Page 72 - Read Online

Page 6 of 15 Miliotis et al. J Cancer Metastasis Treat 2020;6:13 I http://dx.doi.org/10.20517/2394-4722.2020.12

chromosomal regions of the host genome. They identified EBV integration events in 25% (10/39) of the
GC samples analyzed, with some of the integration breakpoints mapping close to known tumor suppressor
[38]
genes . None of their samples had integrated virus close to PD-L1, but studies with bigger EBVaGC
sample sizes are necessary to identify the frequency and significance of viral integration in or close to the
PD-L1 locus. Viral integration in the host genome has been associated with increased PD-L1 expression in
[20]
other virus-associated cancers. Cao et al. identified integrated HPV genomes in the PD-L1 or PD-L2 loci
in three cases from the TCGA Head and Neck Squamous Cell Carcinoma cohort and showed that these
integration events correlated with elevated PD-L1 and PD-L2 expression.

3’-UTR structural variations
SVs in the PD-L1 3’-UTR have also been associated with increased PD-L1 expression in EBVaGC [34,37] .
[37]
Kataoka et al. analyzed RNAseq data from all TCGA cancer types and searched for 3’-UTR disruptions
in PD-L1. The authors identified PD-L1 3’-UTR truncations in 31/10,210 cancer cases and showed that they
correlated with high PD-L1 expression. The highest frequency of 3’-UTR truncations was found in DLBCL
(4/48) and GC (9/415), with a third of the GC samples (3/9) being EBV-positive. Therefore, around 10% of
EBVaGC samples in TCGA were found to have PD-L1 3’-UTR SVs [34,37] . In a follow-up study, Kataoka et al.
[34]
analyzed samples from multiple EBV-associated lymphomas and found that PD-L1 3’-UTR SVs were
significantly more common in EBV-positive compared to EBV-negative lymphomas. They report that
PD-L1 3’-UTR genomic truncations in cell lines and mouse models promote PD-L1 overexpression and
immune evasion, consistent with the patient data .
[37]
POST-TRANSCRIPTIONAL REGULATION
The 3’-UTR contains sequences or structural regions, called regulatory elements, that are important for the
post-transcriptional regulation of a gene. These regulatory elements control binding to miRNAs and RNA-
[39]
binding proteins (RBP), which influence mRNA stability, translation rate, and localization . miRNAs are
short non-coding RNAs that silence gene expression by binding to complementary sequences in the 3’-UTR
of target mRNAs. miRNA-mRNA binding usually triggers mRNA degradation or blocks translation. The
fact that 3’-UTR shortening has such a profound effect on PD-L1 expression in multiple cancers indicates
that PD-L1 is under tight post-transcriptional control .
[37]

3’-UTR short variations
Mutations in the 3’-UTR have the capacity to remove existing or create new binding sites for miRNAs
and RBPs. Some germline and somatic mutations in the 3’-UTR of PD-L1 have been shown to correlate
[43]
with PD-L1 expression in gastric and other cancers [40-44] . Wu et al. analyzed 728 GC samples and found
that the AA and AG genotypes in rs2297136, a germline single nucleotide polymorphism (SNP) located
in the 3’-UTR of PD-L1, were associated with lower PD-L1 protein levels. They reported that the miRNAs
miR-324-5p and miR-362 are predicted to bind to that region of the PD-L1 3’-UTR, but no validation
[44]
experiments were pursued. Wang et al. polymerase chain reaction (PCR)-amplified and sequenced the
3’-UTR of PD-L1 in hundreds of GC and matched normal samples and identified a frequent guanine-
to-cytosine somatic mutation that correlated with increased PD-L1 protein expression. It was shown
that this mutation maps to a seed-binding region for miR-570 and it was proven experimentally that it
[44]
increases PD-L1 expression by disrupting miR-570 binding . To date, most studies looking at PD-L1 3’-
UTR mutations have been low-throughput, with small sample sizes or targeted on specific SNP locations.
There has not been a comprehensive study looking at the frequency and effect of all possible somatic and
germline variants in the PD-L1 3’-UTR in EBVaGC or other EBV-associated cancers. The fact that SVs in
the PD-L1 3’-UTR appear to occur more frequently in EBV-positive than EBV-negative cancers raises the
question of whether short variants in the 3’-UTR could be an alternative or parallel mechanism for PD-
L1 overexpression. Large-scale variant calling studies in gastric and other cancers, including the TCGA
somatic mutation data, have mostly relied on whole exome sequencing data and exclude 3’-UTR sequences.
This has created a gap in our understanding of 3’-UTR variations in cancer in general.