The medical records of 33 patients with suspected UTI admitted to the Second Hospital of Tianjin Medical University (Tianjin, China) from March 2021 to July 2022 who underwent urine mNGS testing were retrospectively analyzed. These patients had negative urine culture results prior to mNGS detection, but they had multiple abnormal urinalysis results and persistent urinary tract symptoms and were still suspected of having a UTI.

Most patients underwent the detection when they had poor responses to empirical antibiotic treatment, and a few patients underwent the detection directly without receiving empirical anti-infective treatment. All of them underwent a concomitant urine culture for comparison with mNGS in the diagnosis of infection and determination of the causative organism. Urinalysis using urine sediment analyzer and voiding symptoms were evaluated before and after anti-infective therapy.

A standard urine collection procedure was performed to obtain a minimum of 80ml of urine, with 3ml being utilized for mNGS examination, at least 20 ml for urinalysis, and 50 ml or more for urine culture. The results of the urine culture were then compared to those of the mNGS examination in order to evaluate any differences in pathogen detection. Patients were instructed to collect clean midstream urine after cleaning the area around the urethra. For patients who need to collect urine from a catheter or nephrostomy tube, the proximal sampling site of the catheter was cleaned and a syringe needle was inserted into the catheter lumen to aspirate urine after clamping the drainage tube for no more than 30 minutes. A total of 33 samples were collected in this study. Urinalyses were performed within 2 hours of collection; urine culture samples were sent to the microbiology laboratory at the Second Hospital of Tianjin Medical University within 2 hours of collection and processed according to recommended microbiological methods. mNGS samples were stored at 4 degrees Celsius after collection. All the patients signed an informed consent form.

The culture, isolation and strain identification of bacteria and fungi from midstream clean urine specimens were completed according to the recommended microbiological methods. The quality control strains of Gram-positive cocci, Gram-negative bacilli and fungi were Enterococcus casseliflavu ATCC700327, Enterobacter cloacae ATCC700323 and Candida krusei ATCC 6258, respectively.

After sample collection, samples were sent to the testing center within 24 hours on dry ice for PACEseq mNGS (Hugobiotech Co., Ltd., Beijing, China). Firstly, the cells in samples were removed by centrifugation, and the supernatant was collected for the subsequent DNA extraction. DNA extraction was performed using the QIAamp DNA Micro Kit (QIAGEN). Secondly, a DNA library construction was performed on the obtained extracts using the QIAseq™ Ultralow Input Library Kit (Illumina). The constructed libraries were assessed for quality using a Qubit fluorescence quantification analyzer (Thermo Fisher) and an Agilent 2100 Bioanalyzer (Agilent Technologies). The quality-qualified databases were subjected to macro-sequencing using the Nextseq 550 platform. Then, the low-quality, low-complexity and small fragment DNA sequences were removed from the raw files after mNGS sequencing; the human sequences matching to the human reference genome (hg38) were then screened out using BWA software. Finally, the remaining sequences were matched to the microbial reference genome database (ftp://ftp.ncbi.nlm.nih.gov/genomes/) (Li and Durbin, 2009). The species information of the genomic targets matched by the above sequences was retrieved, and the specific sequencing reads of all microorganisms and their ratio to the total number of sequences, i.e., relative abundance, were retrieved. In addition, negative controls (sterile deionized water) and positive controls (synthesize fragments with known quantities) were established for each batch of experiments using the same wet lab procedures and bioinformatics analysis as the clinical samples.

Ultimately, the report of mNGS detection provided the species, number of the specific sequencing reads, confidence level of all detected microorganisms (bacteria, fungi, viruses) as well as drug resistance gene testing.

After the results of culture and mNGS detection were returned, at least 3 urology, infectious disease and microbiology specialists worked together to interpret the reports, taking into account the patient’s clinical background, concomitant symptoms, and other microbiological and laboratory findings. They assessed whether the patient has a UTI and gave a final clinical diagnosis, and then jointly decided on a treatment plan.

Urinary symptoms before and after medication were assessed using overactive bladder syndrome score (OABSS). Total score (score: 0-15) and individual scores for daytime frequency (score: 0-2), nighttime frequency (score: 0-3), urgency (score: 0-5), and urgency incontinence (score: 0-5) were assessed. A total score of 3-5 was considered as mild voiding symptoms, 6-11 as moderate, and ≥12 as severe ( Supplementary Table 1 ) (Homma et al., 2006). In addition, patients’ frequency of dyspareunia was assessed as: not at all, a few times, almost always, fairly often and usually (O’leary et al., 1997).

