Evidence for Multiple Subpopulations of Herpesvirus-Latently Infected Cells

Abstract

Kaposi's sarcoma-associated herpesvirus (KSHV)-associated primary effusion lymphomas (PEL) are traditionally viewed as homogenous regarding viral transcription and lineage of origin, but so far this contention has not been explored at the single-cell level. Single-cell RNA sequencing of latently infected PEL supports the existence of multiple subpopulations even within a single cell line. At most 1% of the cells showed evidence of near-complete lytic transcription. The majority of cells only expressed the canonical viral latent transcripts: those originating from the latency locus, the viral interferon regulatory factor locus, and the viral lncRNA nut-1/Pan/T1.1; however, a significant fraction of cells showed various degrees of more permissive transcription, and some showed no evidence of KSHV transcripts whatsoever. Levels of viral interleukin-6 (IL-6)/K2 mRNA emerged as the most distinguishing feature to subset KSHV-infected PEL. One newly uncovered phenotype is the existence of BCBL-1 cells that readily adhered to fibronectin and that displayed mesenchymal lineage-like characteristics. IMPORTANCE Latency is the defining characteristic of the Herpesviridae and central to the tumorigenesis phenotype of Kaposi's sarcoma-associated herpesvirus (KSHV). KSHV-driven primary effusion lymphomas (PEL) rapidly develop resistance to therapy, suggesting tumor instability and plasticity. At any given time, a fraction of PEL cells spontaneously reactivate KSHV, suggesting transcriptional heterogeneity even within a clonal cell line under optimal growth conditions. This study employed single-cell mRNA sequencing to explore the within-population variability of KSHV transcription and how it relates to host cell transcription. Individual clonal PEL cells exhibited differing patterns of viral transcription. Most cells showed the canonical pattern of KSHV latency (LANA, vCyc, vFLIP, Kaposin, and vIRFs), but a significant fraction evidenced extended viral gene transcription, including of the viral IL-6 homolog, open reading frame K2. This study suggests new targets of intervention for PEL. It establishes a conceptual framework to design KSHV cure studies analogous to those for HIV.

Keywords: Castleman’s disease; KSHV; Kaposi’s sarcoma; Kaposi’s sarcoma-associated herpesvirus; herpesvirus; lymphoma; primary effusion lymphoma; scRNAseq; single cell; single-cell RNA-seq.

FIG 1

Cell growth characteristics and adhesion phenotype of BCBL1 cells. (A) Shown are growth curves of BCBL-1 at consecutive passages and the linear fit across all the data (black) (n = 13 passages). Log₁₀ value of cell concentration, in cells per milliliter, is shown on the vertical axis and days postseeding on the horizontal axis. Cells were seeded at 0.25 × 10⁶ cells/ml on day 0. (B and C) Shown is an image (200×) of cultures of either BCBL1 (B) or JSC1 (C) cells that were grown on horizontal fibronectin-coated flasks. The yellow arrows point to some spindle-shaped cells that are flattened out and tightly attached to the fibronectin matrix.

FIG 2

Heatmap of KSHV transcription in PEL. Shown are the results of unsupervised hierarchical clustering of PEL cells (columns) by KSHV genes (rows). Genes are arranged in order from 5′ (bottom) to 3′ (top). A dendrogram on top represents the clustering results, which were based on the binary matrix of detected/undetected genes. A gene was detected if it had at least 2 UMI counts and undetected otherwise. The top heatmap represents this binary signal, whereby detected genes are labeled in blue and nondetected genes are labeled in yellow. The lower heatmap represents the KSHV transcription pattern after log normalization. With this, we can visualize how many genes possessed exactly 1 count.

FIG 3

KSHV-based clustering of PEL. (A) Distribution of the detection rates (nonzero fraction) on the horizontal axis. This is a measure of the zero inflation for each gene. (B) Principal-component analysis (PCA) of KSHV genes. LANA, nut-1, and the vIRF have the most distinct expression patterns, and LANA and nut-1 were the most abundantly expressed genes (PC1). (C and D) The same UMAP plot of all cells. In panel C, the cells are color coded by the plate/library number (plate 1, light blue; plate 2, dark blue; plate 3, light green; plate 4, dark green; plate 5A, light red; plate 5B, dark red; plate 6A, light orange; plate 6B, dark orange; plate 7, light purple; plate 8, dark purple; plate 9, yellow; plate 10, brown), and in panel D the cells are color coded by the cluster designation as identified by unsupervised analysis using K-medoids. Cluster identification and per-plate saturation can be found in Table 3. (E) Mean ± interquartile range (IQR)-adjusted counts for the latency locus in each of the clusters, i.e., the relative expression of LANA and the same for vIL-6 (F) and nut-1 (G).

FIG 4

Association between KSHV and host cell transcription. Heatmap of cellular transcription of BCBL-1 cells after batch correction. The associated SigClust dendrogram is above with annotations of which clusters are significant. The KSHV K-means clustering assignments and cell group identity are mapped between the heatmap and the dendrogram. Panel B shows a volcano plot of genes that were differentially regulated between BCBL-1 and JSC-1 cells, panel C shows a volcano plot of genes that signify differences at the major split (P ≤ 0.001), panel D shows a volcano plot of the left minor split, and panel E shows a volcano plot of the right minor split (P ≤ 0.01). The individual genes are listed in Table S1.