Host-Pathogen Interactions in Human Polyomavirus 7 (HPyV7)-associated Pruritic Skin Eruption

TO THE EDITOR

Certain polyomaviruses, such as JC and BK, are known to cause severe disease in immunocompromised hosts (Moens et al, 2017). Human polyomaviruses 6 and 7 (HPyV6/7) can be shed from skin surfaces without apparent disease (Schowalter et al, 2010; Hashida et al, 2018), while, in immunocompromised patients, they are associated with a pruritic eruption (Ho et al, 2015; Nguyen et al, 2017; Canavan et al, 2018; Smith et al, 2018).

We report the case of a 38-year-old woman who developed intractable pruritus and burning associated with a skin eruption one year after her second renal transplantation (Supplementary Figure S1a-c). Her immunosuppressive regimen consisted of tacrolimus and mycophenolate mofetil (MMF). Histopathology was characteristic of HPyV6/7-associated eruptions (Ho et al, 2015; Nguyen et al, 2017). Immunohistochemistry confirmed expression of polyomavirus large T antigen (LT), and immunofluorescence identified HPyV7 T antigen and HPyV6/7 VP1 expression (Supplementary Figure S1d). Shotgun metagenomic sequencing revealed HPyV7 DNA and the absence of HPyV6 (Supplementary Figure S2). The addition of acitretin and reduction of MMF dose eventually led to resolution of symptoms (Supplementary Figure S1b,c). Detectable viral protein expression was absent in apparently normal skin after 15 weeks on acitretin and MMF dose reduction (late-treatment timepoint) (Supplementary Figure S1d).

Shotgun metagenomic sequencing of skin swabs from pre-treatment and early-treatment (3 weeks on acitretin) revealed that 99.99% of mappable non-human reads were HPyV7, which is a high frequency, even in patients with viral skin disease (Tirosh et al, 2018). Surprisingly, HPyV7 continued to constitute 99.96% of non-human reads at the late-treatment timepoint (Figure 1a). However, the percentage of non-human reads among total reads was 59% at pre-treatment compared to 4.5% at late-treatment (Supplementary Figure S3), suggesting that the overall HPyV7 DNA burden had decreased. Bulk RNA sequencing (RNA-seq) revealed a 940-fold decrease in total normalized HPyV7 transcripts in a late-treatment clinically normal-appearing skin specimen as compared to the more affected pre-treatment specimen (Figure 1b), demonstrating that clinical resolution was associated with a large reduction in viral transcription.

Figure 1: HPyV7 was highly prevalent on affected skin, had significantly reduced gene expression at late-treatment, and differentially expressed minor variants of the agnogene in affected skin.

(a) Heatmap displaying mappable non-human reads from shotgun metagenomic DNA analysis from skin swabs. (b) Total HPyV7 Fragments Per Kilobase Million (FPKM) as determined by bulk RNA-seq. (c) Fraction of Variant 1 and Variant 2 as determined by shotgun metagenomic DNA sequencing and bulk RNA-seq at a highly mutated locus in the regulatory region (RR). (d) Barplots showing the fraction of reads supporting the genbank reference sequence (dark green), “common indels” (overall allele frequency (AF) > 0.025) and “uncommon indels” (AF < 0.025) in HPyV7 DNA from Patient #2 (a published patient), Patient #1 and “healthy” Sequence Read Archive (SRA) datasets, identified by having >1 read mapping to the HPyV7 RR and annotated as coming from healthy patients. Each x-axis tick denotes one SRA run.

During pathogenic infection with some polyomaviruses, the regulatory region undergoes sequence rearrangements and can accumulate insertion or deletion nucleotide variants (indels) that affect transcription factor binding (Ajuh et al, 2018). Comparison of HPyV7 DNA and RNA sequences from our patient to a reference viral genome revealed subclonal polymorphisms scattered throughout the viral genome. Of note, a section of the regulatory region contained a high diversity of indels. Interestingly, an eight base pair deletion variant in this region (Variant 2) was predominantly expressed in the more affected skin compared to the less affected skin, despite representing a minor variant at the DNA level in both samples when compared to the non-indel containing species (Variant 1) (Figure 1c, Supplementary Figure S4a,b). Functional prediction of this variant suggested disruption of a TATA-box, but no other definitive effects on likely transcription factor binding sites could be identified.

Further investigation revealed the novel transcript likely encodes a previously unrecognized agnoprotein. In addition to the expression of various agnoprotein sequence variants at this site, there were also multiple splice isoforms (Supplementary Figure S4c). Human BK and JC polyomaviruses express agnoproteins, with little amino acid identity to the proposed HPyV7 agnoprotein, which have been suggested to regulate multiple viral and host functions (De Gascun and Carr, 2013).

