3.4.1 In silico predictions
To understand more about the variants identified and their LD structure, as well as to prioritize variants for in vitro validation based on their predicted functionality, a series of in silico analyses were conducted. Variants from Aim 1 with evidence of promoter or enhancer histone modifications (e.g. H3K4me1, H3K4me3 and H3K27ac marks), evidence of DNase sensitivity, evidence that they influence transcription factor binding through changes to
transcription factor binding motifs and/or evidence of being an expression quantitative trait locus (eQTL), were prioritized for in vitro functional validation.
For the VEGFA variants identified in Aim 1 (rs1885657 and rs3024987) and two variants in perfect LD with rs1885657 (rs58159269 and rs943070), there was substantial evidence that they exist in a regulatory region. In multiple cell lines, there was evidence of both promoter and enhancer histone marks (H3K4me1, H3K4me3 and H3K27ac marks) (Table 3.1 and Figure 3.1). Briefly, methylation or acetylation modifications, or marks, to histones can influence how accessible the chromatin is to transcription, thereby influencing gene expression. H3K4me1 histone mark is a mono-methylation of lysine 4 of the H3 histone protein and is associated with enhancers and DNA regions downstream of transcription start sites. An H3K4me3 histone mark indicates tri-methylation of the same lysine residue and is associated with promoters that are active or poised to be activated. And, a H3K27ac histone mark indicates acetylation of lysine 27 of H3, and like H3K4me1, a H3K27ac mark is also often indicative of a transcription enhancer.21
These four VEGFA variants also revealed evidence of altered transcription factor binding and/or altered transcription factor motifs, and areas of DNase hypersensitivity (Table 3.1 and Figure 3.1). Variants in active regulatory regions (e.g. promoters and enhancers) often reveal evidence of DNase hypersensitivity. Areas of DNase hypersensitivity indicate loss of condensed chromatin structure, which exposes DNA and makes it accessible for transcription.21,22
No renal endothelial cell lines were analyzed during the ENCODE project. This was a potential limitation to leveraging ENCODE resources because RCC is highly dependent on angiogenesis and host vascular endothelium. In addition, analyzing data from an endothelial
cell type was essential because sorafenib, in part, targets host vascular endothelium. Therefore, examining data from human umbilical vein endothelial cells (HUVEC) was important because it provided insight into regulatory elements within the cell type closest to renal endothelial cells. In HUVECs, there was evidence of H3K4me1, H3K4me3, and H3K27ac marks at rs1885657. For rs58159269, there was evidence of H3K4me3, and H3K27ac marks. For rs943070 and rs3024987, there was evidence of H3K27ac (Figure 3.2).
By Regulome DB, the variant predicted to have the most influence on regulation of
VEGFA was rs59159269. It had a Regulome DB score of 2b (Table 3.2), which means there was evidence of altered transcription factor binding sites, any transcription factor motif, a DNase footprint, and a DNase peak. Variants rs1885657, rs943070, and rs3024987 all had a Regulome DB score of 4, which means there was evidence of a DNase peak and altered transcription factor binding. Collectively, there was sufficient ENCODE prediction data to prioritize these variants for functional validation.
There was only minimal evidence in ENCODE that rs3816375 in ITGAV and rs8047917 in WWOX were in regions that would affect gene regulation (Tables 3.3 and 3.4, respectively). There is no data to provide a Regulome DB score for rs3816375, and the Regulome DB score for rs8047917 was 6. Based on minimal ENCODE evidence in the UCSC Browser, HaploReg and Regulome DB, these two variants were not prioritized for functional validation.
Variant rs30726 was not predicted to be deleterious or cause a harmful amino acid substitution. While a threonine to alanine amino acid substitution results in a change from a medium-sized and polar amino acid to a small-sized and hydrophobic amino acid,23 rs307826 received a score of 0.53 from SIFT, which means the algorithm predicted that the amino acid
substitution would be tolerated.18 Similarly, rs307826 received a score of 0.005 from
PolyPhen-2, which means the algorithm predicted that the amino acid substitution would be benign.17
3.4.2 Variant Effects on Cell Viability
To validate the functionality of rs307826 and to gain insight into whether this non- synonymous variant in a coding region of FLT-4 has an effect on VEGFR3 signaling, vectors containing reference and mutant cDNA were transfected into HEK-293 cells and
subsequently treated with sorafenib. Cell viability assays were conducted to generate IC50
values for VEGFC-stimulated and unstimulated reference cells, as well as VEGFC- stimulated and unstimulated mutant cells.
