Allelic effects on KLHL17 expression underlie a pancreatic cancer genome-wide association signal at chr1p36.33

Abstract

Pancreatic Ductal Adenocarcinoma (PDAC) is the third leading cause of cancer-related deaths in the U.S. Both rare and common germline variants contribute to PDAC risk. Here, we fine-map and functionally characterize a common PDAC risk signal at chr1p36.33 (tagged by rs13303010) identified through a genome wide association study (GWAS). One of the fine-mapped SNPs, rs13303160 (OR = 1.23 (95% CI 1.15-1.32), P-value = 2.74×10^-9, LD r² = 0.93 with rs13303010 in 1000 G EUR samples) demonstrated allele-preferential gene regulatory activity in vitro and binding of JunB and JunD in vitro and in vivo. Expression Quantitative Trait Locus (eQTL) analysis identified KLHL17 as a likely target gene underlying the signal. Proteomic analysis identified KLHL17 as a member of the Cullin-E3 ubiquitin ligase complex with vimentin and nestin as candidate substrates for degradation in PDAC-derived cells. In silico differential gene expression analysis of high and low KLHL17 expressing GTEx pancreas samples suggested an association between lower KLHL17 levels (risk associated) and pro-inflammatory pathways. We hypothesize that KLHL17 may mitigate cell injury and inflammation by recruiting nestin and vimentin for ubiquitination and degradation thereby influencing PDAC risk.

Fig. 1. Overview of chr1p36.33 PDAC risk locus.

a Locus Zoom plot of the variants identified in the meta-analysis, colors are indicative of the linkage disequilibrium (LD) r² in reference to the tag SNP (red 0.8 < r² ≤ 1.0; yellow 0.6 < r² < 0.8, green 0.4 < r² < 0.6, blue 0.2 < r² < 0.4, purple 0 < r² < 0.2), -log10 P-values are calculated using a logistic regression analysis and are not multiple-testing corrected. The gray-dashed line indicates the Bonferroni-corrected P-value significance threshold (5×10^-8) used in GWAS; b UCSC genome browser view of chr1p36.33 with ChromHMM and ATAC-seq annotations in PDAC and normal-derived duct epithelial cell lines; c Zoomed in UCSC browser snapshot showing the candidate functional variants and nearby genes. ChromHMM states are indicated by color: weak enhancer 1 (light green), weak enhancer 2 (dark blue), active enhancer (light blue), active element (purple), active transcriptional start site (TSS, red), bivalent TSS (yellow), polycomb repressed (light purple), quiescent (dark green). Source data are provided as a Source Data file.

Fig. 2. Identification of allele-preferential binding and activity using EMSA and luciferase reporter assays.

a–c Representative EMSA results with PANC-1 nuclear extract and fluorescently labeled 31 bp oligonucleotides with each variant, rs13303010 (n = 4 independent experiments), rs13303327 (n = 3 independent experiments), rs13303160 (n = 4 independent experiments), respectively, centered in the middle. Competitor is the same sequence with no fluorescent label in excess (50, 100X, indicated by the black triangles). Arrows denote the allele-preferential binding. d–f Luciferase reporter assays using rs13303010, rs13303327, and rs13330160, respectively and the surrounding sequence as a promoter to the luciferase gene in three cell lines, number of biological replicates are indicated below the cell line name. g Luciferase assay for rs13303160 and surrounding sequence as an enhancer upstream of a minimal promoter and luciferase gene in three cell lines. For EMSA and luciferase, the risk allele is colored in red. For luciferase, the forward (fwd) and reverse (rev) orientation of the sequence was used. Pink bars indicate the risk allele and blue bars the protective allele. Gray bars are the Empty Vector (EV) control. The number of biological replicates is indicated underneath the cell line in the figure for (d–g). Error bars represent the standard error of the mean (SEM) and significance was determined by an unpaired, two-tailed t-test. Source data are provided as a Source Data file.

Fig. 3. Allele-preferential binding of ELF to rs13303327 in vitro and in vivo.

a in silico transcription factor (TF) binding motif prediction for rs13303327. The risk G allele disrupts the motif. b Representative EMSA images with increasing amounts of recombinant ELF proteins (indicated by black triangles) and fluorescently labeled oligonucleotide (n = 4 (ELF1), 4 (ELF2), 3 (ELF3), 3 (ELF4) individual experiments); arrow denotes binding of interest. The risk allele is denoted in red. c Representative EMSA supershift with an antibody against ELF2 using PANC-1 nuclear lysate (n = 3 biological replicates). The risk allele is indicated in red. d ChIP-qPCR for ELF2 in Hs766T PDAC cell line with primers near or encompassing the SNP. The positive control is a documented region from an ELF2 ChIP-seq in K562 (ENCODE) and negative control is from a quiescent region of chr1p36.33. Percent input enrichment was quantitatively determined using a standard curve derived from input DNA as described in the ActiveMotif protocol. Gray bars indicate IgG, blue bars indicate ELF2, n = 4 biological replicates. e TaqMan genotyping of enriched ChIP-qPCR DNA; A to G ratio was calculated relative to the input ratio. Gray bar is input and blue bar is ELF2 IP, n = 4 biological replicates. Error bars represent SEM. All statistical tests were unpaired, two-tailed t tests were performed. Source data are provided as a Source Data file.

