Association between germline single nucleotide polymorphisms in the PI3K-AKT-mTOR pathway, obesity, and breast cancer disease free survival

Abstract

Purpose

Obesity-related hormones and cytokines alter PI3K-AKT-mTOR pathway activation in breast tumors contributing to poorer disease-free survival (DFS) and decreased responsiveness to tamoxifen and trastuzumab. We hypothesized that single nucleotide polymorphisms (SNPs) in candidate genes in the PI3K-AKT-mTOR signaling pathway may act as genetic modifiers of breast cancer DFS.

Methods

We analyzed the association of 106 tagging SNPs in 13 genes (ADIPOQ, IGF1, INS, IRS1, LEP, LEPR, LEPROT, PIK3CA, PIK3R5, PTEN, TSC1, TSC2, and AKT1) in the P13K-AKT-mTOR pathway with DFS in a sample of 1,019 women with stage I–II breast cancer. SNPs significantly associated with DFS in any genetic model (additive, dominant or recessive) after correcting for false discovery rate (FDR=0.10) were included in Cox proportional hazards multivariable analyses.

Results

After adjusting for race/ethnicity, age at diagnosis, tumor stage and treatment, rs1063539 in ADIPOQ, rs11585329 in LEPR, and rs2519757 in TSC1 were associated with improved DFS and rs1520220 in IGF1 and rs2677760 in PIK3CA were associated with worse DFS. The associations were not significantly modified by type of systemic treatment received or body mass index. The SNPs were not associated with tumor characteristics such as tumor size, lymph node status, nuclear grade, or hormone receptor status.

Conclusion

In this study, germline SNPs in the PI3K-AKT-mTOR pathway were associated with breast cancer DFS and may be potential prognostic markers. Future studies are needed to replicate our results and to evaluate the relationship between these polymorphisms and activation of the PI3K-AKT-mTOR pathway in breast tumors.

Background

Obesity is an established risk factor for postmenopausal breast cancer and expanding epidemiological evidence suggests that obesity is independently associated with breast cancer recurrence and survival regardless of menopausal status [1,2]. Although the modifying effect of obesity on prognostic breast tumor characteristics has been studied extensively [3,4], there are limitations in our understanding of the contribution of genes involved in obesity-related signaling pathways to breast cancer outcomes.

Growth factors (e.g. IGF-1) and obesity-related hormones and cytokines (e.g. insulin, adiponectin) interact with mammalian target of rapamycin (mTOR) through activation of the phosphatidylinositor-3 kinase (PI3K)-AKT pathway [5,6]. Activation of mTORC1 either through upstream insulin or growth factor receptor signaling or genetic/epigenetic abnormalities promotes cell proliferation and inhibits apoptosis through the phosphorylation of downstream targets including p70S6K, and subsequently 4E-binding proteins (4E-BPs), and ribosomal protein S6 kinase 1 (S6K1) resulting in enhanced protein translation [5,6]. PI3K-AKT-mTOR pathway activation has been associated with decreased responsiveness to a wide range of breast cancer therapies including tamoxifen, aromatase inhibitors and trastuzumab [7–14]. Recently, the mTOR inhibitor everolimus was approved in the U.S for the treatment of metastatic breast cancer resistant to aromatase inhibitors [7].

Despite the importance of the PI3K-AKT-mTOR pathway in breast cancer and the link with obesity, there are no studies to date that have evaluated the role of germline variations in the PI3K-AKT-mTOR pathway and breast cancer outcomes. Prior epidemiological studies have shown significant associations between single nucleotide polymorphisms (SNPs) of candidate genes in the PI3K-AKT-mTOR pathway and increased risk of death from bladder cancer [15,16] and recurrence of esophageal cancer [17]. Promising candidates for investigation include obesity related genes involved in the initiation of pathway activation such as INS, IGF1, IRS1, ADIPOQ and genes that represent the core functional components of the pathway such as PIK3CA, mTOR, AKT and the tumor suppressor gene, PTEN. This study utilizes a well characterized cohort of patients with early stage breast cancer with detailed information on tumor characteristics and treatment to determine whether tagging SNPs selected for each of the candidate genes in the PI3K-AKT-mTOR pathway are associated with breast cancer disease-free survival (DFS).

