Genomic rearrangements involving programmed death ligands are recurrent in primary mediastinal large B-cell lymphoma

Primary mediastinal large B-cell lymphoma (PMBCL) is an aggressive disease known to share certain genotypic and phenotypic features with classic Hodgkin lymphoma (CHL) and diffuse large B-cell lymphoma (DLBCL).^1,2  However, the complete landscape of genetic alterations involved in PMBCL pathogenesis has yet to be fully elucidated.³  Among the most common chromosomal alterations in PMBCL are amplifications of chromosome 9p and translocations involving CIITA (16p13.13).^4-7  These aberrations have been suggested to affect tumor-microenvironment interactions resulting in immune privilege.^7,8  Here, we demonstrate that rearrangements involving immune cell anergy-inducing programmed death ligand (PDL) 1 (CD274) and 2 (PDCD1LG2) are recurrent in and characteristic of PMBCL. Furthermore, we show such rearrangements are correlated with elevated transcript levels, and we characterize novel translocations identified using high-throughput sequencing.

We studied 571 primary B-cell lymphoma samples in conjunction with 17 established B-cell-derived cell lines. Using in-house bacterial artificial chromosome probes, fluorescence in situ hybridization (FISH) or FISH combined with CD30 immunofluorescence (in the case of CHL specimens) was performed to characterize the PDL locus.^7,9,10  These cases were also analyzed with Epstein-Barr virus (EBV)-encoded RNA in situ hybridization. Quantitative real-time polymerase chain reaction (qRT-PCR) was performed for CD274 and PDCD1LG2 transcript expression on a subset of cases from the FISH cohort (N = 76). Surface PDCD1LG2 expression of the cell lines was determined via flow cytometry. To characterize the Hodgkin lymphoma cell lines found to be rearranged by FISH, 1 whole-transcriptome sequencing (RNA-seq) library (CHL-derived L-428) was reanalyzed, and 1 new whole-genome library (L-428) and 2 new RNA-seq libraries (CHL-derived L-1236 and nodular lymphocyte predominant Hodgkin lymphoma-derived DEV) were sequenced.⁷  This study was approved by the BC Cancer Agency (REB: H11-00684) and conducted in accordance with the Declaration of Helsinki.

There is increasing evidence that genes involved in tumor-microenvironment interactions play a crucial role in the pathogenesis of B-cell lymphomas.⁸  Two such genes, CD274 and PDCD1LG2, have been implicated in promoting tumor cell immune evasion.^11,12  These ligands have been reported to be overexpressed in solid tumors, including lymphomas, but the mechanisms by which overexpression occurs are incompletely understood.^4,7,13-15  Using an in-house break-apart FISH assay (supplemental Figure 1; supplemental Table 1), we determined the frequency of PDL locus aberration in 17 lymphoma-derived cell lines (supplemental Table 2) and 571 clinical lymphoma samples (Figure 1). This analysis revealed a break-apart frequency of 20% and an amplification frequency of 29% in 125 PMBCL cases. The prevalence of PMBCL break-apart events was significantly higher than in DLBCL, PCNSL, TDLBCL, and FL (P < .05). No significant differences of any noted clinical parameters, including treatment outcome variables (progression-free and overall survival), were observed between PDL rearranged, amplified, and nonrearranged cases.

To study the correlation of genomic rearrangements and copy-number changes of the PDL locus with gene expression, we performed qRT-PCR on 17 cell lines and 76 clinical cases that had fresh frozen material available (supplemental Table 3). These cases were stratified according to the genomic aberration status as determined by FISH. In break-apart positive cases across all lymphomas, PDCD1LG2 transcript levels were significantly higher as compared with neutral (P = .0003), gained (P = .001), and amplified (P = .005) loci (Figure 1; supplemental Table 4). A significant difference was also observed between CD274 transcript levels of rearranged and copy-neutral cases (P = .03), although only a trend was found when compared with copy-number aberrated cases. This reduced level of correlation may be attributable to the heterogeneity in rearrangement anatomy, which disparately affects the PDL loci. The observation that CD274 and PDCD1LG2 expression levels were also elevated (relative to PDL expression in reactive tonsil cells) in cases where no rearrangement could be observed via FISH, suggests alternative mechanisms of deregulation exist. These mechanisms might include epigenetic and microRNA regulatory factors.^16,17  As EBV infection has been implicated in PDL expression, we performed EBV-encoded RNA in situ hybridization (supplemental Table 1).¹⁸  No significant correlation was observed between EBV positivity and either PDL–break-apart or transcript levels, although the small number of EBV-positive cases in our study limits statistical power.

