First-Catch, Mid-Stream and Catheterised urine: A Comparative Study of Male Urinary Microbiome by Expanded Quantitative Urine Culture and Next-Generation Sequencing


 BackgroundNumerous studies have emerged in the past decade investigating human urinary microbiota. Alterations in the microbial composition of urine have been linked to structural and functional abnormalities of the lower urinary tract. There has been considerable variation in the methodology of the studies published so far including the cornerstone of any biomedical analysis: sample collection. The aim of this study was to describe the urinary microbiota of first-catch voided urine (FCU), mid-stream voided urine (MSU) and aseptically catheterised urine in men and find the most suitable approach to urine sample collection for the purpose of male urinary microbiota investigations.ResultsForty-nine men (mean age 71.3 years) undergoing endoscopic procedures in our Department of Urology were enrolled in the study. Each of them contributed three samples: first-catch urine (FCU), mid-stream urine (MSU) and a catheterised urine sample. The samples were subjected to next-generation sequencing (NGS, n=35) and expanded quantitative urine culture (EQUC, n=31). Using NGS, Bacteroidetes, Firmicutes, and Proteobacteria were the most abundant phyla in our population. The most abundant genera (in order of relative abundance) included: Prevotella, Veillonella, Streptococcus, Porphyromonas, Campylobacter, Pseudomonas, Staphylococcus, Ezakiella, Escherichia and Dialister. Eighty-two of 105 samples were dominated by a single genus. FCU, MSU and catheterised urine samples differed significantly using ANOVA in three out of five alpha-diversity measures (p<0.05): estimated number of operational taxonomic units, Chao1 and abundance-based coverage estimators. There were no differences found in Simpson and Shannon indices. Beta-diversity comparisons using the PIME method (Prevalence Interval for Microbiome Evaluation) resulted in clear clustering of urine samples according to the mode of sampling.EQUC detected cultivable bacteria in 30/31 (97%) FCU and 27/31 (87%) MSU samples. Only 4/31 (13%) of catheterised urine samples showed bacterial growth.ConclusionsUrine samples obtained by transurethral catheterisation under aseptic conditions differ from spontaneously voided urine samples and represent a better reflection of urinary bladder microbiota. Catheterised urine is the most appropriate way to sample urine in future studies of urinary bladder pathological conditions and their relation to the urinary microbiota.


Background
The human urinary tract had traditionally been considered a sterile environment unlike other body niches such as the gut, oropharynx or vagina; hence, it was not included in the Human Microbiome Project [1].
However, the last decade has brought evidence of microbial communities residing in the female [2], [3] and male [4], [5] urinary tract. Various microorganisms have been cultured from urine using expanded quantitative urinary culture (EQUC) and much more prokaryotic diversity unearthed using next-generation sequencing (NGS) [6].
Female urinary microbiota has been given a broad attention; it was less so in case of male urinary microbiota (MUM). And while the microbiota of the female urinary bladder overlaps with that of the vagina to the point that a term "female urogenital microbiota" has been suggested [7], little is known about potential differences in the microbiota along the male urinary tract. The bacterial genera detected in rst-catch urine (FCU) are also present in urethral swab specimens; Lactobacillus, Sneathia, Veillonella, Corynebacterium and Prevotella were the most abundant ones according to a study from 2011 [4]. Anecdotic evidence has suggested certain differences between microbial compositions of voided and catheterised urine samples in males [8]- [10]. Differences in beta-(but not alpha-) diversity were reported in a study of urinary bladder cancer patients between voided and cystoscopy-obtained urine samples [8].
Another study reported that urethra and bladder microbiomes did not differ in their taxonomic composition but rather in taxonomic structure (relative abundances of several genera) [9]. And while the prevalence of bacteria in catheterised urine correlated with the degree of lower urinary tract symptoms in men with enlarged prostates, this relationship was not demonstrated on voided samples [10].
Alterations of the urinary microbiome have been associated with functional [11]- [13] and anatomical [10] abnormalities of the urinary tract and with the presence of genitourinary malignant disease [14], [15]. Urge urinary incontinence was associated with higher microbiota diversity [11], [12] and increased diversity also correlated with poor response to anticholinergic (for bladder overactivity) treatment [11]. Changes in the urinary microbiota were reported in patients after spinal cord injury [13] and even among patients with chronic kidney disease where lower diversity was associated with more advanced renal insu ciency [12]. Patients with bladder cancer were reported to have higher microbiota richness compared to healthy controls [14]. Chronic pelvic pain syndrome was associated with a greater alpha-diversity and greater prevalence of anaerobic bacteria than urine of controls without the condition [16]. Conversely, women with intersticial cystitis/painful bladder syndrome had less diverse microbiota than controls [17].
Despite studies looking for and nding differences in composition of the urinary microbiota in a variety of clinical scenarios, an elementary methodological question -i.e. how to collect urine samples to ensure an exact re ection of the urinary bladder microbiota and comparability of future studies -remains unresolved.
The aim of the present study was to describe male urinary microbiota of rst-catch voided urine (FCU), mid-stream voided urine (MSU) and aseptically catheterised urine in subjects without symptoms of a urinary tract infection (UTI) and with a negative standard urine culture result. Secondly, we aimed to identify the most suitable sample collection method for the characterization of bacterial communities residing in the male urinary tract.

