HMG-CoA reductase expression in breast cancer is associated with a less aggressive phenotype and influenced by anthropometric factors

Abstract

Although several studies have reported on the anti-tumoural properties exerted by 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMG-CoAR) inhibitors (statins), the in vivo expression of HMG-CoAR in human cancer has been considerably less investigated. In our study, we examined the immunohistochemical expression of HMG-CoAR in 511 incident breast cancers within the Malmö Diet and Cancer Study in order to explore its relationship to established clinicopathological and tumour biological parameters. Furthermore, the potential influence of estrogen exposure on HMG-CoAR expression was assessed by performing Cox's proportional hazards analyses of the relationship between the use of hormone replacement therapy (HRT), obesity (waist circumference) and tumour-cell specific HMG-CoAR expression. We found that HMG-CoAR was present in various fractions and intensities in the cytoplasm, sometimes with a membranous pattern, but not in the tumour cell nuclei. The expression of HMG-CoAR was associated with a smaller tumour size (p = 0.02), low histological grade (p = 0.001), low Ki67 index (p = 0.004), ERα+ (p = 0.02), ERβ+ (p = 0.005), and high p27 expression (p = <0.001). The incidence of tumours with a high HMG-CoAR-expression was increased among HRT-users, although this was not statistically significant in a heterogeneity analysis. Obesity was significantly associated with a high HMG-CoAR expression assessed both as a high (>50%) fraction of positive cells (relative risk: 2.06; 95% confidence interval: 1.20–3.51), and a strong staining intensity (2.33: 1.08–5.02). In summary, we demonstrate that HMG-CoAR is differentially expressed in breast cancer and that a high expression is associated with prognostically favourable tumour parameters. Moreover, estrogen related life-style and anthropometric factors might indeed regulate HMG-CoAR expression. © 2008 Wiley-Liss, Inc.

Cancer cells express constitutively elevated levels of the enzyme 3-hydroxy-3-methylglutharyl-coenzyme A reductase (HMG-CoAR).1 HMG-CoAR is rate-limiting in the mevalonate pathway, i.e., the cholesterol synthesis.2 However, the non-sterol products of the mevalonate pathway are essential for cell survival.1 HMG-CoAR is required for the synthesis of isoprenoids, which, in turn, are essential for post-translational modification of proteins that regulate cellular proliferation and apoptosis.3 Animal studies have shown that estrogens influence cholesterol metabolism,4, 5 which may be due to an estrogen-responsive region located in the promoter of the HMG-CoAR gene.6 A recent in vitro study on colon cancer cells confirmed that estrogen has a regulatory effect on HMG-CoAR activity.7

The importance of HMG-CoAR in cancer development is further indicated from studies on statins. Statins act as reversible HMG-CoAR inhibitors8 and are widely used in the treatment of hypercholesterolemia. Recent findings propose that statins exhibit anti-tumoural properties as well: In vitro studies on prostate and breast cancer cell lines suggest that lipophilic statins inhibit proliferation via arrest in the G1-phase of the cell cycle, probably through induction of cdk-inhibitors, e.g., p27,9, 10 and a decreased expression of G1-S phase stimulators.11 Some epidemiological studies report an up to 50% reduced risk of cancer among statin users,12–14 whereas others could not support the association between statin use and breast cancer.15, 16

Despite implications of a multifaceted role in cancer as well as estrogen-related activities, little is known about the tumour-specific expression of HMG-CoAR in breast cancer and its relationship to relevant clinicopathological parameters and molecular biomarkers. To address this issue, we analysed HMG-CoAR expression in tissue microarrays (TMAs) representing tumours from 511 incident invasive breast cancers within The Malmö Diet and Cancer Study (MDCS).17 We further aimed to analyse the association between tumour-specific HMG-CoAR expression and hormone related factors such as HRT use and obesity.18

Material and methods

The MDCS

The MDCS is a population-based prospective cohort study initiated in 1991, which enrolled 17,035 healthy women during the years of baseline examinations, 1991–1996. A questionnaire assessed education, occupation, marital status, age at menarche, parity, age at first childbirth, age at menopause, exposure to oral contraceptives (OC) (ever/never), medication, alcohol consumption, smoking- and dietary habits. Anthropometric measurements were conducted during baseline examinations. Information on gynaecological surgery was retrieved from hospital records. Breast cancer cases were ascertained by record linkage with the Swedish Cancer Registry until the end of follow-up, December 31, 2004. Ethical permission for the MDCS was obtained from The Ethical Committee at Lund University (LU 51–90).