Cells were treated in triplicate wells with 12 concentrations of sorafenib (0–30 µM), and cell viability relative to untreated controls (0.2% DMSO only) was determined through the alamarBlue® cell viability assay. First, a statistically significant difference in sorafenib cytotoxicity was observed between reference-transfected and mutant-transfected HEK-293 cells that were not stimulated by VEGFC (p<0.0001). The IC50 for reference-transfected cells
was 7.58 µM, while the IC50 for mutant-transfected cells was 15.45 µM (Figure 3.3). Second,
a statistically significant difference in sorafenib cytotoxicity was observed between reference-transfected and mutant-transfected HEK-293 cells that were both stimulated by VEGFC prior to sorafenib administration (p<0.0001). The IC50 for reference-transfected cells
was 2.02 µM, while the IC50 for mutant-transfected cells was 7.67 µM (Figure 3.3). Finally, a
statistically significant difference in sorafenib cytotoxicity was observed between VEGF- stimulated and unstimulated mutant-transfected HEK-293 cells (p<0.0001) (Figure 3.3).
3.4.3 Variant Effects on Transcriptional Activity
To provide insight into the mechanisms underlying the variant-OS associations observed among intronic variants identified in Aim 1, variants were assessed for their function as potential effectors of regulatory activity using dual reporter gene luciferase assays. Genomic regions containing the reference sequence, or sequence with the variants (or those in perfect LD with the identified variant) were cloned into a pGL4.26 plasmid construct containing a minimal promoter and downstream of the Firefly luciferase gene. Allelic
variation (rs1885657 C allele, rs58159269 C allele, rs943070 G allele, or rs3024987 T allele) were introduced by site-directed mutagenesis. These assays were performed to determine how each variant (or the construct with three variants) affected the transcriptional activity of the minimal promoter.
For the VEGFA variant that associated with shorter OS in patients treated with sorafenib (rs1885657), and the two variants in perfect LD (rs943070 and rs58159269), significant increases in luciferase activity in LPEC cells were observed individually for all three variants, as well as for the “triple variant.” For rs1885657 (T>C), the C allele increased luciferase activity by an average of 48% (p<0.05). For rs58159269 (T>C), the C allele increased luciferase by an average of 70% (p<0.001). For rs943070 (C>G), the G allele increased luciferase by an average of 40% (p<0.05). And, for the triple variant, luciferase was increased by an average of 98% (p<0.01), when compared to the reference alleles (Figure 3.4).
In the TIME human endothelial cell line, increased luciferase activity was again observed. For rs1885657 (T>C), the C allele increased luciferase activity by an average of 57% (p<0.001). For rs58159269 (T>C), the C allele increased luciferase by an average of
80% (p<0.001). And, for the triple variant, luciferase was increased by an average of 99% (p<0.001), when compared to the reference alleles. Luciferase activity was not significantly different between rs943070 and the reference allele in TIME cells (Figure 3.4).
Finally, in the clear-cell mRCC Caki-1 cell line significant increases in luciferase activity were observed individually for rs1885657, rs58159269 and rs943070, as well as for a haplotype of all three SNPs. For rs1885657 (T>C), the C allele increased luciferase activity by an average of 30% (p<0.05). For rs58159269 (T>C), the C allele increased luciferase by an average of 56% (p<0.001). For rs943070 (C>G), the G allele increased luciferase by an average of 35% (p<0.01). And, for the triple variant, luciferase was increased by an average of 70% (p<0.001), when compared to reference (Figure 3.4).
For VEGFA rs3024987 (C>T), which was associated with shortened OS in a combined analysis of both study arms, significant increases in luciferase activity were
observed in all three cell lines. In LPEC cells, the T allele increased luciferase activity by an average of 34% (p=0.0032), when compared to reference. In the TIME cells, the T allele increased luciferase activity by an average of 38% (p=0.0002). Finally, in Caki-1 cells, the T allele increased luciferase activity by an average of 32% (p=0.0001), when compared to reference (Figure 3.5).