Associations of the Transforming Growth Factor β/Smad Pathway, Body Mass Index, and Physical Activity With Breast Cancer Outcomes: Results From the Shanghai Breast Cancer Study

Abstract

The transforming growth factor β (TGF-β) pathway plays an important role in breast cancer progression and in metabolic regulation and energy homeostasis. The prognostic significance of TGF-β interaction with obesity and physical activity in breast cancer patients remains unclear. We evaluated the expression of TGF-β type II receptor and pSmad2 immunohistochemically in breast cancer tissue from 1,045 patients in the Shanghai Breast Cancer Study (2002–2005). We found that the presence of nuclear pSmad2 in breast cancer cells was inversely associated with overall and disease-free survival, predominantly among participants with lower body mass index (BMI; weight (kg)/height (m) ² ) and a moderate level of physical activity. However, the test for multiplicative interaction produced a significant result only for BMI (for disease-free survival and overall survival, adjusted hazard ratios were 1.79 and 2.05, respectively). In 535 earlier-stage (T1-2, N0) invasive cancers, nuclear pSmad2 was associated with improved survival among persons with higher BMI (overall survival: adjusted hazard ratio = 0.27, 95% confidence interval: 0.09, 0.86). The cytoplasmic pattern of TGF-β type II receptor expression in cancer cells was significantly associated with a lower survival rate but was not modified by BMI or physical activity. Our study suggests that the TGF-β pathway in tumor cells is involved in breast cancer prognosis and may be modified by BMI through pSmad2.

Breast cancer represents a group of heterogeneous neoplasms with a high degree of diversity. The outcome of breast cancer varies not only because of the discrepancy of tumor-intrinsic features (classical histological and molecular classifications) but also because of the features of the local microenvironment (reactive stromal cells) and the systemic macroenvironment, such as age, menopausal status, body mass index (BMI), and overall immune status ( 1 , 2 ).

The transforming growth factor β (TGF-β) pathway plays a dual role in human cancer, functioning as a tumor suppressor via its antiproliferative effects in early tumor stages but radically changing to a tumor promoter that enhances tumor growth, invasion, and metastasis in later stages ( 3–5 ). The most widely studied component of the TGF-β pathway in breast cancer is TGF-β type I. It binds with a high affinity to TGF-β type II receptor (TGF-β-RII), which transactivates TGF-β type I receptor to initiate an intracellular signaling cascade by bringing about phosphorylation of 2 receptor-regulated Smads (Smad2 and 3), recruiting the common-partner Smad (Smad4) to form the Smad4-Smad2/3 complexes, and translocating to the nucleus as the transcription factor to regulate expression of target genes ( 4 ). Thus far, the clinical prognostic role of TGF-β pathway components in human breast cancer remains elusive, which may be due to the complexity introduced by a variety of other molecules and systemic factors, such as estrogen receptor α and human epidermal growth factor receptor 2 status, common-partner and inhibitory Smads, early-age onset, and menopausal status ( 1 , 5–11 ). While several studies have been published on the prognostic role of pathway components such as TGF-β receptors, Smad2/3, and other regulatory molecules, their results conflict with respect to the association of cancer survival with expression of TGF-β-RII ( 7–9 ) and pSmad2/3 and pSmad4 in breast cancer cells ( 11 , 12 ). One reason for the inconsistent results, beyond diversities in study population and methodology, might be the small sample sizes involved.