Methods

Study population

The Early Stage Breast Cancer Repository (ESBCR) is a retrospective cohort of 2,409 women diagnosed with American Joint Committee on Cancer pathologic stage I or II breast cancer and surgically treated at The University of Texas MD Anderson Cancer Center between 1985 and 2000. Criteria for eligibility and cohort details have been previously described [18]. Briefly, we selected a subset of the ESBCR population with available formalin-fixed paraffin-embedded normal lymph nodes or blood samples available for DNA extraction for participation in the genotyping project. Detailed clinical information including patient age, race/ethnicity, stage, tumor size, lymph node status, nuclear grade, estrogen receptor (ER) and progesterone receptor (PR) status, and primary treatment (including surgery, radiation therapy, chemotherapy and endocrine therapy) were abstracted from medical charts. Follow-up information for patients in the ESBCR is obtained by direct review of the medical records to determine dates of relapse or death, and linkage to the MD Anderson Cancer Center Tumor Registry, which mails annual follow-up letters to each patient registered at MD Anderson Cancer Center known to be alive to determine their clinical status. The MD Anderson Tumor Registry checks the social security death index and the Texas Bureau of Vital Statistics for the status of patients that fail to respond to letters. We selected a subset of the ESBCR population for participation in the genotyping project enriched to include all African American (n=196) and Hispanic patients (n=208) and a random sample of non-Hispanic White patients (n=986). Of these, 301 patients were excluded because of DNA extraction failure, 54 because of genotype assay failure and 6 because of inadequate clinical information. The final study analysis included 1029 breast cancer patients.

SNP selection and genotyping

We reviewed the literature and selected the following candidate genes of interest in the PI3K-AKT-mTOR pathway: ADIPOQ, IGF1, INS, IRS1, LEP, LEPR, LEPROT, PIK3CA, PIK3R5, PTEN, TSC1, TSC2 and AKT1. Tagging SNPs were selected within and 2-kb flanking each gene using an algorithm by Carlson et al [19] based on data from the HapMap project (www.hapmap.org Version 23) [20]. All selected SNPs met the following criteria: minor allele frequency (MAF) of ≥0.05, Illumina design score >0.04 and r2 >0.8 for binning. Genotyping of the breast cancer cases was performed using Illumina GoldenGate technology (Illumina, San Diego, CA) as part of a larger array of 1,514 SNPs in total. Briefly, genomic DNA was extracted from peripheral blood using the QIAamp DNA Blood Maxi kit (QIAGEN, Valencia, CA) according to the manufacturer’s protocol [21]. Genomic DNA was extracted from formalin-fixed paraffin embedded (FFPE) tissue using the Pico Pure DNA extraction kit [22]. The success call rates for the FFPE and non-FFPE samples were 92% and 99% respectively. Blinded duplicate samples (5%) were included in the array platform and the reproducibility rate was 100%. All genotyping information was analyzed and exported using the GenomeStudio software (Illumina).

Statistical methods

Principal component analysis was performed for detecting population structure and potential outliers within each ethnic group (Golden Helix SNP and Variation Suite, Golden Helix Inc, Bozeman, MT). SNPs with genotype missing in >10% of individuals, or with MAF <0.05 were also excluded. Breast cancer DFS was calculated as the time from the date of first treatment to the date of breast cancer recurrence or date of death due to breast cancer, whichever came first.

To evaluate the relationship between each SNP and DFS, univariate Cox proportional hazard regression analysis was performed using three genetic models: additive, dominant and recessive. Correction for multiple testing was performed on univariate SNP analysis results using Benjamini-Hochberg FDR method [23]. SNPs with Wald statistic Chi-square p<0.1 in univariate analyses for any of the 3 models were selected and pairwise linkage disequilibrium (LD) test was performed. SNPs with R² > 0.8 were excluded. Remaining SNPs were used as input variables for multi-SNP analysis. For each SNP, the model (additive, dominant or recessive) with the lowest p-value was used. Stepwise multivariable Cox regression analyses were performed for clinical/phenotypic variables race/ethnicity, age of diagnosis, nuclear grade, presence of ER or PR, stage, treatment (endocrine therapy) and body mass index (BMI). We categorized body mass index (BMI) based on the cut points for normal weight (BMI <25 mg/kg²) and overweight or obese (BMI ≥25mg/kg²). Variables with Wald statistics p<0.05 in the multivariable model were selected as covariates for inclusion in the multivariable model along with SNPs described above. Race was not significant at Wald p<0.05 but was included in the model to account for any residual confounding. Stepwise analysis was performed using all individuals in the dataset as discovery set. Threshold for significance for SNPs in the multivariable model was p<0.05. Multivariable analyses were also performed stratified to evaluate the effect of individual SNPs by race and BMI. Associations were quantified using hazard ratios (HRs) and 95% confidence intervals (CIs). Kaplan-Meier analysis was performed to estimate survival over time and test homogeneity between strata within a SNP. We used SAS version 9.1 (SAS Institute, Cary, NC) and R (http://www.r-project.org) to perform all of our analyses.

