Screening Biomarkers and Constructing a Predictive Model for Symptomatic Urinary Tract Infection and Asymptomatic Bacteriuria in Patients Undergoing Cutaneous Ureterostomy: A Metagenomic Next-Generation Sequencing Study

Abstract

Objectives: To investigate the clinical diagnostic value of differential flora as biomarkers in patients with symptomatic urinary tract infection (UTI) and asymptomatic bacteriuria (ASB) undergoing cutaneous ureterostomy based on metagenomic next-generation sequencing and construct predictive models to provide a scientific reference for clinical diagnosis and treatment. Material and Methods. According to standard procedures, samples were taken from each patient for routine tests (urine, ureteral stent, and skin swab around the stoma). Cytokine levels in the blood were also detected. Urinary microflora were measured by mNGS, and potential biomarkers for distinguishing UTI and ASB were identified by differential flora. Finally, we generated the predictive models for ASB and UTI using the Lasso method and cytokine levels.

Results: Urine culture was performed for 50 patients with cutaneous ureterostomy; 44 of these patients developed bacteriuria. The incidence of symptomatic bacteriuria was 54.55%. Biomarker analysis showed that Propionimicrobium lymphophilum, Staphylococcus haemolyticus, Stenotrophomonas maltophilia, Ralstonia insidiosa, and Aspergillus sydowii all had good predictive performance and were combined in a single model. The predictive model exhibited good prediction performance (area under the curve (AUC) = 0.8729, sensitivity = 80%, specificity = 83.3%, and cutoff = 1.855). We also identified a significant negative correlation between the weight sum of the abundance for these five characteristic pathogens (Sum_weighted_Reads) and levels of the cytokine IL-6 and IL-1β (P < 0.05).

Conclusion: mNGS had a higher positive detection rate for pathogens in urine samples. The selected differential bacteria can be used as biomarkers of ASB and UTI, and the prediction model has good predictive performance. Analysis also showed that the occurrence of symptoms was related to individual immunity. Combined with the Sum_weighted_Reads cutoff and cytokine levels (IL-6 and IL-1β) of differential flora, it was possible to judge the severity of symptoms in cutaneous ureterostomy patients with bacteriuria and provide new insights for the treatment and intervention of ASB and UTI.

Figure 1

Urine culture pathogen detection spectrum and frequency.

Figure 2

Ven diagram of cultured species of different kinds of samples. The 14 overlapping species of three groups (urine, skin, and stent) were Candida auris, Enterococcus faecium, Enterobacter cloacae complex sp., Proteus vulgaris, Providencia stuartii, Proteus mirabilis, Morganella morganii, Enterococcus faecalis, Staphylococcus aureus, Escherichia coli, Klebsiella pneumoniae, Klebsiella oxytoca, Serratia marcescens, and Streptococcus oralis.

Figure 3

Heatmaps of detection frequencies for 14 species.

Figure 4

mNGS pathogen detection spectrum and frequency. Only twenty top species of bacteria by frequency were displayed.

Figure 5

Metagenomic microbial diversity analysis. (a) Shannon α index diversity analysis. (b) α diversity Simpson index analysis. (c) Weighted PCoA using Bray-Curtis distance. (d) Unweighted PCoA.

Figure 6

ROC curve of the prediction model.

Figure 7

Association analysis of inflammatory factors. (a) Differential analysis of the inflammatory factor IL-6 between the ASB and UTI groups, (b) differential analysis of the inflammatory factor IL-1β between the ASB and UTI groups, (c) correlation analysis between Sum_weighted_Reads and IL-6 levels, and (d) correlation analysis between Sum_weighted_Reads and IL-1β levels.