Study cohort

Among the 17,035 female participants in the MDCS, 622 women were diagnosed with incident breast cancer, 511 of which were invasive cancer and 72 ductal carcinoma in situ (DCIS). In 39 cases, no tumour tissue could be retrieved from the archives. Hence, a total of 511 invasive breast cancers could be used for the correlation analyses. For analyses of HRT use, the postmenopausal cohort of 12,071 women was extracted and all prevalent breast cancer cases excluded. The total number of incident breast cancers in this cohort was 464,389 of which were invasive and the remaining 45 DCIS. In 30 cases tumour tissue was not available. HRT users in the postmenopausal cohort were excluded for analyses of anthropometric factors. Among 9685 postmenopausal women with no prevalent cancer and no HRT use at baseline, 305 women were diagnosed with incident breast cancer.

Ethical permission for our study was obtained from The Ethical Committee at Lund University (Dnr 652/2005).

Medications

HRT-use was assessed in 2 ways. Medications were recorded in a questionnaire using an open-ended question on current use, and in addition, all participants were asked to keep a diary of medications. Information was retrieved from both sources and is able to provide information on current use and non-use. Coding was performed according to the ATC classification system. Use of statins was only recorded for 228 patients in this cohort and could therefore not be used in the risk analyses.

Anthropometric measurements

At baseline, height and weight were measured by a trained nurse, and body mass index (BMI) was calculated as kg/m². Waist- and hip circumference was determined. In former studies, we have analysed several anthropometric measurements in relation to breast cancer risk, and waist circumference was the strongest predictor (unpublished data). Hence, waist circumference was used as marker of obesity in our study. Analyses were repeated using the more frequently applied anthropometric measurement, BMI.

Western blot validation

An immunoblot analysis was performed in order to determine and verify proper target recognition of the poly-clonal anti-HMG-CoA reductase antibody (Catalog # 07-457, Upstate/Millipore, MA). HMG-CoA reductase has predicted molecular weights of 97.476 and 92.020 kDa (Ensembl release 47, Ensembl Peptide ID ENSP00000287936 and ENSP00000340816) possibly corresponding to 2 different splice variants of 888 and 835 amino acids respectively. The antibody is raised against a keyhole limpet hemocyanin-conjugated synthetic peptide corresponding to 13 amino acids at the C-terminus of the full-length protein(s). Four different human tissue lysates in addition to 2 cell lysates were analyzed: Normal breast tissue, breast tumor tissue, normal prostate tissue, prostate tumor tissue, A-431 cells (skin epidermoid carcinoma), and Hep-G2 cells (liver, hepatocellular carcinoma). Twenty μg total human tissue- and cell lysates prepared by the manufacturer (ProSci, CA) were loaded onto a NuPAGE 4–12% Bis-Tris gel (Invitrogen; CA) for resolution by electrophoresis. After separation under denaturing conditions in MOPS-SDS (1 M MOPS, 20.5 mM EDTA, 1 M Tris base, 69.3 mM SDS), the proteins were electro-blotted onto a pre-wetted PVDF-membrane (Immobilon-P transfer membrane, 0.2 μm, Millipore, MA). Transfer was performed using a semi-dry unit (TE77 ECL Semi-Dry Transfer Unit, Amersham Biosciences, UK) for 75 min at a constant current of 0.8 mA/cm² according to the manufacturer's instructions. The membrane was subsequently dried for 2 hr, reactivated, and washed in TBS-T (10 mM Tris base, 150 mM NaCl, 0.05% (v/v) Tween 20, pH 7.5).

After reactivation and washing, the PVDF membrane was blocked in TBS-T (0.1% (v/v) Tween20) supplemented with 5% milk protein (Semper AB, Sweden) for 40 min on a shaker at room temperature (RT). The membrane was probed with the primary antibody anti-HMG-CoAR diluted to a final concentration of 2 μg/ml (1:500) in blocking buffer for 1 hr on a shaker at RT. Following washing in TBS-T (0.1% (v/v) Tween 20), the membrane was incubated with a swine polyclonal horseradish peroxidase-conjugated anti-rabbit IgG antibody (Dako Cytomation; Glostrup, Denmark) diluted 1:3000 in blocking buffer for 1 hr on a shaker at RT. Excess of antibody was washed away as above prior to detection using a SuperSignal West Dura Extended Duration Substrate (Pierce). Chemiluminescence capture and analysis was performed using a G:BOX Chemi HR-16 (Syngene, Cambridge, UK).

