Urinary tract infection (UTI) is one of the most common infections worldwide, especially in women. This is thought to be due to women’s physiological structure and hormonal status. UTIs have high incidence rates and cause an economic burden (Flores-Mireles et al., 2015; Klein and Hultgren, 2020). According to a study from the United States, over 60% of women experienced at least one UTI during their lifetime, and 20–30% of them experienced recurrent UTI (RUTI) within the next 6 months (Foxman et al., 2000). RUTI is defined as the occurrence of ≥ 2 symptomatic episodes of UTI within 6 months or ≥ 3 episodes within 12 months (Kranz et al., 2018). RUTI significantly reduces the quality of life and leads to more medical consultations per year. It is highly prevalent, costly, and burdensome among women of all ages and races (Wagenlehner et al., 2018).

Microbial confirmation by urine culture is important in establishing the diagnosis of RUTI, and further antibiotic susceptibility testing provides tailoring of therapy (Anger et al., 2019). Patients with RUTI are frequently exposed to previous antibiotic treatment. Their bacterial concentration may be low in urine due to the antibiotic treatment. The bacteria of low concentration or slow-growing uropathogens may be missed by standard urine culture (SUC). The negative urine culture results have posed a challenge for antimicrobial therapy.

Bacterial 16S rRNA amplicon sequenced by the high-throughput next-generation sequencing (NGS) technology demonstrates that urine is not sterile, and contains diverse microbiota in both men and women (Nelson et al., 2010; Dong et al., 2011; Fouts et al., 2012; Wolfe et al., 2012). NGS is more sensitive than SUC. McDonald et al. (2017) showed that 100% of symptomatic patients with UTI had NGS positive test results for urinary bacteria, compared with a positive rate of 30% in SUC. Burnett et al. (2021) found that expanded quantitative urine culture (EQUC) could detect more urinary microbes than SUC, and their urobiome compositions were associated with distinct clinical profiles in patients with RUTI. Hochstedler et al. (2022) showed that EQUC detected more uropathogens than SUC from catheterized and voided urine in patients with RUTI. Urinary microbiome detected by NGS in health and disease has been characterized in recent years (Lewis et al., 2013; Pearce et al., 2014; Pohl et al., 2020). The studies on urinary microbiota of women patients with RUTI by NGS are still limited. However, it is a rapid growing field. A consensus to advance urobiome research was formed recently (Brubaker et al., 2021).

Peking University First Hospital is the 1,800-bed tertiary general hospital of Peking University, Beijing, China. Both nephrology and urology departments are the top 1 key discipline in China. Patients with RUTI with negative SUC were usually found in outpatient (Monsen and Ryden, 2017). We have assessed a small group of 10 patients with RUTI with a positive SUC. All of them were a pure culture of single uropathogen ≥ 10⁴ CFU/ml, which was a common circumstance in positive urine culture. The results are in good agreement between urine culture and NGS. Therefore, physicians can treat these patients based on urine culture and antibiotic susceptibility results. NGS does not show an advantage in these patients considering its high cost and complex procedure. The real clinical challenge is the patients with RUTI with negative urine culture, which could show the advantage of NGS. Thus, in this study, we only compared the urinary microbiota between patients with RUTI with negative SUC and age-matched asymptomatic controls.

The demographic characteristics of patients with RUTI and age-matched asymptomatic controls are summarized in Table 1. No significant differences were observed in age, NGS-positive (with successful sequencing result) rate, diabetes, and menopause status between the 2 groups (P > 0.05). The 16S rRNA gene sequencing detected urinary bacteria in 74.4% (67/90) of patients with RUTI and in 71.0% (44/62) asymptomatic controls. A total of 67 patients with RUTI and 44 asymptomatic controls with successful 16S rRNA gene sequencing results were processed for microbiota analysis.

The high-quality clean tags were clustered into a total of 4,390 OTUs. Alpha diversity analysis is shown in Figure 1. Chao1 index and observed species indexes were significantly different between the 2 groups (P < 0.05). The two indexes reflect the richness of species, without considering the relative abundance of each species. Thus, they are easier to show the existence of low abundant species. Shannon and PD whole tree indexes reflect the diversity of species, which are affected by the species’ evenness and richness.

The clustering histogram of an individual sample at the genus level is shown in Figure 2, which showed the communities in each sample. The relative abundance between RUTI group and control group at phylum, family, and genus levels is illustrated in Figures 3A–C. The top 3 dominant phyla were as follows: Firmicutes (relative abundance: 34.8% in RUTI vs. 65.4% in control), Proteobacteria (32.8% in RUTI vs. 8.02% in control), and Bacteroidetes (19.1% in RUTI vs. 12.8% in control). The top 3 dominant genera were as follows: Lactobacillus (24.0% in RUTI vs. 46.3% in control), Ralstonia (23.5% in RUTI vs. 0.242% in control), and Prevotella (14.4% in RUTI vs. 6.82% in control).