Studies indicate that the TGF-β superfamily also has pleiotropic roles in metabolism, energy homeostasis, and adipogenesis, mainly through downstream molecules of Smads that induce obesity and other metabolic abnormalities ( 13–15 ). In human subjects, a significant association between BMI and TGF-β1 concentration in circulating blood ( 16 , 17 ) or adipose tissue ( 18 , 19 ) has been reported. Because of the intrinsic gene-environment interaction between TGF-β and obesity, modulation of the TGF-β pathway has been suggested as a promising treatment strategy for obesity and its comorbid conditions ( 20 , 21 ). In addition to BMI, physical activity appears to be another systemic factor associated with the TGF-β signaling pathway. Studies reveal that physical exercise significantly increases the serum level of TGF-β1 in human subjects ( 22–24 ). Based on current knowledge of the multifaceted roles of the TGF-β pathway in cancer and systemic metabolism, it is plausible that the TGF-β pathway may be associated with breast cancer outcome, and this association may be affected by patient characteristics such as BMI and physical activity. We have reported previously that the cytoplasmic expression pattern of TGF-β-RII and higher nuclear pSmad2 expression in breast cancer cells are associated with decreased disease-free survival ( 25 ). The purpose of the current study was to examine the potential influences of postdiagnostic BMI and physical activity on the association of 2 important components of the canonical TGF-β/Smad pathway, TGF-β-RII and pSmad2, with breast cancer outcomes.

METHODS

Study population

Study participants were enlisted from the second phase of the Shanghai Breast Cancer Study, a population-based case-control study conducted in Shanghai, China. The study design has been described elsewhere ( 26 , 27 ). Through the Shanghai Cancer Registry, a population-based cancer registry, 1,918 eligible breast cancer patients were identified and recruited into the study between April 2002 and February 2005. The participants, aged 20–70 years, had no prior history of cancer and were diagnosed with a primary breast cancer. The cancer tissue sections from 1,045 of these cases were collected from the diagnosis hospitals according to a standard protocol. Briefly, the formalin-fixed, paraffin-embedded blocks were cut into 5-µm-thick sections. The sectioned tissue slides were covered with a thin layer of paraffin, stored in vacuum chambers, and placed in a 4°C cold room to properly preserve the antigenicity of the sectioned tissues.

The Shanghai Breast Cancer Study protocol was approved by the institutional review boards of Vanderbilt University (Nashville, Tennessee) and the Shanghai Center for Disease Prevention and Control (Shanghai, China). Written informed consent was obtained from all participants.

Data collection

Trained interviewers conducted in-person interviews using a standard baseline questionnaire to collect information on demographic characteristics, disease history, medication use, and selected lifestyle factors. Anthropometric measurements, including height, weight, and the circumferences of the waist and hips, were taken 6 months after diagnosis according to a standard protocol by trained interviewers. BMI was calculated on the basis of these measurements and was defined as weight (kg) divided by the square of height (m ² ). Collection of physical activity data has been described in detail elsewhere ( 28 ). Briefly, using a validated exercise questionnaire, participants were asked after diagnosis whether they participated in exercise regularly (at least twice a week), the most common activities in which they participated, and how many hours per week they spent exercising. The intensity level of each activity was assigned a metabolic equivalent of task (MET) score based on the method proposed by Ainsworth et al. ( 29 ). MET-hours/week for each activity were calculated by multiplying the number of hours per week the participant reported engaging in that activity by the assigned MET score. The MET scores for individual activities were summed to derive a total exercise MET score.

Clinical information collected included cancer stage, tumor estrogen receptor α and progesterone receptor status, and primary treatments. The human epidermal growth factor receptor 2 status of cancer cases was evaluated previously by a centralized laboratory ( 30 ). Diagnoses and clinicopathological data were confirmed through a combination of medical chart review and centralized review of pathology slides. Histological types were confirmed by the research pathologist (Y.S.) according to the criteria of the World Health Organization ( 31 ). The histological grade of all cancer slides was determined using the Nottingham histological grading system ( 32 ). Patients with cancer have been followed for survival status and breast cancer recurrence through a combination of record linkages with the Shanghai Vital Statistics Registry and in-person surveys.