Tissue microarray construction and immunohistochemistry

Invasive and sufficient tumour material was obtained from 511 cases. Prior to TMA construction, all tumours were re-evaluated by 1 senior pathologist (LA) and reclassified regarding histological tumour type19 and grade.20

Areas with invasive cancer, whenever possible representing both peripheral and central parts of the tumour, were then marked on H&E stained slides and two 0.6 mm tissue cores collected from each corresponding donor paraffin block and arranged in a new recipient block using a manual tissue arrayer (Beecher, Sun Prairie, WI).

Four micrometer sections were then dried, deparaffinized, rehydrated and antigen retrieval performed by 2 × 5-min microwave treatment in a citrate buffer (pH 6.0). Immunohistochemical processing was performed as described previously for ERα, ERβ, PgR, HER2, Ki67, cyclin D1, and p27.21 A polyclonal anti-HMG-CoAR antibody (diluted1:250 Catalog # 07-457, Upstate) was employed for the present study. All automated immunohistochemistry (IHC) processing was performed using the DAKO Techmate 500 system (DAKO, Copenhagen, DK), except for ERα and PgR, where the Ventana Benchmark system was used (Ventana medical Systems, AZ).

HMG-CoAR was only present in the cytoplasm and the expression was assessed both as the fraction of positive cells (0–1%, 2–10%, 11–50%, and 51–100%) and the staining intensity (negative, weak, moderate, strong). Membranous staining was recorded as absent or present.

In order to assess a potential heterogeneity of the staining distribution of HMG-CoAR, IHC analysis was also performed on 5 full-face tumour tissue sections.

Tumours had previously been grouped into no, low, medium or high expression of Ki67, ERα, ERβ, PgR, cyclin D1 and p27, using 0–1%, 2–10%, 11–50%, and 51–100% positive nuclei as cut-off points.21 HER2 status had been evaluated according to current clinical practice using Herceptest.22

Statistical methods

Correlations between HMG-CoAR expression and other tumour parameters were estimated using the Chi-square test and p-values refer to the linear-by-linear association. For statistical analyses, the HMG-CoAR fractions were recoded into 3 groups (0–1%, 2–50%, and 51–100%) due to small numbers in the groups with 2–10% and 11–50% respectively. Similar recoding (negative, weak, moderate/strong) was performed for the HMG-CoAR intensity due to small numbers in the moderate and strong subgroups. HRT was categorised into combined HRT (estrogen and progestin), ERT (estrogen alone), and no current HRT use. Waist circumference and BMI were categorised into quartiles with the boundaries being 0.71/0.77/0.85 meters and 23/25/28 kg/m², respectively. Each subject was followed until the event of breast cancer, death or end of follow-up, December 31, 2004. The incidence of breast cancer was calculated per 100,000 person-years in different exposure groups. Corresponding relative risks, with 95% confidence interval, were calculated using Cox's proportional hazards analysis. In the multivariate Cox's analysis potential confounders were incorporated and included age at baseline, age at menarche, age at menopause, age at first child birth, parity, use of OC, oophorectomy, educational level, type of occupation, and marital status. Considering that HRT may affect the relation between body mass and breast cancer risk,23–25 all analyses concerning anthropometric variables were limited to women without HRT. In order to examine heterogeneity between different HMG-CoAR subgroups regarding their association to HRT and obesity a case-to-case analysis was performed using an unconditional logistic regression model adjusted for the same covariates as the main analyses. All p-values <0.05 were considered statistically significant.

All statistical analyses were conducted using SPSS version 12 (SPSS, Chicago, IL).

Results

Western blot analysis

In the Western Blot analysis, a single band corresponding to a molecular weight of ∼90 kDa, which most likely represents HMG-CoAR, was detected. The protein was clearly expressed in breast and prostate cancer as well as in Hep-G2 cells. A strong expression was also seen in A-431 cells. There was no detection observed in any of the normal tissues, breast nor prostate (Fig. 1).

