Impact of CYP19A1 and ESR1 variants on early-onset side effects during combined endocrine therapy in the TEXT trial

Background Single nucleotide polymorphisms (SNPs) in the estrogen receptor 1 (ESR1) and cytochrome P450 19A1 (CYP19A1) genes have been associated with breast cancer risk, endocrine therapy response and side effects, mainly in postmenopausal women with early breast cancer. This analysis aimed to assess the association of selected germline CYP19A1 and ESR1 SNPs with early-onset hot flashes, sweating and musculoskeletal symptoms in premenopausal patients enrolled in the Tamoxifen and Exemestane Trial (TEXT). Methods Blood was collected from consenting premenopausal women with hormone-responsive early breast cancer, randomly assigned to 5-years of tamoxifen plus ovarian suppression (OFS) or exemestane plus OFS. DNA was extracted with QIAamp kits and genotyped for two CYP19A1 (rs4646 and rs10046) and three ESR1 (rs2077647, rs2234693 and rs9340799) SNPs by a real-time pyrosequencing technique. Adverse events (AEs) were recorded at baseline and 3-monthly during the first year. Associations of the genotype variants with grade ≥2 early-onset targeted AEs of hot flashes/sweating or musculoskeletal events were assessed using logistic regression models. Results There were 2660 premenopausal patients with breast cancer in the intention-to-treat population of TEXT, and 1967 (74 %) are included in this translational study. The CYP19A1 rs10046 variant T/T, represented in 23 % of women, was associated with a reduced incidence of grade ≥2 hot flashes/sweating (univariate odds ratio (OR) = 0.78; 95 % CI 0.63–0.97; P = 0.03), more strongly in patients assigned exemestane + OFS (TT vs CT/CC: OR = 0.65, 95 % CI = 0.48–0.89) than assigned tamoxifen + OFS (OR = 0.94, 95 % CI = 0.69–1.27, interaction P = 0.03). No association with any of the CYP19A1/ESR1 genotypes and musculoskeletal AEs was found. Conclusion The CYP19A1 rs10046 variant T/T favors lower incidence of hot flashes/sweating under exemestane + OFS treatment, suggesting endocrine-mediated effects. Based on findings from others, this SNP may potentially enhance treatment adherence and treatment efficacy. We plan to evaluate the clinical impact of this polymorphism during time, pending sufficient median follow up. Trial registration ClinicalTrials.gov NCT00066703, registered August 6, 2003.


Background
Adjuvant endocrine therapy significantly prolongs diseasefree and overall survival in women with hormonereceptor-positive early breast cancer, but it is associated with several side effects, which may lead to early treatment cessation [1][2][3]. In the combined analysis of the Tamoxifen and Exemestane Trial (TEXT) and Suppression of Ovarian Function Trial (SOFT) [4], comparing adjuvant exemestane plus ovarian function suppression (OFS) with tamoxifen plus OFS in premenopausal patients with breast cancer, early cessation of OFS and the assigned oral endocrine treatment occurred in 16 % of patients receiving exemestane + OFS and 11 % of those receiving tamoxifen + OFS. Nonetheless, exemestane + OFS significantly improved disease outcome compared to tamoxifen + OFS after 5.7 years median follow up.
The acute onset of menopause induced by gonadotropinreleasing-hormone analogues (GnRHa) is associated with more frequent and severe side effects compared to natural menopause, significantly impacting the quality of life of young patients with breast cancer [5]. The most common side effects associated with early menopause include vasomotor symptoms (hot flashes and sweating), decreased libido, insomnia, and dyspareunia secondary to vaginal dryness. The frequency and severity of hot flashes may depend on the abrupt fall in circulating estrogen levels as observed in several studies among women undergoing a natural menopausal transition [6][7][8], although other factors also play a role [9,10]. While chemotherapy, OFS, and aromatase inhibitors (AIs) directly lower circulating estrogen levels, tamoxifen, a selective estrogen receptor modulator, has both agonistic and antagonistic effects on estrogen signaling [11]. In addition to menopausal symptoms, AIs are frequently associated with joint and muscle pain [12], decreased bone density [13] and risk of fracture [3], which appears to increase with better compliance with AIs [14].