Double-label fluorescent immunohistochemistry for TGF-β-RII and pSmad2

The sequential immunofluorescence double-staining method used has been described previously ( 25 ). Briefly, the breast cancer tissue slides were deparaffinized, followed by antigen retrieval using a pressure cooker. After blocking steps with 3% hydrogen peroxide, 5% normal goat serum, biotin solution, and avidin D solution (catalog number SP-2001; Vector Laboratories, Burlingame, California), the slides were incubated with polyclonal rabbit antibody anti-TGF-β-RII (catalog number E11244, 1:100; Spring Bioscience, Pleasanton, California) overnight at 4°C; biotin-conjugated goat anti-rabbit antibody (catalog number BA-1000, 1:300; Vector Laboratories) for 30 minutes at 37°C; and streptavidin-Cy3 (catalog number 43-8315, 1:100; Zymed Laboratories Inc., South San Francisco, California) for 15 minutes at 37°C. The slides were then incubated with polyclonal rabbit antibody anti-pSmad2 (Ser465/467) (catalog number 9510, 1:200; Cell Signaling Technology, Danvers, Massachusetts) for 30 minutes at 37°C; biotin-conjugated goat anti-rabbit antibody for 30 minutes at 37°C; and streptavidin-FITC (catalog number 43-8311, 1:100; Zymed Laboratories Inc.) for 30 minutes at 37°C. Slides were washed thoroughly between all of the above steps. The cover slips were mounted with ProLong Gold Antifade Reagent with DAPI (catalog number P36935; Invitrogen Corporation, Carlsbad, California), and slides were stored in the dark at 4°C.

The staining protocol was validated by comparing it with conventional single stains using the Dako EnVision+ Kit (catalog number K4011; Dako North America, Inc., Carpinteria, California) and a quality control tissue microarray block. The quality control tissue microarray slides were stained in parallel with each batch of staining samples. Four tissue microarray slides made by our centralized laboratory, which included 182 valid cases of breast cancer, were also stained as a training set before formal study. Consistent staining results were observed with our staining system by comparing built-in control tissues and cell lines. The stained slides were scored blinded with respect to clinical patient data. All slides with inconsistent results were evaluated again by 2 investigators (Y.S. and Q.Q.) jointly, and a consensus score was obtained. TGF-β-RII intensity and pSmad2 intensity were semiquantified using a 4-scale Allred scoring system (see Supplementary Data A, available at http://aje.oxfordjournals.org/ ). The subcellular staining patterns of TGF-β-RII were recorded as either “membranous,” which is beehive-like in appearance, or “cytoplasmic,” which has a cloudy appearance in the cytoplasm of the cells ( Supplementary Data B).

Statistical analysis

Analysis of variance was used for comparing continuous variables and a χ ² test was used for comparing dichotomous variables across the different intensity groups of TGF-β-RII and pSmad2 expression and TGF-β-RII patterns. Outcomes of the study were defined as recurrence/breast cancer-specific mortality (disease-free survival) and all-cause mortality (overall survival). Event-free participants were censored at the date of last follow-up. Five-year rates of disease-free survival and overall survival were estimated using the Kaplan-Meier method. A total of 24 in situ breast cancer cases (stage 0) were excluded from the survival analysis because most cases of in situ ductal carcinoma (47%–83%) do not progress to invasive cancer within 10 years ( 33 ). In addition, 7 cases of stage 4 breast cancer were excluded from disease-free survival analysis. Cox proportional hazards models were used to estimate hazard ratios and 95% confidence intervals with adjustment for other prognostic variables, including age at diagnosis, tumor size, grade, tumor-node-metastasis (TNM) stage, estrogen receptor α, progesterone receptor, human epidermal growth factor receptor 2, radiotherapy, chemotherapy, immunotherapy, and tamoxifen treatment. Analyses stratified by BMI and physical activity were carried out. Physical activity was classified into 3 categories: no exercise (MET score = 0), less than the median amount of activity (<9.5 MET-hours/week), and greater than the median amount (≥9.5 MET-hours/week). All tests were performed using SAS software (version 9.3; SAS Institute, Inc., Cary, North Carolina). The significance levels were set at P < 0.05 for 2-sided analyses.