In the beta diversity analysis, inter-group difference evaluated by PLS-DA is shown in Figure 4. There was a significant divergence between RUTI and control groups (ANOSIM, R = 0.209, P = 0.001), indicating the difference of inter-group was larger than intra-group.

In the multivariate statistical analysis, LDA (LEfSe) was used to analyze the differentially abundant taxa as taxonomic biomarkers of RUTI. The relative abundance of 36 bacterial taxa was significantly higher, and another 24 bacterial taxa were significantly lower in RUTI compared with control (LEfSe analysis, P < 0.05, LDA score > 3). Specifically, Proteobacteria, Burkholderiales, Ralstonia, Prevotella, Dialister, and Corynebacterium were more abundant in RUTI, while Lactobacillus, Gardnerella, Viridans Streptococci, and Ezakiella were significantly reduced in the RUTI group (Figure 5). Cladogram displayed differentially abundant taxonomic clades with an LDA score of >3.0 between RUTI (red) and control (green) groups (Figure 6), which was consistent with the LEfSe result.

The previous study has showed that 94.1% of patients with clinically diagnosed pyelonephritis were treated with antibiotics empirically, but only 55.0% of them had positive urine culture; thus, the diagnostic uncertainty led to inappropriate antibiotic usage (Shallcross et al., 2018). The SUC technique was originally designed in 1956 with the threshold of urinary bacterial count ≥ 10⁵ CFU/ml (Kass, 1956, 1957). However, concerns about the suitability of this threshold for different kinds of UTI (e.g., urethritis, cystitis, and pyelonephritis) were raised (Stamm et al., 1982; Sathiananthamoorthy et al., 2019). Naboka et al. (2020) showed Enterobacteriaceae from the urine of women patients with RUTI at 10² CFU/ml had a wide spectrum of virulence factor genes, which may play a role in RUTI. However, the bacteria of 10² CFU/ml could not be detected by SUC. More sensitive methods are required. Current standards for UTI diagnosis had limitations and reduced the opportunity to improve patient care. Thus, new methods and technologies combined with rigorous clinical research could be used to correct the limitations (Price et al., 2018). NGS for bacterial 16S rRNA gene is more sensitive and can detect more atypical bacteria, anaerobes, or mixed pathogens causing UTI compared to urine culture (McDonald et al., 2017). NGS was applied in the diagnosis of chronic or persistent lower urinary tract symptoms (LUTS) and indicated that urinary microbiota and dysbiosis were possibly involved in LUTS. It may affect patients’ clinical outcomes in a variety of clinical scenarios (Gasiorek et al., 2019).

Since there was no standard protocol for the collection and extraction of bacterial DNA from urine, we followed a previous study (Gasiorek et al., 2019) with a larger volume of urine for DNA extraction to get a better detection rate. About 75% of patients with RUTI and 71% of asymptomatic controls were successfully amplified and sequenced the 16S rRNA gene. Other unsuccessfully amplified samples were repeated for a second round. Those still unsuccessful amplified samples after the second round were considered to be non-detected/negative. The previous study has showed that the detection rates of NGS in urgency urinary incontinence (UUI) and non-UUI urine samples were 63.9 and 65.8%, respectively (Pearce et al., 2014), which was lower than our study. Pearce et al. (2014) sequenced 1 ml of catheterized urine. Our study sequenced 30 ml of voided urine. Thus, our study detected bacteria in a greater percentage of samples. Another study found that bacterial DNA was detected in 64.9% of urine samples by 16S rRNA gene sequencing, while about 90% of them were negative by SUC (Hilt et al., 2014; Pearce et al., 2014). Our results confirmed their findings that SUC may miss some low biomass or slow-growing bacteria in patients with RUTI, which could be detected by 16S rRNA gene sequencing.

The alpha diversity index of chao1 and observed species showed that the RUTI group was less diverse than asymptomatic controls, possibly due to previous antibiotic usage, leading to reduced microbiota diversity or dysbiosis. The antibiotic could affect the gut microbiota, leading to loss of taxonomic and functional diversity (Lange et al., 2016). Another possibility could be antibiotics came after the first dysbiosis and symptomatic episode, and then antibiotics exacerbated dysbiosis. Beta diversity of PLS-DA analysis showed that the individuals in the RUTI and control groups were mainly in two different regions, as shown in Figure 4 (ANIMORM, P < 0.05). The scattered subjects may be due to individual difference, as hormonal status, body mass index, and certain clinical conditions may to some extent affect women’s urinary microbiota (Mueller et al., 2017).