The collected medical records were analyzed descriptively. A 2 × 2 column table was established to determine the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) using the clinical diagnosis as the criteria. Statistical analysis was performed using SPSS (version 25.0) software, and data were plotted using Prism (version 8.0) software. Detection rates of mNGS were compared with conventional culture using McNemar’s test; paired t-test or Wilcoxon signed rank test was used to test paired data. P < 0.05 was considered statistically significant, and all tests were two-tailed. Bonferroni correction is applied to address the issue of multiple testing.

Of the 33 patients enrolled, 22 were males and 11were females, with a mean age of 65.0 ± 12.0 years. Their demographic and clinical characteristics are summarized in Table 1 . 22 of these 33 patients were tested after a poor response to empirical anti-infective therapy, and the remaining 11 patients chose to be detected without empirical antibiotic therapy. The included cases were all patients with clinically suspected UTI, and 25 were ultimately clinically diagnosed with UTIs.

Among the 33 patients, mNGS results were compared with conventional culture results in terms of detection of bacteria and fungi as shown in Table 2 and Supplementary Table 2 . The results of both methods were positive in 10 cases; of the 23 patients with negative urine culture results, only 4 were negative for mNGS, leaving 19 cases with at least one pathogen detected by mNGS; there were no cases with positive results by conventional culture and negative results by mNGS ( Table 2 and Supplementary Table 2 ). mNGS detected at least one pathogen in 29 (87.9%) urine samples, while urine culture detected only 10 (30.3%) cases (87.9% vs. 30.3%, P < 0.001). Compared with the urine cultures, mNGS detected more bacteria (87.9% vs. 27.3%, P < 0.001) and fungi (12.1% vs. 3.0%, P = 0.250). ( Table 3 and Figure 1 ).

The sensitivity and specificity of mNGS for the diagnosis of UTI, were 100% and 50%; 86.2% and 100% for PPV and NPV, respectively. The sensitivity and specificity of conventional diagnostic methods for the diagnosis of infection were 40% and 100%, and 100% and 34.8% for PPV and NPV, respectively ( Table 4 ). And the sensitivity of mNGS for the diagnosis of UTI was significantly higher than that of culture (100.0% vs 40.0%; P < 0.001).

Among the microorganisms (n=81) isolated by mNGS from 33 specimens, 51 were determined to be of high confidence and 30 of moderate confidence. There were 51 strains of bacteria, of which 37.3% (19/51) were Gram-positive and 62.7% (32/51) were negative; 4 strains of fungi and 26 strains of viruses ( Figures 2 , 3 ). A total of 9 bacteria and 1 fungus were detected by culture. In terms of the detection of bacteria, both assays showed that the Gram-negative bacteria and facultative anaerobes were the most common ( Figure 3 ). The most common bacteria detected by mNGS was E. coli, followed by Prevotella and Enterococcus faecalis; the most common bacteria detected by culture was also E. coli. Among the 33 specimens, mNGS detected 23 species of bacteria, 3 species of fungi, and 6 species of viruses. Out of these, 19 bacterial and 2 fungal species were undetected through conventional culture techniques. In addition, all viruses, M. tuberculosis and anaerobes were detected by mNGS ( Figures 2 , 3 ). The mNGS performed significantly better than culture the detection of pathogens.

Among the 33 cases of mNGS detection, 22 patients were tested when their disease did not improve significantly or even worsened after treatment with empirical antibiotics, and 11 patients underwent testing directly. An overview of cultures and mNGS results in detection of bacteria and fungi in the two groups was demonstrated in Table 5 . mNGS was not affected by previous use of antibiotics in detecting pathogens (90.9% VS. 81.8%, P = 0.586).

Of these 33 patients, 24 patients voided through the urethra and 8 patients voided through the catheter (including urinary catheters and nephrostomy tubes) ( Table 1 ). Pathogenetic findings suggested that the majority of microorganisms detected are bacteria, regardless of whether it is mNGS or culture. Fungi detected in culture are only from samples collected through catheters, and mNGS results showed that more viruses are detected in samples collected through catheters. In our study, no virus culture was performed ( Supplementary Figure 1A ). Urine samples tested using mNGS, whether within the range of bacteria and fungi or all pathogens (bacterial, fungal, viral), no negative cases were found in any of the transcatheter-collected urine samples, and the proportion of polymicrobial samples was greater ( Supplementary Figures 1B, C ).

In the detection of bacteria and fungi, 10 (30.3%) cases showed positive for both methods (double positive), 4 (12.1%) cases showed negative for both methods (double negative), while 19 (57.6%) cases had pathogens detected by mNGS only. Notably, there were no cases in which the conventional method showed positive while the mNGS showed negative. Within the 10 double-positive cases, the results of the two tests were compared, and 5 cases were completely matched and 1 was completely mismatched. The remaining 4 cases were found to be partly matched ( Figure 4 and Supplementary Table 2 ). There are very few cases with conflicting results between the two tests, indicating good consistency between them.

