|Home | About | Journals | Submit | Contact Us | Français|
Elevated expression of the prostaglandin synthase cyclooxygenase-2 (COX-2) is commonly observed in many chronic inflammatory diseases and cancer. However, the mechanisms allowing for pathogenic COX-2 overexpression are largely unknown. The gene for COX-2 (PTGS2) carries a common single-nucleotide polymorphism (SNP) at position 8473 (T8473C), in exon 10 that is associated with diseases in which COX-2 overexpression is a contributing factor. We demonstrate that the T8473C SNP resides within a region that targets COX-2 mRNA for degradation through microRNA-mediated regulation. miR-542-3p was identified to bind transcripts derived from the 8473T allele and promote mRNA decay. By contrast, the presence of the variant 8473C allele interfered with miR-542-3p binding, allowing for mRNA stabilization, and this effect was rescued using a mutated miR-542-3p at the respective 8473 site. Colon cancer cells and tissue displayed COX-2 mRNA levels that were dependent on T8473C allele dosage, and allele-specific expression of COX-2 was observed to be a contributing factor promoting COX-2 overexpression. These findings provide a novel molecular explanation underlying disease susceptibility associated with COX-2 T8473C SNP, and identify it as a potential marker for identifying cancer patients best served through selective COX-2 inhibition.
Cyclooxygenase enzymes, COX-1 and COX-2, perform the rate-limiting step in the conversion of free arachidonic acid into prostaglandins and have distinct roles in physiologic and pathologic conditions (Wang and Dubois, 2010). COX-1 (PTGS1) is the constitutively expressed isoform involved in gastric mucosa protection and maintenance of vascular tone. Normally absent in most cells, COX-2 (PTGS2) expression is rapidly induced by pro-inflammatory and growth-associated stimuli. Substantial evidence has shown unregulated COX-2 expression to be a contributing factor in many chronic diseases and cancer (Menter et al., 2010). Numerous studies have demonstrated the benefit of inhibiting COX-2 activity with non-steroidal anti-inflammatory drugs and COX-2 selective inhibitors. However, unwanted cardiovascular side effects associated with long-term COX-2 inhibition has limited their use as primary chemoprevention agents (Menter et al., 2010).
Many genetic variants within the COX-2 gene have been identified and select polymorphisms have shown disease association. Several studies have identified the variant allele of one particular single-nucleotide polymorphism (SNP), T8473C (rs5275), to be associated with increased risk and/or NSAID responsiveness in a number of cancers where COX-2 overexpression has been observed (Campa et al., 2004; Ali et al., 2005; Siezen et al., 2005; Langsenlehner et al., 2006; Shen et al., 2006; Ferguson et al., 2008; Vogel et al., 2008; Gong et al., 2009; Ozhan et al., 2010). The T8473C SNP is located in exon 10 which encodes the COX-2 3′-untranslated region (3′-UTR), and work from our group and others has identified the significance of this region in targeting COX-2 mRNA for post-transcriptional regulation (Young and Dixon, 2010).
MicroRNAs (miRNAs) are small non-coding RNAs, ~22 nucleotides in length, that primarily bind to the 3′-UTR of targeted transcripts and promote mRNA decay and translational suppression (Fabian et al., 2010). Current work has established the critical role miRNAs have in regulating gene expression and the consequences of dysregulated miRNA expression (Croce, 2009). More recently, genetic polymorphisms in components of miRNA networks are being recognized as contributing factors in disease etiology (Ryan et al., 2010). In this study, a common genetic alteration in the COX-2 3′-UTR was shown to impact specific miRNA activity allowing for COX-2 overexpression and disease association.