Results

From the subset of 1029 ESBCR patients an additional ten outliers (2 white, 6 African-American and 2 Hispanic patients) were excluded based on principal component analysis resulting in 1019 women in the final analyses. Of these 72% were non-Hispanic white, 13% were African American, 11% were Hispanic and 3% were of unknown race/ethnicity. Demographic and clinical characteristics of the study population and the univariate associations with breast cancer DFS are listed in Table 1. One hundred six tagging SNPs in 13 genes related to the PI3K-AKT-mTOR pathway were analyzed. Of these 106 SNPs, 27 SNPs significantly associated with increased or decreased DFS in any genetic model (additive, dominant or recessive) were selected with p<0.1 in univariate Cox regression analysis. Only 1 SNP out of the 27 was excluded due to high LD (R²>0.8) (rs5742629 rs7956547 had LD=0.84, and SNP rs5742629 was excluded), leaving 26 SNPs for inclusion in multivariable analyses. In the univariate model, age at diagnosis, tumor stage and treatment were identified as significant clinical covariates for breast cancer DFS (Table 1). After adjusting for these covariates, as well as for race/ethnicity, 5 of the 26 SNPs were significantly associated with breast cancer DFS (p<0.05) (Table 2). SNPs rs1063539 in ADIPOQ (HR = 0.40, 95% CI = 0.19 – 0.86), rs11585329 in LEPR (HR = 0.72, 95% CI = 0.53 – 0.98), and rs2519757 in TSC1 (HR = 0.29, 95% CI = 0.09 – 0.91) were associated with improved DFS and rs1520220 in IGF1 (HR = 1.56, 95% CI = 1.21 – 2.00) and rs2677760 in PIK3CA (HR = 1.43, 95% CI = 1. 07 – 1.92) were associated with worse DFS. The associations were not modified by type of systemic treatment received and the individual SNPs were also not associated with tumor characteristics such as tumor size, lymph node status, nuclear grade, or hormone receptor status. Kaplan-Meier analysis for breast cancer DFS by IGF1 rs1520220 (the SNP with the smallest association p-value), indicated a statistically significant difference (log-rank test p-value=0.0004) between the homozygous wild-type genotype (WW) and the heterozygous and homozygous variant genotypes (WM/MM) which were the genotypes associated with worse DFS (Fig. 3). In analysis stratified by race/ethnicity, 4 out of the 5 SNPs were significantly associated with breast cancer DFS among Whites. None of the 5 SNPs were statistically significant among Black (n=135) or Hispanic patients (n=115), however the sample size for these groups was potentially too small to provide power to detect associations.

Table 1.

Univariate associations of demographic and clinical characteristics with breast cancer disease-free survival (DFS)

Covariate		No Event	Event
		n(%)	n(%)	HR (95% CI) a
Years of survival, mean (range)		10.5(0.2,22.2)	4.7(.2,21.5)
Age at diagnosis, years
	<50	288(38.1)	146(55.5)	Reference
	≥50	468(61.9)	117(44.5)	0.58(0.46–0.74)
Race
	White	550(72.8)	186(70.7)	Reference
African American		97(12.8)	38(14.4)	1.18(0.83–1.67)
	Hispanic	86(11.4)	29(11)	1.12(0.76–1.65)
	Unknown	23(3)	10(3.8)	-
Stage
	I	251(33.2)	51(19.4)	Reference
	II	505(66.8)	212(80.6)	1.87(1.38–2.55)
Tumor size (cm)
	0	35(4.6)	16(6.1)	Reference
	>0–2	462(61.1)	115(43.7)	0.61(0.36–1.02)
	>2	252(33.3)	127(48.3)	1.07(0.63–1.79)
	Unknown	7(0.9%)	5(1.9)	-
Lymph node status
	Negative	459(60.7)	132(50.2)	Reference
	Positive	297(39.3)	131(49.8)	1.41(1.10–1.79)
Nuclear grade
	1	251(33.2)	94(35.7)	Reference
	2	405(53.6)	129(49)	0.86(0.66–1.12)
	3	72(9.5)	25(9.5)	0.98(0.63–1.53)
	Unknown	28(3.7)	15(5.7)	-
ER/PR
	Negative	162(21.4)	72(27.4)	Reference
	Positive	562(74.3)	174(66.2)	0.72(0.55–0.95)
	Unknown	32(4.2)	17(6.5)	-
BMI (kg/m2)
	<25	314(41.5)	111(42.2)	Reference
	≥25–<30	212(28)	82(31.2)	1.09(0.82–1.45)
	≥30	213(28.2)	62(23.6)	0.91(0.66–1.24)
	Unknown	17(2.2)	8(3)	-
Treatment (Hormone and Chemotherapy)
	Neither	162(21.4)	66(25.1)	Reference
	Both	386(51.1)	150(57)	0.97(0.73–1.30)
	Hormone only	204(27)	47(17.9)	0.63(0.43–0.92)
	Unknown	4(0.5)	-	-

