Human endogenous retrovirus onco-exaptation counters cancer cell senescence through calbindin

Increased levels and diversity of human endogenous retrovirus (HERV) transcription characterize most cancer types and are linked with disease outcomes. However, the underlying processes are incompletely understood. Here, we show that elevated transcription of HERVH proviruses predicted survival of lung squamous cell carcinoma (LUSC) and identified an isoform of CALB1, encoding calbindin, ectopically driven by an upstream HERVH provirus under the control of KLF5, as the mediator of this effect. HERVH-CALB1 expression was initiated in preinvasive lesions and associated with their progression. Calbindin loss in LUSC cell lines impaired in vitro and in vivo growth and triggered senescence, consistent with a protumor effect. However, calbindin also directly controlled the senescence-associated secretory phenotype (SASP), marked by secretion of CXCL8 and other neutrophil chemoattractants. In established carcinomas, CALB1-negative cancer cells became the dominant source of CXCL8, correlating with neutrophil infiltration and worse prognosis. Thus, HERVH-CALB1 expression in LUSC may display antagonistic pleiotropy, whereby the benefits of escaping senescence early during cancer initiation and clonal competition were offset by the prevention of SASP and protumor inflammation at later stages.


Supplemental Figure 4. HERVH-dependent expression of CLAB1 in LUSC cell lines. (A)
HERVH-CALB1 expression, assessed by analysis of RNA-seq data, in the indicated LUSC cell lines from CCLE. (B) Diagram depicting the location of PCR primers (black arrows, pointing the end position of each primer) used for the amplification of the canonical CALB1 transcript (exons 1 and 2), the HERVH-CALB1 transcript (HERVH and exon 2) or both (exons 7-10) (left), and expression of the canonical CALB1 or HERVH-CALB1 transcripts, relative to expression of HPRT, determined by RT-qPCR in the indicated LUSC cell lines (right). (C) Diagram depicting the position of the gRNAs (black arrows) used by Cas9-mediated deletion of the HERVH provirus upstream of the CALB1 gene and resulting alleles in LK-2 clones 6E8 and 6A8 (left), and expression of the canonical CALB1 or HERVH-CALB1 transcripts, relative to expression of HPRT, determined by RT-qPCR in the same LK-2 clones and parental LK-2 cells (right). Clone 6E8 sustained a near complete deletion of the 5.4 kb HERVH provirus (with a residual 0.5 kb segment) and clone 6A8 showed extensive deletions with inversion of a residual 0.6 kb and 1.5 kb segments in separate alleles, respectively. In both clones, expression of the HERVH-CALB1 transcript, which accounted for over 90% of total CALB1 transcription in these cells, was reduced by over 1 log.
Average scaled fraction of RNA-seq reads per cell

CALB1-expressing tumours
Average scaled fraction of RNA-seq reads per cell Fibroblast

Myeloid cells Cancer cells
Supplemental Figure 17. HERVH-CALB1 expression correlates with cancer cell-intrinsic chemokine production. Correlation between expression of CALB1 and of the indicated chemokine/cytokine, in RNA-seq from lung squamous or adenosquamous cell lines from CCLE (n=53). Blue and red colors indicate significant positive and negative correlation, respectively, and grey color indicates lack of significant correlation.

Supplemental methods
Single-cell RNA-seq data analyses. All four cell populations (HARA cells, Calbindindeficient HARA 3D5 cells, HARA.LTR7-GFP + and HARA.LTR7-GFPcells) were processed as one batch at the same time, with cell viability between 89% and 93%. Cells were dispersed into single cell droplets using the 10× Genomics Chromium platform, and RNA was isolated and reverse transcribed with the 10× 3'-mRNA Seq (v3)   3D collagen matrix invasion assay. Collagen matrices were constructed as previously described (86). Briefly, monomeric collagen type I of rat tail origin (First Link, was used at an initial 2.0 mg/ml concentration. Antibody (Abcam, Cat #ab150077) was carried out the next day for 2.5 hours at room temperature at a 1:1,000 dilution. F-actin was visualized using AlexaFluor568 phalloidin (Invitrogen, Cat #A22287) at 165 nM followed by 4′,6-diamidino-2-phenylindole (DAPI) counterstain, using NucBlue (Invitrogen, Cat #R37606) at 2 drops/ml for 15 min. Samples were imaged on the Zeiss Observer.Z1 (Carl Zeiss Meditec AG) using Micro-Manager 2.0 software. For enumerating γH2AX stained cells, cells exhibiting DNA damage were identified using a combination of morphological criteria (enlarged cell size and flattened cell shape) and distinctive γH2AX staining (strong pan-nuclear staining, chromatin reorganization and visible foci).
Anti-Calbindin antibody specificity. The anti-Calbindin antibody used here for immunocytochemistry, immunofluorescence and Western blotting (Atlas Antibodies, Cat #HPA023099) is a rabbit polyclonal antibody raised against an immunogen corresponding to amino acid residues 7-96 of the full-length canonical Calbindin. Owing to the N-terminal deletion of amino acid residues 1-57 in the HERVH-driven Calbindin isoform, the latter isoform overlaps with the immunogen by only 39 of a total of 90 amino acid residues, whereas the canonical Calbindin isoform overlaps completely. As a result, although reactive with both isoforms, this polyclonal antibody detects the canonical Calbindin isoform much more efficiently than the HERVH-driven Calbindin isoform.
Reverse transcriptase-based quantitative PCR (RT-qPCR). RNA was extracted from pelleted cells using QIAshredder columns (Qiagen) using the RNeasy kit (Qiagen), and cDNA was synthesized using the Maxima First Strand cDNA Synthesis Kit (Thermo Fisher Scientific).