A bioinformatic approach was used to identify novel miRNAs that could target the COX-2 3′-UTR in the vicinity of the 8473 SNP. Using miRanda target prediction, miR-542-3p was predicted to bind the major allele 8473T with a minimum free energy of −20.29 kcal/mol. The functional significance of this potential miRNA:mRNA interaction was shown using HeLa cells stimulated with IL-1β to induce COX-2 expression, and then transfected with mature miR-542-3p. Northern and immunoblot analysis indicated that miR-542-3p was potent at inhibiting COX-2 mRNA and protein expression (Figures 1a–c). This miRNA-dependent loss of COX-2 was reflected in ~10-fold decreased PGE2 synthesis, nearly to the extent observed with the COX-2 selective inhibitor NS-398 (Figure 1d). The ability of miR-542-3p to attenuate PGE2 synthesis was consistent with specific targeting of COX-2, as mRNA and protein expression of COX-1, mPGES-1, or 15-PGDH was not impacted by miR-542-3p (Supplementary Figures 1a–c).
miRNAs can regulate gene expression, both by influencing translation and by causing degradation of target mRNAs (Fabian et al., 2010), and recent findings have implicated miRNA-mediated mRNA decay to be the predominant mechanism (Guo et al., 2010). To determine if miR-542-3p controlled COX-2 expression through rapid mRNA decay, IL-1β-treated HeLa cells transfected with miR-542-3p were treated with actinomycin D to halt transcription and determine mRNA half-life (Figure 1e). Consistent with previous results, IL-1β promoted COX-2 mRNA stabilization (Young et al., 2009), whereas the presence of miR-542-3p promoted rapid decay leading to a >2-fold reduction in mRNA half-life.
To determine if the effect of miR-542-3p on COX-2 expression was mediated through direct association between miR-542-3p and COX-2 mRNA, miRNA ribonucleoprotein immunoprecipitation was performed to evaluate if miR-542-3p participated in recruiting COX-2 mRNA to the RNA-induced silencing complex (RISC) complex. IL-1β-treated HeLa cells were cotransfected with miR-542-3p or control miR, along with HA-tagged Argonaute-1 (Ago1) expression vector, a central component of RISC that directs miRNA–mRNA interaction (Fabian et al., 2010). The association of COX-2 mRNA and miR-542-3p RNA with Ago1 was assayed by quantitative PCR in immunoprecipitates. As shown in Figures 2a and b, both COX-2 and miR-542-3p were highly enriched in the Ago1 samples, in which miR-542-3p was expressed. COX-2 and miR-542-3p were minimally detected in reactions using control miR, or in IPs using non-specific IgG. Furthermore, knockdown of Ago-1 negates the effect of miR-542-3p on COX-2 mRNA stability (Figure 2c). Taken together, these results indicate that miR-542-3p facilitates COX-2 mRNA association with RISC to promote rapid mRNA degradation.
A key 3′-UTR cis-acting element controlling COX-2 expression at the post-transcriptional level is the AU-rich element (ARE; Dixon et al., 2000). Through its interaction with RNA-binding proteins, the ARE targets COX-2 mRNA for rapid decay and translational suppression (Young and Dixon, 2010). Current work has shown that miRNAs can promote ARE-mediated mRNA decay through a mechanism directing recruitment of RNA degradation machinery to the ARE (Jing et al., 2005). As the binding site for miR-542-3p is ~350 nucleotides from the COX-2 ARE, we sought to determine if there was a functional interaction between these 3′-UTR elements. Luciferase reporter constructs bearing the full-length 8473T-containing COX-2 3′-UTR, or with the ARE deleted (Dixon et al., 2000), were transfected along with miR-542-3p. As shown in Figure 2d, miR-542-3p was equally effective at inhibiting expression in the presence or absence of the ARE, indicating that the presence of the COX-2 ARE is not required for miR-542-3p function. A >2-fold reduction in luciferase mRNA was observed in the presence of miR-542-3p, consistent with miR-542-3p-dependent reduction in luciferase activity (Figure 2e). Furthermore, overexpression of the ARE RNA-binding protein HuR, which can promote COX-2 overexpression by interfering with rapid mRNA decay (Young et al., 2009), did not interfere with miR-542-3p targeting of COX-2 (Supplementary Figure 2). Although these results differ from findings demonstrating the ability of HuR to rescue ARE-containing mRNAs from miRNA-mediated regulation, possibly by interfering with miRNA association with ARE-containing transcripts (Bhattacharyya et al., 2006), they indicate that the ARE and miR-542-3p targeting site are functionally separate post-transcriptional regulatory elements that act independently to maintain tight control of COX-2 gene expression.