RESULTS

The characteristics of the study population are shown in Table  1 . A total of 1,045 breast cancer patients with an average age at diagnosis of 51.4 years were recruited into this study. Approximately one-third of participants (34.3%) were diagnosed at an early stage (TNM stages 0 and 1), more than half (54.5%) at an intermediate stage (TNM stages 2a and 2b), and about one-tenth (11.2%) at a late stage (TNM stages 3 and 4). We found that 62.7% of the breast cancer cases had estrogen receptor α positivity and 30.4% had human epidermal growth factor receptor 2 overexpression. Triple-negative breast cancer (TNBC) accounted for 15% of cases. For cancer treatment, all patients received surgery, and 94.4% received chemotherapy.

Table 1.

Characteristics of Study Participants, Shanghai Breast Cancer Study, 2002–2005

Characteristic	Total No. of Subjects	No. of Subjects	%
Age group, years	1,045
<45		213	20.4
45–49		321	30.7
≥50		511	48.9
Age, years a		51.4 (8.3)
Menopausal status	1,045
Premenopausal		570	54.6
Postmenopausal		475	45.5
Family history of breast cancer	1,045
No		992	94.9
Yes		53	5.1
TNM stage	1,015
0		24	2.4
1		324	31.9
2a		334	32.9
2b		219	21.6
3		107	10.5
4		7	0.7
Histological grade	1,038
1		175	16.9
2		528	50.9
3		335	32.3
Tumor size, cm	981
≤2		437	44.6
>2		544	55.5
ER-α, PR, and HER2 status
ER-α-positive		655	62.7
PR-positive		642	61.4
HER2-positive		313	30.4
Molecular type	844
Luminal A		443	52.4
Luminal B		150	17.8
HER2		125	14.8
Triple-negative		127	15.0
Cancer therapy received
Chemotherapy		986	94.4
Radiotherapy		335	32.1
Tamoxifen		566	54.2

Abbreviations: ER-α, estrogen receptor α; HER2, human epidermal growth factor receptor 2; PR, progesterone receptor; SD, standard deviation; TNM, tumor-node-metastasis.

^a Values are presented as mean (standard deviation).

The pSmad2 positive signal was mainly seen in nuclei ( Supplementary Data A). The expression of nuclear pSmad2 in cancer cells was not associated with postdiagnostic BMI and physical activity (Table  2 ). The high intensity (score 3) of nuclear pSmad2 expression in cancer cells was associated with lower 5-year overall survival ( P = 0.02), with an adjusted hazard ratio of 1.39 (95% confidence interval (CI): 1.03, 1.89), and disease-free survival ( P = 0.03), with an adjusted hazard ratio of 1.34 (95% CI: 0.98, 1.84) (Table  3 ). The association between a high level of nuclear pSmad2 and a poorer likelihood of survival was predominantly seen among participants with lower BMI (<25) and moderate physical activity levels (<9.4 MET-hours/week), although the test for multiplicative interaction was significant only for BMI (for disease-free survival, adjusted hazard ratio (HR) = 1.79, 95% CI: 1.17, 2.76 ( P_interaction = 0.02); for overall survival, adjusted HR = 2.05, 95% CI: 1.36, 3.11 ( P_interaction < 0.001)). We also performed analyses using a BMI cutoff point of 24 for overweight, as suggested by the Chinese Obesity Task Force ( 34 ). Similarly, the association between a high level of nuclear pSmad2 and a poorer likelihood of survival was predominantly seen among participants with lower BMI (<24), although the interaction was not significant ( Supplementary Data ).

Table 2.

Associations of Transforming Growth Factor β Type II Receptor and pSmad2 Expression With Body Mass Index and Physical Activity in Breast Tumor Cells, Shanghai Breast Cancer Study, 2002–2005