Fig. 4. Allele-preferential binding of AP-1 proteins to rs13303160 in vitro and in vivo.

a in silico Transcription factor binding predictions; the risk (G) allele disrupts the motif. b Representative EMSA using TPA-stimulated nuclear HeLa extract and fluorescently labeled oligonucleotide (n = 3 independent experiments). Arrow indicates the allele-preferential binding. c Luciferase reporter assay using DMSO and TPA stimulation and the rs13303160 sequence as an enhancer in the PANC-1 cell line; luciferase activity reported relative to the Empty Vector (EV, gray bars). Pink bars denote the risk allele, and blue bars represent the protective allele. Both alleles were tested in the forward (fwd) and reverse (rev) orientations. Unpaired, two-tailed t tests were performed on the relative luciferase activity of the A/G ratio compared to A/A; n = 6 biological replicates. d Representative EMSAs with increasing amounts of recombinant Fos proteins (from left to right: c-Fos (n = 3 independent experiments); FosB (n = 2); Fos1L (n = 1); Fos2L (n = 1)). e Representative EMSAs with increasing amounts of recombinant Jun proteins (from left to right: c-Jun (n = 2 independent experiments), JunB (n = 2), JunD (n = 2)). Arrow indicates the allele-specific binding. f, g Representative supershift EMSA with antibodies against JunB and JunD (n = 3 independent experiments each), respectively, using both TPA-stimulated nuclear lysate and recombinant protein; Arrows denote the shift in the bands. h ChIP-qPCR in SW1990 PDAC cells for JunB (denoted in blue, IgG in gray) (n = 6 biological replicates) and JunD (denoted in blue, IgG in gray) (n = 4 biological replicates) using 3 primer sets (PS) surrounding the SNP. Positive control (PC) is from a JunB ChIP-seq performed in the CFPAC1 PDAC cell line. Negative control (NC) is from a quiescent region on chr1p36.33; i TaqMan genotyping assay for rs13303160 using immunoprecipitated DNA from the ChIP. The ratio of A to G was determined relative to the quantity of A and G alleles in the input DNA (gray) (n = 6 biological replicates for JunB; n = 5 biological replicates for JunD, blue bars). Red font denotes the risk allele in (b–g). For all graphs, error bars represent the SEM. Unpaired two-tailed t tests were performed. Source data are provided as a Source Data file.

Fig. 5. Analysis of the effects of altered KLHL17, a Cullin3-E3 complex member, expression on cellular growth of PDAC cells in vitro.

a Pancreas GTEx v8 eQTLs for rs13303010, n = 20 (GG), 67 (GA), 218 (AA) individuals from the GTEx portal. Red denotes the risk allele. Linear regression using FastQTL was performed to identify the eQTLs and calculate the P-values (details on the GTEx webpage). The violin plots display the density distribution of the data with the white line indicating the median normalized expression, and the gray box displaying the interquartile range. b Representative immunofluorescence for KLHL17 in the MIA PaCa-2 (n = 2) and PANC-1 (n = 2) KLHL17 overexpressing cell lines. Scale bar = 100 micron. c Peptide Spectra Matches for Cullin3-E3 members identified by KLHL17-FLAG immunoprecipitation (n = 3 independent KLHL17-FLAG IPs, n = 4 independent Empty Vector IPs, ran multiplexed across 2 mass spectrometry runs) and mass spectrometry analysis. d, f Cell counts normalized to 0 h for CRISPRi-mediated knockdown of KLHL17 in PANC-1 (n = 4 biological replicates) and MIA PaCa-2 (n = 4 biological replicates) cells, respectively (left panel). sgNeg (black line) is the negative control targeting a sequence within the same topologically associated domain (TAD). Pink line represent three different sgRNAs; e,g) qPCR analysis of CRISPRi sgRNA efficiency; expression is relative to the sgNeg control and internal HPRT control (right panel) and measured at the start and end of the growth assay (n = 4 biological replicates). h, i Cell counts normalized to 0 h for doxycycline-inducible KLHL17-FLAG overexpressing (pink) and empty vector (EV, gray) control MIA PaCa-2 (n = 4 biological replicates) and PANC-1 (n = 4 biological replicates) cell lines, respectively. For panels (c–g), error bars indicate the SEM; For panels (c–g), unpaired, two-tailed t tests were performed relative to the empty vector IP (c) or sgNeg (d–g). Source data are provided as a Source Data file.

Fig. 6. Gene Set Enrichment Analysis (GSEA) of differentially expressed genes from an in silico KLHL17 knockdown.

a GSEA using the KEGG functional database and the significantly (FDR < 0.05) differentially expressed genes when KLHL17 expression is lower. b GSEA using the Reactome dataset. c GSEA using Biological Processes Gene Ontology dataset. For all three analyses only gene sets with a False Discovery Rate (FDR) < 0.1 are shown. Bars in orange indicate gene sets associated with inflammation or inflammation-related diseases. No gene sets were negatively enriched at this FDR threshold. Source data are provided as a Source Data file; d Working hypothesis for the function of rs13303160 and KLHL17 in PDAC risk. Created in BioRender. Connelly, K. (2025) https://BioRender.com/t14y333.