Common genetic polymorphisms of the genes involved in estrogen production and estrogen target genes have been linked to breast cancer risk, prognosis, treatment response and side effects. One of these genes, the CYP19A1, encodes for the enzyme aromatase that promotes the bioconversion of androgens to estrogens. Genetic variations at the CYP19A1 locus may result in increased or decreased aromatase activity and influence concentrations of circulating estrogens [15][16][17]. For example, the rs10046 and rs4646 variants, located in a 3' untranslated region, were associated with higher estradiol and estrone levels due to increased aromatase activity. Alternatively, these variants could be linked with other gene variants such as the rs749292, which is associated with even higher estrogen levels [17]. A recent review and meta-analysis analyzed the influence of common CYP19A1 polymorphisms on postmenopausal patients with breast cancer treated with AIs [18], indicating a certain heterogeneity between studies.
The estrogen receptor α (ESR1) gene was recently recognized as a low-penetrance breast cancer susceptibility gene. Numerous studies suggest an association between ESR1 gene polymorphisms and breast cancer risk [19]. However, results have been controversial due to heterogeneous data sources, differences in study designs, ethnic background, disease status, and sample size. ESR1 is an important mediator of endocrine pathways involved in breast cancer risk and outcomes, including endocrine treatment response and side effects. Genetic polymorphisms altering the expression of ESR1 have been suggested to affect breast cancer susceptibility [20]. In particular, the restriction enzymes XbaI (rs9340799) and PvuII (rs2234693) have been extensively evaluated. Both are located in the first intron of the ESR1 gene. The association between variant allele T of ESR1 PvuII (C > T) and breast cancer appears to be linked to a higher transcriptional activity of the variant gene [21] and correlated with circulating estrogen levels [22].
A recent meta-analysis [23] found that menopausal status modifies breast cancer risk associated with ESR1 PvuII (C > T), with premenopausal variant carriers being at higher risk, possibly related to differences in circulating estrogen levels [22]. However, an updated meta-analysis restricted the effect to the Asian population [24]. Another meta-analysis of almost 19,000 individuals in eight European centers reported that ESR1 XbaI (A > G) protects against overall fracture risk [25], suggesting an involvement of these polymorphisms in bone metabolism. These ESR1 polymorphisms have also been described to be involved in ovarian hyperstimulation response in assisted reproduction studies [26], further highlighting their role in endocrinerelated mechanisms.
Within the phase III TEXT trial in which 2672 premenopausal women were randomized to adjuvant therapy with exemestane + OFS or tamoxifen + OFS, with or without adjuvant chemotherapy, we prospectively designed a translational research project for blood collection to investigate the effect of selected single nucleotide polymorphisms (SNPs) on treatment efficacy and toxicity. The purpose of the current analysis was to investigate the association of common genetic variants of CYP19A1 (rs10046, rs4646) and ESR1 (rs2077647, rs2234693 and rs9340799) with early-onset vasomotor and musculoskeletal symptoms.

Patients
TEXT is an International Breast Cancer Study Group (IBCSG)-coordinated, randomized, phase III trial that enrolled premenopausal women with histologically proven estrogen receptor (ER) and/or progesterone receptor (PgR)positive early breast cancer. From November 2003 through April 2011, patients were enrolled within 12 weeks from surgery, prior to the initiation of any systemic adjuvant therapy, and randomized to 5 years of exemestane + OFS or tamoxifen + OFS. OFS was achieved by monthly injection of the GnRHa triptorelin; bilateral oophorectomy or ovarian irradiation was allowed after at least 6 months of triptorelin. Chemotherapy was optional and, if administered, triptorelin and chemotherapy were started concomitantly; oral endocrine treatment was started after the completion of chemotherapy, or if chemotherapy was not administered, it was started 6 to 8 weeks after the initiation of triptorelin, to allow for the suppression of ovarian estrogen production.