Variable	No. of Cases	TGF-β-RII Intensity, a %			P Value	TGF-β-RII Pattern, %		P Value	pSmad2 Intensity, a %			P Value
		0–1	2	3		Membranous	Cytoplasmic		0–1	2	3
Body mass index b	954	24.3 (3.3) c	24.3 (3.1)	24.1 (3.4)	0.572	25.1 (3.7)	24.0 (3.1)	<0.001	24.1 (3.4)	24.2 (3.4)	24.2 (3.1)	0.919
<25	622	9.6	42.0	48.4		13.2	86.8		18.8	28.0	53.2
25–29.9	279	9.3	43.4	47.3		21.2	78.8		17.9	30.1	52.0
≥30	53	13.2	35.8	50.9	0.836	34.0	66.0	<0.001	24.5	35.8	39.6	0.411
Physical activity, MET-hours/week	955	7.3 (8.3)	7.5 (10.3)	8.5 (10.6)	0.288	8.9 (10.3)	7.8 (10.3)	0.193	8.7 (10.3)	7.6 (10.1)	7.8 (10.4)	0.531
0	327	11.0	44.7	44.3		13.5	86.5		18.4	27.5	54.1
<9.5	316	7.9	42.7	49.4		17.1	82.1		15.2	32.6	52.2
≥9.5	312	10.6	38.4	51.0	0.295	19.5	80.5	0.114	23.1	26.9	50.0	0.10

Abbreviations: MET, metabolic equivalent of task; TGF-β-RII, transforming growth factor β type II receptor.

^a TGF-β-RII intensity and pSmad2 intensity were semiquantified using a 4-scale Allred scoring system ( Supplementary Data A): 0 (negative), no positive staining or fewer than one-third of cells with a weak fluorescent signal difficult to identify under an X100 field; 1 (weakly positive), more than one-third of cells with a weak fluorescent signal; 2 (moderately positive), more than one-third of cells with a moderate fluorescent signal easily identified under an X100 field, or fewer than two-thirds of cells with a strong fluorescent signal; and 3 (strongly positive), more than two-thirds of cells with a strong fluorescent signal.

^b Weight (kg)/height (m) ² .

^c Values with parentheses are presented as mean (standard deviation).

Table 3.

Breast Cancer Survival According to Transforming Growth Factor β Type II Receptor and pSmad2 Expression, Body Mass Index, and Physical Activity, Shanghai Breast Cancer Study, 2002–2005

Variable	Intensity a or Pattern	Total No. of Cases	Overall Survival						Disease-Free Survival
			No. of Events	5-Year Survival	P Value b	HR c	95% CI	P for Interaction	No. of Events	5-Year Survival	P Value b	HR c	95% CI	P for Interaction
pSmad2 Intensity
All patients	0–2	449	73	0.90					67	0.87
	3	493	109	0.85	0.02	1.39	1.03, 1.89		101	0.82	0.03	1.34	0.98, 1.84
BMI d
<25	0–2	283	35	0.93					33	0.90
	3	326	76	0.84	<0.01	2.05	1.36, 3.11		66	0.82	<0.01	1.79	1.17, 2.76
≥25	0–2	166	38	0.84					34	0.82
	3	166	32	0.88	0.38	0.72	0.44, 1.18	<0.001	35	0.81	0.99	0.85	0.52, 1.39	0.02
PA, MET-hours/  week
0	0–2	147	23	0.90					22	0.85
	3	173	41	0.81	0.07	1.12	0.66, 1.92		35	0.81	0.19	1.07	0.61, 1.86
<9.5	0–2	150	21	0.93					16	0.93
	3	165	34	0.86	0.11	1.60	0.91, 2.84		36	0.81	<0.01	2.01	1.09, 3.72
≥9.5	0–2	152	29	0.85					29	0.83
	3	155	34	0.88	0.62	1.17	0.68, 1.99	0.81	30	0.84	0.98	0.93	0.54, 1.62	0.94
TGF-β-RII Intensity
All patients	0–2	490	100	0.87					92	0.83
	3	452	82	0.88	0.36	0.98	0.73, 1.32		76	0.86	0.40	0.96	0.70, 1.30
BMI
<25	0–2	316	64	0.87					56	0.83
	3	293	47	0.89	0.18	0.82	0.56, 1.21		43	0.88	0.30	0.88	0.59, 1.31
≥25	0–2	173	35	0.87					36	0.81
	3	159	35	0.86	0.75	1.27	0.78, 2.09	0.24	33	0.83	0.95	1.01	0.62, 1.65	0.64
PA, MET-hours/  week
0	0–2	179	37	0.86					35	0.82
	3	141	27	0.84	0.76	1.03	0.62, 1.72		22	0.85	0.43	0.82	0.48, 1.41
<9.5	0–2	160	29	0.88					27	0.85
	3	155	26	0.91	0.67	0.95	0.55, 1.66		25	0.88	0.77	0.93	0.53, 1.65
≥9.5	0–2	151	34	0.85					30	0.81
	3	156	29	0.88	0.39	0.92	0.55, 1.53	0.94	29	0.86	0.74	1.00	0.59, 1.69	0.52
TGF-β-RII Pattern
All patients	1 e	153	21	0.92					20	0.90
	2 f	789	161	0.86	0.06	1.54	0.97, 2.46		148	0.83	0.07	1.59	0.99, 2.57
BMI
<25	1	76	8	0.93					7	0.93
	2	533	103	0.87	0.06	1.68	0.81, 3.48		92	0.84	0.07	1.73	0.79, 3.77
≥25	1	77	13	0.90					13	0.87
	2	255	57	0.85	0.32	1.54	0.77, 3.08	0.88	56	0.80	0.28	1.57	0.80, 3.07	0.92
PA, MET-hours/  week
0	1	41	6	0.90					5	0.90
	2	279	58	0.85	0.35	1.24	0.52, 2.97		52	0.82	0.28	1.57	0.61, 4.03
<9.5	1	53	6	0.96					6	0.94
	2	262	49	0.88	0.17	1.08	0.43, 2.69		46	0.85	0.23	1.03	0.42, 2.54
≥9.5	1	59	9	0.88					9	0.86
	2	248	54	0.86	0.30	1.87	0.89, 3.92	0.63	50	0.83	0.38	1.82	0.86, 3.83	0.92

