Viral associations with kidney disease diagnosis and altered kidney metatranscriptome by kidney function

Graphical Abstract

Common causes of nephrotic syndrome are known to have immune or autoimmune components and are treated with immunosuppression. However, for many of these kidney diseases, the initiating pathogenic event remains unknown. Recently, the combined use of next generation sequencing and pathogen searching algorithms have identified infectious causes for chronic diseases of unknown etiology.¹ RNA sequencing data (RNAseq) from tissue extractions contain both human RNAs (transcriptome) and non-human RNAs from infecting pathogens or commensal organisms (metatranscriptome). If an infectious agent is an important contributor to pathogenesis, RNA of the infectious agent would be expected to be abundant in the kidney during disease onset and with active disease. Next generation sequencing methods also are revealing the complexities of the human microbiome in health and disease. For kidney disease, microbiome studies have been limited to the role of the gut microbiome on kidney function or cataloging the blood or urine microbiomes in the setting of kidney disease or urolithiasis.² However, the microbiome of the kidney (upper urinary tract) has not been examined.

As a discovery strategy, we examined the metatranscriptome from existing kidney biopsy RNAseq data to identify viral infections associated with kidney diseases of unclear etiology. We also used the composite metatranscriptome data on viruses, bacteria, and eukaryote microbes to provide an initial assessment of the kidney microbiome.

RESULTS

RNAseq data from two chronic kidney disease (CKD) cohorts, NEPTUNE³ and the HIV-positive BEAN^4,5 cohort, along with healthy controls were used to assess the prevalence and abundance of microbial RNAs in kidney biopsies (Table S1 and Supplemental Methods). Due to their unique structure and function, the glomeruli (GLOM) and tubulointerstitium (TI) were examined independently and revealed striking differences in viral RNA abundance between the two compartments (Fig 1A). A few virus families were more prevalent in the healthy controls such as Polyomaviridae, and is consistent with studies detecting greater JC virus in urine of healthy individuals compared to subjects with nephropathy.^S1–S4 Some viruses were unique to the GLOM, and overall, the restriction of viruses to GLOM and TI compartments was consistent with known viral tropism (Fig S1A). Similarly in the CKD+HIV group, Retroviridae and Flaviviridae were the most abundant and prevalent viral RNAs and would be expected based on subject infection status (Fig S1B). Only one viral family (Picornaviridae) was more common in the TI, driven by two enterovirus species restricted to the TI (Fig S2).

Figure 1. Viral RNA prevalence and abundance differed by kidney compartment and were associated with specific kidney diagnoses.

A. Heat map of viral family abundance by kidney disease diagnosis. Several viral families were significantly different based on CKD diagnoses. Data are mean transcripts per million reads (TPM), scaled by viral family, with differences between CKD diagnosis determined by ANOVA (P values FDR corrected), *P<0.1, **P<0.01. B. Heat map for prevalence and abundance of papillomavirus RNAs for each kidney disease diagnosis by HPV subtype (uns, unspecified subtype). In the GLOM, association with Papillomaviridae in panel A was driven by a human papilloma virus (HPV) subtype 68. Significance determined by FDR corrected ANOVA, *P<0.05. C. Individual subject data for GLOM RNAs of viruses identified in panel A stratified by CKD diagnosis and number of APOL1 risk alleles (RA). Subjects on immunosuppression at time of biopsy are noted, but no clear association with virus prevalence or abundance was evident. Collapsing FSGS was associated with high prevalence of RNA sequences from Arenaviridae, Nairoviridae and Nimaviridae. See Table S2 for relative risk estimates.

Virus Discovery

Several viral RNAs in GLOMs were significantly associated with specific diagnoses (Fig. 1A). Human papilloma viruses (HPVs) were common in both the GLOM and TI, with the TI possessing a variety of HPVs although all of low abundance (Fig 1B). HPV-68 was the most common HPV subtype detected in GLOMs and predominated in cases of MN. HPV-68 is more frequently associated with cervical cancer in Africans and African Americans,^S5–S8 but there was no significant difference in HPV-68 abundance when stratified by race, sex, age, or APOL1 genotype (Fig S3). Additionally, several viral families were more prevalent in FSGS cases, and in particular cases with the pathologic feature of glomerular collapse (Fig 1C) and may represent new viral families that contribute risk for collapsing FSGS (Table S2). Since associations were only observed at the taxonomic family level, this may reflect the identification of uncharacterized viruses with sequence similarities to various known viruses, as the searching method permitted mismatch alignments.

Glomeruli and tubulointerstitium have different metatranscriptomes

We also assessed the entire kidney metatranscriptome using ecological community analyses. In the CKD cohort, alpha diversity was greater in the TI and significantly different from the GLOM (Fig 2A) and these differences were evident in each domain (Fig 2B). Similarly, beta diversity comparisons (Fig 2C) also identified significant differences between the GLOM and TI for each domain. For the CKD+HIV cohort, alpha and beta diversities were also significantly different between the GLOM and TI (Fig S4). Using differential abundance, the bacterial domain exhibited similarities between the GLOM and TI, whereas the GLOM had more diverse viral and eukaryote microbe RNAs (Fig 2D). The bacterial RNAs were mainly common gut commensals including Proteobacteria, Actinobacteria, Bacteroidetes, and Firmicutes. Although viral RNAs were abundant in both fractions, the TI was dominated by bacteriophages (Caudovirales) common in the gut. In the GLOM, instead of bacteriophages, RNAs of eukaryotic viruses predominated. Eukaryote microbial RNAs also had greater diversity and abundance in the GLOM, which was most abundant in subjects with APOL1 low risk genotypes (Fig S5).