Figure 1. Western blot analysis showing the detection of a single band corresponding to an expected M_w of ∼90 kDa. Lane 1, human breast normal tissue lysate; lane 2, human breast tumor tissue lysate; lane 3, human prostate tissue lysate; lane 4, human prostate tumor tissue lysate; lane 5, A-431 cell lysate; lane 6, Hep-G2 cell lysate (ProSci). Some additional faint background is mainly observed in the cell-line lysates.

Immunohistochemical expression of HMG-CoAR in breast cancer

HMG-CoAR expression was evaluable in 396 of 511 cases. The remaining cases/tissue cores were either lost during IHC processing or lacking invasive cancer. There was no difference regarding important clinicopathological parameters between cases with and without information on IHC HMG-CoAR expression (data not shown). HMG-CoAR was expressed in various fractions and intensities in the cytoplasm. Some cases displayed a membranous staining (Fig. 2). No nuclear staining of HMG-CoAR could be detected and in the normal glandular epithelium, staining was sparse. Seventy out of 396 (18%) were entirely negative for HMG-CoAR. The staining distributions are outlined in Table I.

Figure 2. Examples of immunohistochemical HMG-CoAR staining in invasive breast cancer with negative (a), intermediate (b) and strong (c) expression, respectively.

Table I. HMG-CoAR and Correlation with Tumour Size, Node Status, Grade, HER2 Status and Proliferation (Ki 67)

Tumour characteristics (n)	HMG-CoAR fraction (%)			HMG-CoAR intensity (%)			HMG-CoAR membrane (%)
	0–1 (n = 71)	2–50 (n = 60)	51–100 (n = 265)	Negative (n = 70)	Weak (n = 186)	Strong(n = 139)	Negative (n = 337)	Positive (n = 58)
Size
1–10 mm (95)	14	18	28	14	24	30	24	25
11–20 mm (177)	49	47	43	50	45	42	44	49
21–30 mm (73)	17	22	18	16	20	18	19	14
>30 mm (49)	20	13	11	20	11	10	13	12
p-Value	0.008			0.023			0.66
Node status
Negative (233)	67	61	66	67	62	69	66	64
Positive (123)	33	39	34	21	38	31	34	36
p-Value	0.88			0.61			0.88
Grade
1 (102)	16	27	28	16	25	32	27	19
2 (189)	46	48	48	46	49	47	47	50
3 (105)	38	25	24	38	26	21	26	31
p-Value	0.009			0.001			0.20
HER2 status
Negative (220)	58	69	61	57	59	67	64	49
1 + (92)	31	21	25	32	28	19	25	27
2 + (26)	6	6	8	6	7	8	6	13
3 + (20)	5	4	6	5	6	6	5	11
p-Value	0.63			0.55			0.007
Ki67
0–1% (149)	36	44	39	36	39	42	41	30
2–10% (122)	24	24	37	23	31	38	30	43
11–50% (66)	18	24	16	18	20	14	19	11
>50 % (41)	22	8	8	23	10	6	10	16
p-Value	0.023			0.004			0.30

Full-face section analysis revealed that HMG-CoAR was homogeneously expressed in all of the 5 analysed tumours, 2 of which were entirely negative. In line with previous studies,26 HMG-CoAR expression was accentuated in apocrine epithelium (Fig. 3).

Figure 3. Apocrine breast epithelium with strong HMG-CoAR expression.

HMG-CoAR expression in relation to clinicopathological and tumour biological parameters

The correlations between HMG-CoAR expression and established clinicopathological parameters are shown in Table I. High HMG-CoAR expression (fraction and intensity) was significantly associated with a small tumour size, low histological grade and low proliferation. There was no significant association to lymph node status. The presence of membranous staining did not correlate to any clinicopathological parameters (Table I).

Correlations between HMG-CoAR expression and ERα, PgR, ERβ, HER2, cyclin D1 and p27 are shown in Table II. High cytoplasmic HMG-CoAR expression correlated positively with ERα expression whereas an inverse correlation was seen between membranous HMG-CoAR expression and ERα. There was no significant correlation between HMG-CoAR and PgR whereas both cytoplasmic and membranous HMG-CoAR expression was strongly associated with ERβ positivity. Membranous, but not cytoplasmic, HMG-CoAR expression was associated with HER2 expression. There was a strong significant association between cytoplasmic HMG-CoAR expression and p27, but not cyclin D1.