Abbreviations: BMI, body mass index; CI, confidence interval; HR, hazard ratio; MET, metabolic equivalent of task; PA, physical activity; TGF-β-RII, transforming growth factor β type II receptor.

^a TGF-β-RII intensity and pSmad2 intensity were semiquantified using a 4-scale Allred scoring system ( Supplementary Data A): 0 (negative), no positive staining or fewer than one-third of cells with a weak fluorescent signal difficult to identify under an X100 field; 1 (weakly positive), more than one-third of cells with a weak fluorescent signal; 2 (moderately positive), more than one-third of cells with a moderate fluorescent signal easily identified under an X100 field, or fewer than two-thirds of cells with a strong fluorescent signal; and 3 (strongly positive), more than two-thirds of cells with a strong fluorescent signal.

^b P values were derived using the Kaplan-Meier method.

^c Adjusted for age at diagnosis, tumor size, grade, tumor-node-metastasis stage, estrogen receptor α, progesterone receptor, human epidermal growth factor receptor 2, radiotherapy, chemotherapy, immunotherapy, and tamoxifen.

^d Weight (kg)/height (m) ² .

^e Membranous pattern.

^f Cytoplasmic pattern.

Given that the TGF-β pathway plays a role in both the early and later stages of tumor progression, we performed analyses separately for earlier-stage and later-stage breast cancer. Among 535 earlier-stage (stages 1 and 2 without lymph node metastasis; T1-2, N0) invasive breast cancer cases, increased nuclear pSmad2 was related to a favorable outcome only among persons with a higher BMI (≥25) (for overall survival, adjusted HR = 0.27, 95% CI: 0.09, 0.86). Among the 376 cases of later-stage (stage 2 with lymph node metastasis, stages 3 and 4) breast cancer, increased nuclear pSmad2 was related to an unfavorable outcome among nonoverweight patients (BMI <25) (for overall survival, adjusted HR = 2.29, 95% CI: 1.33, 3.93; for disease-free survival, adjusted HR = 2.19, 95% CI: 1.23, 3.88) ( Supplementary Data ).

