Temporal and diversity characterization of metagenomic viral detection in kidney transplant recipients
Rohita Sinha1, Mikaela Miller1, Christabel Rebello2, Bradley Kinsella2, John Friedewald2, Steve Kleiboeker1.
1Eurofins Viracor, Lees Summit, MO, United States; 2Feinberg School of Medicine, Northwestern University, Chicago, IL, United States
Introduction: Kidney transplant recipients are highly immunosuppressed and thus predisposed to opportunistic infections. Current viral diagnostic methods from plasma detect single pathogens, often by qPCR. However, unbiased method detecting common viruses with sensitivity and specificity comparable to qPCR would allow comprehensive diagnosis. Quantification of donor-derived cell-free DNA (dd cfDNA) is an established method for diagnosis of allograft rejection. Combining dd cfDNA quantification and pathogen detection into a single sequencing workflow represents an opportunity to diagnose two impactful complications in transplant recipients.
Method: A total of 1980 plasma samples from 256 CTOT-08 study subjects were tested. Whole genome cfDNA sequencing was performed on an Illumina platform. Sequence data along with recipient genotype, were analyzed using a bioinformatics pipeline to calculate the percentage dd cfDNA present. Non-human reads underwent reference-assisted assembly and taxonomic annotation using KrakenUneq which was trained (K-mer values of 16, 21, and 31) on an in-house database of ~12,000 viral genomes. The final predictions were made by applying a majority-wins rule. Read assembly (contig) length and read depth were reported along with the viruses detected.
Results: Of 256 subjects, 230 (89.8%) had ≥1 sample positive for viral detection. Of the 1980 samples tested, 979 (49.4%) had ≥1 viral detection(s) with contig lengths from 32 – 125,199 bp. The number of viruses detected was negatively associated with dd cfDNA values. For every additional virus detected, dd cfDNA values decreased 7.8% (P<0.001, 95% CI:4%, 11%). The dd cfDNA values skewed higher in samples with no virus detected (Fig. 1). Torque Teno virus (TTV) was the most common detection (25.6%), followed by BK virus (19.8%), cytomegalovirus (8.3%), adenovirus (2.6%), and Epstein Barr virus (2.6%). Other notable viruses detected included human herpesviruses 1, 2, 6A/B, and 7, JC polyoma virus, human polyomavirus 6 and 8, and gammapapillomaviruses 1 and 9. Concordance with qPCR results was 92.5%, with discordant results <350 copies/mL or contig lengths <100 bp. Analysis of temporal patterns demonstrated the prevalence of viral detection peaked early for all viruses, especially TTV, with declines in prevalence through the study (Fig. 2). Detection of BKV and CMV prior to (1 - 30 days) a clinical diagnosis was documented in 12 of 17 and 7 of 11 samples, respectively. The rate of detection after the clinical diagnosis was 21 of 24 and 4 of 6 for BKV and CMV, respectively.
Conclusion: Metagenomic viral detection combined with dd cfDNA quantification demonstrated sensitive detection of a broad range of DNA viruses, including key pathogens, with a strong temporal pattern. Viral detection was negatively correlated with dd cfDNA values and concordance with qPCR was high. Prospective studies are needed to assess the clinical utility of combined metagenomic and dd cfDNA analysis.