Table II. HMG-CoAR and Correlation with ERα, ERβ, PgR, Cyclin D1 and p27

Tumour characteristics (n)	HMG-CoAR fraction (%)			HMG-CoAR intensity (%)			HMG-CoAR membrane (%)
	0–1 (n = 71)	2–50(n = 60)	51–100 (n = 265)	Negative (n = 70)	Weak(n = 186)	Strong (n = 139)	Negative (n = 337)	Positive(n = 58)
ERα
0–10% (51)	22	10	12	21	13	9	11	23
>10% (335)	78	90	88	79	87	91	89	77
p-Value	0.06			0.02			0.02
ERβ
0–1% (127)	50	37	37	49	38	36	40	35
2–10% (47)	27	22	10	28	15	8	16	4
11–50% (43)	8	14	14	8	12	18	13	13
50% (108)	15	27	39	15	35	38	31	38
p-Value	0.001			0.005			0.04
PgR
0–10% (197)	54	58	53	53	56	52	55	50
>10% (166)	46	42	47	47	44	48	45	50
p-Value	0.82			0.81			0.51
Cyclin D1
0–1% (224)	64	66	56	64	59	32	59	60
2–10% (77)	17	12	23	17	19	40	20	19
11–50% (49)	11	8	14	11	10	45	13	12
>50% (31)	7	14	7	7	12	16	8	9
p-Value	0.39			0.77			0.98
P27
0–1 % (94)	46	27	18	45	27	11	26	20
2–10% (56)	15	12	16	15	13	18	15	17
11–50% (72)	17	27	18	18	20	18	19	19
>50% (153)	22	34	48	22	40	53	40	44
p-Value	<0.001			<0.001			0.48

Risk of HMG-CoAR-defined subgroups of breast cancer in relation to HRT

Potential risk factors for breast cancer among breast cancer cases and the rest of the cohort among postmenopausal women are outlined in Table III.

Table III. Distribution of Potential Risk Factors in Breast Cancer Cases and Rest of Cohort Among Peri/Postmenopausal Women

		Column percent (mean and SD in italics)
Factor (subjects with information: n)	Category	Breast cancer cases(n = 464)	Rest of cohort (n = 11 607)
Age at baseline (years) (12,071)		59.9 (6.2)	60.3 (7.0)
Education (12,035)	O-level college	77	76
	A-level college	4	6
	University	19	18
Type of occupation (11,927)	Manual worker	37	42
	Non-manual worker	56	51
	Employer – self-employed	7	7
Married/cohabiting (12,067)	No	36	34
	Yes	64	66
Age at menarche (years) (11,960)	≤21	21	21
	>12–<15	51	53
	≥51	28	26
Parity (no. of children) (11,841)	0	13	13
	1	22	22
	2	40	40
	3	20	17
	≥4	5	8
Age at first childbirth (years) (12,071)	Nullipara	13	13
	≤02	18	18
	>20 – ≤25	34	35
	>25 – ≤30	25	25
	>30	10	9
Bilateral oophorectomy (12,071)	No	98	98
	Yes	2	2
Age at menopause (years) (11,812)	Perimenopausal	10	9
	≤54	18	18
	>45 – <53	52	53
	≥35	20	19
Exposure to OC (12,056)	Never	58	60
	Ever	42	40
Current HRT (12,030)	No	66	81
	Yes	34	19
Height (m) (12 050)		1.64 (0.06)	1.63 (0.06)
Waist (12,046)	Quartile 1	22	24
	Quartile 2	27	25
	Quartile 3	23	26
	Quartile 4	28	25
Alcohol consumption (12,038)	Nothing last year (teetotaller)	13	13
	Something last year (not last month)	12	13
	Something last month	75	74
Smoking (12,065)	Never	45	47
	Current	26	27
	Ex	29	26

The overall risk of postmenopausal breast cancer was high for women using combined hormone replacement therapy (CHRT). The risk was high for all HMG-CoAR defined subgroups. There was a slightly higher relative risk of tumours with high HMG-CoAR fraction and strong intensity among CHRT users, although the test for heterogeneity was not statistically significant (p = 0.62 and p = 0.72, respectively). The use of ERT was associated with a high risk of tumours expressing membranous HMG-CoAR, and the test for heterogeneity showed that this RR was significantly higher than the RR in relation to membranous negative tumours. Crude and adjusted analyses were similar (Table IV).