Methods
This is a prospective observational study. Its objective is to investigate within-subject diversity of MUM obtained by three different methods of urine sampling and to de ne the most suitable approach to the study of the microbiota residing in the male urinary bladder.

Population
The study participants were recruited among patients undergoing endoscopic procedures for benign or malignant conditions of the urinary tract in the Department of Urology, 3rd Medical Faculty of Charles University and Thomayer Hospital, Prague, Czech Republic. To be included in the study, patients had to have a negative result of standard urine culture preoperatively, no foreign body in the urinary bladder (such as indwelling catheters, ureteric stents or bladder stones) and not have used antibiotic treatment for any medical condition in the past 6 weeks. The study was approved by the institutional review board (ethics committee) and informed consent was obtained from all all participants prior to enrollment.

Urine sampling
Participants were instructed on proper urine sample collection and were asked to provide an FCU and MSU specimens in two separate sterile containers in the morning before surgery. A third specimen was obtained in theatre at the beginning of the procedure after desinfection of the genital area, surgical draping and immediately upon endoscope insertion. A water-based jelly not containing chlorhexidine or any other desinfectant (Optilube, Optimum Medical Solutions, Leeds, UK) was used to minimize the chances of harm to any presumed living intravesical microorganisms. All samples were stored at 4°C before inoculation and processed on the same day with EQUC or frozen and stored at -20°C until DNA extraction.
Expanded quantitative urine culture Only subjects with "no growth" reported on routine preoperative culture were eligible for the EQUC part of the study. Patients with preoperative cultures deemed negative for clinical purposes but reported as "commensal ora" or "suspected contamination" were excluded. Antibiotic prophylaxis, where indicated, was administered after urine samples were obtained. EQUC protocol was based on previous description by Hilt et al. [6]. A 100 µL aliquot of each urine sample was inoculated with a sterile plastic loop onto agar plates (90 mm in diameter) and in liquid broth. EQUC culture conditions were as follows: 1. Columbia blood agar (CBA) incubated at 37°C for 48 hours; 2. CBA incubated at 30°C for 48 hours; 3. CBA incubated at 37°C in a 5% CO 2 incubator for 48 hours; 4. chocolate agar incubated at 37°C in a 5% CO 2 incubator for 48 hours; 5. and 6. Schaedler blood agar at 37°C in a Campy gas mixture (5 % O 2 , 10 % CO 2 and 85 % N) in Oxoid Anaerojar; 7. thioglycolate broth incubated at 37°C for 5 days, then inoculated on CBA and incubated at 37°C for another 48 hours. CBA and Schaedler agar were prepared in the laboratory from dried base (Bio-Rad, Berkeley, CA, USA). 5 % sheep blood (LMS -Labmediaservis, Czech Republic) was added into the medium at 38°C. Thioglycolate broth was prepared from dehydrated powder (Bio-Rad, Berkeley, CA, USA). Chocolate agar was obtained as commercial product (Bio-Rad, Berkeley, CA, USA).
The number of colonies was then counted on agar plates with inoculated urine. Growth was visually detected in the thioglycolate broth and colonies were counted on CBA after inoculation and incubation.
Matrix assisted laser desorption ionization-time of ight mass spectrometry (MALDI-TOF, Bruker Daltonik GmbH, Leipzig, Germany) was used for the identi cation of individual bacteria.