Trial procedures
Targeted adverse events (AEs) were systematically collected, using the Common Terminology Criteria for Adverse Events (CTCAE) Version 3.0, at baseline and every 3 months during the first year of treatment: hot flashes was graded 1-3 (1, mild; 2, moderate; 3, interfering with activities of daily living (ADL)); sweating was graded 1-2 (1, mild and occasional; 2, frequent or drenching); and musculoskeletal symptoms, i.e., myalgia, arthralgia (joint pain), stiffness, were graded 1-4 (1, mild pain, not interfering with function; 2, moderate pain, pain or analgesics interfering with function but not interfering with ADL; 3, severe pain, pain or analgesics severely interfering with ADL; 4, disabling). Depending on institutional procedures, patients may have been systematically asked about targeted AEs during the clinical visit, or targeted AEs may have been recorded in the CRFs based on review of the medical reports.
Protocol amendment 2, dated July 2008, increased the sample size and added the collection of a single whole blood sample for DNA isolation for translational research objectives, i.e., to investigate treatment tolerability and disease outcome according to genetic polymorphisms. Samples and consent were prospectively collected for approximately 600 patients enrolled after the amendment, and approximately 2000 patients enrolled prior to the amendment were asked to re-consent and have samples collected at the next scheduled protocol visit. The translational protocol targeted collection was of 2000 total samples.

Blood collection, DNA extraction and genotyping assays
Venous blood was collected into EDTA-treated tubes provided by IBCSG and either processed and stored locally at −80°C or shipped immediately to the CALGB Pathology Coordinating Office (USA and Canada), for DNA extraction and temporary storage, until shipping to the IBCSG central biomarker laboratory at the European Institute of Oncology for biobanking, DNA extraction (all countries except USA and Canada) and genotyping. Genomic DNA was extracted with QIAamp DNA Blood Kits (Qiagen, Valencia, CA, USA), according to the manufacturer's instructions and extraction was performed by the automated platform "QIAcube" (Qiagen, Valencia, CA, USA).
The germline DNA samples were genotyped for SNPs in CYP19A1 (rs4646 and rs10046) and ESR1 (rs207764, rs2234693 and rs9340799). All samples were analyzed using a real-time sequencing method called pyrosequencing (Diatech Pharmacogenetics S.r.l., Jesi, Italy). The DNA was amplified by polymerase chain reaction (PCR) with biotinylated primers on the Real-Time PCR Cycler "Rotor-Gene TM 6000" (Corbett Research, Sydney, Australia), whereas single-stranded DNA templates were prepared using the PyroMark Vacuum Prep Workstation (Biotage, Uppsala, Sweden). The pyrosequencing analysis was performed on the PyroMarkTM Q96 ID instrument (Biotage). Control samples, representing a complete set of genotypes (wt/wt; wt/v; v/v) for all SNPs, were processed in each run. No patient sample failed genotyping.

Statistical analysis
The analysis included 1967 patients from the TEXT intention-to-treat population (ITT) who gave whole blood for genetic profiling (Fig. 1). The endpoint of early-onset hot flushes/sweating was defined as presence or absence of grade 2 or grade 3 hot flashes or grade 2 sweating reported at the 3 month or 6 month visits after randomization. Early-onset musculoskeletal symptoms were defined as presence or absence of grade 2-4 musculoskeletal symptoms reported at the visits at 3, 6, 9 or 12 months after randomization.
Logistic regression modeling assessed the association of the selected genotypes with presence of early-onset AEs. The model also adjusted for patient characteristics at randomization: age (<45 versus ≥45 years); menstruation status (normal versus irregular versus persistent amenorrhea); body mass index (BMI) (normal (<25), overweight (25-29.9) versus obese (≥30) kg/m 2 )); adjuvant chemotherapy use (yes versus no); treatment assignment (exemestane + OFS versus tamoxifen + OFS); presence of hot flashes/sweating of any grade at baseline; and presence of musculoskeletal symptoms of any grade at baseline. Because concomitant medications may affect the reported AE severity, the impact of relevant concomitant medication use (yes versus no) prior to or continuing at baseline, or introduced during the relevant endpoint time period, was investigated in a sensitivity analysis. Concomitant medications, prescribed for any reason, that might affect the severity of hot flashes/sweating included venlafaxine, SSRIs, clonidine, gabapentin, pregabalin and herbals [27]; medications for musculoskeletal symptoms such as nonsteroidal anti-inflammatory drugs (NSAIDs), glucosamine, corticosteroids, gabapentin, and pregabalin. The analyses also assessed whether the association varied by treatment assignment by including genotype variants-by-treatment interaction in the logistic regression models.
