Functional consequences of PRPF39 on distant genes and cisplatin sensitivity

Abstract

Variation in gene expression has been found to be important in disease susceptibility and pharmacogenomics. Local and distant expression quantitative trait loci (eQTLs) have been identified via genome-wide association study (GWAS); yet the functional analysis of these variants has been challenging. The aim of this study was to unravel the functional consequence of a gene with a local SNP with evidence for local and distant regulatory roles in cellular sensitivity to cisplatin, one of the most widely used chemotherapeutic drugs. To this end, we measured cellular susceptibility to cisplatin in 176 HapMap lymphoblastoid cell lines derived from Yoruba individuals from Ibadan, Nigeria. The 276 cytotoxicity-associated SNPs at the suggestive threshold of P ≤ 0.0001 were significantly enriched for eQTLs. Of these SNPs, we found one intronic SNP, rs17115814, that had a significant relationship with the expression level of its host gene, PRPF39 (P= 0.0007), and a significant correlation with the expression of over 100 distant transcripts (P ≤ 0.0001). Successful knockdown of PRPF39 expression using siRNA resulted in a significant increase in cisplatin resistance. We then measured the expression of 61 downstream targets after PRPF39 knockdown and found 53 gene targets had significant (P ≤ 0.05) expression changes. Included in the list of genes that significantly changed after PRPF39 knockdown were MAP3K4 and TFPD2, two important signaling genes previously shown to be relevant in cisplatin response. Thus, modulation of a local target gene identified through a GWAS was followed by a downstream cascade of gene expression changes resulting in greater resistance to cisplatin.

Figure 1.

Cisplatin IC₅₀ significantly correlates with the local eQTL rs17115814 and PRPF39 expression. (A) Genome-wide association results of 176 YRI LCLs were analyzed to identify local eQTLs at P< 0.01, defined as being within the gene, 2 kb of the 5′ end of the gene, or 0.5 kb of the 3′ end of the gene. Local eQTLs were then analyzed as to whether they were distant eQTLs for more than 10 target genes at P< 0.0001. Finally, the host gene's relationship was evaluated for correlation with cisplatin cytotoxicity at P< 0.05. Cisplatin IC₅₀ is significantly correlated with rs17115814 with an empirical P= 9.99 × 10⁻⁵ (B) as well as with PRPF39 expression at P= 0.01 (C). The intronic SNP is associated with PRPF39 expression at P= 0.0007 (D).

Figure 2.

PRPF39 expression knockdown after siRNA nucleofection. Four different LCLs were evaluated for PRPF39 knockdown after siRNA treatment. Approximately 20% of PRPF39 expression remained 5 h after siRNA was introduced through nucleofection relative to a non-targeting scrambled control siRNA. Expression of PRPF39 rebounded after 29 and 53 h to >75% relative to a non-target control. All values are based on a minimum of three different nucleofections.

Figure 3.

PRPF39 knockdown significantly increases the resistance of four cell lines to cisplatin. After successful PRPF39 knockdown using siRNA, cisplatin cytotoxicity was assessed using the alamarBlue growth inhibition assay. For four different LCLs, IC₅₀ values after successful knockdown of PRPF39 are significantly higher relative to a non-targeting control (19100 P= 0.025; 19252 P= 0.017; 18925 P= 0.069; 18488 P= 0.067), demonstrating that after knockdown cell lines are more resistant to cisplatin. P-values are from a Student's one-tailed t-test based on a minimum of three separate knockdown experiments. When incorporating the four different cell lines into a single mixed effects model, the P-value drops to 0.0001, with the IC₅₀ of the knockdown being ∼34% above the non-targeted control.

Figure 4.

Gene expression changes combined across three cell lines. (A) Change in gene expression 5, 12, 18 and 24 h after PRPF39 knockdown was assessed for each of the 61 target genes, using a mixed effects model that combines all expression data from three cell lines. The X-axis shows fold change relative to a scrambled control, with identical expression having a value of 1. The Y-axis shows the negative log₁₀ of the P-value from the combined model, with P= 0.05 shown with a dashed line. Identification of each gene and the direction of effect are found in Supplementary Material, Table S2. (B) The panel shows that above the time point, the genes significantly (P< 0.05) increased in expression, with the gene closest to the center being the most significant change. Below each time point are the genes significantly (P< 0.05) decreased in expression with the gene, with the most significant decrease being the closest to the center. Genes in bold achieve a P-value <0.05 corrected for the number of genes tested at each time point and those with an asterisk remain significant when corrected for number of genes and time points.

Figure 5.

Combined cell-line model reveals different expression change patterns. After PRPF39 knockdown, ABHD5 expression is also significantly (P< 0.004) knocked down at three time points (5, 12 and 24 h after PRPF39 knockdown). In contrast, TFDP2 shows significant (P= 0.006) increase of expression only at 24 h after PRPF39 knockdown.