Yoo et al. (2021) compared the microbiota of women with acute uncomplicated cystitis (AUC) and recurrent cystitis (RC). They found that most AUCs were caused by a specific pathogen. However, various species other than Enterobacteriaceae were detected in RC (Yoo et al., 2021). Our study showed similar results, suggesting the role of microbiota changes in RUTI. AUC could be detected by SUC, while the circumstances of RUTI/RC were more complex and should be considered to be dysbiosis. The study of RUTI women’s urogenital microbiota is urgently required (Josephs-Spaulding et al., 2021). Thus, our study was performed to compare the women’s urinary microbiota composition with and without RUTI.

Lower thresholds (10²–10³ CFU/ml) were set for suprapubic puncture urine or transurethral catheterized urine in the guideline of UTI laboratory diagnosis (China Health Professional Standards, 2016). The indications of these invasive procedures include urinary retention and surgery, instead of RUTI. The benefit-to-risk ratio and clinical feasibility of transurethral catheterization are low for patients with RUTI who are always from outpatient/clinic. Thus, they are not pragmatic in many clinical settings, and clean-catch midstream urine is still commonly used in China with a threshold of 10⁴ or 10⁵ CFU/ml (China Health Professional Standards, 2016). Some clinically diagnosed patients with RUTI often got negative urine culture results, which hampered their antimicrobial treatment. Urine samples collected by invasive methods (e.g., suprapubic puncture, and catheterization) are considered to reflect the real circumstance of the bladder, while voided urine may be contaminated by urethral orifice flora. Therefore, the two kinds of urine sample collection may cause a potential difference in NGS sequencing. In this study, we compared the microbiota between RUTI and their age-matched asymptomatic controls that provided baseline levels of normal urinary microbiota. Our results provided the differentiation between RUTI dysbiosis and normal urinary microbiota, which was a real-world study considering the possible contamination of midstream urine. The results had a wider scope of application and universal significance when suprapubic puncture urine or transurethral catheterized urine was unavailable, which was common circumstance in outpatients/clinics in China.

From our list of highly abundant bacteria in RUTI compared with asymptomatic controls, Corynebacterium, anaerobes (e.g., Prevotella, Fusobacterium, and Anaerococcus) have been documented to be involved in UTI (Kline and Lewis, 2016; Boyanova et al., 2022). Ralstonia was detected in the RUTI group with high relative abundance (Figure 5). It was undetected in controls. Ralstonia has been reported to cause a variety of hospital infections, namely, bloodstream infection, osteomyelitis, and meningitis (Ryan and Adley, 2014), and its pathogenic role in RUTI has not been reported yet. Thus, further studies are still needed to verify the pathogenic role of Ralstonia in RUTI. Lactobacillus and Viridans Streptococci are the main components of vaginal microbiota. They were found to be more abundant in asymptomatic controls. These bacteria are possibly commensal and non-pathogenic in RUTI. The previous study (Stapleton, 2016) has also showed the protective role of Lactobacillus against UTI. Gardnerella has been detected in urinary microbiota previously (Wolfe et al., 2012; Pearce et al., 2014). Our results showed that Gardnerella was highly abundant in the control group compared to RUTI (Figure 5). A recent study showed that Gardnerella could often be detected in high abundance from midstream urine of asymptomatic European women (Ksiezarek et al., 2021). Gardnerella was more frequently detected in UUI microbiota compared with non-UUI (Wolfe et al., 2012; Pearce et al., 2014). Animal study showed that Gardnerella in the mouse bladder could activate host pathways necessary for uropathogenic Escherichia coli causing RUTI (O’Brien et al., 2021).

Our study had several limitations. First, although 16S rRNA gene sequencing has been used to reveal urinary microbiota in recent studies, it could only identify bacteria well at the genus level and could not detect viruses or fungi. Thus, our results need to be verified by mNGS in future studies. Second, sexual behavior and sexually transmitted infections/diseases could affect urinary microbiota, and we did not have enough information of included patients. Thus, potential confounders could have affected our findings. Third, this is a single-center study located in Beijing, China, and geographical location may affect microbiota to some extent. Thus, future studies including multiple geographic regions are necessary.

In conclusion, our study reported the pathogen list of RUTI compared with asymptomatic controls, which could guide the hard-to-treat patients with RUTI with negative cultures. The urinary microbiota of women with RUTI with negative SUC was different from age-matched asymptomatic controls, suggesting that the differential urinary microbiota may play an important role in RUTI.

The data presented in the study are deposited in the repository of National Genomics Data Center of China (http://ngdc.cncb.ac.cn), with the accession number of PRJCA008664.

The studies involving human participants were reviewed and approved by the Ethics Committee of Peking University First Hospital, approval number: 2021-191. The patients/participants provided their written informed consent to participate in this study.

LH and XL designed the study and wrote the manuscript. XL and BZ enrolled the participant. LH, DW, and CH collected the data and performed the experiments. PL, LS, and HL analyzed the data. All authors read and approved the final manuscript.

PL was employed by the Beijing Yitong Qijun Technology Co., Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.