We first assessed SNP variant effects in an additive genotype model that compared 0 versus 1 versus 2 minor or variant alleles using a one-degree-of-freedom trend test. The minor or variant homozygote effect was assessed in a recessive model that compared the minor or variant homozygote versus the combined heterozygote and wildtype homozygote (reference group) using the chi-squared test. Hardy-Weinberg equilibrium (HWE) for genotype frequencies was tested using the Monte Carlo simulation method [28] to calculate the P value in order to avoid the reliance on the underlying chi-square approximation.
The study is presented in accordance with the Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) criteria [29]. All statistical tests were two-sided, without adjustment for multiple comparisons, and a P value <0.05 in the overall cohort or interaction P value ≤0.10 was considered as statistically significant. For a given sample size, assuming 10 % or 20 % homozygous variant and 43 % and 26 % AE rates, the detectable differences in AE rates for homozygous versus combined heterozygous and wild-type would be in the range of 11 % to 7.2 % (Fisher's exact test, two-sided α = 0.05, power ≥80 %).

Study population
Blood for germline DNA extraction from 1967 consenting women was collected and assessed, representing 74 % of the entire TEXT ITT population of 2660 (Fig. 1). Patients in the analytical cohort were representative of the TEXT trial (Table 1), with the exceptions of race (one country did not participate and some centers with a majority of hispanic ethnicity had low participation rates) and early discontinuation of protocol treatment (retrospective nature of sample collection). Most patients were Caucasian (92 %), median age was 44 years and median body mass index was 24 kg/ m 2 . Adjuvant chemotherapy was given to 58 % of patients.
At baseline, any grade (≥1) of hot flashes and sweating were reported in 8 % and 7 % of patients, respectively, while any grade (≥1) of musculoskeletal symptoms were reported in 15 % of patients ( Table 1).
The reference SNP numbers, minor allele frequencies and genotype frequencies for each analyzed SNP are listed in Table 2. No deviations from Hardy-Weinberg equilibrium were observed. Occurrence and grade of hot flashes, sweating and musculoskeletal side effects during the first year of protocol therapy, overall and by treatment assignment, are depicted in Table 3.
Association of CYP19A1 and ESR1 with early-onset hot flashes/sweating A total of 43 % of patients reported early-onset grade 2-3 hot flashes/sweating during the first 6 months of protocol treatment (42 % of women receiving exemestane + OFS (411/987) and 45 % of women allocated to tamoxifen + OFS (437/980)). Most side effects were reported by the month-3 visit (Table 3). Overall, patients with CYP19A1 rs10046 (C > T) minor variant (T/T) had a 22 % reduced odds of reporting early-onset grade 2-3 hot flashes/sweating (odds ratio (OR) = 0.78, 95 % CI 0.63-0.97; P = 0.03) when compared to patients with the C/T or C/C variants.  (Table 4). A differential effect according to treatment assignment (treatment-by-genotype interaction, P = 0.03) was observed for the association between CYP19A1 rs10046 (C > T) genotype variants and earlyonset hot flashes/sweating. Patients treated with exemestane + OFS and having the T/T variant had a 35 % reduced odds of early-onset hot flashes/sweating ( Adverse events according to common terminology criteria for adverse events (CTCAE) v3.0 of hot flashes and/or sweating reported at 3 or 6 months after randomization; musculoskeletal symptoms, i.e., myalgia, arthralgia (joint pain), or stiffness, reported at 3, 6, 9 or 12 months after randomization. TEXT