Functional characterization of a multi-cancer risk locus on chr5p15.33 reveals regulation of TERT by ZNF148

Abstract

Genome wide association studies (GWAS) have mapped multiple independent cancer susceptibility loci to chr5p15.33. Here, we show that fine-mapping of pancreatic and testicular cancer GWAS within one of these loci (Region 2 in CLPTM1L) focuses the signal to nine highly correlated SNPs. Of these, rs36115365-C associated with increased pancreatic and testicular but decreased lung cancer and melanoma risk, and exhibited preferred protein-binding and enhanced regulatory activity. Transcriptional gene silencing of this regulatory element repressed TERT expression in an allele-specific manner. Proteomic analysis identifies allele-preferred binding of Zinc finger protein 148 (ZNF148) to rs36115365-C, further supported by binding of purified recombinant ZNF148. Knockdown of ZNF148 results in reduced TERT expression, telomerase activity and telomere length. Our results indicate that the association with chr5p15.33-Region 2 may be explained by rs36115365, a variant influencing TERT expression via ZNF148 in a manner consistent with elevated TERT in carriers of the C allele.

Figure 1. Map of TERT and CLPTM1L region.

Recombination hotspots in the CEU population (red line), as well as 1000G combined recombination rate (blue line) across the TERT/CLPTM1L region are shown relative to the CLPTM1L and TERT genes, as well as the grouping of ten highly correlated sequence variants strongly associated with risk of pancreatic, testicular, and lung cancers in the region closest to CLPTM1L. (a) Chromatin interaction analysis paired-end (ChIA-PET) sequencing data from the K562 chronic myeloid leukaemia cell line using an antibody against RNA polymerase II generated by the ENCODE project (https://www.encodeproject.org/) is shown. For each of the ten strongly associated variants, layered H3K4Me1, H3K4Me3, and H3K27Ac chromatin immunoprecipiation (ChIP-seq), DNAse I hypersensitivity sequencing (DNase) and transcription factor ChIP-seq (TF-ChIP-Seq) data from the ENCODE project are shown (b) as displayed by the UCSC Genome Browser (lower panels).

Figure 2. rs36115365 preferentially binds a nuclear protein.

Electrophoretic mobility shift assays (EMSA) with biotin-labelled oligonucleotides containing either rs36115365-C or rs36115365-G in pancreatic (PANC-1) and lung (A549) cancer cell line nuclear extracts. Two specific protein complexes bind the C allele of rs36115365 preferentially in both cell lines and are more strongly competed with unlabeled C probe as compared to unlabeled G probe. Unlabelled competitor was used at × 10 and × 100 (as indicated by gradient symbol). Arrows denote specific protein complexes bound by the C allele of rs36115365.

Figure 3. Regulatory activity for the genomic region containing rs36115365.

(a) Allele-specific luciferase activity was tested for rs36115365 in four cancer derived cell lines from pancreas (PANC-1) (b) melanoma (UACC903) (c) testis (NTERA-2) and (d) lung (A549). DNA fragments containing either the C or G allele at rs36115365 were cloned upstream of a minimal promoter (pGL4.23) driven reporter in the forward (F) or reverse (R) genomic orientation followed by transient transfections and luciferase assays. Luciferase activity was normalized to empty vector. Experiments were performed in triplicate and repeated at least three times; errors bars represent the standard error of the mean (s.e.m.). Supplementary Fig. 4 contains luciferase results for one additional cell line per tumour type (MIA PaCa-2, UACC1113, NTERA-2, 2102Ep and NCI-H460).

Figure 4. siRNA targeting the region near rs36115365 reduces expression of TERT.

(a) Eight double-stranded siRNAs were designed to target the potential gene-regulatory region encompassing rs36115365. (b) Three of eight double-stranded siRNAs targeting the rs36115365 locus can inhibit the gene expression of TERT but not CLPTM1L, ACTB, and GAPDH when transfected into PANC-1 (pancreatic cancer, upper left), A549 (lung cancer, upper right), NTERA-2 (testicular cancer, lower left) and UACC903 (melanoma, lower right). Expression values were normalized to a scrambled control siRNA. Experiments were performed in triplicate and repeated at least three times; error bars represent standard error of the mean (s.e.m.). (c) Allele-specific TERT expression after siRNA targeting (by siRNA3) of the regulatory region encompassing rs36115365. TERT expression was assayed using a quantitative allelic-discrimination TaqMan assay for a synonymous mRNA coding SNP in the TERT gene (rs2736098) that is genetically linked to rs36115365 (r²=0.14, D′=1.0). Two pancreatic cancer cell lines (Panc 05.04, left; and IMIM-PC-1, right) and one lung cancer line (A549, center) heterozygous for both SNPs were assayed. Expression of the allele of rs2736098 that is linked to the C-allele of rs36115365 (labelled ‘C allele') was reduced to a greater extent than the allele linked to rs36115365-G. The degree of knockdown of each allele was normalized to that from a scrambled siRNA control. Experiments were performed in triplicate and repeated three times. Mean measures for three independent experiments are plotted; error bars represent s.e.m.