Effective binding between an miRNA and its mRNA target are essential for miRNA-mediated regulation, and SNPs that exist within these miRNA target sites have potential to impact miRNA binding (Ryan et al., 2010). Although the interaction between miR-542-3p and the wild-type 8473T allele is functional, the variant 8473C allele changes miR-542-3p binding with a predicted minimum free energy of −16.32 kcal/mol (Figure 3a). To test if this polymorphism could alter miR-542-3p activity, 3′-UTR reporter constructs containing either 8473T or 8473C were cotransfected with miR-542-3p into HeLa cells. Comparable to the results in Figure 2b, miR-542-3p downregulates luciferase activity, approximately two-fold in the presence of the wild-type 8473T 3′-UTR (Figure 3b). However, this inhibition was not observed with the 3′-UTR reporter bearing the variant 8473C SNP. To validate these results, miR-542-3p was mutated A→G at the COX-2 8473 site (labeled miR-542-3p Alt), such that it could bind the 8473C-containing 3′-UTR (Figure 3a). As shown in Figure 3b, this mutated miRNA did not impact expression from the 8473T 3′-UTR reporter, whereas miR-542-3p Alt was effective at inhibiting expression from the 8473C 3′-UTR reporter, further demonstrating that the context of T8473C SNP is a critical determinant of miR-542-3p function.
The results indicating that miR-542-3p promoted COX-2 mRNA decay (Figure 1e) suggested that the presence of the variant 8473C allele may promote mRNA stabilization. This was evaluated by determining the half-life of luciferase mRNA containing each 8473 3′-UTR allele in the presence of miR-542-3p. As shown in Figure 3c, miR-542-3p promoted rapid luciferase mRNA decay from constructs bearing the wild-type 8473T 3′-UTR (t1/2=41 min), whereas stabilization was detected in mRNA bearing the variant 8473C 3′-UTR (t1/2=86 min). This two-fold difference in mRNA half-life was dependent on miR-542-3p targeting of the 8473T-containing 3′-UTR, as similar decay rates were observed for each construct when miR-542-3p was absent (data not shown).
Evidence indicating that deficient miR-542-3p binding to the 8473C-containing 3′-UTR was shown using miRNA ribonucleoprotein immunoprecipitation reactions. As shown in Figure 3d, luciferase mRNA bearing the wild-type 8473T 3′-UTR was associated in the Ago1/RISC complex in the presence of miR-542-3p, whereas the 8473C SNP was sufficient to disrupt this association. Reactions using control miRNA or non-specific IgG did not display enrichment of luciferase mRNA in miRNA ribonucleoprotein immunoprecipitations, indicating that miR-542-3p binding and recruitment of Ago1-containing RISC complex is COX-2 T8473C-allele specific (Figure 3d and data not shown).
COX-2 overexpression is commonly observed in colorectal cancer (CRC), and COX-2-dependent prostaglandin synthesis promotes various facets of tumorigenesis leading to worse survival rates (Wang and Dubois, 2010). To determine if a cause-and-effect relationship exists between T8473C SNP and miR-542-3p in CRC, cell lines were genotyped for T8473C SNP and evaluated for allelic COX-2 expression. The CRC cell lines HCA7 and Moser were of T/T genotype, whereas HT29 and LS174T were heterozygotes (T/C) for the SNP; out of 10 CRC lines genotyped, no C/C homozygotes were identified. Allelic mRNA expression was accomplished using a Taqman-based allelic discrimination assay to measure transcript abundance. As shown in Figure 4a, only mRNA containing the 8473T SNP was detected from HCA7 and Moser cells. Interestingly, heterozygote cells HT29 and LS174T showed ~90% of steady-state COX-2 mRNA to contain the 8473C SNP. This suggests that the mRNA derived from the 8473C-containing allele is inherently more stable than the 8473T-containing transcript owing to deficient miR-542-3p targeting.
