Urinary cell-free DNA is a versatile analyte for monitoring infections of the urinary tract

Abstract

Urinary tract infections are one of the most common infections in humans. Here we tested the utility of urinary cell-free DNA (cfDNA) to comprehensively monitor host and pathogen dynamics in bacterial and viral urinary tract infections. We isolated cfDNA from 141 urine samples from a cohort of 82 kidney transplant recipients and performed next-generation sequencing. We found that urinary cfDNA is highly informative about bacterial and viral composition of the microbiome, antimicrobial susceptibility, bacterial growth dynamics, kidney allograft injury, and host response to infection. These different layers of information are accessible from a single assay and individually agree with corresponding clinical tests based on quantitative PCR, conventional bacterial culture, and urinalysis. In addition, cfDNA reveals the frequent occurrence of pathologies that remain undiagnosed with conventional diagnostic protocols. Our work identifies urinary cfDNA as a highly versatile analyte to monitor infections of the urinary tract.

Fig. 1

Shotgun sequencing assay and biophysical properties of urinary cfDNA. a Schematic representation of the ssDNA library preparation protocol used for shotgun sequencing of urinary cfDNA. Key steps include: (i) cfDNA isolation, (ii) DNA denaturation, (iii) ssDNA adapter ligation, (iv) extension and double-stranded DNA adapter ligation, and (v) PCR. b Overview of post-transplant sample collection dates (color indicates pathology, bars connect samples from same subjects). c, d Fragment length density plot measured by paired-end sequencing for different cfDNA types: c chromosomal and BKV cfDNA from representative samples, and d E. coli, parvovirus B19, and mitochondrial cfDNA from representative samples. Fourier analysis reveals a 10.4 bp periodicity in the fragment length profiles of chromosomal and BKV cfDNA but not for E. coli, parvovirus B19, or mitochondrial cfDNA (insets). See Supplementary Data 1

Fig. 2

Urinary cfDNA infectome screening. a Violin plots of BKV cfDNA sequence abundance (in RGE) for subjects with and without BKVN and untested subjects. b cfDNA rank order abundance for clinically reported organisms. In 60% of samples, the bacterial organism detected in culture was the most abundant component of the cfDNA urinary microbiome (rank 1). In one sample, the clinically reported agent was not detected (ND, R. ornithinolytica). c ROC analysis of the performance of urinary cfDNA in identifying bacterial organisms (86 urine samples, AUC area under the curve, n number of positive cultures, see Supplementary Figure 1 for individual ROC curves for these and two additional bacteria). d Viral cfDNA was detected in 66 samples of the 141 samples. cfDNA reveals frequent occurrence of viruses that are potentially clinically relevant (left panel); crosses identify samples belonging to subjects who developed an infection of the corresponding viral agent. Right panel shows boxplots of the viral cfDNA abundance across all samples (right panel). Coloring of points and boxplots by viral taxonomic group; see Supplementary Data 2–5

Fig. 3

Estimating bacterial population growth rates from urinary cfDNA. a Normalized bacterial genome coverage for four representative bacterial species. The coverage was binned in 1 kbp tiles and normalized. Each panel represents a single sample (see Supplementary Data 6), with the exception of C. acnes (asterisk (*)) for which the coverage was aggregated across 99 samples (solid line is a LOESS filter smoothing curve, span = 0.70). The non-uniform genome coverage for E. coli and K. pneumoniae, with an overrepresentation of sequences at the origin of replication, is a result of bi-directional replication from a single origin of replication. The initial and final 5% of the genome is removed for display. b The skew in genome coverage reflects the bacterial growth rate, where a stronger skew signals faster growth. Box plots of growth rates for species in 14 genera grouped by patient groups (at least 2500 alignments, 41 samples, see Methods for definition of pre/post-UTI). Each point indicates a bacterial species in a sample. Triangles indicate culture-confirmed bacteria by genus. Boxplot features are described in Methods. See Supplementary Data 6 and 7

Fig. 4

cfDNA-based antimicrobial resistome profiling. For 42 samples from subjects with clinically confirmed UTI, AR gene profiling reveals the presence of genes conferring resistance to various antimicrobial classes. These data are organized in three sample groups: samples from subjects with vancomycin-resistant Enterococcus (Resistant), samples from subjects with vancomycin-susceptible Enterococcus (Susceptible), and samples from subjects for which vancomycin resistance testing was not performed (Untested). Samples in which fragments of genes that confer resistance to glycopeptide class antibiotics (including vancomycin, red outlines) were detected are marked by red crosshairs. See Supplementary Data 8

Fig. 5

Quantifying the host response to infection from urinary cfDNA. a Proportion of donor-specific cfDNA in urine of subjects that are BKVN positive per kidney allograft biopsy (BKVN) in urine collected in the first 5 days after transplant surgery (Early), urine collected from subjects that are bacterial UTI negative per culture in the first month following transplantation (No UTI), samples collected before or after bacterial UTI (Pre-/Post-UTI), and samples collected at the time of bacterial UTI diagnosis (UTI). The single outliers in Pre-/Post-UTI and UTI groups correspond to the same patient, who suffered an acute rejection episode in the months prior. Low donor fractions in the Pre-/Post-UTI and UTI groups are likely due to increased immune cell, i.e., WBC, presence in the urinary tract; subjects with higher WBC counts have lower donor fractions (inset, red color indicates pyuria). b Absolute abundance of donor cfDNA in the urine of subjects not diagnosed with infection in the first month post-transplant (red line is a LOESS filter smoothing curve, span = 1). Dotted lines connect samples from the same patient. c Genome coverage at the transcription start site, binned by the gene expression level across all samples in the study. FPKM fragments per kilobase of transcript per million mapped reads, an RNA-seq measure of gene expression. See Supplementary Data 9 and 10