The TGF-β-RII-positive signal in invasive breast cancer cells exhibited patterns that were mainly cytoplasmic (795/954 (83.3%)) and sometimes membranous (159/954 (16.7%)) ( Supplementary Data B). Compared with the membranous pattern, the cytoplasmic pattern of TGF-β-RII in cancer cells was more prevalent among patients with a low BMI ( P < 0.001) (Table  2 ) and was associated with lower 5-year overall survival (adjusted HR = 1.54, 95% CI: 0.97, 2.46; P = 0.06) and lower 5-year disease-free survival (adjusted HR = 1.59, 95% CI: 0.99, 2.57; P = 0.07) (Table  3 ). This association was not modified by BMI or physical activity (Table  3 and Supplementary Data ). The intensity of TGF-β-RII in cancer cells was not associated with BMI or physical activity (Table  2 ) or with survival from breast cancer (Table  3 ). However, a high level of TGF-β-RII was associated with better 5-year overall survival among patients with low BMI (<24) ( P = 0.07); the adjusted hazard ratio was 0.64 (95% CI: 0.41, 1.01; P for interaction = 0.04) ( Supplementary Data ). The associations of TGF-β-RII with breast cancer outcome did not differ significantly by tumor stage ( Supplementary Data ).

Additional analyses were performed separately for TNBC (114 cases) and non-TNBC (645 cases). The associations of pSmad2 and TGF-β-RII expression with breast cancer outcomes did not differ between TNBC and non-TNBC cases (data not shown).

DISCUSSION

In this study, we found that a high level of nuclear pSmad2 in breast cancer cells was significantly associated with reduced overall survival and disease-free survival. This association was predominantly seen among participants who had lower BMI (<25) and a moderate level of physical activity (<9.4 MET-hours/week), although the test for multiplicative interaction was significant only for BMI. In earlier tumor stages, however, increased nuclear pSmad2 in tumor cells was significantly associated with a favorable outcome among persons with a higher BMI (≥25). These findings suggest that the presence of nuclear pSmad2 in breast cancer cells may be a prognostic biomarker modified by BMI. The cytoplasmic pattern of TGF-β-RII expression in cancer cells was significantly associated with cancer survival but was not modified by BMI and physical activity.

As an important downstream effector of the canonical TGF-β/Smad pathway, the nuclear accumulation of pSmad2 is considered indicative of active canonical TGF-β signaling ( 11 ). We recently reported that overexpression of nuclear pSmad2 in tumor cells was significantly associated with lower 5-year disease-free survival after adjustment for all known clinicopathological prognostic factors ( 25 ). Our results are consistent with findings from 2 large-scale studies ( n = 623 and n = 574) ( 8 , 11 ), providing strong evidence that nuclear pSmad2 in tumor cells is an independent prognostic biomarker for breast cancer. Given the intrinsic gene-environment interaction between TGF-β and obesity ( 20 , 21 ), we conducted stratified analysis in this study showing that the association between nuclear pSmad2 expression and breast cancer outcome is modified by BMI, with the association predominantly seen among patients who have lower BMI. A similar host-tumor interaction in β-catenin expression and cancer survival was previously shown in a colon cancer study ( 35 ).

Pre- and postdiagnostic obesity are known to inversely affect breast cancer prognosis ( 36–39 ), but whether the body weight of patients has an effect on the prognostic role of the TGF-β pathway is unclear. We found that a high level of nuclear pSmad2 in breast cancer cells was only associated with breast cancer outcome among nonobese breast cancer patients, suggesting that this measure cannot serve as a prognostic biomarker for overweight/obesity patients. This variation may be explained by obesity-related biological changes, such as increased hormone/growth factors (e.g., estrogens, insulin, insulin-like growth factor 1, leptin, adiponectin, and hepatocyte growth factor), as well as an increased level of inflammatory cytokines and vascular regulators ( 40 ), which may be dominant among obese individuals and mask the effects of the TGF-β pathway. Interestingly, we found that overexpression of nuclear pSmad2 in tumor cells was significantly associated with a favorable outcome in patients with earlier stages of breast cancer, as opposed to the later stages, and this association was seen only among persons with higher BMI (≥25). This is consistent with previous findings ( 12 ) and with the well-documented dual roles of the TGF-β pathway in tumor progression ( 3–5 ). This study indicates that the protective role of the TGF-β pathway in earlier-stage invasive breast cancer is significant only among overweight/obese individuals, who have a higher risk of death than nonoverweight individuals ( 36–38 ). The underlying mechanisms are unclear, and additional studies are warranted.

