (51%)0 (0%)

Discussion

Since the discovery of the human UM in 2011 [1, 2], it has been associated with various pathological conditions. Currently, there is not only a lack of description of normal UM, but also a lack of deeper understanding of UM dynamics. However, this knowledge is crucial to draw conclusions about the role of UM in the onset and development of disease. For example Isali et al. [30] recently pointed out that it is difficult to characterize causality in microbiome research of bladder cancer and even with known composition of microbial taxa, these results cannot be readily translated to therapeutic applications. This study aims to uncover the effects of time on the UM composition using repeated sampling of study participants. Furthermore, we want to characterize the overlap of the microbial community from the same subject over a period of several months.

Microbial communities of other human body niches are known to be stable over time [6–8]. For the urinary tract, there has been little evidence on this topic. The first endeavor investigating UM short-term stability on 14 subjects of both sexes and various ages using 16S rRNA gene sequencing found that the voided urinary microbiota composition remained stable on a short-term basis (several days apart) [31]. Eight women were observed longitudinally for three months with daily culture of mid-stream urine (MSU) and periurethral swab; UM was undergoing fluctuations associated with menstruation and sexual activity [9]. In a cohort of 10 women who provided mid-stream urine samples 2.5 years apart, high intraindividual variability was observed with only 29% (range 9–42%) of species overlapping from the first and second sample even though the sampling was performed uniformly in the third week of the menstrual cycle. Of note, the authors chose to use an expanded quantitative urine culture (EQUC) protocol instead of 16S rRNA gene sequencing to assess the microbial diversity [32]. In 13 pregnant women whose mid-stream urine samples were collected in the first and second trimester and subjected to 16S rRNA gene sequencing, the UM was stable with Lactobacillus being the dominant microorganism in the majority of women [33].

In the present study, we show that the number of bacterial taxa overlapping in two urine collections does not usually exceed 50 (Figs. 1, 3B, 4A and 5A) and that the relative abundance of these overlapping OTUs averages between 60% and 70% sometimes even reaching 90–100% (Figs. 4B and 5B). There was no significant difference between men and women in the number of shared OTUs or in the relative abundance represented by these OTUs (Fig. 3A and C). The bladder cancer status of subjects had a slight effect on the number of shared OTUs but did not influence their average abundance (Fig. 3B and D).

Using the Chao1 estimator, we have shown that the richness of the microbial community changes between collections, i.e. there is some variation in the number of resident bacteria in urine samples taken months apart (Fig. 2A). Although Chao1 represents the estimated total count of taxa and calculated difference was only minor. The relative abundance of overlapping OTUs may vary considerably (Fig. 2B) due to the interplay of subject’s health, diet, genetic predispositions and interactions within the microbial community. The phenomenon of abundance variability becomes more significant as the time interval between collections increases (Fig. 5B). In contrast, with longer interval between collections, we do not observe drop in counts of overlapping taxa. Overall, this suggests that while the total richness of microbial community remains constant, the relative abundances of taxa with long-term presence in UM is slowly decreasing. This is probably caused by an ongoing and limited drift in proportional ratios of individual microbial taxa. Interestingly, age of participants, their sex and cancer status did not influence statistics related to overlapping taxa in terms of their count or relative abundance over time. While bladder cancer patients UM have been shown to be different in structure compared to cancer-negative subjects [10], in the current study we are analysing temporal changes of UM composition and diversity within individual participants. Our findings indicate that the temporal succession of UM proceeds at comparable rates in both cancer patients and healthy subjects, suggesting that the presence of bladder cancer does not significantly alter the trajectory of microbiota changes over time. In other words, the temporal variability and stability patterns of UM remain similar regardless of cancer status, indicating that a bladder cancer diagnosis does not fundamentally alter the dynamic behavior of the urinary microbiome. Similarly, neither age nor sex appeared to influence the temporal dynamics of UM, indicating that these demographic factors do not fundamentally affect how the urinary microbiome changes over time.

Missing relationship between the average relative abundance of overlapping taxa and the time between collections for the subset of subjects with collection intervals greater than 15 months indicates that the shorter collection intervals of 3–15 months are primarily driving the overall effect. Notably, there is an overlap in relative abundance values between the two subject clusters and therefore we conclude that, although the subjects tend to cluster, the overlap in values between these two groups suggests a continuum of microbial abundance rather than a strict separation between short-term and long-term collections.