Figure 5. ZNF148 is an allele-specific interactor of rs36115365.

(a) DNA pulldowns were performed using PANC-1 (pancreatic cancer) and UACC903 (melanoma) cell line nuclear extracts with 41 bp biotin-tagged rs36115365 bait oligos. Replicate label-swapping experiments were performed with poly-dAdT competitor. Ratios indicate enrichment for protein binding to either the C- or G-allele. Significantly enriched ratios in both a forward and reverse label-swapping experiment were called as outliers (red: C-allele; blue: G-allele. Outlier cutoff is 1.5 IQRs). Identities of C-allele preferential binding proteins identified as outliers. Boxes in boxplots represent first to third quartiles and whiskers extend to furthest data point still within 1.5 IQRs of either quartile. (b) EMSA was performed with oligonucleotides containing the C or G allele at rs36115365 using nuclear extract from PANC-1 pancreatic cancer cells. The left side (lanes 1–10) shows preferred binding of two protein complexes to the C allele of rs36115365; the right side (lanes 11–24) competition of binding to the C allele by antibodies specific for three zinc finger proteins: VEZF1/ZNF161, ZNF148/ZBP-89 and ZNF281. The gradient symbol indicates 1 or 2 μg antibody used for competition. An antibody to ZNF148/ZBP-89 competes efficiently for binding of a protein to the rs36115365-C oligonucleotide. (c) EMSA showing binding of recombinant human ZNF148 to the C allele of rs36115365 (lanes 1–8). A ZNF148 binding site in the CDKN1A/p21 promoter (p21 Motif, M) was used as a positive control (lanes 9–12), as well as PANC-1 nuclear extract (lanes 13–15). Arrows denote proteins bound by the C allele of rs36115365. The gradient symbol indicates addition of 90, 360 or 630 ng of recombinant ZNF148 to the reaction. (d) rs36115365 alters a predicted ZNF148/CACCC-binding protein DNA recognition motif (Transfac motif M00721). (e) Chromatin immunoprecipitation (ChIP) using an antibody against ZNF148 (black bars) in the A549 lung cancer cell line revealed enrichment of ZNF148 binding at rs36115365 (amplicon #6) over background levels using a nonspecific IgG antibody (grey bars). The location of amplicons tested with qPCR relative to rs36115365 are shown underneath the graph (drawn to scale). PCRs were conducted in triplicate and repeated for each of two immunoprecipitations; error bars represent standard deviation (s.d.) for a representative experiment.

Figure 6. ZNF148 depletion results in decreased TERT expression and telomerase activity.

(a) An siRNA directed against the ZNF148 transcript was transfected into A549 (lung, left) and MIA PaCa-2 (pancreas, right) cell lines, and expression of ZNF148, TERT, CLPTM1L, ACTB and GAPDH were assayed by quantitative PCR. Depletion of ZNF148 resulted in consistent reduction of TERT but not CLPTM1L or control gene expression. Expression values were normalized to those from cells transfected with a scrambled control siRNA. Experiments were conducted in triplicate and repeated three times. Mean measures for three independent experiments are plotted; error bars represent s.e.m. (b) siRNAs targeting ZNF148, TERT, the regulatory region encompassing rs36115365 (siRNA3), or a scrambled siRNA control were transfected into A549 (lung, left) and MIA PaCa-2 (pancreas, right) cell lines, and telomerase activity was measured via a telomeric repeat amplification protocol (TRAP). Negative control represents the TRAP assay performed with no cell extracts added. The internal control represents the 36 bp internal standard. (c) siRNAs targeting ZNF148, TERT, or the regulatory region encompassing rs36115365 (siRNA3) were transfected into A549 (lung, left) and MIA PaCa-2 (pancreas, right) cell lines repeatedly (once every four days), and telomere length was measured after 20 days using quantitative PCR for telomere repeat copy number. Data are normalized to those from a scrambled siRNA control. Experiments were conducted in triplicate and repeated three times. Mean measures for three independent experiments are plotted; error bars represent s.e.m.