According to our database review and sequencing results the coding region of PLK1 is conserved and polymorphisms are located in intronic and regulatory regions. This is in line with general findings with regard to the occurrence rate of genetic variations and especially of SNPs in different gene regions . We considered database-derived polymorphisms with a MAF of at least 1% in Caucasians. Nevertheless, the analyzed databases comprise a relevant number of rare variants of the coding region with MAFs of less than 1% in Caucasian that could have a functional impact on PLK1. Some of these variants reach a MAF of more than 1% in other ethnicities (e.g. rs2230914). Due to the number of chromosomes investigated by sequencing, the probability to detect new, undescribed polymorphisms with a MAF of 1% was 33% only. An adequate probability of at least 90% for detecting new polymorphisms was only reached for polymorphisms with a MAF of more than 5%. Therefore, it is possible that the PLK1 gene still harbors undetected rare variants (most likely non-SNP variations). Furthermore, other databases that were not systematically analyzed for this study might contain additional variations with MAFs above 1% in Caucasians. For example, after completion of our experiments we became aware of a missense variant (rs45569335) with an overall MAF of 0.7%, but with a MAF of 1.2% within the Caucasian subset of the 1000 genome browser .
We selected four candidate SNPs for further investigation, which were either located within the regulatory regions of PLK1 (rs57973275, rs16972787 and rs27770) or showed an association with bladder cancer outcome in a previous study (rs40076). The results of the haplotype analysis implied that it is not necessary to genotype all 4 polymorphisms in future association studies  and that two tagging SNPs, either rs57973275 or rs16972787 in combination with rs27770 would be sufficient to represent the haplo- and diplotype structure of these PLK1 SNPs. Furthermore, the strong linkage disequilibrium suggests rs27770 as the underlying functional SNP in the detected association of the PLK1 intron 3 SNP rs40076 with bladder cancer outcome .
Since the two SNPs 5′ of the coding region are located within previously identified important regulatory regions of PLK1 they were considered eligible candidates for bioinformatic and experimental assessment , and because different software applications access different databases, we used three in silico tools to predict putative TF binding sites. Analysis revealed different allele-specific candidates for both SNPs. Computational approaches for identifying binding sites suffer from high error rates because binding motifs of TFs are typically short and degenerated , therefore, we performed EMSA experiments using three different cell lines to validate the in silico results. Unfortunately, the EMSA results clearly indicated no functional impact of these polymorphisms on TF binding, negating any further evaluation of the SNPs with regard to TF binding activity. Some authors reported that only 33% of all functional promoter variants were found in known consensus sequences or motifs , therefore, we cannot rule out an impact of the two SNPs on binding of other TFs which are not expressed by the selected cell lines. Another mechanism to regulate expression is methylation of CpG islands. Theoretically, the G allele of the promoter polymorphism rs16972787 could be a candidate for allele-specific methylation, however, changes of the methylation status in human malignant cells and tissues have been reported for PLK2 and PLK3 but not for PLK1[12, 43, 44]. A further study suggested that the PLK1 promoter is unmethylated in G0/G1 (PLK1 not expressed) as well as M phase (PLK1 expressed) and regulated during the cell cycle by transcription factors .
Alleles of the 3′UTR polymorphism rs27770 were analyzed with regard to different functional RNA motifs and microRNA binding sites. The analysis revealed neither motif nor target site differences. Until now at least six PLK1 mRNA-targeting microRNAs have been experimentally validated [46–51]. In line with our analysis, the predicted corresponding binding sites do not include the polymorphism. In consideration of the secondary mRNA structure, another key factor for microRNA target recognition is the accessibility of the binding site . Different reports proposed altered microRNA binding because of allele-dependent changes of the secondary mRNA structure due to SNPs outside of the microRNA binding site [53, 54]. Furthermore, alterations of the secondary structure itself can interfere with RNA-binding proteins, which can lead to altered mRNA stability . We therefore investigated the secondary structure of the PLK1 mRNA dependence on rs27770 alleles. Although only one nucleotide was substituted, major changes of the secondary structure were predicted. Reporter assays of the 3′UTR of PLK1 consistently showed statistically significant allele-dependent differences in mRNA stability. In comparison to the G allele, the A allele showed 25% more reporter activity, which faithfully reflects mRNA levels . Our results, as well as a previous bioinformatic comparison of HapMap and dbEST data, support a functional impact of rs27770. However, the results themselves are contradictory, because the previous report predicted an increased expression of the G allele . This could have several reasons. First, an experimental validation of a subset of the predicted candidate SNPs confirmed only 36% of the results and rs27770 was not part of the validation subset. Most of the SNPs (59%) showed no differential allelic expression, but alleles corresponding to 5% of the SNPs were significantly associated with gene expression in the opposite direction. Second, according to the usual practice, we investigated mRNA stability of the 3′UTR by reporter assay, however the complete PLK1 mRNA contributes to the secondary structure. It is therefore possible that the hybrid mRNA of the Firefly luciferase coding region and the PLK1 3′UTR could lead to biased results because of other secondary structures. Third, both results could be genuine, if the effect of the SNP is context-dependent and tissue-related. This is a well-known phenomenon and occurs often in connection with regulatory SNPs, especially if the SNP effect depends on differentially expressed transcription factors and microRNAs respectively .
Finally, genetic variability of the PLK1 gene and rs27770 in particular are interesting candidates for additional studies. Because PLK1 plays an important role in the cell cycle and inhibits apoptosis, it should be investigated whether PLK1 polymorphisms have an impact on proliferation of malignant and non-malignant cells [3, 5]. This would lead to altered expression profiles in cancer tissues and might partly explain the detected variability of PLK1 expression in different cancer entities . In some malignancies like acute lymphoblastic leukemia (ALL) PLK1 expression is highly variable, but expression is not associated with any clinical or biological feature while ALL cell lines respond very well to PLK1 inhibitor treatment . For these malignancies analysis of PLK1 polymorphisms would be an interesting approach to reanalyze genotype-dependent subsets with regards to expression patterns and clinical as well as biological features. PLK1 polymorphisms could be useful with regard to risk as well as outcome studies in Caucasian cancer patients but also in other ethnicities because, according to dbSNP data, rs27770 occurs in other ethnicities as well. Furthermore, functional SNPs in drug target genes may have an impact on targeted therapy. It would therefore be of interest to evaluate PLK1 inhibitor studies with regard to PLK1 polymorphisms, and, because of the predicted impact of the SNP on the secondary mRNA structure of PLK1 and the effects shown on mRNA stability, RNAi based PLK1 inhibitors would be of special interest in this case. It is also well-known that the target secondary structure has a major impact on siRNA and RNAi efficiency [59, 60]. We have no evidence for an interaction of PLK1 polymorphisms with currently clinically evaluated RNAi-based PLK1 inhibitors , but it would be reasonable to analyze the respective binding sites with regard to allele-dependent target accessibility.