A polymorphism at the 3'-UTR region of the aromatase gene defines a subgroup of postmenopausal breast cancer patients with poor response to neoadjuvant letrozole

After cessation of ovarian activity at postmenopause, the aromatase [cytochrome P450 19 (CYP19); OMIM:107910] from bones, adipose tissue and muscle, becomes the key enzyme in estradiol and estrone biosynthesis through the aromatization of testosterone and androstenedione respectively [1]. Aromatase is encoded by CYP19A1 which maps at chromosome 15q21.1 and has a complex structure with a region that contains 10 tissue-specific non-coding upstream exons with separate promoters that regulate transcription in different cells and tissues [2]. Elevated levels of aromatase expression have been observed in breast tumors relative to normal breast tissue [3], and alterations in aromatase expression have been implicated in the pathogenesis of estrogen-dependent diseases, including breast cancer (BC) [4, 5].

The major biological characteristic of BC is that two thirds of cases express estrogen (ER) and/or progesterone receptors (PgR) [2, 6, 7], endocrine manipulation being an effective treatment for these patients. Tamoxifen, a selective ER modulator, has for decades been the gold standard therapy in ER/PgR [+] BC patients [8]. However, in postmenopausal patients, aromatase inhibitors like letrozole have consistently demonstrated to be more effective than tamoxifen [9]. Letrozole is highly specific for the aromatase enzyme and inhibits whole-body aromatization of almost 99%, providing a highly and essentially complete withdrawal of estrogen in postmenopausal women [10].

Neoadjuvant hormonal therapy is a safety option for postmenopausal women with large operable or locally advanced hormone sensitive BC, increasing the possibilities of breast-conserving surgery (BCS) [11]. In this context, letrozole has shown to be more effective than tamoxifen, obtaining higher response and BCS rates following 4 months of preoperative therapy [12, 13]. However, less than 60% of patients will respond to this optimal endocrine therapy [14]. To date, only the presence and intensity of hormonal receptors (ER/PgR) are useful tools as predictive markers in clinical practice. Identification of more accurate markers for more efficient patient selection to exclude non-responsive cases remains crucial.

Genetic polymorphisms in the CYP19A1 gene have been associated with altered sex hormone levels in serum and urine [15‐17], providing an explanation for an elevated risk for BC in relation to estrogen exposure. For example, the tetranucleotide repeat polymorphism in intron 4 (TTTA)n has been associated with BC risk in initial studies [18‐22]. The polymorphism rs10046 in the 3' untranslated region (3'-UTR) also showed inconsistent associations with BC risk [23‐26] and has also been related with tumor stage [25], circulating sex hormone levels [15], HER2 status and disease free and overall survival [27, 28]. rs4646 has been reported to be associated with circulating steroid hormone levels [17] and, as in rs10046, with the HER2 status of the tumor [27]. Hence, it is biologically plausible that the polymorphisms in the CYP19A1 gene may be associated with the response to aromatase inhibitors. Indeed, recently the rs4646 CYP19A1 polymorphism has been associated with letrozole efficacy in advance disease [29]; however, no evidence between CYP19A1 polymorphisms and therapeutic efficacy of aromatase inhibitors in early stage BC has yet been established.

We describe a relationship between the genetic variants (AC/AA) of the rs4646 polymorphism of CYP19A1 and a poor response to neoadjuvant treatment with letrozole in postmenopausal women with ER/PgR [+] BC. The radiological response was measured at the 4^th month of letrozole induction, the observed response rate being 36.8%, similar to the radiological response reported in the P024 trial [12].

At genetic level, mutations in the aromatase gene may lead to a functionally less sensitive aromatase phenotype. Although mutations in the CYP19A1 gene can be generated in vitro, to date no somatic mutation has been found in clinical samples [38]. However, several genetic polymorphisms of CYP19A1 have been reported so far, although the possible functional significance of most of these remains undefined. In the case of BC, population-based studies of common CYP19A1 polymorphisms have generated inconsistent results with regard to their possible association with sex hormone levels, cancer risk, HER2 status or survival [15, 19, 21, 27, 28]. No association between the HER2 or hormonal status of either rs10046 or rs4646 was observed in our series.

Whether aromatase inhibitor agents interact with the different CY19A1 genotypes has not so far been clearly established. Only three previous studies have attempted to demonstrate the effect of certain polymorphisms in CYP19A1 on the efficacy of aromatase inhibitors. In the first, Fasching et al., in a series of patients treated with hormonal therapy for more than 2 years, did not observe significant differences between the rs700519, rs10046 and rs4646 polymorphisms with therapy. Unfortunately, the type of antihormonal therapy was not recorded for this analysis [27]. In the second study, a population-based and in vitro study, Ma et al. revealed reductions in the functional activity of aromatase for four phenotypes resulting from non-synonymous changes [rs700519 (Arg²⁶⁴Cys), rs28757184 (Thr²⁰¹Met), rs2236722 (Trp³⁹Arg), and rs56658716 (Met³⁶⁴Thr)] [32]. These authors found that levels or aromatase enzyme activity decreased dramatically for the Thr³⁶⁴, and also observed a slight decrease in Cys²⁶⁴ allozyme activity. They also demonstrated that the mechanism by which non-synonymous SNPs affect the enzymatic activity is a consequence of an alteration in the enzyme protein level [32]. In the same study, these authors reported that these variants had no significant differences in their affinity to the aromatase inhibitors letrozole and exemestane [32]. Interestingly, the polymorphism rs700519 has also been reported as a prognostic factor in a Chinese study, mainly in a subgroup of premenopausal women [28], but not in the Caucasian population [27]. In addition, it is important to note the differences between these two populations regarding the minor allele frequency of rs700519, being only 3.2% for the Caucasian population (1,257 patients) [27], in comparison with 15.1% in the Asian Cohort (1,136 patients) [28]. In our series, the genetic variants of this polymorphism were underrepresented and not found in any of the cases.