To determine the rearrangement structures that were observed at the PDL locus, we reanalyzed previously published RNA-seq data of L-428.⁷  Additionally, we sequenced the whole genome of L-428 and the whole transcriptomes of L-1236 and DEV. Using genomic breakpoint and fusion transcript predicting algorithms, we mined these data with a specific focus on the 9p24.1 locus (supplemental Tables 5 and 6). We identified 3 novel fusion transcripts within this locus that were validated by Sanger sequencing: CIITA-PDCD1LG2 (DEV), KIAA1432-CLDN14 (L-428), and PDCD1LG2-IGHV7-81 (L-1236) (Figure 2). The fusion transcript observed in DEV was found to be similar to 2 fusions that we had previously reported in clinical PMBCL cases.⁷  Although clearly involving the 9p24.1 locus, the FISH break-apart signal pattern of L-428 was attributed to a translocation 60 kb downstream of PDCD1LG2. The mechanism by which this translocation increases PDL expression is unknown; although KIAA1432-JAK2 fusions have been recently implicated in small cell lung cancer.¹⁹  The chimeric transcript produced by L-1236 was predicted to generate a C-terminal-truncated PDCD1LG2 protein. In this instance, the receptor binding site, coded in the fourth exon, was left intact.²⁰ 

To investigate the effect of these rearrangements on protein expression, we performed flow cytometry on several cell lines with a specific focus on those found harboring 9p24.1 translocations: L-428, L-1236, and DEV. We found elevated expression of PDCD1LG2 in all 3 rearranged cell lines compared with U-HO1 and DOHH-2 (normal loci) (supplemental Figure 2). Furthermore, expression levels in L-428 and L-1236 were determined to be in agreement with the literature.⁴ PDCD1LG2 protein surface levels were highest in DEV and were also appreciably greater than in cell line MEDB-1 (amplified locus). This shows that a PDCD1LG2 epitope is readily detectable in DEV cells at elevated levels corresponding to a CIITA-PDCD1LG2 fusion that may alter T-cell activity states in the microenvironment, as previously demonstrated.⁷ 

By merging the data of this study with the 3 PDL fusions previously reported by our group (Figure 2), we have validated the presence of 5′ fusion transcripts involving PDL in immortalized cells, and we have established the existence of a novel fusion transcript that involves the 3′ end of PDCD1LG2.⁷  As substantially elevated transcript levels were observed in all instances with validated direct involvement of PDL rearrangements, upregulated expression of PDLs may be a consistent mechanism that contributes to pathogenesis. Deregulated transcript expression is believed to arise via juxtaposing either a highly active promoter (CIITA) or enhancing elements (IGHV7-81) adjacent to PDLs.^7,21,22  Other potential selective advantages that these translocations confer may be related to the altered function of the involved fusion partner genes. This would include decreased tumor immunogenicity as a result of reduced major histocompatibility complex II expression from CIITA-associated translocations.⁷  Another potential benefit specific to the 3′ translocation in L-1236 is the loss of the transcript 3′ untranslated region. This may lead to a loss of microRNA binding sites increasing the half-life of PDL transcript and/or the loss of the exons coding for the transmembrane domain region, which may act to solubilize the protein.^17,23,24 

Taken together, our findings establish that rearrangement of the PDL locus is recurrently selected for and that such rearrangements lead to elevated transcript levels and/or the production of chimeric PDL fusion transcripts. Recently, phase 1/2 clinical trials targeting the PDL pathway in solid tumors have shown promising outcomes.^14,15,25  Our data strongly suggest that PDLs and their high-affinity receptor, programmed death 1 (PDCD1), are rational drug targets in a subgroup of B-cell lymphomas characterized by PDL locus aberrations.

This work was supported by grants from the Canadian Institutes of Health Research (111043) (R.D.G.) and the Terry Fox Research Institute (Terry Fox New Frontiers Program in Cancer [1023]) (C.S.), and generous support from the British Columbia Cancer Foundation.

C.S. and S.P.S. are recipients of Career Investigator Awards from the Michael Smith Foundation for Health Research; A.M. is supported by a Mildred-Scheel-Cancer-Foundation Fellowship; and D.D.W.T. is supported by UBC fellowships and the Canadian Hematology Society.

Contribution: D.D.W.T. designed and performed research, and interpreted data; F.C.C., R.S.L., and A.W.M. analyzed sequencing data and contributed to the figures; S.B.-N., B.W.W., K.L.T., A.M., and G.W.S. produced and analyzed data; J.G., R.K., and A.T. assisted with data collection; D.W.S., K.J.S., and S.P.S. provided editorial input; R.D.G. analyzed data, constructed the database, and edited the paper; C.S. designed research, analyzed data, and approved the paper; and D.D.W.T. and C.S. wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Christian Steidl, Department of Experimental Therapeutics, BC Cancer Research Centre, 675 West 10th Ave, Vancouver, BC, Canada, V5Z 1L3; e-mail: csteidl@bccancer.bc.ca.