Table IV. Incidence of HMG-CoAR Defined Breast Cancer Sub-Groups in Relation to Estrogenreplacement Therapy (ERT) and Combined Hormonal Replacement Therapy (CHRT)

Type of tumour	HRT-use	No of cases	Incidence/100,000	RR (CI: 95%)	RR1 (CI: 95%)
All analyses performed in peri/postmenopausal women. *p < 0.05; ** p < 0.01. 1 Adjusted for age at baseline, education, type of occupation, marital status, age at menarche, parity, age at first child birth, bilateral oophorectomy, age at menopause, use of oral contraceptives, height, BMI, alcohol consumption and smoking habits. 2 Reference group in the heterogeneity analysis. 3 Heterogeneity analysis with p < 0.05.
All	No	305	304	1	1
	ERT	29	291	0.94 (0.64–1.39)	0.90 (0.61–1.33)
	CHRT	127	944	3.12 (2.53–3.84)**	3.32 (2.66–4.15)**
HMG-CoAR fraction
0–1%2	No	36	36	1	1
	ERT	1	10	0.28 (0.04–2.03)	0.30 (0.04–2.21)
	CHRT	14	104	2.88 (1.55–5.34)**	2.80 (1.45–5.41)**
2–50%	No	33	33	1	1
	ERT	1	10	0.30 (0.04–2.22)	0.29 (0.04–2.13)
	CHRT	13	97	2.94 (1.55–5.59)**	3.28 (1.65–6.52)**
51–100%	No	131	130	1	1
	ERT	15	150	1.17 (0.68–1.99)	1.06 (0.62–1.81)
	CHRT	54	401	3.12 (2.27–4.28)**	3.21 (2.29–4.49)**
HMG-CoAR intensity
Negative2	No	35	35	1	1
	ERT	1	10	0.29 (0.04–2.09)	0.31 (0.04–2.29)
	CHRT	14	104	2.96 (1.59–5.51)**	2.93 (1.51–5.69)**
Weak	No	101	101	1	1
	ERT	7	70	0.70 (0.32–1.50)	0.61 (0.28–1.32)
	CHRT	35	260	2.60 (1.77–3.82)**	2.69 (1.76–4.01)**
Moderate/strong	No	63	63	1	1
	ERT	9	90	1.47 (0.73–2.95)	1.45 (0.71–2.94)
	CHRT	32	238	3.86 (2.52–5.92)**	4.18 (2.65–6.59)**
HMG-CoAR membrane
Negative2	No	176	175	1	1
	ERT	11	110	0.63 (0.34–1.16)	0.60 (0.33–1.11)
	CHRT	69	513	2.94 (2.23–3.88)**	2.97 (2.21–3.99)**
Positive	No	23	23	1	1
	ERT	6	60	2.69 (1.09–6.60)3	2.20 (0.87–5.57)3
	CHRT	12	89	3.99 (1.98–8.01)**	4.85 (2.27–10.33)**

Risk of HMG-CoAR defined subgroups of breast cancer in relation to obesity

The overall risk of postmenopausal breast cancer was highest for women in the top quartile of waist circumference. The increased risk among obese women (fourth quartile) was significant for HMG-CoAR (fraction and intensity) positive tumours, but not significantly higher than the risk for HMG-CoAR negative tumours in the case-to-case analysis. Obesity (fourth quartile) was associated with a high risk of breast tumours with negative membranous HMG-CoAR staining, confirmed in the case-to-case analysis for the second and third quartiles. Results remained similar in the multivariate analysis (Table V). Analyses were repeated for BMI quartiles and all results were similar compared to those regarding waist circumferences (data not shown).