Tamoxifen and Exemestane Trial, BMI body mass index, IQR interquartile range, HER2 human epidermal growth factor receptor 2, OFS ovarian function suppression   (7) 44 (4) was no statistically significant association between any of the five SNPs of CYP19A1 and ESR1 and early-onset musculoskeletal side effects (  Reports of hot flashes and sweating side effects and of musculoskeletal symptoms according to common terminology criteria for adverse events (CTCAE) v3.0 at time points during the first year of protocol therapy. All data are summarized as number (%) of patients. a Patients without any adverse event data (two patients without hot flashes/sweating and one without musculoskeletal symptoms) were excluded from summary. Unk unknown, Gr grade OFS ovarian function suppression

Discussion
This study provides evidence that CYP19A1 rs10046 variant carriers may face milder vasomotor symptoms under combined endocrine treatment. Notably, the effect was restricted to patients under OFS combined with exemestane (treatment-by-genotype interaction, P = 0.03) and not tamoxifen, after adjusting for patient characteristics and concomitant medications, including the selective serotonin-reuptake inhibitors known to reduce hot flashes/sweating. This finding is in line with evidence from others, linking this SNP to enhanced aromatase activity and higher circulating estrogens [15,17] and underscores a possible relationship between the effect of this variant polymorphism (T/T), hot flashes/sweating and exemestane activity. This result may in fact be related to less effective estrogen suppression by exemestane + OFS in these women as a consequence of higher circulating estrogens compared to patients with wild-type SNPs, although the exact mechanism by which this SNP may affect exemestane efficacy in suppressing the aromatase activity is not known. One study recently reported similar associations in postmenopausal patients with breast cancer [30] enrolled in the TEAM trial: CYP19A1 variants linked with lower estrogen levels were associated with increased risk of early vasomotor and musculoskeletal symptoms under exemestane. The TEAM substudy, however, only included 27 % of the patients enrolled, less than two-thirds of patients, which represents a smaller proportion than is recommended by Simon et al. for evaluating predictive biomarkers [31].
The ELPh trial was designed to address genetic associations with toxicity-related discontinuation of AI therapy for breast cancer [32], including the SNP rs10046. The authors did not specifically report on vasomotor symptoms, but did not find any relationship between rs10046 and toxicity-related treatment discontinuation. In another study, in which the impact of CYP19A1 SNPs with estrogen suppression during letrozole treatment was assessed, the degree of suppression was independent of the SNPs [33].
To our knowledge this is the first study to evaluate the associations between common germline polymorphisms of the CYP19A1 and ESR1 genes and early-onset side effects under combined endocrine treatment in premenopausal patients with hormone receptor-positive early breast cancer. The strength of this translational research is its considerable sample size of 1967 patients, which represents 74 % of women enrolled in TEXT. Furthermore, blood samples were collected specifically for this research, i.e., to investigate treatment tolerability and disease outcome. Women enrolled prior to the amendment were asked to re-consent, but 693 TEXT participants were not assessed due to the retrospective nature of blood collection. As a result, we may have missed some patients who discontinued treatment early, possibly due to treatment-related side effects.