Among the 23 cases with negative conventional test results, only 4 had negative results for mNGS, while 19 had positive results for bacteria and fungi, 6 of which were monomicrobial and 13 were polymicrobial. Among the 19 pathogen-positive patients, 22 species of pathogens were identified by mNGS. In conclusion, mNGS improved the diagnostic efficiency in cases with negative conventional tests ( Supplementary Table 2 ).

Single pathogens were detected by mNGS in 11 (33.3%) cases and mixed pathogens in 18 (54.5%) cases; urine culture detected single pathogens in 10 (30.3%) cases and failed to detect the presence of mixed pathogens. All of the mixed pathogens were detected by mNGS. ( Table 6 and Supplementary Table 2 ).

A total of 3 patients positive for M. tuberculosis were detected by mNGS. Two of these patients, who were diagnosed with high-risk bladder cancer, received intravesical immunotherapy with bacillus Calmette-Guérin (BCG, a live attenuated vaccine for Mycobacterium tuberculosis) as per guidelines (Redelman-Sidi et al., 2014), with specific sequence numbers of 6 and 9, respectively. The remaining patient showed urinary tuberculosis with a specific sequence number of 383 and a positive X-pert test result. This patient with a high suspicion of urinary tuberculosis infection was later diagnosed in an infectious disease hospital.

For the 25 patients with a positive clinical diagnosis, the treatment protocol was adjusted according to the mNGS results in 23 (92%) patients, while the remaining 2 patients continued their current treatment because the causative organisms were covered by the original antibiotic regimens. All patients showed significant improvement in their urinary symptoms and urinalysis after targeted anti-infection treatment ( Supplementary Table 3 ).

A total of 25 patients were clinically diagnosed with UTI. All of these patients received further targeted anti-infective treatment based on the results of mNGS and urine cultures. Multiple urinalyses before and after treatment were collected in 23 of these patients (the other two patients were not reexamined after treatment). The mean value of white blood cells in high power field (WBCHPF) counts before and after treatment were taken for comparative analysis. The results showed a significant treatment effect, with a statistically significant decrease in urine WBCHPF counts after treatment in these 23 patients ( Table 7 and Supplementary Table 3 ).

Changes in voiding are another aspect of assessing the effectiveness of anti-infective therapy. Of the 25 patients, 18 were reviewed for voiding symptoms before and after anti-infective therapy, except for 5 patients who could not be evaluated because of indwelling urinary catheters and 2 patients who underwent percutaneous nephrostomy. 13 of the 18 patients had received empirical anti-infective therapy, and the other 5 patients did not ( Supplementary Table 3 ).

The efficacy of the empirical medication was assessed by comparing the voiding symptoms of the 13 patients mentioned above,7/13 patients had no change in overall OABSS; 1/13 patients had no effect of treatment and showed even more severe symptoms; 5/13 patients showed a slight improvement in symptoms. In these 13 patients, the differences in symptom changes were not statistically significant. Among the 5 patients with symptom improvement, 2 patients still had high scores, 10 and 11, respectively, both graded as moderate, tending to be severe ( Figure 5A ). After mNGS detection, these 13 patients received further anti-infective therapy. 11 out of 13 patients had symptom relief after this treatment, and 2 patients had no change in symptom score; no exacerbation of symptoms occurred. The difference before and after this treatment was statistically significant. The 2 patients with no change in symptom scores had a low score of only 2 points before treatment. The patient whose symptom worsened after previous empirical anti-infective treatment had a significant symptom relief, with a decrease in score from 11 to 2 ( Figure 5B ). Changes in voiding symptoms before and after treatment were compared in 18 patients who received targeted anti-infection therapy in the presence of mNGS detection, and the differences were statistically significant ( Figure 5C ). Similarly, empirical anti-infective treatment was ineffective in relieving patients’ dysuria; however, treatments given on the basis of mNGS detection were effective ( Figures 5D–F ).

The median overall OABSS before targeted anti-infective treatment was 9 (IQR: 5-13), and the median daytime frequency, nighttime frequency, urgency and urgency incontinence scores were 1, 3, 4, and 1, respectively. While the median overall OABSS after treatment decreased to 3 (IQR: 2-6), with median scores of 0, 2, 1, and 0 for daytime frequency, nighttime frequency, urgency, and urgency incontinence, respectively ( Figure 5C and Supplementary Figure 2 ).