Table V. Incidence of HMG-CoAR Defined Breast Cancer Sub-Groups in Relation to Waist Circumference

Type of tumour	Waist quartiles	No. of cases	Incidence/100,000	RR (CI: 95%)	RR1 (CI: 95%)
All analyses performed among peri/postmenopausal women without HRT. * p < 0.05; ** p < 0.01. 1 Adjusted for age at baseline, education, type of occupation, marital status, age at menarche, parity, age at first child birth, bilateral oophorectomy, age at menopause, use of oral contraceptives, height, alcohol consumption and smoking habits. 2 Reference group in the heterogeneity analysis. 3 Heterogeneity analysis with p < 0.05.
All	1	49	164	1	1
	2	79	250	1.51 (1.06–2.15)*	1.49 (1.04–2.14)*
	3	69	212	1.22 (0.85–1.76)	1.17 (0.81–1.70)
	4	102	333	1.80 (1.28–2.53)**	1.78 (1.25–2.53)**
HMG-CoAR fraction
0–1%2	1	6	20	1	1
	2	9	29	1.40 (0.50–3.93)	1.42 (0.50–4.00)
	3	8	25	1.15 (0.40–3.33)	1.20 (0.41–3.51)
	4	13	42	1.85 (0.70–4.87)	2.03 (0.75–5.50)
2–50%	1	6	20	1	1
	2	8	25	1.25 (0.43–3.59)	1.18 (0.41–3.44)
	3	5	15	0.72 (0.22–2.36)	0.64 (0.19–2.13)
	4	14	46	1.99 (0.76–5.19)	1.76 (0.66–4.74)
51–100%	1	19	64	1	1
	2	32	101	1.57 (0.89–2.78)	1.54 (0.87–2.73)
	3	34	105	1.55 (0.89–2.73)	1.49 (0.84–2.63)
	4	45	147	2.06 (1.20–3.51)**	2.02 (1.16–3.51)*
HMG-CoAR intensity
Negative2	1	6	20	1	1
	2	9	29	1.40 (0.50–3.93)	1.41 (0.50–3.97)
	3	8	25	1.15 (0.40–3.33)	1.17 (0.40–3.44)
	4	12	39	1.71 (0.64–4.55)	1.78 (0.65–4.89)
Weak	1	16	54	1	1
	2	21	67	1.23 (0.64–2.35)	1.16 (0.60–2.23)
	3	28	86	1.52 (0.82–2.81)	1.34 (0.72–2.50)
	4	35	114	1.88 (1.04–3.40)*	1.75 (0.95–3.22)
Moderate/strong	1	9	30	1	1
	2	19	60	1.97 (0.89–4.36)	2.06 (0.93–4.60)
	3	11	34	1.07 (0.44–2.57)	1.11 (0.46–2.72)
	4	24	78	2.33 (1.08–5.02)*	2.35 (1.06–5.19)*
HMG-CoAR membrane
Negative2	1	23	77	1	1
	2	47	149	1.91 (1.16–3.14)*	1.87 (1.13–3.09)
	3	43	132	1.62 (0.98–2.69)	1.55 (0.93–2.60)
	4	62	203	2.32 (1.44–3.75)**	2.27 (1.39–3.71)**
Positive	1	8	27	1	1
	2	2	6	0.23 (0.05–1.10)3	0.22 (0.05–1.06)3
	3	4	12	0.44 (0.13–1.44)3	0.41 (0.12–1.41)3
	4	9	29	0.98 (0.38–2.54)	0.99 (0.36–2.70)

Discussion

This is, to our knowledge, the first study on the tumour-specific expression of HMG-CoAR in breast cancer. Our results reveal that HMG-CoAR is expressed in various proportions and intensities in the cytoplasm of the tumour cells, sometimes with a membranous pattern. Cytoplasmic HMG-CoAR expression was associated with more favourable tumour parameters, such as smaller tumour size, lower histological grade, low proliferation and ERα positivity and, furthermore, with the expression of ERβ and p27. In contrast, the presence of membranous HMG-CoAR staining was associated with ERα negativity and HER2 positivity but not to any other clinicopathological parameters. Moreover, there were indications that estrogen related factors such as HRT use and obesity could be associated with the risk of breast tumours with high HMG-CoAR expression.