DNA extraction and PCR
Bacterial 16S rRNA gene was extracted from urine samples using Eligene Urine Isolation Kit (Elisabeth Pharmacon, Brno, Czech Republic) according to manufacturer's instructions. Brie y, urine was vortexed for 15 seconds, centrifuged at 6 000 x g for 20 minutes, the supernatant was discarded and 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 second vortexing, the mixture was incubated for 15 minutes at 65°C. The lysate was cetrifuged at 6 000 x g for 5 minutes. The supernatant was transferred to microtube and 210 µl of MI4 solution added. The lysate was centrifuged for 1 minute at 13,000 x g.
The primers 515F (5-GTGCCAGCMGCCGCGGTAA) and 806R (5-GGACTACHVGGGTWTCTAAT) were used to amplify the hypervariable region V4 of the 16S rRNA gene. Each forward primer was barcoded by a sequence nucleotides designed to multiplexing of different samples [18]. PCR was performed in triplicates, and every reaction contained 5µL of 5xQ5 Reaction Buffer for Q5 High-Fidelity DNA polymerase; 0.25µL Q5 High-Fidelity DNA polymerase; 5µL of 5xQ5 HighGC Enhancer; 1.5µLofBSA (10mg/mL); 0.5µL of PCR Nucleotide Mix (10 mM); 1µL of primer 515F (10µM); 1µL of primer 806R (10µM,); 1.0µL of template DNA and sterile ddH2O up to 25µL. Conditions for ampli cation 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 nished with a nal setting of 72°C for 10 min.

Statistical analyses
Demographic and clinical data were analysed as continuous or categorical variables and reported as median and inter-quartile range (IQR) or percentages as appropriate. Bacterial growth using EQUC was assessed as present or absent and correlation between the three samples from individual patients were evaluated. XLSTAT (Addinsoft, New York, USA) was used for statistical calculations.
The sequencing data were processed using SEED 2.1.2 [19]. Pair-end reads were merged using fastq-join [20]. 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 7.0.1090, and clustered into OTUs using the uparse agorithm [21] at a 97% similarity level. The most abundant sequence from each cluster [22] was assigned to the closest hits at the genus level using the RDP Naïve Bayesian Classi er Version 2.11 method [23] for bacteria. Sequences identi ed as non-bacterial were discarded. The DNA sequences have been deposited at the NCBI SRA under the accession number PRJNA744742.
Analyses of alpha-and beta-diversity were perfomed using the packages Vegan [24], phyloseq [25] and Prevalence Interval for Microbiome Evaluation (PIME) [26] from R language [27]. Details on clinical and demographic data of the study population are provided in Table 1. The phyla Bacteroidetes, Firmicutes, and Proteobacteria represented the majority of sequences in the three types of urine specimens with differing relative abundances. The 10 most abundant genera (in order of relative abundance) were: Prevotella, Veillonella, Streptococcus, Porphyromonas, Campylobacter, Pseudomonas, Staphylococcus, Ezakiella, Escherichia and Dialister representing around 530.000 sequences, approximately 72% of the total sequences obtained after removing errors and contaminants. Figure 1 depicts the relative abundances at genus level in all samples.
FCU, MSU and catheterised urine samples differed signi cantly using ANOVA in three out of ve alpha diversity measures used: estimated number of OTUs (p < 0.05), Chao1 (p < 0.05) and abundance-based coverage estimators (ACE) index (p < 0.05); there were no statistically signi cant differences in Simpson and Shannon indices ( Table 2 and Fig. 2). C. glucuronolyticum in one, E. faecalis in two and S. epidermidis in one patient. All of these four microorganisms were also present in the particular subject´s FCU and MSU samples.
In terms of concordance between FCU and MSU samples, both showed identical EQUC species-level result in 21 cases (68 %). In the remaining 10 patients, the results were discordant as follows: a positive rstcatch with a sterile mid-stream sample (n = 4); same number of detected isolates but different specieslevel composition (n = 3); and different number of isolates and their composition (n = 3). Supplementary table S1 (Additional le 1) shows a complete list of all subjects and their EQUC ndings.
A total of 18 patients had their triplets of samples subjected to both EQUC and NGS. Out of 77 EQUCdetected isolates in these 54 samples, 61 microorganisms (79 %) were also detectable by NGS.