The third study evaluated the efficacy of treatment with letrozole in advanced hormone receptor-positive BC patients with respect to two polymorphisms located at the 3'-UTR (rs10046 and rs4646) and one in intron 2 [rs727479 (G/T)] of the CYP19A1 [29]. The authors found that genetic variants of rs4646 were associated with a greater efficacy of letrozole in terms of time to progression. Hence, patients with the variant genotype (AC/AA) had a three times greater time to progression than the patients with the reference genotype (CC). Furthermore, these authors reported that the frequency of the variant alleles for rs4646 was significantly higher in the responder (61%) than in the non-responder group of patients (40%) [29]. These observations clearly differ from those reported in our series, the genotypic variants of rs4646 being more frequently represented in the non-responder group to letrozole after 4 months of induction therapy (48% vs. 26%). In this regard, there are two main issues that could explain the differences between these two studies. First, and independently of the disease stage, are the criteria for patient selection. Whereas in our series the patients were treated with letrozole as a first line, in the study of Colomer et al. the patients had progressed from a previous treatment with tamoxifen [29]. It is well established that tamoxifen metabolism is influenced by the number of mutant alleles of the gene encoding cytochrome P450 2D6 (CYP2D6), because this enzyme affects the levels of endoxifen, the active tamoxifen metabolite. Thus, for patients who are wild type for CYP2D6, the 5-year disease-free survival outcomes are similar to or perhaps even superior with tamoxifen than with aromatase inhibitors [39]. Therefore, previous treatment with tamoxifen could have genetically selected a population of patients more sensitive to letrozole. Secondly, in our study we used constitutive DNA obtained from peripheral blood for the genetic analysis, whereas Colomer et al. performed the genetic study on DNA obtained from the fixed and paraffin-embedded tumors [29]. The polymorphism studies performed on tumor samples are subject to genetic alterations that can affect the chromosomal region of the gene under study. In this way, CYP19A1 is located in the 21.2 region of the long arm of chromosome 15 (15q21.2) [40], and this region has been reported to be a frequent target of allelic imbalance in advanced breast carcinomas [41], and which could affect the frequency distribution of the allelic variants.

From our series, 65 women underwent surgery after letrozole induction, being age the only difference observed between the operated and non-operated patients. In fact 77% of patients within the non-surgery group were over 78.3 years. We have demonstrated that genetic variants of rs4646 have prognostic value, especially in this group of patients. Indeed, the 6 women who progressed within the group of non-surgery patients were genetically characterized by the variant rs4646 (AC/AA). In this elderly group of patients the decision for undergoing surgery is not always easy and very often both the patient and the clinician opt for a local control of the tumor maintaining the treatment with letrozole. Hence, in this group of patients, in which the tumor is not removed, is where the rs4646 polymorphism identifies the women who progressed, indicating, and despite the small number the cases included in our series, that women with the rs4646 (AC/AA) genotype could benefit from another therapeutic approach.

The regulatory interactions between the ER, growth factor receptors and other kinase signalling pathways could also determine the response to endocrine therapy. In the adjuvant setting, several studies suggest that patients overexpressing HER2 may derive relatively less benefit from endocrine therapy [38]. In the P024 trial, ER [+] tumors that were also EGFR and/or HER2 [+] responded significantly better to letrozole than to tamoxifen (88% vs 21% respectively, P = 0.0004) [13]. The IMPACT study, that confronted tamoxifen with anastrozole, or the preoperative combination of both, observed a similar effect in favour of anastrozole for HER2 [+] tumors (Odds Ratio 58% vs. 22%), although not reaching statistical significance [42]. In our series, as in both randomized trials, no differences in activity were observed relating to HER2 expression among the patients receiving letrozole.

Biological studies of tissues obtained during neoadjuvant therapy have thrown a different perspective on this issue. Short-term estrogen deprivation leads to profound changes in transcriptional profiles, and these changes can be used as predictive tools. Tumor expression of the proliferation antigen Ki67 after 2 weeks of endocrine treatment (tamoxifen or anastrozole) predicts for clinical response and recurrence-free survival [42]. However, among cases on anastrozole, only 7% did not present a reduction in the Ki67 expression at 2 weeks. In many cases, this reduction in the tumor proliferation rate is only modest and could not be sufficient to determine a resistance to the treatment. As expected, early tumor changes are not limited to proliferation markers, but to other genes directly regulated by ER, including the aromatase itself [43]. Other, currently ongoing, approaches have used gene expression profiling techniques able to predict response to endocrine therapy, but the results obtained so far are inconsistent [43]. In our series, expression of proliferating marker Ki67 constituted a factor of poor prognosis in the group of patients undergoing surgery following letrozole induction therapy, suggesting that this marker could identify tumors with a more aggressive behavior.