The mevalonate pathway, including the key enzyme HMG-CoAR, has lately received rising attention due to the suggested anti-neoplastic properties of statins, which exert its main effect on HMG-CoAR. Statins are proposed to have preventive12 as well as reversible effects on tumours.3 The mechanisms by which statins prevent tumour initiation, promotion, and proliferation are not fully understood, however, it is agreed on that targeting cell cycle regulating proteins is central, and that statins are capable of inducing G1 arrest.3, 9–11 Given the proposed association between statins and tumour biology, we hypothesise that the expression of HMG-CoAR in cancer is associated with other tumour characteristics. The observed concordance between high expression of HMG-CoAR and tumours characterised by less aggressive features could seem to be in disagreement with the image of statins inhibiting HMG-CoAR resulting in decreased cellular proliferation. However, epidemiological studies on statin use and cancer incidence have reported a general decreased risk of i.e. breast cancer among statin users without considering breast cancer as a heterogeneous disease with potential different etiology in different sub-groups.15, 16 Furthermore, less well-differentiated cancer cells might be unable to maintain an intact mevalonate pathway. Lastly, cholesterol and its derivates can lead to accelerated degradation of HMG-CoAR,27 and the negative feed-back system makes the immunohistochemical results complex to interpret.

A recent study on a colon cancer cell line indicated that estrogen is capable of down-regulating HMG-CoAR activity and protein expression.7 Estrogen levels in serum can be modified by HRT use and obesity18 which were used as indicators of estrogen exposure in the present study. HRT is a powerful risk factor for breast cancer28 and was associated with all HMG-CoAR defined subgroups of breast cancer, although relative risks were slightly higher in the groups of tumours with strong HMG-CoAR expression. In our study, HMG-CoAR positivity corresponded with other prognostically favourable tumour markers and in a previous work we reported that HRT use was associated with less aggressive breast cancer,21 which supports our present findings. Obesity is another established risk factor for breast cancer among postmenopausal women that do not use HRT,25 and in our study, obesity seemed to be associated with a high risk of HMG-CoAR positive breast cancer. This is in line with the findings regarding HRT use, likewise leading to higher levels of serum estrogen. The disparate associations to established clinicopathological parameters for the cytoplasmic and membranous staining modalities respectively, where the latter correlated positively with HER2 but inversely to ERα, cannot be explained by the descriptive approach used here. Furthermore, the Cox analyses referring to membrane staining are rather difficult to interpret due to a small number of tumours exhibiting membranous staining, giving a limited statistical power. Further studies are needed in order to shed light on the functional aspects of the sub-cellular localization of HMG-CoAR.

Information on medication was available for most participants of the MDCS, however, statins were still infrequently prescribed during the years of baseline examinations and only 228 out of 17,035 female participants were current users of any type of statin. Among statin users, only 8 women were diagnosed with incident breast cancer and further statistical analyses were not possible. Data on statin use subsequent to baseline examinations were not available, which would have been of interest to investigate in relation to expression of HMG-CoAR.

Some methodological considerations need to be discussed. Western blot analysis revealed a single band corresponding to the molecular weight of HMG-CoAR of ∼90 kDa. However, 2 HMG-CoAR variants have been described; a 97 kDa protein localised in endoplasmatic reticulum and a 90-kDa protein localised in peroxisomes29 with both structural and functional differences. The antibody used here has to our knowledge not been used previously in any publication on HMG-CoAR expression in breast cancer. However, in line with Celis et al.26 we observed a distinct staining of apocrine epithelium (Fig. 3).

Furthermore, given the homogenous expression pattern observed on a subset of full-face sections, we believe that the heterogeneity issue is of minor importance when applying the TMA-approach for high-throughput analyses of this marker.

A potential source of misclassification concerns HRT-use. In the MDCS, information on HRT-use was provided from 2 different sources thereby minimising the number of under-reporters. Over-reporting of HRT-use was considered to be rare. Taken together, misclassification on HRT was probably low.

To date, anthropometric studies have not yet agreed on the most appropriate anthropometric variable considering prediction of cancer. However, we have previously evaluated breast cancer risk in relation to 6 different anthropometric measurements (unpublished data), and the strongest association was found for waist circumference. Hence, we consider it reasonable to use waist circumference in our study.

In conclusion, by performing high-throughput analyses of tumour-specific HMG-CoAR expression in 511 incident breast cancers from the MDCS study, we demonstrate that HMG-CoAR is differentially expressed in breast cancer and that HMG-CoAR expression is associated with favourable prognostic clinicopathological parameters. In line with these findings, it was indicated that the incidence of HMG-CoAR high tumours could be associated with obesity and use of CHRT. Further studies on the tumour-specific expression of HMG-CoAR in breast cancer may shed light on its role in tumour progression and as a potential target for chemopreventive and therapeutic approaches.

Acknowledgements

The authors are grateful to Ms. Elise Nilsson for excellent technical assistance in the construction of the TMAs.