^aHR – hazard ratio; CI – confidence interval

Table 2.

SNPs significantly associated with breast cancer disease-free survival in multivariable Cox regression analysis

SNP	Gene	Location	MAFa	Model	p	HRa	95% CIa
rs1063539	ADIPOQ	3UTR	0.11	Dominant	0.02	0.40	0.19–0.86
rs11585329	LEPR	Intron	0.13	Recessive	0.04	0.72	0.53–0.98
rs1520220	IGF1	Intron	0.20	Recessive	0.001	1.56	1.21–2.00
rs2519757	TSC1	Intron	0.05	Dominant	0.03	0.29	0.09–0.91
rs2677760	PIK3CA	Intron	0.45	Recessive	0.02	1.43	1.07–1.92

^aMAF, minor allele frequency; HR, hazard ratio; CI, confidence interval.

In the analysis stratified by BMI (< 25mg/kg² and ≥ 25 mg/kg²) (Table 3), The strength of the association between SNP rs1063539 in ADIPOQ and improved DFS was increased in patients with normal (HR 0.33; 95% CI 0.12–0.90) compared to overweight or obese status (HR 0.63; 95% CI 0.20–2.01). On the contrary, the strength of the association between rs1520220 in IGF1 and rs2677760 in PIK3CA and worse DFS was statistically significantly increased in the overweight and obese group but not in those with normal weight (Table 3). None of the interaction terms were statistically significant (p>0.05).

Table 3.

Association of individual SNPs with breast cancer disease-free survival stratified by body mass index (BMI)^a

BMI kg/m2	SNP	Gene	Model	P	HRb	95% CIb
<25	rs1063539	ADIPOQ	Dominant	0.03	0.33	0.12	0.90
≥2				0.44	0.63	0.20	2.01
<25	rs11585329	LEPR	Recessive	0.27	0.78	0.51	1.21
≥25				0.06	0.65	0.42	1.02
<25	rs1520220	IGF1	Recessive	0.06	1.45	0.99	2.14
≥25				0.01	1.61	1.15	2.25
<25	rs2519757	TSC1	Dominant	0.46	0.48	0.07	3.49
≥25				0.003	0.11	0.03	0.47
<25	rs2677760	PIK3CA	Recessive	0.59	1.13	0.72	1.78
≥25				0.04	1.50	1.01	2.21

^aAdjusted for race/ethnicity, age at diagnosis, cancer stage, and treatment

^bHR, hazard ratio; CI, confidence interval

Discussion

We evaluated the associations between SNPs in candidate genes involved in the PI3-KAKT-mTOR signaling pathway and breast cancer DFS among a large cohort of patients with early stage breast cancer who received long-term follow-up. Evidence was strongest for a role of the IGF1 rs1520220 SNP in reducing breast cancer DFS. Interestingly, the rs1520220 SNP in IGF-1 has been shown to be associated with circulating serum levels of IGF-1 [24,25]. IGF-1 ligand binding with the IGF-1 receptor, leading to activation of downstream signaling pathways including the mitogen-activating protein kinase/extracellular signal-related kinase (ERK) and the PI3K pathway, contributes to IGF-1-mediated breast cancer cell proliferation and metastasis [26]. IGF-1 has been implicated in the survival of triple negative (ER, PR and Her2neu-negative) breast cancer cells [27] and with resistance to endocrine therapy in ER+ cell lines [28]. However, while high circulating serum levels of IGF-1 are a risk factor for developing breast cancer [29] there is conflicting epidemiological data on whether IGF-1 is associated with breast cancer survival or recurrence [30,31]. Given the active development of inhibitory IGF-1 receptor monoclonal antibodies and small molecule receptor inhibitors as a novel cancer therapeutic strategy [32], identifying breast cancer patients with genetic susceptibility to increased signaling via the IGF-1 receptor may play an important role in personalizing treatment.