Discussion
This is the only study to date dedicated speci cally to the investigation of the potential differences among FCU, MSU and catheterised urine in men. We proved that catheterised urine samples differ in microbiota composition from spontaneously voided urine. This nding has important consequences for future MUM studies.
Dong et al. reported on the microbial communities of FCU samples as early as 2011 [4]. The genera detected in his population of men from an STD clinic were also detected among our rst-catch samples and some of them also belonged to the most abundant OTUs: Prevotella, Streptococcus, Veillonella and Staphylococcus, to name the rst four. On the contrary, Sneathia spp. was not detected in our FCU samples at all. Different age of subjects in the Dong study (28 years), different DNA analysis method (pyrosequencing) or the variable region of the 16s rRNA gene (V1-V3) might all explain this difference but there is a substantial degree of concordance between our and Dong et al. data.
Bajic et al. were the rst to note that MSU differs from catheterised samples from the same individual in their study of male lower urinary tract symptoms and suggested that catheterisation was the most appropriate method for sampling the male bladder microbiota. In a study of lower urinary tract symptoms, men with benign prostate hyperplasia were assessed for symptom severity; increase in the symptom score was associated with higher odds of detectable bacteria in catheterised urine, although no speci c genera were associated with the degree of symptoms; no such association was observed when analyzing spontaneously voided urine [10]. Recently, Hourigan et al. reported signi cant differences in beta-diversity (but not alpha-diversity) measures between voided and catheterised urine (n = 14 men). The sample was too small to detect any particular OTU to be enriched in voided or catheterised samples, respectively [8]. Pohl et al. examined MSU and catheterised urine from 14 males of four different ethnic origins and reported differences in relative abunance of several OTUs between voided and catheterised urine [9].
In the present study, we detected a signi cant difference in three out of ve alpha-diversity measuresestimated number of observed genera, Chao1 and ACE indices -demonstrating that catheterised urine displays smaller degree of richness and less variability than voided urine. Beta-diversity comparisons using different stringency levels (removed singletons and doubletons, 100 most abundant OTUs among all samples and the 3 most abundant taxa per sample, respectively Fig. 3A, B and C) for ltering the contingency table under the Bray-Curtis [28] distance resulted in random patterns, no clear grouping between the three urine sampling methods and no statistic signi cance when comparing the distance using PERMANOVA. Clear patterns of microbial community dissimilarities appeared only after ltering the contingency table using the PIME [26] method (Fig. 3D) 4). Clear alpha-diversity differences and the beta-diversity pattern seen in the NMDS after the PIME prevalence ltering con rm previous observations that different sections of the lower urinary tract harbour a speci c microbial community. Urine obtained by catheterisation also displays a less variable community structure, with its samples more similar to each other when considering the signature OTUs identi ed through the Random forest models (Fig. 3D).
The topic of a "core" urinary microbiome is an intriguing one and unlike in females where Lactobacillus, Prevotella and Gardnerella [2][30] seem to be the most commonly represented genera, there is a paucity of data on the dominant OTUs in MUM. While a study on patients from an STD clinic did not nd clear evidence for a "core" microbiota in FCU samples [31], another study employed MSU samples of 11 healthy volunteers and suggested Streptococcus, Prevotella, Veillonella, Peptoniphilus, Campylobacter and Anaerococcus to represent the core healthy MUM [32]. The only work that tried to de ne core MUM based on catheterised urine of males reported Streptococcus to be the dominant genus in males [9]. According to our data, the most common dominant genera in catheterised urine specimens include Pseudomonas, Prevotella and Ferroglobus. While the frequent presence of Pseudomonas and Prevotella in urine seems biologically plausible, Ferroglobus spp. is an anaerobic, Fe 2+ -oxidizing archaeum isolated from a submarine hydrothermal system [33]. There seems to be a misalignment in the process of OTU identi cation here which illustrates one of the weak points of current urinary microbiota investigations: public databases are inadequate for this purpose as they lack urobiome-speci c genomes [34].
Although the urinary microbiome is likely to incur some, yet unde ned changes during lifetime, our data disprove the hypothesis postulated by one group that the genera Jonquetella, Parvimonas, Proteiniphilum and Saccharofermentans only dwell in bladders of people over 70 years old [35]. In our population, these OTUs were detected in 14 patients with a mean age of 66 years and some as young as 39.
The discrepancy in the yield of microorganisms from voided and catheterised urine seen in NGS extends to the study of MUM by culture. We demonstrated that voided urine of asymptomatic men with a negative standard urine culture harbours living bacteria that can be cultured using EQUC. By contrast, catheterised urine from their urinary bladders only rarely yielded a positive EQUC result. Six different sets of culture conditions were employed by our EQUC protocol. Suppmentary table S2 illustrates their respective e cacies.
Hilt et al. were the rst to introduce the concept of EQUC into the study of human urinary microbiome [6].