The combined analysis of TEXT and SOFT [4] showed that adjuvant treatment with exemestane + OFS as compared with tamoxifen + OFS, significantly reduces the risk of recurrence. Although the overall incidence of adverse events and the quality of life were similar in the two treatment groups, between-group differences were observed with respect to specific symptoms. While vasomotor AEs (hot flashes and sweating) were quite frequent and evenly distributed amongst treatment groups, Estimates from univariate logistic regression model. b P value from test of rs10046 variants ((T/T) vs. T/C, C/C) by treatment interaction in logistic regression model (univariable) assessing association between the SNP variants and early-onset adverse events in the overall cohort. c Adjusted for baseline characteristics: age, menstrual status, body mass index, adjuvant chemotherapy use, treatment assignment (for "all patients" cohort), baseline hot flashes/sweating or baseline musculoskeletal symptoms (according to endpoint) and prior to or baseline concomitant medications use and use during relevant time period for the endpoint (yes or no). SNP single nucleotide polymorphism, OFS ovarian function suppression, OR odds ratio, CI confidence interval musculoskeletal AEs were more frequently reported in patients assigned to exemestane + OFS. We did not observe any direct association between the CYP19A1 SNPs and musculoskeletal symptoms, nor any interaction by endocrine treatment. This is in contrast with findings from the TEAM trial [30], but as mentioned they studied a very small proportion of patients. Furthermore, genotyping in that study was performed on DNA extracted from tumor samples. A cross-sectional study of patients receiving AIs [34] found that women carrying at least one 8-repeat allele of the tetranucleotide repeat polymorphism of CYP19A1, associated with higher estrogen concentrations, had lower odds of AI-associated arthralgia. Conversely, they also did not find any association between the rs10046 SNP and arthralgia.
Contrary to findings from case-control studies conducted in different treatment settings, i.e., postmenopausal or premenopausal women with breast cancer treated with tamoxifen alone, we found no association between the three ESR1 polymorphisms and endocrine-mediated side effects (hot flashes/sweating and musculoskeletal symptoms). Postmenopausal Chinese patients with breast cancer carrying an ESR1 rs2234693 CC genotype or rs9340799 AA genotype had an increased risk of AIrelated musculoskeletal AEs [35]. In fact, several studies suggest that the effect of the ESR1 polymorphisms on breast cancer risk is hormone-related and dependent on the woman's hormonal context, showing statistically significant associations mainly in premenopausal women [23]. Likewise, an association with increased mammographic density [36] was shown only in women taking hormone replacement therapy. Possibly, the concurrent OFS by the GnRH analogue triptorelin masked the effect of these polymorphisms due to its complete estrogen deprivation effect. Thus, in the context of adjuvant combined endocrine treatment, these ESR1 polymorphisms may be unlikely to exert their effect.
Musculoskeletal events are a common toxicity, leading to premature discontinuation of AI therapy [37]. In the TEXT-SOFT combined analysis, early cessation of protocol treatment was more frequent among patients receiving exemestane + OFS than among those receiving tamoxifen + OFS. Several studies have investigated the relationship between endocrine treatment efficacy and associated side effects in different settings. Recent findings support an inverse association between the reporting of early side effects under adjuvant endocrine treatment and breast cancer recurrence [38][39][40]. Vasomotor symptoms were associated with improved disease-free and overall survival in the TEAM trial [38] and reduced breast cancer recurrence in the ATAC trial [40], but not in the BIG 1-98 [41] and MA.27 trials [42]. Thus, we cannot exclude that the CYP19A1 rs10046 (T/T) genotype might be associated with reduced exemestane + OFS efficacy: women with this polymorphism possibly lack complete estrogen suppression, despite receiving concomitant OFS. On the other hand, because the rs10046 polymorphism is located in a 3' untranslated region, upstream of the coding sequence, it may interfere with aromatase transcription in a tissuespecific manner, depending on the transcriptional modulators present, thus influencing the degradation rate of the aromatase differently according to tissue and independently from circulating estrogen [9].

Conclusions
This translational study within the TEXT trial for premenopausal patients with hormone-receptor-positive early breast cancer provides evidence that the CYP19A1 rs10046 polymorphism may influence endocrine treatment side effects under combined endocrine therapy. The CYP19A1 rs10046 variant favors lower incidence of hot flashes/sweating under exemestane plus ovarian function suppression treatment, suggesting endocrine-mediated effects that might enhance treatment adherence and potentially impact long-term treatment efficacy. No effect of any other tested SNPs was evident on hot flashes/sweating and no interaction on musculoskeletal symptoms emerged overall. Monitoring of musculoskeletal and bone events, known to occur later during treatment are warranted. Although our results must be considered hypothesisgenerating, longer follow up will allow us to assess the clinical relevance of this finding, in particular its potential impact on disease outcome, and will be the subject of a future report after the TEXT results are further updated.