To test this, CRC cells were transfected with miR-542-3p and evaluated for its effect on COX-2 expression. In T/T homozygote cells, miR-542-3p was able to downregulate~75% of COX-2 protein expression. By contrast, miR-542-3p was ineffective in attenuating COX-2 expression in T/C heterozygote cells (Figure 4b). This effect appeared to be a result of deficient COX-2 mRNA decay mediated by miR-542-3p. As shown in Figure 4c, miR-542-3p promoted rapid mRNA decay in T/T homozygote Moser cells (t1/2=41 min), compared with T/C heterozygote HT29 cells (t1/2=77 min); similar half-lives of 67 and 83 min were observed in control miR-transfected Moser and HT29 cells, respectively. Although demonstrating the effect this SNP has on allelic COX-2 expression in CRC cells, these results indicate transcripts bearing the variant 8473C SNP to be inherently more stable and the underlying source of COX-2 overexpression.
It is conceivable that loss of miR-542-3p could occur during tumor development as observed with other tumor suppressor miRNAs (Croce, 2009) allowing for COX-2 overexpression. This was evaluated by assaying miR-542-3p expression in normal colon and tumor tissue along with CRC cell lines (Supplementary Figures 3a–c). In both normal/tumor pairs and in CRC cells, the average miR-542-3p fold-change was not significant. Similarly, no correlation was observed between miR-542-3p levels and COX-2 mRNA levels in tumor tissue. These results are consistent with miR-542-3p expression in other tumor types (Volinia et al., 2006), indicating that the T8473C SNP genetic variation could be a contributing factor impacting COX-2 levels. To assess this, CRC tumor samples were genotyped for T8473C SNP, and COX-2 levels were assayed. As shown in Figure 4d, there was a significant increase in both COX-2 mRNA and protein levels present in tumors that correlated with C allele dosage. With regard to tumor location, the 8743C allele frequency was greater in distal than proximal colon (60 versus 25%), which is consistent with COX-2 upregulation predominantly observed in distal colonic and rectal tumors (Dimberg et al., 1999; Nasir et al., 2004; Birkenkamp-Demtroder et al., 2005). These results further demonstrate the significance the 8473T allele has in controlling COX-2 levels, and illustrate the profound effect the variant allele has in promoting COX-2 overexpression observed in CRC tumors.
A large amount of data from population studies and clinical trials demonstrate the efficacy of COX-2 selective inhibitors and non-steroidal anti-inflammatory drugs in reducing cancer risk (Menter et al., 2010). However, an accurate measure of identifying patients most likely to receive the greatest chemopreventive benefit from COX-2 selective inhibitor-based therapy has been a limiting factor in that an incomplete understanding of the genetic causes associated with COX-2 overexpression currently exists. The results presented here describe the mechanism by which T8473C SNP functions in COX-2 regulation, and provides a molecular explanation to tumor susceptibility associated with this SNP. Through binding at the 8473 3′-UTR site, miR-542-3p targets COX-2 mRNA for decay, and the presence of the variant C-containing allele disrupted miRNA–mRNA interaction allowing for COX-2 overexpression. It is conceivable that this mechanism may impact expression of other miR-542-3p targets that contain SNPs within the miR-542-3p binding site (Supplementary Table 1). By understanding the function of T8473C SNP, along with other 3′-UTR SNPs shown to affect miRNA activity (Nicoloso et al., 2010; Ryan et al., 2010), these findings highlight the contribution of miRNA-mediated regulation of cancer-associated gene expression and supports their use as markers of risk and personalized therapy.
This work was supported by the National Institutes of Health (R01CA134609) and American Cancer Society (RSG-06-122-01-CNE).