We next compared the DNA sequences from this highly variable region to the skin of 20 healthy volunteers with asymptomatic HPyV7 colonization (“healthy” SRA datasets). We identified a small number of common indels in the healthy volunteers, in contrast to the numerous distinct indels in our patient (“uncommon indels”). Similar distinct regulatory region indels were detected in the skin of a second, previously reported, patient with HPyV7-associated eruption (Canavan et al, 2018), upstream of the agnogene. The variants from this patient were also rare or absent in healthy volunteers, suggesting that these highly mutated viral regions were relatively restricted to diseased skin and may be associated with pathogenicity (Figure 1d, Supplementary Figure S4b).

Pathway analysis of host gene expression in the RNA-seq samples revealed induction of inflammatory genes associated with type I interferon, the antiviral response, and NKG2D ligands (Figure 2a). Consistent with this, there was strong expression of the interferon-induced antiviral protein, MX1, in pre-treatment epidermis, which was absent during late-treatment (Figure 2b). Two of the most upregulated genes included innate antiviral defense proteins, APOBEC3A/B (Figure 2a). These ssDNA deaminases potentially promote generation of indels upon DNA repair (Burns et al, 2013; Green et al, 2016). APOBEC3A/B exhibited cytoplasmic and nuclear expression, most often within or near virally infected cells (Figure 2c). BK polyomavirus has been shown to upregulate APOBEC3B (Verhalen et al, 2016), which can produce viral mutations that alter glycan usage for viral entry and, conversely, facilitate escape from neutralizing antibodies (Peretti et al, 2018).

Figure 2: Interferon and innate immune signatures in pre-treatment skin resolved by late-treatment when there were epidermal CD8+ T cell infiltration and HPyV7-specific antibody production.

(a) Heatmaps illustrating pathways associated with the most highly upregulated genes in affected skin. (b and c) Immunofluorescence staining of MX1 (b, scale bar=165.12 micrometers), APOBEC3A/B, HPyV6/7 VP1 (c, scale bar=165.12 micrometers), HPyV7 T antigen (c, scale bar=154.8 micrometers) in pre-treatment and late-treatment skin. (d) The absolute number of epidermotropic CD8+ T cells/mm as determined by quantification of immunohistochemical staining of CD8 in pre-treatment and late-treatment skin biopsies. (e) EC50 values, representing the calculated serum dilution that gave half-maximal signal in HPyV7 or BKV virus-like protein (VLP) ELISAs. Pre-rash (-93 weeks), early-rash (-28 weeks), pre-treatment (-11 weeks), early-treatment (+3 weeks), late-treatment (+15 weeks), with time 0 representing the initiation of twice daily acitretin treatment. Technical triplicates were run. Error bars represent the EC50 95% confidence interval.

Although the RNA-seq data suggested that affected skin exhibited innate immune activation, the host response was inadequate to achieve viral clearance. Considering the clinical resolution after tapering immunosuppression, we characterized the changes in the host immune response. Immunohistochemistry revealed a moderate CD4+ infiltrate pre-treatment and during late-treatment, but CD8+ T cells infiltrated the epidermis only during late-treatment (Figure 2d, Supplementary Figure S5). An HPyV7 virus-like particle ELISA detected a high titer of circulating antiviral antibodies late-treatment but not at earlier timepoints. In contrast, antibodies to BK virus were present at each timepoint (Figure 2e).

By using multiple orthogonal approaches, these data support a causative role for HPyV7 in this eruption. These results suggest that HPyV7 stimulates an innate immune response, but a protective antiviral response is impaired by treatment with MMF. We further identified a previously undescribed agnogene within the HPyV7 genome prone to mutation during infection. Many of the observed mutations in the HPyV7 genome are consistent with host activation of APOBEC3 enzymes, as has been seen for other polyomaviruses. Taken together, our findings are consistent with the hypothesis that iatrogenic impairment of immunosurveillance enabled abnormal viral proliferation and innate immune activation that induced APOBEC3A/B expression, which in turn led to viral mutagenesis that correlated with increased HPyV7 pathogenicity and clinical disease.

ABBREVIATIONS :

PyV

Polyomavirus

HPyV7

Human Polyomavirus 7

LT

large tumor antigen

VPs

viral structural proteins

MMF

mycophenolate mofetil

RNA-seq

RNA sequencing

indels

insertion or deletion nucleotide variants

VLP

virus-like protein

EC50

50% effective concentration

sFPKM

significant fragments per kilobase per million aligned reads

AF

allele frequency

RR

regulatory region

SRA

Sequence Read Archive

DATA AVAILABILITY STATEMENT

Datasets related to this article can be found at: https://www.ncbi.nlm.nih.gov/nuccore/MG674199.1 (HPyV7 complete genome), hosted at Genbank; https://submit.ncbi.nlm.nih.gov/subs/bioproject/SUB6080707/overview (metagenomic data), hosted at NCBI BioProject; https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE148169 (RNA-seq data), hosted at GEO. There will be no restrictions on data availability.