TGF-β-RII is the ligand-binding receptor for all members of the TGF-β family (TGF-β types I, II, and III). The association between TGF-β-RII and breast cancer outcome is inconclusive ( 7–9 , 11 ). Our study found that it is not the TGF-β-RII intensity in breast cancer cells but its subcellular localization patterns that are associated with breast cancer survival, independent of other prognostic factors. To the best of our knowledge, this is the first finding that the cytoplasmic pattern of TGF-β-RII is more prevalent among patients with low BMI ( P < 0.01) and is associated with lower 5-year survival (Tables  2 and 3 ). Previous studies showed that TGF-β-RII was more abundant in the cytosolic compartment than in the membrane compartment of 4 detected breast cancer cell lines ( 41 ). The cytosolic form of TGF-β-RII, which has a different glycosylation pattern than the wild-type membrane form, did not bind ¹²⁵ I-labeled TGF-β1 but had detectable in vivo and in vitro kinase activity. When MCF-7 breast cancer cells were induced into cell differentiation, the membrane localization of TGF-β-RII and TGF-β response was restored. These findings support observations that the presence of the cytosolic form of TGF-β-RII in breast cancer cells might have important biological and prognostic significance in breast cancer progression. However, unlike pSmad2, the prognostic significance of this cytoplasmic pattern of TGF-β-RII in breast cancer cells is not modified by BMI or physical activity.

Our previous study showed that exercise following breast cancer diagnosis may improve overall survival and disease-free survival for breast cancer patients ( 28 ), a finding that concurred with that of another report ( 42 ). In this study, we found that physical activity does not modify the association of altered TGF-β-RII and pSmad2 expression in cancer cells with cancer survival. Currently there is no evidence that physical exercise affects TGF-β-RII and pSmad2 expression in normal tissue or tumor tissue, except that it significantly increases the serum level of TGF-β1 in human subjects ( 22–24 ). It is possible that physical activity may not be associated with TGF-β-RII or pSmad2 expression in residual cancer cells or with the tumor cells that are disseminated to other organs ( 43 ).

Our study has several noticeable strengths. To our knowledge, this is the largest population-based study to date focusing on the prognostic role of the canonical TGF-β/Smad pathway. The data on anthropometric measurements, physical activity, clinicopathological variables, and cancer treatment information were prospectively collected. The pathology diagnosis and histological grading were reviewed and confirmed in a centralized laboratory. The staining assay methods for pSmad2 and TGF-β-RII were carefully validated before the formal examination of study samples. The stained slides were scored separately by 2 investigators blinded to clinical data, and all slides with inconsistent results were reevaluated jointly to reach a consensus score. A major limitation of this study is the collection of tissue slides from multiple diagnostic hospitals, resulting in the possibility that the degradation of protein antigenicity may have varied despite our use of a standard protocol to collect, process, and store tissue sections to maximally preserve tissue antigens ( 44 ). This potential antigen degradation might have reduced the statistical power of the study. Although our study is (to our knowledge) the largest population-based study of this topic to have been conducted to date, the sample sizes were small in some of the strata. Thus, the power to detect interactions in some stratified analyses was limited. Further studies with larger sample sizes are warranted to confirm our findings.

In summary, we found that nuclear pSmad2 expression intensity and the cytoplasmic expression pattern of TGF-β-RII in breast cancer cells were associated with breast cancer prognosis. The overexpression of nuclear pSmad2 was associated with unfavorable outcomes, especially for patients with lower BMI, but was associated with favorable outcomes for overweight/obese individuals in early-stage invasive breast cancer.