Figure 2. Glomerular and tubulointerstitial compartments have different metatranscriptomes and vary based on kidney function.

Comparison of microbial RNAs in GLOM and TI by domain (bacterial, viral, and eukaryote microbes). A. CKD cohort microbe taxa complexity was determined by alpha diversity assessments for richness (Margalef index: species abundance) and diversity (Shannon index: species abundance and evenness) revealing significant differences between GLOM and TI. B. Shannon and Margalef indices stratified by domain, showing the differences identified in panel A impact all viral, bacterial, and eukaryote populations. Plots are median line, interquartile range boxes, and 95% confidence interval whiskers. C. Beta diversity comparisons using the ecological Bray-Curtis dissimilarity index (a community level comparison) similarly demonstrate differences between GLOM and TI. Statistical analysis was conducted as a PERMANOVA with 999 permutations (P values on graphs, PCo, principal coordinate). D. Differential abundance analysis reporting the number of operational taxonomic units (OTUs) per the lowest assigned taxonomy (p, phylum; c, class; o, order; f, family; g, genus; s, species). Normalization was performed using DESeq2, only OTUs with false discovery rate P values <0.05 are shown. E. Alpha diversity (all domains) using eGFR as a continuous variable. F. Two-way alpha diversity for differences based on HIV infection status and for differences in GLOM and TI. Only subjects with documented tests for infection were included (HIV+, n=7; HIV−, n=32). Box plots show mean line, interquartile range boxes, with 95% confidence interval whiskers. *P=0.017, **P=0.002, Bacterial diversity P=0.084. See Table S3 for additional two-way alpha diversity comparisons.

Metatranscriptome diversity differed with kidney function and pathology

Kidney function also exhibited significant differences in alpha diversity in the composite metatranscriptome based on eGFR (Fig 2E). However, differences between the GLOM and TI in the metatranscriptome diversity (all domains) did not identify any significant differences when stratified by sex, age, APOL1 genotype, or CKD diagnosis by domain (Fig S6). However, when domain compositions were examined in conjunction with GLOM and TI differences, alpha and beta diversity differences were significant for some cohort characteristics using two-way analysis of variance (Table S3). In two-way comparisons, both virus and bacterial domain diversities were different based on pathologic features including glomerulosclerosis and glomerular hyperplasia, and in the virus domain, glomerular hypertrophy. The eukaryote domain diversity was significant for proteinuria, hypertension, and APOL1 genotype, and the bacterial domain for CKD diagnosis. A further examination of TI bacterial species (Fig S7) observed alterations in bacterial populations that were analogous to prior studies describing changes in gut bacteria in subjects with low eGFRs.^S9–S11 In addition, in the setting of HIV infection, diversity in all three domains increased in the TI, but not in the GLOM (Fig 2F). In prior studies of HIV/AIDS patients, several groups have reported a similar expansion of viral populations^S12–S14 and bacteria translocation from the gut.^S15–S16 Taken together, the TI metatranscriptome resembled the gut microbiome in terms of bacterial and bacteriophage composition, and exhibited changes in viral and bacterial populations known to occur in the gut in the setting of CKD or HIV infection. The metatranscriptome of the GLOM fraction, however, was clearly different than TI, with more diversity in viral and eukaryote microbe domains.

DISCUSSION

Although prior studies have examined the microbiome of urine, there have been no assessments of the microbiome of the kidney itself. The remarkable differences in the metatranscriptomes between the GLOM and TI suggest there may be several unique microbiomes in the upper and lower urinary tracts. Recent studies using single cell sequencing methods have observed spatial zonation of immune cells populations in the kidney pelvis, medulla, and cortex, reflecting specialization of antimicrobial immunity in each compartment.^S17–S18 Continued work to understand these cell-type restricted immune functions and compartmentalized infections or commensal populations may advance our mechanistic understanding of kidney disease phenotypes.

Our study focused on glomerular disorders and the abundance of microbe RNAs in the GLOM may reflect the disease process and cellular tropism of the infectious agents. It is possible a similar analysis of tubular disorders may identify a novel population of infectious agents in the TI. The use of RNAseq data may have missed DNA viruses, although if a viral infection is contributing to pathogenesis, there is an expectation the virus would be replicating and generating mRNA. Similarly, old, latent infections also would be unlikely contributors to new disease or flares. Commonly used medications in these cohorts (antibiotics, antivirals, immunosuppression) could alter microbe populations, and appropriately controlled treated and untreated groups would be needed to assess their impact.

The new viral associations with specific CKDs require validation, using virus-specific assays and verification in other cohorts, and experiments to establish causation. Our study identified several new viral families associated with collapsing FSGS and strengthens the concept that glomerular collapse is pathognomonic for a viral infection. The association of HPV-68 in MN was unexpected. The initiating pathogenic events that cause MN are not well understood, but connections with infections or cancer are not new.⁶ Future work simultaneously examining blood, urine, and kidney tissue for these viruses may help clarify their role in pathogenesis. Importantly, linking a viral infection with pathogenesis could enable the use of anti-viral medications or vaccines for treatment and prevention.