In their landmark study, 80 % of catheterised female urine specimens reported as "no growth" at 10 3 by the standard urine culture protocol yielded a positive result using EQUC. The most prevalent genera were Lactobacillus, Corynebacterium and Streptococcus. Another study of female urinary microbiota reported 89 % positive EQUC results from catheterised urine samples of 75 women not reporting signs and symptoms of UTI; most of bacterial species were not detected by standard urine culture. Median of 3 isolates per sample (IQR 1-5) were detected and Streptococcus, Lactobacillus and Corynebacterium (in order of prevalence) were the most prevalent genera [3].
Several explanations can be given for the discrepancy between the yield of bacteria from catheterised urine in our and the abovementioned studies. Anatomical differences in the lower urinary tract of men and women seem the most obvious [36]: a long, twice curved male urethra as opposed to a straight, short and wide urethra in females; one that opens in direct proximity to the vaginal introit and not far from the anus. The vicinity of vaginal environment would also explain Lactobacillus being one of top-three bladder-dwellers in women but not in men. Another explanation for a higher detection rate of microorganisms by EQUC in the Hilt et al. and Price et al. studies [6], [37] might be less stringent enrolment criteria (no growth at 10 3 CFU/mL and absence of UTI symptoms, respectively) compared to our study protocol. Lastly, minor modi cations of our EQUC technique might have in uenced the results.
In the only other work employing EQUC for the study of male urinary microbiome, Bajic et al. detected cultivable bacteria in 96 % of voided urine specimens, a gure strikingly similar to our detection rate; among catheterised urine samples, 29 % were positive in their study [10]. Because they did not perform standard urine culture, no comments on the pre-enrolment microbiological status of the study subjects can be made and, as noted earlier, all of our EQUC study participants preoperative urine samples were reported uniformly negative.
A subset of samples from 18 patients from our population were subjected to both EQUC and NGS in order to nd out whether 16S rRNA gene sequences of the microorganisms cultured by EQUC are detectable by NGS. DNA sequences from 61 of 77 (79 %) EQUC-detected microorganisms were demonstrable by NGS; for 16 isolates, NGS failed to detect their DNA. This is a recognised phenomenon reported in two other studies combining EQUC and NGS for the detection of urinary microbiota [6], [38]. The culture-positive microorganisms were not necessarily the most abundant ones; on the contrary, mean relative abundance of the culture-proven microbes was 9.4 % and varied widely from 0 % to 93.2 %.
When it comes to the limitations of the current study, the following points should be recognized. The taxonomy assignment in high-throughput sequencing amplicon studies rely on a very small segment of the marker gene, considering the whole 16S rRNA gene contains approximately 1600 base pairs; we have used fragments of around 290 bp, therefore the identi cation of the taxa found can be misleading, as evidenced in the case of likely misalignment of Ferroglobus sequences (see above). Multiple copies of the 16S rRNA marker gene can in uence the abundance of the taxa identi ed in the study and no reliable database to correct for this bias exists so far, therefore there might be orders of magnitude errors in the ranking of the genera present in a sample. Clustering sequences into OTUs has the limitation of grouping taxa that could behave differently and multiple copies of the marker gene could separate the same taxa into two different OTUs; moreover, two OTUs belonging to the same genera could grow differently in the same treatment, being different species or even different strains of the same genus. All of these biases become noise in the identi cation of patterns in the microbial composition of a sample. However, these aws are inherent to microbiota investigations in general and this study is no exception. It ought to be noted that subjects of the study were recruited from patients of a urology department and not from a healthy population. This should not be a major issue since the study focuses on intra-individual variations of the microbiota. Finally, some suggest that DNA extraction protocols that employ bead beating for cell lysis give a better representation of bacterial community structure [39]. Our method of DNA extraction relied on a commercial kit that did not employ a mechanical cell disruption procedure. It should be noted, though, that in the context of the existing microbiome literature, bead-beating is far from being standard practice.

Conclusion
In the present comparative study of male urinary microbiome, we demonstrated that urine samples obtained by transurethral catheterisation under aseptic conditions lead to different results on subsequent NGS analysis compared to spontaneously voided ( rst-catch and mid-stream) urine samples. Using EQUC we corroborated previous evidence from female urinary microbiota studies regarding the detection of living microorganisms in samples of urine that are classi ed as "no growth" on standard urine culture.
Catheterised urine is the most appropriate way to sample male bladder microbiota in future studies of urinary bladder pathological conditions and their relation to the urinary microbiota. Availability of data and material Some of the raw data generated and/or analysed during this study are included in this published article as supplementary les. Larger datasets (16S rDNA gene sequences) are available in the NCBI SRA database under the accession number PRJNA744742 (http://www.ncbi.nlm.nih.gov/bioproject/744742).

Competing interests
The authors declare that they have no competing interests.

Funding
Supported by MH CZ -DRO (Thomayer University Hospital -TUH, 00064190). The funding body had no role in the design of the study or interpretation of data.   Boxplots comparing alpha-diversity measures for rst-catch, mid-stream and catheterised urine. See also Table 2.  Figure 3C) and after ltering the most descriptive OTUs based on their prevalence using the PIME algorithm ( Figure 3D).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.