Obesity negatively affects breast cancer prognosis through disrupting inflammatory cytokines and adipokines, such as adiponectin and leptin that have anti-inflammatory and insulin-sensitizing properties [15, 16]. Adiponectin suppresses cell growth, induces apoptosis and angiogenesis [33]. In our study of breast cancer patients, the minor genotype C/C for SNP rs1063539 in ADIPOQ was associated with an improved breast cancer DFS. In a previous study, this genotype of SNP rs1063539 allele in ADIPOQ (which is in LD with ADIPOQ SNPrs3774262) was associated with lower body weight, waist circumference and BMI in healthy female controls [34]. No association was observed in a nested case-control study between SNPs in the ADIPOQ gene including the rs1063539 SNP and breast cancer risk among postmenopausal women [35], suggesting that any modification in energy balance or obesity associated with this genetic variant in the ADIPOQ gene may have a stronger influence on tumor survival than initiation.

The biological mechanisms through which tagging SNPs rs11585329 in LEPR, rs2519757 in TSC1 and rs2677760 in PIK3CA (and the SNPs that they represent) influence breast cancer disease-free survival are not known and the possibility of a gene-gene or gene-environmental interaction cannot be ruled out. However, given the importance of the PI3K-AKT-mTOR pathway in breast cancer prognosis, evaluating the contribution of genetic variants of significant candidate genes to breast cancer DFS may have potential implications for identifying women at increased risk of recurrence or death who can be targeted for novel therapeutic interventions with mTOR inhibitors and IGFR1 monoclonal antibodies.

Our study had several limitations. First, we included only tagging SNPs therefore it is possible that a comprehensive screening approach for each candidate gene may have uncovered additional functionally relevant prognostic SNPs. Second, in this study set with long term follow-up, we were unable to measure serum levels of IGF-1, adiponectin or leptin and test for a correlation between the prognostic tag SNPs identified in IGF-1, ADIPOQ and LEPROT. In addition, we did not perform analyses according to tumor subtype although there was no evidence of an interaction of the significant SNPs with ER and PR status.

In summary, although we cannot rule out that the observed associations between SNPs in the PI3K-AKT-mTOR pathways and breast cancer DFS may have occurred by chance, the results of this study are biologically plausible given the functional significance of the tag SNPs rs1063539 in ADIPOQ and rs1520220 in IGF-1 and the important role that obesity plays in breast cancer survival. Prior to attributing clinical significance to the study findings, replication of these results is warranted with evaluation of the relationship between these polymorphisms and activation of the PI3K-AKT-mTOR pathway in breast tumors.

Fig. 1.

Kaplan-Meier graph comparing breast cancer disease-free survival for IGF1 rs1520220 homozygous wild-type genotype (WW) with that for heterozygous and homozygous variant genotypes (WM/MM)

Abbreviations

PI3K

phosphatidylinositol-3 kinase

mTOR

mammalian target of rapamycin

ADIPOQ

adiponectin, C1Q and collagen domain containing

IGF1

Insulin-like growth factor 1

INS

Insulin

IRS1

Insulin receptor substrate 1

LEP

Leptin

LEPR

Leptin receptor

LEPROT

Leptin receptor overlapping transcript

mTOR

mammalian target of rapamycin

mTORC1

mammalian target of rapamycin complex 1

PIK3CA

Phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit alpha

PIK3R5

Phosphoinositide-3-kinase, regulatory subunit 5

PTEN

Phosphatase and tensin homolog

TSC1

Tuberous sclerosis 1

TSC2

Tuberous sclerosis 2

AKT1

V-Akt murine thymoma viral oncogene homolog 1

SNP

single nucleotide polymorphism

DFS

disease-free survival

ESBCR

Early stage breast cancer registry

ER

estrogen receptor

PR

progesterone receptor

MAF

minor allele frequency

LD

linkage disequilibrium

HR

hazard ratio

CI

confidence interval