At the community composition level (beta diversity), greater variability was explained by the subject identity than by the effect of collection time. Therefore, the differences in community composition within a subject were smaller than the differences between individuals. As reported previously, cancer status explained part of variability in community composition [11], and its explanatory power was comparable to explanatory power of subjects’ sex and collection method [11]. The comparison of MSU samples from both collections revealed a higher number of overlapping OTUs than in the case of the comparison between catheterized urine from the first collection and MSU from the second sampling. This observation is consistent with previous research reporting microbial communities in catheteri [13]. The significant reduction in overlapping OTUs when comparing catheterized urine to MSU may reflect differences in the microbiota associated with the collection technique, which could have implications for understanding urinary microbiome dynamics in clinical settings. Despite the difference in collection techniques in the case of some patients, we observed stable UM across the subjects thus concluding that the collection technique effect was limited in terms of the whole cohort.

We have noted that certain OTUs were observed in both paired samples in more than 50% of subjects (Table 2). These might represent candidate OTUs in the quest for a core urinary microbiota. Whether such phenomenon exists is still a matter of debate but of the taxa listed in Table 2, the following have been consistently reported in previous urinary microbiota studies in both males and females (in alphabetical order): Corynebacterium,* Finegoldia*,* Gardnerella*,* Peptoniphilus*,* Staphylococcus* and Streptococcus [1, 34–38]. Aerobic subgroup of these taxa (like Corynebacterium, Staphylococcus and Streptococcus) is also present in a recent genome collection of bladder-specific isolates and thus their metabolic potential might be assessed [39]. Other taxa like Peptoniphilus are more elusive due to their anaerobic growth. The fact that Lactobacillus does not feature among the top OTUs in our dataset may be explained by the postmenopausal status of the study female participants [40]; indeed, Lactobacillus was detected in both paired samples in only 40% of women.

We analyzed the dynamics of the microbial community in individual sample pairs (Fig. S1) to explain very low community overlap in some cases. In some of the pairs with the lowest overlap, a particular OTU was present with a high relative abundance (> 50%) in one sample and absent in the other. These OTUs were identified as (in alphabetical order) Arcanobacterium,* Enterobacter/Morganella*,* Idiomarina*,* Lactobacillus*,* Pseudomonas and Ureaplasma*.

Some methodological weaknesses of our study deserve attention: 1) Urine pairs collected at shorter time intervals between collections were from younger subjects, while paired samples from older participants were separated by a longer time gap. This may have led to a systematic bias favoring the statistical significance of the trend seen in Fig. 5B. 2) While none of the study subjects had taken antibiotics six weeks prior to the first urine collection, the absence of antibiotic treatment was less rigorously documented at the second urine sampling. Therefore, we cannot rule out the possibility that a shift in the richness or diversity of the UM was due to undocumented antibiotic exposure; 3) patients who harbored urinary stones at first sampling may have been stone-free at second sampling; 4) some of the first urine samples from each pair were collected via transurethral catheterization rather than spontaneously voided mid-stream samples; 5) data on sexual activity and menstrual cycle which may also affect the bacterial community composition were not collected; 6) all of the study participants were Caucasian; hence the results may not be generalizable to other geographical areas. While these shortcomings weaken the statistical significance of some of our results, they strengthen the claim that UM undergoes only limited change in time.

Conclusions

This study represents a comprehensive investigation dedicated specifically to the subject of urinary microbiome (UM) temporal stability. Our dataset (n = 63) consisted of a relatively large group of subjects of one ethnic origin but varying ages and both sexes. Additionally, samples from individual participants were separated by different lengths of time. This approach allowed us to assess the effects of three variables on the stability of human UM. The key finding of this study shows that there is a stable proportion of the taxa present in repeated urine samples, even though their relative abundance may fluctuate.

No significant intraindividual differences in beta diversity were observed between the two urine samples. This result demonstrates the extent to which the composition of the urinary microbiome remains stable over time, thus allowing for a better understanding of its role in urinary tract health and potential implications for diagnostic and therapeutic approaches in urology.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Supplementary Material 2

Supplementary Material 3

Supplementary Material 4